c/EPA
United States      Industrial Environmental Research  EPA-600/7-78-170
Environmental Protection  Laboratory          August 1978
Agency        Research Triangle Park NC 27711
Proceedings:
Symposium on New
Concepts for Fine
Particle Control

Interagency
Energy/Environment
R&D  Program  Report

-------
                 RESEARCH REPORTING SERIES

 Research reports of the Office of Research and Development, U S Environmental
 Protection Agency, have been grouped into nine series. These nine broad cate-
 gories were established to facilitate further development and application of en-
 vironmental technology. Elimination of traditional grouping was consciously
 planned to foster technology transfer and a maximum interface in related fields.
 The nine series are

      1   Environmental Health Effects Research
      2   Environmental Protection Technology
      3   Ecological Research
      4.  Environmental Monitoring :
      5   Socioeconomic Environmental Studies
      6   Scientific  and Technical Assessment Reports (STAR)
      7   Interagency Energy-Environment Research and Development
      8   "Special" Reports
      9.  Miscellaneous Reports

 This report has been assigned to the INTERAGENCY ENERGY-ENVIRONMENT
 RESEARCH AND DEVELOPMENT series Reports in this series result from the
 effort funded under the  17-agency Federal Energy/Environment Research and
 Development Program. These studies relate to EPA's mission to protect the public
 health and welfare from adverse effects of pollutants associated with energy sys-
 tems  The goal of the Program is to assure the rapid development of domestic
 energy supplies in an environmentally-compatible manner by providing the nec-
 essary environmental data and control technology Investigations include analy-
 ses of the transport of energy-related pollutants and their health  and ecological
 effects, assessments of, and development of, control technologies for  energy
 systems, and integrated assessments of a wide range of energy-related environ-
 mental issues
                           REVIEW NOTICE

 This report has been reviewed by the participating Federal Agencies, and approved
 for publication. Approval does not signify that the contents necessarily reflect the
 views and policies of the Government, nor does mention of trade names or commercial
 products constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia  22161

-------
                                                   EPA-600/7-78-170

                                                           August 1978
                             Proceedings:
                Symposium on New Concepts
                    for Fine  Particle  Control
                             Teoman Ariman, Compiler

                             University of Notre Dame
                            Notre Dame. Indiana 46556
                     Grants: R805148 (EPA) and ENG. 77-02016 (NSF)
                           Program Element No. EHE624
                                Project Officers:

           Dennis C. Drehmel                          Morris S. Ojalvo

   Industrial Environmental Research Laboratory   and    Solid and Particulate Processing Program
     Office of Energy, Minerals, and Industry               National Science Foundation
      Research Triangle Park, NC 27711                   Washington, DC 20550



                                  Prepared for

U.S. ENVIRONMENTAL PROTECTION AGENCY                 National Science Foundation
   Office of Research and Development                 Solid and Particulate Processing Program
        Washington, DC 20460                           Washington. DC 20550

-------
                            TABLE OF CONTENTS

                                                                       Page

List of Participants ......................         3

Technical Program    ......................         °

Welcome Address - Joseph C. Hogan  ...............        12

Opening Address - Teoman Ariman    ...............        ^

State of the Art and Objectives of the Workshop - T. Ariman  .  ,        ^

ACOUSTICS IN PARTICULATE-GAS SEPARATION

Pulse-Jet Acoustic Dust  ....................        ^3

Acoustic and Turbulent Agglomeration of Sodium Aerosols -
Experimental Results - William C. Hinds and Eugene F. Mallove.  .        35

New Application of Acoustic Agglomeration in Particulate
Emission Control - David T. Shaw and David Wegrzyn .......        "4
Effects of Electric and Acoustic Fields on Particle
Collision Rates in Aerocollodial Suspensions - Paul D. Scholz.  .         86

HIGH GRADIENT MAGNETIC FIELD IN PARTICULATE-GAS SEPARATION

Engineering Aspects of Dry Magnetic Separation - Robin R.  Oder  .        108

Magnetic Separation of Particulate Air Pollutants -
Dennis C. Drehmel and C. H. Gooding ...............
Research Needs and Opportunities in High Gradient Magnetic
Separation of Particulate Gas Systems - Y.A.Liu and C.G.Lin. .  .        170

High Gradient Magnetic Separation:  Review of Single Wire
Models - F. J. Friedlander ...................        183

GENERAL LECTURE

The Influence of Electrostatic Forces for Particle Collection
in Fibrous Filters - Friedrich LBffler .............        206

ELECTROSTATIC FILTRATION

Electrofluidized Bed for Industrial-Scale Air Pollution
Control - James R. Melcher ...................        237

Approximate Equations for Predicting Electrostatic Particle
Collection - Douglas W. Cooper .................        260

-------
                                                                       Page


Electrostatic Filtration and the Apitron - Design and Field
Performance - Dennis J. Helfritch and Teoman Ariman ........     286

Electrostatic Capture of Fine Particles in Fiber Beds -
Donald L. Reid  ..........................     305

ALL OTHER NEW METHODS

Fine Particulate and Gaseous Emissions Control Experience with
the TRW Charged Droplet Scrubber - R. R. Koppang ..........     320

Research on New Equipment for Cotton Dust Collection -
Albert Baril,Jr. , Devron P. Thibodeaux and Robert B. Reif .....
PANEL DISCUSSION

Remarks  - Dennis C. Drehmel

Remarks  - Morris S. Ojalvo

Closing  Remarks - Teoman Ariman
Dielectrophoretic Air Filtration - Progress and Problems -
J. K. Thompson, R. C. Clark and G. H. Fielding ...........     36l

Fine Particulate Control Using Foam Scrubbing -
Tom E. Ctvrtnicek, H. H. S. Yu, C. M. Moscowitz and
Geddes H. Ramsey ..........................     373

The Anomalous Behavior of Asbestos Fibers - James W. Gentry ....     399

GENERAL LECTURE

High Temperature Filtration-Technical Prospects and as Means
for  Check on Theories - Michel M. Benarie  .............    Ul3

-------
                                 PARTICIPANTS
Ronald B. Adams
Division of Industrial Energy Conservation
ERDA
20 Massachusetts Ave., N.W.
Washington, DC  20545

Ali Aktay
Inland Steel Company
Gary, IN

Jeffrey C. Alexander
88 Beacon St., Apt. 13
Somerville, MA  02143

Teoman Ariman
Department of Aeospace and Mechanical
   Engineering
University of Notre Dame
Notre Dame, IN  46556

Albert Baril, Jr.
Cotton Textile Processing Laboratory
Southern Regional Research Center, ARS
1100 Robt. E. Lee Blvd.
New Orleans, LA  70179

Michael Benarie
Institut National de Recherche
   Chimique Appliquee
16, Rue Jules Cesar 75012
Paris, France

Charles E. Billings
740 Boylston Street
Chestnut Hill, MA  02167

Warren L. Buck
Components Technology Division
Argonne National Laboratory
9700 S. Cass Avenue
Argonne, IL  60439

Kurt Carlsson
AB Svenska  Flaklfabriken
Fack
S-35187 Vaxjo,Sweden
William Charles
Magnetic Engineering Associates
247 Third Street
Cambridge, MA  02142


Douglas W. Cooper
Dept. of Environmental Health Sciences
Harvard University
665 Huntington Avenue
Boston, MA  02115

T. W. Ctvrtnicek
Monsanto Research Laboratory
Dayton Division
1515 Nicholas Road
Dayton, OH  45407

H. P. Dibbs
Air Pollution Control Directorate
Environment Canada
Ottawa, Ont.
Canada K1A 1C8

Dennis C. Drehmel
Industrial Environmental Research
   Laboratory
Environmental Protection Agency
Research Triangle Park, NC  27711

George H. Fielding
Combustion and Fuels Branch
Naval Research Laboratory
Washington, DC  20375

Edward R. Frederick
Technical Operations Manager
Air Pollution Control Association
P.O. Box  2861
Pittsburgh, PA  15230

F. J. Friedlander
Professor
School of Electrical Engineering
Purdue University
West Lafayette, IN  47907

-------
 James B.  Galeski
 Senior Chemical Engineer
 Midwest Research Institute
 425 Volker Blvd.
 Kansas City,  MO  64132

 James W.  Gentry
 Department of Chemical Engineering
 University of Maryland
 College Park, MD  20742

 James A.  Gieseke ft
 Aerosol and Emissions  Technology  Group
 Battelle  - Columbus
 505 King  Avenue
 Columbus,  OH   43201

 Charles H.  Gooding
 Research  Triangle Institute
 Research  Triangle Park,  NC  27709

 Eugene E.  Grassel
 Donaldson Co.,  Inc.
 P.O.  Box  1299
 Minneapolis,  MN  55440

 R.  Duncan Hay
 Magnetic  Engineering Associates
 247 Third  St.
 Cambridge,  MA  02142

 Dennis J.  Helfritch
 APITRON, American Precision  Industries
 1 Executive Park,  Suite  500
 Charlotte,  NC   28287

 William C.  Hinds
 Department  of Environmental Health Sciences
 Harvard University
 665 Huntington Avenue
 Boston, MA  02115

 David  R. Kelland
 Massachusetts Institute of Technology
 NW14-3113
 Cambridge, MA  02139

 Richard R. Koppang
 Environmental Products Development
TRW Inc., Energy Systems Group
1 Space Park
Redondo Beach, CA  90278
 Tuncer M. Kuzay
 Reactor Analysis and Safety Division
 Argonne National Laboratory
 9700 S. Cass Avenue
 Argonne, IL  60439

 George E. Lamb
 Textile Research Institute
 P.O. Box 625
 Princeton, NJ  08540

 William F. Lawson
 Morgantown Energy Research Center -
    USERDA
 P.O. Box 880
 Morgantown,  WV  26505

 C. J.  Lin
 Department of Chemical Engineering
 Auburn University
 Auburn, AL  36830

 Benjamin Y.  H.  Liu
 Particle Technology Laboratory
 Department of Mechanical Engineering
 University of Minnesota
 Minneapolis,  MN  55455

 Y. A.  Liu
 Department of Chemical Engineering
 Auburn University
 Auburn,  AL  36830

 Friedrich Loffler
 Institut fur  Mechanische Verfahrenstechnik
 Universitat Karlsruhe
 7500 Karlsruhe  1, Germany

 E.  Maxwell
 Francis  Bitter  National  Magnet Laboratory
 Massachusetts Institute  of Technology
 170 Albany St.
 Cambridge, MA  02139

 Andrew  R. McFarland
 Department of Civil Engineering
 Texas A&M University
 College Station, TX  77843

 James R. Melcher
 Department of Electrical Engineering
   and Computer Science
 Rm 36-13
Massachusetts Institute of Technology
 Cambridge, MA  02139

-------
Grady B. Nichols
Environmental Engineering Division
Southern Research Institute
2000 Ninth Ave. South
Birmingham, AL  35205

John A. Oberteuffer
Sala Magnetics, Inc.
247 Third St.
Cambridge, MA  02142

Morris S. Ojalvo
Solid and Particulate Processing Program
Engineering Division
National Science Foundation
Washington, DC  20550

Robin R. Oder
Research and Engineering
Bechtel Corporation
P.O. Box 3965
San Francisco, CA  94119

Gaylord W. Penney
Department of Electrical Engineering
Carnegie-Mellon University
Schenley Park
Pittsburgh, PA  15213

Leon Petrakis
Gulf Research and Development Company
P.O. Drawer 2038
Pittsburgh, PA  15230

Michael J. Pilat
Air Resources Engineering
Department of Civil Engineering
University of Washington
Seattle, WA  98195

Donald L. Reid
Battelle-Northwest
P.O. Box 999
Richland, WA  99352

Robert R. Reif
Engineering Physics and Electronics Section
Battelle-Columbus
505 King Avenue
Columbus, OH  43201
Paul D. Scholz
Division of Energy Engineering
College of Engineering
The University of Iowa
Iowa City, IA  52242

David A. Schulz
U.S. EPA
230 S. Dearborn
Chicago, IL  60604

David S. Scott
Department of Mechanical Engineering
University of Toronto
Toronto, Ont.
Canada  M5S 1A4

Michael Shackletown
Acurex Aerotherm
485 Clyde Avenue
Mountain View, CA  94040

David T. Shaw
State University of New York at Buffalo
4232 Ridge Lea Rd.
Buffalo, NY  14226

Norman Surprenant
GCA/Technology Division
Burlington Road
Bedford, MA  01730

William Swift
Department of Chemical Engineering
Argonne National Laboratory
9700 S. Cass Avenue
Argonne, IL  60439

James Wegrzyn
State University of New York at Buffalo
4232 Ridge Lea Road
Buffalo, NY  14226

K. T. Yang
Department of Aerospace and Mechanical
   Engineering
University of Notre Dame
Notre Dame, IN  46556

Hsu-Chi Yeh
Inhalation Toxicology Research Institute
Lovelace Foundation for Medical Education
   and Research
P.O. Box 5890
Albuquerque, NM  87115

-------
               WORKSHOP PROGRAM

          WEDNESDAY, APRIL 20,  1977
8:00 am     Registration

8:30 am     Welcome
            JOSEPH C. HOGAN
            Dean, College of Engineering
            University of Notre Dame
            Notre Dame, Indiana

8:40 am     Opening Remarks
            TEOMAN ARIMAN
            University of Notre Dame
            Notre Dame, Indiana

9:00 am     Acoustics in Particulate-Gas Separation
            Chairman:
            K. T. YANG
            University of Notre Dame
            Notre Dame, Indiana

            "Pulse-Jet Acoustic Dust Conditioning and
            Some New Potential Applications"

            DAVID S. SCOTT
            University of Toronto
            Toronto, Canada

            "Acoustic and Turbulent Agglomeration of
            Sodium Aerosols - Experimental Results"

            WILLIAM C. HINDS  and EUGENE F. MALLOVE
            Harvard University
            Cambridge, Massachusetts

10:30 am    Coffee Break

10:45 am    "New Applications of Acoustic Agglomeration
            in Particulate Emission Control"
            DAVID T. SHAW and JAMES WEGRZYN
            State University of New York at Buffalo
            Buffalo, New York

            "Effects of Electric and Acoustic Fields on
            Particle Collission Rate's in Aerocollodial
            Suspensions"
            PAUL D. SCHOLZ
            The University of Iowa
            Iowa City, Iowa

-------
12:15 pm    Lunch

1:30 pm     High Gradient Magnetic Field in Particulate-
            Gas Separation

            Chairman:
            Benjamin Y.  H. Liu
            University of Minnesota
            Minneapolis, Minnesota

            "Engineering Aspects of Dry Magnetic Separation"

            ROBIN R. ODER
            Research and Engineering
            Bechtel Corporation
            San Francisco, California

            "Magnetic Separation of Particulate Air Pollutants"

            DENNIS C.  DREHMEL
            Environmental Protection Agency
            and
            C. H, GOODING
            Research Triangle Institute
            Research Triangle Park, North Carolina

3:00 pm     Refreshments Break

3:15 pm     "Research Needs and Opportunities in High Gradient
            Magnetic Separation of Particulate Gas Systems"
            Y. A. LIU and C. G. LIN
            Auburn University
            Auburn, Alabama

            "High Gradient Magnetic Separation:  Review of
            Single Wire Models:
            F. J. FRIEDLANDER
            Purdue University
            West Lafayette, Indiana
           THURSDAY, APRIL 21, 1977
8:00 am     General Lecture
            Chairman:
            MORRIS S,  OJALVO
            National Science Foundation
            Washington, D. C.

-------
            "The Influence of Electrostatic Forces for
            Particle Collection in Fibrous Filters"

            FRIEDRICH LOFFLER
            University of Karlsruhe
            Karlsruhe, W. Germany

9:00 am     Electrostatic Filtration

            "Electrofluidized Beds for Industrial-Scale
            Air Pollution Control"

            JAMES R. MELCHER
            Massachusetts Institute of Technology
            Cambridge, Massachusetts

            "Approximate Equations for Predicting
            Electrostatic Particle Collection:

            DOUGLAS W, COOPER
            Harvard University
            Cambridge, Massachusetts

10:30 am    Coffee Break

10:45 am    "Electrostatic Filtration and the Apitron-Design
            and Field Performance"

            DENNIS J. HELFRITCH
            Apitron
            Buffalo, New York
            and

            TEOMAN ARIMAN
            University of Notre Dame
            Notre Dame, Indiana

            "Electrostatic Capture of Fine Particles in
            Fiber Beds"
            DONALD L. REID
            Battelle Northwest
            Richland, Washington

12:15 pm    Lunch

1:30 pm     All Other New Methods

            Chairman
            DENNIS C. DREHMEL
            Environmental Protection Agency
            Research Triangle Park, North Carolina

-------
            "Fine Particulate and  Gaseous  Emissions
            Control Experience with the  TRW Charged
            Droplet Scrubber"

            R.  R. KOPPANG
            TRW,  Inc.
            Redondo Beach,  California

            "Research  on New Equipment for Cotton Dust
            Collection"

            ALBERT BARIL, JR

            DEVRON P.  THIBODEAUX
            Southern Regional Research Center
            Department of Agriculture
            New Orleans, Louisiana

            and

            ROBERT B.  REIF
            Battelle Columbus Laboratories
            Columbus,  Ohio

3:00 pm     Refreshments Break

3:15 pm     "Dielectrophoretic Air Filtration - Progress and
            Problems"
            GEORGE H.  FIELDING
            Chemistry  Division
            Naval Research  Laboratory
            Washington,  D.  C.

            "Fine Particulate Control Using Foam Scrubbing"

            TOM E. CTVRTNICEK
            Monsanto Research Corporation
            Dayton, Ohio

            "The Anomalous  Behavior of Asbestos Fibers"

            JAMES W. GENTRY
            University of Maryland
            College Park, Maryland

6:00 pm     Social Hour

7:00 pm     Banquet

            Speaker:
            TOM PAGNA
            South Bend,  Indiana

-------
10

                              FRIDAY,  APRIL  22,  1977

                  8:30 am     General  Lecture
                              Chairman:
                              ANDREW R. McFARLAND
                              Texas  A&M University
                              College  Station, Texas
                              "High Temperature Filtration - Technical Prospects
                              and as Means for Check on Theories"
                              MICHEL M. BENARIE
                              National Institute for Applied Chemistry
                              Paris, France
                  9:30 am      Panel Discussion
                              "New Directions in Particulate-Gas Separation
                              Research"
                              Moderator:
                              BENJAMIN Y. H. LIU
                              University of Minnesota
                              Minneapolis, Minnesota
                              Panel Members:
                              TEOMAN ARIMAN
                              MICHAEL M. BENARIE
                              DENNIS C. DREHMEL
                              ROBIN R. ODER
                              DAVID S. SCOTT
                  10:30 am     Coffee Break
                  10:45 am     Continuation of the Panel Discussion
                  12:15 pro     Closing Remarks
                              TEOMAN ARIMAN
                  12:30 pm     Lunch
                  1:45-        Guided Tour of the Campus
                  2:30 pm

-------
                                                                              11
                             ACKNOWLEDGEMENT
     The editor of this proceedings  gratefully acknowledges at  every phase of


the Workshop, the interest and continuous support and advise  of Dr.  Morris  S.


Ojalvo, Director of the Particulate and Multiphase Processes  Program of the


National Science Foundation and Dr.  Dennis C. Drehmel of the  Industrial


Environmental Research Laboratory of the Environmental Protection Agency.


Special thanks are due to Mrs. Shirley Wills for her enormous efforts in the


transcription of the taped sessions of the workshop and her expert typing


of most of the proceedings.  I would also like to acknowledge the diligent


work by two of our students at Notre Dame, Mr, Greg Muleski and Mr.  Carl Aumen


in the preparation of the open discussions and panel session   and in editing


of the proceedings.

-------
12
                                 WELCOME ADDRESS


                                 Joseph C.  Hogan
                           Dean,  College of Engineering
                             University of  Notre Dame
                            Notre Dame,  Indiana  46556
         Let me welcome  you  to  the University  of Notre Dame.   I'll not  go  through
     the  Chamber of  Commerce  sales pitch because engineering  schools are known by
     the  faculty they  have  and I hope you are familiar with members of our  faculty.
     You  are more  than welcome to go over to the Engineering  Building, but  you
     will have  to  tread through  a good deal of  mud  to get  there.  This is due to
     the  eight  million dollar addition which is just being started and should be
     completed  by  December  of next year.  If any of you would like to see that
     particular construction, I'm sure K. T. or Teo will be glad  to make arrange-
     ments  for  this  or anything  else you'd care to  see while  you're here.

         We are honored  to have such a distinguished group here.  It is farsighted
     for  NSF and EPA to act as co-sponsors.  Often a field moves  forward in tiny
     increments because the papers and work are presented  at  conferences at random
     times.  I  think that getting the top people in the field together at one
     time is without question the best way to move  the field  forward.  I want to
     congratulate  NSF  and EPA again for their farsightedness  and  for sponsoring
     this program.   I  hope  it will prove to be  an interesting and productive two
     and  a  half days.  Thank  you.
                                OPENING ADDRESS


                                 Teoman Ariman
                               Workshop Director
               Department of Aerospace and Mechanical Engineering
                            University of Notre Dame
                           Notre Dame, Indiana  46556
         Welcome to the Workshop entitled "Novel Concepts, Methods and Advanced
    Technology in Particulate-Gas Separation."  As the Dean said, the workshop
    is sponsored by NSF and EPA and we would like to express our appreciation,
    especially to Dr. Morris Ojalvo of NSF and Dr. Dennis Drehmel of EPA, for

-------
                                                                             13

help with the workshop and continuous support.  We would also like to thank
the members of the planning committee, Professor Benjamin Liu of the
University of Minnesota and Professor K. T. Yang, Chairman of the Aerospace
and Mechanical Engineering Department here at Notre Dame.  We sincerely hope
that everything went well last night and that the transportation and hotel
accommodations are satisfactory.  If there are any complaints, please see
Dr. Yang or Dean Hogan.  If there are any compliments, please see me.

     I would like to briefly introduce the workshop.  As you know, a
number of emission control systems have been used for years—fabric filters,
electrostatic scrubbers, electrostatic precipitators, etc.  In years past,
energy has been quite cheap.  But times have changed and we face an energy
crisis unless we do something.  Furthermore, we face the challenge of
controlling fine particulate emissions.  Taken together, these two demand
that we search out new technologies, and in the past few years, there has been
substantial growth in this area.  The thrust of this workshop is the bring-
ing together of people and their technologies.  We wish to discuss not only
collection efficiency but also energy efficiency.  With the close work of the
planning committee and the NSF and EPA, it was decided to invite three or
four people in each area - a theoretical analyst, an experimentalist and a
design/applications specialist.  By mixing people from universities, govern-
ment laboratories, private laboratories, and industry, we hope to have a
healthy exchange of ideas.  I sincerely hope that the workshop reaches that
goal.

-------
14
                  STATE OF THE ART AND OBJECTIVES  OF THE WORKSHOP

                                  Teoman Ariman
                             Director  of the Workshop
                Department of Aerospace and Mechanical Engineering
                             University of Notre Dame
                            Notre  Dame, Indiana 46556
         Traditionally,  developments  in  particulate control equipment have been
     slow in  starting,  as evidenced by the  fact  that today's basic equipment
     differs  little  from  that  in use 100  years ago.  One major reason for the
     lack of  progress is  that  successful  innovation requires field trials, where
     the  cost of  failure  to  the trial  plant is large.  Therefore, the risk must
     be small, and innovative  can be only a small departure from the traditional
     design.  Moreover, most industrial applications of air pollution control
     equipment are considered  as non-productive  economic burdens on the processes
     to which they are  applied.  As a  result, control devices so far has received
     minimum  attention, and  incentive  for research and development to improve
     their performance  is lacking.

         Fabric  filtration  systems have  been employed in industry for over a
     century  with relatively few technological modifications [1,2].  Indeed,
     there have been in recent years only a few  advances of note:  introduction
     of fabrics capable of withstanding higher temperatures and development of
     pneumatic cleaning techniques.  For  the most part, there was in the past
     little incentive for improvement.  The collection efficiency of the systems
     was  mostly satisfactory and the units  were  inexpensive to purchase.  However,
     with the recent substantial increase in energy costs attention should now
     be given to  operating expenses (e.g.,  energy consumption).  As a result, the
     filtration systems of yesteryear  may not be the best approach for future
     applications.

         Recently a new  method has been  developed for the particulate-gas
     separation through the  fabric filtration.   Some initial analytical and/or
     experimental investigations have  been  performed (e.g., Zebel [3,4], Butterworth
     [5], Ariman  [6-9]  and Penney [10]) for the  particulate-gas separation by fiber
     and  fabric filters in an electrostatic field.  It has been observed that
     the  application of an external electrostatic field improves the efficiency
     of fine  particle collection considerably.  More recently Ariman has begun
     to investigate  the pressure drop  across a flat fabric filter due to the
     collection of electrically charged dust particles.  There was no electro-
     static field placed  across the filter.  Two industrial dust samples, fly

-------
                                                                             15

ash and limestone, and Nomex 14 oz. fabric filter samples were utilized in
the bench scale tests.  It was found that under the same operating conditions
the pressure drop across the dust layer and filter was reduced by 1/2 to 2/3
as compared to the conventional noncharged dust particles case.  An electron-
microscopic  study of the dust cakes for charged and noncharged cases
indicate that charged dust particles deposit in a relatively porous layer
resulting in a low pressure drop across the filter.  The resulting more
porous dust layer also facilitates cleaning of the filter, since it bends
to fall off readily, leaving the cloth relatively clean.  It was also found
that there was considerable improvement in the collection efficiency.

     It appears that this new method has a great potential in particulate-gas
separation by fabric filters and may bring enormous changes in the con-
ventional filtration area.  However, the research performed to date in this
field has only just scratched the surface and a number of unknown aspects
of this highly promising area need immediate attention of researchers
[12-34].  For example the electrical behavior of aerosols, electrostatic
charge on fiber and fabric filters, particulate cake structure, collection
efficiency and pressure drop and a number of other unknown aspects of the
field need to be investigated.

     Another new method in the particulate-gas separation has recently been
initiated by Drehmel and Gooding [35].  The method is directed toward a
combination of magnetic and filtration collection mechanisms.  In its
simplest practical form, the high gradient magnetic separator  (HGMS) consists
of a canister packed with the fibers of a ferromagnetic material (such as
stainless steel wool) and magnetized by a strong external magnetic field [35].
The strong magnetic forces produced near the edges of the fibers are cabable
of trapping fine, weakly magnetic particles with high efficiency.

     The packing density of the separation zone can be adjusted to obtain
a proper balance between conventional filtration phenomena and the HGMS
effect.  Reducing the packing density lowers the pressure drop through the
separator but requires a higher magnetic field for efficient particle
collection.  The overall particle collection efficiency is theoretically a
function of the applied magnetic field, or filter mesh parameters (fiber
diameter and magnetization, packing density, and length of mesh in direction
of flow), of particle parameters (magnetic susceptibility and diameter) and
of fluid parameters (superficial velocity and viscosity).

     To date the high gradient magnetic separation has been applied extensively
to particle liquid separations  [36-39].  The bench-scale experimental work
performed by Drehmel and Gooding [35] appears to be the first known attempt
to apply HGMS to particle-gas separations.  It is known that conventional
magnetic separation is generally useful for the separation of relatively
large particles of strongly magnetic materials such as iron and magnetite.
HGMS, by contrast, makes possible  the efficient separation of micron-sized
particles of weakly paramagnetic materials from liquids at high process
rates.  It can be shown through generalized theory that application can also
be made to particle removal from gas streams.  Magnetic forces can be cost

-------
16
     effective if the control device is properly designed.   One of the most
     important feature of the recent work by Drehmel and Gooding [35]  is the
     dramatic improvement of collection efficiency when a low-strength magnetic
     field was applied.  However,  at present substantial analytical and experi-
     mental research efforts are needed to identify other particle-gas
     separation problems to which high gradient magnetic separation might be
     applicable, to develop analytical models based on the experimental investi-
     gations and to provide economic estimates for specific promising  applications
     such as particulate emission control from basic oxygen surfaces.

          Another recently developed new method in particulate-gas separation
     is based on acoustic coagulation.  Recently there has been a substantial
     growth in research dealing with various aspects of acoustic-aerosol
     interactions.  In the last few years considerable attention has been devoted
     to the area of acoustic propagation and attenuation in aerosols [45].
     Theories have been developed  for both small amplitude [41-42]  and large
     amplitude [43-44]  sound waves, including the processes of mass, momentum
     and energy transfer between droplets and the surrounding gas-vapor mixture.
     It is well known that acoustic streaming can substantially increase the
     rates of convective transport of mass and energy and chemical reactions at
     solid-fluid and fluid-fluid interfaces, provided the sound pressure level is
     sufficiently high [45].  The  aforementioned investigations have indicated
     [46]  that the acoustic streaming mechanism may be of importance in droplet
     evaporation, particle drying, acoustic coagulation,  and droplet or particle
     combustion, especially in those cases where the droplets or particles are
     large enough so that the acoustic boundary layer thickness is much smaller
     than the droplet radius.   In  that case very significant increases in heat
     and mass transfer are possible [47].

          Initial experimental investigations were directed to the use of shock
     waves to coagulate fine particles [48-50].   They also  included the use of
     pulsating combustion burners  for providing the shock or more correctly
     saw-tooth waves,  i.e.,  sound  waves with periodic discontinuities  for
     agglomeration [51].   The pulse output approximating  a  saw-tooth wave with
     a  step wave front  has successfully agglomerated  ZnOo  aerosols to bring
     about a five-fold  increase in mean particle size [51j.   Since no  special
     agglomeration chamber was used,  the experiment was equivalent to  placing the
     pulse-jet in a suitable elbow of the duct transporting the dusty  gas in an
     industrial plant.   Recently Scott and Rennick have experimentally investigated
     several aspects of acoustic-aerosol interactions in  finite-amplitude,
     mutually perpendicular, standing and progressive acoustic fields  [52-55].
     Acoustically generated turbulence and the role it may  play in aerosol
     agglomeration has  been more recently investigated [55-56].

          Another recent experimental program has been initiated by Shaw [57]  and
     aims  to provide information on the experimental investigation of  acoustic
     coagulation in the flowing coagulation chamber, the  evaluation of sinusoidal
     standing wave acoustic system versus  saw-tooth traveling wave  acoustic
     system and on the  search and  testing  of large acoustic generators.

          As indicated  here although in recent years there  has been a  substantial
     growth in research dealing with various aspects of acoustic-aerosol
     interaction, a number of vital investigation areas need to be  identified in
     this  promising field.

-------
                                                                              17
     New concepts and methods in the particulate-gas separation are not
restricted to the stated three major areas.  For example in the recently
developed Charged-Droplet Scrubber (CDS), the scrubbing droplets, usually
water, carry a high electrical charge which is induced during or after
formation [58-59].  The droplets may move under the influence of electric
fields, and the particulate may also be electrically charged.  More
investigation may be needed for the various charged droplet scrubbing
mechanisms toward the control of fine particulate (0.1 - 1.0 vim in diameter).

     In another research effort, an experimental investigation has been
conducted by Scholz [60] on the effects of sonic and electric fields on the
collision rate between suspended submicron sized particles.  Recently a
number of new concepts have been considered for the collection of submicron
particles, specifically cotton dust in the Southern Region Research Center
of the Department of Agriculture [61].  However, these developments are at
the initial stage and a number of unknown aspects of these efforts may need
immediate attention of investigators.

     As indicated above,all three new methods and the other new  techniques
appear in the urgent need of a number of investigations.  In order to identify
new research areas associated with these methods as well as to stimulate
future research activities in these fields, the research workshop on the
novel concepts, methods and advanced technology in particulate-gas
separation was held with the sponsorship of the National Science Foundation
and Environmental Protection Agency at the University of Notre Dame.

     A technical committee consisting of Drs. T. Ariman, B. Y. H. Liu and
K. T. Yang, was  in charge of the technical content of  the research workshop.
Drs. Ariman and Yang have been associate professor and  professor and chairman
respectively at  the Department of Aerospace and Mechanical Engineering of the
University of Notre Dame, while Dr. Liu has been a   professor of the
Department of Mechanical Engineering and the director of the Particle
Technology Laboratory at the University of Minnesota.

     Three groups of researchers were invited by the technical committee;
1.  theoretical analysts; 2. experimentalists; 3. design and application
specialists.  The interaction between the three groups  through presentations
and intensive discussions is emphasized particularly, because very often a
researcher in one group can benefit from the other groups.  It was quite
gratifying that practically all participants attended the workshop in its
entirety and took an active role in every aspect of it.

     The workshop took place on April 20, 21 and 22, 1977 at the University
of Notre Dame Campus.  The number of participants was limited to 56 scientists
and engineers from universities, government and other research organizations,
and industry.  Two general lectures by internationally  known scientists, one
by Dr. M. Benarie of France and another by Professor F. Lbffler  of W. Germany,
and 17 invited lectures by the prominent experts were presented.  The time
allowed for general and invited lectures were 45 and 25 minutes  respectively.
Each presentation was immediately followed by an in-depth discussion of 15
minutes for general and 20 minutes for invited lectures.  The latter part

-------
18
     generally began with a designated discusser for each presentation and then
     a general discussion open to every participant was conducted.   Each
     designated discusser made a considerable preparation for a constructive in-
     depth discussion of each presentation and submitted his written discussion
     well before the workshop date.   In order to stimulate the discussions,  the
     copies of the manuscripts were  distributed to the relevant discussers prior
     to the workshop.  On the morning of Friday, April 20, a panel  discussion on
     "New Directions in Particulate-Gas Separation Research" was conducted.   The
     formal presentations by the members of the panel and concise discussions and
     comments by participants were quite productive in identifying  new research
     topics in the particulate-gas separation area.   The entire workshop sessions
     were tape-recorded and the tapes were extensively utilized in  the preparation
     of these proceedings.

-------
                                                                             19

REFERENCES

 [1]  Billings, C. E. and Wilder, J. Handbook of Fabric Filter Technology.
     Volumes I-IV, GCA-TR-70-17-G, prepared for NAPCA  (EPA) (1970).

 [2]  Air Pollution Control. Part 1, Edited by W. Strauss, 337, Pergamon
     Press, New York (1966).

 [3]  Zebel, G..'J. of Colloid Sci.. 20, 522 (1965).

 [4]  Zebel, G., Staub.  (English Translation), 26, 18 (1966).

 [5]  Butterworth, E., Manufacturing Chemist, 65, Feb.,  (1964).

 [6]  Ariman, T., Rao, K. S., Yang, K. T. and Hosbein,  R. L., "Collection
     of Dust by Fabric  Filtration in an Electrostatic  Field,"  Proceedings
     of the Second Annual  Environmental Engineering and Science Conference,
     p. 555  (1973).

 [7]  Ariman, T. and Tang,  L., "Collection of Aerosol Particles in an  Electro-
     static Field,"  invited lecture, presented at the Symposium on Surface
     and Colloid Chemistry in Air Pollution Control, August (1973) and
     appeared  in the Atmospheric Environment, 10, 205  (1976).

 [8]  Ariman, T. and Lane,  M. J., "On Experimental Determination of Collection
     Efficiency and Pressure Drop in Fabric Filters in an Electrostatic
     Field," to be published.

 [9]  Rao,  S. K. and Ariman, T., "Collection of Dust by Fiber Filters  in an
     Electrostatic Field," Survey Article,  to be published.

[10]  Gaylord,  W. Penny, U.S. Patent No.  3, 910,  779, on an Electrostatic
     Dust  Filter, October  9, 1975.

[11]  Liu,  B. Y. H. and  Pui, D. Y. H., J. Aerosol Science. .5, 465  (1974).

[12]  Liu,  B. Y. H. and  Pui, D. Y. H., J. of  Colloid and Interface  Science.
     49  .  305  (1974).

[13]  Liu,  B. Y. H. and  Pui, D. Y. H., "On Unipolar Diffusion Charging of
     Aerosols  in  the Continuum Regime,"  presented at  the  International
     Conference on Colloids and Surfaces, San Juan, Puerto Rico, June, 1976.
     Particle  Technology Laboratory Pub. No. 311  (1976).

[14]  Fuchs,  N. A.,  The Mechanics of Aerosols. Pergamon Press, Oxford (1966).

[15]  Davies,  C. N., Air Filtration. Academic Press, New York  (1973).

[16]   Frederick, E. R.,  Chem Eng..  68,  107  (1961).

-------
20
    [17]    Kraemer, H.  F.,  and  Johnstone,  H.  F.,  Ind.  Engr.  Chem.  47,  2426,
           (1955).                                                 —

    [18]    Dawkins, G.S., Technical  Report No.  15,  Engineering  Experiment  Station,
           University of Illinois  (1958).

    [19]    Natanson, G. L., Proc.  of the Academy  of Science. U.S.S.R., Physical
           Chemistry Section, 112, 95,  (1957).

    [20]    Lundgren, D.A. and K. T.  Whitby, I and E. C. Fund (1965).

    [21]    Yoshioka, N., Emi, H.,  Hattori, M.,  and  Tamori, I.,  Kagaka Kugaku.
           Chem. Ene. Japan. 32, 815 (1968).

    [22]    Rossano, A.  J., Jr.,  and  Silverman,  L.,  AEC NYO-1594.   Air Cleaning
           Laboratory,  Harvard  School of Public Health.

    [23]    Goyer, G. C., Gruen,  R.,  and LaMer,  V. K., J. Phys.  Chem.. 58,  137,(1954)

    [24]    Silverman, L., Conners, E. W.,  Jr.,  and  D. M. Anderson, AEC NYO 4610,
           Air Cleaning Laboratory,  Harvard School  of Public Health  (1956).

    [25]    Rivers, R. D., ASHRAEJI.  k_t 37, (1962).

    [26]    Whitby, K. T., and Liu, B.Y. H., Aerosol Science. Academic Press  (1966).

    [27]    Gillespie, T., J. Colloid, Sci., 10, 299 (1955).

    [28]    Thomas, J. W., and Woodfin, E.  J., Applic. and Industry.  (AIEE) (1959).

    [29]    Sweitzer, D., Electrets,  Literature  Search No.308, Jet  Propulsion
           Laboratory, Cal. Inst. of Tech., (1961).

    [30]   Dennis, R., Kristal, E.,  and Silverman,  L., AEC NYO-4614, Air Cleaning
           Laboratory, Harvard School of Public Health (1958).

    [31]   Anderson, D.  M., and Silverman, L., AEC  NYO-4615, Air Cleaning
          Laboratory, Harvard School of Public Health (1958)

    [32]    Lapple, C. E., Advances in Chem. Engr.. j}, Academic Press.

    [33]   Havlicek, V., Int. J. of Air Water Pollution. 4., 225 (1961).

    [34]    Silverman, L., Billings, C. E., and Dennis, R., AEC NYO-1592, Air
          Cleaning Laboratory, Harvard of Public Health.

    [35]   Drehrael, D. C. and Gooding, C. H., "High Gradient Magnetic Particulate
           Collection,"  presented at the 82nd National Meeting of  the AIChE,
          Atlantic City, 1976.

-------
                                                                              21
[36]   Kolm, H.,  Oberteuffer,  J.  and Kelland,  D.,  Scientific American.  233.
      47 (1975).

[37]   Oberteuffer, J.,  IEEE Transactions on Magnetics.  MAG-10 223 (1974).

[38]   Melville,  D., Paul F. and  Roath,  S.,  IEEE Transactions on Magnetics,
      MAG-11,  1701 (1975).

[39]   Sala, Magnetics,  Inc.,  "Water Reclamation Using SALA-HGMS Magnetic
      Separation Equipment  and Filtration Processes,  Sala Magnetics,  Inc.
      Cambridge, Mass., (1975).

[40]   Scittm D,  S, (Editor),  Acoustic-Aerosol Interactions, Proceedings of  a
      Technical  Discussion  Session held during 82 National meeting of  the
      AIChE, Atlantic,  City,  1976.

[41]   Davidson,  G. A.,  J. of  Atmospheric Sci.. 32, 8201 (1975).

[42]   Cole, J. E., and  Dobbins,  R. A.,  J. of  Atmospheric Sci..  27. 426, (1970).

[43]   Davidson,  G. A.,  and  Scott,  D. S., J. of Acoustical Society of America,
      53,  1717  (1974).

[44]   Davidson,  G. A.,  J. of  Sound and  Vibration. _38_, 475 (1975).

[45]   Folger, H. S., (Editor), Sonochemical Engr., AIChE Chem.  Engr. Progress
      Symposium  Sci.,  67,  (1971).

[46]   Lyman, F.  A., "Aspect of Sound Attentuation in  Aerosols,"  Research
      Progress Report,  Acoustic-Aerosol Interactions. Edited by D. S.  Scott,
      16,  (1976).

[47]   Lyman, F.  A., "Attentuation  of High-Intensity Sound in a  Droplet - Laden
      Gas," Technical Report. MAE-5 192-T1 (1976).

[48]   Mednikov,  E. P.,  "Acoustic Coagulation  and  Precipitation  of Aerosols,"
      Translation from  Russian by  Consultants Bureau, New York, -(1965).

[49]   Gulyaev, A. I. and Kuznetsov, V., "Coagulation  of Aerosols in Response
      to Periodic Shock Waves,"  Translation from  Russian by Consultants
      Bureau, New York  (1965).

[50]   Jahn, R.,  "Improvements Relating  to Sound Generator,"  U.S. Patent #849,
      504,  (1960).

[51]   Last, A. J., "Acoustic  Conditioning - Sound Sources,"  Research  Progress
      Report, Acoustic  - Aerosol Interactions, Edited by, D.  S. Scott, 13,
      (1976).

-------
22

    [52]  Scott, D. S., J. Sound Vib.. 43, 607,  (1975).

    [53]  Scott, D. S., and Rennick, D. F., J. Aerosol Sci., 5_, 301  (1974).

    [54]  Rennick, D. F., and Scott, D. S., Proc. 6th Inter. Symp. on Non-Linear
         Acoustics, Moscow, 174 (1975).

    [55]  Scott, D. S., and Rennick, D. F., Abstracts 7th Inter. Symp. on Non-
         Linear Acoustics, 43,  (1976).

    [56]  Scott, D. S., "Acoustic Dust Conditioning, Acoustically Generated
         Turbulent and F - A Sound Propagation  Through Aerosols,"  Research
         Progress Report, Acoustic-Aerosol Interactions, Edited by D, S.Scott,
         29,  (1976).

    [57]  Shaw, D. T., "Research Activity Summary on Acoustic Coagulation and
         Acoustic Generators," Research Progress Report, Acoustic-Aerosol
         Interactions. Edited by D. S. Scott, 31,  (1976).

    [58]  Lear, C. W., Knieve, W. F., and Cohen, E., J. of APCA. 25,184, (1975).

    [59]  Koppang, R. R., "Pollution Control Equipment and Technology - A New
         Hybrid EP Concept, The TRW Changed Droplet Scrubber," presented at the
         4th  International Pollution Exposition and Congress, Cleveland, (1975).

    [60]  Scholz, P.O., "Experimental Study of Sonic and Electric Field Induced
         Coagulation in Aerocolloidal Suspensions," Research Progress Report,
         Acoustics-Aerosol Interactions, Edited by D. S. Scott 25,  (1976).

    [61]  Deluca, L. B., Weller, H., and Claassen,  B., Personal Communications,
         Oct.,  (1976).

-------
                                                                 23
        PULSE-JET ACOUSTIC DUST CONDITIONING AND SOME
                 NEW POTENTIAL APPLICATIONS
                             by
                       David S. Scott
        Mechanical Engineering, University of Toronto
                       I INTRODUCTION

      Acoustic conditioning of fine-particle aerosols is a
process by which the mean size of the particles is increased
and their number density decreased through exposure to finite-
amplitude acoustic fields.  This change in the particulate
size distribution is important - for it can allow an increase
in the collection efficiency of a downstream dust collector
and/or a reduction in the overall particulate separation cost
because, with few exceptions, collecting devices retain higher
mass fractions as the particulate size is increased.  The
increased coagulation rate which achieves this rapid change in
the size distribution is the result of manyfold increase in the
particle-particle collision frequency — which in turn results
from dynamic acousto-aerosol interactions.  Experimental evidence,
and straightforward physical arguments, indicate that the process
can be particularly effective on high dust load fine particle
aerosols.
      Although the principle of acoustic dust conditioning (ADC)
has been known for sometime and has been shown to be technically
effective, the process has enjoyed little industrial application.
In my view, this lack of industrial application has been the
consequence — primarily — of economic limitations.   The first
of these is high operating costs due to high specific power

-------
24
    requirements, and the second  is high capital costs.
          Studies performed  at  the  Ontario  Research  Foundation  and
    the  University  of Toronto have  examinedanew approach  to ADC
    — in which  the  following two  complimentary principles were
    explored.
          (i)    Firstly,  the conventional standing sinusoidal
    acoustic field,  used  in  essentially all earlier  acoustic dust
    conditioning work,  was dropped  in  favour of progressive saw-tooth
    fields.   This change  in  acoustic configuration allowed major
    simplifications  in  the agglomeration chamber yielding both
    capital  and operating cost  savings.
          (ii)   Secondly, the normal techniques of sound  generation
    — such as  sirens and  mechanically  activated pistons — were
    discarded  in favour of a resonant  pulse-jet acoustic  source.
    This not only allowed a  direct  conversion of fuel  to  sound  and
    the  use  of  a simple no-moving-part device, but also readily
    produced high intensity  progressive saw-tooth acoustic fields.
          The  results of  this research/development program are
    thoroughly  reviewed in Reference 1.  In summary, the  practical
    results  of  this  program  showed  that significant  —  0(10) — advances
    can  likely  be achieved by this  new process in all  three areas of
    (a)  capital, (b) maintenance, and  (c) operating  costs.  At  the
    same time,  the  mean particle  size  was increased  5-7 fold.   But
    as noted,  the details of this research  are well  covered in
    Reference  1 and it  would be inappropriate to review these here.

                II   THE NEXT STAGE  OF  ADC DEVELOPMENT
          Since the work  described  in  Reference 1 was  performed,
    the  acoustic-aerosol  research with which I have  been  involved
    has  primarily concentrated  on more fundamental acousto-aerosol
    phenomena.   Insofar as the  applied aspects of pulse-jet acoustic
    dust conditioning we  have,  as they say  in the performing arts,
    been between shows.  And this gives us  time to pause  and take
    note of  where we are. To date, it seems to me that research
    on pulse-jet dust  conditioning  has indicated economic and

-------
                                                                25
technical feasibility in principle.   The next stage in the
systematic development of these concepts must examine promising
conditioner/collector systems in the context of likely applica-
tions.  It is this next stage of the development, to which I
address the remainder of this paper.
      But by way of philosophical introduction, I preface my
remarks by saying that it is this stage which is, perhaps, the
most critical.  The history of acoustic agglomeration has been
marred by overzealous researchers who have felt the process to
be a likely panacea for all fine particle collection requirements,
as well as by ambitious entrepreneurs anxious for solution to
specific problems on short-time-scale development programs.  My
own view is that ADC will never be a panacea-type solution to
fine particle removal from industrial sources.  But if well
conceived development programs are carried out with selected
conditioner/collector systems, and with specific applications in
mind (and over development time scales that can allow careful and
systematic evaluation of the relevant parameters) then it seems
reasonable that in specific technical sceneries, ADC in combination
with an appropriate collector, could be more effective than any
alternate approach.  Moreover, once we have identified single
application in which this situation exists, the value of careful
development will have been illustrated.  And by that impetus the
process development can be expanded to further applications.
      But care — and perhaps some luck — is required in the choice
of both the conditioner/collector configurations as well as the
problem to which they will be applied.  And those funding this
development must not be mislead into thinking that conclusive
results can be obtained in a matter of months.
      With this background, I will briefly outline two development
projects with which I am associated, and which have been initiated
within the past few weeks.  Clearly, I can only describe the work
we hope to do rather than present any results.

-------
26
    2.1  ADC/Wet Scrubbers in the Pulp and Paper Industry
          In Reference 1,  wet scrubbers are identified as perhaps
    the most promising of collector systems for use in combination
    with ADC.  To illustrate by example; in the case of a particular
    venturi,scrubber,  3,150 bhp was required to collect 90% of the
    0.2 ym particles while only 60 bhp was required to collect 90%
    of  the 0.5 ym particles.  The implication for ADC is that by
    affecting a 2.5 fold increase in mean particle size (we have
    previously noted 5~7 fold increases) a 50 fold reduction in
    scrubber power requirements would be realized.  While this
    illustration refers to a particular venturi scrubber, all
    scrubbers exhibit  sharp transitions in collection efficiency
    as  a function of particulate size within the size range
    0.1-1.0 ym.
          Of course wet scrubbing can also be used to simultaneously
    remove noisome (noxious/odorous) gases.  And there is some
    evidence that the  presence of a finite-amplitude acoustic field
    can also improve noisome gas scrubbing effectiveness.  In this
    circumstance, the  acoustic field might be considered to act both
    as  a conditioner for particulate collection and as a catalyst
    for noisome gas removal.  Although the latter potential benefit
    is  more speculative, it could be important in applications within
    the Pulp and Paper Industry and so has been included in our
    program.
          A possible deleterious effect on scrubber performance
    would occur if the sound were to rapidly agglomerate the liquid
    droplets in the scrubber itself, thereby markedly reducing the
    surface volume ratio of the scavenging liquids.  Whereas it is
    possible to visualize  a situation in which this occurs — it is
    also apparent that configurations can be designed to minimize
    this effect if it  should prove important, as proper design un-
    doubtably involves some balance between maximizing the droplet
    "swept" volume and minimizing droplet agglomeration.

-------
                                                                27

The research program will proceed as follows.
      (1)  Visits will be made to various pulp and paper mills
for the purpose of familiarization with the practice and
potential of wet scrubber gas-cleaning.  This study will
include an appraisal of (a) existing scrubber installations,
(b) potential new installations both in new mills and as
replacements for existing non-scrubber equipment, (c) potential
for pulse-jet conditioning of existing scrubbers for the
purpose of upgraded performance and/or reduced annual costs,
(d) identification of the mill environment in gas-cleaning
locations for the purpose of evaluating noise isolation
requirements .
      (2)  With the benefit of (1), the experimental program
will be finalized.  In particular, key parameters requiring
examination will be identified and a desired data point grid
constructed.
      (3)  A parametric evaluation of the effectiveness of
pulse-jet acoustic conditioning on particulate collection by a
medium energy venturi scrubber will be performed.  Flow rates
between 1000-3000 cfm will be studied.  A ZnO fume will be
generated on-line and used as the test aerosol.  By this means,
particulate size distributions and densities can be readily
achieved which properly model typical pulp and paper aerosol
effluents.
      (4)  The influence of intense acoustic fields on gas
scrubbing performance in the above facility and at flow rates
of 1000-3000 cfm will be studied.  The test noisome gas will
be S02 and the test scrubbing liquid an Na2C03 solution.
      (5)  If practical from an experimental point of view,
the pulse-jet acoustic field will be subdivided into a dirty
gas stream and a clean gas stream in order to obtain data for

-------
28
    extrapolating  to  the  ultimate  flow  rates which may be treated
    by  the  pulse-jet.   (It  rs  expected  that the pulse-jet will be
    able  to condition a much greater  flow  volume than the 3000 cfm
    generated  in this  program).
    ?kOL&
-------
                                                                 29
particulate laden, gases cannot pass through gas turbines
without massive destruction to the turbine blades resulting in
unacceptably short lifetimes.  And so we are faced with a
particulate-gas separation problem at high temperatures and
pressures that is critical not only for environmental consider-
ations but also for the success of the process itself.
      At the temperatures 0(400°K) and pressures 0(10 atm.)
which are anticipated, most conventional particulate removal
techniques are unsuitable.  Mechanical, or inertial, separation
by cyclones seems to be one of the few systems that can, in
principle, operate at these temperatures and pressures.
Unfortunately, the performance characteristics of cyclones are
not sufficient to meet the gas cleanliness standards set by
turbine manufacturers — although there remains some uncertainty
regarding what these standards should be.  It now appears that
the ultimate criteria may require that there be no more than
0.5 mg/m3 of particulate matter above 2 vim diameter.
      Results of the FBC program at ANL have shown that the mass
loading and size distribution of the particulate matter exiting
from the second of two inertial separators is not far from that
which would be acceptable as turbine inlet conditions.  Moreover,
a fairly straightforward evaluation indicates that if the size
distribution of particulate matter to the entering the cyclone
was increased by perhaps less than a factor of 5, the performance
of the cylcone would be improved such that the effluent would
meet turbine inlet criteria.  For these reasons the group at
ANL have decided to initiate a research program to investigate
the effects of using ADC in this application.  And it is in this
regard that I have become associated with these activities.
      The program at ANL will evaluate effectiveness of ADC in
a bench scale fluidized-bed combustor.  It is proposed that the
sound source be a resonant pulse-jet and that progressive saw-
tooth waves constitute the acoustic field configuration.
      There are several promising aspects of this approach.

-------
30
    For  instance the heat of combustion from the pulse-jet simply
    adds to the overall process heat.  Thus the effective power
    requirements will, in principle, be zero - although the cost
    per heating unit for pulse-jet fuel may be expected to be some-
    what higher than that of coal.  Furthermore, the pulse-jet
    will run at higher power levels due to the elevated pressures
    and  should, in principle, be a more effective sound generator
    than if it were running at ambient conditions.
         Conceivably, there might also be deleterious ADC effects.
    For  instance, very fine submicron particles could be agglomer-
    ated into a size range below that which would be effectively
    removed by the cyclone, but within a range which could do damage
    to turbines.  Evaluations of the mass loading ratio in the
    very fine particle spectrum indicate that this should not be a
    problem — but it will clearly be an aspect which shall be monitored
    during the experimental program.
         Turning to the experimental program itself, one of the main
    difficulties in this initial stage of process evaluation shall
    be "scaling".  Put simply, a pulse-jet is a device which cannot
    be scaled up or down without changing the acoustic characteristic
    - for a pulse-jet is a quarter wavelength device.  Ultimately,
    it is expected that single pulse-jets will generate sufficient
    sound to treat between 5,000 to 10,000 cfm.  But the current
    test plant at ANL yields approximately 81 scfm, or on the order
    of 20 acfm.
         In response to these scaling demands, the current intention
    is to design a pulse-jet that will exhaust into a resonant-manifold,
    The manifold shall be capable of splitting the off-gases into a
    "waste" stream and two "process streams".  Two process streams
    are required in order to examine the affects of ADC on both the
    primary and secondary cyclone performance.  By means of the
    resonant-manifold a large total acoustic power output (Watts)
    can be reduced in total magnitidue to match the amount of flue
    gas being treated.  At the same time, this approach will allow the

-------
                                                                 31
acoustic field to be varied in intensity from approximately
165 db downwards, in order to parametrically evaluate sound
intensity requirements.  It is my feeling that by exhausting
the pulse-jet into a resonant-manifold, and then drawing off
the acoustic power as required for the much smaller scale
dust treatment streams, we will both eliminate the scaling
difficulties as well as allow a much greater degree of
parametric flexibility, than would be the case if we exhaust
the pulse-jet directly into the particulate laden gas.  Of
course the negative of this approach is that it requires the
design, construction and debugging of a portable pulse-jet/
resonant-manifold system.
      Although the experimental program I have just described
is still very much in the planning stage, and alternative
programs may, in fact, be followed — the entire concept of
applying ADC to high pressure, high temperature, particulate
removal sytems is, I think, very exciting.

                  Ill  CONCLUDING COMMENTS
      It is my view that ADC has reached the stage where economic
and technical feasibility have been shown to be feasible in
principle.  We must now carefully select certain applications
where ADC may have clear advantages over alternate systems,
and then perform careful unhurried evaluations of its capabilities
in these technical sceneries.  Perhaps one of the two applications
I have reviewed in this paper — and which will be evaluated over
the next 1~2 years - will be the application that gives us the
first real illustration of an ADC/collector system meeting the
required criteria better than any alternative process.  I hope
other investigators, and laboratories, seek other unique applica-
tions where their special requirements might yield an "application
breakthrough".

-------
32

                                   REFERENCE
    1.   Scott, D.S., "A New Approach to  the Acoustic Conditioning
         of  Industrial Aerosol  Emissions",  J.  Sound  and  Vj-biation,
         Vol.  43,  no. 4,  pps.  607-619, 1975.
   OPEN  DISCUSSION

   Billings:   I think there is a misconception in accordance with what we might
   expect.  We would all agree that agglomeration of some  sort, whether mechanical,
   thermal, or turbulent,is desirable to improve the mechanical collector.  I
   think zinc  oxide, as an aerosol, is a questionable aerosol  to use.

         What you said was that you went down in concentration and up in apparent
   particle size.  Correspondingly, the aerodynamic  diameter parameter would go
   down.  Although this may be 2.0 to 7.0 microns, the fact of the matter is that
   the momentum relaxation time is about 0.1 of the  original value.  This is what
   you should  be measuring.

   Scott: I agree with what you are saying.   One thing that was not brought out,
   and for obvious reasons, is the kinematic viscosity of  the gas.  Aerodynamic
   diameter is a characteristic parameter of the aerosol and the momentum
   relaxation  time involves both the characteristics of the gas and the particles.

         I would like to clarify that all the sizes talked about are aerodynamic
   sizes and are a measure of the momentum relaxation time or of aerodynamic
   diameter.  Results of size distribution were obtained from a cascade impactor.
   The means of measuring size is a means of predicting improved performance
   since collectors,particularly cyclone collectors,  are inertia collectors.

   Melcher: What are the resonance times requires for gas treatment to go from
   0.5 to 2.0 microns?

   Scott: Approximately between 1.5 and 2.5  seconds.   Resonance times must be
   a factor of concentration.   Acoustic vibration constants are inversely pro-
   portional to the concentration.  So you should put the stuff in front of a
   conventional device where the concentration of particles is high.  This is a
   necessary requirement for acoustic coagulation.

   Scott: Is it possible to amend the resonance times by using an adverse density
   gradient?

   Shaw:  If you have an amount of the fine  particles,your acoustic agglomeration
   is small so the count is large.  With a large standard deviation you need
   large particles to act as collectors for  the small particles.  It is a
   complicated function of standard deviation,  mass  concentration and frequency.

-------
                                                                              33

That is why I am skeptical about the recent report by the EPA.  It is based
on the wrong equation, wrong data and, of course, has the wrong conclusion.

       Another point I'd like to bring up is that because of economic
comparison on acoustic agglomerators with conventional devices, comparisons
can only be made for specific applications.   A general statement is too
dangerous.  It can be too optimistic or too pessimistic.

       Something else worthy of note is on the acoustic device, for instance,
sirens; the last reference I could find was published in 1949.  Bell
Laboratories were to build a siren to put on the Empire State Building for
air raid warnings.  Acoustic devices must be studied using up-to-date
information on acoustic agglomoration.

       Finally, I have a question.  By using a pulse generator, isn't a
combustion product used which may get into the system?

Scott: To further elaborate on resonance times, no studies were done on
parametric times for the resonance frequency.  We are now studying an oil
mist aerosol at the University of Toronto (UT) and have found that the change
from uncoagulated to fully coagulated particles with no further change is
very rapid.  Well, under 1 second.  The density was above 1 grain/ft3.  So
we  can assume  that resonance times are not a problem.

       From my work at Argonne National Laboratories, we found that by using
a spiral  configuration it would be easier to modify resonance times simply by
changing  the length of the spiral.  Parametric evaluations would be easier
to  conduct.

       From my work through the Ontario Research Foundation, we found that  .
economics depend mainly on the costs  attendant to acoustic isolation.  Total
capital,  operating and maintenance costs ran $.50 to  $1.50 per cubic foot.
So  it  is  economically feasible as compared to current devices, but it is a
cost which must be added on since it  is a conditioning.  It is within  or
under  costs of current devices.

       Concerning the pulse jet,  the  products of combustion did go into the
gas stream.  This represented less  than 1% of the total.  In  doing preliminary
design work at Argonne on resonant manifolds, I  think it will be possible  to
drop  the  products of  combustion which get into the gas stream down to zero.
A resonant manifold  system can be used in almost any  application.
                                                      •j
Koppang:  What is the actual energy  input in watts/ftj?

Scott: I  don't recall right off hand,  but it  is  in the paper.

Baril: Because there  are maximum  noise levels  in the  U.S., we must be concerned
 that  the  noise from  the  system  isn't  too loud.   Is this  a problem?

 Scott: Not  really.   The  pressure  levels  achieved are  not larger than  those
 from an internal  combustion  engine.   Since mufflers  can  be built  for  cars,  it

-------
34
   should be no problem to build one for the system.  This will be even easier
   since an internal combustion engine operates at a range of pressure levels,
   whereas the acoustic device has a resonant frequency and the muffler can be
   tuned to that frequency.

   B. Liu:  The agglomeration that you observed at a bend was irrespective of
   the pipe.  That would mean that the ortho-kinetic agglomeration is really
   not very effective.  Rather it is based on something that is acted upon
   through turbulence and the acoustics.

   Scott:  Your explanation is very good.  Ortho-kinetic interaction is import-
   and because you can describe it physically and everyone can believe it.  I
   don't think it is the important mechanism.

          One of the things that has come out of our research at UT is that an
   acoustic field itself can generate a turbulent field, even if no turbulence
   existed (we think).  If you have a turbulent field with a flow process, then
   it is clear that an acoustic field adds to that turbulence.  I think there
   is a whole set of messy processes occurring here that increases particle-to-
   particle collisions.  And that's where I'd like to stop until we get some more
   fundamental work in.

          It is interesting, diagnostically, to look at this question though.
   Because we are now using microfilms, hot wires and LD-8's (which go very well
   with the fact that we are using particles) the diagnostic procedures should
   improve the information gathered.

-------
                                                                 35
ACOUSTIC AND TURBULENT AGGLOMERATION OF SODIUM AEROSOLS


    W. Hinds, E. P. Mallove, and M. W. First

    Department of Environmental Health Sciences, Harvard
    University, School of Public Health


ABSTRACT

    Tests of turbulent and acoustic agglomeration of captive
sodium fire aerosols at concentrations of 0.1 to 20 gm/mJ were
conducted in 90 m3 and 0.65 m3 vessels to evaluate these mech-
anisms for direct application air cleaning systems.  Aerosol
mass concentration decay with time was monitored by sequential
filter samples.  Turbulence was generated mechanically with a
51 cm diameter  centrifugal  fan impeller and a reverberant acous-
tic field was created with  an electronic siren.  The effectiveness
of each method  over a range of particle concentrations and power
densities was evaluated by  an agglomeration index, a measurement
of particle  growth based on sedimentation characteristics.  Both
turbulent and acoustic treatment markedly enhanced sedimentation
rate  compared to undisturbed settling.  The effectiveness of
both methods increased with increasing aersol mass concentra-
tion  and increasing power input per unit volume of aerosol.  The
agglomeration index reached 20 for turbulent agglomeration at an
aerosol mass concentration  of 3 gm/m3 and 7 for acoustic agglo-
meration at  14  gm/m3 when using an acoustic intensity  of 115 dB.
Turbulent agglomeration was more effective  than acoustic agglo-
meration for the same mass  concentration and power density con-
ditions.

INTRODUCTION

    A study  is  being  conducted in  the Harvard Air  Cleaning La-
boratory to  examine the  feasibility  of  using direct application
air cleaning as a  treatment method for  high concentrations of
sodium fire  aerosol that would  likely fill  a reactor  contain-
ment  vessel  following a  release  inside  a  liquid metal  fast bree-
der reactor.  Most particulate  air cleaning systems are  dynamic,

-------
36
   flow-through devices where particulate matter is continuously-
   removed from a stream of dust laden gas as it passes  through.the
   device.  A less common,  approach can be applied when  the  dust
   laden gas is confined within a vessel.  This  approach treats  the
   entire gas volume simultaneously,  as a batch  process, to  reduce
   airborne particulate concentrations as rapidly as possible.

       Leakage from the containment vessel is expected to be small,
   (no more than 0.1$ of the containment volume  per day). Because
   it is continuous, significant public health benefits  to the  sur-
   rounding community would accrue from a rapid  reduction in air-
   borne particulate concentration within the containment vessel.
   To reach the design goal,  reducing by a factor of 10  the  two-
   hour integrated concentration,  it  is necessary to reduce  the
   airborne concentration within a matter of minutes by  90%  or  more.
   Although rapid clean up  can be accomplished by a conventional
   air cleaning loop, such  a system would have to be very large
   and expensive to treat a containment volume of 7x10^  m3 in a
   few minutes.  A direct application air cleaning approach, on
   the other hand,  has considerable potential for this situation
   because it treats the entire containment  volume simultaneously
   without the necessity of having to pass large volumes of  gas
   through air cleaning devices.

       Three approaches have  been evaluated  experimentally:   sca-
   venging by inert powder  injection,  enhancement  of coagulation by
   mechanically generated turbulence,  and enhancement of coagula-
   tion by high intensity sound.   The latter two methods are simi-
   lar and are the  subject  of this report.   Inert  powder injections
   have been described in Reference 3-   Both turbulence  and  high in-
   tensity sound enhanced coagulation by applying energy to  the gas,
   throughout its volume, to  create relative motion between  particles

       Because of the complex geometry  of a  containment  vessel, in-
   cluding its machinery, the use  of  a  tuned sonic  coagulation sys-
   tem was not considered feasible and  consequently  no attempt was
   made to evaluate such a  system.  Instead,  random  traveling waves
   in a reverberant enclosure were used.   Mechanically generated
   turbulence was produced  with a  centrifugal blower wheel.  The
   experiments provided some  insight  into  coagulation mechanisms
   and the effectiveness  of two different  methods  operating  under
   similar conditions.   Effectiveness was  evaluated  by monitoring
   the change in aerosol  mass  concentration  as a  function  of treat-
   ment  time  in 0.65  and  90 m3  chambers.   Results  show the effect of
   initial  mass  concentration and  applied  power  density  on coagula-
   tion  rate.

-------
                                                                 37

EXPERIMENTAL PROCEDURES

    1.  Aerosol Generation for Turublent Agglomeration Tests

        Turbulent agglomeration tests were conducted in a 90 m3
sealed chamber, 4 meters in height.  Sodium aerosol was genera-
ted by burning the metal in an electrically heated steel pot.
The sodium pool, 260 cm2 in area, produced a dense white aero-
sol that filled the chamber within 10 minutes after initiation
of a fire.  A one pound sodium pool fire would burn for 20-30
minutes.  Decay in mass concentration with time was obtained
with open-faced MSA 1106B high efficiency filter samples taken
near mid-chamber height.  Ten to 15 samples were taken over a
3 to 6 hour period at 18.4 1pm for 0.5 to 3 minutes.  Less than
1% of the aerosol was removed by sampling.

        Undisturbed,mass concentration decay tests were conduc-
ted in the 90 m3 chamber to serve as a baseline against which to
measure the effectiveness of turbulent enhancement.  The results
shown in Figure 1 were obtained from eight baseline tests fol-
lowing the burning of 1 Ib. sodium.  Initial chamber relative
humidity ranged from 1*4 to 22%.  Analysis of the unhindered set-
tling data showed that this aerosol has the decay characteris-
tics equivalent to that for lJ urn monodisperse particle under
stirred settling.  The actual mass median diameter (MMD) and
geometric standard deviation (GSD) of the aerosol cloud was
determined with the aid of eight-stage Andersen impactors and
an aerosol centrifuge [1].  The size distributions, 20 minutes
after initiation of a 1 Ib. sodium fire for all of the generated
aerosols were found to be nearly the same with a MMD of 2.1 urn
and a GSD of 1.8.  Measurements taken at later times indicated
an equilibrium between sedimentation and growth with MMD stabi-
lizing at about 3 ym.

        Measurements were made of particle density and dynamic
shape factor from the aerodynamic equivalent diameters deter-
mined by aerosol centrifuge and geometric parameters measured in
the scanning electron microscopy [2].  Untreated aerosols aged
from 34 to 376 minutes gave an average particle density of .78
gm/cm3 and average dynamic shape factor of 1.37 [2].

        Quantitative chemical analysis indicated that the aero-
sol particles were porous sodium carbonate with some water of
hydration [3].  During the undisturbed settling tests, aerosol
deposition per unit area of chamber wall was  0.10 to 0.01 of  the
amount  that settled on the floor.

-------
38
     S
     W)
     c
     o
     •H
     -p
     
-------
                                                                 39

    2.  Turbulent Agglomeration Tests

        A 7.5 HP, 3,000 cfm centrifugal blower (New York Blower
Co., type GI, size 20) was placed in the center of the 90 m3
chamber, to induce turbulence.   The blower was started shortly
after peak aerosol concentration was reached.  The time decay
history of five such tests with peak aerosol concentration near
2 gm/m3 is presented in Figure 2.

        The blower was run with and without the scroll casing in
place to distinguish the effect of centrifugal deposition from
the effect of turbulence enhanced agglomeration and sedimenta-
tion.  Operation with the scroll casing off takes away a major
surface for deposition, though some cleaning credit must still
be taken because of particle impaction on the fan blades.  Re-
moving the casing did not greatly change the decay profile from
casing-on performance, as can be seen in Figure 2.  The primary
aerosol decay mechanism was, therefore, by enhanced agglomeration
and sedimentation onto the chamber floor.  Analysis of samples
of wall, floor, and blade deposits indicated that aerosol depo-
sition occurs, roughly, 28% on walls, 65% on floor, and 7% on the
fan blades.  Numerous large spherical particles (d > 100 ym) were
found deposited on horizontal surfaces following each turbulence
test.

        Substantially faster decay rates were obtained with tur-
bulent agglomeration when the peak aerosol mass concentration in
the chamber was above 2 gm/m3.  One such tests in shown in Figure
3.

        By using different pulley combinations on the blower, it
was possible to generate different levels of turbulence in the
chamber.  Power densities, determined from measurements of fan
speed and electricity consumption, are discussed in a later sec-
tion.  Assuming an electrical to mechanical energy conversion
efficiency of 0.80, power .densities in the range of 20-70 watts
per cubic meters of chamber volume were utilized.

    3.  Acoustic Agglomeration Tests

        Efforts to determine the effect of acoustic agglomera-
tion on sodium combustion aerosols began with the search for a
suitable and convenient source of high intensity sound.  The
first sound generating devices employed in this research were
pulse jet engines [4].  A special pulse jet engine constructed
by Ontario Research Foundation (Canada) was unable to provide a
sound field of sufficient intensity for acoustic agglomeration
experiments in the 90 m3 chamber.  These tests were described in
Reference 3.

-------
40
                                    FIGURE 2

                TURBULENT  AGGLOMERATION AEROSOL CLEARANCE TESTS
        U.O
     CO
      g
      60
      d
      o
      •H
      •P
      0]
      k
      •P
      c
      CO
      O
      C
      O
      0

      CO
      01
      cd
      S
      o
      CO
      O
        1.0
                                      O  Blower  Casing On

                                      O  Blower  Casing On

                                      A  Blower  Casing On

                                      D  Blower  Casing Off

                                      0  Blower  Casing Off
                                Baseline       vx
                                  Concentration     >»
                                    Profile Without    -
                                      Induced Turbulence
                                        (From Figure 1)
0.1
       0.01
                          50       100      150     200     250

                           Time  After  Blower Start (Minutes)
                                                         300

-------
                                                           41
S
o
c
o
•H
-p
cd
fc
-p
c

-------
42
           For later tests,  a Federal CJ-24  electronic  siren (a
   type employed in emergency vehicles)  was  used to generate in-
   tense sound fields in a steel chamber having a volume of 0.65 m3
   and an average internal height of 75  cm (Figure 4).   A smaller
   chamber was needed because the acoustic power radiated by a  sin-
   gle CJ-2*I siren will not produce adequate sound levels though-
   out a 90 m3 chamber.  The CJ-24 "siren" is not a gas-dynamic
   rotating siren,  by the usual  definition,  but is a very powerful
   loud speaker powered by a PA-20A amplifier and signal'generator.
   When 60 watts of electrical power were used to energize this
   siren,  sound pressure levels  of 1*15 dB were measured at the  sam-
   pling port shown in Figure 4.   The frequency spectrum was broad
   band with a peak in the 600 to 1,200  Hz range.   Variations in
   intensity did not exceed 3 dB throughout  the 0.65 m3 chamber.
   The siren was used in the "yelp" signal mode,  a complex wave
   form in the audible range which has a superimposed oscillation
   of approximately 5 Hz.   The acoustic  power of the siren was  cal-
   culated from data supplied by  the manufacturer.   The acoustic
   power output of  the siren for  lower sound pressure levels was
   calculated by assuming a  reverberant  field in the chamber with
   constant wall losses.   Acoustic tests were performed at 1^5  dB,
   132 dB,  and 125  dB, spanning  a power  density range from 10 watt/
   mo to 0.1 watt/m3.   Sound pressure levels were  measured with a
   General Radio octave band analyzer in the flat  response mode.

           In a typical test,  between 50 and 100  gm of  sodium metal
   was heated rapidly  to  burning  temperature in a  steel pot and
   allowed to burn  for 10  minutes filling the sealed chamber with a
   dense sodium aerosol.   Open-faced glass fiber  filter samples
   were obtained at mid-chamber height to determine the mass  con-
   centration over  time for  unperturbed  and  siren-on conditions.
   Each filter sample was restricted to  9.2  liters (I.H% of the
   chamber volume)  to minimize depletion of  the aerosol by sam-
   pling.   Up to 11 samples  were  taken giving a maximum concentra-
   tion decrease due to sampling  of 1*1$.  Because  unperturbed and
   siren-on tests were sampled at roughly the same time intervals,
   changes in concentration due  to sampling  were identical for  each.

           Figure 5 shows  mass concentration decay  results  for  un-
   perturbed sedimentation in  the small  chamber and sedimentation
   enhanced by  acoustic agglomeration.   The  curves  for  the  three
   sonic tests  indicate moderate  enhancement  of sedimentation.

   RESULTS

       Data for undistrubed  settling and  turbulent  agglomeration
   tests in the 90  m3  chamber are  summarized  in Table 1.   The basic
   parameters which were extracted  from  the  concentration  decay
   graphs  for each  test are  initial  aerosol mass concentration

-------
                              FIGURE
       0.65 M3 CHAMBER FOR ACOUSTIC AGGLOMERATION EXPERIMENTS
   Power
  Supply
Federal
Elec-
 tronic
Siren
               Light for Illumi-
                nating chamber
Signal Generator
                l-r
122 cm

-------
44
        S
        bO
       C
       O
       •H
       -p
       0)
       !H
       •P
       C
       (U
       O
       c
       O
       u
       oJ
       a
       o
       n
       O
       ^
       OJ
                                      FIGURE 5


                 ACOUSTIC AGGLOMERATION  AND UNDISTURBED SETTLING
                               IN  A  0.65  M3  CHAMBER
                  Arrows Indicate  Time
                     Siren Turned  On
                                            Unperturbed Settling
                                            Siren On
            1CLS7
           0.5-
           0.1
                     10      20     30     UO     50      60      70


                 Time After Peak Aerosol Mass Concentration  (Minutes)

-------
                                                                 45
TABLE 1 - SUMMARY OF UNDISTURBED SETTLING AND TURBULENT AGGLO-
                 MERATION TESTS IN 90 M3 VESSEL
Initial
Aerosol
Mass
Conc'n
(gm/m3)

6.5
3.8
2.4
1.6
1.1
0.084


Time for
C/C0=e-l
(min. )

47
105
162
175
290
592
Mass
Fraction
Removed
in
5 min.
Undisturbed
0.10
0.047
0.031
0.028
0.017
0.008
Mono-
disperse
Equivalent
Size ,d*
(ym)
Settling Tests
9.2
6.2
5-0
4.8
3.7
2.6

Agglomer-
ation
Index
IA

4.4
3.0
?.4
2.3
1.8
1.2


Power
Density
watts/m3

0
0
0
0
0
0
Turbulent Agglomeration Tests
3.4
3-1
2.0
2.0
1.9
1.7
1.6
1.3
1.2
0.94
0.85
3.4
2.1
7.5
2.3
3-8
16
15
12.5
25
13
25.5
0.76
0.91
0.49
0.89
0.74
0.27
0.29
0.33
0.18
0.32
0.18
33.8
43.7
23.1
41.7
32.5
15.8
16.3
17.9
12.7
17.6
12.5
16.1
20.8
11.0
19.9
15.5
7.5
7.8
8.5
6.0
8.4
6.0
27
27
21
27
71
18
18
27
18
18
27

-------
46
   (C0) and time (te) for the mass concentration to reach (C0/e).
   For undisturbed settling,  the initial mass concentration was
   considered to be the peak  concentration reached.  For the tur-
   bulence tests, initial concentration was defined as the measured
   concentration at the time  turbulence was initiated, usually
   within a few minutes of peak chamber concentration.  Data from
   undisturbed settling and acoustic agglomeration tests in the
   0.65 m3 chamber are similarly presented in Table 2 and defined
   in like manner.

       Most mass concentration decay profiles gave straight lines
   when plotted semi-logarithmically, indicating exponential decay
   functions.   Stirred settling conditions occur in the chambers
   due to thermal convective  motion and energy from the agglomera-
   tion mechanisms  themselves.  'It is well known that stirred set-
   tling of monodisperse aerosols produces exponential mass con-
   centration decay.   Even though the sodium aerosol is not mono-
   disperse, it is  extremely  useful to approximate aerosol settling
   characteristics  by the exponential decay of a monodisperse aero-
   sol with 1/e decay time, te,  measured directly from the concen-
   tration data.  The mass fraction removed,  f(t),  in any standard
   time interval (e.g.,  5 minutes,  as in Tables 1 and 2)  is readily
   determined from  the relation,
       f(t)  =  l-e~t/te
(1)
  More  si'gnificantly,  the particle diameter of an  "aerodynamically
  equivalent monodisperse aerosol" with the same settling charac-
  teristics, djf,  can  be  derived  from  the  decay profiles.  We have,
  for a stirred  settling monidisperse aerosol of diameter,  djf,  in
  a  chamber of height, h,
                     Pd#2gt
       c(t)/co  '
      where  g  =  gravitation acceleration
             n  =  gas  viscosity
             p  =  particle density
             X  =  aerodynamic shape  factor


  Equating the bracketed quantity  in Equation  2 with t/te gives
  d* as,

-------
TABLE 2 - SUMMARY OP UNDISTURBED SETTLING AND ACOUSTIC AGGLOMER-
                 ATION TESTS IN 0.65 M3 VESSEL
Initial
Aerosol
Mass
Cone Tn
(gm/m3)
Mass Mono-
Fraction disperse Agglomer-
Time for Removed Equivalent ation
C/CQ=e-l in Size,d* Index
(min. ) 5 min. (ym) IA
Power
Density
watts/m3
Undisturbed Settling Tests
21.
16.
8.
6.
5
5
2
8
13
11
20
23
0
0
0
0
• 32
.37
.22
.20
Acoustic
14.
10.
9.
7.
1».
2
5
0
6
2
3.3
1.9
5.3
7.0
28.5
0
0
0
0
0
.78
.6/1
.61
.51
.16
7
8
6
5
.6
.3
.1
.7
3
4
2
2
.6
.0
.9
.7
0
0
0
0
Agglomeration Tests
15
12
11
10
5
.1
.if
.9
.4
.1
7
5
5
5
2
.2
.9
.7
.0
.4
9.1
9.1
9.1
0.46
0.091

-------
48

   Values for d^ computed for each test are given in Tables 1 and 2.

       A useful parameter for characterizing both undisturbed and
   enhanced agglomeration tests is the "agglomeration index", IA,
   defined here as the ratio of dx to the initial aerosol MMD.
   Eight-stage Andersen impactor data indicate that within the
   first 20 minutes after initiation of a sodium pool fire, the  MMD
   of the aerosol Is about 2.1 ym.  Using this reference MMD to
   form the agglomeration index, we have plotted IA versus initial
   mass concentration for undisturbed settling tests and for turbu-
   lent and acoustic agglomeration tests as shown in Figures 6 and
   7.   No distinction is made in these figures between tests at
   different power densities.   Because of the limited range of our
   power densities, strong statements about their relation to IA
   are not justified.

       Figures 6 and 7 also present trend lines fitted to the data.
   In each case, the trend line is constrained to pass through IA=
   1.0 at zero mass concentration, a constraint which has an obvi-
   ous and strong physical interpretation.   As expected from basic
   coagulation, theory [5] each of the four trend lines shows increas-
   ing agglomeration indices with increasing mass concentration.
   Enhanced agglomeration rate leads naturally to an increased
   average particle size due to agglomeration in a given time peri-
   od.   By adding turbulent or acoustic  energy to the aerosol,
   relative motion between particles is  enhanced and agglomeration
   speeded up.   For the range  of power densities used,  turbulence
   yields a higher agglomeration index than acoustic for the same
   initial mass  concentration.

       For the turbulence experiments  and the acoustic  tests, the
   following results  of multiple regressions  of agglomeration index
   with respect  to initial mass concentration and power denisty
   were obtained.   The point IA=! at C0=0 was  included  in each
   case:

       IA = 0.302  + 4.85  C0 +  0.093  P   turbulent agglomeration

       IA = 1-05 + 0.441  Co +  0.029  P  acoustic  agglomeration

       where  Co  =  initial mass  concentration  in gm/m3
             P   =  power density in watts/m3

   Significance  tests  of the regression  equation coefficients showed
   that both  concentration coefficients  were  significant (P<0.02)
   and  both power  density coefficients were not  significant.  It  is
   apparent  that turbulent  agglomeration is more  effective  than
   acoustic agglomeration for  a given  initial  mass concentration
   and  that turbulent  agglomeration  is enhanced  to a greater  extent

-------
                        'FIGURE 6

-------
50
                                 FIGURE 7


  AGGLOMERATION INDEX AS A FUNCTION OF MASS CONCNETRATION  FOR  ACOUSTIC

      AGGLOMERATION AND UNDISTURBED SETTLING TESTS  IN  0.65 M3  VESSEL
     X
     
-------
                                                                 51
for a fixed fractional increase in initial mass concentration
than is acoustic agglomeration.

DISCUSSION AND CONCLUSION

    The range of agglomeration indices found in our acoustic
experiments is consistent with analogous measurements made by
other researchers.  Although no previous work compares directly
with our research because sound sources were usually employed
only for brief periods (seconds) on flowing rather than fixed
volume systems, Scott [4] and'jVolk and Moroz [5], obtained ag-
glomeration indices from 3 to 9 for sound pressure levels from
120 dB to 165 dB.  Their mass concentrations were in the range
0.5 to 2.5 gm/m3 and sound was continued from  2.5 to 40 sec-
onds .

    By using the concept of "scale of turbulence" noted by
Fuchs  [6], it is possible to gain some insight into why turbu-
lent agglomeration does not show significant increases in effec
tiveness for significant increases in power density.  The in-
ternal scale of turbulence, X0, used by Levich and referred to
by Puchs, is defined for a turbulent medium by,
where v is the kinematic viscosity of the medium and e is the
power density per unit mass of the medium.  For the range of
power densities  employed here the scale of turbulence, Xo, ranges
from 90 ym to 120 urn; a range unlikely to be associated with
large changes in agglomeration index.

    Turbulent motion and high intensity reverberant acoustic
fields induce rapid agglomeration of sodium aerosols.  The _ high-
er the initial aerosol mass concentration, the more effective
are these agglomeration mechanisms which produce enhanced sedi-
mentation.   For  undisturbed settling, sedimentation is similarly
enhanced by  increasing initial mass concentration.  For the same
energy expenditure turbulence appears to be a more effective ag-
glomerating  mechanism than acoustic energy at a fixed mass con-
centration.   Increasing the applied power density of turbulence
or sound produces  smaller  increases in agglomeration index than
does a similar  fractional  increase in initial mass concentration.
Both mechanisms  show promise  for direct  application air clean-
ing systems.  Because of  the  relative ease  of  generating  turbu-
lence  versus generating a  large-scale high  intensity sound field,
turbulent agglomeration may be preferred.

-------
52
   ACKNOWLEDGEMENT

       Support of this research was provided by the  U.S. Energy
   Research and Development Administration under Contract No. E(ll-
   D-2801.

   REFERENCES

   [1]  Tillery, M.  I.,  "A Concentric Aerosol Spectrometer", Am.
        Ind.  Hyg. Assn.  J. , 34:62,  (197*1).                     	

   [2]  Hinds, W., Mallove, E.,  and First,  M. W., "Density and Shape
        Factor of Sodium Aerosol",  Final Report, COO-2803-5, Febru-
        ary 1977, ERDA Contract  No. E(ll-l)-2803.

   [3]  Mallove, E., Hinds, W.,  and First,  M. W., "Direct In-Vessel
        Application  Experiments  at  Harvard  Air Cleaning Laboratory",
        Annual Report,  February  1977, ERDA  Contract  No.  E(ll-l)-
        2801.

   [4]  Scott, D. S.,  "A New Approach to the Acoustic  Conditioning
        of  Industrial  Aerosol Emissions", Journal of Sound and Vi-
        bration, Vol 43,  No.  4,  1975, 607-6T9~!~

   [5]  Volk,  Michael,  Jr.,  and  Moroz, William J.,  "Sonic Agglomera-
        tion  of Aerosol  Particles", Water.  Air,  and Soil Pollution,
        Vol.  5,  1976,  319-334.	—

   [6]  Fuchs. N. A.,  The Mechanics of Aerosols. New York,  Pergamon
        Press, 1964.        ~
  WRITTEN DISCUSSION

  J.A.  Gieseke, L.D.  Reed, and K.W. Lee
  Battelle, 505 King  Avenue
  Columbus Laboratories
  Columbus, Ohio  43201

      Aerosol behavior within Liquid Metal  Fast Breeder Reactor (LMFBR) con-
  tainments is of critical importance since most of the radioactive species are
  expected to be associated with particulate forms, and the mass of radiologically
  significant material leaked to the ambient atmosphere is directly related to
  the aerosol concentration airborne within the containment.  Mathematical models
  describing the behavior of aerosols in closed environments, besides providing
  a direct means of assessing the importance of specific assumptions regarding
  accident sequences, also serve as the basic tool with which to predict the con-
  sequences of various postulated accident  stiuations.  Consequently, considerable
  efforts have been recently directed toward the development of accurate and
  physically realistic theoretical aerosol  behavior models.

      These models have accounted for various mechanisms affecting agglomeration
  rates of airborne particles matter as well as particle removal rates from closed
  systems.  In all cases, spatial variations within containments have been neglected

-------
                                                                               53

and a well-mixed control volume has been assumed.

    The most.recent of these, the HAARM-2 code developed at Battelle's Columbus
Laboratories ?, is an extension of the HAA-3B code developed at Atomics
International .  The HAARM-2 model differs from previous models in that it
allows temporal variation of containment gas temperature, pressure, and tem-
perature gradient normal to the containment walls.  Also, settling velocities
which are dependent upon the morphological properties of.individual agglomerates
are corrected using an experimental dynamic shape factor .  In addition, wall
deposition by thermophoresis is included as an aerosol deposition mechanism
and a calculated particle-particle collision efficiency is employed.  Though
not included previously in any model, terms describing agglomeration as caused
by gas turbulence have been added to the HAARM-2 code so that calculations in
support of this discussion could be made.

GENERAL THEORY

    The theory used to develop the HAARM-2 code is based on solution of the
governing intftgro-differential equation which describes the rate of change of
particle concentration due to various agglomeration and removal mechanisms.
This equation may be written in the following form:
    -  n(x,t) = K   [1/2
                 O
                        -n(x,t) /  
-------
54
     4   3
     rirr  = volume  of  particle with radius  r

     4   3
       '  = volume  of  particle with radius  r1
   n(x,t)  =  the  size  distribution  function

     R(x)  =  the  removal  rate  of  particles  produced  by  gravitational
            settling  to  the floor,  diffusion to  the walls
            (wall  plating) , and  leakage

   S(x,t)  =  represents the source  rate of  particles input  to  the  vessel.

       The first integral in  Equation (1)  represents the formation rate of
   particles between  the sizes x and  x+dx  as a result  of collisions  between
   particles of  volumes  £ and x-£.  Similarly, the  second  integral represents
   the  disappearance  rate of  particles in  the size  range between  x and x+dx
   due  to  collisions  with all other particles.

       The functional form of the  collision kernel  (x>£)  depends upon the
   coagulation mechanisms present  in  a given system.   In an enclosed containment
   vessel, possible mechanisms causing relative  motion between particles,  and
   thus coagulation,  include  Brownian motion of  the particles, gravitational
   settling, and turbulent gas motion.   In most  analyses where more  than one of
   these mechanisms is present,  they  are assumed to be separable  and additive
   such that

       •Kx,£) =  B(x,O + G(x,O  +  T(x,£)                                       (2)

   the  agglomeration  and removal terms will be described in more  detail.

   AGGLOMERATION TERMS

   Brownian  Agglomeration

       Current aerosol models use  the Brownian collision parameter B in a  form
   which can be  written  as
       K  B(x,C)  =     (r
        0
                                  r+a+be    .r'+a+be

                                       r
(3)
  where  a,  b,  and  X are  constants  defining the Cunningham correction factor
  which  accounts for the low Knudsen number effects  present for small particles.

  Gravitational Agglomeration

       The collision parameter for  gravitational agglomeration is based on
  consideration of the relative sedimentation rates  for different sized particles
  and  is dependent on the collection of small particles from the volume swept

-------
                                                                               55
out by a large, rapidly settling particle.
parameter is given by
                                            The gravitational agglomeration
                           ,5) -{| ro(r) - r'a(r')|  •  (r + r')3

                    (r + r1)2] + b(r + r1)2  -| re"Xra(r)-r' e~XrIa(r')|}     (4)
where
    e(x,£) - the particle-particle collision efficiency

     a(r)  = the particle shape factor

       p   = the density of the particle material.

Turbulent Agglomeration

    Particles are caused to collide in a turbulent gas by being suspended  in   ,
a shear flow and by inertial effects cuasing relative motion between .particles .
The shear and inertial components of the turbulent collision parameter are
termed T.. and T?, respectively.  These should appear as T = (T.  + T» )    , but
have been assumed additive as T -  T- + T_ to permit analytical integration.
The expression for the shear induced component of the turbulent collision
parameter is
                              1/2         ,
                                  (r+r')J,                                 (5)

and the inertial component of the turbulent collision parameter is

                                         1/4
                       ATT  1/2T
                         P
                        9n
                                   15n
                  |(r) (r2 + ar + bre"Xr) -a  (r1)
                                                                  ) |
                                                                             (6)
where e  is the rate of turbulent energy dissipation per  unit  mass
(cm2/sec3), and p is the gas density.
Removal Terms

    In Equation  (1) , R(x) represents  the removal  terms.  HAAM-2  accounts for
any of the four  possible mechanisms :   (1) removal at  the  vessel floor  due to
gravitational settling, (2)Brownian diffusion  to  the  walls  (wall plating),
(3) thermophoretic deposition on the  walls,  (4) leakage out  of the vessel.
Mathematically formulated
r(x) = G0 r(r + a + be Xr) + PD ~4 (r + a + be Xr) +
        K                     R  t.
                                                            + T.
                                                                R
                                                                             (7)

-------
56

   where
            2ga(r)p
       G  = —jjr	- ,  the gravitational settling constant
        K
                     ,  the plating coefficient

       R^  = leak rate from the vessel

            A
       TR  = ~-^U ,  thermophoretic deposition coefficient.

   further, h is the vessel height,  A  the wall  area,  A  the distance from the wall
   over which a particle concentration gradient exists, U  the thermophoretic
   velocity, and V the  vessel volume. In the HAARM-2  code the assumption has
   been made that  the final term in the slip correction factor is negligible and
   hence for both  gravitational settling and agglomeration the constant is taken
   as zero.

       The method  of moments was used to solve the governing integro-differential
   equation, Equation (1), and the solution technique has been described
   previously  ' ' .  Basic to this method is the assumption that the particle
   size distribution may be represented by a log-normal distribution at all times
   and the parameters describing the particle population  (number concentration,
   mean size, and  geometric standard deviation) are derived at the end of
   selected time intervals in terms of the moments.

   RESULTS

       A series of calculations were performed to compare the HAARM-2 aerosol
   behavior model  with  the experimental measurements  of Hinds, Mailove, and First.
   Of  particular  interest was the comparison of the  predictions and experimental
   results for the case of turbulent induced augmentation of the agglomeration rate
   since this is the first such comparison ever possible.  Of course, the base
   cases with no added  turbulence are of importance in  establishing a reference
   point for comparison with the turbulent cases.

       The base cases with no added turbulence were calculated in three ways.
   First,  the instantaneous release of sodium oxide aerosol was assumed to occur
   at the  end of the sodium burning period of 20 minutes.  The characteristics
   of the  aerosol  at this time were taken as aerodynamic  mean diameter = 2.1pm,
   effective3density =  0.78 g/cm , shape factor = 1.37  and mass concentration
   =2.0 g/m .   The HAARM-2 code employs mass equivalent  particle diameters and
   makes its own corrections to particle mobility accounting for effective
   density and shape factor combined.  Therefore,  an initial mass equivalent
   radius, r, of 1.4 pm was used (calculated by correcting the impactor determined
   aerodynamic diameter for density and shape factor),  a  geometric standard
   deviation, a ,  of 1.8 was assumed for the distribution, the material density,
   p  , was taken" as that for hydrated Ha2CO_ or 1.5 g/cm  , and a constant correct-
   iBn for density, a,  of 0.33 was assumed.  These values are consistent with
   those reported  by Hinds, Mallove and First for the particles at the end of the

-------
                                                                               57

burning period.

    A second similar case also assumed an instantaneous source and identical
conditions except that the value used for the aerodynamic mean radius was 1.5
ym as reported by Hinds, Mallove and First for times after the end of the fire
which corrected for effective density and shape gives the value of r =2.1.

    A third case followed the procedure usually used to predict aerosol
behavior following a sodium fire.  A constant particle source rate was
assumed over the 20-minute burning period.  In this case the source particles
were-assumed to have a mass mean radius of 0.5 urn, a material density of 2.27
g/cm , and a geometric standard deviation of 2.  The effective density for
the particles was calculated within the program as a function of the number of
particles forming an agglomerate.

    The results of HAARM-2 code calculations for these three cases are shown
in Figure 1.  It is seen that the predicted airborne mass concentrations are
consistently higher than the experimental values.  This is consistent with the
nature of the HAARM-2 code which was intended to provide conservative
estimates for nuclear reactor safety analyses and, therefore, employs procedures
which always lead to slight overestimates of airborne concentration.  Neverthe-
less, the aerosol behavior model shows reasonable agreement with these most
useful data.

    To compare the HAARM-2 code calculations with the experimental results
when the blower was operated, the terms for turbulent agglomeration were added
to the HAARM-2 code.  To simplify the integration for this comparison it was
assumed that the particle-particle collision efficiency, e, which appears in
Equations (5) and (6) was a constant rather than a function of particle size.
This constant is taken as unity in the turbulent agglomeration terms but is
included as a function of particle size in the gravitational agglomeration
term.  The calculations to show turbulent agglomeration effects were based on
a value-for e  , the turbulent energy dissipation rate, equivalent to 45
watts/m  whicS is the mid-point in the range of 20-70 watts/m  reported by Hinds,
Mallove, and First for their experiments._ Other assumptions were the same as
for the baseline case calculations using r = 2.1 ym.

    The results of the calculation for turbulent agglomeration are shown in
Figure 2.  It is seen that the prediction is on the low wide but adjacent
to the range of experimental results.  This is reasonable agreement but will
probably be improved when the HAARM-2 code is modified to include a size
dependent collision efficiency, e, in the turbulent agglomeration terms
and some accounting is made for overestimates resulting from the additive
procedure which was used for agglomeration rates in the  form T=Tj+T2.  The
agreement between theory and experiment is very encouraging and snows promise
of future improvement.

CONCLUSIONS

    The data reported by Hinds, Mallove, and First are very useful as
additional confirmation of sodium aerosol behavior in enclosed vessels and in

-------
58
               PREDICTED AND  MEASURED  AEROSOL BEHAVIOR
             0.01
                     Time dependent calculation within HAARM-2_
                     code, value ranged  from 1.0 to 0.47
                      50    100   150   200  250   300   350
                      Time After  Initiation of Fire, minutes
            FIGURE 1.  PREDICTED AND MEASURED AEROSOL BEHAVIOR

-------
                                                      59
                     Predicted, no turbulence
                Baseline  data
                        Data with turbulence added
                        from Hinds,  Mallove  and
                        First
      Predicted with
      turbulence
0.01
          0     50    100     150   200   250   300
            Time  After Blower Start, minutes
      FIGURE 2.  COMPARISON OF PREDICTED AND MEASURED
               TURBULENT AGGLOMERATION EFFECTS

-------
60
   particular provide a most useful basis for evaluating the effects of turbu-
   lence on agglomeration rates.  The  agreement between the HAARM-2 code
   calculations and experimental results for the baseline conditions without
   significant turbulence is very good with the HAARM-2 code, as expected,
   giving slight overestimates of airborne concentration.  The agreement
   between HAARM-2 code predictions and experimental results for cases with
   significant turbulent agglomeration is quite good but some improvement can
   be expected from scheduled code modifications.

   ACKNOWLEDGEMENT

       The research on which this discussion is based was performed for the U.S.
   Nuclear Regulatory Commission under Contract No. AT(49-24)-0293, Task 7.

   REFERENCES

   [1] Reed, L.D., and Gieseke, J.A., "HAARM-2 Users Manual", BMI-X-665
       (October 31. 1975).

   [2] Hubner, R.S., Vauehan, E.U., and Baurmash, L.. HAA-3 User Report. AI-
       AEC-13038 (1973).

   [3] Kops, J., Dibbets, G., and van de Vate, J.F., J. Aerosol Sci., £, 329 (1975).

   [4] Saffman, P.G., and Turner, J.S., J. Fluid Mechanics, 1, 16 (1956).

   [5] Cohen, R.E., and Vaughan, E.U., J. Colloid Interf, Sci., 35, 612 (1971).
   OPEN DISCUSSION

   Shaw:  I was wondering,  Jim and Bill,  how the turbulence - both with acoustics
   and without acoustics -  affect your cutting into the wall.

   Hinds:   We saw an effect on the deposition on the walls   from turbulence,
   however, the predominant deposition was still on the floor.   About 65% of  the
   material ended up on the floor.   Half  of the remaining 35%  was on the walls,
   and the rest was  trapped directed on the blades  of the device.

   Shaw:  Do you see any significant increase because of the turbulence?

   Hinds:   We had between 1% and  10% without turbulence on  the walls,  and then
   went up several fold because particles large enough to get  in are driven onto
   the wall by turbulence on the  chamber.

   Cooper:   The different coefficients for the concentration for the two different
   cases might suggest  that there were somewhat different particle size
   distributions.  So that,  although the  work shows the turbulent agglomeration
   can be  significant,  perhaps you  have to be careful in making  a comparison  between
   the turbulent agglomeration in the big room and  the acoustic  agglomeration of

-------
                                                                               61
the smaller chamber.  Especially when these concentration coefficients are so
different.

Hinds:  We measured the size distribution in both cases.  In the case of the
small chamber, the burn took place very rapidly, about 2 minutes versus 15
minutes in the other one.

        We waited until the fire ended before we made any measurements or
treatment.  At that time we had roughly the same size distribution.  But it is
a younger aerosol; the aerodynamic sizes were roughly the same.

Benarie:  How did you monitor or control humidity in the chamber?  Because you
are working with a very humidity - sensitive aerosol, the conditions of a few
percent change in humidity can cause a sizeable change in turbulence.  Your
experiments end up to what I am going to say on Friday morning.

        In fiber filtration, we generally do not consider that air in the
gates entering the filter develops a turbulence or anything like a turbulence.
I will show you how this occurs.  The turbulence contributes a large measure
of agglomeration and leads to a better performance of filtration because parts
of the filter are already agglomerated.  So in developing filtration mechanisms
which are most formed in uniform and laminar flow, a turbulent agglomeration
should also be considered.

Hinds:  The effect of humidity is important to a hygroscopic aerosol.  In the
runs that I described earlier, the humidity ranged from 16% to 24%.  We found
that in that range, based on our baseline runs, we got very reproducible
results.  When we went to 35% to 50% humidity, these results were quite
different; by a factor of two in mass concentration.

Scott:  I am very interested in your results.  I noted that the highest sound
pressure levels were 145 dB and I am intrigued that you caught anything there.
You obviously caught something, but you will find that there is a marked increase
in agglomeration ratio up towards 155 dB.

        I think it would be interesting if you would introduce some other runs.
It would be interesting if they were done at the same time and in the same
chamber.  It would also be interesting if you had three conditions.  1) Pure
turbulence  2) Pure acoustic  3) A combination of turbulent and acoustic.  Our
preliminary results show that the third condition is where things really start
to happen.

        My last comment refers to power.  I think you said that "turbulent
agglomeration, for the power requirements, would be satisfactory without the
acoustics."  But we hope that this would only be used in the event of a
catastrophe.  At that point power would not be a criteria at all and should
have no relevance.  Would you please comment on this?

Hinds:  On power requirements, your statement is true.  However, in the case
of an emergency, power may not be available.  But slight changes in efficiency

-------
62

   are really irrelevant in this case.

           On sound pressure levels,  we were aware that these were on the low end
   of the scale and we weren't able to  reach the higher pressures.  But we felt
   that if we could get above 140 dB  we would have something that was worthwhile.
   We did reduce it down to 125 and didn't observe a great deal of difference.

           The combination of systems is worth note and we are looking into it.

   Lamb:    What is the feasibility of artifically increasing the aerosol
   concentration?  I'm thinking of a  case in an emergency where you would blow a
   harmless aerosol which  would scrub.

   Hinds:   If we could artifically increase the concentration we would be much
   better  off.  This is worth considering given the shape of those curves.   In
   terms  of the particular situation  with the reactor,  we have many thousands of
   pounds  of non-radioactive sodium in  the thing.   All  you have to do is open a
   valve,  let sodium burn  and you have  got a lot higher concentration.   But if
   you have an accident with a lot of sodium around,  you just don't want to burn
   more.   So this  may not  be very realistic.

   Baril:   Could you tell  us a little bit about your  sampling techniques.   Where
   in the  chamber  did you  take your samples?

   Hinds:   We sampled about  mid-height  in the chamber through a 1  1/2"  diameter
   opening at about 20 liters/min.  The sampling tube is  4  inches  long  and  has  a
   filter.   We have a valve  that  we could  shut off  to isolate the  chamber.   We
   sampled directly under  the filter.   The samples were then weighed within a
   matter  of seconds after removal to prevent any  inclusion  of  moisture.  The
   samplings were  brief, lasting  between one  and  three  minutes,  and  spaced  about
   10 to 15 minutes apart.   All were  done  at  one height in  the  chamber.

           Initially,  when we burn the  sodium,  we  see stratification.  The  hot
   sodium  rises  to  the top,  but there is enough convective energy  from  the  burning
   sodium  to mix it well.  When the tests  were taken, there'was  no  stratification.

   Oder:    Will  increased  humidity enhance  sedimentation, in which  case, it  is
   very easy to  boil water without fear  of  loss of sodium?

   Hinds:   Yes.   You could  inject steam.  However, if  you have metallic sodium
   present  this  would  be dangerous because of  the evolution  of hydrogen gas.  So
   you have a  tradeoff of  blowing  the thing up  and getting the air back out.

   Loftier: We sometimes use  acoustics  in  liquids for agglomeration.  I would like
   to know  if  there  is any maximum or optimum  power input or perhaps a final size
   of agglomerate which you  can obtain  if you put more  energy into the system?

   Hinds:   The mechanisms  involved in the  scale of turbulence is such that  it
   would tend  to scale surfaces on the order of about 100 microns.  It would also
   tend to  break up  particles  that were long on that order of dimension and you

-------
                                                                               63

would end up with a kind of spherical shape.

        As for turbulence in the break-up of agglomerates in acoustics, I
think David Scott would be more qualified to respond.

Scott:  If you do break up particles in an aerosol, it is clear that it is the
big ones that you are breaking up.  It also makes sense, almost intuitively,
that if you have got some long stringy thing that you are going to break up,
it would break approximately in the middle, rather than chip little bits off
the end.  So it seems that you could bring the agglomerate to too large a
size and break it into approximately equal size particles and you would kick
up the number density which is desirable.

        If you are talking about a solid particle in a liquid, that means you
would be propagating through a nearly incompressible fluid.  For a given
acoustic field and intensity, the wavelengths are totally different.  The
amplitude, which is the critical thing, would be much smaller than in a gas.
These things occur because you do have a dynamic motion.  In a liquid, the
dynamic motion would be much less than in a gas unless it was a bubbly liquid
where you have sort of an analog thing too.

Loffler: A most important thing would be the stress which acts on the agglomerate
and the forces, for example, adhesion forces.  In the case of turbulent flow
you will not get shear stress for particles in the range of one to five
microns which will break the agglomerates.  However, in acoustics it may be
possible.

Hinds:  I would tend to agree.  The situation of mechanically generating
turbulence gives a more intense turbulence than you would find in a normal
turbulent flow.

Yang:   Another note on the relation between the frequency of excitation and
the size that you have.  In turbulence in the case of a liquid in a tunnel
you have an additional mechanism, a Reynolds stability.  You can stretch the
filaments and they will break up.  It depends on the size and adhesion force.

-------
NEW APPLICATIONS OF ACOUSTIC AGGLOMERATORS IN PARTICIPATE
EMISSION CONTROL

    D. T. Shaw and James Wegrzyn

    Laboratory for Power and Environmental Studies, State
    University of New York at Buffalo

INTRODUCTION

    In order to meet the national goal of energy self-independence,
expanded research and development on coal-fired and nuclear power
generating plants have become a national priority.  Numerous new
programs have been established by various Federal agencies to seek
answers to problems associated with how to generate electric power
with low-grade (high sulfur) coal, to improve power plant efficiency,
to increase our capacity in nuclear power generation, and above all
to accomplish all these with minimum environmental degradation.
These new programs lead to the new engineering problems such as the
control or suppression of airborne dust particles in very unconven-
tional situations - the combination of high temperature, pressure,
or corrosive environment - in which the conventional air-pollution
abatement equipment cannot function effectively.  In many of these
situations it appears that acoustic agglomerators possess certain
characteristics which seem very attractive for dust-particle control
in several of these new programs.  These will be discussed first in
the following sections, followed by the descriptions of some large
sonic generators which may be suitable for these new applications.

    Acoustic agglomeration is not a new idea.  As early as 1931,
Patterson and Cadwood [1] studied the effect of sound on aerosol
dispersion.   A great many papers appeared in the literature of the
1950's and these papers were summarized in the well-known book on
sonic agglomeration by Mednikov, published in 1963 [2].   Since then
however,  there has been an almost complete vacuum in which little
interest  has been shown on the sonic agglomeration of aerosol par-
ticles.  Shirokova reviewed acoustic agglomeration in 1973 with
virtually no new experimental data.   A series of papers were pub-
lished in the USSR concerning the theoretical treatment of aerosol
behavior  in  an acoustic field [3-10].   A group at the University of
Ontario proposed the idea of using an acoustic field to enhance
dust collection by filters [11].  At about the same time Scott and
his associates worked on the idea of using a progressive wave for
industrial aerosol emission control  [12-15].   Their work shows that
when properly designed, acoustic agglomerators can be economically
competitive  with more conventional devices for special  appications

-------
                                                                   65
in which conventional  devices are not effective.

    It is important to emphasize here that sonic  agglomerators
cannot compete in applications where the conventional devices
already have been successfully used.  Thus,  in a  recent report
by the Midwest Research Institute [16], it is found that sonic
ajglomerators are too  expensive for conventional  pollution control
because of the high power consumption in the generation of strong
acoustic fields.

    However, the  sonic agglomerators have some characteristics-
which make them uniquely suitable for certain special applications.
These include the acoustic conditioning of fine particulate emis-
sions (particle diameter between 0.2-2ym) the suppression of radio-
active sodium-fire aerosol in hypothetical Liquid Metal Fast Breeder
Reactor (LMFBR) accidents, and the control of particulate emission
under special conditions where conventional  abatement techniques
are not applicable.  The objectives of this  paper are to describe
these new acoustic applications and to identify the predominant
sonic agglomeration mechanisms for each application,

LMFBR SODIUM-FIRE AEROSOLS

    In the development of the Liquid Metal Fast Breeder Reactor
(LMFBR), a major  safety problem concerns the suppression of the
radioactive sodium-oxide aerosol particulates produced in a hypo-
thetical core-disruptive accident.  Assuming that a small fraction
of the airborne sodium aerosol leaks into the environment, radio-
logical analysis  of the site boundary does show  that inhalation
of aerosol particles containing plutonium oxides  may be a major
concern.  This is different from the problem associated with the
Light Water Reactors (WR) in which the airborne thyroid doses
from radioiodine  is of principle concern.  To minimize the leaking
of these radioactive aerosols into the environment through small
cracks, pipe fittings  and so  on, the effective suppression of
these aerosol particles inside the containment building in case
of such an accident is probably the highest  priority in the devel-
opment of LMFBR emergency air cleaning systems.

    This so-called sodium-fire aerosol is produced from the combus-
tion of liquid-metal sodium coolant in air.   Sodium combustion can
take place either rapdily through, for example, a ruptured pressur-
ized pipe; or from a slow spill of sodium on to the reactor floor.
In the former case it  is called a spray fire, and in the later case
a pool fire.  A review covering both types of these fires can be
found in the literature [17].

-------
66
     The  sodium fire aerosol which  consists primarily  of  sodium
 peroxides,  have diameters predominantly in the range  of  2-20um.
 Owing to  the  constant thermal  convective stirring,  these relatively
 large particles stay airborne  for  a  long period of  time, say two
 or three  davs,  with a very high  total  mass-loading  (in the  range
 of 1-50  g/m4).   In the recent  report  entitled, "Evaluation  of Air
 Cleaning  Concepts  for Emergency  Use  in the LMFBR Plants",  twenty-
 four different  air cleaning systems  were investigated and  these
 are listed  in  Table 1.  Among  these,  direct in-containment  acoustic
 agglomeration  is  rated to be number  one in terms of effectiveness,
 and number  four in terms of fabrication, as shown in  Table  2.
                              TABLE  1
System
Number
SR-1
SR-2
SR-3
SR-4
SR-5
SR-6
SR-7
SR-8
SR-9
SR-10
SR-11
SR-12
SR-1 3
Description
Prefilter-HEPA
Deep Bed Graded Media
Sand Bed
Bag Filter
Cyclone
Cyclone plus HEPA
Wet Scrubber
Fluidized Bed
Acoustic plus Cyclone
Mechanical Separator
Settling Chamber
Electrostatic Precipator
Steam Conditioning
System
Number
SD-14
SD-15
SD-16
SD-17
SD-18
SC-19
SC-20

CF-21A
CF-21B
CF-22
CF-23
CF-24
                                            Description
                                          Liquid Sprays

                                          Powder Discharge

                                          Foam Encapsulation

                                          Acoustic Agglomerator

                                          Direct Electro. Precip.

                                          Comb. Acoustic plus Cyclone

                                          Comb. Powder plus Filter



                                          Large HEPA, Charcoal HEPA

                                          Small HEPA, Charcoal, HEPA

                                          Wet Scrubber, HEPA, Char.,HEPA

                                          Elec.Precip.,HEPA, Char.,HEPA

                                          Sand Bed, HEPA, Char., HEPA

-------
                                                                    67
                              TABLE  2

        RANKING OF TWENTY EACS CANDIDATES BY TOTAL SCORE AND BY
              CRITERIA GROUPS - SINGLE  CONTAINMENT PLANT
Overall
Rank
9

14

11
SD-17

SD-15

Sc-19
Total Ranking by Criteria
System
In-con tainment
Acoustic
In-Containment
Powder
In-Cont. Acoustic
Score A B
309.5 1 7

296.4 2 16

304.6 3 8
C
15

11

16
D
16

18

rs
Group
E
16

14

11

F
4

6

9
             +Recirc.Cyc
13
16

SD-14 Liquid Spray
SC-20 Powder + Reef re.
HEPA
296.8
288.7

4
5

10
18

19
14

11
17

13
3

1
14

* A- Effectiveness, B- Reliability, C- Compatibility, D- Credibility,
     E- Flexibility, F- Fabrication
This suggests that  if  R  §  D effort can improve the credibility,
reliability and  compatibility of this concept,.acoustic  agglom-
eration may be one  of  the  top contenders for such emergency
application.

CONTROL OF  FINE-PARTICULATE EMISSIONS FOR STATIONARY SOURCES

    The use of mass weight is an acceptable practice in  particu-
late emission control  since the current regulations are  completely
based on emission mass.   Missing from the current air  quality
criteria is the  particles  sizes, which is probably the most  impor-
tant parameter in the  discussion of environmental effects  of
particulate emissions.   Large particles (diameter larger than 2pm)
constituted the  major  mass fraction of most emissions  sources.
Although they represent  a  significant public welfare problem  by
contributing to  the soiling and unsightly smoke plumes,  these large
particles are not a major  health problem since inherently  they tend
to deposit  in the upper  airways of the human respiratory system.

-------
68
 It is the fine particles (diameter less than 2ym)  that usually
 penetrate very deeply into the pulmonary airways and cause perman-
 ent health effects.   Furthermore,  these fine particles,  when
 released into the atmosphere,  stay airborne for an extended period;
 in contrast to the large particles,  most of which  are quickly
 brought back to the  earth's surface  by either direct fallout or
 by precipitation scavaging.

     Large particles  can be effectively controlled  by conventional
 equipment such as cyclones, filters,  scrubbers, electrostatic  pre-
 cipitators,  etc.   However,  these devices have relatively poor  effi-
 ciencies for the control of fine particles  with diameters ranging
 from 0.2-2um.

     Since acoustic agglomeration has  as its principle advantage  the
 ability to control particles  of any  size without a minimum in  the
 collection efficiency,  it  is  potentially attractive for  use as a
 preconditioner of the conventional  emission control devices for  the
 effective removal of fine  particles.   In other  words,  acoustic
 agglomerators  offer  distinct  advantages in  hybrid-unit applications
 in which the acoustic agglomerator  is  used  upstream from the con-
 ventional devices.   In  this way acoustic agglomerators  can be
 used to precondition the size  distribution  of the  particles which
 permits a more effective emission  control of fine  particles.   The
 purpose of the acoustic  preconditioner is to shift  the maximum of
 the dust particle size  distribution to a larger particle size,
 leading to a more effective removal of the  submicron  particles by
 electrostatic  precipitation.


















«o«ti         ° """"."""P""" for the  production of the

                                             "- ^ -»•< «.„,».

-------
                                                                      69
EMISSION CONTROL  OF  A FLUIDIZED
  STEAM TURBINE COMBINED CYCLE
BED BOILER WITH  A  GAS  TURBINE-
    To meet  the  requirement of producing  adequate amounts of  elec-
tric power with  a  minimum degradation  of  the environment, it  has
been increasingly  important to find  a  way of burning the high-sulfur
coal and at  the  same time, to remove SO-  from the flue gas  before
release into  the atmosphere.  Among  the concepts being developed,
the idea of  fluidized-bed combustion is rapdily gaining favor.   In
this case the high-sulfur coal is  injected and burns in the bed
at a temperature of 1450-1750°F.   The  S02 released during the burn-
ing of the coal  is reacted with lime and  in the presence of excess
oxygen in the bed  forms calcium sulphate.  Fresh, crushed high-
calcium limestone  (dolomite limestone)  is injected continuously
into the bed  and the sulphate lime  is  removed continuously.

    Several  concepts of the fluidized  in-bed combustors have  been
considered for power plant applications  [19-22].  Among these the
ideas of using a pressurized, fluidized bed combustor in a  combined
gas-turbine  and  steam-turbine cycle  is  particularly attractive,  as
demonstrated  in  Fig. 1.  Steam for  operating a turbine is raised  by
                                                        Slack
                   \ Disposal/Utilization
                    or Regeneration
                     (from each Lied)
                                              '•e.g. spent tfolomi'.e (CaSO.j-M5G.
                                                CaC03- MgO)
                            H33t Recovery Unit
                             (Flue Gas)
                                       Boiler Fe&l-
                                       W3ler Pump
                  Figure 1 -Pressurized fluuiizr] lied toiler - power pknt sclitwlSc

-------
70

both the steam tubes immersed in the fluidized bed and by the heat
converter boiler of the gas turbine.  The heat of combustion is
removed from the bed to run directly to the gas-turbine.  Typically
such a fluidized bed uses crushed coal of about 1/4" diameter.  Most
of the fly-a-sh particulate particles are rather large, say lOOym;
however, some smaller particles are also produced which must be
removed from the flue gas to avoid the attack of gas-turbine blades
by such corrosive particulates as sodium salts.

       Cyclones are used to effectively remove large particles.  For
particles in the range of 2-20um diameter non-conventional devices
are suitable for this application.  Filters are difficult to use
because of the high temperature, corrosive environment and because
of the inherent high-pressure drop across the filters.  The electro-
static precipitator is not applicable because under the high pressure
(10-20 Atm in the fluidized-bed boiler) it is difficult to produce
the corona discharge without generating excessive electric arcs.
Again, the use of an acoustic agglomerator, operating either in
standing-wave or traveling-wave mode, appears very attractive.

SOLID-SEED EXTRACTIONS FROM MHD GENERATORS

       Because of the increased cost and reduced availability of
fossil fuels, a strong incentive now exists to substantially in-
crease the electric power plant efficiency which is currently run-
ning between 33-40% depending on whether it is a nuclear or conven-
tional fossil-fuel plant.   Since 1960 new plants have not done
appreciably better in terms of the power plant efficiency.  This
comes primarily from inherent upper temperature limits of about
1050°F imposed on the steam-turbine cycle because of some inherent
thermal and material problems.  Alloy steels lose strength rapidly
above this temperature.   Furthermore, water begins to decompose
at this temperature and it attacks and weakens the boiler tube
walls.  Consequently unless a new thermodynamic unit is developed
we have reached a point of diminishing returns in our attempt to
further increase the thermal efficiency of central power plants.

       Magnetohydrodynamic (MHD) topping units have been developed
as a way to increase the peak temperature and the efficiency of
electric power stations [23-25].  A schematic diagram is shown  in
Fig.  2 to illustrate the important components that have to be
developed for a commercial unit of this type [23].  One of the  key
technical problems in the development of such a coal-fired MHD
powered generator involved the separation and the efficient removal
of the seed material, which is usually either sodium or potassium,
before releasing the flue gas into the atmosphere.  Again, because
of the high temperature and the excessively corrosive environment,

-------
                                                                      71
                    T I I
it
              Fig.2  Schematic diagram of important MHD generator components
                    (Taken from ref. [23])

conventional  devices are difficult  to  use  in  this situation.   A
recent study  shows  that cyclones can be used  to remove most  of  the
large particles  (diameter larger than  lOym).   New techniques must
be found to collect particles in smaller sizes.  Particles are
mostly spherical because of the very high operating temperature at
which MHD generators perform, typically between 2000-2500°C. Figure
3 shows typical  agglomerates collected  from cyclones from a  MHD
generator [23] .
      Fig. 3  Typical agglomerates collected from cyclones from a MHD generator
                (Reference [23])

-------
72
 THEORETICAL ASPECTS

         As discussed  in the preceding  section,  the  interest  in
 various applications  of acoustic  agglomerators  concerns  partic'les
 in the range from 0.2- to about 20pm.   Particles  smaller than 0.2ym
 usually coagulate quickly into a  larger  size; while  particles larger
 than 20um can be effectively removed by  mechanical  devices  such as
 cyclones, scrubbers,  etc.  For particles  of  this  size  region acoustic
 agglomeration offers  some unique  characteristics  as  will be  explained
 later.  Figure 4 shows the variations  of  the  relative  amplitudes of
 the oscillating particles to gas  with  respect to  particle radius for
 various acoustic frequencies.  On the  top  of  the  figure  the  oscilla-
 tion of the particles is also plotted  in  terms  of the  parameter T/T.
2 3 4 6 8 i   Z  3
     Particle radius
                                           6 e 10  20
        Fig.  4  Variation of relative particle amplitudes vs. particle radius
                   for various particle frequencies
 It is seen that the particles vibrate virtually with  the  same
 amplitude as the carrier gas at low frequencies.  As  the  sonic
 frequencies increase  the inertia of the particle becomes  increas-
 ingly evident until eventually a frequency is reached  in  which  the
 particles remain virtually stationary despite the intense  vibration
 of the carrier gas.  In the intermediate frequencies  the  particles
 precipitate in the vibrations to varying degrees, depending  upon  the
 sizes.  For example, if the frequency of the acoustic  field  is
 20 Kc/sec, particles with diameters smaller than 2um  vibrate with
 nearly full amplitude (Xp/Xg = 1), while particles with diameter
 larger than 4um are virtually stable (Xp/Xg= 0).  This differential
 motion of vibrating particles leads to the agglomeration  of  particles

-------
                                                                     73
     Since the inertia of a particle can be conveniently expressed
in terms of T, which can be expressed as:

       .!a*                                      a)
where p  is the particle density, n is the dynamic viscosity of the
carrierp gas and a is the particle radius.  T = 2iT/u= 1/f where w
and f are the angular velocity and frequency of the acoustic field,
respectively.  T/T characterizes the relative effectiveness of the
inertial force during one oscillating cycle.  When T/T » 1 the
inertia is too large for the particle to respond to the acoustis
oscillation.  Thus, the particle can be considered to be relatively
stationary.  When T/T « 1.0, the inertia is negligible and the
particle is oscillating together with the carrier gas.  The degree
to which a particle is carried along with the oscillating motion is
specified by an entrainment factor, y  which is defined by:

          X         ,     _ ,   u
      u  =/= [1 + 4/ (t/T)Y = /                    (2)
           8                    g
where Xp and Xg are the oscillating amplitudes of particles and the
carrier gas, respectively.  Equation  (2) is plotted in Fig. 4.

     Owing to the differential motions, smaller particles are agglom
erated into larger ones and the time  rate of small particle concen-
tration can be written as:
       dt
Where n_  is the small particle concentration.   Integrating  Eq.(3)
and using the  initial conditions n-- at t  = 0,  we have
                                   --


               -K t
      n2 ' n20 e

-------
 74

 where  Kfl  is  defined  as  the  acpustic  agglomeration  constant  given
 by

        K& = a e TT(a]L + a2)2(2Xp)n1f                         (5)

 where  a is  a  filling-up  factor,  n  is  the  concentration of  large
 particles which  act  as  collectors  of small  particles,  Xp is  defined
 in Eq.(2),  al and a2 are  the  radius  of large  and small  particles,
 respectively.  e is  defined as the particles  collision  efficiency.
 To understand the acoustic  agglomeration mechanisms  it  is necessary
 to describe  briefly  the  theoretical  predictions of the  values  of  e.
 The particles of diameter range  0.2-20ym,  the  calculation of e may
 be divided  into  two  subregions.  The first  region  is  for particles
 of diameter  range from  0.2  to about  2um.   For  these  particles  the
 predominant  mechanism for acoustic agglomeration is  the so-called
 ortho-kinetic interaction.  Basically,  orthokinetic  interaction
 occurs  between two particles  of  different  sizes which are moving
 at different velocities  according to Eq.(2).   Such differential
 motion  causes the two particles  to collide  with each  other,  leading
 to agglomeration into one larger particle.  For'this  interaction  the
 particle efficiency e can be  expressed  as  [26,27]:

          e  = K*  (Ke + 0.25)-2                               (fi)
 where          2u a
Typical results of Eq. (6) are shown in Fig. 5.  For particles of
diameter range from 2-20 the predominant mechanism for acoustic
agglomeration is the so-called hydrodynamic interaction.  In this
case particles interact with each other due to the hydrodynamic
interaction forces resulting from the disturbance of the flow field
of one particle to another.  It is important  to point out that
the orthokinetic interaction is not important in this region because
particles larger than 20ym seldom exist in any of the applications
as discussed in this paper.  For this mechanism the particle colli-
sion efficiency e can be determined by using the following expression
for

-------
                                                                            75
 .     _   1.587  . 32.73  .
where B = 	+ —^— +
                               344
                           I      ,1.56
                          a (20/a )
                           e     e
       a -10      a -10
exp  - — - sin  — -
        15          63
                                                                    (7)
          (2u
     p  • and p   are constants such that y = 0 when 1 = 0 and y  = 1/4
     •L      £                          C                    C

     when 1=1 and are determined by iteration for each a .
                       10*
Fig. 5  6 (ratio of'acoustic agglomeration rate to the agglomeration rate
         without acoustic field) vs. particle-radius ratio for various
         acoustic frequencies  (in kHz) for large particle radius  (urn)

-------
76
 Typical results of Eq.(7) are shown  in  Fig.  6.   For  both  types
 of interactions a very large increase in  the  agglomeration  rate
 as compared with the conventional gravitational  coagulation can
 be expected.  However, it is important  to point  out  that  the
 theoretical prediction in both Figs. 5  and 6  are  likely to  be too
 optimistic since in both figures the filling-up  factor a  (as shown
 in Eq.5) is assumed to be one.   In practical  cases a is less than
                   10'
                   10
                     8
                     6
                     4  68io'
6*,o°
       Fig. 6 6 vs. small particle radius for various acoustic frequencies
              (a = 5y, j = lw/cm2)
 one.   The  actual  value depends mostly on the acoustic turbulent
 strength.   No  information,  neither experimental nor theoretical,
 is  available on a at  the present time.  It should be noted also
 that  for control  of small  particles (diameters less than 2ym) a
 high  frequency acoustic field is much more effective; while for
 large  particles  (diameter  between 2-20  m) , a low frequency field
 is  much more effective.

-------
                                                                   77
    The rate of  reduction,  as  indicated  in  Eq.(3)  depends  on  the
 square of particle  concentration.   Thus  the  higher the  particle
 concentration  the more  effective  is  the  acoustic  agglomeration.
 For the LMFBR  applications,  typical  mass  loading  is  high  (range
 from  l-30g/m ),  thus  acoustic  agglomeration  is  expected to be
 very  effective.  For  other  applications  in  which  the particle
 number concentrations are not  as  high  as  in  the case of MHD and
 fluidized bed  boilers,  it may  be  necessary  to  add a  mist  to bring
 the particle loading  to a satisfactory level.   However, the mist
 must  be chemically  and  thermally  stable  under  the high  temperature
 and high pressure environment,  such  that  the mist droplets will
 be removed  completely by the cyclone after  leaving the  acoustic
 agglomerate.   The required  acoustic  intensity  depends on  the
 requirements on  particle agglomeration time, since the  acoustic
 coagulation constant  K   is  directly  proportional  to  the square
 root  of acoustic intensity.

 ACKNOWLEDGEMENT

 This  work was  supported in  part by  EPRI  and NSF.


 REFERENCES

 []]   Patterson  and W.  Cawood, Nature  127,  667 1931.

 [2]   Mednikov,  E.P.,  Acoustic  Coagulation and  Precipitation of
      Aerosols,  English  transl.  by Consultants  Bureau, New York.
[3]



[4]

[5]

[6]

[7]

[8]
Shirokova, N.L., Physical Principles of Ultrasonic Technology
2 (L.D.Rozenberg, ed.)477-539.Aerosol Coagaulation,  Plenum
Press, 1973.
Timoshenko,  V.I.,

Timoshenko,  V.I.,

Timoshenko,  V.I.,
Appl. Acoustics, 1, 173, 1968.

Appl. Acoustics, 1, 183, 1968.

Appl. Acoustics, 1, 192, 1968.
Timoshenko, V.I., Appl. Acoustics, 1, 200, 1968.

Shirokova, N. L., Interaction of Aerosol Particles in a Sound
Field (Candidate's Dissertation), Akust.Inst.Akad,Nauk USSR,
Moscow 1968.
[9]    Mednikov,  E.  P.,  Dokl.Akad.Nauk.SSSR,  183 (2):  382, 1968.
[10]   Mednikov,  E.  P.,  Akust.  Zh.,  14  (4):  582,  1968,

-------
 78
 [11]  Mguyen,  S.  and  Beeckmans,  J.M.,  J.Aerosol  Sci,  5,  133-143.
      1974.

 [12]  Scott,  D.S.,  Proceedings Symposium  on  Control  of  Fine^-Particu-.
      late  Emissions  from  Industrial Sources,  597^621.  San  Francis.co
      1974.

 [13]  Davidson,?.A. and Scott, D.S., J. Acoust.Soc,  of  Amer.  53,
      1717-1729,  1973.

 [14]  Davidson, F.A., J. Sound and Vibration 38,  475-495, 1975,

 [15]  Davidson, F.A.  and Scott,  D.S.,  J.  Aero.Sci,5,  55-69,  1974,

 [16]  Hegarty, R. and Shannon, L.J., Evaluation  of Sonics for  Fine
      Particle Control, EPA-600/2-76-001, January  1976.

 [17]  MacPherson, R.E., Report No. ORNL-TM-1937,  1968,

 [18]  Milliard, R.K., McCormack,  J.D., Postma, A.K.,  Report  No.
      TC-536, Hanford Engineering Devel.  Lab,  1975,

 [19]  Jonke, A.A., Swift,  W.M. and Vogel, G. J., Transactions AIME,
      Vol. 258, 159-167, 1975.

 [20]  Ehrlich, S.  and McCurdy, W.A.  .Proceedings 9th  Intersociety
      Energy Conversion Eng.Conf., San Francisco,   1974.

 [21]  Wright, S.J.,  Proceedings 3rd Int1.Conf.Fluidized Bed Combus-
      tion,  EPA-650/2-73-053, EPA, 1973.

 [22]  Highley, J.  Ed,  Combustion  of Coal in Fluidized Beds,  Proceed-
      ings,  Coal  Research  Establishment, May, 1968.

 [2-3]  Jackson, W.D., ERDA,  Zygielbaum,  P.S.  EPRI, et al,, MHD Power
      Generation   769176,  Status  Report, llth Intersoc.  Energy
      Conversion  Eng.  Conf.,  Sept.,  1976.

 [24]  Office of Coal Research Report  #71,  prepared by MHD Power
      Generation  Study Group, 1972.

[25]  Dicks, J.B.  et al.,  Proceedings,  llth Intersoc. Energy Con-
      version Eng. Conf.  769177,  Sept., 1976.

[26]  Fuchs, N.A., The Mechanics  of  Aerosols, trans,  by R.E. Daisley
      and  M. Fuchs,  Pergamon  Press,  New York, 1964.

[27]  Hidy,  P.M.  and Brock, J.R., Dynamics of Aerocolloidal  Systems,
      Vol.  1,  Pergamon Press, 1970.

-------
                                                                              79
WRITTEN DISCUSSION

W.M. Swift
Department of Chemical Engineering
Argonne National Laboratory
9700 South Cass Avenue
Argonne, Illinois  60439

    Argonne National Laboratory's interest in controlling particulate emissions
with acoustic agglomerators arises from its participation in a research and
development program designed to develop the concept of fluidized bed com-
bustion (FBC) and to eventually demonstrate on a plant scale that the process
is economical and meets pollution standards for stationary power sources.
Simply stated, fluidized bed combustion is a process in which coal is burned
with excess air in a fluidized bed of limestone or dolomite at temperatures
of 800 to 950°C.  The heat of combustion is used to generate steam for process
use or to generate electrical power.  Fluidized bed combustion, which results
in both reduced SO- emissions (by reaction of the SO- with CaO in the fluidized
bed) and reduced NO  emissions (by reason of the low combustion temperatures),
is potentially bothxenergy efficient and economically attractive.

    In advanced concepts of FBC, the combustor is operated at elevated pressure
(e.g., 10 atm).  Energy is extracted from the process by raising steam in
boiler tubes and by expanding the flue gas through a gas turbine to take
advantage of the potentially higher thermal efficiencies offered by a gas
turbine-steam turbine combined cycle.  Use of a gas turbine, however, requires
that the particulate loading of the high temperature-high pressure flue gas
from the pressurized FBC  (PFBC) be reduced to very low levels.  Estimates  of
acceptable dust loading for turbine operation range from 0.05 to 0.0002 grains/
scf.

    Until recently, the problem of high temperature-high pressure flue gas
cleaning  (particularly as  it relates to PFBC) was not adequately addressed.
At  present, however, a fairly intensive effort  is being made to develop
viable  techniques  for sufficient cleaning of  flue gas from a pressurized  fluid-
ized bed  combustor  to allow its use in reliable gas turbine operation.

    Currently,  ANL  is conducting basic supportive studies in fluidized bed
combustion.   One  task in  our program is to  test and evaluate novel  concepts
 for high  temperature-high pressure  flue gas cleaning  in our process  develop-
ment  facility,  which  includes a  6-in.-dia. fluidized bed combustor which  can
be  operated  at  pressures  up  to 10  atm.  Initially,  a review was made of  the
 existing  and  developing particle-gas separation technologies to  identify
 concepts  which  would  be amenable  to  the high  temperature-high  pressure  require-

-------
  80

 ments of pressurized FBC.    One area identified for investigation was
 acoustic agglomeration.   As a result, we entered into a working relation-
 ship with Dr.  David Scott of the University of Toronto, who is also present-
 ing  a paper on acoustic  agglomeration at this workshop.  It is our intention
 to investigate the effect of acoustic dust conditioning on the efficiency of
 particulate removal devices in the flue gas system of the ANL experimental
 combustor.

     Although the agglomerating influence of an acoustic field on an aerosol
 has  been known for slightly over 100 years, the development of acoustic
 agglomeration  (AA) as an aid in the control of particulate aerosols has  passed
 through cycles varying from enthusiasm to almost total disinterest.   When the
 electrostatic  precipitator won the performance/cost war in about 1952,
 interest in acoustic agglomerators in the U.S.  essentially went to zero.   At
 the  same time,  however,  interest in AA was generated in the U.S.S.R., Hungary,
 Poland,  and Japan.   This  interest in foreign countries has continued to  the
 present.  As a result, the U.S.  has been left behind in both developing  this
 technology  and maintaining the necessary scientific interest.

     Some of the factors  which led  to the premature  disinterest in AA were:
 (1)  a poor  understanding of the aerosol mechanics of AA,  (2)  less than
 optimum  configurations in  the designs originally developed for commercial
 application, (3)  a demonstrated  superiority of  electrostatic  precipitators,
 (4)  concern over  the high  power requirements for the generation of the acoustic
 fieldl, and  (5)  potentially high capital cost for acoustic  agglomeration
 systems.

     In the  paper  by  Shaw and  Wegrzyn,  the  argument  is  made that  sonic
 agglomerators  cannot  compete  in  applications where  conventional  devices
 already  have been successfully used,  but  that they  have  some  characteristics
 which make  them uniquely suitable  for certain special  applications.  The
 phrase "uniquely  suitable" must not  be interpreted to  imply  a priori that
 sonic agglomeration  is necessarily the  most  suitable control  technique for
 these special applications.   In  the  example  of  the  LMFBR sodium-fire aerosol,
 although  the acoustic control technique was  ranked  first in effectiveness,
 it was ranked ninth  in overall preference  due to other considerations.  Also
 although  acoustic  agglomeration appears attractive  in  controlling emissions
 from a fluidized-bed boiler to protect  a gas  turbine in  a  gas  turbine-steam
 turbine  combined  cycle,  a  number of  alternative  (and generally more highly
 regarded) control  technologies are being investigated.   Granular-bed filters
 and ceramic filters  (either fabric or porous  sintered) are the more commonly
 mentioned alternatives being  considered for  this application.

    It is interesting to note  in the  ranking of  control methods  by criteria
 for the control of  LMFBR  sodium-fire aerosols  (Table  2  in paper by  Shaw and
Wegrzyn), one of  the problems  facing acoustic agglomeration was  that of
 credibility.  This same problem of credibility exists  in the application of
 acoustic dust conditioning to PFBC applications.  The problem of  credibility
 arises mainly from the lack of experimental correlations or experimentally
verified theory which can be used  to predict either performance or cost in a
 given application.

-------
                                                                              81
    The credibility problem is really not too surprising in view of the
complexity of the equations which govern the aerosol mechanics and the f-act
that the experimental data on acoustic agglomeration are for the most part
of no use for evaluating theory.  If AA is to gain credibility, both the
performance and the energy requirements have to be established (either by
experiment or from reliable theory) for the particular application being
considered and a performance-cost comparison made against competing
technologies.

    A good argument for determining energy requirements can be made in the
case of particulate control from pressurized fluidized-bed combustors.  For
atmospheric fluidized bed combustion, an overall plant efficiency of 36-39
percent is estimated.  In atmospheric operation, the flue gas can be adequately
cleaned using conventional technology.  First generation pressurized FBC
systems will have an estimated plant efficiency of 40-43 percent.  The small
but significant gain in plant efficiency for pressurized systems is at the cost
of considerably more complex development problems (of which cleaning of high
temperature-high pressure flue gas is one).  If cleaning of the high temper-
ature-high pressure flue gas to acceptable particle levels for turbine
operation by AA or any other method is excessively energy-intensive, the PFBC
loses much of its advantage over the atmospheric concept.

    With reference to the theoretical aspects of acoustic agglomeration as
presented by Shaw and Wegrzyn, the following comments are offered.  First,
the authors are to be commended for their  efforts to identify  the controlling
mechanisms in AA and to experimentally verify the respective governing
equations for these mechanisms.  It should be emphasized, however, that, the
agglomeration mechanisms considered by Shaw and Wegrzyn  (i.e., orthokinetic
vibration and hydrodynamic attraction) are but  two of approximately nine mechan-
isms known to have some effect on  the agglomeration kinetics.  The problem is
sufficiently complex that  the mechanisms of AA may depend upon the nature  of
the aerosol, the acoustic  field, and the system geometry.  Thus,  the  con-
clusions of  a particular study may be seemingly contradicted by  the results
of other experiments when  both  sets of data  are equally  valid  but  depict
different mechanisms under different circumstances.  The statements,  there-
fore,  by  Shaw and Wegrzyn  that  the predominant  mechanism is acoustic  agglom-
eration for  particles of diameter  range  0.2  to  2  pm is ortho-kinetic  inter-
action and  the  predominant mechanism for particles of diameter  range  2 to
20  ym  is  hydrodynamic  interaction  should not be accepted without justification
as  general  observations which can  be  applied to all  acoustic-aerosol  inter-
actions.
 REFERENCES

 [1]
      Problem." 	0	 	
                (in preparation).
LENCES

R. Razgaitis, "An Analysis of the High-Temperature Particulate Collection
Problem," Argonne National Laboratory, Argonne, 111., Report No.
ANL-77-14 (in preparation).

-------
82

 OPEN DISCUSSION

 Shaw:  I agree with what Bill said about the nine mechanisms he mentioned.
 The other seven are mostly associated with the turbulence.  The parakinetic
 is also what I term as the filling up of the space.   So I agree with what
 you are saying on the nine mechanisms.   However, I  don't think you can
 really identify the way we are approaching the problem on the mechanism
 beyond these two. We have already identified the linear case; we are
 also looking at the affect of turbulence overall. You have to define some
 parameter in terms of acoustic turbulence which I think will cover other
 mechanisms such as the parakinetic and so on,  because there is no way to
 separate the parakinetic from the turbulence under high intensity.  They are
 interrelated.

      We have done separate things in a nonlinear case,  theoretically and a
 few experiments under high intensity;  but we are far away from matching the
 theory and experiment.

 Yang:   Bill, is the work that you are just finishing available?

 Swift:   No,  it is in a  state  of  being published,  both as an Argonne report
 and an NTIS  document.

 Oder:   Could you comment on the  strength of  coagules that are formed by the
 agglomeration process?   I gather that they are  brought  together by turbulent
 acoustic  forces and they settle  out  in rather undisturbed environments.
 Suppose,  instead,  that  you try to extract these particulates  under turbulent
 conditions,  what is it  that holds them together?

 Shaw:   That  is  a very good question.   So far we have only done experiments
 with particles  of  a very small size, primarily  for the  application of fine
 particulate  control which is  supported by EPRI.   We  have not  done  any work on
 the high  concentration  yet.   We  are  expecting a contract from ERDA on the
 high intensity  agglomeration  of  large  particles for  sodium  aerosol.  The
 small particle  sticking is not a problem.  That comes when  large particles,
 sticking  together,  break apart.   In  that case,  some  kind of salient  condition
 must exist so  that  they stick together.   This is  something  that we  are just
 still speculating,  we have no experimental data to show.  The  breaking up by
 the acoustics will  be important  only when you have very  large  particles.

 Oder:   I  am  sorry,  I still didn't hear what it  is that makes  them  stick
 together.

 Shaw:   Surface molecular  forces.  Just like when you have a particle on an
 impactor, it is  the  same  force holding a  particle to an  impactors plate as
 when the  particles are  sticking  together.

 Helfritch:  You mentioned  that, most often, the use of agglomeration techniques
 applies to cyclones.  There are  two questions that have been brought up
 about the success of that  application.  The apparent specific gravity of the
 agglomerates would be quite low.   Secondly, they may break up.  I was wonder-

-------
                                                                              83
ing if it had actually been tried in a cyclone.

Shaw:  No, we have not tried it.   The only reason mentioning cyclones is
that it can only be used in applications like the Argonne system where they
have a cyclone and want to work it with something else.   Whenever you have
a situation where the pressure drops and pumping is a problem, the cyclone
seems to be the answer.  It requires very little pumping power.

Helfritch:  You mentioned that a high pressure drop across the filter is
part of the reason for projecting that technique.  However, I would submit
that the pressure drop across the filter is typically on the order of pressure
drop across the cyclones.  I cannot accept that argument.

Shaw:  If that is the case, then I am wrong.  When I talked to Argonne
people, they told me that they cannot use filters.  My impression is that the
problem with filters are high temperatures, the other is a pressure drop
problem.

Helfritch: Finally related to that is the fact that high temperature, and
pressures adversely affect performances of cyclones.  That is because of the
viscosity relationship.

Swift:  By acoustically treating the aerosol you can increase the efficiency
of  the cyclone over that of the untreated aerosol even though you have high
temperatures and pressures.

Cooper:  The problem with  the sodium aerosol is  that it clogs  the filter in
a way that is very unusual and very difficult.   So you are not  talking about
the usual few inches of pressure drop and any reasonable phase  velocity;
you are  talking about  much higher  than  that.  Second, there is  a difference
in  degrading the performance of  the cyclone by changing  its collection
efficiency cut point by  the higher temperature and pressure.   And indeed,
the degradation of  the filter is where  it just falls apart.

Grassel:  It seems  to  me  that the  work  that  is going on  now,  is just repeating
work that was done  a number of years  ago  by  the  Bureau of  Mines and  other
places  during  the war.  Yet it does not address  itself  to  the practical
problem of  implementing sonics as  a method  of  treating aerosols.  It does  not
address itself  to  the  problem of  true output,  sound  level  and size.   You
are going to have  to  give up one of  those for  the others;  sound level is what
you are attempting  to  compromise.  You  are  attempting  to use  sound  level
as a method  of  accomplishing  size and true  output.   This represents  itself
 in power.  As  the  power goes  in,  you have two  practical  problems of  application:
 containment of  the sound level  and dissipating it.

      The power necessary to agglomerate particles is very  low.  The power
 necessary to generate a field  of sufficient intensity to collect those particles
 or to agglomerate those particles is very high.

 Shaw:  What you are saying is the power attenuation in the particle is small.

-------
 84

 Grassel:  But the power put into the system is very large.

 Shaw:  You must obtain a level of 160 dB,  That is precisely the reason
 why we think a standing wave kind of thing would be good for fine particle
 control.

      In response to your comments, we are doing something like the whistle
 experiments in the 1950's.  We have to go through this phase to reactivate
 the field.  The thing we are doing different nowadays is that the particle
 sizing techniques are much more improved than the wartime.  I do not agree
 that others have done the same thing.  No one before has measured size
 distribution change in an acoustic field.  The only other related work has
 been done by a Russian who sampled the size by putting in something to collect
 the particles by sedimentation.

      What we are doing in our laboratory is trying to bring us to large
 scale testing to give us credibility and demonstrate to people that it will
 work.   If you do it right it will work.   The problem is how well it will
 work.   This is the thing we are  driving  at.   We have to go through the large
 scale testing which is nothing new.   I emphasize that the state of  our
 siren is  1949.   But we are applying  the  new sizing technology to study this
 problem.

 Benarie:   Let me take just a minute  for  this problem of the cyclone.   Perhaps
 it would  simplify concepts if  I  explained that  we  have  vortex agglomeration .
 for both  cases:  (1)   a cyclone with  a hardware  around  the vortex and  (2)
 when  there is no  hardware  around  it.  It  is  a turbulent agglomeration.

 Shaw:   In no  way  do  we think the  acoustic cyclone  works the best.   In fact,
 from  the  data, we  feel that  some kind of acoustic  thing  working  together
 with a  scrubber, with a very fine mist, for  certain applications  is best.
 With high temperatures you have  trouble maintaining that  mist.

 Scott:  All the work that was done at the Ontario Research  Foundation was
 with a  cyclone.  So  the improvements noted,  from 35 and 40% up  to 75 and 80%,
 were  done in cyclones.  It  is clear  that we want  to go that way again.
 Second, on the power,  that was what the whole project was about.  It was to
 determine  the cost of  the power.  Our evaluation has shown  that if you use
 your fuel evaluation  for the pulse jet and take the cost  of the fuel, then
 the power requirements are way down for scrubbers.  We are  talking about 25c/
 ACFM/year for power.  That is for a pulse jet, which in that case was designed
with an aerodynamic valve. I think I simply must not let  the thing go without
 saying  that cyclones have been used.   Secondly, we have concentrated on power.
That was really the whole topic of our work.  Lastly, as  far as the Bureau
of Mines is concerned, I do not think we are repeating very much.  I would
be happy to see those areas where you think we are.  I think we are doing
 things that are quite different.

Cooper:  Generally, surprisingly little power is required to remove particles
from the gas stream if you just take  the  force on to particles times the
distance they are to be moved.   In some sense we ought to keep that in mind.

-------
That is one of the advantages of electrostatic and magnetic separation as
opposed to other methods which operate on the gas which often has very much
more mass than the particles.  There is an inherent advantage to those
systems which apply the force directly to the particle rather than through
the medium of the gas to the particle.
                                                                              85

-------
 86
EFFECTS OF ELECTRIC AND ACOUSTIC FIELDS ON PARTICLE COLLISION RATES IN
AEROCOLLOIDAL SUSPENSIONS
    P. D. Scholz

    Division of Energy Engineering, University of Iowa, Iowa City, Iowa


ABSTRACT

    An experimental investigation is described that involves the use of an
a.c. electric field and a traveling sound wave field to increase the colli-
sion rate between submicron particles suspended in air.  A brief discussion
of the theoretical collision rates in the free molecule regime is also in-
cluded for the cases where each field is applied separately.  Velocity pro-
files and values of the turbulent intensity in the continuous flow coagulation
tube are presented and discussed.  In addition, the research plan is briefly
outlined.

I.  INTRODUCTION

    The work reported here is based on an experimental project that is under-
way to investigate the effects of an a.c. electric field and a traveling wave
sound field on the collision rates between suspended particles with radii
less than about 0.3 um.  We have chosen to work with particles in the sub-
micron range because of the compelling need to find ways to effectively re-
move these particles from man-made suspensions.  Although we are not dealing
with the direct removal of the particles from the parent gas, the work dis-
cussed here involves the use of two external fields to "condition" the sus-
pension by inducing particle growth of the fine particulate matter through
agglomeration or coalescence.  Following the "conditioning" process, the
suspension is envisioned to be passed through a downstream and conventional
removal device where the resulting agglomerates are to be removed.  The
induced growth is achieved by using the external fields to selectively in-
crease the relative motion and therefore, the probability of collisions be-
tween the suspended particles.

    The motion of the particles under the influence of an a.c. electric field
and a traveling sound wave may be demonstrated in the following way.  Con-
sider a charged particle with a charge*q and mobility B suspended in a gas.
The velocity of the particle in the presence of an a.c. electric field is

-------
                                                                               87
    v = q B E sin u  t,


where ID  is the angular frequency of the field and E is the electric field
vector.  Here we assume that the product of the angular frequency and the
particle relaxation time, T, is much less than unity, i.e., u T « 1.  Now
consider that a continuous train of traveling planar sound waves passes
through the suspension.  The velocity of the particle in this case is [l]

              U
        —           [sin <
        (1 +
where u  is the angular frequency of the wave, U  is the magnitude of the
oscillating velocity of the gas molecules, ty is  the phase lag between the
particle and the oscillating gas molecules, and ft is a unit vector in the
direction of propagation.  The magnitude of the particle velocity,
and the phase lag, <|>, are dependent on the product u> T, where T is dependent
on the size and mass of the particle.  For u T « 1, the particle oscillates
at nearly the same speed as the gas molecules and with very little phase dif-
ference.  On the other hand, for u T » 1, the particle is virtually sta-
tionary.  (For two different particles, one with uaT  « 1 and one with
u>ar  » 1, there is a large difference in the oscillating speeds.  Such a
differential motion is the basis for orthokinetic coagulation.)

    By superimposing the two fields, the resultant motion of the particle is
a function of the two field frequencies and magnitudes, the directions of the
two fields, and of the particle parameters, which include charge, size, and
mass.  It is this resultant motion of many particles submerged in these two
fields which can lead to an increase in the particle-particle collision rate
that is of interest here.

    It should be noted that, whereas acoustic coagulation generally requires
high frequency sources (i.e. , on the order of 10-30 kHz) to set up the dif-
ferential motion in a suspension of fine particles, the present two-field
scheme utilizes full entrainment of all the particles and hence lower sound
frequencies.  This feature has some sound transducer and power supply advan-
tages.  However, this scheme also requires that the particles be charged,
which necessitates the1 addition of some type of charging stage in the system.
It is also noted that a traveling wave sound field, as opposed to a standing
wave field, is considered here following the work of Scott [2].

    For this case of combining an electric field and a sound field to pro-
mote an increase in particle-particle collisions, there is no known theoreti-
cal expression that relates the collision rate to the field and aerosol

-------
parameters.  However, past work  involving derivations of the collision rates
for  electric field  enhanced coagulation and for acoustically induced coagula-
tion can be used as a basis for  considering a new model for the case where
both fields are applied.

     A.  Theoretical Background

        Using the particle Knudsen number to identify the usual aerosol size
regimes (see Hidy and Brock [3], for example), suspensions of particles at
standard conditions with radii in the approximate range 0.01 < R < 1 pm are
classified as transition regime  aerosols.  (Continuum regime aerosols have
particle radii larger than about 1 ym and free molecule regime aerosols have
particle radii less than about 0.01 vim.)  The particle radii of interest here
are  in the transition regime, and herein lies a problem for treating the
collision dynamics  of these particles.  Whereas, continuum mechanics is used
in treating the dynamics of continuum aerosols and kinetic theory is used in
the  free molecule regime, the transition regime is without a single, unified
theoretical base.   Extensions of continuum regime theories and of free
molecule theories into the transition regime for the thermal collision rate
are  available [3,4].  Experimental results show that in some cases these ex-
tensions are reasonably accurate [4],  However, extensions do not exist for
collision rates involving electric or acoustic fields.  Expressions are
available for the collision rate as a function of the field quantities for a
sound field in the  continuum regime (see Mednikov [1]) and for an electric
field in the free molecule regime (see Brock and Hidy [5]).  Rather than try
patching together existing theories and then extending the results into the
transition regime,  we have taken the approach of using a free molecule theory
to construct an expression for the collision rate due to these external
fields.  This approach, which is discussed in the next major section, empha-
sizes the diffusional processes  and also provides a framework for dealing
with the basic collisional dynamics of the suspended particles in the pre-
sence of the two fields.  There  is evidence that diffusional effects are
significant in sonic agglomeration of particles with radii below about 1 ym
[6].  This fact is  reason enough to justify the development and examination
of this free molecule approach.

    B.  Summary of Past Experimental Work

        There is very little experimental data available on the collision
rates of suspended  submicron sized particles subjected to electric or acous-
tic fields.  In the transition regime, Mednikov [1]  reports some experimental
results of acoustic coagulation  in DOF fogs with mean particle diameters of
about 0.28 urn.   The coagulation constant was found to be directly proportion-
al to the product of the sound pressure and the time that the sound field was
applied to the fog.   Details of  the experiment are vague and the accuracy of
the results is unknown.  The work in the past has involved aerosols that were
not very monodisperse, and hence the reported collision rates are not very size
sensitive.  In addition, many of the earlier studies involved time decay
measurements in a closed chamber and were subject to significant diffusion
loss effects to the chamber walls as the coagulation process evolved and the
samples were extracted for size analysis [6]..  Currently,  there are several

-------
                                                                               89
experimental investigations underway that should contribute significantly to
the understanding of aerosol-acoustic field coupling.  Much of this current
work is possible because of significant improvements in the equipment avail-
able to generate and analyze aerosols.  The experimental investigation re-
ported here is a product of this improved technology.

    C.  Paper Organization

        The next section, which is Section II, is a summary of a first order
derivation using kinetic theory to determine the particle-particle collision
rate as a function of the aerosol parameters and the field frequency and
magnitude for an a.c. electric field.  The derivation of the collision rate
for a traveling sound wave is also summarized.  The expressions for these
rates, along with the known expressions discussed earlier, are to be tested
during the experimental investigation, which is underway and which is de-
scribed in the third section of the paper.  Section III also contains the
results of velocity and turbulence measurements taken in the coagulation tube.
In addition, the experimental plan is briefly discussed.  Section IV consists
of a summary of the paper.

II. COLLISION RATES IN THE FREE MOLECULE REGIME

    As mentioned before, an expression for the collision rate between parti-
cles in an aerocolloidal suspension subjected to a combined electric field
and traveling sound field does not exist.  It is our desire to construct a
simplified model to describe the particle-particle collisions in this case.
Because diffusional effects are important in the collisional processes in-
volving submicron particles [6], it was decided to approach the problem using
a free molecule regime model.  However, before solving the more complicated
combined field problem, it was decided to derive the expressions for the col-
lision rates for each field separately to better understand the effect that
each field has on particle collisions.  These derivations have been completed
recently and the following is a summary of that work.  The derived rates are
to be checked with experimentally determined rates during the experimental
investigation which is underway and which is discussed in Section III of this
paper,

    A.  General Approach to Deriving  the Collision Rate in the Free
        Molecule Regime

        In  the free molecule regime,  we  treat the aerocolloidal system as a
dilute mixture of gas molecules and  two  separate, but homogeneously mixed
monodisperse aerosol species with particle radii R^  and Rj and concentrations
n- and n*,  respectively.  For each species,  the particles are considered to
be spherical and  the number density  of either particle species is much less
than  the number density  of the  gas molecules, while  the mass per particle of
either species is greater than  the mass  of a gas molecule.   (This  is equiva-
lent  to a very low mass  loading.)  In this case the  three governing Boltzmann
equations  (i.e.,  one for each type of aerosol species and one for  the gas
molecules)  are uncoupled and  the equations for  each  aerosol  species are  iden-
tical, except  for a  difference  in  subscripts.   An approximation technique  is

-------
 90
 then used to solve one of the aerosol equations  for the particle distribution
 function, say flf  after assuming the form of the distribution function for the
 suspending gas.   (The second species distribution function,  fj,  is the same,
 except for a change of subscripts.)   An expression for the collision rate is
 then determined  by evaluating the collision integral of the  products of the
 two  particle distribution functions  for a specified particle-particle force
 law.   This approach has been discussed elsewhere (see Scholz and Barber [8] ,
 for  example)  and will not be presented here.

     B.   Collision  Rate for an Aerosol with an a.c.  Electric  Field

     In this case we assume that  the  dilute mixture is spatially  uniform and
 that  the particles in each species have an average charge  per particle of
 qitand q j ,  respectively.   The acceleration due to the externally applied
 electric field is
           r
     -r = m~  (E COS
where mr is the particle mass of each aerosol species and E cos o)et is the
magnitude of the z directed, time varying electric field.  In addition, the
gas is assumed to obey a Maxwellian velocity distribution function.

        Following the procedure outlined by Scholz and Barber [8], the Boltz
mann equation for the itn aerosol species may be solved by an approximation
scheme using an expansion with a weighting function equal to some zero-order
approximation of f^.  In this work the zero-order function is taken as the
Maxwellian distribution function.  In addition, the particle-gas collisions
are assumed to obey an inverse fifth power central force law (i.e., the
Maxwell force law).  Using this approach, the approximate particle distribu-
tion function for a three term expansion is
fi ~ Do
                                         \ B2[sin2(o>et -
where
                                                    - tan
                                                          1
                  2ir]T'  -
and where (fi)^ is the Maxwellian function for the i**1 aerosol species.  In
these expressions n^ and m^ are the aerosol particle concentration and mass
for species i, n  and m  are similar quantities for the gas molecules, v
is the particle speed in the z direction (and is a variable) , k is the
Boltzmann constant, T is the absolute temperature of the suspension, AI is a
collisional constant for the particle-gas collisions and equals 0.422,  K is
the inverse fifth power force constant and DQ is an arbitrary constant.  As

-------
                                                                               91
mentioned earlier, the expression for f.  is identical except the i subscripts
are replaced by j.

        It should be noted that this expression for f^ differs from the dis-
tribution function used by Brock and Hidy [5] to determine an expression for
the electric field induced collision rate.  In that work, they expressed the
effect of the electric field as an equivalent drift term.  This term was
then included in the argument of the exponential velocity term of a Max-
wellian velocity distribution function used to construct fi-  The approach
here is more fundamental for we have accounted directly for the effects of
gas-particle collisions in the collision integral when we solved the Boltz-
mann equation for f..

        As mentioned previously, the collision rate is determined by evaluat-
ing the collision integral of the products of the i and j particle distribu-
tion functions for a specified particle-particle force law.  Following Hidy
and Brock [3], the specified force law for the particle-particle collisions
is based on a hard sphere collision model with a Debye-Huckel shielded elec-
trostatic potential acting between the colliding and charged spheres.  The
collision rate, LE, for the special case where the two species are the same
size (i.e., R^ =  R..) and have the same, but opposite average charge per
particle is




where L  is the  self-thermal collision rate and is

                 R2n2   /kT//2
 In  this expression n  is the fraction of  particles  that  stick upon  contact
 and J2 is the charged  particle interaction parameter  and is

              q2  exp(- 2R /A )
      fl - 1 +
                 16ir e


 where e is the dielectric constant of the aerosol,  and
 is the Debye radius.  This case where R± = Rj  and q± = -qj  corresponds to a
 quasi-neutral, monodisperse aerosol and might  represent an initially neutral,
 monodisperse aerosol that is passed through a  bipolar charger in which com-
 plete charging takes place.

-------
 92
     C.   Collision Rate  for an Aerosol with a Traveling Sound Wave

         In  this  case we consider planar sound waves propagating in the posi-
 tive z  direction with a speed of c and an angular frequency of u>a.  The gas
 is  assumed  to obey  the  Wang-Chang and Uhlenbeck  [9] distribution function for
 small amplitude  sound waves propagating in a monatomic gas.  That is,
                     v     5v
            1 + a   1-^ + —^- I sinui
                 oo \ c     3(;2 /     a
                                      M
In  this expression a00 is an arbitrary constant, the g subscript refers to
the gas, and  (fg)M is a Maxwellian distribution for the gas.  The arbitrary
constant, aQO, is evaluated here by setting the z component of the average
speed of the  gas at the piston face (i.e. , at z = 0) equal to the oscillating
piston speed, which, in the absence of attenuation effects, is also the mag-
nitude_of the oscillating gas speed, Ug.  That is, vgz(z = 0) = Ugsinu)at,
where vgz represents the z component  of  the gas speed averaged  over
velocity space.  Although the Wang-Chang  and Uhlenbeck distriubtion function
for the gas is limited to small amplitude oscillations, the function does
appear to have the correct structure since the moments of the distribution
function yield reasonable expressions for the density, pressure, and temper-
ature of the  gas.  The structure of fg is important to this development of
determining f^ since the effect of the sound wave on the suspended particles
is  through collisions of the particles with the gas molecules; that is,
through the collision integral in the Boltzmann equation for the ic^   parti-
cles.  (This  indirect coupling occurs because the mass loading for the parti-
cles has been taken to be very low.)  As  in the electric field case, the
particle-gas  collisions are modeled with an inverse fifth power force law.
Again, using  an approximation technique with a three term expansion to solve
the Boltzmann equation, the particle distribution function is
fi ~
7V" + 5V
izJ [
                              f
                                             cos(ocosB -
where
a = 1 -
                      ,2  2   2C,c[£ + atan(acos6 - x)]
                      t- ~A  ,    JL
                       2A
                                      2    2
                   2    2    2C.c[C + atan(ocos9-
                  C  — A       JL
                     2A'
                        22    2
                      UA CC  + 0r )
                  0) C
                   a
          fir2   25 A2
          oc  - -T— A

             2     2~~
            c  - 3A
                       C-A
                        2
                                           21      ?   2S  2
                                    c  - 3A) -
                                                       ,  and

-------
                                                                              93
                                     2
In these equations 6 = ioa(t - z/c), A  = kT/m.p DI is an arbitrary constant,
and as before (f^^ is the Maxwellian function for the ith aerosol species.
A2 is another collisional constant for the particle-gas collisions and equals
0.436 and Dj^ is an arbitrary constant, as is the phase angle x-  The other
quantities were identified earlier.  As before, the expression for fj is
identical to f^ with a change of subscripts from i to j.

        One property of this particle distribution function is that it exhib-
its a drift phenomenon due to the expUcosS) term.  This stems from the sin9
factor in the gas distribution function, which reflects itself through the
collision integral of the Boltzmann equation.  This factor subsequently be-
comes a coefficient in the set of partial differential equations that are
solved for the expansion coefficients of f..

        As before, the collision rate is determined by evaluating the colli-
sion integral for i-j particle collisions.  We again assume that each species
is charged.  Using the hard sphere Debye-Hiickel shielded potential model to
account for the interaction of the charged particles during collisions, the
collision rate, LA, for the case where the two species are the same size
(i.e., R.£ = R.J) and have the same, but opposite average charge per particle
is           J

          7          2                             2
    LA = D,(0.23776 ^r - 0.99420)  [exp(2?cos0)][cos (ocosG)] L_,
     A    1         A2                                        T

where all of the terms have been introduced before.  (This collision rate
may also be used for uncharged, hard sphere particles by setting q^_ = 0 in
the charged particle interaction parameter, Q.)

    D.  The Behavior of LE/LT and L./L-

        A plot of the maximum ratio of the self-collision rates, LE/LT
and LA/LT» is given in Figures 1 and 2 for R£  = 10~2 ym sized particles.  The
arbitrary constant, D2,, in the electric field  case is eliminated by setting
the ratio LE/LT to unity at zero electric field.  The constant, D2, in the
acoustic case is eliminated by normalizing all the values of the ratio L^/Lf
with the value of the ratio evaluated for Ug = 0, which is taken to corre-
spond to the state where collisions result from thermal motion only.  The
maximum values are plotted, instead of the time averaged values, because in
both cases all points in the suspension will experience the total fluctua-
tions of each field.

        The curves in Figure 1 indicate that at a frequency of lO^Hz the
self-collision rates increase when the magnitude of the electric field in-

-------
94
                    FREQUENCY OF EACH
                    FIELD  IS 103 Hz
     102


     I
                       MAGNITUDE OF THE A-C ELECTRIC FIELD (v/ro)
                        1 - 1 - 1 — I  I  I I  I I _ l     l   l   l  l
104


J
                10'
                                          10
                     MAGNITUDE OF THE ACOUSTIC PISTON SPEED (m/sec)
                                                          10'
                   FIG (1) SELF-COLLISION RATES AS A  FUNCTION  OF FIELD
                           MAGNITUDES FOR Rj = 10'2 pm PARTICLES
              18
              16
 I-
 <
_l
(t
o
 k-
_l
 UJ
             1 4
             1.2
             1 0 -
               10J

                L
                                              LE/LT FOR E = 104 v/m
                    LA/LT FOR Ug = 1 m/sec
                       _L
            FREQUENCY OF THE A-C ELECTRIC FIELD (Hz)
           -I - 1 - 1  I   I I  I I  I - 1 _ I _ I  I
                                                         io5
                                                               l  l  I i
    10<                         io3
             FREQUENCY OF THE ACOUSTIC WAVE (Hz)
                                                                    10"
                FIG (2) SELF -COLLISION RATES AS A FUNCTION OF FREQUENCY
                   FOR R, = 10
                                       'z
                                            PARTICLES

-------
creases above about 10 V/m or when the piston speed increases above about
0.5 m/sec.  At room temperature the mean thermal speed of 0.01 urn sized
particles is about 1 m/sec and, for a particle charge of one electron, the
maximum speed of the particles in an electric field of strength lO^V/m is
about 6 m/sec.  Then, as expected, the values of the collision rates in-
crease above the thermal rates only when the directed motion due to the ex-
ternal fields is on the order of, or larger than the mean thermal speed of
the particles.  The curves in Figure 2 show that as the field frequencies
increase, the ratios of the collision rates decrease and approach unity.
This effect is due to the inertial lag of the particles which increases as
the frequency of oscillations increases.

        At this point, the simple kinetic model based on a dilute model of
hard sphere particles suspended in a gas with the two external fields seems
to be qualitatively correct.  (The collision rates for the case of unlike
particle collisions (that is, for i $ j) are currently being computed.)  The
next step is to construct a similar model and include both fields simultane-
ously.  In the meantime, the rates discussed above are going to be quanti-
tatively checked during an experimental investigation involving self-colli-
sions of monodisperse aerosols.  The experimental program is discussed next.

III.  EXPERIMENTAL INVESTIGATION

    A.  Introduction

        The objective of the experimental investigation is to determine the
collision rate and growth characteristics of submicron airborne suspensions
subjected to an a.c. electric field, to an acoustic field, and to a combina-
tion of these two external fields.  During the study, liquid and solid parti-
cles are  to be used and the following parameters are  to be varied: particle
composition, concentration, and  size; average electrical charge per particle;
and the magnitude and frequency  of each field.  Monodisperse aerosols are  to
be studied primarily, but some simple polydisperse aerosols are to be
examined  later in the project.   At the present time no coagulation studies
have been carried out in the system.  However, the experimental system is
nearly ready for some initial self-collision rate measurements.  The experi-
mental system is described in  the next  section, followed by a discussion of
the velocity and turbulence measurements recently completed.  The last sec-
tion describes the experimental  program.

    B.  Experimental System

        The experiments are  to be  carried out in a continuous flow coagula-
tion system which  is shown  schematically in Figure 3.  The aerosol is  to be
generated by  a combination  of  an atomizer, particle conditioner, and  an
electric  mobility  classifier.  This  subsystem is capable of producing highly
monodisperse  particles  in the  size range from about 0.01 to  0.3 ym.   The
aerosol  generating  subsystem is  discussed  in more detail in  a later paragraph.
The  generated aerosol  is mixed with  filtered air in  the mixing chamber  to  in-
crease the volume  flow rate.   From the  mixing chamber the aerosol  enters  the
                                                                              95

-------
VACUUM
PUMP
                                                                                                                       vo
                                                                                                                       CTN
ELECTRIC AEROSOL
SIZE ANALYZER
                                             r- ELECTRIC
                                             \ FIELD SUPPLY
         MASS FLOW
         TRANSDUCER
            r-ACOUSTICAL
            \ DRIVER SUPPLY
                                                                        AEROSOL
                                                                        ENTRY SECTION
        FROM FILTERED
        AIR SUPPLY
                                                                          AEROSOL
                                                                          MIXING CHAMBER
                                                                          PARTICLE       I
                                                                          NEUTRALIZER-^LJ
                                                                                         T
                                                                  ELECTROSTATIC
                                                                  CLASSIFIER
                                                                  WITH ELECTRIC
                                                                  AEROSOL DETECTOR
                                                    COLLISON
                                                    ATOMIZER
EVAPORATOR/CONDENSER
CONDITIONER
                                                                                                           MASS FLOW
                                                                                                           TRANSDUCER
                          FROM
                          FILTERED AIR
                          SUPPLY
                     FIG (3) EXPERIMENTAL SCHEMATIC OF THE CONTINUOUS FLOW  COAGULATION  SYSTEM

-------
                                                                               97
horizontal coagulation tube through an inlet section.  The continuous flow
coagulation tube is shown in more detail in Figure 4.  The coagulation tube
is about 4.9 m long and is made of 6 inch (inside diameter) Plexiglas tubing.
The tube has two ports for isokinetic sampling probes, two ring electrodes
wrapped around the outside of the middle section, plus an electro-mechanical,
acoustical driver mounted at the end with the aerosol inlet section.  The
acoustical driver is to generate a continuous wave of traveling sound waves
which travel down the coagulation tube and which are absorbed by a fiberglass
mat located at the far end.  The acoustical driver is a James B. Lansing pro-
fessional series compression driver, Model 2440, fitted with a specially
machined, air tight, exponential horn which makes a  smooth transition from
the 2 inch diameter of the driver throat to the inside diameter of the coagu-
lation  tube.  A thin membrane of plastic wrap is to  be used to keep the
aerosol particles out of  the driver.  The acoustical driver is to be driven
by an audio power amplifier with an output up to 60  watts.  A signal genera-
tor is  to be used as the  input  to the amplifier.  The ring electrodes, which
are spaced about 150 cm apart,  are used to generate  the axial electric field.
The a.c. electric power supply  consists of an inhouse built high voltage
amplifier  (0 to 6 kV) driven by the output from a signal  generator over the
frequency range of about  30-400 Hz.  Two moveable, isokinetic sampling probes
are used  (one at a time)  to extract a continuous sample from various points
across  the diameter of  the coagulation  tube at  the entrance and exit planes
of  the  test  section.  The distance between  the  probes is  about  183  cm, which
is  the  length of the  test section within  the  coagulation  tube.  The  extracted
sample  is  diluted with  clean  air and passed  through  an electric aerosol size
analyzer  to  determine  the particle  size distribution.   (More detail on the
probe design and the  analyzer is provided below.)  The aerosol  is  continu-
ously pumped through  the  coagulation  tube and the pressure can  be  varied
about ± 5  inches of water using a vacuum  pump.   As mentioned earlier,  this
feature of a continuous flow minimizes  transient effects  that have been known
 to  cause erroneous results in past  studies  that depended  on time decay
measurements to  determine collision rates.

         The coagulation tube is a critical  component for  it is  important  that
 the flow in the  tube be uniform and relatively free  of  turbulence.   As
 shown in Figure  4,  the flow enters  the tube through  the entry  section which
 introduces the flow uniformly and axially so as to minimize turbulent ef-
 fects.   The distance from the entry section to the  inlet of the test section
 (i.e.,  the location of the upstream isokinetic probe) is 60.96 cm and the
 distance from the inlet to the outlet of the test section is 182.9 cm.   For
 a flow rate of 44 1pm and a tube diameter of 15.24  cm,  the Reynolds number
 is about 430.   Assuming that the flow is totally uniform at the entrance of
 the coagulation tube, the thickness of the laminar boundary layer at the
 test section Inlet is estimated to be about 1.01 cm and at the outlet
 the thickness is 4.04 cm.  Therefore, the flow is not fully developed in
 the test section and there should be a central core of nearly uniform axial
 flow with a uniform speed slightly greater than the average speed given by
 the flow rate divided by the cross-sectional area of the tube.  (.Velocity
 and turbulence measurements have been taken across  the inlet and outlet
 cross-sections using a hot film anemometer inserted through the sampling
 tube ports and the results are to be discussed in a later paragraph.)

-------
98
         AEROSOL -
         DILUTED 1 10
         AT 44
                                  •ACOUSTICAL DRIVER
                                  MASS FLOW
                                  TRANSDUCER
          TO a-c
       ELECTRIC
         POWER
         SUPPLY
                                        •AEROSOL ENTRY SECTION
                                        FINE MESH SCREEN

                                        HONEYCOMB  STRUCTURE

                                        MOVEABLE ISOKINETIC  SAMPLING PROBE
                                         PARTICLE
                                         IVIEUTRALIZER
                       UNDILUTED AEROSOL
                       AT 0.4  ipm  ( TO
                       ELECTRIC  AEROSOL SIZE
                       ANALYZER OR STAGED
                       MEMBRANE FILTER SET)
                                      6 in (ID) PLEXIGLASS TUBING
        TO  VACUUM
        PUMP AT
        ~ 44 1pm
MOVEABLE ISOKINETIC PROBE
( IN  RETRACTED POSITION)
                                     •ACOUSTICAL ABSORBER
               FIG (4) SCHEMATIC OF ENTRY SECTION AND COAGULATION TUBE

-------
                                                                               99
        The aerosol is to be sampled and analyzed at different positions
across the coagulation tube diameter at the inlet and outlet stations of the
test section.  The samples are to be taken with the use of isokinetic probes
(inserted one at a time so as to minimize flow disturbances).   The diameter
of the probes is 1.45 cm which gives a probe flow rate of 0.4 1pm at the
average flow velocity in the tube.  A dilution system (shown in Figure 3) is
used to increase the total flow rate to the analyzer to a value of 4 1pm.
This corresponds to a concentration dilution of about 1:10.

        The aerosol generating equipment is commercially available from
Thermo-Systems Incorporated of St. Paul, Minnesota, and is based on the sub-
micron aerosol generating system developed by Liu and associates at the
University of Minnesota.  (The system is described and discussed by Liu and
Pui in Reference  [10].)  The system consists of an atomizer, an evaporator-
condenser conditioner, and electrostatic classifier with an electric aerosol
detector, and a neutralizer.  The atomizer can generate a moderately mono-
disperse aerosol  from non-volatile liquids (such as DOP) dissolved in ethyl
alcohol and  from  aqueous solutions such as methylene blue.  Suspensions of
pre-sized latex spheres and of carbon  (non-spherical) and  iron oxide  (sper-
ical) particles with a mean particle size in the range from 0.01 to about 0.3
ym may also  be produced by mixing these particles  in very  pure demineralized
water, or in alcohol.   (All of these particles are  to be studied during  the
coagulation  tests to be described later.)  The evaporator-condenser is  in-
cluded to convert the moderately monodisperse aerosols produced by the  atom-
izer  from liquids (like DOP)  into a more monodisperse suspension by an
evaporation  process  followed  by a condensation process.   (The evaporator-
condenser is not  to  be used-for  the solid particle suspensions.)  The elec-
trostatic classifier serves  as  the  final  stage to  select particles, on  the
basis of their electric mobility, to make up a truly monodisperse aerosol.
An electric  aerosol  detector  is  included  so that the particle  concentration
of the generated  monodisperse aerosol  may be determined when  desired.   (As
discussed  later,  the concentration  measurements  are required  in  order to
calibrate  the  electric  aerosol  analyzer used with the  isokinetic sampling
probes.)  A Kr-85 neutralizer is  also  shown in Figure  3  and is  to be used
when a neutral aerosol  is desired for  the coagulation  tests.   Liu  and Pui
 [10]  indicate  that this system is capable of  producing uniform particles from
 0.01 to  0.5 ym at concentrations up to 106  cm~3.  (The manufacturer  indicates
 that concentrations  up  to 107 are possible with  DOP aerosols.)   According to
 Liu and  Piu, the particle diameter  of  the aerosols generated by this system
 may be determined to within an accuracy of about 2% and the concentration
 to within about 5%.   The relative standard deviation of the size distribution
 for aerosols produced by this system ranges  from about 0.04 to 0.08.

         The particle size analyzer is an Electric Aerosol Size Analyzer (EAA),
 Model 3030, manufactured by Thermo - Systems Incorporated and is an instrument
 developed by Whitby and Liu at the University of Minnesota.  The analyzer
 first electrically charges the incoming aerosol particles and then sizes them
 on the basis of  their electrical mobility in the presence of an electrostatic
 field.  The performance of the analyzer is presented in a paper by Liu and
 Pui  [11].  The analyzer has been purchased with two additional voltage divid-
 er chips:  one to divide the 0.02 to 0.22 ym size range into ten arithmetic

-------
  100

  size  intervals of  0.02 ym each, and  one  to divide  the  0.20  to  0.40  ym  size
  range into  ten arithmetic size  intervals  of  0.02 ym each.   (The  standard
  voltage divider  chip divides  the 0.0032  to 1.0 ym  size range into ten geo-
  metrical size intervals  that  vary  in size from 0.0024  ym to 0.438 ym.)  This
  modification is  included in order  to improve the size  resolution of the in-
  strument.   Using the aerosol  generating system described above,  the EAA is
  to be calibrated by connecting  the analyzer directly to the output of  the
  aerosol generator.  It is expected that through a  careful calibration pro-
  cedure, the size analyzer may be used to  measure unknown concentrations in
  the individual channels  to within  an accuracy of about 5%, which is the con-
  centration  accuracy of the submicron aerosol generating equipment discussed
  above.

         A staged membrane  filter set is also to be used in conjunction with
  the isokinetic sampling probe to extract and collect particle samples for
 analysis of particle shape, size uniformity, and cluster configuration.  The
 membrane filters are to be arranged  in order of decreasing pore size from
 0.8 to 0.025 ym.   Photomicrographs of the collected samples are to be used to
 check the EAA results and  to determine cluster configuration during the
 coagulation studies using solid particles.  (In addition,  an attempt will be
 made to borrow a portable diffusion battery and a condensation nuclei counter
 during the  investigation to provide an additional check on the EAA and  on the
 filter- photomicrograph data.)

     C.  Velocity Profiles and  Turbulence  in the Coagulation Tube

         A hot  film  anemometer  has been  used  to  analyze  the  flow through the
 coagulation  tube.  Filtered air  without any  particle matter  was used during
 the tests.   The  turbulence structure  was  initially  unsatisfactory and the
 flow was not uniform.  Following the  addition of  a  screen in the  inlet  sec-
 tion and a honeycomb  structure across the  coagulation tube  at  the inlet
 section, the flow was found to be uniform  and the turbulence fairly  low.
 Velocity profiles at  the  upstream and downstream  locations of  the isokinetic
 probes (i.e., at  each end of the test section) are  shown in  Figure 5.   The
 profiles were taken using a probe mounted  on  a 90°  support sting  that could
 be  rotated and translated in order  to survey  the  cross-sectional  area of the
 coagulation  tube  at each  location.  The probe sting passed through the  iso-
 kinetic probe port  and "0" rings were used to ensure that there was no  leak-
 age  from the coagulation  tube.   The probe  used for  the  measurements was a
 Disa fiber film probe, Model 55R11.   The hot film was operated  in the con-
 stant  temperature mode by a Disa anemometer, Model  55D01.  Using  the square
 wave technique, the frequency  response of  the hot film  anemometer was about
 11 kHz  at a  mean  flow speed of 4 cm/sec.   The anemometer output was not
 linearized and the  system was  calibrated using the  fully developed flow in a
 1.50 cm diameter  tube with an L/D ratio of 80.  The flow rate through the
 calibration  tube  was measured using a Sierra Instruments mass flow transducer,
Model  500-3, which  was calibrated to  within an accuracy of 1%.   A signal of
 510 mV  corresponded to a  speed of 4 cm/sec.  It was found that the system
 could not resolve speeds  less  than  about 0.8 cm/sec.  The calibration curve
was repeatable to within  3% and the velocity profiles within 7%.

-------
                                                              101
                                                 3 E
 VELOCITY PROFILES AT UPSTREAM PROBE PORT
VELOCITY PROFILES AT DOWNSTREAM PROBE PORT
 FIG (5) VELOCITY PROFILES AT UPSTREAM AND
 DOWNSTREAM ISOKINETIC PROBE PORTS.

-------
 102

        The velocity profiles  in Figure  5  show  the existence of a central
 core where the flow is nearly  uniform.   The diameter of  the core at  the up-
 stream port is about 9 cm and  at the downstream port the diameter is about
 6.5 cm.  The profiles are stable and steady.  Flow visualization tests with
 smoke indicated  that the streamlines are straight and  that there is  no rota-
 tion of the flow.

        Frequency correlated turbulence  measurements were also taken using
 a Disa analog correlator, Model 55D70, with a signal generator, an active
 low-pass filter, and an rms voltmeter.   The rms values of turbulence for the
 0-11 kHz frequency range averaged about  3.8% of the mean flow speed  values
 over the core region of the flow within  the coagulation  tube.  The greatest
 turbulent intensity levels were detected in the lower  frequency ranges.  Some
 of the intensity peaks could be correlated at specific wave numbers  with
 characteristic dimensions of the coagulation tube diameter and length and the
 diameter and length of the straws used in  the honeycomb  structure inserted at
 the entrance of the coagulation tube.  The velocity profiles and turbulence
measurements indicate that the air flow  in the central region of the coagula-
 tion tube is relatively uniform and quiescent.  This condition is desirable
 in order to isolate the effects due to the external fields that are  to be
 investigated later.  In this case, the effects of velocity gradients and
 turbulence should be negligible.

    D.  Experimental Program

        The remainder of the investigation is envisioned to be conducted in
 three phases.  The first phase involves  using monodisperse DOF aerosols to
calibrate and check out the system.  During this time,  the wall losses due
to diffusion and gravitational settling  are to be investigated using the
isokinetic probe sampling system.   In addition, the thermal self-collision
rate is to be determined and checked with earlier work [3, 4].   It should be
noted that the results of this thermal self-collision work will extend the
range of the experimentally determined collision rates from a Knudsen number
of about 1 to 6.   In addition, the effect of particle charge is to be exam-
ined by conducting tests without the nuetralizer that is located between the
aerosol generator and the mixing chamber.

        The second phase involves determining the collision rate and growth
characteristics of DOP aerosols subjected to the external fields.   The mag-
nitude of the a.c. electric field is to be varied from about 100 V/m to about
4kV/m and the frequency from about 30 to about 400 Hz.   The power delivered
to the acoustical driver is to be varied from 0 to about 60 watts at fre-
quencies ranging from about 300 'to 10  Hz.   The self-collision rates using
a single monodisperse DOP aerosol are to be determined for each field and for
the two fields together.   It should be noted that regardless of which field
is applied, the number of collisions per unit time between particles of the
same size is expected to exceed the number due to thermal motion alone.  This
increase is due to the superposition of  the thermal motion and the motion in-
duced by either field.    For the very fine particles (i.e.,  particles with
radii less than 0.1 ym)  the thermal motion is relatively large.   When both
fields are applied simultaneously,  the number of collisions per unit time

-------
                                                                              103
between particles of the same size exceeds the thermal rate because of the
differential motion induced by both fields.  Of course, the thermal motion
will also contribute to the collision rate for the smaller particles in this
case.  In addition to the use of mohodisperse suspensions, some simple bimodal
suspensions are also to be generated and studied.  During these tests the
charge on the particles is to be varied by removing the neutralizer located
downstream of the aerosol generating subsystem and by inserting a corona
charging stage in the aerosol mixing chamber.  In all of these tests, photo-
micrographs of particle samples are to be used to check the EAA results for
size uniformity.  The samples are to be withdrawn through the isokinetic
probes and collected by the staged membrane filter set.

       The third phase involves repeating the second phase for different
types of monodisperse particles.  The different materials include latex, NaCl,
carbon, titanium oxide, and iron oxide.  Photomicrography of samples withdrawn
through the probes is to be used to monitor particle size and shape, including
cluster configurations in the case of the solid particle tests.

IV. SUMMARY

    An experimental investigation has been described that involves the use of
an a.c. electric field and a traveling wave sound field to increase the col-
lision rate between particles in an aerocolloidal suspension.  In addition,
a simple free molecule regime theory for the collision rates due to each
field applied separately has been discussed.  Experimentally determined
velocity profiles and average levels of turbulence in the continuous flow
coagulation tube have also been reported.  The plan for completing the in-
vestigation has been outlined.

    The investigation is a basic study of the collision rates between parti-
cles with radii less than about 0.3 urn suspended in air and subjected to
electric and acoustic fields.  The results of the study should contribute to
a better understanding of particle collisions and growth characteristics
under the influence of electric and acoustic fields.  In addition, it is an-
ticipated that the results will show that these  fields can provide an effec-
tive means for preconditioning man-made particulate suspensions so that the
very fine particles may be removed in a subsequent and conventional particle
removal device.

ACKNOWLEDGMENTS

    The author is appreciative of the work of P. H. Paul who made the hot
film anemometer measurements and L, W, Byrd who  designed  the inlet section
for  the coagulation tube.  This study  is supported by a research grant from
the Solid and Particulate Processing Program of  the National Science Founda-
tion.

-------
 10k
 REFERENCES

 [1]   Mednikov, E. P., Acoustic Coagulation and Precipitation of Aerosols,
      translated from Russian and published by Consultants Bureau, New York,
      1965.

 [2]   Scott, D. S., J. of Sound and Vibration, Vol. 43, No. 4, p. 607, 1975.

 [3]   Hidy, G. M. and Brock, J. R., The Dynamics of Aerocolloidal Systems,
      Pergamon Press, New York, 1970.

 [4]   Chatterjee, A., Kerker, M. and Cooke, D. D., J. Colloid and Interface
      Sci., Vol. 53, No. 1, p. 71, 1975.

 [5]   Brock, J. R., and Hidy, G. M., J. Applied Physics, Vol. 36, No. 6,
      p. 1857, 1965.

 [6]   Shaw, D. T. and Tu, K., "Acoustic Coagulation and Precipitation of
      Aerosols,"  Paper No. 15d, Aerosol Science and Technology Symposium,
      AIChE National Meeting, Atlantic City, New Jersey, August 29, 1976.

 [7]   Fuchs, N. A.  and Sutugin, A. G., J. Colloid Sci., Vol.  20, p. 492, 1965.
        «
 [8]   Scholz, P. D. and Barber, D. R., J. Colloid and Interface Sci., Vol. 46,
      No. 2, p. 220, 1974.

 [9]   Wang Chang, C. S.  and Uhlenbeck, G. E.,  "On the Propagation of Sound in
      Monotomic Gases,"   Engineering Research  Institute Report for Project
      M999, University of Michigan,  Ann Arbor, October, 1952.

 [10]  Liu, B.  Y. H.  and  Pui, D.  Y. H., J. Colloid and Interface Sci., Vo. 47,
      No. 1, p. 155, 1974.

 [11]  Liu, B.  Y. H.  and  Pui, D.  Y. H., J. Aerosol Sci., Vol.  6,  p.  249,  1975.
WRITTEN DISCUSSION

Andrew R. McFarland
Department of Civil Engineering
Texas A&M University
College Station, Texas  77843

    The experimental research efforts proposed by Scholz should permit the
achievement of a new level of refinement in data quality.  With reference to
Scholz's Figure 4, the apparatus which is to be employed is as good as the
state-of-the-art has to offer and the flow schematic seems to be well thought
out.  It may be anticipated that a high quality aerosol will result through
uniformization of a Collision Atomizer spray  with both an evaporator/

-------
                                                                              105
condenser and an electrostatic classifier.  It is planned to use a specially
equipped electrostatic aerosol analyzer which is fitted to divide the range of
0.2 - 0.4 micrometers into ten arithmetic intervals, however this resolution
may demand more of the instrument than it is capable of providing.

    With respect to the analytical modeling which is presented, p. 4, the
effort to date has been directed towards the free molecular regime with
separate consideration of the electric and acoustic fields.  The proposed
next step, that of modeling the effects of the combined fields, should provide
an interesting extension to the understanding of aerosol technology.

    The predictions which have been made of the augmentation to thermal
collision rates effected by acoustic and electric effects  (Figures 1 and 2)
show that for particles of 10   micrometers radius  the expectation is that
the rates would be approximately doubled — at least for the operational
conditions which were examined.   Considering^that  it is necessary to have
on the order of 10  collisions to cause a 10   micrometer  droplet growth to
one micrometer, a substantial augmentation is needed.
OPEN DISCUSSION

Scholz:   I have  a  couple of comments  to make regarding the taxation of the
electrostatic aerosol  size analyses.   I cannot  say at this time what the
detection range  would  be.  I  had  an impression  that  the  system is capable
of  producing aerosols  with concentration up to  10& particles per cubic
centimeter.  I do  not  know how good that value  is in terms of different
sizes  and different  materials.  At this point all I  can  say is that we are
expecting to get good  results.

     For the comment on the particle  size  under 100  ym not being terribly
important; I would like to indicate that I used the  100  ym particle size
as  a demonstration to  show the significance of  the thermal motion.  Our
approach is  to  try to  understand  the  interaction of  forces with  the
particles and  to try to get a good handle  on  the basic mechanism taking place
within the aerosol.   It is from this  point of view that  we used  the 100 ym
size particle  as a sample.

Baril:  We have some work going on  in our  laboratory which  is  comparable  to
your  study.   I was wondering  if you are familiar with Professor  Massudu's
work on hammering particles with traveling waves.  I gather  that you  are
going  to use this electric  field to enhance  the traveling wave.  We have
done extensive work in this area.  If you  would increase the number of rings
you could have a much better  effect.   In fact you  can  transport  the particles
down the tube without the acoustical  energy.   You might  even be  able  to  take
 the air energy out and just  use the electric  field.   If  you change the size
 of the rings as you go down,    it will make the particles come to  the
 center of the system.

 Scholz:  I am not familiar with any of this.   However,  I am very much

-------
106

   interested  in  learning more about your research  in  this area.

   Helfritch:  Paul,  is  the  inverse force law the right one to use between a
   charged particle and  a non-charged gas molecule?

   Scholz:  I  do  not  exactly know the answer.  In kinetic theory, the inverse
   power force law is the usual one and seems to give  qualitative effects, if
   you are interested in qualitative interaction.  I used it basically as a
   standard model.  I do not know how bad it is in terms of a neutral versus
   a charged particle interactionalwhether it is second order or primary, I
   do not know.

   Reif:  Back in the 1940's, Pohl did quite a bit of work with sonic dispersion
   of fog.  He reportedly had a system that worked on ships which would
   disperse fog up to 300 meters.  Are you familiar with this work and if so,
  whatever became of it.

   Scholz: I am not familiar with the work,  but there may be some people here
  who are.

  Baril:  I might add something to that,  Bob.   There was also some work done
  at Orly airport by Dr. Bouche with sirens.   If I remember correctly,  this
  all fell by the wayside.   It turned out that the amount of energy and money
  put into these systems was astronomical.

  Shaw:   I can comment  on that.   It was proven that if you use seeding  techniques
  it is much more economical for that particular application than acoustic
  sirens.  Some redesign work was done by  Calspan Corporation.

  Liu:   I am a little mystified  with the  use of  the Debye length for this
  calculation.  It would seem to me that  it is applicable primarily  to  the
  plasmas and colloidial suspensions  where the bipolar ions  are  around  the
  particles.   When you  are dealing  with a gaseous medium,  how does  the  Debye
  length come into play?

  Scholz:   It  is  my understanding that  as long as  the  Debye  length is greater
  than  the mean distance between the  particles,  then the Debye particle theory
  may be applied.  The  other necessary  condition is  that  the  energy  in  the
  electrostatic field of the particle at one Debye  length must be less  than
  the thermal  energy  that the particle  has.  Both of these conditions are
  satisfied in our application.

      In terms of polarization  effects and so forth,  I don't  know.  I  am only
  saying  that  it  is a model.  In terms  of kinetic theory, the  coulomb inter-
  particle force  law  has an  infinite cross section.  In order  to treat  the
  system you have to  cut it  off at some point.  Kinetic theory is based  on
  two-particle interaction.  It  is necessary to cut off the collision inter-
  action length at some  point.  We used the Debye length.

-------
                                                                               107
Liu:  The Debye length depends on the concentration of the charge carriers
and the particles.

Scholz:  The concentration, dielectric strength, the thermal characteristics
are also important.

     Hidy and Brock have a section in their book on the application of the
Debye Potential Theory in aerosols and have used it.

Cooper:  In the experimental set-up, the plexiglass, unless you are careful,
is going to have .areas of localized charge.  When you apply the voltage on
the rings, you will get stress charging which can create more problems.  The
use of isokinetic samplers can also cause problems since it is crucial at
which position they should be isokinetic.  The right angle bends are going
to cause some difficulties.  Precisely when the particle relaxation time
times its velocity is comparable to the dimension of the sampler.  Those
bends look as though they are going to give substantial impaction on the
inside surface.

Scholz:  In reference to the plexiglass, at this time we have only used low
field intensities, and we have not noticed any  localized effects.  It  is
possible that changing to higher intensity might cause some problems.

Cooper:  You may  notice localized areas of intensities because the fields may
be  as high as a thousand volts per centimeter.

-------
 108
             ENGINEERING ASPECTS OF DRY MAGNETIC SEPARATION

                               R. R. Oder*
                 Bechtel Corporation, San Francisco, CA
 ABSTRACT

           Engineering aspects of batch-type and continuously operating
 magnetic gas-particulate separation methods are compared.   The comparison
 is made in conceptual applications to desulfurization of dry fine-
 pulverized pretreated coal.   With pretreatment,  it  is claimed that
 magnetic separability of mineral sulfur forms  occurring in coal can be
 considerably enhanced so that a variety of  magnetic methods could  be
 expected to be  applicable.   In processing fine pulverized  coal,  batch-
 operated HGMS-type  separators offer the advantage of  efficient desul-
 furization of high  coal-value recovery but  can incur  practical difficulties
 associated with packed bed operation.   By contrast, continuously operating
 smooth-bore magnetic  separators have the advantages and simplicity of
 continuous operation  but are not as effective  in desulfurization in high
 gas-velocity applications.   Trade-offs between separator performance and
 total process requirements will be  very important in  future developments
 of magnetic methods for  gas-particulate separations.  New  applications
 of magnetism can be anticipated in  process  industries employing  entrained
 and fluidized bed operations.

 INTRODUCTION

          Two widely different  approaches have been considered for
development  in  the application  of magnetic methods to gas-particulate
separations.  Basically, these  two  approaches provide the researcher
with a choice between batch-operated magnetic filtration (modified
HGMS-type separators)  or continuously operating smooth-bore beam-splitting-
type devices  (e.g., multipole magnetic separators).   These magnetic
 Presently with Gulf Research & Development Company, Pittsburgh, Pennsylvania

-------
                                                                              109
separators can be expected to have widely different capabilities in gas-
particle separations and to exhibit significantly different process
design requirements in high velocity and high throughput applications to
gas clean-up.

          The batch-type magnetic separator envisaged for gas-particulate
separation is similar to the HGMS-type separator recently developed for
wet minerals beneficiation.  An example of the HGMS-type separator is
illustrated in Figure 1[1].  This device utilizes flow  through  a relatively
open and porous magnetized bed which is composed of filimentary strands
of strongly magnetic material such as ferritic stainless steel  wool,
wire mesh, or expanded metal lath.  The magnetized bed  elements provide
a volume distribution of very short range but intense magnetic  capture
forces which can be made specific to particle size.  These  batch-operated
devices have proven commercially effective in wet separations of  feebly
paramagnetic minerals with size distributions occurring into  the  submicron
range  [2].	            	—
                CLEANEO
              j [^EFFLUENT
                   MAGNETIC PARTICLES-
 Figure 1. VERTICAL SECTION THROUGH CENTER   Figure 2. QUADRUPLE MAGNETIC SEPARATOR
        OF HIGH GRADIENT MAGNETIC               (Re(- BO)
        SEPARATOR
           The second approach to dry magnetic separation of gas-entrained
 particulates employs continuously operating smooth-bore magnetic beam-
 splitting devices utilizing multipole magnets of the type presently
 employed in particle physics applications  [3,4],  These devices, which

-------
 110
 have not yet been applied outside of the laboratory, offer the potential
 simplicity of continuous smooth-bore operation.  The magnetic forces
 they generate, however, are weaker and less specific to fine-size
 materials and they are inherently power-intensive devices.  An example
 of a quadrupole magnetic separator is shown in Figure 2[3].

           HGMS-type structures are probably closer to commercial develop-
 ment in dry handling of fine pulverized materials than are continuously
 operated magnetic devices.   The latter type separators, however, are
 deserving of consideration  because of several significant technical
 reasons.  They require no gas flow interruption,  are relatively immune
 to abrasive wear, and are low-pressure drop devices.   In gas-particle
 transport,  entraining velocities can be expected  to be very high because
 of the  large density difference between the suspended particles and the
 fluidizing gas.   Even though modified HGMS-type devices are superior to
 multipole structures from the standpoint of fine  particulates  removal,
 they may be limited  in application because  of  severe  problems  of flow
 distribution and  pressure drop associated with high velocity flow
 through  the magnetized filter bed.   In  applications of  50 000  to 100 000 cfm
 flows, pressure drops of  a  few hundred  inches  of  water  can result in
 drive-fan power requirements  in excess  of megawatts and may well over-
 shadow magnet  capital and operating  costs.  Multipole magnet structures
 appear to offer considerable  potential  for  control  of entrained  and
 fluidized bed  reactor operations.

          The understanding of  two-phase flow  (solids and  entraining
 fluid) through consolidated porous media such as HGMS-type packed filter
beds is  incomplete at present.  Each potential application must be
treated  individually with due consideration for the magnetic particle
density, mean size, and size distribution and for the size and shape of
individual magnetic filter bed elements.  Preliminary engineering
evaluations indicate that quadrupole-type smooth-bore magnetic sepa-
rators may possibly compete  with batch-operated filter-type magnetic

-------
                                                                            Ill
separators in high velocity separations of strongly magnetic micron-size
materials.  To be competitive in separations applications, however, it
is believed that multipole structures will have to be developed with
pole-face field strengths of the order of 100 000 Oe at one foot pole-
opening.

          A comparison of these magnetic separation methods will be
given in a coriceptual application to the removal of dry mineral sulfur
from fine-pulverized coal.

MAGNETIC DESULFURIZATION OF DRY COAL

          Coal has traditionally been cleaned to control mineral matter
content.  Cleaning for sulfur control is now becoming increasingly
important because of governmental regulations of sulfur oxide  emission
in coal combustion flue gases.

          Sulfur occurs in coal both as a discrete mineral  form  (pre-
dominately iron disulfide  [FeS2]) and as "organic sulfur" which  is
believed  to be chemically bound to  the coal structure.  Because  of  the
need to remove both sulfur forms, chemical methods are being developed
to augment more conventional physical cleaning methods which are generally
restricted to removal of coarse  (>35 mesh) mineral sulfur only.  Recent
studies have  shown, however, that chemical cleaning per se  can be
expected  to be expensive  [5].  For  this reason,  there is  renewed emphasis
on upgrading  state-of-the-art physical cleaning  technologies and on
developing new approaches  to combined physical and chemical cleaning
which  have the capability  of preparing a moderate cost clean fuel
without requiring  coal  conversion.

          Coal washability studies  [6]- have  shown  that mineral sulfur
occurs in U.S.  coals  in a  great variety  of particle  sizes.  Most of the
important Eastern  utility  grade  coals  have  significant amounts of  mineral
sulfur with  particle  sizes occurring down into  the micron range.  Any

-------
  112
 comprehensive new coal cleaning method developed for these coals, then,
 will of necessity be required to process coals which have been pulverized
 to sizes generally finer than 100 mesh (75 microns).  There is no such
 commercial coal cleaning method in use today.

           Laboratory investigations have demonstrated the scientific and
 technical feasibility for magnetic desulfurization of coal [7].  Those
 investigations show that the mineral sulfur occurring in coal is para-
 magnetic and is magnetically separable from mineral-free coal which is
 diamagnetic.  That work was not pursued,  however,  because the laboratory
 magnetic separators employed could not be adapted  economically to commercial-
 size applications.   With the recent development of HGMS-type devices,
 however, there is now a magnetic separation method available for adaptation
 to desulfurization of coal with the potential for  economical removal of
 micron-size mineral contaminants in commercial-scale installations.

           Because of this potential,  the  concept of magnetic desul-
 furization of coal  is again being pursued in several laboratory investi-
 gations.   The results reported  here were  derived in an  engineering
 evaluation of dry magnetic  methods for desulfurization  of  pneumatically
 conveyed fine-pulverized  coals.

           The most direct use of magnetic methods  in  dry coal  desul-
 furization would  be  in power plants which burn  fine-pulverized coal.
 The magnetic  separator would be "engineered"  to  fit in  line between the
 pulverizer and the coal burner inlet.  A  preliminary  evaluation of this
 concept  [8],  however, has indicated that  the use of magnetic separators
 in this application may not be practicable because of severe limitations
 imposed by the feeble paramagnetism of the mineral sulfur and  the use of
very high carrier gas velocities (5 000 to 6 000 ft/min).  Estimates
 indicate that capital and operating costs for this direct approach to
magnetic desulfurization may be similar to those of stack gas scrubbing.
These costs are largely associated with the high magnetic field strength
required to capture paramagnetic minerals and with power requirements

-------
                                                                            113
for auxiliary equipment.  Because of this unfavorable cost estimate, the
engineering evaluation presented here was restricted to the use of
magnetic separators in the processing of dry coals which have been
pretreated to make the mineral matter strongly magnetic.

MAGNEX PROCESS

          Several methods have been proposed for converting mineral
matter in coal to strongly magnetic forms  [9].  One new method is the
Magnex process which exposes dry pulverized coal to gaseous iron
pentacrbonyl  [FeCCO)-] under conditions whereby the impurity mineral
matter becomes strongly magnetic and can be separated from the coal by
magnetic separation methods  [10].

          Under  the Magnex reaction conditions, gaseous iron pentacarbonyl
is  substantially decomposed  into free iron and gaseous  carbon monoxide.
It  is claimed that the  liberated iron reacts with  the surface of  the
pyrite mineral to form  a  strongly  ferrimagnetic pyrrhotite-like  sulfurous
mineral.  The reaction  stoichiometry is  assumed to be:
17°-200°C>   l-FeS   •   (^FeS] + 5xCO      (1)
 FeS2   +   xFe(CO)5    -> [(l-y)FeS2]  •

 iron         iron                "iron-rich disulfide"       free
 d-isulfide   pentacarbonyl                                   carbon
 (pyrite)                                                   monoxide

           If x moles of iron pentacarbonyl react with 1 mole of granular
 and completely liberated iron pyrite, it is assumed that a fraction of
 y moles of the outside of the pyrite grain react to give y-Hx moles of an
 iron-sulfide of sulfur to iron atomic ratio of 2y/(y4x).  If one demands
 that this ratio correspond to that of magnetic monoclinic pyrrhotite
 (^Fe S ), then the mole fraction y is determined by the mole quantity of
     7 8
 iron pentacarbonyl used, x (see iron-sulfide phase diagram[ll]).  This
 relationship is shown in Table I where percent conversion of iron pyrite

-------
  114

 to monoclinic pyrrhotite is shown for a coal containing about 1.2 wt.%
 pyritic sulfur and where, for simplicity, it has been assumed that free
 iron is deposited only on the surface of the iron pyrites.

           In the magnex process, it is conjectured that some of the free
 iron also coats the non-sulfurous mineral matter, making it amenable to
 magnetic separation as well.  For this case, the conversions shown in
 Table I are upper limits expected for reactions with pyrites only.  It
 is  claimed that free iron does not adhere to the mineral-free coal
 structure.

           Laboratory quality monoclinic  pyrrhotite (Fe^.)  is strongly
 ferrimagnetic with a spontaneous magnetic moment of  13  emu/gm at  200°C
 which saturates at =4 kOe  [1].    This moment is about 14 000 times
 greater than the  paramagnetic  moment  of  pure pyrite  at  200°C in a 4 kOe
 magnetic field.   Indeed, only  about 0.7% of  the pyrite  need  be converted
 for the magnetic  moment  of the "iron-rich  disulfide"  to predominate that
 of the  converted  mineral.  The resulting impurity  saturation magnetic
 moment  calculated for assumed  selected reaction of iron pentacarbonyl
 with  iron-disulfide  is shown in Table  I.  Magnex has  recommended  32 Ib
 iron pentacarbonyl per ton coal for most Eastern coals, so that with
 this concentration, one can, under ideal conditions,  anticipate a
magnetic moment of 8  emu/gm which saturates in  a 4 kOe magnetic field.

                                TABLE I
            "Ideal" Pyrite Conversion by Iron-Pentacarbonyl
      for a Coal Containing 1.2 wt.% Pyritic Sulfur — Dry Basis
     Iron Pentacarbonyl       Calculated          Calculated
       Concentration         % Conversion       Magnetic Moment
   Ib Fe(CO),/T drv coal     J^F^S^       (emu/gm) @200°C
           0                      0                    o
          20                     36                    5
          32                     59                    8
          40                     73                    9
          55                    100                   13

-------
                                                                             115

IRON CARBONYL PROCESS

          A process flow diagram for an iron-pentacarbonyl process is
shown in Figure 3.  This layout is similar to the Magnex process [10]
except that here fine-pulverized coal is specifically employed and iron
pentacarbonyl is consumed.  In this process layout, carbon monoxide
liberated in the reaction is flared to produce carbon dioxide which, in
turn, is used for penumatic transport purposes.  In a commercial process,
the liberated carbon monoxide might be passed through an iron reservoir
for generation of process iron pentacarbonyl.  In this manner, iron
rather than iron pentacarbonyl would be consumed.  It is to be emphasized
that the hypothetical process described here differs from the Magnex
process primarily in that the coal is fine-pulverized.  When coarser
coals are treated with  the Magnex process, pneumatic transport is not
necessary and simple magnetic separators  adaptable from other areas  of
minerals beneficiation  can be employed  [13].  The hypothetical process
given here  for desulfurizing fine-pulverized coals is chosen specifi-
cally to illustrate  the comparison of advanced magnetic separation
systems such as might be developed for  handling  gas particulate  sepa-
rations in  applications where the particle grain size is  smaller than
that presently considered practicable for physical coal beneficiation.
These applications are  significant and  can include steel  furnace off-gas
cleanup as  well  as coal desulfurization.  Extensions of the principles
illustrated here may also find  applications  in electromagnetic  control
of entrained  and fluidized  bed  reactors.

           The fine-pulverized  coal  is pneumatically  conveyed  to the
 cyclone collector as shown in  the upper left corner  of  Figure  3[5],
where it  is uncoupled from the gas  and  dropped through  a  lock  hopper
 into a fluidized bed reactor where  the coal  is reacted  with gaseous iron
 pentacarbonyl.   The dry treated coal is passed through  a second lock
 hopper at the bottom of the reactor and is pneumatically conveyed to a
 magnetic separator where clean coal and magnetic refuse are separated.
 Lock hoppering is required because of the toxic properties of iron
 pentacarbonyl and carbon monoxide.

-------
 116
          PULVERIZED
          COAL
           FLUE GAS TO COAL PREPARATION
                    Figure 3 IRON CARBONYL PROCESS
                           FLOW DIAGRAM
                                                      CO BURNER
           Pulverized coal is pneumatically conveyed in this process with
 inert  flue gas  which is primarily carbon dioxide.   Under normal operation
 conditions,  between 2 and 5 Ib O>2 gas would be used to convey the coal
 at velocities betwen 2 000 and 5 000 ft/min.  The  use of these conditions
 would  be prohibitive in the application of almost  any type magnetic
 separator  because:   (a)  very high velocity applications will require
 excessive  pressure  drop if a magnetized bed concept is employed,  and
 (b) dilute solids suspensions will require large magnetized cross sections
 in order to  be  able to handle the full process flow.   In order to circumvent
 these  difficulties,  a cyclone separator and bypass fans have been used
 to uncouple  the coal dust  from the main volume of  entraining gas  flow [8].
For the examples to  follow,  pulverized coal is conceptually blown through
the magnetic separator at  velocities  ranging between  the normal settling
velocities for  the  dust  and  transport velocities of 5 000 to 6 000 ft/min.
Gas loadings of four  pounds  of  coal per pound of carbon dioxide are
employed.  The  use of  these  conditions  allows considerable latitude in
the choice of magnet  size.

-------
                                                                            117
          Conceptual applications of batch-operated and smooth-bore
magnetic separators have been compared for coals pretreated by the iron
pentacarbonyl process described above.  In the comparison, a high volatile A.
bituminous Pittsburgh Seam coal containing 1.2 wt.% iron pyrite was the
coal (100 mesh top size and 50 micron mean particle size) that had been
pulverized to 70 to 80% finer than 200 mesh (power plant grind).  With
this grind, mineral sulfur is assumed virtually completely liberated for
this coal.

     A.   Batch-Operated HGMS-Type Separators

          Magnetic capture of pneumatrically conveyed fine-pulverized
"iron-rich disulfide" in HGMS-type separators has been calculated with
the use of a mathematical model adopted from that developed to describe
the capture of paramagnetic minerals from highly dispersed wet-minerals
processes (kaolin clay)  [14].  The model has been modified for this
application to describe  the capture of strongly magnetic materials from
turbulent flows.  The fractional removal of "liberated" magnetic particles
is described by Equation (2) where C_ is the ingress and C is the egress
concentration of these materials.
                           Kl     ,  1
R =
                     1-e" CNR/24  (1+K2)
                    y(d/D)2e (1-e )L
                      9nv
NR is the Reynolds number defined by,
          N  =f                                                      (4)
           K.     n

and C is the drag coefficient defined by  the relationship,

          drag force =  l/8p Cird2v2.                                    (5)

-------
 118
 At  Reynolds numbers  generally below 1/2,  the  drag force  is  given by
 Stokes Law,

          Stokes Law drag force  =  Sirridv.                               (6)

          The  term KZ of  Equation  (2)  is  the  ratio of  particle weight  to
 viscous drag,
                                                                       (7)
and is generally unimportant for  the  flow conditions  considered here.

          The mathematical modeling of Equations  (2)-(7)  is  for "ideal"
unloaded filters and infers that  retention in the magnetic filter  is
dominated by forces which hold the magnetic particle  on the  capture
element.  Capture on "loaded" filter  elements can be  handled empirically
by a variety of methods including adjustments to the  capture surface
strand diameter D.  In the above  equations, M (gauss) is  the magnetic
capture surface magnetization, p  is  the magnetic particle density
(gm/cc), y is the "iron-rich disulfide" magnetic moment (emu/gin),  d is
the magnetic particle diameter (cm),  D is the capture surface strand
element diameter (cm), e is the bed fractional void volume,  L is the bed
depth (cm), n is the fluid viscosity  (poise), and v is the superficial
velocity through the filter bed reckoned for 100% void volume, pf  is
the fluid density (gm/cc), and g  is the acceleration due  to  gravity
       2
(cm/sec ).  The fluidizing gas is carbon dioxide at 200°C where p. =
         -3                        -It.
1.13 x 10   gm/cc and r\ = 2.27 x  10   poise.

          The magnetized bed configuration visualized for the conceptual
application discussed here is a three-dimensional diamond-shaped structure
composed basically of one-inch segments of 1 mm diameter  No. 430 or
No. 446 ferritic stainless steel.  This material and its  three-dimensional
structure and element size have been  chosen on the basis  of  commercial
availability (demister tower packing), corrosive resistance  under  process

-------
                                                                             119
conditions, immunity to abrasive wear,  and effective magnetic removal of
mean-size magnetic particles  [8].   Coals pulverized to power plant
fineness contain particles  ranging  downward from about 150 microns with
a mean particle size of 50  microns  (see Table II).   A choice of thinner
capture surface material, such  as,  for  example,  100 micron diameter
steel wool, would have provided for better removal  of the finest size
components of the coal dust,  but its use would have involved more serious
abrasive wear and would have  led to increased pressure drop (in excess
of 100 inches of water).

          Because of the  strong ferrimagnetism of the "iron-rich disulfide,"
a magnetic field of only  2  kOe  has  been employed.  The magnetic field
strength dependence of both the capture surface magnetization and of the
calculated mineral magnetization are shown schematically in Figure 4.
        SATURATION .
        MAGNETIC
        MOVEMENT

         IEMU/GM)
                                                V  s 430 STAINLESS STEEL
                                                12 '  EXPANDED METAL
                  0               2               «               6
                    Figure 4 MAGNETIC PROPERTIES OF STAINLESS STEEL, AND Fe7S8
           The ideal capture model of Equation (2) predicts virtually
 100%  removal of the coarse "iron-rich disulfide" particulates occurring
 in coals pulverized to power plant fineness.  The broad performance
 characteristic inferred for HGMS removal of "iron-rich disulfide"  from
 pulverized coal is shown in Table II.

-------
 120
                                TABLE II
                  Calculated HGMS  Removal  of  7 MJ emu/gm
               "Iron-Rich Disulfide"  from  Pulverized  Coal

Particle
Diameter
(U)
150
100
50
10
3
1
Particle
Size
Distribution
(% Finer Than)
100
83
50
18
7
3
                                           Mineral Matter  Removed
                                            Velocity  (ft/min)
                                              200        1600
100
100
100
100
100
99
100
100
100
100
99
45
          On the basis of this simple model, it is theorized that HGMS-
type filtration can achieve efficient removals of strongly magnetic
mineral contaminants even at velocities which are large fractions of the
entrained gas velocity.  The calculated capture for fine particle components
less than five micron diameter are shown in Figure 5, where it can be
seen that 100% removal is predicted for particulates down to 3 micron
diameter at flow velocities up to 1600 ft/min.  Ninety-five percent
removal of iron-rich particulates down to 4 micron diameter is predicted
for flows of 6600 ft/min.
                                3              2
                             PARTICLE DIAMETER (MICRONS)
                    Figure 5  CALCULATED HGMS CAPTURE OF Fe?Sg

-------
                                                                            121
          The calculated capabilities discussed above suggest that HGMS
has the potential for 90+% removal of mineral sulfur from fine-pulverized
iron carbonyl treated coals in direct in-line application between the
pulverizer and the burner.  It is estimated that the HGMS-type separator
can achieve this mineral sulfur removal for the reference coal with a
95+% recovery of mineral-free coal.  This type process performance is
not inconsistent with the experience of HGMS wet minerals beneficiation
in the clay industry.  Estimated add-on cost for desulfurization alone
with the iron-pentacarbonyl process described here (including batch-
operated HGMS separators) is less than $7.00 per ton of clean coal and
compares favorably with current costs for high level commercial physical
coal cleaning [15].  Conventional utility coal washing, however, is
generally limited to the processing of coals coarser than 35 mesh.

     1.   Magnet Structure

          The pillbox-type HGMS magnetic structure shown in Figure 1 is
not well suited to high velocity processing.  The top and bottom of the
HGMS structure of Figure 1 form the north and south poles of the magnet
and are solid iron pole pieces.  In the wet application, slurries enter
and exit the magnetized bed through a multiplicity of small diameter
holes bored through  the pole pieces.  The entrance and exit pipes also
promote flow distribution in the relatively dense mineral slurry.
Complete utilization of the entire magnetized volume is essential to the
economic viability of the HGMS concept, so that good flow distribution
and low pressure drop are essential features of the design of entrance
and exit to the magnetized bed.

          It is believed  that neither good flow distribution nor low
pressure drop can be achieved with the magnet structure shown in Figure 1
in high velocity gas-particulate separations.  Use of an iron-clad
Bitter solenoid  [16] with poles removed or of a cross-flow H-frame
structure recently proposed by Allen  [17] would provide better means for
flow distribution and pressure drop control than does the pillbox magnet
structure of Figure  1.  The Bitter-type magnet, with iron poles removed,

-------
 122

would allow for ease  of  access  to  the  magnetized  volume but would have
marked disadvantages  of  requiring  higher currents to achieve fields
equal to  those developed by the magnet structure  of Figure 1 and would
have the  added occupational hazard of  large dipole fringing fields
outside the magnet in the working  environment.

          The magnet  conceptualized for the gas-particulate separator is
designed  around the H-frame magnet structure proposed by Allen and shown
in Figure 6.  This magnet is essentially identical to the pillbox magnet
of Figure 1, except that the iron  on two of the outside edges has been
removed,  and the energizing coils  have been separated in the center and
folded up in a saddle shape so  as  to permit transverse access to the
magnetized volume.  Performance characteristics for this type magnet are
essentially identical to those  of  the  Bitter-type magnet structure, but
dipolar fringing fields  are low.
           SADDLE SHAPED
            UPPER COIL
          TRANSVERSE FLOW
            SADDLE SHAPED
            LOWER COIL
                           FigureG H-FRAME MAGNET

-------
                                                                            123
          The H-frame magnetic separator is different from the pillbox
structure in another important aspect.  In the pillbox structure, both
filtration flow and flushing flow occur along the same length which is
parallel to the magnetic field direction in Figure 1.  Flushing can be
difficult when the pillbox separator employs a deep filter bed.  The
Allen patent teaches filtration flow along the long pole length trans-
verse to the short pole gap opening and flushing flow, with the magnet
de-energized, along the short length parallel to the pole opening.  In
this fashion, flushing time is expected to be less, for equal bed volume,
than for the pillbox separator.  Reduction of dead time in filter cleaning
is important in improving the batch-operated filter on-line time.

          With the use of transverse flow the relative orientation of
magnetized capture surface element, magnetic field direction, and flow
velocity is different from that employed with the pillbox magnet structure.
It is believed, however, that the magnetic capture surface designed for
this study permits good capture of strongly magnetic materials at low
pressure drop in high velocity turbulent flow applications.

     2.   Process Characteristics

          Magnet size to handle process capacity is determined with use
of Eq.  (8).

                              C = 0.72 S EvAe                     (8)
                                        c

In Eq.  (8), C is the process capacity  (8 000 tons coal per day), S  is
the pounds of entrained coal per cubic foot of flue gas  (0.28), E is the
on-line factor for the batch-operated magnet, v is the flow velocity in
the filter bed (ft/min), A is the cross-sectional area of the filter bed
   2
(ft ) for a magnet operating with on-line factor E, and e is the filter
fractional void volume.

-------
 124
                               8            7
                      FRACTIONAL VOID VOLUME, e. V = 1600 FT/MIN
           Figure 7 EFFECT OF BED PACKING DENSITY ON MAGNET SIZE AND PRESSURE
                 DROP AT 1600 FT/MIN. GAS FLOW
          Calculations  of magnet  size requirements (magnetized cross-
sectional area) and pressure  drop across the filter bed are shown in
Figures 7 and 8 as functions  of the bed fractional void volume and
velocity of flow through the  bed, respectively.   The calculation of
magnet size involves use of the on-line factor shown in Figures 7 and 8.
Details of on-line factor calculations can be found in the literature [18],
As is apparent in the figures, significant reductions in magnet size
requirements can be achieved  with use of densely packed filter beds, but
only at the expense of  high pressure drop.   By way of example, a single
2 kOe HGMS-type magnetic separator operating at 60% on-line factor could
handle the entire 8 000-ton-per-day process stream with an 8.3-ft deep
            2
bed of 45 ft  cross-sectional area when packed to 90% void volume.  The
device would develop a  calculated pressure drop of 138-inch water gauge.
At a flow rate of 39 000 cfm, the system would require a 1400 hp auxiliary
fan of 1.05 megawatts peak power  consumption.   This power requirement is
large compared with the 140 kilowatts estimated for magnet operation.

-------
                                                                              125
          High velocity  gas  entrained flow of particulates through
porous media is a very complex phenomenon, and magnetic capture in such
beds will probably have  to be studied and understood on an individual
basis.  The calculations presented illustrate the important point that
the magnetic separator cannot be isolated from the overall process for
planning and estimating  purposes because choice of auxiliary process
operations and conditions may well be more important than magnet optimization.
                                       NORMALIZED MAGNETIZED CROSS SECTION
             200
                   400
                                                    1400
                                                           1600
                                                                  1800
            600    800    1000    1200
                FLOW VELOCITY (FT/MIN I,C = 09
F,9Ure 8 EFFECT OF FLOW VELOCITY ON MAGNET SIZE AND PRESSURE DROP
      AT 90% VOID VOLUME
                                                                        2000
While  the example presented is a severe one, it is encouraging to know
that the  use of HGMS hypothesized for this dry application,  even with
its expensive ancillary cyclones and fans, still appears  to  be economical
in the coal  processing context.  Estimates of the costs of HGMS coal
cleaning  with the Magnex process [5] indicate that magnetic  separation
will be a small component of the capital and operating costs,  and that
this type of coal cleaning can be expected to be accomplished  at add-on
fuel costs for desulfurization of about $7.00 per ton of  clean coal

-------
 126
 produced.  This estimated cost is competitive with that known for stack
 gas desulfurization and with those estimated for more elaborate chemical
 cleaning and coal conversion technologies.

      B.    Quadrupole Magnetic Separators

           Quadrupole magnets are structures which produce a uniform
 magnetic field gradient throughout a relatively large working volume.
 Figure 9 shows a cross-section transverse to the length of a quadrupole
 magnet.   The magnetic field lines emerging from the north poles are
 deflected to the south poles in such a manner that the lines of isofield
 intensity are concentric circles about the symmetry axis [19].  The
 magnetic field strength is zero on the axis and increases linearly to a
 maximum value on the pole tip which is typically 10 to 12 kOe for modern
 iron-return electromagnet quatrupole structures.
                                                                MAOMTK OMTOMHT f KIT
Figurt 9  CROSS SECTION "A-A" OF FIGURE 10
       WITHOUT BEAM SPLITTER
Figur* 10  CROSS SECTION "B B" OF FIGURE 9
        INCLUDING BEAM SPLITTER

-------
                                                                            127
          Figure 10 shows a vertical section through the symmetry axis
of a quadrupole structure in which a "beam-splitter" pipe has been
inserted for the purpose of separating the clean coal product from the
magnetic fraction.  The coal enters the quadrupole field region in the
annular space of inner and outer radii R... and R.,, respectively.  The
inside radius R  is the minimum distance from the symmetry axis for
which the magnetic field strength is sufficient to saturate the magne-
tization of the "iron-rich disulfide."
                                 Hs '    Ho
The outer radius, R,,, is chosen so that N magnets with annular entrance
           2     2
areas, ir(R^  - R  ), can handle the entire process flow
                    C(lb/min) =
where W is the carbon dioxide  to  coal  ratio, PC02  is  the C02  gas  density
 (lbs/ft3), and V  is the gas flow velocity  (ft/min).
          In  flowing  through  the  quadrupole  field  region,  the  magnetic
 component will be  deflected radially  outward toward  the  magnetic  poles
 and will exit through a  take-off  pipe on the outside diameter  of  the
 entrance pipe.  A  small  fraction  of the  clean coal inadvertently  exists
 with  the magnetic  component,  but  the  major portion of the  clean coal
 emerges from  the pipe of radius R at the bottom of  the  quadrupole beam
 splitter.

          The equation of motion  along the vertical  axis is
                              = v  +    PfS                      (11)
                          dt    V0 +    9n                        U '

-------
128
   where p  is the particle density, r is the particle radius, g is the
   acceleration due to gravity, and n is the fluid viscosity.  Buoyancy
   forces have been neglected because of the great difference between
   particle and gas densities.  Stokes Law has been assumed for the viscous
   drag force, and particle-particle interactions have been ignored.  The
   residence time, T,  in the magnet structure of length L is given by the
   relationship
                                  9n
             With similar assumptions,  the radial component of the equation
   of motion is given in Eq.  (13),  and  the time-dependent radial motion is
   given in Eq. (14)  where the particle is assumed to have entered the
   field region on initial radius
  with  zero  radial  velocity.
                                 3H       •
                      V  = %  3R -  67rnrR
                 R(t) = R  + -    (e -a   -  i) + -* t
                              a
                  yH
  In Eq.  (14), Y = — where y is  the "iron-rich disulfide"  ferrimagnetic
  moment,

  and H(R) is equal to H R/FL^.

-------
                                                                           129
          Assuming that all particles which strike  the  inner wall  at
radius R^ are captured, the equation of motion,  Eq.  (14),  and  the  residence
time, T, can be combined to implicitly define a  quadrupole-structure
length, L (r.p.V.), such that all particles of radius greater  than r
will be collected where p is the percent of such particles in  the  annular
region between R. and R^.
                                  KT
                    Vr*p'V =
where

                                W
                    K	=	5-                     (17)
                        ^CO11' (VV *
and
                                              V
 In these  equations, T is  the  solution to the equation f(T) = 0 where
                               2	
      f (T) =  e'aT  + aT - a[ ^- {l-/l-[l-(Hs/Ho)2]P/100} + I/a]       (19)

 The appropriate parameters are:  pCQ  = 1.129 x  10   gm/cc, nco  = 229 x 10
 poise,  y  = 6.9 emu/gm, HQ = 10 000 Oe, W *  .25 Ib C02/lb coal, and p ^4.5 gm/cc
 for "iron-rich disulfide."

           Figure  11  shows the calculated length  requirements for a
 quadrupole magnetic  separator which achieves  100% removal of 10 micron
 diameter  Fe,SQ from  pneumatically  conveyed  pulverized coal.  Performance
            7 8
 at this level is  equivalent  to 90% removal  of all mineral sulfur from
 the coal.

-------
 130
           It can be  seen that a quadrupole structure is not as  efficient
 in removal of fine-sized particles as is HGMS.  This is basically because
 of the low magnetic  field gradient, 656 gauss/cm, produced by the
 quadrupole.  Unfortunately,  the quadrupole magnet requires full field
 strength operation to  produce this field gradient even when processing
 strongly magnetic materials.   With use of the above idealized equations,
 it can be seen that  a  quadrupole of about 16.5 ft length by 1 ft diameter
 opening will be required for  100% removal of 20-micron diameter magnetic
 particulates from a  8000 ton  coal per day process stream flowing at
 velocities of 1600 ft/min.
     1000
             100% REMOVAL OF PYRITES TO 10/J
             PIPE DIAMETER= 1 FT
      01
              100
                      200
                                                     600
       300      400     500
        FLOW VELOCITY (FT/MIN )
Figure 11 QUADRUPLE LENGTH REQUIREMENTS
                                                            700
                                                                    800
          It  is  interesting to note the significant reduction  in quadru-
pole lengths  that  could be achieved if increased field strengths could
be utilized.  With pole tip fields of the order of 100 kOe, quadrupole
magnetic separators would begin to be competitive with HGMS structures
in this application.

-------
                                                                           131

     1.   Magnetic Structures

          Quadrupole magnet costs and power consumptions have been
calculated from design bases developed at the Lawrence Berkeley Labo-
ratory [20].  It is of interest to note that there is a power/field-
gradient trade-off implicit in the design of quadrupole magnet structures.

          In Figure 9, the surface shape of the "ideal" pole for the
quadrupole structure is shown by the dotted lines.  To make the "real"
quadrupole, shown by the solid line, the iron of the "ideal" pole pieces
has been shaved back in order to make room for the electromagnet windings.
The windings are shown as circles in the figure.  Because of the relatively
small  space made available in this shaving operation, the conductor
cross-section per winding is limited so that quadrupoles are inherently
power-consuming devices.

          A conservative design basis has been chosen whereby moderate
pole piece field  strengths can be developed without  excessive power
requirements  (Modification III, pages 8 and 9, Reference  [20]).

     2.    Process Characteristics

           It  has  been noted  that  smooth  bore  magnetic  separators  will be
 inherently low pressure drop devices.  By  way of  comparison,  it can be
 calculated that the power  requirement of all  fans to supply the quadrupole
 magnet installation to handle 8000 T/D coal is only about 100 hp  as com-
 pared  with 1400 hp for HGMS desulfurization.   While quadrupole particulate
 removal performance may not equal that of  the batch-operated device, it
 well could prove to be the case that ancillary process requirements
 might dictate cost/benefit-type compromise in determining overall process
 optimization.

-------
 132

 CONCLUSIONS

          A  technical  and  cost  comparison  for HGMS and quadrupole magnetic
 separators is  shown in Table  III.  The HGMS installation is estimated  to
 be about a factor of 5 less expensive than the bank of 33 quatrupoles
 and consumes about one-seventh  the power.  If auxiliary fan requirements
 are now included, however, the  HGMS installation cost estimate increases
 to about one-fourth that of the quadrupoles, and the total power consumption
 estimate increases to  about one-half that  of the quadrupole installation.
 Including fan requirements, operating costs estimated for the HGMS
 installation are only  about one-half that  of the conventional iron-
return-electromagnet quadrupole structure  (power assumed at 33 mills/kW-hr).

                               Table III
         Conceptual Level Comparison of HGMS and Quadrupole
           Magnetic Separators for Dry Coal Desulfurization
                                                 HGMS        Quadrupole
100% Removal Size (u-esd)
% Removal - All Pyrites
% Coal Recovery
V (fpm)
No. Magnets
Duty Cycle (%)
Field Strength (Oe)
Total Cost ($)
All Magnets
Canisters
Matrices
Power Supply (including
15% uncertainty)

$/ton Clean Coal
Installed Capital Cost
Actual Process CFM
3
96
98
1600
4
50
2000

327 000
175 000
97 000

28 000
627 000
0.03

$1.22/CFM
20
82
97
1 600
33
100
10 000

3 123 000
19 000
o

201 000
3 343 000
0.13

$6.6/CF*

-------
                                                                             133
          It is now interesting to again recall the results of Figure. 11
which suggest that if superconducting or pulsed field quadrupole magnets
could be developed to significantly increase pole tip field strength,
then quadrupole units would be more competitive with HGMS on a cost-
effective basis.  While power consumption could be expected to be sig-
nificantly lower for a superconducting magnet installation, for example,
capital costs cannot be assessed at this time because there is no history
of commercial superconducting magnet installations.  A factor of 10
increase in field strength will result roughly in a factor of 10 reduction
in pole length, but the economy of scale in superconducting installations
is now known.  If capital costs were reduced by a factor of four, operating
costs for the two approaches would be similar.  The point is that one
cannot properly estimate magnetic separator costs alone for comparative
purposes but must estimate the separator costs within the process context.

          The mathematical model of capture in the HGMS device that  has
been employed in this study is a straightforward modification of a model
successfully developed for describing the capture of weakly paramagnetic
materials in the wet beneficiation of kaolin clay.  The modelling reported
here has, accordingly, assumed a modest surface loading of 30 wt%.   While
this assumption may be appropriate to applications such as coal desulfuriza-
tion, where nonmagnetic particulates, in particular fine clay components,
are inadvertently mechanically captured, it may be overly  conservative
in applications  such as steel furnace off-gas  scrubbing where particulate
concentration is very low, but where  the percentage of magnetic particulates
is high.

          In  the magnetic capture  of  very  fine size and  strongly magnetic
materials,  the  continued  capture of  fine particulates may  even be enhanced
by  the  history  of  fine particulate capture—up to  a point.  There are

-------
 134

 local  surface imperfections on  the  capture material which produce highly
 localized capture forces distributed over the entire surface of the
 magnetized bed which are believed to be selective to micron and sub-
 micron size particulates.  Very fine size and strongly magnetic par-
 ticulates which are captured at these force centers may have a tendency
 to create new sites for further fine particulate capture.  Turbulent
 flow aids this process in bringing  the particulates to the capture
 surface from all directions.

          Paramagnetic particles, on the contrary, are believed to be
 captured preferentially on the  leading and trailing edges of the magnetic
 capture material in the regions of  stagnation points for laminar flow  [14].
 Magnetic filter performance degrades continuously with time in the
 processing of these materials (kaolin clay) because of the extremely
 weak forces for capture of paramagnetic materials.  This loss of efficiency
 is due to particle buildup on the capture surface.  In effect, the
 capture of paramagnetic material destroys surface area for collection.
 By way of contrast, the capture of very fine size but strongly magnetic
materials creates rather than destroys new capture sites.  This being
 the case, one would expect HGMS filter performance to be relatively
 uninfluenced by time in the capture of strongly magnetic materials until
 such time as a breakthrough occurs due to turbulent breakup of the
 captured coagules.

          The magnetic desulfurization of coal described here illustrates
an extreme and difficult application of magnetic separation technology.
 In this example, particle size  distribution is very broad, the flow
velocity is very high,  and restrictive pressure drip requirements are
imposed by process considerations.  The performance characteristics
calculated for both HGMS and quadrupole separation indicate an attractive
possibility for the use of magnetic methods in gas-particulate separations
for strongly magnetic particulates.   Some very interesting magnetic
separator development lies ahead.

-------
                                                                            135
ACKNOWLEDGMENTS

          The results reported here are a continuation of work originally
supported by the U.S. Bureau of Mines.  It is a pleasure to acknowledge
technical assistance provided by Messrs. F. A. Karlson, J. Vicory, and
John McCaig, and to thank Messrs. G. H. Dyer and E. L. Ekholm for making
Bechtel resources available for this investigation.

-------
 136

                              REFERENCES

 [1]  "Proceedings of  the High Gradient Magnetic Separation Symposium,"
     Massachusetts Institute of Technology, Francis Bitter National
     Magnet Laboratory, Cambridge, MA, June 22, 1973.

 [2]  Oder, R. R. and  C. R. Price, "Brightness Beneficiation of Kaolin
     Clays by Magnetic Treatment," TAPPI 56. No. 10, October 1973,
     pp. 75-78.

 [3]  Aubrey, W. M., Jr., et al., "Magnetic Separator Method and Apparatus,"
     U.S. Patent No.  3,608,718, issued September 28, 1971.

 [4]  Good, J. A. and  E. Cohen, "A Superconducting Magnet System for a Very
     High Intensity Magnetic Mineral Separator," IEEE Trans, on Magnetics.
     MAG-12. No. 5, September 1976, pp. 503-506.

 [5]  Oder, R. R., et  al., "Technical and Cost Comparisons for Chemical
     Coal Cleaning Processes," American Mining Congress Journal. 63,
     No. 8, pp. 42-49, August, 1977.

 [6]  McCartney, J. T., et al., "Pyrite Size Distribution and Coal-Pyrite
     Particle Association in Steam Coals," Report of Investigation 7231,
     U.S. Department of Interior, Bureau of Mines, February 1969.

 [7]  Kester, W. M., et al., "Reduction of Sulfur from Steam Coal by
     Magnetic Methods,"  Mining Cong. J.  53.  1967, pp.  70-75.

 [8]  Karlson, F. V., "High Gradient Magnetic Desulfurization of Pulverized
     Coal," Bechtel Corporation,  May 1976, unpublished.

 [9]  Ergun, S.  and E. H.  Bean, "Magnetic  Separation of  Pyrite from Coal,"
     Report of Investigation 7181,  U.S. Department of Interior, Bureau of
     Mines, September 1968.

 [10]  Kindig,  J. K.  and R.  L.  Turner, "Process for Improving Coal,"
     U.S. Patent No.  3,938,966,  issued February 17, 1976.   See also:
     J.  K.  Kindig and R.  L. Turner,  "Dry  Chemical Process  to Magnetize
     Pyrite and Ash for Removal  from Coal," Preprint No.  76-F-306,
     presented at SME-AIME Fall Meeting,  Denver,  CO,  September 1976.

[11]  Power, L.  F.  and H.  A. Fine,  "The Iron-Sulfur System," Mins.  Sci.  &
     Engrg. 8.  No.  2, April 1974,  p.  107.

[12]  Besnus,  M.  J.  and A.  J.  P. Meyer,  "New Experimental Data  on the
     Magnetism of Natural Pyrrhotite," Proceedings of International Con-
     ference  on Magnetism,  Nottingham,  England,  1964, pp.  507-511.  See
     also:   F.  K.  Latgering,  "On  the Ferrimagnetism of Some Sulfides  and
     Oxides," Philips Research Reports, Vol.  11,  1956, pp.  190-249.

-------
                                                                                137

                             References (cont'd)

[13]   Porter,  C.R.  and D.N.  Goens,  "Magnex Pilot  Plant Evaluation — A
      Dry Chemical  Process for the  Removal of Pyrite and Ash from Coal,"
      Preprint No.  77-F-352, presented at SHE Fall Meeting, St.  Louis, MO,
      October  1977.

[14]   Oder,  R.R.  and C.R.  Price, "HGMS:  Mathematical Modeling of Commercial
      Practice,"  AIP Conference Proceedings, No.  29, Magnetism and Materials,
      1975,  pp. 641-643.   See also:  Oder, R.R.,  U.S. Patent No. 3,985,646,
      "Method  for Magnetic Beneficiation of Particle Dispersion," issued
      October  12, 1976.

[15]   "Coal  Preparation for Combustion and Conversion," prepared for the
      Electric Power Research Institute, Research Project 466-1, Final Report,
      January  1,  1977.  Gibbs & Hill, Inc., 393 Seventh Avenue, New York, New
      York.

[16]   Bitter,  F., "The Design of Powerful Electromagnets, Part 1.  The Use of
      Iron,  RSI 7.  p. 479, 1936.

[17]   Allen, J.W.,  "Magnetic Separator," U.S. Patent No. 3,819, 515, issued
      June 25, 1974.

[18]   Oder,  R.R., "Magnetic Desulfurization of Liquefied Coals:  Conceptual
      Process Design and Cost Estimation,"  IEEE Trans, on Magnetics MAG-12,
      No. 5, September 1976, p. 535.

[19]  Aubrey,  W.M.  and R.M. Funk,  "The Quadrupole Magnetic  Separator — A New
      Concept  in Magnetic Separation Equipment," presented  at the  32nd Annual
      Mining Symposium, Duluth, MN, January 1971.

[20]  Hubbard, E.L.,  "Design of Strong Focusing Magnets  for Linear Accelerators,"
      Engineering Note, File No. CV-7, Radiation Laboratory, University of
      California at  Berkeley, November 15,  1953.


WRITTEN  DISCUSSION

E. Maxwell
Francis  Bitter National Magnet  Laboratory,*
Massachusetts  Institute of  Technology
Cambridge, Massachusetts  02139

      Dr. Oder has  presented  an interesting scheme  for processing dry coal by
HGMS.  I would like to stress  however,  that there  is really no  experimental data
at this  time  to  support the implicit  assumption  that the separation of pyrite from
coal  suspensions in high  velocity  air  streams will  be as straightforward as
the separation from water slurries.  We have done  a little work on this problem and
can report  that  the problem is more  complex.   Coal  which can  be successfully
processed in  water  slurries by HGMS  does not respond when simply propelled through
a matrix in high air stream.   Agglomeration and  stickness appear to be major
problems.  It may  be that pretreatment of the coal  to enhance the magnetic sus-
ceptibility of the pyrite may help overcome this difficulty but this is yet
to be established.

-------
138
        Successful dry magnetic separation of  coal was  reported  ten years ago
  by Kester et al.;1  they were most successful with an Upper Freeport  coal,
  removing 80% of the pyritic sulfur.  They used a Frantz  Isodynamic  Separator
  which is not a high gradient device but rather a laboratory instrument for
  making a split between a magnetic and a non-magnetic  fraction.  The mineral
  particles slide down a tilted inclined plane.  The magnetic force competes
  with the gravitational force to separate the particles into separate  streams.
  The particles move rather slowly under the influence  of  these  two forces and
  are not air propelled.  It is essentially an open gradient device and there-
  fore in the same general category as the quadrupole separator  but substitutes
  gravity feed for air propulsion.  At the Auburn Conference on  coal desulfuri-
  zation by HGSM, H.H. Murray2 reported on some HGMS separation  of dry  coal again
  using gravity feed.

        Whatever the other problems of magnetically desulfurizing coal  may be,
  the conversion of pyrite to pyrhhotite is a forward step if it can be carried
  out economically.   It would be useful to have more data on the reaction ki-
  netics of the conversion process.  The magnetization of the FeSx sulfides is
  a sensitive function of the composition parameter x, as illustrated in the
  figure.   In his calculation of the magnetization of the converted pyrite Dr.
  Oder assumes the end product is simply Fe?S8.  Actually a continuous  range of
  compositions is possible;   the other stoichiometries have lower magnetization.
  Incidentally,  the composition Te.^ ^S^ does not correspond to Fe?S .

        The temperature dependence of the magnetization is quite complex for some
  of the sulfides,  as shown  in the work of Lotgering.3  From his investigations
  it appears  that rapid quenching of the pyrhhotite from 200°C.  to ambient
  temperature should  result  in a significant  increase in the magnetization.   The
  separation  should  therefore be carried out  in a cooled air stream and not  at
  the Reaction temperature.

        Dr. Oder  mentions that because  of  the toxic properties  of iron pentacarbonyl
  a  lock hopper is required  to contain  it.  How good  must  the lock hopper  be?   What
  is the experience in handling the  relatively  large  quantities  of similar  toxic
  materials under conditions  simulating power plant operation?   Power  plant
  operators are generally reluctant  to  undertake operations which require  chemical
  engineering  expertise because  theirs  is  a different  kind of operation.

        Dr. Oder  points out  that  the  power consumption  of  the fans required  to
  deliver the  air for  propelling  the  coal  is  a  significant  factor in the overall
  cost.  Is it worth considering alternate schemes  for  carrying  out the beneficiation?
  For example, one might  visualize a  process  in which  the  separation was carried
  out in a separator by gravity  feed  and the  coal was  then  pelletized.   In that
  case the processing  could be done near the mine and  the power  plant  would be saved
  the task of  disposing of the reject material  and  paying  for its freight.

-------
                                                                                 139
REFERENCES




[1]   W.M.-Dester et al., Min. Cong. Jour. 53_, 70  (1967).




[2]   H.H. Murray, IEEE Trans. Magn., MAG-12, 498  (1976).




[3]   F.K. Lotgering, Philips Res. Repts., :U, 140  (1952).
*Supported by the National Science Foundation.

-------
140
               20
             o
                    I  I   I  I  I   I  t>?  I  I
                                           I   I  I  I
                            1.05
110
X
1.15
                                                                20

-------
                                                                              141

OPEN DISCUSSION


Oder:  Dr. Maxwell mentions, for reasons other than what I had pointed out,
that the straight use of the magnetic separator in the processing will not
occur.  That is true.  I don't know whether it has to do with agglomeration
or not.  Our point simply is that, just from an economic standpoint, you
need a 20 kilo horsepower magnet to handle the process strain.  Those are
expensive to make.  From that standpoint alone we do not think it is
economical, and suggest looking at alternative approaches such as Hazen .
There are chemical and thermal methods that do the conversions well.  With
regard to the agglomeration problem it is my hunch that turbulent flow through
that device actually aids magnetic separation.  I can not prove it, but it
aids it in two regards.  First is the stripping that was mentioned.  This is
a very difficult gas - patticle separation.  Because you have got a surface
for capture and therefore you can capture magnetic particles.  You can also
mechanically entrain non-magnetic particles.  This is to your disadvantage
with HGMS technology because you eat up captured surface.  If you have got
a turbulent drag force that carries those non-magnetic particles out of there,
then you don't hinder your captured surface characteristic which is good.

     The other point is that when you capture a strongly magnetic particle
on a captured surface, if that particle is itself equivalent  in magnetization
to the capture surface, then you create more capture surface.  The turbulent
flow will bring particles in to the new sight from all directions.  Therefore
the probability of capture  is high.

     Again the feeling is that processes like Hazen  are good.  They have
done this at Debang.  They  have a big scale unit that is running.  They also
have a pilot unit that has  been running for four months.  In  that four
month period, they have brought non-compliance coal into compliance and they
have done this basically so far with coarsely pulvarized coal (only going   to
14 mesh).  They select their coal so the pyrites have a coarse size
distribution.  They  have been clever in that regard.  They choose Upper Freeport
      coals or particular coals that are pyrite, they respond well, and don't
have  to be pulverized.  My  point  is  that,  if you are going to look at mid-
western and eastern  coals,  you are going to have to face  the  problem of
pulverizing it.   To  do that you are  going  to have  the technology  to handle
pulverized coals.  I don't  think  it  will work unless you  convert  to pyrites.

Maxwell:  On  the  role  of turbulence, I  think  if you manage to capture  the
pyrhhotite, since it is more magnetic,you  have a chance  to strip  any coal
attached  to it by turbulence.

      I don't  think captured pyrhhotite  particles will act very effectively as
nuclation centers simply because  the magnetization of pyrhhotite  is not  that
 large.  We have  seen,  for  example,  studies of  captured magnetite  and hematite
 particles.  Magnatite particles have a  magnetization saturation  of  about  one-
 third that of iron and act  as  a nucleation center.   Drum separators are  an
 example.   Hematite on the  other hand, which is probably  down in  order  of

-------
142


 magnitude from magnetite, is perhaps probably more comparable to pyrhhotite
 and doesn't behave in this way.  It collects until it shields the captured
 particles from the rest of the other particles then you no longer capture
 particles.  So I don't really think that pyrhhotite particle would increase
 your effective capture.

 Friedlander:  At higher velocities we have more problems, strangely -enough,
 with liquid systems and we get erratic behavior.  You can make matters worse
 by increasing velocity to the region of unsteady flow.

 Baril:   You are proposing to do a process that people demonstrated could be
 done electrostatically many years ago.   That process has not been successful
 or applied.   The reason is that not all coals are the same.   The objective
 is to remove the sulfur not just the pyrite.  Sulfur occurs  both in the
 form of pyrite and organically formed sulfur which is not released by grinding
 Incidently all pyritic sulfur is not released by grinding in some coals.
 Generally you grind out to 200 mesh to  relieve most of it in most coals.   If
 you don't remove all the sulfur,  you have to process most coals through some
 other operation to meet the emission standards.   So actually you have not
 accomplished anything more than has been accomplished 10 or  15 years  ago.

 Oder:   That  is not true.   We have studied the chemical processes in great
 detail.   We  studied the TRW process and the whole schlemiel.   The idea is to
 look at  the  limit on the organic  sulfur level that you think you must remove
 to get a significant upgrading of our Eastern reserves.   You take a look  at
 that and the cost of chemical cleaning,  projected at $20  per ton,  this implies
 that you must come back to physical cleaning and  do a better job of it,
because  you  are talking  about dollars per ton.  You have to  draw a line in
 there somewhere.   There  are  some  coals  that can be cleaned well  physically,
 some chemically and some  converted  through gasification  and  liquefaction.
But there is no single  technology that  is going to handle it all.

Baril:   Electrostatic techniques  require less  power than magnetic  techniques.
There are existing  large  scale  separators in operation like  the Wabush  in
Canada.   They  are removing magnetite from sand and  upgrading  iron work; it is
commercial.  What is  the  advantage  of going  to magnetic  if there is another
process  which  is  perhaps  cheaper?   It may be more  expensive  for  initial
installation costs;  I do not  know.  But  the  operation  costs would be  lower.

Liu:  On behalf of  all the magnetic people,  I would like  to defend your
criticism about comparison of magnetics and other approaches.  There  is various
documentation  I can show you  that your argument is not necessarily true.  Over
17% of the annual coal production in the U.S. is  low enough in organic sulfur and
less than  0.7%  of  that can't be cleaned magnetically.  It costs less than all
the chemical or physical processes.

-------
                                                                             143
           MAGNETIC SEPARATION OF PARTICULATE AIR POLLUTANTS
                             D.  C. Drehmel
                  U.  S.  Environmental  Protection Agency
              Industrial  Environmental  Research Laboratory
                  Research  Triangle Park,  N.  C.  27711

                                  and

                             C.  H. Gooding
                       Research Triangle Institute
                             P. 0. Box 12194
                  Research  Triangle  Park,  N. C.   27709
INTRODUCTION
In the last several  years, particulate control  technology has been
developed to the extent that methods are now available to control  large
particle emissions from most industrial sources with an efficiency of
greater than 99 percent. Emphasis on particulate control has now shifted
toward the fine particle size range, particularly to particles which
have diameters between 0.3 and 3 micrometers.  These particles are of
primary interest because they tend to remain in the atmosphere for long
periods of time where they may cause atmospheric haze and be transported
to cause a long range adverse health impact.  They also are the particles
which readily enter and deposit in the human respiratory system.

The three major conventional technologies for the control of fine particu-
late emissions are electrostatic precipitation, wet scrubbing, and
fabric filtration.  The cost of applying these control methods and the
fractional collection efficiency obtained vary according to the specific
application.  Recent EPA research in particulate emission control has
followed two complementary courses:  (a) study of the fine particle
collection characteristics of the conventional technologies and (b)
evaluation of viable alternatives to the conventional technologies. Both
of these courses  have as their  goals the enhancement of fine particulate
collection efficiency and the reduction of  the capital  and operating
costs of industrial emission control.

In the  last decade  research and commercial  applications have demon-
strated  that  high gradient magnetic  separation  (HGMS)  is an  effective
and economical method of  removing small paramagnetic  particles  from
selected liquid  streams.   In this study, published  theoretical  and
experimental  reports of HGMS were utilized  to  evaluate  the  potential
success  of  the  process  in  removing  fine, paramagnetic  particles from gas
streams.   Industrial sources of fine,  paramagnetic  particles were
identified, and  the steelmaking basic  oxygen furnace  (BOF)  was  selected
as  the  primary  candidate  for  initial  HGMS evaluation.   A  bench-scale

-------
144
  apparatus was constructed and HGMS experiments were conducted using
  redispersed BOF dust.   With existing theoretical  equations  as a  guide-
  line, the experimental  data were analyzed and favorable operating
  conditions were identified.

  THEORETICAL ASPECTS OF  HGMS

  To evaluate the usefulness and limitations of high  gradient magnetic
  separation, the theoretical basis of the  concept  must  be examined.
  Engineers have long sought a simple but accurate  theoretical  model to
  describe conventional filtration.  The mathematics  of  most  useful models
  quickly becomes unwieldy, necessitating simplifying assumptions. Never-
  theless, the solutions  often yield valuable information about the
  phenomenological  effects  of key operating variables.

  The recent development  of HGMS models has followed  much the same course.
  Watson has published two  papers [1,2] describing  the single-fiber model
  depicted in Figure  1.   The radius of the  cylindrical ferromagnetic wire
  is a, and the radius of the particle is R.   For r»a,  the velocity of
  the particle is V  and  parallel  to but in the negative  x direction.
                                                           0    R
         Figure 1.  Coordinate system for HGMS single-fiber model.

-------
                                                                             145
For r»a, the magnetic field strength is Hg and parallel  to but in the
positive x direction.  Near the wire the field is concentrated, and a
net magnetic force acts on the particle, changing its trajectory.
Watson developed the expression for the magnetic force and incorporated
it in the equations of motion to describe the particle trajectory.

Watson presented an argument that the inertial force terms in the re-
sulting equations are negligible in comparison to magnetic and viscous
force terms [1], and finally obtained for the radial component of motion
                                                 + Cos 201
                                                                      (1)
and  for the angular component
 where r  =  -£-,  normalized  radial  position,
        a     a
            dimensionless

        a  =  radius of wire, m

        t  =  time, s

       VQ  =  fluid velocity, m/s

        0  =  angular position
V  =
       2 x MB R
 m
         n     — magnetic velocity, m/s
         y n a
 where x  = magnetic susceptibility of particle, dimensionless

      M   = saturation magnetization of wire, A/m

      B.  = magnetic flux density, T

-------
 146
           R   =  particle radius,  m

           n   =  fluid viscosity,  Ns/m2

          HQ   =  magnetic field  strength, A/m.
The fluid velocity is assumed to be normal to the wire at some point
upstream where the x-coordinate is large compared to the wire diameter
The solution of equations (1) and (2) can then be used to identify
particle trajectories which result in the particle striking the wire as
illustrated in Figure 2.  The coordinates are normalized by the wire
radius: the initial coordinate corresponding to the maximum y whose
trajectory results in collection,

particles with initial coordinates
                                         , is defined as RC.  All
                                         such that - R
                                                              •  
-------
                                                                              147
To calculate the collection efficiency for an HGMS apparatus containing
many randomly oriented cylindrical fibers, Watson [1] first extrapolated
from the single fiber to a filter element of differential  thickness and
unit cross-sectional area as follows:
                    FAdL  =  total volume of wire,

                    FdL   =  volume of wire/unit cross section of filter
                              enclosure,

                          =  length of wire/unit cross section of filter
                              enclosure,

                    (2R a) =  total capture cross section/unit cross
                      c        section of filter enclosure.

 where  F  =  volumetric  packing density of mesh, dimensionless
                                                          2
       A  =  cross-sectional area of the filter enclosure, m

       L  =  length of the  filter mesh, m

                    2FRrdL
 Thus the quantity —^—  represents the  fraction of particles  captured
 in a differential length of the  filter.  Watson assumed  that 1n  a  filter
 of randomly oriented filters,  only two-thirds  are  oriented  correctly to
 capture particles since  fibers  parallel  to the magnetic  field are
 ineffective.   He also presented an argument showing that the wires in a
 differential thickness affect particles  independently provided the
 packing density does not exceed a critical  value.   Critical  values of F
 for selected values of RC are shown in  Table I.


             TABLE  I.  MAXIMUM VALUES OF PACKING DENSITY  FOR
                          INDEPENDENT WIRE BEHAVIOR [1]
Vm
Vo
400
100
40
10
4
Rc (Case I)
9.45
5.83
4.15
2.39
1.55
Maximum F
0.0092
0.025
0.049
0.182
0.65

-------
148
  The collection efficiency expression was then integrated over the length
  of the filter to obtain
                      p .—^ . ... i    4FRcL
  where P  =  penetration

       Nout=  number of particles out

       Nin  =  number of particles in


  The  quantity (1-P)  represents  the fractional  mass  efficiency  of the
  filter for any  given  size  of particles,  provided that  all  of  the particles
  of that size are  of equal  mass density.

  Other theories  of HGMS have  been advanced  by  Oberteuffer  [3]  and by
  Cummings,  et al.  [4].   Oberteuffer  developed  a force balance  model
  without resorting directly to  the equation of motion.  His result is
  parametrically  similar to  Watson's  but does not yield  direct  quantita-
  tive  efficiency results.   The  major difference between Watson's  theory
  and  that of Cummings,  et al. lies in the calculation of the magnetic
  force on the particle.  In the present treatment (after Watson  [1]) the
  magnetic field  strength in the particle was assumed to be very  nearly
  equal  to the field  strength  outside the particle.  In the strictest sense
  this  implies  that the  potential energy of the particle is zero;  however,
  U is  easily shown  that the  result  is the same as  the more rigorous
  solution obtained  by Cummings,  et al. if the magnetic permeability of
  the fluid  stream  and the particle are nearly  the same  (as will  be the
  case  of paramagnetic particles  in a gas stream).   For the collection of
  ferromagnetic particles this simplifying assumption would be  inaccurate.

  RESULTS OF  PREVIOUS HGMS APPLICATIONS

  High gradient magnetic separation has been applied extensively to particle/
  liquid  separations  [3,5,6,7,8,9,10,11].   The primary commercial application
  has been in  the removal of weakly magnetic color bodies less  than 2 ym
  in diameter from  kaolin clay.  Pilot-scale and laboratory tests have
  been conducted on industrial  waste process water from steel mills and
 electroplating operations,  on nuclear reactor coolants, and on oils and
  hydraulic fluids  (to remove wear particles).   Preliminary experimentation
 has shown promise of application of HGMS to the desulfurization of coal
 and to the reduction of solids  and dissolved  orthophosphate from municipal
 waste.  The latter application  was preceded by seeding  the waste stream
 with a magnetic material that became chemically or  physically associated
 with the pollutant.  Another  recent  novel  experimental  application was
 to the separation of red cells  from  whole blood.

-------
                                                                               149
  Data have been reported from two studies  in  sufficient  detail  to  allow
  comparison to the theoretical  predictions of equation  (3).   In the  first
  case Melville, et al.  [7] studied the high gradient magnetic separation
  of red cells from whole blood.   Red blood cells  are flexible,  biconcave,
  discoid bodies with well-defined dimensions  of 8.5 ± 0.4 ym diameter and
  2.3 ±0.1 urn thickness.  The filter contained 25 ym diameter cylindrical
  stainless steel wire.   Flow velocities varied from 1.6  x 10"4  m/s to
  5.8 x 10'4 m/s, and the magnetic field ranged from 0.6T to  2.4T.  In all
  runs the parameter
was less than 0.2.  The experimental data and
  Case I theoretical  prediction are shown in Figure 3.   The equivalent
  spherical   diameter assumed here for the purpose of modeling was 6.3 urn.
  For any given efficiency, the experimentally required value of
   Vm
  -tp- is about 1  to 1.5 orders of magnitude higher than the theoretical
    o
  value.  Melville attributed this result to a postulated reduction in the
  actual value of Rc which occurred when the bare wire  was covered with
  red blood cells within the first few seconds of filtration.
    l.0r
   o  o oo  o
                                             03 00

                                                  o
                                           THEORETICAL
UJ
z
Ul
OL
       Figure 3.  HGMS theory vs. experimental data:
                  cells from whole blood.
                           removal of red

-------
150

 Oberteuffer [8] reported data for the removal of 10 urn particles of
 cupric oxide from a slurry also containing diamagnetic aluminum oxide.
 Flow velocities ranged from 0.022 m/s to 0.22 m/s and the magnetic field
 ranged from 0.5T to 10T.  Steel wool fiber packing was used. Luborsky
 and Drummpnd [12] reported additional details of the experiment and
 included in their analysis the assumption that the fibers could be
 approximated by an idealized-ribbon geometry (with width, S, and thick-
 ness, 2a).' If only the small edges of this type fiber produce a signifi-
 cant magnetic force on the particles, then for a randomly packed filter
 only one-third of the fibers on the average would be ortented correctly.
 An argument analogous to that presented before then results in the
 theoretical expression
                     P  =  exp   -
                                                                 (4)
 The experimental data are shown in Figure 4. Theoretical penetrations
 calculated from equation (4) with Case I values of RC fall
     1.0 r-
  <
  P  o..h
  u
  ui
  Q.
    001
                                       o
                                                                    o
                           THEORETICAL
                             -
                              3S
                             2B,
                                '=02
                                                                O  0.075,0.12,0.24
       0.1
                             1.0
                                        %
                                          vo
Figure 4.  HGMS theory vs. experimental data:
           particles from a water slurry.
                                                     10
                                                removal  of cupric oxide
                                                                           100

-------
                                                                             151

roughly two orders of magnitude to the left of the data.   Some of the

data corresponding to   2B$   > 0.2 are plotted separately and tend to
                          o
show a slightly enhanced collection.   Luborsky and Drummond [6] were
able to correlate these data with three modifications to  Watson's model:
(1) accounting for the effect on flow potential of particle buildup in
an assumed geometric pattern (the field potential  was assumed to be
unaffected); (2) introduction of an empirically determined factor to
account for the fraction of wires that are active in capturing particles
(in addition to the one-third orientation factor); (3) introduction of
an empirically determined factor to account for mechanical filtration of
particles.

POTENTIAL APPLICATIONS OF HGMS IN PARTICULATE EMISSION CONTROL

Probably the most important parameter to be considered in the evaluation
of potential application of HGMS is the particle susceptibility. Suscepti-
bility of heterogeneous dust particles cannot be predicted accurately
because it is strongly dependent on the distribution of magnetic species
in the dust.  The composition of industrial dust emissions obviously
varies for different processes.  Within a single process category the
composition will vary over a more narrow range with respect to raw feed
and operating variations.  In Table II typical compositions of paramagnetic
compounds are listed for several industrial processes.  Blast furnace
 TABLE II.  POTENTIAL APPLICATIONS OF HGMS IN PARTICULATE EMISSION CONTROL

Industrial Process         References      Typical Dust Composition,
                                                 mass basis

Blast Furnace              [13]        35-50% Fe, 12% FeO, 0.5-0.9% Mn
Basic Oxygen Furnace       [13, 14]    90% Fe203, 1.5% FeO, 4% Mn304
Open Hearth Furnace        [13, 14]    85-90% Fe203> 1-4 FeO, 0.5% MnO
Electric Arc Furnace       [13, 14]    20-55% Fe203, 4-10% FeO, 0.5% MnO
Silico-manganese Furnace   [13, 14]    4-7% FeO, 30-35% MnO
Ferro-manganese Furnace    [13, 15]    6% FeO, 34% MnO
Ferro-chrome Furnace       [13, 15]    7-11% FeO, 3% MnO, 29% total Cr
                                        as Cr203

Coal-fired Boiler          [14]        2-36% Fe203

-------
152
  dust, which has a high reduced iron content, should be relatively easy
  to collect.  Since the dust would likely have residual magnetism,
  cleaning of the filter mesh could be a problem.  Basic oxygen and
  electric-arc steelmaking furnaces are estimated to be prime candidates
  for HGMS control.  Open hearth furnace dust has similar characteristics,
  but most existing open hearth furnaces are well-controlled, and the
  majority of new steelmaking installations will more likely be basic
  oxygen or electric-arc furnaces.  In the ferroalloy industry, processes
  involving combinations of iron, manganese, and chromium have reasonable
  potential for HGMS application although the dust composition of these
  processes is highly variable.  Coal-fired boilers are unlikely to be
  amenable to HGMS application, because the composition of magnetic
  species is relatively low.

  Particle size also varies greatly in different types of dust.  HGMS
  theory indicates that collection efficiency decreases with particle
  size, but experimentation will be required to determine the fractional
  efficiency for particular dusts.  Gas temperature may also be an Im-
  portant factor.  Particle magnetic susceptibility is reduced and gas
  viscosity is increased at increased temperatures.  Both of these trends
  theoretically reduce HGMS efficiency, but the effect could be minor
  provided the efficiency at lower temperatures is sufficiently high.

  DISCUSSION OF EXPERIMENTAL FEATURES

  The bench-scale experiments reported in this paper were conducted with
  an 8.9 cm (3.5 in.) diameter HGMS system collecting dusts from a basic
  oxygen furnace.  An operating temperature of 38°C (100°F) was used.
  Apparatus and procedures were discussed in detail in a previous paper
  [17].  The ranges of important operating parameters are summarized in
  Table III.   Total mass concentrations on the order of 0.01 to 0.1  g/m3
  were produced by the EOF dust redispersion apparatus.   Figure 5 shows
  the typical  particle size distribution at the inlet of the magnetic
  separator as determined by Andersen impactor runs.   The distribution is
  skewed more toward larger sizes than are typical  primary emissions from
  a basic oxygen furnace, indicating that a significant mass percentage of
  the dust was still in an agglomerated form.

  One long term run was made to study the variation in sampling results
  with time and the transient effects of mesh  loading.   The results  are
  shown in Figure 6.  The penetration (1-fractional collection efficiency)
  varied without any pattern for the first 8 hours  of the test and then
  began to rise.   After 10.5 hours the superficial  gas  velocity could no
  longer be maintained at its original value because of the increased
  pressure drop,  and the run was stopped after 11.5 hours.   At the end of
  the run the filter mesh was carefully removed and weighed.  The clean
  weight was  120 g and a total  of 81.8 g of dust had been collected.
  Assuming the total mass collection efficiency was constant over the 11.5
  hour run (neglecting the small decrease in the last 4 hours), these

-------
TABLE III.  RANGES OF OPERATING PARAMETERS

Filter Mesh Number
Filter length, cm
Packing density
Packing orientation
Magnetic field, T
Gas velocity, m/s
Pressure drop, kPa (in. H20)

4
35.6
0.0088
random
0.094-0.308
6.2-8.7
1.2-2.0
(4.6-7.8)
5
20.3
0.0088
spiral
0.094-0.308
5.9-9.3
0.7-1.3
(2.8-5.3)
6
20.3
0.0174
spiral
0.094-0.214
5.9-10.2
1.4-3.2
(5.4-12.8)
7
20.3
0.0132
spiral
0.094-0.214
6.6-10.6
1.2-2.9
(4.7-11.5)
8
20.3
0.0132
spiral
0.094
6.1-7.0
1.2-1.4
(4.6-5.4)
                                                                              U)

-------
    154
   10.0



   8.0



   ao





   40
0   ID


3  o
o


5  0.6
   0.4
   Q2
   Ql
J	'   '
J_
_L
_L
J	L
                                          J	L
                                             _L
                                                         -L
                                             _L
                                                                      _L
     0.10.2  05  I   2    5   10   20  30 40 50 60 70  80    90  95   98 99

                CUMULATIVE MASS PERCENT GREATER THAN INDICATED SIZE
                                                                     99.8993
               Figure 5.  Size distribution of redispersed BOF dust.

-------
  0.08
  007
  Q06
  0.05
  0.04
z
LU
Q.
   0.03
   002
   0.01
MESHES

BQ=Q094T


V0 = 6.lm/s
           0.7-1.0 /xm
                                                                                                                         A
                                                                                                                         H
                                                                                                                            a:
                                                                                                            ui
                                                                                                            tr

                                                                                                            tn
                I
                                                                                     8
                                                                                          10
II
             234567

                                             TIME, hr                                                          £


Figure 6.  Transient loading effects and random variations  in HGMS collection efficiency  and pressure drop.

-------
156
   results  indicate that  the filter  collected  approximately  50  percent  of
   its  own  weight in dust before  the collection  efficiency was  appreciably
   affected.  However, during this time  the  pressure  drop  across the  filter
   also increased by about 50 percent.

   COMPARISON TO  THEORETICAL MODEL

   All  of the filter meshes  used  in  this  study were  made  from a steel wool
   material similar to that  used  by  Oberteuffer, but it was  rolled into a
   spiral configuration in which  essentially all of  the fibers  were  per-
   pendicular to  the field and gas flow.  Reasoning  analogous to that stated
   before implies  that half  the fibers  would have been active in collection.
   Using Watson's  model the  theoretical expression for penetration becomes
                             = exp  -
  Appropriate theoretical values of R. were determined from the case
        M0MS                         c
  where -sn — = 2.  With the low fields used in this work the
  magnetization of the steel wool varied with the applied background field
  in a nonlinear fashion.  Table IV reports the estimated value of fiber
  magnetization as calculated by Magnetic Engineering Associates, a con-
  sultant to the study.  The calculated values of the near field correction
  parameter using estimates of both saturation magnetization and actual
  magnetization are tabulated.
     TABLE IV.  FIBER MAGNETIZATION AND NEAR FIELD CORRECTION PARAMETER

Bo'T

0.094
0.214
0.308
Estimated M


M, A/m

3.2 x
6.8 x
8.4 x
e = 106 A/m

105
105
105

yn M
0
2Bn
0
2.2
2.0
1.7

p M
0 S
2B
0
6.7
2.9
2.0


-------
                                                                               157
At the time these experiments were conducted, equipment to measure the
particle magnetization was not available.  Based on the reported data of
typical BOF dust composition the volume susceptibility was estimated to
be 4 x 10~3 in dimensionless SI units.  Recently, access to a Faraday
apparatus was arranged and three BOF dust samples were processed with
the results shown in Figure 7.  The particulate is much more strongly
magnetic than anticipated.  At 0.094T, the apparent volume susceptibility
with an assumed particle specific gravity of 3.5 is approximately 1.5
dimensionless SI units.

Typical experimental data from this study are presented 1n Figure 8 with
comparison to the corresponding theoretical prediction.  To calculate
the abscissa value for the data, the estimated values M = 3.2 x 105 A/m
and x = 1-5 were utilized.  In general this data set as well as the
remaining data agree more closely with the theoretical prediction than
the liquid system data of Melville and Oberteuffer.  This feature is
perhaps due to enhancement of collection efficiency associated with the
near-field correction.  The near field contribution is apparently important
due to the combination of small particle size, low magnetic field and
high fluid velocity.  Table V summarizes the differences in the values
of these variables for the three experiments.  In view of the significant
difference in the ranges of operating parameters as well as the extra-
polation from liquid to gas systems, discrepancies between this work and
the earlier experiments is not surprising.

Certain other features of the data should also be noted.  Changes in the
abscissa of Figure 8 represent changes in two parameters—gas velocity
and particle size.  Most of the scatter  is believed to be associated
with the measurement of particle size and concentration.  Each data
point represents the average of three or more upstream and downstream
counts with an optical analyzer and sampling/dilution system.  The
individual penetration measurements for a given particle size during a
given run varied by as much as 50 percent from the average.  These
variations were generally larger for 1 to 3 ym particles which appear as
data clustered to the right of Figure 8. The ability to measure low
penetration values for these large particles was also limited somewhat
by low particle counts.  Thus the apparent tendency for the data to
                         V
level out toward larger -n—values could reflect a sampling limitation
                          o
rather than an actual physical trend.  It should also be noted that the
nominal particle diameters indicated by  the optical analyzer are based
on calibration with monodisperse latex spheres.  Since the dust is known
to have contained agglomerated particles (probably of various shapes),
the relationship among the actual physical particle size and shape,
effective particle size for interaction with the wires, and particle
size indicated by the optical analyzer is not clearly defined.

-------
158
          35
          30
          25
       -E?
       I  20
        %
       O

          15
          10
                 O  SAMPLE #1
                 A  SAMPLE #2
                 D  SAMPLE #3
                       0.2
0.4
0.6
0.8
1.0
                            APPLIED MAGNETIC FIELD. OERSTEDS
       Figure  7.   Experimental measurements of particle  magnetization.

-------
                                                                                      159
   1.0
                                   THEORETICAL
   0.1
Z
tc

in
2
   0.01   _
   0.001
       0.01
                               0.1
1.0
                                                                               10
           Figure  8.   HGMS theory  vs.  experimental data:   mesh No.  5,
                       BQ = 0.094T.

-------
                                                                                              ON
                                                                                              o
TABLE V.  SUMMARY OF DIFFERENCES IN THE PRESENT AND PREVIOUS WORK
Parameter
Particles
Particles diameter, ym
Particle susceptibility, SI units
Fiber type, geometry and size
Packing density
Magnetic field, T
Fluid
2
Fluid viscosity, Ns/m
Fluid velocity, m/s

2Bo
Packing parameter

Melville, et al. [7]
red blood cells
6.3
3.0 x 10"6
25 ym diameter cylindrical
stainless steel
0.02
0.6-2.4T
diluted blood
1 x 10'3
(1.6-5.8) x 10"4
0.11-1.6
On/i A 17
. U4-U. 1 /


Oberteuffer, [8, 12]
cupric oxide
10
2.4 x 10"4
20 ym x 200 ym ribbons
stainless steel
0.05
0.5-10T
water
1 x ID'3
(0.022-0.22) x 10"2
0.37-73
.075-1 .5
3S

This Work
EOF dust
0.3 - 3.0
1.5
20 ym x 200 ym
ribbons stain-
less steel
0.0088-0.0174
0. 094-0. 308T
air
2 x 10"5
5.9-10.6
0.17-30
.0-6.7
3T=5'2
^ = 4.5-8.8

-------
                                                                             161
Attempts to compare the effects of individual  parameters to the
theoretical predictions were thwarted primarily by the relatively narrow
range of the individual parameter variations and the scatter in the

experimental data.  The nonlinear relationship of R  and —y—also
                                                            o
complicates individual parameter analysis.  The strongest evidence of
some physical mechanism not accounted for in Watson's theory was seen in
analysis of penetration-vs-velocity data with other parameters constant.
                        V
The theory predicts for ^- <0.2:
                         o
                         P = exp
where k is a constant; i.e., as velocity increases, penetration in-
creases.  In several cases the data tend to show penetration decreasing
with increasing velocity or perhaps a minimum penetration at a mid-range
velocity.  Figure 9 shows one of these cases.  This behavior conceivably
reflects a complicated interaction of high gradient magnetic separation,
inertial impaction (which is neglected in Watson's development), and
reentrainment.

With respect to the magnitude of observed collection efficiencies, the
data in Figure 8 indicate penetrations on the order of 0.01 for particles
in  the 1 to 3 vim range and 0.15 for particles in the 0.3 to 0.5 ym range
with a relatively low applied magnetic field.  In other runs when the
magnetic field or the filter mesh density was increased penetration of
the 0.3 to 0.5 pm particles was easily reduced to 0.05 and below.

CONTINUING WORK

Basically the objective  of this research is to demonstrate on a small
scale  that high gradient magnetic separation may be an effective and
economic technique for the control of fine particle emissions from
selected industrial  sources.

To  satisfy this objective, experiments are continuing on a pilot-scale
HGMS apparatus of approximately 0.7 m3/s  (1500 CFM) capacity.  At least
two levels of the following  parameters will be studied:  superficial gas
velocity, gas temperature, magnetic flux density, particulate source,
and particulate size.

 Particle collection  efficiency will be correlated as a function of
operating parameters either  by an extension of Watson's  theoretical
model  or  if  necessary,  by  a  semi-empirical approach.  The  primary purpose  of

-------
162






z
o
1-
ct
H
UJ
Z
LJ
0.





0.11
0.10
0.09

008
O07
Q06

005



.004
0.03

002
001
0
-
. ° \ °
\
\
\
V
0 \
\
o \
^^
^^^
^^
^^^* ^^^


o
H*
••
	 1 till
                                 8        9



                                VELOCITY, m/s
10
II
    Figure 9.  Anomolous effect of velocity on penetration:  mesh No. 7,
               BQ = 0.094T, 0.3-0.5 ym.

-------
                                                                             163
the correlation will be to develop a reliable basis for designing a
larger scale device with a high probability of success.  A secondary
goal, which is desirable but may prove to be beyond the scope of the
present research, is to develop a model of sufficient completeness and
accuracy to allow elementary optimization techniques to be incorporated
in the scale-up design (i.e., to provide a basis for determining the
incremental cost effect associated with adjustment of the major operating
parameters at constant collection efficiency so that each of these
parameters can be set at its optimum value.)  Although Watson's model
has been useful so far, it may not be sufficient without modification
for at least three reasons:

     1)   The experimental data indicate a velocity effect that is not
          predicted by the model;

     2)   At low fields the fibers are not magnetically saturated as the
          model assumes;

     3)   The particles are ferromagnetic rather than paramagnetic as
          the model assumes.
CONCLUSIONS

The following conclusions have been drawn from the preliminary experiments
and theoretical analysis conducted to date:

      (1)  Fundamental theory predicts the potential success of the
          concept.

      (2)  Preliminary experiments support the potential capability of at
          least 90 percent removal of all submicrpn particulate from BOF
          furnaces.  Collection of submicron particulate might actually
          be enhanced at low magnetic fields.

      (3)  No inherent health and safety  hazards or any other unfavorable
          environmental impacts have been associated with  HGMS application.

      (4)  No prohibitive design and reliability problems have been
          identified.

      (5)  HGMS  is  potentially applicable to several sources of fine
          particulate;  retrofit capability should be enhanced by  the
          high  gas velocities demonstrated in preliminary  experiments.

 ACKNOWLEDGMENTS

 The authors  are indebted to Dr. Larry K. Monteith of North Carolina
 State University  for major assistance in the mathematical  model  in-
 terpretation.

-------
164


  REFERENCES

  1.    Watson,  J.  H.  P.,  Magnetic  Filtration,  J.  Appl.  Phys.,  Vo.  44,  No
       9, September 1973, pp.  4209-4213.

  2.    Watson,  J.  H.  P.,  Theory of Capture  of  Particles  in Magnetic High-
       Intensity Filters, IEEE Transactions on Magnetics, Vol. MAG-11,
       No.  5, September 1975,  pp.  1597-1599.

  3.    Oberteuffer, J.  A.,  Magnetic Separation: A Review of Principles,
       Devices, and Applications,  IEEE Transactions on Magnetics,  Vol.
       MAG-10,  June 1974, pp.  223-238.

  4.    Cummings, D. L., D.  A.  Himmelblau, J. A. Oberteuffer, and 6. J.
       Powers,  Capture  of Small  Paramagnetic Particles by Magnetic Forces
       from Low Speed Fluid Flows,  AIChE Journal, Vol. 22, No. 3, May
       1976, pp. 569-575.

  5.    Kolm, H., J. Oberteuffer, and D. Kelland,  High Gradient Magnetic
       Separation,  Scientific  American, Vol. 233, No. 5, November 1975,
       pp.  47-53.

  6.    lammartino,  N. R., New  Tasks for Magnetism, Chemical  Engineering,
       Vol. 81,  No. 1,  January 7,  1974, pp. 50-52.

  7.    Melville, D., F. Paul,  and S. Roath, High Gradient Magnetic Separation
       of Red Cells from  Whole Blood, IEEE Transactions on Magnetics,  Vol.
       MAG-11,  No. 6, November 1975, pp.  1701-1704.

  8.    Oberteuffer, J.  A., High Gradient Magnetic Separation,  IEEE Transactions
       on Magnetics, Vol. MAG-9, No. 3, September 1973, pp.  303-306.

  9.    Kelland,  D. R.,  High Gradient Magnetic Separation Applied to Mineral
       Beneficiation, IEEE Transactions on Magnetics,  Vol. MAG-9, No.  3,
       September 1973,  pp. 307-310.

  10.   Trindade, S. G.,  and H.  H. Kolm, Magnetic  Desulfurization of Coal,
       IEEE Transactions on Magnetics,  Vol.  MAG-9, No.  3, September 1973,
       pp • «31 U™o Io •

  11.  deLatour, C., Magnetic Separation  in  Water  Pollution  Control,  IEEE
      Transactions on Magnetics, Vol.  MAG-9, No.  3,  September  1973, pp.
      »3 I H~*3 I D»

  12.  Luborsky, F. E. and B. J.  Drummond,  High Gradient Magnetic Separation:
      Theory Versus Experiment, IEEE Transactions on  Magnetics,  Vol.  MAG-
      11, No.  6, November 1975, pp. 1696-1700.

 13.   Katari, V.,  6.  Issacs and T. W.  Devitt,  Trace Pollutant  Emissions from  the
      Processing of Metallic Ores. EPA-650/.2-74-115,  October 1974':	

-------
                                                                               165
 14.  Hedley, W. H., S. M. Mehta, C. M. Moscowitz, A. D. Snyder, H. H. S.
      Yu and D. L. Zanders, Sources and Characterization of Fine Particulate
      Test Dusts EPA-650/2-74-117, November 1974.

 15.  Dealy, J. 0. and A. M. Killin, Engineering and Cost Study of The
      Ferroalloy Industry. EPA-450/2-74-008, May 1974.

 16.  Drehmel,  D. C. and C. H. Gooding, High Gradient Magnetic Particulate
      Collection, 82nd AIChE Nat. Meeting, Atlantic City, August 29-
      September 1, 1976.
WRITTEN DISCUSSION

Duncan R. Hay
Magnetic Engineering Associates, 247 Third St.
Cambridge, Massachusetts  02142

    Because this technique is based on intrinsic magnetic susceptabilitv its
application must be limited to such materials that have magnetic properties.
On a practical level the separation of fine particulates in a large volume
gas stream may only be possible with materials of high magnetic susceptability.

    The experimental results of the Research Triangle Institute (RTI) do support
the validity of the Watson model.  It is especially interesting to note, that by
using the measured (rather than assumed) particle susceptability a much better
fit is found to the theoretical model.

    During the aerodynamic and magnetic field characteristics of the spaces
between the stainless steel fibers is perhaps the most difficult analytical
task,.  The usual model assumes a uniform spacing proportional to the matrix
density and fiber size.  This ignores the actual situation where there is a
large random spacial variation, from touching to many fiber diameters separa-
tion.  It is probable that the ratio of widely spaced to closely spaced
fibers is decreased with increasing matrix density.  This will have a pro-
found effect on the capture of smaller size particles where the viscous
forces are proportionately larger than for large particles.  Investigation
into the specific relationship of capture radius with respect to particle
velocity, the geometric spacing and magnetic field distribution of adjacent
fibers is needed.

    Predicting the performance of a high gradient mangetic separator for air
carried particulates from the results of water mixed experiments has not
been successful.  In the Stokes relationship there is orders of magnitude
difference in the viscosity and density ratio between air and water.  Thus
any scaling would be quite sensitive to the actual experimental conditions.

-------
166

 A steel wool matrix would probably  show  a poorer  correlation  than  an  expanded
 metal  one because  of  greater  displacement effect  of high  density water  on
 individual  fibers.  The  literature contains  some analyses of particle
 trajectories in a  low density medium.  However none has considered local density
 variations  in a high  velocity air stream passing  through  a fiber array  and how
 this may affect the path of very  small  particles in  the  superimposed magnetic
 field  gradients.

    The high magnetic gradient  filtration of  air  carried  ferric oxide par-
 ticulate from steel making processes ia  a very appropriate application  being
 investigated by RTI.  It is also being evaluated  by other groups for  desulp-
 hurization  of an air  powdered coal  mixture.

    Another potential application lies in the filtration  of petroleum gas
 streams containing finely divided iron and sulphur compounds.  A rather unusual
 problem is  found in high temperature utility  steam turbines.  Where iron oxide
 particles exfoliate from the  pipe walls  and are carried with  the stream into
 the turbine causing blade erosion or in  the extreme,  blade rupture.   It is
 possible to concentrate the particles with inertial and magnetic gradients
 and separate this  from the main steam flow into a high gradient magnetic
 separator.

    In all  of these applications the capability of operating  up to 500C is a
 distinct advantage of the Magnetic  Precipitator.  (This designation has been
 suggested by Henry Kolm and seems very appropriate to describe the action of
 high magnetic gradient devices for  air/particulate mixtures.)  Of  course the
 saturation  field for the matrix material will  decrease with increasing
 temperature, (to zero at the  curie  point) which could result  in a  requirement
 to increase the magnetizing field to  compensate.

    The design of magnetic precipitator  for large volume  process rates  most
 conveniently takes the form of several large  solenoids in parallel.   The
 collected magnetics are cyclically  flushed out of each unit in turn by
 turning off the magnet.  A conventional bag house or  spray collector  is needed
 to receive  the blow down.  Because  the buildup time for the magnetics is long
 for the usual low loading rates, flushing may  be needed at 5 or even 20  minute
 intervals and experiment has  shown  flushing is complete in less than  5  seconds
 at process  flow rates.  Because of  this, the  flush may be dumped in a large
 volume receiving tank and the particulate recovery filter operated at perhaps
 5% of the system flow rate.   Otherwise the flushing filter requirements would
 dominate the whole installation which is completely unacceptable.   The  size
 and number  of solenoid magnetic matrix units  would depend on  the particular
 application, with consideration for  flow rate, operating  cost and  capital costs.

    The alternative designs use a moving matrix,  either rotating or oscillating,
where collection of the particulates in the magnetic field is continually
 passes through a field free region  for flushing.  The two principle differences
 compared to the solenoid geometry are decreased magnetic  efficiency and the
 need for moving seals to maintain the pressure difference across the matrix.
 These requirements must be evaluated against  the  need to valve or  direct a large
 volume air  flow from  flushing to a receiver for each solenoid.  Treating the

-------
                                                                               167

flushed mags is the same for each design.

REVIEW OF EXPERIMENTAL RESULTS

    The experimental results on EOF dust obtained by RTI in their laboratory
work using a SALS HGMS system clearly show a velocity dependent penetration
fields up to 3 kilogauss.  The effect is most pronounced for particles smaller
than about 0.7 urn and is field dependent with a minimum in the 1400 to 1700 ft/min
velocity range.  For particles of 1 to 2 y (and larger) diameter magnetic
saturation takes place at 2 kilogauss with almost 100% filtration efficiency.

    A logical explanation for this velocity dependency suggests that at lower
flow rates the particle. flow  is  laminar and  stream  lines           do  not come  close
enough to the fibers for capture in the magnetic gradient.  At the highest
flow rates the fluid drag is stronger than the magnetic attraction and
particles are reentrained.

    A long term run to measure the maximum loading capacity of the matrix
resulted in confirming the usual assumption that the accumulation can equal
the matrix weight.  Insufficient data was taken on the effect of matrix length
to report but penetration appears to be inversely linear with length.

    There is a strong inverse dependency of matrix density and penetration
for particles of less than 0.5y the minimum penetration decreases from 8 to
2% for the smaller particles to the reduction  of average and of the maximum
distance between matrix fibers, but is obtained at a two fold increase in the
pressure drop across the matrix.  Since pumping power is about 10 times more
costly than magnet power it appears there will be a trade off between a maximum
achievable  reduction in the smaller particulate effluent and an economically
acceptable performance.

    One aspect of experiments with air entrained particulates is finding a
satisfactory technique  for introducing properly dispersed material.   So  far,
the use of a fluidized bed has been used to break up agglomerations of particles,
but the maximum size which can be dispersed in near 30 ym and the  total mass
flow which can be entrained is perhaps 50 times smaller than desired  loading
per cubic foot of air.  Other particulate dispersal methods should be developed
to give a much broader  control over experimental particle loading  rates.


OPEN DISCUSSION

Kelland:  When the particulate is on the side of the fiber, the gradient is in
the reverse direction.  Not only does the viscous force take the particle off
but the magnetic  force  tends  to push it away from the fiber also.  However,
there were some experiments done 5 or 6 years ago at MIT by Dr. John  Oberteuffer
in which expanded metal matrices were used in three different orientations
to measure collection efficiency.  Almost no difference in collection efficiency
was found.  How that translates  to the practical devices of today  I am not
sure.

-------
 168


 Gooding:   Charlie  pointed  this  out also;  that his fibers were not round
 but were  usually more  flattened.  The orientation of  the flatness may have
 an effect on  separation  efficiency.

 Friedlander:   I disagree with that and have pictures  to prove otherwise!
 The capture effect change  is not  true because the magnetic force tries to
 hold  them there.   What happens  at higher  velocities,  however, I do not know.

 Olson:  Mr. Gooding has  followed Mr. Watson's analysis very well, however, if
 one considers  a particle trajectory model a drier separation occurs.  In
 particular, the conditions that Gooding and Drehmel considered (Stoke's
 number between .1  and  30)  is a  region on  the low side of transition; from
 a noninertial  region where the  particle inertia is not important to a region
 where the particle inertia is very important.  So, in response, as
 Gooding suggested, there will be inertial impaction in filters over much of
 the velocity and particle  sizes he considered.

      The same would also be expected of  the more conventional high gradient
 magnetic  separator  Robin Oder considered.  In addition, in the Gooding-
 Drehmel work,  the  Stokes drag turns out not to be a very good approximation
 for the viscous forces on  the particle.   They are in  the transition region
 where the Stokes theory  for the flow around a particle breaks down.  So
 there will perhaps be  a  different dependence upon the relative particle fluid
 velocity  rather than a linear one.

 Gooding:   I would  just say that I agree with what you said.  This is something
 we have put some thinking into.   But we have not gone through to try to improve
 the model.

 Friedlander:   It seems to me that the performance goes in the opposite direction
 than what Maxwell  said a little while ago about much worse performance at higher
 velocities.  If I read your graph correctly, improved performance comes
 with higher velocities.

 Gooding:  This is a single component system where you are not trying to separate
 two different  solid particulates in an airstream. It is a much simpler problem
 than the problem where you have two components that you are trying to separate.
 I would not compare them at all.

 Friedlander:  Without  taking away from Watson, I think one has to say that
 Zebel did the basic thing.  We should give credit to Zebel.

 Oder:  I  think we need a knock-down drag-out session just on magnetic separation.
 Now you know why I chose the example I did, I knew the magnetic component you
would be looking at.   I jimmied  up the Hazen process to come out with moments
 in the same ballpark.   We are not that far apart and your flow velocities are
 of the order of 1000 ft/min.   If you can convert coal minerals to a very strong
magnetic component, you get some indirect proof that it should work.

-------
                                                                                169
Gooding:  I think your velocity dependence looks like inertial deposition.
I think there are two things going on, inertial deposition and reentrainment.
I have played with the mathematics a little bit.  We may find that there is
a minimum or optimum velocity.

Oder:  Did you ever run with your magnet turned off?

Gooding:  Yes.  The figure shows no field, and you see the larger particles
are getting as high as 50% collected.  There is a fair amount of uncertainty
in this data.  This is one of the reaons we have not tried to do any more
sophisticated modeling.  Referring back to this slide, we had indications
that some of the larger particles were agglomerated as they impact on the
wire, which would mean that the zero field numbers are suspect.  We have
improved that aspect of our experimental apparatus by building a scaled-up
version. We are also taking data as a baseline at zero field.

     When you are dealing with a ferro-magnetic particle, another force that
you really should consider is one that you do not normally calculate in the
analysis.  If you have a ferro-magnetic particle, particularly a very small
one and it gets impacted on a ferro-magnetic surface which is larger, the
image force due to the moment in the  ferro-magnetic wall surface should be
considered.  In other words,  the particle itself is a source of magnetic field
gradient which makes it stick to the wall.  This operates only at distances
comparable to the diameter of the particle.  It is something which should be
taken into account when you are dealing with ferro-magnetic particles.  This
may have some relation to what you  seek in your zero field measurements.  Be-
cause,  even  though you turn your field off, if  this is steel wool or something
like that, it does have a small amount of residual magnetism.  It will pick
up ferro-magnetic particles.  I believe you are correct.  In our present
experiments we have  designed  it so  that all our zero magnetic field data is
taken before we magnetize the matrix.

-------
 170
 RESEARCH NEEDS AND OPPORTUNITIES IN HIGH GRADIENT MAGNETIC  SEPARATION  OF
 PARTICULATE-GAS SYSTEMS
     Y.  A.  Liu  and  C.  J.  Lin

     Magnetochemical  Engineering  Laboratory
     Department of  Chemical  Engineering, Auburn  University, Auburn, Alabama
     36830
 ABSTRACT

     This  paper  suggests  the major  needs and opportunities in high gradient
 magnetic  separation  (HGMS) of  particulate-gas systems according to three
 categories:   (1)   process and  engineering research oriented towards effective
 applications  of existing devices;  (2) fundamental research aimed at develop-
 ing  a  better  understanding of  the  mechanisms and principles of the separation;
 and  (3) scientific and engineering development of new separation methods and
 devices.
 INTRODUCTION

    High gradient magnetic separation (HGMS) is a new technique that provides
 a practical means for separating micron-size, feebly magnetic materials on a
 large scale and at much faster flow rates than are possible in ordinary fil-
 tration.  The technique is also applicable to separating nonmagnetic materials
 which can be made to associate with magnetic seeding materials.  HGMS has
 already been applied industrially for the wet cleaning of feebly magnetic con-
 taminants from kaolin clay with very low costs and outstanding technical per-
 formance {15, 16, 24, 28, 29, 34-36, 38}.  It is also proving applicable to
 solving many important problems related to energy production (e.g., coal
 destilfurization), environmental control (e.g., particulate collection and
 wastewater treatment), resource recovery (e.g., mineral  beneficiation), and
 other minerals and chemical processing problems.  The significance of the
 HGMS, and its latest engineering and commercial developments have been ade-
 quately discussed elsewhere {16, 28, 29, 35, 36, 38}.

    Because of the necessary costs for drying and dewatering the products in
most wet applications, it is generally believed that the most economical
application of the HGMS is for the dry separation of many particulate-gas
systems, especially those particulates of less than 2 to 3 microns in diameter.
Existing experimental data have already demonstrated the technical feasibility

-------
                                                                               171

of utilizing the HGMS for the collection of the basic oxygen furnace (EOF)
dust {10}, and for the removal of pyritic sulfur and ash from dry pulverized
coal {32}.  These reported results and other unpublished data known to the
authors have suggested that the HGMS appears to hold much promise as an
effective alternative to particulate-gas separation, and further research
and development work is clearly needed.

    The purpose of this paper is to briefly suggest the major needs and
opportunities in the HGMS of particulate-gas systems.  For convenience, the
discussion will be divided into three categories:  (1) process and engineering
research oriented towards applications of existing devices; (2) fundamental
research aimed at developing a better understanding of the mechanisms and
principles of the separation; and (3) scientific and engineering development
of  new separation methods and devices.


APPLICATIONS-ORIENTED PROCESS AND ENGINEERING RESEARCH

    Applications-oriented process and engineering research is needed for demon-
strating  that the HGMS  is an  effective and economical method for separating
particulate-gas systems.  This type of research may  include the development of
magnetic  process for particulate-gas  separation which has been previously
considered to be impossible or uneconomical; and the adaption of the magnetic
approach  to  improve the performance of existing  particulate-gas separation
devices.  Obviously, there are numerous areas or opportunities for applying
the existing HGMS devices to  particulate-gas separation.  Three practical  exam-
ples of such applications related to  energy production, environmental  control
and resource recovery can be  cited here to  illustrate the broad range  of
research  areas or opportunities available.

    Among the many areas in which the HGMS may  have potential  applications to
particulate-gas  separation, the magnetic  removal of pyritic  sulfur and ash
from dry  pulverized coal  is believed  to be  particularly  important and  promising
{19, 21,  24-27,  32, 37, 38}.   Substantial reduction in  both  pyritic  sulfur and
ash contents  in  the HGMS of several  Illinois Basin  coals  has  already been  re-
ported  in a  recent study by Murray {32}.  An interesting  observation from  this
study  is  that  the combined use of several finely divided  filamentary matrix
materials such as steel wool, steel  filings, etc.,  can  significantly enhance
the magnetic removal of both  pyritic  sulfur and ash.  This  observation confirms
the previously recognized importance of  the matrix  characteristics  in  providing
strong  high-gradient magnetic fields  {16, 29,  34,  35, 38}.   Clearly, research
and development  work  in optimum  matrix design  and  composition will  continue
to  contribute  significantly  to advancing  the state  of the HGMS technology,
especially  in  applying  to particulate-gas separation.   Another  important area
which  has not received  much attention is  the optimum use  of the  latest fluidi-
zation  technology in  conjunction with the high  gradient magnetic  desulfurization
of  dry  pulverized coal.  Identifying the  problems  associated  with  fluidizing
pulverized  coal,  determining  the  conditions for achieving good fluidization,
and characterizing  the  performance of the separation are  only some  of the  work
needed.

-------
 172
     Because of the national  interest in  increasing  the utilization of coal,
 more financial support for process  and engineering  research related to the
 magnetic desulfurization of dry pulverized  coal  is  being  provided by several
 governmental and funding agencies.   However,  the results  from a recent survey
 of the supported research projects  on the subject conducted by the authors
 have suggested that the current level of research support does not seem to
 reflect the proper appreciation of  the significance of the HGMS, and the po-
 tential and economics of the magnetic cleaning of dry pulverized coal.  The
 latter has been adequately discussed in  the literature {21,29, 32, 37, 38}.
 Here,  it is sufficient to mention that the  new HGMS technology has provided
 a possible commercial  method for economically separating  the sulfur-bearing
 and ash-forming minerals from dry pulverized coal on a large scale.  It has
 been estimated that a total  of 100  million  short  tons of  U. S. coals per
 year,  low enough in organic  sulfur,  can  be  magnetically cleaned as an environ-
 mentally acceptable,  low-sulfur low-ash  fuel.  This amounts to over 17% of
 the total  U.  S.  production per year {25-27}.  Further, recent analyses have
 clearly indicated  that the costs of high gradient magnetic desulfurization,
 especially of dry  pulverized coal,  appear to be attractive when compared to
 those  of other pyrite removal  processes  currently under active developments
 {23, 37}.   All  of  these aspects will  suggest that increased research support
 for the work on the magnetic desulfurization of dry pulverized coal is most
 likely to produce  the  greatest dividends.
 of ann?vi™e?LWS™<;b¥ DreV and. Goodl'ng {10} has il^strated the potential
 of applying  the  HGMS  to  particulate-gas separation in the important area of
 fhp'S!!?!    *£°ntr?,1' J" Part1cular> the technical feasibility of utilizing
 the HGMS  for the collection of the basic oxygen furnace (BOF) dust has been
 demonstrated.  An  important observation from this work is the dramatic improve-
 ment of particulate collection efficiency resulted from applying a relatively
 low-intensity magnetic field.  Total mass collection efficiencies of 99 9 per-
 SlnrXi  "S wre/™nd with * 3.08 kilogauss field, which compared quite
 favorably with the typical efficiencies of 50 percent or less observed without
 applying  the magnetic field.  However, it was also found from this study that
 the existing theory of HGMS failed to interpret quantitatively the effect of
 some separation  variables such as particle size, field intensity, etc.  Des-
 S™t Jhe Prellminary. nature of this investigation, it is apparent that the
 HGMS holds much  promise of application to the important problem of fine parti-
 rSl2Ha?12S1!n-COntr!!!  e|Pfc1a"y 1n the iron, steel,  ferroalloy,  and other
 related industries.  This follows because a majority of the particulate matters
 in  these industries are sufficiently magnetic to be separated by applying a
 high gradient magnetic field.   The primary need at this  point is to  identify
 clearly several  potential areas of application in these industries by a close
 interdTsciplinary cooperation between the particulate control  practitioners  and
 the magnetic separation technologists.   Further, it is anticipated that many
 technical  problems will  arise in applying the existing HGMS  devices  to some of
 the  selected areas.  This necessitates  some proper process and engineering
modifications of the current HGMS technology.   For instance,  effective means
must be developed to allow the use of the present HGMS devices in  separating
 the high-temperature particulate-gas  mixtures such as  these  off gases  from
 steelmaking furnaces;  and to prevent  the  chemically corrosive  components  such
 as chlorine existing in some parti cu late-gas  mixtures  from corroding  the

-------
                                                                               173
separator matrix.   The major objective of such process  and  engineering  work
should be, of course, on demonstrating that the desired particulate-gas
separation can be carried out effectively by the HGMS at a  low enough cost
to be economically attractive when compared to those of the competing fine
particulate control technologies (1, 11, 12, 31 40}.

    Applications-oriented research has also been initiated  in this laboratory
and elsewhere on utilizing the HGMS to recover the resuable components  from
the fine airborne dusts emitted from industrial operations.  This type  of re-
search attempts to extend the potential applicability of the HGMS for separa-
ting particulate-gas systems in environmental control to an increasingly im-
portant area of resource recovery.  As an illustration, it  is known that fly
ash from coal-fired power plants contains a large fraction  of reusable, fine
maonetic particulates such as magnetites.  Existing analysis has already
suggested that they can be economically recovered by the HGMS, thus reducing
the load on the more expensive electrostatic precipitators  {29}.  Preliminary
work in this laboratory and unpublished prior work in the coal mining industry
have also indicated that the coal ash derived magnetite performs better in
the heavy-media coal cleaning process than does mined magnetite  {26}.  In
view of the large amount of fly ash generated from coal-fired power plants in
the U.S., and the increasing cost of the commercial-grade mined magnetite being
used in coal preparation, further research toward the development of an econo-
mical ash-derived-magnetite coal cleaning process is clearly justified.   It
is also apparent that there are many other areas or opportunities related to
resource recovery,  in which the use of HGMS for separating particulate-gas
systems plays a major role; and they are too diverse to be surveyed here.

    Regardless of the area of application, process and engineering research
oriented towards the effective and economical utilization of the existing HGMS
devices for particulate-gas separation should normally include  the following
steps:   (1)  examination of the evidence of significant magnetic removal  of
particulates from  their carrier gas;  (2)  study of  the quantitative effects of
operating conditions such as field  intensity, residence time, etc. on  the grade
and recovery of the separation;  (3)   study of  the separator capacity and  scale-
up; (4) conceptual  process design and  cost estimation; and  (5)  comparison of
costs and performance with those  of the  competing technologies.   Here, the
recent experience  in carrying out most of these steps  in the development  of
HGMS  processes for  coal desulfurization  and  kaolin  beneficiation reported
in the literature  {21-28, 33, 34, 37-39} is  believed to be  helpful to  those
involved  in  the process and engineering  research on  the HGMS of particulate-
gas systems.

FUNDAMENTAL  RESEARCH

    The  key  to successful utilization of HGMS  devices  for  particulate-gas
separation  is the  clear  understanding of the mechanisms and  principles invol-
ved in relation to specific applications.   Recent experience gained  in the
pilot-scale  and commercial applications  of  HGMS to  coal desulfurization  and
kaolin beneficiation {21, 22, 27, 33,  37-39}  has demonstrated  the importance
of utilizing  quantitative separation  models  in conjunction with the  separation
testing.   In  particular,  the  successful  experimental  verification of two

-------
 174


 simplified mathematical models due to  Bean  {5}, and Oder and Price {39} in
 such  reported applications  has allowed one  to quantitatively predict the trade-
 off of  some operating parameters such  as field intensity, residence time, etc
 so as to optimize the magnetic desulfurization of coal and beneficiation of
 kaolin  {22, 39}.  None of these two simplified models, however, could satis-
 factorily predict the phenomena of mechanical entrapment and matrix loading
 (concentration breakthrough) observed experimentally.

    Similar research related to the above fundamental aspects of HGMS applied
 to particulate-gas separation is, unfortunately, being pursued much less vigor-
 ously.  This may be exemplified by the considerable uncertainties in the theo-
 retical understanding of the quantitative effects of separation variables in
 the recent study of high gradient magnetic particulate collection by Drehmel
 and Gooding {10}.  Clearly, in order to permit a more scientific approach to
 future applications of HGMS to particulate-gas systems, much work is needed
 to develop the necessary conceptual understanding and the proper modeling
 framework.   From a detailed review of the recent development and current status
of the quantitative modeling of HGMS described elsewhere {28},  it may be sug-
gested that such a modeling framework should at least take into account of
the phenomena of mechanical entrapment and matrix loading, as well  as the
characteristics of the particulates to be separated.   A proper approach ac-
counting for such factors also necessitates the conventional  matrix-type
HGMS device to be modeled as a magnetic adsorption column as  suggested by
Kaiser, et al. {17}.   An example of a typical  mathematical  model  based on this
approach can be found from the work in this laboratory {28,  33},  in which'the
successful  application of the modeling results to quantitatively predict the
experimental  data from coal  desulfurization is also demonstrated.   It is be-
lieved that the same  modeling framework can be properly extended  for predicting
the performance of HGMS of particulate-gas  systems.

    The success of the preceding attempts in developing quantitative models
to satisfactorily predict the overall  performance of the separation such as
grade, recovery, etc., can be greatly enhanced by simultaneously conducting
experimental  studies  on some synthetic and well-defined particulate-gas
systems.  Such experiments may be carried out, for example,  by using air-
entrained monodispersed particles with uniformly spaced wires as  collecting
fibers in the separator matrix.   Here, a  number of similar experimental  studies
of the HGMS of particles in water slurries  reported in the literature {6-9,
14, 18} may serve as  good references.   The  experimental  results thus obtained
should provide some helpful  insight into  the mechanisms of particle capture
and buildup,  etc., in particulate-gas  systems.

 EQUIPMENT DESIGN AND DEVELOPMENT

    Research in this  category includes the  improved design and  scale-up of the
existing HGMS devices, as well  as the development of devices  utilizing new
separation  concepts.   The importance of this type of research can  be readily
recognized  by noting  the significant effort currently being devoted to develo-

-------
                                                                            175


ping new separation methods  and devices,  including  those  suitable  for  dry
separation of part1culate-gas systems  {3,  4,  13,  15,  16,  36,  42},   Further,
the commercial  breakthrough  in kaolin  beneficiation resulted  from  introdu-
cing a new level  of HGMS unit in 1969  has  also made it clear  that  expanded
research activity in equipment design  and  development will  offer the grea-
test opportunities in advancing the technology of HGMS of particulate-gas
systems.

    A number of research needs and opportunities  in equipment design and deve-
lopment can be briefly suggested by examing the status of several  recent
technological innovations in HGMS, and their potential applicability  in par-
ticulate-gas separation.  First, it has been generally agreed that improved
or new design, and scale-up of the original production prototype HGMS  unit
have  been made during the past few years; and stainless  steel wool is widely
accepted as a common magnetic matrix.   What has been largely neglected here
is the systematic research aimed at developing alternative matrix  collectors,
especially for use in particulate-gas separation.  Next,  for certain  separa-
tion problems in which the feed stream contains a large fraction of magnetic
particles, technical and economical constraints often necessitate the use  of
continuous HGMS devices to increase the separation throughput. For example,
both moving-matrix Carousel  separator {36} and open-gradient stream splitter
{3, 4} are now being developed for such separations. Conceptually speaking,
the stream splitter is especially attractive because it is a truly continuous
device in which the magnetic particles, rather than being collected on a ma-
trix, are deflected to one side of the splitter by the magnetic field; wher-
eas the Carousel separator contains moving parts, and sometimes does not op-
erate in a truly continuous fashion. Because of the lack of the gradient-
producing collection matrix in the stream  splitter, it has not been proven
practical for separations of fine paramagnetic particles at present.   Thus,
it  has  been  suggested that stream splitting devices are best used  in conjunc-
tion with the high-intensity magnetic fields generated by superconducting
magnets {13}, which appears to hold much  promise of application to separating
particulate-gas systems. Here, the open space  existing in the superconducting
stream  splitting device allows for the use of  the  latest fluidization techno-
logy to enhance the particulate-gas separation. Clearly, further work is need-
ed  to quantitively characterize such an approach.  Alternatively,  there is
also a  need  to develop  a high-intensity,  high-gradient stream splitting device,
without using the  superconducting magnet.  The  recent  work by Allen and  lannice-
11i  {2} illustrates one promising approach to  this development, and this
new design  has yet  to be tested in actual  separations. Finally, perhaps the
most  important need  in  equipment  design and development  is an increase  in
research  support,  as significant  advances can  be made in particulate-gas se-
paration  with the  development  of  the new  generation  of HGMS  technology.


ACKNOWLEDGEMENT

     The authors wish to gratefully acknowledge the financial  support  for their
work  on magnetic  separation  provided  by the  National  Science Foundation, the
Gulf  Oil  Foundation, the Board of Trustees and the Engineering  Experiment
Station of  Auburn  University,  and the  U.  S.  Energy Research  and Development
Administration-  Oak Ridge National Laboratory.

-------
  176

  REFERENCES

  {1}  Abbott, J. H.5 and Drehmel, D. C., "Control of Fine Participate Emissions,"
      Chem. Eng. Progr., Vol. 72, No. 10, p. 47  (1976).

  (2}  Allen, J. W., and lannicelli, J., "Magnetic Separator," U.S. Patent 3,984,
      309, filed Sept. 27  (1974).

  {3}  Aubrey, W. M., Jr.,  Carpiniski, J., Cahn,  D., and Rauch, C., "Magnetic
      Separation Method and Apparatus", U.S. Patent 3,608,718, filed Dec. 20
      (1968).

  {4}  Aubrey, W. M., Jr.,  and Funk, R. M., "The  Quadrouple Magnetic Separator,"
      Paper presented at the 32nd Annual Mining  Symposium, Duluth, Minneasota,
      January (1971).

  {5}  Bean, C. P., "Theory of Magnetic Filtration", Bull.  Amer. Phys. Soc.,
      Vol. 16, p. 350 (1971).

  {6}  Cowen, C., Friedlaender, F. J., and Juluria, R., "High Gradient Magnetic
      Field Capture on a Single Wire", IEEE Transactions on Magnetics, Vol.
      MAG-11, No. 5, p. 1600 (1975).

  {7}  Cowen, C., Friedlaender, F. J., and Juluria, R., "Single Wire Model of
      High Gradient Magnetic Separation Processes I,"  IEEE Transactions on Mag-
      netics, Vol.  MAG-12, No. 5, p. 466 (1976).

  {8}  Cowen, C., Friedlaender, F. J., and Juluria, R., "Single Wire Model of High
      Gradient Magnetic Separation Processes II ", IEEE Transactions on Magne-
      tics, Vol. MAG-12, No. 6,  p. 898 (1976).

  {9}  Cowen, C., and Friedlaender, F. J., "Single Wire Model  of High Gradient
      Magnetic Separation  Processes III", in press, IEEE Transactions on Magne-
      tics, Vol. MAG-13, No. 5,  Sept. (1977).

{10}  Drehmel, D.  C., and  Gooding, C. H., "High Gradient Magnetic  Particulate
      Collection,"  Paper No. 53d, AIChE National  Meeting,  Atlantic City, Sept.


{11}  Englund, H.  M., and  Beery,  W.  T.,  "Control  Technology:   Particulates," APCA
      Preprint Series,  Air Pollution Control  Association,  Pittsburgh, Pennsyl-
      vania (1973).

{12}  Englund, H.  M., and  Beery,  W.  T.,  Editors,  "Fine Particulate Control  Tech-
      nology", APCA Preprint Series, Air Pollution Control  Association,  Pitts-
      burgh,  Pennsylvania,  February (1975).

{13}  Eyssa,  Y.  M., "A  Feasibility Study of  Superconducting Magnetic  Separators
      for Weakly-Magnetic  Ores",  Ph.D.  Thesis,  Nuclear Engineering Department,
      University of Wisconsin, Madison,  Wisconsin, September  (1973).

-------
                                                                               177


(14)  Himmelblau,  D. M.,  "Observation and Modeling of Paramagnetic Particles
      Trapping  in  a Magnetic  Field", M.S. Thesis, Massachusetts  Institute of
      Technology,  Cambridge,  Massachusetts, June  (1973).

{15}  lannicelli,  J.,  "Assessment  of High Extraction Magnetic  Filtration",
      Special  Report  to the National Science  Foundation, available as Document
      No.  Pb240-880/5  from the National  Technical  Information  Service,  Spring-
      field, Virginia  (1974).

{16}  lannicelli,  J.,   "New Developments in Magnetic Separation", IEEE  Transac-
      tions on Magnetics, Vol. MAG-12,  No.  5, p.  436  (1976).

{17}  Kaiser,  R.,  Colton, C.  K., Miskolczy, G.,  and Mir, L.,  "Magnetically
      Induced Separation of Stable Emulsions", AIChE  Symp.  Ser., Vol. 68, No.
      124, "Water",  p. 115 (1971).

{18}  Kelland, D.  R.,  and Maxwell, E.,  "Matrix Loading  in  High Gradient Magne-
      tic Separators", Paper presented  at International  Magnetics Conference,
      London, England, April (1975).

{19}  Kester, W. M.,  "The Effect of High-Intensity Magnetic Cleaning on Pul-
      verized High Sulfur, Bituminous  Coals", Master  Thesis, School  of  Mines,
      West Virginia (1966).

{20}  Lawson, W. F.,  Jr., "The Dynamics of Paramagnetic Particles Near  a Mag-
      netized Wire",  M.S. Thesis, West Virginia University, Morgantown, West
      Virginia  (1976).

{21}  Lin, C. J., Liu, Y. A., Vives, D. L., Oak, M.  J., and Crow, G. E., "Sul-
      fur and Ash Removal from Coals by High-Intensity, High-Gradient Magnetic
      Separation", paper  presented at ACS National Meeting, Fuel Chemistry Di-
      vision, San Francisco,  September  (1976).

{22}  Lin,  C. J., Liu, Y. A., Vives, D. L.,  Oak, M, J,, Crow, G. E., and Huffman,
      E.  L.,  "Pilot-Scale Studies of Sulfur  and Ash Removal from Coals by High
      Gradient  Magnetic  Separation", IEEE Transactions on Magnetics, Vol. MAG-
      12,  No. 5,  p. 513  (1976).

 (23}  Lin,  C. J., and  Liu,  Y. A., "Desulfurization of Coals by  High-Intensity,
      High-Gradient Magnetic Separation:  Conceptual Process  Design and Cost
      Estimation", ACS National Meeting, Fuel Chemistry Division, New Orleans,
      Preprints,  Vol.  22, No. 2,  p. 113, March (1977),

 {24}  Liu,  Y. A., Symposium Chairman and Proceedings Editor,  "Proceedings of
      Magnetic  Desulfurization  of Coal  Symposium:  A Symposium  on the Theory
      and Applications of Magnetic Separation",  Auburn University, March 23-
      25, 1976; published by the  IEEE  Magnetics  Society as a  Special Issue  on
      Magnetic  Separation in the  IEEE  Transactions on Magnetics, Vol. MAG-12,
      No. 5,  September (1976).

-------
178

  (25)  Liu, Y. A., and Lin, C. J., "Assessment of Sulfur and Ash Removal  from
        Coals by Magnetic Separation", IEEE Transactions on Magnetics, Vol.  MAG-
        12, No. 5, p. 538 (1976).

  (26)  Liu, Y. A., and Lin, C. J., "Applications of Magnetic Separation Method
        to Coal Preparation", Invited Chapter,  in "Coal  Preparation",  4th  Edition,
        Leonard, J. W., and Mitchell, D.  R., Editors, to be published  by the
        AIME, New York, New York (1977).

  {27}  Liu, Y. A., and Lin, C. J., "Status and Problems in the  Development  of
        Pilot-Scale High Gradient Magnetic Separation Processes  Applied  to Coal
        Desulfurization", Invited paper,  to appear in Proceedings of Engineering
        Foundation Conference on Clean Combustion of Coal,  Rindge,  New Hampshire,
        July (1977).

  {28}  Liu, Y. A., Lin, C.  J., Vives, D.  L., Oak, M.  J., and Crow, G. E., "Theory
        and Applications of High Gradient Magnetic Separation:   A Review", Invi-
        ted paper,  International  Magnetics Conference, Los  Angeles, June (1977).

  {29}  Luborsky,  F.  E., "High-Field  High-Gradient Magnetic Separation:  A Re-
        view",  21st National  Conference on Magnetism and Magnetic Materials,
        Philadelphia, Pennsylvania, December (1975).

  {30}  Luborsky,  F.  E., and  Drummond,  B.  J., "High Gradient Magnetic  Separation:
        Theory  Versus Experiment",  IEEE  Transactions on Magnetics, Vol. MAG-11,
        No.  6,  1696 (1975).

  {31}  Marchello,  J. M., and  Kelly,  J. J.,  Editors,  "Gas Cleaning and Air Quality
        Control", Marcel  Dekker,  Inc.,  New York,  New York (1975).

  (32}  Murray,  H.  H.,  "Magnetic  Desulfurization  of Some Illinois Basin  Coals",
        ACS  National Meeting,  Fuel  Chemistry Division, New  Orleans, Preprints,
        Vol.  22, No. 2,  p. 106  (1977).

  {33}  Oak, M.  J., "Modeling and Experimental Studies of High Gradient Magnetic
        Separation  with  Applications  to Coal Desulfurization", M.S. Thesis, Auburn
        University, Auburn, Alabama   (1977).

  {34}   Oberteuffer, J.   A., and D. R.   Kelland, Editors,  "Proceedings of High  Gra-
        dient Magnetic Separation Symposium", Massachusetts  Institute of Techno-
        logy, Cambridge, Massachusetts, May  (1973).

  {35}   Oberteuffer, J.  A., "Magnetic Separation:  A Review of Principles,  Devices,
        and Applications", IEEE Transactions on Magnetics, Vol. MAG-10, No, 2,
        p. 223  (1974).

  {36}   Oberteuffer, J.  A., "Engineering Development of High Gradient Magnetic
        Separations", IEEE Transactions on Magnetics, Vol,  MAG-12, No.  5, p.  444

-------
                                                                               179
{37}  Oder,  R.  R.,  "HGMS  Desulfurization  of  Pulverized  Dry  Coal",  Technical
      Report,  Job  No.  90371,  Scientific Development,  Bechtel  Corporation,
      December (1975).

{38}  Oder,  R.  R.,  "High  Gradient Magnetic Separation:   Theory  and Applications",'
      IEEE Transactions on Magnetics,  Vol. MAG-12,  No.  5, p.  428 (1976).

[39}  Oder,  R.  R.,  and Price, C.  R.,  "HGMS:   Mathematical Modeling of Commercial
      Practice", 21st Annual  Conference on Magnetism and Magnetic  Materials,
      Philadelphia, Pennsylvania, December  (1976).

{40}  Oglesby, S.,  Jr., and Nichols,  G. B.,  "Electrostatic  Precipitators",  Chap-
      ter 4, in "Gas Cleaning and Air Quality Control", Marchello, J. M.,  and
      Kelly, J. J., Editors,  Marcel  Dekker,  Inc., New York, New York (1975).

{41}  Riley, P. W., and Watson, J. H.  P., "The Use of Paramagnetic Matrices for
      Magnetic Separation", Paper presented  at the Conference on Advances  in
      Magnetic Materials  and their applications, London, England,  September
      (1976).

{42}  Stekly,  Z. J. J., "A Superconducting  High Intensity Magnetic Separator",
      Paper presented at the 13th International Conference on Magnetics, Lon-
      don, England, April (1975).

{43}  Watson,  J. H. P., and Hocking, D.,  "The Beneficiation of Clay Using a
      Super-conducting Magnetic Separator",  IEEE Transactions on Magnetics,
      Vol. MAG-11, No. 5, p. 1588 (1975),

{44}  Watson, J. H. P., "Magnetic Separation at High Magnetic Fields", Pro-
      ceedings of 6th  International Cryogenic Engineering Conference, p. 223
      (1976).
 WRITTEN DISCUSSION

 David R. Kelland
 Francis Bitter, National Magnet Laboratory
 Massachusetts Institute of Technology
 Cambridge, Massachusetts  02139

      Most of the work done on applying High Gradient Magnetic Separation
 (HGMS) to particulate-gas systems apparently has been in the area of coal
 desulfurization.  For the purpose of this paper as stated in the title it
 would be worthwhile reviewing the work done so far.  Also, the level of
 success achieved should be mentioned and the conditions under which the ex-
 periments have been carried out should be made clear.  Further, a careful
 distinction must be made between particulate-gas systems, in this case dry
 coal cleaning and water slurry separations.

     CoaVwater slurries have been the subject of a considerable amount of

-------
 180


 research which has been going for several years.  It can be said that for
 a given coal in which the inorganic matter has been liberated by grinding or
 perhaps other means, most of inorganic sulfur minerals can be removed by
 HGMS along with some part of the ash-forming mineral matter.  The same cannot
 be said yet for dry separations.

     Several attempts have been made to improve the magnetic properties of the
 impurities in coal so that these might be removed by liberation and subsequent
 magnetic separation.  Siddiqui steam-treated Pakistani coal and noted that a
 small part of the coal could be picked up by a permanent magnet.   This
 fraction had a higher sulfur content than the non-magnetic fraction.   Yurovskii
 and Remesnikov used shorter treatment times on a Soviet coal and reported a
 sulfur reduction in a fraction beneficiated in an induced roll separator.
 Adamov added magnetite to coal ground to minus 3mm and achieved some ash
 reduction by magnetic separation.

     Kester,  Leonard and Wilson at W.  Virginia University found that from 54 to
     of the inorganic sulfur could be removed from untreated coal.   The size
 range was 48 x 200 mesh and the best results were obtained on Upper Freeport
 and Pittsburgh seam coals.   Erratic  results were obtained with a  48 x 0 size
 fraction.  The separations  were performed using a Frantz Isodynamic Separator.

     Ergun and Bean at the Bureau of  Mines did an extensive project  on pretreat-
 ing and beneficiating coals.   Separating untreated coals in an induced roll
 separator, they report reductions  in the pyritic sulfur  of from 7 to  60%
 depending  on the coal.   Three  passes  were made through  the separator.   A  -300
 micron sample of Pittsburgh coal was  pretreated with  a heated  gas in  a
 fluidized  bed and then subjected to  separation in an  Isodynamic Separator.
 Reduction  in pyritic sulfur increased from 27% before pretreatment  to  32  to
 45% afterwards.

     This work seems  to  indicate  that  some of  the  pyritic  sulfur can be removed
 from some  coals provided  that  it can  be  liberated  at  a large enough particle
 size.   All of  these  studies have used  gravity feed  to the  separator and have
 been concerned  at  least with dry separations  if not separations of  material
 from a  gas stream.   Power plants blow fine  coal  (often -200 mesh) into  their
 furnaces.  If  the  pyritic sulfur is liberated,  it might be interesting  to see
 whether magnetic  separation could be  used on  a relatively high  velocity gas
 stream  carrying  coal  to reduce the sulfur content.

     Studies at MIT last year on an Upper Freeport coal blown through a high
 gradient separator showed that this kind of separation is more  difficult  to
 perform than a  corresponding coal-water  slurry separation.  In  fact, some of
 the best wet  separation results we know of were obtained with this  coal before
 the particulate-gas  separations were attempted.  Total sulfur reductions
 reached  57% compared  to about 40% obtained by Trindade on Brazilian coal at
MIT in  1973 and up to 45% by Murray on Illinois and Indiana coals recently.

    It  is important  to distinguish between total sulfur reduction and inorganic
 sulfur reduction since sulfur is found in both forms in coal.  The organic
 sulfur  is generally not removed by physical means such as magnetic separation.
Murray's test result in dry separation of 40% inorganic sulfur reduction would
correspond to a 20%  total sulfur reduction if half the total sulfur occurred
as inorganic, say pyrite.

-------
                                                                                181
   Sulfur reduction by physical separation depends on the type of coal,
its history, grinding, liberation, etc., so it is very difficult to compare
any of these figures.  But the problem of dry separation of impurities from
coal is far from solved and represents a research need and opportunity.

   I cannot be optimistic enough to agree with the authors overstatement
that "HGMS has provided a possible commercial method, for economically
separating the sulfur-bearing and ash-forming minerals from dry pulverized
coal on a large scale."  There are technical problems remaining in dry
separation and it is not clear what kind or how much of pretreatment might
be necessary nor how the HGMS process might be applied here.

   This is a fertile field for research and development and the need for such
should not be obscured by statements which imply that the problem has been
solved and is waiting for commercial development.  This is one area still in
need of research and that research ought to be done by a number of people or
research groups.  These should include magneticians, fluid flow experts,
chemists, and other experts in surface phenomena, electro-statoc effects,
etc.
   There is a reference here to the "commercial applications of HGMS to coal
desulfurization."  We are not aware of any such commercial applications and
should comment that the authors' use of "pilot-scale" often refers to
laboratory experiments.

   The application of modeling to HGMS has been made by several people.
Reference should be made to work reported at Intermag meetings of the IEEE.
In the case of coal, Trindade at MIT used a model for water-coal separations
in 1973 and Clarkson has extended the fluid model to an extended range of
Reynolds numbers and has taken inertia and the boundary layer near matrix
fibers into account.

   Beyond the generalities offered here about research needs, there are
a number of specific topics that could and should be mentioned.

   Agglomeration of particles may be a major problem in particulate-gas
systems especially when a separation is to be performed and an effort has
been made to liberate one particle species from another. Obviously, in
collection systems this may be desirable but the agglomeration of material
needs to be understood in any case.

   The effect of fines on particle agglomeration needs to be known.  Methods
of grinding to reduce fines may be employed and moisture levels varied to
change the effect of the fines content in the particulate system.  Electro-
static charges on particle surfaces have an effect on agglomeration and are
certainly affected by moisture.  The carrier gas, its moisture content and
temperature are all considerations for separation studies.

   For matrix separations the flow characteristics of gas and particles past
matrix materials should be studied over a reasonable range of Reynolds
numbers.  A number of studies of this sort have been done for fluids based
on work such as Zebel's on aerosols but they have generally considered the

-------
182

 fluid incompressible.

     Loading of matrices in a high gradient separator will have to be studied.
 There are now available both batch-type and continuous separators, the latter
 offering a possible solution to short duty cycles in the former when they
 occur.  Since neither coal loss nor inefficient collection of impurities is
 desirable in coal cleaning a clean separation is the goal.  Besides agglomera-
 tion mentioned before, mechanial trapping, collection efficiency, and
 changes in the flow resistance, all functions of loading, affect the quality
 of separation.

     The authors point to the work of Drehmel and Gooding as an illustration
 of the application of HGMS to particulate-gas separation. This is a very
 interesting piece of work and should be continued as a vigorous program.   A
 point not made here though is that in the case of BOF dust all of the
 material wants to be trapped out of the gas stream.   I would call that
 collection rather than separation.  The difference between the two is signi-
 ficant.  Mechanical trapping aids collection and therefore enhances the
 magnetic effect.  In particle separations on the other hand mechanical
 trapping can be a serious problem whether in gas streams or liquids.

     It is stated by the authors that an "observation" is made in Murray's
 work that confirms the enhancement of high gradient  magnetic separation by
 the use of combined matrix materials.  In fact,  Murray reports only the use
 of Frantz screens in his experiments.  Many people have patented matrix
 designs using carpet tacks,  steel wool, ribbon screens,  grooved plates,  etc.
 In our experience, however,  steel wool or stainless  steel wool (not necessarily
 the magnetic 400 series) remains an excellent matrix material.   Expanded
 metal is nearly as effective and is used in larger pilot plant moving matrix
 HGMS devices.  Steel shot,  filings and grooved plates have the disadvantage
 of occupying part of the magnetized volume of the HGMS device and magnetized
 volume is what one pays for.

     Matrices consisting of combinations of material  types have been used  in
 batch and moving matrix HGMS  devices.  We found  this type of matrix to be
 very useful in treating iron ores in a program carried both at our laboratory
 and on location in Minnesota.   It could almost go without saying that matrix
 design constitutes a development used and that companies engaged in machine
 production are not unaware of this need.

 OPEN DISCUSSION

 Liu:   I think I should quickly mention that our  research has been totally
 funded by the RANN Division of NSF.   They have been  very helpful supporting
 magnetic separation,  as Dr.  Oder is obviously aware.   I agree with most of
 the comments made about  magnetic separation.   Much of the data was not
 produced in our paper for this workshop because  approval has not been
 granted by the Electric Power Research Institute.

-------
                                                                               183
HIGH GRADIENT MAGNETIC SEPARATION:  REVIEW OF SINGLE WIRE MODELS
     F. J. Friedlaender

     School of Electrical Engineering, Purdue University
     West Lafayette, Indiana  4790?
INTRODUCTION

     High Gradient Magnetic Separation (HGMS) has become a subject of consi-
derable interest since its introduction into commercial beneficiation pro-
cesses and the hope that it may be a means of removing certain pollutants
encountered in industrial processes and energy generation.  Most systems
employing HGMS which are in use so far have been wet systems but a discussion
of such systems may be useful in suggesting possibilities, as well as diffi-
culties, with the use of air or other gases instead of water.  Here our main
concern will be with the collection process on single fibers which can be
considered the basic "building block" of the more complicated systems in
actual use.

      In HGMS processes a collection of ferromagnetic fibers is commonly placed
in a magnetic field of sufficient strength to saturate the fibers.  Hence,
each properly oriented fiber becomes essentially a line dipole which produces
rapidly changing fields with position in its vicinity.  The resultant field
gradients produce a magnetic force on magnetized particles in the vicinity of
the fiber.  Under appropriate conditions, particles will be swept toward the sur-
face of the fiber and wil 1 be held there as long as the field is present.  The major
competing forces on the particles in such systems are usually the magnetic forces
which sweep the particles toward the fibers and the forces due to the fluid which
carr ies the particles.  In a properly designed system, t he magnet i zed particles
will be retained on the fibers which act essentially as a filter for these particles.

     Here we shall review the various factors that are of importance in the
collection process and discuss the mechanism of particle capture on single
fibers.   In most practical HGMS filters, presently built, water carries the
particles which are to be separated out through a steel wool filter.  The
ability of the magnetized fibers of the steel wool to hold the magnetized
particles will determine the effectiveness of the filter.  Single fiber
experiments to study the collection process are usually carried out with
steel or nickel wires acting as collecting fibers.  The reader will be
referred to the appropriate literature for derivations of various quanti-
tative models.  The emphasis here will be on the presentation of detailed

-------
184
  experimental results of particle buildup on single collecting fibers (nickel
  wire) for several geometrical arrangements of the fiber relative to the flow
  and the applied magnetic field.

  FACTORS INFLUENCING PARTICLE CAPTURE

       It is convenient to consider separately each of the elements entering
  into filter operation.  Such an approach provides considerable simplification
  though sometimes with some sacrifice of rigor.  To begin with there are the
  particles which are to be removed from the slurry.  It is obvious that  these
  particles must be capable of being magnetized.  If the particles are ferro-
  or ferr Imagnetic they will either possess a net magnetic moment or can  be
  magnetized by a magnetic field.  Usually the moment will be large, and, once
  induced by a field, may be permanent.  Such particles  may be conveniently
  removed by more conventional methods and HGMS may pose certain problems.  For
  instance,  attempts to remove the collected particles by flushing out the filter
  after turning off the field may not be fully successful  since the particles
  may, in effect, become permanent magnets.  Paramagnetic particles, on the
  other hand, have no permanent magnetic moment and a field is required to in-
  duce such a moment which is proportional  to the product of the applied  field
  and the susceptibility of the particle.   It is obvious that weakly paramag-
  netic particles require larger fields.  Diamagnetic substances have not been
  removed by HGMS since the very weak magnetic moment induced in them opposes
  the applied field so that weak, repulsive rather than  attractive forces
  result  in  the usual collection region.

       The  magnitude of the force, considered for paramagnetic particles  only
  in the  following, is proportional  to the product of the magnitude of the   .
  induced magnetic moment,  the gradient field and the volume of the particle. '
  For larger particles the relationship becomes more complicated,  partly  because
  the field  gradient can no longer be considered uniform over the entire  parti-
  cle volume.3  But major interest usually centers on small  particle sizes since
  it was  the (successful)  intent of HGMS to obtain sufficiently large gradient
  fields  to  make separation of micron and  submicron size particles possible.
  At those  small  sizes,  the major competing force is the fluid drag which tries
  to carry  the particles along.   Obviously, below some critical  size the  fluid
  force will  always be larger than the magnetic force and  no collection will
  take place.  There are no particular difficulties in calculating the magnetic
  forces  although different approaches are possible and  are still  the subject
  of inquiry.  Gravitational  forces and particle inertial  effects may play an
  increasingly important role in the case  of larger particles.  At the other
  end of  the scale, very small  particles could be subject  to thermal  motion as
  well as other effects.  Electrical  forces play a role  in the capture process
  by influencing particle to particle, as  well  as fiber  to particle,  inter-
  actions.   Finally, magnetic interactions  between particles can become a
  factor  for particles with large moments.

       The  fiber serves  both as  the collection center and  as a means of produ-
  cing the  high gradient field.   It has to have appropriate mechanical  and
  chemical  properties (I.e.,  itshouldnot be corroded by the water usually  used

-------
                                                                               185
as the fluid that carries the particles).   Secondly,  it should  be ferromag-
netic with a large magnetic moment so that high gradients can be achieved.
The shape of the fiber cross-section is the other factor that determines the
magnitude of the gradient.  Hence, small  diameter wires or steel fibers with
relatively sharp edges are used to obtain  large gradients.  Steel with appro-
priate magnetic and corrosion-resistant properties is probably the best fiber
material for many HGMS applications, but nickel can be used and is sometimes
more readily available in a variety of sizes, as needed for experimental
purposes.

     The particle carrier, for instance the water used to make the slurry in
conventional HGMS systems in use today, plays a vital role in the process
since it carries the particle past the collecting fibers.  A determination
of the flow fields and how they affect the collection process is one of the
major problems facing investigators.

     Generally, the fluid flow problem has been considered under various,
simplifying assumptions.  At very small Reynolds' numbers, a viscous flow,
also called creeping flow^S, has been assumed by several investigators and
has been solved by a method due to Lamb.  This solution has been claimed to
give a better approximation of actual flow conditions close to the cylindrical
fiber than other often used solutions.  At somewhat larger Reynolds' numbers
the assumption of potential flow  is made,  but  is probably reasonable only
some distance from the fiber.  Boundary layer effects near the fiber surface
have been postulated 6 to  take   into account the hydrodynamic forces on
particles near the fiber surface.  When considering the particle capture
process, the calculation of forces on particles building up in layers at the
fiber surface has been based on various assumptions6'7'8'^.  In some cases,
predictions of particle buildup seem to follow observed patterns, mothers
they are off by several orders of magnitude.  The actual flow conditions such
as a stagnant region and the formation of eddies and an unsteady flow region,
all downstream behind the cylindrical fiber  seem to make calculations more
difficult, especially at  larger Reynolds' numbers which occur at higher flow
velocities.  Experimental results will be presented to point out such diffi-
culties.  One of the references quoted above deals with the capture of parti-
cles By a dielectric fiber1* due to electric  rather than magnetic forces.  When
the particles are electrically neutral but possess an electric dipole moment
induced by an applied electric field, then the capture mechanism becomes
analogous to that of magnetic dipoles captured on a ferromagnetic fiber.
Since such analogies may  be useful--capture  by electric fields  is receiving
a great deal of attention by workers studying  the behavior of aerosols--
a further discussion of this analogy  is in order.

     Generally,  the  induced electric dipole  moment can be assumed to depend
linearly on the applied field.  The same  assumption can be made  for magnetic
dipoles over a  limited  range of applied magnetic fields.  A paramagnetic parti
cle will have a dipole moment proportional to  applied  field for  most practi-
cally obtainable values of  field.   However,  magnetic  fibers will saturate
with  sufficiently  large fields whereas dielectric fibers will only  show
saturation  phenomena  if  they are  made of  a ferroelectric  substance.

-------
 186

 Electric fields are necessarily limited in magnitude to below electric
 breakdown gradients, but there is no such limitation on magnetic fields.
 (However, the region in which corona discharges take place may be used in
 electric separation processes.)   Hence, magnetic gradients are limited only
 by geometrical  considerations and by the magnetic properties  of the collecting
 fibers.

 EXPERIMENTAL RESULTS

      Collection of  paramagnetic  particles on  a  nickel wire has been studied
 by photographic means and  through the use of  a  hot wire anemometer10»'1»12.
 Here photographic collection  data will  be presented  for the three idealized
 cases that  are  usually discussed  in  the literature.

      Case A has the collecting fiber,flow, and the field all mutually perpen-
 dicular.

      Case B has field and  flow parallel  to each other and  perpendicular to
 the fiber axis.

      Finally, the third  arrangment,  case C, has the  flow along the  axis of
 the fiber with  the  field again perpendicular, as  it  must be if the  fiber-
 field  combination is to  produce a  large  magnetic  field  gradient.  All  three
 cases  are shown in  Figure  1.

     The  details of the  experimental  method are given in references  9-11.
 All  experiments described  here were  carried out with a  nickel   wire as  the
 collecting  fiber C»TrMs = 6900 Gauss  for  nickel) and paramagnetic  particles
 of  manganese pyrophosphate being collected from the  slurry.  The manganese
 pyrophosphate,  J^PzOy- (3HzO)  is  reported  to  have a magnetic susceptibility
 of  186 x  10~6 emu/oested/gram.  Particle  sizes  ranged from about  1 vim  to
 100 vim with most of  the particles  in  the  1-10 ym  range, but an appreciable
 fraction  of the total particle volume  is made up of  large particles.

     Photographs for case A and case  B have been obtained by taking pictures
of  the collecting wire "end on."  Case C shows  the wire with the attractive
 regions on the sides and one repulsive region facing the camera with the flow parallel
 to the wire.  In  all pictures the flow is from top to the bottom except in Figure 10.

     Figure 2 shows  the buildup of particles on a 25 pm diameter nickel
collecting wire, for case A geometry.  Parameters are 4.5 cm/s slurry
velocity and a  12.5  kOe applied field.  The concentration of Mn2P207'(3H20)
 is  0.0264 g/1.  The  time shown below each picture is measured   from the instant
at which  the slurry  first reached the wire (A flow of deionized water was
carried past the collecting wire until a diverter valve changed the flow to
 introduce the slurry.  Flow velocities were controlled by means of an orifice
downstream far  from  the collection region).

     Figure 3 shows the buildup again for case A using a 25 vim diameter wire
and 0.0264 g/1 concentration of Mn2P207'(3H20),  with  a flow velocity of

-------
                                                                        187
                          1
                                              case B
                                               case  C
Figure 1.  The three cases of flow and magnetic field  orientation
           relative to the collection wire.

-------
188


                                    OQ urn   I2"*u sec
     Figure 2.  Buildup for case A--.0264  g/1  Mn2P207*(3HoO^ field
                12.5 kOe, flow velocity  4.5  cm/s,  25 ym diameter
                nickel collecting wire.

-------
                                                                         189

                           100 >r
Figure 3.  Case A—0.26^ g/1 Mr^PzOy (3^0),  Field 7kOe,
           flow velocity 10 cm/s, 25 ym diameter collecting
           wire.

-------
 190

 10 cm/s and a field  of 7 kOe.   Figures  4,  5,  6  show the  buildup  on  a  125  urn
 diameter nickel  wire for case  B for  increasing  velocities  3.7  cm/s  in Figure
 4, 17.8 cm/s in  Figure 5 and 35.0 cm/s  in  Figure  6,  all  for  a  12.5  kOe
 applied field.   In  Figures  7 and 8 a  17-8  cm/s  flow velocity was used, with
 a^3.0 kOe field  in  Figure 7 and a 1.0 kOe  field in  Figure  8.   It was  more
 difficult to obtain  photographs which were adequate for  reproduction  for  case
 B with 25 ym diameter nickel wire.   Figure 9  shows  case  B  with 8 cm/s flow
 velocity and a  12.5  kOe field,  and slurry  concentration  as before.  Finally,
 in Figure 10 collection on  a 250 ym diameter  wire for case C is  shown.

      Theoretical  predictions of buildup for cases A and  B  were given  in
 references  6 and  11  and are reproduced  here in  Figures 11  and  12 to show
 the  good agreement with experiment for  the lower  values  of velocity.  As  can
 be seen by  examining some of the data taken at  higher velocities'?, less
 reliable data is  obtained at these velocities.  Downstream collection "behind"
 the  collecting fiber,  for case  B, appears  to  indicate the  appearance  of a
 stagnant region  that can be expected  to occur at  the lower velocities.  At
 the  highest  velocity (35 cm/s)  collection  becomes unreliable.  Further dis-
 cussion on  this  point  will  be found  in  reference  12.  Case C has  been discussed
 in detail  in  references 13  and  14.

 DISCUSSION AND CONCLUSION

      In  summary  it can  be said  that experimental  results and theoretical
 predictions give  substantial agreement  in  regions in which the assumption of
 potential flow holds,  possibly  modified to allow  for the existence of a
 boundary layer near  the surface of the collecting fiber.   Lamb's  solution
 (creeping flow) may  give  better  predictions in  some cases, but further
 studies  are needed to  settle this claim.  Calculations have usually been made-
 for  the  separation of a  single  component in a slurry, often assumed to consist
of uniform-sized  particles  (spherical),  with no mention of agglomeration.

      Experiments  have been carried out,  essentially duplicating these assump-
 tions,  except that particle size or shape has not often been controlled and,
again, agglomeration  is not mentioned but may play a role.   In  regions in
which the flow may be other than "smooth," i.e., unsteady flow with eddies
"behind" the wire'5, or generally higher velocity (and,  thus,  larger Reynolds'
numbers) flow, collection data   indicate that problems arise which still await
a  solution.  The difficulties with unsteady or turbulant  flow,  both from a
theoretical and from an experimental  point of view may not, however, be a
matter of major concern since regions in which such flow occurs do not seem
to be advantageous in HGMS.   Mixed systems, containing particles with different
magnetic susceptibilities with   the more magnetic particles to be separated
from the slurry need to be  investigated  both theoretically and  experimentally
 in single fiber systems.  Questions of particle to fiber  and particle to parti-
cle adhesion and how they differ in liquid and gaseous carrier  systems need to
be answered.  Considerable differences between the two systems  (wet and dry)
can be expected and have apparently been observed in the  operation of entire
systems^.   Instabilities may occur in the particle mot ion'7 and experimental
observation of particle trajectories  is  needed to test the theoretical

-------
                                                                          191
                                    t = 1066 sec

                                   '
Figure k.  Case B—0.264 g/1 Mr^PzOy(3H20), flow  velocity
           2.7 cm/s, field  12.5 kOe,  125 m diameter  collecting
           wi re.

-------
192
                                         333 >
                                              *
     Figure 5.  Case B—.0264 g/1 Mn2P20y (3H20), field 12.5 kOe,
                flow velocity 17-8 cm/s, 125 ym diameter collecting
                wire.

-------
                                                                         193

                   717 sec
Figure 6.  Case B— .026** g/1 Mn2P2°7* (3H20), field  12-5 k°e>
           flow velocity 35.0 cm/s, 125 ym  diameter collecting
           wi re.

-------
194
           •
       1U6 s«e
                                         1039 sec
       1868 sec
100 an
2632 sec
     Figure 7-  Case  B—.026^ g/1 Mr^O;'(3H20),  field 3 kOe,
                flow  velocity 17.8 cm/s,  125 ym diameter collecting
                wi re.

-------
                                       195
•
Figure 8.  Case B—.0264 g/1  I
           flow velocity 17-8 cm/s,
           wire.
   ' (3H20), field  1  kOe,
   125  nm  diameter  collecting

-------
196
                                        , ' -
                        838 sec
    Figure 9.  Case B—.0264 g/1 Mn2P207' (3H-0), field  12.5  kOe,
               flow velocity 8 cm/s, 25 ym diameter  collecting
               wi re.

-------
                                                                         197
Figure 10.   Case C — .0664 g/1  Mn2P207' (3H20), field  12.5  kOe,
            flow velocity 10 cm/s,  250  ym diameter  collecting
            wire.

-------
198
      THEORETICAL
FROM
PHOTOGRAPH
  Figure 11.  The theoretical buildup for case A compared to buildup
           shown in photographs.

-------
                                                               199
        THEORETICAL
  FROM PHOTOGRAPH
                                   LOWER SECTION IS SHOWN

                                    SCALE OF UPPER SECTION
   THEORETICAL
FROM PHOTOGRAPH
Figure 12.  The theoretical buildup from Luborsky  compared to
          buildup as shown on photographs for case  B.

-------
    200

predictions.  Since many systems in use today have collecting fibers arranged
in other than the idealized geometries of cases A, B, or C some thought has to
be given to the effect of more complicated geometries obtained essentially by
a linear combination of two or more of the  ideal  models.

ACKNOWLEDGMENTS

     This work was supported in part by a grant from the National  Science
Foundation.  The photographs of particle buildup on  a single wire were taken
from a thesis by Carl Cowen.'8

REFERENCES

[1]    Oberteuffer, J. A.,  "Magnetic Separation:  A Review of Principle, Devices,
      and Applications," IEEE Trans. Magn.,  Vol.  MAG-10, p.  223,  June 1974.

[2]    Watson,  J. H.  0., "Magnetic Filtration," J.  Appl.  Phys., Vol. 44, No. 9,
      p.  4209, Sept.  1973.

[3]    Aharoni, A., "Traction Force  on Paramagnetic Particles in Magnetic
      Separators," IEEE Trans.  Magn.,  Vol. MAG-12, p. 234, May 1976.

[4]    Zebel, G., "Deposition of Aerosol  Flowing Past  a Cylindrical  Fiber in a
      Uniform Electric Field," Journal  of Colloid  Science,"  Vol.  20,  p. 522,
      1965.

[5]    Cummings,  D. L.,  Prieve,  D. C.,  and Powers,  G.  J., "The Motion  of Small
      Paramagnetic Particles in a High Gradient Magnetic Separation," IEEE
      Trans. Magn., Vol. MAG-12,  pp. 471-473,  Sept.  1976.

[6]    Luborsky,  F. E.,  and  Drummond, B.  J.,  "Build-up of Particles  on Fibers
      in  a  High-Field-High  Gradient  Separator," IEEE  Trans.  Magn.,  Vol. MAG-
      12,  No.  5, p. 463, Sept.  1976.

[71    Luborsky,  F. E.,  and  Drummond, B.  J.,  "High  Gradient Magnetic  Separa-
      tion:  Theory vs.  Experiment," IEEE Trans. Magn.,  Vol.  MAG-11,  No.  6,
      p.  1697,  Nov. 1975.

[8]    Stekley,  Z.  J.  J., and Minervini,  J. V.,  "Shape Effect  of the Matrix on
      the  Capture  Cross  Section of  Particles  in High  Gradient Magnetic  Separa-
      tion," IEEE  Trans. Magn., Vol. MAG-12, No. 5, p. 474,  Sept.  1976.

[9l    Cowen, C.  C., Friedlaender, F. J.,  and Jaluria,  Rajiv,  "Single  Wire Model
      of  High  Gradient  Magnetic Separation Processes  II,"  IEEE Trans.  Magn.,
      Vol.  MAG-12, No.  5, p.  466, Sept.  1976.

[10]   Cowen, C.  C., Friedlaender, F. J.,  and Jaluria,  Rajiv,  "High  Gradient
      Magnetic  Field  Particle Capture on  a Single  Wire," IEEE Trans.  Magn.,
      Vol.  MAG-11, No.  5, p.  1600,  Sept.  1975-

-------
                                                                               201
[11]  Cowen, C. C., Friedlaender, F. J., and Jaluria, Rajiv, "Single Wire Model
      of High Gradient Magnetic Separation Processes  II,"  IEEE Trans. Magn.,
      Vol. MAG-12, No. 6, p. 898, Nov. 1976.

[12]  Cowen, C. C., and Friedlaender, F. J., "Single Wire Model of High Gradient
      Magnetic Separation Processes III," IEEE Trans. Magn., Vol. MAG-13, No. 5,
      Sept. 1977-

[13]  Birss, R. R., Gerber, R., and Parker, M. R., "Theory and Design of Axially
      Ordered Filters for High Intensity Magnetic Separation," IEEE Trans.
      Magn., Vol. MAG-12, No. 6, p. 892, Nov. 1976.

[1ft]  Uchiyama, S., Rondo, S., Takayasu, M., and Eguchi, I., "Performance of
      Parallel Stream Type Magnetic Filter for HGMS," IEEE Trans., Magn.,
      Vol. MAG-12, No. 5, pp. 895-897, Nov. 1976.

[15]  Batchelor, G. K., "An  Introduction to Fluid Dynamics," Cambridge Univer-
      sity Press, Cambridge, England, 1970.

[16]  Luborsky, F. E., Private Communication.

[17]  Lawson, W. F., Jr., Simons, W. H., and Treat, R. P., "The Dynamics of a
      Particle Attracted by a Magnetized Wire," J. Appl. Phys., to be
      pub)I shed.

[18]  Cowen, C. C., "High Gradient Magnetic Separation:   Single Wire Studies,"
      Ph.D. Thesis, Purdue University, December 1976.
WRITTEN DISCUSSION

W. F. Lawson, Jr.
Morgantown Energy Research Center
U.S. ERDA
P.O. Box 880
Morgantown, W.V. 26505

     As Professor Friedlaender points out, most HGMS experiments have been
concerned with the separation of particles using a liquid carrier media.  In
viscous fluids, particle inertia and the Earth's gravitational force can
generally be disregarded when describing the particle dynamics.  This is the
low Stokes number regime where Watson's theory1 is applicable.  The agreement
betwee.n  gross filter efficiencies calculated using Watson's theory and re-
sults of actual filter experiments is generally unsatisfactory. Luborsky
states in his review2 that from the mass of material captured in randomly
packed magnetic filters it was clear that Watson's theory would have to be
extended to account for particulate accumulation on individual fibers.  Thus,

-------
202
  the need arose to examine the particulate buildup on single wires.  Subsequently,
  theories were developed based on trajectory models and concurrently the photo-
  graphy of particulate buildup on single wires in an aqueous medium was begun
  at Professor Friedlaender's lab.  The agreement between these theories and
  experimental observations has not been entirely satisfactory and,  as noted in
  the article under discussion, only the slower flows substantiated  the
  theoretical predictions.

       The problem with the theories may be the result of the complexity of the
  physical situation.  Professor Friedlaender correctly indicates many of the
  difficulties, in particular those associated with the description  of the fluid
  flow.  Different flow fields around the cylindrical fiber have been used in
  single fiber theories and have yielded quite different quantitative results.
  In addition it appears that there are conditions of practical interest for
  which the relative velocity between the particle and the fluid medium become
  large.  For such non-Stokesian particles one should of course use  the empirical
  formula for the fluid drag in the equation of particle motion.

       It is perhaps worth emphasizing again that the above theories are based
  on a particle trajectory model.   Although Himmelblau3 has observed some HGMS
  trajectories, neither these observations nor other experiments provide satis-
  factory particle dynamical data  to check this basic model.   As Kolm^ has
  advocated and Professor Friedlaender suggests,  careful observations of the
  particle dynamics are needed to  check the trajectory models and to provide a
  better understanding of the basic processes involved in particle capture.

       What of particulate-gas separations using HGMS?  Professor Friedlaender
  has noted some of the potentially different aspects of dry  separation as com-
  pared to wet HGMS.  Drehmel and Gooding5 have demonstrated the efficiency of
  dry magnetic separation in cleaning simulated iron furnace  emissions.   However,
  in many instances there was not  even qualitative agreement  with fiber efficiency
  predictions based on Watson's theory.   This is not surprising since Stokes
  numbers for their experiment ranged from about 2.5 to 250	consistently above
  the threshold marking the onset  of inertial effects.   A general treatment  of  the
  effects of particle inertia and  the Earth's gravitational force when vectorially
  combined with the magnetic  and  fluid  forces is given in Reference 6,  but
  these results are still experimentally untested.   Calculations of  single fiber
  capture distances and particle trajectories and more details on the qualitative
  differences between the inertial and noninertial cases of particle dynamics
  can be found there.   The two factors which make inertial effects manifest  in
  dry HGMS are the reduced viscosity of  the medium and the typically higher
  flow velocities used in gas filtration.   The reduced fluid  mass density can
  cause the Earth's gravitational  pull to  be more significant than with denser
  liquids.

       Moreover,  it is not yet clear that  a satisfactory filter efficiency can
  be derived from single fiber trajectory  calculations.   In fact,  some pre-
  liminary work on multiple fiber  filter theory indicates  that the effect of
  neighboring fibers  in large arrays can be quite significant.   Shown in Figure
  1  are two pairs of  capture efficiency  curves for a long  fiber presented as  a

-------
  1.0
    .8
C5
a.
«t
C_J
                                                                          INTERIOR CELL OF
                                                                          MULT I FIBER ARRAY
                                                                  	 SINGLE FIBER
     0  L_
       .0
FIGURE  1
           J_
           ,5
1.0
1 .5
 LOG TT-
2.0
                                         2.5
                                         3.0
                    Capture efffency of an  interior cell  of  a multifiber filter based on 97% and 33%
                    void  filter fractions using flow and  magnetic field descriptions of a squar^
                    multifiber array (solid  lines) and an isolated single fiber (dashed lines).
                                         3.5
                                                                                                 ro
                                                                                                 O

-------
   204

 function of Watson's parameter V^V^.  The capture efficiency is defined here
 as the fraction of particles starting from the boundary of an interior cell
 which is captured by the enclosed fiber.  The geometry is Professor
 Friedlaender's case B.  The pair of curves labelled 99% are (a) the capture
 efficiency of single interior cell of an infinite square array of parallel
 fibers (the solid line) and (b) the capture efficiency of a single isolated
 fiber (i.e., a single fiber in a non-interacting array), calculated by the
 method of Zebel8 (the dashed line).  The efficiencies and fiber spacing in
 the arrays are based upon a 99% void fraction in the filter.  The pair of
 curves labelled 97% are the analogous cases of (a) and (b) above for 97%
 void fraction.  In analyzing the information contained in these curves, one
 should be aware that the flow has been treated as potential flow, the viscous
 force is assumed to be Stokes drag, and that the square matrix geometry
 analyzed is probably not the most practical design for a filter.  However, the
magnitude of the difference between the isolated fiber and the multifiber
 efficiencies with even loosely-packed arrays certainly indicates a need for
 caution in applying single fiber results to predict raultifiber efficiencies.

     We agree with the lists of research needs in fundamental HGMS work
 compiled by Professor Friedlaender in these proceedings and by KolnP.  It is
our belief, however, that priorities must be established in the investigation
of high gradient magnetic capture mechanisms.  First and foremost, the
dynamics of the single fiber particle trajectory model in both inertial and
non-inertial regimes should be checked experimentally.  The trajectory model
has been the cornerstone of most recent buildup and filter efficiency
calculations and its validity has not been directly investigated.  Next,  in
view of new multifiber data presented here,  the nature of particle capture in
geometrically - arrayed multifiber filters should be studied.   Information
provided in these areas could enable designers to make more intelligent
decisions balancing engineering and economic considerations in future commer-
cializations of HGMS.

REFERENCES

 [1]  J.  H. P. Watson,  J.  Appl.  Phys., 44,  9,  p.  4209 (1973).

 [2]  F.  E. Luborsky, AIP Conf.  Proc., 29_,  p.  633 (1976).

 [3]  D.  A. Himmelblau,  M.S. Thesis (MIT,  Cambridge,  1974)  (unpublished).

 [4]  H.  H. Kolm,  IEEE Trans. Magn.,  MAG-12.  5,  p. 450 (1976).

 [5]  D.  C. Drehmel and C.  H. Gooding, "High  Gradient Magnetic  Particulate
     Collection,"  82nd AIChE Nat.  Meeting,  Atlantic City,  Aug.  29 - Sept.  1,
     1976.

 [6]  W.  F. Lawson,  Jr.,  W.  H.  Simons, and  R.  P.  Treat,  J.  Appl.  Phys.,  48
     p.  3213 (1977).

-------
                                                                               205
[7]   W.  H.  Simons  and  R.  P.  Treat,  private  communications.  Drs.  Simons  and
     Treat  provided  the multifiber  capture  efficiency  data  shown  in  Figure 1.
     A full account  of their work will  be published  at a  later  date.

[8]   G.  Zebel,  J.  Colloid Sci.,  20, p.  522  (1975).
OPEN DISCUSSION

Ariman:  If you go back through all the HGMS presentations,  I have a couple of
things to say.  First of all, the workshop presentations and discussions so
far have been very exciting.  In the morning sessions we did have very
interesting comments from specialists in different fields.  For example, in
the morning, acoustics was the main area of presentations.  We had stimulating
discussions from researchers in the acoustics,filtration and HGMS areas.  But
when we cover the HGMS area it was quite interesting to see that, all of the
comments during the discussion presentation were made by HGMS workers.

    I would like to make the following comments for the interest of the
participants who are experts in the HGMS.  It seems to me there is a strong
emphasis for the basic reference in this field to Zebel's work.  I think, it
is an excellent work.  Quite a few of you have been doing a lot of work using
his model.  It is a single fiber, single particle model.  Professor Friedlander's
statement that Zebel did not only consider dielectric particles and conducting
fibers, but he looked also at both conducting and nonconducting particles.
For one of the cases, he obtains a closed form solution for the particle
trajectory equation.  But for the other case he has a numerical solution for
the particle trajectory.

    He also looks at both viscous and non-viscous flow cases.  He has published
a total of four articles on  this subject.  Two were coauthored by one of his
associates, Walkenhorst.  They also did some experimental work.  But they  talk
of a very special type of fiber filter.  In other words, this fiber had a
special formation and a lot  of distance between the fibers.  That single
fiber model will hopefully be suitable for that particular  filter.  However,  I
really do not know how good  that single wire model would be for  the very
complicated filter which is  used in the HGMS process.

    On the other hand we have modified and extended the single fiber model to
include the interference effects between the  fibers.  In the model there are
an infinite number of rows and infinite number of fibers on each row.  We  found
out that when the distance between the fibers is somewhat small, there was
quite a bit of difference between Zebel's model and our model.   Furthermore,
a relatively recent  publication, Ionia and his co-authors did both analytical
and experimental work and their experimental  results  compare very well with our
analytical model findings.   Furthermore  they  concluded  that when porosity  of
the filter  increases, there  is a decrease  in 'the collection efficiency. Those
are  the results we obtained  .  We found  the same result   through our  analytical
model  prior  to  their publication.

-------
206
THE INFLUENCE OF ELECTROSTATIC FORCES AND OF THE PROBABILITY
OF ADHESION FOR PARTICLE COLLECTION IN FIBROUS FILTERS


    F. Loffler

    Institut fur Mechanische Verfahrenstechnik,
    Universitat Karlsruhe
INTRODUCTION

    Fibre filters have long been used industrially to sepa-
rate solid or liquid particles from gases . The number of pub-
lications on this topic is correspondingly large. The answer
to some important questions are nevertheless obscure to the
present day. This is particularly true for the technically
interesting particle size range above 0.5 - 1 ym and for
Reynolds' numbers larger than 0.1. Exceptionally striking
differences between theoretical calculations and experiments
have been observed in these regions.

    Generally speaking the distances between the fibres in
these filters are large compared with the particles. The
particles are therefore not deposited by a screening action
but must be moved to the fibre surface and retained there. It
has already been shown | 1 |  that the deposition efficiency   is
the product of the collision efficiency n times the probabi-
lity of adhesion h.
       = n • h                                             (1)
    One important reason for the above mentioned discrepancy
is the fact that the significance of the probability of ad-
hesion has been rather neglected, i.e. it was assumed that
all particles striking a fibre would also stick there. However,
this is frequently not so. Many theoretical papers on trans-
port processes, on the other hand, do not take sufficient
account of physical realities and simplify the problem unduly.

    In order to elucidate some of the open questions on trans-
port and adhesion processes we have carried out a three-part

-------
                                                          207
research programme: theoretical and experimental studies of
single fibres, model filters and industrial filters. Some of
the results will be discussed below.

THEORETICAL MODEL

    The deposition efficiency is calculated on the basis of
a model cylinder on which the incident flow is at right
angles.
           U0,M,P
          Fig.(l) - Model deposition on a  fibre
    The Brownian motion  can be  neglected  in  the  particle  size
 range D  > 0.5  ym under  consideration  here.  In this  case  the
 other p&ssible  transport mechanisms  include:  inertia,  gravi-
 ty  and electrostatic  forces.  Let  the collision efficiency in
 these cases  be  defined as  (cf.  Fig.  1):
                                                           (2)
 where  y   is  the  origin of the  coordinate  of  the limiting

-------
 208
 particle  trajectory in the  undisturbed flow in front of the
 cylinder,  i.e.  the  trajectory  of  particles  which still just
 touch  the  fibre.  All particles of size D which arrive in the
 band  2yQ will  strike  the  fibre.
     In order  to  calculate  the  collision  efficiency  it  is
necessary  to  find  the  limiting particle  trajectory  by  solving
the  equation  of  motion of  the  particles.  The  general formula-
tion is: the  sum of  all the  forces  shall equal  zero:
    Fd + F. + Fg + p^



where Fd = the  force of  resistance, F. =  the  force of  inertia,
Fg = force of gravity, Fel = the electrostatic  force.

    In the general case  it follows from equation  (3) that


    nTES = f (*, Re, W, Fr, Nel)                             (4)


          p D 2-UQ
where ijj = ^ P p      the inertia parameter  (Stokes'' number) ,



,_ _ VDf'
                 the Reynolds number, W = ' -  the density ratio,



             Froude's number and Nel an electric charge para-
meter.
    Nel is tne ratio of the electrostatic force at a certain
point   to the drag force. Various modifications of the elec-
trostatic effect are in principle possible, depending on
whether the particles, the fibres, or both of them are charged
and whether external fields are applied. Assuming that both the
particles and the fibres are charged, i.e. that Coulomb forces
act,

-------
                                                               209
    N . = N
                     Q ' * -                          (5)
where Q is the fibre charge per unit length, q = particle
charge, e  = influence constant =8.86 • 1CT12 Coul/V-m.

    The differential equation of the particle trajectory can
only be solved numerically. The required condition of contact
is that the particle surface must touch the surface of the
fibre.

    Most authors simplify the calculation according to equation
(3) by entering only a single transport effect and estimating
the overall effect by superimposing various partial ones. Par-
ticles are also frequently treated as mass points, and their
geometrical size is additively superimposed as what is some-
times called the interception effect. The following results
will show that both these simplifications lead to errors.

    The solution of the equation of motion involves a con-
siderable problem: the velocity field for the flow round the
fibre is required. There are no closed solutions for the
Navier-Stokes equations for the region of Reynolds' numbers
0.1 < Re < 100 which is relevant for many practical, industrial
cases. Suneja and Lee  |2| worked out a numerical solution for
discrete Re numbers  (1, 10, 50). All that is otherwise
available are approximate solutions, e.g. Lambs' solution for
the viscous flow round a single cylinder, or that of Kuwabara
for a fibre system who neglected the forces of inertia com-
pletely in the flow; this solution therefore holds strictly
only  for Re = 0.

    For Re > 100 the equations for potential flow are often
used  for single  cylinders. Any decision on how closely these
various approximations reproduce the true flow round single
fibres, even layers of fibres, must be arrived at by careful
experiments.

-------
 210
 SOME  NUMERICAL RESULTS FOR THEORETICAL SINGLE FIBRE COLLISION
 EFFICIENCIES
    A.   No  electrostatic forces
(Nel  = 0)
         We  calculated particle trajectories for Re > 50 with
 the potential  flow field plus an approximate equation for the
 boundary layer flow |3|. For Re < 1  we used Lambs flow field.
                    XT 2  t 6 810 2  4 6810° 2  4 6810  2  4 6810
                                      18uDf
         Fig.  (2) - Theoretical single  fibre  collision
                    efficiencies  nT  (Inertia  effect  only)
    Fig. 2 shows collision efficiencies nT calculated  for  an
inertia effect only-i.e.(without gravity) for a particle den-
sity p  = 0,86 g/cm  . In the high Reynolds number-range only
the curves for Re =  50 and 500 are plotted. The curves for
Re = 1, 10, 50 represent results of Suneja and Lee  J2|. We
see that their and our results for Re = 50 are in fairly good
agreement.

    For Re = 500 we  found good corrrsponda nee with  the results
of Langmuir and Blodgett in the range of fy > 2. Langmuir and
Blodgett used only the potential flow field and their  results

-------
                                                           211
are not plotted in the figure. For ijj < 2 our values are some
what smaller, which is due to the influence of the boundary
layer flow.

    It is possible to fit our numerical results  for 50 < Re
500 with good accuracy by the following equations:


    nT = n.,  (*» Re) + n2  (*, Re, p/pp)                   (6)
                      2
               f.,(Re)i|i  +  f2(Re)  i|i  +  f3(Re)


                                  1/2
    "  -
     f  (Re)  =  - O.O133  In  Re + O.9 31
     fJ(Re)  =  + 0.0353  In  Re - 0.36
     f^ (Re)  =  - 0.0537  In  Re + 0.398


     The  curves of Suneja  and Lee show that for \l> < 1  nT de-
 creases  with  decreasing Reynolds number, whilst for ip > 1  we
 observed the  reverse tendency. This  is because to keep ty
 constant we need a bigger particle  size D  with decreasing
 Re  i.e.  with  decreasing velocity U  . This^results in a higher
 interception  effect; therefore nT increases.

     The  same  effect we found also in our calculations in the
 range  Re < 1  using Lambs  flow field.

     Fig. 3 shows the results for the Reynolds-numbers 0.1, 0.2
 and 0.66. There  are two groups of curves. The dotted lines
 give the results which have been calculated by assuming the
 particles to  be  mass points  ("without interception effect )
 whilst the full  lines give the results when the geometrical
 size of  the particles was allowed for.

-------
212
                      1.2

                      1.1

                      1.0

                      as

                      0.8

                      0.7

                    |  0.6

                    -0.5

                      0.4

                      0.3

                      0.2

                      0.1
Inertia effect
               Pp'2.164g/cm3
                without
                interception
               Pp' 0.86 und
               2.164 g/cm3
                      0.1  0,2  0,4 OJ50.81
          4 6 8 10  20
                                                  100
            Fig. (3) - Theoretical  single fibre collision
                       efficiencies ru,  (Inertia effect,
                                     ,'T
                       without gravity,Lamb's flow field)
    From this example it will be  seen that the negleot of  the
particle dimensions leads to considerable errors, particu-
larly  in the range i|> > 1 . The correct contact condition must
be used  from the very start.

    The  influence of gravity was  calculated for a vertical
flow downwards,  i.e. for a flow in  the same direction as
gravity  is  acting.

    Fig.  4  shows curves plotted for different Reynolds-numbers
and for  two particle densities p  =  2.164  g/cm3 and 0.86 g/cm3.
The dotted  lines have been calcinated as  before and the particle
dimension is neglected.

    We see  that  for the correct contact condition we have  high
values for  the collision efficiencies r\  .  The increasing  rimc
with decreasing  Reynolds number or  decreasing particle density
is again  due to  the fact that for constant  ty we have bigger
particle  sizes D  for decreasing  Re or p  .

-------
                                                                 213
             Fig.  (4) - Theoretical single fibre collision
                        efficiencies TITS
                        (Inertia effect , with gravity ,
                         Lamb's flow field)
    In our opinion this effect of gravity has not been con-
sidered sufficiently till now.
    B,
With electrostatic forces
          We must expect considerable effects in the case of
coulomb forces. For this case we calculated the influence of
electrostatic forces by inserting them into the equation of
motion and solving the complete equation numerically. Most
important are the results for the range Re < 1 which will be
discussed in the following.

    Fig. 5 shows some calculated particle trajectories round a
cylindrical fibre for different particle sizes in the diameter
range 1 - 10 ym.

-------
214
                                 Lamb's flow field
            Fig.  (5) - Theoretical particle trajectories
                       (Lamb's flow field. Re = 0.63)
    The upper two examples show trajectories which just pass
the fibre. All particles of the same size  starting at a point
1/1OOO D  closer to the axis of symmetry will touch the fibre
surface. This touching trajectory is called the limiting tra-
jectory.

    The top picture shows the case without electrostatic forces,
Here collision is only due to inertia. Small particles pass
round the fibre, i.e. they have small collision efficiencies.
The 10 urn-particles are only little influenced by curved flow
around the fibre.

-------
                                                          215
    In the centre we see the corresponding particle trajecto-
ries for charged particles and fibres. These have a product of
the charges Q  • q = 10~28 Coul2/cm. This value is approximately
equal to our experimental values.

    The enormous influence of the electrostatic forces can be
seen clearly especially for the smaller particles. Their colli-
sion efficiencies are increased. Obviously there exists a mini-
mum collision efficiency at a particle size of about 5 urn. The
path of the 1O ym-particle is nearly unaffected by the electro-
static forces.

    The focussing of the particles on the axis of symmetry be-
hind the fibre is a most interesting effect. This nonhomogenous
distribution of particles behind the fibre has some influence
of course on the deposition in the following layers of fibres.
This is in contrast with any theory which assumes homogenous
particle distribution.

    In the lower picture we see the touching and thus limiting
trajectories. The larger particles which exhibit little influ-
ence by the electrostatic forces strike the fiber on its front
side. The smaller ones first pass the fibre and are then attrac-
ted to its far side.

    From the limiting particle trajectories we calculate the
theoretical collision efficiencies, shown in fig. 6.

    In fig. 6 the single fibre collision efficiency HTES nas
been plotted, as a complete solution of equation (3), against
the inertia parameter ty for a particle density pp = 2.164 g/cm
and for Re = 0.2. The parameter of the curves is the charge pa-
rameter NQg. The above, correct contact condition was assumed
for the curves marked "with interception effect", i.e. the-.geo-
metrical size of the particles was allowed for, while they were
regarded as mass points in the curves "without  interception
effect".

    The curve for Ngg = O holds when there is no charge. As
NQg rises, the collision efficiency increases quickly for
small ijj values. However, the curves for Ngg pass through a
minimum at about ij; = 1 and then approach the curve for Ngq = 0
asymptotically. This means that above ty = 1 the influence of

-------
216
the Coulomb force becomes  less  than that of  the  force of
inertia. The total collision efficiency in the region of  the
minimum is less than that  produced by electrostatic  attraction
alone, i.e. inertia and the force of gravity counteract the
electrostatic force. If the force of inertia and the force
of gravity were to be neglected, the electrostatic attraction
by itself would give a collision efficiency  independent of ijj.

    This shape of the curves shows that partial  solution  for
the equations of motion obtained by neglecting individual
forces cannot be superimposed.  That means that all the effec-
tive forces must be entered in  the equation  of motion, which
should really be a matter  of course but has  generally not been
done in the past.

    In order to cover the  additional influence of electrostatic
forces as against the forces of inertia and  gravity, the
difference between HTES an^ tne collision efficiencies was
formed for the case HTS witn no charge.
    nE   nTES ~ nTS
(7)
    This difference can be approximated by
       =  1 .22(2 - In Re)  . N
     E    (	|jj_ , 1 .5 +  1   "  Qq

          2/Re
    This equation determines  for the  first time the  dependence
of the additional collision efficiency TIE on the parameter  fy,
Re and N. Past publications always  gave. TIE no more than HE =
const f(N_), which is  fundamentally incorrect.

-------
                                                                217
                             t.6
                Fig. (6) - Theoretical single fibre
                           collision efficiencies
                           (Inertia with gravity with
                           electrostatic forces. Lamb's
                           flow field)
DEPOSITION EFFICIENCY MEASUREMENTS

    Comprehensive experiments with model filters were carried
out as a check of the collision theories |5|.

    Fig. 7 shows a schematic diagram  of the apparatus.

    The model filters consisted of several layers of parallel
fibres. The fibre diameter was Df = 38 ym. The distances bet-
ween the fibres varied between 1T5 ym and 1600 ym in the direc-
tion of the flow and between 115 ym and 275 ym at right angles

-------
218
to this direction  so  that  the  percentages  of the fibre volume
ranged between 0.3 and  8 %.

    Monodisperse NaCl particles  and paraffin wax particles in
the range of about 1  to 10 ym  were  used.
                  Kr-85
                  p-ray source
                                                      filter
                   rotor
                                                  exhaust
             Fig.  (7) - Schematic  diagram of experimental
                        apparatus

    For the production of the NaCl-particles we  used a
spinning-top-Generator as shown  in Fig.  7. The geometric  stan-
dard deviation a  was for all size distribution  <  1.12. The
Kr 85-3 ray source served for reducing the electrostatic
charge of the particles in certain experimental  runs.

    Paraffin wax particles were  produced by  a Sinclair-La Mer-
Generator. These particles are almost spheres and  practically
without electrostatic charge; their size distributions have
geometric standard deviations 1.09 < o   < 1.12.

    Particle size distribution and concentration are measured
continously and directly in the  undisturbed  flow by  means of
a scattered light instrument. This instrument developed in our

-------
                                                            219
institute does not require the  extraction of a sample and is
also capable of measuring high  particle concentrations up to
106 Part./cm3.
    The velocity of the  incident   flow was varied between 25 cm/s
and 200 cm/s. This results in a range from Re = 0.63 to Re = 5.2,
1.4
1.3
12
1.1
1.0
09
0.8
07
06
05
0.4
03
0.2
0.1
                                              Symbol
                           0.0 01 02  05  1
                                               5  10
                                       18 M OF
                   Fig. (8)  - Single fibre deposition
                              efficiencies for NaCl
                              particles, particles and
                              fibres charged
    These  measurements determine the deposition efficiency
 i.e.  the product of the collision efficiency n and the  proba-
 bility  of  adhesion h. In Fig. 8 the experimental single, fibre
 deposition efficiencies 


-------
220
    The shape of the curves is very reminiscent of the theo-
retical curves in Fig. 6. The strong maximum in the region
ty < 1 for particle sizes Dp < 3 ym is very remarkable. The
deposition efficiencies measured in this region can only be
explained by the effect of Coulomb forces. This is also shown
by the theoretical curve for Q-q = const = 7.8 • 109 eQ2/cm and
for Re = 0.63 which has been entered for comparison as sequen-
ces of two dots and a dash.
                               9   2
    The charge product 7.8-10  eQ /cm was confirmed by
measuring the charge on particles and fibres; e.g. for 2.7 ym
particles we measured about 1100 e  charges, the fibre charge
per unit length was in the order of magnitude of 106 - 10'e^cm.

    The fact that electrostatic forces significantly affect the
curves, especially for particles smaller than 3 ym, was also
demonstrated by changing the charges on particles and fibres.
        Fig.  (9) - Single fibre deposition efficiencies
                   for NaCl particles.
                   (Particles and fibres discharged)

-------
                                                                 221
    In Fig. 9 some single fibre deposition efficiencies have
been plotted against the inertia parameter iji for three different
Reynolds-numbers. The corresponding velocities were in the
range 25 cm/s to 100 cm/s. Compared with the results in Fig. 8
we find a drastic change in deposition especially for values
of ty < 1, i.e. for fine particles < 3 ym. But presumably in
this range particles and fibres were not completely discharged;
in the range ijj < 0.2 the experimental values of deposition
efficiencies are somewhat higher than the theoretical collision
efficiencies. We suppose this to be an effect of remaining
charges.

    Particles of 5 ym and more were about equally deposited
whether they were charged or not. This observation again con-
firms the theoretical prediction of a decrease in the influence
of the electrostatic charge as the effect of inertia increases.

    In the region i|j > 1 it is striking that the deposition
efficiencies become smaller with increasing Re. This contra-
dicts theoretical expectations. We explain it by an increase
in the elastic rebound of the crystalline NaCl particles
striking the fibre which is equivalent to a drop in the pro-
bability of adhesion. This process will be dealt with in the
next section.

    A completely different behaviour from that of the NaCl
particles is shown by the softer paraffin wax particles.
These electrostatically neutral particles were deposited much
better than the NaCl particles at large inertia parameter ip.

    As we see in Fig. 10 the deposition efficiency decreased
steeply with dropping i|>. The Reynolds number had practically
no observable influence between Re = 0.63 to Re = 2.63. This,
incidentally, is also in agreement with the theoretical pre-
diction when the size of the particles is taken into account.
A decrease in the deposition efficiency was only measured
for Re = 5.26, corresponding to Uo = 2 m/s, and for smaller
Reynold's numbers at ip > 2-3, and here, too, the explanation
is a decrease in the probability of adhesion.

    With regard to the influence of the filter fibre volume
fraction packing density a it may be noted that no dependence
was found for a < 3 %. The fibres behaved like single fibres.

-------
222
A  slight rise in the deposition efficiency was only measured
at a  = 5.4  %.
                              o
                              I-
                              0
1  6810










o











PflRRFFIN
PRRTICLES






















('(
1C





00
[ O 0
lo.c





o
8
fY





c
<
(1
v\
-, it

•o ji


u
1
d
,
()°
"^ ^)
^
'' DA
,,/<
1 0
l'» /
/
>1!/

L
3«
R
n/
/ >
!
'y
y f





^
P
x


/
E




«5i
M
/
c >
c
/

=0.6




•=.&/REzS.6
/





5






































                                 o-'
io-'       10°       lo1
 INERTIfl PBRflMETER-PSI
                                                                  10'
Fig.(10) - Single  fibre  depo-
           sition  efficiencies
           for paraffin  wax
           particles.  Fibre Vo-
           lume  fraction
           a = 0.017
           Fig.(11) -
     Comparison of
     single  fibre de-
     position efficien-
     cies  and theories
     for paraffin wax
     particles
    In Fig. 11 the experimental values of single fibre depo-
sition efficiences for paraffin wax particles for all fibre
volume fractions and all  Reynolds-numbers have been plotted.
The full lines represent  theoretical calculations of Suneja
and Lee  |2| when they are  recalculated for Re = 0.6 and
Re = 5.6.

-------
                                                          223
    When a particle strikes a fibre surface it can only stick
if, firstly it does not rebound and, secondly, if it is not
subsequently detached.

     Earlier studies |8|  have shown that once particles are
deposited on a fibre they stick very tenaciously. What is
crucial therefore is the impact process when the particle
strikes the fibre which is an elastic-plastic impact invol-
ving elastic forces of resilience which may cause the particle
to rebound.

    The speed at which the start of rebounding may be expec-
ted can be estimated by means of an energy balance |3,9|.

    From this we obtain with the assumption that the deforma-
tion at the contact point is inealstic and that the main ad-
hesion mechanism is due to van der Waals forces:
    u° = ^  ^7^  =  Lifshitz-van der Waals energy
           h*  =  h/2 ir where h = Planck's constant
           z   =  minimum distance of surfaces on
            0     contact * 4-10~° cm
           H   =  micro-hardness of the softer material

           p   =  particle density


If we assume the deformation to be an elastic one we also ob-
tain a Uc a  (k-DD)~T porportionality. The coefficient of
restitution involves the material properties but it is sub-
stantially unknown  for microscopic particles. We are there-
fore currently working on a project for determining the im-
pact number.

-------
224
    From estimates  for k  and other material parameters we ob-
tained, for example,  for  Dp  = 10 ym,  U  = 2-15 cm/s and for
   = 5 ym, U  = 4 -  10 cm/s.
 D
     However, these estimates do  not  supply the distribution
 of the probability of adhesion but merely the start of the
 rebound. More accurate information on  the probability of ad-
 hesion must be obtained experimentally.  We therefore deve-
 loped a method for the direct determination of the probabi-
 lity of adhesion by the high-speed cinematograph |1,9  of
 particles striking a single fibre. We  regard this direct me-
 thod as more accurate and reliable than  the comparison of
 measured deposition efficiencies with  theoretical collision
 efficiencies as suggested recently by  Stenhouse et al. 1 10 1.
                              Btta -Source
                                          Elutnotor
                      Lead Sheat
                    / Xenon -lamp  Da
                            irk Field Condenser/ !
                               Housing
                                           Test Fiber, adjustable
                                                   big
          Fig.  (12)  - Scheme of apparatus  for  determining
                      probability of adhesion
     Fig.  12 shows a scheme of the apparatus  by  which we ob-
 served the particle trajectories.

     In the apparatus a free jet, loaded with particles, im-
 pinges vertically downwards upon a fibre. The observed space

-------
                                                                 225
is illuminated with a 250 W  xenon lamp which is pulsed to three
times the power during the periods of film exposure.

    In order to reduce the electrostatic  charge on the particle
impinging on the fibre,  from a  0  ray  source, ions are genera-
ted, which accumulate on the particles. An electric field is
applied at right angles  to the  direction  of flow and this
defelcts the charged particles  sufficiently to stop them from
colliding with the fibre.

    The particle trajectories for determining the adhesion pro-
bability are recorded by the arrangement  shown in Fig. 12 in
which the high-speed film camera  is at right angles to the
axis of the fibre  (max.  speed 30.000 frames/second). The par-
ticle velocity can be controlled  between  3 cm/s and several m/s,
and particles as small as 3  ym can be recorded.
                     IP
                                          Polyamidfibre 20 pm
                                           Quartz - particles
                                          • Glass - spheres
                                              70  80 cm/s 100
                                                        U0
         Fig.  (13)  - Probability of adhesion for quartz
                     particles and glass spheres on a  2O  ym
                     polyamide fibre

-------
 226
     Muhr's  theoretical  curve  for Re  =  0.66 was  calculated with
 Lamb's  flow field.  The  comparison with the readings  and  with
 the  Suneja  and Lee  curve  for  Re  = 0.6  suggests  that  at Re = 0.66
 Lamb's  flow field evidently provides unduly  small  collision
 efficiencies.

     Below ty =  0.2 the readings are higher than  would be  expec-
 ted  from the pure effect  of inertia. This can,  however,  be ex-
 plained with the Stenhouse theory |6|  through elecstrostatic
 forces  between uncharged  particles and charged  fibres.

     Nguyen  and Beeckmans  |7|  derived an empirical  equation from
 measurements,  but this  provides  much smaller deposition  effi-
 ciencies than  our measurements with  paraffin wax particles.
 These authors  also  found  a strong dependence on the  Reynolds
 number  and  the percentage of  the fibre volume.  Presumably the
 discrepancy arises  from the fact that  Nguyen et al.  based the
 fitting of  their equation mainly on  measurements with crystal-
 line particles. In  these  cases the material property results
 in differences in the probability of adhesion - rather as with
 our NaCl measurements.  We therefore  consider the application'
 of this empirical equation to other  practical cases  to be
 highly  problematical.

    In  the  current  state  of knowledge  we regard Suneja and
 Lee's theory for the calculation  of  collision efficiencies in
 the transition region of  Reynolds' numbers O.5  < Re  <50  as very
 suitable as  long as there are no  electrostatic  forces. In the
 case of Coulomb forces  their additional effect  can be estima-
 ted with equation (8) according  to Muhr.


PROBABILITY OF ADHESION

    It has already been pointed out in the introduction that
particles transported to  the surface of the fibre must also be
retained there. That this is by no means always the case is
particularly clearly shown by the experiments with NaCl par-
ticles but also by those with paraffin wax particles. The
literature contains numerous references to inadequate adhesion,
e.g.  the experiments of Gillespie and of Whitby are typical.

-------
                                                           227
100
h/"/.
60
40
20
n
\
\
^


— A
•

V.
\ .'
X.
•
dp = 5,1 pm
dp = 7pm

A A
— — — """" "•
•»
•

^H*
>1^
•

roruiii
Polyiai






n - Oil
rud- Fibre 20pm
                        0   10   20  30   tO  SO   60  70
                                                   U0 fan Is
                (14) - Probability of adhesion for droplets
                       of paraffin oil

    By way of example the probability of adhesion h has been
plotted in Fig. 13 against the speed of incident flow UQ for
various kinds and sizes of particles and types of fibres.

    Rebounding starts at about 5-15 cm/s. As the speed in-
creases, the probability of adhesion drops quickly. In accor-
dance with theoretical expectations the probability of adhesion
is less for D  = 10 ym than for D  = 5 ym, and the particles also
rebound more Itrongly on the hardir glass fibre than on the poly-
amide fibre. The values for the start of rebounding and the trend
with respect to particle size and fibre hardness correspond fair-
ly well with theoretical estimates.

    The lower adhesion probability with glass spheres is presum-
ably an effect of the contact geometry. Irregular quartz par-
ticles may have multi-point or area contacts while only single-
point contacts are possible with the very smooth glass spheres.

    The readings for the probability of adhesion for oil drop-
lets shown in Fig.  14 demonstrate that 7 pm droplets are more
prone to rebound than 5 ym droplets. It may be surprising  that

-------
 228
 at  30  cm/s  as many  as  35  to  50%  should  rebound. So  far we  ex-
 plain  it by the  fact that in this  velocity  region the droplets
 react  as if they were  elastic. At  higher velocities  the plastic
 deformation of the  droplets  and  therefore the  contact area are
 larger which results in a higher probability of adhesion.  Still,
 this behaviour will have  to  be examined in  more detail. Current-
 ly we  are also examining  the probability of adhesion of paraf-
 fin wax particles.  Here,  too, first results indicate rebounding
 which  had already been recognisable, at any rate at  higher velo-
 cities and  with  larger particles,  during the deposition experi-
 ments  on model filters.

     Our own studies with model  filters and real filters,  just
 like the work of other authors,  have shown  that the  problem of
 the probability  of  adhesion  is not only important for single
 fibres but  also  for layers of fibres. With  model filters |5|
 and NaCl particles  very similar  distributions  of the probabili-
 ty of adhesion were found as  for quartz particles on single
 fibres, even when the  percentage of fibres  by  volume was 8%.
 The reason  must  be  that in many  cases the acceleration zones
 for particles between  the  fibres are often  smaller than the
 mean fibre  distances and  that the particles, after a first
 rebound, will therefore strike subsequent fibres at approxi-
 mately the  same  velocity. Raczynski et  al |11|, for instance,
 have reported on studies  of  the  deposition  of quartz particles
 in glass sphere  filters. These authors, too, found probabili-
 ties of adhesion of less  than 1, even for 1  pm particles.

     So far  no final description of the probability of adhesion
 is possible  because the investigations have not yet been con-
 cluded. Nevertheless existing results already explain many of
 the phenomena observed in the laboratory and in practice, and
 they also yield useful suggestions for the development or
preparation of filters. Some questions are still left open but
the need to allow more thoroughly for the probability of ad-
hesion to solve practical problems is nevertheless evident.

-------
                                                                229
FINAL OBSERVATION

    The theoretical and experimental investigations presented
here assume simple models: a single fibre and a model filter
consisting of parallel fibres. In reality filter practice is
quite different, at least with respect to filter structure.
Nevertheless we selected this way not only because it corresponds
to the procedure in the classic filter theory but also because
it was possible to achieve very informative detail on the basic
phenomena with the influences separated.

    In a third step, of course, investigations of real filters
are necessary and these we have started already. The aim of this
new work is to find out the relationship between model ex-
periments and real filters. In doing so the material properties
of the particles are to be allowed for, especially with respect
to adhesion properties.

-------
230
REFERENCES

|1| Loffler, F., H. Umhauer: Eine optische Methode zur Bestim-
             mung der Teilchenabscheidung an Filterfasern.
             Staub - Reinhaltung der Luft 31, 51-55 (1971)

|2| Suneja, S.K., C.H. Lee: Aerosol filtration by fibrous fil-
             ters at intermediale Reynolds numbers.
             Atm. Environment 8, 10-81-1094 (1974)

|3| Loffler, F., W. Muhr: Die Abscheidung von Feststoffteilchen
             und Tropfen an Kreiszylindern infolge von Trag-
             heitskraften.
             Chemie-Ing.-Tech. £4,  510-514 (1972)

|4| Muhr, W.: Theoretische und experimentelle Untersuchungen
             der Partikelabscheidung in Faserfiltern durch Feld-
             und Tragheitskrafte.
             Dissertation Universitat Karlsruhe,  1976

|5| Loffler, F., W. Muhr: A study of the deposition of partic-
             les in the 1-10 microns range in model filters.
             Filtration - Separation V1_, 172-178  (1974)

|6| Stenhouse,  J.I.T.:  The influence of electrostatic  forces
             in fibrous filtration.
             Filtration - Separation 11, 25 - 26  (1975)

|7| Nguyen, C., J.M.  Beeckmans:  Single  fibre capture efficien-
             cies of  aerosol particles  in real and model fil-
             ters in  the intertial  interceptive domain.
             J. Aerosol Sci. 6,  205-212 (1975)

|8| Loffler, F.: Blow-off of particles  collected  on filter
             fibres.
             Filtration - Separation £, 688-696 (1972)

|9| Loffler, F.: Adhesion probability in fibre filters.
             Clean Air  fit, 75-78  (1974)

|10|Stenhouse,  J.I.T.,  D.C.  Freshwater: Particle  adhesion in
             fibrous  air filters.
             Trans.Instn.Chem.Engrs. 54, 95-99 (1976)

-------
                                                                             231
[11]  Raczynski, B., J.C. Guichard:  Penetration des aerosols dans les lits
      fixes de billes de verre. Froc. 1st World Filtration Congress, Paris,
      14-17, Mai 1974.
WRITTEN DISCUSSION

Hsu-Chi-Yeh, Ph.D.
Inhalation Toxicology Research Institute
Lovelace Biomedical and Environmental Research Institute, Inc.
P.O. Box 5890
Albuquerque, New Mexico  87115

    The effects of electrostatic forces and the coefficient of adhesion on
particle deposition in fibrous filters are among those areas which are still
not well understood in filtration theory.  Professor Lb'effler's paper will
surely contribute to further understanding of these two interesting areas of
filtration and will serve to stimulate future studies.

    Two parameters are important in the design of fibrous filters:  the
efficiency of collection and the pressure drop across the filter.  Practical
fibrous filters usually have high porosity, with the inter-fiber distance
large as  compared to the fiber diameter, and their fiber packing density is
normally less than 0.1.  Fibrous filters can be grouped into two classes :
very fine filters of high efficiency with fiber diameter on the order of 20
ym or less and moderate to low efficiency filters with larger diameter fibers
of up to about 200 ym.  High efficiency filters are normally used to remove
submicron particles and are operated with relatively low gas velocity of a
few centimeters per second.  Therefore, the Reynolds number (Re) is normally
less than 1.  For moderate to low efficiency filters which may be used as
pre-filters in ventilating systems, higher gas velocities of about 1-3 m/sec
may be used and the Re could go up to about 50.  For filters operated at low
flow velocity, i.e., Re < 1, those deposition mechanisms due to diffusion,
inertial impaction and interception are  understood, and various modern
filtration theories '  can be used to predict the collection efficiency with
good agreement with experimental data when inhomogerieity of a real filter is
taken into account from the measurement of the pressure drop across the filter.
Still not well understood in filtration by fibrous filters are deposition
mechanisms such as electrostatic attraction, gravity effect, acoustical
loading effect, the phenomenon of particle rebound at higher flow velocity,
etc.

-------
 232
     The first  step in solving the basic problem of filtration is the calculat-
 ion  of  the flow field associated with filters.   Because of the complicated
 structures of  a real filter,  a filter model has to be used.  Numerous,such
 modeljs  and their calculated flow fields, such as an isolated cylinder , a cell
     1  ' ,  a staggered-array model  and a fan model ,-have been reported.
        I  dl/tT.TO ^VlA 0^4 r*r»n^>nA	•*«•*»«««* .^J —. 1 ._ *..«. J 1_._ V_l_  	1 Ti •	 *\ O _1_
model'
 Figure  1  shows  the  staggered-array model used by Yeh  and Figures 2-3 show
 the  flow  field  calculated by solving full Navier-Stokes equations numerically
 for  two Reynolds  numbers, 0.4 and 30, with fiber packing density of 0.1.  The
 results of  these  flow field calculations indicate that both the Lamb's flow
 and  potential flow  describe poorly for a system of cylinders.  They also suggest
 that Kuwabara's flow field is a good approximation to the staggered-array
 model for Re less than 1, even though the Kuwabara's flow was derived from
 Re=0.   The  same conclusion has been obtained experimentally by Kirsh and Fuchs .
 It has  been shown that the cell model by Kuwabara and the fan model both
 describe  adequately the dependence of pressure drop upon fiber packing  density
 and  face  velocity for Re  < 1.   For Re > 1, there is still no analytical solution
 of flow field in  either a system of cylinders or an isolated cylinder.
                       Fig. (1)  Staggered-Array Model
    Many current theories of aerosol filtration by  fibrous  filters  have  been
based on or modified from the Kuwabara-Happel flow  field  for  the  cell  model.
Various theories which take into account two or more deposition mechanisms
simultaneously have been developed for calculating  the single fiber efficiency
of a filter.  These are diffusion with interception ,, inertia with intercept-
ion2, 3, 10 and diffusion plus inertia plus interception .   It would be interesting
to comparec. results of inertia with interception presented here with other
theories, but the theoretical treatment of inertia  with or without  interception

-------
                                                                                              233
2.5



  2
• Kuwabara's CeU Model,

A Happel's CeU Kodel,
Q Klrsh & Fuehs1 Fan llodel,
V Splelnan & Coren's I!odel,
O Lai* "8 Flow,
O Potential Flow,
                        	1	

                        ' 0.05999, 0.2076
                         0.1,205 , 0.6835
                        • 0.05999, 0.2076
                        • 0.05999, 0.2076
                        • 0.05999, 0.2076
                        • 0.05999
                         0.05999
                                                        T
                                                           RE = 0.4
 w rotcmiai rxow,         T- v*vsf7r
	 Theoretical Results for Staggered-Array Model
             Figure 2:   Comparison of Streamlines  for Various  Theories

                          (Re =0.4 and Fiber Packing  Density  a  = 0.1)
            KUWABARAS  FLOW,  ^=0.107
   0       I
                                                                         10      II
      Figure  3:   Streamlines for  Re  =  30 and Fiber  Packing Density  a =  0.1

-------
 234


 in his paper is not very clear.  Also, some parameters such as  interception
 parameters and fiber packing density are needed to explore other theories
 but are not listed, making comparison impossible.  In addition, there is
 some doubt about the validity of his Figure 2.   If it is for the inertial effect
 only without interception, then the single fiber efficiency curve should not
 go beyond the maximum of 1.   If the figure represents inertia with inter-
 ception, then the method for obtaining the curves which represent the results
 of Suneja and Lee   should be mentioned, for their original curves for Re = 1,
 10,  and 60 (not 50 as quoted in the paper) represent results of inertia with-
 out  interception.

     The application of electrostatic forces in  fibrous filter filtration is a
 novel way to enhance filter  efficiency.   As mentioned in Professor LBffler's
 paper,  there are several forms of electrostatic forces depending upon whether
 the  particles,  the fibers, or both of them are  charged or whether an external
 electric field  is applied.   In the paper,  Professor LOffler showed some in-
 teresting effects of coulomb force on filtration (i.e.,  both particles and fibers
 are  charged), such as nonhomogenous distribution of particles behind the fiber.
 On the  other hand,  it also raises some questions which need to be answered.
 In the  case  of  charged particles  with fibers either charged or uncharged,  the
 electrostatic force (or electric  charge  parameter)  will  change with time as
 charged particles continue to be  collected on the fiber.   What the effects of
 this  would be are as yet  unanswered.   According to  Fig.  6,  his theory predicted
 that,  for  Stokes  number < 1,  the  physical  size  of the particle is immaterial
 (i.e.,  independent  of the interception parameter) and this needs to be checked
 experimentally  using aerosols of  different densities.   In experiments where
 both  particles  and  fibers were discharged,  he showed  that for paraffin wax
 particles  (Fig.  10  in the paper),  the  Reynolds  number had practically no
 influence between Re =  0.63  to  Re  - 2.63,  while for NaCl  particles  Fig.  9),
 the Reynolds number  clearly  showed influence on deposition efficiency,  even
 for Stokes number <  1 where  rebound is not  a problem.  The  reason for this
 discrepancy  is  unknown.   Also it would be  interesting to  compare Figs.  9
 and 10 of the data with the  theoretical prediction  shown  in Fig.  3  and 4.

    The coefficient  of  adhesion (sometimes  called accommodation  coefficient)
 is not only  important in  high velocity filtration.it  is also  important  in
 sampling and/or sizing  devices such as impactors.  Professor  LOffler's  experi-
mental approach,  direct observation of particle rebound using a  high-speed
cinematography  technique, would also be useful  in investigating  rebound
problems in other devices.  The probability of adhesion in  filtration  is an
area where further studies are needed.

-------
                                                                                235
  REFERENCES
[ 11   Stenhouse, J. I. T., P. J. Lloyd and R.  E.  Buxton,  "The retention  of
      large particles (> 2 ym) in fibrous filters," Amer.  Ind.  Hyg.  Assoc.  J.
      37J7): 432-436, 1976.

[21   Stechkina, I. B., A. A. Kirsh and N. A.  Fuchs, "Investigations of  fibrous
      filters for aerosols.  Calculation of aerosol deposition in model  filters
      in the region of maximum particle breakthrough," Kolloidn.  Zh. 31_: 121-
      126,  1969.

[3]   Yeh,  H. C. and B. Y. H. Liu, "Aerosol filtration by fibrous filters - I.
      Theoretical." J. Aerosol Science 5: 191-204, 1974.

[4]   Lamb, H., Hydrodynamics. 6th ed., p. 77, Cambridge University Press,
      London, 1932.

[5]   Kuwabara, S., "The forces experienced by randomly distributed parallel
      circular  cylinders or  spheres in a viscous flow at small Reynolds  numbers,"
      J. Phys.  Soc. Jpn. 14:  527-532, 1959.

[6]   Happel, J.,  "Viscous flow relative to arrays of cylinders," A.I.Ch.E. J.
      5.: 174-177,  1959.

[7]   Yeh,  H. C.,  "A  fundamental  study of aerosol  filtration by fibrous filters,"
      Ph.D. Thesis, University  of Minnesota, Minneapolis, 1972.

[81   Kirsch, A. A. and N. A. Fuchs,  "The fluid flow  in a system of parallel
      cylinders perpendicular to  the  flow direction at small Reynolds numbers,"
      J. Phys.  Soc. Jpn. 22(5): 1251-1255, 1967.

[9]  Stechkina,  I. B. and N. A.  Fuchs,  "Studies on fibrous aerosol filters - I.
      Calculation  of  diffusional  deposition of aerosols in fibrous  filters,"
      Ann.  Occup.  Hyg. £:  59-64,  1966.

[10] Harrop, J. A.  and J.  I. T.  Stenhouse, "The theoretical prediction of  inertial
       impaction efficiencies in fibrous  filters,"  Chem. Eng. Sci. 24: 1475-1481,
       1969.

[11]  Suneja, S.  K.  and C.  H. Lee,  "Aerosol filtration by fibrous filters at inter-
      mediate Reynolds numbers  (< 100),"  Atmos. Environ. 8:  1081-1094,  1974.


 OPEN DISCUSSION

 Lbffler:  Thank you for the interesting comments,Dr.  Yeh.  You mentioned  that
 the Kuwabara flow field in the  experiment of  Fuchs, confirmed  the validity  of
 Kuwabara flow field.  We should not forget that Kuwarbara flow field assumed
 that the Reynolds number is very near to zero.  If you deal with  flow fields
 with Reynolds numbers between 4 and 30, I doubt  if this  is possible.  Because
 we suppose that these Kuwabara  flow fields are tested by Fuchs  in  a very low

-------
236

 Reynolds  number range  (about  .001)  and  is  only  tested by  a measure  of direct
 force.  The direct  force with a velocity and pressure distribution  in the
 very near neighborhood  of  the fibers  is important.  As  for particle
 collection we  need  the  velocity distributions in  the whole flow  field.

     Perhaps I  should point out that I talked a  lot about  theories.  This
 is in no  good  accordance with the time  we  spent for experiments.  We have
 spent a very short  time, about 1 1/2  years, with  the theories, and
 about 10  years doing experiments.   That means we  believe  the theories are
 very helpful but we can not find a  solution to  the problem of filtration
 only from the  theoretical  point of  view.   We need experiments, especially
 since in  industrial filters the situation  is quite different from the
 theoretical models.  That  is  why we do  no.t like to emphasize the theoretical
 views.  We should look  more into the  material behavior.

     You mentioned there is a  difference between the influence of Reynolds
 number to collection efficiency between parafin wax particles and filter.
 That is just the point.  What  I meant was  the material behavior.  For
 sodium chloride particles, we  are sure  to  have an elastic rebound.  Because
 we could  recalculate adhesion  probability, we could theorize the same
 behavior  from  the probability  of adhesion.  We measured the beginning of
 rebound of  a particle with a velocity in the range of 5 to 7 centimeters/
 second.   In our experiments with sodium chloride  particles, and multifilter
 the  lowest  velocity has been  25 centimeters/second.  It is in a range where
 adhesion  probability is about  80%.  This is our explanation for the
 difference  between  sodium chloride particles and wax particles.

     One very important change  is the  electrostatic behavior of particles at
 the  fiber.   From the theoretical point of view,  I can only expect there
 must be something going on during the build-up.   We do not have a
 theoretical  solution.  We have tried  to find experimental solutions!

-------
                                                                            237
ELECTROFLUIDIZED BEDS FOR INDUSTRIAL SCALE AIR POLLUTION CONTROL


    J. R. Melcher

    Department of Electrical Engineering and Computer Science
    Massachusetts Institute of Technology


ABSTRACT

    A review is given of the fundamental and practical development of the
electrofluidized bed, beginning with application to the control of submicron
particualte.  Models and corroborating experiments establish fundamental
particle collection coefficients, extend two-phase fluidized bed theory to
account for bubble bypassing and describe the performance with ac excitations.
Practical developments in the control of oil ash and asphaltic fume are de-
scribed.  Bed particle scale electromechanics are elucidated by recently
developed models that are the basis for current efforts to apply the bed to
controlling coal ash.

INTRODUCTION

    As it is applied to air pollution control, an electrofluidized bed  (EFB)
is a  shallow bed of particles, fluidized by the polluted gas, with an electric
field applied by means of electrodes.  Typically, the gas-entrained pollutant
particles are electrically charged  by ion impact prior  to entering the  EFB.
Stressed as  they are by the electric field, the bed particles act as collec-
tion  sites for  the pollutant.  By design, they do not carry a significant net
charge.  The fluidized, usually semi-insulating, particles are subjected to
an electric  field that is imposed either transverse to  the flow (the cross-
flow  configuration) or, by means of screen electrodes and the distributor
plate, co-linear with the flow (co-flow configuration).
    The  following sections  present  an overview of fundamental and developmen-
tal efforts  aimed at applying the EFB to large-scale air pollution control.

FUNDAMENTALS OF SUBMICRON PARTICULATE COLLECTION

    A major  incentive for using the EFB  is  the extremely short gas residence
time  required  for collection of submicron particulate.  A remarkably good
model for plug-flow  beds pictures the bed particles as  being much like  the
electrodes of  an ESP, with  the electric  field terminating over one side of
each  particle  and originating on the other.   Instead of the  Deutsch model

-------
238


used  for  turbulent  flow  ESP's  [1],  the  "local mixing model"  is  invoked with
the assumption  that on the average  the  fluid mechanics  supplies  the  full  sur-
face  of a  bed particle with  the  gas-entrained particulate.   The  electric  field
simply brings the particulate  from  the  local volume to  the bed particle sur-
face.  According to .this model [2], a bed  having unfluidized height  I  , bed
particles  of mean radius R,  a  superficial  gas velocity  U and an  imposed elec-
tric  field E collects particles  of  mobility b with the  efficiency
    n = 1 - exp  (-
u  _ Sir    o /bE\
M  = ~0~ - IT \TTJ
     cj     ^  U
         (1)
where c is a coefficient of order unity.

    This model is valid provided that bubbling and reentrainment are not sub-
stantial.  The first of these conditions is promoted by the several screens
used to impose the electric field in the co-flow configuration.  These tend
to prevent the growth of bubbles.  The second is not a problem for liquid
aerosols such as OOP (used in fundamental studies).

    With the objective of testing this model in a way that would separate
out less fundamental (but perhaps essential) phenomena, extensive efficiency
experiments have been carried out using an apparatus much like that shown in
Fig. 1.  For the data shown in Fig.  2, the cross-flow configuration shown in
Fig. '  is replaced by a co-flow bed.  Efficiencies for OOP aerosols having
two different mobilities are represented, the superficial  velocity varies
from 0.8 to 2.5 m/sec.   For all  of the measurements, i  = 6 cm and the bed
particles are sand,  R = 1  mm.   The solid curve is Eq. T.1)  with c = 1.
Fig 1 - Facility for funda-
mental room temperature
tests using di-octyl phthalate
aerosol.
FLUIDIZED
,'  BED

-------
                                                                239
         0
05  bE/U —
Co-flow efficiency as a function of bE/U for two  different mobilities
and sizes of participate.

-------
240
      The residence-time advantage of the EFB is emphasized by recognizing that
  for an ESP having electrode spacing s, the Deutsch equation for the efficiency
  takes the form of Eq. (1), but with the collection coefficient K,  replaced by
  AbE/sU.  Because 3irc/8 is essentially unity,  this means that the unfluidized
  height of an EFB required to give the same efficiency as an ESP of length I
  is fcQ = fc(R/s).  Typically, bed particles  have  a radius R on the order of
  1  mm, so that for an electrode spacing of 10  cm the length of the  EFB (not
  counting the charter) is theoretically 100 times less than that of the EFB.
      Initial  tests were aimed at establishing  this residence- time advantage of
  the EFB.  Subsequently reported results [2] (typified by Fig.  2) corroborated
  the model  underlying Eq. (1) in the submicron range where the electrically
  induced collection is dominant.  Note from Fig.  2 that without an  applied
  electric field, the bed is between 20 and 30% efficient as a collector.   With
  bubble bypassing minimized and the bed relatively well  controlled  in a plug-
  flow mode, these tests are the best available basis for establishing the para-
  meter c.  For highly fluidized beds,  it is found that c = 0.9 with variations
  of about 15%.  For packed or almost packed plug-flow beds,  c is  about unity
  and varies with test conditions by 25%.   The  local  mixing model  for dilute bed
  particles  predicts c = 1.  These tests were performed with sand  at relative
  humidities of 50 - 90%,  so that the particles were semi-insulating and hence
  the collection field placed under control  [1,3].

      Unless baffled to promote bubble  breakup, gas fluidized beds do not  expand
  uniformly  as the fluidizing velocity  increases and the  resulting effects of
  nonuniformity in the gas-solid distribution is not accounted for in the  plug-
  flow model.   This is especially true  in  the cross-flow  beds where  the screen
  electrodes are not used.   Models have been developed  and  correlated with ex-
  periments  carried out in an apparatus such as that of Fig.  1.  These combine
  the particle-scale collection model tested in the plug-flow experiments  with
  two-phase  models that have been developed  for the transfer  characteristics of
  conventional  fluidized beds.   It is important to  recognize  that  the structure
  of beds  can  be strongly  influenced by an applied  electric  field.   In what will
  be termed  the "bed electromechanics,"  particles  tend  to form "chains"  or
  "strings"  while gas flow causes a  fluttering  motion instead of the  typical
  bubbling one [3].   These  effects have important  implications for bed mixing
  and particulate elutriation.   However, high submicron collection efficiencies
  can be obtained at relatively low  electric field  Intensities where  electro-
  mechanical effects are not of significance.   Thus,  in this  work, the Davidson
  model  is adopted for a conventional bubble having  the 'diameter Dk  (in which
  the particle density is  low)  interacting with the  dense phase.   Part of  the
  gas passes through a dense region  at  about the minimum  fluidization  velocity
 Umf Wh1le  the reminder bypasses the system in the form of bubbles.  There is
 a  continual exchange of particles between phases accounted for by the Davidson
 model.

     Two possible extremes can be used to model the penetration of particulate
 through the bed as a whole.  The dense phase can again be viewed as evolving
 uniformly in the flow direction (the plug-flow model).  Because of bubble agi-
 tation a complete mixing model is also plausible.  Both models have been carried
 through 1n  this study, with the dense phase plug-flow model clearly found to
 be more representative of what happens.

-------
                                                                            241
    The plug-flow bubbling model  shows  how parameters  not  represented  in the
collection coefficient Kn  come into play.   The predicted efficiency  is


                                                                        (2)
            lip IIin


where m, and n^ are the roots of
and
       _ 9 (VVf)  !W_
    C<   2    Db     U'Umf

The collection coefficient K, is the same as defined with Eq. (1), U and A-
are respectively the superficial velocity and fluidized bed height under tne
conditions of interest while U f and i f are these quantities under conditions
of minimum fluldization.  The mist Important parameter is D., the bubble dia-
meter.  The quantities necessary to evaluate these expressions are of course
averages.  The bed height 1s a fluctuating quantity and even in the shallow
beds of interest here, bubbles grow significantly as they pass through the
bed.
    If the rate of gas interchange between the bubbles and the dense phase is
large, which according to Davidson's model means that c  * », the efficiency
predicted by Eq. (2) reduces to that for the single-phase plug-flow model, Eq.
(1).  The most Important aspect of the bubbling model is its prediction of ef-
ficiencies less than 100% no matter how large the collection parameter.  In
the limit K, •*• », the efficiency approaches the limiting value
,*!-[,-Y].^
                                                                        (3)
 Using  experiments  to evaluate c  , it is found that over a wide range of fluidi-
 zation states,  this parameter is mainly a function of the bubble diameter Db.
     In connection  with cross-flow experiments without baffles, the two-phase
 model  has  been  compared with extensive tests aimed at showing the dependence
 of collection efficiency on electric field intensity, bed particle  size,
 static bed height, particle mobility and superficial gas velocity, with em-
 phasis on  the degree of bypassing through the bubble phase.  The bed cross-
 section was 10  x 10 cm.  In order to fix the electrical characteristics of
 the bed, the relative humidity was controlled at 902. The fluidized particles
 were glass beads of 0.5 mm and 1 mm diameter as well as sand with mean diameters
 of 0.8 and 2 mm.  The distributor plate was a 30 mesh plastic screen sandwiched
 between two insulating perforated plates with large holes.

-------
242


    Efficiency measurements over a wide range of parameters are summarized
in Fig. 3, where the parameter is K, from Eq. (1).   The solid curve is  Eq.
(1), and hence the limiting efficiency if bubbling  is ignorable.   A correction
is made on the efficiency for mechanical  effects.   Unfluidized bed heights  of
1Q = 4,6 and 8 cm are represented by the  data.

    That the experimental efficiencies for these bubbling beds do  not correlate
with the parameter K, is evident from Fig. 3.  In effect, the bed  height  be-
comes an additional parameter in correlating experiments carried out under
different conditions.  The data for the 6 cm beds is summarized in Fig. 4,
where the ratio of experimentally observed to theoretically predicted efficien-
cies is shown as a function of the collection coefficient.   The theoretical
efficiency is given by Eq. (2).

    The tendency for the model  to predict the observations  best at high values
of K,  is also seen in results for the other two  bed heights.   Correlation with
the model  over the full  range of Kj  is best for  the deepest bed.

    Details of this work on effects  of bubble bypassing are given  in a forth-
coming article [3].
    100-
    80-
Fig.  3 - Composite of efficiencies for cross-flow beds  as  a  function of col-
          lection parameter.   Widely varying states of fluidization are represented.

-------
     EXP
        .4
        0
                           i  i  i

u
If
KEY
0.35
6
•
.= 6 CM
0.75
7
a
1.0
8.8
H
1.35
9
0
1.75
II
a
3.0
20
a
                                                                             243
                                                                   J	L
5      10
20
                                      K
Fig. 4 - Measured efficiency normalized  to  theoretical efficiency as a  function
         of collection parameter.
SUBMICRON PARTICULATES COLLECTION WITH AC EXCITATIONS

    Savings from a reduced need for power conditioning equipment is an obvious
advantage of being able to use alternating voltages to excite the EFB.  But,
perhaps a more important motive comes from the difficulty encountered when
conventional ESP's are used to collect highly resistive materials.  Poor per-
formance is linked to the dc energization of the ESP.  The collection of highly
resistive ash results in the buildup of net charge and an attendant interference
with the imposed precipitating field.  Such effects are not expected in alter-
nating field devices.  To be effective with an ESP required to collect submicron
particulate, alternating fields must have an extremely low frequency (on the  or-
der of 1 Hz).  (Laminar flow ac ESP performance is worked out in detail by Alexan-
der [5].)  Can the EFB operate with 60 Hz excitations?  More fundamentally, what
are the implications of the frequency dependence of the efficiency for the basic
collection process?
    With beds of varying particle size and states of fluidization, and with applied
fields of different amplitudes, experiments have been carried out using the ap-
paratus of Fig. 2.  Efficiencies were measured as a function of frequency.  Both
cross-flow and co-flow configurations were tested over the frequency range of
30 - 5000 Hz.
    A correlation of the cross-flow data for the former experiments 1s given  in
Fig. 5.  In order to emphasize the particle-scale collection process to the ex-
clusion of bubble bypassing or bed scale mixing, the beds were tested at low
overall efficiencies, 1n a range where the efficiency 1s proportional to the
single particle rate of collection r.  Thus,

-------
 244
         -  r
            nf
                     -

     rdc  "  rnf
ndc * nnf
                                                                         (4)
where  <>  indicates  the  time-average with  ac  excitation, while  nf  and  dc  denote
no  field  and with dc  excitation  respectively.   It  is  this  normalized  efficiency
that is plotted  in  correlating the data of Fig.  5.  To correlate  the  data,  the
angular excitation  frequency u is normalized to  an angular cutoff frequency
<*>e  • \/9TT/8(cs  bE /R), where E  is the  peak applied electric  field and s  is  the
velocity  gradient adjacent to the equator of a given  particle  (with the  local
flow approaching the  particle along a  polar  axis).  This parameter can be ap-
proximated  by  s  = 3U  f(l-Kb)/R (2-3K+3K5  - 2K6)  where K =  R/R   is the ratio
of  particle radius  to the radius of a  unit cell  representing the  particle voidage.
The solid curves represent asymptotic  predictions predicated on the period  of
the applied field either being very low or very  high  compared  to  the  time re-
quired for  particulate  to traverse a unit cell.
                                             effic1enc* as a fu"ction of norma-
          n«                    »ndj"9f.fluency theoretical asymptotes.  Solid
        lines-cross-flow theory, dashed lines-co-flow theory. Rightmost of these
        curves for u>e/W|n = 1, leftmost for u>e/u>m = 0.75.

    Details of the asymptotic models and specifics of the experiments will  be
available in the literature in the near future [6].  It seems clear that such
frequency response measurements are a way of probing the mass transfer process
at the bed particle scale.   From a practical  point of view,  it is  the fact
that the 60 Hz efficiency is well  below the critical  frequency in  all of the
tests that is most significant.  In this low frequency range, the  efficiency
m                    ^ - - - -	   —.. v- • • v • w*i • t ^^ M\> 11 w^  i u i iy
is as predicted by the models for the dc excitations  with E
                                                           dc
                                                       As

-------
                                                                                245
the field reverses, collection occurs first on one side and then on the other
side of a collection site.  The effective electric field is therefore the
average of E .jcos ut| over a full period.
    Experiments, correlated with theoretical models, have been carried forward
to understand how ac excitations can be used in the corona, ion-impact charging
section.  The general disposition of the charger just below the distributor
plate  is as in Fig. 1.  At some penalty of power required, the charging can
be accomplished at 60 Hz.  This makes it possible to operate the entire EFB
system, the bed and the charger, from a transformer.  As an example, Fig. 6
shows  the efficiency with the bed in this entirely ac excited type of operation.
Details are given elsewhere [5].
Fig. 6 - Efficiency with 60 Hz on EFB _?
and charger.  R = 0.5 mm, b = 1.2 x 10
(m/sec)/(volt/m).  Fluidized:  i
U = 1.3 m/sec.  Packed:   «-. = 7.5
= 10
cm,
cm,
U = 0.5 m/sec.
                                                         PACKED
                                                       a FLUIDIZED
 COLLECTION  OF OIL ASH
                  UR
                      2     3
                     IN UNITS OF
                     10* V-SEC/M
     A  natural  application of  the EFB  for large-scale gas cleanup is to the
 products  of oil  combustion.   Experiments have been conducted on a 100 cfm by-
 pass on flue  gas prior  to the economizer in  the M.I.T. Central Utility Plant.
 Loss of heat  in  the  connecting pipes  cooled  the gas to about 350° F.  Tests
 were conducted in both  the co-flow and cross-flow configurations using 1 mm
 glass  beads and  0.8  and 2 mm  sand.
     The oil ash  was  entirely  in the submicron range with loadings of the order
 of 0.01 to  0.06  grains/scf.   Unlike the OOP, the oil ash is a dry material.
 Could  the EFB be used to collect an essentially dry material and what residence
 time for  the  particles  was consistent with retaining a reasonable efficiency?
 A typical test of efficiency  as a function of time is shown in Fig. 7.  For
 this case,  a  medium  state of  fluidization prevailed and the configuration was
 co-flow.  The open circles represent  collection with no charging and no applied
 field.  Because  the  bed was at a sufficiently high temperature that the humidity
 did not render the particles  semi-insulating, there were appreciable "micro-
 fields" associated with patches of charge on the particles created by frlctlonal
 effects.  Thus,  the  open half circles which  represent data for collection with
 charging  but  without a  field  applied  to the  bed, show a considerably higher

-------
246
   efficiency than found in experiments such as represented by Fig.  2.   The  solid
   data points are with both charger and applied field on.
uu
80 '
60 (
40 (
20
0
	 1 	 • 	 r— 	 i—i 	 r-
1* *••••• t ^ 0
3 9 o 9 99 9 9
3ooo°°oo0
°Oo°O 99(1
°oo-
1/ '
                                  5          10
                                     Time (hour)
                                                2Z  24
? -
            Efficiency  as  a  function of time for co-flow EFB on bypass from M.I.T.
            Utilities while  burning low sulfur oil/ Open data points are for no-
            charging and no  bed  field.  Half circles  are with charging but no field
            Closed circles are for charger and field on.

      It is clear that the bed can operate for substantial  periods of time at
  this mass loading.  Either using a once through or a reprocessing system for
  the sand, the EFB appears to be an economically feasible approach to a utilities
  scale operation.  However, there are questions that must be confronted by a
  development.  The M.I.T. plant is required to use low sulfur oil and considerable
  excess air, so for the  tests conducted the combustion products were extremely
  light and free of carbon.   What is the effect on bed electrical  losses of col-
  lecting combustion products of high sulfur oil?  Losses  with the tests reported
  here were insignificant.

      A sketch of a co-flow EFB incorporated into a stack  installation 1s shown
  in F1g. 8.  The top view shows, to scale,  the cross-section of a stack, with
  the EFB constructed around its base.  Flue gas is fed into the annular EFB
  through eight ducts,  passed through the bed and then into the stack through
  ducts in the side walls.  The dimensions are for a 106 cfm unit.  A typical
  pressure drop for 90 - 95% efficiency would be 10 cm H20.

  CONTROL OF ASPHALTIC  FUMES IN REPROCESSING OF ASPHALTIC  HIGHWAY

      Recycling of asphaltic pavement is  being developed by the Warren  Brothers
  Co.  To be economical,  the process  should  be carried out  in adapted conventional
  plants.  A major impediment to the  conversion of existing plants is the forma-
  tion of fine partlculate smoke when crushed pavement is  processed through  the
  conventional aggregate drier.   In the  drier,  the residual asphalt In  the old
  crushed pavement begins  to crack  and release hydrocarbon  vapor which  subsequently
  condenses into droplets.

      Devices currently available for collection of this submicron material  are
  ill suited.  The asphaltic by-products  are  not easily removed from the  plates
  of an ESP.  Fabric filters that will catch  the submicron  particulate  subsequently

-------
                                                                             247
                                                                      EFB
Fig. 8 - A 10° cfm EFB
constructed around the
base of a stack.
                                                                        SAND IN
 are fouled.   Wet wall  electrostatic  precipitators, charged droplet scrubbers
 and high energy Venturi  scrubbers  all  create a water pollution problem.  The
 EFB is well  suited to  this  application,  not only because of its high efficiency
 in the submicron range,  but because  the  material used to collect the particulate
 is easily processed through the  EFB  and  conveniently incorporated into the
 asphalt hot mix product.   In fact, in  this and other related operations oiling
 of the sand is  a desirable  step  in the plant process.
     A prototype EFB of cross-sectional area 1 ft  has been tested on both a
 small-scale prototype  asphalt recycling  system and on a bypass to a full-scale
 asphalt plant operating  in  the recycle mode (Warren Brothers plant #135, Rich-
 mond, Va.).   In both of  the tests  reported here, a bag-house was used ahead
 of the EFB to catch large dust particles from the drier.  The submicron material
 easily penetrated this bag-house.  Tests were performed on a batch basis using
 sand of 2 mm median diameter.  The charger was of the plate wire type.  Tem-
 peratures ranged up to 500° F.

-------
 248

     The results of three separate tests  of the  collection efficiency of the
 EFB on the small-scale recycling system  are reported  in  Table  1.   Experimental
 results are shown along with  the theoretical  efficiencies predicted by the
 plug-flow model and by the two-phase  bubbling model.
EFB OPERATING PARAMETERS
Particulate
Loadi ng
(mg/sm3)
U
(m/sec)
*o
(cm)
E
(kv/m)
EFFICIENCY (%)

Experimental Theoretical
Plug-flow
Model
Bubbling
Model

167
82
44
1.7
1.8
1.8
10
10
13
465
250
500
96 99
93 94
94 99
96
91
92
 Table 1   Summary  of  the  collection performance of the EFB; comparison of experimen-
          tal  and  predicted efficiencies.


 After removal of  dust  by the bag-filter, the mass median diameter of the smoke
 ranges from 0.3 to 0.5 ym, with a geometric standard deviation of about 2.

    An important  parameter is  the amount of time that sand can be used before
 it  must be  replaced  because of a drop  in collection efficiency.  The mechanism
 for this drop seems  to be channeling of the gas through spouts in the bed, oc-
 curing after  the  sand  becomes  viscous  due to excessive oil collection.  In
 Table 2, tests under similar operating conditions are compared.  In.the two
 cases for which the  oil  loading is 0.7%, the measured efficiency for the oily
 sand  is  essentially  the  same as for the clean sand.  In the case of 1.2% oil
 loading, a  drop in efficiency  is observed.
EFB OPERATING PARAMETERS EFFICIENCY (%)
U E JL Fresh Bed Oil
(m/sec) (kv/m) (cm) Particulate Experi-
Loading mental
(mg/sm )
1.8 500 13 44 94
1.8 500 13 44 94
1.7 465 10 167 96
Coated Bed
Particulate Oil *
Loading Loading
107
200
233
0.7%
0.7%
1.2%
Experi-
mental
94
95
85
  Oil loading is defined as the ratio of the weight of collected oil  to the bed
  total weight.

Table 2  Comparison of collection performance of fresh and oil  coated beds.

-------
                                                                              249
    Pressure drop usually accounts for the largest share of EFB  operating
costs.  In these tests, the pressure drop across the EFB, including  the  dis-
tributor plate, was less than 18 cm O.   With the bed operating properly,
this drop is essentially that across the  distributor plate plus  what drop
is required to support the bed.   As spouting occurs because of bed overloading,
there is an associated loss of pressure drop that can be taken as an external
indication of loss in efficiency.
    Results from the Richmond tests are shown in Tablets, where  400~cfm  is  fil-
tered with particulate loadings  varying between 4 mg/m  and 41 mg/m  .  Parti-
culate was observed to have a mass median diameter of 0.35 urn with a geometric
standard deviation of 1.4


   Efficiency Measuring       EFB Operating Parameters      Efficiency
         Device
                            E(kv/m)    U(m/sec)    H
Mass Monitor
Andersen Impactor
400
400
1.9
2.1
10
11
99.3
99.1
Table 3  Summary of collection performance in Richmond.

    This work will be reported in more detail in a forthcoming article [8].

ELECTROMECHANICS OF ELECTROFLUIDIZED BEDS

    In the application of the EFB to submicron particulates control, it is
possible to separate the electrical interaction between  the bed particles and
the charged particulate from the electrical interaction  between bed particles.
It is the latter that is termed the bed "electromechanics."  The propensity
of the field for altering the state of fluidization by inducing interparticle
stringing is discussed in previous publications [3].
    If the bed is to be applied to collection of both submicron and super-
micron particles, it is necessary to understand the implications of the sur-
prisingly large electrical effects at the bed particle scale.  To what extent
can these forces be relied upon to prevent loss of large particles from the
bed?  Because the field can turn a bubble into a "fritter," the turbulence
within the bed is considerably reduced by the field.  Each bed particle is
subjected to a reduced amount of agitation because the "wake  mixing" associated
with the classical bubble is reduced or eliminated.  To  what extent are these
effects useful in achieving EFB operations in a mode where bed particles are
agglomerates of the particulate, the self-agglomerative  mode?

    Two types of fundamental experiments have been carried out with the objec-
tive of having a physical picture of the interparticle force.  In the first
of these, the EFB (in a cross-flow configuration) is in  the annul us between
the cylinders of a Couette viscometer, as shown in Fig.  9.  The shear stress

-------
250
             SUPPORT
               SYSTEM
                     INNER
                      CYLINDER
                   x FLUIOIZED
                       PARTICLES
                     OUTER
                       CYLINDER
                    PLASTIC
                                  200
  100
                                    50
jgzo
LJ
QL

w 10

CC
<
UJ g.
I 5
                                          1
                                                     I  ,1  I	I
                                                2        5      10     20
                                                 VOLTAGE  (KV)
        Cross-flow EFB in viscometer.  Outer cylinder rotates and inner one is
        fixed.  Voltage applied between inner and outer cylinders results in
        shear stress at inner cylinder as shown.

tends to a common asymptote at high field strengths over a range of shear
rates.  In this high field regime, the electrical force is large enough that
the bed is electropacked.  Each particle chain transmits its force through
friction to the electrodes, so the shear stress can be used as a measure of
the interparticle force.
    For the particles typical of the EFB applied to air pollution control,
polarization forces  (associated with the permittivity of the particles) are
too small to be responsible for the observed effects.  Also, such observations
as the dependence of the electromechanical effects on applied voltage negate
such polarization forces as the basis for what is observed.  A theoretical
model has been developed in which the force between particles is accounted for
in terms of the concentration of the electric field near the contact caps
between particles.   In this region there is a strong constriction of the
current as it passes between particles.  The correlation between theory and
experiment supports  the contention that it is this field concentration that
is responsible for the large electromechanical effects and their dependence
on the applied voltage.  This work is reported in more detail in a forth-
coming article [9].

-------
                                                                               251
    Further experiments allow deduction of this interpartjcle force from bed
characteristics.  In the first of two experiments in this category, the pres-
sure drop at incipient fluidization is measured as a function of the applied
voltage.  In the second, the minimum voltage required to suspend a bed against
an upper screen is measured as a function of the gas-flow rate.   Both experi-
ments are performed using the semi-insulating particles typical  of the sand
at high relative humidities with the cross-flow configuration.  Typical re-
sults for this latter experiment are shown in F1g. 10, which shows the super-
ficial velocity at which particles fall from a pinned position as a function
Of voltage.  The inset shows an idealization of the curves, identifying two
regimes.  In the low-field regime, the velocity 1s above U -, and so there is
enough pressure drop to support the bed as a whole.  In this regime, failure
is because particles at the bottom interface are not held in equilibrium.  In
the high-field regime, the electric field makes it possible to support the
bed as a whole and failure involves the entire bed.  Correlations between the
bed measurements and a theory based on interparticle forces due to current
constriction and a contact cap determined in size by electrical  breakdown are
not only successful in predicting the dependence of the parameters, but in
giving quantitative predictions as well.  This work will be reported in the
literature [10].
Fig.  10 - Air velocity required to  support a  cross-flow electropacked bed against
          an upper stream as a function of the  applied voltage.

-------
252

COLLECTION OF COAL ASH

    Work is being carried forward by Alexander (Ph.D.  research)  to apply the
EFB to the control of coal ash.  Experiments are on a  100 cfm facility that
ranges to 600° F in temperature.  Two approaches are being investigated.   In
the first, the bed particles are coal, which is used as a collection  site
for ash in a process that builds up particles of about twice  the original  mass
before the coal and ash particle is removed from the system.   Combustible ad-
ditives are used to promote stable agglomeration.   A picture  of  one of these
composite products from the EFB is shown in Fig. 11.  A typical  collection
efficiency for the bed is shown in Fig. 12, where the  advantage  of using  the
EFB for the submicron collection is apparent.  Note that a fluidized  bed  with-
out electrification can be used for removal of particles above about  3 ym.  In
the system being developed, two beds are used.   In the first, the large particles
are removed and in the second, charging and application of a  field results in
good efficiency for collection of the fines.

    In the system using coal as the seed material, the output of the  bed  is
to be injected into the combustor where the ash.cracks away and  the coal  and
additive are burned.  Provided that the combustor is one of relatively high
ash retention, this approach can be shown to lead to good overall  efficiency
with the ash removed from the combustor.  The fluidized bed combustor is  a
natural  candidate for this mode of operation.

    The second system under development starts  the collection process by  "seeding"
the bed with foreign particles.  As the agglomerates are built up,  they are
removed, treated, and a fraction broken into particles  suitable for reinjection
into the bed.   Thus, the bed is operated in a self-agglomerative mode.  Detailed
studies are being made of the interplay between parameters contributing to  this
process.
                               *   %**
F19-
       - Composite ash and coal particles built up in the EFB.   Coal  particle
         alright (2 mm diameter) is seed for agglomerates shown at left.   The
         third particle from the left has been crushed.

-------
    100
     80    -
-   60
 
-------
  254


  to the general application of the EFB to air pollution control have been
  indicated to be held allowable.  A third relating to the coal ash collection
  is pending.

  REFERENCES

  [1]  White, H.J., Industrial Electrostatic Precipitation, Addison-Wesley, Pub.
      Co., Reading, Mass., p. 164,1963.

  [2]  Zahedi, K. and Melcher, J.R., "Electrofluidized Beds in the Filtration
      of Submicron Aerosols," APCA Journal, Vol. 26, p.  345, 1976.

  [3]  Johnson, T.W. and Melcher, J.R., "Electromechanics of Electrofluidized
      Beds," I&EC Fund., Vol. 14, p. 146, 1975.

  [4]  Zahedi, K. and Melcher, J.R., "Collection of Submicron Particulate in
      Bubbling Electrofluidized Beds," I&EC Fund.(in publication).

  [5]  Alexander, J.C., "Frequency Characteristics  of Electrofluidized Beds in
     the Collection of Submicron Particulate," M.S.  thesis, Oct.,  1975.

  [6]  Alexander, J.C. and Melcher, J.R.,  "Alternating Field Electrofluidized
      Beds in the Collection-of Submicron Aerosols," I&EC Fund, (in publication).

  [7]  Final  Report to ESEERCO on Electrofluidized  Beds in the Filtration of
      Submicron Particulate, Continuum Electrotnechanics  Group. M.I.T., T976".

 [8]  Zieve, P.B.,  Zahedi, K., Melcher, J.R.,  and  Denton, J.F., "Electrofluidized
      Beds in the Filtration of Smoke Emissions  from an  Asphaltic  Pavement Re-
      cycling Process,"  Env. Sci. & Tech., submitted for publication.

 [9]  Dietz, P.W. and Melcher, J.R.,  "Momentum Transfer  in Electrofluidized Beds," AIChE
      Symposium Series Book on Air, submitted  for  publication.

[10]  Dietz, P.W. and Melcher, J.R.,  "Interparticle  Electrical Forces  in Packed
      and Fluidized Beds,"  I&EC Fund.,  submitted for publication.
 WRITTEN DISCUSSION


 Robert B.  Rief
 Battello Colombus  Laboratories,  505 King Avenue
 Columbus,  Ohio   43201


     My experience  with  electrofluidized beds differs somewhat from Jim
 Melcher's  experience in that  in  my work the beds were used to supply powder

-------
                                                                             255

to electrostatic coating and printing processes which required moving
particles larger than 10 microns through or out of the beds.   Particles finer
than 10 microns and especially finer than 1 micron were avoided  because the
surface molecular forces or Van der Waal forces cause the fine particles to
stick to the larger bed particles.

    I have been very interested in Jim Melcher's work and have been following
it for some time.  Although a number of questions can be raised on the basis
of the information mentioned in this paper, these questions have been answered
in papers which have been published earlier.  I strongly recommend these
papers for further reading to anyone interested in electrofluidized beds.

    I find a problem with only one part of the paper.  The statement is made
that "Typically, bed particles have a radius R on the order of 1 mm, so that
for an electrode spacing of 10 cm the length of the EFB (not counting the
charger) is theoretically 100 times less than that of the ESP."  The factor
"100" was obtained by equating the K coefficients in the Deutsch equation and
in the efficiency equation of the electrofluidized bed.  A factor of 2  was
omitted from the K factor of the Deutsch equation.  Thus, the length of the
equivalent ESP should be 200 times rather than 100 times longer than the
length of the EFB.

    Since I have no further comment about this work, I shall spend the rest
of the alloted time in discussing another new air cleaning system which is
called an electroinertial air cleaner.

    Inertial air cleaners consisting of a bank of straight flow tubes, each
tube having a swirl means to cause dust to be centrifugally thrown outwardly.
toward the tube wall, are well known for engine air cleaning application:
Dust near the tube wall is drawn from the main stream by a scavenger fan and
cleaned air emerges from the core zone of the tube through a small diameter
take-off tube extending into the downstream end of the flow tube.  However,
the centrifugal separator action is relatively ineffective on particles smaller
than 5 microns; therefore, electrostatic separator action was added in electro-
inertial devices to enhance overall collection efficiency.  In one arrangement,
a fine wire at a potential of approximately 15-20 kV is extended through a
1-1/2-inch-diameter tube on the tube centerline with the tube wall at ground
potential to establish a flow of negative ions.  The resultant negative charges
on the particles and the radial electrical field from the ionizer wire to the
tube enhance outward migration especially of the smaller particles thereby
improving overall collection efficiency.

    Unfortunately, the electrostatic separator action causes some particles
to precipitate  and adhere rather  strongly to the tube side wall.  Therefore,
the tube usually must be periodically rapped or vibrated in the radial and/or
axial direction  in order to dislodge the particles sufficiently to permit the
scavenger air  to carry them away.  A rapper equipped tube is shown in Figure 1.
Under some circumstances the particles are jarred with such force as to be
reentrained into the clean air stream.  Even when properly applied, rapping or
vibrating requires special shock mounting of the tube; mechanical wear on the
mounts is a problem.  Therefore,  the use of rappers as a dislodging expedient

-------
                                                                         a-.
FIGURE  1.   ELECTROINERTIAL  UNIT WITH RAPPER

-------
                                                        257
Steel Tube
                              Dirty Air
                                 Exit
                               Clean Air I
                                 Tube
FIGURE 2.  ELECTROINERTIAL UNIT WITH
           GLASS LINED TUBE

-------
 258

 is not entirely satisfactory.

     While working with a unit  of this type under Contract No.  DAAE07-72-C-
 0293*, a glass  lining'was tried in the tube as shown in Figure 2.   During trials
 without vibrating the tube,AC  fine test dust was passed through a  1.5-inch-
 diameter by 6-inch unit at a flow rate of  40 cfm including 4 cfm of scavenger
 air.   No measurable amount of  dust accumulated in the tube.  The fact  that no
 dust  accumulated in the glass-lined tube represents  a major advance in the
 technology.  The overall efficiency was about 92 percent.   For  comparison,  with
 a  vibrated unlined tube with comparable voltage on the wire, the overall
 efficiency is generally about  99 percent.   As  an inertial unit,  with no applied
 voltage,  the same size unit  would remove about 80 percent of the dust.  The
 electrical contribution with the glass-lined tube was significant—increasing
 the  efficiency  from 80 percent  to 92  percent.

     The low material  retention  in the glass-lined tube is believed  to  be
 unique to  "leaky" insulators such as  glass.  In the  precipitator section,  large
 numbers of ions are generated but only part  of the ions are effective  in
 charging the dust.   Once the dust is  deposited on the tube, excess  ions bombard
 the dust and inhibit  its release from conductive surfaces.  However, the excess
 ions  accumulate on insulators and prevent  the  deposition  of the  dust.   If  the
 insulator  is too  good,  the potential  will  rise on the surface  and the  corona
 from  the wire will be  shut off causing  the efficiency to  drop.   "Leaky"
 insulators,  such  as glass, let some current  pass  so  that  the charging  process
 continues  but some  charge accumulates on the surface.   This surface  charge,
which  is the same  polarity as the  charge on  the dust,  repels the dust  as it
approaches and  prevents  it from making  good  contact with  the surface.   As  a
result,the surface molecular forces,  which act over very  short distances and
cause  fine particles to  adhere tenaciously when good  contact is made, are  not
effective.   The loosely  held dust  thus moves along the  charged surface  of  the
tube and out through the scavenger system.

       Recently U.S.patent4,010,Oil was issued  on this device  and further  work
has been done for  the USDA on other modifications of  the basic electroinertial
design for removal of fine cotton dust.
 USA TACOM.


 OPEN DISCUSSION

 Oder:  The question has to do with bubbles  and the  suppression of  bubbles
 with the electrofluidized bed.  What  is the fundamental  mechanism  for  the
 creation of bubbles and why should the  electrostatic  field  suppress  that?

 Melcher:   The basic reason for the bubbles,  is instability  in  fluid. Usually
 that is how it is approached in the literature of fluidized beds.  It  is
 unstable,  and the nonlinear equilibrium that results  is  the bubble.  The

-------
                                                                            259


other question, how is it that the electrostatic field seems to fill in the
bubble?  Again this is the sort of dependency in this self precipitation
time constant.  Particles do not have net charge, they do not go off the
wall, and they do not annihilate each other.  But they are polarized and that
now has a new time constant.  That time constant is now in the steady state
configuration in competition with the turbulence and, at a certain point, is
short enough that it overcomes the inherent mixing time constant and they
form strings.  These strings are held together by this capped force which
we are talking about.  I can use an electric field to support a static bed
very deep in spite of the fact that a field that large will not support
water at an altitude of a fraction of a millimeter.  The force between the
particles is a very.very strong one and it can be transmitted to the walls.
That is why you get the overshoot.  It can change the effective shear stresses
as well.   Students in industry are now working with  S02  on the type of
approach.

Oder:  Let me further my question. You are not really getting rid of a
fundamental instability.  You have really put a gas distributor in there.
In the bed where you have this string of particles do you have mass bed
mixing or is  the bed stable?

Melcher:  The bed  is indeed mixing.  I have a movie with me  if any one wants
to see it.  With the field on, the bed mixing is much slower.  The way  in
which  the particles are  treated by the bubbles  is different  because the
bubble goes  through and  you get strong mixing in the  trail.  What the  field
does is  eliminate  that trail,  it  eliminates a lot of  the  agitation  in  the
bed.

Koppang:  I  was wondering what your mechanisms  are  for  cleaning your bed.

Melcher:  I  appreciate your question.  Two  advantages of  the thing  are,
 first  of all, the  savings you get in  the pressure  drop  and next,  it flows
 like a fluid. You put a hole in  this  thing and it  runs out.  You can  run
 this thing  on a  continuous  basis.  Generally you let it run down  through a
 hole and you put  a little  in.  It is  a very slow rate at which you  do  that.
 They use more oil  and sand  than we can supply.   For certain kinds of
 applications I can see  for  fly ash you might agglomerate and recycle the
 whole  thing.

           On the fly ash we have  two approaches.  One is to self  agglomerate
 the fly ash on itself.   The other is to agglomerate on coal.  In either case
 we want to recover the heat  value of additive of the coal injected into the
 combustor.   A combustor like a fluidized bed combustor has a high ash
 retention,  so that taking all the ash out of the combustor is the grand plan.
 Also I might mention we are working on recycling seed for MHD generators,
 where we have a local pollution problem in our MHD experiment.  The sodzum
 carbonate that we put in as seed turned out a submicron particulate all
 over the campus and an EPA warning.  We've gotten ourselves a 70 inch pressure
 drop scrubber which gets 70% of  the stuff.  We are trying to collect it our-
 selves since it is valuable.  It is a nice application; we  can put it right
 back into the combustor.

-------
 260
 APPROXIMATE EQUATIONS FOR PREDICTING ELECTROSTATIC
 PARTICLE COLLECTION
     Douglas W.  Cooper

     Department of Environmental Health. Sciences, Harvard University


 ABSTRACT

     A new approach to calculating approximate single-collector collection
 efficiencies is presented that eliminates the need to calculate particle
 trajectories, for particles of negligible inertia.  The general case of a
 central force, F = k/rm  (m i 0), is solved for spherical and cylindrical
 collectors having collection efficiencies either much greater than or much
 less than one.  For those values of  m for which efficiencies have already
 been evaluated by others (using trajectory integrations), the results agree
 quite well,  except for the low-efficiency case  of a charged particle and an
 uncharged collector, where the central force approximation breaks down.
 The approach taken here emphasizes the significance of the aerosol con-
 centration profile, n(r).  It shows that increased  collection efficiency will
 result from improved mixing of the  aerosol in the vicinity of the collectors,
 perhaps through induced turbulence.   The  collection efficiencies are  ex-
 pressed in terms of  K = ws/vo>  the ratio  of the particle radial velocity at
 the collector surface (due to the central force) to the mean gas velocity far
 upstream from the collector.   The collection efficiencies  are as follows:

 Collector                       K « 1                    K » 1

 cylinder                          n K                       (TT K)1/m

 sphere                           4K                        (4K)2/'m

 The results are exact for coulombic attraction (charged particle, charged
 collector, point and/or line charge approximation), if inertia is negligible
 and if interception is not important enough to produce critical trajectories
 which strike the  collector  upstream  from the rear stagnation point.  Also
 discussed are the effects of other particles, other collectors, and other
 fields, with respect to the collection efficiency of a single collector.

INTRODUCTION

    Recent emphasis on the control of respirable dust, airborne particles
with aerodynamic diameters s 3 um, has led  to increased interest in using

-------
                                                                          261

electrostatic forces to enhance the collection in filters and scrubbers.  To
predict the enhanced collection efficiency of these  control devices exactly
is very difficult.  Often one can determine whether or not the addition of
electrostatic forces to a conventional scrubber or  filter would be a major
improvement through the use of approximate equations for collection
efficiency.

GOALS

    The goals of this work are to make it easier to calculate approximate
collection efficiencies of control  devices (such as  scrubbers and filters)
which employ electrostatics to augment their usual collection mechanisms
and to clarify the role of electrostatics in such devices.  The alternative
to such simple calculations is often the rather difficult determination of
particle trajectories for the variety of conditions and particle sizes of
interest,  an expensive endeavor.

ANALYSIS

    A.  Particle Motion

        Denote the particle velocity by u and the gas velocity by v.  The
well-known equation of motion for a particle governed by Stokes1 law with
the Cunningham slip correction is

    m du/dt  = -3nu d (u - v)/C + F                                     (1)

in which m is particle mass (g),  t is time (s), u is gas viscosity (poise),
dp is particle diameter, C is the Cunningham  slip correction (approximately
1 + 0.16 um/dp at NTP), and F is the vector sum of any external forces
acting on the particle (dynes).  Fuchs[l] discussed this equation's range of
application in detail.  It is generally applicable  to particles with diameters
from 0.1  urn to 10 urn.

    Equation (1) can be simplified by introducing particle mobility, B,  given
by

    B = C/3TT|_id                                                       (2)

for spherical particles.  The  mobility is the terminal velocity of the particle
with  respect to the gas per unit applied force.  The mobility is also the ratio
of the particle aerodynamic characteristic time, T, to the particle mass:

    B=T/m                                                           (3)

where, for spheres,

    T = Cp  d 2/18u                                                    (4)
           P P
 (p is the particle density).
   P
    Equation (1) then becomes

-------
 262

    T du/dt = -(u-v) + BF.                                             (5)

    HThe first simplification we make is to eliminate the inertial term
 T du/dt.  We distinguish between the "control device" and the "collectors"
 (fibers, droplets,  etc. ) within the device.  Let the mean flow velocity (or
 the flow far upstream from the collectors) be vo and let the collector's
 smallest dimension be  L.   Then  T du/dt becomes negligible when

    vQT/L«l                                                       (6a)


    BFT/L«1                                                      (6b)

 under which conditions

    u = v*+ B?                                                        (7)

    Inequality (6a)  means that the particle inertia is insufficient to cause it
 to deviate appreciably from the streamlines in the vicinity of the collector.
 Inequality (6b) means the particle inertia is insufficient to prevent the
 particle from almost instantaneously achieving its terminal velocity with
 respect to the fluid due to the force  F.  Equation (7) indicates that under
 these assumptions the particle motion is  the motion^of the gas plus the
 particle terminal velocity with respect to the gas, w = FB.

    B.  Particle Collection

        Figure 1 shows a collector of arbitrary shape in an aerosol flow of
 free-stream velocity VQ and free-stream concentration n .  The number of
 particles collected per unit time at steady-state will be tRe integral over
 the collector surface of the particle velocity normal to the  surface times
 the concentration:
             s s
    N = -d>   n u  •  d                                                   (8)
         •                                                             l°'
where Ag is the surface area (cm  ), u  is the particle velocity (cm/s ),
and ng is the particle concentration at tine surface (cm~3).  In steady state,
the number of particles removed. per unit time from the volume (spherical
or cylindrical) within R2 is also N,
    N =  -    n    -  d                                                  (9)
    Evaluation of the preceding integrals requires concentrations, velocities,
and surface areas.  One simplification is that at the  surface of the col-
lector the normal component of velocity  for the gas must be zero, so
Equation (8) becomes
    Ns = - £  ng ws- dAg                                              (10)
           As

-------
                                                                        263
               >
     Fig.  1  Collector of arbitrary shape in aerosol flow of free-stream
             velocity  v  and concentration n  .
                   '
   C.  Collection Efficiency - Coulombic Attraction Only

        Coulombic attraction is the attraction between one point charge (the)
particle) and another (the collector) or between a point charge and a
line charge (a cylindrical collector).  The charges are opposite in sign
and are treated as being located at the centers of the  particle and the
collector.  The force between the point charges decreases with the inverse
square of the distance (r'2); the force between the particle and the line
charge decreases inversely with distance (r'1)-  This force will often
predominate over the other electrostatic forces.  Equation (1) becomes
easy to evaluate because, as Levin demonstrated (see Fuchs [1]):  for
forces, such as gravity or coulombic attraction,  for which the divergence
is zero; that is, for

    V.F=O                                                         <">

the concentration of particles in an inertialess aerosol  subjected to that
force is constant along particle trajectories.  Thus, the surface con-
centration at the collector will be the same  as the free  stream concentration


     8    O

which means

-------
264

    N = n w A                                                        (13)
         oss                                                       l  '

for a spherical or cylindrical collection surface,  where A  is the surface
area.                                                    s

    If  Ac is the cross-sectional area of the collector in a plane perpen-
dicular to the mean flow,  v ,  then the collection efficiency of a single
collector is defined as
           oVoAc                                                     <14)

For a cylindrical collector and couloumbic attraction alone, the collection
efficiency becomes, from equations (13) and  (14)

    Tl = TT W  /v                                                        (15)

and the collection efficiency is

    n = 4wg/vo                                                        (16)

for a sphere.  Usually,  efficiency is expressed in terms of the ratio,  K,
of the electrostatic force on the particle at the collector surface (F  ) to
the drag force (Fj^) associated with a velocity vo, but in fact

         F      w  /B      w
    K _   s    _  s	s_  ^                                      ,J7j
          D     vo         vo

Note that the coulombic force  decreas.es as rapidly as the collection area
increases; thus,  the  collection rate, N, is independent of  size, even
though the collection efficiency  is dependent  on size.   If particulate capture
by interception can be correctly modeled as an increase in the effective
collector size (increased outward from the surface by a dimension equal to
the particle radius),  interception will not increase the rate of capture,
unless ws(r=R-lr ) at the collector rear  is smaller than v(r=R+r ) there.
              f                                               p.
    It is  also significant that the rate of capture  nowsA8» is independent of
the flow  profile,  and so is  the efficiency.

    Further, if a surface such as RI in  Figure 1 can be drawn surrounding
an irregular  particle, and if  that surface  is  completely intersected by
particle  trajectories (collection efficiency  much greater than one should
suffice), the collection rate will also be independent of shape, for coulombic
attraction, which agrees with  a result of Levin (see Fuchs  [1]).

    D.  Collection Efficiencies - Central Forces Other than Couloumbic

        For the other electrostatic forces,  it is not generally true that the
divergence is zero.  Thus, particle concentrations are not constant along
trajectories.  Instead,  for a central force

-------
                                                                         265
   F=k/rm
in steady-state  with no gas flow, the rate of particle passage through a
(spherical or cylindrical) surface surrounding the collector would be con-
stant at each distance r (cm):

    N = n(r)w(r)A(r)
If the concentration is nQ  at rSRQ, then
                  m A(R )
                    -             "                                 (20)
    For the case in which the velocity v  is much less than w  , equation
(20) should hold near the collector.   Thus for a spherical collector:
                    m   R  _     _   m-2

and for a cylindrical collector

     n(r)   _ , _ r_vm-l                                               (22)
      no       Ro

assuming w » v .  Often collection is put in terms of the free-stream
streamline farthest from the flow centerline that is cleaned, a distance R*
 such that
    N  = n v  IT R*2                                                   <23a)
         o o
 or
    N = n v  2R*L                                                   (23b)
         00      C
 for a sphere or a cylinder (of length Lc), respectively.  The assumption
 ws> > v  implies  that R*> > R,where  R is the collector radius, which
 makes plausible the assumption

    n(R*) '= n(RQ) = nQ  .                                             (24)

 The concentration at R* is approximately the free-stream concentration.

    For a sphere, this implies
N = n v n R*2 = n(R)w(R)A(R)                                     (25a)
     o o                    ,
                2 = nw
                                ,
                             m-£    R,        7
                                        r)HTTR2)                    (25b)
                                     R

-------
 266

 Thus,
    R* = (4Bk/v )
                  I/
                    m
                o'
(26)
    One obtains the collection rate from equation (23a) and the collection
 efficiency from

    TI = (R*/R)2 =  (4Bk/Rmvo)2/m                                    (27)

 which is equivalent to
    T] =  (4w /vJZ/m = (4K)2/m
            s  o
(28)
 This is a general expression for strong central force collection efficiency
 for a sphere, derived without reference to a specific flow or a specific
 force         ( m / 0 »

    For a cylinder, a similar analysis  yields

    Tl = (R*/R) =  (TTBk/Rmvo)1/m = (nK)1/m                         (29)

 a general expression  for strong central-force collection efficiency for a
 cylinder.  This approach is believed to be original.

    In the other limit,  w « v , the collection velocity is much less than the
 free -stream velocity.  Because the collection, velocity is low compared with
 convection  (vo), the concentration near the collector  surface  should be quite
 nearly the free-stream concentration:
        •
    ns =  V                                                         (30)

 in which case the  collection rate becomes:
     •
    N  =  n w A   .                                                   {->i\
          oss                                                     1-5 -U

 The collection efficiency for a spherical collector will be

    r, =  4wg/vo  =  4K                                                (32)

 and for a cylindrical collector

      =  TTW/V   =  nK                                                (33)
               O
    Equations (32) and (33) may also be correct for  w ~v   when the gas is
sufficiently turbulent to mix the particles well  in the vtcin°ty of the collector,
because the equations are based on approximating  n(R)  by  n  .  Turbulence
might, thus, raise collection.                              °

    Intermediate cases (K = w  /v ~ 1)  should  have efficiencies intermediate
between the  K» 1 and the  K %< 1  cases studied above,  because the

-------
                                                                          267


concentration at the collector surface should be between  no  and the n(R)
calculated for  K» 1.

    Approximate efficiency formulas have been obtained here without recourse
to trajectory calculations and without assuming particular flow profiles,
assumptions having  been made about concentration/ profiles instead.  The
equations apply to any force such that  F = k/rm.

RESULTS:  COMPARISON WITH OTHER EFFICIENCY CALCULATIONS

    Table 1  is based partly on a similar table by Whitby and  Liu  (2). The
first column has the collector shape.  The second indicates which force is
being isolated; the subscript  Q  means the collector is charged and the
subscript q means the particle is charged.   The  third column shows the
exact or approximate form of the radial  dependence of the central  force.
The fifth column has the efficiency calculated by the approximations above
for K  range shown in the fourth column.  The last three columns give the
efficiencies obtained by others, the flow conditions they assumed,  and the
reference,

    Where there are  other determinations with which to compare them, the
approximations made here agree rather well (except for  Foq at K«  1,
where the central force expression does not remain valid).

    It can be noted, in general, that as K  increases, the efficiency goes
from being  linear in K to being  sublinear in  K,  except  for coulombic attrac-
tion,  which means that as steps are taken to raise  K,  a  law of diminishing
returns sets in,   TI increasing  less  rapidly with  K.

DISCUSSION

    This simplified approach for the calculation of the  single collector elec-
trostatic collection efficiency neglects differences in flow profile, collection
by particle  interception, and particle  inertia. These aspects, and  others,
are discussed next.

    A.   Flow Profiles

         Kraemer  and Johnstone [3] and others [4, 8] have found  that for coul-
 ombic  attraction without interception far inertialess particles, the collection
 efficiency is  independent of whether the flow profile is potential or  viscous.
 Our analysis shows coulombic  collection of inertialess particles (even with
 interception) to be independent of the flow profile whenever  the critical
 particle trajectory strikes the  collector directly in the rear (rather than
 when part of the collector is not  intersected  by particle  trajectories).

    For non-coulombic attraction,  the results of Kraemer and Johnstone
 and others  indicate  some dependence of collection efficiency on flow  profile.
 Results [3] of calculations for collection of uncharged particles  on  a charged
 spherical collector were  nearly indistinguishable for potential and  viscous
 flows.  For charged spherical particles, and an uncharged  spherical

-------
                                 Table 1.   Efficiency Determination Results
                                                                                 IS)
                                                                                 c^
                                                                                 CO
 Collector
               Force
          _	Collection Efficiency Determinations
          SJroxl-      ?ther  Efflclency Determinations
           matlon	Efficiency            Flow
                                                                                               Ref.
 Sphere
                 Qq
                          ~2
                                all
          UK
                                      potential
                 QO
                                         IK
Cylinder
                 0q
                            	                 [33
                            potential                  [3]
                            potential and  viscous      [3]
                "QQ
                QO
                0q
all
all
4K_

irK
                                                      2.9(K)°'35     potential
                                                                     viscous
                                                                     viscous
                                                      nK
         irK
         UK)1/*"
            ii
             irK	
             (_JfK_)
             v2-ln Re;
             (6,rK)1/3
                            potential
                            all(r_,
                           	p .
                            potential
                            all(rp/R«l)'
viscous*
potential"
 viscous
( experiment)
                                K«  1    n K
                                                      2.3(K)1/2
                        ...I3J
                           [3]
                        _.C5J
                                                                                                C5]
•Natanson's condition was  that  capture  could occur  at  the rear of the cylinder:
  -BF(rp+R)/v(rp+R)^l at 6=0.
[5]
C5]

[6]
C6]
[7]

-------
                                                                          269

collector, the Kraemer and Johnstone [3] calculations showed that the
collection efficiencies were two to three times higher for potential flow than
for viscous  flow ( the streamlines are farther from the collector in vis-
cous flow),  for  K  in the range  10'4 to  10. Other profiles might pro-
duce greater differences  in efficiency for the  same  K,  especially flows
at sufficiently high collector Reynolds number to produce a turbulent wake
behind the collector.

    B.  Particle Interception

        Calculations [3] for spherical collectors and spherical particles
showed interception to  be negligible for K> > rp/R and dominant for
rp/R»K, for inertialess particles.

    C.  Particle Inertia and Impaction

        Particle inertia can lead to impaction  of the particles, the particles
may also delay significantly in reaching the electrostatic terminal velocity.

    Grover  and Beard [9] calculated coulombic collection efficiencies for
charged cloud drops and  charged particles. They did this by trajectory
determinations, using flow fields which were  calculated by solving the
Navier-Stokes equations,  rather  than using the viscous or potential flow
approximations.   (They found  that the critical trajectories for inertialess
particles did indeed strike the collector directly in the rear).  The expression

    n  = 4K                                                         (34)

was found by them to be correct within  a factor of two for collector Reynolds
numbers less than 100 and for inertial  parameters

    Stk  = vQT/R<0.2                                                (35)

for the range of particle  and droplet charge levels  studied.  This confirms
the restriction on the impaction parameter:

    Stk«l  .                                                         (36)

The same work indicated that as  Stk becomes much larger than 1,  the
collection efficiency became dominated by impaction.

    That raises another interesting point.  The particle collection rate for
the collector with combined impaction and electrostatic attraction will be
very approximately the sum of the two  collection rates:

    N  = n  w A   +  n  v  nT(Stk)A                                   (37)
          s s  s      o  o ' 1      c
where  r\. (Stk)  is the collection efficiency due to impaction alone.  The
collection efficiency becomes

-------
  270


    1 =  rTTST-  +  Vstk)                                        (38)
           o o  c

 The electrostatic efficiency will decrease as vo  increases, and the impac-
 tion efficiency will increase.  Although this means there is a minimum col-
 lection efficiency at some non-zero vo,  it is misleading.  The particle
 collection rate  should increase monotonically as vo is increased, even
 when electrostatic attraction dominates impaction.   For the same gas volume
 flow  rate and the same total collector area, this would indicate total col-
 lection can be increased by making the entire collection zone of the control
 device have a smaller cross-sectional area perpendicular to the mean flow
 and a longer dimension parallel to the mean flow.  This should  not be pursued
 to extremes,  however,  because the pressure drop across the colle'ctor will
 also increase as the geometry is changed in this way.

    D.  Determining Whether Electrostatic Forces are Important

        Collection efficiencies are not generally additive,  but the effect of  two
 or more collection mechanisms should be  greater than the largest of any of
 them and less than their sum.  The removal of particles from  the gas means
 creating a net particle velocity perpendicular to the mean flow  sufficient to
 cause the particles  to strike and be held by the collection surface.   The
 total  collection  surface will be the collector surfaces plus the inner surface
 of the device's collection zone,  that volume that contains the collectors. A
 first approximation for comparing the magnitudes of the collection mechanisms
 is to estimate the effective velocity at the  surface of a typical collector due
 to each of the mechanisms; these approximate velocities are as follows:
 - impaction: v  Stk
 - interception:  v  r /R
                 op
 - gravitation: g T

 - diffusion: D/R (D is particle diffusivity,  cm2/s)
 - electrostatics:  w
                    s
    If the estimated electrostatic velocity, w is similar to or larger
than the largest of the other estimated velocities, then it should be included
in the efficiency estimate.

    The magnitudes of the particle migration velocity  w  due to electro-
static forces will  clearly depend upon details of particle charge and size
and collector charge and size.  Figure 2 shows these velocities [13] as a
function of particle size under the following assumptions: the particles,
if charged,  are charged to saturation in an electric field of  10 kV/cm; the
collector, if charged, produces a field of  10 kV/cm  at its surface and is a
100 (am spherical collection (not grounded).  The particles are unit density
spheres of the  indicated diameter,  For the space charge case,  the particles
were assumed  to be charged as above, monodisperse at the size indicated,
and having a concentration of  lg/m3.  The migration velocities calculated in

-------
                                                                                 271
this way ranged from  ~10  cm/s  to  ~10 cm/s  for particle diameters
from 0. 1 to 3 |_im.  In decreasing order the velocities were as follows
(at 1 urn diameter):  charged particle and  charged  collector (coulomb
attraction), charged collector and  uncharged particle, charged particle
and uncharged collector,  space charge repulsion.
              O
                 10s
                 I02
                  I01
              E
              u


              t   10°
               O
               I
              i   '°" \ x
                   -'
                  10
                  1C'4
                               "---..X
                                      /
^
 •  I COULOMB FORCE-CHARGED PARTICLE IN
               A FIELD.
 A  2 CHARGED PARTICLE WITH UNCHARGED
     COLLECTOR.
 D  3. IMAGE FORCE-CHARGED COLLECTOR WITH
             UNCHARGED PARTICLE.
 *  4. CHARGED PARTICLE IN SPACE CHARGE FIELD
	I	1
                               3            10         30
                                   PARTICLE DIAMETER ,  Jim
 Fig.  2  Migration velocities  (particle velocity at collector surface  wg)
          calculated under conditions indicated in text.

-------
  272


    E.  Other Fields Superimposed

        To a first approximation,  the addition of a gravitational field should
 produce an increase in collection rate due to a gravitational collection rate
    Ng =  n Ahvg= ^h8                                              (39)

 the product  of the mean concentration, the horizontal area of the  collectors
 and duct,  and  the gravitational settling velocity of the particles.  The situa-
 tion is actually more complicated, especially where the flow is either pre-
 dominantly upward  or downward; gravitation can then act as  though it changes
 VQ, allowing the particles more or less time to be in the vicinity of collectors.

    An external electric field will not only produce a net migration velocity
 equal  to the  product of the particle charge  (q ), the field (E) and  the particle
 mobility (B), but the field will also polarize tfie collectors, which will in
 turn modify  the field in their vicinity.  Zebel [10]  gives the following equa-
 tion for the radial component of field in the vicinity of a cylinder  of dielectric
 constant  e  placed in a  field perpendicular to  its axis,  a field  having a
 strength Ec:


                      ~)  EQ cos 9                                  (40)
                      r
The angle 0  is measured from the direction of E .  The directional nature
of the field produced means that the integration   °

      2TT
   •

      J    s D  r    c                                                  \^^/
      o

is complicated.  To a first approximation,  the  collection efficiency should be
between what it would be for a field of strength Eo alone (integrating only
over the region where  q E  cos  0  < 0)


       o       P

and twice this, what it would be for a field of strength 2E ,  the maximum
value of E ,  achieved where the collectors are conductors,  e  -»°°.
           r                                                 c
    Nielsen [8] studied the combination of coulombic attraction, gravitation,
and external electric field for the collection of inertialess particles on
circular cylinders.  He found the collection efficiency to be the same
"under many conditions for a wide  variety of flow profiles, which includes
potential, Oseen,  and stationary-vortex flows." As one expects  from the
constancy of particle concentration along trajectories, Nielsen found the
deposition density on the cylinder to be uniform for the  case  of coulombic
attraction alone.   Combining coulombic and external electrical fields did not

-------
                                                                          273

necessarily improve  collection.  Addition of gravitation and an external
electric field gave maximum collection when they were perpendicular and
minimum when they were aligned.

    If collectors are inserted into a region having an electric field, the
presence of the collectors will alter that field.  If the field is maintained by
applying a constant voltage difference,  for example across parallel plates,
then the field in the gaps amid  the collectors will be larger than it would be
without the collectors.  If the voltage difference V is applied across two
parallel plates or grids, the spacing of which is H,  then the electric field
will be V/H if there are no collectors between the plates.  If cylindrical
.collectors are-placed between the grids, with the collector  axes  parallel
to the plane of the grids,  then the electric field in the gaps amid the  col-
lectors becomes [11]:

    E  = (V/H)/(l-a/4vT)

in which expression a is the cylindrical collector volume fraction.  Thus
the field is increased in the gaps as the volume fraction of collectors
increases.

    If,  on the  other hand, the external field is due to fixed charges (rather
than fixed voltages),  the  effect becomes the opposite:  the more collector
volume between the charges creating the field, the  less field in the gaps
amid the collectors.  This shielding is similar to that of a "Faraday cage
and can be quite nearly complete when the collectors are conductive.

   -F.  Cooperative Effects: Many Particles

    If the particles are all of like sign, they will repel each other, a
 phenomenon called "electrostatic scattering" or  "space  charge repulsion.
 It can be shown that the force on an individual particle near a  conductive
 collector is just the same as  coulombic attraction would  be if the collector
 had charge located at its center  equal to the product of the collector volume
 and the aerosol charge per unit volume. Thus, the efficiencies of collection
 for spheres and cylinders are 4K and TiK, respectively.

     If there are only charged particles in a  circular  duct, and the particles
 are all of the same sign in charge,  then the electric field at the inner
 surface of the duct is given by the volume integral

     E = [ n>q dV]/2TTR,L                                             (43)

     E = — R  /2                                                      (44)
           P  d
 where q   is  the mean particle charge and R^ the duct radius. The
 depositiBn rate at the inner surface of the  duct will be

     N = nq~~BEA/2 = nq B~Sq  2nR,L  /2                               (45)
          ^p     d       p   p    d d

-------
 274

 or

     N =n qp B2nRd Ld/2                                            (46)


     The presence of particles of opposite charge,  collectors of opposite
 charge,  or even collectors which are uncharged will all serve to lessen
 collection due to space charge repulsion at the duct  inner surface.  Correct
 calculation of the space charge  field within a filter or  scrubber requires
 taking into account in  detail the geometry,  particle  charge concentration,
 collector volume fraction, collector charge,  and collector dielectric constant.
 This effect may be important when, for example,  non-conductive fibers
 collect unipolarly  charged aerosol particles.

     If the particles are of equally positive and negative charge,  then
 they will coagulate more readily, but the net field at the inner surface  of
 the  duct will have a zero contribution from particle charge.

   G.  Cooperative Effects:  Many Collectors

     In general, if there are many collectors (droplets, fibers, etc. ), the
 collection efficiency of the assembly is

     E = 1-exp ( WA/Q)                                                (47)

 where

     WA =  £(w A  ).
           ^   s si


 that is, WA is the  sum of the products of the  collection areas and the
 migration velocities.   Q is the volume flow rate through the collector.  The
 presence of other collectors will change somewhat the  gas flow in the
 vicinity of a given collector, so  that its collection  efficiency will be some-
 what different from that calculated assuming potential  or viscous flow.  As
 we have  seen,  the collection efficiencies for electrostatic effects have not
 been strongly influenced by details of the flow profile.  If the collectors are
 charged to the same sign, this will enhance the space charge repulsion
 effect.  If they are of mixed charge, that will decrease space  charge but
 enhance collection  on the collectors,  because the electrical fields will be
 somewhat greater than those calculated by considering the isolated
 collector..

   H.  Cooperative  Effects:  Loss of Collectors

   In filters, the collection surfaces are held fixed with respect to the  gas
 stream to be cleaned.  In scrubbers, both collectors and particles move
with the gas.  Melcher and Sachar [12] analyzed many different charged
particle-charged droplet configurations.  The rates of  change  of con-
 centration of particles  and collecting droplets were put in terms of
characteristic  times {times for approximately 1/e decrease in concentration).

-------
                                                                           275


For particles,

    t  = 'l/4nq2Bn                                                   <«)
    cL
is appropriate for particle neutralization due to agglomeration in a bi-polar
aerosol or for particle concentration decrease, due to electrostatic
scattering,  for a unipolar aerosol.   For droplet neutralization (if bipolar)
or scattering (if unipolar) a  similar time can be calculated:

    tR = 1/4, Q2Bdnd                                                (50)

where the subscript d indicates these are the mobility and number  con-
centration of the droplets. In order for the droplet charge or concentration
not to be greatly diminished before collection is completed, tR should be
large in comparison to the time characteristic of coulombic collection
    t  =  l/4TTqQBNd.

Similarly, if the charge on the particles is bi-polar, then for nearly complete
collection,  ta must be greater than t ,  which in turn must be much greater
than the residence time of the particles in the collection zone.
 CONCLUSION
    It has been shown for inertialess particles undergoing collection due to
 central forces that the collection efficiency for a single spherical collector
 is (neglecting interception):

    n=  (4K)2/m                                                     (52)
 for K » 1.  where K is the ratio of the particle terminal velocity at the
 collector surface, due to a force which varies radially as 1/r  ,  and v  i
 the free -stream velocity.   The corresponding equation for a cylindrical
     n^K)17"1                                                      (53)

     When K < < 1,  the approximate collection efficiencies are

     „ '• «                                                            (»«

 for a spherical collector and

     T, = TTK                                                          <55>

 for a cylindrical collector.  These approximation^ should hold for any flow
 profile and any central force of the form F  - k/r  .

-------
  276


  REFERENCES

  [1]    Fuchs, N A.,  Mechanics of Aerosols. Pergamon Press,  New York
        and London (1964).

  [2]    Whitby,  K T. and Liu, B.Y.H. ,  "The Electrical Behaviour of Aero-
        sols", Aerosol Science. C.N. Davies, ed. ,  Academic Press, New
        York and London  (1966).

  [3]    Kraemer,  H.F. and Johnstone, H.F. "Collection of Aerosol Particles
        in Presence  of Electrostatic Fields", Indus. Eng.  Chem.  47_,  2476
        (1955). Correction in Indus.  Eng. Chem. . 48,  812 (1956).

  [4]    Nielsen, K.A.  and Hill, J.C.,  "Collection on Spheres with Electrical
        Forces", Presentedat the Sixty-Eighth Annual Meeting,   A.I.Ch.E.,
        Los Angeles, November 1975.

  [5]    Natanson,  G. ,  "Deposition of Aerosols by Electrostatic Attraction
        Upon a Cylinder Around Which They are Flowing", Dokl.  Akad,  Nauk. USSR
        11£,  696-699, (1957).
  [6]   Lundgren, .D. A. and Whitby,  K.T., "Effect of Particle Electrostatic
       Charge on Filtration by Fibrous Filters",  Indus. Eng.  Chem. Proc.
       Design fc  Dev. .  4_,  345-349 (1965).

  [7]   Yoshioka, N., et al. ,  Chem. Eng.  (Japan), 33., 1013 (1969).

  [8]   Nielsen, K.A.,  "Collection of Inertialess  Particles on Circular
       Cylinders with Electrical Forces and Gravitation", Presented at the
       Sixty-Ninth Annual  Meeting,  A.I.Ch.E., Chicago,  November 1976.

  [9]   Grover, S.N. and Beard, K.V.,  "A Numerical Determination of the
       Efficiency with Which Electrically Charged Cloud Drops and Small
       Raindrops Collide with Electrically Charged Spherical Particles of
       Various Densities", J. Atmos. Sci. . ^£;2156-2165 (1975).

[10]    Zebel, G. , "Deposition of Aerosol Flowing Past a Cylindrical Fibre
       in a Uniform Electric Field", J.  Colloid Sci. .  20.  522  (1965).

[11]   Havlicek,  V., "The Improvement of the Efficiency of Fibrous Die-
       lectric Filters by Application of an External Electric Field", Int.
       J. Air & Water Pollut. 4, 225-236 (J961).

]12]   Melcher,  J.R. and  Sachar,  K S. , "Charged Droplet Scrubbing of
       Submicron Particulate",  Draft Final Report to EPA for Contract
       No. 68-02-0250, July 1974.

[13]    Cooper, D. W. , "Fine Particle Control by  Electrostatic Augmentation
       of Existing Methods, " Presented at the Sixty-Eighth Annual Meeting,
       A.P.C.A., Boston, June 1975.

-------
                                                                              277
ACKNOWLEDGEMENT

Support by the National Institutes of Health through NIH Grant 5-S07-RR-
05446-15 is acknowledged.
WRITTEN DISCUSSION

Kenneth Andrew Nielsen
Union Carbide Corporation
South Charleston, West Virginia  25303

   The particle flux method used by Dr. Cooper, which was developed
independently by Chalmers  [14] and Dukhin and Deryagin [15], and the stream
function method for determining particle trajectories [4, 8, 16, 17], which
was  first used by Whipple  and Chalmers [18], are two powerful techniques for
analyzing particle collection behavior when particle inertia can be ignored.
Both methods can only be applied, however, to solenoidal particle fields
resulting from incompressible flows and solenoidal forces.  Such forces include
the  gravitational force and coulombic and external electric field forces on a
charged particle.  These electrical forces are given by QpE, where  Qp  is
the  particle charge and  E  is the electric field intensity present in the
absence of the charged particle and, from Maxwell's equations, E is solenoidal.
If the flow, force, and hence particle fields are solenoidal, then for two-
dimensional systems stream functions for the fields exist.  The stream function
for  particle trajectories, which provides a complete description of particle
motion, is then given by superposition of the flow and force stream functions.

   In the particle flux method, the solenoidal property of the particle field
makes it possible to calculate the particle flux at the collector surface
without having to solve for the particle concentration profile in the flow,
because particle concentrations are constant along the particle trajectories.
This gives limited but useful information about the particle motion, namely,
the  collection efficiency.  The most useful application of this method is in
obtaining results for three-dimensional systems, for which the stream function
method can not be used.  Whenever both attractive and repulsive forces are
present at the surface of  the collector (such as for a combination of coulombic
and  external electric field forces), however, the method must be applied with
caution to avoid erroneous results.  From just a knowledge of the force
distribution at the collector, it is not possible to distinguish between the
flux of particles coming from upstream and the flux of particles on closed
trajectories originating elsewhere on the collector.  This determination
requires a knowledge of the flow field and the types of limiting trajectories
that occur.  Although the method was originally proposed for point particles,
it can also be applied to  collection with interception, using the force

-------
278

distribution at the surface surrounding the collector that is defined by
the centers of particles at capture [17].

   The results found by Dr. Cooper for the coulombic force case, given by
collection efficiencies of -4Kq  for spheres and  -irKqq  for cylinders, where
Kqq  is the coulombic force parameter (negative for attraction), are remark-
able in their generality. Furthermore, the result for spheres has been
verified experimentally [3] for measurements over a wide range of parameter
values (10""-* < -Kqq < 10).  That these results hold for all steady incom-
pressible flows was shown by Nielsen and Hill [4,16] for spheres and Natanson
[5] and Nielsen [8] for circular cylinders.  That both results are insensitive
to both collector shape and flow profile was shown by Levin [18] and Nielsen
[17].  In addition, Nielsen also noted that the rate of particle collection
on fibers should be unaffected by accumulation of previously captured particles
(provided net fiber charge is not significantly affected).  Coulombic force
influence on particle dendrite formation was also examined.

   Using both the stream function and particle flux methods, Nielsen [8, 17]
has obtained particle trajectories and collection efficiencies for the
external electric field force on a charged particle for cases in which the
force acts alone or in combination with the coulombic force and gravitation.
Collector geometries used are spheres, prolate and oblate spheroids, disks,
circular and elliptical cylinders, ribbons, and collectors of arbitrary shape
as well.  Flow fields considered are potential, Stokes, Oseen, stationary-
vortex, cellular, and arbitrary steady incompressible flows.  Results were
found for arbitrary external electric field, flow and collector orientations
and include the effects of particle interception. In general, the results
show that for any given collector shape and given flow and field orientations,
for any combination of the forces the collection efficiency is the same for
all steady incompressible flows when interception effects are negligible.  In
cases in which the coulombic force is sufficiently dominant, the collection
efficiency is given by -4Kqq/Vo  for sphere-type collectors and  -irKQq/Vo
for cylinder-type collectors independent of the collector shape, of the flow
field, and of the orientations of the collector and external field.
VQ = (1 + 2Kexcos9+ Kex2)1/2 is the ratio of the particle approach speed
far upstream to the free-stream velocity, where  Kgx is the external electric
field force parameter (see Table 3) and 9  is the angle between the flow and
external field.  For the external electric field force alone, the efficiency
is given by  nmaxKex/vo> where  Hmax* the collection efficiency for infinite
field strength, is a function only of collector geometry, dielectric constant,
and orientation of the collector with respect to the external field.

   The problem of calculating collection efficiencies for nonsolenoidal
electrical image forces is much more difficult than for the coulombic force
case since the particle concentration profile is not uniform and changes with
flow field.  When analytical descriptions of a flow field are available it is
a simple task to obtain efficiencies via numerical integration of particle
trajectories, but the simple flow fields normally used are not very represen-
tative of real flows and, hence, the results are often less than satisfying.
Dr. Cooper's approximate analyses represent the first attempt to use the

-------
                                                                             279

particle flux method to obtain results that reflect an "average" influence
of flow field.  Indeed, if such results were available for the whole range
of image force strengths they would be very useful.

   In Table 1, Dr. Cooper has assembled results of previous studies for
comparison with his high and low efficiency asymptotic results.  Unfortunately,
Table 1 contains several inconsistencies, misinterpretations, and omissions
of pertinent results.  The apparent difference in results obtained by Kraemer
and Johnstone [3] and Natanson [5] for force  FQQ  for cylinders comes from
parameter definitions that differ by a factor of two.  For force  FQq  for
spheres, Nielsen [4] and Kraemer and Johnstone obtained identical results for
the viscous flow case and in their analyses used a force expression that
goes to 2r~5  for large r; hence their parameter definitions differ from
Cooper's by a factor of two.  Natanson1s [5] results for force  FQQ  for
cylinders were derived for the low and not high efficiency case.  In order to
avoid confusion for the reader and to include additional results,a revised
version of Table 1 is given in Table 2, with results reported on the basis of
the electrical force parameter definitions appearing in Table 3, which have
been used by Kraemer and Johnstone and Nielsen.

   Compared to the coulombic force, the electrical image force (FQQ) between
a neutral particle and a charged collector is a weak, short range force.
When both forces occur, the image force can be neglected (for spherical
collectors) if the particle-to-collector charge ratio is much greater than
the volume ratio.  However, for droplet collectors charged to near the
Rayleigh limit by electrical atomization the image force can be important,
but efficiencies of micron sized particles are much less than needed for high
efficiency results to be valid.  As indicated in Table 2, a number of
investigators have obtained results for the charged-collector image force
case.  In addition, for intermediate parameter values for spheres, results
which do not follow a power law are available for potential and Stokes flows
[3, 2, 16].  For the high efficiency results (K»l), the limiting trajectory
lies far from the collector, where flow fields all degenerate to uniform flow.
Hence these results are independent of flow field and, similarly, independent
of collector geometry.  The low efficiencyresults (K«l), which agree with
those of Dr. Cooper, are said to be valid for all flows, but this results
from ignoring radial variations of the force near the collector and hence
making the particle field artifically solenoidal.  In the exact case results
should vary with velocity gradient at the surface; that Kraemer and Johnstone's
numerical results for potential and Stokes flows do not merge even at very
low parameter values supports this.

   The electrical image force (FQ ) between a charged particle and a neutral
collector has the unique property of becoming infinite at the collector sur-
face, although it drops off very rapidly with distance.  It is negligible
compared to coulombic interaction (for spheres) when particle charge is much
less than collector charge.  For micron sized particles collected on droplets,
parameter values are typically much less than one.  However, the force will
always exert some influence on trajectories passing very near the collector
surface.  The high efficiency limit results are independent of flow field,

-------
                       Table 2.   Comparison of Collection Efficiency Results
Parameter Cooper's Other Studies
to! lector horce r Dependence* Range Results Efficiency
Sphere FQq r"2 all -4K -4K
F ~~5 ix >> i /Aif\2/5 /15npl/\2/5
'00 ' * -1 I™; \ 3 K;
K « 1 4K 4K
rl 9 /R 1 R 9 /K
r U *•»••» 1 /OI/\^'3 / •!• ^1» I/ \ ^ / ^
Oq 2 \2 " 3 * ' ^ 4 ^J
,- .002iK5.1 - (20. 2K)'353
= 2r~
.002£K£.l - (2.50K)1/2
for r » 1
K« 1 8K (^K)17-*
(2K)l/2
Cylinder FQQ r"1 all -nK -i^K
F¥»~ I/ ^^ 1 /_ I/ \ ^/ ** / *5 AlX \ I/ ^
OQ ' i\^^i \T^<^/ \^r^/
.01 11 fc I/ ">•> 1 f*eV\i/c- f/ll/\*-tt-
i-Qq ir-i; K ^>^i ^K; ^K;
= r"2 - - (5.29K)1/2
,forr»l K < .03 - (2.25K)1/2
.001
-------
Table 3.  Electrical Force Parameter Definitions (K)
                                                                             281
                  SPHERE                       CYLINDER
                         Coulombic Force (F_ )*
                  Charged-Collector Image Force
                                                    . **
                        o c   o
Charged-Particle Image Force (FQ )
        2
n /]	if DD^II       \ • f-   f /
                                                      CQr
                      External Electric Field Force
                       (all collector geometries)
Q    collector charge
 w
Qn   particle charge
                               collector radius
                               particle radius
p   collector charge per unit length        UQ   free stream velocity
      fluid dielectric constant
      particle dielectric constant
      collector dielectric constant
 C     Cunningham-Millikan  slip correction  factor
                          M.   viscosity
                               uniform external electric
                               field intensity
 *  For  collectors  of  different  geometry  R  is  an  arbitrary  collector  reference
   length  (an  equivalent  radius,  for  instance)  upon which the  definition  of
   collection  efficiency  depends.
 ** For cylinders  multiply  by 2 to  get Kraemer and Johnstone's definition.

-------
282
  but not necessarily independent of collector shape.   As noted by Dr.  Cooper,
  it is not valid to use the large distance force approximation near the
  collector.  The literature low efficiency results reflect the actual  force
  near the surface but, as Davies [23]  has pointed out,  approximations  made in
  their derivations could produce serious limitations  on their usefulness.
  Nevertheless, for cylinders the one half power law for viscous flow,  which
  appears valid over the entire range of parameter values, has received experi-
  mental [6,7] and numerical [22] confirmation.   For spheres the results for
  Stokes flow show little deviation from a one half power law [3,4].

     When particle inertia is present,  second-order effects can create  very
  complex particle motions,  and analytical solutions are generally not  avail-
  able.   For spherical collectors,  however,  collection efficiencies for the
  combined influences of particle inertia and  electrical forces have been ob-
  tained by Nielsen and Hill [4,  23]  through numerical integration of particle
  trajectories using potential and  Stokes flows.   The  electrical forces con-
  sidered are the coulombic  force,  both electrical image forces,  and  external
  electric field forces, with the external field  parallel to the flow.  Results
  for perpendicular fields and for  prolate spheroidal  collectors are  given  by
  Mukherjee and Hill [24].  These results show that characterizations made  on
  the basis of inertialess analyses (such as flow independence of coulombic
  force  collection) can not  be extended into the  inertial realm.   Although
  maximum collection is favored by  large particle inertia when electrical forces
  are absent, collection with electrical forces often  reaches  a maximum for
  negligible particle inertia.  In  general,  the steeper  the velocity  gradient
  at the collector, the greater the effect inertia has.   The statement  by
  Dr.  Cooper on the additivity of purely inertial and  purely electrical collec-
  tion efficiencies (Equation 38) simply is  not borne  out by these studies
  even in approximate form.

     Although analyses of particle  collection  on  single,  isolated droplets  and
  fibers are very similar, the  dynamic  characteristics of droplets in wet
  scrubbers and fibers in filters are very different.  In wet  scrubbers,  drop-
  lets follow the bulk flow  and only affect  a  localized  region.   Shadow or
  distortion effects caused  by  the  particle  field interacting  successively  with
  several droplets is minimal,  and  isolated  collector  results  may be  a  good
  approximation.   Whereas, in filters fibers continually sweep the bulk flow,
  and  upstream fibers can cast  particle shadows on downstream  fibers  or  other-
  wise distort the particle  field such  that  simultaneous  analysis  of  multiple
  fibers is really required.  This  is a difficult task,  however,  for  which  new
  insights  into particle collection behavior,  such as  found  in Dr.  Cooper's
  approximate analyses,  may  be  required.

  REFERENCES

  [14]  Chalmers, J. A., "The Capture  of Ions  by  Ice Particles," Quant. J.
        Royal Meteorol. Soc.. ^3, 324  (1947).

  [15]  Dukhin, S. S. and Deryagin, B.  V., "Method of  Calculating the Precipi
        tation of Disperse Particles from a Stream on  an Obstacle (Russ.)",
        Kolloid Zh.. 20, 326 (1958).

-------
                                                                             283
[16]   Nielsen,  K.  A.  and  Hill,  J.C.,  "Collection of  Inertialess  Particles
      on Spheres with Electrical Forces,"   Indus.  Eng.  Chem  Fundamentals',
      15,  149 (1976).

[17]   Nielsen,  K.  A., "Turbulent Mixing with Chemical Reaction and  Collection
      of Inertialess  Particles  on Cylinders and  Spheroids  with Electrical
      Forces and Gravitation,"  Ph.D.  Thesis, Iowa State University,  Ames,
      Iowa,  1977.

[18]   Levin, L. M.,  "Electrostatic Precipitation of  Aerosol  Particles  from a
      Flow Upon Large Bodies,"   Izv.  Akad.  Nauk  SSSR (Ser. Geofiz.), ]_,
      1073,  (1959).

[19]   Smirnov,  L.  P.  and  Deryagin, B. V.,  "Inertialess  Electrostatic Deposi-
      tion of Aerosol Particles on a  Sphere Surrounded  by  a  Viscous Stream
      (Russ.)," Kolloid Zh.. 2£, 400  (1967).

[20]   Cochet, R.,  "Evolution d'une Gouttelette d'eau Chargee dans un Nuage a
      Temperature  Positive," Annales de Geophysique, 8^, 33  (1952).

[21]   Knutson,  E.  0., "Approximate Formulas for  Electrostatic Collection of
      Aerosol Particles by Spheroids," J.  Colloid Interface  Sci., 54,  453,
      (1976).

[22]   Stenhouse, J.  I. T., "The Influence of Electrostatic Forces in Fibrous
      Filtration," Filtration & Separation. 11,  25 (1974).

[23]   Davies, C. N.,  Air  Filtration,  Academic Press, New York, 1973, p.  85.

[24]   Mukherjee, S.,  Nielsen, K. A.,  and Hill, J. C., "Effects of Electrical
      Forces on Particle  Collection by Spheres and Spheroids,"  Presented  at
      the Ninth Annual Meeting  of the Fine Particle Society, Menlo  Park,
      California,  August  25-26, 1977.

-------
284
 OPEN DISCUSSION

 Friedlander:  We have heard a criticism with high gradient magnetic separation
 theory.   It may be interesting to point out that there is no central force
 in the HGMS theory and that complicates matters tremendously.   This is
 very clearly demonstrated.   Calculations can be made rather readily in  the
 case of  the electric field.  In the case of the magnetic  field,  there are
 always dipolar interactions which never have central forces.   That  is what
 makes matters so much worse.  On the other hand, in the situations  so far
 reported,  the collection agent is usually stationary.   That will change as
 one goes to gaseous streams.

 Cooper:  A quick comment.   One of the ways you  can pursue it,  at least  to
 an approximate level is  to  find the directional force  and the  area  of the
 particle collectors which indeed collect.   You  then have  some  idea  of what
 the velocity is on the surface of that area.  You have to make an assumption
 about the  concentration  profile.

 Loffler:   I agree completely with your remarks  and that you expect  from
 theory to  get some ideas on the most important  inputs  and so on.  That  is
 the same point of our theoretical work.

      There is one point  on  your assumptions and simplifications  that  may  cause
 a  problem.   I agree as long as particles are below .1  ym  but as  I have  shown
 earlier  they may range between .1 and  1 ym. You should think of  the
 supposition you used  to  simplify your  equation.  Another  thing is that  all
 these comments also refer to the interception effect,  and  I must  insist that
 interception is not an impact  force  mechanism.   Interception is only  the  con-
 tact  between particle and collector.   You might  speak  of  interception if  the
 mass  densities of  particles  in the fluid are equivalent. Then, you have some
 effect, but it is  not  an impact  force  mechanism.   Sometimes it is important,
 especially  in your case  where  the particle  is collected in  the rear.

 Melcher:  On the  question of impaction and  interception,  the paper by Brogan
 &  Beer doing for  the  spherical case  what Professor Loffler has done for the
 cylindrical  case,  may  be of  interest.  Essentially you want to take collection
 efficiency  as  a  function of velocity.  For  fairly weak electrical forces
 you are going  to have  decreasing collection efficiency.  Then you will have
 the typical  impaction  curve and you will get a combined thing.  It will come
 off about 1  or  a  little  higher  if you  want  to take interception into account.
 If you have  a very  strong electrical force, you will eventually have the
 collection  efficiency  asymptotically approach the impaction collection
 efficiency.   It is a little misleading, however, because if you look at
 the collection rate, it  turns out to be equal to the concentration on the
 surface times  the velocity at  the surface times collecting area of the sur-
 face plust  the impaction contribution.

     The rate of collection will remain almost independent of velocity.
There is a Reynolds number correction or if we get too high, there is a problem

-------
                                                                              285
of collecting particles with   great inertia.  The last thing we have is the
freestream concentration which will not change,  The velocity will go up.
As it turns out, with the same collectors and volume flow, you want to
arrange the geometry that as long as you can handle the pressure drop, the
velocity is high.  That suggests that going from a stubby collection area to
one that is small and long, both having the same volume flow.  Indeed then
you can get the impaction.

     The other point is that, after a while, the particles get so massive
that you just cannot pull them in with electrical forces.   Impaction does
work well.

Liu:  I understand your assumption, the divergence of the force will be equal
to zero, which leads to the conclusion that the concentration of particles
along the trajectory remains constant.

Cooper:  That is for the first part of the paper where I deal with special
cases of Coulombic force, gravitational force and uniform electric field. We
stopped that development at that point, then took away that assumption and
dealt with central forces in general.   The rest of the paper deals with
a generalized central force.

Liu:  The first part of your paper would also be applicable to the magnetic
situation,  because you have a uniform particle concentration along the
trajectory even for the case of dipole magnetic particles.  Because the
divergence of the field will be zero.

-------
286
 ELECTROSTATIC FILTRATION AND THE APITRON/
 DESIGN AND FIELD PERFORMANCE

    Dennis J. Helfritch

    American Precision Industries Inc.

    and

    Teoman Ariman

    Notre Dame University


 ABSTRACT

    The electrostatic augmentation of fabric filtration has been examined
 for many years.  Most often these studies have been directed toward the
 effects of electrostatics on the efficiency of particle collection,  and it has
 generally been shown that efficiency increases with the addition of electro-
 statics.  It is  shown in this  paper that electrostatic augmentation also
 gives  rise to substantial increases in filtration rate per unit fabric area.
 An industrial fabric filter utilizing electrostatics is described, and its
 performance is demonstrated on several typical applications.

 INTRODUCTION

    The effects of electrostatics on the fabric filtration process have been
 investigated and discussed for decades.   In 1930, N. J. Hansen discovered
 that the filtering efficiency of a wool filter pad was much improved when
 powdered with ground resin because the resin possessed a substantial
 static charge  (1).   This discovery was used to improve the particulate
 collection efficiency of military gas  masks.  Since that time, many studies
 have been made concerning  the effects of charged particles,  charged
 fibers, externally generated electrostatic fields, or combinations of these
 factors.

    Frederick (2) studied the interactions between naturally charged dusts

-------
                                                                           287
and fabrics.   Lundgren and Whitby (3) investigated the effect of particle
charge on the filtration efficiency of uncharged fibers.  Walkenhorst and
Zebel (4) applied an electrostatic field across a fabric filter.  Inculet and
Castle (5) describe a device in which particles are charged and a field is
applied across the filter.  The work of Lundgren and Whitby  (3),  shown in
Figure  1, illustrates the magnitude of the influence of electrostatics on
fabric filter collection efficiency.
                             PARTICLE
                              CHARGE
                          10          30         100
                     FILTRATION  RATE, CFM/SQ.FT.
300
         Figure (1) - Particulate penetration through a filter as
                     a function of  particle charge and filtration
                     rate (from Ref.  3)

   Mathematical models  for the efficiency of fabric filters have  been pre-
sented by several investigators.  Natanson (6) gives equations for the
collision efficiency between a charged particle and a dielectric fiber.
Zebel adds an  external electrostatic field to the same problem and presents
solutions to the governing equations of particle motion. Ariman and Tang
(8) extend Zebel's  results by including the effects of neighboring fibers.

-------
288
     Virtually all of these investigations, both experimental and theoretical,
  indicate that a substantial increase in collection efficiency can be realized
  when particles are charged or fields applied.  This increased collection
  efficiency is the result of induced polarity of the fibers or particles,  which
  effectively increases the collision cross section between a fiber and an
  approaching particle.   Experimentally, this  effect appears to obey  the
  equation (1, 3)

          fl = single fiber collision efficiency = OC (K)^

          OC = 2. 3  (ref.   1)         K = image force parameter
              1.5  (ref.   3)
                                     = ( gc - D/( £c +  DQ2

        Cc  = dielectric constant fiber            /A  = gas viscosity
       Qp  = particle charge                    Dp  = particle diameter
       D£  = fiber diameter                    *yo  - gas velocity

  It is obvious from the  results so far reported that electrostatic augmenta-
  tion of fabric filters yields improvement in collection efficiency, but very
  little information concerning the effects of electrostatics on filter resis-
  tance is available.  It  will be shown in this paper that electrostatics can
  have a profound effect on the flow resistance of fabric filters.  For example,
  it will be shown that by simply charging particles, the filtration rate per
  unit fabric area of a given fabric filter can be increased by a factor  of four.
  This implies that with the aid of electrostatics it might reasonably be
  expected that fabric filters can  be reduced in size by a factor of four,
  while achieving better collection efficiency.

  BASIC STUDIES

     The results  of recent bench scale tests  can best be used to illustrate the
  variation of filter resistance with particle  charge.  At constant face velo-
  city, suspended dusts  were filtered by initially clean fabrics.  The dusts
  were charged prior to deposition by means of passage  through a corona
  discharge.   Degree of charge was controlled by the magnitude of the field
  strength, and pressure drop was recorded 'after a specified weight of dust
  had been deposited.

     Filter resistance, or pressure drop, decreases as the degree of particle

-------
                                                                         289
charging increases, as is shown in Figure 2.
                                      NOMEX

                                    15 FPM   FACE VELOCITY

                                      A LIMESTONE
                                      D SILICA
                            2   4  6   8  10  12  14   16
                           VOLTAGE  GRADIENT, KV/IN.
              Figure (2) -  Pressure drop vs. charging voltage
                           gradient for the deposition of lime-
                           stone and silica dusts on -Nomex felt

Each point of this figure represents the pressure drop resulting from the
deposition of 25 grams of dust at a 15 foot per minute face velocity.  In
particular, for. the silica dust, it can be seen that filter resistance is
decreased by a factor of 8.   Although there are quantitative differences in
the magnitude of the effect of charging dusts  prior to filtration, the effect
is qualitatively identical for all dusts tested.

Fabric type exerts some influence on the electrostatic filtration process,
but the relative  differences among fabrics are small, as is shown in Figure
3.   Charging dust particles always results in a substantial reduction of
pressure drop,  regardless  of fabric type.

-------
290
                                     LIMESTONE

                                   15 FPM  FACE VELOCITY

                                     O ORLON

                                     D POLYESTER

                                     A NOMEX
                          0  2  4  6  8   10  12   14  16
                           VOLTAGE  GRADIENT, KVA/IN.
               Figure (3) - Pressure drop vs. charging voltage
                           gradient for the  deposition of lime-
                           stone dust on three felted fabrics
   The variation of filter resistance with face velocity is shown in Figure
4.  Each point of this figure represents the pressure drop resulting from
the deposition of 25 grams of dust at the face velocity given.  Pressure
drop increases with face velocity as expected,  and it can be seen that
particle charging results in a two-fold increase in face velocity at an equiv-
alent pressure drop.

-------
                                                                       291
                o.
                o
                (£.
                O
  10

   9

   8

   7

-  6
                         LIMESTONE/NOMEX

                            O- 0  KV/IN.
                            A — 7  KV/IN.
                            o-IO  KV/IN.
                            •~I3  KV/IN.
                            4   6  8  10  12  14  16  18
                          FACE  VELOCITY, FPM
              Figure (4) - Pressure drop vs. deposition face
                          velocity for limestone dust on Nomex
                          felt as a function of charging voltage
                          gradient

FILTER DESIGN

   In order to apply electrostatics to the fabric filtration process, efforts
were initiated in 1971 for the design of a practical device.  Various con-
figurations were tried,  and the geometry of Figure 5 was ultimately
chosen as most acceptable.  This device is designated the electrostatic
fabric filter or "Apitron".  During normal  operation, incoming airflow
enters the  precipitator  section in the area below the  walkway.  It is
deflected downward and distributed among the tubes by means of an
inclined, expanded metal grating.  The flow then passes upward, through

-------
292
  the tubes, where the particulate is charged and where most is  deposited.
  Flow continues upward past the tubes, into and through  the bags, where
  the final filtration of the charged particulate takes place.  Clean air exits
  the unit at the  exhaust,  located at the top of the baghousing section.
                         BAG HOOKS ABOVE
                                              11
           CORONA WIRE HOOK
           ON H.V. G^ID. 5.PRIN
           TENSIONED  8,
           INSULATED  AT TOP
INSULATOR
 . /CORONA /
 \  WIRt   /

"\    /
   \    ™
                IGH VOLTAGE
                TO GRID
            Figure (5) - The Apitron industrial scale electro-
                         statically augmented fabric filter

    Periodically, the deposit of particulate is cleaned from the tube  and
 fabric.  Six bags and tubes are cleaned of deposited dust at one time.  This
 cleaning is initiated when an electrical pulse from the  control box opens one

-------
                                                                        293
of the1 twelve diaphragm valves for 1/10 second.  The blow pipe connected
to the valve is then pressurized and compressed air jets downward from
six nozzles, each directly above a tube and concentric to a corona wire.
The jet of air flowing downward through the tube entrains and mixes with
a secondary airflow,  and the tube is swept clean of deposited dust by the
mixture of high velocity air.  The secondary airflow, passing from the out-
side  to the inside of the bag, snaps the bag inward and dislodges dust
deposit from it.  Typically, a single bag and tube would be cleaned in this
manner once every five minutes.

PERFORMANCE

   When no high voltage is applied, the fabric filter behaves as a conven-
tional continuous cleaning, pulse type  baghouse.  As voltage is applied,
filter resistance decreases as shown in Figure 6.
                          10  15  20  25 30
                                KILOVOLTS
35  40  45  50
             Figure (6) - Pressure drop vs. charging voltage
                         for the Apitron collecting silica dust

-------
294
  This decrease in resistance continues as voltage is increased because
  particles become more highly charged and because more particulate is
  deposited on the metal tube.  This behavior is similar to that shown in
  Figure 2 for the bench scale tests.

     Figure 7 shows a recorder tracing of pressure drop during the collec-
  tion of silica at a 15 CFM/Sq.Ft.  filtration rate.
                                20 30  40
                                MINUTES
                   Figure (7) -  Pressure drop vs. time,
                               showing the effect of a
                               high voltage shut down

  The applied voltage was shut down at the 11 minute point and restored at
  23 minutes.  Under normal conditions, with applied voltage and particle
  charging,  the pressure drop was 1 inch W. G.  The pressure drop increased
  to 6 inches W.G. after the voltage was shut down and returned to 1 inch
  following high voltage  restoration.  Discontinuities in the tracing were
  caused by bag cleanings, which  continued during the voltage shutdown, and
  it can be seen that each cleaning caused a decrease in the pressure drop
  as expected.

  Field testing  of the Apitron unit was initiated in April,  1976.  All tests are
  carried out on a 1000 CFM, trailer-mounted pilot plant unit,  shown in
  Figure 8.   A  slipstream of process gas is withdrawn from exhaust ductwork
  and is then filtered by the pilot plant unit.  Generally,  this operation is
  carried out in parallel with existing emission control systems,  often bag-
  houses, and comparative performance can be easily defined.  Each field

-------
                                                                          295
test is run for two weeks or longer,  and stable operation is achieved
before data is taken.  Occasionally,  dilution air or water spray is added
to the pilot plant inlet in order to reduce temperature or to affect some
degree of gas conditioning.
                                         AMERICAN
                                         PRECISION
           Figure (8) - Apitron mobile pilot plant at asphalt
                       batching test location

   The results are shown in Table  1, and Figures  9 and 10 are examples
 of typical data.  The performance of parallel operating baghouses is shown,
 and  it can be seen that air cloth ratios from four to seven times higher than
 that of conventional equipment can  be achieved.  In addition, collection effi-
 ciency is generally improved,  resulting in a 1/2 to 1/10  reduction of parti -
 culate emissions on a weight basis.

-------
                                                                                          VO
                                    TABLE  1
Application
Asphalt Plant Rock Dryer
Gypsum Rock Dryer
Phosphate Conveyors
Electric Steel Furnace
Asbestos Wallboard Finishing
Cement Kiln
Air Cloth
Ratio *
CFM/
Sq. Ft.
12.5
13.5
13.0
8.0
15.0
8.0
Inlet
Loading
Grain /
ACF
32
10
7
.13
.32
3.5
Ou tie t
Loading
Grain /
ACF
.002
.001
.0008
.0008
.0004
.004
Parti culate
MMD
Microns
>40. 0
7.0
12.0
2.5
Fibrous 'N-3.
2.8




0

* Measured at a 3-inch W. G.
  Pressure Drop

-------
                                                        297
           STEEL  FURNACE  FUME


           .13 GR./CU. FT. INLET



      H.V. OFF  A          H.V. ON  0


      .0016 GR/CU.FT.     .0008 GR./CU. FT.


      OUTLET             OUTLET
  10
5:
 •
z
o
on
o

UJ 4
o: n
o

CO
UJ

£2
       EXISTING


       BAGHOUSE
         2    4     6    8    10    12


         FILTRATION  RATE, CFM/SQ.  FT.
 Figure (9) - Results of pilot plant testing on

             fume from an electric steel

             furnace, showing Apitron per-

             formance with and without electro-


             static augmentation

-------
298
                                 ASBESTOS- CEMENT
                                 .32 GR./CU.  FT.  INLET
                                 .0004 GR/CU.  FT. OUTLET
                               5            10            15
                      FILTRATION  RATE, CFM/SQ. FT.
             Figure (10) - Results of pilot plant testing on dust
                          from the ventilation of an asbestos
                          cement wallboard trimming operation

 CONCLUSIONS

    Two basic conclusions can be drawn from the given results:  high
 collection efficiencies can be expected from the Apitron, and high air
 cloth ratios can generally be achieved.  As previously discussed,
 improved collection efficiency results from the electrostatic augmentation
 of the filtration process because of polarization effects between charged
 particles and fibers.

   A more surprising result is that of the demonstrated increases in filtra-
 tion capacity per  square foot of fabric when charged particulate is collected.
 This occurs because the particulate deposits on the  fabric in a more open,
 loosely packed structure, and this difference in structure can be observed
 under magnification.  The structure is more permeable  and, hence,

-------
                                                                          299
allows greater filtration rates.

   For a given flow rate of gas,  the filtration capacity,  or air cloth ratio,
directly determines the overall dimensions of a fabric filter because it
specifies the area of fabric needed.  The method and frequency of removal
of the deposited particulate from the fabric can effect air cloth ratio
requirements.  For example,  a  fabric filter utilizing a  shake cleaning
mechanism might require an air cloth ratio of 2, while  a jet pulse fabric
filter would yield adequate performance on the same application at an air
cloth ratio of 5.  With the addition of electrostatics, the air cloth ratio
could be increased to 14.  Typically, an air cloth ratio  is selected such
that the pressure loss of the fabric filter will fall between three and six
inches of water. Figure 11 shows the air  cloth ratio capabilities of three
types of fabric filters and the  resulting floor space  requirements.

       FLOOR SPACE REQUIREMENTS
                  VS.
          AIR CLOTH  RATIO
       |/
-------
300
  The exact air cloth ratio required depends on the application,  but, in
  general, when the ratio is  in the high range for one type of filter, it will
  also be in the high range for the other two types.

     It has been seen that the electrostatic augmentation of the fabric filtra-
  tion process  yields improved collection efficiency and reduced space
  requirements. The Apitron system can be regarded as a new  category
  of particulate control device, offering performance unattainable with more
  conventional  equipment.

  REFERENCES

  (1)   Davies, C.  N., "Air Filtration", Academic Press, London, 1973;
       pp. 106-108.

  (2)   Frederick,  E. R., Chemical Engineering, June 26,  1961; pp. 107-
       114.

  (3)   Lundgren, D. A., and Whitby, K. T. ,  I & EC Process Design,
       Vol. 4, October, 1965; pp.  345-349.

  (4)   Walkenhorst,  W. and  Zebel, G., Staub-Reinhalt, Luft, Vol.  24,
       1964; pp. 444-448.

  (5)   Inculet,  I. I. and Castle,  G. S. P., ASHRAE Journal, March,
       1971; pp. 47-52.

  (6)   Natanson, G.,  Dokl,  Akad, Nauk USSR, Vol.  112, 1957;  pp. 696-
       699.

  (7)   Zebel, G.,  Journal of Colloid Science,  Vol. 20, 1965; pp. 552-543.

  (8)   Ariman, T.  and Tang, L., Atmospheric Environment, Vol. 10,
       1976; pp. 205-210.
WRITTEN DISCUSSION

T.M. Kuzay
Argonne National Laboratory
Argonne, Illinois  60439

    The authors are to be  congratulated for  bringing to our attention a field

-------
                                                                                301
 tested  filtration technique with much reduced pressure drop  feature compared  to
 a conventional bag-house design.

    A central theme of  the paper hinges on  the gains achievable  in lowering
 the overall pressure drop in the practical  bag-house type filtration systems
 by the  application of electric fields to such systems.  This reduction is
 achieved with equal or  better collection efficiencies when compared with a
 similar system of the non-charged case.  Qualitatively it is asserted in the
 paper that the smaller  pressure drop feature is due to dust  cake being more
 porous  in electrostatically charged systems and the increased efficiency is
 possible because collision cross section between the particle and the collector
 (fiber) is larger due to induced polarity.

    Ramifications of the paper are enhanced further when the nation is turning
 to recognize the energy shortage and the conservation as an effective measure.
 The smaller pressure drop in a filtering system for the same flow rate and dust
 loading, but with equal or better collection efficiency, affords a direct reduc-
 tion in the fan power necessary to drive the system.  For the proposed system,
 however, economics should also be studied to correlate the cost  aspect of the
 increased hardware and  the maintenance of the fields.  I presume the work being
 preliminary in nature will be expanded and encouraged to cover such matters in
 the future.

    As  noted earlier, data for the paper have been obtained essentially from  a
 field tested pilot plant and to a limited extent from a bench scale model.
 Therefore, only qualitative discussions are possible.  To increase accuracy
 and for comprehensive data, bench top models need be built.  In  this report,
 for example, a comparison of Figure 2 and 4 shows that at the same face velocity
 (15 FPM) and for the same cloth and dust variety (NOMEX/LIMESTONE) the data
 points  shown in Figure  4 could be shifted to higher absolute pressure drop
 values  resulting in less favorable pressure drop ratio between the non-charged
 and the charged filtration cases.

    Examination of Figures 2,3 and 4 shows also that the pressure drop in the
 charged filtration system levels off after approximately 10 kV/in applied
 voltage gradient for all fabric and dust varities presently tested.  Is the
 mechanism for such an applied potential gradient threshold well  understood?
 Existence of such a "saturation" point is important in the system optimization
 for power reduction.  Furthermore, if such a threshold, indeed,  exists, then
 this value might be considered for normalizing the data.

    The data presented  in Figure 9 raises questions in two respects and the
 author's comments are welcome.   First, is the given dust loading figure at
 outlet  the same for both systems?  Second, is indeed the single  data point
 shown for the bag-house a valid data point?  Or, were there some extraneous
 effects such as humidity difference between the two gas streams as the paper
mentioned as a general  comment.

    Figure 10 is perhaps a more direct comparison of the improvements possible
with a  charged system over a similar non-charged system.  It appears that even
 the non-charged APITRON has been a favorable design over an existing

-------
 302


 conventional   bag-house  system in  terms  of  its  pressure  drop characteristics.
 Consequently,  Figure  9 should  be interpreted with this difference in mind.

     In  conclusion, while I  find the  authors'presentation most interesting and
 far reaching  in  its energy  saving  potential in  engineering  applications,  I
 would like  to  emphasize  the necessity  for more  comprehensive and  well-controlled
 research  in fundamental  assessment and understanding  of  the system presented.
OPEN DISCUSSION

Helfritch:  I would like  to reply to Dr. Kuzay's  comments.

    First of all, I think we are referring to Fig. 2 and A in the manuscript.
These figures show that at the same face velocity with the same flow and dust,
the data points shown in  Fig. A should be shifted to a higher absolute pressure
drop. Those two sets of data were not taken under identical conditions.  The
one parameter that was different between the two was the relative humidity.
This is also a factor that I have not talked about, but is being looked at.
I think that accounts for the difference you see in pressure drop at given
filtration  rates.

    Examination of Figures 2, 3 and 4 show also the pressure drop levels off
at approximately 10 k volts/inch.  I do not know if there is a threshold' but
it is possible.  We could not usually go much beyond the voltages I have
shown because that is the spark level point.  If we were able to carry the
voltage gradient out farther, I do not know whether we would reach some sort
of a threshold or not.  At the inlet the grains per cubic foot would be the
same for both systems.  What is shown in Fig. 10 is the outlet concentration.

Liu:  Could you clarify the voltage gradient?  Is this the voltage of the
wire divided by the distance of the wire from the cylinder?  Or is this
actually the voltage of the wire?

Helfritch:  It is not a good representation.  The actual voltage gradient is
not linear at all.  It is just the average voltage.

Nichols:  You showed a variation in pressure drop with applied voltage.  Have
you conducted any tests where you operated at a lower voltage than the corona
inception?

Helfritch:  No.

Nichols:  There is a potential for the formation of dipoles that may lead to the
same type of mechanism.   If that is the case, you may want to put a pipe down
the middle of your bag rather than having an auxiliary charger if only the
dipoles are necessary.  Secondly, have you made any attempt to determine the
relative collection in the precipitator versus what it is in the bag filter?

-------
                                                                                303
 Halfritch:  On occasion we have done that and it varies quite a bit with the
 dust.  The best I can say it is between 20 and 80 percent, depending on
 conditions and the type of dust we are using.  It is not a very efficient
 precipitator.   We run very high velocities through those pipes.

 Nichols:  I wonder if it is the charging of particles or whether it is the
 formation of dipoles that is important for the reduction in the pressure drop.

 Helfritch: No, it is the charging of the particles.

 Nichols:  But  it still could be forming a dipole that leads to the low pressure
 drop.  They would not, necessarily,  have to be charged,  they could form dipoles.

 Helfritch:  There is no field to do  that.

 Nichols:  How  about from the previously deposited material?

 Helfritch:  But that requires a charged material.

 Nichols: The charge is on there when you charge the  particle,   If  all  you
 need  is to form a dipole,  all you would need  to do is form a field within the
 filter  by having metal fabric in the bag with a pipe in  the middle,  If  you
 form  the dipoles at the slow face velocities,  you would  still  have time  for
 these to align and give the low pressure drop.   I do not know whether  it is  the
 fact  the particles are charged  or the fact  that they have formed the dipoles
 that  leads to  the low pressure  drop.

 Lamb:   I have  seen similar effects but  we do not control the charge on the
 particles.   We do  not charge  the particles, but  apply a  field  in the plane of
 the fabric by  just putting electrodes in the fabric  itself.

 Penny:   Do you have figures on  the resistivity  of  the dust at  which this
 comparison was made?   Precipitators  are so dependent  on  this resistivity.

 Helfritch:  We have never measured resistivity.  We have no equipment for- that.
 We have  noticed a  strong effect  that  we interpret as  high resistivity problems.
 We have  also noticed  strong corona quenching effects when we go to very small
 particles.   In these  cases, we generally have to run at  lower voltages and
 currents  than  otherwise.  However, even  in these cases, particularly in Table
 1 for the  case of  the cement kiln and the electric furnace fume, we were way
 down in voltage and current but we still got improvements in performance.  I
 attribute  low back voltage and current to resistivity of the dust.

 Cooper:   It may happen that changing  the resistivity of the dust cake would
 change the particle size distribution.  This might be expected if you have 20
 to 80 percent collection in the electrical case beforehand. If you had a
 coarser  dust that would be consistent with a lower pressure drop per weight
per unit area on the filter.  It also might suggest to some degree  the different
structure in the cake.

-------
304
  Helfritch:  Again I allude to the bench scale-testing where none of the
  particulate was collected in the precipitator section,  Dust particles were
  deposited in the charged state.  Basically in our APITRON we are getting sort
  of the same thing.


  Reif:  With  the configuration you are using, I do not think there would be
  any evidence to suspect that there are polarization effects present.  Any
  particle polarized upstream probably would "unpolarize" by the time it got
  into the filter.


  Nichols:  The space charge would still have an electric field in the bag.

  Helfritch:  You are talking about a non-ionizing field; you are saying you
  produce a dipole before you produce corona.  You do not have a charged
  particle so you do not have a space charge field.

  Nichols:  But you can have a field set up where you can induce dipoles to
  determine whether it is a charge or dipole that cause this effect.   It is
  possible, but you did not do it.

      In Figure 6,  the dotted portion of the curve should look steeper,   I
  would  expect that there would almost be a vertical drop in pressure before the
  corona,  unless there is something else going  on.

      Looking at Figures 2,  3 and  4,  there is an area where  there is  no
  corona generated.

  Grassel:   Does this  affect  the cake structure?   Does it affect  the  depth of
  deposit  on the fabric  filter  and does it affect  the strength of the
  agglomerate that  comes off?


  Helfritch:       It does affect the  cake structure.   The depth of particulates
  penetrate into the filter would  probably have to be reduced  because most of the
  pressure drop  you see  across  a felted filter  is  due to  particles that  have
  penetrated the  filter.

-------
                                                                              305
ELECTROSTATIC CAPTURE OF FINE PARTICLES IN FIBER BEDS



    D.  L.  Reid

    Applied Engineering & Development Section, Battelle-Northwest


ABSTRACT

    Experiments at Battelle-Northwest led to the discovery that charged
particles could be efficiently collected by a bed of knitted fibers having
a very high bed volume -- fiber volume ratio and very low pressure drop.
The described concept depends on unipolarly charged particles being
deposited on fibers of like charge.  Results are presented for the system
efficiency for submicron particles as a function of velocity, particle
resistivity and bed depth and porosity.

INTRODUCTION

    The capture of charged fine particles in highly porous fiber beds was
discovered by A. K. Postma and W. K. Winegardner near the end of a study on
the collection of charged submicron particles by oppositely charged spray
drops.  To demonstrate the enhancement of collection by the charged spray
drops, the negatively charged particles were passed through a dry system
which consisted of a spray chamber and, fortuitously, a polypropylene fiber
demister.  Surprisingly, a rather dramatic increase in the collection effi-
ciency was observed rather than an expected lower efficiency.  This obser-
vation inaugurated an investigation which eventually reduced the concept to
practice and ultimately a patent.

    Since the Demister® pad produced little or no attenuation of submicron
particles without field charging, some phenomenon other than natural depo-
sition processes was involved.  Subsequent tests revealed that the collec-
tion mechanism depended on the development of an induced space charge within
a bed of electrically non-conducting fibers.  This is simply accomplished by
passing the aerosol through a negative corona field and then through the
fiber bed.  The deposition of charge on the fibers instantaneously creates
a negative space charge field within the bed.  Then if a highly electrically
resistive fiber is used and the particle is electrically non-conducting, the
bed field charge is maintained at a magnitude sufficient to produce a high
removal efficiency.

-------
306

  THEORY OF CONCEPT

      The mathematical modeling of the concept, by D. L. Lessor, is too
  extensive for this presentation so only a brief explanation is presented.
  Since the electrical field in the bed is induced by the deposition of
  charge, efficient operation of the system depends on the development of
  electrical fields of like polarity sufficient to overcome the coulombic
  repulsion force of a single fiber within times shorter than the residence
  time of the particle within the bed.

      As the charged aerosol enters the bed of like charge, the interaction
  between one particle and the fiber nearest to it is a repulsion force.
  However, the charge distribution on the whole bed produced an electric
  field force that drives the charge from the bed.   This bed scale field
  apparently has a force sufficient to divert the charged particlefrom the
  normal flow lines and onto the fibers.   Thus, the fiber bed concept is
  effective when the bed scale field force is significantly greater in mag-
  nitude than the repulsion force of a single fiber.   The following calcu-
  lation compares the magnitude of the bed scale field in the gas flow direc
  tion (z), E ,  with the radial electric  field, E .  caused by the nearest
  fiber.                                         r

      Consider a fiber filter of thickness 2h in the  direction of the gas
  flow and infinite in the transverse direction.   If  the fiber length per
  unit of pad volume, £v, is uniform and  the fibers carry uniform charge  per
  unit length,  q£, the macroscopic charge density,  t,,  is
  This charge density serves  as  a  source for the  bed  scale  electric  ffeld
  whose local  average over several  fiber spacings has a  component  E   given  by

             0 C. 2
      E7  =  -^r-  (for -I
  where e is  the  permitivity of the  bed,  and  z  is measured  from  the midplane
  of the bed.   Positive values  are in  the direction of the  gas flow.

      The radial  electric  field from the  nearest charged fiber is


      E   .  J!s-
       r     2nre

  where r is  the  distance  from  the center of  the fiber.

      Then the  ratio  of these forces at the fiber surface is

-------
                                                                             307



    rr  -  2nVfz  =  2B^

                                         o
where rf is the fiber radius and B is nrf £v,  the  fraction  of the  bed
occupied by fibers.

    This ratio is much larger than unity for the conditions used in  the
fiber bed concept.  For example, with an occupied  bed fraction of  0.05 and
a ratio of bed thickness to fiber radius of 500, then Ez/Er is 50.

    A computer program has been developed which  calculates  steady-state
conditions for simultaneous charged aerosol deposition,  charge conduction
on the bed, charge leakage off the bed edges,  and  effect on deposition by
the field from the retained charge.  This modeling work  indicates  that

    *  electrically resistive beds retaining substantial  charge provide
       enhanced filtration efficiency and

    *  the bed scale electric field effect is  a  dominant colle'ction
       mechanism for submicron aerosols.

EXPERIMENTAL PROGRAM

    Based on initial observations variables thought to influence system
efficiency were (1) particle-size, concentration,  resistivity and  charge
level, (2) gas-velocity, temperature and composition, and (3) fiber  bed-
resistivity, thickness, porosity and dust surface  coverage.  A complete
block experimental program involving all variables would require more than
a practical number of tests for proof of the concept.  Therefore,  the
initial program was designed to look principally at the bounding condi-
tions over some reasonable operating range.  Table I lists  the test  condi-
tions for the parameters of interest.

                      TABLE I.  System Test Conditions

        Parameter                               Range

    Particle Size                         0.06 - 3 ym
    Aerosol Concentration                 10 - 250 mg/M3
    Particle Resistivity                  108 -  1013 ohm-cm
    Particle Charge Level                 Near Saturation
    Gas Velocity                          0.25-2  m/sec
    Gas Temperature                       20 - 130 C
    Gas Composition                       Air and  Combustion Products
    Fib.er Resistivity                     Conducting and Nonconducting
    Bed Thickness                         7.5 -  30 centimeters
    Bed Porosity                          Vpid Fraction - 0.94 - 0.98

-------
 308
 TEST SYSTEM

     Figure 1  is  a schematic drawing  of the system  which  consists  of a  7
 foot long 24  inch diameter pipe,  a short  corona  charging section  with  an
 effective cross-sectional  charging area equal  to that  of the duct, a 3 foot
 long section  between the charger  and fiber bed to  preclude any direct  field
 charge effects from the charge section, and a  bed  frame  for the knitted
 fiber mats.   The inlets and outlets  to the bed were  24 inches diameter.
 Since the bed was hand  packed, the bed face dimension  was made 28 inches by
 28 inches to  extend the experimental  run  time  before edge leakage became
 the predominant  source  of  downstream particles.  The frame was made with
 lucite to accommodate up to a  12  inch thick bed  and  was  electrically iso-
 lated from the rest of  the system as  was  the charging  section.  The system
 was built in  sections for  future  alteration of configurations and distances
 between principal  components and  was  held  together with  electrically insu-
 lated external straps.   A  dispersion  plate was placed  in the 12 inch diam-
 eter inlet at the transition point to produce  the normal velocity profile
 at the three  principal  sampling points  at  positions  1, 2 and 3.

 FIBER BED VOID FRACTION

     The void  fraction of the fiber bed  was  defined as  that volume of the bed
 not occupied  by  the  fiber  volume  and  was determined  from the bulk weight of
 the fiber mats,  density of the fiber  per unit  length and the fiber diameter.
 The knitted mats  were supported within  the  frame by  a  grid of teflon coated
 wires  having  about a  4  inch wire  spacing.    The 15 cm thick bed was packed to
 ^0.96  void fraction  (VF).  The 7.5 cm bed  had an ^0.94 VF.  For the 30 cm
 thick  bed, the first  15  cm had a  VF of MJ.975 and the  last 15 cm ^0.96.  The
 lower  fiber density  in  the first  half was used to increase depth of penetra-
 tion and  thus increase  the  load to bed  pressure drop ratio.   This  accomplished
 both goals, but appeared to be detrimental with respect to efficiencies in the
 later  stages  of a run series as well  as length of operating  time before a
 decrease  in efficiency was observed.   Visually, the depth of penetration
 appeared  to be directly  related to particle resistivity for  the submicron
 aerosols  — increasing with increasing resistivity.  Particle size was  not a
 primary variable.  Interest was in the 0.1 - 0.3 ym size range except for
 fly ash which was collected downstream of a two-stage ESP.

 PARTICLE GENERATION AND CHARACTERISTICS

    Four particles were used in this  series of experiments to highlight any
 change  in efficiency with  particle resistivity — namely, amonium  chloride,
magnesium oxide,  sodium oxide,  and fly ash.  The fly ash, obtained from a
 plant burning  low-sulfur Western coal, was resuspended by aspiration.   The
 NH4C1 particle was generated by air sparging of HC1  and NfyOH solutions with
 interception of the gas  streams at the inlet to the system.   The particle
 generation rate was controlled  by varying  both the  concentration of  the
 solutions and  the sparge air flow rate.  The MgO  and Na20 particles  were
 generated by metal vaporization and oxidation.   The metal was  heated  induc-
 tively  in a boat  enclosed in a  tube.   An inert gas  forced the metal  vapors

-------
                             WIND TUNNEL SCHEMATIC
 AEROSOL    DISPERSION
GENERATOR      PLATE
SAMPLING POINTS
                                                                                 EXHAUST
                    SUPPORTS    CHARGE SECTION     FIBER BED
                                     FIGURE 1
                                                                                          to
                                                                                          O

-------
310
  out the end of the tube where ignition to the oxide occurred.  Since a
  technique was available for varying the particle size, some tests were
  made with MgO particles having an AMD of less than 0.1 urn.

      An eight-stage Anderson sampler was inserted upstream of the charge
  section to obtain the aerodynamic mean diameter for each generation in
  which the conditions were altered.

  TEST PROCEDURE

      After establishing the desired particle generation rate mass samples
  were collected on membrane or fiber glass filters at three positions:
  upstream of the charger and upstream and downstream of the bed.  The high
  particle and field charges upstream of the bed produced odd results that
  defied interpretation and precluded the use of much of the data for deter-
  mining bed efficiency.  Consequently, the results reported are for the
  system efficiency.  Later tests have shown good agreement between bed and
  system efficiencies -- apparently the result of cutting the current flow in
  the charge section by one-half.

      Generally, operation was continuous until the end of the day or until
  apparent breakthrough (whichever came first) except during adjustment of
  generation rate and flow.  Usually, two consecutive measurements were made
  before changing the flow and/or concentration.  Low aerosol concentration
  runs generally required two days of operation.  When this occurred, the
  system was shut down over night.  Prior to particle generation the second
  day, the charger was turned on to impose a field charge on the bed and a
  reentrainment test made at 1.5 m/sec flow.  There were no apparent degrading
  effects even though the bed charge was at zero over night.  For a series of
  measurements, the flow rate was varied from low to high and back td lo»\» or
  in the reverse order.

  RESULTS AND DISCUSSION

      Typical results obtained with the 15 cm and 30 cm thick beds whose
  void fractions were described earlier, are shown in Figure 2.  The decon-
  tamination factors are plotted as a function of velocity.  The data
  approximate the expected relationship with the log of the Df varying
  linearly with a power of the velocity.  For the 15 cm bed all data were
  considered to be from the same population for the least squares fit repre-
  sented by the line.   The coefficient of determination was 0.9 and all data
  fell within the 2a error limits.

      For the 30 cm bed, the log-log plot of the data suggested a real
  although small difference in the system efficiency for NfyCl particles.
  The NH4C1 data were segregated from Na20 and MgO for the linear regres-
  sion analysis.  The average efficiency for NH4C1 was lower than that of
  MgO or Na20 by about 2 to 5 percent over the range of 0.5 - 1.5 m/sec face
  velocity.  This is the first data that might confirm the model's prediction
  that efficiency should decrease with decreasing resistivity.  Since the

-------

99.8


99.5

LU
* 99
^j-
o
"Z.
n 98
o
i
1 95
0
o 90
o
80
50


- 500


-uT 200
i— i
r—t
•of 100
o
t—
0

-------
312
  bulk of the data does not indicate a like trend, final  judgment is reserved
  until confirmed.  Also bearing on the reservation is the lack of any dis-
  cernable difference in the bed field charge, inferred from current measure-
  ments, between the four aerosols used.   Thus, it appears that something
  more subtle than particle resistivity was responsible for the lower effi-
  ciency observed and the threshold for this effect would occur at a lower
  resistivity than the 108 ohm-cm of NfyCl.  The anomalously low efficiencies
  for NfyCl and MgO at 0.5 m/sec face velocity were the last samples taken in
  each series when the operating log noted edge leakage and were not included
  in the analysis.

  8 cm Thick Fiber Bed

      Since the efficiency appeared to be influenced by fiber density as  well
  as thickness, a run series was made with the NfyCl aerosol  using a 7.5  cm
  thick bed packed to a void fraction of  ^0.94.  In this  test series it was
  planned to obtain successive measurements at one velocity until  break-
  through at the bed edges became apparent.   In an effort to increase the
  number of consecutive measurements at the high aerosol  concentrations used,
  a folded fiber mat was inserted at the  bed-frame interface to bring the
  fiber density at that point nearer to that in the rest  of the bed.   This
  produced a marked improvement in total  run time before  edge breakthrough
  was indicated at a much higher pressure drop than in previous tests.  Five
  successive measurements were made for all  flowrates except 0.5 m/sec  where
  twenty measurements were made.   The particle size for this series of  tests
  hovered around 0.25 pm AMD and a mass concentration near 75 mg/M3.

      A typical equation developed from the least squares fit of the effi-
  ciency versus velocity was

      ES  =  102.8 - 10.28 V

  where

      Ej  =  system efficiency in percent and,

      V   =  bed face velocity in m/sec and
       2
      r   =  0.986, correlation coefficient.

  The plot of the decontamination factor  vs. velocity (Figure 3) shows  the
  average values for each velocity along  with  the 95 percent confidence
  levels.   Since the average efficiency (99.95  percent) at 0.25 m/sec flow
  was higher than the measurement sensitivity,  the  lower  value of  99.93 per-
  cent was used for curve fitting.   The slope of the line was highly  influ-
  enced by the high Df at the lowest velocity  tested.   This  was the first
  indication that the log of the decontamination factor may  not be linearly
  related to some power of the velocity for  all  conditions.   More  tests at
  lower velocities will  be required to define  the relationship.  But  first

-------
                                                                  313
                    DECONTAMINATION FACTOR
                    7.6  cm POLYPROPYLENE BED
                        0.946 VOID FRACTION
o
UJ
o
99.9


99.8



99.5


99


98
CO
95


90


80
                  1000


                   500
              100


               50
              o
              o
                   10
                                      SENSITIVITY LIMIT
                                  NH4CI-0.25MmAMD

                                   Dp - (16.24) V"
                            BEDAP /
                         Es = SYSTEM EFFICIENCY

                        ® = NUMBER OF DATA POINTS
                                                        1.0
                                                           o
                                                             CM
O
a:
o
LU
O£
                                                           CO
                                                           UJ
                                                           Of.
                                                           Q_
                                                           OQ



                                                           I
                                                           O
                                                  0.1
                                                       0.05
                     0.1                    1.0        3.0

                       SUPERFICIAL GAS VELOCITY, m/sec
                           FIGURE 3

-------
314
  some measurement method other than mass, if available, will  be required to
  provide adequate sensitivity at the lower velocities.

      Comparatively, this thinner <^8 cm bed with a lower void  fraction
  exhibited an efficiency about equal to the thicker beds at 0.5 m/sec flow
  and at 1.5 m/sec flow was about 5 percent higher and lower than the 15 and
  30 cm thick beds.

  Loading Run

      Continuous loading runs were made to examine loading and pressure drop
  characteristics of the bed.  Generally the system was  operated continuously
  during the day and shut down over night.  After being  shut down,  the corona
  charger was turned on before air flow was re-established and aerosol  gen-
  eration effected.   A typical  run made with fly ash having an AMD  of 2 vm
  and particle resistivity near 1013 ohm-cm is shown in  Figure 4 to highlight
  the higher efficiency obtained with aerosols having AMD's greater than 1  ym.
  Also included in the figure are the load and pressure  drop as a function of
  time.   Since the load-pressure drop relationship is a  function of the par-
  ticle  resistivity  and bed void fraction the onset of an increase  in the
  pressure drop and  the rate of change may be different  than that observed
  for fly ash.

      The lower efficiencies observed for the first few  samples after restart-
  ing the system the first  five  times are as yet unexplained.   Since this was
  not observed in other loading runs, it is thought to be peculiar  to this
  series of tests.   Comparable efficiencies for the 8 cm thick bed  at a face
  velocity of 1.5 m/sec were >99 percent for the 2 um AMD fly  ash and -v,90
  percent for the submicron aerosols.   This higher efficiency  may be due,  in
  part,  to particle  attenuation by natural  processes such as impaction  and
  interception.   The long  period of operation before an  increase in the bed
  AP was observed reflects the  high void volume as well  as  the high load
  capacity that can  be obtained especially with beds of  higher void fractions.

  SUMMARY AND CONCLUSIONS

      The efficiency and design flexibility of the fiber bed for the collec-
  tion of submicron  particles suggest many practical  applications for rela-
  tively economical  particle control  that is  not attainable with presently
  available equipment.   The deposition of a negatively charged  particle on a
  negatively charged fiber is apparently dependent upon  the magnitude of the
  ratio  of the  bed scale electrical  field to  the electrical  field of the
  individual  fiber —  both created by the deposition of  charge  carried  by  the
  aerosol.   Other principal  observations were as follows:

      *   Velocity was  the  prime variable with the efficiency decreasing
         approximately linearly with  increasing velocity.

      "   Particle resistivities  over  the range  tested  produced  no consistent
         trend  in collection efficiency.   However,  some  degradation  in

-------
                           POLYPROPYLENE FIBER BED
                       7.5 cm THICK - 0.94 VOID FRACTION
                FLYASH  - 2pm AMD  - 10*3 OHM - CM  RESISTIVITY
                            1.5 m/sec FACE VELOCITY
0
100


 99


 98


 97
            °-° - ^
oo°o°0
                                                                        0  o
                L
        INDICATES OVERNIGHT STANDBY


       '     ill	J	1	L.
           4    8    12   16   20    24    28    32    36   40   44   48   52   56

                                      TIME IN HOURS
   0.40
o 0.36
 CM
^ 0.32

D_
< 0.28
                                                                                    o
                          J	I	I	L
  J	L
                                                                 J	L
                                   FIGURE 4

-------
316
         efficiency might be expected at some lower resistivity -- the thres-
         hold, of which, is yet unknown.

      *  Particle size had no apparent effect on efficiency for AMD's less
         than 1 urn.   A significant increase in the efficiency was shown for
         aerosols of 2-3 urn AMD — ascribed to both electrostatic effects and
         natural deposition processes.

      *  Fiber electrical resistivity is essential for efficient particle
         control in the submicron range.  Any fiber that maintains a high
         electrical  resistance during operation should be applicable.

      *  Bed thickness had a pronounced effect on collection efficiency espe-
         cially at the higher flow rates.   However, small changes in the
         fiber volume density produced significant changes in the efficiency.
         Apparently the bed thickness and fiber density can be tailored for
         the aerosol condition.

      *  Dust concentrations between 10-250 mg/M3 were used without any sig-
         nificant effect on bed efficiency.  As might be expected, data
         reproducibility was better at the higher concentrations.

      "  Particle charge level  was not varied intentionally.   For these
         initial investigations, a near saturation charge was desirable.
         The fiber beds had no  affinity for uncharged particles at the
         higher velocities and  showed <\-8 percent retention at a 0.25 m/sec
         flow.

  The test  results substantiate the influence of the bed scale electrical
  field forces  for collecting submicron particles.  The computed Ez/Er ratio
  for the beds  tested ranged from 20 - 70 without major changes in the
  efficiency.
 OPEN DISCUSSION

 Melcher:   I would  like  to  set  the  trend of  the discussion if I could.  I use
 a rule of  thumb  in dealing with electrostatics and  that is  that there is no
 such thing as a  repulsion  force.   The reason I say  there is no such  thing as
 a repulsion is you really  ought to account  for where  the charge is.

     You have the filter, with  charged particulate passing through it, the
 question arises:   how is it  that the collection sight collects anything? Let
 us  look at a calculation, call  the  half spacings, S, my contention would be that
 one way to look  at it is the particulate  gets into  the matrix, then, in a
 van de Graffe sense, with  the  gas  as the  belt, and  this (the collectors) as the
 resistic load generating its own electric field.  It  is that electric field that
 then allows it to  be collected on  sight,  As a calculation, let us say we
 have  Pmas a density per  unit volume, that would mean if we have a  charge on

-------
                                                                              317
each of the particles in there, we would have a charge per unit volume which
would be  q.

    Gauss' law says that coming out of this region is an electric field which
is equal to  S  times the charge density, or eQE actually.  So if you look
at a particle out in front we find that it is experiencing this sort of
polarizing field.  Not only does it have a charge, q, but it is also in an
ambient field.  That means the particle with charge  q is generating a field
that would equal  R -5^ — , where R is the radius of this spherical particle.
So the situation is such that the particle experiences the ambient field and out
of the particle comes its own field.  Certainly if plus particles are going
to get collected on that surface, we have to have  E lines coming in.  So you
see how you might solve the problem.  Just look at it as a conducting sphere
in a uniform field carrying that charge.  You would find that it is a matter
of whether there are any  E  lines coming in.   It depends on if  C^~*~ R) is
less than the  E  field coming in, which is  -^a-   where  3  is the
concentration factor.  You can then cancel out the charge and e ,  Let us also
                                                               o
view the packing as being determined so we get the particle size as much as
possible by the spacing between particles, aR.  In other words I am normalizing
the spacing to the size of the particle.  In that case the number density is
^=lparticle ^  We make thg condition t^j- the particles will collect if
    a r
, < (3S)(4Tr)
       3
     Ra

    What this tells you is that the thickness of the filter must be bigger than
the size of the fiber particles, which is a very good approximation.  In order
for it to compete with an electrostatic precipitator, you want this self-
generated electric field to be larger than what you have in the ESP.  It will
do much better in resonance time than ESP.  To me it is an electro-packed
bed with a self -generating field, where this is a big resistor that drains
off current.  One thing that bothers me about the experimental work is how
much it can be related to something else when in fact you are not held
accountable for where the images are.  If the load is on the edges then the
field is not actually one dimensional, it is actually quite complicated and
will depend on the size.  If you had screens clearly the field would be
longitudinal and you would have the thing well under control.  I would worry
how relevant it is to scale-up.

    There is a patent by Cole in which he packs between screens particles which
are semi- insulating.  He sends the current through and collects in this mode.
What he has done is to apply the field rather than to self- generate it.
That would be my view of it.

Reid:  I did not do the mathematical modeling.  This was done by D. L. Lessor
This is true if you have a uniform charge throughout the bed and a uniform
fiber length per unit volume.  At this position in the bed plane, essentially
your charge effects are zero.  However, this is not the condition encountered
in the fiber bed.  If you look at it with respect to this and you look at the
ratio of electrical force in the bed versus the electrical force at the surface
of the fiber, you end up with a relation of  ±   .  These are just physical

-------
  318

 dimension of the system but  it  does  give  an  approximate  relationship  of  the
 bed scale field  forces  with  respect  to  the force  at  the  fiber  surface itself,
 The system,   as  we have run  it,    varied from ratios  of 3  to  10,  We
 interpret the repulsive forces  as  acting  on  the particle itself and diverting
 them from the normal  flow lines that you  would expect  in the system.   That
 would mean a diversion  of submicron  particles from their normal flow  lines
 to  allow them to deposit on  the fiber itself.  The magnitude of the bed  scale
 field effect overcame the repulsion  forces of the fiber  and allowed it to
 deposit.

 Melcher:   This,  by the  way,  is  the configuration  in which Whipple & Chalmers
 say en  isolated  particle will not  collect.   This  is that "backward configuration,"
 One that  we  say  collects just as well.  But  one had better put something into
 his model that adds turbulence, or has  something  about the wake in it,

 Reid:   He has a  very  extensive  mathematical model development of this  system.
 It  is worthy of  a paper in itself  and I hope it will be  published in  the not
 too distant  future.

    On  the upstream face of  the 12 inch bed, it appears  that it is a  completely
 closed  structure.  It should have  an extremely high pressure drop in  this
 system.   Under these conditions the pressure drop at the  time we terminated
 this  test is about 1.2  inches of water.  The particles deposited completely
 around  the surface area of the  fiber not in any one particular orientation,
 front or  rear.   The initial modeling work predicted the particles would all
 deposit on the rear face of  the fibers.  He had to revise his mathematical
 modeling  on  the  basis of  his observations.  The particles deposited around
 the fibers.

 Cooper:   The thing that  intrigued me most was that if you shut it down, waited
 overnight, turned the gas on the next day with a corona charger, you'could
 get the thing back up to  the condition where it was catching charged particles.
 That suggests to me that  something other than the particles themselves was
 bringing  the  charge to  the fibers.

    There are two possible solutions:  1)  You are some how getting it from the
 ions and  2) Maybe we are  seeing some sort of tribo effect in the plastic. That
 fact really puzzled me.

Reid:  If we run on uncharged particles for the bed we see a positive leakage
current, not a negative one.   That suggests that the bed is producing a positive
charge.  It  is very weak charge we are producing under the hydro-electric
 effect, principally since we have such an open structure.

    It is  true that all we had to do is turn on the charged section,  although
we could not use filtered air, and all air has particles in it.  So whether
we are looking at ion deposition or charged condensation nuclei count particles
 that are deposited on the fiber, we do not know.   We produced approximately
an order of magnitude less field force in the bed than when we were generating

-------
                                                                               319
particles.  We do not know whether it is ion deposition or deposition of
condensation nuclei that have been charged.  We know the system is about 95%
efficient for condensation nuclei particles.

Cooper:  I am still puzzled on how a negatively charged particle is deposited
on a negatively charged bed.  The strong coulombic repulsion, it would seem,
would not allow a deposition.

Reid:  The key is the ratio of the field forces.

Liu:  I would like to suggest another point.  Your device is very similar to
a space charge precipitator,except you have localized, immobilized centers
on which the space charge can be held stationary.  So your figures depicting
repulsion are still valid except that in half the electric field you have
additional condensation centers which would act as impaction centers for the
particle.  For the case of the space charge precipitator, the particle would
have to be driven all the way to the driving surface before it could be
collected.  It is a combination of Melcher's fluidized bed device plus the
space charge precipitator.

Reid:  That's pretty much right.

Friedlander:  I am a little confused.  Where is the energy coming from?

Melcher:  It is a self-induced space charge.  It is a gaseous van de Graffe.

Reid:  The particles are the source of the energy.  There is an imposed charge
put on the particles upstream of the system.

-------
 320
 FINE  PARTICULATE  AND  GASEOUS  EMISSIONS  CONTROL  EXPERIENCE  WITH
 THE TRW  CHARGED DROPLET  SCRUBBER
     R.  R.  Koppang

     TRW Inc., Energy  Systems  Group


ABSTRACT

     A unique, hybrid  electrostatic precipitator  for  the collec-
tion of particulate and gaseous emissions has been developed and
commercialized by TRW.  The TRW Charged Droplet  Scrubber  (CDS)
combines high energy  wet scrubbing collection for micron  size
particulate with electrostatic precipitation for submicron
particulate collection.  The  gas cleaning process steps of
water droplet atomization, acceleration, particulate collection
and  demisting are performed through applied principles of
electrostatics within a gas treatment equipment  length of eight
feet.   Other CDS operating advantages include:   no moving parts
in the  gas stream; low sensitivity to particulate physical
properties; and low liquid-to-gas ratios.  Particulate collection
efficiencies exceeding 99% and 96% have been accomplished in the
micron  and submicron  region,  respectively.  HC1  gaseous emissions
control  from 90 to 99% have been obtained in pilot scale work with
minor modification of the equipment internals.   S02 flue gas
desulfurization also  appears  promising.

INTRODUCTION

     Recent, more stringent air pollution controls require high
collecting efficiencies of submicron particulate to meet stack
opacity  regulations.  The simultaneous control of gaseous and
particulate emissions with one unit operation is highly
desirable.  Such collection must be accomplished with equipment
which minimizes capital costs and significantly  reduces energy
consumption.

     The  CDS employs a unique engineering concept which provides
for more diverse and  economic applications of electrostatic
precipitators in particulate emission control.   Both large and
small gas flow rates  can be economically controlled with
modular preassembled CDS equipment.

-------
                                                               321

    The CDS is a hybrid particulate collector using two
parallel mechanisms; high energy water scrubbing and electro-
static mechanisms.  These mechanisms are further augmented by
the gas conditioning effects of droplet evaporation.  Larger
particulate is captured through high energy collisions with
optimally sized water droplets.  The benefits of the small
equipment size associated with wet scrubbers and the low energy
requirements of electrostatic atomization and precipitation are
thus obtained.

    The CDS is less sensitive than ESP's to particulate
physical properties.  The electrostatically sprayed water
droplets create highly concentrated space charge fields and
the evaporative losses are sufficient to improve the stability
of high voltage breakdown.  The walls of the plates used to
collect charged particulate and demisted charged water droplets
are evenly wetted and the material continuously flushed as a
slurry.  Its small  size minimizes plant modifications, where
space is usually  limited, and permits factory assembled equip-
ment to 100,000 cfm capacity.  Low liquid to gas ratios, less
than 1 gal/1000 cfm, minimize waste water treatment and the
operating and capital costs associated therewith, and frequent-
ly allow waste water to be returned to other points in the
factory production  processes without secondary treatment.  In
many applications  the CDS discharge gas is not saturated thus
allowing less expensive fan and duct materials and minimizing
the frequency of  steam plumes.

    These attributes of the CDS have found wide appeal among
users in a  number of basic industries, such as pulp and paper
or iron and steel,  as well as  several specialty applications.

    Described below are some of the more unusual operating
principles.  This is followed  by  a discussion of recent develop-
ment and commercial experience with equipment configured  for
improved performance on submicron particulate and gaseous
emissions.  A previous paper,  Reference 1, reported on
earlier pilot and commercial experience on a  first  generation
design.

OPERATING  PRINCIPLES

    The CDS operating  principles  are briefly  summarized below.
More detailed explanations may be found in References  1 and  2.
The CDS accomplishes the  major process  steps  for particle
collection simultaneously within  a  small  treatment  zone using
electrostatic principles  for droplet  atomization,  charging,
acceleration, particulate collection  and  demisting.

     In the use  of electrostatic atomization  the CDS is unique  to
chemical  process equipment  design.  Electrostatic  spraying

-------
322

 results in well dispersed water droplets which,  when collected,
 evenly  distribute water films on collector plates.   Within the
 CDS water distribution and atomization are accomplished by an
 array   of metering tubes arranged along a conducting, water
 carrying, high voltage energized electrode,  Figure 1.  Local,
 high voltage gradients are created at the tips of the tubes
 located equidistant between grounded parallel collecting
 electrodes (plates).

     During the engineering development of the CDS various  types
 of  electrostatic atomizer designs,  hollow tubes  and  rods,  were
 considered.   Currently a tube  design  has been selected which
 provides sufficient water flow rate for good droplet spatial
 distribution,  optimal  mean droplet size for  high collecting
 efficiency,  good plate water film coverage,  and  acceptable
 evaporation  rates.   High speed (1 second)  photographs of the
 electrostatic  spraying phenomenon for a voltage  gradient of
 5 kv/cm equivalent to  corona breakdown in air are shown in
 Figures 2 and  3.   Without high voltage,  the  water jets issuing
 from the tubes do not  become atomized because of the low feed
 pressure (under 10"  we).   As voltage  is increased above
 2-3  kv/cm, the water column becomes highly unstable  bursting
 into a  convoluted filament as  it searches for a  minimum gradient
within  the space  charge field.   A highly filtered DC (±5%)  power
 supply  is required to  obtain stable electrostatic spraying at
maximum operating voltage.  Electrode operating  rms  voltages
 (35-55  kv) are typical  of  conventional  dry electrostatic precip-
 itators  having similar  electrode geometry.

     Figure 2 is a  time  sequence  photograph (15 exposure) which
recorded  the position of  the water  column as  it  adjusted to  the
droplet  space  charge; from this  photograph the excellent space
distribution of water droplets can  be noted.  The Figure 3
photograph recorded the early  stages  of  disintegration  of  a
portion of filament that separated  from  the main  column; charged
droplets  are sprayed off the ends of  the  column and  from kinks
in the middle.  Filament break-up processes,  including  local
surface separation where surface charge  forces exceed tension
forces  (Rayleigh instability),  can  also be observed  in  both
pictures.  Typical droplet sizes were in  the  range from 120  to
180p m model (geometric) diameter,  and  300 to 400u m mass-mean
diameter.  Droplet size is influenced by  nozzle tip  geometry and
water flow rate.  The maximum velocity component seen for both
120p droplets  and 180p droplets was about  30 m/sec.

    Altering the physical properties of the spray liquid by
adding a  surfactant to reduce surface tension or soluble salt
to increase conductivity have only  secondary effects on spray
patterns.

-------
INSULATOR
SCRUBBED GAS
  DISCHARGE
TO ATMOSPHERE
   D-C
  POWER
                     DUST LADEN
                      GAS FLOW
                                                                      ELECTRODE
                                                                  (HIGH POS. POTENTIAL)
                                                           CHARGED
                                                           DROPLET
                                                             SPRAY
                                                           COLLECTOR
                                                             PLATE
                                                         (NEG. POTENTIAL)
                                                             WATER/DUST
                                                               SLURRY
                                                              CARRY-OFF
            Figure 1   Charged  Droplet  Scrubber Operating Principle and  General
                        Component Arrangement  for a Single Electrostatic
                        Spraying Stage
                                                                               OJ
                                                                               ro

-------
                                                                                         OJ

Figure 2  Droplet Formation by electro-
          hydrodynamics spraying,
          expsoures at 1/15 second
          intervals
Figure 3  Droplet formation,  filament
          breakup

-------
                                                                325
     For example,  increasing conductivity from nominally 300 to
 10,000 micromhos/cm requires a reduction in electrode voltage
 of 2-4 kv,  but does not substantially change the characteristics
 of the spray.   Apparently,  the improved conductivity decreases
 the relaxation time for droplet charging.

     Particulate collection  in the CDS is accomplished by both
 direct collision  of high energy water droplets and electrostatic
 precipitation.  Direct collision and agglomeration is the
 dominate mechanism (efficiency)  for particulate greater than one
 micron,  depending upon particle density and wetability.   Water
 droplets,  instantaneously charged and accelerated,  reach terminal
 velocities  approaching 100  fps with a velocity trajectory which
 is substantially  cross flow to the particulate velocity.   The
 resulting high relative velocity (cross flow mass transfer)  and
 optimally generated water droplet size (100-200 microns)  result
 in target collecting efficiencies as high  as 99+% for 2-3u sized
 particle.

     Two  electrostatic particulate charging mechanisms are used
 to achieve  submicron particulate collection.   Charge transfer
 to the particulate is accomplished through gas ion current
 generated at the  tube tips  by corona discharge (conventional
 ESP) and by grazing encounters of the particulate with water
 droplets.   In  the latter case,  charge transfer is the result of
 field  distortions during the  droplet-particle  encounter which
 result in a droplet surface electrical breakdown.   Such  surface
 electrical  breakdown can effectively occur at  surface to
 particle distance of up to  2y  for a O.lp particle (e.g.:  a
 single encounter  will  transfer sufficient  charge  to precipitate
 a  particle  at nominal  operating  conditions).   The optimum
 relationship between particle  and droplet  size, Figure 4,  has
 been estimated  from theoretical  considerations.   The  optimum
 droplet  size is about  proportional  to the  particle  size,  but
 always a factor of  ten larger.

 EQUIPMENT DESIGN

    The  previous  section described  the  equipment  configuration
 and operating parameters  for a single  chamber  (module) and
 electrostatic spraying  stage.  CDS  overall efficiencies are  a
 function of process  conditions and  equipment design.  Signifi-
 cant process conditions  are particle  size and composition.
 Physical properties  of  importance are  specific gravity,
wetability, and to  a lesser extent, high electrical resistivity.

    In considering overall equipment efficiency for a specified
process the two most significant CDS design parameters are:
Number of stages operated (Figure 5) and the plate spacing - gas
 flow velocity relationship,   (Figure 6).  Both of these parameters

-------
  4.0
 o


<"

*o
10
  2.0
z
LU


P
u
   1.0


   0.8
   0.6
<0.4

O
 0.2
 0.1
   0.1
                0.2
                                S, = .5 MICRONS
0.4    0.6  0.8  1.0         2.0


      PARTICLE RADIUS (MICRONS)
                                                                 DROPLET RADIUS
                                                                  S] = 100 MICRON



                                                                  S =60 MICRON
                                                                    .SQ = 30 MICRON.
                                                                   4.0
                                                                           6.0  8.0 10.0
        Figure  4   Theoretical Droplet-Particle Removal  Effectiveness

                   Relative to a 30 micron Radius Droplet
                                                                                            OJ
                                                                                            ro

-------
   100
    98  -
    96
u
z
u
)—I
u
I—1
Un
r^
w
                                                     Silica, d   = 2. 3)U

    80
                                                    3               4

                          NUMBER OF STAGES

      Fiaure  5   Effect of Series Electrostatic Spraying Stages  or. Overall

                 Collecting Efficiencies
OJ
ro
-J

-------

99 9
99 8
7 995
W
i — ;
(J QQ
a 99
t-i QQ
ri ' °
O
^ QS
H 95
O
Cd
3 90
^
0
u
80-







- - -
i
•

! , :
i
:.:-! :;'
i
! ' "
j
i
o



: : i
J
i
::i •''
'. \ . .
j
^f
8
:






^ C


: " ' . ; i
• \ '• i \
\
. '•!..!.
3?
r
3
PLATE
6


ox-"X<


2^'
D
SPACING
' •
. .
!
X^ -50 KV
; i • I ^^^'
^
'• \ '• '
', :
- ': ' \
. \
,. INCH., .at 4 fps
5







4-

o
+ 40 KV :
1

;
i
i
 . 06
08
. 10
. 12
. 14
. 16
                                             . 18
                                            . 20
                SPECIFIC COLLECTING AREA,  ft /cfm
                                              22
                                                                                  ho
                                                                                  OO
Figure 6  Effect of Plate Spacing, Residence  Time,  and Voltaae
          on 3-Stage Collection Efficiency  for  1.8p  Mass  Mean'
          Mineral Dust

-------
                                                                329

are fundamentally bounded by the collector plate treatment
height of eight feet, which was selected to meet the attractive
commercial design requirements of preassembled, truck shippable
structures and access for maintenance from a single work plat-
form.  Depending upon the characteristics of the process gas,
which in turn affects the final plate spacings, this dimension
allows either three or four electrical spray stages in series
and, primarily, sets the baseline collecting efficiency.  Greater
efficiencies are achieved by concurrently selecting a plate
spacing and gas velocity.

    The collector plates can be arranged on 5 inch to 8 inch
centers in increments of 1 inch to make the first efficiency
adjustment.  Operating voltages are set for spark rates of
100 to 250 arcs/min.  Greater spark rate frequencies allow
momentary field collapse and gas bypass.  The possible benefits
of higher voltage are consequently limited by the controlling
spark rate criteria.

    Gas flow capacity for a single module is governed by the
plate spacing, the plate width and the gas velocity.  Again
considering the limitations imposed by contemplated truck
transportation, the  collector plate width has been fixed at
10  feet.  This selected design dimension results in a  nominal
modular flow capacity of 2400 cfm at a base design plate
spacing of 8 inches  and a gas velocity of approximately
6 feet per second.   All pilot CDS  (1000 cfm nominal rating)
incorporate the basic CDS electrical configurations consequently,
there is no electrical scaling required to achieve full scale
equipment design from pilot test data.  In full scale equipment
the flow capacity is controlled by preselected groups of modules
arranged in parallel to conform to the standard equipment model
sizes.

RECENT DEVELOPMENT AND OPERATING EXPERIENCE

    The operating experience  (pilot and commercial) with the
first generation three stage CDS has been previously reported,
References 1 and 3.  This work identified a need  for more
economical equipment for independent and simultaneous fine
particulate and noxious gases emissions control.  Fortunately,
these requirements need not be mutually exclusive (if the system
is wet).   Indeed, similar changes  in equipment configuration  and
operation  are  required.

     •  Higher  liquid-to-gas ratios  (longer  liquid surface
       residence times).

     •  Extended, wetted  surfaces for absorption/space
       charge  collection.

-------
330

     The first change to the baseline three stage design was to
 add a fourth stage within the same envelope.  This is possible
 because the requirement for higher specific plates areas
 dictates closer plate spacing which, in turn, allows closer
 stage spacing without space charge interactions.  A four stage
 commercial pilot is shown in Figure 7,  with its auxiliary
 equipment.

     Simulation tests with submicron iron oxide fumes substanti-
 ated improved collection efficiencies of the four stage unit.

     The following pilot and commercial  applications experience
 has subsequently been acquired on fine  particle control.

     •  Pilot scale urea prilling tower.

     •  Commercial scale flue gases containing varying amounts
        of S02-

FINE PARTICLE CONTROL

     Typically,  the most difficult emission  control  requirement
is   to  meet stack discharge opacities when  substantial  fractions
of  micron sized particles  are present.

     Urea  prilling is  a  direct contact evaporative process  for
crystallizing urea droplets.   Drying is  accomplished by a  long
gravity free-fall between  the injection  point and the collection
point.  Rising  drying air  entrains  sub-micron urea  particulate
and emits  fume  to the atmosphere.   Aerodynamic mass  mean
particle  sizes  are in the  range'of  .6 to 1.2  microns,  (Figure  8)
Dust loads  are  typically in the  range of .012  and .038 gr/dscf,
substantially out of  compliance  as  opacities  exceed  60%.   The
urea is very  hydroscopic and  soluble  in  water.

     A four  stage  pilot  scale  TRW CDS was  installed at a typical
site  for manufacturing  urea by the prilling process.
The  pilot CDS was  essentially  identical  in its  internal design
to  a  full scale,  commercial unit, thus,  no performance scale-up
is  necessary.  Previous experience indicated  that placing  the
collector plates  on 6 inch  centers would  achieve the  desired
performance requirements.  The performance was varied by
adjusting the gas velocity between the collector plates.
Breeching between the CDS and the process was  accomplished
through a length of about 150 feet and 12 inch pipe.  No attempt
was made to extract an  isokinetic sample  since the micron
sized particulate should not stratify.  The EPA Method 5
procedure was used to measure inlet and discharge particulate
concentrations.

    The test results have been generalized in a plot of

-------
                                                                   331

   Feed Water
   Isolation


                                        Ports for Measuring
                                      /'Stratification Effects
                                         41 x 9' Collector Plate
                                              Electrostatic Spraying
                                              "tages
Low Ripple DC
Power Supply
      Figure 7  Pilot CDS, Full Scale Four-Stage Module Use for Fine
                Particulate and Gaseous Emissions Development Program

-------
                                                                                                        u>
                                                                                                        OJ
                                                                                                        10
i
u
Q
s
INLET SAMPLES
   © 1-1
   V 1-2
     1-3
     1-4
     1-5
                                                                  CDS
                                                                  OUTLET, .0031 GR/SGF
                                                                  TEST NO.  13. 6/13/76
   0.1
                      2   5   10   20   30 40 50  60  70  80   90  95  98
                       COMMULATIVE WEIGHT FRACTION LESS THAN CUT  DIAMETER
     Figure  8   Urea Prilling  Tower Particle  Size  Distribution

-------
                                                                333
efficiency versus specific plate area (velocity), Figure 9.
Efficiencies can be varied from 70% at a specific plate area and
gas velocity of about .063 and 8.8 fps,  respectively to 95% by
derating velocity.  The most cost-effective operating point on
the curve is near its knee at about a .08 specific plate area.
The pilot CDS stack discharge was substantially clear at this
operating point.  A downdraft configuration was also tested at
the design efficiency to determine if performance improvements
could be achieved.  About a 2 to 3% efficiency increase at a
lower electrode waterflow rate (about 25% reduction) was
measured.  The higher efficiencies are attributed to the fact
that the lower water flow rates result in larger residence times
of the water columns in the electrostatic field.  This results  in
a smaller mean droplet size and improved  collection efficiency of
submicron particulate (see Figure 4).  This configuration may
also offer some significant economies for ground level installa-
tions by reducing the amount of duct-work required, particularly
for an induced draft fan configuration.  Higher electrode
operating voltages  (38 to 40 kv) improved efficiency at the
design gas flow velocity.  Utilities requirements at the design
point were:

    •  Electrical,  .4 watts/acfm

    •  Water,  .46 gpm/1000 acfm

    Process gas flange-to-flange pressure drops were nominally
0.5 inches we and recycled water conductivity varied from  200 to
2400 micromhos/cm.

GASEOUS EMISSION CONTROL

    The disadvantages of a wet system (disposal of  a sludge)  can
be turned to advantage if simultaneous,  selective absorption  of
noxious gases  can be effected.  A two-part development program
has been completed  to evaluate the  feasibility of using the CDS
to absorb the  chemically active  (HCl) and inactive  (SO2) gases.
Chemically active systems require high mass transfer rates
through the gas film boundary layer, inactive systems require
improved mass  transfer rates into the liquid surface.  Typical
process conditions  are combustion flue gases containing 400 to
2000 ppm noxious  gases such as S02-

    For the reactive HCl system investigation a modified pilot
scale CDS  (four electrostatic spray stages and plates on 5  inch
centers) was used.  A hydrogen chloride  high pressure cylinder
was connected  to  a  trace gas injector installed within a
mixing chamber.   The hydrogen chloride gas was distributed
through three  horizontal tube runs  containing 15  equally spaced
orifices.  Sampling ports were  located in the vicinity of  the

-------
334
     >1
     u

     <1>
     •H
     O
     •H
     4-1
     W
•°6     -07     .08     .09     .10     .11     12
      Specific Collecting Area, A/Q, ft2/acfm

 987           6               5      4.5
             Collecting Velocity, fps
                                                               .13
          PROCESS SPECIFICATION

               Dust Load, gr/dscf
               Temperature, °F
               Humidity, %
               Particle Size - mean, y
                             - % < ly
               Particle Composition

          CDS DESIGN AND OPERATION

               Plate Spacing, inch
               Gas Velocity, fps
               Transformer - Voltage, KV
                           - Power, va/cfm
               Number of Stages
                                      0.012 -  0.038
                                         Ambient
                                           100
                                        0.7 -  1.2
                                         30 -  60
                                          Urea
                                         50 - 55
                                        0.4 - 0.8
                                            4
        Figure  9   CDS  Pilot  Performance  Urea  Prilling  Tower

-------
                                                                 335
inlet to the CDS where stratification of the injected gas was
anticipated to be small.  The CDS discharge sampling points were
located at ground level just before the CDS induced draft fan.

    The primary procedure for determining hydrogen chloride
absorption rates was to sample simultaneously at the CDS inlet
and outlet.  The sample was dry gas metered and drawn through a
train consisting of three fritted glass impingers in a sodium
hydroxide solution.  This solution was then analyzed for chloride
by titration of a known silver nitrate solution.  The correlation
between metered to sampled inlet hydrogen chloride concentrations
was usually within ±20%.

    Tests were initially run to determine if sufficient wet
contacting and absorption could be accomplished using electro-
static water spraying and plate water films.  The tests were run
at two space velocities (3.5 and 5 fps) resulting in liquid to
gas ratios of 1.26 and .9, respectively.  The absorption is
substantially removed (order of magnitude) from equilibrium
saturation suggesting that the system is mass transfer limited.
The data indicate that the absorption efficiencies are sensitive
to inlet HCl concentration with the higher efficiencies being
accomplished at the lower inlet loads. Figure 10.  This would
suggest that mass transfer rate is limited in part by the water
droplet film.  Higher liquid rates improve efficiency about 26%,
in proportion to the increase in liquid to gas ratio.

    To improve collection efficiencies by increasing the absorp-
tivity of the liquid, a weak sodium hydroxide solution was
applied through the plate wall wash system.  The liquid to gas
ratio for the plate wall wash is a factor of three higher than
the electrode water so the comingled electrode spray water does
not substantially dilute the concentration of sodium hydroxide.
Total wetted surface area is not affected by this change so
improvement in absorption must be attributed to improved liquid
film mass transfer.  This configuration allows recirculating the
sodium hydroxide liquor through the wall wash system until a
substantial fraction of the NaOH is consumed.  Further efficiency
improvements of up to 95% were achieved.

    Interfacial CDS velocities are similar to those used in
packed bed absorbers allowing a simple modification to affect a
low pressure drop series equipment arrangement.

    The current pilot and commercial configurations of the CDS
have about 2 feet of open space between the gas distribution
baffles and outlet flange within the lower section.  This void
space can be converted to a prescrubber by the addition of a
non-plugging type of packing.  In this case, the upper stage

-------
336
   99,6

   99.0
0°
 * 9o.O
6
U2  95.0
u
z
o
         90.0


         co.o

         70.0

         60.0
         50.0
           2000
SYM
V
A
0
Q
»
VELOCIiY
_fZi_
5
5
3.3-S
5
5
i-ACMNG
ft
2
1
0
0
0
                                         V
                                               A
                                         Oo.5%NaOH
                                                    0.90
                 1000         500      300    200
                   INLET CONCENTRATION, ppmV
100
PROCESS  SPECIFICATION
    HC1  Concentrations,  ppmV
    Temperature,  °F
    Humidity,  %
                                     180  -  1000
                                      80  -  185
                                      20  -  30
CDS  DESIGN  AND OPERATION
     Plate Spacing,  inches
     Gas Velocity,  fps
     Transformer -  Voltage,  kv
                 -  Power, watt/cfm
     Number  of  Stages
                                     3.5  -  5
                                       50
                                     0.9  -  1.2
                                         4
     Figure  10   HCl  Absorption Efficiencies for Several
                 Equipment Configuration and Operating Variables

-------
                                                                      337
                                                      COKC-OVEN GAS SCRUBBING
                                                      KILN FLUE GAS SCRUBBING
                                    Pscy MM HG
Figure  11  Preliminary Data Showing Unusual  S02-Absorbing Capability
            of CDS

-------
 338

 electrostatic spraying discharge water should be  separated  from
 the  post-particulate scrubbing section to minimize  the  suspended
 solids  clean-up for the recirculated sodium hydroxide solution.
 Liquid  to  gas ratios must be increased to about four, with  the
 packed  bed being operated  at a L/G of 3.  The total system
 pressure drop with the maximum packing depth of 2 feet  was
 1.5  inches we.   These test results, Figure 10, demonstrate  that
 HC1  absorption efficiencies of 99+% are achievable  at 5 fps space
 velocity and  HCl input of 1000 ppmV.

    SO,  has an extremely low solubility in water when an
 absorBant  such as an alkali salt is absent.  Routine testing of
 CDS  controlling particulate emissions from various  combustion
 process off gases containing S02 have indicated absorption
 efficiencies  which substantially exceed saturation  by an order
 of magnitude  (Figure 11).   These data suggest that  electro-
 static  phenomena may be enhancing the reactivity at the liquid
 interface.  Hypothesized chemistry includes the formation of
 activated  states of S02 and 02 in the corona fields.  Research
 and development to understand and exploit this chemistry is
 currently  in  progress.

 REFERENCES

 1. R. R. Koppang,  "POLLUTION CONTROL EQUIPMENT AND
    TECHNOLOGY,  Broader  Availability of Electrostatic
    Precipitators - A New Hybrid EP Concept,  The TRW Charged
    Droplet  Scrubber", TRW Inc.,  Energy Systems Group, 1975.

 2. c. W. Lear,  "Charged Droplet Scrubber for Fine Particle
    Control:   Laboratory Study",  TRW Systems Group,
    EPA-600/2-76-249a.

 3. W. F. Krieve  and J.  M.  Bell,  "Charged Droplet Scrubber
    for Fine Particle Control:   Pilot Demonstration", TRW
    Defense  and  Space Systems Group,  EPA-600/2-76-249b.
WRITTEN DISCUSSION

Michael J. Pilat
Department of Civil Engineering
University of Washington
Seattle, Washington

   The paper presents the description and performance of a space charged
scrubber-precipator named the "TRW Charged Droplet Scrubber." My comments

-------
                                                                             339
concerning the paper are mainly in the form of questions; these questions
being posed in order to better understand the scrubber system.  My questions
are as follows:

    1.  On page 3 a voltage gradient of 2.3 kV/cm is mentioned as necessary
to achieve electrostatic spraying.  Then on page 6 is it mentioned that in-
creasing the water conductivity from 300 to 10,000 micromhos/cm requires a
reduction in the electrical voltage of 2 to A kV.  Why is this voltage re-
duction necessary?
    2.  In Fig. 4 what are the definitions of P^) , A^) , P(SQ), and A(SQ)?

    3.  In Fig. 5 what are the units and definitions of A/Q?

    4.  The water to gas flowrate ratio of 0.46 gal/1000 acf seems somewhat
low (page 14).  Is this the water requirement for one stage, or is the water
usage not including water recycling?

    5.  The electrical utility requirement is listed as 0.4 watts per acfm.
Does this electrical usage include power required for water pumping, the  fan,
and the high-voltage power supplies?

    6.  Is the particle collection efficiency (either overall or at a- specif ic
particle size such as at 0.5 micron diameter) of the CDS sensitive to the inlet
particle mass concentration?  The collection efficiency data presented  (Fig.
5, Fig. 6, and Fig. 9) only mentions the area prilling concentration in the
0.012 to 0.038 gr/scf.

    For a comparison with the TRW CDS  system, let us look at the U. of W.
Electrostatic Scrubber  (UWES) .   A schematic illustration of the UWES  is provided
in Fig. 1.  The UW Electrostatic Scrubber  involves the use of electrostatically
charged water droplets  to collect air  pollutant particles electrostatically
charged to a polarity opposite  from the droplets.  The particles are  electro-
statically charged  (negative  polarity) in  the corona section.  From the
crona section  the gases and charged particles flow into a scrubber chamber into
which electrostatically charged  water  droplets  (positive polarity) are  sprayed.
The gases and  some entrained  water droplets flow out of the spray chamber into
a most eliminator consisting  of  a positively charged corona section in which
the positively charged water  droplets  are  removed  from the gaseous stream.

    The general  layout  of  the UW Electrostatic  Scrubber pilot  plant  (Mark 2P
model) is  shown  in  Fig.  2.  The system (in the  direction  of gas  flow)  includes
a gas cooling tower,  an inlet test duct with  sampling port, a  particle  charging
corona section (corona  no.  1),  a charged water  spray  tower  (tower no.  1),
a particle charging corona section  (corona no.  2),  a  charged water  spray tower
 (tower no.  2),  a positively charged  corona section to collect  the positively
 charged water droplets,  an outlet  test duct with sampling port,  and  a fan.
The pilot  plant  is  housed in  a 40  foot long trailer and  can be easily trans-
 ported to  emission  sources.

     In Fig.  3 the particle collection efficiency as a function of  particle size

-------
         340
GAS INLET
                                                                                GAS  OUTLET
              CORONA
          (PARTICLE CHARGING)
 CHARGED WATER  SPRAYS
(COLLECTION OF  CHARGED PARTICLES
 BY OPPOSITELY  CHARGED WATER  DROPLETS)
MIST ELIMINATOR
                         Fig.  (1) UW Electrostatic Scrubber

     is presented  for emissions from an electric arc steel furnace.  Note that the
     specific collecting are.a of the UWES is  .05 ft2/acfm which is somewhat less
     than the .08  to  .22 ft /acfm presented by Koppang for the CDS system.  Of
     course this illustrates the difference in the particle collection operating
     mechanisms utilized, the UWES using charged water droplets to collect the
     particles and the CDS having a greater collection plate area.  Fig. 4 and
     Fig. 5 present data obtained at a pulverized coal-fired power plant.  Fig. 4
     illustrates the increased collection efficiency obtained at the higher water
     charging voltage.  Fig. 5 presents the effect of water to gas flowrate ratio
     on the collection efficiency.

        For additional information on the UWES system, see the references listed
     below:

     [1] Pilat, M.J., S.A.  Jaasund, and L.E. Sparks (1974) "Collection of Aerosol
        Particles by Electrostatic Droplet Spray Scrubbers" Envir. Sci. & Tech.
        8 340-348.

-------
INCOMING
 GASES
                       INLET TEST DUCT
                           SPRAY TOWER NO. 2
                                CORONA NO. 1
                                                            EXHAUST
                                                             GASES
                                                                   SECTION  A-A

                                                             CROSS SECTIONAL VIEW OP
                                                          THREE PASS HORIZONTAL  SECTION
                                                                                                   OUTLET TEST DUCT
                                                                 MIST
                                                              ELIMINATOR
                              SPRAY TOWER NO. Z
SPRAY TOWER  NO. 1
                                                                                             CORONA NO. 2
                                                             ELEVATION  VIEW
                                                                                                                         FAN
                             Fig.  2   General Layout  of  Electrostatic  Scrubber Pilot Plant  (Mark 2P Model)
                                                                                                                             UJ

-------
     342
         [2]  Pilat, M.J.  (1975) "Collection of Aerosol Particles by Electrostatic
             Droplet  Spray Scrubber" APCA Journal 25 176-178.

         [3]  Pilat, M.J.  and D.F. Meyer (1976) "University of Washington Electrostatic
             Spray Scrubber Evaluation" Final Report on Grant No. R803278,  EPA Report
             No. EPA-600/2-76/100 (NTIS No. PB 252653/AS).
 Particle
Collection
Efficiency
              100
               95
               90
               85
               80
               75
   I IMIIII| I  I |||  |  |  i

- Electric Arc Steel  Furnace

               Test 23
                      s
                      /
                      Test  22
                                                 Gas Velocities  (ft/sec)_
                                           Corona section   4.8
                                           Spray towers     2.4

                                           SCA =0.05 ft2/acfm
  Test
  No.
   22

   23
            Water  to Gas
               Ratio
            (gal/1000acfO

                17.2

                16.8
Overall Mass
     Eff.
    98.6

    96.4
Corona
Voltage
 (kV)


  70

  70
Liquid
Voltage
 (kV)


  10

  10
' MHIllll I  Mil  I  I  I  I I  lilililil  i i.i..ml MM!   i  III,  III,I,[7
                 .1
                            •2     .3    .4   .5  .6   .-8  1.0        2.

                                        Particle Diameter (microns)
                                                      3.   4.  5. 6.   8.  10.
               Fig.  (3)  Collection Efficiencies With Similar Operating Conditions
                         At Electric Arc Steel Furnace

-------
           100
            95
Particle     90
Collection
Efficiency
            85
            80
—  Cool-Fired  Boiler
    Test No. I
    Overall Eff. = 99.555
                    SCA = 0.05  ft2/acfm
                  IIMM
                                                    'Cool-Fired  Boiler  Test  No. 2
                                                           Overall Eff.  =  98.1%
                                 Notes:  I. Water droplet  charging  voltage
                                           = 20  kV on test no. I and  15 kV   -
                                           on  test  no. 2.
                                        2. All other pilot  plant operating   —
                                           parameters  essentially  unchanged
                                              Illllllll I  Illllllll Mill  I  I  I  I  I Illllll'
                                                                                 343
              0.1
           .2     .3   A  .5  .6   .8  I.           2.

                        Particle  Diameter  (microns)
                                                        3.   4.  5.  6    8.  10.
            Fig.  4   Influence of Water Droplet Charging  Voltage on Particle  Collection
                     Efficiency
            100
            95
Porlicle     90
Collection
Efficiency
   (%)
            85
            80
            75
 TTTT M 11 11 l i TTT  I   I  I  I I  I I I  I I III iTTnT I TTT| I I  I I |   T
                   I     I    I   I  I   I  I I
                        "	*	
               Coal-Fired Boiler Test No.I
               Overall  Eff.  =  99.556
               Outlet Cone.  =  .002 grains/scf
          Cool- Fired Boiier  Test  No. 3
            Overall Eff. = 96.1%
1 1  1 1 1 1 1 1 1
                               ,
            1 1  1 1 1  1  i   I  i 1  1  1 1 1 1  1 1 1 1 1
r
 Notes:  I. Water  to  gas  ratio (gal./    ~
           1,000 acf ) = 5 82 in  test  ~
           no.I  and  2.36  in  test
           no. 3
        2. All other  pilot plant
           operating  parameters
           essentially  unchanged

            SCA = .05 ft2/acfm

     LUiu.lillll  I  I  l  I. Mill
              0.1
           .2     .3    rt   .5 .6'  .8   I.          2.

                        Particle  Diameter   (microns)
                                                                         3.    4.   5.  6.    8.  10.
             'Fig.  5   Influence of Water to Gas Flow Rate  Ratio on Particle Collection
                       Efficiency

-------
 344
RESEARCH ON NEW EQUIPMENT FOR DUST COLLECTION-7'

    A. Baril, Jr., D. P. Thibodeaux, and B. J. Claassen

    U. S. Dept. of Agriculture, ARS, SRRC, New Orleans, La.-'

    R. B. Reifl/

    Battelle Memorial Institute, Columbus, Ohio

INTRODUCTION

    OSHA cotton dust standards and the possible occurrence of
byssinosis in cotton textile workers has focused attention on
the dust problem in cotton textile mills and related opera-
tions where cotton and cottonseed are processed.  One of the
most urgent problems facing cotton textile mills is that of
removing resplrable dust from the work environment and main-
taining the dust level within prevailing OSHA standards with-
out a disturbing economic unbalance and with no increase in
energy consumption (I-1!).

    A wide variety of systems is currently being used to
clean the air in textile mills that process lint cotton. Most
of these are custom designed, and use proprietory equipment
such as bag filters, rotary drum filters, condensers, V cells
etc. (5), but many systems have Just evolved as a result of
additions and changes needed at the moment.  The better sys-
tems have several stages, including a waste separator that
removes lint and large trash particles, a dust separator
that can reduce dust content of the air to one mg/m3 by some
form of filtration, and a fine dust separator capable of re-
moving respirable dust (less than 15 aim).  The effective con-
trol and removal of dust in this size range is quite diffi-
cult and expensive.  Presently, there are few, if any, air
handling systems in the textile industry capable of econo-
mically and efficiently collecting and removing respirable
dust.  For the past several years, our laboratory has spon-
sored and carried out research to such a system.  The first
two devices described were developed by Battelle Columbus

-------
                                                                     345
   Laboratories  under  contract to U.S.D.A., and the third one
   was developed at  the  Southern Regional Research Center in
   New Orleans.

   Wet Wall Electroinertial Unit

       The first device,  called a wet wall electroinertial unit
   (WWEU) air  cleaner, was  developed to meet the requirements of
   high operating efficiency,  low maintenance, low operating
   cost, and simplicity  of  design.  To achieve high efficiency
   with low pressure loss,  an  electrostatic precipitation system
   was the best  candidate.   A  wet wall was desirable to flush the
   precipitated  material  away  and to minimize maintenance.
   Inertial effects were  added to assist in moving material
   toward the wall and to improve the cleaning action of the
   water on the  wall.
 air miet
                       • 3 B75in ID i 10m
                        Lucite luDe
                      • Insuiotor
                       Water inlet
    These features were best  incorporated in the design shown
in Figure 1.  The unit consists  of a concentric wire-in-tube
                               precipitator.   The charging
                               wire  is  located axially in a
                               stainless  steel tube in a ver-
                               tical orientation.  Air enters
                               tangentially  at the upper end
                               of  the tube (through a duct
                               with  a flattened  cross section)
                               and acquires  a rotating move-
                               ment.  When high voltage is
                               applied  to the wire, an ioni-
                               zing  corona discharge charges
                               the dust in the air.  As the
                               air flows  through the tube,
                               the charged dust is driven to
                               the wall by the radial elec-
                               trical field.   Water from the
                               upper water inlet flows down
                               the wall and  flushes the pre-
                               cipitated  dust into the water
                               outlet at  the  lower end of the
                               tube.  The rotational movement
                               of  the air also induces rota-
                               tional flow of the water,
                               which assists  in uniformly
                               wetting  the surface of the
                               tube.  Cleaned air is expelled
     3B7m ID t 26in
     Stainless steel lube
    30m 00 x 60in
      Bross tube
                       0008m Diameter
                      • Corono wire
                      . 'nsulatar
Waier outlet
                         exhaust
                                   from  the  tube's  lower end.
Fig. 1.  Wet Wall Electroinertial
         Unit

-------
346

    Tests of the WWEU were  conducted with  4  in. and  8 in.
diameter units at  flow rates ranging from  100 to  800
ft-Vmin.  Particle  collection efficiency of  these units was
measured with atmospheric dust, AC  fine test dustz/  (mass
mean diameter (MMD) = 12 jam), artificial cotton dust (MMD =
4.0 jam), and ootton dust drawn from the processing area of
a card  (MMD = 3.0jum).  Efficiency  data with atmospheric
dust was obtained by isokinetic-sampling downstream of the
unit with an optical particle counter.  Sample preparation
and testing procedures with the AC  test dust and artificial
cotton  dust have been described previously (6).  Essentially,
known quantities of these dusts were fed to  the WWEU and all
particles that escaped were captured on an absolute filter
installed downstream of the unit.   Preliminary results are
also included on tests with an 8 in. WWEU that cleaned dust
laden air drawn from a full scale cotton card operating at
20 Ib/hr.

    Tests were conducted with the 4 in. diameter WWEU to
determine the effects of tube length, air flow rate, and
voltage polarity on efficiency of particle collection.
Tube lengths of 1, 2, and 4 ft were tried.   Shearing of
water droplets from the wall at higher velocities made the
practical upper limit on flow rate  about 200 ftVmin.  Re-
sults of the efficiency tests with  the 2 ft  unit are inclu-
ded in Table I.  Runs made at zero  potential and zero cur-
rent with the rather coarse AC dust show efficiencies in the
BQ% range due merely to centrifugal forces.  Increasing the
voltage to -30 kV raises the efficiency to 992.  Which
diminishes to 98.5$ when +33 kV is  applied.  Tests with the
fine artificial cotton dust Indicate a somewhat lower effi-
ciency  (98.1 and 96.15? respectively) for 100 and 200
ftVmin.

    Table II contains the data for tests with the 8 in. dia-
meter, 66 in.  long WWEU.   These data show that the 8 in.
unit can be scaled up from the 4 in. WWEU, maintaining the
same efficiency levels for a higher flow rate.  While scal-
ing up the capacity of the units, the velocity and field
strength were maintained at the same values.   In this case,
particle collection efficiencies are still better than 95?.
Differences in efficiencies with the artificial and real
cotton dust for zero voltage are no doubt caused by the
relative fineness of real cotton dust,  which diminishes the
effects of centrifugal forces.   More recent studies with the
8  in.  unit to clean air from an actual textile card show
that the unit  operates quite satisfactorily under realistic
test conditions.

-------
                                                                 347
                          TABLE I

           TOTAL MASS EFFICIENCY OF THE 4 IN. WET
               WALL ELECTROINERTIAL UNITS./".'
Test Circulation Poten- Cur-
Dust Rate tial rent
ftVmin

AC fine
AC fine
AC fine
AC fine
AC fine
Cotton dust
Card trash
Cotton dust

100
200
100
200
100
100
200
200
kV

0
0
-30
-30
+ 33
-30
-30
-30
mA

0
0
2.4
2.1
— —
2.2
2.2
2.1
Dust
Feed
Rate
g/min

1.0
2.0
0.2
0.2
1.0
0.83
0.67
0.62
Pres-
sure
Drop
in.
H20
1.25
3.75
1.25
1.25
1.25
1.25
3.75
3.75
Efficiency
%

81.2
87.6
99.0
99.0
98.5
98.1
99.7
96.1
a/  Length of WWEU was 2 ft.
b/  Water flow was 0.25 gal/min.
                         TABLE II

           TOTAL MASS EFFICIENCY OF 8 IN. WET
                WALL ELECTROINERTIAL UNITa/
Test Dust

AC fine dust
Card trash
Card trash
Artificial
cotton dust
Artificial
cotton dust
Card output^./
Card output!!/
Card output*!/
Poten-
tial
KV
-60
-60
0

-60

0
0
-58
+60
Cur-
rent
ma
^mm
-7
0

-7

0
0
-6
+5-8
Water
Flow
g/min
0.46
0.46
0.46

0.46

0.46
0.64
0.65
0.65
Air
Flow
rt -vmin
800
800
800

800

800
800
800
800
Efficiency
*
97.6
98.4
99.8

99.4

80.2
30.3
99.9
96.9
a/ Length of WWEU was 66  in.
b/ Tested at SRRC on output of V-cell  filter.

-------
  348
   High  Speed  Rotary  Drum Air Cleaner

      The  air cleaning systems  of most  textile mills  utilize a
   rotary drum (low speed)  filter that  allows  the unit to con-
   tinuously self-strip the accumulated  filter cake.   This pro-
   cedure allows  for  a  low pressure drop in the air circulating
   system but  leads to  inefficient dust  removal, particularly at
   the respirable dust  range.  Investigation of these  cleaners
   with  the intention of increasing their efficiency  led to ap-
   plication of the principle that movement of a collection
   surface  normal to  the direction of the air  flow through it
   increases the  effective  interception  cross  section  of the
   collector surface.   If the collector  surface moves  at a
   velocity n  times that of the  air stream, the number of par-
   ticles impacted on the surface is  increased by a factor of
   between  n and  n +  1.
                               Air Inlrt
                      12" x 12 x 17"
                      Luclte enclosure
   Plow orifice
RLower Intake
          I
                                The apparatus
                            shown in Figure 2
                            was constructed to
                            evaluate this con-
                            cept.  A hollow
                            aluminum cylinder
             ,	was mounted coaxi-
              eSSeSSeS I I  I]ally on a drive
              	rrHshaft on one end
                            and was connected
                            via a large bearing
                            OOOOOOOO
                            lOOOO OOOO
     Pig.  2
High Speed Rotary
Drum Air Cleaner
                                   lolorv drun  tO  Si  3
                                         meter suction  pipe
                                         on  the  other end.
                                         The surface of the
                                         aluminum  cylinder
                                         was perforated
  , .                                     with  holes arrang-
ed in equally spaced rows parallel to the axis  of the  cy-
linder.  Filter materials were wrapped around the drum to
cover the holes, and the drum was  rotated to  increase  the
effective cross section of the fibers.   In  operation,  air
was drawn in through the holes and out through  the 3 in.
diameter suction pipe by a high pressure blower as the cy-
linder was rotated.  The efficiency of the  rotary drum unit
was evaluated for several configurations, such  as hole size,
open area, rotational speed, flow rate,  and cylinder dia-
meter.  The drums were evaluated with and without filter
media on the surface, and some of the drums were  evaluated
with electrostatic assist.

-------
                                                                349
    Initially, the efficiency of the rotary drum unit was
evaluated by using a particle monitor to measure the parti-
culate concentration in the air in the room outside the
drum and in the air at a point Inside the drum.  During
this work, some variation was observed depending on the
position of the sampling probe inside the drum.  These data
generally reflect relative dust loading differences in the
portion of the air flow sampled and do not indicate overall
efficiency for the total air flow.  The true efficiency of
the unit was measured with AC fine test dust with a fiber-
glass filter installed downstream of the rotary drum to col-
lect all of the dust that passed through it.

    The  first drum tested was a 4.5 in. OD  x 10 in. long
cylinder with a 5/32 in. wall.  Sixty  1 in. diameter holes
were drilled in ten equally spaced rows, with six equally
spaced holes in each row to provide a  total open area of
0.327 ft2.  The partial efficiency was remarkably high
with the sampling probe inside the drum at  a position on
the axis of the drum and 3 in. from the bearing.  As shown
in Table III, from 58? to 93? of the atmospheric dust was
removed  from the air passing through this area when the
drum was rotated at a surface speed of 3000 ft/min with
air flow rate of 65 ftVmin.  The efficiency increased
with drum speed up to about 3,000 ft/min; no further im-
provement was attained at 4,000 ft/mln.  The efficiency
decreased as the air flow rate was increased to 200 and
300 ft^/min.  Particle counts outside  the drum showed
increased dust concentration In the air around the drum.

    With a layer of polyester filter material covering
the drum, the partial efficiency of the system was general-
ly higher than that of the plain drum  in the smaller
particle size ranges.  The efficiency  in the larger parti-
cle size ranges is quite good O90JS),  but is not signifi-
cantly better than for the plain drum.  Covering the drum
with a PC (Pneumafil Corp.) second stage filter gave
almost the same efficiencies as the aplin drum, whereas
covering with a PC first stage filter  resulted in a de-
crease in efficiency of the plain drum.

    Two types of electrostatic systems were evaluated in
conjunction with the rotary drum with  1 in. holes.  In the
first system, electrical fields were applied to the drum
from a 1/4 in.  hardware cloth cylindrical  grid, spaced
1 in. from the drum.  With potentials  up to 20 kV on the
grid, the efficiency of particle removal with the plain

-------
 350
                        TABLE III
 PARTIAL EFFICIENCY OF ROTARY DRUM WITH 1-IN. DIAMETER HOLES
Drum
Speed
ft/min
Flow
Rat e,
tVmin
               Cover
Efficiency, percent
ToT5       > 5.0
micron      microns
ft
1,000     65      None                  H.l
2,000     65      None                 H2.3
3,000     65      None                 58.2
4,000     65      None                 51.21

1,000    200      None                  3.1
2,000    200      None                  9.3
3,000    200      None                 16.1

    0     65      E2B-P01B             16.9
1,000     65      E2B-P01B             63.8
2,000     65      E2B-P01B             75.7
3,000     65      E2B-P01B             82.6

    0     65      PC, first stage       1.1
1,000     65      PC, first stage      29.3
2,000     65      PC, first stage      41.6
3,000     65      PC, first stage      51.8

    0     65      PC, second stage      3.0
1,000     65      PC, second stage     48.9
2,000     65      PC, second stage     59.3
3,000     65      PC, second stage     61.1
                                          23.5
                                          85.9
                                          93.1
                                          93.5

                                          12.7
                                          55.4
                                          60.8

                                          49.1
                                          95.4
                                          95.4
                                          94.6

                                           3.9
                                          81.4
                                          83.0
                                          86.9

                                          32.7
                                          90.6
                                          9.3.5
                                          93.4
drum increased to the level achieved with filters covering
the drum.  With the external field and filters on the drum,
the increase in removal of submicron particles was about
10% to 2058.

    In the second electrostatic system, a series of 0.005
in. diameter charging wires were spaced at about 2 in. in-
tervals parallel to and 1 in. from the drum.  Data in
Table IV show that the partial efficiency was higher with
auxiliary electrostatic charging.  With the drum station-
ary, the charging system alone removed 56!? to 73% of the
atmospheric dust at 65 ftVmin; however, when the flow
rate was increased to 200 the efficiency decreased
although not as much as without the electrostatic assist.

-------
                                                                351
At low flow rates, the good performance of the E2B-P01B fil-
ter was improved considerably by adding the electrostatic
assist.

                         TABLE IV

     PARTIAL EFFICIENCY OF ROTARY DRUM (1-IN. -DIAMETER
        HOLES WITH AUXILIARY ELECTROSTATIC SYSTEM)
Drum
Speed
                                            Efficiency
 Flow
 Rate
ftVmin
                   Cover
           Poten
            tial
          > 0.5
          micron
> 5.0
microns
ft/min

    0
    0

1,000
2,000
3,000
4,000

1,000
2,000
3,000

    0
    0
1,000
2,000
3,000
                               kV
            65
           200

            65
            65
            65
            65

           200
           200
           200

            65
            65
            65
            65
            65
None
None

None
None
None
None

None
None
None

E2B-P01B
E2B-P01B
E2B-P01B
E2B-P01B
E2B-P01B
-20
-20

-20
-20
-20
-20

-20
-20
-20

-10
-15
-15
-15
-15
                                          55.9
                                          11.7
                                          62
                                          76
                                          85
                                          45,
                                          49
                                          75.6
                                          49.0
                                          14,
                                          40,
                                          84,
                                          89,
                                             73.0
                                             52.8

                                             82.7
                                             90.1
                                             96.1
                                             95.1

                                             66.6
                                             79.6
                                             77.3
                                             20
                                             46
    ,6
    ,9
                                          92.3
                                             96.3
                                             96.3
                                             96.4
    The second drum tested (45 in. OD x 10 in.long with
5/32 in. wall) had 897 holes 1/4 in diameter drilled in 39
equally spaced rows with 23 equally spaced holes in each
row to provide a total open area of 0.305 ft2.  This drum
was considerably more efficient than the drum with 1 in.
holes for removing atmospheric dust at lower speeds.

    A third cylinder (same dimensions in the first 2) was
made with additional 1 in. diameter holes to provide more
open air:  91 holes were arranged in 13 rows of 7 holes
each to provide an open area of 7 holes each to provide an
open area of 0.496 ft2.  A flow rate of 65 ft3/min produced
a gas velocity of about 130 ft/min through the holes in
this drum, as compared with velocities of 200 to 210 ft/min

-------
 352

 with the  first two.   In general, performance of the drum with
 91  1-in.  diameter holes was erratic.  Efficiency usually im-
 proved when  surface speed was increased and worsened when air
 flow was  increased.   However, the overall performance was
 inferior  to  that of the drums with fewer or smaller holes.

    Tests with AC fine dust were run with the drum enclosed
 in  a 12 x 12 by 17-in. Lucite box with a 4 in. diameter air
 inlet in  the center of the top of the box, as shown in
 Figure 2.  A 1 in. diameter spherical deflector disk was
 mounted about 6 in. below the air inlet to distribute the
 incoming air around the box.  AC fine dust was fed into a
 4 in. diameter pipe connected to the inlet of the box by
 means of a small rotary platform feeder.  Efficiency was
 determined by collecting material that passed through the
 drum on a 1/2 in. PF  105 Fiberglas filter downstream of the
 drum.  Data  from these tests run on the drum with both the
 1 in. and the 1/1 in. holes is shown in Table V.  For the
 drum with the 1 in. holes, the mass efficiencies agree
 quite well with the partial efficiencies for atmospheric
 dust.  The mass efficiencies for the drum with the 1/4 in.
 holes do not agree with the partial efficiencies obtained
 with the optical counter since there is no evidence of a
 maximum at the low flow rates.

                         TABLE V

      MASS EFFICIENCY OF THE ROTARY DRUM WITH AC DUST
Drum
Speed
ft/min
3,000
3,000
3,000
325
805
1,530
3,000
Hole
Diam. ,
in.
1.0
1.0
1.0
0.25
0.25
0.25
0.25
Flow
Rate,
ftVmin
65
65
65
65
65
65
65
Cover

None
None
PO-1B
None
None
None
None
Poten-
tial,
kV
0
-20
0
0
0
0
0
Efficiency
i
90.0
93.0
98.0
57
72.3
80.8
90.3
    The various measurements and observations made while
working with the rotary drum Indicate that separation of
particulates occurs externally and internally.  Some par-
ticulates are rejected from the air stream at the surface
of the drum;

-------
                                                                353
as a result, the partlculate concentration in the air
around the drum increases.  At the same time, particulate
concentration inside the drum varies at different points.
Lowest particle concentrations generally occur along the
axial center of the drum.  At first the air stream was
sampled along the axial center of the drum to avoid partl-
culates generated by the bearing or introduced by air
leaks through the unsealed bearing on the exhaust end of
the unit.  However, as work progressed, evidence indicated
that particulate concentration in the air inside the drum
was higher in off-axis positions.  A functional design thus
must provide means for removing the dust-laden air outside
the drum and for separating the clean air from the dust-
laden air Inside the drum.

Fluid Electrode Precipltator

    In general, the best way to remove fine dust from an
air stream is to use conventional electrostatic precipita-
tors.  Although the efficiencies of these devices are quite
high, there are inherent drawbacks that do not make them
acceptable in continuously operating air systems.  One of
the major drawbacks to self-cleaning electrostatic precipl-
tators is that they must be shut down in order to clean
themselves.  The development of the fluid electrode preci-
pitator is an attempt to alleviate this problem.  The fluid
electrode precipitator illustrated in Figure 3 consists of
an array of electrodes in a casing to charge, direct, and
collect fine dust.  Tungsten wires serve as discharge elec-
trodes for all units tested, auxiliary driving electrodes
are charged metal plates or screens that can be inserted
between the fluid electrodes, and the collection electrodes
are formed by vertical tubes that release a laminar flow of
water falling through the height of the casing.

    The precipitator operation is quite simple.  The  fall-
Ing columns of grounded water act as cylinders in cross
flow and create vorticies for enhanced particle collection.
Auxiliary electrodes and discharge electrodes can be  posi-
tioned in arrays to direct charged dust into the grounded
fluid.

    The use of free-falling  fluid results in a system that
is  less complex than rigid surfaces with wetting devices.
Shear forces due to air  flow over a wetted surface can
severely disrupt the thin layer  of  fluid flow.  A cylin-
drical fluid column is a  stable  shape in cross flow,  and

-------
   354
   Upper water Inlet
is deflected very little by drag forces at velocities up to
1000 ft/min.  The fluid is deflected by electrostatic at-
traction, but this can be compensated for by auxiliary
electrode placement.
                               The fluid electrode precipi-
                               tator concept was applied to
                               experimental model units, and
                               collection efficiency of pre-
                               cipitator array configurations
                               was measured in a cotton dust
                               test facility.  This apparatus
                               consists of a 9 x 9 in.
                               straight tunnel 9 ft. long,
                               containing traversable samp-
                               ling probes located upstream
                               and downstream of the preci-
                               pitator test section.  Flow
                               straighteners near these
                               locations allow isokinetic
                               sampling through probe tips
                               of various inlet sizes. Dust
                               analysis based on projected
                               area diameter was performed
                               with an optical particle
                               counter.  A mass monitor was
                               used to analyze dust mass
                               concentration.
Pig. 3.
Fluid Electrode
Precipitator
                                    Two series of tests, one
  at  low capacity and one at high, were performed with  cotton
  dust from different sources.  For low capacity tests., air
  was drawn through a model-size cotton carding machine run-
  ning at 2 Ib/hr.  The air was prefiltered to remove lint
  and large particles and was then passed through the test
  tunnel.  In the high capacity tests, fine dust released by
  mechanically tapping a V-cell nonwoven filter was used.
  The filter was already loaded with lint and dust from two
  production cotton carding machines.  The air drawn through
  the filter and passed through the test tunnel contained
  fine dust that was similar to the prefiltered air from the
  model card.

      Samples of dust from each of the two dust sources were
  collected for testing.  They were dispersed in an electro-
  lyte and analyzed for size on a volume basis.  Model card
  dust had a mass mean diameter of 3.6>i with a geometric stan-
  dard deviation of 1.6u.  The cotton dust released by tapping

-------
                                                                355
the loaded filter varied from 4.5 to 6.7ju mass mean diameter
with a geometric standard deviation of about 2.  All test
concentrations were in the range of 0.5 to 1.5 mg/m3, which
is typical of cotton mill fine dust concentration.

    Table VI contains the operational data and collection
efficiency for a typical configuration tested.

                         TABLE VI

           EFFICIENCY OF TYPICAL FLUID ELECTRODE
                PRECIPITATOR CONFIGURATION
Velocity
ft/min
250
350
500
Field
Strength
kV/cm
6.2
5.9
4.9
Dust Supply
Concentration
mg/m-i
0.8
0.5
0.8
Count
Efficiency
1.5u-10u
%
75
81
60
Mass
Efficiency
%
86
93
78
    In the design, development, and testing of model fluid
electrode devices, some qualitative findings are notable
concerning fine cotton dust collection or other dust col-
lection.  The establishment and control of fluid electrodes
requires careful balancing of flow rates and use of smooth
tubes to assure laminar flow in the fluid electrodes.  In
practice, fluid can be piped from a central reservoir
through a branching network to fluid electrode locations.
Units of practical size would require Intermediate flow
tubes at approximately 1  ft intervals of vertical fall to
catch fluid, decelerate it, and return laminar conditions.

    Once properly adjusted, an array of fluid electrodes
can operate to continuously remove dust, with low mainte-
nance.  Dust entrained in the fluid can continue to circu-
late through the system until its accumulation warrants a
fluid cleaning cycle.  Fluid cleaning would not require
precipitator shutdown.  Applied to textile mills, con-
tinuous fine cotton dust  removal of 85$ would greatly reduce
levels of fine airborne dust.  For other industrial dusts,
possibly with collection  fluids other than water, similar
applications could be developed.

-------
 356

     Further research is being conducted to apply  findings  to
 large prototype units,  to increase efficiencies with  deeper
 or more optimally selected electrode  arrays,  and  to extend
 principles to other industrial situations  requiring con-
 tinuous fine dust collection.

 REFERENCES

 [1]   National Institute for Occupational Safety and Health,
      "Criteria for a recommended standard...  Occupational
      exposure to cotton dust," U.  S.  Department of Health,
      Education, and Welfare,  NIOSH 75-118,  1974.

 [2]   Anderson,  C.  D.,  "Federal regulations  and citations
      for cotton dust,"  Conference  Proceedings on  The  Engi-
      neering Control of Cotton Dust,  sponsored by Clemson
      University, July 29.  1972.  PP. 21-26.	

 [3]   Wakelyn, P. J., "Composition  of  cotton dust:  possible
      agents  causing byssinosis," Conference Proceedings on
      the Engineering Control  of Cotton  Dust,  sponsored by
      Clemson University.  July  29,  1972.  pp. 77-99.

 [4]   Merchant,  J.  A., et  al.,  "Dose response  studies  in
      cotton  textile workers,"  J. Occup.  Med.. vol. 15-3,
      pp.  222-230,  March 1973.

 [5]   Barr, H.  S.,  Hocutt,  R. H., and  Smith, J. B., "Cotton
      dust  controls  in yarn manufacturing,"  U. S. Department
      of Health,  Education,  and Welfare,  Cincinnati, Ohio,
      Report  No.  HSM 99-72-M,  March 197M.

 [6]   Thibodeaux, D.  P., Baril,  A., Jr.,  and Reif, R., "A
      Wet-Wall Electroinertial  Precipitator:  A Highly
      Efficient  Air  Cleaner for Cotton Dust," Conf. Rec.
      1976 Annual Meeting IEEE  Ind. Appl. Soc., pp.333-339.

I/  Presented at the Dust Workshop, Notre Dame University,
    Notre Dame, Ind., April 20-21,  1977.

2_/  One of the facilities of the Southern Region,  Agricul-
    tural Research Service, U.S. Department of Agriculture.

3_/  Employee of Battelle Columbus,  Columbus, Ohio.

4/  Names of companies or commercial products  are  given
    solely for the purpose of providing  specific informa-
    tion; their mention  does not imply recommendation  or

-------
                                                                        357
endorsement by the  U. S.  Department  of Agriculture  over
others not mentioned.
WRITTEN DISCUSSION

Dennis J. Helfritch
Apitron Division, American Precision Industries
Buffalo, New York, New York  14225

    The authors should be congratulated for the  innovative  designs which are
described in their paper.   These new devices have been designed  for  the
removal of respirable dusts in the textile industry,  but could  equally well
be applied to scores of other applications.

    A general comment concerns the reported efficiencies of the three novel
devices.  The authors state»that the indoor dust content of textile mills
can be reduced to one mg./m , but do-not indicate the dust  collector  effi-
ciency required to achieve one mg./m .  Therefore, one does not know  if the
devices described are satisfactory for this service.   The most  efficient
conventional device, the baghouse, is used extensively in the asbestos
industry and typically achieves 99.99% efficiency on  fibrous dusts less
than 15 microns.  I do not know why the devices  achieving 95%-99% efficiency
as described in the paper are considered improvements.

    A few specific comments, regarding data and  hardware can be made  at
this point.  Two devices utilize corona wires, and I  would  think  these
wires would require a method for cleaning deposits which are almost sure
to accumulate over periods of time.  The Wet-Wall Electronertial  unit is
described as having low operating costs, but the corona input is  approximately
500 watts/1000 CFM.  Including perhaps a 1-inch  pressure drop and water
pumping, it seems that power consumption is relatively high.

    According to Table III, in some cases the efficiency of the High  Speed
Rotary Drum actually decreases when the 1-inch diameter holes are covered
with fabric.  According to Table VI, the efficiency of the  Fluid  Electrode
Prescipitator increases when velocity is increased from 250 to  350 FPM
and field strength is decreased from 6.2 to 5.9  KV/cm.  The data  in Tables
III and VI is contrary to general principles and should be  examined more
closely.

AUTHORS' WRITTEN RESPONSE

    The level of "one mg/m3" quoted from the paper refers to total dust of
all sizes.  The efficiencies of present cleaning systems used in  textile
mills to produce these levels can be greater than 90% for total mass  removal
of lint, trash, and dust.  However, these systems have very poor  efficiencies
(less than 50%) for capturing respirable dust (less than 15 ym).

-------
  358

     Regarding the question of possibly using the baghouse, it is our understand-
 ing that the textile industry has had some experience here and that the peculiar
 nature of cotton lint,  trash, and dust is such that baghouses are not satis-
 factory.  These deficiencies are particularly in the areas of automatically
 cleaning and maintaining the bags and in the tremendous areas required by the
 high air capacity of textile mills.   Our experience with cotton dust is that it
 is significantly different in character from asbestos dust and no valid analogies
 between the two can be  drawn.

     The question about  cleaning of trash collected  on the corona wires is well
 taken.   Our experience  with running the two devices in question is that this
 has not been a problem.   From the standpoint of the W.W.E.  unit the vibration
 which the wire experiences in operation tends to keep it clean.  As regards the
 power requirements,  our understanding is that a conventional  precipitator could
 require 50 to 500 watts/1000 CFM with the air velocity below  500 ft/mia.   The
 W.W.E.  unit operates at  500 watts/1000 CFM but handles air  velocities between
 2000 and 3000 ft/min. 2Another advantage of the W.W.E.  unit is that its collection
 surface is about 10,  ft  /1000 CFM,  whereas conventional precipitators require
 from 50 to 1000 ft /1000 CFM.   These  factors (higher operating velocities and
 smaller collection areas)  put the W.W.E.  in a very  favorable  position compared
 to  a conventional precipitator or other air cleaner.

     With regards to  the  Reviewer's comments concerning anomolies in the efficiency
 of  the  rotary drum as given in Table  III,  it is true that in  some instances the
 efficiency of the drum did  decrease when the 1-inch diameter  holes  were covered
 with fabric.   This has been explained  by the fact that  apparently the particle
 counter was  registering  particles  that  were being knocked off  of the back side
 of  the  filter fabric.  Concerning  the  apparent  problems  posed  in interpreting
 Table VI,  it  should  be clarified  that  this  data is  not  presented for all  the
 same experimental  configurations,  thus  explaining the reason  for the seeming
 inconsistency  in  the way the  efficiency followed the principles  of  physics.
 This data was merely presented as  a sample  of the types  of  efficiencies obtained
with the  fluid  electrode precipitator.


 OPEN DISCUSSION

 Baril:  As  far  as  we know we do not have  to  clean the corona wire.   We have
 not  had  any problem on the wet wall electro-inertial unit. The rotating drum
was  looked at as an air cleaner , also  as an electro-inertial unit.

Reif:  For  the  rotary drum  the dust goes in  the bottom of the box, most of it
does not enter  the holes in the drum.  The comment about the fiber filter not
seeming  to improve the effectiveness of the device is true.  These types of
fiber filters are used currently in the textile industry which do not have very
high efficiencies.  That is one reason why we were looking at them.  If you
look at them with  the atmospheric dust type test that we were using, their
efficiencies are just a few percent by themselves.  The base velocities are
about 200 ft/min.  So they are very poor filters actually.  The  first stage
filters were designed to pull out long cotton fibers, the second stage filters

-------
                                                                             359

are generally of higher quality but still are not particularly good filters.

Baril:  Concerning ourselves with table 2,    the artificial cotton dust at
zero potential and zero current has    an efficiency percent of about 80.2%.
Then right under it is a card output (this is at zero current), its efficiency
is 30.3%.  The content of larger particles is different and you are seeing
the effect of electro-inertial input.

Billings:  I do know you had card type dusts and this analysis is a serious
problem.  I am concerned what you will do with this collector and how you
will distribute the dust, and what you are talking about in terms of the
actual staple, yarn and textile weaving.  Are you basically going to have a
ventilation system which sucks all of the air out of one end of the building,
cleans it and redistributes it?  Are you going to to have local exhaust
ventilation on every dust source?  How are these going to be configured in
attempting to meet the standard?

Baril:  You have asked a question that people in the air handling textile
industry suggest that the best thing they could do is get out of that industry
and service some other industry.

    This is a very serious problem and the people who service the textile
industry are few and far between (one is Albert Baril).  To answer this question
let us go back to a question asked by Dennis Helfritch.  He said you have units
and this and that, we readily agree  that you have a lot of units and they work
very well.  We could go to a bag unit which is very efficient.  EPA has looked
at some of this.  The estimate comes back that the baghouse must be about the
size of the textile mill.  This is one of the problems.

    Right at the moment, we do not know what we are going to do.  The government
is going to set some standards.  The Department of Agriculture is going to  try
like mad to get the standard high.   If we get that, we are not in as much
trouble as you  propose.  If we do go to the 200 bin it will  take us from 5
to 7 years to get there.  If we set  a standard that will bring it down to 200 in
5 years, who is going  to buy a piece of machinery that will bring it down to
500 in a year?

    What it looks like right now is  that a lot of the weave rooms have air
which comes in at a high level and escapes completely through the floor and
is taken out and cleaned or it sweeps across the room and does this.  To
increase that system  to handle what  the bin could possibly  take, I  then think
we might have to have  terrific exchanges of air.  We would  have  to  bring  this
air down to someplace  where we could actually clean it.  This incurs a cost
exceeding  $2 billion.  This was worked  out by TRI Princeton.  Right now this
section  is really up  in arms.  One possible way  to  go is to build individual
containment units around many machines.  They have  already  been  successful  in
doing this on the other  end of  the textile industry; that  is  cards,  enclosing
them  and capturing  them.

    We  also  enter a second  problem which we  have not addressed at  all

-------
 360

 and that is capturing the dust so you can clean it.   You do have leakage into
 the system and you can only capture a percentage of  the total dust,  and by
 cleaning,  putting it back into the room,  you are also leaking some of the un-
 wanted particles into it.  The capture system has to be worked on too.

     We have been working on an entirely,  completely  new system for textile
 mills which puts all the systems together in an integrated  machine.   In other
 words, you start off at one end by putting in cotton puffs  and you get  yarn  out
 of  the other end of the machine.  This machine could be totally enclosed.  We
 are doing  research on that line.

 Billings:   It is important to  mention one other federal agency which comes
 into the discussion and that is the FDA.   If you use an electro-static  precipitator
 that involves any kind of ionization you  are going to make  ozone.  There are
 already ozone standards with regard to this  kind of  technology.

 Baril:   We have  been monitoring the ozone levels.

 Benarie:   I have some information  about rotary filters  that are  used, for
 instance,  under  a British patent  on submarines.  There are three  possibilities
 for  building rotary filters.   The  mechanism  of the rotary filter is  quite  easy
 to explain.   The impaction mechanism on the  filter,  on  the  individual fibers,
 instead  of being impacted individually by the  vector  of  the air  stream  is
 impacted by the  vector  of the  whole movement within  the  airstream which can  be
 10 times as large,  at  say 3000 rpm.   The  fibers  can be  oriented either
 radially or      longitudinally and the stream can be from inside  toward the
 outside or in  reverse.  Now  the reverse system seems  much better because the
 cake is  thrown away automatically  by centrifugal  force.  Rotary air  filters
 are made which are used in the electro-chemical  industry  and are quite good.

     I have some more  information about  it  if you are  interested.

 Baril:  We  built a model  that  the British have and it was designed primarily
 to remove  radioactive particles from  the stream.  We built 3 units and
 approached  it  3 ways.  All three were good, but we felt the fluid electrode
was the best.  We have not given up on  that.  We still have it and work on it.

-------
                                                                             361
DIELECTROPHORETIC AIR FILTRATION:   PROGRESS AND  PROBLEMS


     J. K. Thompson, R. C. Clark,  and G.  H. Fielding

     U. S. Naval Research Laboratory (Code 6180),  Washington,  D.C.  20375
ABSTRACT

     Dielectrophoretic  air filters  under  development  at  the  Naval  Research
Laboratory,  as  well  as  those previously  reported  or currently  manufactured,
are  discussed  with  a  view to  establishing  successful  design  principles.

INTRODUCTION

     Our  burgeoning  air and gas filtration technology concentrates  largely  on
the  removal  of  particulate material  from the gaseous discharges  of  industrial
processes.   This  is legally  necessary  and highly  desirable from an  environ-
mental  health  point of view.   There  is,  however, another  air filtration chal-
lenge:   that of  cleaning  the  air drawn into, and  circulated within,  occupied
buildings,  vehicles  and other enclosures.   The  aerosols of  principal  concern
are  of several  types:   (1)  the airborne  pathogens  that  can  cause cross-in-
fection in  homes,  hospitals,  schools,  theaters,  etc.;  (2)  the  miscellaneous
airborne  debris  that  affects  the operation of  delicate electrical  and elec-
tronic  equipment  of many  kinds;  (3)  the  common dirt  and dust  that  deposits
on  surfaces and requires control as  a  matter  of esthetics and good  housekeep-
ing;  (4)  the  particulates  formed  in  motor  vehicle operation;  and   (5)  the
industrial  airborne  refuse which is  not removed  at its  sources.   Other common
aerosols  may in time be thought sufficiently  objectionable  to require removal
from the  ventilation air of  buildings.   Such  aerosols include  the  sulfate/
sulfuric  acid originating in  the burning  of sulfur-containing fuels  (1),  the
terpene-derived blue haze  produced by vegetation (2), and the metal-containing
particles produced by vegetation (3).

      The  means presently  available  for high-efficiency  filtration of ventila-
tion air appear  to  have sufficient  disadvantages  to severely limit  their use.
Such means  include  HEPA  filters,  very  high  quality  glass fiber  media,  and
electrostatic   precipitators.    Their  disadvantages  are  either  high  cost  or
high pressure  losses, or both.

      We believe  that a promising means  of handling present and future prob-
lems  in  ventilation  air  filtration  is  the  technique  of  dielectrophoresis.

-------
  362

 Equipment of  this type  is  inherently  simple, low  in  cost,  and very  compact.
 Since a dielectrophoretic air filter is a field-operated rather  than  a  current-
 operated device,  its  power consumption  is  near  zero.   Additionally,  there  is
 no corona discharge, ozone production  is zero, and  sparking between  electrodes
 can be eliminated.

      A brief  historical  review of dielectrophoresis and dielectrophoretic air
 filtration is desirable, both to give  credit where  due  and to assist others  in
 avoiding duplication of ineffective  or  non-optimized designs.

      The word dielectrophoresis was  coined by Pohl  to describe the accelerated
 separation of solids from  liquids in  nonuniform  electric fields (4).  Me have
 continued to use the same word to describe  the accelerated separation  (filtra-
 tion)  of  solid  or  liquid  particles from  air in  nonuniform  electric  fields.
 Such  an unusual  descriptor helps  nonspecialists  to  understand  that we are not
 talking about electrostatic precipitation.

     Let us first describe and discuss briefly  the dielectrophoretic process.
 uielectrophoresis  is  the movement  of a  particle in a nonuniform electric field
 the direction of  movement  usually being toward  the most intense  part of the
 field.   This  mechanism  is conveniently  applied in air filtration by passing the
 air through a thin  layer of  air  filter material which  is  sandwiched  between
 metal  screen  electrodes.  The  air moves  through  the sandwich  perpendicular to
 its plane,  and experiences no  resistance other than that of  the filter medium
 itself,  which may be  a commercial   glass   fiber  mat or  a  slab of  open-pore
 plastic  foam.  Since every  glass  fiber  or  every element of the  foam  locally
 distorts  the  applied  electric  field,  the mat or  slab contains  an enormous num-
 ber of  field  innomogeneities.   Since tne glass fiber and the plastic foam have
 dielectric constants  greater  than  that  of air,  the field is  more intense close
 to  a  fiber or foam element than remote from it.   Therefore  airborne particles
 within  the  filter  medium experience  a  dielectrophoretic  force.   This force is
 approximately  described  by:
Thus,  the  dielectrophoretic  force F  is  proportional  to the square of  the  ap-
plied  field  E  and, to the cube  of  the particle radius r; the  dielectric con-
stants of the medium, KX, and of the particles,  K2,  are also  involved.
NRL STUDIES
     The NRL studies  have  employed a uniform geometry comprising a  flat  glass
fiber mat, U.64  cm thick,  separating two steel wire screen electrodes.   Vari-
ables have been  the composition of  the  mat (three as-received commercial air
filter media), voltage  (0  to 7kv,  equivalent to 0 to 11 kv/cm), air velocity
and aerosol (OOP of 0.3 and  U.8 micrometer  diameter, and a fine fly  ash).  The
dielectrophoretic effect is  illustrated  oy the Dielectrophoretic Augmentation
Factor (DAF).   This is the  ratio  of the aerosol  percent  penetration (100% -
percent aerosol  retention)  at zero voltage  to  the percent penetration at the

-------
                                                                             363

voltage of  interest.   Tables  1 and 2 present data  for  one  filter medium (the
Farr Go's HP-100) at several voltages and air speeds, and for two OOP particle
sizes.  The manufacturer's  rated  air  speed  for  the HP-100 medium is 20 cm/sec
(40 fpm),  at  which speed the  initial  resistance  is 1 cm wg  (0.40  iwg).   The
DAF's at 7 kv and 20 cm/sec  are 21 for 0.3 micrometer OOP and 28 for 0.8 micro-
meter OOP.


                                   TABLE 1

             DAF VS AIR SPEED AND VOLTAGE FOR  HP-100 FILTER  MEDIUM;
                          AEROSOL 0.3  MICROMETER OOP


   Air Speed                                Applied Voltage, Kv
    cm/sec                         2            3.5         _5_       _J_


      3                             8            19             95          330
      5                             3            13             39          120
      8                             3            11             28          100
      13                             2              6            13           42
      18                             259           27
      26                             2              4              6           14

      3*                             2              I              \            I
      46                             1236
                                   TABLE 2

              DAF  VS  AIR  SPEED AND VOLTAGE FOR HP-100 FILTER MEDIUM;
                           AEROSOL 0.8 MICROMETER OOP
    Air Speed                                Applied Voltage, Kv
     cm/sec                        2            3.5          —5_
       3                           30           110            300         1100
       5                            6            30             95          360
       8                            4            18             50          170
      13                            3            10             20           50
      18                            2             6             13           35
      26                            2             4              8           18
      37                            235           11
      46                            1237

-------
  364
       It should be noted that the dielectrophoretic force does not operate alone
 in  a filter medium.   There are also the  ever-present  aerodynamic  forces- and
 in  some cases, forces  due  to  net electric charges on the particles, the filter
 medium, or  both.   Thus, a complete analysis of  the  performance  of dielectro-
 phoretic air filters has not yet been made.   It is clear that the relationship
 between the dielectrophoretic force and  the  applied  field  squared is likely  to
 hold  regardless  of  aerodynamic  factors.   However,  since both  the  aerodynamic
 forces and the dielectrophoretic forces  are  affected  by  both  particle size and
 filter element size, the total  interaction may be difficult to predict.

      We offer the hypothesis that the  dielectrophoretic  force  is effective only
 at  shorter range than  the  aerodynamic forces  and  that therefore,  the former  is
 operative only when there  is pronounced  evidence  of aerodynamic effects  in the
 absence of the electric field.

      The interplay  between the  aerodynamic  and  the  dielectrophoretic  forces
 is  complex.   For example,  aerosol particles  of  a given size which  are  moving
 through a  filter  medium deposit  inertially  on a  small  filter element more read-
 ily  than on  a large one.   However,  the electric field gradient  is stronger
 adjacent to a large filter element  than to a small one.   There would seem  to
 be a conflicting  requirement here:   small filter  elements for  inertial deposi-
 tion of particles,  and large elements  for dielectrophoretic  deposition.    In
 fact,   however,  small   elements  seem  to  be  desirable  for  dielectrophoretic
 deposition  also;  only  with small  elements  do  the aerodynamic  effects allow
 particles  to pass close enough  to the  elements  for the dielectrophoretic  foam
 to act effectively.

 OTHER  UIELECTROPHORETIC AIR FILTERS

     As far  as we know, the first dielectrophoretic  air filter, the "Electro-
 Polar  ,  was constructed  by Sproull and others  at the  Western Precipitation
 Corporation  in Los Angeles  during the  early 1950's (5).   This was a full-scale
 device  with  an air  flow rate of  3500 cfm.  The two filter media used were typ-
 ical  commercial glass  filter materials of good  quality.  It was tested by the
 Harvard University  Air Cleaning  Laboratory  for the Atomic  Energy Commission.
 The  reported results were mediocre (6,7), although in  our view the performance
 was  better  than reported.   The dc  voltage  used for  the Electro-Polar  device
 was  15 kv  for a one-inch  medium.   This  provided  a gradient  of  5.9  kv/cm.
 Influential  in  the  choice  of voltages  was  the  desire   to  avoid both  corona
 discharge  on  the expanded-metal  electrodes  and  a   spark  discharge  through
 the  filter medium.                                                          3

     A  report by  Thomas and Woodfin  in 1959  described a  dielectrophoretic air
 filter  of  unusual design  (8).    The spaces  between   a  series of charged  and
 grounded metal  plates,  oriented  parallel to  the  air flow as in an  electro-
 static  precipitator,  were  packed with  glass  fiber.   A  filter one  inch  deep
 in the  direction  of  the air  flow showed "great improvement" when the  voltage
was  applied.   The performance of a filter  14.5  inches deep at a  voltage gra-
dient  of  9.3 kv/cm was improved by a  factor of  over five at air  velocities

-------
                                                                             365


of 400-500  fpm.   No further mention  of  the Thomas-Woodfin device is found  in
the literature.

     Havlicek's valuable  theoretical  dielectrophoresis  paper of  1961 reported
on the effectiveness of  electric  fields  parallel  and perpendicular to  the  di-
rection of  air  flow  (9).   The former was  found to be preferable on the  basis
that  it  allows the maximum field gradient to  occur at points where the  rela-
tive particle-to-fiber velocity is at a  minimum.   Thus  the two effects  act  to-
gether to maximize particle  deposition.    In fields perpendicular to air  flow,
deposition  is  correspondingly minimized.   Havlicek's  theorem indicates that
the geometry  of Thomas and Woodfin's filter was  not such as  to  optimize fil-
ter efficiency.

     Rivers' 1962  paper  provides  a  useful  theory  review  and,  in  addition,  de-
scribes a  unique  air filter which  may  have some  dielectrophoretic  properties
(10).   The medium is a  multilayer  paper;  the electric field  causes a  separa-
tion  of  the layers.   The modest improvement  due to electrification has been
attributed  to  separation  of the  paper  layers,  probably  attributable to  elec-
trostatic repulsion.  Rivers'  device  also  places  the  electric  field perpendic-
ular to the air flow direction.

     The  next  example  in our historical   review is that of Walkenhorst  and
Zebel  in  1964  (11).   Their filter medium  unit was ten  layers  of  nylon  hosiery
fabric between  screen  electrodes;  several  of the  units were placed in  series.
Using  aerosols  of  crushed coal or quartz,  good filtration performance  was  ob-
tained in an electric field.   Remarkably,  a similar filter, exposed to  the  NRL
OOP aerosols, showed no significant particle deposition.

     Finally,  a currently  manufactured  air filter  of  dielectrophoretic  de-
sign  is  the  "Fiberstatic",  by  Filterlab  of  Houston,  Texas.   The design  of
this  device is a  high voltage screen electrode  having  a  0.6  cm polyurethane
foam  mat  on each side, all sandwiched between  two grounded screen electrodes.
In a  test with  a OOP aerosol no appreciable filtration efficiency was
observed.

     My  colleagues and  I would  now  like to  discuss the  historical cases  we
have  cited  in order  to assist  future  development  of  dielectrophoretic  air
filters.   We  consider  that  the Electro-Polar of Sproull, et al was the best
of  those  we have reviewed, as well as being  the  earliest.  Why,  then,  did  not
the  Electro-Polar  perform  in  consonance  with  the potential  indicated  by  the
NRL  data.   Perhaps it was an  idea  whose  time had  not  yet come.   More  signif-
icantly,  the  problem  may have  been  the  unavailability  of glass  filter mats
thinner  than 1.3 cm  (0.5 in).  The  thicker media available required  a  higher
voltage  to attain an  efficient voltage  gradient, but the  higher  voltage  in-
creased  the probability  of  the undesirable  corona discharge.   Forced to  use
a  low voltage  gradient,  the  Electro-Polar team  then were hurt by  the  square-
law  dependence of dielectrophoretic  force on voltage  gradient.   Another fac-
tor  may have  affected  the operating voltage  chosen.  This was  the increased
probability of  a  spark  discharge  through  the  filter  medium as  the  voltage

-------
 366

was  raised.   Spark discharges produce  holes  in the filter mat, and may ignite
flammable  materials deposited in  the  filter.   It is possible  that the sharp
edges  of  the expanded  metal  lath electrodes were likely sources of  both corona
and  spark  discharges.   This problem  has  not  yet been fully solved, but an NRL
development  may  eliminate  this difficulty.   In any case, steps should be taken
in  dielectrophoretic  filter design to  allow  operation  at  the highest practi-
cable  voltage gradient, taking advantage of the  square-law relationship.

     Turning attention to  the  Rivers  and  Thomas/Woodfin devices,  we believe
that  they  demonstrate  the validity of  Havlicek's  axiom that dielectrophoretic
filters  should  be  designed for  parallel  alignment  of  air flow  and electric
field.  This should be considered an established principle of dielectrophoretic
filter construction

     Walkenhorst  and  Zebel's filter,  with the stacked nylon  hosiery medium,
has  the  property of  good performance  with solid  aerosols of  crushed coal  or
quartz,  and  negligible performance with  the  NRL  OOP aerosols.   Further,  the
relatively coarse nylon filaments  (10  to  15 micrometers) would not be expected
to be  well  suited as  dielectrophoretic filter  fibers.   We suggest reconciling
the Walkenhorst/Zebel  and  NRL  data as  follows.  The  nylon  medium  per se is in
fact  ill  suited  for  dielectrophoresis because of its  large  fiber diameter.
However, as  a few solid aerosol  particles deposit  on  the fiber, probably by an
aerodynamic  mechanism,  the sharp points and edges of the deposited solid parti-
cles serve to produce electric field inhomogeneities.   Dielectrophoretic action
then  increases  progressively.   With  OOP  aerosols, some OOP droplets may  also
deposit  aerodynamically,  but then yield  no  sharp  points  or edges.   In fact,
they  simply  spread  smoothly over  the  nylon fibers,  increasing  their diameter
slightly.  If the above reasoning  is  correct, widely  differing performance can
be expected  from a large-fiber or large-element dielectrophoretic  filter,  de-
pending on whether  the test  aerosol is  a  liquid or certain solids.  It is  pos-
sible, also,  that the satisfactory performance of Walkenhorst/Zebel filter  with
the coal or  quartz  aerosols  was  in part dependent  on  particles larger than the
0.3 and 0.8 micrometer OOP.

     While we have no  test data  on the Fiberstatic filter with  a solid aerosol,
it is  plausible that this  filter performs  well in the  manufacturer's hands,  and
that the test  aerosol  is  a  solid.  Thus, the test pattern  would correspond  to
that of  the  Walkenhorst/Zebel  filter.  This  could be expected since  the  ele-
ments  of  the reticulated  polyurethane  foam are also  very large compared  with
the glass filaments of the NRL  filters.

CONCLUSIONS

     We believe  that  dielectrophoretic  air filtration has  valuable character-
istics that  will  contribute  toward very wide  application in the future.   How-
ever, even the "first generation" NRL filters require  careful  consideration  of
the basic phenomena to  insure  satisfactory  performance.  Consideration  of  the
design of prior  and current dielectrophoretic  air filters is  helpful  in  pro-
ducing new designs.

-------
                                                                             367
REFERENCES

(1)  R.E. Weiss et al,  Science 195:979-980 (11 March 1977).

(2)  R.A. Rasmussen and F.W. Went, Proc.  Natl. Acad. Sci.  USA 53:215 (1964).

(3)  W. Beauford et al, Science 195:571-572 (11 February 1977).

(4)  H.A. Pohl, J. Appl. Phys. 22:869-871 (1951).

(5)  Persona]  communications from W.A.  Sproull.

(6)  C.E. Billings, R.  Dennis, and L. Silverman, Harvard Air Cleaning Labora-
     tory, Boston, MA,  Report NYO-1592  (1954).

(7)  L. Silverman, Indust. Hyg. Quarterly:183-192 (Sept. 1954).

(8)  J.W. Thomas and E.J. Woodfin, AIEE Trans. Pt. II 78:276-278 (1959).

(9)  V. Havlicek, Int.  J.Air and Water  Poll.  4:225-236 (1961).

(10) R.D. Rivers, ASHRAE Jnl:37-40 (Feb.  1962).

(11) W. Walkenhorst and G. Zebel, Staub.  24:444-448 (1964).
 WRITTEN DISCUSSION

 E.R.  Frederick
 Air Pollution Control  Association,  P.O.  Box 2861
 Pittsburgh,  Pennsylvania  15230

     In accepting this  assignment  reluctantly,  I reminded  Professor Ariman  that
 my experience has been confined mostly to  the  practical aspects  of   particulate
 control by fabric filters  without electrical augmentation.  While quite some-
 time  ago,  limited studies were  conducted with artificial  electrification of
 an experimental fabric filter-1-, most  of my work has been  restricted  to  consider-
 ations of the effects  of natural  charging  in both the media and  particulates.
 But because of the very outstanding benefit realized by augmenting  the  charge
 on the medium and my personal views concerning critical importance  of electro-
 static involvement in  the  filtration  process,  I can be completely sympathetic
 to the Thompson et al  observations.  Hopefully, studies such  as  this will
 receive further attention.

     I would like to take this opportunity  to ask  the following questions that,
 perhaps, I should have had answered from my own earlier evaluations  if  only
 they had been made under more controlled conditions at that time.

-------
 368

     1.   How was the aerosol generated and conducted to the filter and was it
 charged in the process?

     2.   What,for example,  is the ultimate fate of the collected product?

     3.   What form does it  take as it deposits or forms on the collecting
 surface(s)?

     4.   How much longer can the electrified field filter  operate than the non-
 electrified variety?

     5.   Since  we might expect DOP to coalesce, particularly in the field,  does
 this occur and to what extent?

     6.   If the DOP mist coalesces on the  collecting fibers,  does the  oil  form
 on - and then  run from - the surface without  affecting, adversely,  the filter's
 performance for extended periods of  service?

     7.   If this is a realistic appraisal  of the process,  can it be applied to
 other mists or mist-like emissions?   And  if so,

     8.   What are the limitations,  considering the flammability and explosive-
 ness of  some mist-like solvent emissions?

     The  above  questions are  based  primarily on my suspicion  that,  like effective
 natural  charging,  electrical augmentation may lead  to  improved aggregation or
 agglomeration  of the collected product.  Most often, when  this is  accomplished
 ideally,  the nature  of the deposited  cake is  most favorable  for optimum col-
 lectability, for most  effective cleanability  and  for highest  efficiency.   With
 a condensable  liquid mist, this effect  should  be easy to  detect.

     Professor  Penney has provided  significant  substantiation  of some of our
 experiences  through  his  fundamental studies by  relating the combined effects
 of particle  charge  and  fields  with the deposited product. He  has  shown that
 "impact"  charged particles form a  chain-like  deposit on a  fabric without the
 use  of any high  voltage on the  particulate  or on  the collecting surface.   In
 another  study, again without  external potential impressed  on  the filter fabric,
 corona-charged particles also  deposited in  a  chain-like aggregated manner  on
 one  and  only one of  the fibers  in  a composite,  two  fiber filters.

     Figure 1 is  a photomicrograph  showing a blend of two different  (wool and
acrylic)  fibers  about  3 ym in diameter.  The  fact that one fiber remains
clean while  the  adjacent fibers  and only these electropositive  fibers  collect
 the negatively charged particles as a porous aggregate is  extremely significant.
 If this  is the kind  of open deposit that can develop because of electrostatic
 forces,  it would seem  likely that  the exceptional performance of the NRL
dielectrophoretic filter could  also be attributed to the formation of  such
a super porous deposit.

     In the simulated filtration  tests that we carry out without charge aug-
mentation, natural electrostatic charges,  when suitably balanced, offer a most

-------
                                                                                                                                         369
FIGURE t  AGGREGATED DEPOSIT OF PRECHARGED  PARTICUL*TE ON A  FILTER FIBER
                         O	
       FILTRATION CONDITIONS:  A/C = 5.4
                                 DELTA P LIMIT = 6 IN, W..C.
                                 TEMP. = 150 °F
                                 CLEANING - HORIZONTAL SHAKE

	  RUN NO.. 41 - FABRIC 16 [_SP. DA..  (55) fl. SILICONEJ. PERM. = 36 CFM/FT2@ 0.5 IN. WX..


	RUN NO.. 40 - FABRIC 15 CFll_ DA.. (?)J. PERM. = 39 CFM/FT2 S O.S IN.. W..C.


	RUN NO. 42 - FABRIC 18 QSP. P-E.. NAPPEDJ PERM. = 38 CFM/FT2® 0.5 IN.. W..C.
                             20
                                                       40                       60
                                                       TOTAL COLLECTED SOLIDS, G
                                                                                                          80
                                                                                                                                   103
FIGURE 2 EXPERIMENTAL FILTRATION OF ELECTRIC FURNACE DUST

-------
370
  desirable means for achieving the highest level of filtration performance.
  The evidence implies that when a particulate will form a porous deposit or
  aggregate, this most desirable change in the cake can be achieved during the
  filtration process by a favorable balance of the electrostatic properties
  of the particulate and of the collecting fabric.  While most participates can
  form porous deposits, not all dusts aggregate so easily that any but the more
  active (electrostatically) fabrics can promote the change.  Extraordinary
  persuasion seems possible by using media of the most suitable electrostatic
  polarity, charge intensity and, sometimes, of perferred charge dissipation
  rate.

      On page 12 of the Helfritch paper "Electrical Filtration and the Apitron/
  Design and Field Performance," results are given for pilot tests on electric
  furnace (steel) dust.  The curves indicate that electrical augmentation
  increases filtration performance (rate and efficiency) very significantly.  The
  curves obtained in our non-electrified filtration tests also carried out on
  electric furnace (steel) dust have a similar appearance indicating, possibly,
  the same kind of improvement in deposited cake.  Thus, Figure 2 shows a plot of
  the quantity of electric furnace solids collected as a function of bag pressure
  drop (limited in the tests to 6 inches w.c.) for three oolvester fabrics used
  in the writer's filtration study of an electric furnace (steel) dust.  The
  conditions in the three evaluations were essentially identical and the fabrics,
  although of different construction, were of about the same permeability.
  Significant variations were noted in electrostatic properties among the three
  media  with the filament Dacron fabric (#15) being quite electropositive, the
  napped polyester fabric (#18) being most electronegative with highest intensity
  and the spun Dacron fabric (#16)  of intermediate polarity and intensity.

      Professor Penney?s reproduction of nodules or "dingle-berries" in experi-
  mental filter tests also seem to  reaffirm the fact that charges play a very
  important role in aggregate formation.  The nodules were produced on high cover
  (fuzzy),  not filament (smooth),  fabrics with a strong electric field.  In
  subsequent tests,  it was found that the nodules formed more quickly in a region
  where  there was little field compared to the strong field area.  Nodule for-
  mation can be induced by electrostatic forces and their retention to the fiber
  seems  to  be due to this same force.

      It is clear that not all of  the facts concerning charge related particulate
  aggregation are known,  but it does  seem evident that the porous deposit - low
  flow phenomenon resistant type of cake formation could account for the super
  performance of the NRL filter.

      Whatever the specific mechanism by which electric charges and fields
  influence the collection process, it is apparent that if the advantages of
  electrical effects are  to be generally utilized, certain specific properties of
  filter media and particulates, not  now considered significant in commerical
  fabric filtration circles,  need to  receive more serious consideration.   Cer-
  tainly, the electrical  resistivity  of both the particulate and the medium have
  to be  considered worthy of determination routinely.   Quite likely,  oversight of
  this property in the past could have been a factor contributing to the poor or
  non-reproducibility of  data noted in Thompson's review of "Other Dielectro-

-------
                                                                             371

phoretic Air Filters."

    That high resistivity is a problem limiting the charging of certain fly
ashes and, thereby, restricting their collection by electrostatic precipi-
tation, is well established.  It is also clear that the resistivity of other
participates vary from relatively low (108  £2 cm) to extremely high
(> 1014 n cm) values.  Since particulates and fabrics of high resistivity
cannot be charged, except naturally, it is quite possible that some incon-
sistencies in performance and the non-reproducibility of results indicated
with the various dielectrophoretic devices reported by Thompson et al can be
attributed to such variations.

    Our experience with many filter fabrics provided by a number of different
suppliers indicates conservatively, that up to 20 percent of the commercial
media offered to the baghouse users carry residual finishes which reduce the
resistivity of the fabric to 1010 fl/Q  or less.  It is not too difficult to
imagine, then, that different lots of supposedly the same product have
performed differently in dielectrophoretic trials, just as they function
differently in non-electrified filtration operations.

    Product differences among manufacturers introduce another variable and
represent a problem that is more the rule than the  exception.  When to this
uncertainty is added the strong competition to cut prices and, in some instances,
a policy of interchanging fibers of a generic type under one specification, it
becomes evident that reproducibility in the performance of filter media cannot
be achieved without identical duplication.

REFERENCES

[1]  Frederick, Edward R.,"Some Effects of Electrostatic Charges in Fabric
     Filtration," J. Air Poll. Control Assn., 24, 1167  (1974).

[2]  Penney, G.W., "Collection of Electrically Charged Particles in Filters,"
     J. Air Poll. Control Assn., 26, 58 (1976).
OPEN DISCUSSION

Penney:  There was quite a  flurry of activity among manufacturers on this about
25  years ago.  They were trying  two^stage precipitators.  Nothing was ever
published  about  it.  There  was a big difference whether you could charge the
corona dust,  then you could get  fairly high efficiency.  But you were in
trouble if the efficiency was not high enough because you would have space
charge troubles.

     There was the remark about  a filter giving a high efficiency on quartz
or  coal dust  and not on DOP.  That  is to be expected.  My experience has
told me that  it  is difficult to  disperse the quartz without having  the particles
strongly charged.

     The same thing happened with Rivers when he presented the paper at Penn

-------
372

 State.  He noticed  that his artifically dispersed dust gave a high efficiency,
 but  the atmospheric air which has had a long time to get naturally discharged
 showed a low efficiency.  It seemed a question of whether or not your test
 particles are charged or not.

 Frederick:  Do you  think that the fact the particles were solid rather than
 oil  spheres contributed to their collection efficiency?

 Penney:  I do not know.  I think it is more than likely to be the dispersion
 of particles.

 Friedlander:  This paper is quite interesting; it shows the similarities and
 differences from the HGMS.  First of all Zebel's paper is not quoted by the
 authors.  This paper shows very clearly that it is easy in the electric case
 to get confused  between the dipolar and the charged cases.   You can not keep
 them apart.  In the magnetic case, there is no confusion because there are
no monopolar magnetic charges available.

    In this case you have the problem of electric dipolar.   Magnetics people
do not have magnetic monopoles because we do not have monopoles.  We need
large currents of low power if we use superconducting currents.   Many appli-
cations would be well suited for permanent magnets.   The ferro-electric
does not really get high enough electric radiation.

    I think that the electro-ferretic people can learn from  HGMS.   Because
all the things discussed in this paper have been thoroughly  investigated
in HGMS.  Now you could go back and see where you could use  this information
the best.

-------
                                                                  373
COLLECTION OF FINE PARTICULATES USING A FOAM SCRUBBER
    Tom E. Ctvrtnicek. H. H. S. Yu*, and C. M. Moscowitz
    Monsanto Research Corporation, Dayton, Ohio

    Geddes H. Ramsey, EPA, Industrial Environmental Research
    Laboratory, Research Triangle Park, North Carolina

ABSTRACT

    This paper addresses the subject of application of foam scrub-
bing to fine particulate control.  Diffusion and sedimentation
appear to be the primary mechanisms of particle collection.  Fun-
damental theory for these two mechanisms is discussed.

    Principles of foam scrubbing were experimentally verified for
particles between 0.18 and 1.0 vim diameter on a bench (2.5 acfm)
and small pilot scale (500 acfm).  The results of these experi-
ments are presented.  In either scale the experimental results
agree well with the theory.  Comparisons of economics indicate
that the foam scrubber can be competitive with conventional col-
lection devices including electrostatic precipitators, fabric
filters, and high energy scrubbers in both capital investment
and operating costs.

ACKNOWLEDGEMENT

    This work was supported by the EPA under contract 68-02-1453.
Appreciation is expressed for EPA's continued interest in the
foam scrubber.

INTRODUCTION

    The basic principle of the foam scrubber is illustrated in Fig-
ures 1 and 2.  Gas stream contaminated with the particulate matter
is forced through a screen wetted with a surfactant solution.  As
the gas passes through the wet screen openings, foam is generated
and the gas is enclosed in a multitude of small bubbles, Figure 2.
The distance which a particle must travel to reach a collection
surface is now greatly reduced by the bubble enclosure.  The foam
*Present address:  APT, San Diego, California.

-------
  374

                                                      •SCREEN WIRE
                                               V
                                         AEROSOL » » » »
 Figure  1.   The  basic principle     Figure  2.    Aerosol  enclosure
             of foam  scrubbing                   in  a  foam bubble


proceeds through the main chamber of  the scrubber  where the par-
ticles are given time to migrate to collide with the bubble wall.
As the foam  is broken, particles retained  on  the bubble walls  are
collected in the surfactant  liquor.

    Monsanto Research Corporation  (MRC) initiated  work  on foam
scrubbing in the latter half of 1974  under a  contract with IERL-RTP,
EPA.  The contract involved  theoretical evaluation,  a bench-scale
verification, and a  small pilot scale demonstration  of  the foam
scrubber.  Results of this work form  the basis  of  this  paper.

    The next -section deals with theoretical aspects  of  particle col-
lection by foam.  The results obtained on  a bench-scale (2.5 cfm)
and small pilot  scale  (500 acfm) foam scrubber  units are then  pre-
sented.  Finally, the economics of  the foam scrubber are  presented.

    In this study, a scrubber system  using wet  screen foam gen-
eration was considered (see Figure  1).  It should  be noted that
other methods of  foam generation may  result in  conclusions differ-
ent from those stated in this paper.  Methods of foam generation
other than the wet screen are not a subject of  this  paper.

THEORETICAL ASPECTS OF PARTICLE COLLECTION BY FOAM

    Several governing equations could be applied to  describe the
process of foam  scrubbing.  These include  the equation  of contin-
uity, momentum (Navier-Stokes), energy, and diffusion.  As is
often the case in practical  situations, exact descriptions of
the phenomena using these equations are characterized by  rigorous
formulas too difficult to solve mathematically.  As  a result,
many simplifications must be applied.

    Let us first consider the idealized case of bubble  formation as
shown in Figure  2.  Aerosol with constant  particle concentration,
C0, rushes into  a foam bubble as it is being  formed.  The bubble
containing the aerosol detaches at  time t  = 0.  As long as the cross-
sectional areas  of the foam generator and  the foam scrubber are sim-
ilar, the mass of aerosol stream will move with essentially the same
velocity before  and after the formation of a bubble.  It  is therefore
reasonable to assume that the aerosol inside the bubble and the
foam bubble surface progress with the same velocity  through the

-------
                                                                  375
 scrubber.  Consequently,  the  impaction  forces active  in this  sys-
 tem may be assumed  to be  negligible.

    Now let us  consider diffusion of particles enclosed in  the
 foam bubble.  Assuming an isothermal system  (i.e., gas tempera-
 ture - bubble film  temperature)  and a very small,  spherical bubble,
 if the motion of  gas is negligible inside the bubble  after  the
 bubble has been formed, then  from the set of governing equations,
 Pick's Second law applies:


          |£ =  D72C                                       (1)
          d L

 Assuming  uniform  aerosol  concentration  and  transformation of
 Equation  1  into spherical coordinates,  we obtain:
             3t        I,   3r
                          _  _  (
n>  J
 The  diffusion  coefficient,  D,  for  small  particles  may  be  ex-
 pressed  by1:


          D ' 60-  [ X  + A(r->  +  Q<  (r->  e~brP/A~|      (3)
                  PL       p        p        J
                                              _22       2
 where     k1 =  Boltzmann's constant (1.38 x 10  kg m/s °K)

          T =  temperature

          y =  viscosity  of  gas

          r  =  radius of  diffusing  particle

          X =  mean free  path of  the gas

     A,Q',b =  empirical  constants


    The solution of Equation 2 will give the particle concentration
distribution inside a bubble as a function of time and radius for
various particle sizes, assuming that the particulate cloud is
monodisperse and, at given concentration, the coagulation of par-
ticles is negligible.

     The  partial differential  equation,  Equation 2, can be solved
 by the separation-of-variables method.   To solve  Equation 2 the
 initial  and boundary conditions must  be  summarized:

-------
  376

          (a)  C(r,0) = C   initial  concentration  at
                       t =  0 is constant                 (4)

          (b)  C(R,t) = 0 on the surface of  the
                       sphere at t >  0                   (5)

          (c)  C(r,t) 7* », r —»• 0                         (6)

The general solution will have a form as  follows:
C(r,t) =  (cie-"2Dt + C2ea2Dt)
                      fc  1
                      N
                                                         (7)
                                           i          i
                                      + cv
where  a, d, Ci, Cs, Ci* = constants

                      J  = Bessel functions
                       n


By use of boundary conditions, Equations 4, 5, and 6, and  lengthy
derivation and transformation, Equation 7 becomes:
                                     ,mi,  ,
                      n=l


This equation gives the  instantaneous particle concentration
inside the bubble.

    The instantaneous rate of loss of particles to the bubble
surface is now given by:
    q(R,t)  = -D(4uR2)  J£
     7.
>-(^)  Dt
  K            (9)
                                 E-(^)
                                   R
                               e
                         r=R          n=l

The total loss of particles in time t from time t =  0  is:
= /q(R,
                   t)dt  =
                          8R3C

                                    _    _
                                                         (10)

-------
                                                                   377
Hence, the particle  collection efficiency, E (%), in time t may
be determined as  follows:
  E  =
         Q
      4/3nR3C
100 =  —
                  1 - e
                                                 x 100
                                                          (ID
  noLo:
          n=l
   '- 6"
Inspection of  Equation 11 shows that bubble diameter,  R,  is of
primary  importance to collection efficiency.   It  appears  as a
square term  in the denominator of the exponential term.   For
illustration,  particle collection efficiency,  Equation 11 was
plotted  for  bubble size of 1.0 mm diameter and three particle
sizes, 0.01,  0.1,  and 1 ym diameter, as a function of foam resi-
dence time,  Figure 3.  Collection efficiency by diffusion only of
0.01 um  particles  within a reasonable practical residence time
(say 5 seconds)  is very high as compared with  a 1.0 ym particle.
                            PRESSURE - 760 mm Hg
                            TEMPERATURE-23°C
                            BUBBLE DIAMETER-1.0 mm
                                     40
                               RESIDENCE TIME, s
                                           50
                                                60
      Figure 3.   Theoretical  particle collection efficiency

-------
   378
     It  is  obvious  that some of the enclosed particles could reach
 the  bubble surface by sedimentation.   The number of particles
 deposited  per unit time inside the bubbles by sedimentation can
 be expressed  as  UR2)CVS,  where Vs =  particle settling velocity.
 For  simplicity,  it is assumed that the concentration of particles,
 C, inside  the bubble is uniform.   Using this expression we can
 formulate  the coefficient  of absorption by sedimentation (ratio
 of the  rate of particles deposited to the total number of particles
 in the  bubble),  as,  as follows:
                             TrR2CVo       3V
                                  s        s
                       ft  ~ "i"^^™™^^""^^^"^^^  S «w^^«^             / T f\ \

                        S    (4/3irR3)C   4R
The settling velocity can be expressed by1:


                   d£p«9   /   «7»     01 -bd /2A'
              Vs =
where    p  = particle density

Combination of Equations 12 and 13 gives:
                                        -bd /2X
                                           p
Typical sedimentation velocities of particles  (Equation 13) and
coefficients of absorption by sedimentation  (Equation 14) are
listed in Table 1 for a bubble diameter of 1 mm and a particle
specific gravity of 1 g/cm3.

-------
                                                               379
          Table 1.  TYPICAL SEDIMENTATION VELOCITY AND
                    COEFFICIENT OF SEDIMENTATION
                    (for R = 0.0005 m, p  = lg/cm3)



Vg, m/s
V 1/s
d (ym)
0.01
6.5 x 10~C
0.0001
0.1
8.8 x 10~7
0.0013
1.0
3.5 x
0.052
10~5

    Evidently, for particles larger that 1 ym, sedimentation
would significantly enhance the collection of particles during
the foam scrubbing process.  For 1 ym particles, the table indi-
cates that about 5.2% of the remaining particles could be re-
moved by sedimentation mechanism every second, assuming uniform
particle concentration within the foam bubble.  Because of the
diffusion process, the particle concentration profile, Equation 7
would not be uniform throughout the bubble (i.e., leaner near
the wall) and, therefore, the actual sedimentation coefficient
would be somewhat less than indicated above.

    The combined particle collection efficiency  (including dif-
fusion and sedimentation) was determined in a step-by-step cal-
culation.  Although the method may not be absolute, it provides
a tool for estimating foam scrubber collection efficiency.

    The computational results are presented in Figure 4, a through
i, for various bubble diameters, particle sizes, and particle den-
sities.  For convenience, the diffusion collection curves are also
shown in these figures.  As evidenced by the collection curves,
the enhancement of collection efficiency by the sedimentation
mechanism is far more important than the diffusion mechanism
for the larger particles.  However, for the particles smaller
than 0.1 vim,  the sedimentation mechanism of particle collection
is negligible.

     It was mentioned earlier that the impaction mechanism of col-
lection in the foam scrubber may be assumed negligible.  To eval-
uate the impaction one would need to analyze flow development condi-
tions within the foam bubbles applying the Navier-Stokes equation.
This is not an easy task.

-------
   380
                                     BUBBLEDIAMETER-0.5nm
                                     PARTICLE DENSITY - Ig/cm3
                                     PRESSURE-ATMOSPHERIC
                                     TEMPERATURE 23 ° C
                              d -l.Oum
                               P   _T^-
                                   DIFFUSION AND SEDIMENTATION
                            	DIFFUSION ONLY
                            • CURVE IDENTICAL WITH CURVE FOR
                              DIFFUSION AND SEDIMENTATION

                                 j_
 30      «     SO
    RESIDENCE TIME, s
                           (a)
                                            3   «
                                                                               p • 0.01pm •
                                                                                 10      20
                                                                                     BUBBLE DIAMETER- 0.5 mm
                                                                                     PARTICLE DENSITY-2 o/cm'
                                                                                     PRESSURE-ATMOSPHERIC
                                                                                     TEMPERATURE - 23 ° C
                                                                                         DIFFUSION AND
                                                                                      "~ SEDIMENTATION
                                                                                  	DIFFUSION ONLY
                                                                                  • CURVE IDENTICAL WITH CURVE FOR
                                                                                   DIFFUSION AND SEDIMENTATION
                                                                                               30      40     50
                                                                                                RESIDENCE TIME, s
                                                                              (b)
                                       BUBBLE DIAMETER- 0.5 mm
                                       PARTICLE DENSITY- 3g/cm3
                                       PRESSURE - ATMOSPHERIC
                                       TEMPERATURE-23° C
    0       10     20
                 DIFFUSION AND
             	SEDIMENTATION
            	DIFFUSION ONLY

        • CURVE IDENTICAL WITH CURVE FOR
         DIFFUSION AND SEDIMENTATION
-i	1	L_	i	
30      «      50      a
RESIDENCE TIME, i
                          (c)
                                          BUBBLE DIAMETER- 1.0mm
                                          PARTICLE DENSITY- Jo/cm3
                                          PRESSURE - ATMOSPHERIC
                                          TEMPERATURE - 23 »C
                                         DIFFUSION AND
                                         SEDIMENTATION
                                         DIFFUSION ONLY
                                    • CURVE IDENTICAL WITH CURVE FOR
                                     DIFFUSION AND SEDIMENTATION
                         »      40      50
                          RESIDENCE TIME, s
                                                                                                              BUBBLE DIAMETER-1.0 mm
                                                                                                              PARTICLE DENSITY- Ig/cm'
                                                                                                              PRESSURE - ATMOSPHERIC
                                                                                                              TEMPERATURE - 23 ° C
                                                                                                                 DIFFUSION AND
                                                                                                                 SEDIMENTATION
                                                                                                             	DIFFUSION ONLY
                                                                                                          CURVE IDENTICAL WITH CURVE
                                                                                                          FOR DIFFUSION AND SEDIMENTATION
                                                                                              30     «      50
                                                                                                RESIDENCE TIME. 5
                                                                             (d)
                                                                             _-	BUBBLE DIAMETER-1.0mm
                                                                                      PARTICLE DENSITY - 3o/cm3
                                                                                      PRESSURE-ATMOSPHERIC
                                                                                      TEMPERATURE -23°C
                                                                                      IffUSION AND SEDIMENTATION
                                                                                	DIFFUSION ONLY
                                                                                                       • CURVE IDENTICAL WITH CURVE
                                                                                                        FOR DIFFUSION AND SEDIMENTATION
                                                                      30      40      SO
                                                                       RESIDENCE TIME, s
                          (e)
                                                                             (f)
Figure   4.     Theoretical  collection  efficiency   (cont'd  next  page)

-------
                                                                       381
                   	DIFFUSION ONLY

                  • CURVE IDENTICAL WITH ClKVE fOB
               t MtiM DIFFUSIONANp_SEpl«NHltON
               KSIKNCITINE. 1
                (g)
                                                  ffiSIHICI 1IW. 1
                                  (h)
                                     8UB8LE OIAWETER -2.0mm
                                     PARTICLE DENSITY - la/cm'
                                     PBESSUW-ATMOSPHfHIC
                                       DIFFUSION *f*0
                                       SEDIMtflTATION
                                     	DIFFUSION WHY
                                  (i)

    Figure  4.   Theoretical collection efficiency (concluded)

    However, as the  first approximation, the  order of magnitude
of relaxation time for gas flow inside the  bubbles to reach a
steady  state can be  estimated  from an analogous case of  a gas
bubble  rising through a liquid,  Figure 5.   This will provide
some  indication of significance of the unsteady state on the
overall collection process.  Let us determine how fast the steady
flow  profile can be  attained in the bubble  illustrated in Figure 5
                                    LIQUID
           ••••• ^^
           o o
           O LJ-J
           UJ _J
           Q; CD
           — CO
           O ID
             CO
V,
                                               (AT t »  0)
              GAS BUBBLE
             Figure  5.   Flow profile of a rising bubble

-------
  382


    From the result of the analysis of Couette flow between the
annulus with one moving wall, the value required for the flow to
reach 95% of final velocity profile from the start was investi-
gated by H. S. Yu and E. M Sparrow:2
                   | = 0.32 Re                           (15)


where    x = entrance length

         R = channel width, or in this
             case, radius of bubble

        Re = Reynolds number

The time required for flow to reach 95% of final profile is
therefore:
              t =  - = 0.32R(Re)/V                      (16)
                -  Vb              b
     For a bubble 1  mm in diameter,  rate of bubble rise of 0.30 m/s,
 and a  Reynolds  number of 100,  Equation 16 gives t = 0.05 second.
 That means the  gas  flow in a  1 mm bubble would reach the steady
 state  almost  immediately,  say in  less than 0.1 second.   The motion
 of  gas,  however,  would persist as long as the bubbles are rising
 through the liquid  layer.   By the same reasoning, as soon as the
 foam bubble is  out  of the liquid, the internal gas motion would sub-
 side immediately in the same  relaxation time  (<0.1 s)  due to
 absence of surface  film motion.

     For in the  case of a screen foam generator,  initial disturbance
 of  gas movement from the foam generation process would be expected
 to  subside in the same short  period of time.   This would tend to
 indicate that due to insignificant  relative motion between moving
 bubbles and gas,  and short flow relaxation time,  contribution
 of  impaction would  appear insignificant.   This supports the
 assumption previously made.

     This concludes  the discussion of the major mechanisms that
 may be involved in  particle collection by the foam scrubber.
 These  should be reevaluated for some specific conditions of
 scrubber operation  when other  mechanisms might come into play.

-------
                                                                 383

EXPERIMENTAL EVALUATION OF FOAM SCRUBBER

     Theoretical evaluations were most helpful in designing the
experimental test unit shown in Figure 6.   Many experiments were
performed using this unit to learn more about foam generation,
foam destruction, foam flow characteristics, foam stability, foam
bubble size, and particle collection efficiency.  It would be
difficult to cover all of these tests in this presentation and
the reader is referred to the references for additional details.3/4

    Four surfactants were evaluated  in our experiments.  They are
listed in Table  2.  Most of the particle collection experiments
were performed using Tergitol surfactant.

    The foam destruction chamber was built to enable testing of
various foam destruction techniques.  Specifically, five foam
destruction techniques were investigated.  They included liquid
spray, thermal destruction, ultrasonic destruction, compressed
air spray, and high speed rotating disk.

      The  high  speed rotating  disk was  found  to  be  the  most effec-
tive  device  for  foam  destruction because of  low energy require-
ments  and minimum deterioration  of  surfactant liquor qualities
for a possible  recycle.   Consequently,  this  device was selected
and used  in  particle  collection  efficiency  evaluations.

     Two types of aerosol particles were also tested.   They were
 OOP (dioctyl phthalate)  and polypropylene glycol.   Relative to
 water the two particle types possess different wettability pro-
 perties,  the dioctyl phthalate being essentially nonwettable and
 the propylene glycol being wettable and miscible with water in
 all proportions.  The experimental aerosols were generated by
 means of a condensation aerosol generator previously described.5
Monitoring of particles  was done by different particle counters
 including a nuclei condensation counter, an optical counter
 (Royco),  and an electrical mobility analyzer.  These  instru-
ments have a combined capability to measure aerosol particles
 in the size range from 0.003 ym to about 30 ym and were selected
 depending on the particle size range investigated  at  the time.

      Collection efficiency measurements were made  with different
 scrubber residence times.   The scrubber residence  times were
 varied by changing the scrubber length.  At the same  time the
 air flow velocity through the scrubber was maintained constant.
 This  arrangement permitted performance of collection  measure-
ments with identical  foam characteristics.   The results of
 collection efficiency experiments  are summarized in graphical
 form in Figures 7 and 8.

-------
384
            ROTAMETER
        HUMIDIFI CATION
           CHAMBER
      COMPRESSED
     AIR REGULATOR
     —01
                                                                    EXHAUST
                                     FOAM
                                   DESTRUCTION
                                   CHAMBER
              Figure  6.    Experimental foam  scrubber
                                                                           DRAIN
                  Table  2
SURFACTANT  TYPES
Surfactant
Tergitol TMN

Sterox NH



Alkanol DW





Aerosol OT
(751)


Manufacturer
Union Carbide
Corp. ,
Plastics Div.
Monsanto Co. ,
Inorganic
Chemicals Dlv.

E.I. du Pont
de Nemours &
Co., Inc.



American
Cyanamid Co. ,
Industrial
Chemicals, Div.
Composition
Trimethyl nonyl
polyethylene
glycol ether
Alkyl phenol
ethylene
oxide adduct

Sodium alkyl-
aryl sulfonate




Dioctyl ester
of sodium sul-
fosuccinic acid

Chemical
type
Nonionic

Nonionic



Anionic





An ionic



Concen-
tration, %
90

99.5



28





75



Comments
Wetting and re-
wetting agent
General purpose
detergent
(melting point
30-35°C)
General use
detergent; du
Pont says it is
in tight supply
at the present
time
Fruity pungent
odor



-------
                                                                                385
         <  wo-
         I
         a.
         >-
         ¥
            60
Figure  7.
             20
              FOAM CHARACTERISTICS
              250 MESH SCREEN   ,  ,
              AIRFLOW 1.137x10, m3/s
              WATER FLOW 5 83 cnT / s
              BUBBLE SIZED 81 mm
 RELATIVE HUMIDITY
    45*  -5*   PARTICLE SIZE
    o   •   0 18 -0.32 jjm
    o   •   0.32-0.56pm
    A   A   0.56 -1.00 jjm
                   10
                 20
                               30
                                     40
                                           50
                                                 60
                                                       70
                                                             80
                                   RESIDENCE TIME, s
       Foam scrubber collection efficiency  of  polypropylene
       glycol  425  using 2%  Tergitol foam with  rotating  disk
       for foam destruction
     in
     <£
     CO
     8
     tx.
     •<
     Q_
100
         80
     *   60
     o
     iZ  40
     o

     y  20
     o
     o
Figure 8.
                 FOAM CHARACTERISTICS
                2* SOLUTION      ,  ,
                AIRFLOW 1.137x10, m/s
                WATER FLOW  5.83 cm3 Is
                                                            TERGITOL-
           PARTICLE SIZE
           o • 0.18 - 0.32 jjm
           o • 0.32 -0.56 urn
           « • 0.56 -1.00 jum
                 10
                20
                               30
                              40
50
                                                   60
                                                  70
                                                                  80
                            RESIDENCE TIME, S
       Collection of  OOP  aerosol  using  foam  scrubbing with
       rotating  disk  for  foam destruction

-------
   386

      Because  the collection of  fine particles  in  foam takes place
 mainly by diffusion and  sedimentation,  the collection efficiency
 is  a strong function of  scrubber residence time.  Both Figures 7
 and 8 demonstrate curves asymptotically approaching 100% collec-
 tion with an  increase of scrubber residence time, which agrees
 with the theoretically predicted curves reasonably well  (see
 Figures 4, a  through i).

      Bubble diameter in all cases was about 0.8 mm.  Somewhat
 higher collection efficiencies  than those predicted have been
 observed using polypropylene glycol (wettable) aerosol especially
 at  the lower  scrubber residence times (Figure  7).  This indicates
 that the compatibility between  particles and the  foam (as for
 example suggested by the aerosol wettability)  is  an important factor
 to  consider in optimization of  foam scrubber design and operation.
 Figure 8 shows several additional data  points  obtained with the
 OOP aerosol and different surfactants including Tergitol, Sterox,
 and Aerosol OT.  These points further support  the factor of
 aerosol/foam  compatibility since some changes  in  collection
 efficiency were noted due to the change of surfactant.  Even
 though the collection efficiencies were found  to be generally
 lower than those observed using Tergitol foam, proper surfactant
 selection might improve OOP collection  and approach the effici-
 ency that was observed for polypropylene glycol aerosol.

      A comparison of the collection data presented  in Figures  7
 and 8 suggests  that there might be somewhat different collection
 rate controlling factors in the cases of OOP  and  polypropylene
 glycol aerosols.  While  the small size  particles  showed better
 collection efficiency  than large size particles using DOP
 aerosols, this  trend was found  to be reversed  for particles of
 polypropylene glycol.

     Figure 7  also includes two sets of  data measured with dif-
ferent humidity in the scrubber.  Two relative  humidity levels,
5% and 45%,  were investigated using glycol aerosol and Tergitol
foam.  More humid input air increased the particle collection by
about 1 to 2%  for residence times of 13  and 26  seconds.  Since
the  accuracy of the particle concentration measurement is be-
lieved to be about 10 to 20% it is not certain whether the effect
of humidity is significant or not.

     The  effect of bubble size  on collection efficiency was also
 evaluated.   The 60  mesh screen was used to generate coarser foam
 with the  other operating parameters  kept constant.  The size of
 cells for the 60 mesh screen was photographically determined to
 be about  3.9  mm.  The  observed collection rates are presented in
 Figure  9 .   The collection  rates are  significantly less than the

-------
                                                                387
values determined for the small-celled foam (0.83 mm).  The DOP
aerosol showed a collection rate of about 40% with a particle
size between 0.56 and 1.0 mm and 80 second residence time  (75%
collection was observed with the small-celled foam).  The glycol
achieved about 84% collection under similar conditions  (98% with
small-celled foam).  Two data sets are also displayed using a
large-celled 1% Alkanol foam.  While this foam improved collec-
tion of DOP aerosol, the collection of polypropylene glycol
aerosol was essentially unchanged.

    The operating conditions which have been used in our experi-
ments give an air  (or foam) velocity in the scrubber of about
0.15 m/s  (0.5 ft/s).  It is postulated that a higher average
velocity might influence the collection rate, especially at the
point of foam generation where higher turbulence could enhance
the effect of collection by impaction.  Consequently, the ve-
locity was increased by a factor of three to about 0.45 m/s
(1.5 ft/s).  At the same time the scrubber was lengthened to
5.5 m  (18  ft) so that a 13 second residence time remained un-
changed.   The collection results for both DOP and the polyglycol
were essentially the same as for the 13 second residence time at
the lower  air velocity.

      This  supports  the  theoretical  evaluations presented earlier
which  indicated  that the  collection rates during  the  foam  forma-
tion  will  be  negligible and  that  the majority of  particles will
have  to  be collected by diffusion and  sedimentation mechanisms.

     There does,  however,  appear  to be  a point where  collection
efficiency is enhanced  by  capture of particles in the foam genera-
tion  step.  This  occurs at high liquid-to-gas ratios  causing  the
foam  scrubber to  behave similar to  a wet scrubber.  Figure 10
shows  the  effect  of distilled water sprayed  on the 250  mesh
screen in  relation to collection  efficiency  of DOP particles  in
the size range between  0.7 and  1.0  ym.  No collection of parti-
cles  is  seen  for  liquid-to-gas  ratios  below  5,130 cm3/m3  (38.4
gal./I,000 ft3).   Thus, the  collection efficiencies for foam
scrubbing reported above primarily  represent the mechanisms of
diffusion and sedimentation.

      Experience  from bench scale  operation was used to  scale-up
the foam scrubber to a  small pilot  unit with a capacity of 500
acfm.   The objectives of  this  scale-up may be summarized as
follows:  (1)  demonstrate  the foam scrubber on a  larger  scale,
 (2) test it using an industrial aerosol particle,  (3)  gain ex-
perience in scale-up, and  (4) verify a  potential  for  multiple
surfactant liquor recycles.  It may be  said  that  all  these ob-
jectives were met with  a  considerable  degree of  success.

-------
  388
      (/>
      s 100
      o
      o
      0.
         60
         40
      Us
         20 -
      o
      o
                               2 % TERGITOL FOAM EXCEPT WHERE SHOWN OTHERWISE
 Figure 9.
             POLYPROPYLENE GLYCOL AEROSOL {-5
                         DOP AEROSOL!^*
                                                   PARTICLE SIZE

                                                   • 0.18-0.32jum
                                                   • 0.32-0.56ym
                                                   * 0.56-1.00jjm
                                   _L
                                          _L
                10      20     30     40     50
                                  RESIDENCE TIME, s
                                                60
                                        70
              Foam scrubber collection  efficiency  of polypropylene
              glycol  425 and OOP  using  large-celled (3.9  mm)  foam
 /l
 <
 oo
    100
g  80
oc
<
>-
CJ
    60



g   40
Ll_
u_
UJ

I   20
S   o
                        BENCH SCALE
                         OPERATION
                  20
                             40
                            60
   GALLONS / 1000 ft

80          100
Figure 10.
               2.0
                                                             12.0
                                                                       14.0
         4.0       6.0       8.0       10.0
           WATER ROW RATE, I03 cm3/m3
Collection  efficiency values for distilled  water
sprayed on  the  screen versus water flow rate, OOP
particle  size:   0.7  - 1.0  \im

-------
                                                                  389

     The scrubber was scaled-up and operated with no difficulties
for several weeks.  Collection efficiencies were measured using
an industrial aerosol, a flyash obtained from a stoker coal-fired,
power plant, Nucla, Colorado.  The schematic of the small pilot
scale foam scrubber is shown in Figure 11.

     As  shown in  Figures  3  and  4,  the particle collection effi-
ciencies follow curves that steadily increase with  time  and
asymptotically approach 100%.   After a certain time,  the gain  in
particle collection efficiency  that would  result with a  further
increase of  residence time  becomes rather  negligible, and gets
within the range  of experimental error.  To operate the  pilot scrubber
in this  region would greatly reduce the possibilities for observing
the effect of influential variables on the scrubber collection
efficiency.  It was decided that high scrubber collection efficiencies
obtainable with longer scrubber residence  times would be sacrificed
for the  purpose of having a more flexible  pilot test  facility, and
the scrubber was  thus designed  for short residence time.

     The scrubber residence time was actually measured and
determined to be  20.5 seconds.  Other operating conditions are
listed below:
             gas  glow       0.260 m3/s          (550  acfm)
             liquid flow    6.81 x  10 4m3/s    (10.8 gpm)
             scrubber AP    1.84 kPa            (7.4  in. H20)

Collection efficiencies for submicron particles were measured
and are  illustrated in Figure 12.   They ranged between 50.0% and
75% by count for a particle size range of  0.056 pm to 1.0 ym.
Corresponding theoretical collection efficiencies for particles
in the size range between 0.1 pm and 1.0 ym and for the residence
time of  about 20  seconds are 32% to 67% by count (see Figure  4h).

     The data in Figure 12 are average collection efficiencies
over the period of 20 hours .  The  surfactant concentration during
the 20-hour experiment was kept at 0.25% by weight.   This is a
significant improvement over the bench-scale phase where about 2%
surfactant solutions were needed to generate satisfactory foams.
In addition, the  surfactant liquor requirements were  reduced from
5.17 x 10~3 m3 per m3 of gas  (38.4 gpm per 1,000 cfm) to 2.64 x
10~3 m3  per m3  of gas (19.6 gpm per 1,000 cfm).

     During the 20-hour run, the surfactant liquor was recircu-
lated.   As a consequence of the recirculation, the concentration
of particles in the scrubbing liquor steadily increased to a
level of 2 kg/m3.   No difficulties were observed in generating
foam with the surfactant liquor containing this concentration of
solids.

-------
                                 10 PMfCL MANOMETER

                                PIEXICIASS SECTION


                               —EC-SCRUBBER
                                                                                           00
                                                                                           VO
                                                                                           O
Figure 11.   Pilot-scale  foam  scrubber  process layout.

-------
                                                                      391
                     SHADED AREA - OUTLET CONCENTRATIONS
                     SHADED + UNSHADED AREAS • INLET CONCENTRATIONS
                        0.08 0.1       0.2   0.3  0.4   0.6  0.8 1.0
                           PARTICLE DIAMETER, dp, pm
Figure 12.
Average particle collection  efficiency data  (12 runs)
Percent collection  shown inside the  unshaded  area.
Scrubber residence  time 20.5 seconds.

-------
392

  FOAM SCRUBBER ECONOMICS

       Based on the experience from operating the foam scrubber
  on bench and small pilot scales,  an economic evaluation of this
  device was performed.   The economics were compared with those for
  an electrostatic precipitator,  a  high energy wet scrubber, and
  a  fabric filter.

       Figures 13  and 14  show capital and operating costs for the
  three conventional devices as well as the foam scrubber.5

  CONCLUSIONS
  Darticlfrnn" ^ sedimentation aPP*ar to be the most important
  particle collection mechanisms active in a foam scrubber.  Theo-
  retical evaluations of these mechanisms indicate that over 90%
  collection efficiencies are possible in foams with bubbles 1 mm

  of ^eo  S af ab?Ut 4° t0 6° S6COnd residence time.  ThI remits
  of theoretical evaluations seem to agree well with the results
  obtained by experiments in a bench and a small pilot scale foam
  are comoar^ "^ ^ °^ati^ costs of foam scribing
  M^in?P i   4-   r0? the conventional particle collectors in-
  cluding electrostatic precipitators, fabric filters, and hiah
               ™^6"-  At Present,  however, the foam slubber
               be1called commercially available and fully provln
              o?%hr?ntS *" K^  tO test  the Performance^'
              of the  foam scrubber  in industrial applications.
 REFERENCES

 1.  Fuchs, N. A. , The Mechanics of Aerosols, A Pergamon Press
     Book, The Macmillan Company, New York 1964.

 2.  Yu, H. S., and E. M. Sparrow.  Flow Development in a Channel
     Having a Longitudinally Moving Wall.  Journal of Applied
     Mechanics.  37 (2) :498-507, June 1970.             FP^ea

 3.  Ctvrtnicek, T. E, T. F. Walburg, C. M. Moscowitz, and H. H. S
     Yu, Monsanto Research Corporation, Dayton, Ohio, Application
                              '
                                                    ,
                       t0 Fine' ^tide Control, Phase I, Contract
                   ,  May 1976, MRC-DA-556, EPA-600/2-76-125.

 4.   Ctvrtnicek,  T.  E. ,  C.  M.  Moscowitz, S. J. Rusek, L.  N. Cash,
     Monsanto Research  Corporation, Dayton, Ohio, Application of
     Foam Scrubbing  to  Fine Particle Control,  Phase II, Contract
     No.  68-02-1453,  December  1976.

 5.   Liu,  B.  Y. H.,  K.  T. Whitby,  and H. H. S. Yu.   A Condensation
     Aerosol  Generator  for  Producing Monodispersal  Aerosols in the
     Size  Range of 0.036 to 1.3 Microns.  (Presented at the 6th
     Inst.  Conf.  on Condensation Nuclei.  Paris. May 1966.)
     J. de  Recherches Atmos.  (Paris.)   No.  3:397-406 (1966).

-------
                                                                                        393
                                • CAPACITY. ACFM
       i
               FOAM SCRUBBER
               RESIDENCE
                                                            aECTROSTATIC
                                                             PRECIPITATOR
                                                           | HIGH ENERGY
                                                           L-WET SCRUBBER
                                              FABRIC FILTER

                                              HIGH ENERGY WET SCRUBBER

                                              ELECTROSTATIC PRECIPITATOR

                                              FOAM SCRUBBER
          1.0
                               10.0
                                                   100
                                                                      1000
                                    CAPACITY, m'/s
Figure  13.   Capital  cost  for particulate  control  devices
              105
           s
           o
           «c
           85
           0.
           o
                 T-|—I  J
                                 CAPACITY, ACFM
                            10
                                                                 106
           , FOAM SCRUBBER - SURFACTANT MAKEUP (-2%)
               (40s)
                                       CASE A
; FABRIC
- FILTER
                                 HIGH ENERGY WET SCRUBBER
                      FOAM SCRUBBER -NO
                       SURFACTANT MAKEUP
                                 (40$
                 - ELECTROSTATIC
                  PRECIPITATOR
                                                                   10.0  -
                                                                   i.o
                                    CAPACITY, nr/s
 Figure 14  .    Operating  cost for  particulate  control  devices

-------
  394
WRITTEN DISCUSSION

George E.R. Lamb
Textile Research Institute
Princeton, New Jersey  08540

    The main problem of the technique, according to the data shown in Figure
12 of the paper, is the low collection efficiency.  If the method is to be
acceptable for pollution control applications, the efficiency will have to
rise to the order of 99%. We can estimate the bubble size required to achieve
this by using a relationship found at TRI in connection with fabric filtration
We have found that capture efficiency for a particle passing through a pore
was a function of the log of r /s where r  is the Brownian radius of a
particle,  and s is the pore diameter.

    The Brownian radius is given by


              yr
     r
      e
 where   n  is  the  viscosity  of  air
         1  the mean free  path in  air
         r  the particle radius
         t  the residence  time

    Table I gives  values  of  r   for various particle sizes for a residence time
 of 10 seconds.  Figure 1  shows  values of r /s where s is 0.5, 1 and 2 mm, where
 s is  now the bubble diameter.   Capture efficiency values were obtained from
 Figure 4a,d and g  at 10 second  residence times and all points in Figure 1
 seem      to fall on a common curve.  The curve indicates that for 99% efficiency
 re/s  must  be about  0.3.

    Table  II gives  bubble diameters (in pm) that will satisfy this requirement
 at 10, 20  and  100 second residence times.  Since the capture of one-pm
 particles  is aided  by sedimentation, we may consider the bubbles for 0.5 urn
 particles  to be the  largest  permissible and we see that at 100 second residence
 time, thesg  are 1/4 mm.  The formation of bubbles having a surface tension of
 70 ergs/cm  requires pressure drops shown in Table III.  We see that formation
 of 1/4 mm  bubbles only requires a theoretical pressure drop of about 7 inches
 of water and is thus  within tolerable limits.  The values in Table III were
 calculated on the assumption that the bubbles are already joined to neighbor-
 ing bubbles as they form,  i.e.,  that each bubble wall serves as the common
wall to two adjacent bubbles.  If the bubbles are formed individually, the
pressure drops in Table III must be doubled,  since a bubble has two surfaces.

-------
                                                     395
                TABLE I

     BROWNIAN RADIUS OF VARIOUS PARTICLES
Particle radius r
(urn)
1
0.
0.
0.
0.

5
1
05
01
Brownian radius, r
(vim)
16
23
66
114
482
               TABLE II

BUBBLE DIAMETERS GIVING  99%  CAPTURE  EFFICIENCY
Particle
radius
(pm)
1
0.5
0.1
0.05
0.01
Residence time,
seconds
10
51
78
220
380
1600
20
72
110
311
540
2260
100
160
250
700
1200
5000
                   TABLE III
   PRESSURE  DROP REQUIRED FOR FORMATION OF
             BUBBLES OF 70 ERGS/CM2
Bubble Diameter
(mm)
1
0.5
0.25
AP
(mm H20)
42
84
168

-------
                                                                                                                                      00
                                                                                                                                      s
o u

-------
                                                                             397
    The paper makes no mention of the effects of temperatures which must be
limited to somewhere short of 100°C, and presumably limits application of the
technique.
OPEN DISCUSSION

Cooper:  I was struck with the lack of time dependence of the low collection
efficiency values, for instance, in the DOP aerosols,page 16.  One of the
things being suggested is that there is the occurrence of another generation
mechanism.  We have the formation of very fine sodium chloride particles
by the breaking of the bubbles.  Maybe the device is better than we think
because when the bubbles break, they are putting some  of the material back
into the air.  I was wondering if you had a comment on that.

Ctvrtnicek:  You mentioned that the collection efficiencies were high on the
500 cfm unit.  I tried to explain that the unit was purposefully designed to
have low collection efficiencies, that we wanted to be on that steep end of
the curve.  Our results were somewhat better than the theoretical collection
efficiencies.  We extended the time for deposition and still could not reach
900 gauss collection efficiency. Whatever we achieved was not achieved through
any extraordinary effort to optimize the device.  We were primarily after a
demonstration of the model.  We never optimized bubble generation, or the
geometric structure.  For example, in the bench scale, we were generating
foam destruction on the order of 10  particles per cubic centimeter.  When
we went to the pilot scale, we cut that down by a factor of 10.  I believe
that it is very likely that, by proper geometry of the disc and impaction
surfaces,this can be reduced even further.  The same goes for the foam
generation.  We really did not try to optimize it at all.  Once we got a
decent foam we could use to make experiments, we just performed the experiments.
There were no optimization efforts at all.

Lawson:  You mentioned that the collection efficiency might be enhanced if
you had smaller bubbles.  Unfortunately, one of the undesirable effects,
with the understanding that there has been no real effort to optimize the
situation, is that if you go to smaller bubbles, you probably need even more
surfactant to get the total bubble volume that is required.  That is a possibility
because the wall surface of the bubble will probably not go down nearly as
fast as the  size of the bubble.  You will end up making smaller bubbles
requiring even more surfactant.  So surfactant recovery will really be
important in  that case.

Ctvrtnicek:  That is a good point.  However, when you can believe in recycles,
surfactant really does become  a factor.  In operating this unit we used semi-
horizontal and vertical modes.  I believed that the  device worked much better

-------
398
   in  the vertical mode because you were continuously getting a drainage of
   the foam.  As  the foam becomes drier, it is easier to destroy.  You can
   reduce the energies for destruction and make it even more effective.

-------
                                                                399
"THE ANOMALOUS BEHAVIOR OF ASBESTOS FIBERS"


   J. W. Gentry


   Department of Chemical Engineering, University of Maryland


INTRODUCTION

   Asbestos is a term used to include a number of fibrous silcate
minerals including amosite, crocidolite, chrysotile, tremolite,
and anthophylite.  Because of its inflammability it is widely
used for insulation in clothing and construction, for brake-
linings, and as a binder with cement and asphalt.

   Recently there has been evidence showing a 50-300% increase
in cancers of the digestive system among insulation workers  [1].
A significant increase of lung cancers  [2]  especially mesothe-
lioma among workers in the asbestos mining has been found.  In
this case a significant synergistic effect was demonstrated for
smokers.  Finally, an examination of cancerous tissue indicated
the pressence [3] of longer fibers (fiber length > 5y); although
no similar studies have been carried out for healthy tissue.

   This latter effect provides some insight into the reasons for
increased interest into asbestos fibers.  Probably the major
emphasis to date has been in the detection of fibers.  First the
concentrations are very low, in the order of 105 to 10b particles/
liter for rainwater.  This corresponds  to a weight fraction of
107 - 106.

   Typical concentrations of^fibers in  air at rural locations are
of the order of 10  fibers/m .  Such concentrations require a
lengthy sampling time which has so far  precluded automated
instrumentation  [5].  Furthermore, as is discussed, below only a
fraction of the fibers can be asbestos.  In recent years the
separation of fibers has become recognized as an important prob-
lem.  For example, are only long fibers carcinogenic?  The point
is important in that the general population is typically exposed
only to short fibers whereas industrial workers are exposed to
a wide spectrum of particle sizes.  There is, also, the need to
separate fibers from spherical particles in industrial regions.
Typically fibers provide only a small fraction of the total

-------
 400
 fibers.

   One point  that  should be  emphasized  is  the difference between
 the mineral fibers.  Amosite and crocidolite are rod-like, they
 appear rigid, and  their shape can be approximated by a prolate
 spheroid.  Chrysotile which  is the most widely used fiber con-
 sists of hollow, bent tubes  which appear to be flexible.
 Anthophylite  appears to be mostly plate-like resembling shingles.
 For most measurements amosite or crocidolite are recommended.

 GENERAL PROPERTIES

   Before the specific properties of non-spherical particles are
 discussed, it is well to outline the steps required for a
 systematic development of spherical particles.  Condensation and
 evaporation are, of course,  less important than for spherical
 particles, since non-spherical particles are solids.  Similarly
 the problem of coagulation has been less studied, primarily
 because of the much more complicated expressions and the
 stochastic nature of the collision process.  Other than these
 differences the problems are similar.   First, the aerosols must
 be generated.  The question  is what is  the analog of a monodisperse
 distribution of spherical particles.  Next the particles must be
 measured.  The methods of measuring non-spherical particles by
 microscopy, by aerodynamic behavior and by optical methods are
 related to those of spherical particles.  In this lecture,
 emphasis will be focused primarily on the aerodynamic behavior
 because it is the most thoroughly investigated.  Finally there
 are the separation processes - electrostatic precipitation,
 filtration, gravitational settling in an elutriator all of which
 are just beginning to be studied.

GENERATION

   The ideal generator would produce a uniform, monodisperse cloud
of fibers with a known and controllable charge distribution.   A
monodisperse fiber would be one having the same shape and the
 same dimensions.  For the approximation to this fiber one would
have an amosite fiber which can be approximated by a prolate
 spheroid.  The monodisperse distribution would then be amosite
 fibers having the same length and diameter.  Unfortunately we do
not have the analog of latex particles or a Sinclair-La Mer
generator for non-spherical particles.

   Many of the current asbestos generators which work by scraping

-------
                                                                 401
a packed cake of ground asbestos fibers are completely unaccept-
able for experimental studies of the properties of fibers in
that the distribution is very broad consisting primarily of
aggregate rather than the necessary individual fibers.

   The most successful generator [6] currently available uses a
vibrating bed with the flow rates through the bed at sub-fluid-
ization velocities.  The asbestos fibers are ground and dried
prior to placing in the generator.   Mixing the asbestos with
small metal spheres (Nickel) proved unsuccessful in that the
fibers produced by the generator were coated with a thin layer
of nickel dust.  The generator described above has been proven
to give a narrow particle size distribution with the diameter
of the fibers ranging from 0.2-0.4y and the lengths from 1 to
10y.  The mean value of the aspect ratio  (the ratio of major
to minor axis) was approximately 10.
                                                            2
   The generator can  produce fibers in the concentration 10
104 fibers/cm*.  Examination of the fibers indicate that
approximately 15% of the weight of the fibers is in agglomerates
but approximately 85% of the weight is in individual fibers.
The particles are mostly shorter than 5y and are relatively highly
charged in comparison to spherical particles.

   It has been suggested that non-spherical particles can be gen-
erated using solutions which yield non-spherical crystals on
evaporation  [7].  Also, certain types of spores are oblong and
could be used to model non-spherical particles.  Neither of
these suggestions have been successfully developed.

DETECTION

   The only currently reliable method of determining the size
and number of asbestos particles is electron microscopy.  The
key points that have been raised regarding microscopy of non-
spherical particles can be  summarized as follows:

       1. Not all  fiberous  particles are asbestos.

       2. Electron microscopy gives only a 2 dimensional pro-
jection.  In order to unambiguously characterize the  fiber it
should be examined from three dimensions.
       3. Nuclepore filters are the best current technology  for
collecting a  sample.

-------
 402
   The  first  result was demonstrated by  Spurny and co-workers  in
 their measurement of air  in  the mountains near the source of
 the Ruhr.  They examined  the elemental composition of each
 apparent  fiber  (particles with an aspect ratio >  3).  Any fiber
 which contained Fe, Si, and  Mg was classified as  a possible
 asbestos  fibers.  These experiments showed that no more than 5%
 of the  sampled particles  could be asbestos.  Similar results
 were found for air sampled from Munster  a city of 200,000 with-
 out heavy industry.

   In analyzing asbestos  fibers, the prevailing assumption is
 that fibers lie flat on the  surface, so  that the  particle di-
 mensions  can  be obtained  directly from the projection.  Doubt
 has been  raised regarding this point, and studies with spores
 (1 < B  <  3) have shown some  particles aligned with the surface.
 Because of the non-random orientation of particles, Cauchy's
 theorem cannot be applied to obtain a particle size and the
 particle  must be viewed from three angles.  The work required
 is painstaking and tedious.   The hypothesis of the fibers
 lying flat has not been adequately tested.

   Three  methods have been examined for  collecting samples -
 filtration through membrane  or nuclepore filters, electrostatic
 precipitation, and thermal precipitation.  The results of a
 comparative study are discussed extensively in a  recent review
 article [5].  Two conclusions were drawn from the article -
 nuclepore filters are superior to membrane filters in the
 preparation of the sample and the filters require considerably
 less time per experiment  than do the thermal or electrostatic
 precipitator.

AERODYNAMIC DIAMETER

   Perhaps, the most extensively studied aspect of non-spherical
 particles is  their behavior  in free fall.  In comparing experi-
mental measurements with  theoretical calculations - use is made
of the implied assumption that the particles may be treated as
prolate or oblate spheroids.

   In discussing results  for non-spherical particles it is
necessary to define the aerodynamic diameter and the dynamic shape
 factor for a non-spherical particle.

   The aerodynamic diameter D   of a particle is defined as a
                             ae

-------
                                                                 403
sphere of unit density having the same settling velocity as the
particle at low Reynolds number.  The aerodynamic diameter is
related to the volume equivalent diameter D  and the dynamic
shape factor K by equating the settling velocities giving:

   v     1 pCDae      _  1 p De K _
    s   18 nF(Dae,vs)   18 n F(De,vs)


where the volume equivalent diameter is the diameter of a
hypothetical sphere having the same volume  as  the particle.
The function F(D,V )  is related to the drag coefficient (CD)
and Reynolds numbe?  (Re) by
               T)(*
   F(D,V) = CD If

   For the important limiting case of continuum mechanics and
creeping flow the aerodynamic diameter is related to the
equivalent diameter by

          D2
   K - £--§-

       P°Dae

   For spheroids and cylinders, the theoretical shape factors
have been found for motion parallel to the polar axis K, , and
perpendicular to the polar axis K22.  For a prolate spheroid
these results are
   Kll
            B -1                        'B -1
where B is the aspect ratio  (the ratio of major to minor axis).
Similar expressions exist for cylinders and oblabe spheroids
[8].

   Several key points should be stressed regarding the shape
factors for non-spherical particles.  First the shape factors
are a function of the aspect ratio alone.  Secondly, the ratio
of K_2/K   is bounded between 1 and 2.  The function K22 is a
slowly increasing function of B.

-------
 404
    The  effective  or  experimental  value  of  the shape  factor can
 be  calculated  by
         F.2TT

           66  [K.. + K0J  f(0)  sin9
where  f(6)  is  the angular distribution of  the  particle  orientation.
It  is  precisely the difficulties with the  form of  f(6)  which  is
the subject of much uncertainty and which  accounts for  the  curious
behavior of non-spherical particles that have  been observed
experimentally.  These effects can be summarized in  the table below.

                              Table 1

             1. Random-Orientation Assumption
             2. Cans Diffusion Solution
                 a. Gallily-Cohen Simulations
                 b. Comparison with Random-Orientation  Assumption
             3. Collection Efficiency Measurements
                 a. Spiral Centrifuge
                 b. Inertial Impactor
                 c. Filters
                 d. Elutriators

   The basic result of all experimental measurements have been that
the particles  tend to acquire a preferential orientation, usually
in the optimum position for penetration, in any flow field  where
there are inhomogenieties in the flow field.

   For the  special case of free-fall of particles  in a  stationary
medium, the shape factor can be computed if the functional  form
of f (0) is  known.   By suitable experiments, experimental value of
the aerodynamic diameter can be measured and compared with  theory.
Assuming creeping flow and a continuum, one has
   D
    ae
18nV
      cosa
           P
            o
The product V cosa is the vertical component of the settling
velocity where a is the angle of descent of particles.  As is
known, non-spherical particles do not descend vertically, but
at an angle a [9] which can be related to the ratio of the shape

-------
                                                                405
factors for movement parallel and perpendicular to the major axis
and the angle of orientation 0.  The angle a is defined by
   tana =
          - —
          (K22/Kn+tan 6)
   In a recent note [10] ,  comparisons were made of calculations
for the aerodynamic diameter using the assumption of random
orientation, the assumption of an averaged orientation, and a
stochastic  simulation where the particle was subject to fre-
quent changes in orientation with the probability chosen to
correspond to Cans solution [12] .  It was found that the average
value of the aerodynamic diameters were similar for all three
calculations, but that the scatter of measurements or the
standard deviation of the aerodynamic diameter was less for the
numerical simulations than found for the assumption of random
orientation.  It should be stressed that the two calculations
are based  upon  different assumptions, and that the simulation
gives a lesser degree of randomness manifested by a lower
standard deviation than does the assumption of random orientation.
The discrepancy indicated here emphasizes the need for experi-
mental measurements to check the assumption of random orientation
(which is the prevailing assumption) and the need for looking
at variances of  stochastic processes rather than mean quantities,

   Quanitative experiments in which the aerodynamic properties of
non- spherical particles have been examined include sedimentation
experiments in silicone oil, experiments with the spiral centri-
fuge, experiments with an inertial impactor, and experiments
with nuclepore filters.

   The spiral centrifuge is widely used for particle separation
and analysis of polydisperse aerosols.  The particles are removed
by centripetal force as a function of their aerodynamic diameters,
Particles having a smaller aerodynamic diameter are deposited
on a metal strip lining the walls of the spiral nearer the gas
inlet then larger particles.  The advantage of the instrument
is that there are distinct regions of deposition depending on
the aerodynamic diameters.

   Experimental measurements  (primarily by Stober [13] and co-
workers) have found the following important results:

-------
 406
        1. The aerodynamic behavior of clumps of monodisperse
latex spheres differ depending on whether they are ball-like or
chain-like clusters.  Specifically

    D   a N '    for chain aggregates
     cl6

and D   a N '    for clusters.
     ae

        2. Experiments with aerosols generated from an exploding
wire in which the particles showed a high degree of agglomeration
could be correlated with similar expressions as the latex
particles [14].   Two separate regions were found with R  for the
lower values of N correlating according to the chain-liice clusters




        3. Experiments with asbestos fibers  (amosite and
crocidolite which both provide rod-like fibers) showed that the
aerodynamic diameter could be correlated with expressions of the
form [13]


    Dae a al Dl (L/D)b

where D is the diameter of the fiber,  a was found to be approxi-
mately 2 and the exponent  b ranged between 0.1 and 0.2.  It
should be noted that there was considerable scatter in the data
and that the L/D ratio used in this study was approximately 10.

    In summary the spiral centrifuge can be used to separate non-
spherical particles, however the separation is strongly related
to the diameter and a weak function of the aspect ratio indicating
that other methods will probably be necessary in order to deter-
mine the length.  It should be noted that the deposition length
in the spiral centrifuge is a calibrated quantity.

OPTICAL METHODS

    The optical properties of fibers are not only effected by
the physical properties of the fiber but the angle of orientation
as well.  Although the problem of back-scattering from an arbitrary
prolate spheroid has been solved for a conducting spheroid at
direct incidence [15], the problem for a general orientation has
not been.  Further as pointed out in the previous section the

-------
                                                                 407
orientation of the fibers are generally not known.

    Several investigators have measured the fiber properties with
a Royco.  Early experiments reported good agreement with micro-
scopic  [16] examination for large fibers, but recent experiments
show that the distribution of the "apparent [17] optical diameter"
is significantly broader than the distribution of the diameter,
length, or aerodynamic diameter determined from measurements of
the individual particles.  The location of the maximum for the
aerodynamic diameter and the "optical" diameter almost overlapped.

    By modifying the Royco so that higher velocities could be
obtained Gentry and Spurny demonstrated an apparent narrowing of
the distribution.  This effect was attributed to the particles
acquiring a less random orientation.  Their experiments suggest
that forcing the particles to acquire a preferred orientation
may be the only method for measuring the particle distribution
by light scattering.

    Three methods have been suggested for imposing a preferred
orientation on the particles.  These methods are electrostatic
forces, magnetic fields, and by increasing the flow rate.

    Generally speaking, all the basic problems - the rate unipolar
charging of the particles, the distribution of charges on the
particles, the mobility of particles in an electric field and
the charge distribution are unsolved.  Preliminary measurements
of doublets of latex particles [18] in an electrical field show
an alignment with the particles aligned parallel to the field.
Qualitative experiments indicate that fibers tend to be more
heavily charged than spherical particles.

    The use of magnetic field to align particles has been suggested
by Timbrell [19] as a means of particle separation.  The method
has been tested in quiescent liquids.  However, the use of the
method for particles in air has  not been tested as yet.  There
are several drawbacks in the proposed method which can be in-
dicated as follows:

        1. The fibers may take orientations parallel or per-
pendicular to the field, including both orientations.

        2. The preferred orientation can not be predicted in the
absence of experiment, and

-------
 408
         3. When  fibers are cleaved along the major axis, the
preferred orientation may be opposite of the parent fiber.

    It should be pointed out that for asbestos fibers no treatment
is necessary and that it has been shown that non-magnetic fibers
can be treated so that they can be aligned in the field.

    There is little doubt that the presence of a small gradients
can induce alignment in fibers.  The results are striking, as
demonstrated by the following set of photographs taken for a
series of nuclepore filters in series.  The filters range through
the sizes 8.0y to 0.2y.  The length and diameters of the fibers
were measured.

    What is striking about this measurement is that the particles
deposited on the last stage penetrated a filter with openings of
0.6y diameter.  Since the average dimension of the fibers was a
length of 2.3y and a diameter of 0.4y, only if the fibers were
aligned with the gas stream lines could they penetrate to the
backing filter.

    Similar results have been found for a cascade impactor where
almost 80% (by mass) of the particles penetrate to the backing
filter although measurements of the aerodynamic diameter (cal-
culated from electron micrographs of the fibers)  indicate most
of the particles should be deposited on the 4th or 5th stage.

SUMMARY

    The characterization of asbestos fibers (or non-spherical
particles)  provide a number of significant differences to the
characterization of spherical particles.

    For efficiency measurements of spherical particles it is
necessary to measure only the relative intensities of mono-
disperse particles.   For fibers there are no monodisperse particles,
and the amount of light scattered depends on orientation as well
as the length and diameter.

    For spherical particles the distribution can  commonly be
characterized by a log-normal distribution.  For  non-spherical
particles there are additional dimensions that must be accounted
for in the distribution.   In addition average parameters such as
the optical diameter can not be characterized by  a simple dis-

-------
                                                                 409
tribution.

    There, however, have been significant discoveries that indi-
cate the directions of the future.  These include

          (i) The development of a generator capable of giving
reproducible distributions of individual fibers,

        (ii) The demonstration that only a small fraction of
fibers in ambient air are asbestos,

       (iii) The demonstration that the tendency of fibers to
align with the gas stream lines in the presence of small grad-
ients is  sufficiently great that inertial impactors are useless,
and

        (iv) The proof that electrical, magnetic, or differences
in flow gradients can impose a preferred orientation on the
fibers.

REFERENCES

 [1]  Hallenbeck, W. H. and C. S. Hesse, Reviews on Environmental
     Health, II,  (1977).

 [2]  Selikoff, I., Environmental Health Perspect., £:  299-305
      (1974).

 [3]  Borow, M. , A. Conston, L. Livornese, N.  Schalat,  Chest, 64_,
     641-646  (1973).

 [4]  Hesse, C. S.  and W. H. Hallenbeck, Atmospheric Environment,
      (1977).

 [5]  Spurny, K. R., J. W. Gentry,  W.  Stober,  "Sampling and Analysis
     of Fibrous Aerosol  Particles", Ed. D.  Shaw (1977).

 [6]  Spurny, K. R., C. Boose, and  D.  Hochrainer,  Staub-Reinhalt
     Luft,  3^, 440  (1975).

 [7]  Kerker, M.,  in Assessment of  Airborne  Particles,  ed.  by T. E.
     Mercer et al., Charles Thomas (1972).

 [8]  Fuchs, N. A., The Mechanics  of Aerosols, Pergamon Press (1964)

-------
 410
[9]  Cans, R. , S. Ber. Bayer. Akad, Wiss.y 41, 191  (1911).

[10] Gentry, J. W., "Comments on a Paper by Gallily and Cohen",
     to be published, J. Colloid and Interface Sci.,  (1977).

[11] Gallily, I. and A. Cohen, J. Colloid and Interface Sci.,  56,
     443 (1976).                                               —

[12] Cans,  R.,  Ann. Phys. (Leipzig), 86, 628-656  (1928).

[13] Stober, w., H. Flachsbart, D. Hochrainer, Staub, 30, 277,
     (1970).                                   	  —

[14] Kops,  J. A. M., G. Dibbets, L. Hermans, J.F. v.d. Vate, J.
     Aerosol Sci., £, 329 (1975).

[15] Schultz, F., "Scattering by a Prolate Spheroid", ERI, Univ.
     of Michigan (1950) .

[16] Addingley, C. H.,  Ann.  Occup. Hyg., j),  73 (1966).

[17] Gentry, J. and K.  Spurny, "The Effect of Orientation on the
     Optical and Collection Properties of Fibers", GAF
     Jahreskongress, Bad Soden (1976).

[18] Boose,  C.  and D.  Hochrainer, "Orientation of Elongated
     Particles  in an Electric Field",  GAF Jahreskongress, Bad
     Soden  (1974).

[19] Timbrell,  V.,  Ann. Occup. Hyg.,  18, 299 (1975).

-------
                                                                            411

OPEN DISCUSSION

Gentry:  The reason I presented more problems than solutions, is because
there are more problems than solutions.  Secondly, i. > 3   and should be 6.
                                                   D

Friedlander:  At the bottom of page 9, about the use of magnetic fields, no
statement is made about what the magnetic and dielectric fields are.

Gentry:  What has been suggested here is using the property that the
particles will fall at different angles as a method of separation.  A series
of studies have been carried out that investigated this type of thing.
First by dropping particles without a magnetic field and then with a magnetic
field to align them.  The particles were aligned in aqueous solutions which
solidified.  We looked at the solidified mass and found that the particles
aligned either parallel or perpendicular to the field, depending on which
asbestos was used.  It was not merely a matter of whether the magnetic field
was on or off.  For example, by using permalite from Uganda and permalite
from Uganda and  from South  Africa,  there were different preferential
orientations.

Oder:  I have been involved in a lot of mineralogical studies with asbestos,
and generally speaking, asbestos is associated with iron ore processing. The
magnetic properties of asbestos are usually associated with the impurities
attached to the surface.  When I looked at your slides, the thing that
struck me was that the crystals seemed to be very clean.  Nothing was on
the surface at all.

Gentry:  Let me point out that the asbestos, which is pure  (at a commercial
grade) to start with, is not of the same magnitude of particle generation
as that which is usually dealt with.   Our generator output  is about 13
milligrams  to .13 milligrams per cubic meter.  This is probably why the
particles are so clean.

Oder:  Since you have cleaned off the  fibers, the fibers  are not as  they
would  occur in nature, but  as attached to magnetic particles.  The magnetism
left over would probably be diamagnetism for fibers that would have very high
substitutional iron  or paramagnetic depending on  the  iron composition.

Friedlander:  I would go further than  that and say that just by taking  a
chemical analysis, you can  not tell.   An example  is that the alpha  and
gamma  phases of  the  iron would be quite different.  Or maybe for  iron oxide,
one case would be parallel  and the  other ferromagnetic.  So just  knowing the
composition does not tell you  the answer either.

Gentry:   If you  use  chrysotile,  it  will cleave along  the major axis.  The
fibrils will  then have a different  magnetic  orientation.

Friedlander:  The paper says:  "The presence of  small gradients will  produce

-------
412
  alignment of fibers."  What gradients are you talking about? Flow field?
  Magnetic field?

  Gentry:  Gradients of flow field.


  Billings:   Isn't asbestos of several different types?  Aren't some quartz
  and  some serpentine or magnesium?
  Gentry:  In our study, we defined  three types of asbestos.  They are
  chrysotile, crocidolite and amosite. Other people have carried out studies
  with tremolite and other asbestos types, but all basically consist of iron,
  magnesium and silicon.  Silicon dioxide is not considered asbestos.

  Oder:  Isn't there a dependence between biological activity and fiber length?

  Gentry:  It is speculated that way and studies have been done.  It has
  always been associated with longer particles.  It is an interesting point
  now that carcinogens are important.  People have suggested that only particles
  over 5 microns have any carcinogenic effect.  The size dependence is deter-
  mined by autopsy.   If you look at a lung with a lesion, you will only find
  particles of over  5 microns.

  Billings:  I would like to comment on your measurement technique.  There is
  a tendency for particles to bounce on the impactor stages and if you are
  not using some type of grease, there is a high probability that these
  particles just bounced all the way through.  Once they bounce, they will go
  all the way to the back-up filter, because you do not have an efficiency
  going from zero to 100%.  If goes from zero up to 90% then goes back down.

  Gentry:  I am fairly certain that for the most part there was a backing
  filter of petroleum jelly on each stage.  This was used as a substrate
  for collecting particles and acts somewhat as an impactor.

      The 90 to 95%  of the particles that look like asbestos were predominately
  calcium sulfate.   The analysis of chemicals was obtained by x-ray phoresence.

  Lbffler:   Do the particles coming out of the fluidized bed carry any
  charge?  What is their behavior?

  Gentry:  They are  charged.  We put the particles through a decharger to
  study them.

-------
                                                                 413
HIGH TEMPERATURE FILTRATION - TECHNICAL PROSPECTS AND AS MEANS
FOR CHECK ON THEORIES.

    Michel M. Benarie

    Institut National de Recherche Chimique Appliquee,B.P. 1
    Vert-le-Petit,France.
ABSTRACT

    Adding temperature as a variable to the classic test parame-
ters ( particle size and face velocity ) demonstrated that the
flow field in fibrous filters.even at low fiber Reynold's number,
shows departures from laminarity.Two phenomena might be involved:
a. unsteady flow at capillary entrances and branchings and b. se-
condary flow.
    These turbulence-like phenomena improve the filter performance
in the sense of lowering pressure drop and increasing efficiency.
    It is surmised that increasing the turbulence within the mat
could result in improved industrial filters.

O.INTRODUCTION

    Filtration through fibrous media at elevated temperatures is
a well established practice in the cement and other industries.
From the environmental point of view this process has the advan-
tage that buoyancy of the hot gases is preserved and provides
better dispersal.Nuclear technology also uses high temperature
filtration.Considering its actual and potential importance,it is
amazing that but so few papers were published about filtration at
higher temperatures : First et al.(1),Billings et al.(2),Spaite
et al.(3) .Thring and Strauss (4),Pich and Binek (5),Strauss and
Lancaster (6),Dyment (?).But when you get down to experimentation
and notice that tests which normally take half an hour need a
whole day or more at high temperature then the amazement stops.
Some of the difficulties encountered are: the nature of the aero-
sol and its dispersal; problems in attaining thermal equilibrium
within the apparatus; difficulties in attaining gas-tightness;
delay  of cooling of the apparatus to enable the change of the

-------
414
  filter in view of weighing...)•It goes without saying,that when
  "high temperature" is not 50O°C for research purposes but just
  10O°C for technical assessment as described in Sect.  3.2, the ex-
  perimental obstacles become less forbidding.

      Besides its technical importance,  tests at elevated tempera-
  tures allow deeper comprehension of the real behavior of fibrous
  filters.Given the well known inaccuracies of the filter testing,
  it is not surprising to see many a theory confirmed or  at least
  not being contradicted by experiments  made in the three-dimensio-
  nal efficiency-particle size-face velocity space.But  with one
  more dimension added,namely temperature,the challenge on the mo-
  del-concept becomes harder.Dimensional analysis  with respect to
  temperature gives a hitherto unsuspected  insight in the physics
  of filtration.Discrepancies which seemed  previously incidental,
  become apparent and this way our understanding about  what happens
  might be  improved.

      Two aspects of the behavior of filters at elevated  tempera-
  tures are of concern here :  their pressure drop and their effi-
  cacity.The subject has been reviewed more or less recently in a
  very comrehensive way by several authors  : Chen (8),Rich (9),
  Spurny (1O) and Davies (11).Thus we assume that the theory of
  aerosol filtration by fibrous filters  as  well as the  basic ter-
  minology  are well known and therefore  no  details will be given
  here.

  1.PRESSURE DROP AT ELEVATED TEMPERATURES

      There are four lines of  basic approach for the  understanding
  of the pressure drop of filters:

      a.The  channel theory,  b.the drag  theory,  c.the fiber drag
  and ineraction (  e.g. the Kuwabara(l2) flow field ) and d.the
  geometrical theory of pore structure (13).

      Even  the hydrodynamically     most elaborate of  these theo-
  ries as those of  Happel (14), Fuchs and Stechkina (15)  or Spiel-
  man and Goren (16) and which at room temperature are  experimen-
  tally well verified,show one common feature :  the pressure drop
  is directly proporttnal to the dynamic viscosity of the carrier
  gas.The reason is the more or less implicit use of  Stokes1  lami-
  marity criterion,i.e. that the streamlines have up  and  downstream
  symmetry.

-------
                                                                  415
    Our experiments show against all theoretical expectation that
at variable temperature the pressure drop does not follow exactly
the increase of the gas viscosity: Figs 1 and 2 show that the
pressure drop of the media tested either in air or in argon in-
creases less steeply than does the dynamic viscosity.For all ex-
perimental detail,see Quetier (17).
11
1 III


-
(vnVM H 0 . pale..
'f ii Tn n^^' s ft^4o
fs
4* _
/$r taic.
^* /S+
Awow^r x X
j^r ^f "~~ —
s*t
to*/*
\ \ 1 i ~"

1&0

160


144

120

100


40

T 300° 400° 500° 600° K
60
g. (1 ) - Increase of the pres-
sure drop with tempe- SO
rature of a sintered
I 111
_ _
calc.
"T 6 Kfa , ""•
1^.U A *+^
*W^Tfn n^U ^^x
— xx^ —

^
x
_ ^^^ _
If.** i
Vs 10 CW/S
calc.
x/e*P--
x^X
x^*
>^ _
*'
>X/*5c-m/t -
»• ' I / [
           glass filter,average
           pore diameter  40 - 9O
           micron,at constant
           face velocity,in air
           and argon.
  T 300°
400°    500°  600«K
Fig. (2) - Increase of the
           pressure drop with
           temperature of a sin-
           tered glass filter
           of average pore dia~
           meter of 2O - 4O mic-
           ron,as a function of
           face velocity.

-------
416
      Two different mechanisms are susceptible to be involved in
  such an effect:

      a.  For short tubes the pressure drop is  increased by  a kine-
  tic energy term caused by the end effect involved in  establishing
  the steady laminar flow:
            +  a?T                                          (2)

      While the first  term of  Eqs.(l)  or  (2)  indicates  the  linear
  relationship between  A^and  ny  as  already mentioned, the second
  term yields Af prop T~    because of o /T  = const.

      b.  An other  mechanism which could be made  responsible for
  the  pressure drop  deficiency effect  is the  secondary  flow arising
  in curved pipes. From White's (18)  experiments  it may  be computed
  that pressure drop caused by secondary flow alone would increase
                    1/4
  approximately as T   .

      Although we  lack experimental  evidence  to  decide  if one  or
  the  other,  or both mechanisms  concurrently  might be  involved, an
  estimation  could be  tempted  to  assess their relative  importance.

      Gaseous viscosity  may fairly well approximated by the formu-
  la
 where   b   is a constant generally comprised between O.6 and 1.5.
 Eq.  (3) is  valid over the range  from 2O°C to 4OO°C with an error
 not  beyond  2%. The value of  b   e.g. for air is  O.768. For ex-
 perimental  checks  ( e.g. as in the Figs 1 and 2 ) we always use
 the  exact experimental values of gas viscosity. For the purpose
 of dimensional analysis alone we assumed that the viscosity of
 air  increases with the  3/4  power of temperature.

      Considering that e.g. Figs.  1 and 2 show at 6OO°K about 6%
 less pressure drop than required by the pure viscosity term alone,
 the  solution of the following equations ( 4a and 4b) may yield
 x, the fraction of energy dissipated by one or the other of the
 mechanisms.

-------
                                                                 417
" • WO \ .^_ j   _ f-^ i ^p* t   "XI  ' * I \  T~ f               .              /  k
                                            1/4             (4b)

    The exponent  y  has the value of  -1  for the end effect
mechanism and 1/4 for the secondary flow mechanism. As for 60O°K,
T/To«2, from 4a and 4b one obtains about 57% in the first and
63% in the second case. Without any pretention to accuracy,one
may state that both phenomena seem to play a rather important
role in the pressure drop of fibrous media.They may accessorily
explain.why so often the deposition on the lee side of the fibers
can take place — for which laminar-flow theories offer no ready
explanation.

    Empirical pressure drop relationships give obviously a nearer
fit to experimental data.Whitby (19) introduced an empirical func-
tion for the dependency of the drag coefficient on the Reynolds'
number.Kimura and linoya (2o) fitted the following expression to
a great number of experimental data:
with

    The temperature dependency of the pressure drop  in  the Kimura-
linoya equation  is:

*  = A + Bf1/8 + CT3/4                                       (6)
In order to fit experimental facts, an empirical term wi
 steep variation than T '    had to be introduced. Eq. (6)
                                                         with
 less steep variation  than  T  '    had  to  be   ntroduce.  q. (6)  fits
 also quite well  our experiments  at all  temperatures  and  two  dif-
 ferent  test  gases  ( air and  hydrogen) .Fig. 3  shows the example
 of  a carbon  fiber  mat  tested in  air.

    The exponent in the Kimura-Iinoya equation  allows  a  further
 estimate : all  over the range  1 $ T/TQ 4 2  ,  the function
 (T/T0)~1'8 may  well bfe approximated  by   (T/TO)    --HT/TO)~   .
 This could be taken as an  indication that  about one  third of the
 energy  dissipation by non-laminar  mechanisms could be  attributed
 to  secondary flow  and the  remaining  two thirds  by the  unsteadi-
 ness at capillary  entrances  or ramifications.

-------
418
                      2500 -
                     2000 -
                     1500
                     1000
         Fig.  (3)  - Experimental  fit  obtained by
                     Eq.(?) in the  case of carbon
                     fiber mats of variable thick-
                     ness,tested with  air.

-------
                                                                 419
2. A RAPID THEORETICAL REVIEW OF THE DEPENDENCY OF FILTER
   EFFICIENCY ON TEMPERATURE.

    This is a short brush-up of ideas already developed in the
quoted basic references.

    The collection efficiency  E   of a filter mat is given by
the rather well known expression ("log-penetration law") :

                                                            (7)

where it is assumed that ZE  , the collection efficiency of a
single fiber is obtained by adding up single fiber efficiencies
due to separate mechanisms :

ZE = Ed + Er + E. ...                                      (8)

E  = single fiber efficiency by diffusion
E  = single fiber efficiency by direct interception
E. = single fiber efficiency by inertial impaction.

    We do not take into consideration electrical forces because
the main experiments were made with earthed semiconducting carbon
fiber mats. Deposition by gravitational forces was neglected as
not significant.

    Single fiber efficiency  by diffusion was calculated by many
authors on basis of slighltly different assumptions , but all the
results may be brought  to the form  :
Ed = B ( r)X                                             (9)

with  x  either -1/2 or -2/3 or -1. Anyway, E   as a function of
a positive power of  T  , must increase in the same sense .Follow-
ing Eq,(7),the overall efficiency must grow in the same direction.

    As for  E. .Whitby's   (19) approximation represents fairly
well all previous experimental results :

EA = exp [-(log St)2]                                       (10)

where  St, the Stokes' number or inertial parameter is defined
by:

-------
  420


 St -.-?**0" *4
      15/r? df


    When  the temperature  increases  ,  E.  decreases,  provided  that
 St<1.  Thus by Eq.(?)  the  overall efficiency  varies  also  in  the
 same  direction.

    According to well  established concepts  of the majority of
 the authors quoted, E   is  independent  or  only very  slightly  de-
 pendent on the temperature.

    The experimental and technical  interest of  filter  tests  at
 variable  temperature follows  from these  basic theoretical consi-
 derations.In order to  improve a  filter systematically,a model
 must  fit  its real behavior.Theory allows  to compute  individual
 fiber efficiencies,but  unfortunately  ( or fortunately  ) we do
 not have  just one or two theories,but  rather  a  dozen able to yield
 quite divergent numerical  results.Tests  at  variable  temperature
 provide a  ready means  to get  insight  in  the prevailing mechanism.

    To assess fiber efficiencies experimentally, the  tests have
 to be run at constant  temperature with particles of  uniform  size
 classes.This v/as done  and  published by about  a  score of authors.
 Before the automatic particle spectrum measurements were known,
 to gain insight in the  filtration mechanism at  constant tempera-
 ture of a few filter mats was already a complex research project.
 Because of the time,expenses and care involved.it was out of ques-
 tion to study filtration mechanism of industrial filters this
way.

    As it will be shown below,tests executed  at variable tempe-
 ratures and not necessarily over a very wide  range,represent an
easy means to get insight  into the filtration mechanism even with
mechanically dispersed particles of non-uniform size.The tests
thus become easier and more rapid.

3. FILTER EFFICIENCY AND TEMPERATURE: EXPERIMENTAL.

    3.1. Particle size dependency of the single fiber efficiency.
    The filter mats tested were of carbon fibers, df =     ,
£= O.969 and L = 3.9 mm .The test aerosol was obtained by atomi-
zing and drying of a NaCl solution; the filtration velocity in
all tests was 1OO cm s~ .The particle counting and sizing was

-------
                                                                  421
done by an automatic "scintillation"  ( flame photometric  )  ana-
lyzer manufactured by Sartorius  ( Gb'ttingen  ) and developed by
Binek (21).The apparent fiber efficiencies were computed  from
the experimentally measured mat  efficiency by Eq.(7).

    The results obtained at 2O°C, at  1OO°C ( = 393°K)  and at 2OO°C
{ = 493°K) are represented in the Figs.4  ,5  and 6, which  also
show ( dashes ) the theoretical  single fiber efficiencies calcu-
lated by one of the most sophisticated theories : stepwise  cal-
culation of particle trajectories following  Dawson  (22) using
the Spielman-Goren (16) flow filed  for interception  and inertia;
the diffusion term after Stechkina  and Fuchs (23) as  exposed in
detail by Davies (11) p. 74 ff.
              ncT
                       \Theor«ttc.al

                        \
                        bmtflt
        Fig.  (4) - Single  fiber  efficiencies  as  function
                   of the  particle  size,carbon  fiber  mat  at
                   2O°C;face  velocity  1OO  cm  s~  .

-------
 422
Fig.(5)
Fig.(6)
                    L*>I a.f
\ K
 \
 \
                                  200°C
Single fiber efficiencies  as  a function of the particle size;

carbon fiber mat,face  velocity 1OO cm s~ .

-------
                                                                  423
    Looking,  as usual, only at the results obtained at room tem-
perature ,Fig .4 , theory and experiment are in rather good fit.Both
curves have the same overall trend and the fact that there is
no exact coincidence should not bother us too much, as some of
the constants of the theories are still empirical and depend on
fiber orientation and other factors.

    Up to this point,the theory seems adequately confirmed.The
situation changes when looking at the results obtained at higher
temperatures.Figs. 5 and 6.The'minimum requested by the theory
fails to show up.At first sight,we would be inclined not to trust
the experiment as it often happens.But another representation
of the same results ( Fig. 7) does not allow this easy explanation.
Up to the particle diameter of O.13/tm, one would assert that the
theory is well confirmed because neither computation,nor the ex-
periment can have any claim to very high precision.Conjecturing
a somewhat different flow fi,eld and computational formula, theore-
tical single fiber efficiencies differing by a factor of five are
often obtained ( see Davies.l.c. p.64).Due to this reason,we may
1
-10
a
u
-J
V
V
_Q
.,*
*
vn
8
c


a
.0




--
- '
-


L0« V
«r




•--




»' i
•m




FKtei*.




00' 1
.07





--

:

<>•«






L--



«• 2






n-.~. s



flo* 3^
.15









0' 10
^.•m









<>• 2







ThMv.


00* 31








._.

0» 40
Dr-







,__

o* a<








TltMV.

90* a
.4









xf \
V1"



;





»• 2

)








00- I
rs^


)
i
~~




0' It
*m
3
:


--




»• 2

-



TH.OV.
-
-


00*
Fig.(7)  - Single  fiber efficiencies with  particle  size  and  tem-
          perature  as parameter;carbon  fiber mat,face velocity
          1OO cm  s~ .

-------
  424
disregard on the  first sight  that  from 0.2 yum  diameter  on,  the
experimental fiber  efficiencies  are  two  to  five  times higher
than  the theoretical ones. More  significant  is the  trend  :  at
0.2 and 0.4 yum,there appears  a clear  rise in the experimental
efficiency, unpredicted by the theory. At 0.75yu.m,the computed
fiber efficiency  shows already the decrease  requested by  the
inertial mechanism,while the  experimental results show  a  clear
minimum for the intermediate  temperature.

    These findings, considered in conjunction  with  our  former
hypothesis that the pressure  drop cannot be  fully accounted for
by the laminarity concept, seem  to indicate  that the flow field
at the basis of the computations does not adequately take into
account non-laminar conditions which  seem to exist  within the
filter.

    Extrapolating from this point on, we might presume  that a
filter in which turbulence is prevalent, will  show  less pressure
drop and a better efficiency  than a purely laminar  one.Therefore,
in order to develop new concepts in fibrous  filtration.unsteady,
turbulent or secondary-flow conditions should  be1 sought.

    3.2. Tests with polysdispersed particles.

    The filters employed in these tests were the sintered glass
ones already mentioned and described in the captions of Figs. 1
and 2. The test aerosol was a mechanical dispersion of silica
dust,  approximately log-normally distributed between 1   and 15m
with the mass-median diameter of 1.5  m and a geometric standard
deviation of 4  .

    The  X  signs in Fig.  8 show a case of decrease of the overall
efficiency with temperature ( slope approximately -3/4  ) .consequent-
ly the prevailing effect of inertial impaction.The  signs  0  per-
tain to a "filter" made up of a paralell bundle of  glass  fibers,
all arranged somewhat like a brush aligned in the sense of the
flow.In that way,impaction was kept at minimum and  the particles
could  be stopped only by diffusion. As expected.the overall ef-
ficiency is rising with a  slope but slightly different from  7/8.

    The signs  +  in Fig.  8 show a case where both mechanisms
seem to be involved,with a minimum of the efficiency near 4OO°K.

-------
                                                          425
       90
      3-
       '°
      (U70

      "u

      it 6o
      9)
        s0
      o
      a*
       40
       30
                i
i
              300°      400°   500°  600° K
Fig. (8)  -  Tests with polydispersed silica dust,

-------
  426
LIST OF SYMBOLS.




d,  fiber diameter




d   particle diameter




D   capillary diameter




Ed  single fiber efficiency by diffusion




E.  single fiber efficiency by inertial impaction




Er  single fiber efficiency by direct interception




ET  overall collection efficiency of the filter




g   acceleration of the gravity




L   thickness of the filter mat




Ap  pressure drop




Q   volume flow




St  Stokes'  number ( see Eq.1O)




T   absolute temperature




v   (  flow ) velocity




VQ  face velocity ( velocity of the upstream flow )




£   porosity of the filter mat




^   density  (  gaseous...)




 ^  density  of the particle




rn   dynamic  gas viscosity

-------
                                                                  427

 REFERENCES

 (1) First, M.W..Graham,J.B..Butler,G,M..Walworth,C.B. and Warren,
     R.P.,Ind. and Engr.Chem.,Vol.48,p. 696, 1956.

 (2) Billings, C.E..Silverman,  L.,Dennis,R. and Levenbaum.L.H.,
     J. Air Poll.  Contr. Assoc. Vol.5,  p.159,  1955  ; Vol.8, p.53,
     1958 ; Vol.10, p.318, 196O.

 (3) Spaite.W..Stephen,D.G. and Rose,A.H.,J. Air Poll. Contr.  Assoc.
     Vol.10, p.  215,  196O.

 (4) Thring.M.W. and Strauss W..Trans.Inst.Chem.Engr.,Vol.41 ,  p.
     519, 1963.

 (5) Pich.J. and Binek,B.,"Aerosols,Physical Chemistry and Appli-
     cation" ,Proc. 1st Nat. Conf. on Aerosols.Czehoslovak Akad.Sci.
     Prague pp.  257-264, 1965.

 (6) Strauss.W.  and Lagcaster,B.W..Atm.Env.,Vol.2,  p.135, 1968.

 (?) Dyment.J.."Assessment of  Air Filters  at Elevated  Temperatures
     and Pressures" - Filtration Society's Conference .London-, Sept.
     23-25, 1969.

 (8) Chen,C.Y.."Filtration of  Aerosols  by  Fibrous Media".Chem.Rev.
     Vol. 55,  p. 595, 1955.

 (9) Pich,  J.,"Theory of Aerosol Filtration by Fibrous and Membrane
     Filters".Aerosol Science,C.N.  Davies,ed..Academic Press,Lon-
     don , pp.  223-285, 1966.

(OO) Spumy, K. ."Grundlagen der Filtrationstheorie  in  aerodisper-
     sen Systemen  und die realen Ultrafilter".Fortschr.-Ber.VDI-Z.
     Ser. 3 Nr.  17, 1967.

(11) Davies, C.N.,  "Air  Filtration".Academic Press,London-New  York,
     1973.

(12) Kuwabara, S.,"The Forces Experienced  by Randomly  Distributed
     paralell  Circular Cylinders or  Spheres in Viscous Flow at
     Small  Reynolds Numbers",J.Phys. Soc.  Japan Vol.14,p.  527,1959.

(13) Piekar, H.W.  and Clarenburg, L.,"Aerosol  filters  - the tor-
     tuosity factor in fibrous  filters".Chem.Eng.Sci.Vol.22,p.765,
     1967.

-------
 428
(14) Happel, J.  "Viscous flow relative to  arrays  of cylinders",
     Am. Inst. Chem.  Engr.  J.,Vol. 5, p. 174,  1959.

(15) Fuchs, N.A.  and  Stechkina,  I.B. "A note  on the theory of
     fibrous aerosol  filters", Ann. Occup. Hyg. Vol.  6,  p.27,1963.

(16) Spielman, L.  and Goren,  S.L. " Model  for  pressure drop and
     filtration  efficiency  in fibrous media",  Env.  Sci.  & Techn.,
     Vol. 2, p.  279,  1968.

(17) Quetier,  J.P. "  Etude experimental des  effets de la tem-
     perature  sur les mecanismes  de  la filtrationVThesis,Paris
     1977.

(18) White, C.M.,Proc. Roy. Soc.  Vol. A  123,  1929.

(19) Whitby, K.T. " Calculation of the clean  fractional effici-
     ency  of low media density filters" Am. Soc.  Heating,Refrig.
     Air Cond. Engrs. J., p. 56,  Sept. 1965.

(2O) Kimura, W.  and lionya, Chem. Eng.(Tokyo),Vol.  23, p.792,1959.

(21) Binek.B.  "Szintillationsspektrometer-Analysator fur Aerosol-
     teilchen",Staub  Vol.2O, p.184,  1960.

(22) Dawson, S.V. "Theory of collection of airborne particles
     by fibrous  filters" Sc.D. Thesis,Harvard  School of Public
     Health, 1969.

(23) Stechkina,  I.B.  and Fuchs,  N.A. "Studies  on  fibrous aerosol
     filters.I.Calculation  of diffusional  deposition of aerosols
     in fibrous  filters," Ann. Occup. Hyg. Vol.9, p.  59,  1966.
WRITTEN DISCUSSION

K.T. Yang, Chairman
Department of Aerospace and Mechanical Engineering
University of Notre Dame
Notre Dame, Indiana 46556

   Dr. Benarie should be congratulated for a most stimulating talk on a very
timely subject.  His qualitative analysis on the temperature effects on both
the pressure drop and collection efficiency should be very useful in inter-
preting experimental data for high temperature filtration.

-------
                                                                            429

    My comments are primarily directed toward the following issue:  Do the
same physical mechanisms at low temperature for the pressure drop and col-
lection efficiency behavior operate also at high temperature?  If so, a
variable property analysis is all that is necessary to predict the behaviors
at high temperatures, provided that the physical mechanisms can be properly
modeled at low temperatures.  Professor Benarie seems to think that this is not
the case,and other effects,such as capillary entrances, secondary flows and
the like are operative even at moderately high temperatures.  However, I would
like to suggest that uncertainties in low temperature behaviors in pressure
drop and collection efficiency far out-weighs the uncertainties in the variable
property analysis to predict the high temperature behaviors.  In other words,
if we had an adequate quantitative theory for low temperature filtration
through fibrous media, the variable property analysis may be quite adequate for
high temperatures, with possible exception at very high temperatures.

    One evidence of my speculation is the correlation of the pressure drop data
shown in Fig. 3 of Dr. Benarie's paper, which, as I undertsand, covers a wide
range of temperature and is based on a variable property analysis on viscosity
and density based on the Kimura-Iinoya equation for the pressure drop at room
temperatures.

    A second evidence can be observed in Figs. 4,5 and 6 in Dr. Benarie's paper,
which show the experimental fiber efficiencies over the size range at three
different temperatures and one face velocity.  These are the solid curves in
the figures.  Also shown are the best theoretical predictions at low tempera-
tures, modified with variable properties at high temperatures, and these are
the dotted curved in the figures.  If we had, a good quantitative theory for
single fiber collection efficiency at room temperatures, then such a theory
would give up a prediction close to the solid curve instead of the dotted
curve in Fig. 4.  Now suppose that we use this curve as a base and extend it
by a variable property analysis to 100°C, we would obtain a curve in Fig. 5
which would fit the experimental data even better than the experimental best-
fit curve.  Furthermore, this curve would also exhibit a minimum around O.OSum.
If the same procedure is extended to 200°C, the case of Fig. 6, the resulting
curve would be much closer to the experimental data than the theoretical curve
indicated, even though in this case, there is a consistent deviation from the
experimental data points, which seems to indicate that additional physical
mechanisms are indeed at play at this temperature level.

    Based on these observations, it certainly appears that more studies, either
analytical or empirical, are needed to provide a better predictive theory for
low temperature filtration before any meaningful predictions can be made at
elevated temperatures.

-------
430

  OPEN DISCUSSION

  Benarie:  My point was that these experiments should shed some light on
  the theories of the dependence, of filtration on temperature.  They were
  analyzed in three dimensions rather than two.  I also pointed out that my
  suspicion is that some turbulence of significant duration is important.
  None of the actual filtration theories takes into consideration these facets
  of filtration.  The next step should be to introduce these mechanisms, into
  filtration techniques in general, then we shall see a temperature dependence.

  Billings:  It seems to me that there is a fair amount of confusion among
  people in filtration about what happens in filtration.  First of all, the
  phenomena themselves are not very well understood.  They are not very well
  documented and not very well analyzed.  The first thing desired is a
  phenomonological model.  Before you can have a phenomonological model, though,
  you have to have observations.  What we observe is macroscopic.  We go in
  and measure the concentration coming into  a   black box.  Then we measure
  the concentration coining out of the black box.  We get back to a theory on
  individual fibers and individual fiber efficiency.

       If I then ask one of the noted professors to explain this, they would
  sit down and start with a continuum theory, assume a potential flow, and
  imagine a streamline with a particle on it, and a fiber in the path of the
  particle.  By doing a trajectory calculation, they can figure out where it
  intercepts the fiber.  A lot of assumptions were made up to this point,
  the biggest being the matter of the particle-fiber interaction.  If you
  look carefully at the underlying assumptions that have gotten us to this
  point, you will see that there are no direct measurements, by eye,of the
  interaction between the particle and the fiber. It all goes back to
  Dr. Friedlander's experiments in liquids - that is not a gas-solid case, it is
  a liquid-solid case.  As you well know, the forces that come into play in
  liquids between two surfaces are orders of magnitude different from the
  forces that come into play in gas-solid interaction.

       From our phenomonological model we can derive an analytic model.  Given
  the analytical model, what does one do?  You say that the particles will be
  stopped by the obstacle.  Now we have a fiber efficiency and can build an
  analytic model.  The first assumption is the agglomeration coefficient. I
  have given this approach to the problem many times before.  It is a macroscopic
  basis that is dealt with.  But it must be approached on a microscopic basis.
  The third one is a molecular basis.   What you have to do is ask how effective
  that interaction is and what is the influence of temperature on that inter-
  action.  The same would be true of the viscosity.  Instead of using the
  horizontal approximation, we have to look at several vertical coefficients
  on the equation of state at the molecular level.

  Benarie:  By past experiments it has been shown that filtration theories
  are not the crux of the problem.  Perhaps all we have to do is introduce
  some new concepts.  We ignore what was previously introduced.

  Liu:  Most of the theories developed are for a fiber of uniform size.  If

-------
                                                                            431

you want to translate the single fiber efficiency to a multi-fiber filter,
you assume the filter to be composed of uniform fibers.  In the real filters,
the fibers are not uniform.  So if you try to compare theory with experiment,
you must try to find a particle size that will give you a good fit to the
data based on the pressure drop.  A different particle size may be required
to fit the collection efficiency data.  The question to answer is:  How do
you determine the correct particle size to use?  That determines how much
you can respect your theory.

Benarie:  In this case, it was a very uniform fiber size.  It was determined
that it was almost one size.  I would like to add that,in the theory, it
has been found that a number of different fibers sizes are acceptable.

Friedlander:  It is interesting to point out, even though we are  using
liquid systems, that we are not only having problems correlating the types
of existing models in the unsteady flow or turbulent flow region, but
there is also a problem at the downstream end, especially at the lower
velocities.  So there seems to be problems at both ends of the model.

Lbffler:  I am afraid I did not explain clearly enough my approach.  My
approach is not to consider the theory but to go into the details to analyze
the fundamental phenomena.

     One comment to this problem of Dr. Benarie:  I can assume that in his
experiments, the difference between experimental and theoretical values
should be an effect of electrostatic forces.  This is because you used
sodium chloride particles produced by atomizing.  All these particles have
charges.  The tendency of this electrostatic effect of increasing 10 or 20%
is minimal.  This would mean that increasing the viscosity would create an
effective decrease.

Benarie:  Therefore we do first the electro-conducting filter.  Your explanation
does not correlate with the pressure drop.  I consider the pressure drop
easier to measure and more indicative of turbulence than the deficiencies that
are being criticized,because electrostatic effects, adhesion and impurities
can be changed by temperature.  I need not grade my arguments,but if I woud, I
would give myself a  9 out of 10 for keeping pressure drop and coefficients,

Oder:  I think Dr. Benarie gets a W for candor.

Cooper:  One thing that strikes me is that the experimental results are
substantially higher than the theoretical, above a few tenths of a micron.
Which would probably suggest that we are not having accommodation or
rebound on the fiber.  One of the things I wanted to know was what would be
the Reynolds numbers of the fibers in the flow.  Could we expect a wake?

Benarie:  Do not be  surprised if the results are 2 or 5  times different from
the theory.  Because if you take the models   ,97 of them,
and compared the different flow fields within the same theoriesj discrepancies
among theories  are present.  So a person chooses the theories they desire  to
make the best fit.

-------
432
       The Reynolds number will be discussed in the panel discussion.  If you
  have a tube, what is the Reynolds number; the thickness of the wall or the
  tube diameter?  Why is the fiber Reynolds number so important?  There is not
  a great difference.  With a porosity of 90% there is a factor of about 2
  we have to deal with.  The right Reynolds number is around unity.  Alone, the
  Reynolds number can not explain the turbulent development.  But we can cover
  that in the panel discussion.

  Billings:  My point is that you are assuming that the Sutherland approximation
  to a Kuwabara approximation to the flow field is not going to lead to a
  set of problems.  I think we need to look at the accommodation coefficient for
  the molecular material on the solid.   It is at the molecular level that the
  effects come in.  The viscosity arises because of the accommodation coefficient
  of the solid for the gas.  That is a non-stick assumption on the solid which
  is probably not true.  I think you have to look for temperature dependence
  on the equation of state to look for the change in the viscosity.

  Benarie:   We analyzed the viscosity effect and about six different possible
  effects.   The non-slip coefficients explain only about 1% of the observed
  effects.

-------
                                                                              433
                              PANEL DISCUSSION



Subject:        "New Directions in Particulate-Gas Separation Research"

Moderator:      Benjamin Y. H. Liu

Panel Members:  Teoman Ariman, Michael M. Benarie, Dennis C. Drehmel,
                Robin R. Oder and David S. Scott.


B. Liu:  First of all I would like to congratulate Dr. Ariman for the excellent
job in the organization and guidance of this workshop.  The purpose of the panel
discussion is to look at the new directions in particulate-gas separation research.
I think the panel is well represented with experts in the various research areas.
It will be a ten minute presentation by each member of the panel.  Then a
general discussion with the participation of everyone will be held.

      I would like to begin the discussion by giving a few of my general comments,
informal questions, and suggestions and finally some of my opinions.

      First of all with regard to magnetic separation, I really don't know much
about magnetic separation but one thing that has bothered me is that no one in
this conference has dealt with the problem of energy needed in order to grind a
particle down to the size to separate out the mineral impurities and the sulfur.
So I would very much like to see Dr. Oder address the question of the energy
that you have to put in the system to grind a particle to two or three microns
in order to separate out impurities.  With regard to electrostatic separation I
would like to say that, if my memory serves me right, a few years ago electro-
statics was the whipping boy for the blame of the experimentalist whenever he
could not make an experiment agree with the theory in aerosol studies.  Usually
the first thing you point out is electrostatics.  So I am very happy to see that
now at least we are putting electrostatics to good use and I think the electro-
static is no longer a black magic as it once was.  It has begun really to become
a science.  People are beginning to understand and study it and know methods of
measuring the charge and neutralize the charge and handle the electrostatic
problem much better than they did before.  So,I am very happy to see that ad-
vances have been made in the electrostatic area.

      Now for the case of the acoustic interaction I think Dr. Shaw had pointed
out clearly that the problem seems to be  that we don't have a good handle on the
theoretical coagulation coefficient for particles in the acoustic field.  We
know that it is very expensive to perform full-scale testing of these ideas.  There-
fore, sound theoretical understanding of  the basic phenomena involved is very,
very important.

      So at this point I would like to put out a pitch to EPA and NSF for the
importance of the fundamental research. I think the reason  that we have to devote
attention to the basic research is that by understanding fundamental phenomena

-------
  434


 at a microscopic level we can avoid the very expensive and costly mistake that
 we could make by using a larger scale apparatus.   So I am very glad to see
 that the NSF and EPA are joining hands in sponsoring this workshop.   Obviously
 EPA has its own mission to meet so therefore they cannot sponsor too much of
 the fundamental research whereas NSF certainly is in the business of support-
 ing basic research.   I would like to see in the future that more fundamental
 university research  programs be supported because it is very important in the
 long run.   Furthermore the fundamental study may  turn out to be also the most
 economical way of solving some of these problems.

      So these are some of my very brief comments.   At this point I would like
 to turn to Dr.  Benarie.

 Benarie:   I begin this discussion with an answer  to a question.   You see quite
 clearly that my purpose  of presenting some research in three-dimensional channel
 filtration was  to rock the boat  a little bit.   I  continue rocking the boat with
 the following.

      One  of the most  important theories of pressure drop is the channel  theory.
 We do not  have  a channel theory  of  efficiency;  we  speak only about fibers.  I
 would like to draw attention to  the fact that  there are holes between the fibers.
 Because the air passes through the  holes,  the  particles are captured by  the
 fibers.   I find that  by  drawing  just  the holes  in  front of the channels,  and,
 instead of looking at the fibers, by  looking at the channels of  air  coming in,
 you immediately have  turbulent agglomeration.  We  have no. where  taken into
 account the turbulent agglomeration within the  filter.

      Next,  we have here  the  effect  of  secondary flow,  because it  disturbs  us,
 we don't take it  into account.   Thirdly,  at all beginnings and  ends  of channels
 we have an end  effect which  is turbulent-like;  this  also   is not  taken into
 account.

      I  suggest  that what we  need  theoretically is  a  channel  theory of sections.
 My  second  suggestion  is  that we  try filters with different  shaped  fibers  to try
 to  enhance  in-filter  turbulence  and take  a step forward  in meaningful turbulent
 filtration  by mass.   Currently we do most  filtration through  fibrous media
 which has not straight but curved sides.  More advanced  technology would allow
 us  to make  any micro-fiber shape.  A turbulent pressure drop  is less  energy-
 wasting than a  laminar pressure drop.  Perhaps we  can get more efficient and
 more energy conservative fiber shapes.  A number of fiber  capsules with rough
 and smooth  fibers  are available  to us for  full testing.  So  exterior  fiber
 geometry still has the potential to bring a renaissance to this old  art of
 fibrous filtration.

Drehmel:  I guess  like Dr. Billings I have to start with the microscopic level
and work backward.  We of course at EPA are mission oriented.  We have to pro-
vide demonstrated  technology to control particulate emission from stationary
 sources.  And unlike  some of the other goals that we might pursue such as
understanding the  theory better we quite often get ourselves into the position
 of knowing a piece of hardware works and then asking ourselves, what happened.

-------
                                                                             435

     In order for us to make some sense of what is happening with the equipment
and also to understand how it will fit into the total picture of particulate
control equipment,we do get involved with theory.  But our goal is to provide
technology.  So let us spend my few minutes looking at where problems are, and
some of the solutions we have taken so far.

     So far we have three types of technology that are available:  filtration,
scrubbing, and electrostatic precipitators.  We know that under certain circum-
stances at least an efficiency of 99.9%  exits.   We have heard  people mention  today
that filtration efficiency was 99.99%.  Collecting the particles isn't necessarily
the problem.  It is the circumstances under which you have to catch these
particles that give us some difficulty.  The precipitators have problems with
resistivities that are too high or that are too  low, the pressure drop in
scrubbers is too high and the filters are  too big.  And so research to this
point has taken the attack one of trying to improve standard technology and
also to go out and find new types of  technology  that could be brought to bear.
The types of research we have looked  at are pre-charge chambers or 2 stage pre-
cipitators in order to get around resistivity problems.  We looked at condition-
ing with agents such as SO, or a sodium carbonate.  We have looked at baghouse
problems where we have size limitation.  We are  looking at high velocity types
of filtration where we go to air cloth ratios of 3 to 10.  We  like to look at
an air cloth ratio of 1000.  We also  would like  to look at media that were not
limited to less hostile environments.  We  would  like to look at media that
could withstand high temperatures, we would like to go to 800°C, not just a
couple hundred degrees.  We would also like to be able to deal with environments
that have acid gases, sulpheric acid, HF,  we have found a great deal of work
on that.

     Since the point has been brought up about filter geometry, we have funded
research looking at smooth fibers vs. trilobal* fibers.  And we found in fact
that the trilobal fibers worked much  better.  For scrubbers to get around the
pressure drop problem, we have looked at the device called the flux force con-
densation scrubber.  We bring to bear condensation on the particulate which
helps remove particulate both by growing it and  also by bringing to bear
diffusional and  thermal mechanisms.

     So this is where some of our research has been oriented on  conventional
devices and trying  to take advantage  of  new mechanisms.  We have funded research
in several areas, we have looked at sonics and spent some money  on magnetic,
but primarily we  have looked at the combination  of electrostatic and other
forces.  And for  example we had funded some research at MIT to look at systems
with charged droplets and particulates.  That  is where the  time  constant  theories
of Dr. Melcher came from.  We asked Doug Cooper  when he was still at GCA  to
look at evaluating  electrostatically  augmenting  devices.  And  you have also
heard about  these other projects with Battelle Northwest  and TRW.  We have also
talked  to  George Fielding  to do some  work  with us.  We are  currently funding
Mike Pilat  to  look  at his  charged droplet  scrubber.  We  felt that electrostatics
provides a  very effective force  in the control  technology  of removing fine
particles.

     As to where we have research needs  in the future, rather  than talking

-------
  436

 about some of the theoretical research needs (I am sure it will be brought up
 more by the other members of the panel) I would like to take a different attack
 and say that primarily we have talked about particulate-gas separation from a
 conventional point of view.   We have talked about them in terms of a stationary
 source with temperatures or  pressures that we are used to looking  at.   We find
 that the suspended particulate,  where we have problems in not meeting  particulate
 standards,  is primarily due  to sources that are not enclosed sources.   They are
 what we would call either fugitive emissions or  fugitive dust emissions.
 Fugitive emissions would be  the  type of thing where a  source is not completely
 contained,  not properly hooded.   Because of the process we have had some
 difficulty  with an open electric hearth,  for instance,  where we have some
 emission around the site or  sintering machines which are  not completely enclosed.
 We  would also have fugitive  dust emissions where dust  has been deposited from an
 industrial  process which is  now  a roadway in a plant or it is part of  an open
 mine face,  or it  is where work is being done in an ore  deposit.  Even  more
 troublesome are the types of things  where we have asbestos from brake  linings
 or  we have  particulate  from  the  use  of  tires.   We find  that  all these  things
 contribute  to this suspended particulate  and we find a
 this  area.   We need to  have  research oriented to  the control  of particulate
 from non-enclosed  sources.   Also  we  just  heard  that we  are going to  use more
 coal.   One  of the  ways  we go about using more coal  is to  gasify it or  burn  in a
 fluidized bed combustor.   In either  case you can make these processes  much more
 efficient if  you can use  a gas turbine.  A gas  turbine  operates at high pressure
 and  it  would  be necessary to  remove  the particulate  in  advance  to  the  gas  turbine.
We are  talking about particulate  control at  1500°F  (800°C) and  10 atmosphere
 pressure.  Again we suddenly  find that  there  is no technology available.  We
have various  types of cyclones that can be used or have been  talked about.  They
do not  provide control  technology below 1 micron.  In order to provide particulate
control for both the turbine and environmental protection we will need research
in this area.  So  I would see that-much of the applied research for the near
future will be in line with conventional devices.  We will work with conventional
devices to make them solve the problems but innovative research has a large
application to areas where we have no technology; to the high temperature, high
pressure control problems and to the fugitive dust emission area.

Oder:  Coming from an engineering firm, I came prepared to address  many of the
topics already treated by Dennis.  So now that he has given his presentation,
I am free to come back to magnetic separation and other areas.

     Before doing  that, however,  I would encourage the  funding  agencies to
 fund scientists and engineers so as  to  bring  these people  into  the heart of the
 problem together.  The  scientist, from  the  engineer perspective, is apt to be
paranoid and  to think that the engineer is  limited in his vision and is a
negative influence.  The  scientist is concerned that the program may be stopped
at an early stage because of peripheral constraint that the scientist  is not
 even interested in.  I  think that this viewpoint  is wrong.  I also think that
overconcern with engineering economics can be a negative influence.  So I would
like to see a program that studies feasibility from a creative perspective at
a very  early  stage in any developmental program and magnetic separation is one
area that certainly could use that kind of insight.

     Now let me turn to magnetic separation.  Gas-particulate separation in

-------
                                                                             437

this area is new.  Basically, the only experimental evidence has been gathered
by Drehmel and Gooding.  So other than their work, there is conjecture and
a great deal of that.  The technology of HGMS that they are studying has been
successfully developed commerically for wet mineral beneficiation.  However,
in the area of dry separations it is in its infancy.

      Now I want to also remind you that the magnetic technology is severely
limited in its applicability.  It doesn't have broad general applicability. It
is usable only in systems which can be magnetized.  The stronger the magnetic
material the more suited it is to magnetic separation.  Now with that idea in
mind I would also suggest to you that when looking for applications of the
technology in gas-particulate separations you must address the problems
associated with high velocity processing.  In order to be able to afford HGMS
technology high material throughputs are required.  So whether it be in areas
of removing mineral matter from coal or whether it be removal of suspended
particulate, such as in the case for the steel making furnace off gases, you
need to have large capacity systems to even think of the technology in terms
of air-particulate separation.

       There  is a plus  going  for magnetic separaston which hasn't been mentioned
here at all.  In those areas where the materials are strongly magnetic and
where magnetism  is iron based, the technology can be employed, in principle,
up to 1400°F, which is the cure temperature of iron.  So in principle you are
talking about potential applications up to that temperature.  So if one has
applications where magnetic forces will work, he is not limited to low temperature
applications only.

      Now with regard  to the question about the cost for pulverizing; first of
all, I did not want the emphasis of my talk to be on coal cleaning.  Rather
I wanted coal cleaning to serve as an example of an area where one could
potentially make strongly magnetic materials so that an application of magnetic
methods of gas-particle separation similar to the steel furnace application
could be studied.  But in that particular context, the costs breakdown the
following way:   pulverizing  including crushing is about $2 or $2.40 a ton based
upon a conservative design.  Coal separation might be done for less. The capital
cost for the desulfurization circuit considered here including all reactors and
including the magnetic separator, is about $105 per ton per hour of clean coal
product.  Operating costs are estimated at $7 per annual ton of clean coal  for
desulfurization  only.  And if one wants to store  the clean coal,  then the cost
for compacting the pulverizing coal  is about $4 per ton.  So, the costs for
pulverizing  are  expensive, but I don't think that is going to be a major con-
sideration in limiting the application to dry processing of 200 mesh coal.

       Now let me say something about fibers.  The magnetic separation technology
has perhaps  been studied more extensively from the standpoint of  choice of
collection material  than has  the electric filtration and that surprises me.   It
is known that the shape of fiber, the local surface curvature and the deployment
of the fiber in  the  packed bed is crucial and extremely important to magnetic
separation.  Kaolin  processor have  learned how to selectively remove particles
on the basis of  their  size.   The bulk strand diameter  is important  in collect-
ing large particles  over 8 microns  in size and fine size surface  imperfections
on the strand surface  are  important  to the removal of materials in  the  0.25
micron size  range.   This can be confirmed by etching  the imperfections  away,

-------
 438

 I will now  list areas  in  gas-particulate  separation  that I think are very
 interesting.

      I would  like  to  see more  information about  the strength of the coagula
 that are formed utilizing the agglomeration technologies.  The idea is that
 electric or magnetic separations might be aided with use of appropriate seed
 material.   For example, nonmagnetic particulates  could be seeded with magnetic
 particulates to form a coagulum to be removed magnetically.  What is the
 strength of that coagulum?  What strength turbulent  forces would break it up?
 With regard to fluidized  beds I feel there is something very fundamental that
 hasn't really  been  touched on and that is the stability of fluidized beds.  I
 do  feel naively, but cannot prove to you, that it is possible to introduce
 electric and magnetic  forces to promote stability in fluidized bed processing.
 By  stability I mean no bubbles, no mixing in the bed whatsoever.  Particles
 which are absolutely fixed in space would be useful  as filter elements.

      I am  curious  to  know if anyone has  tried the use of magnetic forces to
 suppress arcing in  electrostatic precipitators?  One thinks of capacitor
 suppression of electric arcing  in an automobile ignition system and of magnetic
 methods for suppression of arcs in submarine battery systems.  These devices
 are magnetic in nature and they blow the  unwanted arc away.  Just a curious
 point, there might be  some magnetic way to suppress  that arc in electrostatic
 precipitators.

      The last comment that catches my imagination is the use of foaming
 technology  to  capture  fly ash from coal combustion products in flue gases.  I
 wouldn't suggest the same foaming agents  that have been discussed here.  But if
 you could precede the  precipitators with  some kind of a foaming agent that was
 liquid at the  operating temperature (a few hundred degrees Fahrenheit) and you
 could blow  large bubbles  (a quarter of an inch or so in size) then you could
 encapsulate many millions of micron size  particles per bubble.  Then by captur-
 ing and destroying those bubbles, you would have a very nice way of capturing
 particulates which are otherwise very difficult to control.  That is my list of
 suggestions.

 Scott:  I obviously would be glad to talk about acoustic dust conditioning but
 I think I will not.  Rather I will title my comments something like "Notes of
 a Workshop Watcher".  And I want to discuss two themes under the "workshop
watching idea".

      First is the theme of technology transfer.   I  think at this meeting we
 have a set of  people who are working to better understand the fundamental
mechanisms of  those processes with which we deal in particulate-gas separation.
Needless to say, to do that right they are isolating various sub-mechanisms
and so,  necessarily they can examine only a few.   It is, therefore, very
difficult to thoroughly determine the macro-mechanism of any given process.
 But it is clear that this fundamental micro-mechanism research is the touch-
 stone of all the work we are doing.  And  I agree entirely with Professor Liu
 that this must be encouraged because it is the only way we can really achieve
breakthroughs  in the second level.

-------
                                                                             439

      The second level is when people perform large scale evaluations, or
overall evaluations, of total processes.  Things like power requirements,
resident times, capital costs, collector efficiencies, etc. are examined. I
guess the comment I would make here is that these have to be done as completely
and honestly as possible, using similar criteria in each research group -
particularly on collection efficiency.  I think there is obviously an area of
research that we haven't talked about very much here, and that is the develop-
ment of techniques for particle sampling, in particular, very fine particle
sampling.  And maybe I could suggest that Grady Nichols or Michael Pilat (who
at least in times past, I think, have worked on fine particle sizing) might
make a comment later on.

      The next step is to commercial operation.  This is where technology
transfer becomes more difficult.  Let us presume that we have a process which
has been developed and works well.  Somehow it has to be used.  And the fact
is, that the best way for it to be used is by our free enterprise system; that
is, you have to have manufacturers which are picking these things up and carry-
ing them on.  But then we are into proprietary interests, practical marketing
constraints, etc., which sometimes are not at all scientific but are nonetheless
very, very real.  The technology transfer from fundamental research to second
level, or more applied research is a process which works.  It works from con-
ferences like this.  But once  you get the process into equipment manufacturing
- with sales, maintenance, and many other forces at play - then I am not very
sure how it works.  That is why I introduce the question.  I am really suggest-
ing that in the next hour, there might be some comments on technology transfer
barriers.  Maybe it is not as serious as I think it is.  But I think the
difficulty of commercializing the university (as distinct from industrial)
scientists' ideas is very real.

      My second theme under this "Notes of a Workshop Watcher" discussion, is
a theme on the interrelation of processes for accomplishing gas-particulate
separation.  This comes  from two observations of the  last couple of days.  The
first observation is many approaches and perhaps the most promising approaches -
how one judges the most  promising approaches is difficult - are hybrid, or they
involve physical phenomena of several different types.  My own field is acoustics
which, of course, is a conditioning process and thus  demands that there be
some other collector.  The point is that there are many systems.  I think it is
unlikely that any single process as it was defined at this conference will find
its way to wide industrial application - unpollinated by other ideas and
processes.

      A further observation related to  the interrelation of processes is that
new particulate-gas separation  applications are popping up like mushrooms after
a warm spring  rain.  I  think Dr. Drehmel has mentioned  this, and others have
seen it.  But  it is something that I  have certainly been struck by during this
workshop.  And  these new applications are coming with obviously a set of new
requirements.   I think  a lot of processes are hybrid  and will  continue to be
hybrid.  Moreover,  applications emerging which will yield  requirements we
cannot anticipate.  All  leads to a comment about the  credibility of a process.
I  think  that none of the processes we have been discussing, are yet credible.
I  think, in  fact, that we don't even  know how  to judge  them so that  they could

-------
 440

 be  defined  as  credible.   For  example, we don't  even know yet what  the dust load
 requirements at  the  inlet of  a gas  turbine are!  We don't know yet what  the
 cotton  processing requirements will be.  So how can we say  that processes de-
 signed  to satisfy these as yet undefined requirements are credible or not?  I
 think the question is not the credibility of the process, rather it is the
 credibility of the study.

      Finally, I come back and wrap all this up.  First of  all, we need  funda-
 mental  work which is credible.  Secondly, we need results - especially those
 that  analyze the macro-processes to be as complete as possible.  In that sense
 it  is as important to identify the  faults of the process as it is  to identify
 the good parts.  Perceived negatives may turn out to be strengths  after  the
 applications are clarified.   Thirdly, we have to identify all the  criteria.
 I have  mentioned several:  power, dust load, size distribution.  If Jim Melcher
 is  going to be at the next conference you know  for sure you have to evaluate
 residence times for him to give you his ear.

      My last  comment with regard to all of this is that the pride in one's
 work  is obviously essentialy  for good work.  I  suppose it is only  natural that
 those who are  working on  a particular process have more faith in that process
 than  do their  peers working on other processes.  We are all at fault, sometimes,
 in  this.  We've got to be careful that this natural pride does not change into
 something that sounds more like shrill advocacy.  I think intense,well -
 articulated discussion on the interpretation of the physics of processes is very
 productive.   Yet I think we must avoid a posture of shrill  salesmanship for
 specific processes. As such,it could be very counterproductive for all of us.

 Ariman:  I  think I would like to concentrate on mostly electrostatic filtration
 in  this short  presentation.   I am sure you noticed that there are several
 interesting things happening and many more difficult problems remain to be
 solved  in electrostatic filtration.  When you talk about electrostatic filtration
 you have to differentiate two areas.  Namely electrostatic  fibrous filtration
 and electrostatic fabric filtration.  If you noticed in this workshop and
 literature we really are concentrating on a mathematical modeling of electro-
 static  filtration by fibrous filters.  Professor LHffler, Dr.  Zebel and number
 of other investigators have been looking into the problem of a single fiber
 and a single particle and are trying to make a mathematical model for it.
 Hopefully maybe for specially built fibrous filters this model could be used.
 However, when you look at the fibrous filters you use in your air conditioner,
 furnace, etc., we are really facing a much more complicated situation than a
model with a single fiber and a single particle.  On one hand,  we have a very
 basic and also quite mathematical single-fiber model.   On the other hand, we
 have  almost thing but a black box as Dr.  Billings indicated.

      Furthermore,  we have been doing a lot of testing in particulate collection
 with  fabric filters.   Then, by mostly using the experimental data we have been
 developing empirical or semi-empirical formulations for the collection efficiency
 and pressure drop in fabric filtration.   It seems to me the time has come,
 particularly when an external electric field does exist, to take a systematic
 approach to the very difficult problems of fibrous filtration,  in the following
way (1) a single fiber model  (2)  a group fibers in the same row (3) fibers in

-------
                                                                             441

a number in parallel rows (4) a staggered array model of fibers (5)  consideration
of fibers in perpendicular directions (6) a three-dimensional model  which would
be the most realistic.  I readily accept that the stated models and  particularly
the last two are quite complicated but I feel that this proposed way could be a
reasonable approach to the analysis of real fibrous filters. Hopefully, with
some modifications and depending on the series of it for fibrous filters, a
corresponding approach could be utilized for fabric filters.

      I would also like to see experimental work hand in hand with the analytical
investigations.  It is certainly very attractice to carry out pure theoretical
work in fibrous or fabric filtration with fourth-order, sixth-order, ordinary,
partial, linear, non-linear differential equations.  We also have powerful
computers which give us a lot of interesting numbers.  We also have a capable
draftsman to make nice figures; so we can get some excellent publications out
of all these in reputable journals.  However, I do not think this pure
theoretical approach with very little relation to the real life of filtration
is a healthy one.  Therefore, we should make sure that whatever mathematical
model we propose, we provide a very valuable support to it and verification of
the model with a suitable experimental program.

      On the other hand yesterday, I am sure you noticed a remarkable decrease
in the pressure drop in the  electrostatic filtration by APITRON of the API
and our bench scale test equipment at Notre Dame.  However, some of my colleagues
tell me "well, look, you are trying to convince us that you save some energy by
reducing the pressure drop,  but how about the additional energy you are  putting
into the system by charging  the particles or charging the filter or both?  What
happens to that?"  Well, if  I may have a couple of transparencies, let me show
you something.  Based on our bench scale experimental results, and the full
scale field tests by Dr. Helfritch in industrial dust filtration, we obtain a
substantial decrease  in the  pressure drop.  First of all, why  is this happening?
We studied the dust cake under the microscope and we have found out that the
dust cake differs quite a bit from the noncharged cake.  The charged cake looks
much more porous than the noncharged case.  When you do have more porous dust
cake, you will have less resistance against the flow and you don't need  too much
power to pull air or  gas through  the system.  Hopefully, we will be able to
produce  some  three dimensional pictures  and publish  them for a better  under-
standing.

      Now  if  I go back  to a  power situation for  the  bench scale  testing, we have
up  to 80% decrease  in the pressure drop.   In  our  bench  scale experiment  we have
usually  a corona  formation by a DC charger around  5.5  or 6  kilovolt per  inch
 (kv/in.).   In our  experiments we  obtained  the best  results  between  the 7 or  8
kv/in.   In  the graph  of  power, you can  see that  it  is  less  than  half a watt.
Now I am not  speculating  that will be  the  case in the  full  scale application,
but  at  least  our  bench  scale experiment  shows that  we  use almost all  the power
 for  the  exhaust  fan system  and almost nothing for electric  energy.  Therefore
 energy  savings  up to  80% could be realized.    In 1976 approximately one
 billion dollars  was spent for the power required in fabric  filtration in
 this country.   Therefore,  it appears  that  it  is  very worthwhile  to  conduct
 further research in the electrostatic  filtration.   Now the  second aspect I would
 like to discuss  is the  effect  of  the fabric  structure.   We  all know when energy

-------
442
        ASBESTOS
      8 -
                                    No Charge
                WOOL  (16 oz.)
                   5          10
                VELOCITY (ft./min.)
              ORLON (16 oz.)
                                       No Charge
                  5           10
                VELOCITY  (ft./min.)

-------
                                                        443
   UNCHARGED ASBESTOS CEMENT DUST
DEPOSITED ON FABRIC - 5X MAGNIFICATION
    CHARGED ASBESTOS CEMENT DUST
DEPOSITED ON FABRIC -  5X MAGNIFICATION

-------
444
     FLY ASH

   8h-
o
 CM
X
 -  4
    0
     0
                 V = 7KV/in.

                 T= 81° F
                                      CW=22gr.
                                      RH = 46%
                                         DACRON (22g)
                                         26 gr.

                                         53%

                                       ORLON (I6oz.)
                                 22gr.

                                 48 7o

                               WOOL (I6oz.)
                      22 gr.
                      45%
                  POLYPROP.dSoz.)
                                     10

                         VELOCITY  (ft./min.)
15

-------
                                                                            445

was very cheap nobody cared about possible savings through a lower pressure.
The second transparency shows the effect of fabric type for coal fly ash dust
samples and you can see that there is a considerable effect of fabric type on
pressure drop.  So I think experimentally we should look into the fabric
structure in filtration with or without electrostatics.  I should point out
that Dr. Drehmel and his associates were the first to study the effect of
fabric structure in the filtration process.

     I would like to mention one more important aspect of filtration, humidity.
If we do not control the relative humidity it seems to me there is quite a bit
of difference in experimental efficiency and especially pressure drop results.
In the full size outdoor experiments or testing in the field we have very little
way to control the relative humidity in the air.  It may rain a day just prior
to an experiment or you may run your experiment after the rain and there may
be some substantial changes in the relative humidity.  Until my visit in 1973
to Dr. Drehmel in EPA, I was not too aware of the importance of the relative
humidity factor.  Incidentally, a large number of researchers have been con-
ducting bench scale or full scale test results in the laboratory or in the
field without too much attention to the relative humidity.  However, it is seen
that the relative humidity affects the pressure drop a great deal for a number
of dust samples we have tested with our bench scale equipment. It appears that
a reaglomeration of particles prior to the filtration phase takes place at the
higher humidity level.  As a result it can be seen under the microscope that we
usually have more porous dust cakes than we do have at lower humidity levels.
As indicated by Professor LOffler, this reaglomeration is due to the change in
adhesive forces between the particles themselves and between particles and
filter surface.

     This time instead of bench scale testing we used miniature single bag
dust collector designed by Dr. Helfritch of American Precision Industries as
in the figure.  There is just one single bag with jet pulse cleaning.  With
this single bag equipment we were able to measure the pressure drop for a
number of dust samples with different fabric bags at different humidity levels.
Let me show you a couple of results.  In this figure you see the pressure drop
versus velocity for the three different humidity levels at 24%, 46%, 84%, for  furnace
fume dust.       Again you see  a linear decrease in the pressure drop with an
increase in the relative humidity.  Now, this really shows that we should be
aware of the  fact that if we do not control the humidity conditions in our
tests, it might be quite different to compare test results.  Secondly, it is
possible to condition the gas by water spray not only  to cool off the hot gas
but also to provide a decrease in pressure drop.

B. Liu:  Before we begin the general discussion from the audience, I think
Dr. Robin Oder would like to make additional statements regarding magnetic
separation.

Oder:  Yes, I want to address the question of power consumption in magnetic
separation.   There seems to be a reluctance to accept  the magnetic technology
on the pretense that large amounts of energy are required.   I want to make  the
suggestion  that you make that comparison fairly.  For  example, in the use of
electrostatic precipitators, one might estimate how much energy is consumed

-------
446
      Duct
  Adjusta
  Area
  Nozzle
     Dusty
     Gas
                               FIGURE
                       Miniature Single Bag Pilot

                                   Cleaning Air
                  Jet
              Cleaning
              Air Flow
                  I


Flow -
Control
Valves!
/ 9
_x \


r>
r
ible
t
y



J

ST*
 Hopper
Collected
 Dust
                                                            3-Way
                                                            Solenoid Valve
                            Ck
                                    .•*\
                                     b  *cr
                                   Pulse    Int.

                                    Cleaning
                                    Control
                t
            Pressure  Flow
            Drop      Rate
                                                            Regulator
                                                        i.Compressed Air

-------
  30
                            V= 10 ft/min


                            V =  5 ft/min
                                                      447
o
 CM
X
   20
CL
O

-------
 448

 in terms of kilowatt hours per ton of material that you capture.

      Now, the only commercial technology in high gradient magnetic separation
 is in the processing of kaolin clay.   As I showed you Wednesday,  the separations
 are 500 kilowatt magnets.   That sounds excessive, but if one puts the power
 consumption on a per ton basis, it is not so bad.  Those machines operate in
 a variety of modes depending upon what the clay processor desires.   These
 magnets operate in a range of 9 to 100 kilowatt hours per ton of  material pro-
 cessed.  If you put that in dollars and cents and use 2.4c per kilowatt  hour
 which is an industrial rate, that is  21 cents to $2 per ton.   That is not a
 lot of money.  So I think that if you are worried about power consumption for
 magnetics,  then put these estimates on a comparative basis with your technology.

      Now the point to pulverizing is  that liberated mineral impurities can be
 removed from coal with magnetic separators or any other technology for that
 matter.  It is felt that by pulverizing from 70 to 80% through 200 mesh,  which
 is called power plant grind, one can  remove mineral sulfur from about 80 to
 90% of our eastern coal.   So extraordinary pulverizing to 1 micron size  prob-
 ably will not be required  in physical cleaning.   Such extensive grinding is for
 the future, for selected coals.   So you do not have an extraordinary cost for
 pulverizing coal in that particular case.

 B.  Liu:  I  would like to ask if any panel member would like to comment on the
 remarks by  any other member.

 Benarie:  Yes,   just a brief comment.   Measuring costs of taking  out substances
 is  not a good measure of particulate  separation  because as  everyone knows if
 I  take out  one kilo of substance,  it  means 99.999%  purities cost  a  lot more
 than 50%.   Secondly,  your  price is  to be  compared with flue gas desulfurization.
 As  you know the EPA made a very important  400 page  report about the cost compari-
 son of 6  possible methods  of  flue  gas desulfurization.  After analyzing
 al], the cost is  between  0.6  and  2.4 cents  per  kilowatt  depending  on the
 method.

 Oder:  Yes,  that is  an  operating  cost  figure.   Costs  for  flue  gas  separation
 have been estimated  at  16  dollars per ton  of  coal fed  into  the plant,  That
 is  expensive.   And  furthermore,  you have  to  realize  that  flue  gas desulfuri-
 zation is tied  to the  operating  factor  of  the  plant.   When  the plant  is down,
 you are still  paying for the installation.   But  if you are  not running, you
 are not burning  clean  fuel.  I think  that  is a significant  point.

 Drehmel:  We are entering  the area of comparison between  different  types  of
 control devices and  I  think it is a very difficult area  to  consider.  Not only
 can we  use  the  types of measures already mentioned but of course we also  use
 dollars per kilowatt.  People want to know what is going  into capital equipment.
We can also talk about how much  it is going  to cost the consumer.   That is
 just the  economic basis of the comparison.  We have talked about power.  We
might want  to talk about resident time or total space and floor space.  The
ways in which you could compare a device are so diverse from one application
 to the next that the only really fair way to compare is to take an actual
 circumstance and say, "I've got this type of aerosol with this type of gas and

-------
                                                                             449

these types of retrofit problems."  Is it a retrofit or I can redesign with
some new installation ?  What can I do under these circumstances?  Look at
the actual design.  As soon as you start talking in generalities, about watts
per CFM, dollars per CFM, you are going to get into trouble.

Liu:  I would like to open this up to the floor now.

Baril:  Do you envision that this cleaning of the coal takes place at the
sight of burning?

Oder:  I really didn't intend for that to be coal.  I was trying to get into
gas-particulate but we have done an extensive study for the Bureau of Mines
where we considered the chemical cleaning of coal.  This is short of coal
conversion such as gasification.  This would include, however, physical con-
cleaning and some modest amount of hydro-thermal-chemical operations such as
ditioning of fine coal dust. Dry coal cleaning employing pulverizing might
best be developed at the power plant location unless economical methods are
developed for compacting dry pulverized coal.  With this technology coal
cleaning could be located remote from the power plant site.

Baril:  My second question is, unless you could do it and do it very well to
meet the standards, you would not replace the necessity of having to clean the
exhaust from a coal burning plant.

Oder:  It has been estimated that about 21% of the mid-Western and Eastern
U.S. bituminous coal reserves can be brought into compliance at the 1.2 Ibs
S02/mm Btu level utilizing physical cleaning technologies with the capability
of removing 95% of the mineral sulfur in the coal.  Chemical cleaning, with
the ability of removing 40% of the organic sulfur, is estimated to have the
potential for bringing about 38% of these-reserves into compliance.  The re-
maining coals might be brought into compliance by a combination of technologies
employing physical & chemical coal cleaning, coal blending and flue gas desul-
furization.  The extent to which this mix is employed will depend critically
upon component economics and upon sulfur emissions standards which must be
met.  The proper mix of technologies will most likely be determined on a case
by case basis.

Billings:  It seems to me the next question would be:  What is the iron ore
benefit to that process?  You have half of the weight percent which is iron
that you are going to get out of it and you ought to be able to sell that for
$10 to  $20 per ton or something like that.

Oder:  We have not credited any of the processes.

Billings:  Is it a major credit?

Oder:  Actually the sulfur could be a major credit  in Brazil.  The first
studies by Sergio Trindade in coal cleaning with magnetic separators looked
at coals from Brazil.  Brazil has a lot of dirty coal but doesn't have much
sulfur.  Trindade's emphasis was not environmental, but rather to pursue
recovery of pyrites because of  their sulfur content.  The next major appli-
cation  of magnetic separation,  interestingly enough, could occur in processing
of rich iron ores.  And  these ores are mined and beneficiated for the purpose
of iron ore recovery.

-------
 450

 Billings:   I don't  know what  ore goes  for  but  it  has  to  be  $10  to  $20  a  ton.
 You start  adding  that  up to a hundred  million  tons  a  year of  iron  ore  for  the
 1985 capacity in  the U.S.,  then  you  start  looking at  5 million  tons of coal
 and it  sounds like  you could  meet the  iron ore requirement  from this
 beneficiated material.

 Oder:   It  could be,  but we didn't consider it.

 Billings:   Well,  if  you got a hundred  million  tons  of iron  out  of  that process,
 it  would make it  a very attractive way.

 Oder:   Iron and sulfur could,  in principle, be  recovered as values in  physical
 coal cleaning.  However,  I feel  that annual recoveries on the order of 100
 million tons  of iron are  out  of  the question since  the 1973 market for coal
 supplied to electric utilities was itself  only  380  million  tons.   If 21% of
 this utility  coal had  been cleaned completely of  mineral sulfur  using an
 iron carbonyl process  (32 Ibs  iron carbonyl/ton of  coal), then  I estimate
 about one million tons  of iron at most would have been available as iron-rich
 disulfides.   While this  is not insignificant, it  falls far  short of the 100
 million tons  per year  quoted by  you.

 Walston:  I agree with what Robin has  said - but  just to keep us honest and I
 have vested interest in coal - if we can take out the pyrite, the main goal
 in  the  U.S. is S0» reduction.   But we  still will  have a particular problem
 which I am  sure everyone is aware of.  I just wish  to make  that comment and
 also that a couple of  times during this workshop my name has been mentioned
 in  connection with some HGMS model work.   I want  to mention that Professors
 Richard Treed and William Simons were full collaborative investigators in that.

 Melcher:  Just a suggestion regarding a question  that came up about compari-
 sons among  different clean-up techniques.  It is a pretty simple matter to
 improve the efficiency or pressure drop by using something akin to effective-
 ness factor of a filter that was defined by Chen back in the mid '40's where
 you used like a negative log of  the penetration divided by pressure drop.
When you use augmentation of some sort, rather than using pressure drop perhaps
 you should  use total power input.  That gives you a clear indication of what
 you have done.  You  can always improve efficiency by putting in more layers;
 of  course,  it doesn't help you in the pressure  drop.  And there has been a
 lot of  talk about improving the  process this way  or that way and it seems
 reasonable  to look at  it in terms of the power  input.

 Oder:   That is just  the message  I was  trying to give on the comparison
 between the magnetic filter bed  and the quadrapole.  They are diversive
 approaches  to the same problem.   One has excellent  technical ability from the
 standpoint  of  separation but poor performance from  the standpoint of pressure
 drop and total power requirement.  The other one has moderate to poor  technical
 performance,  high power requirements but no pressure drop problem.  I  am say-
 ing that you  have to look at the total device within the process context.

 Melcher:  In  the magnetic separation in both the  two types,  the bed and the
 quadrapole  design, you end up loosing  some of the coal into the slip stream

-------
                                                                             451

of the coal off the quadrapole or you deposit some in your filter bed.  I
think Billings reported deficiencies with magnet off as high as 50%.  You can't
tolerate a 50% loss to your coal if you are going to run that data.  I can see
it as a clean-up device where you would like to get 50%, but if you are going
to separate, you don't want that to occur.  How much do you think you can
tolerate in terms of loss of coal, for instance?

Oder:  I will tell you what can be achieved in commercial practice for kaolin,
then I will tell you what we estimated for coal. Again I emphasize that coal
is a calculation and an estimate where problems associated with fine clay
inclusion, have been ignored.  For the processing of kaolin clay which is a
wet mineral beneficiation, where 90% of the particles are finer than 2 microns
in size a mineral impurity is removed that has a bimodal distribution, 1/4
micron and 8 micron in particle size.  The magnetic  process operates in a
97% kaolin clay yield.  That is one of the main reasons that the technology
has won out over such alternative technologies as differential flocculation and
flotation.  Now the coal that we used for the Bureau of Mines survey is the
same coal that has been considered in the paper presented here.  It is a coal
that is chosen on the basis of several levels  (1) the organic sulfur level was
low enough for the conceptual study such  that  if you removed most  of the
pyrites sulfur you would make a coal that comes into compliance  (that seems to
be significant) and (2) we chose a coal that cannot be  cleaned with a present
day cleaning circuit.  We chose a coal that had a particle  size distribution
that corresponds very closely to 200 mesh grind.  The feeling was  if we grind
to 200 mesh, then we will do a very good  job of liberation.  Given that size
distribution, we can estimate the liberation in 200 mesh  grind.

      On  the basis of that estimate, the  HGMS  technology  in the  example that  I
gave should receive about a  95% Btu recovery.  The  quadrapole  I  think had maybe
80 to 85% Btu  recovery.   It  is  in  the paper.   There  is  obviously a liberation
problem;  you will have  to pulverize  the coal to eliminate the mineral matter.

Melcher:  You  are saying  then that perhaps  something like 10%  of your coal  is
lost.

Oder:   Look,  no  one has done this.   I am  guessing that  it is 10%.   If you have
problems  of coalescense of  the  particles  like  I have been listening to  here,
the  loss  could be a great deal  more.   If  one pyrite particle (even though it
is ground to  liberation)  has 2  or 3  coal  particles  struck to it  because of
electrical attraction,  when  you take out  the pyrite,  you might take out a lot
of coal.   The estimate is based strictly  on a  simplified, theoretical calculation.

Nichols:   I thought  I would  make one comment about  your magnetic quenching
 sparking  in your electrostatic  circuitry.  Anyone who is going to  do research
 in this area  should  consider the fact that the energy is stored in the electrode
 system rather than  at the power supply.   So your magnetic quencher has  to be
 distributed throughout the whole corona wire conduction electrode system.   The
majority of energy that goes into this part is stored energy that is in the
 distributive capacitance of the system.

-------
  452


 Oder:   Yes, I wouldn't know how to do it.   I would suggest either couple it
 inductively or put in a permanent magnet.

 Nichols:   Yes, that is one area they might do research in.  It has to be in
 a box.

 Friedlander:   First,  last night I gave it  some thought and I guess this exist-
 ing system has some merit to it,  because it quenches quickly,  it  localizes  the
 arc, but  I also would like to consider any method  that you always quench after
 you stop  the  arc.   In magnetic method you  use nothing to quench unless you  stop
 it. So you still  have a problem  there.  I want to get back to something.   My
 feeling is quite strong about the many technologies of the agglomerations
 sticking  together.  It is really  important in two  ways,  (1)  To avoid  where  you
 don't want it and  (2)  To use it in seeding.   Two years ago I read about water-
 based technology of seeding water with iron oxide  to remove bacteria  viruses.
 Apparently it works very well and I have not heard much  more about it.   The air
 system,!  haven't heard anything about it yet except here.   There  are  some
 problems  that can be  solved that  way.   But I think a lot of  work  has  to be  done
 to  understand the system and the  technology  to  make it work.   The magnetic
 technology is a peculiar one.   There are not that  many particles  around today.
 But what  about  seeding magnetic particles,  then getting  that magnetic particle
 to  marry  another magnetic particle?   It  is not  the usual agglomeration problems
 where you have  to get  a lot of  them together.   If  you  don't  have  high densities,
 you only  need one magnetic  particle to remove particles  so you  only have to
 marry two particles and not a  large number of particles.   The  time constant may
 not be  that bad.

 Liu:  I would like  to  change the  topic just  slightly.  I  think  I  would  like to
 ask Don Reid  about his  electrostatic  augmented  filters.   I am just wondering:
 eventually the  filter will  clog up  and you have  to  replace it,  don't  you.   I
 just wonder what type of  application you think your  equipment would have.

 Reid: We  haven't done any extensive studies  of the  cleaning techniques.  Again,
 we do know that we can  clean it back  to  the  original pressure drops, which
 indicates that we have  a  clean bed.  It has  a high  loading capacity.  We have
 operated  some of these  filters for  60 hours  and only had 0.1 inch  increase in
 pressure  drop.  Of course the length of operation time before cleaning  is going
 to be a function of concentration of the aerosol.

Liu:  So you are intending  to use this as a clean up filter.  You wash  it after
 it gets clogged.

Reid:  You have to  clean it.  It is not one you just throw away.

Liu:  I see.

Billings:   I wonder if Professor Loffler would give us the benefit of his think-
 ing about  the strengths of aggregates or agglomerate particles.  What the forces
are, what  the orders of magnitude are, that kind of thing.  Would that be a
fair question?

-------
                                                                             453

L'dffler:  I think it would be a topic for another meeting if you talk about
agglomeration processes between various particles.  I could only give a very
brief review of the main mechanisms.  We could list all the forces which are
possible, we can go back to non-material binding, such as Van der Waals forces,
electrostatic forces, or any other dipole interaction.  We have worked with a
material which is like a liquid (water or something like water) where there
are chemical forces, surface tension forces and you might have some solid which
is crystallized.  Those are some of the mechanisms.  As far as magnitudes:  The
highest of course is ? solid which might also be an effect of surface diffusion.
The next range is the liquid which also acts over large distances.  The
magnitude below that you get the van der Waals forces.  The order below that,
electrostatic forces.  As long as the particles are in direct contact the
distance must be in the range of several angstrom units.  But electrostatic
forces are long range forces.  To give you an idea about the orders of magnitude:
For a 1 ym particle, the van der Waals adhesion force, to a surface would be
one million times the gravity.  That should give you an estimate on the amount
of energy needed to deagglomerate.  This means that it is normally impossible to
deagglomerate very fine particles only by shear forces in a gas flow.  It is
easy through the use of impact forces.  This way you can deagglomerate particles.
We found that in deagglomeration, impact onto a hard surface, you need velocities
approaching 200 meters/second.

Oder:  Professor Friedlander asked if anything was occurring in the
seeding technology for wet processing with magnetic separators
and the answer is "yes".  It bears on what you are saying.  There is a great
deal proceeding from the standpoint of industrial development for processing
the effluent water from municipal and industrial complexes and it is a technology
that utilizes the seeding technology.  Coagules are formed in a normal
flocculation chamber where you change the pH so as to make the particles
coagulate and form on the surface of magnetic iron oxide particles which are
about 20 microns in diameter.  Now what has been found is that the magnetic
capture force for these coagules in high velocity flow through that packed bed
is so great that it is shear force that is breaking up the coagules.  The process
rate through the magnet is limited by the breakup of the coagules and not by
the availability of a magnetic force to hold the seed material on the capture
surface.  In that case we have a seed material that is rough in shape, about
20 microns in diameter, and that can capture four or five times its volume in
entrained materials.  Studies have been made of capture of iron oxide particles,
oil agglomerates and even bacterial systems which are the products of tertiary
and secondary sewage processing.  You can capture and hole these up to a point,
but as the flow velocity increases eventually viscous forces will break the
coagules.

Billings:  I think that there seems to be a general concern, both on the gas
side and the liquid side, to attempt to understand something that is pretty
fundamental.  That is, what is the force of adhesion between two fibers, similar
and dissimilar, and how does that compare what those forces are?  First, where
do they come from?  Second, how does one speculate about the order of magnitude?
Then you can begin to compare those forces between the particles with the other
forces that come into play, like the fluid mechanical forces.  The forces of
adhesion between particles in gases at normal temperatures and pressures are
enormous.  They are very much higher than in liquids.  It becomes very difficult

-------
 454

to go from one fluid to the other because the adhesion forces dominate in the
gas side.

Oder:  Is that because of double layering in the liquid?  Do you have shielding
of the charge in some parts?

Billings:  From an engineer's point of view, the adhesion force of the particles
in a gas without the Van der Waals, without electrical, without sintering, with-
out actual contacts is primarily the surface tension of water.  It is
essentially a humidity effect.  It is demonstrated very easily by varying the
humidity and examining the force between the particle and the surface.  The
point is that tension force is higher than anything else.  The adhesion force .
is of the order of 5 dynes.  The weight of the particle is on the order of 10
dynes.  The drag force = SiiyV^ is of the order of 10"^ times the velocity.  So
the velocity has to be of the order of 10^ meters/sec in order to blow that
particle off the surface.

Oder:  Question. Do you know of, or does any one know of, dispersing materials
to be used in dry systems?  Wet particulates can be stabilized in fluids with
detergents effectively.  One adds polyphosphates so that the particulates be-
come stabilized by charge repulsion.  Are there equivalent materials for use
with dry material to promote stability in separation in particles?

Liu:  An answer to that is no.  A number of chemical warfare people have tried
for many years to stabilize the aerosols so that they would not flocculate and
coagulate.  I think the efforts have all been negative.  You do not have an
equivalent double layer situation in a gas system.  When they come together,
particles simply stick.  I think one of the biggest problems with gas separation
is you have to deal with very high concentrations of particulate in order to
reduce the volume of gas in the process.  That is conducive to coagulation.
Coagulation has to be looked at very carefully before one can say anything about
the economic feasibility in magnetic separation.

Scott:  I just want to return to the electrostatic filter.  Question:  It is not
my field, but what is the physics or mechanism that allows this particular pro-
cess to occur with much lower pressure drops.  I'm particularly intrigued by
one of the slides where you had low pressure drop, then you turned off the
electric field and the pressure drop went up; then, magically, you turned on
the electric field and I think it dropped again.  I wonder if I misinterpreted
that or could someone give me a little bit more physical interpretation of that
part.

Ariman:  Dennis, would you please make some comments on the last part of the
question?

Helfritch:  When the electric field is turned off, we do know we have briefly
a precipitated section that the aerosol has to pass through before it is filtered
by the fabric. Some of it is taken out.  Second, when dust deposits on the
fabric in the charged state, it deposits in a different way, more  porously
and openly.  When the charge is turned off, the dust then can deposit in its

-------
                                                                             455

normal fashion, which is a fairly tightly packed structure with a lower
permeability.  We clean the fabric periodically.  The entire unit is cleaned
every 5 minutes.

Ariman:  Let me add a couple of things to his comments.  We do have a more
porous dust cake when we have an electric charge on the particles.  In the
pictures you will see just the particles are charged.  We do have charges on
particles, and it looks like they keep their charge at least for a while de-
pending on the filter.  In bench scale experiments once one particle lands on
the filter the next one because of the charge of the first particle lands some-
where else on the filter.  However, the second one would land next to the first
particle in the case of non-charged particles.  Therefore for the charged case
there is a nonuniform distribution of the landing of particles on that particular
filter.  As I indicated earlier there are little hills, valleys, and bridges.
If you are not careful in carrying the dust cake to the microscope, you can damage
the cake structure.  It is very fluffy and porous.  We hope to shed more light
into it in the near future.

Liu:  I would like to suggest one of the more important mechanisms might be
due to the poly dispersity of the dust.  If you do not have an electrostatic
charge, then the fine particles would have a tendency of filling the pores be-
cause this is where it can be collected by interception.  Electrostatically
charged aerosols tend to be collected at the entrance of those pores, probably
keeping those pores open so that the fluid can go through.

Cooper:  One of the things that struck me was decrease in the flow resistance
of the dust cake as humidity increased and adhesive forces which might tend to
pack it tighter and electrical forces which may decrease by humidity.  Those
results are surprising.  The humidity is allowing us to build up a more open
cake simply because once we get a particle striking another, it stays in a
forward position rather than going into a backward position.

Ariman:  First of all, let me clarify one thing.  The results provided here
were not for the charged cakes.  They were obtained by the use of single
miniature bag testing devices.  We really tried to compare our bench scale
testing which uses flat fabric with a regular single bag, but in a small scale.
So we did not have any charge.  Also we did have greater complexity when we
had electric charge because of the interaction between humidity and the electric
field.  Now I'll go back to your first question - how come the humidity was
causing a decrease in the pressure drop? It is my understanding, that we do
have a change in the adhesive forces between the particles when relative
humidity varies.  Also the same thing applies for the forces between the
particles and the fiber filter surface.  It looks like there is an increase
in agglomeration process of the particles before the landing on the filter.
Instead of having very tightly packed dust cakes we have quite porous ones.

Penny:  There is one factor that wasn't mentioned in connection with humidity
and that is  the question of resistivity.  For the information of the porous
cake the surface has to be in equipotential.  If the surface is equipotential

-------
456
  then the field will concentrate on any projection.  It turns out that requires
  a certain amount of conductivity for the surface to be in equipotential.  So it
  seems to me that when we mention humidity, we ought to mention the kind of
  material and how its resistivity varies from humidity.  In other words, it is
  not sufficient to have charged particles in that cake - you really want that
  charge conducted.  If you do it right you build up these particles into almost
  like fiber like clouds.  But that requires high enough conductivity.

  Ariman:  I agree with Professor Penney. Resistivity is another area that I was
  going to suggest as equally important with relative humidity.  I think we should
  look into the resistivity very, very closely and for bench scale testing, we
  should find a way to measure the resistivity of each dust particle we have been
  using.  We should also try to incorporate resistivity into the mathematical
  model of filtration in an acceptable way.

  Loffler:  What is needed is to go into more detail, not in theory but in
  experiment, to get more information on the material qualities.  All the effects
  that were observed rely on very specific materials and conditions.  For example,
  we did measurements on low humidity using a very similar apparatus as you did
  and found that in this case the particles formed a cake.   There was a continuous
  variation in the permeability of the cake.   That means the cake was compressed.
  There was an increasing pressure drop across this cake and this could be
  theorized by going for a short moment with a higher velocity and with a lower
  velocity.  You would obtain a lower permeability.  I would speculate that in
  your case one of the possible reasons is with increasing humidity you get a
  liquid which is at the contact point and is without a higher adhesion force iri
  stronger agglomerates.  This type of cake perhaps does not reduce permeability.

  Dr.  Friedlander:  Basically this research turns out to strongly involve the
  property of solid materials.  On the other hand, those of us at the materials
  research laboratories, are told that any work that considers such things can
  not be considered to have anything to do with money that is given to the study
  of materials which is also contradictory because we end up right at this point
  finding that where the lack of knowledge is extremely strong is in the under-
  standing of the solid materials involved.  The NSF says that this is not so.
  But yet it is interesting to find that the interaction between these two may be
  extremely important.   It should be encouraged rather than discouraged.

  Liu:  Dr. Ojalvo, I would like to hear your comments on that.

  Ojalvo:   I certainly agree and it takes workshops like this to bring this sort
  of thing out.   Eventually this reaches the people at the top of the NSF ranking
  and decision-making and it has been argued strongly enough, forcefully enough,
  and a number of times.  It is good this sort of work can be brought together.

  Billings:  It seems like Dr. Friedlander's  comments should be modified in the
  following way.   That we really aren't interested in the bulk properties of
  material.  To a great extent, these are inconsequential.  We really are interested
  in interfacial property.  And that has nothing to do or very little to do with
  the material science area.

-------
                                                                             457

Friedlander:  Surface studies are the big area - Purdue University does surface
study and get well paid.  This is precisely the interaction of surface study on
the solid material, the oxide.  And here we find a different viewpoint but
exactly the same property, when you get down to real basics, we are talking
about the same thing they are.  Maybe they know a lot more than we do except
maybe it doesn't apply to this function.  We are saying the same thing but for
a different situation.

Lamb:  We have reviewed reports by Lo'ffler on a phenomenon which he has re-
ported reducing the pressure drops by a factor of six.  The important thing
is that he describes this and nobody knows why it is happening.  I am somewhat
discouraged that he lists several papers on the theory of fiber filtration that
always come to deal with a single fiber being struck by a single particle.
They always deal with conditions that can be described as fictitious.  The
situation is one in which we are coming to know very precisely whatever happens.
The filter is always clean and we never see what happens when the second
particle comes along and is faced with a different situation.  Perhaps it is
time that we devoted efforts to this very messy situation.

Ariman:  I agree with you, Dr. Lamb, that is one of the major points that I
have tried to stress in my presentation.  The time has come to go from this
single fiber, single particle approach step-by-step to the real filtration with
close coordination of experimental work.

Liu:  The time is getting late.  Some people have to leave for the airport so
I think we should close at this point.  So before we do that I would like to
ask Dr. Ojalvo and Dr. Drehmel if they have some closing remarks.

Drehmel:  It seems to me that one thing that we really haven't addressed is
how good our theories are.  Do we believe them and what happens when that
second particle comes in?  I agree with Dr. Lamb - there are a lot of other
theories that can be proposed.  So far as comments on this workshop are
concerned, I think that we owe a great deal of thanks to Dr. Ariman for organiz-
ing this conference and getting the facilities for us.  I think the attention to
details that he has given us on this workshop has just been fantastic.

Ojalvo:  I would like first to make some general remarks about the National
Science Foundation (NSF) and then conclude with information about my program
and its relationship to this workshop.

    The basic purpose of NSF is to initiate and support scientific research
(including social science and engineering) and programs with scientific research
potential - such as science education.  To accomplish this purpose NSF is
organized into seven directorates, each headed by and Assistant Director, under
the Director of the NSF.  The first three directorates are concerned with basic
research - Mathematical and Physical Sciences and Engineering; Astronomical,
Atmospheric, Earth and Ocean Sciences; and Biological, Behavioral and Social
Sciences.  The other four are:  Science Education; Research Applied to National
Needs  (RANN); Scientific, Technological and International Affairs; and finally
a Directorate for Administration.  The first three directorates  (basic research)
are the heart of NSF and account for 70% to 75% of its total budget.

-------
 458

     There  are  five divisions  in the Mathematical  and  Physical  Sciences  and
 Engineering  Directorate  -  Physics; Chemistry; Mathematical and Computer Sciences;
 Materials  Research;  and  Engineering.   The  Engineering Division consists of  three
 sections:  Electrical  Sciences  and Analysis; Engineering Mechanics; and the
 section with which I am  associated -  Engineering  Chemistry and Energetics.
 The  program  which  I  direct, Solid and Particulate Processing,  is one of five
 in this section.   The  other four are  Heat  Transfer; Engineering Energetics;
 Thermodynamics and Mass  Transfer; and Chemical Processes.

     The four major areas in my  program are:  Particle and Particle System
 Characterization;  Particle Generation and  Size Modification; Interfacial and
 Colloidal  Phenomena; and finally, the area of interest  in this workshop -
 Particle Processing  and  Separation.

     The program is quite new  having been initiated two  years ago.  One  of the
 first activities was to  hold  workshops.  There was much interest from mineral
 engineers, so  the  first  workshop was  on Mineral Processing.  Then, the  Fine
 Particle Society conducted a  workshop on Particle Technology.   One of the
 conclusions  of  that workshop  was that particulate - gas separation was  an area
 that could be  explored in  a separate  workshop by  itself.  So here we are doing
 just that!

     I want to  congratulate Dr. Ariman for  conducting  a  very fine and productive
 workshop.  I feel  that quite  a bit has been accomplished.  I would like  to
 have the Proceedings include  all the  papers, invited  lectures,  discussions and
 comments presented here  so that we at NSF  (more specifically the Solid  and
 Particulate Processing Program) and others in the technical community can use
 it as guidance  to  focus  our efforts on the promising  research  opportunities in
 particulate -  gas  separation.

 Ariman:  Thank you, Dr.  Drehmel and Dr. Ojalvo, for the nice comments you have
 made concerning the workshop.  First  of all, I would  like to thank our
 distinguished  lecturers, Professor Loffler and Dr. Benarie.  They were kind
 enough to come from Europe on such a  short notice.  I also would like to thank
 the  lecturers and  the designated discussers for their fine contributions and
 certainly the  important  contributions of all the other participants through
 open  discussions.

     It is proper also to express my appreciation and  thanks for the great help
 to the members of  the organizing committee, Professors Ben Liu and K. T. Yang.
 Also special thanks to our sponsors NSF, EPA and host institution University of
Notre Dame.  Without their help, there was no possibility of having this
workshop.   Very special  thanks to Dr.  Ojalvo of NSF and Dr. Drehmel of the EPA,
 for  their continuous support  from the beginning and for their wise advice and
 suggestions.   So,  on behalf of the organizing committee and the University of
Notre Dame, we are very  grateful to them.  I hope it was a useful workshop
 for  every participant and I would like to congratulate  all participants .for
 their very valuable contributions.

-------
                                TECHNICAL REPORT DATA
                          (Pleate read Inunctions on the reverse before completing)
 1. REPORT NO.
 EPA-600/7-78-170
2.
                            3. RECIPIENT'S ACCESSION NO.
      AND SUBTITLE  proceedings: Symposium on New
 Concepts for Fine Particle Control
                            5. REPORT DATE
                            August 1978
                                                       5. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)

 Teoman Ariman, Compiler
                            B. PERFORMING ORGANIZATION REPORT NO.
9 PERFORMING ORGANIZATION NAME AND ADDRESS
 University of Notre Dame
 Notre Dame, Indiana 46556
                            10. PROGRAM ELEMENT NO.
                            EHE624
                            11. CONTRACT/GRANT NO.
                            EPA Grant R805148; NSF
                             Grant ENG. 77-02016
 12 SPONSORING AGENCY NAME AND ADDRESS
 SPA, Office of Research and Development(*)
 Industrial Environmental Research Laboratory
 Research Triangle Park, NC  27711
                            13. TYPE OF REPORT AND PERIOD (
                            Proceedings; 1/77-7/78
          (COVERED
                            14. SPONSORING AGENCY CODE
                             EPA/600/13
 is. SUPPLEMENTARY NOTES Project officers are: Dennis C. Drehmel (EPA), Mail Drop 61,  919
 541-2925, and Morris S. Ojalvo (NSF). (*) The National Science Foundation is co-
 sponsor.
 i6. ABSTRACT rpj^ rep0rt documents presentations made during a symposium on novel
 concepts, methods, and advanced technology in particulate/gas separation. The sym-
 posium ,  held at the University of Notre Dame and sponsored by the National Science
 Foundation and the Environmental Protection Agency, was held both to identify new
 research areas and to stimulate future research activities. The presentations
 included two general lectures by internationally known scientists (Dr. M. Benarie of
 France and  Professor F. Loffler  of West Germany) and 17 invited lectures by other
 prominent experts.
17.
                             KEY WORDS AND DOCUMENT ANALYSIS
                 DESCRIPTORS
                                           b.lDENTIFIERS/OPEN ENDED TERMS
                                                                    c.  COSATl Field/Group
 Pollution
 Dust
 Flue Gases
 Measurement
 Separation
 Air Cleaners
               Pollution Control
               Stationary Sources
               Particulate
13B
11G
2 IB
14B

13A
 9. DISTRIBUTION STATEMENT

 Unlimited
               19. SECURITY CLASS (This Report I
               Unclassified
21. NO. OF PAGES

   459
                                           20 SECURITY CLASS fllils page)
                                           Unclassified
                                        22. PRICE
EPA Form 2220-1 (9 73)

-------