EPA-600/2-77-201
September 1977
Environmental Protection Technology Series
                 GENERATION AND SIMULATION
 OF  METALLIC  PARTICULATE AIR POLLUTANTS
                   BY ELECTRIC ARC  SPRAYING
                            Industrial Environmental Research Laboratory
                                 Office of Research and Development
                                U.S. Environmental Protection Agency
                            Research Triangle Park, North Carolina 27711

-------
                     RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental Protec-
tion Agency, have been grouped into nine series. These nine broad categories were
established to facilitate further development and application of environmental tech-
nology. 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 ENVIRONMENTAL PROTECTION TECHNOLOGY
series. This series describes research performed to develop and demonstrate instrumen-
tation, equipment, and methodology to repair or prevent environmental degradation from
point and non-point sources of pollution. This work provides the new or improved tech-
nology required for the control and treatment of pollution sources to meet environmental
quality standards.
                             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 Information
Service, Springfield, Virginia 22161.

-------
                                               EPA-600/2-77-201
                                                 September 1977
         GENERATION  AND SIMULATION
OF METALLIC PARTICULATE AIR POLLUTANTS
           BY  ELECTRIC  ARC SPRAYING
                              by

                   B. Linsky, R. Hedden, M. Naylor, and F. Dimmick

                        University of West Virginia
                       Morgantown, West Virginia 26506
                          Grant No. R801858
                         ROAP No. 21ADM-025
                        Program Element No. 1AB012
                     EPA Project Officer: Dennis C. Drehmel

                    Industrial Environmental Research Laboratory
                     Office of Energy, Minerals, and Industry
                    Research Triangle Park, North Carolina 27711
                            Prepared for

                    U.S. ENVIRONMENTAL PROTECTION AGENCY
                      Office of Research and Development
                         Washington, D.C. 20460

-------
                                  PREFACE






     The growing concern about fine particles was barely detectable among




air pollution specialists in the late 1950's.  It surfaced in the emission




inventory of the San Francisco Bay Area Air Pollution Control^District




Technical Report of 1958.  By the early 1970's recognition of the need to «




control fine particles separately had been established.  It became U.S.




Environmental Protection Agency policy to deal with these microscopic and




sub-micrometer particulates differently than larger ones from a technological




control viewpoint, from an undesirable effects viewpoint, from a regulatory




viewpoint, and from a measurement viewpoint.




     The fine particles that were technically available for research and




engineering development were either surrogates for realistic fine partic-




ulates, such as DOP, or were stale dusts that were redispersed in a gas




stream.  Some of the surrogates came from aqeous solutions or suspensions




and were high in water vapor content as they entered the air or other gas




streams.  Caught in a dust collector weeks earlier and redispersed, the




metal or metal oxide fine particulates were  not fresh even though it is




generally recognized that fresh metal fumes  are often more reactive than




stale ones.




     There seemed to be a growing demand for a fresh dry fine particle gen-




erator of realistic metallic oxides such as  a fine particle generating





                                     ii

-------
 system with a fresh, dry particle output somewhat similar to that from  an



 electric arc furnace.  It was hoped that the mass per unit time (mass



 emission rate) and mass per unit volume of gases (mass volumetric concen-



 tration) would be sufficient to allow a realistically high dust-in-exhaust-



 gas  stream for testing bench scale or small pilot scale dust collector to



 "improve the breed" of such collectors.  The U.S. Environmental Protection



 Agency having an immediate need for such a fine particle generator because



 of work they were doing and extramural work they were having done for them



 on charged droplet fine particle scrubbers and other fine particle control



 equipment research and development purchased identical equipment to the



 metallizer generator when the ongoing findings of this research grant indi-



 cated a resonable possibility of success in mass volumetric concentration



 and  mass emission rate of fine particles.



     This report contains, as appendices, the theses and problem reports




 of the graduate students who did most of the research work on this project.



 As will be noted, most of the electron micrography and all of the elemental



 and  chemical compound analytical work were done by a specialist contractor,




Walter C. McCrone Associates.  Much of the fine particle sizing was also done



by the McCrone organization.
                                    ill

-------
                                  ABSTRACT






     This research program was a pragmatic experimental activity that was




conceived to provide a long-needed research tool for those concerned with




very fine particles and their technological control.  In order to try to




provide a generated output with an appropriate mass and concentration of




fresh dry fine metal oxide particles for bench or pilot scale collector




research and development work, electric arc generators of fresh fine par-




ticle aerosols were modified and tested with zinc metallizing wire and with




steel welding and  metallizing wire at West Virginia University.




     Work with two electric arc aerosol generators is reported; one generator




employed a single consumable electrode of welding wire and the other generator




employed two consumable wire electrodes of a commercially available electric




arc metallizer.  The generated aerosols were exhausted into a duct system




and sampled using membrane filters.




     The single electrode generator produced 0.67 grams per cubic meter of




0.1 micrometer diameter fresh iron oxide agglomerated particles as sampled




by an Andersen Stack Sampler.  The mass emission rate with an average of 1.95




grams per minute varied within a plus or minus 12 percent range.  The basic




particles that formed the 0.1 micrometer agglomerated particles were charac-




terized to be log normally distributed with a mass median diameter of 0.0167




micrometer as observed and measured by electron microscopy.





                                     iv

-------
      The  two  consumable electrode aerosol generator produced agglomerated




submicron particles  (as measured by scanning electron microscopy) formed




from 2-20 nanometer  diameter basic particles (as measured by transmission




electrode microscopy).  Mass volumetric concentrations ranged from 0.7 -




2.0  grams per cubic  meter for zinc oxide aerosols.  The mass emission rate




averaged  8.6  grains per minute for the zinc oxide aerosols and 7.7 grams per




minute  for the iron  oxide aerosols.  Selected area electron diffraction




analyses  found the zinc oxide to be ZnO and the iron oxide to be FejD..




The  two electrode aerosol generator was further tested and validated for




reproducibilities of total mass volumetric concentration and basic particle




diameter  distributions.  Variables of operation were investigated to deter-




mine their effect on the mass volumetric concentration of the fresh dry fine




particle  aerosol produced by the generator.




      This  report was submitted in fulfillment of Grant No. R-801858-01-1




by West Virginia University under the sponsorship of the U. S. Environmental




Protection Agency. This report covers the period July 20, 1973 to January 15,




1975, and work completed outside of the grant period through April 29, 1977




was included.

-------
                                CONTENTS






                                                            Page



PREFACE	ii




ABSTRACT	iv




FIGURES	vii




TABLES    	viii




CONVERSIONS FROM METRIC TO ENGLISH	ix




ACKNOWLEDGMENTS    	   x




     A.  INTRODUCTION 	   1




     B.  CONCLUSIONS AND RECOMMENDATIONS	6




     C.  EXPERIMENTAL PROCEDURES  	  11




             Apparatus	12




             Sampling    	17




             Analytic Techniques  	  19




             Experimental Designs 	  19




     D.  RESULTS AND DISCUSSION	21




APPENDICES




     1.  THE HEDDEN REPORT	1-i




     2.  THE NAYLOR REPORT	2-i




     3.  THE DIMMICK REPORT	3-i
                                   vi

-------
                               FIGURES


Number                                                       Page

   1     Simplified diagram of aerosol generator
           apparatus (Hedden) 	 13

   2     Metallizer equipment used by Naylor 	 15

   3     Shrouded head metallizer equipment used
           by Dimmick	16
                                  vii

-------
                                 TABLES


Number                                                             Page

   1     Summary of mass volumetric concentration and emission
           rate with approximate range of agglomerated particle
           size indicated (Redden)  .   '.	22

   2     Relative basic particle size frequency distribution
           in percent (Hedden)   ,	24

   3   •  Zinc oxide mass concentration mean and variance
           summary (Naylor)   	 25

   4     Means and variances of zinc oxide basic particle size
5

6
7
8
Analysis of variance for Latin Square experimental
design mass volumetric concentration data (Dimmick) .
Statistics for agglomerated samples (Dimmick)
Statistics for very fine basic particles (Dimmick) .


. 28
. 29
. 30
. 31
                                   viii

-------
         CONVERSIONS FROM METRIC TO ENGLISH
1 meter            -  1010 A = 3.28 feet

1 meter/sec        =  3.28 feet/sec
                3                          3
1 cubic meter (m ) =  35.315 cubic feet (ft )

1 m3/min           =  35.315 ft3/minute
                                           _3
1 gram             =  15.43 grains = 2.2x10   pounds
        3                          33
1 gram/m           =  0.44 grain/ft = 0.0624 pounds/1000 ft

1 gram/min.        =  0.138 pounds/hr.

°C                 =  (°F-32)/1.8
                         ix

-------
                               ACKNOWLEDGMENTS

     The cooperation of several present and former materials identification
specialists at West Virginia University is recognized and appreciated.  They
are named in the component reports and theses.  Assistance beyond the ordi-
nary level of equipment and service supplier's advice and guidance are also
appreciated and recognized from Wall Colmonoy, Flamespray Industries, and
McCrone's firms.

     There were also inumerable discussants by phone and at various air pol-
lution specialists' gatherings, ranging from a Gordon Research Conference
through an Air Pollution Control Association session to several EPA supported
conferences on fine particulates.  It would not be possible to name all of the
contributors of ideas, experiences, cautions, etc.

     Considerable financial support, other than that which was provided in
the research grant itself, was in the form of U.S. Environmental Protection
Agency traineeships for graduate study from the Office of Air Pollution
Manpower Development, as is shown in the component reports and theses.

-------
                                 Section A




                                INTRODUCTION






     The need for a generator of fresh fine particle aerosols and the lack




of a suitable generator led to the proposal of research and development of a




fine particle aerosol generator for use in fine particle air pollution con-




trol research and development.  The general purpose of this research was to




prepare and characterize high mass volumetric concentration, fresh fine




metallic oxide particle aerosols.  In particular, the electric arc process




of particle generation was chosen as most likely to provide the desired




product.  This required characterizing the fine particle aerosols and the




mass concentrations produced by an electric arc aerosol generator.




     Two electric arc aerosol generators were investigated and developed




by graduate students at West Virginia University under the guidance of




Professor Benjamin Linsky.  Robert Redden, having developed an initial single




consumable electrode aerosol generator, reported results on a modified ver-




sion of his generator in his Masters report.  Michael Naylor and Fred Dimmick,




working with a commercially available metallizer employing two consumable




electrodes, reported results in their Masters reports.  These three Masters




of Science reports are the basis of this summary report; each of the Masters




reports is attached as an appendix.




     In order to place into proper perspective the information from these

-------
Masters reports, a short developmental history from Hedden's  first  consumable




electrode electric arc aerosol generator  to the current  form  of  the metalllzer




is presented.  Again, differences in equipment configurations, differences in




aerosol sampling methods and differences in analytic techniques  are presented




tp provide insight into the background of the information.








     During his graduate study, Robert Hedden developed  a number of very fine




particle generators which used an electric arc welder system. His  systems




changed based on his experience; one of his systems used two  consumable wires,




another used consumable wire and a non-consumable fixed  electrode and another




used one consumable wire and a non-consumable rotating electrode.  He




examined the last system in detail; it consisted of a consumable wire feed




as the anode and a relatively non-consumable tungsten electrode  as  the cathode.




Carbon steel welding wire was used for the anode.  Runs  were  limited to 30




minutes due to the tungsten cathode being consumed after that length of time.




     Modifications of Hedden's system were developed in  1973-74  by Professor




Linsky, Robert Hedden, Dr. William Fischer, Arland Johansen,  Michael McCawley,




Rasool Nekooi and Dr. Richard Sears.  This system consisted of using a Hobart




continuous electric arc welder with two consumable wires from 2  separate heads




each with its own wire feed motor and motor speed regulator.   This  system was




unstable because of inadequate control over several factors including the




guidance of the wires; for example, sometimes gaps between the wires were




greater than 0.25 inch.  Data collected using this system was insufficient




to establish conclusive results.




     In December, 1974, a commercial metallizer was found on the industrial




consumer market that seemed to have suitable properties.  The Electrospray

-------
metallizer unit (Wall Coltnonoy Corp., Model VT-500) was selected after exam-




ining alternatives such as Model 10E metallizing gun of METCO, Inc.




     The Electrospray metallizer was put into operation by Professor Linsky,




Michael McCawley, Joe McCauley, Dr. Sears and Anthony Angotti.  After estab-




lishing the generation system consisting of the metallizer, a collection




barrel and an exhaust system for the aerosol, Robert Redden, John Garbak,




Don Stone and Mike Naylor proceeded to calibrate the equipment used in the




process and tried to establish stability on generation runs.  Even though




reliable stability on runs was not established, periodic stable runs allowed




some sampling.




     Early work with the metallizer, which initially consisted of generating




aerosols with metallizer steel wire  (coated with copper), was followed by a




change to 15 gauge zinc metallizing wire.  Continued difficulty in estab-




lishing reliably stable generation runs occurred using the zinc wire.




     In December, 1975, Naylor, Stone, and Dimmick made modifications and




ran several successful runs with the zinc metallizing wire on the aerosol




generator.  Naylor collected samples in January, 1976 for his report.




     During the summer of 1976, Dimmick, Stone and Craig Repp worked on aerosol




generation with steel welding wire and copper coated steel metallizing wire.




After a few modifications and additions, Dimmick took samples with welding wire




and metallizing wire.








     Differences in generator configurations, in sampling methods, and in




analytical techniques are evident when comparing the three Masters reports.




The obvious differences between the  Hedden design and the commercial




metallizer design may overshadow the subtle differences in  the  generator




                                      3

-------
configurations of Naylor and Dimmick.  For example, Naylor positioned the




separation and collection barrel vertically while Dimmick positioned the




barrel horizontally; Dimmick employed an electrode tip shroud and Naylor




did not.  Redden's power source was a constant voltage source; Naylor and




Dimmick's was not.  Also, Hedden's arc chamber was above atmospheric pressure




and Naylor's  and Dimmick's was below atmospheric pressure.




     Differences in sampling methods center around two points:  (1) methods




used and  (2) proficiency of each experimenter.  Redden used an Andersen




Stack Sampler followed by a membrane filter.  Naylor and Dimmick used mem-




brane filters to collect particle diameter count and mass volumetric concen-




tration samples.  While Naylor's mass volumetric concentration samples were




collected generally under isokinetic conditions, Dimmick's were not and




Redden did not report an isokinetic estimate.




     Differences in analytical procedures may have led the research astray.




In particular, Redden employed typical electron microscopy techniques of




particle measurement including dispersion of particle aggregates.  Although




Hedden explained that chained and agglomerated particles should be expected




from electric arc production of aerosols, Naylor and Dimmick, through the




McCrone organization, measured particles in the 2-20 nanometer range while




an awareness of an apparent agglomerated-particle aerosol grew when properly




loaded membrane filters were evaluated from scanning electron micrographs.




Also, the inherent differences in sampling methods led to other differences



in analytical measurements.




     Despite the differences in equipment configuration, sampling techniques




and analytical methods, the three Masters reports have fulfilled the objectives




of aerosol characterization and apparatus investigation.  Certainly some

-------
questions have been answered, some have been partially answered and many more




have been raised.  The report that follows should be taken in the light  of




the differences and similarities between each Masters report.

-------
                                  Section B




                       CONCLUSIONS AND RECOMMENDATIONS'






     Hedden concluded that  based on  the results of limited experimental




tests the feasibility of  employing an electric arc to produce aerosols of




very fine iron oxide particles  from  a feedstock of consumable wire has been




demonstrated.   More than  30 grains per minute  (1.95 g/min) of discrete,




spherical particles with  a  mass median diameter of 0.0167 micrometers were




repeatedly produced within  a range of plus or minus 12 percent.  Also, a




conservative examination  of the mass distribution data obtained  from the




Andersen Stack Sampler particle sizing technique indicated that  approx-




imately 75 percent of the particles  have an  effective diameter less  than




0.1 micrometers.




     Hedden recommended:




     (1)  With modification, the unit might  be adapted for the production




          of aerosols of  non-conductive refractory materials.  This  might




          be accomplished by replacing the wire  torch with another non-




          consumable electrode  holder, establishing an arc between them,




          and guiding the non-conductive feedstock  into  the  high-intensity



          arc.




     (2)  Elimination of  the tungsten electrode  as  one  side  of  the arc could




          be accomplished by utilizing two wire  torches  and  establishing an




          arc  between the two consumable wire  "electrodes."

-------
     (3)  If agglomerating effects prove undesirable for a particular study,


          an Ion generator or polonium grid could be employed to neutralize


          the charged particles.


     (4)  Optimization of main operating parameters - arc voltage, wire feed


          rate, and wire diameter - could significantly improve the feed-


          stock-to^particulate conversion ratio of 4.75 percent.




     Naylor concluded that for at least one operating condition, the generator


can produce very fine agglomerated zinc oxide particles that are reproduc-


ible with respect to both the distribution of the basic particle diameters


and mass volumetric concentration.  Therefore, this metallizer is considered


worthy of continued attention by researchers as a device for generating very


fine particles.

                                                                       3
     The generation of mass concentration ranging from 0.67 - 2.05 g/sm


(1.5 - 4.7 grain/scf) was reproducibile at three of four operating conditions.


The mass volumetric concentration was sensitive to changes in the operating


conditions.  The mass emission rate was 4.7 - 14.5 gram/minute.


     Basic particle sizes ranged from 6.0 - 10.8 nm (0.006 - 0.0108 jim).  At


one operating condition, the basic particle diameters' mean and standard


deviation were reproducibile, while at another the basic particle's mean,


standard deviation and cumulative frequency distribution diameter was repro-


ducibile.  Comparatively, the mean particle diameters from these two operating


conditions were significantly different.
                \

     Virtually all of the basic particles were agglomerated into chains or


clusters of various sorts.  The agglomeration initiated immediately after


basic particle generation at the electric arc.  Its fast rate undoubtedly

-------
resulted from the high initial  number  of  basic  particles per volume,   their




very fine particle size,  sonic  agglomeration  from the noisy arcing process,




and thermal turbulence of the gas   stream near  the arc.




     At the one condition for which mass  concentration was not reproducible




the mean basic particle diameter was and  is not considered to  be reproducible.




     In summary, Naylor recommended:




     (1)  Improve the research  tool.




          a)  Further development  of equipment  to improve operational




              quality.



          b)  Investigation of  other metals and expansion of generating




              and sampling program.




     (2)  Further study of the  agglomeration  of very fine particles.




          a)  Define appropriate  dimensions for describing the agglomerates.




          b)  Based on these dimensions,  determine whether or  not the metallizer




              generates predictably reproducible distributions of sizes.




          c)  If predictable reproducibility  is established determine if  there




              is a significant  correlation between agglomerate size  and such




              variables as chamber-duct retention time,  number of very fine




              basic particles per  unit volume,  and diameters of the  very




              fine particles.




          d)  Related to "c" above, determine if there  is a  significant




              correlation between  very fine basic particle diameter, agglom-




              erate size and mass  volumetric  concentration.




     (3)  Use the research tool; for example, modify various  experimental




          collection equipment  designs to achieve and  improve  the effective-




          ness of small particle control.





                                     8

-------
     Dimmick concluded that the electric arc generator produced an aerosol


of agglomerated iron oxide fine particles in mass volumetric concentrations
                  3
averaging 1.4 gm/m  and mass emission rates averaging 7.8 grams per minute.


The generator showed some controllability of mass volumetric concentration

and variability of very fine basic particle populations with respect to some


operating variables.  The analysis of thei production of the mass volumetric

concentration of the aerosol showed the mass volumetric concentration to be

dependent on the wire feed rate and the open voltage across the electrodes.


     The iron oxide fine basic particles,which composed the agglomerated


particles having an estimated size of 0.4 um, have been characterized to

have means ranging from 5.4 nm to 9.1 nm (0.0054 pm to 0.0091 ym).  Differences

in size means and size distributions between sets of operating conditions


indicated that changing the operation variables changed the size character-

istics of the pppulation of fine particles.

     A simple comparative test for discerning the effect, if any, of changing


the exhausting rate of the arc chamber on the aerosol's particle size was


inconclusive.  The agglomerated particle size was found not to change with an

increase in exhausting rate.  However, the very fine basic particle mean size


was found to be greatly affected by the increase in exhausting rate.  This


change in generating properties indicated a need for further experimentation.


     Dimmick recommended:

     (1)  Further investigation of the evidence found in the simple compara'-


          tive test.

     (2)  The compressed air-jet that atomizes and quenches the melted and


                                      9

-------
          vaproized metal should be studied to  ascertain its  effect  on




          the aerosol's particle size mean and  distribution.   The  open vol-




          tage would require very precise control if  its effect  on an aerosol's




          particle mean size or size distribution were to be  investigated;  a




          more appropriate power supply is needed.




      (3)  In the application of this aerosol generator,  an on-line con-




          tinuous mean size and size distribution analyzer, such as  a Whitby-




          Liu mobility analyzer, would provide  the necessary  "before control




          equipment" and "after control equipment" data  in an engineering




          evaluation of a piece of air pollution control equipment.   Better




          control of the separation of the larger particles within the barrel




          could also help approximate real life circumstances.








     The overall conclusion is that electric arc generation of fine  particle




aerosols provides an adequate emission rate at  high enough mass volumetric




concentrations to test fine particle control equipment.   The  fine  particles




are formed by agglomeration from many 2-20 nanometer sized particles with the




metallizer generator.




     An overall recommendation is to provide continual monitoring  of the




aerosol's particle size mean and distribution while further investigating




any effects of the many independent variables present in the  system.  Although




development is still needed, the electric arc generation of fresh fine metallic




oxide particle aerosols provides a generating means for those interested in




research and development of fine particle air pollution control technology.
                                     10

-------
                                 Section C




                          Experimental Procedures






     The two aerosol generators  that were developed by Redden and by Naylor




and Dinmick use the energy  from  a direct current electric arc to heat materials -




steel and zinc -  in the production of fine particle aerosols.  The heat (up




to 7000°K) was provided from an  arc between an electrode and the material to




be melted (direct arc heating) or by an arc between two electrodes (indirect




arc heating).  In both direct and indirect heating, charged electrodes initiate




an intense electric arc when brought into close proximity to each other.  The




result of this heat generating arc is the melting  and vaporizing of the




electrodes - particularly the anode.




     In both generators, a  nozzled gas  (compressed air) jet converged about




the zone of melted and vaporized materials.  This  jet had the effect of atom-




izing the melted  materials  into  molten droplets as well as facilitating the




condensation of the vaporized materials into very  fine particles.  These very




fine particles, being newly formed, were considered to be fresh.




     Simplistically, two products are produced by  the electric arc aerosol




generators : the  solidified molten droplets from the atomizing spray and the




solidified condensate of metallic vapors.  Both generating apparatus incorp-




orated settJing   chambers to separate the much larger particles from the product.




Presumably, the fresh metallic fine particles were not affected by the gravi-




metric and centrifugal separation even though it became evident that the high




                                     11

-------
particle number concentrations and the presence of abundant electric charges




agglomerated the very fine basic particles into larger agglomerated particles -




thereby increasing the effective size of the particles.  This assumption does




not exclude the possibility that discrete, very fine particles remain in the




gas stream.








Apparatus



     Hedden's final apparatus (Figure 1) consisted of an adjustable, constant




voltage direct current power supply, a consumable electrode wire feed assembly




with control panel, an arc chamber, a non-consumable  tungsten electrode with




cooled holder and shielding gas supply system, and a system to connect the




arc chamber to an air mover upstream and a settling chamber and exhaust system




with sampling port downstream.




     The power supply utilized in Hedden's apparatus, conventionally used to




provide power for semiautomatic welding processes such as gas shielded open




air and submerged arc, was a direct current, self contained motor driven




generator unit rated for a 100% duty cycle.  This rating allowed continuous,




uninterrupted production of power under full load.  Once set, the output of




a constant voltage power supply has essentially the same voltage no matter




what the welding current may be.  The voltage could be varied over a range




of approximately 10 - 50 volts.




     The automatic welding head assembly consisted of two main components:




(1)  the wire feed assembly with torch, and (2) remote control panel.  The    ,




wire feed assembly combined an insulated wire supply reel, guides, drive motor




with gear reduction box, and feed rolls, that constantly feed a consumable




electrode wire through the air-cooled torch to the arc.  The remote control





                                     12

-------
u>
                           AIR
                          FLOW
         EXHAUST
                          VACCUM
             POWER SUPPLY
                                                                        WIRE REEL
£
                                                           WIRE TORCH

                                                         ARC CHAMBER
                                TUNGSTEN
                                ELECTRODE
                                HOLDER
                                                                 SHIELDING
                                                                       GAS
                                                                    SUPPLY


                                                                    AIR
                                                                    MOVER
SHIELDING
GAS NOZZLE
                                                                                             COOLING
                                                                                               WATER
                                                                                                TANK
                            Figure 1. Simplified diagram of aerosol generator
                                       apparatus (Hedden)

-------
panel was electrically connected to both the welding machine and the non-




consumable electrode holder cooling water supply through solenoid valves.




     The arc chamber was fabircated from a 33 centimeter length of 12.7




centimeter (5 inches) diameter steel pipe.  The attachments between the arc




chamber and the wire feed assembly were designed to allow adjustment in all




axes to obtain perfect alignment between the feed wire electrode and the




non-consumable electrode.  A shielding gas nozzle and supply system provided




for displacing the air surrounding the non-consumable electrode thus preventing




rapid consumption of the tungsten by oxidation .




     Air flow through the system was provided by a "Gelman Hurricane Air




Sampler" set on low speed.  The ductwork leading from the arc chamber incor-




porated a settling chamber and elutriation column to separate the molten




globules of steel from the gas stream.  The duct leaving the elutriation




column made a 180 degree turn before passing the sampling port and nozzle.




An adequate length of straight duct was established to allow optimum position-




ing of the sample probe - a minimum of 10 duct diameters downstream and 5




duct diameters upstream from any disturbance to air flow.




     The Naylor and Dimmick aerosol generating apparatus consisted of an




electric arc  two consumable electrode metallizer that incorporated an elec-




tric arc and a compressed air jet to spray melted and vaporized consumable




wire electrodes, a 55 gallon oil drum that was used as the arc chamber and




was connected to a supplemental air supply cleaned by a HEPA  (High Efficiency




Mr Particulate) filter, a draft fan with exhausting ductwork from the oil




drum, and a cloth filter particulate collector.  (For Naylor's see Figure 2,




for Dimmick's, Figure 3)  The .metallizer system supplied the necessary power




and wire feed delivery.   The ductwork and fan provided for exhausting and





                                     14

-------
                                     Phenolic spray head Cross-section

                                  I           E
                                                           lid of
                                                            barrel    i
A) Power  source of metallizer
B) Control console
C) Dual  wire spools
D) Wire  straightening, drive and  conduit  feed mechanism
E) Flexible conduit for wire
F) Flexible compressed air line - main air
G) High amperage  electrical cable
H) 55 gallon barrel
I) Phenolic spray  head
J) Dilution air from Hepa filter
K) Flexible compressed air line -  secondary air
L) Exhaust duct -  6  inch diameter
M) Sampling  port
N) Venturi
O) Stack sampler
                  Figure 2. Metallizer equipment used
                             by Naylor
                                 15

-------
Wire
dispenser
               Secondary
      Arc gun  air
     Secondary air
     connection
                                        Electrode tip

                                             gas jet


                                        •Electrode wire
                                       Lid
      Figure 3.  Sfiroulded Read metallizer equipment
                 used by Dlmmick
                      16

-------
sampling the generated aerosol.  The oversized baghouse cleaned the particle




laden gas stream of most of the particulates before returning the process air



to the atmosphere.




     The metallizer of the Naylor and Dimmick apparatus is used commercially




to provide coatings on metal or to fill in or build up worn areas on shafts,




etc.  It consisted of a non-constant voltage power source which converts 3




phase 220 volt alternating current to a direct current that is variable from




0 to 40 volts, up to about 400 amperes.  A console was located on top of the




power source from which air pressure to spray gas jet, consumable wire delivery




rate, and arc open-circuit voltage are adjusted.  Flexible conduits conveyed




the wires  from reels passing  through the console to a phenolic spray head




that guided the wires into fixed position electrode tips.




     The spray head was fastened to the lid of a 55 gallon barrel  with the




electrodes extending through a slot into the barrel.  The arcing takes




place in the barrel.  The larger particles of the metallized wire are directed




by the high pressure air jet to the other end of the barrel where they are




deposited.  The dispersed air  jet stream and a low pressure air stream enter-




ing through a HEPA filter conveyed the smaller particles through an exhaust




duct past the sampling port to a cloth filter collector and exhaust blower.









Sampling



     Hedden's system was allowed to operate for approximately one minute to




establish steady-state conditions for sampling with steel welding wire.  During




this period the temperature .of the exhaust gas stream stabilized producing




a constant velocity past the sampling nozzle.  A ten minute sampling period




then commenced.





                                     17

-------
     The sampling train consisted of a one-half inch (12.7 mm)  stainless steel

hook nozzle attached to an Andersen Stack Sampler, in series with an aluminum

Millipore filter holder supporting a 47 mm diameter  Gelman type GA-4 membrane

filter to separate the fume from the sample gas stream.

     A calibrated orifice was used to indicate instantaneous sample flow rate,

with total flow rate recorded on a dry gas meter.

     The fine particle aerosol mass concentration was determined gravimetri-

cally.  Also, each sample of collected fume; was resuspettded in 2 ml of a 1%

nitrocellulose solution and dispersed by placing the test tube in an ultra-?

sonic cleaner; the fume collected on the membrane filter was carefully

scraped from the membrane and placed directly in 2 ml of the nitrocellulose

solution and dispersed.  A disposable pipette was used to place one drpp of
                                                                       i
the solution on an uncoated electron microscope grid.

     The metallizer generator was operated for at least two minutes before

samples were taken either for mass conversion analysis or particle size dis-

tribution analysis.  Randomly ordered samples were taken with a shut-off

between each sampling run.  Naylor took samples using zinc metallizing wire

and Dimmick took samples using steel metallizing and welding wire.

     Sampling for mass concentration data consisted of pre-trial determinations

of isokinetic sampling conditions and then the actual sampling.  After adjust-

ing the stack sampler for isokinetic sampling, samples were taken such that

at least 10 mg of mass was deposited on the tared filter.

     Sampling the aerosol for subsequent particle size distribution and

analysis consisted of estimating isokinetic conditions and then collection

on appropriate membrane filters with sampling times of three, five and seven

seconds.   Then, the filters were enclosed in a plastic filter holder  and


                                     18

-------
were shipped by United Postal Service to McCrone Associates, Inc. in Chicago




for analysis by scanning electron microscopy  (SEM), transmission electron




microscopy  (TEM), and selected  area electron  diffraction SAED).









Analytic Techniques




     Hedden used gravimetric and flow measurement  techniques to determine a




aerosol mass concentration.  Hedden had electron micrograph from an RCA




electron microscope  at West Virginia University to measure particle diameter




counts with the use  of a semi-circle graticule.                         '




     Naylor and Dimmick used gravimetric and  flow  measurement with an RAC




Stak sampler to determine  aerosol mass concentration.  Naylor  and Dimmick




both recieved  fine basic particle measurements made by McCrone Associates




using TEM  and  SEM.   McCrone Associates also performed chemical identification




using SAED.  Naylor  characterized the agglomerated particles from McCrone




and Dimmick measured the agglomerated particles using the enlarged SEM




photographs.









Experimental Designs




     Hedden investigated the performance characterisitcs of the aerosol




generator  under semi-continuous operation.  A series of three  controlled




test runs  were made  and a  portion of the resultant fume was collected.  All




tests were conducted under identical conditions of arc voltage, wire feed




rate, shielding  gas flow  rate, and dilution  air flow rate.  The test program




was structured to determine the mass rate  of  production of iron oxide fume,




the particle size (not cecognized at first) distribution of that fume, and




the degree of  reproducibility of production rate and size distribution.





                                     19

-------
     Naylor investigated the reproducibility of the metallizer aerosol




generator using zinc metallizing wire.   He used four settings with the




generator to obtain data for particle size measurement and mass volumetric




concentration determination. Naylor varied the consumable wire feed rate, open




circuit voltage and the gas jet pressure.




     Dimmick characterized the aerosols' particle size under four different




operating conditions using steel welding wire.  He also performed a simple




comparative experiment investigating the effect of increasing the exhaust flow




within the ductwork.  Dimmick designed an experiment to investigate the




effect of generator variables - wire feed rate, open circuit voltage, and




gas jet pressure - on the aerosols' mass volumetric concentration.
                                     20

-------
                                 Section D




                           Results and Discussion









Mass Volumetric Concentration and Emission Rate (Hedden)




     With the mass of particulate in the sample gas stream determined gravi-




metrically, the average mass volumetric concentration of iron oxide fume in




the sample gas stream was 0.292 grains/scf (668.1 mg/m ).  Assuming that the




mass volumetric concentration of particulate matter in the total gas stream




is the same as for the sample gas stream, the average total particulate pro-




duction rate was calculated to be 30.08 grains/min (117.0 grams/hr).  Total




particulate emission rate, given in common units of measurement, is listed




in Table 1.




     Another performance factor directly related to the rate of production




is the conversion ratio of consumable wire feed stock to very fine particles.




The wire consumption rate was 5.43 Ib/hr (2463 gram/hr).  The average partic-




ulate production rate was 0.258 Ib/hr  (117.0 g/hr).  Therefore,  5.43 pounds




(2463 g) of wire was required to produce 0.258 pounds (117.0 g) of very fine




particles for a conversion ratio of 4.75 percent.




Particle Sizing (Hedden)




     Two methods were employed, with differing results, to determine particle




size distribution: (1)  an electron microscope and (2) an Andersen  Stack




Sampler.



     Because of the high magnification,  and consequently small field,  of  the




electron microscope, photographic plates having a final magnification of




87,600 times were made of a minimum of  three fields of each  grid.   Individual




particles of iron oxide were spherical; and individual spherical particles





                                     21

-------
          TABLE 1. Summary of mass volumetric  concentration and emission rate with approximate range
                                 of agglomerated  particle size indicated(Redden)
Approximate
range of
particle
STAGE size (urn)
0 > 30
1 9.2-30
2 5.5-9.2
3 3.3-5.5
4 2.0-3.3
5 1.0-2.0
6 0.3-1.0
7 0.1-0.3
8 < 0.1
Filter < 0.1
(1) TOTALS
RUN
Collected
weight
(mg)
2.5
0.4
0.4
0.8
1.3
2.7
8.1
16.4
9.0
91.6
1
Cumulative
percent
total
100.10
98.13
97.83
97.53
96.93
95.95
93.92
87.84
75.53
68.77
133.2
RUN
Collected
weight
(mg)
1.5
0.3
0.2
0.2
0.7
3.1
7.7
11.2
12.8
86.1
2
Cumulative
percent
total
100.00
98.79
98.55
98.39
98.23
97.66
95.16
88.94
79.89
69.55
123.8

(2) Sample volume
(scf)
(3) Mass volumetric
concentr. (mg/scf)
(4) Volumetric flow
rate (scfm)
(5) Emission rate
( grams /min)

7.23

18.42

103

1.90









7.21

17.17

103

1.77








RUN
Collected
weight
(mg)
2.4
0.5
0.4
0.5
1.5
3.3
8.4
17.4
13.4
103.4
3
Cumulative
percent
total
100.00
98.42
98.09
97.83
97.50
96.51
94.34
88.82
77.38
68.57
152.1


7.17

21.21

103

2.18








OVERALL
Mean
weight
(mg)
2.13
0.4
0.33
0.5
1.17
3.03
8.07
15.0
11.73
93.7
136.0
Standard
deviation
(mg)
0.50
0.08
0.09
0.25
0.34
0.25
0.29
2.72
1.95
7.22
14.42


7.20

18.89

103

1.95

-

2.1

-

0.17
NS
NJ

-------
composing hirger agglomerates were easily distinguished.  The agglomerates




wcro not. MM.slgned a composite or equivalent particle size but the basic




pnrtlcU'8 composing an agglomerate were classified individually.




     With a total of 12,355 basic particles sized the raw data was grouped into




10 classes with a basic particle size interval of 0.01 micrometers each to




simplify calculation of various statistical values.  Three basic particles




exceeded 0.2 micrometers at 0.22, 0.25 and 0.24 micrometers.  A simplified




method for calculating the arithmetic mean and standard deviation of the




basic particle size distribution was employed and relative particle size




frequency was presented in Table 2.




     The arithmetic mean size of the basic particles was found to be 0.022




micrometer with a standard deviation of 0.0163 micrometers.  Using log-




normal probability paper revealed an indication that the basic particle size




distribution was log normal.  The median, sometime known as the mass median




diameter  (HMD) divides the frequency distribution  (by mass) in half and was




found to 0.0167 micrometer.




     The second method involved the relatively new Andersen Stack Sampler




used in conjunction with a back-up membrane filter.  Combining the mass distri-




bution data with calibration information  supplied by the manufacturer,




cumulative particle  (agglomerated) size mass distributions can be obtained




and were presented in Table 1.  Although  this mass distribution data indicates




approximately 69 percent of the sampled particulate passed through the




Andersen Stack Sampler without collection, the previous data presented con-




cerning the size of the discrete basic particles suggests that essentially




all the particulate matter should have passed through the sampler without




impacting on the collection plates.



                                      23

-------
Table  2.   Relative basic particle size frequency distribution in percent(Hedden)
STAGE
0
1
2
3
4
5
6
7
8
Filter
Mean
Standard
Deviation
Range
Particle Size, micrometers
0.005-
0.015
46.3
43.0
54.3
50.5
43.0
46.3
47.4
44.3
48.0
42.8
46.6
3.73
11.3
0.015-
0.025
22.0
22.4
19.1
19.2
23.3
22.3
20.5
21.2
21.6
22.1
21.4
1.38
4.2
0.025-
0.035
13.6
17.7
12.3
14.4
17.1
16.4
17.3
15.1
14.9
15.7
15.5
1.72
5.4
0.035-
0.045
7.7
6.9
7.0
6.6
8.7
5.3
6.6
8.8
6.3
8.3
7.2
1.15
3.5
0.045-
0.055
4.5
4.5
3.5
4.2
3.1
3.7
3.4
5.1
6.2
4.5
4.3
0.92
3.1
0.055-
0.065
2.1
2.6
2.1
2.2
2.2
2.6
2.7
2.1
1.3
3.6
2.4
0.59
2.3
0.065-
0.075
1.5
1.5
0.9
1.1
0.7
2.3
0.7
1.1
0.6
1.4
1.2
0.52
1.7
0.075-
0.085
1.3
0.4
0.2
0.8
0.6
0.3
0.3
1.0
0.4
0.8
0.6
0.36
1.1
0.085-
0.095
0.1
0.2
0.1
0.6
0.3
0.2
0.4
0.2
0.3
0.4
0.3
0.16
0.5
0.095
and>
0.8
0.9
0.4
0.1
0.9
0.6
0.7
0.9
0.4
0.4
0.6
0.27
0.8

99.9
100.1
99.9
99.7
99.9
100.0
100.0
99.8
100.0
100.0
100.0


-------
     Two methods were used to estimate the reproducibility of the relative

particle size distributions	both using the mass distribuiton data obtained

from the Andersen Stack Sampler and  the back up membrane filter.  The results

of the three runs are presented in Table 1.  The arithmetic mean and standard

deviation for the weights found for  each of the three runs on each stage of

the sampler and membrane filter   are shown.  The second estimate of agglomerated

size distribution reproducibility was obtained by calculating the arithmetic

mean and standard deviation of the cumulative relative mass distribution for

each of the various size intervals as shown in Table 2.



Mass Volumetric Concentration and Electron Diffraction (Naylor)

     Naylor stated that the mass  concentration results were more meaningful

when the statistical parameters of mean, standard deviation, and coefficient

of variation were  calculated.  The coefficient of variation is one indicator

of reproducibility or precision of mass volumetric  concentration generation.

These parameters are summarized in Table 3.
                   Table 3.   Zinc oxide mass  concentration
                             mean and variation summary  (Naylor)
             Condition
       (1) Mean (g/sm3)                   1.03      0.68      2.05       1.12

       (2) Stand.  Dev.  (g/sm3)            0.20      0.05      0.14       0.10

       (3) Coefficient 100x(2)/(l)         19%      8.4%      6.9%       8.7%
           of  variation
                                      25

-------
     Naylor tested the null hypothesis that the mean mass concentrations

were equal using standard statisitcal methods.  These hypothesis tests

support the conclusion at the 5% level of significance that the mean of

Condition 3 was greater than the mean of Condition 1, and that the mean of

Condition 1 was greater than the mean of Condition 2.

     Selected area electron diffraction results, as received from McCrone

Associates, were compared to standard electron diffraction information (ASTM)

and were  found to indicate that the particles generated were ZnO.



Particle Sizing (Naylor)

     The electron microscopy photographs of the zinc oxide samples were all

similar in terms of particle appearance.  Particle size means and variances

were presented in Table 4.  SEM photographs show agglomerates of different

shapes and sizes.   The agglomerates range in size from less than 0.1 vim to

over 2 pm and appear to be interconnected.  The shapes vary from spherical

to chainlike.  These agglomerates are composed of very fine particles as

shown in the TEM photos of sample 101.
             Table 4.  Means and variances of zinc oxide basic
                       particle size data (Naylor)
     Sample #

     Condition II

         5 (nm)

         S2 (nm2)   14.6

         S (nm)

     Sample Size
101
1
6.0
4.6
3.8
297
125
1
10.2
29.2
5.4
280
133
2
9.1
35.6
6.0
350
134
2
10.0
41.1
6.4
280
104
3
10.8
28.1
5.3
376
128
3
10.4
22.0
4.7
281
                                     26

-------
     The basic particles mean diameters for samples 101 (D,0,) and 125 (D,25)





of condition 1 were named   significantly different by inspection.  The




values of the pairs D    and D     for Condition 2, and D  , and D  . for con-
dition 3 were analyzed by a statistical method for significant difference.




This analysis indicated that the  two replicates for Conditon 2 have diameter




means and variances  that were not significantly different.  The analysis also




indicated that  the means and variances for Condition 3 were not significantly



different.




     An important question is whether the means and variances for basic




particle diameter populations generated under different conditions are




different.   Since the mean of Condition 1 cannot be estimated within a rela-




tively tight range,  only the relationship of Condition 2 to Condition 3 is




examined.  Hypothesis testing demonstrated that the mean basic particle




diameters for Conditions 2 and  3  are significantly different.  Thus changing




the operating conditions apparently changed the mean basic particle diameters.








Mass Volumetric Concentration and Electron Diffraction (Dimmick)



     The Latin  Square analysis (Table. 5)  found  the wire feed rate  to be sig-




nificant at  the 0.01 significance level.  This implied the wire feed rate




had an effect on the resultant  mass volumetric concentration.  The variables




of open voltage and  jet pressure  were not found to have any effect on the mass




volumetric concentration of the aerosol even at the 0.1 significance level.




The analysis of variance statistics for replication was also calculated and




found to indicate a  significant difference between trials at the  0.05 signi-




ficance level but not the 0.01  significance level.
                                     27

-------
     Table 5.  Analysis of variance for Latin Square experimental design
               mass volumetric concentration data (Dimmlck)
Variable
Wire feed
rate
Open Voltage
Main air- jet
pressure
Replication
Experimental
error
Total
df
3
3
3
1
21
31
Sum of
Squares
1.473
0.319
0.106
0.335
0.952
3.185
Mean F , F rt ... F n ni
cal a=0.05 a=0.01
Square
0.4908 10.86 3.07 4.87
0.1065 2.35 3.07 4.87
0.0352 0.77 3.07 4.87
0.3349 7.39 4.32 8.02
0.0453

     The SAS GLM analysis of the combination of the Latin Square and another

set of data found the results to significantly describe the variability of

all the data.  With this total set of 40 observations, a statistical procedure

calculated F statistics for determining the effect or noneffect of the indepen-

dent variables in the experimental model.  The procedure found the variables

of wire feed rate, open circuit voltage, and sample replication to have a

significant effect on the outcome:  mass volumetric concentration of the gas

stream.

     The  SAED analysis showed two possible resultant materials.  Two different

phases of iron oxide were possible:  magnetite and maghemite.

Particle Sizing (Dimmick)

     The size means of the agglomerates (Table 6) were found to be 0.41 ym

(S=0.53 jjm)  for sample 715 and 0.40 ym (S=0.54 ym) for sample 714.  These


                                     28

-------
means were also found to be equal; this implied that they were from the same




population.  The coefficient of variances were rather large, thus the agglom-




erated particle distributions representative of the aerosols were not "mono-



disperse."
                6.  Statistics for agglomerated samples  (Dimmick)
pie No.
714
715
Mean
(ym)
0.40
0.41
Standard
Deviation
(ym)
0.54
0.53
Standard Error
of the Mean (ym)
0.03
0.03
Coefficie:
Variance
135
129
      The  simple  comparative  analysis  of  increasing the duct exhausting




 velocity  from 1300 fpm to  1850  fpm showed  only  a decrease in the agglomerated




 particle  size from 0.41 ym to 0.40 ym.   This  difference was not significant




 at a 0.1  level of significnace.



     Fine basic particle statistics were presented in Table 7.   Statistical




 tests indicated a variety of  results.  The within group t-test analysis




 showed that the sample means  in all the sets were equivalent within groups.




 The between group  t-test analysis  showed a variety of equal and unequal




 means.  The within group very fine basic particle size distributions for sets




 2, 3, and 4 were found  to be  equivalent.   Size distributions between groups




 were found to be all different at  an  alpha level of 0.05.  This implied that




 at three of four different operating  conditions three different fine basic




 particle size distributions were generated.  The t-test analysis between the




 two comparative samples 712 and 716 showed sample 712 to have a significantly




                                      29

-------
Table  7.  Statistics for very fine basic par tides (Dimmick)
Set Number
1
2
3
4
5
6
Sample No .
600
601
602
603
604
605
607
611
609
612
716
712
Number of
Particles
386
446
434
369
368
356
296
274
298
297
449
497
Mean
(nm)
8.00
7.83
7.53
8.45
8.32
8.31
9.05
8.57
7.86
8.13
8.05
6.42
Standard
Deviation
(nm)
5.35
4.56
5.23
5.37
5.51
5.41
5.49
5.40
5.40
5.37
5.21
4.79
Standard of Error
of the Mean (nm)
0.27
0.22
0.25
0.28
0.29
0.29
0.32
0.33
0.31
0.31
0.25
0.21
Coefficient of
Variance (%)
66.9
58.2
69.5
63.5
66.3
65.1
60.5
62.9
68.8
66.0
64.7
74.6

-------
smaller very fine basic particle size than sample 716.




     An overall characterization of the fine particle generators is presented


in Table 8.  This table list basic particle size statistics for the basic
              Table 8.  Hedden-Naylor-Dimmick summary of results
Limit of microscope resolution  (nm)


Width of diameter count  interval  (nm)


Mean basic particle diameter  (nm)


Range of basic diameters  (nm)


Standard deviation  (nm)


Cumulative Frequency Distribution

                                   3
Mass volumetric concentration  (gm/m )

                                   3
Mass volumetric concentration  (gm/m )


Mass emission rate  (gm/min)


Percent conversion  (%)
                                            Redden         Naylor       Dimmick
                                           (iron oxide)  (zinc oxide)  (iron oxide)
5
10
22
5->100
16.3
log normal
0.67
.61-0.75
1.95
4.8
1
1
6.0-10.8
1-50
3.8-6.4
normal
1.21
.67-2.05
8.57
	
1
1
6.4-9.1
1-720
4.7-5.5


1.38
0.81-2.16
7.73
8-15
particles - not the agglomerated particles.  Redden did not directly measure


what may be presumed to be the aerosol; that is, he measured agglomerated


particles by inference from the Andersen Stack Sampler and not directly


from micrographs.  Naylor indicated a range of sizes for the agglomerated


particles from less than 0.1 ym to greater than 2 ym.  Dimmick measured the


agglomerated particles and found an average size of 0.4 ym.  The table also


shows various production characteristics related to the efficiency of mass
                                      31

-------
conversion from wire feed to particle aerosol.




     A question that should be opened for discussion and research is centered




around the physical explanation of the electric arc generation of fine particle




aerosols.  What are the relations and factors involved in the dichotomous




formation of solidified molten droplets and agglomerated solid fine particles?




Also, what effects on the chemical processes involved in the aerosol production




does the dichotomous formation have?
                                     32

-------
             ELECTRIC ARC GENERATION
           OF POLYDISPERSED IRON OXIDE
            AEROSOL IN AN AIR STREAM
                    PROBLEM
       Submitted to the Graduate School
                     of
           West Virginia University
In Partial Fullfillment of the Requirements for
the Degree of Master of Science in Engineering
                     by

           Robert E. Hedden, B.S.
                  Morgantown
                 West Virginia
                     1972

-------
                             ABSTRACT






   Freshly formed metal and metal oxide particles in the sub-micron




size range are known to have different behavior characteristics




than those of collected, classified, aged, and sedispersed participates.




In the past, significant quantities of freshly formed ultrafine particles




have not been available for research work in air pollution control




techniques, instrument calibration, emission simulation, animal




inhalation or particle behavior studies.  This has caused considerable




difficulty in translating laboratory research to field application.




   An experimental aerosol generator that supplies reproducible




amounts of spherical, solid, fresh particles of metals and metal




oxides directly in a hot or cooled gas stream is described in this




report.  The generator employs a direct current power source to create




an arc between a relatively non-consumable tungsten cathode and a




consumable feed-stock of wire.  The feed-stock is vaporized in the high




current density of the electric arc and particles are formed on cooling




by a gas stream moving past the arc area.




   •Rate of continuous production of 0.1 micrometer diameter fresh iron




oxide particles exceeds 100 grams/hour.  The moving gas stream was




sampled using a high temperature Andersen Stack Sampler probe followed




by a membrane filter to obtain mass concentration.  The size, shape




and particle size distribution of the particles were determined using




an electron microscope.




   Other operating parameters, performance characteristics, and




development background are discussed.

-------
                         ACKNOWLEDGEMENTS






   The author IB grateful to all those who have contributed to the




successful preparation of this report.  Special appreciation is




extended to his graduate advisor, Professor Benjamin Linsky, for his




sincere interest in this study; and to Frank Noonan, then Director of




Air Pollution Control Engineering Laboratories at West Virginia




University, for his laboratory assistance and constructive criticism.




Gratitude is also expressed to Dr. Howard W. Butler, Chairman of




Mechanical Engineering, and Hasin T. Gencsoy, Professor of Mechanical




Engineering, for making special equipment and facilities of their




department available for this study.  The cooperativeness of




Don Garletts and Harold Martin of the Mechaincal Engineering




Laboratory, and Fran Culler of the Civil Engineering Laboratory, during




the mechanical design and fabrication of the apparatus was sincerely




appreciated.




   Special thanks is extended to Dr. M. R. Friedman of the




Anatomy Department, West Virginia University School of Medicine,




under whose guidance electromicroscopist Betsy Walker obtained




numerous electronmicrographs of the material produced in this




experiment.  Further thanks is expressed to J. Reginald Dietz,




Director of Research and Development for National Steel Corporation,




Weirton, West Virginia, under whose direction R. E. Brien, Senior




Research Metallurgist, prepared electronmicrographs of material




collected from preliminary experiments.

-------
   The author also appreciated the encouragement  and assistance




given by Benjamin Euseblo and Bruce Harris,  of what was then the




National Air Pollution Control Administration, Cincinnati,  Ohio.




   Several training grants,  5 T01 APOOOO 9-07 and 5 T01 APOOOO 9-08,




from what is now the Air and Water Programs  Division of the




U. S. Environmental Protection Agency made this study possible.  The




author is further indebted to the U. S.  Bureau of Mines, Morgantown,




West Virginia, for supplying the helium used in these experiments




and to Two Thousand, Inc. for the use of the Andersen Stack Sampler.
                               1-iv

-------
                           TABLE OF CONTENTS

                                                                       Page

      ABSTRACT	l_ii

      ACKNOWLEDGEMENTS 	  1-iii

      LIST OF TABLES	1-vi

      LIST OF FIGURES	1-vii

I.    INTRODUCTION	1-1

II.   LITERATURE REVIEW	1-5

III.  APPARATUS	1-10

IV.   EXPERIMENTAL	1-25

        SAMPLING PROCEDURE 	  1-25

        RESULTS	  1-29

        Mass Concentration and Rate of Production	1-29
        Particle Size Distribution 	  1-34
        Reproducibility of Production Rate and Relative
          Particle Size Distribution 	  1-49

V.    CONCLUSIONS	1-52

VI.   LIST OF REFERENCES	1-55

      APPENDICES	1-58

        APPENDIX A.  Selected Bibliography of Readings Concerning
                       Electric Arc Discharges 	  1-A-l

        APPENDIX B.  Developmental Background of the Tungsten
                       Electrode Holder  	  1-B-l

        APPENDIX C.  Operating Procedure for Arc Aerosol
                       Generator	1-C-l

        APPENDIX D.  Electron Microscope Particle Count Raw Data .  .  .  1-D-l

        APPENDIX E.  Statistical Calculations  	  1-E-l


                                  1-v

-------
                             LIST OF TABLES

                                                                      Page


Table   1.  Operating and sampling test conditions                    1-27

Table   2.  Weight (mass) distribution of particulate matter
              retained by the Andersen Stack Sampler and
              backup membrane filter                                  1-30

Table   3.  Mass concentration and total production rate of
              iron oxide particles                                    1-31

Table   4.  Electron microscope particle count data                   1-36

Table   5.  Relative particle size frequency distribution in
              percent from the data of Table 4                        1-37

Table   6.  Cumulative relative particle size distribution
              from data of Table 5                                    1-43

Table   7.  Arithmetic mean and standard deviation of the
              cumulative relative mass distribution
              from data of Table 2                                    1-46

Table   8.  Arithmetic mean and standard deviation of the weight
              (mass) distribution from data of Table 2                1-51

Table D-l.  Electron microscope particle count raw data               1-D-l

Table E-l.  Transformation for calculating the arithmetic mean
              and standard deviation from data of Table 4             1-E-l

Table E-2.  Tabulation for calculating median and geometric
              mean from data of Table 4                               l-E-2

Table E-3.  Calculation of arithmetic mean and standard
              deviation of the relative frequency within each
              particle size interval from the data of Table 5         l-E-3

Table E-4.  Calculation of arithmetic mean and standard
              deviation of the cumulative relative mass
              distribution from data of Table 7                       l-E-7

Table E-5.  Calculation of arithmetic mean and standard
              deviation of weight (mass) distribution from
              data of Table 8                                         l-E-8

                                  1-vi

-------
                            LIST OF FIGURES

                                                                      Page

Figure   1.  Sketch of the components of a direct current arc .... 1-3

Figure   2.  Graphical representation of thermal plasma in
               the arc region	1-3

Figure   3.  Simplified diagram of aerosol generator apparatus  .  .  . 1-11

Figure   4.  Photographs showing assembled and exploded view
               of the non-consumable electrode holder 	 1-15

Figure   5.  Photographs showing general arrangement of arc
               chamber apparatus  	 1-19

Figure   6.  Simplified pictorial sketch of the arc chamber
               and ductwork system	1-23

Figure   7.  Histogram representing particle size data
               from Table 5	1-38

Figure   8.  Distribution curve representing data from Table 5  ... 1-38

Figure   9.  Ogive representing particle size data from Table 6 ... 1-39

Figure  10.  Cumulative distribution data from Table 6 plotted
               using normal probability scales  	 1-39

Figure  11.  Cumulative particle size distribution data from
               Table 6 plotted using log-normal
               probability scales	V.  .  . 1-40

Figure  12.  Cumulative particle size distribution data from
               Table 7 plotted using log-normal
               probability scales 	 1-47

Figure B-l.  Sketches showing developmental configurations of the
               tungsten electrode-arc chamber apparatus 	 l-B-3

Figure C-l.  Simplified cross-sectional, plan view sketch showing
               location of components inside arc chamber  	 l-C-3

Figure C-2.  Calibration curve of wire speed dial versus wire speed
               of 0.045 inch diameter wire	l-C-5
                                  1-vii

-------
                        I.  INTRODUCTION








   The generation of significant quantities of solid aerosols with




reproducibility as to mass, shape and particle size distribution is




an essential  sine gua non for research in air pollution control




techniques, instrument calibration, emission simulation, animal inhalation




and particle behavior studies, and in related fields of research (1,2,3,4).




Therefore, the gas-solid phase system of a test aerosol is usually made




up of components to simulate an actual natural or industrial situation.




   Redispersion of collected, aged, and classified powders, the largest




present source of significant quantities of solid aerosols, requires




large energy inputs to overcome adhesive and agglomerating forces inherent




In bulk powders (5).  Even then, experience has shown that laboratory




redispersions of particles exhibit different behavior characteristics




than freshly formed particles (6,7).  Significant quantities of freshly




formed particles, synthesized in the same manner that they are produced




industrially, have not been available for research work.  These factors




have caused specific difficulties in translating laboratory control




engineering research and medical research to field applications and




diagnosis.




   One industrial process that is adaptable to the primary formation of




solid aerosols in the laboratory is the heating of materials by the




energy from a direct current electric arc.  The heat is provided by an




arc between an electrode and the material to be melted  (direct arc




heating) or by an arc between two electrodes (indirect arc heating).
                                 1-1

-------
Because the source of most of the heat is nonchemical, electric arc
discharge heating is especially desirable in controlling contamination
of the heated material.
   The arc discharge occurs when two electrodes are brought into close
proximity to one another such that a high electrical resistance is
developed at their boundary and the tips begin to glow.  This glow
indicates the emission of electrons from the cathode, which because of
high emission temperatures, up to 7000 K (8,9,10), produces ionization
of the air.  The bombardment of electrons received by the anode causes
it to become white hot and to erode either by melting or vaporization.
(Refer to Figure 1.)
   In the actual direct current arc, the molecules of air dissociate.
They lose some of their orbiting electrons and form a mixture of posi-
tive ions and electrons.  This mixture of electrically charged atoms,
and electrons, with their negative charge, is called thermal plasma
(Figure 2).  As a result, the ionized air becomes conductive to elec-
tricity and current flows without a mechanical connection between
electrodes.  Although a more detailed discussion of what happens in an
electric arc is beyond the scope of this paper, a bibliography has been
included in Appendix A listing specific readings which examine arc dis-
charges .
   In industry, the intense heat produced by electron activity in the
electric arc is utilized for melting and refining of ferrous and nonfer-
rous metals, and for the production of refractories.  As the electric arc
furnace is one of the most difficult to control sources of particulate
emissions in industry, it is an appropriate and adaptable method  for  the
simulation and primary formation of solid aerosols in the laboratory.
                                1-2

-------
    Cathode (-)
Anode (+)
                Direct Current
              Electromotive Force
Figure 1.  Sketch of the Components of a Direct Current Arc
           Free Electrons
  Q.o
                                        o
                                    o
 Figure 2. Graphical Representation of Thermal Plasma
                 in the Arc Region
                  1-3

-------
   Using the technique of direct arc heating, an electric arc aerosol




generator has been designed and built and has undergone limited tests




to evaluate and maximize its performance characteristics and the




properties of the aerosol produced.  The generator appears suitable for




production of particles of metals, oxides of metals, and other refractory




materials, utilizing a constant potential, direct current power supply.




An arc is sustained between a relatively nonconsumable tungsten cathode




and a consumable feedstock of wire.  Commercially available steel




welding wire is the consumable feedstock used in the present investigation.




The wire feed is maintained at a constant rate by a modified automatic




welding head mechanism with a variable speed controller.  A specially




designed, water-cooled holder allows simultaneous manual advancement and




rotation of the tungsten cathode to compensate for and maintain uniform




erosion.  The continuous arc discharge is contained in a 13 inch (33 cm)




length of 5 inch (12.7 cm) standard steel pipe through which passes the




entraining gas stream.
                                1-4

-------
                     II.  LITERATURE REVIEW






   Many methods have been utilized to produce solid aerosols in




significant and reproducible quantities for gas cleaning, animal




inhalation and other aerosol research.  In the past, aerosols of




ultrafine solid particles made from materials to reproduce actual




environments have not been available; this has caused difficulty




in translating laboratory research to practical applications.




   There are two basic  types of aerosols:  (1) aerosols which are




homogeneous with respect to size, shape, and specific gravity, and




(2) heterogeneous aerosols which simulate particles generated in




nature or by an industrial process.  Homogeneous aerosols are used




for fundamental research and product testing and have uniform




properties in order to  minimize the number of variables which must




be considered.  This facilitates characterization of factors involved




in a study and structuring standardized tests for determining accept-




ability of control and  test equipment and working atmospheres.




Although aerosols produced by homogeneous aerosol generators have




many inherent advantages, the output concentrations from these gener-




ators is in the range of 0.1 to 10 milligrams per liter with flow




rates of 1 to 4 liters  per minute; further, dilution ratios of 10 to




100 are required to prevent agglomeration  (11, 12).  Consequently,




these units are not adequate for the evaluation of gas cleaning equip-




ment, where large volumes with high cdncentrations of particulate




matter are required, as the concentration would be too low for realistic




results.



                                1-5

-------
   For most applied and engineering research, heterogeneous aerosols


are used.  Silverman and Billings (13) have reviewed many of the


methods and problems associated with generating solid aerosols in the


laboratory for a variety of materials.  A more complete treatise by


Silverman on the generation of heterogeneous aerosols can be found in


section 12.4.3 of the Air Pollution Handbook (14).


   Of the many methods discussed by Silverman and Billings, only the


redispersion method is capable of producing large quantities of solid


aerosol for large scale laboratory testing.  Large quantities — grains


to grams to tons — are available with a wide variety of physical,


electrical, and chemical characteristics for resuspension in a gas
               /

stream.


   The major difficulty in dry dispersion of particles is limiting


agglomeration.  This limitation can be minimized by employing feeders


utilizing techniques discussed by Olive (15) and by considering


dispersion principles as discussed by Fuchs (16) to overcome bulk


powder adhesive and agglomerating forces.   However, the lower limit


of particle diameter is 0.1 micrometers (17) and almost all aerosols


produced from dry powders seem to be heterogeneous with respect to


particle size.


   The method which most closely approximates an actual industrial


situation is that of direct attrition.  "Fresh" aerosols can be


produced by various mechanical and electro-mechanical means such as


friction, evaporation, impact, explosion and condensation and burning.


Small hammer and ball mills have been employed to produce fine


materials on the order of 1 micrometer diameter.  And a number of


                                1-6

-------
eorapauJes, most notably manufacturers of air filtration equipment,




produce size reduction equipment usable as aerosol generators  (18).




There are many other primary methods reported in the literature but




none have been capable of achieving wide acceptance because constant




and reproducible production rates are very difficult to achieve in




the submicron size range.  At best, it is usually hoped a constant




volumetric particle feed rate can be achieved but the resulting




concentration and particle size distribution must still be determined




by measurement.




   The problems of reproducibility appear to be directly attributable




to two major factors:   (1) the control of the composition of the




feedstock, and  (2) the control of input energy, whether it be mechanical,




chemical, electrical, and so forth.




   In the aerosol generator under present consideration the feedstock




is "common" welding wire available with various metal bases (aluminum,




iron, chromium, copper, and nickel) in a wide range of trace metal




compositions.  Feed rate is controllable by one of a number of




commercially available, constant speed, wire-feed motors.  The energy




input is electrical and is supplied by a constant voltage  (CV) direct




current power supply.  An electrical potential is created between  a




consumable wire feedstock (the anode) and a relatively non-consumable




tungsten electrode (the cathode).  When the anode and cathode  are




brought in close proximity to each other, an intense electric  arc  is




initiated which melts and evaporates a portion of the feedstock




material.  After leaving the vicinity of the electric arc, the




vaporized metal rapidly condenses producing a desirable ultra-fine





                                 1-7

-------
metal fume which can be conveyed by an air stream to sampling equip-
ment, control devices, inhalation subjects, and so forth.
   Although the use of an electric arc has been previously adapted for
production of submicron particulates, vaporization of materials by this
method has been primarily to obtain quantities of submicron particles
as an end'product in themselves.  For example, Holmgren, et. aJ. (19)
used various electric arc systems to generate particulates to be col-
lected for later study of the particles using chemical analysis, x-ray
diffraction, chemical reactivity tests, sinterability tests and electron
microscopy.  This and similar work by Harris, et. al. (20) has been
performed to determine commercial applications and production methods
for particulate materials.  No consideration was given to maintaining
the gas-solid phase system as a viable tool in itself.
   The TAFA Division of the Humphreys Corporation (21) has developed a
commercial carbon arc processing system capable of producing spherical
particles.  However, this system requires preliminary size reduction of
the feed material as the feedstock must be in powder form.  The end
product in this system, once again the particles themselves, is produced
by melting the powder (as compared to vaporization and subsequent con-
densation) whereupon each discrete particle assumes a spherical config-
uration.  It is apparent that the size and mass of the spheroidal
particles produced is directly limited by the size characteristic of
the powder fed.  Also, as opposed to a workable aerosol generator,  the
design emphasis of this unit is on the generation and immediate col-
lection of the particles themselves.  The inclusion of air  or  other gases
is usually undesirable for their purposes.
                                1-8

-------
   In addition to carbon arcs, high voltage condenser discharge through




thin foil and wire (22), and high voltage sparks (23) have been used to




produce metallic aerosols for study in chambers.  Truitt et. al. (24,25),




under several studies for the United States Atomic Energy Commission,




have used two "spark generators" which differ only in the power source.




A Tesla coil was used as a power source for one and the high-frequency




starter attachment on a welding machine supplied the power for the other.




The electric "spark was set up across the axis of a 2-in. diameter glass




pipe" between various metal electrodes.  However, the design of the spark




chamber limited aerosol generation to about one minute because of the




mechanical limitations  of electrode consumption.  To avoid the problems




of non-uniform and non-sustained generation, tanks, plastic bags, and




stainless steel drums were used as holding chambers from which relatively




small volumes, at rates up to 2 liters per minute, were withdrawn.  Not




only are mass concentrations and flow rates low and suitable only for




small scale laboratory  studies, but the problems of maintaining a con-




stant aerosol particle  size distribution by achieving a steady-state




between agglomeration and preferential settling of these larger particles




must be considered.  Even then, the mass concentration of the aerosol




will decrease due to settling.



   In the present apparatus, the undesirable, as well as the desirable,




characteristics of other uses of electric arcs have been integrated




into the prototype of a continuous, steady state, and reproducible source




of relatively large quantities of fresh metallic aerosol.
                                 1-9

-------
                         Ill.  APPARATUS


   The apparatus, diagrammatically sketched in Figure 3, consists of

 an adjustable, constant voltage, direct current power supply, a wire

 feed  assembly with control panel, a non-consumable (tungsten) electrode

 holder with cooling system, an arc chamber, a shielding gas supply

 system,  and a ductwork system connecting the chamber to an air mover

 upstream and a settling chamber and exhaust system, with sampling

 port, downstream.

   The conventional use of the power supply  utilized in the present

 apparatus is to provide power for semiautomatic welding processes such

 as, gas  shielded, open arc, and submerged arc.  This "constant voltage

 (CV)" type of power source is a direct current, self-contained motor

 driven generator unit rated for a 100% duty cycle.  This rating allows

 continuous, uninterrupted production of power under full load.  Various

 instruments and controls are mounted on the front of the sheet metal

 enclosure of the motor-generator unit.  In addition to an ammeter and

 voltmeter, a 110 volt receptacle, fuse holder, remote control receptacle,

 and electrode and ground terminals are located on the welding machine.

 Controls include start-stop pushbuttons and a field rheostat to regulate

 the output voltage.  The voltmeter located on the machine reads open

 circuit voltage (while not welding) and arc voltage (while welding).

   Once  set, the output of a CV power supply has essentially the same

voltage no matter what the welding current may be.  The voltage can be
   iflobart Motor Driven Constant Voltage Welder for Automatic Welding,
Model MC-900, Hobart Brothers Company, Troy, Ohio.
                               1-10

-------
                                            WIRE REEL
                                                         CONTROL
                                                          PANEL
                                                               SHIELDING
                                                                     GAS
                                                                  SUPPLY
                               WIRE TORCH

                             ARC CHAMBER
                             SETTLING
                              CHAMBER
                                                       SHIELDING
                                                       GAS NOZZLE
                              TUNGSTEN
                              ELECTRODE
                              HOLDER
POWER SUPPLY
                                                                 COOLING
                                                                  WATER
                                                                    TANK
Figure 3.  Simplified diagram of aerosol generator  apparatus

-------
set over a range of approximately 10-50 volts; there is no current

control on a CV type machine as the welding current flow is determined

by the wire feeder.  That is, as the wire feed speed is increased, the

machine automatically supplies additional current, at the same constant

voltage, to maintain the arc (26, 27).  Thus, the wire feed speed

control knob "sets" the welding current.  The welding current is read

on either the ammeter on the welding machine or the ammeter on the

remote control panel for the "automatic welding head."

   The automatic welding head assembly^ consists of two main components:

(1) wire feed assembly with torch, and  (2) remote control panel.  The

wire feed assembly combines an insulated wire supply reel, guides, drive

motor with gear reduction box, and feed rolls, to constantly feed a

consumable electrode wire^ through the air-cooled torch to the arc.  The

consumable electrode wire used in the present apparatus is solid steel

(because an iron oxide fume was desired) and bare, except for a very

thin coating of copper to prevent surface oxidation.  The consumable

wire is maintained at a positive (+) potential with respect to ground.

The remote control panel is electrically connected to both the welding
   ^Hobart Automatic Welding Head, Model AI-22, Hobart Brothers Company,
Troy, Ohio.

   3The steel wire used is Hobart, Type 18 (E 705-G).  The spool was
further designated "6719-67, MIG 18B, .045."  A typical wire  trace
element composition by per cent is given  (27)
                     C               .12
                     Mn             1.90
                     Si              .80
                     P               .020
                     S               .020
                     Mo              .50

                                1-12

-------
machine and the non-consumable electrode holder  cooling water  supply



through solenoid valves.




   The front of the control panel  contains  the following  controls



and instruments (28).




   •'••  JLuse ~ This 4 ampere fuse fuses  the  input power to the  control



       panel from the welding machine.




   2.  Ammeter - This ammeter indicates the amount  of current  flow



       at the arc.




   3.  Wire Speed Control - This controls the speed of the wire through




       the torch.   (Figure C-2 is  the calibration curve for this




       control).




   4.  Voltage Control  - This controls  the  amount of voltage across




       the arc.




   5.  Voltmeter - This voltmeter  indicates the  amount of voltage




       across the arc.




   6.  Pilot Light - This light indicates when power is available at




       the control panel.




   7.  Wire Feed Control - This switch  controls  the direction  of travel




       of the wire through the torch.




   8.  Inch Button - By depressing this button,  the wire  inches through




       the torch, forward or reverse, depending  on  the position of the




       "wire feed control."




   9.  Start and Stop Button - These start  and stop the wire feed motor




       and therefore start and stop the arc.




  10.  Purge Button - This button  purges the gas through  the shielding




       gas nozzle.




                                1-13

-------
  11•  On-Off Toggle Switch - This switch controls power to the control




       panel.




  12.  Water Valve - This valve controls the on and off cycle of water




       to the non-consumable electrode holder.




  13.  Gas Valve - This valve controls the on and off cycle of gas to




       the shielding gas nozzle.




   The non-consumable electrode holder, shown in an exploded view in




Figure 4, was designed and fabricated by the author.  Materials of con-




struction include stainless steel, steel and brass as indicated in the




diagram.  The non-consumable electrode is a 1/4 inch (6.35 mm) diameter




tungsten arc welding electrode with a ground finish and 2 per cent thoria




content.  The ground finish is smoother and therefore makes better




mechanical and electrical contact than the swaged type.  Thoria is added




to tungsten electrodes to improve arc stability, to make initiation of




an arc easier, and to increase the amount of current carried per electrode




diameter (29).




   Several preliminary configurations of the tungsten electrode holder-




arc chamber combination were constructed to investigate the feasibility




of developing a workable semi-continuous model capable of achieving the




desired result.  Many of the problems encountered in the construction and




operation of the models described in Appendix B have thus been eliminated




in the construction of the present holder.  Final design of the tungsten




electrode holder involved three prime considerations:   (1) compensation




for electrode erosion, (2) electrical connection between power supply and




electrode, and (3) dissipation of heat build-up.





                                1-14

-------
Figure 4.  Photographs showing assembled and exploded view
           of the non-consumable electrode holder
Figure 4a.  Assembled view
Figure 4b.  Exploded view
      PARTIAL PARTS LIST

 1  Electrode guide (S)
 2  Sliding seal components
 3  Tungsten electrode
 4  Feed screw (B)
 5  Electrode holder body (M)
 6  Cooling water fittings (B)
 7  Rubber "0" ring
 8  Packing retainer (A)
 9  Silicone rubber packing
10  Feed screw guide (S)
11  Rubber bellows
12  Jam nut and lock washer (M)
13  Compression spring
14  Electrical connector
15  Rotating contact disks (B)
16  Plexiglas insulator
17  Feed screw knob
       Letter in parenthesis indicates material of construction:
aluminum (A), stainless steel (S), brass (B), and mild steel (M)
                              1-15

-------
   Among other reasons, tungsten was chosen as a "non-consumable"




electrode for its inherent property of having the highest melting




point of all metals, 6170 F (3410 C) (30).  But even with its




refractory nature, the intense heat of the arc, the availability of




air for oxidation, and the physical forces in the arc cause erosion




of the electrode.  The major portion of the erosion is due to melting




of the electrode and subsequent removal of the molten material from




the arc region by gravitational force, arc blow, and the moving gas




stream.  It is theorized a much smaller portion of the tungsten is




lost by evaporization as its boiling point is 10,700 F (5927 C) (30).




No trace of discrete tungsten particles could be detected on examin-




ation of the electronmicrographs of the collected fume as condensed




tungsten has a morphology similar to that of iron oxide.




   To advance the electrode into the arc, a manually operated screw




feed with a pitch of 20 threads per inch  (7.87 threads per cm) was




employed.  To compensate for the increased erosion rate on the down-




stream side of the electrode due to arc blow, the electrode was fixed




to the feed screw causing the electrode and screw to rotate together.




This arrangement allowed a relatively rounded point to be maintained




in the arc throughout the tests.




   Transfer of power from the stationary power supply to the  rotating




feed screw was accomplished through a sliding contact between two




4 inch (41.7 cm) diameter spring-loaded, brass disks  (see Figure 4).




A thin coating of graphite paste was placed between the disks for




lubricating purposes and to improve conductivity.  During the test  the
                               1-16

-------
  feed screw was turned approximately four revolutions per minute; upon

  disassembly no trace of arcing was observed between the contact

  surfaces of the disks.

     To take advantage of electron flow, the tungsten electrode is

  grounded (-) making it the cathode of the direct current circuit.

  The characteristics of electron flow, negative to positive potential,

  cause the consumable wire (anode) to be bombarded with electrons from

  the relatively non-consumable cathode thereby concentrating the heat

  in the wire to be vaporized.  At a current flow through the arc region

  of 150 amperes, the ratio of current density at the consumable wire

  versus that for the non-consumable tungsten electrode is approximately

  30 to 1.  Current density is defined as the current flow in amperes

  divided by the cross-sectional area of the conductor.  The current

  density for the wire is,

                   current _  150 amperes x 4
                    area      IT (0.045 inches)2
                                                2
    current density (wire) = 94,300 amperes/inch

                          (= 14,620 amperes/cm2).

  The current density in the tungsten electrode is,

                   current _ 150 amperes x 4
                    area     ir (0.25 inches)^

current density  (tungsten) = 3056 amperes/inch2

                          (= 474 amperes/cm2)

     As this is a direct current arc, the power  (P) consumed in the  arc

  is the product of electro-motive force  (E), in volts,  and current  (I),

  in amperes.
                                 1-17

-------
Therefore,




                P  =  El




                P  =  23 volts x 150 amperes




                P  =  3450 watts or 11,775 Btu/hr




                  (=  49.44 kg-calories/min).




   This relatively high rate of heat input to the system must be




dissipated by one of the various methods of heat transfer: convection,




radiation, or conduction.  Air flow through the arc chamber past the




arc region at the rate of 103 scfm (2.92 m3/min) supplies significant




cooling for the overall system.  However, the major source of concern




for heat build-up to adverse levels is the tungsten electrode holder




itself.  An easily replaceable 1/8 inch (3.175 mm) thick steel disk




(see Figure C-l) is used to protect the exposed end of the electrode




holder from arc splatter and to reflect the radiant energy of the arc.




To dissipate the heat build-up in the tungsten electrode, a portion of




which is conducted to the electrode holder through intimate metallic




contact, the entire electrode holder is water cooled.  Cooling water




is supplied by a self-contained tank and pump assembly , through a




solenoid valve located on the control panel assembly to the inlet port




on the electrode holder.  The return water flows unrestricted back to




the holding tank which acts as a heat sink.




   No attempt was made to cool the arc chamber itself as no problems




of dimensional stability due to heat build-up were encountered.  The




arc chamber was fabricated from a 13 inch  (33 cm) length of 5 inch  (12.7  cm)
    Hobart Circoolator, S-3568B, Hobart Brothers Company, Troy,  Ohio.



                                1-18

-------
Figure 5a.
Electrode holder being inserted
into mounting block of arc chamber
 Figure 5.
            Figure 5b.  Arc chamber mounting
                        arrangement with electrode
                        holder in place
Photographs showing general
arrangement of arc chamber apparatus.

-------
diameter, Schedule 40, steel pipe.   The pipe axis was supported in




a horizontal plane by steel supports (see Figure 5) attached to the




chamber by stainless steel clamps;  the steel supports were then




bolted to a 20 x 24 inch (50.8 x 61.0 cm), 1/2 inch (12.7 mm) thick




aluminum plate.  (The wire feed assembly with torch are also bolted




to this base plate through insulators.  This arrangement assured




positive orientation between wire feed torch, tungsten electrode




holder and arc chamber.)  The attachments between the chamber and




the base plate, and between the wire feed assembly and base plate,




are designed to allow adjustment in all axes to obtain perfect




alignment  between the feed wire and the non-consumable electrode.




   The feed wire torch passes through the chamber wall 90 degrees




to both the center line axis and the vertical tangent of the circular




chamber.  A concentric phenolic collar insulates the positive potential




of the wire torch from the ground (-) potential of the chamber, base




plate, and electrode holder.  The tungsten electrode holder passes




through an aluminum mounting block, milled to the contour of the arc




chamber and bolted in place, on the same axis as the wire torch.  A




sliding fit exists between the mounting block and electrode holder,




the latter being held in place by set screws.  Various adjustments




allow the arc to be maintained on the center line of the chamber.




Refer to Appendix C, Figure C-l for a detail of the positioning of the




various components inside the arc chamber.




   A 2 x 4 1/4 inch (5.1 x 10.8 cm) view window is located on  the  top




of the chamber to allow visual evaluation of tungsten electrode erosion




                                1-20

-------
and observation of arc shape.  The window is a standard lift-front,
lens holder from a weldor's helmet modified to fit the chamber.  A
Shade Number 12 welding lens is used in the holder as protection from
the injurious rays of the arc.  The lift-front feature of the lens
holder permits raising the window for a clear view of the interior
of the chamber and the opening allows insertion of tools for making
adjustments within the chamber.
   Another salient feature of the arc chamber is the shielding gas
nozzle and supply system.  The shielding gas is supplied by a standard
steel cylinder containing approximately 244 standard cubic feet
(6.91 m3) of gas under a pressure of 2200 psig (154.7 kg/cm2).  To
reduce cylinder pressure to a usable level and assure constant flow,
a single stage regulator and flowmeter is used and is mounted on top
of the cylinder.  A valve is used in the line to control flow through
the flowmeter which is graduated in standard cubic feet of helium per
hour (cfh).  The gas flows to a solenoid on the control panel which
shuts off the gas supply after the arc is extinguished, thus conserving
the inert gas.  From the solenoid, the gas flows to a ceramic nozzle
located in the arc chamber which directs the gas at the tip of the
tungsten electrode.  Effective shielding of the tungsten electrode dis-
places the air surrounding the electrode; preventing rapid consumption
of the tungsten by oxidation.  However, effective shielding is a dif-
ficult problem.  Turbulence in the inert gas stream itself tends to
pull in air; this problem is compounded by the relatively high velocity
air stream being forced past the arc region.  In an attempt to compensate
for gas stream turbulence and maintain effective coverage of  the electrode
                               1-21

-------
tip, an unusually high gas flow is used - 100 cubic feet per hour



(47 1/min).  Because of the relatively high rate of electrode consump-



tion experienced it is still doubtful whether adequate coverage was




obtained.



   The usual choice of inert gas for shielding during tungsten inert



gas. (TIG) arc welding is between argon and helium.  Each gas has its



own particular advantages for different types of welding processes and



materials.  In the present apparatus, helium was the choice because of



its availability.  However, the choice of helium over argon appears



advantageous for several reasons:  (1) helium allows a 40 percent



greater heat input per unit length of arc because of its higher ioniza-



tion potential, and (2) it gives a more stable arc when using direct



current power.



   Figure 6 is a pictorial diagram of the arc chamber and its associated



ductwork.  Air flow through the system is provided by a "Gelman Hurricane



Air Sampler" set on low speed.  The air discharge of the blower passes



through a 1 1/8 inch (28.6 mm) orifice, which limits flow to a nominal



100 cubic feet per minute (47.2 liters/second), through a detachable



muffler into a 4 inch (10.2 cm) diameter horizontal duct.  All ductwork




throughout the system is fabricated from 22 gauge (0.794 mm) galvanized




steel sheet.  A transition piece connects the 4 inch (10.2 cm) diameter



duct to a 2 foot (0.61 m) length of 5 inch (12.7 cm) duct leading  directly




to the arc chamber.  From the chamber, a 2 foot (0.61 m) length of 5  inch




round duct diverges into a 10 inch (25.4 cm) diameter section over a



distance of 2 feet (0.61 m) and intersects a vertical duct at a 90 degree




angle forming a tee.  The bottom of the tee is closed by a removable

-------
t
KJ
U)
       ELUTRIATION
         COLUMN
       10"O.D.
       VIEW
       WINDOW
                                                          EXHAUST
SETTLING CHAMBER
                                                                AIR
                                                               MOVER
                                                                  TO POWER SUPPLY
                  Figure 6.   Simplified pictorial sketch of the arc chamber  and  duct work  system

-------
cap; the top of the tee converges over a 2 foot (0.61 in) length to




a 3 inch (7.62 cm) diameter duct.




   To separate the larger molten globules of steel from the gas




stream, the ductwork leading from the arc chamber incorporates a




settling chamber and elutriation column section in its design.  The




actual velocity in the 10 inch (25.4 cm) diameter vertical duct is




220 feet per minute (67.1 m/min) or 3.7 feet per second (1.12 m/sec).




At this velocity steel particles larger than approximately 1/8 inch




(3.175 mm) diameter are effectively eliminated from the gas stream by




gravitational force.  Further, inertial separation takes place in




this section as the gas stream is caused to make a 90 degree change




in direction.




   A view window, again adapted from an arc welder's helmet, is located




on the vertical portion of the tee in such a manner as to allow observa-




tion of the arc by sighting through the inside of the 5 inch (12.7 cm)




diameter duct.




   The 3 inch (7.6 cm) duct leaving the elutriation column makes a




180 degree turn, and passes a 1/2 inch  (12.7 mm) sampling nozzle.  An




adequate length of straight duct was established to allow optimum




positioning of the sample probe — a minimum of 10 diameters downstream




and 5 diameters upstream from any disturbance to air flow.  The 3 inch




(7.62 cm) duct then diverges into a 4 inch  (10.2 cm) duct, makes a




180 degree turn and exhausts the contaminated gas stream to the atmos-




phere through a roof ventilator.




   Temperature sensors (bimetallic helix, dial thermometers) were placed




in the duct at locations where significant  temperature  changes were



                               1-24

-------
expected to occur.  A thermometer was also placed in close proximity




to the sampling port.  Location of thermometers is indicated on




Figure 6.



                           *     *     *




   A detailed procedure for the assembly and operation of the electric




arc aerosol generator apparatus is found in Appendix C.
                                1-25

-------
                         IV.  EXPERIMENTAL



   To Investigate the performance characteristics of the aerosol

generator under semi-continuous operation, a series of three con-

trolled test runs were made and a portion of the resultant fume was

collected.  All tests were conducted under identical conditions of

arc voltage, wire feed rate, shielding gas flow rate, and dilution

air flow rate.  Refer to Table 1 for a complete listing of test

conditions.  The test settings for voltage and wire feed were estab-

lished on the bases of limited preliminary studies and are not claimed

to be optimized values.  The test program was structured to determine

the mass rate of production of iron oxide fume, the particle size

distribution of that fume, and the degree of reproducibility of pro-

duction rate and size distribution.


SAMPLING PROCEDURE

   After the arc was initiated according to the procedure outlined in

Appendix C, the system was allowed to operate for approximately one

minute to establish steady-state conditions for sampling.  During this

period the temperature of the exhaust gas stream stabilized producing

a constant velocity past the sampling nozzle.  A ten minute sampling

period then commenced.

   The sampling train consisted of a one-half inch  (12.7 mm) stainless

steel hook nozzle attached to an Andersen Stack Sampler-5, in series
   5Andersen Stack Sampler manufactured by 2000 Inc., 5899 South State
Street, Salt Lake City, Utah 84107.
                               1-26

-------
Table 1.  Operating and sampling test conditions


OPERATING PARAMETERS         English Units        Metric Units

  Wire Feed Rate            200 inches/min      5.1 meters/min

  Wire Diameter             0.045 inches        1.14 mm

  Cathode Feed Ratea        0.2 inches/min      5.08 mm/min

  Cathode Diameter          0.25 inches         6.35 mm

  Arc Chamber Pressure     +0.72 inches W.C.   +1.34 mm Hg

  Shielding Gas Flow Rate   100 cubic feet/hr   47.2 liters/min

  Air Flow Rate             103 scfm            2.9 cubic meters/min

A power supply voltage control setting of 23 volts produced a current
flow through the  arc which fluctuated between 150 and 165 amperes.


SAMPLING PARAMETERS

  Sample Time                       10 minutes per run

  Sample Rate               0.72 cfm @ 70F      20.41 1pm @ 21C

  Exhaust Gas Temperature   167F                75C
    at Sample Port

  Sample Nozzle Diameter    0.25 inches        6.35 mm

  Ambient Temperature       70F                21C


    aCathode  (tungsten electrode) was manually advanced approximately
four turns per minute to compensate for  erosion.
                                1-27

-------
with an aluminum Millipore filter holder supporting a 47 mm diameter




Gelman Type GA-4 membrane filter to separate the fume from the sample




gas stream.



   The Andersen Stack Sampler is approximately 2.75 inches (6.99 cm)




in diameter and 5.0 inches (12.7 cm) in length, and weighs 3 pounds




(1.36 kg).  Construction is of stainless steel to withstand high




temperatures and resist corrosion.  The sample contains nine stages




which are numbered 0 through 8, the number 8 stage having the same




characteristics the number 2 stage has.  The stages are 2.5 inches




(6.35 cm) diameter gold plated brass plates with asbestos gaskets




between the plates for spacing and sealing.  Each stage has 300 air




jets  (holes) arranged in concentric circles, the holes decreasing in




size moving from 0 through 7 (31).  As the air flow through each stage




is the same, the velocity of air through the air jets increases as hole




size decreases.  When the velocity imparted inertia of a particle is




sufficient to overcome the aerodynamic drag, the particle will leave




the gas stream and impact on the collection surface.  Therefore,




selective impaction of a particle on a particular plate depends on its




aerodynamic dimensions - size, shape and density - and their relation-




ship to gas velocity (32).




   A calibrated orifice was used to indicate instantaneous sample flow




rate, with total flow rate recorded on a dry gas meter.




   Fume concentration was determined gravimetrically.  After the sample




was collected on the pre-weighed collection plates of the Andersen




sampler and the back-up membrane filter, the plates and filter were




                                1-28

-------
dried at 248 F  (120 C) for two hours.  The plates and filter were then

removed to a. dessicator containing indicating type silica-gel and

allowed to come to room temperature.  When the plates and membrane fil-

ter were removed for weighing they were immediately placed on the en-

closed balance pan and weighed.  A small container of indicating-type

silica-gel was placed in the balance enclosure.  After the gravimetric

determinations were made for Run 1, the collection plates were cleaned

using minimal amounts of distilled water and a rubber policeman.  The

wash water from each plate was collected in separate test tubes, and

the particulate allowed to settle for one week; the bulk of the water

was then decanted and the remaining water evaporated in a drying oven.

Each sample of collected fume was then resuspended in 2 ml of a 1%

nitrocellulose solution  and dispersed by placing the test tube in an

ultrasonic cleaner.  The fume collected on the membrane filter was

carefully scraped from the membrane and placed directly in 2 ml of the

nitrocellulose solution and dispersed.

   A disposable pipette was used to place one drop of the solution on

an uncoated electron microscope grid.  A separate grid was prepared

using the solution containing particles from each collection plate

and the membrane filter from Run 1.  The 200 mesh grids (specimen

screens) were bombarded by electrons and the projection recorded on

photographic film for later particle size determination.  An RCA

Type EMU-4 electron microscope was used and produced a primary

magnification of 29,200 times.
   6No. 1120 Nitrocullulose Solution, 1%, 30 ml bottles. Ernest F.
Fullan, Inc., P.O. Box 444, Schenectady, New York.
                               1-29

-------
                Table 2.  Weight  (mass) distribution  of  particulate matter  retained

                           by Andersen Stack  Sampler  and back-up membrane filter


STAGE
0
1
2
3
4
5
6
7
8
Filter
TOTAL
RUN 1


Weight
(grams)
.0025
.0004
.0004
.0008
.0013
.0027
.0081
.0164
.0090
.0916
0.1332


Percent
Total
1.88
0.30
0.30
0.60
0.98
2.03
6.08
12.31
6.76
68.77
100.01
Cumulative
Relative
Frequency
(Percent)
100.01
98.13
97.83
97.53
96.93
95.95
93.92
87.84
75.53
68.77

RUN 2


Weight
(grams)
.0015
.0003
.0002
.0002
.0007
.0031
.0077
.0112
.0128
.0861
0.1238


Percent
Total
1.21
0.24
0.16
0.16
0.57
2.50
6.22
9.05
10.34
69.55
100.00
Cumulative
Relative
Frequency
(Percent)
100 . 00
98.79
98.55
98.39
98.23
97.66
95.16
88.94
79.89
69.55

RUN 3


Weight
(grams)
.0024
.0005
.0004
.0005
.0015
.0033
.0084
.0174
.0134
.1034
0.1521


Percent
Total
1.58
0.33
0.26
0.33
0.99
2.17
5.52
11.44
8.81
68.57
100.00
Cumulative
Relative
Frequency
(Percent)
100.00
98.42
98.09
97.83
97.50
96.51
94.34
88.82
77.38
68.57

Co
o

-------
 Table 3.   Mass concentration and  total  production rate of iron oxide particles




CD      (2)      (3)        (?)         (5)      (6)      (5)       .(8)        (9)

RUN
1
2
3
AVE
Table 1
Sample
Time
(mins)
10
10
10
10

Total
Sample
Volume
(scf)
7.23
7.21
7.17
7.20
Table 1
Sample
Weight
(grams)
0.1332
0.1238
0.1521
0.1364
15'43%
22880
Mass Concentration
(gr/scf)
0.284
0,265
0.327
0.292
(mg/m )
649.8
606.3
748.2
668.1
Table 1
Total
Gas
Flow
(scfm)
103
103
103
103
©X{6)
.008570
Total Particulate
(gr/min)
29.25
27.30
33.68
30.08
(Ibs/hr)
0,251
0.234
0.289
0.258
.06480
3.890
Production Rate
(g/min)
1.90
1.77
2,18
1.95
(g/hr)
113.8
1Q6. 2
131.0
117.0

-------
RESULTS


   In all test runs to determine performance characteristics,


identical apparatus and operating and sampling conditions were


maintained as listed in Table 1.



Mass Concentration and Rate of Production


   The mass of particulate in the sample gas stream was deter-


mined from the gravimetric differential between the tare weight of


the clean Andersen collection plates and membrane filter and the


weight of the plates and filter after sampling.  The weight


differential for each of the tests runs is shown in Table 2.


Combining the information from Table 2 with knowledge of the total


gas volumei'sampled and the sample time, the mass concentration of


particulate matter in the sample gas stream was calculated.  Average


mass concentration of iron oxide fume in the sample gas stream was

                            Q
0.292 grains/scf (668.1 mg/m ).  Assuming the mass concentration of


particulate matter in the total gas stream is the same as for the


sample gas stream, the average total particulate production rate was


calculated to be 30.08 grains/min (117.0 grams/hr).  Additional


information concerning mass flow rate and mass concentration cal-


culated for each run for both the sample and total gas streams is


presented in Table 3.  Total particulate production rate, given in


various common units of measurement, is also listed in Table 3.


   Another performance factor directly related to the rate of pro-


duction is the conversion ratio of consumable wire feed stock to


ultrafine particles.  For the feed rate of 200 inches  (5.1 m) per


                               1-32

-------
minute of 0.045 inch  (1.14 mm) diameter mild steel wire, 12 thousand

inches (305 m) are consumed in one hour.   (Refer to Table 1).  As

there are 2210 inches per pound  (25.46 m/kg) of wire  (33), the wire

consumption rate was

            12.000 inches/hr      ,
            2,210 inches/lb   =  5'43 lb/hr  <2'46 k§/hr>-

   From Table 3, the average particulate production rate was

0.258 lb/hr (117.0 g/hr).  Therefore, 5.43  pounds  (2463 g) of wire was

required to produce 0.258 pounds  (117.0 g)  of ultrafine particles for

a conversion ratio of 4.75 percent.  Although no complete mass balance

was performed, the remaining portion of consumed metal can be accounted

for by the following observations.  A relatively large portion of the

feed stock was not vaporized as  some material melted  and formed a

puddle in the bottom of  the arc  chamber.  A significant number of

relatively larger particles, approximately  1000 micrometers, settled

out in the settling chamber and  elutriation column section of the

ductwork.  The loss of smaller particles is due to the electrophoretic

and thermophoretic forces acting on the particles suspended in the gas

stream.  These forces cause the  smaller particles to  migrate to the

walls of the ducting and attach  themselves  through various adhesive

and electrostatic forces.  A thin reddish-brown coating was observed

on the inside of the ductwork after completion of the test program.

   It is also interesting to note the number of discrete spherical

particles generated at the above production rate.  Considering a mass

mean diameter of 0.022 micrometers and specific gravity of approximately
                               1-33

-------
5.2 (34), the numerical generation rate is calculated in the

following manner.
     Volume of single sphere = i^d3
                               6


                             " !ir(0.022)3 = 5.58 x l(f6 pm3
                               6

     Specific gravity of Fe203 =5.2
                                                         rt   O
     Weight of single sphere = 5.58 x 10"6 pm3 x ICT12 cm ^m  x 5.2 g/cm3

                                        —18
                             = 29.0 x 10~   grams

     Generation rate = _ 117.0 g/hr _ ,, .    .18     .  .   „
                       - TH - ~ 4 x 10   particles/hr
                       29.0 x 10    g/particle


     This is more than four quintillion particles per hour.
Particle Size Distribution

   Two methods were employed, with differing results, to determine

particle size distribution:  (1) an electron microscope and

(2) an Andersen Stack Sampler.

   Because of the high magnification, and consequently small field,

of the electron microscope, photographic plates were made of a mini-

mum of three fields of each grid.  The photographic plates were en-
                               ^ *
larged and printed on 8 x 10 inch   (20.3 x 25.4 cm) photographic paper

giving a final magnification of 87,600 times.  The electron micrographs

obtained in this manner have a two-dimensional appearance and discrete

particles are readily discernible.  Individual particles of iron oxide

are spherical* and individual spherical particles composing larger

agglomerates were easily distinguished.  The particle size


                               1-34

-------
distribution data presented in Tables 4 and 5 and graphically




represented in Figures 7, 8, 9, 10, and 11 are based on direct




counts of discrete, spherical particles.  That is, agglomerates are




not assigned a composite or equivalent particle size but the




particles composing an agglomerate are classified individually.




A specially constructed semi-circle graticule was used to facilitate



particle sizing.




   Examination of the raw particle count data found in Appendix D




lists thirty fields for which particle size determinations were




made.  The number of discrete particles observed on each 2.275 x




2.875 micrometer field ranged from 177 to 742 with an arithmetic




mean of 412.  A total of 12,355 particles were individually sized.




   In order to simplify calculation of various statistical values,




the raw data was grouped into 10 classes with a particle size




interval of 0.01 micrometers each.  Although the following mathe-




matical descriptions of particle size were calculated on the assump-




tion that all particles were classified in one of ten equal-interval




classes, this was not exactly the case.  As a small number of particles




exceeded the upper boundary of the tenth class (0.105 micrometers),




the tenth class was actually open-ended and contained all particles




0.095 micrometers in diameter and larger.  This assumption introduces




a slight error in the following calculations, but this error is




relatively insignificant as only 30 particles of the 12,355 particles




counted exceeded 0.105 micrometers.  Three of these particles  exceeded




0.2 micrometers at 0.22, 0.25 and 0.24 micrometers and were found  on




stages 5, 6, and 7, respectively.




                               1-35

-------
                             Table 4.  Electron microscope particle count data'
STAGE
0
1
2
3
4
5
6
7
8
Filter
TOTALS
RANGE
Particle Size, micrometers
0.005-
0.015
522
351
709
431
773
592
345
436
749
818
5726
493
0.015-
0.025
248
183
249
164
419
285
149
209
336
421
2663
270
0.025-
0.035
153
145
161
123
308
210
126
149
233
300
1908
185
0.035-
0.045
87
56
91
56
157
68
48
87
98
159
907
111
0.045-
0.055
51
37
46
36
56
47
25
50
97
86
531
72
0.055-
0.065
24
21
27
19
39
33
20
21
20
68
292
49
0.065-
0.075
17
12
12
9
13
29
5
11
9
27
144
20
0.075-
0.085
15
3
3
7
10
4
2
10
7
15
76
13
0.085-
0.095
1
2
2
5
5
3
3
2
4
7
34
5
0.095
and>
9
7
5
1
16
8
5
9
6
8
74
15
TOTALS
1127
817
1305
851
1796
1279
728
984
1559
1909
12355
1181
1
LJ
               Refer to APPENDIX D for raw particle count data.

-------
Table 5.  Relative Particle Size Frequency Distribution in Percent from data of Table 4
STAGE
0
1
2
3
4
5
6
7
8
Filter
Mean
Standard
Deviation
Range
Particle Size, micrometers
0.005-
0.015
46.3
43.0
54.3
50.5
43.0
46.3
47.4
44.3
48.0
42.8
46.6
3.73
11.3
0.015-
0.025
22.0
22.4
19.1
19.2
23.3
22.3
20.5
21.2
21.6
22.1
21.4
1.38
4.2
0.025-
0.035
13.6
17.7
12.3
14.4
17.1
16.4
17.3
15.1
14.9
15.7
15.5
1.72
5.4
0.035-
0.045
7.7
6.9
7.0
6.6
8.7
5.3
6.6
8.8
6.3
8.3
7.2
1.15
3.5
0.045-
0.055
4.5
4.5
3.5
4.2
3.1
3.7
3.4
5.1
6.2
4.5
4.3
0.92
3.1
0.055-
0.065
2.1
2.6
2.1
2.2
2.2
2.6
2.7
2.1
1.3
3.6
2.4
0.59
2.3
0.065-
0.075
1.5
1.5
0.9
1.1
0.7
2.3
0.7
1.1
0.6
1.4
1.2
0.52
1.7
0.075-
0.085
1.3
0.4
0.2
0.8
0.6
0.3
0.3
1.0
0.4
0.8
0.6
0.36
1.1
0.085-
0.095
0.1
0.2
0.1
0.6
0.3
0.2
0.4
0.2
0.3
0.4
0.3
0.16
0.5
0.095
and>
0.8
0.9
0.4
0.1
0.9
0.6
0.7
0.9
0.4
0.4
0.6
0.27
0.8

99.9
100.1
99.9
99.7
99.9
100.0
100.0
99.8
100.0
100.0
100.0

    Refer to APPENDIX E, Table E-3 for calculations.

-------
  Figure 7.  Histogram representing particle size  data  from Table 5
60*

, percent
.e- m
0 0
i i
1 30-

-------
l''lKur<» ().  Ogive representing  cumulative distribution data from Table 6
 c 100""
 eg

£  90"

 en
 8  80-
 g
 CJ
 Vi
 0)
 PL.
   70-


   60-
I  40
0)

°"  30 —
a)  -ju—i
 (U
 >
 CO
   20-


   10-
                    Median
      CO
      n
      a)
      4J

      §
      o
      Vf
      u
      Q)


      I
      M
      T3
      H
      O
   .01





Figure


.10-


.09-


.08


.07-1


.06


.05-


.04-


.03-


.02 -


.Ol-
                    I

                  .02
                         '      I      i^     I      IT

                        .03    .04    .05    .06    .07    .08


                         Particle diameter, micrometers
                                                .09   .10
                lO.  Cumulative distribution  data  from Table 6

                     plotted using normal probability scales
           0
                      Mediani
                              I '  I '  I   '  §  ' I T
                              |50  |70  I   90 95 98J99.:

                             40  60  80         99
                                                         99.99
1  2  5 10 20 30


Cumulative relative  frequency,  percent less than


                  1-39

-------
Figure 11.  Cumulative particle size distribution data from
            Table 6 plotted using log-normal probability scales
 CO
 n
 
-------
   The raw particle count data of Appendix D has been somewhat



reduced arid presented in a more usable form in Table 4.  A simpli-



fied method for calculating the arithmetic mean and standard devia-



tion of the particle size distribution as described by Cadle (35)



is employed.  Calculation of the arithmetic mean, d, is based on



the following equations,



     _        n              n

     d = c 1  E u f  +d  I  If
           n £-^ i i    on ^-^ i



     d~ = cu + dQ



where c is the class interval, n is  the  total number of particles



counted, u. is the class mark of the new scale  (-1,0,1,2. . .8) ,



d  is the class mark corresponding to u    0, and f , is the frequency
 o                                     i          x


for each class.  Using  the transformations found in Appendix E,



Table E-l, the arithmetic mean is,



     d - .01 / 2763  \ + 0.02
             / 2763 \ + 0.

             VL2.355/
     df = 0.022 micrometers



     The equation for standard deviation  (s) employs the same notation



and is calculated using the following equation.
                          -2*
     a - c / i  y u. f, -i
              ,  £-cl _ JL  -I-
     Using the values of Appendix E, Table E-]





     s - 0.0:
                 ^12355/   \12355



     s = 0.0163 micrometers
                                1-41

-------
   In order to present the particle size distribution data in


the several graphical forms of Figures 9, 10 and 11,  it was


necessary to transform the numerical distribution into a relative


frequency distribution expressed in percent.  The results of this


conversion are presented on Table 6.  The histogram and distribu-


tion curve of Figures 7 and 8, respectively, indicate a highly


positive skewed distribution based on the particle size range


examined.  As the lower limit of resolution of the electron micro-


scope was approximately 0.005 micrometers (50 A), no information


can be presented on particles that may have formed with diameters


less than 0.005 micrometers.


   The range of relative frequency found on each of the specimen


screens for each size interval is also presented on Figure 8 and


indicates the size distribution of discrete particles collected on


each plate of the Andersen Stack Sampler and the membrane filter is


approximately the same.  Standard deviation of the relative frequency


within each size interval is presented in Appendix G.  The range of


standard deviations also seems to indicate similar particle size
                                      s

distributions.


   The cumulative frequency distribution tabulated in Table 6 is


presented as an Ogive in Figure 9 and, more conventionally, using


normal probability and log-normal probability scales in Figures 10


and 11 respectively.  The log-normal scales of the latter reveal a


relatively straight line indicating a log-normal particle size


distribution.


                               1-42

-------
Table (>.  Cumulative distribution of data from Table 5
Class
boundaries
L
U
Percent
in
Class
Percent less than
upper class limit

0.005
0.015
0.025
0.035
0.045
0.055
0.065
0.075
0.085
0.095
0.015
0.025
0.035
0.045
0.055
0.065
0.075
0.085
0.095
0.1053
46.3
21.6
15.4
7.3
4.3
2.4
1.2
0.6
0.3
0.6
46.3
67.9
83.3
90.3
94.9
97.3
98.5
99.1
99.4
100.
        Refer  to  page  34  for  explanation  of upper  limit for
this  class.
                           1-43

-------
   The median and geometric mean are calculated using the procedures

presented by Freund and Williams (36).  By definition, the median

divides the frequency distribution in half, e.g. there will be the

same number of particles above the median size as there are below.

For the present frequency distribution containing 12,355 ranked

observations, the median corresponds to the calculated particle size

of the 6178th particle.  For the grouped data of Table 4 the median (M)

is calculated using the following equation (37),
where L is the lower class boundary of the interval containing the

median value, c is the class interval, J is the difference between

the median observation and the cumulative frequency of all classes

preceeding the class containing the median, and fM is the frequency of

the class containing the median.  Using the tabulated values of

Appendix F, the median is

                    M -  015 + O.Q1 (6170-5726)
                                    2666



                    M = 0.0167 micrometers.


   The median, sometimes known as the mass median diameter  (HMD), has

been graphically inserted in Figures 7 and 9.  Inspection of Figure  9

confirms the calculated value of the median as the ogive curve repre-

senting the cumulative frequency distribution passes through coordinates

of 50 percent cumulative relative frequency and 0.0167 micrometer

particle diameter.               -11

-------
   Geometric mean  (G) is calculated based on the following

relationship  (38),



                       log G = I log x-  (f)
where x  is the class mark of an interval, f is the frequency of

that class interval, and n is the total number of observations

(£f).  Again using the  tabulated values found in Appendix F, the

geometric mean is calculated as follows,



                  Log G - "27,715.1 = 2". 24322
                           12,355


  antilogarithm "2.24322 = 0.01751


                      G = 0.0175 micrometers

   The geometric mean has been plotted on Figures 10 and 11 and at


the "50 percent less than" point intersects the frequency dis-

tribution curves as expected.


   By inspection of Table 4, the mode for this particle size dis-

tribution occurs in the 0.005-0.015 micrometer diameter interval.


   As stated previously, two methods were employed to determine

particle size distribution.  The second method involved the relatively

new Andersen Stack Sampler used in conjunction with a back-up membrane


filter.  Combining the  mass distribution data of Table 2 with cali-


bration information supplied by the manufacturer, cumulative particle


size mass distribution  can be obtained and is presented on Table 7.


Although this mass distribution data indicates approximately


69 percent of the sampled particulate passed through the Andersen


                               1-45

-------
             Table 7.   Arithmetic mean and  standard  deviation  of  cumulative relative mass distribution
                       from data of  Table 2
STAGE
0
1
2
3
4
5
6
7
8
Filter
Approximate
range of
particle
size3
(micrometers)
30 and larger
9.2-30
5.5-9.2
3.3-5.5
2.0-3.3
1.0-2.0
0.3-1.0
0.1-0.3
less than 0.1

Upper
Class
Limit
(micrometers
unknown
30
9.2
5.5
3.3
2.0
1.0
0.3
0.1
0.1
RUN 1
RUN 2
RUN 3
Cumulative relative mass
(percent less than
upper class limit)
100.10
98.13
97.83
97.53
96.93
95.95
93.92
87.84
75.53
68.77
100.00
98.79
98.55
98.39
98.23
97.66
95.16
88.94
79.89
69.55
100.00
98.42
98.09
97.83
97.50
96.51
94.34
88.82
77.38
68.57
Mean
cumulative
relative
massc
(percent less
than)
100.03
98.45
98.26
97.92
97.55
96.71
94.47
88.53
77.60
68.96
Standard
deviation0
(percent)
	
0.330
0."386
0.437
0.652
0.872
0.631
0.603
2.188
0.518
c
s
X
	
.0033
.0039
.0044
.0066
.0090
.0066
.0068
.0281
.0075
-p-
           3  Farrell Yeates, it. at., "Calibration and Evaluation of the Andersen Stack Sampler", Technical
     Report No. FY70 MCR-1, submitted to Two Thousand, Inc., Salt Lake City, Utah (April, 1970), p. 18.
           b  From Table 2.
           c  Refer to APPENDIX E, Table E-4 for calculations.

-------
Figure  12.   Cumulative particle  size  distribution data from

             Table 7 plotted using  log-normal probability scales
   
-------
Stack Sampler without collection, the previous data presented concerning
the size of the discrete particles suggests essentially all the
particulate matter should have passed through the sampler without
impacting on the collection plates.  As the particles were produced in
a direct current arc, it was expected that chains and agglomerates
would form, thereby increasing the effective aerodynamic size of the
particles.  In fact, a limited number of these formations were observed
on the electron micrographs.  However, since the electron microscope
sample preparation technique was designed to disperse any particle
aggregates to facilitate counting, the effective size could not be
confirmed by direct observation.
   The Andersen Stack Sampler utilizes aerodynamic sizing to equate
the size of all particles sampled	irrespective of shape or density	
to the aerodynamic equivalence of unit density spheres.  Vomela and
Whitby (39) have made measurements on aggregates consisting of from
10 to 300 particles having a mass mean diameter of 0.05 micrometers
and concluded the fluid drag on agglomerates of particles is nearly
the same as that of a sphere having a volume equivalent to that of
the aggregates.  For example, a mass consisting of one hundred,
0.022 micrometer diameter particles is aerodynamically equivalent  to
one particle 0.1 micrometers in diameter.  At present, the only
calibration data available from the manufacturer is for unit density
spheres and is listed on Table 7.
   The mass distribution developed in table 7 and presented  graph-
ically in Figure 12 is plotted as if the particles were spherical
and of unit density.  In reality this is known not to be  the case  as
                                1-48

-------
the specific gravity of iron oxide  (Fe203) is in the range of

5.12 - 5.24 (34), thereby increasing the relative inertial force

of spherical iron oxide particles when compared to the same size

unit density spheres.  That is, inside the sampler an iron oxide

sphere will overcome the aerodynamic drag force and leave the turning

gas stream  (and impinge on the collection plate) at a much lower

velocity than would the same size particle of unit density.


Reproducibility of Production Rate  and Relative Particle Size
Distribution

   The production rate of particulate matter for each of the three

runs was calculated and presented on Table 3 with the arithmetic

mean calculated at 30.08 grains/min (1.95 g/min).  As an indication

of variability about the mean, the  standard deviation (s) was

calculated  using the following equation,
                             n
                                    -  x)'
                                  n
where x  represents  an individual  observation, x is the arithmetic

mean, and n  is  the total  number  of observations.
29.25
27.30
33.68
-0.83
-2.78
3.60
.6889
7.7284
12.9600
                                       m.3773
                               1--49

-------
Therefore,
                      \i
21.3773    - 2.68 grains/min
   Two methods were used to estimate the reproducibility of the




relative particle size distribution	both using the mass distri-




bution data obtained from the Andersen Stack Sampler and membrane




filter.  The results of the three runs are presented in Table 2.




   Table 8 presents the arithmetic mean and standard deviation




for the weights found for each of the three runs on each stage of




the sampler and the membrane filter.  As a further, and possibly




more meaningful, comparison of the standard deviations of the various




stages with each other, the last column on Table 8 presents the




standard deviation as a percent of the arithmetic mean of the weights.




The mean and standard deviation of the total particulate mass captured




during each run is also given and is another indicator of the repro-




ducibility of production rate.




   The second estimate of size distribution reproducibility is obtained




by calculating the arithmetic mean and standard deviation of the cumu-




lative relative mass distribution for each of the various size inter-




vals.   The results are presented on Table 7.  Again the last column




represents the standard deviation as a percentage of the mean, in this




case the mean cumulative relative mass.
                                1-50

-------
Table 8.  Arithmetic mean and standard deviation of the
          weight  (mass) distribution from data of Table 2
STAGE
0
1
2
3
4
5
6
7
8
Filter
TOTALS
RUN 1
RUN 2
RUN 3
Collected weight
(milligrams)
2.5
0.4
0.4
0.8
1.3
2.7
8.1
16.4
9.0
91.6
133.2
1.5
0.3
0.2
0.2
0.7
3.1
7.7
11.2
12.8
86.1
123.8
2.4
0.5
0.4
0.5
1.5
3.3
8.4
17.4
13.4
103.4
152.1
X
Mean
Q
weight
(mg)
2.13
0.40
0.33
0.50
1.17
3.03
8.07
15.00
11.73
93.70
136.00
Mean
s
Standard
•a
deviation
(rag)
0.4496
0.0812
0.0943
0.2449
0.3398
0.2493
0.2867
2.7178
1.9482
7.2171
14.4202
1.3629
_s_a
X

0.2110
0.2030
0.2857
0.4898
0.2904
0.0822
0.0355
0.1561
0.1660
0.0770
0.1060
0.1997
       a  Refer to APPENDIX E, Table E-5 for calculations,
                           1-51

-------
                          V.  CONCLUSIONS






   On the basis of the results of the limited experimental tests




reported herein, the feasibility of employing an electric arc to




produce aerosols of ultra-fine iron oxide particles from a feed-




stock of consumable wire has been demonstrated.   Although little




attempt was made to optimize the operating parameters,  more than




30 grains per minute (1.95 g/min) of discrete, spherical particles




with a mass median diameter of 0.0167 micrometers  was  repeatedly




produced within a range of plus or minus 12 percent. Due to




agglomeration, a limited number of aggregates and chains of particles




were formed thereby increasing the mean effective aerodynamic size




of the particle distribution.  Even so, a conservative  examination




of the mass distribution data obtained from the Andersen Stack




Sampler particle sizing technique indicates approximately 75 percent




of the particles have an effective diameter less than 0.1 micrometers.




If the agglomerating effects prove undesirable for a particular study,




an ion generator or polonium grid could be employed to  neutralize the




charged particles.  The rate of particle production is  believed to be




essentially uniform as the consumable wire was fed at a constant rate




and the power supply was regulated at a constant voltage.  It is felt




the feedstock-to-particulate conversion ratio of 4.75 percent could




be significantly improved through optimization of the main parameters:




arc voltage, wire feed rate and wire diameter.  The interior dimensions




of the electrode holder coupled with an erosion rate of 0.2 inches




(5.08 mm) per minute limited length of continuous production to




                               1-52

-------
approximately 30 minutes.  Optimization of the above parameters




complemented by improved shielding of the tungsten electrode should




serve to reduce the erosion rate of the relatively non-consumable




tungsten, thereby increasing the time interval between start-up and




replacement of the tungsten electrode.  A redesign of the electrode




holder through addition of an automatic advancing mechanism capable




of accepting longer tungsten electrodes could greatly improve the



endurance of the generator.




   The morphology of the iron oxide generated in this study closely




approximates the characteristics of emissions from various heavy




industrial processes:  open hearth and basic oxygen steel making,




scarfing, welding, plasma and oxygen-acetylene slab cutting, and of




course, the electric arc furnace.  As the behavior of collected and




re-dispersed iron oxide particulate matter varies significantly from




that of "fresh" iron oxide fume, the primary aerosol formation tech-




nique demonstrated here may prove invaluable for control equipment




research.  The practical implications of a device such as this are




much more far-reachiag. With no modification to the basic generator




apparatus, essentially all metals and metal alloys could be vapor-




ized in the electric arc producing oxides of most metals.  Among




hazardous metals of current interest because of their physiological




effects are lead, chromium, beryllium, vanadian, tin and copper.




To further increase the research applications, special cored wires




can be produced to introduce additional contaminants into a system




at a controlled rate along with a primary metal fume.  The basic




system is also amenable to introduction of gaseous constituents for





                               1-53

-------
study of possible or known synergistic relationships.




   With modification, the unit could also be adapted for the production




of aerosols of non-conductive refractory materials.  This could be




accomplished by replacing the wire torch with another non-consumable




electrode holder, establishing an arc between them, and guiding the




non-conductive feedstock into the high-intensity arc.  Conversely,




elimination of the tungsten electrode as one side of the arc could be




accomplished by utilizing two wire torches and establishing an arc




between the consumable wire "electrodes".  This modification will




inherently eliminate the problems of tungsten electrode erosion




(requiring periodic replacement) and potential contamination of the




solid phase of the aerosol.  Ability of the generator to operate




continuously would also be greatly enhanced.
                               1-54

-------
                        LIST OF REFERENCES
(1)  Lodge, James P., "Production of Controlled Test Atmospheres",
     Air Pollution. Arthur C. Stern, ed., Vol. 2, Academic Press,
     New York (1968) pp. 465-483.

(2)  Amdur, Mary 0., "The Physiological Response of Guinea Pigs to
     Atmospheric Pollutants", International Journal of Air Pollution,
     Vol. 1 (1959) pp. 170-183";

(3)  Fuchs, N. A., The Mechanics of Aerosols. The Macmillan Company,
     New York (1964) p. viii-x.

(4)  Silverman, Leslie, "Experimental Test Methods", Air Pollution
     Handbook. Paul L. Magill, Francis R. Holden and Charles Ackley
     eds., McGraw-Hill, New York (1956) pp. 12-1-12-2.

(5)  Fuchs, op. cit.. pp. 367-377.

(6)  Ibid, p. 374.

(7)  Silverman, op. cit.. p. 12-20.

(8)  Udin, Harry, Funk, Edward R., and Wulff, John, Welding for
     Engineers, John Wiley & Sons, Inc., New York  (1954) p. 141.

(9)  Holmgren, J. D., Gibson, J. 0., and Sheer, C., "Some Character-
     istics of Arc Vaporized Submicron Particulates", Ultra Fine
     Particles. W. Kuhn, ed., John Wiley, New York  (1963) p. 130.

(10) Amick, James and Turkevich, John, "Electron Microscopic Examina-
     tion of Aerosols Formed in a Direct Current Arc", Ultra Fine
     Particles. W. Kuhn, ed., John Wiley, New York  (1963) p. 147.

(11) Whitby, K. T., Lundgren, D. A., and Petersen, C. M., "Homogeneous
     Aerosol Generators", International Journal of Air and Water
     Pollution, Vol. 9 (1965) p. 265, 276.

(12) Silverman, Leslie and Billings, Charles E., "Methods of Generating
     Solid Aerosols", Journal of Air Pollution Control Association,
     Vol. 76, No. 6, (1956) p. 82.

(13) Ibid, pp. 77 ff.

(14) Silverman, "Experimental Test Methods", op. cit., pp. 12-19-12-38.

(15) Olive, T., "Solids Feeders", Chemical Engineering. Vol. 59,
     No. 11 (1952) pp. 163-178.


                               1-55

-------
(16)  Fuchs,  op.  cit.  chapter VIII.

(17)  Silverman,  "Methods  of Generating  Solid Aerosols", op.  cit.
     p.  77.

(18)  Silverman,  "Experimental Test Methods", op.  cit., pp.  12-20,
     12-22-12-23.

(19)  Holmgren et.  al.,  op. cit.. p.  129.

(20)  Harris, V., Holmgren, J. D., Korman,  S.,  and Sheer,  C.,
     "Arc Decomposition of Rhodonite",  Electrochemical Society  Journal.
     Vol. 106, No.  10 (Oct. 1959) pp. 874-876.

(21)  Tafa Division, Humphreys Corporation,  180 North Main Street,
     Concord, New  Hampshire 03301.

(22)  Karioris, Frank  G.,  and Fish, Birney  R.,  "An Exploding Wire Aerosol
     Generator", Journal  of Colloid  Science. Vol. 17,  (1962)  pp. 155-161.

(23)  Silverman,  "Methods  of Generating  Solid Aerosols", op. cit..  p.  81.

(24)  Davis,  R. J.,  Truitt, J.,  and Gill, J. S., "Aerosol  Studies in the
     Reactor Safety Program at  Oak Ridge National Laboratory",  Journal
     of  the  Air  Pollution Control Association, Vol.  18, No. 10
     (Oct. 1968),  pp.  675-676.

(25)  Truitt, J.  and Davis, R. J., "The  Function of Condensing Steam
     in  Aerosol  Scrubbers", Reactor  Chemistry  Division, Oak Ridge
     National Laboratory, Oak Ridge, Tennessee, unpublished report
     of  research sponsored by the National Air Pollution  Control
     Administration (1970).

(26)  "Operation  Instructions for Electric  Motor Driver Constant Voltage
     Welder  for  Automatic Welding, Model MC-900, Specs 4136, 4210".
     Hobart  Brothers  Company, Troy,  Ohio.

(27)  "Hobart Welding  Wire and Flux", Bulletin  EW-382 (NWSA-630).
     Hobart  Brothers  Company, Troy,  Ohio.

(28)  "Installation-Operation-Maintenance Manual for Hobart Automatic
     Welding Head,  Model  AI-22, Specs 3974A."  Hobart Brothers Company,
     Troy, Ohio.

(29)  Welding Data  Book. Robert  N. Williams, ed., Industrial Publishing
     Company, Cleveland,  Ohio  (1967) pp. A/72-A/73.

(30)  Rose, Arthur  and Rose, Elizabeth,  eds. The Condensed Chemical
     Dictionary. Van  Nostrand Reinhold Company (1967) p.  982.


                               1-56

-------
(31)  Yeates, Parrel, et. al.. "Calibration and Evaluation of the
     Andersen Stack Sampler", Technical Report No. FY70 MCR-1,
     submitted to Two Thousand, Inc., Salt Lake City, Utah
     (April 1970) p. 1.

(32)  Andersen, A. A., "A Sampler for Respiratory Health Hazard
     Assessment", AIHA Journal. Vol. 27 (March-April 1966) pp. 160-165.

(33)  Welding Data Book, op. cit.. p. A/79.

(34)  Rose, op. cit.. p. 412.

(35)  Cadle, Richard D., Particle Size. Reinhold Publishing Corporation,
     New York (1965) pp. 20-22.

(36)  Freund, John E. and Williams, Frank J., Modern Business Statistics,
     Prentice-Hall, Englewood Cliffs, New Jersey  (1958) pp. 60 ff.

(37)  Ibid., p. 62.

(38)  Ibid., pp. 66-68.

(39)  Vomela, R. A. and Whitby, K. T., "The Charging and Mobility of
     Chain Aggregate Smoke Particles", Journal of Colloid and Interface
     Science. Vol. 25, No. 4 (December 1967) pp.  568-576.
                               1-57

-------
                        APPENDICES

APPENDIX A   Selected Bibliography of Readings Concerning
             Electric Arc Discharges
APPENDIX B   Developmental Background of the Tungsten
             Electrode Holder
APPENDIX C   Operating Procedure for Arc Aerosol Generator
APPENDIX D   Electron Microscope Particle Count Raw Data
APPENDIX E   Statistical Calculations
                           1-58

-------
                           APPENDIX A
Selected Bibliography of Readings Concerning Electric Arc Discharges

Aidala, Joseph B., "Electric Heating", in Chemical Engineers' Handbook,
      Robert H. Perry at. at., eds., McGraw-Hill Book Co., New York
      (1963) pp. 25-41—25-42.

Fleming, John Ambrose, "Electric Lighting", in Encyclopaedia Britan-
      nica, llth ed., vol. 16, Encyclopaedia Britannica Inc., New
      York  (1910) pp.  659-666.

Jackson, Clarence E., "Welding", in Encyclopedia Americana, vol. 28,
      Americana Corp., New York  (1969) pp.  599-603.

Rose, David John, "Conduction  [of Electricity] in Gases", in
      Encyclopedia Britannica, vol. 8, Encyclopedia Britannica Inc.,
      Chicago  (1970) pp.  206-208.

Thomson, Joseph John, "Electric  Conduction", in Encyclopaedia
      Britannica, llth ed., vol. 6, Encyclopaedia Britannica Inc.,
      New York (1910) pp.  884-887.

Udin, Harry, Funk, Edward R. and Wulff, John, Welding for Engineers,
      John  Wiley and Sons Inc.,  New York  (1954) pp.  114-115 and
      136-169.
                               1-A-l

-------
                            APPENDIX B



                Developmental Configurations of the
            Arc Chamber and Electrode Holder Apparatus

   Thoughts of utilizing an electric arc to create large, reproducible

quantities of ultra-fine particles was precipitated by the author's

experience with various welding and arc-cutting techniques, and

knowledge of the characteristics of the resulting process emissions.

Tungsten inert gas  (TIG) welding techniques were first developed in

the 1930's and have been refined for primary use in the aircraft     *

industry for welding magnesium and aluminum.  A refractory,

"non-consumable", tungsten electrode is used to initiate an arc

between the base metal and itself, and filler wire is manually fed

into the molten puddle.  Helium and/or argon is directed at the weld

area to reduce oxidation of the tungsten electrode and molten metal

in the weld.  Tungsten arcs are also used as a heat source for cutting

of metals.  Another relatively recent development in the welding

industry is the metal inert gas (MIG) process.  It was not until 1948

that the first consumable electrode welding equipment was patented.

This technique has now seen wide use in both automatic and semiauto-

matic applications.

   The electric arc aerosol generator apparatus discussed in this  report

is a combination of equipment from these two welding processes.

An automatic wire feed and torch assembly is used to feed consumable

wire into an arc established between itself and a tungsten

electrode using the constant potential MIG welder to supply the

electromotive force.  In order to establish the feasibility of

                                1-B-l

-------
proceeding with the design of a semi-continuous model, several

preliminary arrangements were tried.

   Figure B-l, pictorially illustrates, in a simplified manner, the

evolutionary configurations investigated which lead to the design

and construction of the arc chamber and electrode holder used to

produce the results reported in this paper.  Configuration I utilized

a standard, rack mounted consumable electrode welding head and feed

motor assembly (the same units used in the final configuration) to

direct the wire at the tip of a one-eighth inch diameter tungsten

electrode.  The tungsten electrode was held in place be a compression-

type, one-eighth inch brass tubing fitting (Swagelok  Part No. 200-1-2)

bored through to allow the electrode to slide in and out of the

chamber.  By tightening and loosening the compression nut, the fitting

acts as a chuck to lock the electrode in position.  The pipe thread

end of the fitting was place in a tapped hole in the arc chamber on

the same centerline axis as the wire torch.  The electrical hookup

was greatly simplified as the tungsten electrode was placed at

negative potential and the whole arc chamber could be safely and

easily grounded.  On the other side, passage of the wire torch through

the chamber wall had to be insulated.

   This configuration met with only limited success as the rate

of tungsten melting and erosion was excessive, and continuous

operation was limited to a maximum of thirty seconds.  Two changes
         "Swagelok" fittings are manufactured by Crawford Fitting
Company, Solon, Ohio  44139.
                                l-B-2

-------
Figure B-l.
Simplified sketches showing developmental configurations
of the tungsten electrode-arc chamber apparatus
                             CONFIGURATION I
                            CONFIGURATION II
                                  CONFIGURATION III
                               l-B-3

-------
were deemed appropriate to retard erosion and thereby increase




operation time:  (1) increase electrode diameter, and (2) supply




a water-cooled electrode holder to dissipate heat.




   Configuration II used the same wire torch arrangement, but the




non-consumable electrode was increased in size to 0.25 inches in




diameter and placed in a specifically constructed water-cooled holder.




The holder was fabricated from an eight inch length of one inch stan-




dard pipe with a steel plug welded in one end.  The plug was threaded




and held a compression-type, stainless steel tubing fitting (Swangelok




Part No. 400-1-316), again bored through to act as a water tight chuck




to hold the tungsten electrode.  The other end of the holder contained




inlet and outlet fittings for the cooling water supply.  An aluminum




electrode holder mounting block was machined to fit the curvature of




the chamber and bolted in place in such a manner as to locate the




electrode centerline on the same axis as the consumable wire center-




line.  The holder was held in place by a set screw.




   This configuration operated satisfactorily for up to two minutes




before the build-up of molten wire and tungsten on the compression




fitting forced a shut down.  After melting one of the stainless steel




fittings beyond use, a replaceable steel disc was placed over the end




of the protruding electrode, as shown in Figure B-l, Configuration II,




to protect the fitting and holder body from arc splatter and molten




build-up.  This addition did little to increase length of operation,




as the same problem of build-up of molten metal recurred, this  time




on the disc protecting the holder.






                               l-B-4

-------
   To eliminate the problem of molten metal build-up it became




obvious the arc chamber and its attachments could be rotated 90 degrees




allowing the molten metal droplets to fall harmlessly to the bottom




of the chamber.  This change in orientation required substantial




changes in the chamber mounts and fabrication of a special insulated




mounting system for the feed motor and wire spool.  The arrangement




has been denoted Configuration III in Figure B-l.




   With this change,  length of operation was doubled, but this




only meant continuous operation for a maximum of four minutes.  It




became apparent the limiting factor was excessive tungsten melting




and erosion.  Therefore,  to increase endurance, a holder had to be




designed to allow  advancement of  the electrode into the arc to




compensate for melting  and maintain uniform erosion.  A water-cooled




holder was finally designed and fabricated which satisfactorily incorporated




these features.  This unit, shown in Figure 4, was used successfully to




obtain the results of this report.
                                l-B-5

-------
                             APPENDIX C

 Operating procedure for Arc Aerosol Generator

 1,  Assemble apparatus as shown in Figures 3, 5 and 6, omitting
     inlet duct.

 2.  Insert assembled cathode holder into arc chamber mounting block
     to blue line on holder and tighten set screws.

 3.  Place heat shield over end of cathode holder and tighten set
     screw by inserting alien wrench through view window opening.

 4.  Install shielding gas nozzle through view window opening.

 5.  Advance cathode by turning knob until tip is centered in gas
     nozzle outlet. Refer to Figure C-l.

 6.  Make electrical, shielding gas and water supply connections.

 7.  Turn on power to Control Panel only.

 8.  Set WIRE SPEED control dial to desired setting.  (Refer to
     Figure C-2 for calibration curve for 0.045 inch diameter wire).

 9.  Clip off approximately 1/2 inch from protuding wire to remove
     any oxidized metal.

10.  Place WIRE FEED switch in DOWN position.

11.  Press INCH button until wire (anode) protrudes approximately
     1/2 inch (12.7 mm) from torch tip.  (See Figure C-l).

12.  Adjust arc chamber on its slide mounts until gap between anode
     and cathode is approximately 1/8 inch (3.175 mm) and tighten
     hold down screws. (See Figure C-l).

13.  Spray interior of arc chamber with welding splatter anti-stick
     spray.

14.  Install inlet duct.

15.  Close view window and secure with tape.

16.  Turn on exhaust fan switch.

17.  Turn on main power switch at buss and press START button on
     power supply.

                                 1-C-l

-------
18.  Adjust VOLTAGE control dial on control panel to desired open
     circuit voltage. Voltage is read on power supply voltmeter.

19.  Open valve on shielding gas tank.

20.  Depress PURGE button on control panel and adjust gas flow to
     desired rate.  Release button.

21.  Turn on power to air blower (low speed) and water pump.

22.  Depress PURGE button approximately one second and immediately
     press START button on control panel to initiate arc.

23.  For continuous operation while conducting tests, manually turn
     feed screw on cathode holder to maintain tungsten tip* centered
     in the axis of the shielding gas nozzle outlet.

24.  To cease operating, place WIRE FEE3 switch in OFF position.

25.  Press STOP button on power supply.

26.  Allow shielding gas and cooling water to flow approximately
     5 seconds for each 10 amperes of current flow, e.g. for a
     current flow of 150 amperes, flows should continue for 70
     seconds after the arc is extinguished. Alternately, allow
     cooling water and air to circulate until air temperature
     immediately downstream from arc chamber cools to 100°F (38eC).

27.  Press STOP button on Control Panel.

28.  Close valve on shielding gas tank to conserve gas.

29.  Turn off all power.
                                l-C-2

-------
Figure C-l.   Simplified  t-ross-sectionnl,  p.lan view sketch
              showing location of components  Inside arc chamber
O                                             GAS SO
  Dashed lines denote
    view window opening
                            1-03

-------
    Figure C-2.   Calibration curve of wire speed dial versus
                 wire speed of 0.045 inch diameter wire
 10 _



  9 .



  8 .



  7 .
g
03 r-
a5
  4 -
  3 -
  2 -
     0
 I            I            I

100         200         300

       Wire speed,  inches/minute
I

400
500
                       l    I     1    I    i    i    I

                            5                       10

                       Wire speed,  meters/minute
                            1-0-4

-------
Table D-l.  Electron microscope particle count raw data
W Q
. O i-J

3
4
2
3
2
2
2
3
0
0
1
0
5
7
4
3
5
0
TOTAL
372
493
262
354
267
196
315
668
322
235
200
416
539
620
637
726
237
316
                                                                             M
                                                                             X

-------
                        Table D-l  (continued).  Electron microscope particle  count raw data
W Q
O i-J

2
2
1
1
3
5
5
0
1
2
5
1
74
2
7
7
TOTAL
304
247
177
205
368
411
723
366
470
651
516
742
12355
565
1236
412
NJ

-------
                           APPENDIX E
Table E-l.  Transformation for calculating the arithmetic mean and
            standard deviation from data of Table 4
Class mark
di

0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.10
frequency
f

5726
2663
1908
907
531
292
144
76
34
74
u

-1
0
1
2
3
4
5
6
7
8
uf

-5726
0
1908
1814
1593
1168
720
456
238
592
u2f

5726
0
1908
3628
4779
4672
3600
2736
1666
4336
                   n=12,355
22763    233,451
                               1-E-l

-------
                            APPENDIX E
Table E-2.   Tabulation for calculating median  an  geometric
            mean from data of  Table 4
Class
boundaries
L

.005
.015
.025
.035
.045
.055
.065
.075
.085
.095
U

.015
.025
.035
.045
.055
.065
.075
.085
.095
.105
Class mark
Xi

0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.10

log x±

"2.00000
"2 . 30103
2". 47712
1.60206
1.69897
1.77815
I. 84510
1.90309
1.95424
I. 00000
frequency
f

5726
2663
1908
907
531
292
144
76
34
74
Cumulative
frequency


5726
8389
10297
11204
11735
12027
12171
12247
12281
12355

-log x±(f)

11452.0
6127.6
4726.3
2360.1
1433.2
811.1
409.7
220.6
100.4
74.0
                               n=12,355
227,715.1
                           l-E-2

-------
                     APPENDIX E
Table E-3.  Calculation of arithmetic mean and standard
            deviation of the relative frequency within each
            particle size interval from the data of Table 5

X
i
46.3
43.0
54.3
50.5
43.0
46.3
47.4
44.3
48.0
42.8
465.9
22.0
22.4
19.1
19.2
23.3
22.3
20.5
21.2
21.6
22.1
213.7
13.6
17.7
12.3
14.4
17.1
16.4
17.3
15.1
14.9
15.7
154.5

X

46.6










21.4










15.5










x.-x
i
-0.3
-3.6
7.7
3.9
-3.6
-0.3
0.8
-2.3
2.0
-3.8

0.6
1.0
-2.3
-2.2
1.9
0.9
-0.9
-0.2
0.2
0.7

-1.9
2.1
-3.2
-1.1
1.6
0.9
1.8
-0.4
-0.6 0.
0.2

_
(x -x)
i
0.09
12.96
59.29
15.21
12.96
0.09
0.64
5.29
4.00
14.44
124.97
0.36
1.00
5.29
4.84
3.61
0.81
0.81
0.04
0.04
0.49
17.29
3.61
4.41
10.24
1.21
2.56
0.81
3.24
0.16
0.36
0.04
26.64

s

3.726










1.386










1.720









s
~
X
0.0799










0.0647










0.1109











-------
Table E-3 (continued).
Calculation of arithmetic mean and standard
deviation of the relative frequency within
each particle size interval from the data
of Table 5

x
i
7.7
6.9
7.0
6.6
8.7
5.3
6.6
8.8
6.3
8.3
72.2
4.5
4.5
3.5
4.2
3.1
3.7
3.4
5.1
6.2
4.5
42.7
2.1
2.6
2.1
2.2
2.2
2.6
2.7
2.1
1.3
3.6
23.5

X

7.2










4.3










2.4










x -x
i
0.5
-0.3
0.2
0.6
1.5
-1.9
-0.8
1.6
-0.9
1.1

0.2
0.2
-0.8
-0.1
-1.2
-0.6
-0.9
0.8
1.9
0.2

-0.3
0.2
-0.3
-0.2
-0.2
0.2
0.3
-0.3
-1.1
1.2

_
(x -x)
i
0.25
0.09
0-04
0.36
2.25
3.61
0.64
2.56
0.81
1.21
11.81
0.04
0.04
0.64
0.01
1.44
0.36
0.81
0.64
3.61
0.04
7.63
0.09
0.04
0.09
0.04
0.04
0.04
0.09
0.09
1.21
1.44
3.17

s

1.146










0.921










0.593









s
"=
X
0.1591










0.214










0.247










                              l-E-4

-------
Table E-3 (continued).
Calculation of arithmetic mean and standard
deviation of the relative frequency within
each particle size interval from the data
of Table 5

X.
i
1-5
1.5
0.9
1.1
0.7
2.3
0.7
1.1
0.6
1.4
11.8
1.3
0.4
0.2
0.8
0.6
0.3
0.3
1.0
0.4
0.8
6.1
0.1
o.2
0.1
0.6
0.3
0.2
0.4
0.2
0.3
0.4
2.8

X

1.2










0.6










0.3










x —x
1
0.3
0.3
-0.3
-0.1
-0.5
1.1
-0.5
-0.1
-0.6
0.2

0.7
-0.2
-0.4
0.2
0.0
-0.3
-0.3
0.4
-0.2
0.2

-0.2
-0.1
-0.2
0.3
0.0
-0.1
0.1
-0.1
0.0
0.1
<

/„ _x\ ^
i
0.09
0.09
0.09
0.01
0.25
1.21
0.25
0.01
0.36
0.04
2.40
0.49
0.04
0.16
0.04
0.00
0.09
0.09
0.16
0.04
0.04
1.15
0.04
0.01
0.04
0.09
0.00
0.01
0.01
0.01
0.00
0.01
0.22



0.516










0.357










0.156









s
""^
X
0.430










0.596










0.521










                              l-E-5

-------
Table E-3 (continued)
Calculation of arithmetic mean and standard
deviation of the relative frequency within
each particle size interval from the data
of Table 5
xi
0.8
0.9
0.4
0.1
0.9
0.6
0.7
0.9
0.4
0.4
6.1
X
0.6









x.-x
0.2
0.3
-0.2
-0.5
0.3
0.0
0.1
0.2
-0.2
-0.2

(x..-x)2
0.04
0.09
0.04
0.25
0.09
0.00
0.01
0.04
0.04
0.04
0.64
s
0.267









s
X
0.444










                                             210.788  2.867

                                          Mean 1.079  0.287
                            l-E-6

-------
                          APPENDIX E
Table E-4.  Calculation of arithmetic mean and standard deviation
            of the cumulative relative mass distribution data
            of Table 7
STAGE
1

2

3

4

5

6

7

Xi
98.13
98.79
98.42
295.34
97.83
98.55
98.09
294.47
97.53
98.39
97.83
293.75
96.93
98.23
97.50
292.66
95.95
97.66
96.51
290.12
93.92
95.16
94.34
282.42
87.84
88.94
88.82
265.60
x
98.45

98.26

97.92

97.55

96.71

94.47

88.53
x -x
-0.32
0.34
-0.03

-0.43
0.29
-0.17

-0.39
0.47
-0.09

-0.62
0.68
-0.05

-0.76
0.95
-0.20

-0.55
0.69
-0.13

-0.69
0.41
0.29

— 2
.1024
.1156
.0009
0.2180
.1849
.0841
.0289
0.2979
.1521
.2209
-0081
0.3811
.3844
.4624
.0025
0.8493
.5776
.9025
.0400
1.5201
.3025
.4761
.0169
0.7955
.4761
.1681
.0841
0.7283
s
0.330

0.386

0.437

0.652

0.872

0.631

0.603
s
X
0.0033

0.0039

0.0044

0.0066

0.0090

0.0066

0.0068

                           l-E-7

-------
Table E-4 (continued).
Calculation of arithmetic mean and
standard deviation of the relative mass
distribution data of Table 7
w
a
CO
8



F



Xi
75.53
79.89
77.38
232.80
68.77
69.55
68.57
206.89
X
77.60



68.96


x -x
-2.07
2.29
-0.22

-0.19
0.59
-0.39

(x^x)2
4.2849
5.2441
.0484
9.5774
.0361
.3481
.1521
0.5363
s
2.188



0.518


s
X
0.0281



0.0075



                            l-E-8

-------
                           APPENDIX E
Table E-5.  Calculation of arithmetic mean and standard deviation
            of weight  (mass) distribution from data of Table 8
w
o
3 g
Cfl Pd
0
1
2
3

1
1
2
3

2
1
2
3

3
1
2
3

4
1
2
3

5
1
2
3

6
1
2
3

Xi
2.5
1.5
2.4
6.4
0.4
0.3
0.5
1.2
0.4
0.2
0.4
1.2
0.8
0.2
0.5
1.5
1.3
0.7
1.5
3.5
2.7
3.1
3.3
9.1
8.1
7.7
8.4
24.2
X
2.13
x. -x
0.37
-0.63
0.27

0.40
0.00
-0.10
0.10

0.33
0.07
-0.13
0.07

0.50
0.30
-0.30
0.00

1.17
0.13
-0.47
0.33

3.03
-0.33
0.07
0.27

8.07
0.03
-0.37
0.33

(Xj-x)
0.1369
0.3969
0.0729
0.6067
0.0000
0.0100
0.0100
0.0200
0.0049
0.0169
0.0049
0.0267
0.0900
0.0900
0.0000
0.1800
0.0169
0.2209
0.1089
0.3467
0.1089
0.0049
0.0729
0.1867
0.0009
0.1369
0.1089
0.2467
s
0.4496
s
X
0.2110

0.0812
0.2030

0.0943
0.2857

0.2449
0.4898

0.3398
0.2904

0.2493
0.0822

0.2867
0.0355

                             l-E-9

-------
Table E-5 (continued).   Calculation of arithmetic mean and standard
                        deviation of weight (mass) distribution from
                        data of Table 8
w
o

-------
              VERY  FINE  PARTICLE

          GENERATION BY  ELECTRIC  ARC:

             SAMPLING AND ANALYSIS
                    PROBLEM

       Submitted to the Graduate School
                      of
           West Virginia University
In Partial Fulfillment of the Requirements for
the Degree of Master of Science in Engineering
                      by

              Mike Naylor, B. S.

                  Morgantown,
                West Virginia

                     1976
                       2-i

-------
                          ABSTRACT






       Established methods of generating useful reproducible con-



centrations of very fine particles for use in research possess



various limitations.  The development of an improved electric arc



metal generating method used a new apparatus:  a commercial electric



arc metallizer.  It is capable of generating very fine particles of



the metals available in the form of metallizing or welding wire.



The generation, sampling and analysis of zinc oxide particles are



examined.



       In addition to studying the effects of changing the arc



metallizer's operating variables of air spray pressure and rate of



wire consumption, the reproducibilities of total mass concentration



and distribution of particle diameters were tested.



       Particles were examined on Nuclepore  0.2 (j,m membrane filters




by electron microscopy and electron diffraction.  Particles were



counted by diameter.  Two replicate samples for each of three arc



metallizer operating conditions were analyzed.  One condition is




not reproducible with respect to the mean diameter of sampled parti-




cles.  The other two conditions are reproducible with respect to




diameter means and variance.  One of these conditions also has




particle diameter distributions that are shown not significantly




different.  The mean diameters of the conditions are significantly




different, indicating that by changing the air spray pressure and




wirefeed rate, the particle diameters are changed.




       The diameters of the particles generated at specific  conditions




                               2-ii

-------
are distributed approximately normally.  Mean diameters range from


6 to 11 nm  (0.006 to 0.011 |j,m).


       Virtually all of the very  fine particles agglomerated,


forming interconnecting chains.   This phenomena apparently resulted


from combined effects of  the very fine particle diameters and the


large initial concentration of the particles.


       Mass concentrations were measured with Millipore 0.45 |j,m


membrane filters.  Four different operating  conditions were defined,


with three  of them being  the same as the ones used  for particle


sizing.  These yield mean mass concentrations ranging from 0.65

             3
to 2.0 gram/m.


       For  three of these conditions, the  replicate mass concentra-


tion coefficient of variation  (C.V-) is less than 9%.  The fourth


has a C.V-  of 19%.  This  is the same condition for  which the particle


diameter means are not reproducible.  The  mass concentrations gen-


erated at different conditions are found significantly different by


testing.  The mass concentration  is increased by increasing the air


spray pressure or the rate of  wirefeed  or  both.


       Since the influence of  the operating  conditions has been


demonstrated as well as varying levels  of  reproducibility for each


condition,  this metallizer is  considered worthy  of  continued develop-


ment and potential utilization by researchers as a  device for


generating  very fine particles.
                               2-iii

-------
                      ACKNOWLEDGEMENTS






       The investigator is grateful to his graduate advisor,  Pro-




fessor Benjamin Linsky, who introduced him to the topic of fine




particle generation.   He offered suggestions and encouragement in




preparing for and writing this report.




       For their research assistance, appreciation is extended to




Dr. D. R. Sears, then Director of the Air Pollution Laboratories at




WVU, and to graduate students, John Garbak, Don Stone and Fred Dimmick.




The investigator thanks Fran Culler of the Civil Engineering Labora-




tory and Don Garletts of the Mechanical Engineering Laboratory for




fabricating equipment parts.




       The investigator acknowledges the suggestions from Charles




Ghastin and Ed Morantz, sales representative for and developer of,




respectively,the Wall Colmonoy electrospray metallizer.




       The investigator expresses thanks to Richard Shimps of McCrone




Associates for his advice on obtaining appropriate samples for electron




microscopy.



       The investigator appreciates the critical review of the sta-




tistical analysis by Dr. E. Z. Damewood of the Industrial Engineering




Department at W.V.U.




       Several types of financial assistance made  this research




possible.  The metallizer and electron microscopy  services were




funded by Environmental Protection Agency  Research grant  R-801858-01-l>




which was administered by Bennis Drehmel.   The  investigator  received




a study traineeship from EPA  guant T-90Q487-10-0,and tuition compensa-




tion for one semester from  the  Civil  Engineering Department  at WVU.






                              2-iv

-------
                         TABLE OFCONTENTS



  ABSTRACT	2-ii

  ACKNOWLEDGEMENTS	2~iv

  LIST OF FIGURES	2-viii

  LIST OF TABLES	2-ix

  CONVERSION FACTORS, EQUIVALENTS, AND ABBREVIATIONS  	   2-x

  I.  INTRODUCTION	   2-1

 II.  REVIEW OF LITERATURE	   2-3

III.  EXPERIMENTAL	   2-6

      A,  GENERATION	   2-6

      B.  SAMPLING	   2-7

          1)  General Considerations 	   2-7

          2)  Description of Sampling Components	2-11

          3)  Filters	2-12

          4)  Calculation of Nucleopore Efficiency 	  2-15

          5)  Sampling Procedure for Mass Concentration  ....  2-16

          6)  Sampling Procedure for Electron Microscopy ....  2-21

          7)  Sensitivity of Results to Deviations from
              Isovelocity Sampling	  .  2-22

      C.  ANALYTICAL TECHNIQUES	2-24

          1)  Mass Concentration Analytical Techniques 	  2-24

          2)  Particle Sizing and Identification Techniques  .  .  2-25

      D.  SELECTION OF EQUIPMENT OPERATING CONDITIONS FOR
          SAMPLING	2-26

 IV.  RESULTS	2-28

      A.  BRIEF EVALUATION OF EQUIPMENT OPERATION  	  2-28

      B.  SAMPLE NUMBER IDENTIFICATION 	   2-29


                                2-v

-------
                                                                   Page




          1)  Mass Concentration	,	    29




          2)  Electron Microscopy and Electron Diffraction ...    29




      C.  CHEMICAL COMPOSITION 	    29




          1)  Electron Diffraction Data	    29




          2)  Assay of Wire	    30




      D.  MASS CONCENTRATION	    31




          1)  Means Testing	    32




          2)  Analysis of Variance	    32




      E.  PARTICLE SIZING BASED ON ELECTRON MICROSCOPY 	    33




          1)  Analysis	    36




          2)  Computation of Means and Variances	    36




          3)  Reproducibility Among Replicates 	    37




          4)  Differences Between Conditions 	  .    37




          5)  Frequency Distributions  	    38




              i)  Computation of Normal Distribution  	    47




              ii) Goodness of Fit	    47




      F.  AGGLOMERATION	    48




      G.  COMPARISON WITH HEDDEN'S RESULTS	    50




  V.  A.  CONCLUSIONS	    52




          RECOMMENDATIONS	    53




 VI.  LIST OF REFERENCES	    56




VII.  APPENDICES




      Appendix A	    A-l




      Appendix B	    B-l




      Appendix C	    C-l






                                  2-vi

-------
                                                           Page




Appendix D	,	   2-D-l




Appendix E	   2-E-l




Appendix F	   2-F-l




Appendix G	   2-G-l




Appendix H	   2-E-l
                            2-vii

-------
LIST OF FIGURES
Figure
1
2
3
4
5
6
7
8
9
10
11
12
13
14
H-l
H-2
H-3

Schematic of EPA-RAC Sampling Train 	 . , .
Actual Stack Sampling Flow System 	 	 	
Schematic of Sampling Point and Venturi Meter ....
Approximate Relationships of Sampling Nozzle ....
Typical Form for Mass Concentration Data 	
Plot of Line of Equal Bias as Function of R and K . .
Comparison of Cumulative Frequencies for Condition 2
Comparison of Cumulative Frequencies for Condition 3
Observed Frequency of Sample 134 and Normal
Frequency ....... 	 .
Observed Frequency of Sample 133 and Normal
Frequency 	
Arithmetic Probability Plot of Sample 128 Diameters .
Log Probability Plot of Sample 128 Diameters ....
SEM Photograph of Sample 133 (5000 X) 	
SEM Photograph of Sample 133 (10,000 X) 	
TEM Photograph of Sample 101 (300.000 X) 	
Page
2-8
2-9
2-10
2-13
2-14
2-18
2-19
2-23
2-41
2-42
2-43
2-44
2-45
2-46
2 -H-2
2 -H-2
2-H-3
        2-viii

-------
                       LIST OF TABLES


Table
                                                                 Page
  1    Summary of Nuclepore  Impaction, Interception,
       Diffusion and Total Collection Efficiencies  ......  2-16

  2    Summary of Possible Sampling Bias  ...........  2-23

  3    Zinc Wire Sampling Conditions .............  2-27

  4    Identification of Mass  Concentration Samples   .....  2-29

  5    Identification of McCrone  Samples  ...........  2-29

  6    Electron Diffraction  Data-Sample 104   .........  2-30

  7    Diffraction Data  — Sample 133   ............  2-30

  8    Standardized  Zinc Oxide Mass Concentrations  for
       All Samples  ......................  2-31

  9A  Zinc Oxide Mass Concentration Mean and Variation Summary 2-31

  9B  Mean Mass Concentration Matrix   ............  2-32

  10A  Zinc Oxide Particle Size Data  .............  2-34

  10B  Zinc Oxide Particle Size Data  .............  2-35

  11    Means  and Variances of  Zinc Oxide  Particle Size Data  .  2-37

  12    Cumulative Observed and Expected Frequencies for
       Condition 2  ......................  2-39

  13    Cumulative Observed and Expected Frequencies for
       Condition 3  ......................
  14    Hedden-Naylor Data  ..................   2~50
                               2-ix

-------
      CONVERSION FACTORS. EQUIVALENTS AND ABBREVIATIONS



acfm = actual cubic feet per minute


acmm = actual cubic meters per minute


°C = (°F - 32)/1.8


1 cm = 0.03281 ft.


cm/sec =» 1.97 fpm


1 cubic meter (m3) = 35.315 cubic feet  (ft3)


g/acf - gram/actual cubic feet

    3
g/sm  = gram /standard cubic meter


1 grain = 0.0648 gram


1 grain = 1/7000 pound


grain/acf = grains/actual cubic foot

                           3
2.29 grains/scf = 1.00 g/sm


HEPA = high efficiency particulate air


mg/acf = milligrams/actual cubic foot

             -3            -7
1 nm = 1 x 10   |j,m - 1 x 10   cm


SAED = selected area electron diffraction


SEM - scanning electron microscopy


scfm = standard cubic feet/minute

  3
sm  = standard cubic meters


TEM - transmission electron microscopy

           -4        +3
|j,m = 1 x 10   cm = 10   nm
                              2-x

-------
                        INTRODUCTION




       This report examines a method of producing in the laboratory


reproducible concentrations of freshly made very small, metal oxide


particles.  The method represents a continuation of a long range pro-


ject of the Air Pollution Engineering Laboratory, the goal of which


is to develop a valid research tool to generate very fine freshly


made particles for extended periods of time and to characterize these


particles.  The results of this project are applicable to the fields


of respiratory, control equipment, instrument design, vegetation and


cloud physics research.


        The particle generator simulates  general industrial  electric


 arc processes.   The concentrations of particles generated may be


 typical of those to which workmen are exposed.


        A coonercially available electric arc metallizer has been


 four.i. that is potentially capable of delivering reproducible very


 sr^i.11 particles evolved from at least a dozen different metals.


 Metals in wire form are utilized by the machine as consumable elec-


 rr^ces.  "The wire passes through the high heat zone of the arc,


 beccning molten.  Air pressure then sprays the molten particles onto


 tie surface being coated."*1 '


        The process evolves metal vapor at high temperatures, which cools


 in an air stream and condenses  into very fine particles commonly


 called metallurgical smoke or fresh metal fume.  The nucleation of


 vapor is facilitated by the intense concentration of ions  produced


 around the arc.  Most of the original particles are believed to be

                                                                (2)
 formed by condensation on ions rather than by self nucleation.



                               2-1

-------
The unit pMrtlcJ.es are electrically charged.  The complexes consist


of long chains, frequently of hundreds of fine particles, most of


which are too small to be resolved by the optical microscope so


that electron microscopy must be used to observe their structure


and form; electron diffraction shows the composition and nature of

                                                             (2)
their surface regions more sensitively than Xray diffraction.  '


       The particles generated were analyzed for mass concentration


and characterization, and distribution of sizes.


       A problem report describing the historical development of


the equipment in this project is presently being written by Don


Stone.
                              2-2

-------
                    REVIEW OF LITERATURE


       The entire subject area of very fine particles is growing

rapidly.  Those working in the field know that the published litera-

ture has fallen far behind the real developments.  This was clearly

depicted during the Aerosol Measurement Workshop at Gainesville,

Florida, in March 1976.

       This literature review covers a few topics relating to very

fine particle generation, sampling and analysis.

       For years, Professor Linsky has pushed for recognition of

small particles in air pollution separate from total emitted par-

ticulates.  In 1958, he presented a paper explaining the need for
                                                            /3\
such a listing by particles sizes in an emissions inventory.     He

has also pointed out the lack of acceptable methods for producing

fine fresh metal particles as distinguished from "stale particles."

                     (4)
        Robert Redden    developed a series of modifications of gene-

 rators of very fine particles that utilized an electric arc welder

 with a consumable wire feedstock as an anode and a relatively non-

 consumable  tungsten cathode.

        Redden"s literature review, which was not comprehensive,dis-

 cussed the following limitations in particle generation to date (1972)

        1)  The homogeneous aerosol generators yield con-
            centrations that are too dilute for air pollu-
            tion control device research and development.

        2)  Higher concentrations can be achieved by re-
            dispersing large quantities of previously pro-
            duced and collected powder but these particles
            lack the quality being "fresh" and are agglom-
            erated .
                               2-3

-------
       3)  Direct attrition methods — friction, evapora-
           tion, impact, explosion and combustion approach
           the situations in the industrial environments
           but do not generally achieve particles smaller
           than one micrometer.  Submicrometer ranges avail-
           able are not reproducible due to control of feed-
           stock and input energy.

       4)  Electric arc generators have been used pre-
           viously but particle characterization was given
           little consideration

       5)  Finally, a. high voltage continuous discharge
           method lacks the features of high concentration
           and runs lasting an hour or more.' '

       Hedden's equipment provided data which established reprodu-

cibility of mass concentration.  For a particular set of operating
                                                                3
conditions he indicated an average concentration of 0.65 gram/sm

(1.49 grain/scf) at standard conditions of 25°C and 760 mm.  He also

studied particles collected on one filter by electron microscopy.

Due to limitations of time and money, he was not able to investigate

the reproducibility of particle diameter sizes by count nor the in-

fluence of operating parameters on mass concentration and particle

size.  He used a carbon steel welding wire.  Runs were limited to

30 minutes as the tungsten electrode would be consumed by that time.

       Due to limited facilities, Hedden prepared the samples col-

lected on membrane filters for electron microscopy analysis by dis-

persing the particles on a suitable grid.  The dispersion probably

altered agglomeration characteristics.

       A major revision of Hedden's modifications was developed  in

1973-74 by Hedden, Professor Linsky, Dr. William H. Fischer,  Arland

Johansen, Michael McCawley, Rasool Nekooi, and Dr. Sears.   '  The

revised apparatus was an electric welder which used two  consumable
                              2-4

-------
wires Instead of one.  This machine  suffered  stability  problems due


to inadequate control of  the wires meeting  and  forming  an arc.  Gaps


of greater  than fc  inch resulted in unstable runs.   '  Mass  concen-


trations were measured for varying voltage  and  rates  of wire  feed


but, due to equipment limitations, the data was insufficient  to


establish conclusive results.


       A method was cited by Brain^  '  in 1974 for  producing ferric


oxide by combustion of iron pentacarbonyl,  He  observed concentra-

                                                                 Q
tions in an animal exposure chamber  ranging from 1.00 to 4.00 g/m


and noticed agglomerates  of particles about O.G2(j,m  in  diameter averaging


0.4 (J.m in average  count diameter.  He  said  the  diameters of the


agglomerates were  distributed  log normally.   While he argued  that


the iron oxides provide excellent test particles for  animal exposure,


his technique was  obviously limited  to iron oxide.


       Hedden's size analysis  of iron  oxide particles also  indicated


a log normal distribution of diameters by count.  However,  neither


Hedden nor  Brain used statistical techniques  of goodness to fit


tests to check whether their observed  size  distributions were rea-


sonably close to the generally claimed log  normal  distributions.


       This report examines a  method of producing  metal oxide


particles to determine its acceptability with respect to continuous


generation, influence of  operating conditions and  reproducibility.
                              2-5

-------
                    EXPERIMENTAL METHODS






                         GENERATION




       The metallizes,  manufactured by Flame Spray Industries,     has




many commercial applications.   It is widely used to provide coatings




on metal or to fill in and build up worn  areas on shafts,  etc.   One




use for zinc wire in a continuous production line is to replace the




rust proof coating along the machined or rolled threaded ends of elec-




trical conduit that had  been galvanized  before the thread cutting.




       The metallizer operates in a manner similar to an electric




arc welder except that the product is blown onto another piece of




material that is not connected electrically to the electrodes or




power supply.  The consumable wires are energized by the high




amperage low voltage power source forming a plasma arc through




which current will flow.  The heating volatilizes more wire while




the compressed air upstream dissipates and directs the condensing




particles.  A steady state of an arc generating fine fresh metal-




oxide particles is established.




       The specific metallizer used in this research (Fig.  1)  con-




sists of a power source which converts 3 phase 220 volt alternating




current to direct current that is variable from 0  to 40 volts  and




up to about 400 amps; a console on top of the power source  from




which air pressure, wire feed rate, and open  circuit voltage are




adjusted; flexible conduits which convey wire, from reels  secured




to the console, to a phenolic spray head that  guides the wires  into




fixed position electrode tips.  Electrical cables  transfer  the direct




current from the power source to the wires at the  spray head.
                              2-6

-------
       The spray head was fastened to the lid of a 55 gallon barrel,



with the electrodes extending through a slot into the barrel.  The



arcing takes place in the barrel.  The larger particles of the



metallized wire are directed by  the high pressure air stream to the



other end of the barrel where they are deposited.



       A low pressure air stream entering through a HEPA (high effi-



ciency particulate air filter) and the dispersed compressed air convey




the much smaller particles  through an exhaust duct to a cloth filter



collector and  exhaust blower  (baghouse system).  The entire layout



of generation, exhaust and  sampling is shown in Figure 1.



       The  size of the deposited particles  in  the barrel not drawn



into  the exhaust stream  can be approximately calculated by treating



the barrel  as  a cyclone.   As  estimated by the  formulas and computa-




tions  in Appendix A,  the barrel  bottom would collect 50% by weight




of the particles 7 jj,m in diameter and 95% by weight of those 30 M-m




in diameter.





                           SAMPLING




General Cons iderat ions



       To  assure  the best  representation of the very  fine  particles,




sampling was  done under  iso-velocity conditions.   The  procedure



followed recommendations of the  EPA Method Five^ '  where possible.




The  recommended  sampling train,  as interpreted by Research Appli-




ance  Company,   '  is  indicated in Figure 2.   This was  modified as




shown in Figure  3.



       The  membrane  filter holder was seated on top of the sampling
                              2-7

-------
                                     Phenolic spray head Cross-sect ion

                                  I          E    G F G   E           I
                                  |	
A) Power source  of metallizer
8) Control  console
C) Dual  wire  spools
D) Wire  straightening, drive and conduit feed mechanism
E) Flexible  conduit  for  wire
F) Flexible  compressed  air line - main air
G) High amperage electrical cable
H) 55 gallon barrel
I) Phenolic spray head
J) Dilution air  from  Hepa filter
K) Flexible  compressed air  line - secondary air
L) Exhaust  duct - 6 inch diameter
M) Sampling port
N) Venturi
O) Stack sampler
                   Figure 1.   Layout of equipment
                                2-8

-------
N>
I
vo
      3)
      4)
      5)
      6)

      3
      9)
     1GJ
     11)
     12)
     13)
     14)
     15}
     16)
     17)
     19
     19)
     20)
     21)
Probe
Cyclone "\ ,_..  .   .
Flask   J Eliminators
Particulate filter
Impingers (Greenburg-Smith)
Thermometer
Check valve
Umbilical cord
Vacuum gage
Course  flow adjust valve
Fine flow adjust  valve
Oiler
Vacuum  pump
Filter
Dry gas meter
Orifice tube
Incline manometer
Solenoid valves
Pitot
Thermocouple
Pyrometer
21
                                Figure 2.   Schematic of EPA-RAC sampling train

-------
                         T)16
I
M
O
                                            13
       1) Probe
       2) Filter  holder
       3) Umbilical cord
       4) Impingers
       5) Vacuum gage
       6) Course flow adjust valve
       7) Fine  flow adjust valve
       8) Oiler
       9) Vacuum  pump
      10) Filter
      11} Thermometer
      12) Dry gas meter
      13) Orifice  tube
      14) Incline manometer
      15) Solenoid valves
      16) Wet bulb thermometer
r    \
     14
                                   Figure 3,   Actual stack sampling flow system

-------
port, itself on top of the horizontally positioned duct.  The inten-




tion was for the filter to trap all sizes of particles reaching it,




differing from the arrangement in Figure 2 whereby the first impinger




collects larger particles.  A second difference in procedure was in




the method of moisture measurement.  A wet bulb thermometer was




placed in the duct to determine the dew point.  This was preferred




to the method of monitoring the water levels in the impingers and




measuring the weight gain of the silica gel impinger, as small incre-




mental increases were anticipated.






Description  of Sampling Components




       The sampling train (Figure 3) consisted of  the filter holder,




filter cord  to impingers, impingers, cord to the  sampler, and the




RAG  stack sampler.  Manometers were placed at the  HEPA filter, venturi




meter, and baghouse, in addition to the two contained in the sampler




for  monitoring orifice pressure drop and stack-velocity pressure.




Also, thermometers are placed on the barrel lid near the spray head,




2  inches downstream of the  sampling port, and  at  the baghouse.




       Sampling Port.  A  %  inch opening, 11 diameters from exhaust




 port on lid.



        Probe.  The S shaped straight-inlet nozzle was inserted into




 the air stream 1/3 diameter vertically from the  port entrance.




       Filter holder.  The  Gelman  filter holder  accepts 49 mm filters.




The  nozzle assembly is directly connected  to  the inlet  of  the holder.




       Filters.  Millipore  filters, type HA,  .45 (jtm  pore diameter,




were used to  determine the mass concentrations  which  required collection
                               2-11

-------
times of several minutes.   General Electric Nuclepore  filters were




employed for short duration particle sizing sampling.  The filters




are discussed further in the next sections.




       Impingers.  Moisture in the sampled air was removed by the




series of four 1 liter impingers, chilled by an ice bath.  The first




two impingers each had  100 ml distilled water,  the third was dry




and the fourth 200 grams of dry silica gel.




       RAG Stack sampler.   The RAG Stack sampler Model 2343(5»^




consists of all elements  downstream of the' umbilical eord in




Figures 2 and 3.  It satisfies the specifications of the EPA




Method V.  '




       Venturi Flow meter.  A constriction was built into the ductwork




about 5 diameters downstream from the sampling port  (Figure 4).



                                                 (9)
By a generally accepted stack sampling procedure,    velocity tra-




verses along two perpendicular diameters at the sampling port loca-




tion were tabulated as a function of the pressure drops measured by




the manometer across the venturi.  The velocities were averaged and




corresponding flow rates were calculated.  The data  and  calculations




are presented in Appendix B.  A plot of exhaust flow versus venturi




pressure drop is shown in Figure 5.






Filters




        Nuclepore filters (0.2 (jjn pore diameter) were selected




for determining the distributions of particle diameters  and for




identifying the particles for the following reasons.   Their smooth




surface and better controlled pore size and frequency  make them






                              2-12

-------
                 :M
                    D=6 inches
                    J	
       11D
        •barrel
1) Direction  of air  stream from  chamber
2) Sampling nozzle and filter holder with cord to impinger
3) Thermometer
4) Venturi constriction with taps for static pressure
5) Incline manometer
      Figure 4.   Schematic of sampling point and venturi meter
                               2-13

-------
      450
     400
      350 -
o
     300 -
s:
x
UJ
      250 -
     200
                   .1
.2
.3
.4
.5
           Venturi  pressure drop (inches water)
       Figure 5.   Exhaust flow versus venturi pressure drop
                       2-14

-------
Ideal for analysis by electron microscopy and electron  dlffrac-




tlon-  '  '  '     The Millipore  (Type HA, 0.45 (j,tn pore diameter)



filters were selected for determining the mass concentrations of



the particles sampled.  The Millipore filter's greater thickness



(150 urn) than the Nude pore  filters  (12 ^m) and its random fibrous



structure allows the greater retention volume needed for several



minutes of exposure in the particle stream.




       The added advantage of the Nuclepore  filters, which have



simple straight through pores compared to the more "tortuous path"



configuration of the Millipore membranes, is that they most closely



approximate the popular model of  filters used to predict particle



collection efficiency.    '






Calculation of Nuclepore  Efficiency



       One mathematical model for the study of membrane filters



assumes a flat surface with equidistant parallel straight through



pores.  Since Nuclepore   filters  more closely resemble this model



than other membrane filters the theory has been applied and corrected



                   (14)
by Spurny, et. al.



       Three mechanisms:  interception, diffusion, and impaction,



contribute to the  overall efficiency  Efc.   If the impaction efficiency



is labeled as E.,  the diffusion efficiency as Ed, and the interception



as E , then the respective efficiences are individually calculated




and combined as follows:




       Et =E.+Ed+0.15 Er  -EiEd - 0.15 EI Er




In Appendix C, the individual efficiencies are defined, and  appropriate




                              2-15

-------
values assumed are listed and efficiences are calculated.  The

following table summarizes the computed relationship between effi-

ciences and particle size.
Table 1.  Summary of Nuclepore  Impaction, Interception, Diffusion
          and Total Collection Efficiencies (Pore Diam. = 0.2 (j,m)

Part. Dia.
( fim)
.02
.04
.10
.15
.20
.30
E.

impact.
.00
.00
.012
.028
.05
.447
.15E
r
intercept .
.00
.054
.11
.13
.15
1.00
Ej
d
diffusion
1.00
1.00
1.00
.993
.962
.853
E
t
total
1.00
1.00
1.00
1.00
1.00
1.00
Sampling Procedure for Mass Concentration

       The following steps outline the method followed by the experi-

menters when sampling for mass concentration.  Before sampling:

Measure atmospheric pressure at the equipment location.  Oessicate

the Millipore filters for two days and obtain tare weight to nearest

0.1 mg.

       1)  Equipment should be set up as in Figure 1.  Metallizer

has been running at a specific condition for at  least two minutes

and baghouse has been turned on.  Millipore filter has been placed

into holder.

       2)  Adjust flow at venturi meter.  For a  desired  flow  of 250

acfm (7.08acmm), for example, the manometer should indicate a pressure

difference of 0,2 inches according to Figure 5.

       3)  Measure the velocity pressure of the  exhaust  stream at


                               2-16

-------
the sampling nozzle  location.   Continuing  the  above  example,  this has




been measured  as  0.12  inches  for 250  scfm.   Compute  stack velocity by




Vstack= 234 (298 x °' 12>^fPm - 140° fPm (see Appendix B) .   Convert this to
 a nozzle  flow yielding same velocity.  Since the nozzle has an 1/8 inch dia-



 meter, with area of 0-000085 ft2 the required nozzle  flow is  0.12 scfm.



       4)  Make a preliminary run.  Since isovelocity sampling is



planned, the ratio of  inlet nozzle velocity to duct velocity should



be in the range of 0.9-1.1  (4).   A preliminary run is necessary to



check the rate of sampler  pumping.  Consult Figure 6 for an approxi-



mate setting for the orifice pressure drop.  For .12 scfm,  k, (orifice) =



0.07 inches. This setting  should be checked and manipulated after the



probe has been inserted into the gas stream, described in step 6.



       5)   On a form similar to  Figure 7, record HEPA filter  APQ ,



barrel (chamber) T, Average dry  gas meter T at start,  Meter CF start,




and venturi AP-



       6)   Two to three persons  are necessary.  One person manually



inserts probe into sampling port on a signal.  A second person turns



on sampling pump and timer at same signal.   A third person gives the



signal.   All three monitor and note  stack temperature, baghouse



temperature, the temperatures and pressures  in step 5 as well as




the metallizer load voltage response, amperage, air pressure  and




quality of  generation.



       7)   Upon signaling  the end of  the preliminary  sampling period,




about 2 minutes, remove the  probe and turn  off the pump.




       8)   Remove the  filter from the holder  and store in a covered
                               2-17

-------
E
>»—
u
N
N
O
.20

.19  -

.18  -


.17-

.16-


.15  -

.14  -

.13-

.12-

.1 1  -

.10  -

.09  -

.08  -


.07  -

.06  -

.05
                i     i     i    I     I     i    i     i     i
               .01   .02   .03  .04  .05 .06  .07  .08  .09  .10
                  AH  of orifice (inches water)
        Figure 6.  Approximate relationship of sampling nozzle
                  flow to AH orifice of  sample
                       2-18

-------
     Wire type 	          p
               •	•    bar-
                             Wet bulb Temp.

 1.  Sample #

 2.  Barrel T(°C)
  \234


25.  I ^
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19-
20.
21.
22.
23.
24.
HEPA AP
Pduct=Pbar-@/13.6
Venturi AP
Duct T (°C)
Bag T
Pitot AP
Wire feed rate
Amps
open
Vload
MAP
SAP
AH orifice
CF start
CF end
ACF
ATIME
CFM = @/@
Tmeterstart
T
meter ,
end
Nozzle Vel.
Vduct
              Figure 7.   Typical form for Recording
              Mass  Concentration Sampling Data
                             2-19

-------
petri dish.  With compressed air,  blow the probe and holder inlet free



of settled dust.



       8)  Note CF end.   Approximate the isovelocity parameter by



performing following computations.



       Nozzle vel. = ACF/ATIME •(!/.000085 fpm)



       Stack Vel. = 234 (•12x29-92 xTstaclloK^/Pbar'  ^See APPendix B)


       P,   is the absolute total pressure at the sampling port.
        D fit IT


       I  =Nozzle Vel/Stack Vel.



This is only approximate since it ignores moisture and meter temper-



ature.  But it is adequate for sampling.  If the computed I is within



the range of .9-1.1 then assume the sampler is properly adjusted and



a sample may be taken by repeating the above steps.  The sampling



time should be designed to yield approximately a lOmg to 20mg



deposit on the filter.



       If the computed I is outside of the range, the $H on the pump



must be changed in a direction that would bring the I ratio closer



to unity.  To achieve the acceptable ratio, another preliminary



sample should be taken if the deviation is significant, otherwise



take a sample and adjust the AH as necessary.



       9)  During one of the sampling periods, the wet bulb tempera-



ture should be measured and recorded.  As a computation in Appendix D



shows, the maximum moisture influence at the range  of operating



conditions was 4.4%, and thus this factor is not important  in esti-



mating the isovelocity factor.



       10)  At the end of all sampling  for the day,  place filters



with samples that are considered  satisfactory  in a dessicator, laying
                              2-20

-------
in the petri dish with the lid removed.  After about 2 days, weigh




them to the nearest 0.1 mg.   (Weighing was not necessarily done in



a dry environment.)






Sampling Procedure for Electron Microscopy




       The goal in sampling with the  Nuclepore filters was to obtain




a deposit with a countable number of particles.  When the loading




is too intense, individual particles are not distinguishable.  When




loading is too light, the array of particles is too low to provide




statistically satisfactory results.  Isovelocity conditions are




strived for but are difficult to confirm due to the short sampling




time.  Particle size sampling is usually done  concurrently with mass




sampling.   Nuclepore filters are not dessciated and tared.  The




following steps outline the method followed.




       1)  Satisfy the first  3 steps on page 2-16. Place the  Nuclepore




           filter into holder with glossy side toward sample stream.




       2)  Start the pump with the probe not in the exhaust stream.




           After about 5 seconds, insert the probe and direct the




           nozzle toward the  stream and hold for an interval of 1




           second to 4 seconds.




       3)  Repeat the above steps, using different holding  times.




       4)  After collecting several filters, examine  them visually and




           with  ari optical microscope.  The filters  should show  a




           slight shadow or darkening  due to the deposit.    At  a




           magnification of several hundred times, distinct particles




           should be apparent along with ample clean  area.
                               2-21

-------
Sensitivity of Results to Deviations from Isovelocity Sampling

       According to Peterson, aerosol sampling is subject to two

biases:  velocity and inertial impaction.  Bias is defined as

 (13, pg F-4).

       "Volumetric particle concentration or size distri-
       bution determined from probe sample divided by
       actual array of particles which existed in aerosol
       cloud."

       He described work by Lundgren and Calvert in which the two

sampling biases are referred to as R and K.

       R s V /V
            o  c
       K = (Cp V D  )/(18|j,D )

where  C = the Cunningham Correction factor =

       1 + 1.62xlO"5/D  +((.55) xlO"5/D ) EXP(-.67 x 10-+5 D )
                       P                P-                  P
       V  = Duct Velocity = (1400 fpm) = 711 cm/sec

       V  = Probe inlet velocity

       D  = particle diameter (cm)
        P
       p  = particle density,

       p, = viscosity = 185 x 10"  poise (at 25°C)(ref. 16)

       D  = probe inlet diameter = 1/8 inch  = .32 cm
        c

       Peterson developed the following graph, modified from Lundgren

and Calvert data, that indicates "typical experimental sampler inlet

bias for a straight inlet probe" plotted as a function of R and K.

       Peterson said, (13, p. 1-4)

       "As is quite often true in aerosol studies the .  .  .K
       is of primary importance.  The range of maximum interest
       is the K value range from0.01 to 1.0.  Below0.01  the
       sampling bias is low enough to be relatively  unimpor-
       tant.   Above 1.0 the sampling bias is equal  to the
       velocity ratio (R)."

                              2-22

-------
          R   1.0
             0.001         0.01        0.1

                     Value of  K, Dimensionless
1.
       Figure 8.  Plot of lines of equal bias as function of
                  R and K (13, pg. L-5)
       The above formula for K and graph of R and K are applied to

predict the biases to which the particle sampling is subject.  Since

there are small fluctuations in the duct velocity pressure, after

sampling has commenced, it is assumed  the actual R could vary from

0.8 to 1.2,  The following table summarizes the bias study  for which

computations are given  in Appendix D.
       Table 2.  Summary  of Possible  Sampling Biases  Duct
                 Velocity - 711  cm/sec
Particle
Diameter
cm
1 x 10~7
•*• ** _/i
i x 10 7
—A
i x 10 7
—A.
i x 10 7
-6
1 x 10 ?
-6
1 x 10 ?
1 x 10"1
1 x 10 -6


|jiin
1.00
1.00
1.00
1.00
0.01
0.01
0.01
0.01


c
1.16
1.16
1.16
1.16
22.34
22.34
22.34
22.34


PP
5
5
1
1
5
5
1
1


R
1.2
0.8
1.2
0.8
1.2
0.8
1.2
0.8


K
.039
.039
.0077
.0077
.0001
.0001
.0001
.0001
Bias (1.00
is left of
^ sign)
< 1.03
> .97
< 1.01
> .99
= 1.00
- 1.00
= 1.00
= 1.00
                              2-23

-------
       For the most conservative assumptions, the maximum error,



according to the Peterson article, would be 3%.  If discrete parti-



cles on the order of 0.01(j,m (lOntn) are present, the sampling probe



would show no significant bias to them.



       Another bias that might occur during the particle sizing



sampling would be a misalignment of the probe with respect to the



exhaust stream.  The filter is exposed while the probe is being posi-



tioned during which some misorientatien necessarily occurs, Limited data



by Lundgren and Calvert indicate an angular difference of 90  between



inlet and stream flow for 2 |j,m particles gives a bias of less than


   (13")
5%.      Therefore, the placing and removing of the probe probably



causes little bias in the observed particle sizes and distribution.



This problem could be expected to be insignificant in the mass con-



centration samples.  An additional bias occuring is due to the de-




positing of particles  along the walls of the probe and filter holder



upstream of the filter.








                    ANALYTICAL TECHNIQUES



Mass Concentration Analytical Techniques




       Compute the mass concentration  of the sampled air by dividing



the weight gain of a Millipore filter  by the standardized  volume of




air pumped by the sampler,  Dessicate  the filters before and  after



use for approximately 48 hours.  Weigh the  filters on an analytical




electro balance (approximate precision of +0.5mg).  The balance



scale indicates weights to the nearest 0.1  mg.




       The mass concentration analyses are  done by the  experimenters.
                               2-24

-------
Particle Sizing and  Identification Techniques
       Store  the  filtered  samples  individually  in  labeled, capped petri
dishes.  Send to  Walter  McCrone  Associates,  Chicago,  Illinois, the
contractor  for microscopy  analysis.   McCrone examines  the samples by
transmission  electron microscopy (TEM),  scanning area electron micros-
copy  (SEM), and selected area electron diffraction (SAED),
       McCrone prepares samples  for scanning electron microscopy by
a replication process.  See Appendix E.  All of the original features
on the filter surface, i.e., pore  openings, and aggregates are pre-
served.
       SEM photographs are most  informative because of their three
dimensional representation.  Objects closer to the electron beam
collector appear brighter than those farther away.   The depth of
field of the  SEM  is  300 to 500 times that available in a light micro-
                                 (12)
scope of the  same magnification.       Photographs  from McCrone indi-
cated magnifications ranging  from  5,000  x to 11,000 x.  SEM is thus
useful for studying  particle  shapes, porosity, aggregation and surface
features.
       The TEM allows  inspection of individual particles by magni-
fications of  up to several hundred thousands times but sacrifices
the SEM's depth of field.
       Using  the  TEM,  particles  are counted  and sized  by "hand" by
                                          (14)
observing about six  areas  of  the deposit.
       High energy electron diffraction  of an individual particle
yields photographs showing a  set of dark and light concentric bands,
whose spacings and relative intensities  depend upon the structure
                               2-25

-------
 uncl  composition of the specimen.  The pattern formed by scattered


 electrons  is Mcanned by a recording microphotometer.  The calibrated


 output  shows experimental diameters of circles from specific reflect-


 ing  planes.  These diameters are compared with the  ASTM diameter,

                                                          (14)
 for  the specified plane, to identify the sample material.




     SELECTION OF EQUIPMENT OPERATING CONDITIONS  FOR SAMPLING


       The metallizer manufacturer recommends that the electric arc


be kept as small as possible without 'spitting1  to achieve the most


efficient metallizing.     The experimenters consider this condition


as stable and observe that stability results in the following charac-


teristics :



       1.  Particle generation is smoother and quieter.


       2.  The indicated amperage remains steady.


       3.  The load voltage is either equal to or a few volts


           less than the open circuit voltage setting, and does


           not oscillate about this level by more than one volt.


       The experimenters assume that smaller particles and repro-


ducibility will result from stable operating conditions.


       The metallizer can be readily adjusted for rates of wire feed,


air pressures to the arc (MAP and SAP) and open circuit voltage.


The exhaust flow rate can also be adjusted by a damper and valve in


the exhaust ductwork.  It should be noted that the open circuit


voltage is considered a dependent variable of the wirefeed rate.



The wirefeed rate establishes the amperage across the arc.



       With due consideration for the funds  available for electron
                               2-26

-------
microscopy analysis, the following conditions were selected to examine

the zinc wire fume.



           Table 3.  Zinc Wire  Sampling Conditions

            Wirefeed
Condition	Rate	open    load   Amps   MAP   Flow   SAP
1
2
3
4
30
30
50
50
21
21
23
23
21
21
22
22
30
30
70
70
70
50
70
50
250
250
250
250
20
20
20
20
 Notes:   Wirefeed rate- percent of full rate,  V0   - open circuit
 voltage,  V1   ,- voltage across arc,  MAP - main air pressure-psi,
 SAP- secondary air pressure,  Flow- Exhaust flow in duct-scfm.
                               2-27

-------
                           RESULTS






           BRIEF EVALUATION OF EQUIPMENT OPERATION




       As of April 1976 the experimental apparatus was considered




moderately reliable by the experimenters.  The equipment would not




operate consistently satisfactory at established stable conditions.




Specifically, many runs of zinc wire were interrupted when either




of the wires jammed in the flexible conduit due to various friction




affects.  Also, the observed stabilities of steel wire at several




operating conditions and of the zinc at condition four (Table 3)




were not repeatable on different days.




       It is important to note that the metallizer by itself is




noisy and the use of the large particle collection drum adds to




the noise.  This noise, while not measured, interferred with communi-




cation between research personnel in a high ceiling room.  If not




corrected, it would reduce research work effectiveness, possibly




interfere with the other research work nearby and could be expected




to create stress in animals if the particles were used in animal




quarters.  The noise can undoubtedly be corrected by  inherent design




changes and by additional noise control materials.
                             2-28

-------
                SAMPLE NUMBER IDENTIFICATION
Mass Concentration

       The zinc oxide samples shown in Table 4 were analyzed for
mass concentration.


       Table 4.  Identification of Mass Concentration Samples
          Sampling Condition            Sample Numbers
                  1                   13, 19, 34, 45, 46
                  2                   31, 32, 33, 44
                  3                   15, 18, 35, 36
                  4                   37, 38, 40


Electron Microscopy and Electron Diffraction
       The zinc oxide samples shown in Table 5 were sent to Walter
McCrone Associates for SEM, TEM and SAED analysis.
       Table 5.   Identification  of McCrone Samples
          Sampling  Condition           Sample Numbers
                  1                       101,  125
                  2                       133,  134
                  3                       104,  128
                     CHEMICAL COMPOSITION
Electron Diffraction Data
       Selected area electron diffraction results, as received from
McCrone Associates,  are given in Tables  6 and  7.  According to Richard
Shimps of McCrone Associates, these tables indicate that the particles
generated were  ZnO.
                               2-29

-------
      Table 6.  Electron Diffraction Data  -- Sample  104
      SAED Calculated diameters      ASTM  Card S-0664
      of Circles from Specific       (^nO) diameters
      Reflecting Planes              of Circles  from Specific
                                     Reflecting  Planes
Angstrom (0.1 nm)
2.814
2.592
2.463
1.903
1.622
1.498
1.378
Angstrom (0.1
2.816
2.602
2.476
1.911
1,626
1.477
1.379
nm)







       Table  7.   Diffraction  Data --  Sample  133

       SAED Calculated  diameters       *>™ C*f  S:°664  _
       of  Circles from  Specific        ^n0  diameters of
       Reflecting Planes               Circles  from Specific
       	_	Reflecting Planes
          Angstrom (0.1 nm)              Angstrom (0.1 nm)
              2.82                          2.816
              2.60                          2.602
              2.50                          2.476
              1.92                          1.911
              1.635                         1.626
              1.483                         1.477
              1,392                         1.379

Assay of Wire
       A typical assay of the Wall Colmonoy zinc wire  (Platt #302)
15 gauge (.145 cm diameter)is
                    Zinc      99.99$  +•
                    Iron         .0015%
                    Cadium       .0015%
                    Lead         .002% Maximum
                              2-30

-------
                     MASS CONCENTRATION
       Mass concentration sampling data as observed are presented
in Appendix F.  Concentrations  expressed as mg/acf are converted to
mg/scf at the standard conditions of  760 mm mercury and 25° Centi-
grade.  Table 8 summarizes  the  standardized concentrations expressed
     /  3
as g/sm  .
           Table  8.   Standardized Zinc Oxide Mass
                     Concentrations for all Samples
CONDITION
REPLICATE




1
#
13
19
34
45
46
3
g/sm
0.87
1.26
1.00
1.20
0.80
2
#
31
32
33
44

g/sm
0.74
0.67
0.62
0.63

3
#
15
18
35
36

g/sm
1.91
2.04
2.02
2.25

4
#
37
38
40


g/sm
1.03
1.09
1.23


        The  concentrations in Table 8 are more meaningful when the
 statistical parameters  -- mean,  standard deviation,  and coefficient
 of variation (C.V.)  are calculated.   The coefficient of variation
 is one  indicator of  reproducibility or precision of  concentration
 generation.   These parameters are summarized in Table 9A.
          Table  9A.  Zinc Oxide Mass Concentration
                     Mean and Variation Summary
               Condition                  1      2        3       4
        (1)  Mean  (g/sm3)                  1.03   0.68     2.05    1.12
        (2)  Mean  (grain/scf)©x 2.29     2.36   1.56     4.69    2.56
        (3)  Stand,  dev.  (g/sm3)           0.20   0.05     0.14    0.10
        (4)  Coefficient  (3, /(J)x 100)        19%   8.4%     6.9%    8.7%
            of variation
                              2-31

-------
          Table 9B.  Mean Mass Concentration Matrix
                       Wirefeed 30           Wirefeed 50
       MAP 50          0.68 cond. 2          1.12 cond. 4
       MAP 70          1.03 cond. 1          2.05 cond. 3

       Row 4 of Table 9 indicates that the C.V.'s of conditions 2,
3, and 4 are less than 9% while condition 1 has a C.V. of 19%.
Condition 1 is apparently much less reproducible.

Means Testing
       A question relative to the values in Table 9B is whether the
means are significantly different.  The mean of condition 1 (X..)
is fairly close to X, but two operating variables (wire feed rate
and MAP) are different.  A statistical method to test the other
                                              C\g\
pairs of means is the Smith Satterthwaite testv   .  This method
is utilized by hypothesis tests applied (see Appendix G-l).
       These hypothesis tests support the conclusion at the 5%
level of significance that X^ > X^, X- > iL and X- > X,.  This
implies that the mass concentration of zinc oxide particles is
increased when air pressure or wirefeed rate or both are increased

Analysis of Variance
       Another statistical analysis applicable to the mass concen-
tration data is the analysis of variance.  This  is generally  a more
efficient and more powerful systematic approach  of distinguishing
experimental (random) error from variation due to different operating
           (19)
conditions.   '  However, appropriate experimental design to  facili-
tate the analysis of variance would have required samples  from

                              2-32

-------
several more wirefeed rates and MAP's instead of several replicates




for each condition.  Application of the two way classification




analysis of variance as described on pp. 274-278 of ref. 19 yields




results in contradiction to the hypothesis tests.  But the contra-




diction is presumably caused by the insensitivity of the test at




the minimum degrees of freedom of the two variables.








        PARTICLE SIZING BASED ON ELECTRON MICROSCOPY




       The electron microscopy photographs of the zinc oxide




samples were all similar in terms of particle appearance.  SEM




photographs in  Appendix H  of sample 133, show agglomerates of dif-




ferent shapes and  sizes.   The agglomerates range in size from less




than 0.1 p,m to  over 2 p,m and appear to be interconnected.  The




shapes vary from spherical to chainlike.  These agglomerates are




composed of very fine particles as shown in the TEM photos for




sample 101 (Fig. H-3).  These particles, approximately circular




and presumably  spherical,  have diameters of about 10 nm  (.01 fj,m).




These were the particles sized and counted by the McCrone organiza-




tion.  Since the agglomerates were interconnected, McCrone's group




made the assumption that there was no meaningful method  to count




them for some appropriate  dimension.  The agglomeration  is dis-




cussed more extensively later in the report.




       Table 10 is a modified form of the zinc oxide particle




sizing and count data as it was received from McCrone Associates.




This table indicates the number of observed particle diameters in




each one nanometer wide interval.  Particles were not counted by






                              2-33

-------
                                       Table 10A.   Zinc Oxide Particle  Size Data
N3
I
Size Range
nm
20 -» (a)
19-20
18-19
17-18
16-17
15-16
14-15
13-14
12-13
11-12
10-11
9-10
8-9
7-8
6-7
5-6
4-5
3-4
2-3
1-2
0-1
Total
Sample
number
2
0
0
1
1
2
3
2
9
13
8
13
23
32
36
41
15
15
26
23
32
297
101
percent
.67
-
-
.33
.33
.67
1.01
.67
3.03
4.37
2.69
4.37
7.74
10.77
12.12
13.80
5.07
5.07
8.75
7.74
10.77
99.93%
Sample
number
24
2
7
17
18
24
16
22
34
22
22
20
35
19
16
12
17
12
20
11
6
376
104
percent
6.36
.53
1.86
4.52
4.78
6.38
4.25
5.85
9.04
5.85
5.85
5.31
9.30
5.05
4.25
3.19
4.52
3.19
5.31
2.92
1.59
99.90%
Size Range
nm
20 •*
19-20
18-19
17-18
16-17
15-16
14-15
13-14
12-13
11-12
10-11
9-10
8-9
7-8
6-7
5-6
4-5
3-4
2-3
4-2


Sample
number
26
7
3
1
11
9
9
9
29
5
13
9
31
20
31
20
8
36
2
1
	
280
125
percent
9.29
2.50
1.07
1.07
3.93
3.21
3.21
3.21
10.36
1.79
4.64
3.21
11.07
7.14
11.07
7.14
2.86
12.86
0.71
0.36
	
99.99%
       (a) In samples 104, 125, 128,  133,  134 there were no more than 3 particles per 1 nm interval > 20nm.
           Particles were observed up to 50 nm.   In this report particles > 20 nm are treated mathe-
           matically as belonging to  an interval  from 20-22nm.  The investigator realizes that this
           assumption significantly influences the computation of statistical parameters.

-------
                                   Table 10B.   Zinc Oxide Particle Size Data
I
U>
Ul
Size Range
nm
20 -» (a)
19-20
18-19
17-18
16-17
15-16
14-15
13-14
12-13
11-12
10-11
9-10
8-9
7-8
6-7
5-6
4-5
3-4
2-3
*2
Total
Sample 128
number percent
19
2
5
2
4
7
9
13
43
17
33
14
11
25
19
17
23
17
1
0
281
6.76
0.71
1.78
0.71
1.42
2.49
3.20
4.63
15.30
6.05
11.74
4.98
3.91
8.90
6.76
6.05
8.19
6.05
0.36
0.00
99,99
Sample
number
25
5
6
2
11
12
8
17
27
9
29
9
21
15
29
45
21
41
13
5
350
133
percent
7.14
1.43
1.71
0.57
3.14
3.43
2.29
4.86
7.71
2.57
8.29
2.57
6.00
4.29
8.29
12.86
6.00
11.71
3.71
1.43
100.00
Sample
number
31
8
6
11
9
13
3
10
8
15
24
5
8
11
4
11
25
49
25
4
134
percent
11.07
2.86
2.14
3.93
3.21
4.64
1.07
3.57
2.86
5.36
8.57
1.79
2.86
3.93
1.43
3.93
8.93
17.50
8.93
1.43
280 100 . 01
               (a)  In samples 104, 125, 128, 133, 134 there were no more than 3 particles
                    per 1 nm interval > 20 nm.  Particles were observed up to 50 nm.  In
                    this report particles > 20 nm are treated mathematically as belonging
                    to an interval from 20-22 nm.  The investigator realizes that this
                    assumption significantly influences the computation of statistical
                    parameters.

-------
mass or surface area.


Analysis

       It was intended that the analysis of the particle size data

would answer the following questions.
                           «
       1)  Are the particle size statistical parameters (mean,

           variance, distribution) from each replicate sample

           significantly different?

       2)  Are the statistical parameters from samples differing

           by one variable significantly different?

       3)  Do any theoretical statistical distributions reasonably

           fit the observed distributions?

Such an analysis would give a quantitative measure of the repro-

ducibility potential and effect of the operating variables in addition

to a characterization of the generated particles.


Computation of Means and Variances

       The mean diameter for each sample can be calculated by

       _    n
       D -  S d.f
           i-1   1


where  d. = midpoint of size interval

       n  = number of intervals

       fi = decimal fraction of total within interval
        n
        E f, = 1.00
                               2-36

-------
       The variance  of  the  sample diameters  can be computed  from
        2    n   2     -2
       S  =  S x/f   - D^

                 L i
           Table  11.   Means  and Variances  of zinc Oxide Particle Size Data
Sample #
Condition II
D (nm)
S2 (nm2)
S (nm)
Sample Size
101
1
6.0
14.6
3.8
297
125
1
10.2
29.2
5.4
280
133
2
9.1
35.6
6.0
350
134
2
10.0
41.1
6.4
280
104
3
10.8
28.1
5.3
376
128
3
10.4
22.0
4.7
281
Reproducibility Among Replicates



        The mean diameters  for samples 101 (D,-,,)  and 125  (D125) of



condition 1  are significantly different by inspection.



        The values  of the pairs D^-  and D, •,*  f°r  condition 2, and



^in4 an(*  ^128  ^or  con<*ition 3 are closer and  are  analyzed by a


statistical  method for significant difference in  Appendix G-2.



        This  analysis indicates that  the two replicates  for condi-



tion 2  have  diameter means and variances that are not significantly



different.   The analysis also indicates that  the  means  and vari-



ances for condition 3 replicates  are not significantly  different.





Differences  Between Conditions



        An important question is whether the means and variances for



particle  diameter  populations generated under different conditions



are different.   Since the  mean of condition 1 cannot be estimated



within  a  relatively tight  range,  only the relationship  of condition
                                2-37

-------
2 to condition 3 is examined.   However,  conditions 2 and 3 differ by
two variables, and if the mean sizes are different there is a lack
of data to distinguish variable influence.
       Hypothesis testing in Appendix G-3 demonstrates that the
mean particle diameters for conditions 2 and 3 are significantly
different.  Thus changing the operating conditions apparently
changes the mean particle diameters.

Frequency Distributions
       Another statistical parameter that provides useful informa-
tion is the frequency distribution.  The observed distributions are
presented above in Table 10.  The following two tables and two
graphs indicate the cumulative frequency for conditions 2 and 3.
In addition typical histograms representing observed frequencies
and frequencies expected from the appropriate normal distribution
are plotted.
       The normal distribution provided a better fit to the observed
distributions than did the log normal.  This was found by calcu-
lating the maximum differences between the observed and expected
(theoretical) distributions.  This  is examined statistically  in
the next section.
       It was also attempted to decide tfhich of the two fit better
by plotting the data onto log normal and normal probability paper.
By this method, data should plot as a straight line.  However,
this method was inconclusive as indicated by Figures  13 and  14.
                               2-38

-------
       Table  12,   Cumulative Observed and Expected Frequencies
                   for  Condition 2
                                              Maximum  Diff.    (a)
                                               (133)    (134)

d
ran
: 2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

z

-1.23
-1.06
- .90
- .74
- .58
- .42
- .26
- .10
+ .06
+ .23
+ .39
+ .55
+ .71
+ .87
1.03
1.19
1.35
1.52
1.68
(J, « 9.6
F(d)133

.014
.051
.168
.228
.357
.440
.483
.543
.569
.652
.678
.755
.804
.827
.861
.893
.898
.916
.930
a - 6.2
F(Z) :

.109
.145
.184
.230
.280
.337
.400
.46
.52
.59
.65
.71
.76
.81
.85
.88
.91
.94
.95

F(d)1;

.014
.104
.279
.370
.407
.42
.46
.49
.51
.59
.65
.67
.71
.72
.77
.80
.84
.86
.89
                                                       .140


                                               ,103
(a)  The maximum absolute difference between observed (F(d)) and
    expected frequency (F(Z)).
                              2- 39

-------
      Table 13.   Cumulative Observed and Expected Frequencies
                 for Condition 3

                 j,  * 10,44    a - 4.9
                                              Maximum  Diff.    (a)
                                               (104)   (128)
di
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Z
-1.93
-1.72
-1.52
-1.31
-1.11
- .91
- .70
- .50
- .29
- .09
+ .11
+ .32
+ .52
+ .73
+ .93
+1.13
+1.34
+1.54
+1.75
+1.95
F104
.016
.045
.098
.130
.175
.207
.250
.300
.393
.446
.505
.563
.654
.712
.755
.818
.866
.911
.930
.935
F(Z)
.0268
.043
.064
.095
.134
.181
.242
.309
.386
.464
.544
.626
.698
.767
.824
.871
.910
.938
.960
.974
F (d ) .. ,
-
-
.004
.064
.146
.206
.274
.363
.402
.452
.569
.630
.783
.829
.861
.886
.900
.907
.925
.932
                                                      .0844
                                               .0688
(a)   The maximum absolute difference between observed (F(d)) and
     expected frequency (F(Z)).
                             2-40

-------




«0
i_

(D
E
o
\l^
c
rt
c
i_
/«)
VL*
+•>
(U
E
rd
T>
/ii
vU
lj
-H
C.
rt
IU
Q.



,

cv -
19 -

1S -
1 7 -
16 -
15 -

14 -i

13 '

12 -
11 -
10 -
9 -
S -
7 -
6 -

5 -
4 -
3 '
— — . 	 o-^a— i
D Cum. freq. O AD
Normal dist.
O Cum. freq. ° ^
Sample 134
O IA
A Cum. freq.
Sample 133 o ^
O C&

O Q A

O D A

a\
D A
CQ A
DO A
D OA
D Gk
f
a A o

Q 0
za o
A oa
\ 	 n 	 	 	 • — 	 	 —
O 	 LJ 	 1 	 1 	 1 1 1 T i '
O 10 20 30 40 50 60 70 SO 90 10
: Cumulative percent
Figure 9.  Comparison of cumulative frequencies for
           condition 2 (table 12)
                   2-41

-------





'in

I
O
c
cti
* w
c
c_
"£
E
Til
"y
OJ
Q_





20 -
19 -

18 -
17 -
*
16 '
1 5 -
14 -

13 -
12 -
11 -
10 -
9 -
8 -
7 -

6 -
5 -
4 -
3 (
ZJLJ
D Cum. freq. CD
Normal dist.
O Cum. freq.
Sample 12fc
A CD
A Cum. freq.
Sample 104 ADO
ADO
AD 0

AD 0
A D
ADO
O
DO
tE O
CZLO

DO
DDA
CD A
1 DA i i
J LJ*-i .1 i l i i i i t
0 10 20 30 40 50 60 70 S»0 90 10
             Cumulative  percent
Figure 10.  Comparison of  cumulative frequencies for
           condition 3 (table 13)
              2-  42

-------
to
I
18 -
17 -
16 -
4) 15 -
T)
^ 14
13 -
E 12 -
C
v- 11 -
Percent in interva
-*t3(t>b-(fl®^CD(9O






•••^M^























	








































— — Normal











distribution




















f 17
- 16
- 15
• 14
- 13
• 12
- 11
• 1O
• 9
8
7
6
5
4
3
2
1
                2.5  3.5  4.5  5.5  6.5  7.5  8.5  9.5 1O.5  11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5


                            Particle diameter, midpoint of interval  (nm.)
                     Figure 11.  Observed frequency of  sample 134 and normal frequency

-------
NJ
JS
-P-
     E
     c
(U

c
c
-H
c
     (D
    Q_
14 -
13 -
12
11 -\
10
 9 -

 7 -
 6
 5 -
 4 -
           3 -
           2 -
           1 -
                                                                           Normal distribution
-14
-13
-12
- 11
-10
- 9
- 8
- 7
  6
- 5
- 4
- 3
- 2
  1
               2.5  3.5 4.5  5.5  6.5  7,5   8.5  9.5 10.5 11.5  12.5 13.5 14.5 15.5  16.5 17.5 18.5 19.5
                            Particle diameter, midpoint  of interval  (nm.)
                      Figure 12.   Observed frequency of sample 133 aad normal frequency

-------
c
4)
U
£_
CD
Q.
3J

E
rj
U
95




90 -





80



70



60



50



40


3O



2O J





10 -




 5 -
   024
                      T	1	1	1	1	1	1	1	

                      6   8   10  12   14   16   18  20   22
            Particle  diameter  (nanometers)
          Figure 13.   Arithmetic probability plot of sample

                     128 diameters
                        2-45

-------
 c
 (D
 U
 t_
 
U
95



90 -




80


70


60


50


40


30


20 -




1 0 -



  5 -




  2 ~


  1 -


0.5 -


0.2 -

0.1
                —T~

                 3
                   -i	1	1	1—i—i—

                    5   6   7  S  9  10
20     30
                Particle  diameter  (nanometers)
        Figure 14.  Log probability plot of sample 128 diameters
                       2-46

-------
Computation of Normal Distribution



       The cumulative normal distribution can be constructed from



statistical tables by assuming values of jj, and a for the population



of diameters  (d^.  Tables ^ 9^ of the standard normal distribution

                                                        d. - n

give values of F(Z).  F(Z) is the probability that (Z. = —	) < Z
                                                     1    C7     —



Goodness of Fit



       Figures 9 and 10 show by inspection that the normal distri-



butions provide an approximate fit to the observed data.   Two of the


statistical methods for testing goodness of fit are the Kolmogorov-



Smirnov (K-S) test(19'20^ and the chi square test.(19'21^  The


investigator used both methods and selected the K-S test for this



report.  The K-S test is a more powerful test for continuous dis-


                  (25}
tribution fitting.   '


       The one sample K-S test concerns the agreement between an



observed distribution and a specified continuous distribution.  The


two sample K-S test concerns the agreement between two observed



distributions.  The test is (19, p. 222) "sensitive to population



differences with respect to location, dispersion or skewness."



Both tests are based on the maximum absolute difference between



the two cumulative distributions.


       In Appendix G-4 the one sample test is applied  to  samples



133, 134, 104, and 128.  Only the distribution of diameters  from



sample 104 statistically fit an appropriate normal distribution.



The two sample test is also applied to the replicates.  The  dis-



tributions of diameters for samples 133 and 134 of condition 2  are
                              2-47

-------
not significantly different, but the distributions for samples 104
and 128 are different.


                        AGGLOMERATION
       An important question is where do the agglomerates originate?
The available information does not directly answer this question.

One reason why this question is important is that the agglomeration
might be a peculiarity of only the generating equipment, the drum
and duct system, the sampling processes, or of some combination of
them.
       Particle agglomeration is discussed in references 2, 7, 22
                                    (22)
and 23.  According to Green and Lane     the general equation "for

the rate of coagulation (agglomeration) of an aerosol" is


                                     -2
       where
           A
       1 +	 = Cunningham  correction  factor

       A = .9 for smokes

       Ji = mean free path

       Jl  = 9 x 10~6 cm

       r = particle radius  = mean value of  r,  and r2 (cm)

       r^ = radius of particle  1  (cm)

       •^2 = radius of particle  2  (cm)
                                     3
       n = number of particles  per  cm
                                        23
       N = Avogadro's number =  6.0  x 10

       R = gas constant = 8.3 x 10   for
                              2-48

-------
       T = 293° K



       T) - 1.82 x 10    poise



       S * sphere of influence/particle jradius

                        (r,+r )2

Green indicates that %[— — - ], wMoh has a minimum of 1, has an average

                           12       r

value of 1.27 when the range of radii ~- in a uniformly polydiaperse aerosol is 1-8 .



       Then ^=ffL(l+f)n2  (1.27)                (2)





                = K (l+^-)n2 (1.27)                     (3)



Green lists a table giving values  of K for ferric  oxide and magnesium


                                                -9   3
oxide, among others, which are  .66 and .83 x 10   cm /sec, respectively.



By integrating  the  above  equation, it  follows




       ^  - ^ = 1.27 K (1 + ^)t                       (4)

            O                                   _y

For  the zinc oxide  particle we  can let r = 5x10 cm (5nm)
        (1 -H-) -1+18-19


                       -9
and assume K =  .7 x  10  .   To  estimate n , we  can assume  concentration^
             o          ^f      O

       1  g./m  =  1 x  10   g/cm  and
                           43
       S.G. =5; C = nQ  5.|- nr    g/cm



       nQ = 3.83 x 10    particles/cm
3
Computing the time  to  reduce  the initial number of  particle n ,  to
                      -4
       tj, = 1.55 x  10   sec



This implies that the agglomeration takes  place rapidly and well



before the particles reach the  sampling probe,  an elapsed time



of about 0,3 seconds.  The agglomeration apparently commences close



to the arc.
                              2-49

-------
       The agglomeration rate is substantially dependent upon n  and

particle size. If either variable were changed to give a combined
            3
change of 10  (toward smaller concentration and larger size) the

t, would be altered to approximately the elapsed time.
 %
       Another factor in the agglomeration is that the particles

are initially charged. (2,9)      Aggregates of charged particles

show pronounced tendency to be oriented in a chainlike fashion,

indicating polarization.  Since particles in arc smoke are highly

charged, chain formation is not unexpected.
              COMPARISONS WITH HEDDEN'S RESULTS

       Although earlier particle generator  work at W.V.U. by Redden*

involved iron oxide, it is nevertheless interesting to compare some

of his results to the zinc oxide results contained in this report.

       The following table compares mass concentration and particle

size data.

                Table 14.   Hedden-Naylor Data
                                               Hedden      Naylor
                                             (iron oxide) (zinc oxide)

    Limit of microscope resolution (nm)          5           1

    Width of diameter count interval (nm)       10           1
                            3
    Mass concentration (gm/m )                  .63         .67-2.05

    Mean particle diameter (nm)                 22        6.0-10.8

    Range of diameters (nm)                   5->100         1-50

    Standard deviation (nm)                     16.3        3.8-6.4

    Cumulative Frequency Distribution       log normal     normal'
                             2-50

-------
       Hedden apparently did not observe agglomerated particles.



This was presumably caused by his procedure of resuspending the



samples in solution and dispersing, rather than due to the fact



that the metal oxides examined were different.
                                2-51

-------
                         CONCLUSIONS



By a method described above, very fine zinc oxide particles were


generated by a commercial electric arc metallizer fed by two wires


and sampled using out-of-stack membrane and Nuclepore  filters.


At four different combined settings of air pressure to the arc and


rate of wire consumption, mass concentrations were measured.  At


three of these settings particle diameters by count were measured


by electron microscopy.  The mass concentration means ranged from

               3
0.67-2.05  g/sm  (1.5-4.7 grain/scf).  The mean particle diameters


ranged from 6.0-10.8 nm  (0.006-0.0108 |im).


       The generation of mass concentration was more precisely re-


produced at three of the conditions.  The mass concentration was


sensitive to changes of the operating conditions.


       Two samples from each of three conditions were observed by


electron microscopy.  At the one condition for which mass concentra-


tion was not reproducible, the mean particle diameter was and is


not considered to be reproducible.  At a second condition,  the


particle diameters' mean and standard deviation were reproducible,


while at a third the mean, standard deviation and cumulative fre-


quency distribution were reproducible.


       The mean particle diameters of the second and third  condi-


tions were significantly different.  The distribution of  particle di-


ameters from each sample is apparently approximated by an  appropriate


cumulative normal distribution.  However, only one sample distri-


bution, from the third condition, has a good  fit, statistically.



                             2-52

-------
       Virtually all particles were agglomerated.  The agglomeration

initiates immediately after particle generation at the electric arc.

Its fast rate results from a high initial number of particles per

volume and very fine particle size.



                       RECOMMENDATIONS

       The Flamespray model VT500 electrospray metallizer has

demonstrated several important capabilities.  For at least one

operating condition, it can generate very fine agglomerated zinc

oxide particles that are reproducible with respect to both the

distribution of particle diameters and mass concentration.

       Therefore, this metallizer is considered worthy of continued

attention by researchers as a device for generating very fine par-

ticles.  This attention would include, but not be limited to:

   A.   Improving the research tool.

       1)  Further development of equipment to improve
           operational quality

           a.  Improve the wire feeding of the softer metals
               to prevent interruptions of particle genera-
               tion.

           b.  Determine requirements for maintaining con-
               sistently stable operating conditions.

           c.  Reduce the ambient noise levels caused by
               this metallizer.

       2)  Investigation of other metals and  expansion
           of generating and sampling program.

           a.  Utilize different wire types and diameters.

           b.  Determine the sensitivity to additional variables
               such as exhaust flow and secondary air pressure.
                              2-53

-------
        c.  Redesign the sampling program to facilitate
            an analysis of variance of mass concentration.
            Instead of two different wirefeed rates and
            MAP (main air pressure), these could be in-
            creased to 6 different ones.   However, retain
            the present practice of collecting two repli-
            cate samples for electron microscopy analysis.

    3)  Investigation of methods of sampling and measure-
        ment compatible with preserving freshness of
        particles.  This would necessitate a definition
        of freshness and a method for preserving the
        sample during collection, during storage and
        transportation, and during analysis by electron
        microscopy, electron diffraction, x-ray diffrac-
        tion, etc.

    4)  Review of present sampling and measurement methods
        to determine if they reasonably utilize the best
        available technology.

    5)  Further study of the agglomeration of very fine
        particles.

        a.  Define appropriate dimensions for describing
            the agglomerates.

        b.  Based on these dimensions, determine whether
            or not the metallizer generates predictably
            reproducible distributions of sizes.

        c.  If predictable reproducibility is established
            determine if there is a significant correla-
            tion between agglomerate size and such variables
            .as chamber-duct retention time, number of very
            fine particles per unit volume, and diameters
            of the very fine particles.

        d.  Related to c, determine if there is a signifi-
            cant correlation between very fine particle
            diameter, agglomerate size and mass concen-
            tration.

B.  Using the research tool.

    1)  When a better understanding of this agglomeration
        is reached, modify the experimental equipment
        design to achieve predictable levels of exposure
        to mass concentrations, particle diameters, and
        agglomeration sizes found in various types of
                           2-54

-------
    occupational and community atmospheres.

2)  Modify the experimental equipment design to achieve
    effectiveness of small particle control-equipment.

3)  Investigation of the respiratory effects of fresh
    particles.

    a.  What  is the fate of agglomerated particles
        collected at various  locations  in  the  lung?

    b.  What  is the relationship  between respiratory
        effects and particle  age?
                         2-55

-------
                      LIST OF  REFERENCES
 1.   Wall Colmonoy Corporation,  "Directions  for Operating the Model
     VT-500 Colmonoy Electrospray Metallizer," Detroit,  Michigan,  1973.

 2.   Green, H.  L.  and Lane,  W.  R.,  Particulate Clouds;  .Dusts, Smokes
     and Mists, 2nd edition, Van Nostrand Company,  Co.,  Princeton,
     N.  J., 1964.
 3.   Linsky, B., Smith,  G.,  "Sources of Information on Air Pollution,
     Proceedings National Conference on Air  Pollution,  Washington,
     B.C., Nov., 1958.

 4.   Hedden, Robert, Electric Arc Generation of Polydispersed Iron
     Oxide Aerosol in an Air Stream,Problem  Submitted to WVU,
     Morgantown, West Virginia, 1972.

 5.   Johansen,  Arland, Unpublished Report, WVU, 1974.

 6.   Sears, D.  R., "Fresh Metal Fumes Project -2,"  Laboratory Log
     at WVU, 1975.

 7.   Brain, Valberg, Sorokin, "An Iron Oxide Aerosol Suitable for
     Animal Exposures," Environmental Research, Vol. 7, No. 1, Feb.,
     1974.

 8.   "Method 5  - Determination of Particulate Emissions from Sta-
     tionary Sources," Federal Register. Vol. 36, No. 27, 12123, 1971
     pp. 24888-24890.

 9.   Research Appliance Company, "Instructions for RAG Stack Sampler,"
     Gibsonia,  Pa., 1971.

10.   Joy Mfg. Co., "Methods for Determination of Velocity  .  .  .
     Volume of Gases," Bulletin No. W.P. 50, 1970.

11.   Davies, C. N., Air Filtration. Academic Press, N. Y., 1973.

12.   Silverman, L., Billings, C., First, M., Particle Size Analysis
     in Industrial Hygiene, ACGIH and USAEC, Academic Press,  1971.

13.   Air Sampling Instruments Committee, Air Sampling Instruments  for
     Evaluation of Atmospheric Contaminants, ACGIH, 4th  Edition, 1972.

14.   Shimps, Richard, Walter McCrone Associates, series  of Private
     Communications, Jan. 1976-Mar. 1976.

15.   Spurny, Lodge, ejt.al., "Aerosol Filtration By Nucleopore Filters,"
     Environmental Science and Technology. Vol. 3,  1969,  pp.  453-64.
                               2-56

-------
16.  CRC, Handbook of Chemistry and Physics, Van Nostrand, 1975.

17.  "Electron Diffraction," Encyclopedia of Science and Technology.
     McGraw Hill, 1974.

18.  Miller, Richard, Platt Industries, Private Communication,
     January 23, 1976.

19.  Miller, I., Freund, J., Probability and Statistics. 12th
     Printing, Prentice Hall, Englewood Cliffs, N.J., 1965.

20.  Kim, P. J. and  Jennrich, R.  I.,  "Tables of the Exact Sailing
     Distributions of the  Two Sample  Kolmogorov-Smirnov Criterion,"
     Selected Tables in Mathematical  Statistics, Barter and Owen, ed.,
     Vol. 1, Markan  Pub. Co., Chicago, 1970.

21.  Mercer, Thomas, Aerosol Technology  in  Hazard  Evaluation, ACGIH,
     Academic Press, N. Y.,  1973.

22.  Herdan, G.,  Small Particle Statistics, 2nd Edition, Academic
     Press, N. Y.,  1960.

23.  Hidy and Brock, ed.,  Topics in Current  Aerosol Research,
     Part 2, Vol.  3, 1st  edition, Pergamon Press,  N. Y., 1972.

24.  Danielson,  John A.,  Air Pollution Engineering Manual. U. S.
     Department  of H.E.W., No.  999-AP-40,  Cincinnati,  Ohio,  1967.

25.  Damewood,  E.  Z.,  West Virginia University Industrial  Engineering,
     Private  Communication, April 16, 1976.
                               2-57

-------
                         APPENDIX A
             COLLECTION EFFICIENCY OF THE BARREL

       An empirical formula by Lapple^   ' p^    '  is  "sufficiently
accurate for an engineering estimation  of many  cyclone  applications."
He defines D   as  the diameter of those particles collected with 50
percent efficiency.  After determining  D  by the formula  following,
the collection efficiency of  particles  D can be  determined by  a
graph showing cyclone efficiency versus D /D
                                         p  pc

       J)   ;s I i i ii mi rfA»-»n»-i«^ I

       where p, = gas viscosity = (183x10 x.672x10~3)lb mass/ sec  ft.
             b = cyclone  inlet width =  0.08 ft
            N = effective  turns =  %  (For  cyclones  may vary from  %-10)
              e
            V. = gas  inlet  velocity = 44 fps
            p  = particle density  = 5.6(s.g.) x62.4
                                   = 349 lb/ft3
              b, V.,  are  estimated  as spread and velocity of
              compressed  air stream reaching arc
             fe x  (183 x  10"4)(.672  x 10"3) (.08))^ ,  .    .Q-5 ,
           = V	2(%)44(349)n                 7"
V
            = 7.31 x 104 cm = 7.3 |j,m
50% by weight of the 7.3 urn particles are collected.
For 95%  efficiency D /D c = 4 ,  according to Figure 48, pg. 95,
re£.  24.
        D  = 4 x 7.3 tun = 29 JimJSr 30
         P
                             2-A-l

-------
                         APPENDIX B



              COMPUTATION OF STACK GAS VELOCITY.


              VELOCITY TRAVERSES, AND STACK FLOW





       According to reference 10, the velocity at a stream line



in a flue V , can be determined from the following
           s


       Using standard pitot tube



       V (fps) = 3.90   ^•^•xi^xHT
                          s
       P  = absolute pressure  in flue in inches mercury
        s

       Gd = specific gravity of gas referred  to that of air



       H  = velocity head  in inches of water  = AH


       T  = flue  gas temperature in degrees Kelvin assuming Gd - 1
        s
           = 3.90 (^~- £HTgJ *fps  -  234
                             2-B-l

-------
                         APPENDIX B
               Table B-l.  Velocity Traverses for Different
                           Venturi Pressure Drops (6)
Venturi AP .20 inches water
% of duct
diameter
6.2%
25
75
93.8%
6.2%
25
75
93.8%
Pitot AP
in water
.09
.12
.12
.09
•10
.12
.12
.08
v
fpm
1201
1387
1387
1201
1266
1387
1387
1133
Pitot AP
in water
.10
.12
.12
.09
.11
.12
.11
.08
v
fpm
1266
1387
1387
1201
1328
1387
1328
1133
               (1) v = 1298 fpm      2
           for 6" duct Area » .197 ft
               (2) Q = 256 cfm
Venturi AP =
% of duct
diameter
6.2%
25
75
93.8%
6.2%
25
75
93.8%

Venturi AP =
% of duct
diameter
6.2%
25
75
93.8%
.31 inches
Pitot AP
in water
.11
.17
.18
.14
.15
.18
.18
.12
v = 1596
.40 inches
Pitot AP
in water
.14
.20
.24
.18
water
v
fpm
1328
1651
1699
1498
1551
1699
1699
1387
fpm Q = 314
water
v
fpm
1498
1791
1962
1699

Pitot AP
in water
.17
.18
.19
.14
.17
.19
.17
.12
cfm

Pitot AP
in water
.17
.23
.24
.18
                                                       v
                                                     1651
                                                     1699  vertical
                                                     1746  traverse
                                                     1498

                                                     1651
                                                     1746  horizontal
                                                     1651  traverse
                                                     1387
                                                       v
                                                      fpm

                                                      1651
                                                      1921 vertical
                                                      1962 traverse
                                                      1699

                                                      (continued)
                            2-B-2

-------
Venturi ftp =  .40 inches water  (continued)
% of duct
diameter
6.2%
25
75
93.8%
Pitot AP
in water
.22
.24
.22
.15
V
fgia
1879
1962
1879
1551
Pitot AP
in water
.23
.24
.21
.15
V
fpm
MBlriMMIIBH
1921
1962
1835
1551



horizontal
traverse

                v  = 1795    Q - 354 cfm

 (1)   Velocities are calculated by using "standard air tables"  and
      assuming standard conditions of 760 mm and 70 F.  This  intro-
      duces  only a small error.

 (2)   Flows  are calculated by the formula

      Q = v  •  A where A = n(3/12)2 = .197 ft2

      The values of Venturi AP and Q are plotted in Figure 5.
                              2-B-3

-------
                        APPENDIX C
             EFFICIENCY OF NUCLEOPORE FILTERS
Definitions of Individual Efficiencies:
1)  Partial efficiency of impaction
    Ei
    where


    E'±  = 2 STK/T + 2
      e  =


      P  =  filter porosity - .094    (ref.  21)


    STK  (   Stokes number) = m^/(6T|rRo)


      m  =  mass of single particle (gm)


      r  -  radius of particle (cm)


      44


      Assume specific gravity -  1  (conservative)



      m = 4/3 nr3
                              2-C-l

-------
    Table C-l.  E. for Values of Particle Diameter
                                         STK
D
urn
.1
.15
.2
.3
r
-4
xlO cm
.05
.075
.10
.15
m
gm
1.25x10
.42x10
1.00x10
1.41x10
                             -16
                             -15
                             -15
                             -14
                                   6.74x10
                                   1.51x10
                                   2.7 xlO
                                   2.43x10
                                         -3
                                         -2
                                         -2
                                         -1
.0089
.020
.0358
.322
.012
.028
.0497
.447
2)  Partial efficiency of diffusion
    ED = 1 - .81904 exp [-3.6568 ND
    where
    N  -
                                       -
    L = filter thickness (cm) = 12 x 10   cm
                                        2
    D = aerosol particle diffusivity (cm /sec)
    P = filter porosity = 0.094
R  = pore radius =0.1x 10
   = 3 cm/sec
                                cm
Table C-2
           E
                 for Values of Particle Diameters
Diam.
urn

.02
.04

.10

.15

.20

.30
D
(ref . 11)
-4
1.4 x 10
3.6 x 10 "5
-6
6.8 x 10
-6
3.4 x 10 °
-6
2.2 x 10
-'4
1.24x 10

ND

52.6
13.7

2.58

1.29

.84

.47

ED

1.00
1.00

1.00

.993

.962

.853
                            2-C-2

-------
3)  Partial efficiency of Interception
    ER =
    NR = r/RQ   if NR> 1   efficiency is 1



     r = particle radius
                -4
    R  =0.1 x 10   cm
     o
Table C-3.  E  for Values of Particle Diameter
                     E
D
urn
.04
.10
.20
.21
4.
10 "^cm
.02
.05
.10
.105
NR
.2
.5
1.00
1.05
ER
.36
.75
1.00
1.00
.15ER
.054
.11
.15
1.00
 The  partial  efficiencies  are  tabulated and combined in Table 1.
                             2-C-3

-------
                         APPENDIX D
                COMPUTATIONS FOR K IN TABLE 2
   D                   C
    P


1 x 10"4      1 + 1.62 + .0000677 = 1.16


1 x 10~6      1 + 16,2 + 5.14 = 22.34
                              2-D-l

-------
                         APPENDIX E

          PROCEDURE FOLLOWED BY McCRONE ASSOCIATES

      FOR ELECTRON MICROSCOPY AKD ELECTRON DIFFRACTION


       Shimps outlines analytical procedures  practiced by McCrone

Associates in the following excerpted letter.   (14)

           "The sampling procedure is our standard  one for
       both the SEM and TEM examinations.

            For the TEM, a square of filter  3 mm on a side
       is cut out and placed sample side down on a TEM grid
       with a. carbon substrate.  This preparation is then
       placed on a stainless steel mesh support on a cold
       finger in a condensation washer where  the filter
       dissolves and thus deposits the sample material on
       the carbon substrate.

            For the SEM, a square of filter is cut and placed
       on adhesive on an SEM stub.  First carbon and then
       gold is vacuum evaporated on the specimen so that it
       may be examined in the SEM."
                            2-E-l

-------
                         APPENDIX F



                   MASS CONCENTRATION DATA





       The following -1 tables represent a condensed tabulation of



most of the data observations for each  sample during the sampling.



Each -1 table is followed by a -2 table which includes net sample



weight, mass concentration at actual conditions  (gm/acf) and mass



concentrations at  standard conditions  (gm/scf ) .



       Although moisture in the  air stream was monitored, sampled



volumes were not corrected for it as the correction was small.  The



following calculations  give conservative estimates  of the corrections.



       a.  given:   T      = 30°C
                    Rel Humid - 25%
                    p.    = 28.4 in.
                     bar
                    Vapor Pressure =  31.824 mm= 1.25  in   (ref.  16)
                                              V P x(R.H )
                    Percent Moisture in air	'•—'• -v-•'—1Z = R =  1.10%

                                                   bar

        b.   given:   Same as a) except Rel. Humid - 100%


                            », ., <.      1.25 x 1007,   .  ,„
                    Percent Moisture = 	2874	= 4<4'°



 Condition a was probably not exceeded during the sampling.   The



 equation for the isovelocity factor I is
        I =
            ATime VSTACK
                    STACK





            (   IX   ) x (   C   )








                             2-F-l

-------
where  P__ = 29.92 inches
        O ±.U

            - 298°K
       V     - standardized meter dry volume
       V     - volume of water vapor
        WSTD

       VSTACK = STACK ga

       Anozzle - •°00085
       P,    = Atm P at dry gas meter

       T     = average dry gas meter temperature
        mAVG
       R = V    /V     =0
            WSTD  "^TD
       I1 = approximate I

       C = Correction factor

Tables -1 also provide the I' and C numbers.  In tables  -2, actual

concentrations are standardized by multiplying by 1/SC where
       ACF x        x     .  = ^ x sc
               STD     meter
                               2-F-2

-------
Table F-1A.  Zinc Sampling Data as Observed



2
3








4
5



Pitot AP = .12
Sample #
Date
Pbar
Chanib T
HEPA AP
PSTACK=©-@/13.6
Venturi AP
STACK T(°C)
Wirefeed
Amps
MAP
AH or if
ACF
A Time(min.)
acfm
T (°F)
mavev '
Nozzle vel.
Stack vel.
l' - @/


-------
Table F-1B.  Zinc Sampling Data as  Observed
      Pitot AP = .12     SAP = 20 psi
Sample #
Date
Pbar
Chamb T
HEPA AP
PSTACK in'
Vent. AP
Stack T(°C)
Wirefeed
Amps
MAP
AH orif
ACF
A Time (rain. )
acfm
mave
Nozzle vel.
Stack vel.
l'-
Rel. Humid
C
33
2/14
29.51
20°
1.09
29.43
.20
20
30
30
50
.06
.816
6.0
.136
75
1600
1401
1.15
20%*
.99
34
2/14
29.51
20°
1.09
29.43
.19
23
30
30
70
.058
.779
6.0
.129
77
1527
1409
1.09
20%*
.99
35
2/14
29.51
28
1.09
29.43
.21
25
50
70
70
.055
.476
4.0
.119
79
1400
1416
.99
20%*
1.00
36
2/14
29.51
28
1.09
29.43
,21
26
50
70
68
.055
.487
4.0
.122
78
1432
1418
1.01
20%*
1.00
37
2/18
28.54
31
1.0
38.47
.20
28
50
70
50
.055
.395
3.0
.132
66
1549
1439
1.07
20%*
1.03
38
2/18
28.54
29
.8
28.49
.20
24
50
70
50
.052
.539
4.0
.135
64
1585
1431
1.11
20%*
1.00
Table E-2B. Standardized Mass Concentration Data
I = l' x C
Net wt. gain
gm/acf
1/sc
gm/scf
1.14
.0143
.0175
1.013
.0177
1.08
.0218
.0280
1.013
.0284
.99
.0268
.0563
1.018
.0573
1.01
.0305
.0626
1.018
.0637
1.10
.0113
.0286
1.025
.0293
1.11
.0163
.0302
1.025
.0310
                           2-F-4

-------
Table F-1C.  Zinc Sampling Data as Observed
      Pitot AP = .12     SAP = 20 psi
Sample #
Date
Pbar
Chamb T
HEPA AP
PSTACK
Vent AP
Stack T
Wirefeed
Amps
MAP
AH orif
ACF
A Time
acfm
T
mav
Nozzle vel
Stack Vel.
I'
Rel. Humid
C
Table E-2C.
40
2/18
28.54
30
.8
28.48
.20
25
50
70
50
.05
.478
4
.120
63.5
1406
1434
.98
20%
1.01
Standardized Mass
I = I'x C .99
Net weight
6) gm/acf
7)> 1/sc
8) gm/scf
gain .0162
.0339
1.025
.0347
44
2/18
28.54
23.5
.8
28.48
.20
22
30
30
50
.05
.471
4.11
.115
63
1348
1424
1.06
20%
1.00
45
2/25
29.11
27
.70
29.06
.20
28
30
25
70
.04
,766
6.0
.128
79
1502
1427
1.05
25%
1.01
46
2/25
29.06
30
.70
29.06
.215
28
30
25
70

.355
2.79
.127
80
1497
1427
1.05
25%
1.00
Concentration Data
1.06
.0082
.0174
1.025
.0178
1.06
.0260
.0339
1.0056
.0341
1.05
.0081
.0228
1.0056
.0229
                            2-F-5

-------
                         APPENDIX G


             C-l.  HYPOTHESIS TESTS OF THE MEANS

                 OF THE MASS CONCENTRATIONS




       The Smith Satterthwaite test requires no assumption in


equality of variance, which makes it a powerful test.  It does


assume that the observed values in a given sample are normally


distributed, however.


       Generally:

        0*  ^1 = ^"2






       Reject HQ if t' > ta


 where ^- = true but unknown condition #1 population of mass concentrations


       HQ = Null hypothesis


       H, = First alternative hypothesis



              X, - Xo
       t'
              2   S
                                      nl    °2
       vi = degrees  of  freedom = 	«	•>       o     •?
                                 (S2/n )Z    (S^ In2r
        t    = value  in tables  for t as function of v and a
        a
         a =  significance level = probability of rejecting the

              null hypothesis when it is true.
X, = observed mean of sample


     standard deviation from


     number of observations in sample  #1
         S   = standard deviation from sample #1
                            2-G-l

-------
Specifically, for X;L and KZ (see Table 9A)
       Ho:
       H^  p^ > |j,2



       a- = 0.05



       t* =3.2     v = 2.2




       *  * 2"8
                    (Table IV, ref . 19)
       **>*.
       H  is rejected.  Therefore we conclude jj,- >
Considering X, and X_
       Ho:
       a = 0.05



       t* - 7.43    v = 6.94
       t  - 1.9
        o;


       t' > t
             at
       H  is rjected.  Therefore we conclude p,_ > (j,




Continuing with X2 and X,
       HQ:
       01 = 0.05



       t' = 18.4    v = 3.8



       t  - 2.15
       HQ is rejected.  Therefore we conclude |ju >
                            2-G-2

-------
      G-(2.  HYPOTHESIS TESTS OF THE MEANS AND VARIANCES


               OF REPLICATE PARTICLE SIZE DATA





                  (19)
       This methodv  ' employs a simpler hypothesis procedure than



described in G-l because the sample sizes are well over 30, thus



eliminating the need for a degrees of  freedom calculation.  The



first criterion for replicate reproducibility is  equality of the



true mean diameters.



       Generally  HQ:  jj^ = |j,2      (DI > D2)





                  ^1"  ^1 ^ ^2


                  |j,1 = true but unknown diameter  of replicate 1 >

                       population



                  D- = mean of replicate  1  (best  estimate of (j,^)





                   *         1  9

                  Z  =
        Reject H   if  Z* >
 For condition 2  (see Table  11);



        D1 = 10.8       Sx2  -  35.59      nL  =  350



        D2 = 9.1        S22  =  41.14      n2  =  280





        z* =       -
/35.6 .  41.1 )

»            /
              350    280



        For a - 0,05      Z    =  1.96
        Hn cannot be rejected.   Therefore we conclude ^ =
                             2-G-3

-------
       For condition 3:
Dl =

D2 =
*
Z =
*
Z >
10

10

0.

z
.8

.4

92

In
Sl
2
S2

Z /o
a 12


= 28.

= 21.

= 1.


14

96

96


nx = 376

n2 = 281

a = 0.05


       Hn cannot be rejected.  Therefore we  conclude  (j,,  =  M-O-



       A second criterion for replicate reproducibility  is equality



of the true variances.  Generally


              2     2
       H0:
              2     2
                j4 a
Reject HQ if F >  F^/2 (n^l,
             2   2
 where F = S- /S2



       n = sample size



       Applying this to conditions with means that are not  signifi-



cantly different, for condition 2



       Sj2 = 41.14    nx-l - 279



       S22 - 35.59    n^J = 349


       v - 41.14 _ i 16

       F " 35.59 ~ 1'16


   for a =0.02     F ._(279, 349) - 1.32
                    a/2

                                                                   2   2
       F < Py/2 and H~ cannot be rejected. Therefore we conclude cr,  =
-------
               G-3.  HYPOTHESIS TESTS OF THE MEAN
                DIAMETERS OF CONDITIONS 2 AND 3

       Generally, the analysis is similar to the ones for replicate
reproducibility.    The  means  and variances for each condition are
first averaged.
       For condition 3
T *.    n    376(10.8).  + 281(10.4)
Let    Dl         376 + 281
       DX = 10.6
         2   376(28.14) + 281(21.96)
        1  =       376 + 281
       S-j2 = 25.50
       Sl  = 5.05
For  condition  2
       n  - 9.13(280) +  9.99(350)  _  g  6
       U2 ~        280 +  350
        <,  2   280(41.14)  + 350(35.59)  _ ,«
        S2  - - * - 280  + 350        - 38'
        S2 = 6.17
        HQ:  nx =
                  DD
*            2      _ i
                -   0 32
        »                    _  .    _ 3 17
        Z  =— - -        ~
For  a =  0.05      Za/2 = 1-96
        Z* > Z ,2 and HQ is rejected .   Therefore we conclude ^  #
The  mean sizes of particles from conditions 2 and 3 are therefore
considered to be significantly different.
                              2-G-5

-------
          G-4.  KQMOGOBOV - SMIRNOV FITNESS TESTING






       Applying the one sample test to the samples in condition 2



(Table 12)



       D    (sample 133) =0.103
        max     r


   for a = 0.05   Da = 1.36//n   n = sample size = 350   D  = 0.073
                                                          a

       D    > Dry    Sample 133 does not fit the given distribution
        max   -«       r


at  5% level of significance



       D    (sample 134) = 0.140



   for (X = 0.05   n = 280    % = 0-081



       D    > Dry-   Sample 134 does not fit the given distribution
        max


at the 5% level of significance.



       The two sample test is also applied at a - 0.05.



       D  =0.12                                  reference  (20)



       D    =  (0.37-0.23) = 0.14
        max    x


       D    > D  .  Therefore Samples 133 and 134 have significantly
        max    ot


different distributions .



       Applying the one sample test to the condition 3  samples



(Table 13)
       D       =0.0688   n - 376    a =0.05   D  =0.070

        raax!04



       D    < D
        max    a


Sample 104 fits the normal distribution



       D       =0.0844   n = 281   D  =0.081   a =0.05
        max128                      a




            > D ; Sample 128 does not fit  the distribution.
                            2-G-6

-------
       Applying the two sample test to condition 3 samples

       D^ =0.12 at a =0.05                                       (19)
       Dmax = (°-83

       Dmax < Da» SamPles 104 and 128 do not have significantly

different distributions .



             G-5.  SUMMARY OF STATISTICAL TESTS

       The above statistical tests, applied to mass and size count

data are summarized.

       Mass Concentration-Comparing Conditions

          M-! > M-2        M-3 > 1*2       ^3 > ^1

       Particle Sizing

          Replicate Reproducibility

          Condition 1    Condition 2   Condition 3
                         a!33 = CT134   a!04 =

          Differences between conditions

                         fi2 * M-3

          Frequency Distribution

          a)   Closeness of fit to normal distribution
              133 and 134 do not fit
              104 fits but 128 does not
          b)   Closeness of fit to each other - Reproducibility
              133 and 134 are different
              104 and 128 fit
                            2-G-7

-------
                         APPENDIX H
               ELECTRON MICROSCOPY PHOTOGRAPHS

       The following photographs were selected by the investigator
as typical of all the photographs of the 6 samples analyzed by
McCrone Associates.
       The reproductions are not altered with regard to magnifica-
tion.  The prints are reproduced from negatives made by Copycat,
Inc., Morgantown, using the McCrone photographs.
       Figures H-l and H-2 indicate the light colored chain like
agglomerates against the porous Nucleopore filter background.  The
"dark holes" represent the 0.2 |j,m pores.
       Figure H-3 represents the composition of part of an agglomerate.
The particles here are dark and are distinguished by light borders.
                            2-H-l

-------
 Figure H-l.  SEM Photograph of Sample 133
              (5000X).  1 cm.  = 2 urn.
Figure H-2. SEM Photograph of Sample 133
             (10000X) . 1 cm. - 1 \im.
                    2-H-2

-------
Figure H-3.  TEM Photograph of Sample 101
             (300,OOOX)
                    2-H-3

-------
           ELECTRIC ARC GENERATION

                      OF

          FRESH IRON OXIDE AEROSOLS
                    THESIS

       Submitted to the Graduate School
                      of
           West Virginia University
In Partial Fulfillment of the Requirements For
The Degree of Master of Science in Engineering
                      by

        William Fred Dimmick II, A.B.
                 Morgantown,
                West Virginia

                     1977
                      3-i

-------
                            ABSTRACT




      ELECTRIC ARC GENERATION OF FRESH IRON OXIDE AEROSOLS





     The development of a fresh metal oxide fine particle aerosol




generator was needed to provide necessary true-to-life conditions for




testing air pollution control and measurement equipment.  A generator




of fresh aerosols was considered to be applicable to toxicological




studies.  An electric arc metallizer employing dual  consumable  electrode




wires was developed for fresh metal oxide aerosol generation with the




guidance of Professor Benjamin Linsky.  The present  engineering problem




consisted of characterizing the aerosols produced by the  generator when




metallizing steel feed stock and investigating the influence of operation




variables on the conversion of mass from solid wire  to aerosol  particles.




     Generation of fine particle aerosols was done using  a  commercially




available electric arc metallizer to spray melted and vaporized feed stock




electrode wires into a 55 gallon barrel.  The spray  generated aerosol




was exhausted through a duct work.




     The generated aerosols were sampled on membrane filters for sub-




sequent particle size count analysis using scanning electron microscopy




and transmission electron microscopy.  The same samples were analyzed




for chemical composition by selected area electron diffraction.




     Membrane filters were also used to capture mass samples for deter-




mining the mass concentration of the aerosol in the exhausting duct works.




The mass concentration was used as the dependent variable  in a Latin




Square designed experiment where operation adjustments of  the  equipment




as manufactured were the independent variables.  These independent var-




iables were the wire feed rate, the open voltage across  the arc,  and the
                                3-ii

-------
pressure of the compressed air jet used to form the spray.


     The data shows the aerosols to be composed of agglomerated particles


formed from very fine magnetite particles.  The agglomerated particles


had a mean size of 0.4 ym by count whereas the very fine particles


have mean count diameters ranging from 0.0054-0.009 ym (5.4-9nm).  At


three different operating conditions three different very fine particle


means were found.  Each of these conditions produced an unique particle


size distribution.


     The mass concentration Latin Square experimental design found the


wire feed rate and the open voltage to influence changes in the mass


concentration of the aerosol.  A maximum mass concentration was found to

                       3
be approximately 2 gm/m  .  The variability produced by replication of a


data set was found to be  significant.


     In conclusion, the generator was found to produce very fine ferrous


particles that agglomerated into submicron sized particles.  Evidence


suggests that, with further characterization of the agglomerated very


fine particle aerosols, the generator will produce true-to-life fresh


aerosols with reproducible properties.  The aerosol mass concentration


was considered to be adequate for testing of air pollution control and


measurement  equipment.
                                 3-iii

-------
                        ACKNOWLEDGMENT






     The author wishes to graciously thank everyone involved in




the completion of the research.   His major professor and advisor,




Professor Benjamin Linaky, introduced him to the field and concepts of




fine particles.  His colleague Mike Naylor taught him the set up and




operation of the aerosol generating apparatus.   Don Stone and Craig




Repp, fellow graduate students,  helped in the collection of samples.




Craig Repp and the author designed and built a metallizing wire dispenser




to deliver feedstock from 50 pound wire coils to the generating apparatus.




     The author expresses gratitude for helpful suggestions from




Mr. Charles Ghastin of the Wall  Colmonoy Corporation and Mr. Richard




Shimps of McCrone Associates.  Suggestions and questions from Dr. S. R.




Borbash of the Industrial Engineering Department at West Virginia




University guided the researcher's statistical analysis.




     The author was a graduate air pollution trainee under Environmental




Protection Agency grant numbers  T900567010 and T900567020.  The financial




support for this research was funded as part of Environmental Protection




Agency grant R-801858-01-1.




     The author thanks his wife, Linda, for her patience during  the




time of his presently completed  studies.
                                3-iv

-------
                         TABLE OF CONTENTS

                                                           Page
ABSTRACT	3_i]L

ACKNOWLEDGEMENT	3_iv

LIST OF TABLES	>vll

LIST OP FIGURES	3_ix

LIST OF SYMBOLS, ABBREVIATIONS, AND CONSTANTS  ....   3_x

INTRODUCTION  	   3-1

STATEMENT OF ENGINEERING OBJECTIVES   	   3-2

REVIEW OF PERTINENT LITERATURE	3-3

PROCEDURE	3-6

    Equipment and Materials  	   3-6
    Operation of Generator Apparatus  	   3-10
    Sampling	   3-11
    Phenomenon Characterization Techniques  	   3-11
    Experiment:  Analysis of Phenomenon  	   3-13
    Data	3-16

ANALYSIS OF DATA	3-30

    Agglomerates	3-30
    Fine Particles	3-33
    Mass	3-46

DISCUSSION OF RESULTS	3-51

    Agglomerates 	   3-51
    Fine Particles	3-51
    Chemical Composition  	   3-53
    Mass Analysis	3-54

CONCLUSIONS	3-57

BIBLIOGRAPHY	3-60

APPENDICES	3-62

    Appendix A	3-63
    Appendix B	•	3-64
    Appendix C-l  Particle Sizing Pictures	3-65
    Appendix C-2  Sizing of the Agglomerates   .   .   .   .3-66
                             3-v

-------
TABLE OF CONTENTS (continued)
    Appendix C-3
    Appendix D-l
    Appendix D-2
    Appendix D-3
    Appendix D-4
    Appendix D-5
Mass Concentration Data
Latin Square ....
T Test Procedure
ANOVA  	
                            ....    3—68
                            ....    3-81
                            .   .   .   .    3-84
                            .   .   .   .    3-88
Stepwise Procedure 	    3-89
GLM and Residual Plots	3-92
    Appendix E	3-100
                               3-vi

-------
                          LIST OF TABLES


                                                             Page

 1.  Hobart steel metallizing wire (HB 18)  Assay .   .   .   .3-10

 2.  Aerosol generator operating conditions
       for particle size analysis	3-14

 3.  Comparison study operating conditions 	   3-14

 4.  Determination of orthogonal data
       collection points 	   3-15

 5.  Agglomerated data corrected for magnitude
       of magnification:  sample 715	3-17

 6.  Agglomerated data corrected for magnitude
       of magnification:  sample 714	3-17

 7.  Very fine particle distributions:  set 1	3-18

 8.  Very fine particle distributions:  set 2	3-20

 9.  Very fine particle distributions:  set 3	3-22

10.  Very fine particle distributions:  set 4	3-23

11.  Very fine particle distributions:
       samples 712 and 716	3-24

12.  Selected area electron diffraction data  .....   3-25

13.  Diffraction lines for magnetite
       and maghemite	3-25

14.  Latin square mass concentration
       data summary (gm/m3)	3-26

15.  Orthogonal mass concentration
       data summary (gm/nr)	3-27

16.  A listing of all the mass concentration data   .   .   •   3-28

17.  Two means equality test hypothesis
       decision-making criteria   	   3-31

18.  Statistics for agglomerate samples 	   3-33
                                 3-vii

-------
LIST OF TABLES (continued)
19.  Statistics for fine particle distributions   .   .   .    3-40

20.  Within group t-statistics for
       fine particle samples 	    3-41

21.  Between group t-statistics for
       fine particle samples    	    3-42

22.  Within group chi-square statistics  	    3-44

23.  Between group chi-square statistics 	    3-45

24.  Analysis of variance for latin square
       experimental design   	    3-47

25.  SAS ANOVA for latin square experimental
       design data	3-48

26.  SAS GLM table for orthogonal data	3-49

27.  SAS GLM table for all of the data	3-50

28.  Raw data for agglomerated particle count  ....    3-67

29.  Mass data for first latin square experiment  .  .  .    3-71

30.  Mass data for second latin square experiment .  .  .    3-75

31.  Mass data for orthogonal experiment	3-79

32.  Formula and definitions for latin square
       analysis of variance  	   3-82

33.  Calculation table for latin square design ....   3-83

34.  SAS T Test table	3-87

35.  Stepwise regression procedure for
       dependent variable Y  	   3-90

36.  SAS GLM table including residuals	3-93

37.  Relating wire feed rate and percent
       mass converted	3-100


                                 3-viii

-------
                         LIST OF FIGURES

                                                           Page

 1.   Equipment set up with close up  of
       arc gun assembly	    3-8

 2.   Example of data sheet	3-12

 3.   Set 1  frequency distributions	3-35

 4.   Set 2  frequency distributions	3-36

 5.   Set 3  frequency distributions	3-37

 6.   Set 4  frequency distributions	3-38

 7.   Sample 712 and 716 frequency distributions    .   .   .   3-39

 8.   Histograms of agglomerated particles   	   3-52

 9.   SEM picture sample 715	3-65

10.   SEM picture sample 714	3-65

11.   Explanation of data sheet and mass
       calculation symbols   	   3-70

12.   Graph:  wire feed rate versus standard
       residual	3-96

13.   Graph:  open voltage versus standard
       residual	3-97

14.   Graph:  predicted outcome (Yhat)
       versus standard residual 	   3-98

15.   Graph:  order of presentation versus
       standard residual  	   3-99

16.   Wire feed rate calibration	3-101
                                3-ix

-------
          LIST OF SYMBOLS,  ABBREVIATIONS AND CONSTANTS
ANOVA



GLM



T Test



SS



df



Yhat





SEM



TEM



SAED



lines


 d A



HEPA



R#




Pbar



P
 duct



HEPA A P



Venturi ^ P



Pitotflp





Duct T



Bag T



V
 open



AMPS
= analysis of variance



• general linear model



= a statistical test based on the t-distribution



= sum of squares



= degrees of freedom



= predicted outcome from an estimated descriptive

  equation



= scanning electron microscopy



= transmission electron microscopy



= selected area electron diffraction






* diffraction lines in angstroms (A)



= high efficiency particulate air



= reynolds number



« ambient barometric pressure




= pressure in exhaust duct




= pressure drop across HEPA filter



= pressure drop across calibrated venturi



= velocity pressure differential as measured by a

  standard pitot tube



= temperature in exhaust  duct



= temperature within bag  house



= open voltage across electrodes




= electrical current  in amperes
                                3-x

-------
LIST OF SYMBOLS, ABBREVIATIONS AND  CONSTANTS  (continued)
MAP



SAP



AE orifice



CF start



CF end



4CF



dT



T
 meter-start






T
 meter-end






 duct



STP
     3         3
 gm/m  = gm/sm
    acf



 fpm




 8



 P



 P



 d
main air jet pressure at machine



secondary air supply pressure



manometer differential on RAC sampler



dry gas meter reading before sampling



dry gas meter reading after sampling



sample volume



time of sampling in minutes




temperature of RAC thermometer at beginning of

sampling




temperature of RAC thermometer at end of sampling





velocity of gas stream in exhaust duct



standard temperature (273 °K) and pressure

(1 atmosphere)



grams per standard (STP) cubic meters



grams per actual cubic feet



feet per minute



acceleration due to gravity



density



dynamic viscosity



circular diameter
                                 3-xi

-------
LIST OF SYMBOLS, ABBREVIATIONS, AND CONSTANTS (continued)
lA = 10 10m - 10 8cm



grains/ft3 x 2.288



(*F-32) x 5/9



ft3/mlnute x 0.0283



diameter
  -7       -4        -1
10  mm =10   ym = 10  nm
       = grams/m



       -*C

          3
       = m /minute



       = a physical quantity
                                3-xii

-------
                          INTRODUCTION




     The generation of fresh dry metal aerosols for experimental




analysis, whether it is testing fabric filters for particle penetration




or examining the effect of inhaled cadium oxide particles on hamster




kidneys, is a necessity for proper similitude to many industrial events.




The known discrepancy between the occurrence of brass fever after ex-




posure to "fresh" zinc oxide fumes and the lack of occurrence with




exposure to stale, redispersed zinc oxide fumes (14) expresses the




validity of discriminating between fresh and stale particles.  Again,




the fact that the emission of fine particles as air pollutants has re-




ceived increasing attention in the past nineteen years (7) implies the




concurrent need of fine particle aerosol generators for evaluation of




engineering designs.  Eugene Grassel  (_4_) states recent scrutiny of fine




particles in the atmosphere is producing an aerosol generator market.




     A characteristic of fine particulate aerosols is the small rate




at which the particles settle due to  the force of gravity.  In Appendix A,




it is shown that a 0.5 ym diameter solid spherical particle with a unit




density of one gram per cubic centimeter falls less than one-half cen-




timeter in one hour  (approximately four and a half inches per day).




This property virtually eliminates the possibility of fine particles




settling quickly by gravity to earth's surface.  Presumably, the par-




ticles are blown about the atmosphere by winds.




     Fine particles can absorb and scatter light.  Aerosols with par-




ticles in the size range of visible light extinguish light  (8).




Waggoner and Charlson  (18) state that blue hazes in mountainous areas




may or may not be due to scattering by particles.  There  should be  no





                                3-1

-------
haze unless there are particles between the observer and the mountain




If the mountain is closer than 10 km (7 miles).   They also say that




light scattering increases with humidity,  but for relatively hygrophobic




systems the increase may be very slight.  Certainly extinction increases




with an increase in the light extinguishing particle number in the




atmosphere.



     As shown in Appendix B, a unit density spherical particle of diam-




eter 0.1 ym weighs one thousand times greater than a unit density spher-




ical particle with a 0.01 ym diameter.  Any air pollution emission cri-




terion that relies only on measurement by weight is not including all




the proper physically descriptive information.  The importance of char-




acterizing an aerosol by its particle size distribution has been known




in industrial hygiene for decades.  Measurement of a fine particulate




aerosol without considering the size distribution is incomplete.






               STATEMENT OF ENGINEERING OBJECTIVE




     Further research with the Electrospray aerosol generator  (as de-




scribed in Equipment and Materials) was needed to provide more  infor-




mation leading to a better engineering characterization for  the uses




of the equipment and its output.  A statistical-design approach was




incorporated with the object of furthering the knowledge  of  the var-




iables of solid feed material to particle mass concentration conversion.




These variables as represented in this research were  (1)  the open




voltage across the arc,  (2) the wire feed  rate,  and (3)  the pressure




of main gas jet which impinges upon the arc  to form the metallizing




spray.  All of these variables were obtainable from adjustment mech-




anisms routinely supplied as part of  the metallizer.  The dependent






                                3-2

-------
variable  (the outcome) was the aerosol mass concentration.




     Further characterization of  the aerosol produced by the Electro-




spray aerosol generator was also  an objective.  Naylor  (10) reported




on the results of Electrospray production of zinc oxide particle aerosols




from zinc consumable  electrodes.   In the current research, carbon steel




wire was  used to produce  iron oxide aerosols for fine particle size




distribution analysis.  Both the  aerosol characterizing analysis and




the mass  conversion analysis are  analyses in the sense  that they examine




a system  by looking at some of its elements and their relations.  A




simple comparative experiment also observed the effect  on aerosol char-




acteristics of faster gas and particle removal from the chamber (barrel)




where the gas and the particles received direct line-of-sight radiation




from the  arc.






                 REVIEW OF  PERTINENT LITERATURE




     Holmgren et_.al_.  (6)  reported a technique for production of fine




particles in a high  intensity arc. The feed material was incorporated




in  the anode and was  vaporized as a consequence of the  extreme temper-




atures produced by the arcing.  They explained that due to a sudden




drop in temperature  of the  vapor, as it was carried away from the arc,




fine particles were  formed  by condensation.




     Particles that  ranged  from 100 A  to 1000 A  (as measured by elec-




tron microscopy) were produced.   Mass  production  rates  of  10-20 Ib/hr




 (76-152 gm/min) under a practical running  condition  were  reported  for




generation of oxide-type  materials.




     Holmgren et.al.  reported  the production of Fe203  particles with




an  'equivalent sphere diameter1 of 658 A  (65.8 nm,  0.0658 ym).  They
                                 3-3

-------
described the aerosol particles us roughly spheroidal with a distinct




trend toward a hexagonal outline.   They stated that dispersion about




the mean varied with the physical  properties of the feed material.




     Pfender and Boffa (12) described a method of generating fine




particle aerosols in a high intensity arc.  Their apparatus included




transpiration cooling of the anode to facilitate the production of




smaller particles.  They stated that with low intensity arcs  coagulated




and chainlike particles with poorly controlled size distributions are




produced.  They referenced Holmgren .et.al^ to state that better re-




sults are obtained with high intensity arcs.




     Pfender and Boffa (12) reported generation of monodisperse




spheroidal shaped particles with a diameter ranging from approximately




100 to 1000 A as measured by a Whitby analyzer.  They did not report




the criterion used to determine monodispersity nor the mass production




rate.




     Formenti et^.a.1^.  (3) reported the production of fine metal oxide




particle aerosol from diffusion burning of volatile metallic chlorides




in a hydrogen oxygen flame.  They reported generating particles with




a range of diameters pf  100-2000 A.  They stated  that  the  diameter




of the particles and their morphology are controlled by (1)  the  tem-




perature of the flame, (2) the concentration  of  chloride  injected into




the burner, and (3) the residence time of the chloride  vapor  in  the




flame.  They reported production of Fe~0. particles  in  a  mixture of alpha




and gamma phases.  They did not report aerosol size  distribution dis-




persity or mass production rate.




     Grassell (4) discussed the development of aerosol  generators for







                                3-4

-------
Industrial research with respect to fine particles aerosols.  Two basic




criteria were used to evaluate the propriety of an aerosol for filter




 medium testing.  These criteria were  (1) sufficient quantity of the




aerosol for testing a 1000 cfm system and (2) the aerosol should re-




flect problems of both penetration and cleanability observed in real




life situations.  Grassell reported the need and development of an




aerosol generator for testing air cleaners exposed to diesel exhaust.




The generator built from an acetylene burner was relatively stable with




day to day reproducibility of filter media loadings varying by a. factor




of 10 when using repetitive settings of controls.




     Robert Hedden  (5) working with Linsky developed an arc generator




of a fine particle aerosol.  A single  consumable electrode functioning




as an anode was vaporized by the heat  generated by the arc.  The




vaporized metal condensed as it was quenched by a passing air stream




to form fine particles.  The iron oxide particles were sized by electron




microscopy to have an average diameter of 22  nm (220 A) .  The mass




production rate was found to be 1.95 gm/minute.  Hedden characterized




by weight the aerosol as containing 75% particles with effective diameter




less than 0.1 ym.



     Mike Naylor  (10) working with Dr. Richard Sears under Linsky




reported experimental results on a two consumable wire electrode,




electric arc aerosol generator.  Naylor produced zinc oxide aerosols




with the generating apparatus.  The apparatus consisted of a commercially




available electric arc metallizer, a settling barrel and an aerosol




exhausting system.  The same generator unit was  employed to produce




aerosols for this report.




                                3-5

-------
     The consumable electrodes were melted and vaporized in the intense


heat of the arc and subsequently the vapor condensed into fine particles


in an air stream.  These fine particles were found in agglomerated


chains as well as individually.  Naylor reported a range of average


diameter sizes from 6-11 nm as measured by transmission electron

                                                        3
microscopy (TEM).  A mass concentration of 0.65-2.0 gm/m  was found by

                             3
Naylor at approximately 7.8 m /minute.  This is a production rate of 5-


15.6 gm/minute.

     Amson (1) stated that even though erratic behavior of the con-


sumable-electrode arc system is not rare, choosing a suitable power


source and ambient gas with suitable settings for the open-circuit


voltage and the wire electrode feed speed  often leads to the system


operating in a 'quasi-steady state' mode.  He reported that the elec-


trode voltage drop plays a very small part in the behavior of the con-


sumable electrode system.  A decisive part is played by the electrode


stick out only when high resistivity electrodes are employed.  The


electrode stick out is the portion of the electrode wire from the current


transfer tip to the arc.



                           PROCEDURE



                     Equipment and Material


     The aerosol generating apparatus consisted of an Electrospray

metallizer that used an electric arc and a compressed air jet to spray


melted and vaporized consumable wire electrodes, a 55 gallon  oil  drum


connected to a supplemental air supply cleaned by a HEPA  (High  Efficiency


Air Particulate) filter, a draft fan with accompanying  duct works  from
                                3-6

-------
the oil drum, and a cloth filter particle collector (See Figure 1).




The metallizer system supplied the necessary power and material feed




delivery.  The duct work and fan provided for exhausting and sampling




the aerosol.  The baghouse cleaned the particle laden gas stream of




most of the particulates.




     The metallizer  (19) consisted of a power source, dual wire feeding




system, and an electric arc "gun." The metallizer power supply was a




three phase, 230 volt, 60 cycle AC to DC solid state full wave rectifier.




The DC supply could be varied from 20 to 40 volts.  The wire feeding




system consisted of a reproducibly variable drive motor connected to




a pair of interlocking pinch rollers  (with tension adjustments), wire




straighteners, and wire conduits for  simultaneous delivery of the dual




feedstock wire electrodes to the arc  gun.




     The electric arc gun provided the transfer of electric current to




the consumable feedstock electrodes and atomization and quenching of




the products of the arc.  The gun also provided a stationary structure




for proper  electrode alignment and consistent impingement of the main




gas (compressed air) jet on the arc zone by the spray nozzle.




     The arc gun was mounted through  a cutaway in the lid of the oil




drum.  A tip shroud which reduced entrainment of particles through  the




arc zone was mounted on the arc side  of the lid.  A spacer bar between




the gun assembly and the barrel lid permitted connection of a  second




air supply.



     The 55 gallon barrel was set horizontally in a rack to facilitate




connection with the metallizer. The barrel's lid clamp was used  to  hold




the lid "air tight" against a tape gasket.  An arc gun opening,  a






                                3-7

-------
       Wire
       dispenser
                        (top  view)
Electrospray
                      Secondary
             Arc gun  air
             head     adaptor
            Secondary air
            connection
                                               Electrode tip
                                               Main gas jet

                                               Electrode wire
                                               Shroud
                                               Lid
                        ttop view)
Figure 1. Equipment set-up with closeup of arc  gun  assembly.
                             3-8

-------
filtered diluting make-up air  inlet, and an exhaust air outlet were




provided in the lid.




     The barrel functioned as  a  settling chamber for the larger particles.




Particles larger than  30 ym  in diameter were collected with an efficiency




of approximately 95% by weight (10).  The  separation occurred from




gravity and a  centrifugal force  created by the 180 degree rotation of




the  gas stream upon entering the exhaust duct.




     The exhaust duct  began  one  third the  distance from the bottom of




the  barrel.  After  leaving the barrel, the duct was immediately directed




for  a  straight run  of  at  least 20 pipe diameters to the exhaust system




to the baghouse.  At a point 10  pipe diameters from the last duct




disturbance,  the sampling port was located.  A calibrated venturi flow




meter  was placed between  the sampling port and the exhaust pipe system




to the baghouse.  After the  venturi, a damper was placed for adjustment




of the total  air  stream flow rate.  The  cloth filter and fan completed




the  exhaust duct  system.




     The  sampling  train consisted of an  RAC  stack sampler with Greenburg-




Smith impinges connected to the sampling  probe.  A S-hook nozzle




was  connected to  a  membrane  filter holder.  The filter holder retained




an appropriate filter  for collecting the aerosol samples.




     The  stack sampler provided for adjustment of probe tip entrance




velocity  as well  as a  measurement of dry gas volumetric flow.  The




impinger  system cooled and then dewatered the  gas for measurement by




the  dry gas meter.



     Membrane filters  were used to collect samples  for  subsequent




particle  count or mass analysis.  Membrane filters  with 0.45  ym pore
                                 3-9

-------
sizes were used for mass analysis and filters with 0.20 ym pore sizes

were used for particle count analysis.  The 0.20 ym membrane filters

were also chosen because they provided good surface properties for elec-

tron microscopic particle counting.

     The wire used in this research was stainless steel welding or

metallizing wire.  Hobart steel welding wire (HB #18)  was used for pro-

duction of fine particle aerosols for particle count and morphologic

analysis.  Wall Colmonoy wire #10 was used to produce fine particle

aerosols for mass analysis.  The following was an assay of the Hobart

steel wire (16):
               Table 1.  Hobart steel metalizing
                         wire (HB 18) assay


               Fe                         96.84%
               Mn                          1.90%
               Si                          0.65%
               Mo                          0.45%
               C                           0.12%
               P                           0.02%
               S                           0.02%
                Operation of Generator Apparatus

     The Electrospray control unit was operated with spray settings

within the range recommended by Wall Colmonoy (19).  The settings for

the wire feed rate varied within 30-60 (see Appendix E for units), the

open voltage ranged from 20-31 volts, and the main jet pressure ranged

from 40-60 psi.  The experimenter set the control variables and began

operation.  If the apparatus ran without sputtering  (an audio character-

istic of inefficient operation) for two minutes, the apparatus was con-

sidered operating stably.  The closed voltage (voltage across arc during

operation) was not used as a criterion for stability.


                                3-10

-------
                            Sampling




     After generation for at least two minutes, samples were taken




either for mass conversion analysis or particle size distribution




analysis.  Consecutive randomly ordered samples were taken with a




two minute shut down between each sampling run.




     Sampling for mass analysis consisted of pre-trial determination




of isokinetic sampling conditions and then the actual sampling.  After




adjusting the stack sampler for isokinetic sampling as suggested by the




accompanying operation manual  (13) , samples were taken such that at




least 10 mg of mass was deposited on the tared filter.  All necessary




data was recorded before and during the trials on a form such as in




Figure 2.




     Sampling the aerosol for  subsequent particle size distribution




analysis consisted of estimating isokinetic conditions and then collection




on appropriate membrane filters with sampling times of three, five and




seven seconds.  Then, the filters were enclosed in a plastic filter




holder and were shipped by United Postal Service to McCrone Associates,




Inc. in Chicago for analysis by scanning electron microscopy (SEM), TEM,




and selected area electron diffraction (SAED).





             Phenomenon Characterization Techniques




     The sizing of the fine particles was done by McCrone Associates




through TEM with a particle size count of an enlarged photomicrograph.




The fine particles counted by  size were believed to agglomerate  into




larger particles prior to sampling.  The size limits  that the McCrone




laboratory considered appropriate for adequate TEM characterization




ranged from less than 2 nm to  greater than 20 nm by increments of 1 nm.
                                3-11

-------
 Wire Type:                Pbar-           Dry Bulb TemPerature:
 Date:                         '           Wet Bulb Temperature:



     Sample #



     Barrel Temp CO



71   HEPA  Ap
     p     — p    _   *•*•«

      duct    bar    13.6


     Venturi A p



     Duct T CO



     Bag T CO



     Pitot A p



     Wire Feed Rate



     Amps



     V
      open


     MAP



     SAP



     A  H orifice



     CF start



     CF end



4 I   A  CF



     A Time
5
     T
      meter-start


     T
      meter-end


     Nozzle  Velocity



     CFM (       /
     V     - 234
     Vduct   "*
                 V   duct







                Figure 2.  Example  of  data sheet





                                3-12

-------
SEM was used to check samples for the proper laoding  (particles per




area) necessary for TEM particle sizing.  SAED analysis was performed




to determine phase characteristics of the particulate matter.




     Sizing the agglomerates of fine particles characteristic of the




aerosol was done by the experimenter with SEM photomicrographs.  SEM




negatives were enlarged noting the actual enlargement onto Kodak RC




paper as suggested by Mr. Richard Shimps of McCrone Associates (17).




The particles were sized with an uncalibrated clear plastic metric ruler.




A minimum of 150 particles were counted by size and then corrected to




actual size with the known magnitude of magnification.




     Mass determination analysis was done by the experimenter weighing




the exposed sample membrane filters on an analytical  balance accurate




to + 0.5 mg.  With the mass determined by the TARE method and sampling




volumetric flow  (STP) determined, an iron oxide mass  concentration was




found.





              Experiment:  Analysis of Phenomenon




     Four sets of samples were taken at four different operating con-




ditions for particle size analysis.  The conditions were shown in




Table 2.



     Two sets of samples were taken under identical conditions with the




exception that the gas stream velocity within the duct work was varied.




These conditions were shown in Table 3.




     Mass samples were taken to conform to a Latin Square experimental




design.   The Latin Square design is explained in Appendix D-l.  Two




Latin Square designs were randomly chosen and operating conditons




(variables) were designated to the statistical design structure.  The
                                3-13

-------
        Table 2.   Aerosol generator operating conditions
                  for particle size analysis
     Sample
     Numbers
      600
      601
      602

      603
      604
      605

      607
      611

      609
      612
      Condition
      Set Number
       Wire Feed
         Rate
                        25
                        25
                        25

                        25
                        25
                        25

                        50
                        50

                        50
                        50
       Open
      Voltage
                     20
                     20
                     20

                     20
                     20
                     20

                     22
                     22

                     22
                     22
      Main Jet
      Pressure
       (psi)

         60
         60
         60

         45
         45
         45

         60
         60

         45
         45
        Table 3.   Comparison study operating conditons
Sample
Number
 715
 716

 712
 714
Condition
Set Number
Wire Feed
  Rate
                 50
                 50

                 50
                 50
 Open
Voltage
(volts)

  22
  22

  22
  22
Main Jet
Pressure
  (psi)

  60
  60

  60
  60
Velocity of
Gas Stream
   (fpm)

   1300
   1300

   1850
   1850
                                3-14

-------
samples for the structure were randomly ordered and then the samples

were taken.  Two trials, the second trial a replicating trial, were

performed; the trials occurred on different days.

      Another set of mass samples were taken to supplement the Latin

Square design.  A collection of orthogonal points were selected about

an observed maximum point in each of the operating variables; a wire

feed rate of 60, an open voltage of 29, and a main jet pressure of 45

were considered midpoints.  A unit "distance" was added to and subtracted

from each midpoint to develop a resulting eight orthogonal samples as

follows  in Table 4.
              Table  4.  Determination of orthogonal
                       data  collection points
                    Wire Feed    Open Voltage
                      Rate           (Volts)
Midpoint value         60
Unit  distance           5

Resulting  points
   Additive            65
   Subtractive         55

Resulting  samples
   Condition
29
 1
30
28
           Main Jet
           Pressure  (psi)
45
 5
50
40
1
2
3
4
5
6
7
8
65
55
55
65
65
65
55
55
28
28
30
30
30
28
30
28
40
50
40
40
50
50
50
40
 These resultant sampling conditions were randomly order  and then samples

 were taken.
                                 3-15

-------
                              Data




     The aerosol size distributions found for the agglomerated fine




particles of sample 714 and 715 were presented in Tables 5 and 6.  The




percentages of the total number of particles and the cumulative per-




centages were also given.  The raw data uncorrected for the different




magnifications were presented in Table 27.




     The fine particle size distributions found for the various sets




of operating conditions were presented in Tables 7, 8, 9, 10, and 11.




The percentages and cumulative percentages were also given.




SEM pictures of the agglomerated very fine particles were shown in




Appendix C-l.




     In Table 12, the results of the SAED analysis were presented.  In




Table 13, ASTM electron diffraction lines were given for magnetite




(Fe-,0.) and maghemite (r-Fe^,).




     The mass concentration data from the two Latin Square designs




was presented in Table 14.  The mass concentration data from the ortho-




gonal collection of data was presented in Table 15.  Table 16 contains




a listing of all the mass concentration data.  The     data and method




of mass concentration calculation for the Latin Square and the orthogonal




designs were presented in Appendix C-3.
                               3-16

-------
      Table 5.  Agglomerated data corrected for
                magnitude of magnification:  sample 715

Class    Frequency     Cumulative     Percent      Cumulative
Mark                   Frequency                   Percent
(pm)

0.09        141           141           47.0          47.0
0.28         56           197           18.7          65.7
0.43         31           228           10.3          76.0
0.61         23           251            7.7          83.7
0.78         10           261            3.3          87.0
0.96          7           268            2.3          89.3
1.1          11           279            3.7          93.0
1.3           4           283            1.3          94.3
1.5           4           287            1.3          95.7
1.7           3           290            1.0          96.7
1.8           1           291            0.3          97.0
2.3           1           292            0.3          97.3
2.6           8           300            2.7         100.0
      Table  6.  Agglomerated data  corrected for
                magnitude  of magnification:   sample 714

Class     Frequency      Cumulative   Percent       Cumulative
Mark                     Frequency                   Percent
0.09         156           156           55.3          55.3
0.27           41           197           14.5          69.9
0.46           19           216             6.7          76.6
0.64           17           233             6.0          82.6
0.82           14           247             5.0          87.6
1.0             8           255             2.8          90.4
1.2             7           262             2.5          92.9
1.4             3           265             1.1          94.0
16             3           268             1.1          95.0
17             6           274             2.1          97.2
23             2           276             0.7          97.9
2*5             1           277             0.4          98.2
26             1           278             0.4          98.6
27             4           282             1.4          100.0
                                 3-17

-------
                   Table 7.  Very fine particle
                             distributions set 1
 Class
 Mark
 (nm)

 1.5
 2.5
 3.5
 4.5
 5.5
 6.5
 7.5
 8.5
 9.5
10.5
11.5
12.5
13.5
14.5
15.5
16
17
18
19
21
,5
,5
,5
,5
 Class
 Mark
 (nm)

 1.5
 2.5
 3.5
 4.5
 5.5
 6.5
 7.5
 8.5
 9.5
10.5
11.5
12.5
13.5
14.5
15.5
16.5
17.5
          Frequency
18,
19.
21
13
53
53
37
33
23
18
10
18
19
19
17
21
 4
 7
 7
 9
 3
 6
16
             Frequency
             8
            37
            55
            65
            45
            20
            15
            25
            55
            16
            40
             7
             9
            13
            13
             2
             2
             3
             6
            10
            Sample 600

             Cumulative
             Frequency
    13
    66
   119
   156
   189
   212
   230
   240
   258
   277
   296
   313
   334
   338
   345
   352
   361
   364
   370
   386

Sample 601

 Cumulative
  Frequency


     8
    45
   100
   165
   210
   230
   245
   270
   325
   341
   381
   388
   397
   410
   423
   425
   427
   430
   436
   446
                  Percent
                                           Percent
 1.8
 8.3
12.3
14.6
10.1
 4.5
 3.4
 5.6
12.3
 3.6
 9.0
 1.6
 2.0
 2.9
 2.9
 0
 0
 0.7
 1.4
 2.2
                                                  .5
                                                  .5
Cumulative
Percent
3.4
13.7
13.7
9.6
8.6
6.0
4.7
2.6
4.7
4.9
4.9
4.4
5.4
1.0
1.8
1.8
2.3
0.8
1.6
4.1
3.4
17.1
30.8
40.4
49.0
54.9
59.6
62.1
66.8
71.8
76.7
81.1
86.5
87.6
89.4
91.2
93.5
94.3
95.9
100.0
                                            Cumulative
                                            Percent

                                               1.8
                                              10.1
                                              22.4
                                              37.0
                                              47.1
                                              51.6
                                              54.9
                                              60.5
                                              72.9
                                              76.5
                                              85.4
                                              87.0
                                              89.0
                                              91.9
                                              94.8
   95
   95
   96
   97.8
  100.0
                                                 3
                                                 7
                                                 .4
                                 3-18

-------
                       Table 7.   Continued

                            Sample 602

 Class       Frequency       Cumulative       Percent       Cumulative
 Mark                        Frequency                      Percent
 (nm)

 1.5            24              24              5.5            5.5
 2.5            79             103             18.2           23.7
 3.5            57             160             13.1           36.8
 4.5            21             181              4.8           41.7
 5.5            25             206              5.8           47.5
 6.5            43             249              9.9           57.4
 7.5            20             269              4.6           62.0
 8.5            20             289              4.6           66.6
 9.5            22             311              5.1           71.7
10.5            13             324              3.0           74.7
11.5            33             357              7.6           82.3
12.5            16             373              3.7           85.9
13.5             2             375              0.5           86.4
14.5             6             381              1.4           87.8
15.5             9             390              2.1           89.9
16.5             9             399              2.1           91.9
17.5            12             411              2.8           94.7
18.5             2             413              0.5           95.2
19.5             9             422              2.1           97.2
21              12             434              2.8          100.0
                                 3-19

-------
                   Table 8.   Very fine particle
 Class
 Mark
 (nm)

 1.5
 2.5
 3.5
 4.5
 5.5
 6.5
 7.5
 8.
 9,
10.5
11.5
12.5
13.5
14.5
15.5
16,
17.
18,
19.
21
 Class
 Mark
 (nm)

 1.5
 2.5
 3.5
 4.5
 5.5
 6.5
 7.5
 8.5
 9.5
10.5
11.5
12.5
13.
14.
15.5
16.5
17.5
18.
19.
21
  .5
  .5
  ,5
  ,5
Frequency
    2
   45
   58
   34
   29
   18
   13
   18
   30
   11
   16
   14
   12
   19
    7
    7
   11
    4
    7
   14
             Frequency
    3
   51
   66
   18
   33
   19
   13
   14
   39
    8
   21
   12
    9
   10
    5
    7
    7
    5
    9
   19
 distributions set 2
Sample 603

 Cumulative       Percent
 Frequency
      2
     47
    105
    139
    168
    186
    199
    217
    247
    258
    274
    288
    300
    319
    326
    333
    344
    348
    355
    369

Sample 604

 Cumulative
 Frequency


      3
     54
    120
    138
    171
    190
    203
    217
    256
    264
    285
    297
    306
    316
    321
    328
    335
    340
    349
    368
 0.5
12.2
15.7
 9.2
 7.9
 4.9
 3.5
 4.9
 8.1
 3.0
 4.3
 3.8
 3.3
 5.2
 1.9
 1.9
 2.9
 1.1
 1.9
 3.8
                                 Percent
             Cumulative
             Percent
                                                              0.
                                                             12.
                                                             28.4
                                                             37.6
                                                             45.5
                                                             50.4
                                                             53.9
                                                             58.8
                                                             66.9
                                                             69.9
                                                             74.2
                                                             78.0
                                                             81.3
                                                             86.5
                                                             88,
                                                             90,
                                                             93-
                                                             94,
                                                             96
                                                            100.0
             Cumulative
             Percent
8.8
13.8
17.9
4.9
9.0
5.2
3.5
3.8
10.6
2.2
5.7
3.3
2.4
2.7
1.4
1.9
1.9
1.4
2. A
5.2
0.8
14.6
32.6
37.5
46.5
51.6
55.2
59.0
69.6
71.7
77.4
80.7
83.2
85.9
87.2
89.1
91.0
92.3
94.8
100.0
                                 3-20

-------
                       Table 8. Continued

                            Sample 605

 Class       Frequency       Cumulative       Percent       Cumulative
 Mark                        Frequency                      Percent
 (nm)

 1>5             4                4             1.1           1.1
 2'5            41               45            11.5          12.6
 3.5            56              101            15.7          28.3
 4-5            42              143            11.7          40.1
 5-5            21              164             5.8          46.0
 6-5            19              183             5.3          51.4
 7-5            17              200             4.7          56.1
 8-5            17              217             4.7          60.9
 9-5            29              246             8.1          69.1
10-5             9              255             2.5          71.6
H-5            17              272             4.7          76.4
12.5            12              284             3.3          79.7
13.5            12              296             3.3          83.1
14.5            11              307             3.0          86.2
15.5             5              312             1.4          87.6
16.5             8              320             2.2          89.8
17.5             8              328             2.2          92.1
18.5             4              332             1.1          93.2
19.5             8              340             2.2          95.5
21.             16              356             4.4          99.9
                                 3-21

-------
 Class
 Mark
 (nm)

 1.5
 2.5
 3.5
 4.5
 5.5
 6.5
 7.5
 8.5
 9.5
10.
11.
12.5
13.5
14.
15.
16.
17.
18.
19.
21
.5
.5
,5
.5
.5
.5
,5
,5
                Table 9.  Very fine particle
                          distributions  set 3
                         Sample 607

          Frequency       Cumulative       Percent
                          Frequency
 Class
 Mark
 (nm)
 1
16
47
40
20
24
 6
12
11
 9
19
18
10
12
11
 9
 5
 6
 6
14
           Frequency
      1
     17
     64
    104
    124
    148
    154
    166
    177
    186
    205
    223
    233
    245
    256
    265
    270
    276
    282
    296

Sample 611

 Cumulative
 Frequency
 0.3
 5.4
15.8
13.5
 6.7
 8.1
 2.0
 4.0
 3.7
 3.0
 6.4
 6.0
 3.3
 4.0
 3.7
 3.0
 1.6
 2.0
 2.0
 4.7
                              Percent
             Cumulative
             Percent


               0.3
               5.7
              21.6
              35.1
              41.8
              50.0
              52.0
              56.0
              59.7
              62.8
              69.2
              75.3
              78,
              82,
              86,
              89,
              91
              93
              95
                                                           99.9
             Cumulative
             Percent
 2.5
 3.5
 4.5
 5.5
 6.5
 7.5
 8.5
 9.5
10.5
11.5
12,
13,
14,
15,
16,
17.
18.
19.
21
              13
              54
              46
              12
              25
               8
               4
               7
               7
              28
              14
              10
               7
               8
               5
               5
               1
               5
              15
                 13
                 67
                113
                125
                150
                158
                162
                169
                176
                204
                218
                228
                235
                243
                248
                253
                254
                259
                274
 4.7
19.7
16-. 7
 4.3
 9.1
 2.9
 1.4
 2.5
 2.5
10.2
 5.1
 3.6
 2.5
 2.9
 1.8
 1.8
 0.3
 1.8
 5.4
                                   4.7
                                  24.4
                                  41.2
                                  45.6
                                  54.7
                                  57.6
                                  59.1
                                  61.6
                                  64.2
                                  74.4
                                   79.
                                   83.
                                   85,
                                   88.
                                   90,
                                   92,
                                   92
                                   94
                   5
                   2
                   7
                   6
                   5
                   3
                   ,7
                   ,5
                                   99.9
                                 3-22

-------
                  Table 10.
 Class
 Mark
 (nm)

 1.5
 2.5
 3.5
 4.5
 5.5
 6.5
 7.5
 8.5
 9.5
10.
11,
12,
13,
14,
15,
16,
17.5
18.
19.
21
 Class
 Mark
 (nm)

 1.5
 2.5
 3.5
 4.5
 5.5
 6.5
 7.5
 8.5
 9.5
10.5
11.5
12.5
13.5
14.5
15.5
16.5
17.5
18,
19,
21
Frequency
   18
   55
   22
   16
   26
   22
   20
   18
   18
    7
   12
    9
    8
    6
    8
    5
    6
    8
    3
   11
             Frequency
    16
    47
    23
    19
    22
    21
    26
    17
    13
    11
    12
    12
    11
     8
     5
     4
     7
     8
     6
     9
 Very fine particle
 distributions set 4
Sample 609

 Cumulative       Percent
 Frequency
     18
     73
     95
    111
    137
    159
    179
    197
    215
    222
    234
    243
    251
    257
    265
    270
    276
    284
    287
    298

Sample 612

 Cumulative
 Frequency


     16
     63
     86
    105
    127
    148
    174
    191
    204
    215
    227
    239
    250
    258
    263
    267
    274
    282
    288
    297
 6.0
18.4
 7.3
 5.3
 8.7
 7.3
 6.7
 6.0
 6.0
 2.3
 4.0
 3.0
 2.6
 2.0
 2.6
 1.6
 2,0
 2.6
 1.0
 3.6
                                              Percent
 5.3
15.8
 7.7
 6.3
 7.4
 7.0
 8.7
 5.7
 4.3
 3.7
 4.0
 4.0
 3.7
 2.6
 1.6
 1.3
 2.3
 2.6
 2.0
 3.0
Cumulative
Percent


   6.0
  24.4
  31.8
  37.2
  45.9
  53.3
  60.0
  66.
  72,
                                                                .1
                                                                .1
  74.4
  78,
  81,
  84,
  86,
  88,
  90,
  92.6
  95,
  96,
 100.0
Cumulative
Percent


   5.3
  21.3
  28.9
  35.3
  42.7
  49.8
                                                                 .5
                                                                 .3
                                                                 .6
                                                                 .4
                                                                 .1
   58
   64
   68
   72.3
   76.4
   80
   84
   86.8
   88.5
   89.8
   92.2
   94.9
   96.9
   99.9
                                  3-23

-------
           Table 11.  Very fine particle distributions
                      for samples 712 and 716
.5
.5
 Clans
 Mark
 (nm)

 1.5
 2.5
 3.5
 4-5  i
 5.5
 6.5
 7.5
 8.5
 9-5
10.5
11.5
12.5
13.5
14.
.15.
16.5
17.5
18.5
19.5
21
 Class
 Mark
 (nm)
 1.5
 2.5
 3.5
 4.5
 5.5
 6.5
 7.5
 8.5
 9.5
10.5
11.5
12.5
13.5
14.5
15.5
16.5
17.5
18,
19,
21
          Frequency
 26
148
 63
 39
 24
 19
 23
 24
 21
 20
 14
 11
 13
 10
 11
 11
 10
  4
  2
  4
             Frequency
              17
              49
              66
              47
              35
              28
              19
              23
              18
              18
              15
              19
              23
              17
              12
               9
               6
              11
               9
               8
             Sample 712

              Cumulative
              Frequency
     26
    174
    237
    276
    300
    319
    342
    366
    387
    407
    421
    432
    445
    455
    466
    477
    487
    491
    493
    497
Sample 716

 Cumulative
 Frequency


     17
     66
    132
    179
    214
    242
    261
    284
    302
    320
    335
    354
    377
    394
    406
    415
    421
    432
    441
    449
                  Percent
                                                5.2
                                               29.7
                                               12.6
                                                7.8
                                                4.8
                                                3.8
                                                4.6
                                                4.8
                                                4.2
                                                4.0
                                                2.8
                                                2.2
                                                2.6
                                                2.0
                                                2.2
                                                2.2
                                                2.0
                                                0.8
                                                0.4
                                                0.8
                                            Percent
Cumulative
Percent

  5.2
 35.0
 47.6
 55.5
 60.
 64.
 68.8
 73.6
 77.8
 81.8
 84.7
 86.9
                                                             .3
                                                             .1
 89
 91
 93
 95.9
 97.9
 98.7
 99.1
.5
.5
.7
                                                           99.9
                                             Cumulative
                                             Percent
3.7
10.9
14.6
10.4
7.7
6.2
4.2
5.1
4.0
4.0
3.3
4.2
5.1
3.7
2.6
2.0
1.3
2.4
2.0
1.7
3.7
14.6
29.3
39.8
47.6
53.8
58.1
63.2
67.2
71.2
74.6
78.8
83.9
87.7
90.4
92.4
93.7
96.2
98.2
100.0
                                  3-24

-------
    Table 12.  Selected area electron
               diffraction data
Sample      602      604      611      609
lines
d A
^» **




Sample
lines
d A
v* c\







1.478
1.615

2.094
2.526
2.97
712

1.257
1.320
-
1.628
1.721
2.106
2.54
2.96

1.485
1.615
1.710
2.10
2.526
2.97
716

-
-
1.487
1.618
1.713
2.101
2.53
2.97

1.489 1.476
1.615 1.615
1.706
2.10 2.11
2.526 2.517
2.962 2.961










     Table 13.  Diffraction lines for
                magnetite and maghemite


   Magnetite, Fe-jO^             Maghemite, y-Fe

   ASTM Card 19-629             ASTM Card 15-615
lines

d A




•

1.266
1.328
1.485
1.616
1.715
2.099
2.532

1.008 s
1.702 s
2.089 s
2.521 s
2.950 s

s = lines marked with s
           2.967             include those of the
                             spinel structure
                     3-25

-------
              Table 14.  Latin  square  mass concentration
                         data summary  (gm/sm3)
                    Main Air Pressure (psi)
      25
8>    27
ca
•M x-\
r-l CO
O 4J
0)
p.
O
      31
              40
                        45
50
55
0.92
A
1.19
B
1.51
C
1.70
D
1.11
B
1.28
C
1.77
D
1.05
A
1.37
C
1.74
D
1.07
A
1.31
B
1.03
D
0.81
A
1.48
B
1.18
C
                                                       Replicate  1
                       Wire Feed Rates:

                           A  -  30
                           B  -  40
                           C  -  50
                           D  -  60
                    Main Air  Pressure (psi)
              40
                         45
50
55
       25
0)
00

2~  27
rH CO
O -U
> H

gS  29
       31
1.00
A
1.50
B
1.43
C
1.67
D
1.22
B
1.55
C
1.38
D
1.19
A
1.46
C
1.83
D
1.13
A
1.92
B
1.81
D
1.15
A
1.38
B
1.26
C
                                                        Replicate 2
                                 3-26

-------
             Table 15.   Orthogonal mass concentration
                         data summary (gm/m3)
                     Main Air Pressure (psi)

-------
Table 16.  A listing of all the mass concentration data
Sample
Number
Mass Concen.
(gm/m3)
Temp . in
Exhaust
Duct (°C)
Wire
Feed
Rate
Open
Voltage
(Volts)
Main Air
Pressure
(psi)

563
702
562
566
581
554
558
705
550
707
564
565
704
529
703
561
560
576
701
706
573
700
571
575
2.16
2.13
1.92
1.83
1.81
1.77
1.74
1.71
1.7
1.7
1.67
1.55
1.52
1.51
1.51
1.5
1.48
1.46
1.44
1.44
1.43
1.41
1.38
1.38
82
90
62
80
80
85
80
95
72
85
80
77
80
68
72
71
70
76
72
85
68
70
58
76
50
55
40
60
60
60
60
55
60
55
60
50
45
50
40
40
40
50
45
55
50
45
40
60
31
30
31
27
25
29
27
30
31
28
31
27
28
29
30
27
29
25
28
28
29
30
29
29
. 55
50
50
50
55
45
50
40
40
50
40
45
50
40
40
40
55
50
40
40
40
50
55
45
                            3-28

-------
Table 16.  Continued
Sample
Number
Mass. Concen.
(gm/m3)
Temp . in
Exhaust
Duct (°C)
Wire
Feed
Rate
Open
Voltage
(volts)
Main Air
Pressure
(psi)

555
546
548
572
553
568
549
569
570
547
556
552
559
582
557
551
1.37
1.31
1.28
1.22
1.19
1.19
1.18
1.15
1.13
1.11
1.07
1.05
1.03
1.0
0.92
0.81
76
73
74
65
73
57
75
58
54
64
60
53
70
59
57
60
50
40
50
40
40
30
50
30
30
40
30
30
60
30
30
30
25
31
27
25
27
31
31
27
29
25
29
31
25
24
25
27
50
50
45
45
40
45
55
55
50
45
50
45
55
40
40
55
           3-29

-------
                         ANALYSIS OF DATA





                           Agglomerates



     In the comparative experiment samples 715 and 714 were chosen as



representative of the agglomerates found in the samples.  The agglomerates



measured by the experimenter were found to have means of 0.41 ym



(S=0.53 ym) and 0.40 ym (S=0.54 ym) as shown in Appendix C-2.  The follow-



ing analysis used the student t-test for determining the equality of ex-



pected populations of two different samples.  The implication was that



the means were equal if the difference between the two means was equal



to zero.



     The null hypothesis that the difference between the two means was



equal to zero:




                         V  Uru2=0



was tested against the alternative hypothesis that the difference was



not equal to zero,



                         HA:  uru2XO.




To determine which hypothesis was true, the t-test statistic was used.



This statistic was determined by the following general formula:
                          cal   S-
                                 xp



where x..-x_ was greater than or equal to zero and S-  was the pooled




standard deviation as defined in Appendix D-2.  The t   1 value was  com-



pared to a tabulated t  .^ (v) statistic value where the a-  error  was



equal to 0.10 and the degrees of freedom  (v) were set equal to  infinity



because the sample sizes were much larger than 30.  As  shown in Appendix D-2,
                                 3-30

-------
the hand calculations  showed  the  t value  to be 0.194.  The t     (v-«)



value was found  to be  1.645.




     The criteria for  hypothesis  test was as  follows:






Table 17.  Two means equality test hypothesis decision-making criteria
                    If  tcal -  ta/2>  then Ho is accePted
                                    and H  is rejected.
                                         A






                    If  'cal  >  '0/2 f  then Ho ls reJected


                                    and H  is accepted .





Because tcal <  t  ,2,the null  hypothesis, H :  u -u =0, is accepted.  This





implied that the  two sample means came from the same population; this was



tested for 90%  confidence.



     The above  student t-test was only applicable if the variances of



the samples came  from  the same population.  The F-statistic was used to



determine if the  variances  came  from  the same population.  Where
                           cal    2


                                S2


       2                                  •
with S*  equal to the variance  of sample  1 and F   , having an F-distri-
      1                                         cax



bution with degrees of freedom  v.. for  s-  and v_ for s_ was compared to a




tabulated F (v, ,v_) and found to be less  than or equal to the F  value,
           a  1  2                                             ex



the variances were considered to be from  the sample population.  If the



variances were different,  the Smith-Satterthwaite  test (9) was used to



determine the equality of  two means .



     The F    was found to be 1.05 and F  (v. ,v_) was not found in any
          cal                           a 1  2



                                 3-31

-------
statistical table.   But,  if both of  the degrees of  freedom (v.,  and v_)




approach large values (>  100),  values of Fa(vi»v2)  approach 1.00.   It was



assumed that F  n   was greater  than  F  and therefore the Smith-Satterthwaite
              cal                    ct



test was needed for the t-statistic  test for equality of means.



     The hypothesis testing using the Smith-Satterthwaite t-statistic is



the same as the equivalent variance  t-eest with the exception that  the



t-value was found by the following equation:
                               Xrx2
                        "cal
                               nl     n2
and that the degrees of freedom (df)  were found by the following equation:
        df
s 2     2

!i_ + !2_


nl    n2
                                   V1
                                         V1
     The
t  1  value was found to be 0.225 with degrees of freedom 576.
The t  ,„ value was still 1.635.  The outcome of the hypothesis that the




means were equal was not changed by the fact that the variances were not



equal.



     The coefficient of variance was also calculated for each sample dis-



tribution and was presented in Table 1. along with the means  (x), standard



deviation (S), and standard errors of the mean for each sample.  The co-



efficient of variance was an indicator of the monodispersity of a  size



distribution (2) and is given by the following formula:
                                 3-32

-------
                         C.V. -  S/x




The coefficient of variance for  the agglomerate size distributions of




samples  714 and 715 were 1.35 and  1.29,  respectively.  Any sample size




distribution with a coefficient  of variance  less than 0.10 (2_) was con-



sidered  raonodisperse.







           Table 17.   Statistics  for agglomerate samples
mple No.
714
715
Mean
(ym)
0.40
0.41
Standard
Deviation
(ym)
0.54
0.53
Standard Error
of the Mean (ym)
O..Q3
0.03
Coeffi
Varian
135
129
                           Fine  Particles




      The  fine  particle size  distribution  data was broken down into six




 sets  of data contingent on the  six sets of  conditions.  Two basic tests




 were  performed on  the  fine particle size  distribution data: (1) within




 group t-test for equality  of means and between group t-test for equality




 of means  and  (2) within group and  between group chi squared test for




 equality  of frequency  distributions.  Plotted frequency distribution were




 shown in  Figures 3,  4,  5,  6, and 7.




      The  SAS T-TEST  and MEANS (15) procedures were employed to determine




 relevant  statistics  for the  data.   The MEANS procedure calculated the means,




 standard  deviations, standard errors  of the mean, and the coefficients of




 variance  as presented  in Table  19.  The T-TEST procedure calculated the




 F-statistic for comparison of two  sample  variances and the t-statistic for




 determination  of mean  equality.  A summary  of the t-test statistics for the




within group analysis was  presented in Table 20.  A summary of t-test






                                 3-33

-------
statistics for between group analysis was presented in Tables 20 and 21.



The tables were structured for presentation of statistics concerning the



intersecting samples.



     The statistic in the middle of each block was the t-value calculated



for the two mean t-test where
                              s-
                               xp



with the hypothesis that the means of the intersecting samples were equal.



The box above the t-value was checked if the variances of the two samples



were considered to be unequal; this implied that the t-value had been



determined by the Smith-Satterthwaite t-statistic.  The value beneath



the t-value was the probability that the t-value could be exceeded due



to random error in the sampling.



     The equality of the frequency distributions were tested by the use



of the SAS FREQUENCY (15) procedure.  Chi-squared statistics were determined



by the FREQUENCY procedure and presented for within and between groups



in Tables 22 and 23 respectively.
                                 3-34

-------
U)
             150'
F  100
R
E
Q
u
c
Y

     50
                         A
                                    Q
                               I
                                         ©
1.5
3.5   H'
                                                                                            Sample
                                                                                            •600
                                                                                           ©601
                                                                                           £602


                                         7-5   *-5   "?•?   »-5  U-S  12.5
                                              Class Mark  (nm)
                                                                                                   17.5
                                                  Figure 3.  Set 1 frequency  distributions.

-------
u>
         150 1-
F  100
R
E
Q
u
E
N
          504
                                                                                    Sample
                                                                                    •603
                                                                                   ©604
                                                                                   & 605
                                  «   «   *
                                       f   •
                                                                   »   f   A
                                                                                                       2.1
                                                 Class Mark  (nm)

                                            Figure 4.  Set 2 frequency distributions.

-------
10
 I
CO
           150
F  100
R
E
Q
u
E
N
C
Y

     50
                                                                                          Sample
                                                                                          •607
                                                                                          O611
                                  <»
                                  •
                                                                    o
3-5   4-5
7-ST  8.5   <».f
     Class Mark   (nm)
                                                                              l&S
                                                                                               17.5
                                               Figure 5.  Set 3 frequency distributions.

-------
LJ
OJ
c»
           150
F   100
R
E
Q
u
E
N
            50
                                                                                          Sample
                                                                                          • 609
                                                                                          0612
                                                                O

                   1-5   3L-5  3-5
                               5.5   fe-5
                              13.5   if,5 i£*
      Class Mark  (nm)

Figure  6.  Set 4  frequency distributions.
17-5

-------
150-
F 100
R
E
Q
u
E
N
C
Y

   50

                                                                            Sample
                                                                            • 712
                                                                           ©716
                           Q
                                0
                                     •
                                         4
                                                                  ©

                                                                     •
                                                   1
                                                                      1 - 1
                                                                                     o
                                                                                     1
                                          -5   *S   \tS   U.S
                                          Class Mark (nm)
                                     Figure  7. Samples 712 and 716  frequency distributions.

-------
                               Table 19.  Statistics for fine particle  distributions
Set Number
1
2
3
4
5
6
Sample No .
600
601
602
603
604
605
607
611
609
612
716
712
Number of
Particles
386
446
434
369
368
356
296
274
298
297
449
497
Mean
(nm)
8.00
7.83
7.53
8.45
8.32
8.31
9.05
8.57
7.86
8.13
8.05
6.42
Standard
Deviation
(nm)
5.35
4.56
5.23
5.37
5.51
5.41
5.49
5.40
5.40
5.37
5.21
4.79
Standard of Error
of the Mean (nm)
0.27
0.22
0.25
0.28
0.29
0.29
0.32
0.33
0.31
0.31
0.25
0.21
Coefficient of
Variance (%)
66.9
58.2
69.5
63.5
66.3
65.1
60.5
62.9
68.8
66.0
64.7
74.6
u>

-p-
o

-------
Table 20.  Within group t-statistics for fine particle samples
                600
                601
 602

_



^
0.496
0.62



D
1.28
0.20
0.919
0.36
-
                 603
  D
 1.04

 0.30
611
  609  Sample Number
  Q
 0.628

 0.53
612
                604
  605

-



[umber
D
0.302
0.76




D
0.328
0.74
a
-0.026
0.98
-
                           716
-5.02

 0.001
                                                 Sample Number
                                                     600
                                                     601
                                                     602
Sample Number
                                                     603
                                                     604
                                          605
Sample Number



    712
                               3-41

-------
                          Table  21.   Between Group  t-statistics  for  fine particle samples
 i
-P-
          Sample

          Number
            600
            601
            602
          Sample

          Number
            603
            604
            605
603
607
604
605
607
611
609
612
                                                                                      716
611
609
612
716
712
a
-1.43
0.15
a
-1.71
0.09
a
-1.74
0.08
a
-0.30
0.76
a
-0.59
0.55
a
-0.62
0.54
a
1.40
0.16
a
-1.08
0.28
P
1.06
0.29
n
0.74
0.46
a
0.43
0.67
a
0.40
0.69
a
1.05
0.29
a
0.70
0.49
a
0.67
0.50
5.76
0.001
5.29
0.001
5.27
0.001
712
a
-1.14
0.25
-l775
0.08
a
-2.46
0.02
a
-0.80
0.42
-1.36
0.18
a
-2.03
0.04
a
0.78
0.44
GZf
-1.33
0.19
a
-2.06
0.04
a
-2.52
0.12
Of
-3.17
0.001
a
-3.80
0.002
a
-1.36
0.18
-1.91
0.06
a
-2.57
0.01
a
0.34
0.73
3f
-0.07
0.94
a
0.83
0.41
a
-0.33
0.74
GZf
0.80
0.42
a
-1.53
0.13
a
-0.15
0.88
-0.69
0.49
0
-1.51
0.13
sf
4.56
0.001
a
4.63
0.001
CS'
3.35
0.008

-------
                                              Table 21.  Continued
        Sample

        Number
          716
          712
             600
601
602
603
604
605
607
611
609
612
-0.15
0.88
4.56
0.001
-0.69
0.49
a
4.13
0.001
a
-1.51
0.13
3.35
0.008
a
1.05
0.29
5.76
0.001
a
0.70
0.49
5.29
0.001
a
0.67
0.50
5.27
0.001
a
2.50
0.01
6.86
0.001
a
1.29
0.20
5.53
0.001
a
-0.50
0.62
3.79
0.001
a
0.20
0.84
4.54
0,001
u>
I
Sample

Number
          607
          611
                      609
612
716
712
a
2.67
0.008
a
1.59
0.11
a
2.06
0.04
a
0.98
0.33
a
2.50
0.01
a
1.29
0.20
6.86
0.001
5.53
0.001
                         Sample

                         Number
                                                                           609
                                                                           612
                           716
                           712
a
-0.50
0.62
a
0.20
0.84
3.29
0.001
G3f
4.54
0.001

-------
         Table 22.   Within group chi-squared statistics
           Sample
           Number
            600
            601
            602
          600
601
602
—




67.2
0.0001
-


44.6
0.0008
99.98
0.0001
-
Sample
Number
 607
           Sample
           Number
            603
            604
            605
          603
604
611
605
—




14.8
0.733


-

Sample
Number
609
6.98

0.994
17.93
0.527
-
612
7.154
0.993



Sample
Number
712
                                   716


                                  71.77
                                  0.001
                                3-44

-------
Table 23.  Between group chi-squared statistics
Sample
Number
 603
 604
 605
607
6L1
609
612
29.8
0.055
51.7
0.0001
28.5
0.074
56.1
0.0001
73.9
0.0001
45.6
0.0006
49.8
0.0001
47.9
0.0003
46.3
0.0005
45.1
0.0007
51.8
0.0001
39.7
0.0036
Sample
Number
 609
  612
607
611
77.7
0.0001
67.7
0.0001
109.9
0.0001
96.6
0.0001
                         3-45

-------
                              Mass




     The Latin Square data was organized and presented In the Latin




blocks in Table 14.  The Latin Square analysis of variance calculations




were carried out in Appendix D-l with a summary presented in Table 24.




The data was also analyzed using the SAS ANOVA procedure.  This data




was shown in Table 25.  This table was explained in Appendix D-3.




     The data taken about an expected midrange of operating variables




was analyzed by the SAS GLM (15) procedure and was presented in Table 26.




     All of the mass data, that is, both the Latin Square and the




orthogonal data sets were analyzed by the SAS GLM and STEPWISE pro-




cedure.  The SAS GLM table was presented in Table 27.  The STEPWISE




procedure used a statistical method to develop a "best-fit" linear




model for the prediction of the dependent variable-mass concentration




of aerosol within the gas stream.  In Appendix D-4, this procedure was




described.




     The "best-fit" linear model was analyzed for its aptness with




respect to its predictive ability using the SAS GLM procedure and




associated calculations for model residual analysis.  This procedure




was described in Appendix D-5.
                                3-46

-------
Table 24.  Analysis of variance for latin square experimental design
           mass data
Variable
Wire feed
rate
Open Voltage
Main air-jet
pressure
Replication
Experimental
error
Total
df
o
3
*
3
1
21
31
Sum of
Squares
1.473
0.319
0.106
0.335
0.952
3.185
Mean
Square
0.4908
0.1065
0.0352
0.3349
0.0453

F
cal
10.86
2.35
0.77
7.39


Fa-0.05
3.07
3.07
3.07
4.32


Fo-0.01
4.87
4.87
4.87
8.02


                                 3-47

-------
Table 25.  SAS ANOVA for latin square experimental design mass data
Dependent Variable:   Y,  Mass Concentration
Source          Sum of Squares      df       Mean Square
Model
Error
Corrected Total
                                                            F Value
2.23 10 0.22
0.95 21 0.045
3.18 31
R-Square C.V. *
0.70 15.4
Std. Dev. Y Mean
0.21 1.38
4.94
PR>F
0.0010
Source
df
                              ANOVA  SS
F Value
PR>F
Wire feed
rate
Voltage
Air Pressure
Replication
3
3
3
1
1.47
0.32
0.11
0.33
10.86
2.36
0.81
7.35
0.0002
0.1004
0.5042
0.0131
                                3-48

-------
          Table 26.  SAS GLM table for orthogonal data
Dependent Variable:  Y, Mass Concentration
Source          df      Sum of Squares
Model            3          0.26
Error            4          0.14
Corrected Total  7          0.40
                          Mean Square    F Value
                             0.08
                             0.03
                              2.41
                              PR>F
                              0.2079
R-Square
0.64
Std. Dev.
0.19
C.V.
11.8
Y Mean
1.61
 Source
 Wire  feed
   rate
 Voltage
 Air Pressure
df
Type I SS
                                              F Value
                                          PR>F
1
1
1
0.15
0.054
0.054
4.20
1.51
1.51
0.1099
0.2864
0.2864
                                 3-49

-------
          Table 27.  SAS GLM table for all of the data


Dependent Variable Y, Mass Concentration

Source          df

Model           13

Error           26

Corrected Total 39
Source

Wire feed
  rate

Voltage

Air Pressure

Replication
Sum of Squares Mean Square
2.81 0.22
1.09 0.042
3.90
R-Square C.V.
0.71 14.4
Std. Dev. Y Mean
0.21 1.43
f Type I SS F Value
5 1.94 9.21
4 0.37 2.22
3 0.16 1.27
1 0.33 7.87
F Value
5.12
PR>F
0.0002




PR>F
0.0001
0.0946
0.3063
0.0094
                                3-50

-------
                     DISCUSSION OF RESULTS






                        Agglomerates




     The size means of the agglomerates were found to be 0.41 ym




(S=0.53 ym) for sample 715 and 0.40 ym (S=0.54 ym) for sample 714.




These means were also found to be equal; this implied that they were




from the same population.  The coefficient of variances were rather




large, thus the agglomerated particle distributions representative  of




the aerosols were not "monodisperse."




     The simple comparative analysis of increasing the duct exhausting




velocity from 1300 fpm to 1850 fpm showed only a decrease in particle




size from 0.41 ym to 0.40 ym.  This difference was not significant  at a




0.1 level of significance.




     The distribution of the agglomerates was not normally distributed




about the mean as shown in Figure 8.  This fact indicated an inadequacy




of this analysis because the t-test for equality of means assumed the




sample distributions to be normally distributed.  Therefore, the analysis




of the agglomerate distributions was, at least, a characterization of




the aerosol.  While this was true, the analysis still served to illus-




trate the t-test for equality of mean technique.




                         Fine Particles




     The within group t-test analysis showed  that the sample means  in




all the sets were equivalent within groups.   T-test  hypothesis  and




implications were outlined in Appendix D-2.   Set  2 t-test analysis




showed a very small difference within this group  of  operating  conditions.




The variance of this set of samples were also considered  equal by  the




F-test as set forth in Appendix D-2.  Certainly,  these  three  sample
                                3-51

-------
           \s»
           1*0
 Particle
 Count
                              Sample  714
             o a.  4 fe  8 to 12 N \fe 10 20 *# #a 21 30 37.

                        Size  range
          (So
          loo
Particle
Count
                              Sample 715
                                                 n.
             9 2. H
                        Size range
       Figure  8  Histograms of agglomerated particles
                                   3-52

-------
populations (set 2) were representative of the same aerosol parent



population.




     The between group t-test analysis showed a variety of equal and




unequal means.  Table 21 showed that the means of set 1 were all different




than the means in set 3.  Three of the four means in the combination of




set 3 and  set 4 were .different.  Five of the six means in the combination




of set 1 and set 4 were equal.  The majority of the means of the combi-




nation of  set 1 and set 2 were unequal.




     The t-test analysis between the two comparative samples 712 and




716 showed sample 712 to have a significantly smaller very fine particle




size than  sample 716.  In fact, sample 712 was found to have a mean




much smaller than other samples taken.  The increase in the exhausting




rate of the barrel was the  only known difference in the operating con-




ditions and most likely had the observed effect on the means and variances




of the fine particle size distributions.  Even with this information,




the simple comparative experiment was not considered to be conclusive




because of the small number of samples.




     The within group very  fine particle size distributions for sets




2, 3, and  4 were found to be equivalent.  Size distributions between




groups were found to be all different at an alpha level of 0.05.  This




implied that at three of four different operating conditions three




different  fine particle size distributions were generated.




                      Chemical Composition




     The SAED analysis showed two possible resultant materials.   Two




different  phases of iron oxide were possible:  magnetite and maghemite.




Mr. Richard Shimps of McCrone said:  There is difficulty in deciding






                                3-53

-------
which one may truly be present,  or  if  both are present.   This stems




from the fact that the  strongest diffraction lines of each material




are the same since the strongest lines of maghemite are those which




include the spinel structure.  (17)  Magnetite was considered to be present




with a possibility of maghemite  being  present.




                         Mass  Analysis




     The Latin Square experimental  design was analyzed by two analyses




of variance methods; as expected, the  Latin Square hand calculation and




the SAS ANOVA procedure resulted in the same values for the statistics.




The F statistic in ANOVA was used to determine if the particular variable




had any effect on phenomenon of  gas stream particulate mass concentration.




The test was explained in Appendix  D-3.




     The Latin Square analysis found the wire feed rate to be significant




at the 0.01 significance level.   This  implied the wire feed rate had an




effect on the resultant mass concentration.  The variables of open




voltage and jet pressure were not found to have any effect on the mass




concentration of the aerosol even at the 0.1 significance level.  The




analysis of variance statistics for the effect of replication was also




calculated and found to indicate a significant difference between  trials




at the 0.05 significance level but not the 0.01 significance  level.




     As explained in Appendix D-3,  the SAS ANOVA procedure also  analyzed




the variance of the data to determine if the variability of  the  data




was a consequence of random error  (noise) in  the collection  of  the data




or a consequence of the changing of the experimental  variables.   The




F test value for this test was found to conclude that the  variability




was a consequence of the experimental variables at  the  0.01  significance



level.





                                3-54

-------
     The SAS GLM procedure found this particular set of operating points


to insignificantly describe the variability of the orthogonal data.   Even


though this was found, the wire feed rate was found to be almost signi-

ficant at the 0.1 significance level.


     The SAS GLM analysis of the combination of the Latin Square and


Orthogonal data found the data to significantly (a = 0.01) describe  the


variability of all the data.  With this total set of 40 observations,


the SAS GLM procedure calculated a F statistic for determining the effect


or noneffect of the independent variables in the experimental model.


The procedure found the variables included in the model to have a signi-


ficant effect on the outcome:  mass concentration of the gas stream.


That was, the GLM procedure determined that the wire feed rate, the open


voltage across the arc, and the sample replication variables affected the


results of the experiment.

     The SAS STEPWISE procedure determined a "best-fit" linear equation


for the prediction of mass concentration in the gas stream.  The equation
was
              Y =  -0.81 + 0.01  (WP) + 0.048  (V)

                                        3
where Y  is  the mass  concentration (gm/m ), W  is  the wire feed rate


 (range 30-60), and V is the open voltage  (range 25-31 volts).  By the


SAS GLM  procedure  the intercept beta  value  (-0.81) was significant at


the 0.11 level.  The wire feed  rate beta  value (0.019) was  significant at


the 0.001 level and  the open voltage  beta value  (0.048) was significant


at the 0.01 level.  It was found that the prediction of mass concentration


was not  to  be predicted by the  gas jet pressure variable.   The variability


in the data brought  about by the replication factor was ignored  in this
                                 3-55

-------
analysis because it was not a controlled independent variable.




     The analytical method of investigating the aptness of the  model




developed by the SAS GLM procedure by plotting a standardized residual




was presented in Appendix D-5.  The difference between the observed out-




come and the predicted outcome - the residual - at a collected  data point




was expected to be zero assuming a perfect fit of the model. A real




model was expected to have residuals scattered evenly about zero for




all the data points used to determine the model.




     Plots of the independent variables against the standard residual




were expected to evaluate nonlinearity of the regression function:  the




model.  Plots were shown in Appendix D-5 (Figures 12 and 13) and were




interpreted as showing that the model, indeed, fitted a linear  function.




     A plot of the predicted values of the model against the standard




residual was expected to evaluate nonconstancy of error variance.  If




the standard residual increased or decreased as the predicted value




changed, the error variance was considered not to be constant through*-




out the model's predictive range.  As shown in Appendix D-5  (Figure 14),




this was the case with the predictive model determined by the SAS GLM




procedure.  The implication of this fact was that the error variance was




larger when the predicted value was larger; this meant that  the variability




of gas stream mass concentration can be expected to be greater when




larger mass concentrations are being considered.




     A plot of the collection order with respect to the time sequence




of the data against the standard residuals was  expected to  evaluate




nonindependence of the error terms.  The implication of this was  that




the ordering of the sample collection was not random.  The  residual  plot
                                3-56

-------
in Appendix D-5  (Figure 15) showed the sampling order to be random.






                             CONCLUSION





     The Iron oxide aerosol has been characterized by count with a particle




size mean of 0.4 ym.  The  inherent limitations in this analysis, e.g.  choice




of magnification, nonnormality of sample distribution, indicates that  the mean




is probably less than 0.4  ym.  It should be noted that most of the particles




were measured to be less than 0.1 ym.




     The iron oxide fine particles that formed by agglomeration into the




aerosol particles have been characterized to have means ranging from 5.4nm




to 9.1 nm  (0.0054 ym to 0.0091 ym) .  The size means and size distributions




of samples at given operating conditions were found to be equal.  This im-




plied that a specific operating set up generated a particular population of




fine particles.  Differences in size means and size distributions between




sets of operating conditions indicated that changing operation variables




changed the characteristics of the population of fine particles generated.




     The simple  comparative test for discerning the effect on size of




changing the exhausting rate of the arc chamber was inconclusive.  The ag-




glomerated fine  particle aerosol size was found not to change with an increase




in exhausting rate.  However, the fine particle mean diameter size was




found to be greatly affected by the increase in exhausting rate.  The evi-




dence of this change in generating properties indicated a need for further




experimentation.




     The analysis of the production of the mass concentration of  the aerosol




showed that a linear expression aptly fitted the data.  The predictive model




showed the mass  concentration to be a function of  the wire feed rate and  the




open voltage across the electrodes.  The model predicted a range  of 1.0  to








                                   3-57

-------
1.8 gm/m3 with a concurrent increase of wire feed and voltage.  The gener-

ation of a particular mass concentration should therefore be predictable

given adequate preliminary characterization for at least this material.

     With respect to possible suggestions for further development of the

Electrospray aerosol generator, there are two areas that require further

work.  The first area coneerns the control of variables, such as the wire

feed rate, the open voltage, and the main comressed air-jet.  The second

area concerns application precautions.

     The wire feed delivery should be smooth - free of delivery impedance.

The wire straighteners should control the curvature of the wire for con-

sistent contact area at the arc.  It was felt that the straighteners and
                                                              ;/•
the pinch rollers imparted an oscillating tension to the wire and conse-

quently an oscillating arc contact area.  This behavior was observed as a

sound changing in pitch during operation and as an oscillating - up and

down - motion of the wire as it left the electrode tips.

     The open voltage would require very precise control if its effect

on the aerosol's particle mean size or size distribution were to be inves-

tigated.  The compressed air-jet that atomizes and quenches the melted and

vaporized metal should be studied to ascertain its effect on  the aerosol's

particle size mean and distribution.  Possibly, variables related  to the

agglomeration of the very fine particles could be studied.  Also the effect

of the temperature of the particle laden gas stream from the  barrel to  the

sampling port could be studied.

     In the application of this aerosol generator, an on-line continuous

mean size and size distribution analyzer, such as a Whitby-Liu mobility

analyzer, would provide the necessary before control equipment  and after



                                   3-58

-------
control equipment data in an engineering evaluation of a piece of air


pollution control equipment.  Control of the separation of the larger


particles within the barrel could also be used to help approximate real


life circumstances.


     In conclusion, the  electric arc generator produced an iron oxide


aerosol of agglomerated  fine particles in mass concentration averaging

        3
1.4 gm/m  .  The generator showed controllability of mass concentration and


variability of fine particle populations with respect to operating variables.


The resultant conclusion of this report is  that the generator is controllable


and, with detailed characterization of the  aerosol below 0.5 ym that the


generator would fill  a much needed  service  in production of standard, fresh


aerosols  for  engineering evaluations  of bench or pilot scale air pollution


control equipment  and source  sampling (stack testing) equipment.
                                    3-59

-------
                             BIBLIOGRAPHY
     1.  Amson, .1. D.; "Electrode Voltage in the Consumable-Electrode
Arc System."  Journal of Physics D;   Applied Physics. Vol. 5, 1972.

     2.  Davies, C. N.;  Aerosol Science, Academic Press, 1966

     3.  Foment!, M., Julliet, F.,  Meriavdeau, P., Teichner, S. J.,
and Vergnon, P.; "Preparation in a Hydrogen-Oxygen Flame of Ultrafine
Metal Oxide Particles" in Aerosol and Atmospheric Chemistry, edited by
G. M. Hidy, Academic Press, 1972, pp. 45-50.

     4.  Grassel, Eugene E.; "Aerosol Generation for Industrial. Research
and Product Testing," in Fine Particles;  Aerosol Generation, Measure-
ment, Sampling, and Analysis, edited by Benjamin Y. H. Liu, Academic
Press, 1976, pp. 145-173.

     5.  Hedden, Robert; Electric Arc Generation of Polydispersed
Iron Oxide Aerosol in an Air Stream, problem report submitted to West
Virginia University, Morgantown, West Virginia, 1972.

     6.  Holmgren, J. D., Gibson, J. 0., and Sheer, C.; "Some Character-
istics of Arc Vaporized Submicron Particles, "  Journal of the Electro-
chemical Society, Vol. Ill, No. 3, March 1964, pp. 362-369.

     7.  Linsky, B., and Smith, G,;  "Sources of Information on Air
Pollution," Proceedings, National Conference on Air Pollution, Washington,
D. C., November 1958.

     8.  Magill, P. L., Holden, F. R., and Ackley, C; Air Pollution
Handbook, McGraw-Hill Book Company, 1956, section  6.

     9.  Miller, I., and Friend, J. E.; Probability and Statistics for
Engineers, Prentice-Hall, Inc., 1965.

    10.  Naylor, Mike;  Very FineParticle Generation by Electric,Arc;
Sampling and Analysis,  problem report submitted to West Virginia  University,
Morgantown, West Virginia, 1976.

    11.  Neter, J. and Wasserman, W.; AppliedLinear Statistical Models,
Richard D. Irwin, Inc., 1974.

    12.  Pfender, B. and Boffa, C. V.; "Generation of Ultrafine Aerosols
with a Transpiration Cooled Anode in a High Intensity Arc,"  The Review
of Scientific Instruments, Vol. 41, No. 5, May, 1970, pp.  655-657.

    13.  Research Appliance Company; "Model 2443 Staksampler:   Description,
Operating Instruction, and Procedures," Gibsonia,  Pa.,  1972.

    14.  Rohrs, Lloyd; "Metal-Fume Fever from  Inhaling  Zinc Oxide," AMA
Achieves of Industrial Health. Vol. 16, No. 1, July  1957,  pp.  42-47.
                                   3-60

-------
     15.  SAS INSTITUTE, INC.; A USER'S GUIDE TO SAS 76. Raleigh,  North
Carolina, 1976.

     16.  Sears, D. R.; "Fresh Metal Fume Project-2," Laboratory Note
Book at West Virginia University, Morgantown, West Virginia 1975.

     17.  Shimps, Richard; Walter McCrone Associates, private communi-
cations from April to December 1976.
                                       I
     18.  Waggoner, A.  P., and Charlson, R. J.; "Measurements of Aerosol
Optical Parameters," in Fine  Particles:  Aerosol Generation, Measure-
ment,  Sampling  and Analysis,  edited by Benjamin Y. H. Liu, Academic Press,
Inc.,  1976, pp.  511-535.

     19.  Wall  Colmonoy Corporation;  "Model UT-500 Colmonoy Electrospray
Metallizer, Direction  for  Operating,"  Detroit, Michigan, 1973.

     20.  Western Precipitation  Group  -  Bulletin WP-50;  "Methods for
Determination  of Velocity  Volume,  Dust,  Nd Mist Content  of Gases,"
Joy Manufacturing Company, Los Angeles,  California.
                                     3-61

-------
APPENDICES
      3-62

-------
                           Appendix A
     Assuming  that a particle  is  spherical and  that its Reynolds number is
small enough to  indicate  laminar  flow,  it is possible to calculate the
terminal velocity of the  particle by  the following expression:
                              d
                         Vt
where  d  is  the  diameter of  the  particle, p   is the density of the particle,
Pf is  the  density  of  the fluid,  g  is  the acceleration due to gravity, and
Vi is the dynamic viscosity of  the  fluid.
     Assuming  a  particle of  diameter  0.5 ym with a unit density of
       3
(Igm/cm )  in an  atmospheric  fluid  (20 *C,1 atmosphere), the formula yields
            V   =  (5  x 10~7m)2(l  x 103Kg/m3-  1.2kg/m3)(9.81m/s2)
             t                 1.82  x  10~3Kg/m   s
            V  =  1.4  x 10~ m/s  =  0.48  cm/hr
            V  =0.19  in/hr  =4.58  in/hr
                  V  •  d   p
                     DV

            R | .  (1.4 x 10"6m/s)(5  x  10"7m)(l  x lQ3Kg/m3)
                                1.82  x  10~3  Kg/m  s
            R # =  3.8 x  10~7
A particle with above assumptions will have a terminal velocity  of  less
than one half centimeter per  hour; this particle has  a Reynolds  number small
enough to indicate laminar flow; this  satisfied the second  assumption
necessary for use  of this terminal velocity expression.

                                3-63

-------
                           Appendix B



     The mass of an assumed spherical particle is given by  the  following


expression:


                        M = p (4Trr3/3)


                               3
                        M = pird /6


where r is "the radius of the particle, p is the density of  the  particle,


and d is the diameter of the particle.


     The ratio of the masses of two spherical particles with  different


diameter but equivalent densities is given by:



                        M = M1/M2



                        M - (Pird13/6)/(pTrd23/6)



                        M = dx3/d23


with the assumption of equal density spherical particles with diameters


of 0.1 pm and 0.01 ym, the larger particle is 1000 times heavier than the


smaller particle at the same point on the Earth's surface.
                               3-64

-------
            Appendix C-l  Particle Sizing Pictures







     The following pictures were presented to show the SEM pictures




used to measure the agglomerated particles.
                Figure 9.  SEM picture sample 715
                Figure 10.  SEM picture sample 714
                              3-65

-------
            Appendix C-2  Sizing of the Agglomerates




     The agglomerated particles were measured with a clear plastic



metric ruler.  The photomicrographs were enlarged such that a measure-



ment of 2 mm would approximately correspond to 0.2 ym.  The maximum



horizontal - across the photograph - and the maximum vertical distance a



particle covered were measured and averaged to give an average particle



diameter.  With known magnification, the actual size was determined by



the following expression:




                 ,,,,,.     	Image Size	
                 Real Size « 	e	
                             Magnitude of Magnification



Groups ranging from less than 2 mm to greater than 30 mm by increments



of 2 mm were used in counting the number of particles within a range.



Class marks were assigned to each group interval.  The raw data is pre-



sented in Table 28.  The SAS MEANS procedure determined the statistics



presented in Table 18.



     The photographs from the SEM negative were not exceptionally clear;



that is, the particles did not have sharp edges.  This effect would be



a factor in both sample measurements; therefore it was felt the results



were not affected by this extraneous variable.
                               3-66

-------
      Table 28.  Raw data for agglomerated particle count
Sample #
Magnifier  (1) Negative
           (2) Enlargement
Less than 2mm
2 to 4 mm
4 to 6 mm
6 to 8 mm
8 to 10 mm
10 to 12 mm
12 to 14 mm
14 to 16 mm
16 to 18 mm
18 to 20 mm
20 to 22 mm
22 to 24 mm
24 to 26 mm
26 to 28 mm
28 to 30 mm
Greater than 30 mm
Class Mark
  1 inni
  3 mm
  5 mm
  7 mm
  9 mm
 11 mm
 13 mm
 15 mm
 17 mm
 19 mm
 21 mm
 23 mm
 25 mm
 27 mm
 29 mm
 31 mm
714
5000
23/10.5
Count
156
41
19
17
14
8
7
3
3
6
0
0
2
1
1
4
715
5000
23/10
Count
136
56
31
23
10
7
11
4
4
3
1
0
0
1
0
8
                               3-67

-------
             Appendix C-3  Mass Concentration Data




     The mass concentration data is presented in the following three



sets of tables.  The top listings in each table are some of the data



taken during sampling.  The bottom listings are the calculated values



for determining the mass concentration (STP) and indicator of isokenitic



conditions (10,13).



     After finding the net weight of a mass sample, STP mass concen-



tration can be found by dividing the ratio of the mass collected to the



volumetric flow through the filter by a correction factor sc.  This



factor compensates for the^ differences in temperature and pressure re-



lative to standard temperature (298*K) and pressure (29.92 in.) by the



following formula:




           v      v         PMeter     298 °K

            STD ~  Meter    29.92 in.X T      =  Meter x  sc'
                                        Meter



This calculation assumes the negligible influence of water in the duct



gas stream (10,20); the relative humidity was found to be less than 20%.



Symbols are explained in Figure 11.



     The isokenitic factor was approximated by the ratio of the nozzle



velocity to the stack velocity:(!').  It was expected to provide an adequate



target for approximating isokenitic; as can be seen in the concentration



raw data tables, the samples were not taken isokenitically.  The sample



nozzle velocity was mostly greater than the gas stream velocity.  The



approximation  (I1) is transformed into the isokenitic relation  (I):
                               3-68

-------
             l _  (TSTACK/TSTD)(l'STD/1>STACR)(Vm STD * Vw STD>
                         (VelST   )(ATlme)(Area of Nozzle)

by multipling approximation (I1) (10) by the correction factor c which

is given  in  Figure 11.
                                3-69

-------
       V    is STP volumetric flow
        o J.U


       V      is volumetric flow (cfm)
        Meter


       P      is Atmospheric pressure at stack sampler  (in.Hg)
        Meter


       T      is Average temperature of stack sampler thermometer  (*F)
        Meter


       ap HEPA is Ap across HEPA filter (in. H20)



       Pilot Ap is standard pilot  p at point of sampling (in.  HJD)



       SAP is secondary air supply (psi)


       Stack T (TgTACK) is temperature of duct  (*C)



       P      is gage pressure of sampling duct  (in.Hg)
        o JLA.wJx


       Wire feed rate is 30-60 (corresponding to 37  to  86 gm/min)



       MAP is main air jet pressure  (psi)


       Open Voltage is voltage across arc prior  to operation (volts)


       CF is dry gas meter volumetric flow  (ft  )


       Time is time of sampling  (sec)


       Area of Nozzle is (1/4 •  l/!2)2(1T/4)  (ft2)


       Nozzle Velocity is  CFM   area of nozzle  (ft/min)
                               /29.!

                            V PSTJ
Stack Velocity is 234  / f^^   (273 + !__....) (PiloUp)

                          STACK          blACK
       Isokenetic approximation  is N°zzl
                   vv
                                     Stack Velocity
       C =  TSTACK       Pbar      „ _,_ „.
            p -    •   ^ - • (1 + R)
             barStack     meters


       R = % water in gas stream
Figure 11.  Explanation of data  sheet  and mass calculation symbols



                               3-70

-------
Table 29-  Mass data first latin square experiment
        (1) Pitot AP - 0.065     Date  7/5/76
           SAP = 20 psi          (2) P,   = 28.98 in
Sample #
(3)
(4)
(5)



(6)
(7)

(8)
(9)





Ap- HEPA
Stack
Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
CF
Time
T
Meter
Nozzle Velocity =
Stack Velocity =
Isokenetic estimate (I1)
C
I = I' x C
gm/acf
gm/sm
ftr.
529
0.5
28.92
68
50
40
29
0.359
1 min
76
1053
1121
0.94
1.15
1.08
0.0428
1.51
0.97
546
0.5
28.92
73
40
50
31
0.412
1 min
77.5
1209
1129
1.08
1.16
1.24
0.0371
1.31
0.968
547
0.55
28.93
64
40
45
25
0.372
1 min
79.5
1091
1114
0.98
1.13
1.11
0.0315
1.11
0.964
548
0.6
28.93
74
50
45
27
0.399
1 min
80.5
1170
1131
1.03
1.16
1.19
0.0362
1.28
0.96
                        3-71

-------
                  Table 29i   (Continued)








           (1) Pltot AP • 0.065    Date 7/5/76


               SAP - 20 psi         (2) Pbar - 28.98  in Hg
Sample #





(3)Ap HEPA





<*> 'stack - (2) - (3)/13'6




(5) Stack T ( C)




    Wire Feed Rate




    MAP




    Open Voltage




(6) CF




(7) Time




    T
     Meter




(8) Nozzle Velocity =




(9) Stack Velocity =




    Isokenetic estimate (I1)




    C




    I = I1 x C




    gm/acf



      /  3
    gm/sm




    sc
549
0.6
28.93
75
50
55
31
0.364
1 min
81.75
1068
1133
0.93
1.16
1.00
0.0334
1.18
0.96
550
0.65
28.93
72
60
40
31
0.383
1 min
83.5
1124
1127
1.00
1.15
1.15
0.0481
1.70
0.957
551
0.65
28.93
60
30
55
27
0.386
1 min
84
1132
1107
1.02
1.10
1.12
0.0228
0.81
0.956
552
0.7
28.93
53
30
45
31
0.401
1 min
87
1176
1096
1.07
1.08
1.16
0.0297
1.05
0.951
                            3-72

-------
                  Table 29. (Continued)
           (1) Pitot AP = 0.065    Date 7/5/76
               SAP = 20 psi
(2) p    28.98 in Hg
     bar
Sample #
553
                                             554
555
556
(3)
(4)
(5)



(6)
(7)

(8)
(9)






Ap HEPA
c»-™~i, = '*/ ~ \3J/13.6
otacK
Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
CF
Time
T ( F)
Meter v '
Nozzle Velocity =
Stack Velocity =
Isokenetic estimate (I1)
C
I = I' x C
gm/acf
/ 3
gm/sm
sc
0.7
28.93
73
40
40
27
0.389
1 min
88.5
1141
1129
1.01
1.14
1.15
0.0336
1.19
0.949
0.65
28.93
85
60
45
29
0.390
1 min
88.25
1144
1148
1.00
1.18
1.18
0.0502
1.77
0.950
0.65
28.93
76
50
50
25
0.382
1 min
88.75
1121
1133
0.99
1.15
1.14
0.0388
1.37
0.950
0.65
28.93
60
30
50
29
0.380
1 min
89
1115
1107
1.01
1.10
1.11
0.0303
1.07
0.947
                            3-73

-------
                  Table  29.   (Continued)


           (1)  Pltot AP  = 0.065     Date  7/5/76
               SAP = 20  psi          (2) Pfear = 28.98 in Hg
Sample #
557
558
559
560
(3) Ap HEPA
<4> PStack ' (2) - (3)/13-6
(5) Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
(6) CF
(7) Time
T
Meter
(8) Nozzle Velocity «=
(9) Stack Velocity =
Isokenetic estimate (I1)
C
I » I' x C
gm/acf
/ 3
gm/sm
sc
0.65
28.93
57
30
40
25
0.370
1 min
89.75
1085
1102
0.98
1.08
1.06
0.0260
0.92
0.946
0.65
28.93
80
60
50
27
0.378
1 min
90
1190
1140
0.97
1.16
1.13
0.0493
1.74
0.946
0.65
28.93
70
60
55
25
0.400
1 min
90.25
1173
1124
1.04
1.13
1.08
0.0291
1.03
0.945
0.70
28.93
70
40
55
29
0.383
1 min
90.5
1124
1124
1.00
1.04
1.04
0.042
1.48
0.945
                            3-74

-------
   Table 30.  Mass data second latin square experiment
           (1) Pitot AP = 0.06
               SAP = 20 psi
Date 7/11/76
(2) P    = 29.05 in
     bar
Sample #
 561
562
                                                       563
564
(3) Ap HEPA
(4) PStack = <2) ~ <3)/13-6
(5) Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
(6) CF
(7) Time
T
Meter
(8) Nozzle Velocity =
(9) Stack Velocity =
Isokenetic estimate (I')
C
I - I1 x C
gm/acf
gm/sm
sc
0.5
29.01
71
40
40
27
0.389
1 min
80
1141
1089
1.06
1.15
1.22
0.0425
1.50
0.966
0.5
29.01
62
40
50
31
0.350
1 min
81
1026
1006
1.02
1.12
1.14
0.0543
1.92
0.964
0.5
29.01
82
50
55
31
0.340
1 min
82
974
1097
0.89
1.18
1.05
0.0611
2.16
0.962
0.5
29.01
80
60
40
31
0.352
1 min
82
1033
1094
0.94
1.17
1.10
0.0473
1.67
0.962
                            3-75

-------
                  Table 30.  (Continued)
           (1) Pitot AP * 0.06    Date 7/11/76


               SAP - 20 psi       (2) Pu   - 29.05 in Hg
                                       bar
Sample I                           565       566       582      568
(3) Ap HEPA
(4) PStack * (2> - (3)/13'6
(5) Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
(6) CF
(7) Time
T
Meter
(8) Nozzle Velocity =
(9) Stack Velocity =
Isokenetic estimated (If)
C
I = I1 x C
gm/acf
, 3
gm/sm
sc
0.5
29.01
77
50
45
27
0.369
1 min
80.5
1082
1089
0.99
1.17
1.16
0.0440
1.55
0.965
0.5
29.01
80
60
50
27
0.335
1 min
80
983
1094
0.90
1.18
1.06
0.0517
1.83
0.966
0.5
29.01
59
30
40
25
0.393
1 min
82
1153
1061
1.09
1.11
1.20
0.0283
1.00
0.962
0.5
29.01
57
30
45
31
0.375
1 min
82
1100
1057
1.04
1.10
1.14
0.0338
1.19
0.962
                            3-76

-------
                   Table  30.   (Continued)


            (1) Pitot  AP  =  0.06      Date 7/11/76
               SAP -  20  psi         (2) PI    .  29.05  in  Hg
Sample #
                                    569
570
571
                                                                 572
(3) Ap HEPA
(4) PStack = (2) - (3>/13'6
(5) Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
(6) CF
(7) Time
Meter
(8) Nozzle Velocity -
(9) Stack Velocity =
Isokenetic estimated (I1)
C
I = I' x C
gm/acf
, 3
gm/sm
sc
0.5
29.01
58
30
55
27
0.368
1 min
82
1080
1059
1.02
1.10
1.12
0.0327
1.15
0.962
0.5
29.01
54
30
50
29
0.374
1 min
82
1097
1053
1.04
1.09
1.13
0.0319
1.13
0.962
0.5
29.01
58
40
55
29
0.357
1 min
82.25
1047
1059
0.99
1.10
1.09
0.039
1.38
0.961
0.5
29.01
65
40
45
25
0.361
1 min
83
1059
1070
0.99
1.12
1.11
0.0346
1.22
0.960
                           3-77

-------
                  Table 30.  (Continued)
           (1) Pitot AP = 0.06




               SAP » 20 psi
Date 7/11/76




(2) Pbar - 29.05 in Hg
Sample #
573
581
575
576
(3) Ap HEPA
(4) PStack - (2> - (3>/13-6
(5) Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
(6) CF
(7) Time
TMeter
(8) Nozzle Velocity «
(9) Stack Velocity =
Isokenetic estimated (I1)
C
I - I1 x C
gm/acf
/ 3
gm/sm
sc
0.5
29.01
68
50
40
29
0.360
1 min
83.5
1056
1074
0.98
1.13
1.11
0.0406
1.43
0.959
0.5
29.01
80
60
55
25
0.309
1 min
83.75
906
1094
0.83
1.17
0.97
0.0513
1.81
0.959
0.5
29.01
76
60
45
27
0.372
1 min
84
1091
1087
1.003
1.16
1.16
0.0392
1.38
0.958
0.5
29.01
76
50
50
25
0.356
1 min
84
1044
1087
0.96
1.16
1.11
0.0413
1.46
0.958
                           3-78

-------
        Table 31.  Mass data orthogonal experiment
           (1) Pitot AP = 0.05
               SAP = 20 psi
 Date 8/9/76
 (2) Pbar =28.91 in Hg
Sample #
700
701
                                                       702
703
(3) Ap HEPA
(4) PStack - (2> - <3>/13'6
(5) Stack T ( C)
Wire Feed Rate
Map
Open Voltage
(6) CF
(7) Time '
T
Meter
(8) Nozzle Velocity =
(9) Stack Velocity -
Isokenetic estimate (I1)
C
I = I' x C
gm/acf
/ 3
gm/sm
sc
0.4
28.88
70
45
50
30
0.236
77/100
72
986
899
0.91
1.16
1.06
0.04
1.41
0.975
0.4
28 . 88
72
45
40
28
0.236
72/100
72
989
961
0.97
1.17
1.13
0.0408
1.44
0.975
0.4
28.88
90
55
50
30
0.165
50/100
72
1015
968
0.95
1.23
1.17
0.0603
2.13
0.975
0.4
28.88
72
45
40
30
0.216
71/100
72
989
892
0.90
1.17
1.05
0.0428
1.51
0.975
                            3-79

-------
                   Table 31.   (Continued)
            (1)  Pitot AP = 0.05
                SAP - 20 psi

 Sample  //

 (3) Ap  HEPA

 (4) PStack=  (2) -  (3)/13'6
 (5) Stack T  ( C)
    Wire Feed Rate
    MAP
    Open Voltage
 (6) CF
 (7) Time
    T
     Meter
(8) Nozzle Velocity =
(9) Stack Velocity =
    Isokenetic estimate (I1)
    C
    I  =  I'  x  C
    gm/acf
      .   3
    gm/sm
    sc
 Date 8/9/76
 (2) P    = 28.91 in Hg
      bar
704
705
706
707
0.4
28.88
80
45
50
28
0.217
60/100
72
948
1061
1.12
1.20
1.34
0.043
1.52
0.975
0.4
28.88
95
55
40
30
0.205
72/100
72
970
835
0.86
1.25
1.08
0.0485
1.71
0.975
0.4
28.88
85
55
40
28
0.236
67/100
72
1008
1033
1.03
1.21
1.25
0.0408
1.44
0.975
0.4
28.88
85
55
50
28
0.192
65/100
72
1008
867
0.86
1.21
1.04
0.0481
1.70
0.975
                           3-80

-------
                     Appendix D-l  Latin Square





     The Latin Square Design is considered an economic, balanced design.




Extraneous variables can be tested for any effect on the dependent




variable in this design.  The variability produced by the extraneous




variables is separated from the variability produced by the main indepen-




dent variables  (9).  A thorough presentation is included in references



9 and 11.




     A F-statistic  is used to test the null hopothesis that the effect




of the variable is  equal to zero.  The independent, extraneous, and a




replicate statistic are found.  Table 32 presents short cut formulae and




definitions of  terms necessary for the calculation of the F-statistic (9).




This statistic  is  compared to a tabulated F-value with the specified




degrees  of freedom and confidence limit.  If the calculated F-value is




greater  than the  chosen tabulated F-value, then the effect of the variable




is not equal to zero.



     Table 33  shows calculation table and Latin Square Summary table.




Tables 24 and  25  show  the  equivalence of the SAS ANOVA procedure for




analysis of variance and  the  short cut Latin Square procedure; the F values




are essentially equal.
                                 3-81

-------
Table 32.   Formula and  definitions for latin square

           analysis of  variance
Source of
Variation
Treatments
(wire feed rate)
Rows
Columns
Replicates
Error
Total
Degrees of
Freedom
n-1
n-1
n-1
r-1
(n-1) (rn+r-3)
rn2-!
Sum of
Squares
SS(TR)
SSR
SSC
SS(Rep)
SSE
SST
Mean
Square
SS(TR)
n-1
SSR
n-1
SSC
n-1
SS(Rep)
r-1
F
MS(TR)
MSB
MSR
MSB
MSC
MSE
MS (Rep)
MSE
SSE
(n-1) (rn+r-3)

       C  -  (T  ...)2/r-n2
SS(V
                      - c
               k=l
     SSR = — J T2    - C
          r«n E   i..
                                         r » Number of replicates
                                       n = Number  of  factors
     SSC  = — J T2   - C
          r«n Z  j..
               1=1
     SST

                               -C
     SSE =  SST  -  SS(TR) -  SSR -  SSC -  SS(Rep)
                         3-82

-------
                                Table 33.  Calculation table for latin  square  design
                    Replicate 1 (gm/sm )

                        MAP (psi)
                                                                         Replicate  2  (gm/sm )

                                                                             MAP  (psi)
                40
                    45
50
55
40
45
50
55
to
 I
oo
01
4-1
     o
    >
         25
         27
         29
         31
A
0.92
B
1.19
C
1.51
D
1.70
B
1.11
C
1.28
D
1.77
A
1.05
C
1.37
D
1.74
A
1.07
B
1.31
D
1.03
A
0.81
B
1.48
C
1.18
                                           4.43
                                           5.02
                                           5.83
                                           5.24
                             w
                             4->
                             c-H
                             O
                                                         O
                                                        >
                                   25
                                  27
                          29
                                  31
A
1.00
B
1.50
C
1.43
D
1.67
B
1.22
C
1.55
D
1.38
A
1.19
C
1.46
D
1.83
A
1.13
B
1.92
D
1.81
A
1.15
B
1.38
C
2.16
                                                                 5.49
                                                                 6.03
                               5.32
                                                                 6.94
              5.32
                  5.21    5.49    4.50
                                       5.60    5.34     6.34     6.50
              Wire Feed  (Rate  %)

                   A-30
                   B-40
                   C-50
                   D-60
                                           Rep 1

                                          3.85
                                          5.09
                                          5.34
                                          6.24
                              Z Rep 2

                               4.47
                               6.02
                               6.60
                               6.69
                                         ET...  44.30

-------
                 Appendix D-2  T Test Procedures'





     The student t-test is based on a random variable having the student-t




distribution with n-1 degrees of freedom.  Where n is the sample, x is




the sample mean, s is the sample standard deviation, and >n is the expected




mean of a normal population, the value of the t-statistic is given by
The testing of the hypothesis concerning two means with unknown population




variances is based on the t-statistic.  If the variances are equal by the




F-test, the t-statistic that is used for decision-making is the Student-t.




     The t-statistic for comparing two means is found with the formula






                           x      \    »
C3.X S"~
xp
where
i Kn2(ni+ v 2)
Sxp (n + n )[(n -1)S2 +
(1)
£
|
2 | m
1 (n ~1)S 1 J
    The value of 6 is set equal to zero when testing the hypothesis that




the means are not different.  If the variances are unequal, then the Smith-




Satterthwaite t-statistic is used (9) where
                                                              (3)
                                3-84

-------
                   and
                          S2   S2       (S2/n )2     (S2/n )



                          nl   U2    '    V1    +   V1




     The equivalent mean hypothesis  tests in this report are based on



the assumption that 6=0.




                         V  V v2 -  « - o




                         H. :  V.- Vn X  0
                          A    12




     The criterion for rejecting H  , where ta/2 is  equal to the tabulated



t-value with (n..+ no~2) degrees of freedom and Type I error equal to a,




is given by



                         Reject H :  V-- V  * 0




                         if t    < - t  a/2
                             cal



                         or t  , > t a/2                         (5)
                             cal



                                      2 2
     The F-statistic given by F    = 8.^82  is used to determine equality




of sample variances.  If F  ..is less than a tabulated F statistical, then




the variances are equal.  If the variances are unequal, then the Smith-



Satterthwaite t-statistics are used  in  the hypothesis test.



     The following are example Agglomerate T Test Calculations.




             Given:   x - 0.41 ym    s  = 0.53 jam    ^ = 300




                      x  - 0.40 ym   s  - 0.54 ym   n  = 282
                       £•              £*              £•
                                3-85

-------
             with equation (1)


                               0.41 ym - 0.40 urn
f

»
                            (300) (282) (300+282-2) _ 11/2

                      (300+282)[ (294) <0.53)2+(281)(0.54)2] J
             t  . = 0.225    df - 300 + 282 - 2 - 580
              cal
             with equation (3)
              ,   _     0.41 ym ~ 0.40 ytm
              cal                  '
                    ./(0.53)2   (0.54):
                    V    300       282
             t'   =  0.225
              cal
      2         2               22         22
             from equation (4)


              ...   f(0.53)2   (0.54)      . r (0.52)/300)     (0.54)/282)
             dt  ~ \  300   *   282   /   ' l      299      *      281     J


             df -  576.3


     In either case, equivalent variances or not, the t  , value is 0.225.
                                                       cal
The df are approximately 580; this degree of freedom for all practical

calculations is equal to infinity.  With the tabulated t value (alpha

equals 0.10 and df equals infinity) being 1.645, the hypothesis test  (5)

shows the means not different.

     In Table 34, the SAS T Test Calculates both the general t-statistic

and the Smith-Satterthwaite t-statistic.  The differences  in the calculated

by hand and calculated by computer t-values are a consequence  of rounding

and truncating errors.
                                3-86

-------
                  Table 34.  SAS T test table
Variable:  X, agglomerated particle diameter.
REPP
   1
   2
 N
300
282
                Mean
                0.41
                0.40
STD Dev.
0.53
0.54
                            STD Error
                             0.036
                             0.032
Minimum
 0.087
 0.091
Maximum
 2.61
 2.74
                            Variances
                             Unequal
                             Equal
                                             DF
                               PROB>:T!
                                   0.1938   575.5   0.8464
                                   0.1939   580.0   0.8463
For HO:  Variances are equal.  F1 - 1.05 with 281 and 299 DF
                        PROS  > F' = 0.565
                                3-87

-------
                      Appendix D-3  ANOVA


     The SAS ANOVA procedure uses a matrix manipulation technique to

determine the analysis of variance in a collection of samples.  The

procedure finds the analysis of variance sum of squares for each variable

in the selected model as well the sum of squares of the error in the data.

A comparison of the average drop in the sequential sum of squares to the

average drop in the sum of squares of the error terms is used to determine

if the independent variable helps explain the variability of the data.

     The F-statistic in formula expression is


                        F - SSR/p-1
                            SSE/n-p


where SSR is the sequential sum of squares for the particular variable,

SSE is the error sum of squares for the data, p is the number of variables

in the selected model, and n is the number of samples.  This statistic

is shown to the right of the variable names in Table 25.  The ANOVA pro-

cedures also calculates the F-statistic associated with examining the

data fit of the model.  This statistic is shown to the right of the model

source in Table 23.  The term below this statistic, PR>F, gives the

probability that the model is fitting noise.  Other statistics are shown

in the ANOVA table.
                              3-88

-------
                Appendix D-4  Stepwise Procedure





     One of the search methods for arriving at a "best" set of inde-




pendent variables from the total variables available is the stepwise




regression method  (15,11).  This set of variables will best describe




as well as predict  the relation between the independent and dependent




variables.  The stepwise procedure determines the variables which have




an observed effect  on the outcome at a certain level of significance.




The variable which  explains most of the variability in the data is first




added to the set of best  independent variables.




     After the first variable is chosen, a second best variable is chosen




(with the assumption that the first variable explained a portion of the




variance in the data), and then a statistical test determines if the




first variable  (actually, any and all prior variables chosen) should be




eliminated from the set of "best" variables.  If both of the choices are




kept in the set, the procedure continues to add to and check for deletion




of best variables  until all  the variables in the set are "best" with a




certain confidence. The  SAS default confidence is 90%.  The deleted




variables do not significantly explain the outcome therefore they are




dropped.



     The SAS  STEPWISE procedure calculated the best variables that des-




cribe  (predict)  the collection of mass concentration data.  The analysis




printout format  is presented in Table 35.  This stepping procedure took




four steps.  The wire feed rate variable was added first and  the best




variable set contained  the wire feed rate and open voltage variables.
                                3-89

-------
               Table 35.  Stepwise regression procedure for dependent variable Y, mass concentration
          STEP 1     Variable WF Entered
R Square =0.43
u>

VO
o

Regression
Error
Total

Intercept
WF
STEP 2 Variable V Entered

Regression
Error
Total

Intercept
WF
V
DF
1
38
39
B Value
0.52
0.02
R Square
DF
2
37
39
B Value
-0.81
0.019
0.048
Sum of Squares
1.69
2.21
3.91
STD Error

0.0037
= 0.54
Sum of Squares
2.09
1.81
3.91
STD Error

0.0034
0.017
Mean Square
1.69
0.06

Type II SS

1.69

Mean Square
1.05
0,05

Type II SS

1.63
0.4
F PROB>F
29.1 0.0001


F PROB>F

29.1 0.0001

F PROB>F
21.4 0.0001


F PROB>F

33.3 0.0001
8.1 0.0070

-------
                                             Table 35.   (Continued)
        STEP 3     Variable P Entered
R Square = 0.543
vo

Regression
Error
Total 39

Intercept
WF
V
P
STEP 4 Variable P Removed

Regression
Error
Total

Intercept
WF
V
DF
3
36
39
B Value
-1.05
0.019
0.048
0.0047
R Square
DF
2
37
39
B Value
-0.81
0.01
0.04
Sum of Squares
2.12
1.79
3.91
STD Error

0.003
0.017
0.006
- 0.53
Sum of Squares
2.09
1.81
3.91
STD Error

0.003
0.017
Mean Square
0.71
0.090

Type II SS

1.6
0.40
0.03

Mean Square
1.05
0.04

Type II SS

1.6
0.40
F PROB>F
14.3 0.0001


F PROB>F
-
33.1 0.0001
8.2 0.0070
0.56 0.4605

F PROB>F
21.4 0.0001


F PROB>F

33.3 0.0001
8.1 0.0070
        All variables  in  the model are  significant at the 0.1000 level.

-------
              Appendix D-5  GLM and Residual Plots





     The SAS GLM (General Linear Model) procedure functions much as the




 SAS ANOVA procedure.  The GLM procedure is more powerful than the ANOVA




 procedure because it can be applied to virtually any data set; the ANOVA




 procedure requires the data to be balanded.  Since the GLM procedure




 outputs the needed statistics, the GLM procedure was used to calculate




 residuals for standardized residual plots.




     A residual is the difference between the predicted out come from




 the regression model developed from the observations and the observed




 out come at the same point.  A standardized residual is the residual at




 a point divided by the square root of the mean sum of squares (the




 standardized deviation) of the error.  The standardized residuals from




 a well fitting model should be evenly dispersed about the zero residual




 line because they are normally distributed variables if the model ad-




 equately describes the independent-dependent variable relationship  (11).




     The GLM table is presented in Table 36.  The model F-statistic




value indicated that the model is fitting valid independent variable




 induced variability.  The F-value for the wire feed rate and open voltage




 indicates that these variables significantly influenced the outcome. The




next table presents the parameter (Bi) estimates for the linear model:




Y = BQ + Bj^ (WF) + B2 (V).  The t-statistic for acceptance of the beta




parameter is indicative of the confidence in the predictive ability of




the variable.  The next table shows the observed data, the predicted data




using the model and beta parameters, and the residuals.  The  standardized




residuals are plotted against various variables in Figures 12, 13,  14, and




15.





                               3-92

-------
                                    Table 36.  SAS GLM table including residuals
          Dependent Variable Y, mass concentration

          Source
          Model
          Error
          Corrected Total
CO
VO
CO
          Source
          WF
          V
DF
 1
 1
Type I SS
   1.69
   0.40
                       Parameter

                       Intercept
                       WF
                       V
DF Sum of Squares Mean Square
2 2.09 1.05
37 1.81 0.04
39 3.90
R-Square C.V.
0.54 15.5
STD Dev. Y Mean
0.22 1.43
F Value
21.4
PR > F
0.0001




F Value
 34.6
  8.1
PR > F
0.0001
0.0070
DF
 1
 1
Type IV SS
   1.63
   0.40
F Value
 33.3
  8.1
PR > F
0.0001
0.0070
timate
-0.81
0.02
0.048
T for HO:
Parameter = 0
-1.64
5.77
2.85
PR > !T!
0.1085
0.0001
0.0070
STD Error
Estimate
0.49
0.003
0.017

-------
Table 36.  (Continued)
5ERVATION
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
OBSERVED
VALUE
1.51
1.31
1.11
1.28
1.18
1.70
0.81
1.05
1.19
1.77
1.37
1.07
0.92
1.74
1.03
1.48
1.50
1.92
2.16
1.87
1.55
1.83
1.00
1.19
1.15
1.13
1.38
1.22
1.43
1.81
PREDICTED
VALUE
1.54
1.44
1.15
1.44
1.64
1.83
1.06
1.25
1.25
1.73
1.35
1.15
0.96
1.64
1.54
1.35
1.25
1.44
1.64
1.83
1.44
1.64
0.96
1.25
1.06
1.15
1.35
1.15
1.54
1.54
RESIDUA
-0.03
-0.13
-0.04
-0.16
-0.460
-0.13
-0.25
-0.20
-0.06
0.03
0.02
-0.08
-0.04
0.10
-0.51
0.13
0.25
0.48
0.52
-0.16
0.11
0.19
0.04
-0.06
0.09
-0.02
0.03
0.07
-0.11
0.27
          3-94

-------
                     Table 36.  (Continued)
liKVATJLUN

31
32
33
34
35
36
37
38
39
40
OBSERVED
VALUE
1.38
1.46
1.41
1.44
2.13
1.51
1.52
1.71
1.44
1.70
PREDICTED
VALUE
1.73
1.35
1.49
1.40
1.68
1.49
1.40
1.68
1.59
1.59
RESIDU

-0.35
0.11
-0.08
0.04
0.45
0.02
0.12
0.03
-0.15
0.11
Dependent Variable:  Y

      Sum of Residuals                          0.00
      Sum of Squared Residuals                  1.81
      Sum of Squared Residuals - Error SS      -0.00
      First Order Autocorrelation               0.15
      Durbin-Watson D                           1.69
                               3-95

-------
U)
1
2

S
JF

N i
D
A
R
D

R
E
S
I
•D
u
A
L

O
o



© ©

O
° • & O 
-------
3
2
S
T


D L
A
R
D
w 0
i n
vo «•
"•si tj1
S
I
D -1
U
A
L
-2


-3
*
Q LEGEND:
Q O One
O Observation
4. Two
Observations
i
° 0
*" -
(D
0 © A ®
§ S .,
O O ^ ^
^ S
e o °
^ o ©

r
O
0


* 9
e
* A » •• • t * •
25
26
27          28
       OPEN VOLTAGE
29
30
31
             Figure 13 Graph:  Standard Residual versus Open voltage

-------









u>
1
VO
oo











3

2
S
T

N
D
A
R
D
R °
E
S
I
D1
-L
U
A
L

-2

-3


0 LEGEND:
0 One
- - - -• ® O Observation
& Two
Observations
©
o
" 	 - "•*•
o
o o © & ®
© ° <* ^ ® « «
& . Q . . P. .O. . . .

®® /a
® O * 6> S
© ° °

©


O
-
O

1* . 	 * 	 - • 	 1 	 . 	 1 II 1
1.0    1.1    1.2    1.3    1.4    1.5    1.6    1.7    1.8
                    PREDICTED OUTCOME (YHAT)
1.9
 Figure 14.  Graph:  Standard Residual  versus Predicted Outcome(Yhat)

-------
3

2
S
T
A
N
D L
A
R
D
CO n
' R U
VD K.
SO E
S
I
D _±
A
L
b
e
0 e



e
»



	 • . e
o o
o
C3 .4*.
0
O
k • - • -


^
        €
.    0   .

    ©   «
                                                              o
                                                                            LEGEND:
                                                                             ©One
                                                                            Observation
                                                                             £i. Two
                                                                            Observations
                                                                      O
                                                                      O
                                          0
                                              O
                                                          8
-3
          1   2   3   4   56    7   8    9  10  11  12  13  14  15  16
                                  ORDER OF PRESENTATION
             Figure  15. Graph:  Standard Residual  versus Order of Presentation

-------
                           Appendix E




     The wire feed rate is a unitless dimension on the metallizer unit.



The rate of feed is dependent on the wire feed rate setting and the



amount of material that is being feed.  Time trials at various settings



with Walcomonoy #10 wire were used to graph a wire feed rate.  This



graph is shown in Figure 16.



     Using the developed equation for the converted mass of the aerosol:





                   Y = -0.81 + 0.01 (WF) + 0.048 (V)





a relation between increased metallizer wire feed rate and the percentage



of converted mass can be found.  Assuming the open voltage to be 31 volts

                                       3

and a ductwork volumetric flow of 5.6 m /min., the predicted percent



conversion for the various wire feed rates is given by the following



expression:



            J rmwrainT. . 1-0-81+0.048(31) + 0.01 (WF)] 5.6
            /o conversion = •*	1	.  Q^ .—•=—-—*—"——— x ±00
                           mass input @ wire feed rate, WF







Table 37.  Relating wire feed rate and percent of mass converted
Wire Feed
Rate
30
40
50
60
Percent
Conversion
14.8%
11.6%
9.6%
8.3%
                              3-100

-------
gm
mln
         90
         80
         70
         60
         50
         40
         30
         20
         10
                  30
40
50
60
                                    WF
                Figure 16.   Wire Feed  Rate Calibration
                                 3-101

-------
                                 TECHNICAL REPORT DATA
                          (Phase read luanictions on the reverse before complctin/:)
 1. NtPORT NO.
                            2.
 4. TITLL AND SUOTITt.fc
 Generation and Simulation of Metallic Particulate
    Air Pollutants by Electric Arc Spraying
                                 3, RECIPIENT'S ACCESSION NO.

                                 B."REPO"RT DATE
                                  September 1977
                                 6. PERFORMING ORGANIZATION CODE
 7. AUTHOHIS)
 B.Linsky, R.Hedden,  M.Naylor, and F.Dimmick
                                                       8. PERFORMING ORGANIZATION REPORT NO.
 9, PERFORMING ORGANIZATION NAME AND ADDRESS
 University of West Virginia
 Morgantown, West Virginia  26506
                                 10. PROGRAM ELEMENT NO.

                                 1AB012; ROAP 21ADM-025
                                 11. CONTRACT/GRANT NO.
                                                        Grant R801858
 12. SPONSORING AGENCY NAME AND ADDRESS
 EPA, Office of Research and Development
 Industrial Environmental Research Laboratory
 Research Triangle Park, NC  27711
                                 13. TYPE OF REPORT AND PERIOD
                                 Grant Final; 7/73-1/75
                                  COVERED
                                 14. SPONSORING AGENCY CODE
                                   EPA/600/13
 15. SUPPLEMENTARY NOTES IERL-RTP project officer for this report is Dennis C. Drehmel,
 Mail Drop 61, 919/541-2925.
 16. ABSTRACT The report gives results of efforts to provide a generated output with an
 appropriate mass and concentration of fresh, dry, fine metal oxide particles for bench
 or pilot scale fine particulate collection research and development work. The work
 involved two electric arc aerosol generators: one using a single consumable electrode
 of welding wire; the other, two cornsumable wire electrodes of a commercially avail-
 able electric arc metallizer. The generated aerosols were exhausted into a duct sys-
 tem and sampled using membrane filters.  The single electrode generator produced
 0.67 g/cu m of 0.1  micrometer diameter iron oxide particles as sampled by an Ander-
 sen Stack Sampler.  The mass emission rate with an average of 1.95 g/min varied
 within a + or - 12%  range.  The double electrode  generator produced submicron parti-
 cles (measure by scanning electron microscopy). Mass volumetric concentrations
 ranged from 0.7 to  2.0 g/cu m for zinc oxide aerosols. The mass emission rate aver-
 aged 8.6 g/min for  the zinc oxide aerosols and 7.7 g/min for the iron oxide aerosols.
 (The zinc oxide wasZnO andlhe Iroh oxide,TFe3O4.TThe double electrode generator
 was further tested and validated for reproducibilities of total mass volumetric concen-
 tration and basic particle diameter distributions. Variables of operation were inves-
 tigated to determine their effect on the mass volumetric concentration of the "aerosol.
 7.
                              KEY WORDS AND DOCUMENT ANALYSIS
a.
                 DESCRIPTORS
                     b.lDENTIFIERS/OPEN ENDED TERMS
                         c. COSATI Field/Group
Air Pollution
Simulation
Aerosols
Metal Powder
  lectric Arcs
Metallizing
Zinc Oxides
Iron Oxides
Air Pollution Control
Particulate
Electric Arc Spraying
Metallic Oxide Particles
Particle Collection
13B
14B
07D
11F
20C
13H,11C
07B
 3. DISTRIBUTION STATEMENT

 Unlimited
                     19. SECURITY CLASS (ThisReport)
                      Unclassified
                         21. NO. OF PAGES
                             334
                     20. SECURITY CLASS (Thispage)
                      Unclassified
                         22. PRICE
EPA Form 2220-1 (9-73)
                  3-102

-------