EPA-600/2-77-060
February 1977
Environmental Protection Technology Series
                PROCEEDINGS:  SEMINAR ON  IN-STACK
                  PARTICLE SIZING FOR PARTICULATE
                          CONTROL  DEVICE  EVALUATION
                                   Industrial Invirinmeirtal iesearcli 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
 Protection Agency,  have been grouped into five series. These five  broad
 categories were established to facilitate further development and application of
 environmental technology. Elimination of traditional grouping was consciously
 planned to foster technology transfer and a maximum interface in related  fields.
 The five series are:

     1.    Environmental Health Effects Research
     2.    Environmental Protection Technology
     3.    Ecological Research
     4.    Environmental Monitoring
     5.    Socioeconomic Environmental Studies

 This report has  been  assigned  to the ENVIRONMENTAL PROTECTION
 TECHNOLOGY series. This series describes research performed to develop and
 demonstrate instrumentation, equipment, and methodology to repair or prevent
 environmental degradation from point and non-point sources of pollution. This
 work provides the new  or improved technology required for the control and
 treatment of pollution sources to meet environmental quality standards.
                    EPA REVIEW NOTICE

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

-------
                                         EPA-600/2-77-060

                                         February 1977
             PROCEEDINGS:  SEMINAR

      ON IN-STACK  PARTICLE  SIZING  FOR

PARTICULATE  CONTROL DEVICE  EVALUATION
               Douglas Van Osdell, Compiler

                Research Triangle Institute
                     P.O. Box 12194
            Research Triangle Park, NC  27709
             Contract No. 68-02-1398, Task 32
                    ROAP No. AAS90
               Program Element No. 1AB012
           EPA Project Officer: D. Bruce Harris

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

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

-------
                                FOREWORD

     The Environmental Protection Agency, through the Industrial
Environmental Research Laboratory, Research Triangle Park, N.  C.  (IERL-
RTP) sponsored a seminar on In-Stack Particle Sizing for Particulate
Control Device Evaluations.  The seminar was organized by the  Process
Measurements Branch and held at IERL-RTP on December 3 and 4,  1975.
     The seminar was chaired on December 3 by D. B. Harris, and opening
remarks were presented by J. A. Dorsey, Chief, Process Measurements  Branch,
IERL-RTP.  Nine technical papers were presented, followed by a general
discussion of in-stack particle sizing.
     The December 4 session was chaired by L. E. Sparks.  Seven technical
papers were presented, as well as a panel discussion of in-stack use of
diffusion batteries and further general discussion.  Questions were asked
and answered during and following each presentation.
     This document is composed of edited versions of the  speakers'  transcripts
from the seminar.  Some were extensively edited into technical paper format,
while others remained conversational in tone.  Visual aid material presented
by the speakers is included.
                                   ii

-------
                               CONTENTS

                         Wednesday, December 3


                                                                    Page

Foreword	     ii

Opening Remarks (J. A, Dorsey, IERL-RTP) 	      1

Use of Particle Size Data (L. E. Sparks, IERL-RTP)	      3

Calibration of Several In-Stack Cascade Impactors (Wallace B.
Smith, Southern Research Institute)  	  	      6

Substrates and Impactor Cut Points (A.  Kishan Rao,  Midwest
Research Institute)	     66

Experiences in Using Cascade Impactors  for Testing  Electrostatic
Precipitators (Joseph D. McCain, Southern Research  Institute)   . .     84

Outline of Field Experience with Cascade Impactors  (Dr.  Seymour
Calvert, Air Pollution Technology, Inc.) 	    108

Field Experience with Cascade Impactors:  Quality Control  of
Test Results (Dr. David S. Ensor, Meteorology Research,  Inc.).  . .    118

Experience in Particle Sizing of Petroleum Industry Particulate
Emissions (R. L. Byers, Exxon Research  and Engineering Co.)  ...    135

"How to Weigh It, Once You Have Collected It!" (Dr. Colin J.
Williams, Cahn Instrument Co.)	    148

Cascade Impactor Data for Elemental Analysis (T.  A. Cahill,
University of California)  	    154

General Discussion - Wednesday ....   	    169
                                   iii

-------
                            Contents (Cont.)

                          Thursday, December 4
                                                                    Page
New Techniques for Particle Size Measurements (William B.
Kuykendal, IERL-RTP) 	    183

Advanced Particle Sizing Techniques (Pedro Lilienfeld, 6CA
Corporation)	    209

Light Scattering Particle Sizing Techniques (Dr. C.  Y. She,
Colorado State University)	•  •  •    220

General Discussion of Cascade Impactors or Alternative Devices
(D. B. Harris, IERL-RTP) . . .,	    239

Field Experience with Cascade Impactors for Baghouse Evaluation
(Reed Cass, GCA Corporation)	    253

Low Pressure Impactors for In-Stack Particle Sizing (Dr. M.  J-
Pilat, University of Washington) 	    264

Submicron Particle Sizing Experience on a Smoke Stack Using the
Electrical Aerosol Size Analyzer (Gilmore J. Sem, Thermo-Systems,
Inc.)	    276

Panel Discussion on Diffusion Batteries (Joseph D. McCain,
Southern Research Institute) 	    301

Panel Discussion on Diffusion Batteries, The Meteorology Research,
Inc., Extractive Sampling System for Submicron Particles (Dr. David
S. Ensor, Meteorology Research, Inc.)  	   314

Panel Discussion on Diffusion Batteries (Dr. Seymour Calvert, Air
Pollution Control Technology, Inc.)  	   322

Panel Discussion on Diffusion Batteries (Pedro Lilienfeld, GCA
Corporation) 	   327

General  Discussion - Thursday  	   330

Seminar Attendees  	   337
                                    IV

-------
                          ,   OPENING REMARKS
       J. A. Dorsey, Chief, Process Measurements Branch,  IERL-RTP
HARRIS:  We should get this program started.   Jim, would you come up
and give us a start?
DORSEY:  I think that one reason for this seminar is to introduce the engi-
neering people present to the individuals responsible for measurements.   The
engineering people have said to the measurements people, "I  want to watch;"
and quite often the measurements people have  said, "Well, you can't." Then
they have gone out to the source and made a series of tests  with cascade
impactors and diffusion batteries and have come back and told the engineers,
"This is the particle size distribution."  When the engineer says, "How do
you know?"  The response is, "Well, I know because I did it."  And so,
since we wouldn't let them watch, some of them said, "Let's  get them all
together and make them at least tell us what  the devil  they  are doing
out there."
     I really think that ought to be the theme of this  whole seminar.
It is not to discuss all the things that we might do someday, but rather,
what we are doing, and what's the meaning of  what we're doing.  Les [Sparks]
has the opening paper, and I think that it should also be called an opening
comment in that it sets the whole tone for this seminar by answering
the questions:  What are we getting impactor  data for?   What are we going
to use it for?  There is a great deal of relevance to that in that most
of us here are not getting it simply because  we want to publish a paper on
how you use a cascade impactor, or what the wall losses are, or what
happens if you have a substrate reaction problem.  That is not the goal of
the majority of people here.  The goal is to  get usable, useful, and, I
think, definable (what does it mean?) size distribution data that is mean-
ingful to the control device evaluator.

-------
     I don't want to take any more time with general comments.   I will
reemphasize the point that all during the seminar  I hope everybody here
continually asks--not what kind of data can  I  get, or what could I do
to build a better impactor, or conduct a program that's the ultimate
in making size distribution measurements—but rather, what do we need
to do to evaluate control devices and are the measurement techniques,
are the technologies, is the hardware we are using giving us the answers
that we want?  Is it giving the engineer what he is looking for?
     In substance, my opening statement is:   throughout this seminar,
let's all continually ask and reiterate the  question—is what I'm reporting
on, is what I'm doing, relevant to what the  engineer wants and what the
engineer needs to evaluate a control  device  or to  improve the operation
of a control device?

-------
                       USE OF PARTICLE SIZE DATA
         L. E. Sparks, Particulate Technology Branch, IERL-RTP
HARRIS:  Les Sparks will start us off.  Les is really the guy that
started this whole thing because he kept saying, "What are you telling
me?"  So he is going to tell you now what he needs to know, which he
hasn't been able to do to me yet in three years.
SPARKS:  Thank you Bruce.
     Basically, in the Particulate Technology Branch, we're using cascade
impactor data for two purposes—equipment design and device comparison.
     We sponsor work on developing design methods for scrubbers,  preci-
pitators, and baghouses.  Our design models are based on the assumption
     i
that you can predict the overall collection efficiency of a device,
knowing the collection efficiency as a function of particle size  and the
particle size distribution. Now I feel that we can predict the collection
efficiency of a given size particle, at least for scrubbers and precipitators,
That is, we can predict the collection efficiency of a half micron particle
under a given set of operating conditions in a scrubber and a precipitator.
     Then the problem is integrating these individual particle penetra-
tions over the particle size distribution, and here's where we get into
problems:  you have got to go out to the stack and stick a probe  in there
and get a size distribution that means something.  And, unfortunately,
particle control devices are extremely sensitive to particle size distri-
bution, and that means the mass mean particle diameter and whatever other
parameters it takes to characterize the size distribution.  Unfortunately,
I think a lot of stuff got into the literature that said mass mean diameter  .
is all  you worry about and geometric standard deviation and everything else
do not really matter.
     I  have used the mathematical models that our contractors developed
to make some estimates of what happens when size distributions change by
10 percent.  All the calculations were based on log-normal particle size
distributions.  For a scrubber, a 10 percent change in the geometric

-------
standard deviation could mean a factor of 2 or more difference  in  pressure
drop required for a given efficiency.  For precipitators,  the same change
in geometric standard deviation could mean a factor of 30  percent  difference
in the specific collector area required to maintain a  given  collection
efficiency.  These are huge design differences.  Thirty percent in a  large
specific collector area precipitator for power plants  is millions  of
dollars; and a factor of 2 in pressure drop for a large source  is  a signifi-
cant number of dollars per year of operating life.  So that's the  first
reason why I worry about a few tenths of a milligram here  and a few tenths
of a milligram there on the stage weights.
     The other use that we have for impactor data or size  distribution  data
is for comparing devices.  Is Device A a better device than  Device B  or
is Operating Condition A a better operating condition  than Condition  B?
And here again it's the same problem, control devices  respond to the  size
distribution more than our measurements do.  If you've got 30 percent
errors and uncertainties in your penetration curves, which are  probably
reasonable to assume, then you need a dozen data points or so to find a
factor of 2 difference in collector efficiency, and as everybody here who
takes data knows, getting a lot of data is expensive.   A large  field  test
program costs a. lot of money.  I'd like you people to  reduce the amount of
money you need to do it.
     It doesn't matter whether we try to do these device evaluations  in
the laboratory or in the field, we need good measurements.  And if we do the
evaluation in the laboratory, we need some kind of consistent aerosol gener-
ator—which introduces its own problems.
     I guess another use for particle size distribution measurements  is
to define source variations.  Can we find out when the source varies  and
what the source variation is?  Based on the impactor data that were taken to
characterize an aerosol generator recently, I wonder if the  impactor is
introducing more noise than the aerosol generator.
     Whether we're using impactor data for equipment design or equipment
comparison, we need to be able to tell people  in our reports that  the what-
much-what scrubber is no darn good or it's only as good as every other

-------
scrubber you can buy.  Or whether doubling the size of a precipitator
or going to a strange electrode configuration on a precipitator is
really a good idea.  And at the moment we are not able to do that very
well because of the uncertainty in our size distribution measurements.
I hope that, by the time everybody leaves here, they'll be able to go back
and make those 1-percent errors in mass mean and 1-percent errors in
geometric standard deviation size measurements.

-------
          CALIBRATION OF SEVERAL IN-STACK CASCADE IMPACTORS
            Wallace B. Smith, Southern Research Institute
HARRIS:  Now that everybody has a definition of the problem that we are
trying to  solve, we'll start into work on the impactor.  For the last
three years we have been contracting with Southern Research Institute to
try to figure out what this wooly animal was doing to us.  I keep hoping
Wally's  [Smith] going to be able to tell me exactly what it does,
but every  time he does, he says that it does it under this consideration
and we have three other things that it's doing that we don't understand.
The first  year we sprayed them (impactors) out.  "We said, "Well, these
things work nicely, so we're just going to go out and see how well they
work  in  the field."  After that disaster, we went back inside the
laboratory to  see how they work.  So Wallace Smith of Southern Research
is going to outline for us all the data that they got under: well-control!ed
conditions in  the laboratory* with Les1 1-percent accuracy.

SMITH:   The procedures used to perform this calibration study were as
follows:   a fluorescent, monodisperse aerosol (ammonia fluorescein) was
generated  using a vibrating orifice aerosol generator (VOA6).  The aerosol
was sampled with each impactor at flow rates which are within the bounds
specified  by the manufacturer.  The impactors were then disassembled
completely and all the internal surfaces washed in separate, measured  volumes
of 0.1N  NH^OH.  The mass accumulated on each surface was measured by
absorption spectroscopy using a Spectronic 88 spectrometer.  Thus, wall  loss
and stage  collection efficiency data were obtained for all particle sizes
tested.  The calibration aerosols ranged from about 1 to 15 ym in diameter.
It is difficult, to produce smaller particles than 1 wm because nonvolatile
impurities in the solvent contribute significant errors.  The impactor
flow rates were not varied during these tests.  Also, it was necessary to
sample for very long periods in order to obtain reliable data when smaller
particles were used.

-------
     The substrates that were used were glass fiber, Vaseline, bare metal,
and Teflon.  The bare metal and Teflon were unsuitable, and experiments
with these were soon discontinued.
     This work was done under contract to the EPA, with Bruce Harris as
the project officer.  All these data and more will be included in our final
report to him.
     Each figure that is shown will identify the impactor, the individual
stage, the jet velocity, the manufacturer's value for the stage D^Q, and
the Reynolds number for the jet.
     The first impactor for which data are shown is the Andersen Stack
Sampler, or Mark III.  Gelman Type A glass fiber substrates were used,
and the impactor was operated at a flow rate of 0.5 ACFM.
     Figure 1 - Stage 1 - Andersen
     The manufacturer's D™ is 12 urn is for our test conditions, and the
data show that to ,be accurate.
     Figure 2 - Stage 2 - Andersen
     The manufacturer's D5g is 7.5 ym.  Again that is very close.  Notice
however, that the stage efficiency peaks near 90 percent and falls off for
larger particles.  This is a form of nonideal behavior which has previously
been observed by us and by others.  One would prefer the efficiency curve
to rise to 100 percent and maintain that value for all larger particles.
     Figure 3 - Stage 3 - Andersen
     The manufacturer's D5Q for this stage is 5.1 urn.  Figure 3 shows that
the calibration data agree quite well with the predicted D^Q.  Jet velocity
is 1.3 m/sec; the Reynolds number is 78.  Again, the experimental efficiency
curve for this stage rises sharply to a value between 80 and 90 percent,
then drops to lower values for very large particles.
     Figure 4 - Stage 4 - Andersen
     This figure shows our calibration data for stage 4 in the Andersen
impactor.  The .manufacturer's D™ is 3.4 urn, while our calibration data
show a D50 of approximately 4 ym.  This is still reasonably good agreement.

-------
  100
    90
    80
 u  70

 OJ
•1-4


•H  60
U-l
>w
w



 §  50
•r-l

4J
 U
 0)  40
 o
 u
    30
   20
   10
          I  I  I  I  I
1—I  I   I I  I
      .5 .6.7.8.91.0        2      3    4   5  6 7 89 10



                   Particle Diameter,  Micrometers
                      20
Figure  1.  Andersen  Impactor, glass  fiber substrates, stage 1

          V. = 0.4  m/sec, Re = 45,  calculated Dcn = 12 ym.
                                             su
                             8

-------
  100




   90




   80
<#>
 u  70
 c
 0)
14-1
w
•H

-P

O
o
u
60




50




40




30




20



10
          i  i  i
                                                  I  i
      5  .6.7.8.91.0        2      3    4   56789 10


                   Particle  Diameter,  Micrometers
                                                              20
   Figure 2.  Andersen  Impactor, glass fiber substrates, stage 2.
             V. = 0.7  in/sec, Re = 57, calculated Dcn = 7.5 urn.
              J                                 »>u
                             9

-------
100




 90




 80
<*>
u  70
c
0)
•H

•H  60
w

c  50
4J
u
o
u
   40
   30
   20
   10
         I   I  I  II
r—i—i  i  I  I  r
            i  i
             i i
      5  .6.7.8.91.0        2      3   4   56789  10


                  Particle Diameter, Micrometers
                         20
  Figure 3.  Andersen Impacton, glass  fiber substrates, stage 3.

            V. = 1.3 ffl/sec, Re = 78,  calculated Den = 5.1 ym.
             J                                 OU
                             10

-------
  100



   90



   80
u  70
c
•r-l

U
W
-P
o

rH
iH
o
U
   60
§  5°
   40
   30
   20
   10
              I  1 T
             J	I
     .5  .6.7.8.91.0        2      3    4   56789 10


                  Particle Diameter, Micrometers
                                                                20
   Figure 4.   Andersen Impactor, glass fiber substrates, stage 4.
             V.  = 2.0 m/sec, Re = 99, calculated D™ = 3.4 pm.
              J
                               11

-------
      Figure  5  - Stage 5 - Andersen
      The  calibration data shows that stage 5 performs very well.  The
 calibration  curve rises to approximately 90 percent with a fairly broad
 peak, and then drops off for larger particles.  The manufacturer's value
 for  the D5Q  is 2.2  yin for our test conditions, whereas the calibration
 curve shows  approximately 2 ym.  This is still acceptable agreement.
      Figure  6  - Stage 6 - Andersen
      This figure shows incomplete calibration data for stage 6.  Although
 the  manufacturer's  D5Q is 1.1 ym for this stage, our calibration data
 indicates that the  D5Q will be much smaller than 1 \m.  Recall, however,
 that the  data  taken using monodisperse aerosols generated by the VOAG are not
 considered reliable below 1 un.  These data do, however, indicate that the
 DCQ  for this stage  will be significantly smaller than the manufacturer's
 published value.
      Figures 7 and  8 - Stages 7 and 8 - Andersen
      Figures 7 and  8 show our laboratory calibration data for stages 7 and
 8  in the  Andersen impactor.  In this case, the jet velocities are 17.1 and
 28.5 m/sec,  respectively.  We now consider this value to be undesirably
 high.  Notice  that  the stage efficiency peaks at about 70 percent.   In a
 later figure we will show data indicating that jet velocities in excess
 of about  10  m/sec result in significant scouring and particle bounce when
 glass fiber  substrates are used.  Again we have insufficient data to actually
 establish a  D5Q for this stage.  Rao  has used a different technique
 wherein he dispersed PSL beads and used an optical particle counter  to
 determine the  stage collection efficiency.  We intend to use that technique
 to complete  these curves, but at the moment all that we have is an  indication
 that the  stage D5Q's for the lower stages are significantly lower than  the
 manufacturer's published values.
      Figure  9  - Andersen Wall Losses
     Wall   losses were determined, as we stated earlier, by washing  all
 the  internal surfaces of the impactor and using the absorption  spectroscopy
 technique  to determine the amount of mass collected on each surface.   Figure
 9  shows the wall  loss data for each location  inside the impactor and as
a function of  particle size.  The majority of the particulate  lost  inside

                                    12

-------
  100
   90
   80
c
0)
•H
u
•H
U-l


w

C
o
•H
4J
U
r-l
o
u
   70
   60
   5°
   40
   30
   20
   10
                                           I  I   I  I  I  I
       —o
                                              I  I  l  I  I
     .5  .6.7.8.91.0       2      3    4    56789  10


                   Particle Diameter,  Micrometers
                                                                20
    Figure 5.  Andersen Impactor,  glass fiber substrates, stage 5.
              V-  =  3.5 m/sec, Re  = 130, calculated D5Q = 2.2 ym.
                               13

-------
  100
   90
   80
u  70
c
u
•H
U-l
14-1
w
60
§  5°
4J
o
o
u
   40
   30
   20
   10
            I  II I
I	I
                                             A I I
      5 .6.7.8.91.0        2      3    4   5  6 7  8 9  10


                  Particle Diameter, Micrometers
                                                             20
   Figure 6.  Andersen  Impactor, glass fiber substrates, stage 6
             V. = 9.3  m/sec, Re = 210, calculated D5Q = 1.1 ym.
                                14

-------
  100
   90
            I  I  I
  1  I I I
<#>
   80
 o  70
 c
 QJ
•H
 O
   60
W
§  5°
•H
U
r-l

O
U
   40
   30
   20



   10
           J	I
I	I
I  i
      5  .6.7.8.91.0        2      3    4    56789 10


                   Particle  Diameter,  Micrometers
                  20
    Figure 7.  Andersen Impactor, glass fiber substrates, stage 7.
              V. = 17.1 m/sec, Re = 284,  calculated D5Q = 0.69 um
                               15

-------
  100
   90
<#>
   80
 o  70

 0)

U-t

W
O
u
   60
c  50
O
•H
4J
o
  40
   30




   20




   10
             TTT
O
                                              1  I  I  I
      5  .6.7.8.91.0        2      3   4    56789  10


                  Particle Diameter, Micrometers
                   20
    Figure 8.  Andersen Impactor,  glass fiber substrates, stage 8.
              V. =  28.5 in/sec, Re = 475, calculated  D5Q = 0.44 ym
                                 16

-------
                         Total
                         Nozzle
                         Inlet Cone
70
60
50
40
30
ov> 20
v
01
(0
0 10
H
H
1 5
2
1
0.5
0.2
0.1
0.
	 7n







•





































































































































































•
1 Hi

















60
50
40
30
20
dp
10 W
tn
o
^
5 H
H
fO
&
2
1
0.5
0.2
0.1
5 1 1.5 2 3 4 5 6 7 8 9 10 15 20
Particle Diameter, Micrometers
 Figure 9.  Andersen wall losses.

-------
 the  impactor is found in the nozzle and inlet cone.  Also, the losses
 are  much more substantial for the larger particles.  Although we feel
 that these data are qualitatively correct the samples were not taken
 isokinetically.  This could result in larger than normal losses in the
 nozzle  and perhaps the inlet cone of the impactors.
      Figure 10.-Andersen Summary
      Figure 10 shows the calibration data for the Andersen impactor in
 the  classical method--collection efficiency vs.  square root of the
 Stoke's number.  The symbols are related to particular stages, and the
                                                                       2
 solid curve represents experimental results published by Ranz and Wong.
 According to the theory of Ranz and Wong, the value of the Stoke's number
 for  50  percent efficiency is about 0.38 for round jets.  The data shown
 in Figure 10 indicate that this is an idealization and that,  in fact, each
 stage may have a different value for the square root of the Stoke's number
 at 50 percent efficiency.
      Figures 11-18
      Figures 11-18 show similar data for the University of Washington Mark
 III  Cascade Impactor.  This impactor was tested under the same laboratory
 conditions.  The gas flow rate was 0.5 cfm.  The aerosol density was
         3
 1.35 g/cm .  The impactor was operated at ambient temperature and pressure.
 Again,  as in the previous data, the impactor was not operated under
 isokinetic conditions.  This was primarily because of the difficulty
 in supplying enough aerosol to allow isokinetic sampling in a reasonable
 period  of time.
      Figure 11 - Stage 1 - University of Washington
      The manufacturer's D5Q for stage 1 is 34 ym.  Our calibration data  show
 that  the D™ is actually closer to'15 wn.  For this particular impactor
 geometry, the inlet cone serves as a single large  jet for  stage  1.   To
 explain the calibration data, we have surmised that perhaps the air  in
 the jet does not expand to the full diameter of the inlet  column  so  that
 the effective jet diameter is somewhat smaller than the actual physical
 diameter of the inlet column.  The University of Washington impactor was
run with greased  substrates.
                                    18

-------
CO
          100
           90
           80
        170
        4)
        -H  60

        •H
        >H

        W  50


        o


        ti  4°
        01
        5  30
           20





           10





            0
                                                      A

                                                      o
                                                              I     I     I     I
0   0.1   0.2  0.3  0.4  0.5  0.6   0.7  0.8  0.9  1.0  1.1  1.2  1.3  1.4  1.5   1.6  1.7



                            Stokes No.     *^F



                     Stage   123   4567



                             O  D  A V  O  •  •
                                                                                                  1.8   1.9
                        Figure 10.  Summary—Andersen  Impactor, glass  fiber substrates.

-------
  100
   90
   80
u  70


0)
•H


•H  60
§  50
•H
4J
U
QJ  40
o
u
   30





   20




   10
         !  I  I I I
         I	I
i  I
      5  .6.7.8.91.0        2      3   4   56789  10



                  Particle Diameter, Micrometers
            20
Figure 11.  University of Washington Impactor, greased substrates,

           stage 1.  V. = 0.9 m/sec,  Re = 1073, calculated D^n =

           34 urn.     °                                   bu
                                 20

-------
     Figure 12 - Stage 2 - University of Washington
     Figure 12 shows calibration data for stage 2 of the University of
Washington impactor.  In this case the manufacturer's Dcn is 13 ym,
                                                       DU
while the calibration data show the D5Q to be 10 ym.
     Figure 13 - Stage 3 - University of Washington
     In this case the manufacturer's D5Q is 5.6 ym.  There is some scatter
in our calibration data, but nevertheless the indication is that the measured
^50 1S very near tne manufacturer's calculated D50.
     Figure 14 - Stage 4 - University of Washington
     Stage 4 is remarkably well-behaved.  The efficiency rises to well
above 90 percent and remains at a high value for a large range of particle
sizes.   Again .however, the calibration data indicate that the D5Q is
significantly smaller than the 2.7 ym value given by the manufacturer.
     Figure 15 - Stage 5 - University of Washington
     Stage 5 is a well behaved stage.  The stage collection efficiency
rises to a value near 90 percent and exhibits a very wide peak before the
efficiency turns down for large particles.  The velocity of 12.6 m/sec
is approaching a value where the data indicate that scouring and bounce can
occur.  Although we do not feel that our data are especially accurate
below 1 ym, the indications are that the actual D™ is somewhat lower than
the manufacturer's published value of 1.4 ym.
     Figure 16 - Stage 6 - University of Washington
     The calibration data for stage 6 are incomplete because of our
inability to calibrate stages below 1 ym.  Apparently the peak efficiency
is well above 90 percent, even though the jet velocity is 24.2 m/sec.  This
may indicate that the use of grease as an impaction substrate allows
operation at higher jet velocities.  A second peak  is observed for large
particles.  The reason for this second peak is not  understood at this time.
     Figure 17 - Wall Losses for the University of  Washington Impactor
     Wall losses measured for the University of Washington impactor are
similar to those measured for the other impactors.  The losses are much
more significant for large particles, and the bulk  of the material was
found within the nozzle and the inlet column.
                                    21

-------
     100
      90
      80
   o  70
   c
   Q)
   •H
   U
U-l
U-l
U

c
o
      60
      50
   4J
   o
   0)  40
   o
   u
      30



      20



      10
            i  I   I I  I
        ,5 .6.7.8.91.0        2      3    4   5  6  7  89  10


                     Particle Diameter, Micrometers
                                                                20
Figure 12.  University of Washington Impactor,  greased substrates,  stage
        2.
                  =  1.5 m/sec,  Re = 562, calculated DSQ = 13
                               22

-------
     100
      90
      80
   o  70
   c
   0)
   •H

   -H  60
   U-l

   W

   c  50
   o
   •rH

   O
   o  40
   ,-H
      30
      20
      10
            l   I
i   I  I  I I  I
                                              O
                                               J   1  I  I  I  I
         5  .6.7.8.91.0        2      3    4  "5  6 7 89 10


                      Particle  Diameter,  Micrometers
                     20
Figure 13,  University of Washington Impactor, greased substrates, stage
           3.  V. = 4.1  m/sec, Re = 650, calculated D5Q = 5.6 ym.
                                  23

-------
   100
    90
    80
  o 70
  c
  OJ
 -r-i

 •H 60
 W
 -1-4
 4J
 U
  o
  U
50



40



30



20



10

                                                              O
                                                I  I   I
       ,5  .6.7.8.91.0        2      3    4   56789 10


                    Particle  Diameter,  Micrometers
                                                              20
Figure  14.  University of Washington  Impactor, greased substrates,  stage
           4.  V.  = 5.0 rn/sec,  Re  =  264, calculated D5Q = 2.7 pm.
                               24

-------
   100
    90
    80
 u  70
 c
 u
 •H
 U-l
 U-l
 W
60
 §  5°
 -P
 u
 0)
 o
 u
40
    30
    20
    10
       C)
            I   1
          I   I  I	I
                                           I   I  IT
                                   O
                                                   o
                                               I   I  I  I I
       5 .6.7.8.91.0        2      3   4   56789 10


                   Particle Diameter, Micrometers
                                                             20
Figure  15.  University of Washington  Impactor, greased substrates, stage
           5.  V. = 12.6 m/sec, Re = 377, calculated D5Q = 1.4 ym.
                              25

-------
    100
     90
  <#>
     80
  u  70
  c
  OJ
  u
  •H
  W
60
  §  50
  u
  0)  40
  rH
  O
  u
     30
     20
     10
                                    T	T
                I  l  I
                                                        O
                                           II  ill
        5.6.7.8.91.0        2      3   4    5  6789 10


                    Particle Diameter, Micrometers
                                                              20
Figure 16.  University of Washington  Impactor, greased substrates,  stage
           6.  V.  = 24.2 m/sec,  Re = 522, calculated DSQ = 0.69 urn.
                                 26

-------
                            •    Total
                            H    Nozzle
                            A    Inlet
70
60
50
40
30
* 20
W
(0
o 10
H
rH
rt 5
2
1
0.5
0.2
0.1
0
.- 7n


-
































































'


'








-



















































































t






1





















60
50
40
30
20
o\o
10 01
cn
0
5 H
rH
(0
2
1
0.5
0.2
0.1
.5 1 1.5 2 3 4 5 6 7 8 9 10 15 20
       Particle Diameter,  Micrometers
Figu-re 17.  University of Washington wall  losses.

-------
     Figure 18 - Summary - University of Washington Impactor
     Figure 18 presents stage collection efficiencies vs. square root
of the Stoke's number for the University of Washington impactor.  These
calibration data again show that each stage deviates from the ideal
theory of Ranz and Wong to a different degree.  This means , that for the
purposes of data reduction, a different calibration constant would have to be
used for each stage to calculate the DgQ.
     Figures 19-26 - Meterology Research, Inc. Impactor
     With the exception of the inlet cone the MRI impactor is very similar
to the University of Washington impactor.  Figures 19-26 show calibration
data for the MRI impactor.  These will not be discussed in detail.   The
impactor was operated at 0.5 cfm flow rate with greased substrates.  Jet
velocities, Reynolds numbers, stage D5g's, as well as calibration data, are
given for each stage.  Also, as was the case for previous impactors, we
were unable to empirically establish the D™ for the last stage because of
an inaccuracy in our calibration technique below 1 ym.  It is possible,
however, to notice that the trend repeats wherein the highest efficiency
attained for a given stage is somewhat higher for greased substrates than
for glass fiber substrates.  This phenomenon should be interpreted with
the fact in mind that we did not accumulate large quantities of material
in the stages.  Therefore the grease and the glass fiber substrates might
be slightly more effective in our calibration studies than in field tests
                                                             Q
where large amounts of material would be collected.  Lundgren   has done
tests wherein he did find that the collection efficiency of stages
deteriorated with time, dropping from perhaps 90 percent to as  low as
70 percent in half an hour.
     Figures 27-34 - Sierra Impactor
     The next series of figures, 27-34, show calibration data for the
Sierra impactor.  This impactor is of a radial slit design and  employs
glass fiber collection substrates.  The impactor was operated at 0.5 cfm
flow rate.   It can be seen from the calibration data that the stage
collection  efficiencies peaked at much lower values than for the round
impactors,  for which  data  were previously shown.   At this time, we do not
                                   28

-------
       100
CO
CO
      o

      0)
     •H
      u
     -H
     W

     o
      O
0.1  0.2  0.3  0.4   0.5  0.6  0.7   0.8  0.9  1.0  1.1
                                                                 1.2  1.3  1.4  1.5  1.6  1.7  1.8  1.9
                                Stage    1234567
                                        o  a  A  v <> • •
               Figure 18.  University  of Washington, Impactor,  greased substrates,  summary
                           of calibration data.

-------
   100
    90
  (HP
     80
  o  70
  c
  
-------
   100
    90
    80
 o  70
 c
 0)
 u
 -H
 M-l
 4-1
 w
60
 §  50
 O
 u
    40
    30
    20
    10
                 i I
                                     I   I  I   I  1  I I
                                           JTT I   i
      .5 .6.7.8.91.0        2      3   4    56789  10



                   Particle Diameter, Micrometers
                                                             20
Figure  20.  Meteorology Research,  Inc., Impactor, greased substrates,
           stage 2.  V. = 1.1 m/sec, Re = 341,  calculated D5Q = 12 urn.
                               31

-------
   100
    90
    80
 <#>
 o  70
 c
 03


 3  60
 M-I
 M-J
 W



 §  5°
 •H
 U
 O
 u
    40
    30
    20
    10
      .5.6.7.8.91.0       2      3    4    56789 10


                    Particle Diameter,  Micrometers
20
Figure  21.  Meteorology Research, Inc., Impactor, greased substrates,
           stage 3.  V. = 3.2 m/sec,  Re = 411, calculated D5Q = 4.5 urn.
                                32

-------
  100
    90
 <#>
    80
 o  70

 0)
•H
 O
•H
U-l
w

c
O
•H

O
 O
 u
   60
    50
    40
    30
   20
   10
               i  t t
                                    O
                                                              O
                                            I  I   I  I  I I
       5 .6.7.8.91.0        2      3   4    5  6  7  8 9  10


                   Particle Diameter, Micrometers
                                                                 20
Figure  22.  Meteorology  Research, Inc.,  Impactor, greased substrates,
           stage 4.   V.  = 9.0 m/sec,  Re = 684, calculated D5Q = 2.1
                                33

-------
   100
    90
       ,5  .6.7.8.91.0        2      3   4   56789  10

                   Particle Diameter, Micrometers
20
Figure 23.  Meteorology Research, Inc., Impactor, greased  substrates,
           stage 5.  V. = 18.2 m/sec, Re » 973, calculated Dgo « 1.2
                               34

-------
  100
   90
<*>
   80
 u  70
 c
 0)
•H
 U
•H
U-4
•4-1
W
60
g  50

•H
JJ
O

-------
CO
      <#>
      M
      w
70
60
50
40
30
20
10
2
1
0.5
0 2
0.1

















































• Total
B Inlet
A Nozzle






I

*












	
1
— 	
-









'














































































































70
60
50
40
30
20
10 n
M
o
5 rH
H
(13
3:
2
1
0.5
0.2
0.1
0.5 1 1.5 2 3 4 5 6 7 8 9 10 15 20
                                   Particle Diameter, Micrometers
                            Figure 25.  Meteorology Research, Inc., wall losses.

-------
CO
       100
        90
        80
      I
        70
        60
•H

-------
 know  if  this  is due to our operating the impactor at a nonoptimum
 flow  rate,  or if the nonideal behavior is a function of the radial
 slit  design.
      Figures  27, 28 and 29 - Stages 1, 2 and 3 - Sierra
      Figures  27-29 show that the collection efficiencies for these stages
 peak  somewhere at or below 50 percent efficiency.  It is interesting to
 notice that this behavior cannot be blamed on the jet velocities, which
 are well within the bounds, that we have found tolerable for round jet
 geometries.   It is difficult to establish a D5Q for these stages because
 the collection efficiency curves may not reach 50 percent collection.
      Figures  30. 31 and 32 - Stages 4. 5 and 6 - Sierra
      Although the jet velocities are much higher for these stages
 efficiency  curves are improved with maximum efficiencies in excess of
 80 percent  for each stage.  The stage DCQ'S, however, are significantly
 lower than  those given by the manufacturer or those which would  be
                                   4
 predicted by  the theories of Marple  or of Ranz and Wong.
      Figure 33 - Wall Losses - Sierra
      Figure 33 shows wall losses for the Sierra cascade impactor.  As
 was the  case  with previous impactors, the wall losses are more severe  for
 large particles, and a large fraction of the particulate was caught in the
 nozzle and  inlet cone.  John 01 in, who designed the Sierra radial slit
 impactor, has  shown me some data wherein he measured wall  losses which
 were  much lower than this.  For that reason, we should reemphasize that
 these data were not taken isokinetically and, hence, the wall losses might
 be larger than would be experienced in actual field testing wherein
 isokinetic sampling would be done.
      Figure 34 - Summary - Sierra
      Figure 34 is the graph showing collection efficiency for all  the
 stages of the  Sierra impactor vs. the square root of the Stoke's number.
Also,  the theoretical  expression for the collection efficiency as  a
function of Stoke's number taken from the theory of Ranz and Wong  is  shown.
                                    38

-------
  100
   90
<#>
   80
u  70
c
•H
O
-H
U-l
4-1
W
60
§  50
•H
-P
u
o  40
O
U
   30
   20
   10
              I  l  I
           i  I  i
                                        l  I  I  I  I
                                          •Q—*"i  i  i	i  i
      .5  .6.7.8.91.0       2      3    4    5  6 7 89 10


                   Particle Diameter,  Micrometers
                                                             20
Figure 27.  Sierra Cascade  Impactor,  glass fiber substrates, stage  1.
           V. = 1.3 m/sec, Re = 555, calculated DKn « 15 ym.
                                           ^50
                                39

-------
   100
    90
    80
 0)
 •H
    70
    60
 4-1
 >w
 W

 §  50

 •H
 -P
 u
 Q)  40
 o
 u
    30
    20
    10

i     iiiIITT
      .5 .6.7.8.91.0        2      3    4    5  6 7 89 10


                   Particle  Diameter,  Micrometers
                             20
Figure 28.  Sierra Cascade Impactor, glass  fiber substrates, stage 2
           Vj =2.3 m/sec, Re = 571, calculated D50 =8.2 ym.
                              40

-------
  100
   90
<#>
   80
o  70
•H  60
w


o
•r-l
-P
U
0)
rH
rH
o
u
50
40
   30
   20
   10
         I T  I
                                                  I  I
                                            I   I  I  I  I  I
      .5  .6,7.8.91.0        2      3    4   5  6 7 89 10

                   Particle  Diameter,  Micrometers
                                                              20
Figure 29.   Sierra Cascade Impactor, glass fiber substrates, stage 3.
            V. = 5.4 m/sec,  Re = 763, calculated D5Q =  3.7 ym.
            0
                              41

-------
  100
   90
<#>
     ,5 .6.7.8.91.0        2      3    4   5  6 7  89  10

                  Particle Diameter, Micrometers
20
 Figure 30.  Sierra Cascade Impactor,  glass fiber substrates, stage 4.
            V. = 9.6 m/sec, Re = 773, calculated D5Q = 2,3 ytn.
                               42

-------
  100
   90
<#>
   10
     ,5 .6.7.8.91.0        2      3   4   5  6 7  89 10

                  Particle Diameter, Micrometers
20
 Figure 31.  Sierra Cascade  Impactor, glass fiber substrates, stage 5.
            V. = 17.1 rt/sec, Re = 778,  calculated D  = 1.1 pro.
                                43

-------
  100
   90
dip
   80
   70
-H

-H  60
c  50
o
•H
•P
o
a)  40
o
u
   30




   20



   10
         I  I  I  M
          i  i  i  i
-J—T
I  I  I  I
           O
      ,5  .6.7.8.91.0        2      3    4   5  6 7  89 10


                   Particle  Diameter,  Micrometers
                20
 Figure 32.  Sierra Cascade Impactor, glass fiber substrates, stage 6.
            V.. » 42.1 m/sec, Re « 1308, calculated  D5Q = 0.58 ym.
                             44

-------
Ul
     W
     (0
     o
rH

rH



I
           0.1
              0.5
                                                         Total

                                                         Nozzle

                                                         Inlet
                            1.52       3    456789 10

                            Particle Diameter,  Micrometers
15
20
                                                                                        W
                                                                                        CO
                                                                                        o
                                                                                        (fl
                                 Figure 33.   Sierra wall losses.

-------
Oi
       100





        90





        80


     <*>



     £ 7°
     U



     -X  60

     •rl
        50
     o
     o  30
        20
        10
                        A  D
                                             A

                                             a
                                                          A  D
                                      ill	i
             0.1  0.2  0.3  0.4  0.5  0.6  0.7  0.8  0.9  1.0  1.1  1.2   1.3  1.4  1.5  1.6  1.7  1.8  1.9

                                 Stokes Number         T
                              Stage   123456

                                    O D A V O  O


                  Figure 34.  Sierra Cascade  Itnpactor, glass fiber substrates, summary of

                              calibration  data.

-------
     Figures 35-43 - Brink Cascade Impactor
     Figures 35-43 show data for the Brink cascade impactor.   We have
modified this impactor to include:  (1) an in-line cyclone, (2) a zero
stage which is included in the cyclone above the first stage furnished
by the manufacturer, and (3) a sixth stage which is located downstream
from the smallest stage furnished by the manufacturer.  The design flow
rate for this impactor is 0.12 cfm; our tests were done at about 0.03 cfm.
With these modifications, the Brink impactor can be run at a low flow
rate and gives longer sampling times for better averaging  where dust  or
particulate loadings are heavy.  Also, the Brink impactor is the only
impactor for which we have experience with both grease and glass fiber
substrates.  Consequently, included in our calibration data are results
taken using both types of substrates.  Data taken with glass fiber
substrates are shown as open circles while data taken with grease substrates
(greased aluminum foils) are shown as closed circles.
     Figure 35 - Stage 0 - Brink
     Figure 35 shows data for the Brink impactor used with both glass
fiber and greased foil substrates.  As shown in the figure, the manufacturer's
DCQ is about 9.5 ym.  Our calibration data with ammonium fluorescein
however, show  the  D5Q to  be  approximately 8.5  ym.
     Figure 36 - Stage 1 - Brink
     This figure shows calibration data for stage 1, which is the stage
with the largest cut point furnished by the manufacturer.   The manufacturer's
DJ-Q for our test conditions would be 5.4 ym, while our calibration data
indicate  the  D5Q would be  slightly  smaller than  this.  We  do  not  have
adequate resolution to give the D5Q more precisely.  Notice that the peak
efficiency for the greased substrate is near 100 percent while the peak
efficiency for the glass fiber substrate is slightly over 90 percent.  In
this case our data indicate  that  the D™  for both substrates  was
approximately the same.
     Figure 37 - Stage 2 - Brink
     Figure 37 shows our calibration data for stage 2 for the  Brink
impactor.  In this case, once again the greased  substrate  shows a maximum
collection efficiency well in excess of 90 percent, while  the glass  fiber

                                    47

-------
100
    .5  .6.7.8.91.0       2      3   4    5  6  7  89  10

                Particle Diameter, Micrometers
20
Figure  35.  Modified Brink Impactor, glass  fiber substrates, stage 0.
           V. = 1.4 m/sec, Re  = 470, calculated D5Q = 9.5 ym.
                              48

-------
  100
   90
   80
<*>
o  70
c

•H

•H  60
§  50
•H
-P
o
0)  40
o
u
   30




   20



   10
      ,5  .6.7.8.91.0       2      3   4    5  6  7 89  10


                  Particle Diameter, Micrometers
                20
 Figure 36.  Modified Brink  Impactor,  glass fiber substrates, stage 1.
             V. = 2.9 m/sec, Re = 470,  calculated D
             J
50
   =  5.4
                               49

-------
  100
   90
   80
 o  70

 0)
•H

•"  60
w

g  50

-H
•P
O
o  40
O
u
   30



   20



   10
      ,5  .6.7.8.91.0        2      3   4    5  6  7  89  10


                  Particle Diameter, Micrometers
20
Figure  37.  Modified Brink Impactor, glass fiber substrates,  stage 2.
           V. = 5.8 ra/sec, Re =  659, calculated Dcn = 3.2 ym.
            3                                 ou
                                50

-------
substrate shows a maximum collection efficiency slightly under 80 percent.
Also, the DgQ for the stage which employs the greased foil substrate is
near 3.2 ym as given by the manufacturer, while the D5Q for the glass fiber
substrate is moved toward smaller particle sizes somewhat below 3.0 ym.
     Figure 38 - Stage 3 - Brink
     The trend established in Figure 37 can also be seen in Figure 38.
The maximum efficiency for the greased foil substrate is near 100 percent
while that for the glass fiber substrate is below 80 percent.  The D50 for
the greased substrate is near that given by the theory of Ranz and Wong
and by the manufacturer while the D5Q for the stage which employs the glass
fiber substrate is shifted toward smaller particle sizes and is approximately
1. 5 ym.
     Figure 39 - Stage 4 - Brink
     This figure shows calibration data for stage 4 for the Brink impactor.
Although our aerosol generator was not adequate for good calibration below
1  ym, that is, to determine the Dg0 for these stages, once again it can
be seen that the peak efficiency for the greased foil substrate is going
to be much higher than that for glass fiber.
     Figures 40 and 41 - Wall Losses - Brink
     Wall losses as shown for the Brink impactor follow the trends
demonstrated in the case of the other impactor calibration studies.  There
is sufficient scatter in our data that we are not able to determine if
wall losses are larger in the case of glass fiber or greased substrates.
It is clear, however, that in both cases a significant or large fraction of
the total wall losses is due to losses in the nozzle of the impactor.
     Figures 42 and 43 - Summary - Brink
     Figure 42 shows the graph of collection efficiency vs. square root
of Stoke's number for the Brink impactor with glass fiber substrates.
It can be seen that the lower stages with the higher jet  velocities never
attain efficiencies near the ideal value of 100 percent.  In fact, stage 6
appears to never achieve efficiencies over 50 percent.  It can also  be
seen that each stage has a characteristic calibration constant which  should
be used  in data reduction to calculate the cut point for  that  particular stage.
                                    51

-------
 100
  90
  10
    ,5 .6.7.8.91.0        2      3    4    5  6  7 89 10

                 Particle  Diameter,  Micrometers
20
Figure 38.  Modified Brink Impactor, glass  fiber substrates, stage 3.
           V. = 9.2 m/sec, Re = 833, calculated Dgo = 2.2 ym.
                            52

-------
   100




    90




    80
 u  70
 c
 (U
 •H


 •H  60
 w


 §  50
 •H
 -P
 O
 0)  40
    30  —
    20




    10
      .5 .6.7.8.91.0        2      3    4   5  6 7  89 10


                   Particle Diameter, Micrometers
20
Figure  39.  Modified Brink Jmpactor, glass fiber substrates, stage 4.
           V. = 20.3 m/sec, Re =  1236, calculated D5Q = 1.2 ym.
                              53

-------
01
      dP
      to
      W
70
60
50
40
30
20
10
2
1
0.5
0 2
0.1
0 Total
H Nozzle





















































i



|







i




























































•




.






















































'

















70
60
50
40
30
20
10 (a
U)
o
5 rH
IS
2
1
0.5
0.2
0.1
0.5 1 1.5 2 3 4 5 6 7 8 9 10 15 20
                                   Particle Diameter, Micrometers
              Figure 40.  Modified Brink  Impactor wall  losses, glass  fiber substrates.

-------
en
01
70
60
50
40
30
<*> 20
to
W
° 10
flj CJ
t^ «•*
2
1
0.5
0.2
0.1
0.
9 Total
B Nozzle
7n














































































1










•


















•

















•
m


.































i
1









"


















60
50
40
30
20
OlP
10 w
ui
0
P™1
rH
2
1
0.5
0.2
0.1
5 1 1.5 2 3 4 5 6 7 8 9 10 15 20
                                   Particle Diameter, Micrometers
               Figure 41,   Modified Brink Impactor wall  losses, greased  substrates.

-------
cn
OS
                                                                 D
                                             I    I     I    I    I     I    I     I     I    I     I    I
0.1  0.2  0,3  0.4  0.5   0.6  0.7  0.8   0.9  1.0  1.1


                          Stokes No.   rf


                    Stage    1234567

                            O  D  A  V O  • •
                                                                    1.2  1.3  1.4   1.5  1.6  1.7  1.8  1.9
                    Figure 42.   Modified Brink  Impactor, glass fiber substrates, summary  of
                                 calibration data.

-------
Ol
        100

         90

         80
       
-------
     Figure 43 shows similar data for the Brink iropactor with greased
substrates.  Although in this case, the stages in general  exhibit higher
peak collection efficiencies, they do not all  have the same calibration
constant.
     Figure 44
     Figure 44 shows data from our laboratory, which helps to illustrate
the reason that the stage collection efficiencies in some cases do not
reach 100 percent and, also, the reason that the stage collection efficiencies
turn down for large particles.
     Figures 44A and 44C each show round jet impactor configurations
where the D^Q for each stage was 1.8 ym.  In Figure 44A the jet velocity
was 11.4 m/sec while in the case shown in Figure 44C the jet velocity was
4.2 m/sec.  These jet stages were used to collect 2.8 ym diameter ammonium
fluorescein particles.  In the case shown in Figure 44A, 73 percent of the
particles were collected although ideally 100 percent would have been
collected.  In the case shown in Figure 44C, 92 percent of the particles
were collected.  Also notice that in Figure 44A the deposited patterns
of particulate are not sharply defined but are blurred and smeared on
the substrate while those in Figure 44C are nice, circular, compact deposits.
Figures 44B and 44D show stages that were downstream of those shown in
Figures 44A and 44C.  In the case shown in Figure 44B, the D™ for that stage
was 0.38 ym.  For the configuration shown in Figure 44D, the D5Q was 0.8 ym.
Thus, we would expect that 100 percent of the particulate would be caught in
both cases.  The jet velocity in Figure 44B was 45.1 m/sec, and only 79 per-
cent of the particulate was caught.  From experiments such as these we have
concluded that, when using glass fiber substrates to collect hard, bouncy
particles, jet velocities on the order of 10 m/sec are the maximum allowable
for adequate particle collection.  Also it can be noted that wall losses were
much more significant in the cases where we had higher jet velocities and
lower collection efficiencies for each stage.
                                    58

-------
a. Vj = 11.4 m/sec,73%

  COLLECTION
                                             b. Vj =45.lm/sec,79%
                                               COLLECTION. 32% WALL
                                               LOSSES IN a ANDb
          c. Vj =4.2m/sec, 92%
            COLLECTION
                                   d. Vj =9.5m/sec, 94%
                                     COLLECTION. 4% WALL
                                     LOSSES  IN c AND d
Figure  44.
  Particulate deposition patterns for  different flow
  rates.   In all cases  the particles were 2.8 ym
  diameter ammonium  fluorescein spheres.   a.  D5o=1.8  ym,
  b.   D50=0.83 ym, c.   D50=1.8 urn, d.   D50=0.38 ym.
                                   59

-------
     The data presented above show that impactors should actually be
calibrated to determine the cut points, rather than relying on
theoretical predictions.  Maximum allowable jet velocities appear to be
less than 20 m/sec for glass fiber substrates and somewhat higher for
greased substrates.  This also depends on the type of ^articulate sampled.

References
1.   Rao, A.K., Doctoral Thesis, University of Minnesota, 1975.
2.   Ranz, W.'E. and J.B. Wong, Ind. Eng. Chero. 44(6). 1952.
3.   Lundgren, D.A., vL._Ajjr. Poll ut... .Control_ASSQG^ 16, 225, 1967.
4.   Marple, V.A., Doctoral Thesis, University of Minnesota, 1970.
                                    60

-------
DISCUSSION

ENSOR:  Did you measure the actual hole diameters in the impactors.   We
did some analysis last year which indicates that the cut is very sensitive
to the diameter of the holes.  Sometimes the manufacturing can be off and
can cause some shifts in efficiency.
SMITH:  In some of the impactors we did.  We measured them on the Russian
impactor.  I believe we measured them all.  Do you remember, Joe?
MCCAIN;  We measured the Brink, the Andersen, University of Washington,
the Russian.  I don't know whether we measured them on the MRI.
SMITH:  In most cases, they showed less than 10 percent error in diameter.
RAO:  How did you measure the hole diameter?
SMITH:  With an images-shearing eyepiece on a microscope.
RAO:  We found that, if you measure the diameter by pressure drop, you  can
get much more accurate measurements.
SMITH:  More accurate measure or more predictable performance?
RAO:  Yes, it's a good measure—more accurate.
SMITH:  We found what we thought was so little error, although there is a
very  strong dependence on jet diameter, that we didn't do anything with it.
For example, we didn't feel that we needed to correct any of the calculated
Den's because of the small errors that we found.
BLANN:  Is this roll over in this curve something new?  Or has this
been  seen before?
SMITH:  We published some data to that effect in the Report  21 of ours to
EPA, which got considerable circulation,  I don't know how much was in the
literature before that.  That was in 1972.   It's not surprising, when
you think about the details of what is going on.  The larger particles
hit with a lot of momentum and bounce.  I think  it  has  become  commonplace
knowledge in the last year or so.
                                    61

-------
MCCAIN:  The expectation for that kind of behavior was  realized before
we first started this stuff.  Some people said  when we  first  started
impactor work; "It'll never work, look at a sandblaster."  The thing
that's remarkable here is that they've been any good.
	:  You made the comment several  times, "This  is  a  well-behaved stage."
Well, define well-behaved stage.
SMITH:  By that I mean that it went up to near  100 percent (collection)  and
then collected essentially 100 percent of the larger  particles over a
pretty broad range.  So if upstream from that,  you have a  stage with a  D5Q,
say, of twice that D™, that stage will collect any particles of  a size
that would cause the efficiency to roll back down. If  you have got a
sharply peaked stage efficiency, then you must  have another stage upstream
with a slightly larger D5(s.  If you overlay the stage efficiencies, you
would hope that, before one turns down to large particles, the other would
pick up and be collecting large particles.  The perfect stage would be
one that goes up to 100 percent and stays there for all large particle  sizes. '
BOLL:  I wonder if you could tell me what your  particle material  was and
would you elaborate on why the bare metal and the Teflon were totally
inadequate.
SMITH:  The material was ammonium fluorescein,  and we picked that material
because of a paper that was published.  I can't remember the author's  name.
RAO:  Stober and Flachsbart.
SMITH:  That's right.  It is nonhydroscopic, whereas  ordinary Uranine  is
hydroscopic, and we wanted a material that was  bouncy to simulate fly  ash.
There have been studies similar to this done with oils, and in  those cases
glass fiber, bare metal, Teflon, and certain materials  worked well as  sub-
strate materials.  Most of our work is done at  power plants, and fly ash  is
a bouncy material.  We wanted to simulate that  behavior.  In the case of
Teflon, the material doesn't deposit right under the  jet; it just blows
around and settles where there are eddies on the surfaces, and much of it
ends up downstream on the backup filter.  Essentially,  the same thing happens
with bare metal when you have a bouncy material; it just blows  right off.

                                     62

-------
BOLL:  Then your basis for saying that is observing a dune structure on
substrate.
SMITH:  Right, and if the stage had a predicted efficiency of 100 percent
and we got 10 percent, we would say that that's a reason for not using
it also.   The material just blows off, and a lot of it ends up downstream.
OLIN:  Wallace, you plotted the collection efficiency of the incident
aerosol.
SMITH:  That's right.  The collection efficiency was calculated from a
knowledge of  the amount of material which was captured on a given stage
and  of the amount  of  material which penetrated that stage.
OLIN:  So, therefore, if the preceding had been a normal impactor acting
in the cascade configuration, many of these particles that do bounce and
are  shown  on your  efficiency curves would be removed before they got there.
SMITH:  Some of them would be removed before they got there.
OLIN:  Yes, some of them.
SMITH:  These are  curves for a particular stage and actually the efficiency
of the stage doesn't enter into the data reduction as it is done; normally,
you  just  put  in the D^Q.  The assumption is that above a certain size,
100  percent are caught.  I'm not quite sure what effect this has on the
calculated size distribution, certainly it would shift.
OLIN:  So  if your  first stage is quite efficient, then most of these
impactors  did show that you've been removing a good portion of those large
particles  or  if you had a preseparator, such as cyclone preseparator, you
would be  removing  those particles before they had an opportunity to bounce.
SMITH:  Well, that is sort of true, but there is a size at which the first
stage has  zero collection efficiency and it may be like 5 or 6 microns, so
those particular sizes would reach the downstream stages where they must
be caught.
HARRIS:  The problem  is that each one of these stages has an efficiency.
It's a problem then of spacing the cut points to that a stage upstream of
it is collecting   50 percent of the particles but the other stages are
collecting 0 percent of this.  That means 50 percent of the particles are
going to go flying through.
                                   63

-------
 SMITH:   The  problem is  that  large particles which penetrate past the
 stage where  you  would like them  to  be caught are on  the negative slop of
 the efficiency curve of downstream  stages, and there is a decreasing
 probability  that they will be  caught as  they go downstream.  Thus,  they
 may well  end up  on  the  backup  filter, where a single large particle can
 be counted as a  large number of  small particles.
 SPARKS:   In  analyzing these  stages, stage by stage,  this is fine, but have
 you done  any work where you  put  in  a known size distribution with an
 assembled impactor  stages and  came  out with results  on the size distribution
 as you  interpret it and relate that to what you expect.
 SMITH:   The  impactors were all run  completely assembled; i.e., we did not
 disassemble  any  to  do the analysis  stage by stage.   They were operated as
 a whole and  the  analysis was done stage  by stage.  However, the size
 distributions were  essentially all  monodisperse.  We never had a polydisperse
 aerosol  of known size distribution.
	:  You haven't tried to  get the empirical  data  on known  size  distri-
butions; see what the results would be on the basis of analysis of that?
SMITH:  Right, we have done a little bit of that theoretically, and we also
compared the performance of different impactors in the field.
HARRIS:  We've done the same thing on the wind tunnel; Bill [Kuykendal]
attempted that.
DICK:  Could you tell us how much material you were depositing per jet?
SMITH:  Well, I really can't.  In some cases, with the larger particles it
is fairly easy to deposit a total of approximately a milligram.  For the
smaller particles, the analysis by washing the impactor and determining
the mass was done with absorption spectroscopy.  We were specifically
avoiding the very long run times necessary to deposit enough material  so
that we could weigh it.
     The amount of bounce you can get probably depends on  how much you
put down.
                                    64

-------
SMITH:  I think it definitely does.
MCCAIN:  The deposits were always several particle layers thick.  I don't
think there was anything that was done  in which the deposit was essentially
monolayer.
SMITH:  For example,  in the case of tests using very small particles
would run the  impactors for two days  to catch e'nough material for analysis.
But, as I said in my  opening remarks, Lundgren has shown that the collection
efficiency is  a function of time just as you suggested.  He showed curves
with grease substrates where the efficiency changed, I believe, from around
90  percent to  70 percent over  a 20  to 30 minute period.
                                     65

-------
                   SUBSTRATES AND IMPACTOR CUT POINTS
                A. Kishan Rao, Midwest Research Institute
RAO:  Primary fine participate emissions from industrial  sources  are
one of the EPA's major control targets.   For this reason, EPA is  interested
in characterizing particulate emissions  from various  sources  and  in evalu-
ating emission control devices.  Cascade impactors of various designs have
been used for a number of years for physical and chemical characterization
of aerosols.  They are well suited for in situ source testing because of
their simple construction and operation, compactness, ability to  withstand
high temperatures and pressures, and perhaps most importantly, because they
classify particles according to their aerodynamic size.   Additionally, they
have much better size separation properties than competing techniques (e.g.,
cyclone separators).  Impactors, when used properly,  can  provide  reliable
particle size distribution data.
     Because of their wide use, impactors are the subject of considerable
theoretical and experimental investigation.  Ranz and Wong  developed a
                                             I
simple theory in 1952, which is generally used in predicting impactor per-
                                  2 3
formance.  Recently, Marple et al. '  developed a numerical calculation
procedure more accurate than the Ranz and Wong theory.  These theories how-
ever, are limited to impactors of simple geometry operating under idealized
conditions.  It will be shown in this paper that, for impactors such as the
Andersen sampler, these theories are inadequate, and experimental calibrations
are essential if reliable data are to be obtained.
     It is well known that solid particles may not stick to collection
surfaces unless the collection surfaces  are coated with some adhesive.  In the
event of particle bounce, reentrainment, and deaggloraeratton or breakup, the
impactor loses its ability to classify particles according to size, resulting
in erroneous and misleading size distribution data.  The conventional method
of reducing particle bounce and reentrainment is by coating the collection
surface with viscous oils or with other adhesive materials.  However, when
samples are to be analyzed for chemical  composition, use of adhesive coatings
is not recommended for fear of sample contamination and  possible interference

                                    66

-------
with analysis techniques.  In recent years, fibrous filters have gained wide
acceptance as collection surfaces because they facilitate sample handling and
appear to reduce particle bounce.4  However, the real collection character-
istics of these filters had not been investigated prior to this study.
     This paper presents the main results of an experimental study which
was aimed at defining some of the nonideal collection characteristics of
inertial impactors.  Specifically, the problems of particle bounce and
"blow-off", or reentraihment, were studied.
     In this study, two single-stage and two cascade impactors were used.
Figure 1 shows a schematic diagram of the laboratory impactor.  The jet-
throat has a 60° conical entrance.  The jet-to-plate distance is approxi-
mately one jet diameter.  Figure 2 shows a schematic diagram of the Bureau
of Mines impactor.  The design of this impactor is similar to the laboratory
impactor with the exception of jet-to-plate distance, which in this case,
is 3.44 times the jet diameter.  The Lundgren impactor shown in Figure 3
is a four stage, rectangular jet impactor, and it is designed such that all
of its stages are geometrically and dynamically similar.  The Andersen
sampler shown in Figure 4 is a six stage, cylindrical, multihole impactor
designed for atmospheric sampling.
     Figure 5 shows some of the equipment used in this study.  The vibrating
orifice aerosol generator and the polystyrene latex  (PSL) aerosol generator
were used to produce the test aerosols.  Precautions were taken to produce
monodisperse aerosols of spherical, dry, and electrically neutral particles.
Four types of collection surfaces were tested.  They were:  (a) oil-coated
smooth surface  (glass plate in the case of single stage  impactors and stain-
less steel plates in the case of the Andersen sampler);  (b) uncoated sur-
faces; (c) Gelman Type A glass fiber filter; and  (d) Whatman No. 41 filter
paper.  The oil was either Dow Corning 2000 fluid  (with  60,000 cs viscosity)
or petroleum jelly.
     The collection efficiency curves of single-stage  impactors were deter-
mined as follows.  The PSL aerosols were-generated and were sampled at  a
known flow rate with the test  impactor, which was.preassembled with  one of
the collection surfaces to be tested.  Collection efficiency was determined
                                    67

-------
                                     JET DIAMETER = 1.32 mm
                                    THROAT LENGTH = 1.32 mm
                              JET TO PLATE DISTANCE = 1.24 mm
                        i-AEROSOL  INLET
COLLECTOR BASE
                                        COLLECTION SURFACE
                                          TO FLOW METER
                                                       7
  10  20 mm
SCALE
             Figure 1.  Laboratory impactor.
                           68

-------
                        JET DIAMETER = 1.384 mm
                      THROAT LENGTH  = 2.768 mm
                 JET TO PLATE DISTANCE =  4.76 mm
                    BIG INLET NOZZLE
                    TEST INLET NOZZLE
                    COLLECTION SURFACE
                        SUPPORT WIRE
                        COLLECTOR BASE
                        TO A STANDARD TEE
                                           0   10 mm
                                            SCALE
Figure 2.  U.S. Bureau  of  Mines  Irapactor.
                      69

-------
AIR INLET     I St. STA6E NOZZLE
I st. STAGE ROTATING DRUM
                       Figure  3.   Lundgren  Impactor.
                      Figure 4.  Andersen Sampler.
                                      70

-------
   Compressed
   Air
             PSL Aerosol
             Generator  Aerosol
                        Solution—*
        Rotameter

        Valve

I   \^S  Pressure  Regulator

i*'.'.'.' iV'ii' . Dryer

        Absolute Filter
      Vibrating Orifice
    -  Monodisperse Aerosol
      Generator

                  Isokinetic
                  Sampler
Andersen
Sampler
Compressed
Air
                                                                 Vent
                                               Optical Particle
                                               Counter
                                                                                Compressed Air
_+. Vacuum.
   Pump
       Figure  5.   Schematic  diagram of an experimental  set  up.
                                         71

-------
by measuring the aerosol  concentration upstream and downstream  of  the  im-
pactor with an optical  particle counter.   Stokes number was  varied by
varying the flow rate through the impactor for a given size  particle.
      The Lundren impactor and the Andersen sampler were  calibrated con-
currently with oleic acid particles tagged with uranine  dye  tracer. The
flow rates through these impactors were 112.3 and 28.3 1/min,  respec-
tively.  The collection efficiency and wall loss of each  stage  as  a function
of aerodynamic diameter were obtained by analyzing the collected material
by fluorometric techniques.
      Particle bounce in the Andersen sampler was determined by sampling
methylene blue (solid)  or PSL aerosols and measuring the  collection efficiency
with the optical particle counter.  The aerosol flow rate of this  impactor
was kept constant at 28.3 1/min except for a few special  cases  and
particle size was varied to obtain various aerodynamic diameters.   Details
of the experimental procedure are available in Reference  5.
      A.  Calibrations with Oil Particles or Oil Coated  Plates
      Figure 6 depicts the collection characteristics of the laboratory
impactor with different collection surfaces.  It can be  seen that  the  col-
lection efficiency of the Dow Corning oil-coated plate  and petrolatum
jelly-coated plate follow a single curve.  This efficiency curve is in
good agreement with Marple's theory.
      Figure 7 shows the collection characteristics of the Bureau  of Mines
impactor with various collection surfaces.  A good agreement between oil-
coated plate efficiency and theoretical efficiency is evident.
      Figure 8 presents the experimental results for the Lundgren  impactor
operating at a constant flow rate of 112.3 1/min.  For comparison
also shown in this figure are Marple's theoretical curve and Lundren's
experimental curve published previously.   Theoretically, if the stages of
an impactor are geometrically and dynamically similar, their collection
efficiency curves should be similar.  For such an impactor, the collection
efficiency curves of different states (i.e., efficiency plotted versus the
square root of the Stokes number).should fall on a single curve.   The pre-
sent data adequately define a smooth "S" shaped curve which is in  good
                                    72

-------
                          i	1	1	1	r
                              Laboratory Importer
                      Oil Coated Clou Plain
                 |00j_  VII VVUIHU VlUti rUB
                   A DOW Conk* Oil, I.I /in PSL
                   $HHn«UI 11971), II ,im PSL
                 „, 4> Pltrotatum Mly, 1.1 ,1111 PSL
                   OPlMHotum JtHl. I.O/lmPSL
                  80-
                     6km Fib«r Filter
                   _ OI.I /m> PSL
                     0.79 pin PSL
                     & 794 phi PSL
                 1,0
                                       Unconled 6lois PI on
                                   s     S 1.10 /in PSL
                                        Q I.OII/imPSL
                                 ""J	     I
                     35    .40
  Figure  6.   Collection characteristics of the
                laboratory impactor.
              100
                            Bureau of Mines Impactor
                                                     .70
Figure 7.    Collection characteristics  of the Bureau
               of Mines  Impactor.
                                 73

-------
 100
O-i
Figure 8.   Calibration curve of the Lundgnen  Impactor.
                        74

-------
agreement with both theoretical efficiency curves and Lundgren's experimental
curve.  This agreement between theoretical predictions and experimental
calibrations indicates that the theory can accurately predict the performance
of impactors of this type.
     Figure 9 illustrates the collection efficiency curves of the Andersen
sampler operating at a flow rate of 28.3  1/min.  These calibration
curves were obtained with oil aerosols and stainless steel collection
plates.  For comparison Marple's theoretical curve corresponding to a
jet Reynolds number of 3000 and jet-to-plate distance of five jet diameters
is also plotted in Figure 9.  It can be seen that the efficiency curves  of
different stages have different slopes and the curves do not coincide.   The
experimental efficiency curves have smaller slopes and are shifted towards
higher Stokes numbers relative to the theoretical efficiency curve.  This
result means that for this type of impactor theoretical prediction of the
performance would result in considerable error,  and experimental calibrations
are essential if reliable data are to be  obtained.
     The reasons for disagreement between the theory and experimental
calibrations of the Andersen sampler are the following:
     1.  The theoretical curve is developed for a simple geometry
         impactor such as a single jet impactor with jet-throat
         having a conical entrance and operating at a jet Reynolds
         number of 3000.  The Andersen sampler is a multijet, straight-
         hole impactor with jet Reynolds numbers between 84 and 400.
     2.  In the case of multijet impactors, it is usually assumed that
         flow through each jet is equal.  However, in the Andersen
         sampler this assumption does not seem true, especially for
         Stages 1 and 2.
     The effectiveness of an adhesive coating to prevent bounce may decrease
with stage loading.  To investigate this possibility, several tests were
performed with the Bureau^of Mines impactor.  Because of the large jet-to-
plate distance of this impactor (3.44 jet diameters) slight changes  in the
jet-to-plate distance with load were not expected to affett its performance.
The impactor plate was coated with a thick layer of petrolatum jelly and
1.1 vim PSL aerosol was used as the test aerosol.  The results are  shown
in Figure 10.  In this figure, we see no-  indication of bounce with
loading.  However, the overloaded surface has a  much higher collection
efficiency than the fresh surface.
                                    75

-------
  100
   90
   80
   70
   60
U

OJ
uj  50
O
i  40
o
   30
   20
   10
                        i
       MARPLE et ol.
       THEORY
rO	1
                                     ANDERSEN SAMPLER (28.3 Ipm)
                     n
0.4     0.5     0.6     0.7     0.8
                                                      1.0     l.t
       1.2
         Figure 9.  Calibration curves of the Andersen  Sampler.
                            76

-------
              100


              90


              j



             *TO

              -sol
             .1
             •5
             230

              20

               10
   I     |     I
Button of Minn Importer ._„
             ^--t*
        * TNOHCW »wtw» nil* •««*»lta»»l *•'«*>• '
        O 0» Cww,o M tMM «l«t fifcw HIM
               T40   .45   .50  .55   .60   .65  .70
                            ^TK
   Figure 10.   Effect of  Load on the  Impactor
                 Performance
                          AERODYNAMIC DIAMETER , f
Figure 11.   Collection Efficiency of the Andersen  Sampler
              with Various Collection Surfaces
                           77

-------
     On examination of the overloaded surface it was found that the
adhesive actually rose to the top layer of the collected particles  by sur-
face tension forces.  These results suggest that the adhesive selected
should have good wettability characteristics and the aerosol  desposition
rate should be such that there is time for the adhesive to rise up  to the
precipitate surface before any subsequent particle impingement.  The
increased collection efficiency of the overloaded surface also suggests
that a correction may be necessary for the impactor constants if they were
originally obtained with fresh smooth surfaces, but are actually used with
rough surfaces or surfaces having significant particle loading.
     B.  Analysis of Particle Bounce and "Blow-Off"
     Figures 6 and 7 show the collection characteristics of the laboratory
impactor and the Bureau of Mines impactor with various collection surfaces.
In these figures, we see that for solid particles the collection efficiency
curves of the uncoated glass plate follow the efficiency curves of the oil-
coated plates only up to 20 to 50 percent efficiency.  At higher Stokes
numbers, collection efficiency curves of the uncoated glass plate drop
sharply.
     Collection efficiency curves of filter surfaces are qualitatively dif-
ferent from the efficiency curves of oil-coated glass plates.  At low Stokes
numbers for which oil-coated plates have zero or low collection efficiency,
the efficiency of the glass fiber filter is much higher than the efficiency
of oil-coated glass plates.  At higher Stokes numbers where the efficiency
of the oil-coated plate is near 100 percent, the efficiency of glass fiber
filters.attain a plateau of 75 percent due to the limiting effects of the
surface filtration effect.  Whatman filter paper, being harder and less
porous, has a lower collection efficiency than a glass fiber filter over
the entire Stokes number range.
     Results of particle bounce and blow-off studies with the Andersen
sampler are shown in Figure 11, in which the collection efficiency as a
function of aerodynamic diameter is plotted for each stage with oil-
coated stainless steel plates, uncoated stainless steel plates, and the
glass fiber filter.  The data points used to draw each of these curves are
                                    78

-------
not shown for the sake of clarity.  The shape of the curves in this figure
is similar to the shape of the curves in Figures 6 and 7 adding to the
generality of the results described earlier.
      Figure 11 shows that particle bounce with uncoated plates can signi-
ficantly lower the collection efficiency and is especially serious for
stages which collect small particles at high jet velocities.  It appears
that, in the presence of bounce, efficiency is not controlled by Stokes
numbers alone.  Particle size, jet velocity, and other factors affecting
particle adhesion also affect the collection efficiency.  Glass-fiber-filter
collection surfaces reduce particle bounce significantly, but change the
characteristic curves in such a way as to reduce the size resolution of the
impactor.  Compared to a smooth surface, the glass fiber filter becomes
increasingly efficient as the particle size decreases.  This improvement
is due to the changes in the relative roughness of the filter surfaces as
can be seen in Figure 12.  For upper stages which collect larger particles,
the relative roughness of the filter surface is low and the filter behaves
very much like a smooth surface.  However, for lower stages which collect
relatively smaller particles, the fibers and the void spaces of filters are
big compared to the particles, and the particles have a greater chance of
being collected when moving close to the filter surface.
      The following conclusions can be drawn:
      1.  For impactors such as the Andersen sampler  (six stage, ambient
          type) whose design and operation do not satisfy all the
          assumptions of the Marple theory, the theory cannot predict
          the performance.  The experimental stage cut-off  diameters of
          the Andersen sampler are significantly different  from the
          theoretical stage cut-off diameter.  This result  suggests
          that using the Ranz and Wong theory or the Marple theory to
          obtain stage cut-off diameters of these impactors for calcula-
          ting the.size distributions results in  considerable  error.
      2.  The performance of an impactor is significantly affected by  the
          nature of its collection surface.  Glass fiber filters,
          generally used to reduce particle bounce, shift the  collection
          efficiency curves and decrease the sharpness of  cut.  The magni-
          tude of the shift is different for different stages.  This
          result suggests the need for calibrating the  impactors  under con-
          ditions in which they are used.
                                      79

-------
      (A)
(0
CO
o
          (B)
(D)
               Figure 12.  Electron  micrographs of the glass  fiber  filter surface showing the relative
                           size of particles and fibers,   (a)  0.481  ym PSL; (b) 2.75 urn methylene
                           blue; (c)  3.30 vm methylene blue;  (d)  3.81 urn methylene blue.

-------
     3.   Particle bounce is a function of both particle diameter and
         jet velocity.  When particles are bouncing the collection
         efficiency is not a function of Stokes diameter,  but also
         depends upon the jet velocity.  Therefore, if bounce is to
         be minimized, it is preferable to use multiple jet impactors,
         which have lower jet velocity for a given cut size.
     4.   The performance of impactors changes as the collected parti-
         culate material piles up on the collection plate.   The stage
         cut-off diameters decrease as the stage loading increases.


References

1.   Ranz, W.E., and O.B. Wong, Impaction of Dust and Smoke Particles,
     Ind. Eng. Chem., 44:1371 (1952).

2.   Marple, V.A., B.Y.H. Lui, and K.T. Whitby, Fluid Mechanics of  the
     Laminar Flow Aerosol Impactor, J. Aerosol Sci., 5;1 (1974).

3.   Marple, V.A., and B.Y.H. Lui, Characteristics of Laminar Jet Impactor,
     Env. Sci. Tech., 8:648 (1974).

4.   Hu, J.N.H., An Improved Impactor for Aerosol Studies Modified
     Andersen Sampler, Env. Sci. and Tech., 5^:251  (1971).

5.   Rao, A.K., An Experimental Study of Inertial  Impactors, Ph.D.,
     Thesis, University of Minnesota, Minneapolis, Minnesota (1975).

6.   Lundgren, D.A., An Aerosol Sampler for Determination of Particle
     Concentration of Size and Time, J. Air Poll.  Cont. Assoc., V7:225
     (1967).
                                    81

-------
DISCUSSION

HARRIS;  Essentially it says  that  if the  particle  is not caught where we
want it to be caught, it's  going to be  on the  filter if it's not somewhere
else.
RAO;  Right
SMITH;  In that case, you would be a lot  better  off throwing away your
backup filter catch.
RAO:  Yes.  However, the filter catch can give an  estimate  of  the magnitude
of bounce.  If bounce is severe, we would be better off eliminating  the
entire run than to get data,  which results in  distorted size distributions.
MCCAIN;  It doesn't mean much.   Do you  have enough information on the rate
of buildup of efficiency change with buildup of  particles on the stage to
estimate how much the effective cutoff  of the  stage will change as you do
some sampling in a reasonable situation?
RAO;  The study concerning  the effect of  load  on impactor performance was
the result of a half day's  work.  It was  done  along with Dr. Alex Berner
of University of Vienna, Austria,  who first reported the capillary  action
of the precipitate.  Although the  present data are limited, they  clearly
show another nonideal characteristic of the impactor,  which needs some  study.
ENSOR;  Mercer reported a number of years ago  that efficiency  increases  as
you get buildup because you reduce plate-to-collector  distance.   Is  the
increase in efficiency due  to decreased jet-to-plate distance  with  buildup?
RAO;  Yes, the efficiency increases with  decrease in jet-to-plate distance.
However, the impactor chosen here  has a jet-to-plate  distance  of 3.44 jet
diameters.  This spacing is large  compared to  the buildup  on the plate,  and
therefore, decreased jet-to-plate  distance and should  not  have contributed
significantly to the observed increase in collection.
HARRIS;  One problem we have to worry about is that the capillary action
does not stop at the top layer—it just keeps  on going into the air.
                                    82

-------
	:  How did you apply oil or grease to the Andersen?
RAO:  By dissolving Dow Corning oil in toluene and spreading a few drops
on the plate with a fingertip.  The toluene is allowed to evaporate by
placing the coated plates in a clean hood.
                                     83

-------
               EXPERIENCES IN USING CASCADE IMPACTORS  FOR
                   TESTING ELECTROSTATIC PRECIPITATORS
             Joseph D. McCain, Southern Research Institute
HARRIS:  Now we'll ask Joe McCain of Southern Research Institute  to talk
to us about electrostatic precipitators.
MCCAIN:  This presentation will  primarily be a discussion of problems
that we have commonly encountered in field sampling programs on  ESP's
on coal-fired power boilers and  the significance some of the problems
might have.
     The first problem that we have is the location of the sampling ports
(Figure 1).  The typical sampling locations that are provided around pre-
cipitators, at least at locations where we have had to sample, very seldom
are anything like 14 duct diameters downstream and 5 diameters upstream
from flow disturbances.  Generally the sampling locations are, instead,
only one or two diameters from a disturbance.  This means you have to
sample an awful lot of points in order to get a decent traverse,  if the
velocity distribution is such that it is even possible to do so.   Sampling a
lot of points with impactors becomes very expensive.  We generally compromise
by reducing the number of points we sample in order to hold costs down.
That automatically introduces errors into the data, particularly for
large particles.
     Figure 2 shows a typical size distribution taken in the field on a coal
fired boiler, in terms of relative amounts of material in logarithmic size
intervals.  The particular ESP inlet data shown in Figure 2are not typical
of most installations on pulverized coal  boilers.  The distribution in
Figure 2 is deficient in large particles because on this particular occasion
the samples were taken downstream of a mechanical collector,  which  had
removed the bulk of the large particles that would normally be present.  The
error bars shown are one standard deviation about the mean for a number of
runs made on this particular boiler.  In this instance, in order to conserve
time, one impactor run was made  at each of 16 inlet ports with a four-point
traverse of the duct being made  on each run.  So Figure 2 shows 16  inlet runs

                                    84

-------
\
               UPPER
             ELECTROSTATIC
             PRECIP1TATOR
            wu
Figure 1.  Typical locations of test points for control
         device evaluations.
                   85

-------
 10,000
   1000
5
o
en
a
s.
o>
E
    100
     10
1—I  I  I 11II
                          I   I  I  I  MM	1    I   I  I  I I  IN
      0.1
                                          INLET DATA
i    i   I  i  i 11
                                    I    I   I  I  I 111
                 I                      10

                 PARTICLE DIAMETER ,
                                              I    1   I  I  I I  I I
                                                            100
Figure 2.   Average of the inlet impactor data for Naughton tests.  The
           error  bars show ± 1" standard deviation.
                                  86

-------
averaged.  Part of the variation in the data is the result of measurement
errors, and part of it may well be the result of variations in concentrations
across the 16-port traverse.  Mass train measurements, made at the same
time that the data shown in Figure 2 were being taken, showed something  like
a 20- to 30-percent standard deviation in the total particulate loading, and
that would be reflected in the impactors.  Perhaps the size distribution
changes that are occurring might be expected to show somewhat greater changes
than encountered in mass train measurements.
      Figure 3 shows outlet data, at the same location, with standard devia-
tion bars.  For fine particles, the scatter is relatively small.  For larger
particles, it becomes very large, approaching nearly 100 percent at 10
microns.  There are several reasons for this; we'll go over a couple of
them in a moment.  There are two sets of error bars here, the smaller set
represents another phenomenon which we will come back to later.
      Figure 4 illustrates a major problem in measurements at the exit of
precipitators collecting dry particulate.  The precipitator plates are
cleaned periodically by rapping.  This leads to periodic changes in the
size distribution with large swings in particulate concentration, especially
in the large sizes.  Figure 4  is a real-time particle monitor trace over a
48-minute period during a 1-hour rapping cycle on this particular precipitator.
Each of the spikes is a rapping puff.  They are quite evident in the 0.6 to
1.8 micron size range and somewhat less noticeable in the 1.8-3.5 ym range
although they're still evident.  It's not particularly evident that they're
there in the larger sizes; however, they really are there as well.
      Figure 5 shows the result of 10-minute integrations of the particle
counter output over a 10-hour  period.   It shows concentrations of particles
as measured with real-time monitors in 6- to 12- and  12- to 24-micron size
ranges and shows one of the reasons for an extreme amount of scatter in the
data obtained with impactors at the exit of precipitators.
      About once an hour a major peak in concentration resulting from the
rapping of the last field of the precipitator was  expected.  And  in fact,
we see that roughly once an hour we get such a  peak.  The magnitude of  that
peak is quite variable.  (This sample monitor was  operating at  a  fixed  loca-
tion.)  About 10 o'clock we got a very  large concentration  spike, and again
at about 3:30 we got a tremendous peak.
                                    87

-------
    10
                        T	1—I   MM
                           T	1   I   M I II
o
en
a
x
o
o>
x1
£
•o
                             i
   0.5
   0.2
               RAPPING TEST -OUTLET
   O.I
                     I    I   I   I  I  I  1
                                                    J.
                                »   Mt  i i li
    O.I
Figure 3.
                 0.2
0.5        I          2

 PARTICLE DIAMETER,j
10
        Average of the outlet Size distributions.  The larger error bars
        show ±1 standard deviation of the parti cul ate catches.  The shorter
        error bars show the fraction of the uncertainty which can be attri-
        buted to the scatter in the blank weight gains.

-------
              45
    u
    w
    en

    co
    w
    I-)
    u
    M
    H
    PU
CO
    o
    H
    55
    S3
    O
    o

    w
    H
    <
    i-J
               0

              50
  0

 50
  0

500
                                                          _    ..  L I  I
                    48      42       36      30        24       18      12       6         0


                                                  TIME,  MINUTES


                      Figure 4.   Typical data segment from large  particle system,  Aug.  6,  1975.

-------
   7.0
   6.0
    5.0
iu

§  4.0

I
x
OT
UJ
I-
oc
2
   3.0
   2.0
    1.0
                               •    6-12  yiti

                               O   12-24  yra
                                                         HI	*|SOOT
                                                                  BLOWING
      9=00    10=00   11=00    I2--00    1=00   2=00   3--00   4=00    5=00    6=00


                                      TIME, hours
  Figure 5.  Particles/minute  vs.  time  for  large particle system on Aug. 5,
             1975, rappers on.
                                   90

-------
      Cascade impactors was sampling at the same time to provide mass data
as time-integrated averages.  These were the data that were shown in Figure
3(an average of three simultaneous impactor runs).  The outlet loading
at this particular precipitator, a 99.95-percent efficiency percipitator,
was so low that the impactor integration times required to obtain reasonably
weighable quantities were about 20 hours with the impactor flow rates which
were selected.  Each of the three impactors performed a 24-point traverse
of the exit plane twice during the 20-hour sampling period.
      The magnitude of these rapping puffs is spatially dependent.  A large
part of the rapping losses tend to be confined to lower parts of the ducts
and as a consequence, the results that one obtains with the impactor
depends on the phasing of the sampling locations with the rapping of the
precipitator plates.  A significant amount of scatter could be introduced,
particularly in large particle sizes as a result of that effect.  It gives
one a tremendous headache in trying to analyze the data and determine what
the efficiency of a precipitator is for large particles.  One calculates
that it ought to be very, very high.  Figure 6 shows relative percentages
of the total exit emissions resulting from the puffs for various particle
sizes.  At about 20 microns and up essentially everything  appears to be
in the rapping puffs.  At 2 or 3 microns, most of the emissions appear to
be material which has gone through uncoilected rather than being collected
material reentrained  in the gas stream when the plates are rapped.
      That material, which was rapped off the plates and introduced  into the
gas stream in the puffs at this particular installation, had  a size  distri-
bution as shown in Figure  7.  It was approximately log normal, with  a mean
size of about 15 or 20 ym.  With a mechanical collector preceding this
precipitator, there shouldn't have been much at 15 to 20 ym going in.  The
material in the rapping puffs appears to be agglomerates,  which  leads to
another problem in interpreting the data.  At the inlet to the  precipitators
we had primary particles—for the most  part individual spherical  particles.
At the outlet the larger particles were agglomerates  comprised  of much
smaller particles.
      Stage one of the Andersen impactor, which we were using in this case
for the outlet measurements, collected  particles  having diameters greater
                                    91

-------
       100
        80
     O
     cc  crt
     u.  60

     V)
     u.
     u.
     5  40
        20
1     '    T
                    1,1,1,1
          8         12

       PARTICLE  DIAMETER,
                                                 16
20
Figure 6.   Contributions  of rapping  puffs by different particle
           diameters.
                               92

-------
IO
CO
       tu
       M
       V)
       O
       til
       O
       O
       z
       <
       CO
       CO
       UJ
0.01

0.05
 O.I
 0.2
 0.5
   I
   2
   5
   10

  20
  30
  40
  50
  60
  70
1    I
                                    M
            80
            0.2
1    |  I  I  I  I
                     I    I    I   L  1 i L
                                      I    I   I   I  I  I  I
                                I     I   i
                    I                                   10
                          PARTICLE  DIAMETER, um
                                       50
99.99

99.9
99.6

99
98

95
90

80
70
60
50
40
30
20
                                                                                            <
                                                                                            cr
                                                                                            UJ
                                                                                            5
                                                                                            cc
                                                                                            O
                           Figure 7.   Cumulative mass distribution of a  rapping  puff.

-------
than 16 ym (aerodynamic diameter)  or about  10  urn Stokes  diameter.   Micro-
scopic examination revealed that there  were no primary particles on stage
one with a diameter larger than  7  ym and, indeed, very few  particles with
7 urn diameters.   On Stages 1,  2, and 3  of the  impactor samples  from this
test, the bulk of the material appeared to  be  in the  form of agglomerates
of 1 or 2 ym particles.  The agglomerates were roughly spherical in shape
and in the form of very tight, compact  groups.
      Stage 4 of the impactor has  an aerodynamic diameter cut size  of  about
6 ym (Stokes1 diameter of about  3.5 microns).   The particulate catch of that
stage appeared to be dominated by  primary particles of the  expected size.
      Figure 8 shows the measured  fractional efficiency  of  the precipitator.
The bars represent efficiencies  calculated  from the impactor data  using
the outlet means minus one sigma and inlet  means plus one sigma and vice-
versa to get art approximate maximum uncertainty range which might  apply
to the efficiencies.  The solid  line is the measured  curve, taken  directly
from the data.  The dashed line  is the  result  obtained by subtracting  out
the material introduced by the rapping  puffs.   Theoretically there  should
be nothing larger than 5 to 10 microns  present in the outlet of that pre-
cipitator.
      Data collection, at least  in our  experience, has been greatly hampered
by problems with substrates.  We have been  using the  Andersen impactor for
the most part for our outlet data, although we've made several attempts to
use the University of Washington impactor with greased substrates.   We have
been frustrated in trying to get grease that doesn't  lose weight.   We  thought
we had found that glass fibers that worked  satisfactorily until we ran some
tests at several plants at which we got the sort of behavior shown in  Figure
9.  This figure shows the results  of pulling filtered,  particulate-free flue
gas through an impactor loaded with glass fiber substrates  (specifically
Andersen impactors at a flow rate of 0.5 ACFM).  Flue gas temperature is
shown along the ordinate, and the weight gain  per stage  of the impactor
as a result of a gas phase component reacting  with the  substrate material
along the abscissa.  Pre-1974, we had performed similar  experiments and
gotten results at high temperatures, of a small fraction of a milligram
                                   94

-------
  0.01
         OUTLET  I
                                                                   80
                                                                 10.0
                           PARTICLE DIAMETER,
         OUTLET  2
      0.2
U.UI
0.05
o.i
0.2
*
f
- 0.5
F
> 1
V
: 2
»
i 5
10
on
1 1 1 1 1 i 1 1
1
— JL
T
I {


—
i I I i I I 1 1
i I i i i i i i
T I M '
if
1
• } :
«
t
* •*• <
fl
•
V
•••
•VH
•^
k
>
muni
•
^^M
^^™
WITHOUT RAPPING "~
—
1 1 ill!
yy.yy
99.9
99.8

99
98
95
90
8O
   1.0
PARTICLE  DIAMETER,um
10.0
                                                                        o
                                                                        UJ
                                                                        o
                                                                         UJ
                                                                         a
                                                                         o
                                                                         o
Figure 8.  Fractional efficiencies for rapping and non-rapping tests.
                                 95

-------
CD
H-
CD
LU
               -  COAL  FIRED POWER BOILERS
2  —
1  T-
0
                                       400               500
                                    GAS TEMPERATURE,  °F
   Figure 9.  Anomalous weight increases  of Andersen glass fiber impaction substrates at different
             flue gas  temperatures.

-------
weight increase from such a reaction.  The filter manufacturers about June
of  '74 changed the process by which they make glass fiber filters.  Where-
upon we immediately started having difficulties.  This same difficulty that
we  are having with the Andersen substrates in this case would be present
in  all glass fiber backup filters and Method 5 filters.  Backup filters will
tend to have problems in impactors anyway from material which bounces, re-
sulting in oversize material arriving at the backup filter and contaminating
it.  In one case  in a study done by Lawrence Livermore Labs, 98 percent
of  the weight on  the backup filter was the result of oversize material and
only 2 percent of the weight was the result of material that should have
been there.  This was with an impactor run with Nucle-pure polycarbonate
material  as impactor substrates.
ENSOR:  Was that  a stack sample?
MCCAIN:   Yes, a stack sample at a coal-fired boiler.  The backup filter
was meaningless,  in that case, for efficiency calculations.  It had no size
which you could really  assign to it.   If you get substrate reactions going
with glass  fiber  backup  filters as was shown in Figure 9, then you don't know
what's happening  either.  Six milligrams is way in excess of the amount of
material  you might deposit on many stages and completely masks the catch.
This effect is, we believe, a result of an SOp  reaction with the substrate
material.  Once we found this to be  taking place, we set off on a search
for a substrate material—glass fiber—that would not exhibit  such an effect.
Figure shows the  results for  four of the materials that we tried.  This
is  not an exhaustive  list; we've tried a lot more.  Teflon did real well;
unfortunately, we found  that  it doesn't do very well as a substrate.
Quartz  (Pallflex  quartz) fibers do very well, but they are so  fragile  that
they are  almost impossible to use as a substrate or a  filter media for
gravimetric work. Gelman has made a limited batch of  an improved quartz
material, which has about the same tensile strength as the standard  glass
fiber material.   We have tried samples of  that, and it is quite good for  this
application but it is not yet a commercially available product.   A sample
small batch was made  primarily  for pilot studies for  EPA, and  Gelman retained
only a limited quantity for  their own  purposes.  They  may market  it  in the
future.   If they  do,  it will  probably  be a  satisfactory  backup material  and

                                  97

-------
     1,5
     1,0
 CD
 "
     300

    TEMPERATURE, °F
400
500
Figure 10.  Anomalous weight gains  of various 47 mm dia. glass fiber filters at
          different temperatures.  (60 minute samples at flowrates of 0.25 ACFM),
                                 98

-------
substrate material.  It will have to be tried in the lab as a substrate
material to-demonstrate that it does effectively reduce bounce problems.
The materials shown in Figure 10 are all relatively unsatisfactory because
of the reaction affect.  The data were obtained with 47 -mm filters,
somewhat smaller than the size used in most of the impactors that use
glass fiber filters, although the Sierra does use 47 -mm.  The weight
changes are temperature sensitive for all of the materials; higher
temperatures lead to greater uptake.
      We have tried several experiments to alleviate this problem.  One
of them is to pull filtered flue gas through the material before we use it
to precondition it.  Figure 11 shows weight increases of Andersen substrates,
once again resulting from pulling filtered flue gas through untreated
material, resulting in weight increases of about 1.5 to 2 mg per stage.
After running filtered flue gas through the material for a period of
about 24 hours, then desiccating it, weighing it, using it as a
substrate, the reaction affect was  reduced to approximately one quarter of
a milligram per stage, which is a much more acceptable level, although it
still introduces some  uncertainty in the data.  That leads us back to the,
curve on the outlet data at the power plant previously shown.   (See Figure
3).
      The small error  bars  shown on Figure 3 are that part of the uncertainty
or scatter in the  data at this particular plant, which can be ascribed to
the sulfate uptake  (S02 reaction) of the glass  fiber filter media sub-
strates  used in these  runs.  The large error bars are the la limits of the
data  from the average  of the three  runs; the small  ones  are that  part of  it
which would result  from the S02 uptake.  One can see that the problem that
we have  in this particular  instance with the S02 reaction is very small with
respect  to the effect  that  it would have on the data.  Figure 12  shows the
correlation that was obtained by comparing gravimetric weight increases
of substrates that  had nothing but  filtered flue gas pulled over  them  and
sulfate  determinations made by wet  chemical techniques on those same sub-
strates.  Sulfate  represents essentially all of the excess weight pickup
that those substrates  had.  This particular data was obtained at temperatures
                                     99

-------

-------
       BULL RUN STEAM PLANT
                     I
                                              SAN  JUAN STEAM PLANT
                               3        4        5

                             WEIGHT  GAIN, MG,
Figure  12.  Comparison of weight of sulfate on blank Andersen Impactor
           substrates and observed anomalous weight increases.
7
                            101

-------
of about 625° F at a plant  burning  a  low-sulfur Western  coal.   It's hard to
believe that sulfuric acid  condensation would occur at those temperatures;
on the other hand, S(L reactions  with basic  sites  on  the filter materials
could very well account for this.   This same phenomenon  has been  noted
several years ago to take place with  filter  media  used in ambient air
sampling.  (SCL reacting with  basic sites  on the filter  media.  )   As I
say, we've tried greases and other  impactors in an attempt to  get data
without having problems with bounce.   Table  1 shows results from  blank
runs made with two of the greases that we  have tried—we are in the process
of trying a lot more—this  gives  typical  results showing problems we've
had with a couple.  We ran  filtered coal-fired boiler flue gases, one run
at 300° F and one at 260° F through the impactor.  With  Dow Mollikote ILL
compound, the average loss  per stage  was  0.5 mg with  a backup  filter
gain of 1.86 mg.  This filter material is  material that  we would
expect on the basis of S02  uptake along to gain about 0.2 mg.   Presum-
ably some of this increase  resulted from  material  that had somehow either
been eroded, blow off, or outgasped from  the grease and  wound  up  being
caught on the filter.
      With Dow silicon high- vacuum grease,  the losses were somewhat erratic.
This is perhaps because the same  amount of grease  wasn't used  on  every  stage.
The average loss per stage  with  it  was about 2.5 mg.  The backup  filter
gained almost 3 mg when SO^ was  expected  to  add about 0.2 mg.   The greases,
at least the ones we've tried, when applied  in quantities provide enough
grease to wick up into the  particulate layers, just do not stay around.
1.  Forrest, J.  and Newman, L., Sampling and Analysis of Atmospheric Sulfur
    Compounds for Isotope Ratio Studies, Atmospheric Environment, 7,
    pp.  561-573, 1973.                          	
                                  102

-------
                                TABLE I
                            DOW
                         MOLYKOTE
                       111 COMPOUND
SUBSTRATE
Temperature
Flowrate
Sample Duration

STAGE
  1
  2
  3
  4
  5
  6
  7
Filter

AVERAGE  LOSS
PER  STAGE
NET  TOTAL  LOSS
FILTERS - UNPRECONDITIONED GELMAN TYPE A (OLD  TYPE)
EXPECTED FILTER WEIGHT CHANGE - = 0.2 mg
   DOW SILICONE
HIGH VACUUM GREASE
300°F
0.46 ACFM
60 Min.
WEIGHT CHANGE (mg)
-1.1
-0.74
-0.34
-0.36
-0.46
-0.32
-0.46
+1.86
0.54 mg
1.92 mg
280° F
0.45 ACFM
60 Min

-4.06
-1.74
-3.60
-3.76
-1.32
-1.90
-0.64
+2.68
2.43 mg
14.34 mg
                                   103

-------
DISCUSSION

ENSOR:  Are those prebaked?
MCCAIN:  These were prebaked;  the material  was  put  on  the  plates,  baked
in an oven for a period of several hours  at 400°  F  (a  temperature  higher
than they were actually operated at)  and  subsequently  weighed  and  used.
We have tried in the lab about 20-odd materials,  about six of  them looked
good, two of the best-looking  ones were these  two.
ENSOR:  How do you apply that  grease?
MCCAIN:  In this case, it was  just put on the  fingertip and smeared around
right out of the tube.
	:  The important point is that your  application is very,  very important
and hard to control.
MCCAIN:  We have applied it at other  times in  the field in a suspension  in
benzene:  Put on with a eyedropper and allowed to flow and then baked out
at 400°.  Again, we got negative weights  at the end of actual  runs when
there was visible particulate  on the  stages.  You could see that you ought
to have a lot of weight gain but the  results were negative weights.  We  tried
several applications; these two were  just smeared on with  a fingertip.  They
weren't intended to be used as particulate runs,  but just  to see whether
or not we were going to have'a problem with losses, and we did.
	:  Do you think that it might.be a function  of time at all?  In
other words, if you baked them longer, you might see this.
MCCAIN:  We tried this in the  lab. We baked out, weighed, rebaked, reweighed,
went through this treatment about four times.   Several, including  those  in
Table 1, tended to come to a stable weight in the lab.  Out of the 20 we
tried, about six tended to come to a  stable weight.  Those are the six that we
were wanting to try in the field.  As I say, these were two of those six,
and their behavior with flue gases is somewhat different.
	:  You say these were baked more than 4 hours?
MCCAIN:  Yes.  At 400° F, a higher temperature than the application.
                                 104

-------
OLIN:  This loss must have been a function of how much you put on;  if
you put less on, you get less off.
MCCAIN:  Yes, yes, it does appear to be a function of how much you  put
on.  Roughly 30  mg were applied  to  the  stages in each of these
cases.  This is the University of Washington impactor; that's about what
Mike recommends.  At least, that's what  the manual says.
OLIN:  You put about the same on for each of the different greases?
MCCAIN:  Yes.
OLIN:  Same total weight?
MCCAIN:  Same total weight.  This (Molykote III) held up a good deal better
than this  (Dow high vacuum grease).  Neither of them was particularly
satisfactory.  If you are collecting a  few tenths of a milligram or even in
this case a few milligrams of material, you would have difficulties in
interpreting the results.  ;
HARRIS:  We've tried some GC types of greases that give conflicting data.
Some of them that Joe's  run that we had no problem with, he's had some
losses with.  Intuitively, it would seem to be a good candidate to use
because the GC materials are usually defined at a certain temperature
rating, and that temperature rating is  something like the point at which one-
tenth of one percent of  their mass, so  that they don't mess up the GC.  So
that would be intuitively a good source for it, but so far we haven't got
data that tend to agree  on it.  We  have a few runs on polyethylene glycol
under laboratory conditions that indicate it stabilizes very quickly and
you have no losses at 400° F.
	:  Joe, what  is the percent grease in benzene in this first one here?
MCCAIN:  Probably on the order of 5 to  10 percent weight.  No more than 10
percent, and it was not  a solution—it  was a suspension., an emulsion
obtained with ultrasonic agitation.
CALVERT:  Joe, when you  show that curve on the continuous monitor, what
were you using to get that data?
                                 '  105

-------
MCCAIN:  A Royco particle counter with an extraction  and conditioning
system.  It did not provide data on absolute concentration.   It was
qualitative; you could look at concentration variations  with  time.   We  have
done a couple of tests; one specifically a semicontrolled experiment in
which we ran that system and obtained what I think is pretty  good quanti-
tative agreement between the impactors and particle counter.   But it was  in
a situation where we could predict what our transport losses  were in the
probe and predict the transfer characteristics of the conditioning system.
In this particular instance, we weren't able to do that.  The data are  not
adequate.
	:  What was the change in the filter in '74 and was it  only with
Gelman?
MCCAIN:  I don't know whether it was.  The Gel man material, Type  A,
specifically, pre-Oune '74, was made in a sulfuric acid solution.  That
was changed to a hydrochloric wash, and  it  all went to  pot at that point.
point.  I don't know if all of the manufacturers changed at  that  time or if some
were using hydrochloric prior to that.  I do know that we've  been on the
telephone off and on for some period of time now trying to get information
from every manufacturer of glass fiber filters we could find, and none of
them are using sulfuric acid currently.
	:  One other point about your SOp data is that at  that temperature,
your sulfuric would be mainly in the form of SO.,, and it could go through
the same type of reaction that the S02 does.  So you can't say that all that
is due to S0£ conversion; it could well be  collection of SO.,.
HARRIS:  One thing you want to look at is the point that the highest reaction
conversion that he has was at a very, very  low SOp concentration, and so
the SO., would be very little.
MCCAIN;  In that case the S03 was indetectable, and we  had 50 ppm SOp.  We
didn't expect to have any problems, we thought the S0»  content was so  low
that we wouldn't expect any problems at all, but it was one of the worst
cases we have ever seen.
	:   Why do you suppose?  Have you tried exposing the filter  paper  to
pure S02?
                                     106

-------
MCCAIN:  That is one of the next things to be done in the laboratory.   This
has all been done as preliminary work to a field test in each case to
find out whether or not we were going to be able to obtain data and what to
do to obtain data in that case.  We're setting up a conditioning chamber
in the laboratory using S02.
SPARKS:  I think that,  for completeness,  it should be pointed out that some
of our experience, some GCA ran and some experiments Bruce [Harris] ran
in-house, there were problems with filter weight loss on blank runs.
HARRIS:  Structurally weaker.
SPARKS:  Structurally weaker, so there has been an increase in structural
integrity so we don't lose weight anymore.  But we gain weight.
MCCAIN:  Those losses were relatively small compared to these gains.
	:  They were of the same order as GCA's anyway, the same order as your
reduced weight.
MCCAIN:  We never found any at all in excess of a tenth or two-tenths  of
a milligram.
HARRIS:  I think that the  thing that you might be able to generalize is
that the weight loss, Type A thing, seems to be a function of handling.  It
was something that you  had a little bit more control over.  If you had some-
body in your field crew that was the meticulous type, you could control it
a little bit more; somehow we can't control all those little buggers running
around in the stack.  It's just another indication that we had a problem
with the weight loss, which was one reason we asked them  (filter manufacturers)
to try and do something about the structural, state of these, so they did
it for us.  They fixed  us  real good.  But, you know, a lot of what we're
discussing here is:"Has anybody got any great ideas of how to cure all these
things ?"
                                    107

-------
             OUTLINE OF FIELD EXPERIENCE  WITH  CASCADE  IMPACTORS
             Dr.  Seymour Calvert,  Air Pollution  Technology,  Inc.


HARRIS:  We'll now head out to the West Coast  to get our next  two
speakers to tell  us what the problems are out  there.   Fortunately,  they
have to be the same problems that  we have on the East  Coast.   First,
Seymour Calvert will tell us all the woes that you  get into  at the  tail
end of scrubbers.  Among other things, they're wet.
CALVERT;  I hope that there will be plenty of  time  for questions, so
what I am going to do is essentially outline the area  and talk about
troubles.
      Just to sketch in the backgrpund from which I  am going to  speak,
the experience we have had has mainly been concerned with scrubbers and  fine
particles.  This means that we have gotten into  inlet  conditions that
range from hot and dry gas to saturated and wet  gas, and when  I  say wet
here, what I mean is gas containing entrained  drops  of liquid  mist.  The
inlet gas conditions are variable; it depends  on whether the gas comes
right from the source and goes in  the scrubber,  or whether it  goes  through
some type of quencher or precleaner first.  The  outlets are generally
saturated and generally wet, which implies that  the entrainment  separator
isn't working very well.  Sometimes we have to sample  after a reheater
so that, even though what came out of the scrubber was saturated and  wet,
what conies out of the reheater and what we sample is below saturation.
The types of cascade impactors we've used are  the Andersen, the  Brink,
and the APT--that's Air Pollution  Technology—built impactor and the
UW (Pilat) Mark III impactor.
      Generally, the systems and procedures we've used are in accordance
with the EPA draft guidelines for cascade impactor use.  The sampling
trains are comparable to the EPA trains, give or take a few parts.   We've
used some special trains and procedures that I will allude to a little
bit later on.  Sometimes we've had to combine cascade impactor measurements
with EPA Method 5 measurements.  This becomes  important where the  purpose
of the task involves determining whether the source in in compliance with
                                    108

-------
emission standards or not, and generally we have to take simultaneous
samples at various points in the system.
     The major points to consider really have been touched on by other
people and definitely are covered in the draft guidelines that Bruce
Harris has sent out,  so I will  skip  very quickly through those.  They
include choice of the impactor components, substrate coatings, filters,
methods of use, sampling points, procedures, data acquisition methods,
measuring gas velocities, temperatures, pressures, what kind of substrates
and sample treatment, what kind of weighing? I guess we're going to hear
more about that.  How to handle impinger and drier (gas drier) catches,
what kind of data and forms to use, what kind of field laboratory and
other work facilities do you need, where to do the data reduction, what
to do in the field, what to do at home, and what kind of process data
should you acquire.  These things are more or less general and I'm not
going to spend much time on them.
     As to problems, have we got problems!  The first one relates to
sampling position, and  I call this one, in my notes, "Picking the best of
poor alternative sampling positions."  This has been touched on before.
You never get a good sampling position.  There are problems with particle
bias, with getting a good gas flow determination, and there are variations
in gas velocity.  There are problems because of the probe size and shape,
which is required in order to cope with the geometry of the situation.
There are problems with the cascade  impactor orientation, which  is
necessary because of where the duct  is, where the sampling points are,
where the platform or nonplatform  is located.
Coping with Process Variability
     "How simultaneous  can you get or do you have to get?"  This has been
touched on.  Joe McCain talked about the problems of matching  inlet and
outlet sampling times.  Generally, we're concerned with the determination
of the efficiency for highly efficient devices which means that  the outlet
loading is going to be  low, so that you have to sample  for a  lot longer
time on the outlet or at a lot higher sampling rate, or in general you
need a bigger sample volume on the outlet than on the inlet.   Another
                                  109

-------
problem in process variability is:  "Can you replicate operating
conditions?"  Again, Joe touched on that with regards  to such  things
as rapping in the precipitator.  There is also the problem of  process
variability where the source has its ups and downs.
Optimizaing the Test Cost
      Here, the question is:  How much sampling apparatus do you need to
get a fast turn around time?  What crew size do you need?  The problem is
that you have to contend with plant shutdowns.  The plant will shut down.
I don't think we've ever been on one that there wasn't a breakdown.  I
used to be cheap (frugal).  We tried to do tests with  maybe two people,
thinking:  "Okay, we'll spend longer out there, but this will  really  be
economical."  You're asking for it.  If you do that, you're sure to
run into a plant shutdown, you'll have to bring your people back all  the
way back from Provo, Utah, or Casper, Wyoming, or someplace.  Sure as heck
the next time you go back there, there will be a blizzard.  You have  to
face the cost of all that thermal underwear.  So, we came to the conclusion
that it was cheaper to use a bigger crew, buy more equipment,  get in  and
out fast and keep down the possibility of being shutdown.  Weather can also
do it to you.  So there are acts of man, and acts of God.
Transportation for Apparatus
      When you have to go a long way from home, there  is a question of
how do you get the apparatus there:  Should you drive, truck it, fly it,  '
do without it?  Then you run into an optimization problem.  There is one
of:  Is it better to have a van for the purpose of hauling equipment out,
or do you want to have a mobile laboratory where you can actually work
in the van?  Our decision has been to just use the van for transportation
and use laboratory facilities, or get laboratory facilities onsite, some
way.
Reliability of Apparatus
      How much redundacy is good?  After several years of hanging by
our fingernails with meager data because meters failed, we now use
three flow meters in series.  Maybe its enough.  We use a rotameter,
we use a dry gas meter, and we use an orfice meter in series.
                                   110

-------
We check them against one another.  Usually that's alright.  Two sets of
temperature-measuring instruments, extra pressure-measuring instruments,
manometers, magnet heater gauges, etc.
Sampling Pump Leaks
      For everybody else, sampling pumps don't leak.  Ours always leak.
They may start out not leaking, but they sure wind up leaking, and
again the redundant meters on both sides of the sampling pump are a
big help in detecting that kind of problem.
Heater System Burn-outs
      We generally have  to heat the impactor or the probe of the sampling
line between impactor and filter.  I'll get into this more later.
We've had lots of trouble with the heaters burning out.  Lately we've
been using blanket type  heaters where  the heating wires are impregnated
in silicon rubber sheets, and these have been pretty good.  We're out of
most of our troubles there,  but they  do burn out; variacs burn out and
so on.
Corrosion
      We have  run into most  of our corrosion problems when sampling gases
which contain  fluorides.  If you  want to get that sinking  feeling, take
apart your expensive UW  impactor  and  see the pits drilled  into the
substrates and through the  foils  and  then to the impaction plates.  It
might make Mike  Pilat happy, but  definitely not me.   It even  happens to our
APT  impactors, and so we have two impactors that we have machined out of
Teflon  for just  such festive occasions.   Incinerator  gases are where we
get  into trouble. People are burning PVC wire, or  occasionally  they run
into a  little  Teflon wire which  gets  hydrofluoric acid in  addition to
hydrochloric acid into  the  gas,  and  these give you  trouble.   Stainless
steel is just  no good  for those.
                                       Ill

-------
Low Pressure Ducts
      We've sampled downstream from a scrubber with  an  induced  draft1  fan
at -110 in. water column,  and we lost one  pitot  tube and  they wouldn't
tell me what else had been sucked into that duct.  It was terrible,  and
then there's the question  of whether the sampling  pump  can cope with it
and whether your apparatus can cope with it.   Is everything build  so it
will work properly with that kind of high  negative pressure, vacuum?
Large Ducts
      Here we're generally talking about power plant sampling situations
where the duct might be 15 or 20 or 30 ft  across.  I know that  the
Southern Research people have run into this very often.   There  are problems
with just handling the equipment; probes that extend across that length or
up and down that length are hard to handle.  Just  lifting the apparatus
and then being able to determine the position of impactor or the pitot
tube or whatever it is dangling at the end of a  16 ft long pipe is
difficult.  If the gas velocity is fairly  high and there  is some
flexibility to the apparatus, the question is, Where is it; where  is the
end of it?
Crew Pollution
      We were prepared for ammonia because we sampled urea filling towers,
and we were prepared for SOp and, son-of-a-gun,  when we went up to a large
foundry, the guys came down wi;th carbon monoxide,  i.e., with headaches.
They ran into very high CO exposure, so we had to  invest in masks  for that
too.  High temperatures can be a problem;  obviously  particles  also—the
respirable range and even the gritty type  particles.  Entrainment can be  a
problem.  If you have ever had the experience of sampling downstream of
a scrubber on a coal drier, it's a big thrill to be  coated with the black
mud that comes spewing out of the exhaust.  Noise  can be a problem, too.
Entrainment
      I have talked some about entrainment that  can  get onto the people.
The entrainment inside the duct is also a  big problem.   We have developed
a precutter, which is essentially a one-plate one-stage impactor that has
enough volume to handle most situations.   However, we've run into
situations where it was just swamped.  There was so much water coming
                                     112

-------
through that we couldn't handle it with our precutter and we couldn't
handle it with a little cyclone.  We invented what is known as the
"Beer can entrainment separator."  I can discuss it with you over a cold
Coors or something.  It was fashioned on a western sampling trip out of a
you-know-what and essentially it amounts to using a big baffle.  Here we
get into compromises which I'll talk about later.  The question is:  can
you live with that kind of sample bias?
Particle Size Changes
    Actually, that's what I'm going to talk about mostly.  The particle
size is not  constant.  We have  a friend who is a world traveler and he can
gain or lose 50 Ibs.  He would  go on a big trip around the world and
gain 50 Ibs  and go on a crash diet and lose it and he had suits for various
stage weights.  You never knew  exactly what size the guy was until you saw
him; and particles are like that.  Water condenses on them or evaporates
dependent  on temperature, pressure, and partial pressure on whatever this
volatile material  is.
Correlation  of  Cascade Impactors with EPA Method 5 Samples
    Now we have to be concerned with sample bias with losses in the
impactor and what  we  call  "mystery particles" which  are  caught in  the
impinger.  We've  run  into  this  case where something  winds  up in the
impinger.  We've  had  as high  as 30 or 40 percent of  the  total  particulate
found  in the impinger, an  ordinary Greenburg-Smith type  impinger  following
an UW  impactor  with the final  filter.   If you have to work with plant
people or  control  agency  people who  are not as  sophisticated  as us guys,
they will  say;" Well,  it's  the  fine  particles that get  through the filter."
You try to tell them  that the filter stops everything and  they don't
believe you. They make remarks about those long-haired research  types  and
so forth,  but this has so far proven  to be attributable to previously
unknown volatile  materials  that were  present  in the  samples.   So  far we
have been  able  to rationalize what  has  happened,  but it has  taken work.
                                     113

-------
Electrostatic Charge Effects
     (1)  Joe McCain mentioned that you can  get  electrocuted.
     (2)  The impactor measurements are affected.   We've  run  impactors with
and without radioactive charge neutralizers  in a precutter type  assembly.  A
preneutralizer at the impactor has an effect on  the particle  size  distri-
bution measurements.
Cleaning and Checking Cascade Impactor Jets
     How do you know that they are clean? How do you  clean them?   We're
going to have to hurry here because of the time.
Drying Samples of Volatile Material or Volatile  Samples
    Again, its a discouraging experience to  find out that the guys were
out in the field three days sampling a urea  prilling tower scrubber,  they
put the samples in the oven,  went through the usual procedure, and all
the sample evaporated.  Then  you find out that urea is volatile  and
so is ammonium nitrate and other things that might be  in  there.   It's all
very interesting and educational, but it sure isn't profitable.
Coping with Particle Size Changes
    I'll just run through this very quickly.  One type of problem is  due
to the fact that you cannot take your choice of  sampling  position. The
condition of the gases is variable.  You have hot gas  that may go through
a quencher before it gets into the scrubber.  It may then go  through  the
scrubber, and it is wet when it comes out.  If it does not go through
a reheater, it will go to the fan and come out wet. There may be a reheater
so it will be dry when it goes into the fan.
    As Les Sparks mentioned in the beginning, the first big question  is:
"Why do you need the data; what are you trying to do?"  If the purpose is
to develop or validate a model for the scrubber, then  you want to know
what the scrubber saw.  What size were the particles at the end  of the
inlet of the scrubber and at the outlet of the scrubber?  If you don't
have a sampling poing at these locations, and you have to get a sample
at the point where the gas is dry, the question  is how much did the
particles grow?  Did they grow because they nucleated condensation and
they all grew to about the same size as in a cloud chamber (or condensation
                                     114

-------
nuclei counter)?  Or did they grow because they go into solution and
lower the partial pressure of the water, so its an equilibrium type
relationship?
    We have used two approaches.  One has been to run the impactors
heated so that the particles are dried out at the inlet and at the
outlet of the scrubber.  Alternatively, we run the impactors unheated
so that we can match inlet and outlet conditions.  There was one case
where we couldn't get the wet particle size because there wasn't a way
to sample it.  We took  samples from the inlet through a humidifying flask,
which was a very low efficiency impinger having a large diameter tube to
bubble the gas through; took the sample through another flask simply to
insure that it reached  equilibrium; and then we split the sample and put
it into one heated  impactor and one unheated impactor.  There was a
difference between  the  wet particles;  i.e., the particles in the unheated
impactor  as measured with  the  unheated impactor and in the particles
measured  by the  heated  impactor.
     I  think  I'd  better  just stop here.   I'll be pleased to answer
questions whenever  there is time.
                                      115

-------
DISCUSSION
        What is the difference in  that  size  distribution  in  the mean
size as you go from wet to dry or vice  versa?
CALVERT:  This run was about 1.5 times, another  one  about  1.6.   There were
some other runs.  Here's one at about 50 percent.  You're  going  from about
1.3 to about 2 microns mass median diameter between  the  two  runs.
	:  What kind of dust is that?
CALVERT:  This is borax.  We've taken similar  data on  a  composite  dust which
was diatomaceous earth, which contained sodium chloride.   We found consider-
able particle growth with that.  Apparently, there was more  of a nucleation
effect with that one.  In other words,  the particles tended  to be  fairly
uniform in size, which would indicate a nucleation of  condensation effect
rather than just an equilibrium growth.
SMITH;  What kind of source did you do  your charge neutralizing test on?
How large was the effect?                                         x
CALVERT;  That was a titantian dioxide  test dust,  and  it was on a  charged
drop type scrubber, and I can't remember how big the effect  was.  We've
just gotten a draft report on it, and i'll check it  for  you  later.  There
was a definite effect on running with and without a  charge neutralizer
just ahead of the cascade impactor.  Also, while I mention that, we also
used a charge neutralizer on the dust feeder and found a definite  effect
on the efficiency of the scrubber as measured  with that test dust, due to
electrostatic charger on the test aerosols.
BOLL;  Your data on the borax implie to me a bimodal distribution, and
yet you drew a straight line through it.  I'd  be curious as  to why you
did that.
CALVERT;  We just characterized the dust with  that simple distribution.
I would say you are right and it is bimodal.  It is  a combination of a
condensation fume and dust from a rotary kiln.  There are definitely two
sources;  large dust and small dust.  I think  that we were not  that certain
of the lower cut points at that time, and I think this was prior  to the
time when we had calibrated our impactor.  By  the way, we too have

                                    116

-------
calibrated our impactors, and we used polystyrene latex and a light
scattering type counter.  We have used that technique, and if we  were
doing it again I think that we might be more fastidious.
BOLL:  The reason I asked is that this type of curve seems to be  a rather
typical occurrence and the straight line seems to be a rather typical
reaction.  Of course there is a big difference in what your collector  is
going to do depending upon what the truth is.  Unless you have some  reason
to believe the cut sized wrong and the data are wrong and the straight
line is right, for heaven sake we need to know.
CALVERT:  Yes.  We draw the straight lines; however, where we draw them
depends upon what we want to do.  If we were interested in collection
efficiency on the small sizes, we would bias the line toward that end  of
the  curve.  Another problem in here is that (I guess we didn't use a
precutter on this one)  if we use a precutter you have the influence  of
that on the upper stages of the impactor.  It's collecting the bigger
particles and that's a  superimposed effect.
BYERS:  On the determination of the efficiency of your control equipment,
obviously you're looking at particle size distributions and their impact
on the determination of the efficiency.  What would you consider a
significant change in particle size distribution in terms of the mean
diameter?  What would have an impact on the accuracy  of your efficiency
determination?  In other words, what is the kind of accuracy that you're
looking for on efficiency determinations and what would be a significant
size distribution change which would impact on that?
CALVERT:  If you're talking about overall efficiency,  I think Les Sparks
gave the answer that 10 percent change  in mass median diameter can have
a doubling effect on the pressure drop  requirement.   We generally are
not  looking at overall  efficiency nor overall  penetration  over the whole
size distribution.  We  are looking at penetration as  a function  of particle
size,  so that we've been more concerned with  "what  is the  effect?" than
"what  is the error?" and I am afraid that  I  cannot  give you  a nice clean
answer to your question.
                                       117

-------
                FIELD EXPERIENCE WITH CASCADE IMPACTORS:
                     QUALITY CONTROL OF TEST RESULTS
                                                         *
               David S.  Ensor, Meteorology Research, Inc.

HARRIS:  Now I'll ask Dave Ensor to come up and describe  some of his
feats of magic and give  us some of the results of the work.   I guess it's
basically some of the work at Nucla.  Some is going to be on some substrate
work on that thesis.  He is one of the boys who has managed  to go out to
this place in Colorado about three miles past the edge of the earth, so
he's a world traveler.
ENSOR;  Thank you, Bruce.  As Bruce mentioned, we were doing some work with
the Electric Power Research Institute at Nucla, Colorado.   The quality of
particle size distribution data obtained with an in-stack cascade impactor
depends on the magnitude of weighing errors and weight changes of the
collection substrate due to handling and chemical reactions  with reactive
gases.  Procedures using both blank impactor tests and control collection
discs were developed.  These procedures were used during  a recent evaluation
of a fabric filter on a utility boiler.
INTRODUCTION
      Background
      Cascade impactors  are becoming widely used for the  determination of
particle size distributions by mass for control device evaluations.  During
the last year, it has become apparent that attention must be paid to
chemical and physical changes of the collection surfaces.  Fegley et al  (1975)
examined the sensitivity of the test results to the weighing errors and
the impactor construction tolerances.  It has become mandatory to check
impactor results for weighing errors and undesired weight gains or  losses.
*
 Coauthored by Robert C. Carr, Electric Power Research  Institute.
                                    118

-------
    Scope of Paper
    The scope of this paper includes:
      •  description of the MRI cascade impactor and substrates,
         report of the quality control efforts from a recent field  test,
         indication of possible areas of improvement in testing.
MRI IMPACTOR SYSTEM
    Introduction
    In a particle size distribution measurement program, there are  many
important aspects of instrumentation:
      '  number and type of impactors,
      *  particle collection substrates,
      *  type of analytical balance,
      '  support equipment such as cleaning facilities and dessication
         chambers.
    Description of Impactor
    The MRI cascade impactor is an annular jet-collector type, similar to
that reported by Cohen and Montan (1967).  A cut-away drawing of  the
instrument is shown in Figure 1.  The body of the device consists of  quick
connect rings supporting jet plates, collection discs, and a built-in
holder.  The design permits flexibility in application to various sampling
situations.  For example, the same impactor is used with three special jet
stages to sample particulate matter with a light aircraft.
    The impactor has been used with sampling rates from 0.15 to 2 cfm
depending on the jet plates selected.
    One very important aspect of the impactor is the collection disc. The
disc is a lightweight metal stamping (730 mg) of 2-mil-thick 316 stainless
steel.  The disc is used only once, permitting a permanent record of the
test.  Also, this approach lends itself to the preparation of substrates
for chemical analysis (Ensor elb al., 1975).
    The surface of the collection disc is prepared with a solution  of high
vacuum grease in toluene.  The solution is painted onto the disc.  It has been
found that the thickness of the coating is important in the performance  of
the impactor.  A common error is to apply too much grease.  After air drying,
the discs are heated at 400° F for 4 hours to remove volatiles.  The collection
discs are handled with clean forceps by the edge to prevent conformation
and weight changes.
                                    119

-------
           Nozzle
          Jet  Plate
          Collection
             Disc
                                          1st Stage
                                           "O" Ring
                                            Filter
Figure 1.  Assembly drawing of Model 1502 Inertial Cascade
           Impactor.
                          120

-------
     The filter is held by Teflon contact surfaces and is backed by a porous
metal plate.  Teflon washers are used to prevent loss of the filter to the
contact surfaces ,  and a tared aluminum  foil  disk  is  used to weigh the filter
and Teflon washers.
     Weighing
     The weighing of the collection discs is a critical part of the whole
test.  The collection disc is designed to fit the weighing chamber of a
Cahn 4100 Electrobalance.  The Cahn balance offers the advantage of being
portable (it is normally transported in a carrying case as hand luggage),
and is relatively noncritical as to the stability  of the support table.  It
is desirable to do the  weighings at the motel to avoid disruptions due
to low frequency plant  vibration.  Often a separate room is rented solely
for the weighing.  The  discs and filters are desiccated for 24 hours before
weighing to stabilize the water content.
     One of the elements of success with any balance is to thoroughly know
the balance and to develop optimum weighing procedures.  A weighing by sub-
stitution method was developed.
TEST PROGRAM
     Introduction
     Impactor data quality control procedures were used in an evaluation of
a fabric filter on a utility boiler in a project for the Electric Power
Research Institute.   The requirements for the quality control checks were
motivated by:
        •  the reports of Smith et a!  (1975) of reactions of glass
          fiber filter  mats with stack gas
        •  the stainless steel collection disc with a grease coating
          had not been  thoroughly field tested
        •  the low outlet concentrations required  long test times and
          light stage weight  increases
     Test Plan
     All of the quality control tests were conducted with impactors  used
at the outlet of the control device.  Two types of tests were used.
                                    121

-------
     Controls — Collection discs were prepared normally, transported to
 to the test site, but not mounted into an impactor.  Since the outlet im-
 pactor was operated without the last jet stage, the collection disc nor-
 mally used at the last stage became the control with little disruption of
 the test.  The control collection disc was a good check of the performance
 in weighing of the samples.  The tests of weighing repeatability must be
 conducted with the collection discs under field conditions during the
 normal pattern of work.  Thus, problems with static charges, balance
 adjustments, and handling can be identified.
     Blank Tests — Blank tests were impactors prepared in the normal
 manner which sampled only filtered stack gas.  These runs should identify
 problems from chemical reactions of the substrate with stack gas, loss of
 substrate from vaporization,or abrasion,-and contamination of the;.substrates
 from leaks, or during assembly and disassembly.  The blank runs also had
 at least one control disc.
     It was recognized that the physical and chemical nature of the greases
 are important in the performance of particle collection.  A large number
 of greases and adhesive materials were screened in a laboratory study last
 spring; three materials showed promise:
     1. Apiezon-L high vacuum grease:   hydrocarbon with an average
        molecular weight of 1300,
     2. Dow Corning 111 Compound Silicone Lubricant,
     3. Hooker Fluorolube GR660 Fluorocarbon Lubricant.
 RESULTS
     Introduction
     The actual test program was conducted in two two-week phases to allow
 the'evaluation of the control device for various conditions of coal and
 to adjust the procedures between tests.  All weighings were conducted in
 the field.
     Controls
     The Phase I results for the controls are shown in Figure 2.  The
 standard deviation of the weighings are within the precision claimed by
 the manufacturer.  The extreme deviations were isolated to only a few
days of the test program.  Since all the weighings were conducted by
a single individual, it was suspected that perhaps the balance calibration

                                    122

-------
                                                     PHASE  1
PO

CO
              
-------
procedures required modification, that static charge effects were
severe on some days, or the weighting tasks of the program should be modified
to reduce operator fatigue.  The results for the Phase II weighings are
shown in Figure 3.  The work was conducted by the same individual as
the Phase I test.  The improvement in standard deviation was thought to
result from frequent checks of every couple of hours, instead of twice each
day, of the calibration of the instrument and a reduction in the total
number of working hours by improved scheduling.  In future tests, it is
planned to preweigh the collection discs in the laboratory and weigh only
the final samples in the field.  This modification in procedure will
reduce the possibility of operator fatigue influencing the results.
    Blank Tests
    1.  Phase I Tests
    The results for the Phase I blanks are shown in Table I.  The Dow Corning
111 Compound and the Fluorolube GR660 grease both had weight losses, while the
Apiezon vacuum grease had weight increases.  The random nature of the
Apiezon weight increases led us to suspect that reactions with stack gas
were not the primary reason for the changes.  Examination of the plates
revealed that contamination may have caused the increases in weight.
During the Phase I tests, the impactors were scrubbed with acetone, followed
with a rinse with clean acetone.  After results were known from the first
blank tests, more attention was paid to the impactor clean up.  The reduction
in weight change in run 56 is attributed to the greater care taken  in clean-
up.  Work with the Dow Corning 111 Compound and Fluorolube grease was
stopped because of the instability of the materials.
    2.  Phase II Tests
    The Phase II blank tests were planned with the objective of testing  the
improved cleanup procedures, testing at different times to detect  reactions
with stack gases, and to test a Apiezon-lfylar  preparation for  trace element
sampling.  The trace element preparation described by Ensor et al  (1975)
involves the placing of an Apiezon-coated Mylar washer on the  collection
disc.                             j
    The cleanup procedure was modified by an ultrasonic  clean  in  soapy
water, a rinse in acetone, and a final rinse in clean acetone.

                                  124

-------
                                               PHASE  2
ro
01





-------
                                     Table  1.   BLANK IMPACTOR  TESTS  -  PHASE  I
ro
en
Run No.
Date
Start Time
Substrate

Sample Time, min.
Temp., °F
Gas Volume, b DSM3
HoO, Percent
562 / ppm
1
9/17/75
1152
Apiezon

123
253
2.94
4.4
600
2
9/17/75
1448
111 Compound

121
253
2.85
4.4
640
3
9/18/75
0810
36
9/23/75
0910
Fluorolube Apiezon
6R 660
119
256
2.75
4.4
630

120
260
4.12
3.0
750
56a
9/25/75
1550
Apiezon

120
210
2.52
4.6
750
63a
9/26/75
1746
Fluorolube
6R 660
119
218
2.84
5.3
680
Change in Collection Disc Weights, mg
Stage
1
2
3
4
5
6
7 .
Filter0
No. of Blank Disc
Average Weight
Charge of Blank Disc
Standard Deviation
a. boiler at half-load

0.335
0.135
0.27
0.12
0.085
0.21
0.01e
0.075
6
0.192
0.097

b. dry standard conditions, 21.1

-0.02
-0,34
-0.31
-.035
0.02
-0.38
0.0156
0.28
6
-0.230
0.108

°C, 760 mm Hg

-0.115
0.05
-0.075
-0.21
-0.17
0.22
0.01e
0.225
6
-0.050
0.160
d.
e.

0.68
0.23
0.15
0.03
0.09
O.OOe
-0.01s
-0.49
5
0.236
0,259
Reeve Angel 934
controls

0.16
0.09
0.06
0.05
0.05
O.Oie
0.01e
0.23
5
0.0820
0.0466
AH


-0.23
-0.17
-0.27
-0.28
-0.24
-0.066
-0.07e
0.25
5
-0.238
0.0432


c. corrected to 3 percent 0? (preliminary results)

-------
    The results of the Phase II blank runs are shown in Table  2.  The
average weight increase for eight blanks was 0.05 mg.   An  analysis of
variance reported in Table 3 indicated that the weight increases were
random and not related to a specific parameter.  Inspection of the plates
revealed that some contamination existed—such as fibers from  the inlet
filter and fly ask.
    One confounding aspect of these blank runs was the changes in stack
temperature caused by load changes and ambient temperature (the stack
is uninsulated).  Smith et al (1975) found that the weight increases on
blank runs were related to the stack temperature.  However, it is suspected
at the low temperatures under consideration, that this was not an important
variable.  It should be pointed out that these results are for a specific
low temperature and low SOp source.  The results cannot be extrapolated
to other sources with the existing information.  Additional field tests will
be required to determine the changes in weight of blanks on other sources.
    Additional improvements in procedures are planned—such as protection
use of a portable clean bench.
CONCLUSIONS
    The following conclusions are evident from the current study.
    1.  All aspects of the particle size measuring operation need to
be considered for compatibility and efficiency.  This includes the  people
performing the various duties, the impactors, substrates, balances,
support equipment, and procedures.
    2.  The test plan should include both controls and blank tests  to
evaluate the quality of the data.
    3.  The results of the controls and blanks should be carefully
analyzed to improve procedures of the impactor tests.
    4.  The average weight increase of 0.049 mg  for eight blank tests  should
be considered specific only for the source tested.  Evaluation of the
Apiezon-L coated 316 stainless steel collection  disc will be required for
other sources.
ACKNOWLEDGEMENTS
    The data reported in this paper was obtained as part of Contract RP534-1
for the Electric Power Research Institute.  The  support for the impactor
development and the writing of this paper was  obtained  from Meteorology
Research, Inc., internal research funds.
                                    127

-------
                                       Table 2.  BLANK IMPACTOR TESTS - PHASE II
ro
00
Run No.
Date
Start Time
Substrate

Sample Time, min.
Temp., °F
Gas Volume, b DSMd
HpO, Percent
S02, ppmc

Stage
1
2
3
4
5
6
yd
Filter
No. of Blank Disc
Average Weight
Charge of Blank Disc
Standard Deviation
a. boiler at half-load
49a
11/14/75
1000
Apiezon

0
205
0
—
490


0.08
0,01
0.01
0.00
-0.01
0.00
0.00
+0.036
6

0.015
0.033

b. dry standard conditions, 21.1
50a
11/14/75
1055
Apiezon

10
205
0.182
1,4
620


0.31
0.05
0.01
0.06
0.01
0.03
-0.01
+0.306
6

0.078
0.115

°C, 760 mmHg
7
11/17/75
0925
Apiezon

60
240
1.17
5.9
680
Weight

0.05
0.08
0.18
0.03
0.03
0.03
0.00
0.09e
6

0.067
0.059


c. corrected to 3 percent 02 (preliminary results)
42
8
11/11/75 11/10/75
0920
Apiezon

120
210
1.86
5.5
620
0800
Apiezon
Mylar
0
230
0
— -—
720
9
11/10/75
0849
Apiezon
Mylar
10
230
0,155
1.5
720
59
11/13/75
0936
Apiezon
Mylar
60
220
1.07
8.2
690
16
11/12/75
0843
Apiezon
Mylar
120
220
1.82
4.4
640
Change, mg

0.13
0,035
0,06
0.05
0,03
0.03
0.01
0.09e
6

0.056
0.038
d.
e.
f.

0,02
0,07
0,065
0,06
0,03
0.03
0.00^
-1.33f
6

0.046
0.022
control s
Reeve Angel
Whatman 41

0,05
0.05
0,05
0.03
0.03
0.09
0.02
-1.60f
6

0.052
0.020

934 AH


-0.02
0.00
0.025
0.02
0.04
0.50
0.01
-0.06f
6

0.114
0.199




0.07
0.06
0.01
0.01
0.15
0.03
0.02
-0.70f
6

0.0325
0.026




-------
         Table  3.  ANALYSIS OF VARIANCE OF PHASE II BLANK TESTS
Source
of Variation
Sample Time
Substrate
Residual
Total
Sum
of Squares
240.77
5.33
3163.37
3409.48
Degree
of Freedom
11
1
35
47
Mean Square
21.89
5.33
90.38
72.54
F
0.030
0.073
1.24

NOTE:   The weight change in mg x 100 was  used.
                                   129

-------
REFERENCES

1.   Cohen, J.J.,  and D.M.  Montan,  1967:   Theoretical  considerations,
     design, and evaluation of a cascade  impactor.   Am.  Ind.  Hyg.  Assn.
     J_. 28, 95-104.

2.   Ensor, D.S.,  T.A. Cahill,and I.E.  Sparks,  1975:   Elemental  analysis
     of fly ash from combustion of a  low  sulfur coal.   Paper  75-33.7
     presented at the 68th  Annual Meeting of the Air Pollution  Control
     Association,  June 15-20,  Boston, Massachusetts.

3.   Fegley, M.J., D.S.  Ensor, and I.E. Sparks, 1975:   The propagation
     of errors in particle  size distribution measurements  performed using
     cascade impactors.   Paper 75-32.5  presented at the 68th  Annual Meeting
     of the Air Pollution Control Association,  June 15-20, Boston, Mass.

4.   Smith, W.B.,  K.M. Gushing, 6.E.  Lacy, and  J.D. McCain, 1975:
     Particulate sizing  techniques for  control  device  evaluation.
     EPA-650/2-74-102a.
                                   130

-------
DISCUSSION

SMITH;  What would be a typical stage weight, that is particulate catch,
at the outlet?
ENSOR:  Well, probably on the order of a milligram or so depending on
the stage.  Again, I'm still screening the data, but we're getting on  the
order of a total of 5-10 milligrams.
SMITH:  What did the deposits look like visually?  Did they look as if
the grease wicked up in the traditional fashion?
ENSOR:  Well, that is an interesting thing.  We found that what the grease
does is extremely important.  When you're very careful and paint it on,
generally it doesn't cause problems.
SMITH:  As far  as particle retention, though, does it wick up through
the layer of particulate?
ENSOR:  I would suspect that that is not a significant thing.  We've noticed
that the deposits with this type of preparation are generally fairly fragile
and may be self-adhering.  But again, we're not using very heavy stage
loading, and teis business of  retention of greases is an interesting
point.  Tom Cahill has started using Apiezon-L with Lungren samplers and
has done a fairly extensive retention study using trace elements as
tracers, and he has found that this Apiezon grease is probably the most
effective thing we've ever fourvd.   It's more effective than Dow Corning
and in comparison to other materials.  But I think, certainly, the efficiency
with collection time is something that bears looking  into.
HARRIS:  You've had some indications of side collection on the mounds
anyway with that Apiezon.
ENSOR;  Yes, sometimes you'd see the mounds break, and in the collection  it
would look a little funny.
	:  What's  been your experience with that electrobalance in  trans-
porting it around.  Can you weigh something before taking it and weigh it
again and be sure that the balance  is correct?

                                    131

-------
ENSOR:  Well, generally it has been pretty reproducible.   We worked up
some procedures on our own and felt that they weren't terribly satisfactory.
I finally went to Colin Williams of Cahn and told him about the problem,
asked him what would be the optimum technique for weighing these sub-
strates, and he suggested weighing by substitution.   We feel that it's
working real well for us.  Like any other equipment, you have to exercise
an element of precaution and check it over.
CALVERT;  On that point, we have one and we've had nothing but trouble.
MCCAIN;  We've got and we've had very good success with them.
	:  Were those model 4100s?
MCCAIN;  No, we've got a G-2, the next one up from that with a digital
readout and then the automatic, and they've all performed highly satis-
factorily, and they've traveled all over the country under airplane seats.
BYERS;  Let me ask you this,  if you had your choice at a field location,
where you could use an analytical balance:  in a laboratory or your portable
electrical balance; which one would you use?
ENSOR;  Probably the Cahn.  You just don't find analytical balances in
the field.
	:  To carry a microbalance around is just not practical.
HARRIS:  Well, Seymour has had some success with this, carrying a micro-
balance around to use.
CALVERT:  We carry both.  We carry a Sartorious.  The Cahn is great as far
as not being susceptible to vibration, and when it works it's fine.  But when
it doesn't, there you are way out somewhere with no balance.  We always
have the Artorious as a backup, and we've had to use it many times.  We've
gone through the procedures that you found successful, and it didn't work on
ours.  I don't know what the latest is.  We keep getting correspondence
and new pieces to put in and try.
	:  What is your stage weight generally?
                                    132

-------
CALVERT;  I don't know.
ENSQR;  I think that thw size is very important.
CALVERT:  They are not much different from yours.
HARRIS;  I think that Colin is going to go into this a little bit more.
One of the things that's going to come out is that, if you've got some
idea of what your stage weight is, the electrical balance can be turned  to
that area.  It is not really a linearly sensitive type of instrument, and
you can tune it to a little bit finer degree at each end of the range.   If
you set it up that way, then you have better luck.
ENSOR:  I think there is a kind of moral here.  I'm kind of a do-all  as
both a manufacturer (of equipment) and a practioner.  One of the most
discouraging things to me is to go out and to talk to somebody and find
out that they have had problems for years with a piece of equipment and
they didn't say anything.  I think that one of the most important things
we do is to talk to Cahn to try to find out what's going on and to see
if there is a better procedure.  I think that this goes for all of the
equipment that we use.  The manufactures can build something that we can
use once they know what we need.
HARRIS:  One of the things we hope to do is to feed some of this back to
the manufacturers and see what they can do.  One of the things that has  come
out repeatedly is the problem of sampling points and how we're always not
in an ideal position with respect to the diameters upstream and downstream
and so forth.  You might be interested in getting a copy of the report that
Bill Kuykendal has had on contract from TRW on gas flow measurements, total
volumetric flow.  One of the things that is indicated there is that, for the
gas flow itself, the closer we are to a controlled disturbance like an elbow,
the better we understand the flow patterns.  It may be a skewed profile,
but it's defined.  In a rectangular duct, you get a nice smooth profile—all
of it sqished down to the bottom but it's there—and it is a nice regular
thing.  A few diameters downstream you may have eddies where the velocity
varies widely.  The problems that we have had in finding nice  sampling
positions may not have been that much of a problem.  It hasn't been defined
in terms of what it does to particles.  But since many of the  particles,
                                    133

-------
especially those in the fine particle region, follow the gas flow streams
anyway, it may mean that we haven't had that much of a problem.   We may
have been fortuitous in not being able to get to these ideal sampling
positions.
BYERS:  Well, there is a good paper, I can't give you the exact reference,
but it's by Oxley, where he is taking velocity profiles and concentration
profiles in ducts where he has been less than 1 equivalent diameter from
a bend or something upstream and he showed what the concentration variation
is at each equal area.  I think that it is a very useful piece of informa-
tion if you cannot get that ideal location.
HARRIS:  Yes, an extrapolation of the material they have in the study is
that for firms trying to get a measurement of total volumetric flow, without
going to some major, elaborate thing, it's probably better at an elbow
and use a single probe designed for that area, eliminating all of this
multi-point sampling.  It may get a better piece of information and we
may end up with the same sort thing with dust particles.
BOLL:  That means you have to make a presurvey of the points you're going
to pick for it to be representative of the total.
HARRIS:  Yes , that's right, you have to.  But as far as I'm concerned, if
you're going out on a field testing and you haven't made a presurvey, you've
blown the first week anyway.  You know there is no way you're going to go
cold to the site and say, "gee, I've been to a power plant before and they
won't give me any problems".  They don't build two generators at the same
power plant the same, much less two different plants.
BOLL:  I agree with you, that's exactly right, up and down, through a duct,
when you talk about small particles.  It may even be true crosswise.  However,
when you're talking about 100 feet crosswise, I'm not so sure it's still
true.  Sometimes you can get different firing at this burner than at that
burner and the gas flows are not uniform.
HARRIS:  Yes, there are limitations to the idea, and I  think that this  is
just one area where the presurvey comes in.  You may be able to  use only  three
probes down through there for the three burners and you don't have to do  a  700
point traverse to try to establish a profile.  All of this  needs to be  taken
into consideration.

                                   134

-------
   EXPERIENCE IN PARTICLE SIZING OF PETROLEUM INDUSTRY PARTICULATE EMISSIONS
              R. L. Byers, Exxon Research and Engineering Company


HARRIS:  Lee Byers of Exxon will go through some of his experience next.

BYERS:  Particulate size measurements in the petroleum industry are of con-
cern with respect to the three categories of particles.  The first category
is catalyst fines, which are emitted from catalytic cracking units; these
particles are typically oxides of silica and aluminum.  Coke fines (essentially
100-percent carbon), emitted from fluid coker units, represent the second
category.  The third category consists of fly ash from oil-fired furnaces
and process heaters.  The composition of the gas streams containing particulate
matter from each of these categories is characteristic of most combustion
gases:  about 12 percent CO^, 6 percent 0,,, less than 0.5 percent CO, with
the remainder of the gas being nitrogen and water vapor.  The gas streams
are typically 400° to 600° F and near atmospheric pressure.
       I would like to share with you some of the experience we have had with
various methods for sizing particles and indicate what we feel is a preferred
approach for sampling and analyzing particles from the categories just mentioned
       Size characteristics of catalyst fines have been of greatest interest
since catalytic cracking units are one of the seven major stationary sources covered
by EPA regulations.  Figure 1 represents the results of about 30 different par-
ticle size measurements taken at three different refinery locations.  The fig-
ure compares size distribution obtained by microscopic analysis of membrane
filters and by inertia! impaction.  Microscopic analysis was undertaken to
determine if the large percentage of particles below 1.0 micron, as measured
by impaction, was real.  This large "percent (about 40 percent) of submicron
Material was not expected based on process considerations.  It was felt that
the high percentage of fines might be due to reentrainment in the impactor.
       Before comparing these results further, I should indicate how the
filter and impactor samples were taken.  The impactor samples were taken
with an Andersen impactor located in-stack, between the probe nozzle and
                                        135

-------
                  10.0
CO
en
               u
               '§

                a
               O
UJ
»-
LJ
                   1.0
0.2


  0
                          I  I  I  I   I  I/  I   I    I   I   I  I  I   I   I    I   I    I  I  I   I  I
                      01  0.1      125 10  20
                                      40   60

                                       % < Dp
80  90 95    99     99.9  99.99
                Figure  2.   Fluid coker  particulate emissions  size distribution:   impactor
                            vs. membrane filter.

-------
sampling probe of an EPA Method 5 train.  No cyclone was used ahead of
the impactor and, of course, there was no need for the heated oven of
the Method 5 train.  The Andersen plates were greased with a 2 percent
solution of grease dissolved in benzene or toluene.  The membrane filter
samples were taken out-of-stack by replacing the cyclone, filter, and
the heated oven by two 47 mm stainless steel  filter holders, which
attached directly to the end of the sampling probe.  One of these filter
holders contained a blank filter and was used to establish isokinetic
flow through the sampling train.  The second filter holder contained the
filter which was used to obtain the sample.  The membrane filter holders
were heated; the sample filter was exposed for varying Tenths of time in
order to obtain samples which consisted of only a monolayer of particles.
     As Figure 1 shows, microscopic analysis of membrane filters results in*
a signficantly different size distribution for catalyst fines than obtained
by an Andersen impactor with greased plates.  The microscopic results show
a mass median diameter of about 2.5 microns, with about 5 percent of the
particles on a mass basis being less than 1.0 micron; corresponding values
for the impactor measurements are approximately 1.3 microns and 40 percent,
respectively.
     These results satisfied our hypothesis directionally, that is, that
the impactor was undersizing the particles.  More support was found when
the size distributions from the two sampling methods were plugged into a
mechanical model we have developed for predicting the fractional collection
efficiency cvf jet ejector wet gas scrubbers.  The model contains a
correction factor which only has physical meaning when its value falls
between 0 and 1.0.  When using membrane filter data along with scrubber
performance data, realistic values of this constant were found.  Using
impactor size distributions resulted in values of the constant which
exceeded 1.0.  Although this in itself may not prove that membrane filter
size measurement is more representative of the true particle size then
impactor measurements, it does tell us that, if we want to predict scrubber
performance, either for purposes of design of new scrubbers or optimization
of existing scrubbers, we should use membrane filter particle size data.
                                   137

-------
ENSOR:  It seems like you are comparing two different types  of diameters—
one a D5Q diameter and the other some kind of aerodynamic diameter.
BYERS:  The diameters shown in Figure 1 are the physical  diameters of the
particles for both sizing methods.   The microscopic sizing is done by
comparing the projected area of a single particle to the  equivalent area  of
a circle.  This is an actual physical measurement.   In the case of trapactor
sizing, since the density of the catalyst is uniform and  well known, it is
possible to calculate the actual physical D5Q.  Thus, the two size distribu-
tions are comparable since they are both measurements of  the same property,
the physical diameter of the particle.
SMITH:  Did you do a microscopic analysis of the impactor catches?
BYERS:  We looked at the individual impactor stages microscopically and
confirmed that there were larger particles on a given stage than would be
expected based on the impaction equation, when a value of the impaction
parameter for a collection efficiency approaching 100 percent was used.  We
did not determine a size distribution analysis per stage, however, since
there were too few particles present to give a statistically reliable
particle count.
MCCAIN:  If you eliminated the backup filter from your percentage distribution
on your impactor,  what would happen?
BYERS:  On the backup filter we have about 10 percent of the total weight
collected in the impactor.
MCCAIN:  That would not pull the curves  (in Figure 1) much closer.
CALVERT:  Do you have uniform distribution across the membrane filters?
BYERS:  Yes.  We look at filters with different exposure times and  choose
a filter on which the particles are not agglomerated and where the  distribu-
tion is uniform.  Some filters do have high concentrations at  the center of
the filter; these were not used for size analysis.  High concentration of -
particles at the center of the filter can be minimized if the  probe is kept
short and if it is cleaned just prior to taking the membrane filter sample.
                                138

-------
RAO:  Were the particles spherical?
BYERS;  No, they tend to be irregular in shape but more spherical  than
needle-like.
      Figure 2 shows that comparisons between membrane filter and impactor
size distributions for coke particles were quite similar to those found for
catalysts fines.  For coke particles the membrane filter data give typical
mass median diameters of about five microns with about two percent less
than 1.0 micron.  Corresponding values frr the impactor data are 2.6 microns
and 20 percent.
      As implied earlier, comparison of membrane filter data with impactor
data is only meaningful when the particulate matter is of uniform density
and the density is known.  Figure 3 is an example of how one can be misled
if he uses a membrane sample to size particles of nonuniform and unknown
density.  The figure shows typical size distributions of fly ash emitted
from oil-fired boilers.  The data were taken for two fuels:  fuel A,a
"dirty" (high Conradson Carbon Number) fuel, and fuel B, a relatively
clean fuel.  There the differences between the mass median diameters
measured by the microscope vs. impactor ranged from five to ten fold.  This
is  a significantly greater variation than was obtained by the two sizing
methods when the particles were catalyst or coke fines.  The obvious
explanation is that fly ash contains a large number of hollow particles
which, when counted as particles having the same density as solid particles,
results in a size distribution biased toward large particles.   In fact,
the microscopic and impactor size distributions of Figure 3 are not even
comparable.  Only when particles density is uniform and known is it pos-
sible to meaningfully compare the two size sampling methods.
      Much has been said earlier today concerning the use of various
substrates when collecting a sample by means of an impactor.  We have  done
a number of measurements to assess the relative effectiveness of grease
vs. paper substrates in reducing bounce and reentrainment.  Figure  5  shows
results of instack Andersen samples for both grease and paper substrates.
The paper substrates show about a 50 percent increase in mass median
diameter and about 10 to 15 percent less mass smaller than  1.0  micron
                                   139

-------
   10.0
       _   n i   i   I  i    I   i    I   i  i   I  I   i  r.  i    i    i  I  I
e
u
E

 a
Q
o:
ai
o
I-
    1.0
                        MEMBRANE FILTER  >^ /
                                       /  IMPACTOR SAMPLE
    0.1
     0.
           I  I I  I  I   I    II    I   I  I   I  I   t   I   II
                                                     I  1  I  I  I
01   0.1     12    5  10  20    40  60    80   90  95    99     99.9  99.99

                               % < D
   Figure 1.   Cat cracker particulate emissions  size distribution:
               impactor vs. membrane filter.
                                140

-------
than found when using grease substrates.  When these substrates  were compared
for out-of-stack samples, similar results, shown in Figure 6, were obtained.
The overall size distribution for the out-of-stack samples was smaller,
however, for both sampling methods due to the loss of the larger particles
in the probe.  This latter effect was confirmed by the data shown in Figure  7
which compares in-stack vs. out-of-stack particle size measurement using
paper substrates.
      On the basis of the results we have obtained to date, we find paper
substrates are preferable to grease in obtaining reliable size information.
Paper substrates also have the advantage of being stable at temperatures
up to 900° F; most greases are not stable above about 400° F.
ENSOR:  Were you running replicates of these tests?
BYERS:  Yes.  The data shown in Figures 5, 6 and 7 are for single tests, however.
ENSOR:  Single tests?  You do not know what your variation was between
identical  tests?
      Did you say that the difference between these different curves is a
difference in firing conditions?
BYERS:  The  different curves for any given figure represent data taken at
constant firing  conditions.  The difference in  the curves shown in Figures
4 and 5, however, is due to a change in firing  conditions betwen those two
tests.  Both curves  in Figure 4 were at the same firing  condition; likewise
for  Figure 5.
RAO;  When you collected the sample, did you  put the  probe in the same
location for in-stack samples?
BYERS;  Yes.
      All  of the sampling work  I have just described  was done under  typical
stack conditions of  1 atmosphere pressure and 300°-600°  F.   In  one  of our
test  programs it was necessary  to sample a process stream, upstream  and
downstream of a  particle collector, where the gas  pressure was  three
atmospheres  and  the  temperature was about 300°  F.  Seymour Calvert was  talking

                                    141

-------
    10.0
  2
  2
  o
  1
  LU
  5
  <  1.0
  o

  ui
  f-
  e:
     0.2
     Ill  II
                                         II  I   I    II    I  I   I  I  I
0.01   0.1
                          5  10  20   40   60   80  90  95    99     99.9  99.99
Figure 3.  Oil -Fired Boiler Fly Ash Size  Distribution:  Impactor  vs.
           Membrane Filter.
    10.0
      0.01   0.1     125  10  20   40   60    80  90  95     99     99.9  99.99
Figure 4.  Oil-Fired Boiler Fly Ash  Size Distribution:   In-Stack Impactor,
           Grease  vs.  Paper Substrate-
                                  142

-------
    10.0
 u
 1

 Q.
 o

 «t
 ce
UJ
_l
o
•I-
o:
<
a.
I _ I - 1  I   \ - LJ
           I  I  I   I  I   I   I    I    I  I   I  I  I
      0.01
                                                                    99.9  99.99
       Figure  5.   Oil-fired boiler  fly ash size distribution:
                   in-stack impactor,  grease vs. paper substrate.
                                143

-------
this morning about the CO content in certain gas streams to be sampled.
In our process stream we had 5 percent hydrogen, about 7 percent CO,
1000 ppm H2S.  We wanted to sample this stream for both particle con-
centration and size.  The sampling port we used to do the job is shown in
Figure 8.
     A pressure chamber 10 inches in diameter was designed.  It was made
that big merely to accommodate an Acurex stainless steel filter holder.
Figure 8 shows an Andersen sampler in place in the port.  We were able
to keep the particle collection device hot by using a continuous purging
stream from the chamber.  This purge stream was sent to a flare header
because the gas was toxic and potentially explosve.  The pressure in the
sampling probe was dropped across a needle valve downstream of the particle
collector.  Downstream of the valve we used a regular back half of the
EPA Method 5 train.  The whole secret of doing this high-pressure sampling
is to do things in the proper sequence.  With the needle valve shut, the
pump is turned on and about 20 inches of vacuum is pulled on the impingers.
Then, very gradually, the needle valve is opened  until isokinetic sampling
is achieved.  If the needle valve is opened too fast the impingers become
pressurized and the impinger seals are broken.  But with care, it is possible
with this system to sample at high pressures.
                                   144

-------
-p.
O1
                                                                                TO SAMPLING
                                                                                   TRAIN
                                                                             TO WASTE GAS
                                                                               HEADER
                          Figure 8.  High pressure sampling  port.

-------
DISCUSSION

BOLL:  Lee, you said that when  you  used  a  bare  substrate, you  had  seen
evidence of bounce.   Could you  say  what  that  evidence  is?
BYERS:  Yes.  We took the whole metal  plate and put  it under a scanning
electron microscope  and then took pictures of the  particles at about  5,000
X magnification. The sampling time  was not long enough for the particles
to pile up into the  typical peaks;  the deposit  consisted of individual
particles.  There were not enough particles to  get a size distribution on
each stage, but there were enough to give  an  idea  of the range of  particles
that landed on the stage.  On bare  metal plates, where a five  micron  particle
should have been the upper limit of the  size  of particle present,  based on
100 percent collection efficiency,  we saw  particles  up to seven microns.
BOLL:  If efficiency of the upper stages were only 95  percent  for  that
seven micron particle., you would still expect to find  some of  them down-
stream.  But you found more than that expectation.
BYERS:  Yess that's  right.
HARRIS;  Some of the early work on  the Brink, looking  at  bare  metal versus
greased substrate and using the backup filter as your  means of seeing whether
you are getting penetration through these.  Because  in this case,  it was
fairly large fly ash we were using.  You could  visually see the difference
between what penetrates through the impactor  back  stages  to the filter
and what doesn't.  That is, in the^ greased stage there was  a  practically
white filter, and, say, in the ungreased the  filter was dirty.  This is
m aningful since you were using the same instrument.
BYERS:  In summary,  we have found that particle size distributions based
on cascade impaction have significantly  smaller (40 to 50 percent) mass
median diameters than distributions obtained by microscopic analysis of
membrane filters.  The percent mass less than 1.0 micron for  impactor
size distributions is significantly greater  (5 to 10 t.mes)  than for
membrane size distributions.  Paper substrates  give about 50  percent
larger mass median diameters and 0.1 to  0.5 as  much mass below  1.0
                                  146

-------
micron as grease.  When sampling conditions permit, membrane size distributions
are recommended for use in calibrating impactor size distributions.   Paper
substrates are recommended over grease substrates and membrane filters when
impactor sampling is required due to temperature limitations or the presence
of nonuniform density particles.
                                   147

-------
             "HOW TO WEIGH  IT,  ONCE  YOU  HAVE  COLLECTED  IT!"
                                                          *
             Dr.  Colin J. Williams,  Cahn Instrument  Company


HARRIS:  Another important  consideration in the  collection of airborn
participate matter is the precision  and  accuracy of  the weighing  procedure.
How do you weigh the material  once you've collected  it? We have  invited
Dr. Colin J. Williams, an employee of a  major microbalance manufacturer,
to speak on this subject and to tell us  how to minimize our weighing
problems.

WILLIAMS:  Thank you, Bruce.  In his introduction this  morning, Bruce
Harris stated that, "Most of the people  here  have been  active in  in-stack
sizing problems"  Well, that does not include me! My only involvement
in this field results from  working with  people such  as  Dr. Ensor  of MRI
and Dr. Calvert of APT, and talking  to others about  their particular appli-
cation problems, and reading a few reports published by EPA and SoRI.   I
do feel, however, that your weighing problems are basically no different
from those of anyone el,se.   The concerns of a microbiologist weighing
the eyeballs of a plankton  are the same  as the  in-stack particle  technologist
weighing an impactor plate  loading,  and  that's what  I'm here to talk about.
      Let me first of all  define the problem  as  I see it.  There  are various
designs of cascade impactors available commercially, as well as others still
in the research stage.  Collection plates or  substrates of various configura-
tions exist, varying widely in weight and physical size.  The  typical
sample loading is very small with respect to  the weight of  the  impactor plate.
A suitable balance must, therefore,  possess  adequate specifications such  as
load and sensitivity, and adequate features  such as  a large weighing compart-
ment and large stirrups to  support the plates.   There is  one  other most
important criteria which I  will introduce in  the form of  a  quotation from
a paper by Messrs. Byrd, McCain, and Harris  (1973):   "...balances that
operate well in the laboratory environment at required sensitivity, are
*
 Presently with Perkin-Elmer Corporation.
                                  148

-------
not uncommon.  However, the choices are severely limited when the balance
must be transported to remote field test sites, set up in an industrial
environment and still be expected to perform to the required precision.
Those balances that are readily transported and have the necessary sensi-
tivity generally have a maximum capacity of only a few grams."  An accurate
assessment indeed, and one that probably resulted in all heavy impactor
plates being discarded in favor of light weight designs.  I believe the
heaviest are of the order of 1.2 grams.  The heavier they are, the more
unsuccessful one will be in measuring a small weight difference (between
the unloaded and loaded plate weights).
      So, although the semimicro balances of Sartorius and Mettler serve
the purpose adequately in a conventional laboratory (presuming their
delicate knife-edge suspensions are protected from deleterious vibrations),
their inability to be transported with ease render them quite unsuitable
for transportation and field use in mobil labs or motels.  Only electronic
microbalances, with their totally different beam and sample suspension
designs, can withstand the rigors of field use.  The manufacturers of such
microbalances rather appropriately offer foam lined metal carrying cases.
The decision to use electronic microbalances is almost universal amongst
impactor technologist.  Maximum permissible sample weights of up to five
grams and sensitivites down to 0.1 micrograms mean that they provide
adequate specifications.
      The accuracy of any sample weighing depends on various factors such as
calibration  stability within the electronics of the balance  (this should
be carefully checked during the evaluation of a microbalance); accuracy of
the digital  readout display (commonly ± 1 count on a ± 20,000 count display);
and accuracy of the calibration weights (class M weights, traceable to the
National Bureau of Standards, are typically better than 0.05  percent
accurate).  An additional burden of responsibility lies with  the calibration
weights when they are used in substitution weighing.   It is  usually safe
to say that accuracy is not a limiting factor in a weighing  procedure.
Precision is far more important.
                                   149

-------
      Is substitution weighing in  fact  necessary  or even  desirable?  The
answer is yes.   Such a technique allows weighing  of samples to a finer
resolution than when one chooses a weight  range large  enough  to display
the full sample wight.  Most microbalances offer  a series of  weight
ranges, including 0-1,000 mg, which  has a  resolution of 0.1 milligrams.
The range is large enough to exceed  the exeed  the weight  of,  for example, a  732.1
mg impactor plate, but the resolution obtained is insufficient, then, considering
the total wight of sample collected.  If the following substitution  pro-
cedure is used, resolution to 0.01 mg can  be obtained.  Set the weight range
to 0-200 mg, for example; place 600  mg  of  calibration  weights (preferably
a 500 mg and 100 mg weight) as substitution weights, on the sample weight
tray; and counterbalance (tare) with 600 mg on the tare weight tray.  Zero
the readout display with the electrical zeroes.
      Remove the substitution weights  from the sample weight tray and replace
them with the impactor plate.  Considering the relatively heavy weights
of these objects, remove and replace with  great care.  Do not jab  down-
wards with forceps when attempting to pick weights off the weight  tray and
do not drop weights onto trays. The readout should now display something
like 132.13 mg, which, added to the  substitution  weight of 600 mg, means
the impactor plate weight is 732.13  mg. With  great care, this procedure
could be taken a step further, using substitution weights totaling 720
mg and weighing on the 0-20 mg weight range where the  resolution  or
readibility is 0.0001 mg.  An alternate approach  as far as tare weights  and
substitution weights are concerned is  to  use blanks  (impactor plates).
Then, the bouyant forces on the sample  and counterweight, and on  the sub-
stituted plate and actual sample plate, will  be the  same.
      Precision or repeatability is  the most  important figure of  merit for
any microbalance.  If a sample weight  reading  cannot  be  repeated  within
certain performance specifications,  it  was not worth  the  effort of weighing
it in the first place.  Precision, expressed  as a standard deviation, is
actually based upon 11 consecutive weighings.   The balance manufacturers'
literature should be consulted and the  best precision figures (best that can
be expected) calculated and compared to actual results.   Strictly speaking,

                                  150

-------
only when such an exercise results in weight values for which the calculated
precision equals, or is better than, rated precision for the chosen
sample weight and weight range should one assume one has an acceptable,
repeatable, and accurate weight.
      Good weighing technique is desirable for accurate weighings, but no
special skills are required other than an understanding of the various
weighing steps and a certain amount of practice.  Of course, not even those
highly skilled in the art of microweighing can expect success unless the
conditions are suitable.  Dr. Ensor experienced severe problems during
an early field test until he realized that he was scuffing his shoes on
the nylon carpet of the motel-laboratory.  Severe electrostatic charges
can easily be transferred to the microbalance weighing mechanism across
the surface of the operator's body.  In such instances, a radioactive
ionizing source such as Polonium-210, an alpha emitter, (manufactured
for example by Nuclear Products Company, South El Monte, California) is
highly recommended.  It should  be placed on the floor of the balance
weighing compartment throughout the weighing procedure.  For impactor
collection surfaces made of membrane filters, such as those of Sierra
Instruments (Carmel Valley, California), it would also be advisable to
momentarily hold the ionizing unit close to each side before weighing.
      The topic of sufficient roominess of the weighing compartment when
weighing large diameter foils has already been touched on.  However, there
is one other point I would like to make.  In a recent report by Bruce
Harris, the statement was made: "(It is desirable) to have a balance with
a large enough weighing chamber to accomodate the "Pi!at" foils without
having to fold them."  I agree  completely.  Even slight bending of metal
foils incurs the risk of chipping off some dried particulate matter, as
well as increasing the chance of contamination by touching the foils.
      And may I quote again:  "...insensitive to vibration,  if it  (the
microbalance) is to be used in  the field."  The phrase  "insensitive  to
vibration" is ambiguous and often misunderstood.  A microbalance  is  per-
fectly capable of withstanding  the rigors of transportation without  any
arrestment of the beam or suspensions (this is not the  case, of  course, with

                                   151

-------
with conventional  beam balances as stated previously).   No  damage  results
when a microbalance is transported to a field test site, mobile  lab,  or
motel, and set up within a few minutes by the technician.   However,
if the environment is inherently noisy with low frequency vibrations  (from
air conditioners in a mobile lab or from an industrial  plant nearby),
weighing could be seriously hampered.  A simple test involves placing one's
fingertips lightly on the balance table or bench.   If vibrations can  be
felt, somewhat like a pulse, then probably a new,  less  noisy, environment
will need to be located.  In this context, a microbalance quite  obviously
is sensitive to vibrations even though it will not suffer damage as a result
of them.  The rule is to choose as sturdy a support table as is  available,
one that is apart from disturbances which might develop inertia! forces
in a sample during weighing and lead to slow and nonprecise weighing.
      In conclusion, I wish to make the claim that some modern electronic
microbalances can provide the answers to the impactor technologists'
weighing problems.  Considering all of the information I have acquired relating
to your weighing needs, I do believe that sufficient capacity, sensitivity,
                             j
accuracy, and precision are attainable, presuming you follow the recommended
procedures.  Given the opportunity, it's always possible that a tune-up job
can be done to your microbalance, by a skilled service engineer, of course,
and then the performance could possible exceed quoted specifications.
      Gentlemen, thank you for the opportunity of participating in this in-
stack particle sizing seminar, and I trust everyone now feels confident
enough to return to their labs to accurately weigh the eyeballs of a plankton!
                                   152

-------
DISCUSSION

MCCAIN:  We alleviated our problem at SoRI in handling large substrates,
particularly those from the University of Washington, in the following
manner.  Their roils are rather stiff and probably could not be flattened
for reuse if they were folded or bent to facilitate weighing!  So a
chamber was designed and positioned underneath a Cahn balance.  It was
necessary to remove the weighing mechanism compartment from the body of
the balance and drill a hole through the base of this metal can.  Another
hole needed to be drilled through the balance chassis, and both of course
to support the foils, and since the new weighing chamber was constructed
of clear plastic, the operator could see if the stirrups had stopped
swinging about before taking a weight reading.
WILLIAMS;  Apparently SoRI is quite happy with the modifications that Joe
McCain has just described.  However, they are not modifications that most
users would welcome undertaking, for fear of damage to an expensive micro-
balanced.  Joe also tells me that the present setup is hardly suitable
for field use, since a  lack of rigidity  in the added weighing chamber
and the increased distance between suspended foils and their point of sus-
pension to the beam result in significant pan swing.  This problem presents
itself in the form of increased weighing time and reduced precision.
      Weighing chamber  size and other features are not the same for all micro-
balances.  The user is  advised to measure his specific needs against  a
balance's specifications prior to choosing one.
                                    153

-------
              CASCADE IMPACTOR DATA FOR ELEMENTAL ANALYSIS
                  T.A.  Cahill, University of California
     Presented by D.S.  Ensor, Meteorology Research,  Inc.  - Co-Author
HARRIS:  Tom Cahill  of Crocker Labs cannot be here so Dave Ensor will  pre-
sent his talk.   Crocker Labs is using cascade impactors  to get data on
elemental analysis.

ENSOR:  Thank you.   The elemental  analysis of cascade impactor samples
using ion-excited X-ray analysis (IXA) was summarized.   The IXA system and
the results of inter!aboratory comparisons were described.  The elemental
penetration as a function of particle size through a scrubber was used as
an example of practical application.
ACKNOWLEDGMENT
      The scrubber results were obtained under EPA Contract 68-02-1802.
The analytical  system has been developed and used with the support from the
California Air Resources Board, National Science Foundation, Research
Applied to National  Needs; and the United States Navy.
      This paper was written with funds from Meterology Research, Inc.
INTRODUCTION
      The desirability of having information on the elemental content of
particulates as a function of aerodynamic size has long been accepted by
scientists concerned with the biological and physical impacts of atmospheric
aerosols.  However,  particle sizing devices deliver little material for
subsequent chemical  analysis, so that difficulties are presented in the
generation of such information.  Use of destructive methods, such as wet
chemistry, atomic absorption spectrophotometry, emission spectroscopy, etc.,
compound the problem when information on more than a few elements is needed.
                                  154

-------
ELEMENTAL ANALYSIS SYSTEM
     Recent advances  In  the accuracy and sensitivity of nondestructive
X-ray-based analytica-l- procedures  have greatly alleviated these problems.
In particular,  energy dispersive X-ray spectrometers have allowed
scientists to analyze for dozens of elements at one time from a sample
of total mass of  1  mg,  and extension to  atl_ elements  sodium  and
heavier  in a single analysis  has been accomplished.  Surprisingly, costs
have also been  reduced at the same time.
     These methods  however,  have,  limitations inherent to any X-ray
based  technique and limitations unique to  energy dispersive systems:
      •   Light elements are difficult to make quantitative due
         to particle size effects and filter enmeshment effects
         because of X-ray attenuation.  These mainly affect
         elements  lighter than about potassium for  typical
         cascade impactor samples.
      •   In energy dispersive  systems, the  necessity of seeing
       •  all elements, a  great advantage, results also in
         difficulties  in  seeing weak elements in close
         proximity with abundant elements  (ratios of greater
         than  1,000  in mass).   Interferences  cause,  problems
         for the complicated samples  seen in stack  effluents,
         although  with care they are  rarely serious and
         generally limit  only  a few elements.
     Once these limitations are understood, these  methods posses enor-
mous advantage  in that they allow  one to gather great amounts of informa-
tion  in  short sampling times  with  no predisposition about which elements
are present and in what  amounts.   The nondestructive nature allows
reanalysis to higher  sensitivity precision when the.-preliminary data
are interesting,  an important advantage.
     At  the Crocker Nuclear Laboratory of  the University of  California,
Davis, these methods  are being applied to  samples  collected  by Meteorology
Research, Inc.  Ion beams are used to excite the X-rays, which allows
one to perform  an analysis of a single deposit from an  impactor with no
loss ,of  sensitivity.  Any element  sodium or  heavier can  be seen if  present,
although automatic'reduction  codes normally  on'ly print  out 40 elements,
ignoring all rare earths and  most  actinides.  The  most important information
gained is often that  some element  is not present  (to  a  stated  level),  a
result fervently  to be desired in  many cases.
                                   155,

-------
     A schematic of the ion-excited X-ray analysis (IXA)  system at the
University of California at Davis is shown in Figure 1.
     An 18 Me V alpha beam from the cyclotron passes through remotely
readable graphite collimators and impinges on the thin target which was
mounted at an angle of 45 degrees to the incoming beam.   The sample or
target is mounted into a 35 mm slide.   The target slide changes is opera-
ted under real time computer control.   Beam spot uniformity is acheived by
the use of a diffusion foil (6.4 ym Al) and different sized target collim-
ators.  The beam is then collected by a Faraday cup and integrated to a
precision of about 2 percent to give the total  charge, Q, that passed
through each sample.  X-rays that pass through an active filter and a
                                                                   2
25 micron Be window are converted into electrical pulses by a 10 mm  x
3 mm liquid nitrogen-cooled Si (Li) detector and associated pulsed optical
feedback circuitry.  Data are accumulated in a Digital Equipment Corp.
PDP-15/40 computer with Nuclear Data 2200 Analog-to-Digital Converter's
integral to the system, giving a spectrum of characteristic X-rays and
a smooth background.
     An example of an X-ray spectrum is shown in Figure 2, with a small Cu
peak next to a Zn peak of 18 mg/cm^.  This system has mean accuracy in  all
interlaboratory intermethod comparisons for all elements Al and heavier
of 1-03 ±.0.09, since January 1973.  Over 30,000 analyses have been made to
data, yielding over 5000,000 positive elemental determinations and very
many upper limit determinations.  The automatic data reduction procedure
is shown in Figure 3.  A nice feature is that any sample, no matter how
lightly or heavily loaded, is awlays run at optimum count rate for the
detector, as essentially unlimited X-ray excitation is available from the
accelerator.  Also, elements hydrogen through fluorine are not being routinely
extracted from ambient air samples (with elastic alpha scattering).  An
example of the alpha spectrum is shown in Figure 4 of NaCl on mylar although
this is not being done for stack samples at this time.
     The system accuracy has been evaluated in a recent interlaboratory
comparison reported by Camp et a! (1975).  The results are summarized
in Figure 5 for standards made up of trace elemental  solutions deposited  in
a filter.  Each square is one laboratory, one method.  The value  on  the

                                   156

-------
                                                                        TO CONTROL ROOM
                   ALPHA
                  DETECTOR
                                AMP
                                               TO
                                              PDP-15
                                             TO
                                           VACUUM
                                            PUMP
      FARADAY
        CUP
en
             DIFFUSION
             VACUUM
             PUMP
                                                                   u
                                                                   n
DIFFUSION FOIL
     CYCLOTRON
       BEAM
SWEEPING
 MAGNET
                                                                        PULSE
                                                                        SHAPING
                                                                     ELECTRONICS
                                                                                      TO  SWEEPER


                                                                                      TO  PDP-15
          Figure 1.   Ion-excited x-ray analysis system at the University of California, Davis.

-------
      40,000
      30.000 . -
00    20, 000 - -
     10,000 --
                                                                    Run 07005   4/10/73
                                                                         16 Mev a 's
                                                Channel
                                  Figure 2.  Example of x-ray spectrum.
                                                                                                --200
                                                                                                -- 150
                                                                                                -- 100
                                                                                                --  50
                                                                                               500

-------
 o
U
o
U
o
U
6400


3600


1600


 400






6400


3600

•s

1600



 400





6400


3600



1600



 400
   90000
§  40000
U


    10000
         f;
                                        URBAN AEROSOL SAMPLE
                                        VS. MYLAR BLANK
                                        SAMPLE MINUS BLANK VS.

                                        GENERATED BACKGROUND
                                      RESULTANT SPECTRUM WITH

                                      BACKGROUND REMOVED
                                       CORRELATION SPECTRUM
                                JL
                                         10
                                           12
  2468


                        Energy (KEV)

Figure 3.  Example of automatic readout system.
14
16
                                   159

-------
cr>
o
                                                                         1	,


                                                             FILE 00256   SUBGROUP  1
                                                             NaCl on 0. 15 MIL MYLAR
                                                                                   480         560
                          Figure 4.   Example of  a scattered alpha spectrum.

-------
                                 RATIO, MEAN RESULTS VS STANDARDS
cq  4
    0
                                                            UofC
                                                            Davis
                                                          1.01 +0.18

                0.4           0.6           0.8           1.0           1.2           1.4

                 Figure 5.  Results  of  interlaboratory comparison with solution standards.
1.6

-------
figure is the mean accuracy and scatter of all labs reporting data.  The
UCD analysis-average results are marked with an arrow.
     The results of analysis of a rock sample are shown in Figure 6 and
a fly ash sample  is  shown  in Figure  7.  The increases scatter between the
laboratories indicates the importance of sample preparation.  Analysis
techniques requiring some disruption Of the granular sample during pre-
paration had wide scatter.  The IXA analysis method, requiring little
preparation, was very consistent
APPLICATION TO CONTROL DEVICE EVALUATIONS
     The IXA analysis method was applied to determine elemental  efficiencies
in the evaluation of a TCA scrubber (Ensor et al , 1975).  The source testing
was conducted using a scacade impactor as described by Ensor et al (1975).
Inlet and outlet tests were made on the scrubber to allow the determination
of the penetration of the elements through the scrubber.  An example of
the results are shown in Figure 8.  The mass penetration has been plotted
on the graph for reference.  The striking implication of the elemental
results is the large difference in penetration depending on chemical nature
of the aerosol.  Si and Al probably existing as insoluble oxides have a very
low penetration through the scrubber, whereas the soluble elements have much
higher penetration in the submicron region as well as a suggestion of a bi-
modal penetration, which may be due to entrainment through the scrubber
mist eliminators.
     In Table 1, the elemental  penetration is compared to the mass penetratibn.
Again, the large 'difference between penetration of Al and Si and the other
elements is apparent.  It also appears that the scrubber is a source of
sulfur-containing aerosols, possibly in part from gas to solid phase reactions.
The soluble elements, such as Cu, Zn, and Cr, appear to be generated from
the evaporation of the scrubber liquor.
SUMMARY
     The use of IXA for aerosol analysis iand control device evaluating has
been briefly discussed.  The advantages of simple preparation and non-
destruction of the sample make it attractive for source test samples.
                                   162

-------
                                            RATIO,  MEAN RESULTS VS STANDARDS
o\
CO
CO
tt

4 •



2 •

0

















0.








4








0.
.-.














UofC
Davis
0.95
0.16
1













6 0.8 l.O 1.2 1.4 1.6
                        Figure 6.   Results of Inter!aboratory comparison for ground rock.

-------
                                         RATIO, MEAN RESULTS VS STANDARDS
en
            CO
            cq











0.





4 0.6



-

0.





8

UofC
Davis
0.95 _+ 0. 38
f


I -i -I . i
1.0 1.2 1.4 1.6
                  Figure 7.   Results  of interlabpratory comparison for fly ash.

-------
        10'
           10
PARTICLE DIAMETER,
                                                               75-288
Figure 8.   Example of scrubber  penetrations for selected elements
           as particulate matter in the flue gas.
                             165

-------
         Table 1.    PENETRATION OF THE ELEMENTS THROUGH
                    THE SCRUBBER FOR DECEMBER 10, 1974
Element
Al
Si
S :
K
Ca
Ti
V
Cr
Fe
Ni
Cu
Zn
Br
Pb
For All Elements
Total Mass
Penetration
0.029
0.033
3.4a
0.043
0.059
0.073
0. 14
I.10a
0. 18
0.95
2.9a
1.5a
0.28
0.64
0. 108
0.074
Average Outlet
Concentrations
Mi croc rams /DSm3
326
658
1030
50
508
96
27
57
1500
33
668
501
5. 7
120


a
 Penetrations greater than 1 indicate generation of particles from evaporation
 of the scrubber liquor.
3Dry Standard, 21. 1 °C,  760 mm Hg
                                   166

-------
REFERENCES

Ensor, D.S., B.S. Jackson, S. Calvert, C. Lake, D.V. Wallon, R.E.
Nilan, F.S. Campbell, T.A. Cahill and R.F. Flocchini, (1975):   "Evalua-
tion of a particulate scrubber on a coal-fired utility boiler," EPA-
60012-15-074.
Ensor, D.S., T.A. Cahill and L.E. Sparks (1975):  "Elemental analysis
of fly ash from combustion of a low sulfur coal", presented at the 68th
Annula Meeting of the Air Pollution Control Assoc., Boston, Mass,  June
15-20.
Camp, D.C., A.L. VanLehn, J.R. Rhodes, and A.H. Pradzynski (1975):
"Intercomparison of trace element determinations in simulated and  real
air particulate samples", X-Ray Spectrometry, 4_, p. 123.
                                    167

-------
DISCUSSION
        What grease did you use?  Apiezon?
ENSOR:  Yes.  It's probably about the very best thing you can find.
	:  It doesn't give background readings?
ENSOR:  No, it's pure hydrocarbon.  One intriguing thing was that when we
started on this Tom [Cahill] used Apiezon M in some of the screening tests.
I later sent up some of the greases that we were actually going to use.
Apiezon M apparently has Bentonite in it, and all kinds of trace elements
showed up.  We had some pretty hurried telephone conversations about what
to do with the background.  Eventually we used the Apiezon L, which is
apparently pure hydrocarbon.
	:  I noticed that copper was high in penetration.  Do you think that
was interference in the instrument, or something in the background or is
it real?  I can understand the sulfur.
ENSOR:  We found at this source that a lot of the scrubber liquor was being
entrained.  We were seeing the actual liquor itself.
HARRIS:  Thank you.
                                   168

-------
                     GENERAL DISCUSSION - WEDNESDAY
HARRIS:  Well, that's the formal program for today, and I'm hopeful  that
we could use the remaining time we have, an hour and a half or so to kind
of bounce around some of the problems that we've raised.  If there are any
problems that you have maybe some of the people can help with of just any-
thing in general that involves this whole problem.  Seymour is going to
have to cut out, I figured he might be able to go through the rest of his
list.  We would ask that for this type of thing, if we could have you
identify yourself a little bit, so that we can give you credit for or blame
for it when we get the transcripts going if there are any questions.  Any-
body want to lead off?
BOLL:  I'm Dick Boll from Babcock & Wilcox.  You will recall that sometime
earlier Dr. Calvert and I were debating the question of whether or not a
temperature drop occurs within each impactor stage due to the pressure drop.
He suggested that the temperature drop might cause condensation on the parti-
cles, causing them to stick to the target in spite of high jet velocities, and
I maintained that no temperature drop,is possible because the expansion is
isehthalpic, not isentropic.  In the meantime, we have put our heads together
and come up with a hypothesis that is consistent with both points of view.
It is this:  Within the flowing gas jet itself, the temperature does, indeed,
drop due to isentropic expansion.  However, after the gas makes the turn
around the target and its velocity is decreased by turbulence and friction,
the temperature rises back up again to quite close to the initial value.
The temperature of the pile of deposited material on the target is, of course,
the initial "stagnation" temperature, since all of the gas within this pile
is moving with essentially zero velocity.  Thus, since the temperature drop
might be substantial in cases of high-velocity jets (of the order of 50° F),
condensible materials, e.g. sulfuric acid, might condense upon the  particulate
and act as a glue to hold the particles together at the instant of  impaction.
Thereafter, one would expect the "glue" to revaporize, but adhesion once
established might be maintained anyhow.  Interestingly, this theory seems to
explain the Southern Research observation that particle bounce occurs when
ammonia additive is on but stops when it is turned off.

                                   169

-------
MCCAIN:  No, it was the other way around.
HARRIS:  No.  That conforms.   You're using  it as  a  surface  conditioner when
it's on—it's j.ust like you're using the acid as  a  surface  conditioner when
it's condensing in the jet--and once it's  deposited on  there,  it  doesn't
really matter what goes on.
BOLL;  All theories should have a converse, which is also  true.
HARRIS:  Well, we got a theory out on the table anyway.  Any comments
or anything else they want to go over.   Problems?
	:  I was ju't wondering about the problem of trying to get  efficiency
data on control equipment and the question of agglomeration influencing
the efficiency data, what people have done in terms of  looking in to see
that fraction coming out.  Is it what we thought it was going in, or
defining that it's not, looking at pictures of it and that sort  of thing.
Has there been much done in that direction?
HARRIS:  Looking at the catch on the individual stages  and seeing if it's
an agglomerate or if its still wet coming out. I, as a  personal  comment,
I would tend to think tha;t you would have an awful  lot  of  trouble trying
to look at  a catch of an impactor stage and determine which one of those
particles,  other than to assume that the inertia! separations that we are
operating these things under are separating out the particles in a reasonable
manner.  I was a little concerned with Lee's use of Coulter analysis for
analyzing the stage efficiency or the stage sizing, on  those things to because
most of the Coulter Counter techniques rely on getting a very good dis-
persion of  particles.  In fact, it's such a good dispersion that breaks
everything  up that's agglomerated.
BYERS:  Let me relieve your concern.  I've never done that thing in my work,
there  are people in other parts of the company who  requested me  to give
them data from my impactors, and they used the Coulter Counter for their
own purposes.  But, I would agree with you.  There  is no way to  reconstitute
the particles in the distribution which existed  in  the gas phase, once you
have them in the liquid phase.  That's what the  Coulter Counter  requires, so
I wouldn't  put any confidence in a Coulter Counter  distribution.

                                  170

-------
HARRIS:  At least not the kind which analyzes the separate stage.
BYERS:  To answer your question, though, one way that we could do  it is
to sample on a membrane filter, on the inlet and outlet of the control
device.  Then just look for agglomeration if that's a problem that you're
worried about.
HARRIS:  Membranes are going to give you a problem because you're  not going
to catch anything that's less than the particle size in the pore size of
the membrane filter.
BYERS:  A 0.2 micron pore size just about catches everything.
MCCAIN:  If you're getting many one or two micron particles, almost nothing
but 1 and 2's, in large quantities on a stage that's not supposed  to collect
anything smaller than about 10 or 15 microns , you've got a  pretty good
indication that you've got substantial agglomeration.
	:  You've seen something like that?
MCCAIN:  Yes.  At some precipitators we've seen that to be the case.  Some
people from Lawrence Livermore Laboratories  (I think they used one of your
impactors, Mike [Pilat])  have  found  the  same thing.  They attributed it at
that  time, in a preprint I read, to finite collection efficiency for all
particle sizes on all impactor stages.  They concluded that it (the impactor)
didn't work.  That is, it was actually collecting one micron primary parti-
cles  on stage zero or stage one, the first stage.
	:  We've done considerable work, both optical and scanning electron
microscopy on collection plates, and you can, for certain types of emission
sources, characterize the particles as long as you don't overload the stage.
Even  those that are overloaded, with significant mounds, some types you can
see all blended together.  I think that a lot of it depends on your coating
techniques to put on a conductive surface.   If you put too thick a layer
of conducting material, it masks the individual particles.
HARRIS:  One thing that wasn't mentioned was that some of the work you have
done with the electrostatic effect, just -on  the effect of the collection
material itself, influencing the electrostatics, not so much as whether you

                                   171

-------
are neutralizing ahead of time but whether you  had  a  conducting  or  non-
conducting substrate that you were putting on.   If  I  remember  rignt,  the
conducting substrate gave you more problems than you  would  with  the non-
conducting substrate like fiberglass.   Due to electrostatic effects.
MCCAIN:  I have a slide that shows that.   There can be some electro-
static effects.  I would hate to draw that one  from memory.
	:  This is data on, insulated surface versus a  grounded surface.
MCCAIN:  Yes, let me find the right one.   (Figure 1).  Pardon  the way
this is presented.  If I were doing it over I think I'd do  it  the other
way.  We took the ratio of concentration  observed with the  neutralizer to
the concentration observed without the neutralizer instead  of  the other
way around.  I would have been, philosophically, a  little better to put  it
on the other basis.  This was done with an optical  particle counter using
a Brink impactor in two or three configurations, one  of them was just a
simple probe.  First what happens with only the probe with  and without a
charge neutralizer at a fairly low flow rate, 0.1 CFM.  Very little effect
about 2 microns, a substantial effect a 1 ym and then back  down. This all
was done at the exit of the pilot scale electrostatic precipitator, operating
at moderately high efficiency.  The (red) diamonds are what happens to the
Brink impactor under the same circumstances showing relative exit concen-
trations from stage 2 using metal substrates.  For. some sizes, at that flow
rate, there is a trememdous change without the neutralizer  apparently due
to electrostatic effects in the impactor.   This is glass fiber substrates, and
you can't see much difference.  Now why glass fibers  would  do what they did
toward reducing the electrostatic effects, is somewhat uncertain.
	:  That's just greased metal?
MCCAIN:  That's just bare metal, grounded.  Just a stainless steel plate.
Now, the same system and for the probe and impactor, just to see whether
or not the data was real.  (Figure 2) That is, v/as what we were getting
with the particle counter real?  We took the same system (probe and impactor)
and measured it with the particle counter and then did it again gravimetrically.
Figure 2 shows the comparitive results.  One curve shows the predicted results
based on the probe and impactor data using the particle counter.  The particle

                                   172

-------
            2,0
            1,5
     CD
     t_>
     ^
     3
CA)
     LU
            1,0
                                                        —3  FOOT LONG, * INCH  I,D.  PROBE
                                                      0—
                                EXIT OF STAGE 2 OF BRINK  IMPACTOR
                                WITH METAL  SUBSTRATES,
                                EXIT OF STAGE 2 OF BRINK  IMPACTOR
                                WITH GLASS  FIBER SUBSTRATES,
                                    *
,6
                                                 I

                       ,2
                   Figure 1.
         1,0
PARTICLE DIAMETER,
                                                            1,8
Electrostatic effects  in probe and partial blank  impactor.  All
flow rates 0.1 cfm.

-------
     2.5
3   2.0
tu
CD
CJ
     1,5
     1,0
                                                       T
T
T
                                              LINE  WITHOUT POINTS - PREDICTED FROM OPTICAL

                                                                      DATA

                                              LINE  WITH POINTS  -  MEASURED GRAVIMETRICALLY
                                           PARTICLE DIAMETER^M
   Figure 2.   Typical effect of electrostatic  probe losses on observed  particle concentrations.   Brink
              impactor with 3 foot, 1/4 inch I.D, probe and metal substrates at a flowrate of 0.1 cfm.

-------
counter running out of detection ability at about 0.3 ym.  The second curve
shows what we measured gravimetrically and was pretty much what we had
predicted from the particle counter data.  So it would appear that electro-
static effects can be important, at least with bare metal substrates.  The
effect was a factor of about 2%.  That is, the concentration as observed
with the neutralizer was 2% times what it was without the neutralizes at
the exit of stage 2 of the impactor.
SMITH:  Do you remember what the DSQ for that stage was?
MCCAIN:  I can't recall at the flow rate.  It's probably, at that flow
rate, about 2.5 or 3 microns.  I don't really know.
	:  Does this mean that electrostatically charged particles are collected
at higher or lower efficiency?
MCCAIN:  With the neutralizer we measured a high concentration downstream
of the stage so it was collecting particles; these particles are definitely
too small to be collected on that stage by impact!on.  They are below the
cut point, for that stage operating at that flow rate.  So they should have
been penetrating.  We were collecting them by electrostatic deposition.
HARRIS:  Essentially the electrostatic was enhancing the collection more
than it should.
MCCAIN:  As one might expect.
RAO:  The experiments I have done show the same effect.  If you don't
neutralize the charge on the particles you have higher collection.  Some
of the experiments of Dr. Lipkin at New York University Medical Center on
cyclones show the same thing.  If you have charge on the particles and you
don't neutralize them, the curve is shifted to the left and the slope
is lesser.
MCCAIN:  The surprising thing was the fact that glass fiber substrates didn't
show any effect at all.  It didn't seem to make any difference, which  is very
satisfying when we were using glass fiber substrates in stacks downstream
of precipitators for which we have no satisfactory charge neutralizer  to put
in the stack.  Whether a grease layer may do the same thing as glass  fibers,
we don't have any data.
                                  175

-------
        You would tend to think that,  because  it was  nonconductive,  the
glass fiber would allow an electrical  charge  to  build  up on  it and  it would
tend to repel the incoming particulate,  opposing the inertia! collection.
MCCAIN:  In another size which we didn't measure, there may  have  been an
effect like that.
	:  Where it would enhance collection on  the conductive  surface.
SPARKS:  Joe, on those tests you ran at  the pulp mill  with bare metal sub-
strates, did you use charge neutralization?
MCCAIN:  No.  We don't have anything we  can put  in that will stand  the
flue conditions.
SPARKS:  Then why do you use bare metal  substrates?
MCCAIN:  It sticks, the material  sticks  very  well (so  greases are not needed
for particle retention).  We also ran, just for  your  satisfaction,  Les,
both bare metal substrates in the University  of  Washington  impactors and
glass fiber substrates in Andersen impactors. We'll  see  what the results
look like.  They are not analyzed yet, but we did both and,  hopefully,  they'll
both look the same.
SPARKS:  You mentioned, why would you expect  them to  look the same if  this
is the behavior you see with and without charge  neutralizat.on.
SMITH:  This is very limited; this is an afternoon's  work that  you're
seeing here.
MCCAIN:  We were looking to see if there was  potential effect.   Yes, there
is a potential effect.
HARRIS:  Would you expect with a high moisture content that you find in a
pulp situation that you're going to have the electrostatic effect?
MCCAIN:  We don't have enough information.  We don't know.
SMITH:  To our knowledge there is not a suitable source available  in the
temperature range, Seymour has looked into that  and I b'elieve he said he gave
up, right?  Temporarily, anyway.

                                   176

-------
CALVERT:  Temporarily; we're still at it but it gets complicated if
the temperature goes very high.
HARRIS:  It's kind of hard to carry your own nuclear reactor around with you.
CALVERT:  Yes, the licensing problem gets sticky.  They keep taking us
round and round on that.
	:  Joe, what was your material on that?
MCCAIN;  The aerosol?   It was OOP.
	:  The unneutralized particles that you were catching, were they being
collected where you expected them to be or were some of them on the wall
or what?
MCCAIN:  We did not look to see where they were winding up.  They shouldn't
have been collected at  that stage, particularly here.  This size should not
have been going out on  that stage and (they) were without the neutralizer.
We didn't look further  downstream; we stopped and used that part of the
impactor with the particle counter.
	:  Yes, but what you're saying, you're putting a space charge into
something that is grounded with, you know, some of your smaller particle
sizes,  .4 of a micron,  may be diffused to the wall.
MCCAIN:  Oh, you mean where they went out?
	:  I'm curious where the particles that didn't get through are being
collected.
MCCAIN;  Keep in mind that this was one afternoon.
	:  The electrical  velocity should be low in comparison with the gas
velocity, where it should really just influence  it right at the impact!on
point.  This is my opinion.
HARRIS:  Yes, one thing, of course, this is with the Brink and the Brink
does have probably more of an eddy problem than  some of the others do,
just because in the design of those jets there is a lot of free space in
behind  them, they can fill up with eddies that you could have some diffusional
losses  with relatively  low air velocities.

                                  177

-------
        And those particles close to that surface would  tend  to  be  collected?
HARRIS:  Yes, when you clean up a Brink impactor one of the  things  that
you do is that you clean up that cone because it tends  to  get  coated.  Sey-
mour, do you have anything?  We didn't get a chance to  get through  all
of your problems, do you have any you want to kick around?
CALVERT:  Nothing that I have to get unburdened.  I think  that what we
found with a charge neutralizer was that the size distribution measured
without the charge neutralizer was bigger than measured with the charge
neutralizer.  Is that what you found?
MCCAIN:  That would be the trend of this ; you would be  taking  out small
particles on stages ahead of the stage in which they should  be deposited.
	:  Taking out particles you shouldn't be?
MCCAIN:  Right, we were collecting, then, too soon' in the Brink.
	:  What temperatures can you use the charge neutralizers  to, then.
You were talking about sampling at 600° at one time this morning.
MCCAIN:  Yes, what we were using when we used charge neutralizers was
polonium.  It won't really go above ambient—you wouldn't want to. put it
in the stack at all.  Krypton in a sealed source would probably take
stack temperatures, the problem is getting the licensing - getting
a configuration that is suitable and having it accepted by the AEC and
getting all the required AEC and State licenses.
HARRIS:  Somebody looked at nickel and the possibility of using it.
MCCAIN:  Nickel, though, is an alpha source and of very low energy, and
the slightest little dust layer on the source and it's going to be ineffec-
tive.  You've got to have something with reasonable energy to put it in the
stack, where you're almost certain to build up at least a monolayer of dust
on it.  Unless you've got a reasonable energy, the ions won't get out.
HARRIS:  One thing, I have brought over some copies of the reports that I
alluded to earlier.  It's there if anybody wants to pick  it up.  There are
some extra copies here of the last two of the reports that Southern put out.
There are a few others here that you are welcome to take the number off of
and see if you can get the information, I'm out of other  copies.
                                   178

-------
SMITH,:  One problem that seems to me to be really horrible and I  guess
Seymour could, it belongs to him, is sampling wet sources.  I guess  Seymour
and Les maybe could comment on where that stands.  I would hate to have
to sample the outlet of a scrubber or the inlet if you can't really
actually measure what's going in and what's  coming out.   Some of it's
being created in the scrubber and some of it is being evaporated. Where
does that technology stand?
CALVERT:  We use, generally we use a precutter that has a cut diameter,
depending on the sampling rate, of around 8 microns aerodynamic size.
Maybe a couple more, maybe a couple less.  We essentially, then,  throw
away the top part of the particle size distribution, say that we're  interes-
ted in particles smaller than 2 to 3 microns aerodynamic diameter.
Generally the penetration for high-efficiency scrubbers has dropped  off  on
to practically nothing at let's say 2 and a  half to 3 microns.  So that
we're saved by that fact, we can cut off the high end of  the size distri-
bution and not influence the computations that we're going to be making
later.
     We've run into, as I mentioned, some situations where the sampler was
literally deluged.  From what the fellows told me there was just water
sloshing all over the place and there was just no escaping it, so, essentially
we put a cylindrical shroud out as, it was either a beer  can or a welding
rod can, sort of prosaic.  Sort of this way  and then with holes punched
in the bottom so that there tended to be an  induced gas flow through it
and out through the downstream side and then the impactor sticking into  the
side of it.  And we just had to forget about being isokinetic, having the
probe pointed in the right direction.  There was just no  other way to do it.
We did some computations about what the bias would be at  a particle size
of 3 or 4 microns.  We concluded that it wouldn't  be too  much.  You have to
put up with that error.
SMITH:  How about evaporating particles that had nucleated water?
                                   179

-------
CALVERT:  We have a lot of trouble with filters loading up.   One of the
mysteries is why water will get out to the filter and  cause  a very high
pressure drop on the final filter and not appear on any of the stages  in  the
impactor.  We have this happen.  But it happens very frequently, so when
we get a bad one that way and if we don't want to heat the impactor, then
we have to use a separate filter and heat the final filter,  keep it dry.
MCCAIN:  Have you tried glass-wool plugs?
CALVERT:  No.
HARRIS:  Why don't you want to heat the impactor?
CALVERT:  Because we don't.  For instance, if we're in a situation where
we don't want to heat the impactor, we want to get the wet size or one we
did recently, we were concerned with coke oven emission, we  were concerned
that a significant part, fraction of the particles, might be organic matter
that was volatile.  We just plain didn't know in advance. So here we are
with hot gas going into the scrubber and cold, say 135° F gas coming out
of the scrubber, saturated.  So we wanted to have the inlet  sample the
same temperature as the outlet sample and what we did was pull our samples
out of stack, and pull the sample through a probe to the imapctor and have
enough cooling in there so that the impactor temperature at  the inlet
was approximately the same as the impactor temperature on the outlet.
HARRIS:  How do you ascribe your efficiencies, what you're actually sizing
on the outlet is not the particles, not primary particles, but the growth
particle, it's the water.  How can you then?
	:  Do you weigh your filters wet?
CALVERT:  Well, if we are, all right, that's another problem and if we want
to do that then we, what we have to do is determine a wet particle size.
Depends on how the scrubber is built, if it has a presaturater on it, quencher,
something of that sort and then the gas goes into the scrubber and we want
to separate the efficiency of the scrubber from the efficiency of the pre-
saturater.
                                   180

-------
     We'll sample the gas out of the presaturater and at the scrubber inlet
cold so that we get wet particle size, and we'll also get wet particle
size on the outlet.  If there ha,s been additional particle growth in the
scrubber due to condensation, we're stuck,  and  we run into these too, and
there we back off and do the same thing and measure the size in and the size
out dry and heat the impactors on both sizes.   But we're left with then,
the necessity to rationalize what we find by some type of theoretical  model.
HARRIS:  Is the heating of the impactor with the precutter on it to knock
oijt the large rain clouds?  Is that by itself sufficient to evaporate these
smaller than 5 micron particles or do you have  to put some kind of section
where you specifically add energy into that thing to evaporate before it
gets into the impactor?
CALVERT:  We've done it both ways.  There have  been some cases where
apparently we weren't getting enough heat transfer through the impactors
so we put a long section of, perhaps, something like 1-inch-diameter tubing
between the precutter and the impactor and put  a heater on that and then
also heated the impactor.
BYERS:  Does the Brink sampler that's designed  to hold deep dish cups below
the plates help with the problem?  I think that it was designed to sample
liquid droplets.
HARRIS:  The biggest problenfi I guess is whether you're trying to size the
droplets or whether you're trying to size the particles.
ENSOR:  A philosophical problem more than  ...
HARRIS:  Well, in .your case, if you're going after a model, if it's a dry
inlet scrubber,  then you've  got  to  size  both,  as Seymour said,  on  the inlet
and outlet.  If it's a wet presaturated scrubber, then you have to go
wet to evaluate the droplets.
CALVERT:  The problem isn't with retaining  the  drops on  the stages but, and,
as a matter of fact, if we have that kind of situation we're more  likely
to use glass fiber paper substrates, and they will blot  up the liquid.  But
the problem comes in in the liquid getting  through to the final filter  and

                                   181

-------
decreasing its porosity or permeability and  giving  very  high  pressure  drops,
so it will cause the, well, the sampling period to  be  cut  as  short  as  30
seconds.  We've had that happen.   There's still a little bit  of  sample in
there.
BYERS;  Could you separate that final  filter out of the  impactor and heat
it separately?
CALVERT:  Yes, that's what we do in that kind of situation.   The big
problem, the part of my notes that I didn't  get to, was  facing the  fact
that you didn't get a clean answer.  There are some problems  that
you just can't resolve and this is one of them. We've gone  round and
round on this dilemma—should we measure the particle  wet  or dry—and  there
is just no perfectly correct experimental way that  will  give  you the
information you need to validate the mathematical model  without  some
assumptions, some hypothesis tied in.
HARRIS;  The guys from GCA will be here tomorrow; we'll  see  if they have
horror stores about what goes on in baghouses that  the scrubber  and
ESP guys got today.
CALVERT;  This is what's called: facing the  fact that  you  can't  compensate
for everything, at least not in a reasonable way;  and  trying to  get a  dry
particle sample from a urea scrubber outlet; trying to get as-is particle
size distribution for sulfuric acid mist; trying to get impactor conditions
that are exactly the same as in-stack conditions.   We're hoping  that the
size selectivity of the particle control instrument will not cause the inlet
and outlet characteristics to have different composition and/or characteris-
tics.  It can be important if particle growth is important in your mathe-
matical model for scrubber penetration.  Or  not knowing, again,  whether
hydration or dehydration changes the morpholoty and/or subsequent suscepti-
bility to wetting.
      Error—you just can't have everything  in an absolutely unequivocal
way.
                                   182

-------
              NEW TECHNIQUES FOR PARTICLE SIZE MEASUREMENTS
    William B.  Kuykendal, Environmental Protection Agency,  IERL/RTP*
SPARKS:  The first speaker today is Bill Kuykendal of the Process
Measurements Branch and he is going to tell us all about weird and
wonderful things that are going to solve all of our problems.

KUYKENDAL:  I am in a group of process engineers who are interested in
developing control techniques for pollutants and emissions  from stationary
sources.  We are, therefore, interested in being able to evaluate the
fractional efficiency of particulate control devices and, hence, our interest
in particle sizing techniques.
     Because of this interest, we undertook a study to evaluate several
different devices developed by EPA and others to measure particle size
distributions.  The intent in our study was to determine how the instruments
compared with one another rather than an absolute determination of accuracy
for each device.  The devices and instruments which will be described are
listed  in Table 1.
     Figure 1 shows the Southern Research  Institute series cyclone used
in-stack.  In addition to size distribution information, our group feels
that determination of chemical composition as a function of size is
desirable.  The use of a set of cyclones offers several advantages in this
regard  over impactors.  First, and of prime importance, a large amount of
sample  can be collected for analysis.  Also, since substrates are not used,
the problems discussed in earlier papers are not encountered.  Three cyclones
are used which gives four size fractions,  one for each cyclone plus one for
the back-up filter.
     Figure 2 shows the PILLS-IV by Environmental Systems Corporation.  It
is an in-stack instrument with a purge box, a probe containing a gallium-

*Coauthored by Charles H. Gooding, Research Triangle  Institute
                                   183

-------
                          TABLE 1.  INSTRUMENTS
Instrument

Brink Impactor


Andersen Impactor


Series Cyclone


PILLS IV


In Stack Beta Impactor
Piezoelectric
Microbalance
Impactor
        Company

Monsanto Envirochem Inc.
St. Louis, Mo.

Andersen 2000
Atlanta, Ga.

Southern Research Institute
Birmingham, Ala.

Environmental Systems Corp.
Knoxville, Tenn.

GCA Corp.
Bedford, Mass.

Celesco, Inc.
Irvine, Ca.
  Status

Commercially
Available

Commercially
Available

R&D
Commercially
Available

R&D
Commercially
Available
arsenide laser,  the detection  optics,  and  the  processing electronics.
Figure 3 shows a schematic of the instrument.   The control  electronics
causes the laser to pulse.  The detection  optics then looks at two angles

in the forward scattering mode, ratios these angles, and then puts out the

size distribution which is a function of the intensity ratio.  That is then

output on a paper tape.  It takes about 5  minutes to process a sufficient

number of particles to compute the size distribution.

     Figure 4 shows the IBC-Celesco instrument.  It was principally developed

for ambient air monitoring.  It is a 10-stage cascade impactor that uses

piezoelectric microbalance technology for sensing the mass on each of the
stages of the impactor.  Since this device was developed for ambient air

sampling, it could only be operated for a  very brief period of time because

of the higher dust loading in the wind tunnel.  The sample enters a 3-way
valve.   Purge air (filtered room air) is pulled through the valve during

the nonsampling mode.   To sample, the valve is flipped over to the sampling

                                    184

-------
Figure 1.   Southern Research  Institute  series  cyclone.
                        185

-------
Figure 2.  PILLS-IV.
            136

-------
       LASER
       AND
       OPTICS
CONTROL
ELECTRONICS
                       COLLECTION DETECTORS
                       OPTICS     AMPLIFIERS
-SAMPLING
 VOLUME
             SENSOR HEAD
                                     SIZE
                             RATIO    ANALYSIS  READ-
                             CIRCUIT  CIRCUIT   OUT
                                                                   ptd)
                                                           -»- METER
                               CONTROL UNIT
                   Figure 3.  PILLS-IV schematic.
                               187

-------
Figure 4.   IBC-Celesco piezoelectric microbalance impactor.
                        188

-------
mode, and the particulate-laden gas impacts on the sensing crystal.  There
is a reference crystal behind the sensing crystal to compensate for changes
like temperature variation.  The gas then passes on out through the flow
control, flow metering, and blower.  The electronics ratio the two signals
              ^'
from the sensing crystal and the reference crystal, convert frequency to
voltage, and indicate the change in frequency . which then becomes directly
related to the mass of the particulate on each stage.  In this fashion,
we are able to get mass concentration as a function of size for each of
the ten sizing increments.
     Figure 5 shows the GCA beta impactor.  This is a device that is very
much in the research and development stage.  It is a beta-sensing impactor
which is located in-stack.  It requires a 20.3-cm (8-inch) port for insertion
into the stack.  There are seven elements in the cascade impactor.  Figure
6 is a schematic presentation of one of these seven sampling chambers.  The
flow of the particulate is down through the sizing orifice and onto the
impactor surface.  In this case a mylar tape is used.  The flow passes around
the tape cassette through the holes on down to the next stage.  A carbon-
14 beta source is located below the tape.  A reference detector gauges the
thickness of the tape.  Then as the particulate is collected, the detector
gauges the new thickness, which then is a measure of the particulate mass
per stage.  In this fashion we can get real-time, in-stack, continuous
measurement of the size distribution.
     Table 2 is a summary of the instruments considered, giving for each
the operating principle, size range, number of size  intervals, the approxi-
mate measurement concentration range,  instrument flow rate, data  rate,
and approximate cost.
Experimental Program and Evaluation
     Figure 7 is a photograph of the wind tunnel used in  the  experimental
program.  We can maintain accurate control over  the  dust  feed and the
volume of gas through it, as well as humidity and temperature.  We can vary
temperature up to 205°C (400°F).  The tunnel can also be doped with gases to
simulate stack conditions when that  is required.
                                   189

-------
vo
o

-------
Figure 6.  Schematic of one of the sampling chambers of
           the GCA beta impactor.
                         191

-------
                                            Table 2.   INSTRUMENT  SPECIFICATIONS
ro
Instrument


Operating
Principle
Size Range
(micro-
meters)
(unit
density)
Number of
Size inter-
vals
Mass Cone.
Range (g/m3)
Instrument
Flowrate
(1/min)
Data Rate
Approx.
Cost
Brink
Impactor


Manual Impactor
with cyclone
<0.4 to >14




9


0.1 to 6

1.4


manual
$2. 5K

Andersen
Impactor


Manual Impactor
with cyclone
<0.4 to >8.3




10


0.02 to 3

20


manual
$3K

Southern
Series
Cycl one



Manual cyclone
<0.7 to >3.5




4


0.1 to 25

20


manual
R&D

ESC
PILLS IV

Single Particle
Dual angle
light scattering
0.3 to 3.0




10


•3 C*
10*3 to 10°

NA


batch 5-15 min
$35K

6CA
Beta
Impactor

Impactor with
beta detection
of mass
0.3 to >6.5




7


0.3 to 20

9


real time
R&D

Celesco
Piezoelectric
Impactor
Impactor with
piezoelectric
detection of
mass
0.09 to >35




10


50x1 O"6 to 0.08

0.2


batch 10-15 min
$15K

      particles/cc

-------
Figure 7.  Wind tunnel
        193

-------
     Table 3 shows the wind tunnel  conditions  as  operated.  Two  particu-
late concentrations were used:   high (0.86 g/m )  and low (0.08 g/m3).
These were obtained by varying  the tunnel  air velocity and  the dust feed
rate.  In all cases we were using air at room temperature.

                    Table 3. WIND TUNNEL CONDITIONS

                               High Concentration        Low  Concentration

     Velocity (in/sec)                9                        27
     Temperature (a€)               27                        27
     Relative Humidity            50-55%                    50-55%
     Gas Composition               Mr                       Air
     Dust Feeder kg/hr               8.3                       2..3L
     Concentration (g/m3)            0.86                      O..Q8

     Since the purpose of this  study was to compare the various  techniques
available, it became necessary  to select a baseline methad.  Because of
the considerable experience with the Brink and Andersen impactors;, these
devices were selected.,  Figure a presents the data taken far reference
purposes.  The results from the Brink and Andersen impactor runs? at the
low concentratians; have been plotted cm a single curve on a lag/lag scale.
There is a certain amount of scatter in the data, but considering that data
were obtained with two different instruments, the agreement is goad.  The
curve was generated by a least squares curve fit.  At the high concentration,
a similar curve was obtained and is shown in Figure 9.  The two curves
are extraordinarily close in shape.  This convinced us that our dust feeder
was giving a reproducible size distribution.  On each of the  remaining
plots, the solid line is the least squares curve fit for the  Brink and
Andersen impactors and the points are the data for the sizing device
being evaluated.  This allows a direct comparison with the impactor data.
                                   194

-------
    10'
    10*
    10'
a>
o
•o
    10'
                             BRINK DATA AVG.
                           A ANDERSEN DATA AVG.
                           I DATA RANGE
     0.1
1.0            10
 DIAMETER, fj.m
100
  Figure  8.   Brink  and Andersen  impactor  data
              (low concentration).
                       195

-------
    10'
    10
o
    10
    I0";
    10'
                           o BRINK DATA AVG.

                           A ANDERSEN DATA AVG

                           I DATA RANGE
     0.!
10            10

 DIAMETER,
100
 Figure 9.  Brink  and Andersen impactor data
             (high concentration).
                     196

-------
     The data from the Southern Research Institute series cyclbne are pre-
sented in Figures 10 and 11 for the low and high dust loadings, respectively.
The agreement with the Brink and Andersen impactor data is very good.  In
addition, this device has the advantage that it can collect a large quantity
of sample for later analysis without overloading.
     Figure 12 gives the information from the  IBC-Celesco piezoelectric
impactor.  Since this instrument was developed for ambient air sampling and
was operated in an extractive fashion without benefit of a dilution system,
we were only able to test  it at the low concentration and then only for a
duration of approximately  7 seconds.  The instrument has ten sizing
intervals.  We saw nothing in the first three sizing intervals, and we sus-
pect that this was because of deposition in the probe used to extract the
sample.  This is borne out in the fourth size  interval which reads low
with respect to the three  subsequent stages indicating that a portion of
this particulate is also separated in the probe.  Stages five, six, and
seven all collected particulate.  The shape of the curve is similar to the
baseline curve with the values consistently high, indicating a possible
calibration error in the Celesco instrument.  The last three sizing
intervals did not detect any particulate.  When the baseline data were used
to calculate the expected  loading on these stages, it resulted in mass
concentrations that were below the detectable  limit of the instrument.
     Figure 13 gives the low concentration information using the PILLS
light scattering instrument.  It is obvious tnat the PILLS does not pro-
duce the same type of response as the inertial devices.  The data at
the high concentration with the PILLS, shown in Figure 14, bear this out
Superimposing the PILLS curve at the high concentration on the results
at the low concentration shows quite good agreement, indicating that the
response of the instrument appears to be
     Figures 15 and 16 show data from the GCA  in-stack beta instrument.
The data seem to follow the pattern of the instruments evaluated earlier.
                                   197

-------
      10'
      icr
      10
  i
      io
      10=
      10"
       O.I
1.0
10
100
                       DIAMETER,
Figure 10.  Comparison  of Series Cyclone data  with
            impactor  data (low concentration).
                     198

-------
        10
       10'
   o

   S
       id2
       \0'
         0.1
                               • AVERAGE CYCLONE
                                   DATA

                               I DATA RANGE
1.0           10

  DIAMETER, fj.m
Figure 11.   Comparison of Series Cyclone data with
             impactor data (high  concentration).
                       199

-------
       10'
       IOC
       10"
    o»
    o
    •a    -2
        10
10'
        IO"4
         OJ
                             • AVERAGE DATA POINT

                             I DATA RANGE
                1.0           10

                  DIAMETER, /urn
100
Figure 12.   Comparison of piezoelectric microbalance
             impactor data with  impactor (low concentration)
                           200

-------
    10'
    10'
    10
      -2
Q

o>



5
2
•o
    10"
    10
      AVERAGE PILLS DATA



    I DATA RANGE
    I0's

      0.1
1.0            10



  DIAMETER, p.m
100
 Figure 13.   Comparison  of PILLS-IV data with

              impactor data (low concentration)
                      201

-------
     10
     IOC
     10
  o


  S
  5
  •o
     10
     10
      -3
     lO'4
        • AVERAGE PILLS DATA


        I DATA RANGE      _
       O.I
1.0            10


  DIAMETER, fj.m
too
Figure 14.   Comparison of  PILLS-IV data with Brink

             impactor data  (high concentration).
                       202

-------
       10'
       10"
       10
      10
        ,-3
      10'
        0.1
                           o  AVERAGE OF BRINK RUNS

                           •  AVERAGE  OF GCA RUNS

                           I  DATA RANGE
1.0           10


 DIAMETER, /j.m
100
Figure 15.  Comparison of in-stack Beta impactor data
            with  Brink impactor data (concentration =
            0.24  g/m3).
                       203

-------
      10'
      IOC
      10'
      10
        -2
      10'
                            AVERAGE OF BRINK RUNS

                            AVERAGE OF GCA RUNS

                            DATA RANGE
      10"
0.1
                      1.0           10

                        DIAMETER,
100
Figure 16.  Comparison of in-stack Beta  .impactor data
            with Brink impactor  data  (concentration =
            0.86 g/m3).
                      204

-------
Note that the high concentration in these tests is 0.86 g/m , and the low
concentration is 0.24 g/m3.  The sensitivity of the instrument would not
allow us to test it at the earlier conditions.  Upon examining the instru-
ment, we found a considerable amount of particulate matter on the wall
in the upper stages.  This would account, at least in part, for the low
readings on the first stage.  In the runs at the higher concentrations
with this in-stack device, the agreement with the impactor data becomes
better.
Summary
     I believe we can fairly say that  all the instruments are useful in
measuring particle size.  The devices  that  employ inertia! size separation
all seem to agree reasonably well with one  another, and therefore can be
used in exchanges of data with each other.   In the case of the PILLS
instrument, which is based on single particle optical light scattering,
it would not be recommended that  it be used in direct comparisons with data
obtained using  inertia!  separators.  I should point out that
of the two different principles of operation, we are making no distinction
as to which is  right; one merely responds to a different  set of para-
meters than the other.     „
                                   205

-------
DISCUSSION
        Series cyclones, were those used in the vertical or horizontal
position?
KUYKENDAL:  They were used vertically.'
BYERS:  This PILLS-IV instrument.  That is a single particle counter?
KUYKENDAL:  It is a single particle counter, that's right.  We were relatively
sure that we were counting single particles.  It uses a gallium-arsenide
diode laser, which pulses at 1  kilohertz.   We were counting particles
only about once every hundred or thousand pulses so we were relatively sure
that there were only single particles in the viewing volume.

OLIN:  You seem to be using the Brink and the Andersen as a standard.
I was wondering if you really felt that to be true.  With all  the problems
we heard about yesterday, I was wondering if maybe you would prefer
to believe the PILLS data as opposed to the impactor data.
KUYKENDAL:  From a convenience standpoint only, I would prefer to believe
the PILLS data to be correct.  The PILLS is a much simpler instrument to  •
use.  However, the accuracy of the PILLS is subject to question.  Keep in
mind that our tests were conducted in a wind tunnel at ambient conditions.
Therefore, there were no acid gases to react with the substrate.  Also,
we used coated substrates which we feel reduced the particle reentrainment
problem.  In general then, we feel that the impactor data accurately
characterizes the aerodynamic diameter.
	:  What particular calibration did you use?  Did you use Southern's
calibration?
KUYKENDAL:  We did not make the correction for the smaller sizes.
MCCAIN:  Well, our calibration of those impactors would tend to  increase
the discrepancy between the PILLS and the others.  And then there's the
question of which one you believe.
                                  206

-------
SMITH:  Those are differential velocities shown; and it turns out that  if
you make the correction it might not make so much difference as you think,
because it would be dividing the DM by a bigger log D; so you tend to just
move down that differential curve.  It's really not going to just change
your size distribution by a tremendous amount.
	:  Even though your D5Q on the last stages is way less than the
calibration that the manufacturers say it is?
SMITH:  The error will not be as much as you might think on the
differential plot, because you don't move those points horizontally on  a
differential plot—they'll move over and down.
MCCAIN:  We have a little more information on the PILLS too, running
at Fluidyne, and there we had other particle counters plus impactors to
                                ,
compare it with.  The PILLS showed the same disagreement, it looked almost
exactly the same as the data that you obtained.
KUYKENDAL:'  It seems ,to be that the PILLS is not corresponding, not con-
forming to the same particle size as the others.  Whether that can be
corrected by calibration or not is not clear.   It is indicating a change
in the right direction; maybe that's sufficient.  I don't know.
	:  Has anybody looked at the flow field in that?
KUYKENDAL:  That's something we plan to do, but it hasn't been done.
MCCAIN:  From the Fluidyne data, where we ran it at two different locations,
one with a high velocity gas stream and one with a low velocity gas stream,
the disagreement was much worse in the low velocity location than in the
high  velocity location.  That is the sense of things that you would expect
if there is a sweeping of view volume with purge air.
KUYKENDAL:  There appears to be another thing that leads us  to that con-
clusion.  The data rate that we would anticipate with the particle  size
concentrations that we think that we should have in that tunnel are much
lower than the PILLS actually sees.  It's only  counting one  in a  thousand.
The high concentration count should be about one in every 10.
                                   207

-------
 BYERS:   I  think we ought to remember that  we  should  not  expect the same
 thing from the PILLS-IV and impactors--they're  not measuring  the  same
 parameters.
 KUYKENDAL:  That's right.   I want  to make  that  point a  little more
t  —-—.-,,-.,  , _ ._. ,
 emphatically.   We didn't go into this test with the  idea of establishing  truth
 and then comparing the other instruments  to that truth.   We went  into the
 tests with a great deal  of background in  the  Andersen and Brink impactors.
 We wanted  to see how the instruments that  we  had available would  compare.
 If, for example, you could use the PILLS  downstream  of  a control  device
 and the Brink upstream and just compare the data? Well, I can safely
 say now that you can't.   That is not to say that if  you  would take the PILLS
 and go upstream and downstream that that  fractional  efficiency would  not
 have meaning.   We are not prepared to make judgements at this time.
 BYERS:   Joe, could you quantify those high and  low velocities?
 MCCAIN:  It was several  meters and about  a meter per second;  I can't  recall
 the exact  numbers.  I believe that the design value  for the PILLS instrument
 was two meters per second or greater gas  velocity.   In  the low velocity
 case it was definitely under the design value for the PILLS.   I think that
 we were on the order of 8 in high  velocity and  about 1  in the low.
 	:   What was the wind tunnel  velocity?
 KUYKENDAL:  In our tests it was higher than that; it was 8.7.
 SPARKS:  We've just got time for one more  question Bill.
 SHE:   Bill,  how does your flow diameter compare with your total volume?  May-
 be the beam is seeing only part of the particle?
 KUYKENDAL:  The volume is cylindrical and  I don't recall exact dimensions,
 but I expect I can check and let you know.
 SPARKS: Thank you Bill.
                                   208

-------
                   ADVANCED  PARTICLE  SIZING TECHNIQUES
                     Pedro  Lilienfeld,  GCA  Corporation
SPARKS:  As Bill  said,  GCA  has  been working on  some advanced methods of
particle sizing and  they  have a representative  here to talk about the work
they've been doing.

LILIENFELD:  We have a  few  areas in which we are working in the field and
related to that which is  of interest  here.  The description that Bill Kuy-
kendal just gave  with the tests bears upon one  of these developments.  The
instrument that was  shown previously  is  the result of a R&D program for EPA,
in which we tried to develop an instrument capable of performing a real time
size distribution measurement in emission sources, and the objective was to
build an instrument  that  could  be inserted into the stream and perform the
measurement in the stack.
     The general  philosophy now is that  this is probably the only reasonable  way
to perform any such  measurements of size distribution and that extraction of
samples for that  purpose  is probably  unacceptable from all points of view.  So
the general approach we used for this device was to separate the particles col-
lected at each stage by means of beta absorption as shown on one of the slides
you saw before.
     The original  objective of  the program was  to be able to operate at tem-
peratures typical  of stacks—perhaps  up  to 500°F.  The design of the device
would be capable  of  withstanding that temperature.  During'the development
of the instrument, it became apparent that there was one element that mitigated
against being able to operate at elevated temperatures and that was the beta
detector.  The technology of beta detectors is  somewhat of an alchemy at this
point, especially as it refers  to geiger tubes.  Several manufacturers had
assured us that there would be  absolutely no problem in producing detectors
that would operate up to  500°F.
     We obtained  such detectors and tested them up to those temperatures;
they all failed within  one  hour.  So, we pursued the development, but with
the awareness that we would not be able  to obtain a detector very easily
that would stand  the temperature of interest, and the device that was tested

                                   209

-------
at EPA was essentially a room temperature device,  perhaps  150-160°F.   It
served to demonstrate the principle, but the objective of  operating in a
stack had not been achieved.  We are pursuing, at  present, the idea of a
so-called electronically quenched geiger tube beta detector,  and that approach
does show promise and perhaps that is going to be  the solution.
     As mentioned during Bill Kuykendal's talk, the substrate material which
we used for these tests was mylar, but we have the intention  of using
a very fine stainless steel foil for high temperature tests.   Of course,
the mylar wouldn't take temperatures above about 200°F.
     There were some interesting problems in the development  of that device,
and I'll just mention a few of them.  One is the problem of coating the
foil.  This is an advancing foil, if that wasn't made clear before.  It
advances continuously and thus, the collection occurs in the  form of a
streak or line, and of course, we had to coat the  substrate with some
adhesive to be able to retain the particles.  The  problems of how to coat
this foil and maintain the adhesive on while it is being unrolled and
rolled over were things which we had to resolve and are more  of a mechanical
nature than anything else.  We also had the intention of having the foil
go back and forth, shifting sideways each time so  that we could make a
better utilization of the total area of the foil.   We haven't gone this far
in the present prototype, but we have the provisions to go the next step.
     The signal from each of the seven impaction detection channels consists
of two pulse trains from the two beta detectors at each stage, one the
reference and the other one measuring the collected mass.   These pulse
trains are put into a frequency-to-voltage converter, each one of them
and the outputs of these into a log-ratio module,  the output of which
is directly proportional to the mass concentration as measured by the
device because of the moving nature of the substrate.  Because of the
different nozzle sizes required to impact different sizes at each stage,
each of the streaks of collection was different in size and we had to adjust
for that in terms of the detection geometry.  We used a flow control  that
was based on running the last stage critical.  We  haven't evaluated entirely
to what extent errors result from bounce off of particles at that  last
stage, of course, considering that by then the particles will only be the

                                   210

-------
very smallest ones, a few tenths of a micrometer, and. that bounce off under
those conditions may be minimal although the velocity is high.
     The system is built such that it can be inserted into a stack in a
horizontal position and then lowered down by 90° once the device is inside.
We felt that the use of any kind of bends or 90° nozzles or goosenecks was
defeating the purpose of measuring the particulate size distribution
representatively.  I think the plot that Mr. Cass later on will present
will bear out the problems associated with such 90° bends and gooseneck
nozzles.
     We are, at present, also  in the course of developing another device
for another group at EPA.  That .is what is possibly called the cheap ver-
sion of that which was described previously.  It is again the intention
to measure the size distribution in real time and use inertia! methods
to separate the particles.  The device that we are developing at present
is a virtual impactor, a cascade virtual impactor, where we would use some
means to detect or to evaluate the mass collected at each of these stages,
perhaps, or possibly, a simple and very straightforward means of detecting.
The three methods of detection,of the collected, or the separated fractions
at each stage are:
     a)  An optical method, by which we would measure the light
         transmission or extinction through each of the stagna-
         tion volumes of the virtual impactor.
     b)  Use of the electrical contact method on which the IKOR
         instrument is based,  by extracting from each of the
         virtual impactor stages a so-called token flow and
         making that flow pass over a contact electric detector
         and deriving the signal from that and taking it to the
         outside.
I should clarify that the optical method that I mentioned before is based
on the use of optical fiber bundles so that the sources and detectors would
be on the outside of the chamber rather than on the inside exposed to the
high temperature.  Fiber-optic bundles can be made now in a very compact
manner and this technique could be applied.
     c)  The third approach that appears at present as the most
         promising because of  its simplicity and reliability  is
         the measurement of the  pressure  drop developed  across a
         filter collecting the mass of particulate at each stage
         with the  token  flow,  as  it  is  called,  as  the  driving flow
         for that  pressure  drop.
                                   211

-------
That token flow is typically on the order of 1  to  10  percent of the  sample
flow and creates a small, low pressure drop across  the  filter  at each of
the virtual impactor stages.  As a  result of having this  very  low  flow
rate, these filters do not tend to  clog and the collection  occurs  as a
cake, such that the pressure drop versus mass per  unit  area is a very
linear function over a very, very wide range.  By  differentiating  the slope
of this pressure drop increase across the filter,  one can determine  the
concentration at any time.  By simply measuring the pressure drop  increase
between two instants in time, one can determine what  the  average concen-
tration was over that period.
     This method appears extremely  promising as I  said  because it  is
very simple and all that is required are some pressure  taps carried  to the
outside of the stack and whatever pressure sensing or display  devices on
the outside, magnehelics or anything that one desires.  And such an
instrument can then be calibrated for the particular  dust and  the  size
characteristics of each stage.  Our first indications are,  that because
of the size separation which occurs with the virtual  impactor, the sensi-
tivity to different types of particulates in terms of their pressure drop
effects are lessened.  That is, it  is more a size  function  than a  composi-
tion function.  Of course, that will only apply to dry  particles.   When
you get to liquid ones, that is going to change radically.
     I have two viewgraphs, Figure  1 shows a schematic  representation
of a three-stage device as described above.  The gas  inlet  is  at  the top
and the virtual impactor consists of a conical stagnation volume.   The
first stage has its filter and we measure the pressure  drop across it.   The
second stage is similar, and then the third stage, in this  particular case,
is a total filter.  This device happens to have only three stages, but
the technqiue can be extended, of course, to any number of  them.
     The next slide shows a typical run performed measuring just the pres-
sure drop on two of these stages.  We determined the mass collected on the
filter at various intervals and it appeared as if the pressure drop increased
linearly with the mass over this range.  We're talking  about  concentrations
on the order of somewhere between  .05 and 0.1 grams  per  cubic meter.  This
particular run was performed with Arizona road dust.

                                       212

-------
                              GAS INLET
  NOZZLE
                                     ORIFICE
           BACKUP FILTER
                                                  TOKEN FLOW
                                                 TOKEN FLOW
                            i- MAIN GAS FLOW
Figure 1.   Schematic  representation of  three-stage device.
                                 213

-------
S-


cO
3


0)
JC
o
c
0.9
0.8
0.7
0.6
0.5
0.4
0.3
 0.2
 0.1
                       I    I
                                                   .03
                                                           AP,
                       _i	»_
    Q   4   8  12  16  20  24 28   32   36   40  44  48  52  56  60

                                Time,  minutes



                     Figure 2.  Data for  test 23.
                                214

-------
     This particular development is of interest because it may afford a
technique for sizing in real time with a minimum amount of effort and cost
and yet relying on an inertia! separation technique, which most probably
is more acceptable than most other approaches.
     We have explored some of the problems that one faces in the design of
virtual impactors, the main ones being losses at the nozzle exits and at
the inlet of the stagnation volume.  There are critical dimensions that
one must maintain to minimize the losses and also the token flow itself has
a very significant role in the degree to which the losses are important.
We are in the process of optimizing this virtual impactor from all points
of view, including the token flow and we are also in the process of
writing a computer program that will take into account the effect of the
token flow on the size distribution measurements.  It is obvious that, if
the token flow is more than about 1 percent of the main flow, the size
distribution as measured by the device will no longer be rigorously that
of the aerosol that one is trying to measure because a certain fraction is
collected on each of the stages as a result of the token flow and not as
a result of the inertia! properties of the aerosol.  But one can perform
a mathematical inversion of the observed size distribution to come back
to one with which the measurement has been made.  We are in the process
of developing a program, which, if it is rigorous must take into account
the true shape of the cuts at each of the stages.
     So far, we have found for the three stage device that I showed before
that the losses can be brought down to the order of 5 to 6 percent by mass
with respect to an independent probe.  That is probably fairly acceptable
for most applications.  Another advantage of the virtual impactor as pre-
sented here is that, in addition to the pressure drop information, there
is the mass information obtained on the filters themselves which can be
weighed after the test has been completed.  The amount of material that one
can collect is very significant, it is not limited by bounce off and reen-
trainment as is typical in most jet-to-plate type impactors.
     I want to mention one more area that we have been working which
bears upon impaction, and that is work on the coating material for impact!on
substrates if a jet-to-plate device is used.  This interest comes from our
commercial instrumentation for the measurement of respirable dust in
industrial environments.  We found that the mechanism by which particles  are
                              215

-------
collected on wetted or adhesive substrates is far more complex than was
generally assumed.  If one calculates the depth of penetration of a typical
particle into a viscous substrate, one finds that, let's say for a 10
micrometer particle, the penetration is much less than the diameter of the
particle itself.  This applies to typical velocities, even up to sonic
velocities and for typical viscosities of the coating.  Consequently,
one must conclude immediately that after a monolayer has been deposited,
there should be no further action of the coating to retain the particles.
But this is really not the case in practice.  So we made a number of
observations and measurements and determined that, depending on the type
of coating one uses, there is a capillary type motion that tends to coat
the particles immediately as they are being collected; that is, the upper
surface of the collected layer is rewetted by capillary migration and thus
provides an adhesive layer for further particles to be collected on.
     The critical condition seems to be that if the rate of arrival,
of collection by impaction of particles on the substrate exceeds the rate
of capillary migration, then particle reentrainment and loss ensues
immediately.  However, if the rate of collection is less than the coating
rate, then the particles tend to collect without significant losses.  The
problem here is obviously one of compromising between a coating that has low
viscosity to permit this capillary migration to occur rapidly and the
opposing requirement that it should have high viscosity to prevent cratering
as a result of the air jet impacting on that coating, i.e., the coating
should not be spread away from under the jet as a result of the impact of
the air.
     One has to deal with a number of complex phenomena and I don't want to
go to deeply into these.  The treatment of this subject involves the study of
Bingham fluids, non-Newtonian flows, etc.  But we have found that an interest-
ing effect results from mixing a fairly fluid adhesive such as petroleum
jelly and paraffin oil with a small amount of dry particulates, dry  dust.
This produces a rather unique mixture which exhibits a low viscosity in
terms of capillary migration and a high viscosity on a macroscopic  basis  to
prevent the cratering as a result of the jet impingement.  This is work
which is probably in some way relevant to all jet-to-plate impactors, and I
think that further research should be done in that area.
     I think that I have talked enough.  Yes?
                                   216

-------
DISCUSSION

RAO:  When you change the filter on the virtual impactors, how does the
calibration affect you?  Do you have to calibrate again?
LILIENFELD:  You mean depending on what filter medium one uses?  Or is
that the question you asked?
RAO:  Yes, or for the same filter media.  If you change it, does the cali-
bration change?
LILIENFELD:  The calibration is essentially a function of the buildup of
the particles.  They build up as a cake, so the filter material itself,
or  the filter itself, plays a very secondary role.  All it does is give you
an  initial pressure drop before the particles are collected, but that
initial pressure drop is of no interest, it is the increase above that
pressure drop that we're looking for.  So the material of the filter
has little relevance unless it is the type of filter that would prevent
the formation of the cake.  I don't know if that is the answer to your
question.
OLIN:  What concentration ratio were you able to achieve with the virtual
impactor?
LILIENFELD;  Concentration ratio?
OLIN:  The ratio of the flow rate that goes through the receiver tube to
the flow rate around.
LILIENFELD:  The token flow fraction with respect to the total flow.
OLIN;  Yes, do you call that the token flow?
LILIENFELD:  No, the token flow is that secondary flow one extracts through
the stagnation volume.  We have explored fractions between 1 and about 12
percent.  We have found that the losses seem to decrease, the nozzle losses
seem to decrease as you increase the token flow.
OLIN:  That's been the observation of LBL and ERC.  One to seven is what
they claim.
LILIENFELD:  Yes, I know they have.  Well, there is apparently not an  optimal
ratio.   Because so far it is a monotonic behavior.  That is, we increase the
                                  217

-------
 token flow and we see decreasing losses.  However, at some point, you have
 to compromise.  You have to stop because the errors associated with collecting
 particulates resulting from the token flow then become too significant.
 OLIN:   I have one more question.  How often would you have to change the
 filter  in the token flow?
 LILIENFELD:  Well, that depends, of course, very much on the concentration
 and  it  depends on the dimensions of the filter and it depends on the token
 flow that you end up having.  We calculated that, for concentrations on
 the  order of 10 grams per cubic meter, for typical situations you could
 run  the filter for perhaps one or two hours without having to change it.
 If you  make the filter larger than a few centimeters in diameter, per-
 haps you can run 8 to 10 hours.
 OLIN: .In an automatic system you would have to have some kind of a tape
 feed.
 LILIENFELD:  Well, that depends on how often you do want to change it.  If
 you're  concerned about running let's say for a whole day, without changing
 anything, you can do that without advancing anything.  It may be easier
 to change the filters once a day then to have an advance mechanism which
 is complex and can break down.
 	:  If you change filters, using the same type of filter, but put a
 new  filter on, is the calibration mass sensitivity calibration the same?
 LILIENFELD:  Yes, that is true.
 	:  So any variations in the filter have no effect.
 LILIENFELD:  No, it is a totally negligible effect.  As a matter of fact,
 we have tried different types of filter and have found very little effect
 on the  average.  As long as the filter does not have significant depth, to
 the  degree that particles would be collected by two different mechanisms,
 first by penetration into the filter, and then by the formation of a  cake.
 Then you would see a discontinuity in the curve of the pressure drop  versus
 collection.  But if you use any type of tight filter, Teflon filters  for
 example would be applicable in this case for high temperatures, you don't
 have to worry.
	:   So you use membrane type filters?
                                   218

-------
LILIENFELD:  We also use the glass fiber filters, and in that case,  you  may
have a very short initial difference in the behavior when the particles
are penetrating into the fibers.  As soon as the cake forms, then you  get
back to the original calibration, which then is the same as the membrane
filter.
ENSOR:  How well do you need to know your pressure drop to obtain a
reasonable resolution in your concentration?
LILIENFELD:  That depends on a number of factors.  It depends on the con-
centration.  If the concentration is high, the rate of change of pressure
drop is high too.
ENSOR:  Well, I understand that.  What range are you talking about?
LILIENFELD:  You're talking typically in the range of an inch of water
per hour, or so.
SMITH:  What does the stage collection efficiency curve look like for
a virtual impactor?
LILIENFELD:  It is very similar to a jet-to-plate impactor.  It is slightly
less sharp and has a rather different impaction parameter.  It's shifted
for the same dimensions, it is shifted with respect to that which is
obtained for the jet-to-plate impactor.
	:  is it necessary to introduce your token flow as a part of the  main
gas stream or can you introduce it from the outside so you don't tend  to
extract part of your sample stream?
LILIENFELD:  It appears that it is necessary to extract it as a fraction
of the total flow, because it facilitates and provides a general drift
of the particles into the stagnation volume.
	:  I was thinking in terms of an annulus that would have the air
introduced there to give you the same flow.
LILIENFELD:  Well, it may be possible to come up with a geometry with
comparable effects.  I don't really know.
SPARKS:  Thanks, Pedro.

                                  219

-------
               LIGHT-SCATTERING PARTICLE SIZING TECHNIQUES
         Dr. C. Y. She, Dept.  of Physics, Colorado State  University
SPARKS:  Colorado State University has been working  on a  grant  on  optical
techniques for particle sizing, and Dr.  She of Colorado State University
is here to talk about how that works.

SHE:  I would like to say, first of all, that I have enjoyed this  very
successful conference very much.  Very informative indeed!   I thought
maybe in addition to a brief progress  report on the  six-month old  pro-
gram at Colorado State, I would comment on the light-scattering technique
in general and share with you my opinion of the potential  of this  technique.
     When I asked myself what one would like in particle  sizing, and being
an inexperienced worker in this field, I decided (Figure  1)  that perhaps
one would like to do it in real time and be able to  measure  the particle
size without the knowledge of other properties, such as the  refractive
index in the optical case.  If possible, one would like to be able to
process a lot of particles, in other words, to handle a high concentration
of particles.  Fourthly, one would like to be able to size the  particles
in a wide range, mainly from submicron to 10 micron, if that is possible.
     One can think of light-scattering techniques simply  as  follows.   One
should have a light source, maybe a laser in this case, and  focus  it down
to a volume V.  Now, send the particles through this focal volume  and
you will see a current pulse on the detector.  Hopefully, by measuring the
height of the pulse you learn something about the particle size.
     In fact, all of the commercially available devices,  besides the  PILLS
that Dr. Kuykendal talked about today, use a single detector arrangement
like that.  Of course, the height of the current pulse you detect depends
not only on the particle size, but also on the refractive index.  There-
fore, there may be some ambiguity with this kind of arrangement.  One way
to get around this problem is to use two detectors as Dr. Kuykendal has
shown, and there you take the current ratio and it turns  out that the current
ratio in the forward direction is more or less independent of refractive
                                  220

-------
              I.   IN REAL-TIME
    II.   INDEPENDENT OF REFRACTIVE INDEX
        III.  WITH HIGH CONCENTRATION
       IV.  OVER A WIDE RANGE OF SIZES

Light scattering satisfies (I) better than other
techniques
The intensity-ratio technique, in particular,
satisfies both (I) and (II); it has potential  to
be better in (III) and (IV).
     Figure 1.  Requirements for sizing.
                     221

-------
index.  One thing you want to make sure of is  that you always  have  single
particle detection.  You want to avoid multiple particles.   You can see,
if you have a concentration n, the number of particles in the  focal volume
on the average would be N, which equals n times the viewing volume  V.   If
you want to avoid multiple particle scattering, you would limit IT to
much less than one, thereby restricting the detectable concentration to be
less than the inverse of the focal volume you can achieve.
     If you want to be able to exceed this limit of detectable concentration,
you have to come up with a technique that will allow you to detect  multiple
particles.  We have some ideas along these lines, and I'll  be  talking
about them.
     The range of sizable particles is limited by the theory of scattering.
In general, there is not much one can do, but I will also breifly talk
about some methods by which perhaps one can expand that range  a little more.
     The second viewgraph (Figure 2) is a schematic that looks quite
similar to the one Bill [Kuykendal] used.  You have two detectors here
at different angles; you amplify the current pulses, take the  ratio by
dividing, and then send that divided pulse into a pulse height analyzer
or something like it.  As a result, you will get a particle size distri-
bution as shown.  On the X-axis you have current ratio, which  indicates
the size of the particle, and on the vertical  axis you have the number of
counts corresponding to a particular pulse height.  A curve like that,
with proper calibration,  will give you directly the particle size distribution
in real time.
     The two-angle intensity ratio technique was first pointed out  by Hodkin-
son in a paper in Applied Optics around the early 60's.  He pointed out that
if you take the ratio, the result will be independent of refractive index.
The intensity ratio is plotted here versus particle size (Figure 3).  This
parameter alpha, is the ratio of the circumference, ir times the diameter
of the particle, divided by the wavelength you use.  The range of  particle
sizes that you can measure depends on the two angles chosen.  If you choose,
for example, the 10° and 5° angles in the forward direction, you are dealing
with the curve marked 10°:5° on Figure 3.  The other abcissa scale is
                                  222

-------
                 Optical   Module
          i --- ~ -- * --- • -------- T   i —
           Two-angle Ratio  Mode  Shown       j   I
ro
ro
CO
Electronic    Unit
                                                               I,(t)       Integrator
                 Correlofor
                    or
                Multi-channel
               Analyser  (MCA)
                    or
                Probability
                 Analyser
                            Figure 2.   Schematic of  an optical particle  counter.

-------
            0.40
            0.33
Particle Diameter, ym
    1.21             2.42
    0.98             1,96
     T
                                         30^10°
                                         20°:!0°
                                         20°; 15°
                                         I 5°= 10°
                                         I00=50
         O    2
Wavelength (x)
 •«-  0.6328 pm
    0.5140 pin...
                                 16    18
Figure 3.  Intensity ratio versus  particle  diameter.   [After
           Hodkinson].
                              224

-------
particle diameter at particular wavelengths.  This corresponds to 0.4
microns, for example, if you use a Helium-Neon laser at X=0.6328 pm,  to
say something like 3 microns.  If you want to extend the range to larger
particles, you have to be able to detect scattering at smaller angles.
If you want to size smaller particles, you want to have a technique to
senstively measure the small changes in the intensity ratios and that,
of course, is not very easy.
     This viewgraph (Figure 4) is taken from a paper by Kreikebaum and
Shofner.  They designed the PILLS IV.  They plot several curves similar
to the one that I showed; except in this case, they take into account the
complex part of the refractive index.  Of course, for most particles  of
interest, you are dealing with the complex refractive indices, and they
have shown, more or less, that the ratio technique is also insensitive  to
that.  In fact, if your particles are somewhat glossy and have complex
indices, you can get a smoother curve than those corresponding to particles
with real refractive index.  Looking at this sort of curve, you would con-
clude that the accuracy of particle sizing by this technique is around  10
percent independent of the knowledge of refractive index.
ENSOR:  What's the vertical axis on that graph?
SHE:  The scattered intensity ratio.
     This viewgraph (Figure 5) summarizes the achievement up to this  point.
As I said, Hodkinson was the one who pointed out this technique theoretically,
and in fact, a fellow by the name of Gravatt at NBS has done considerable
work to put this idea into practice over the last few years.  He was  able
to size, by using different lasers, particles from 0.2 to 4 microns.  He was
using 10° and 5° as compared to the  PILLS IV which uses 14° and 7°.  The
focal volume Gravatt selected was 1.5 x 10   cc.  An improvement, you
see, of the PILLS IV over Gravatt's is basically this.  They think that they
can work with a smaller focal volume like 2 x 10"  cc, to size higher
concentrations.  The concentration that the PILLS IV claims to be able
to handle is 106 per cc, but Gravatt, working on a research program at  that
time, did not try to make a very small focal volume.  He could count only
104 particles per cc.  I want to point out that if you multiply the

                                   225

-------
ro
ro
       (f)
       \-
       o
       UJ
       O
       O
          1.0  L
          0.8   -
          0.6
      LJ
      jl  0.4

      o

      o  0.2

      o
      H
      <
              X= 0.904pm
              Q, =7°
              02 = 14°
              n = 1.6- i n
                               1.0
2,0
3.0
4.0
5.0
     Figure 4.  Scattered light  intensity  ratio for  particles of varying diameter and refractive
                index.   [After Kreikebaum  and  Shofner].

-------
Range
Angles
Beam Dia.
Depth of
View
Viewing Vol.
n
N"
T
GRAVATT
0.2-4.0 ym
10°/5° etc.
0.14 ram
1.0 mm
l.BxlO"5 cc
104/cc
0.15
10 ys
KREIKEBAUM
and
SHOFNER
0.2-3.0 ym
14°/5°
0.02 mm
- 0.6 mm
2xlO"7cc
106/cc
0.20
ty
Improvements Desired
1.  Increase-the detectable concentration n.
2.  Increase the sizeable range.
Figure 5.  Comparison of Gravatt and PILLS-IV optical
           particle counters.
                        227

-------
concentration with the focal  volume,  in both these,cases,  you  get  a  number
like N = 0.1 or 0.2.  So, on  the average,  you have much  less than  one
particle in the viewing volume.   This is done purposely  to avoid the pro-
blems of multiple scattering.   If you want to do better, than  you  have  to
be able to handle two overlapping pulses.   That is the only way  you  can
increase the detectable concentration.  Therefore, improvements  are
possible:  you want to be able to increase R, and you want to  extend
the sizable range also.
     Let's look at the lower  part of the viewgraph (Figure 6)  first. Let's
assume that you somehow get two particles  into the viewing volume  at the
same time.  Two pulses like that; they overlap one another.  If  you  devise
electronics in such a way to  sample the height of the pulses only  when  the
slope of the pulses change signs, in other words when the slope  changes
from positive to negative, you measure the height of pulses for  a  short
duration of time.  If you indeed have such electronics then you  can
separate two pulses as shown.   Therefore,  two overlapping pulses cause
a problem no longer* and you  only have to  worry about three overlapping
pulses.  So, instead of N = 0.2, such as in PILLS IV, you can  now  handle
H = 0.6.  This is a problem of statistics.  The probability of having three
particles for N! = 0.6 is about 2 percent which you can afford  to lose.
Similarly, in the PILLS' case, they try to avoid two overlapping particles.
When W = 0.2, the probability of having two overlapping  particles  is also
about 2 percent.  Basically if you can handle two overlapping  pulses,
you can push the value of the N from  .2 to .6 and that is a factor of  three
improvement.  I will show you an electronic circuit  that can do  that.
     Looking ahead, there is  really no fundamental reason that one cannot
push N" even higher.  Suppose  you know how to separate six overlapping  pulses.
The only thing that you have  to worry about is if you have seven overlapping
pulses, or higher.  In that case, you can handle N of three and  that would
increase the detectable concentration by a factor of 15.   I think that
we will be quite happy to achieve that now; if you can do  better, you can
increase the concentration even more.
     Here is an electronics circuit (Figure 7) that we use to handle two
pulses.  Basically, we have two photomultipliers, integrators,  pulse in-
verters, and a divider.  We send the divided pulse into a  pulse height
                                   228

-------
Number of
Particles in
Focal Volume
(N)
1
0.8
0.6
0.3
0.2
0.1
Probability of n particles entering viewing volume
simultaneously, P(n)
P(0) P(l)
3.68 x 10"1 3.68 x 10"1
4.49 x 10"1 3.59 x 10"1
5.49 x IO"1 3.29 x 10'1
7.41 x IO"1 2.22 x 10"1
8.19 x IO"1 1.64 x IO"1
9.05 x IO"1 9.05 x 10"2
P(2)
1.84 x 10"1
1.44 x 10'1
9.88 x 10"2
3.33 x 10"2
1.64 x 10"2
4.52 x IO"3

P(3)
6.13 x IO"2
3.83 x IO"2
1.98 x IO"2
3.33 x 10~3
1.09 x 10"3
1.51 x 10
          Detection of two overlapping pulses will  increase N  by
          a factor of 3.
          Why not detecting multiple pulses?

             For 1=3, P(7) = 2.6 x 10"2; N t by 15

             For 1=6, P(ll) = 2.25 x 10"2; I t by 30
Figure 6.   Increase of allowable particle concentration by detection and
           separation of overlapping particles.
                                   229

-------
rv>
oo
o
        1.  Photomultiplier  at  5°
        2.  Photomultiplier  at  10°
        3.  Input amplifier
        4.  Inverting amplifier
        5.  Isolation amplifier
        6.  Signal delay  circuit
        7.  Denominator bias circuit
 8.  Analog divider
 9.  Comparator
10.  Inverter and attenuator
11.  One shot
12.  Electronic sampling switch
13.  Impedance matching amplifier
14.  Pulse heighth analyzer
                            Figure 7.  Circuit diagram of two pulse equipment.

-------
analyzer to get a probability distribution.  I will go into details when
people are interested, but now let me just say that what this circuit will
do is to open this switch whenever the slope of the pulse goes from posi-
tive to negative.  When that happens, it closes the switch and samples the
pulse height for a very short time,  like 2 microseconds.  These elements
are all commercially available integrated circuits.  So, whenever you have
two overlapping pulses, the switch will open to get readings for the height
of the individual pulses.
	:  Are you able to get clean enough signals from the electronics to
handle it?  Noise doesn't bother it?
SHE:  Well, in some cases it does and in some it doesn't.  It depends on
how much intensity you have.  For the large particles, it's no problem;
for the small particles you have to  be very careful.
     This is a viewgraph  (Figure 8)  from a theoretical paper published in
Applied Optics by Steve Chan and myself.  This shows the concept of multiple
particle sizing.  You see a laser is focused down to a volume, and the
scattered light is viewed by a photomultiplier.  Suppose for a moment you
only allow one particle at the most  into your focal volume.  Tt is the
transit time, the time it takes the  particle to go through the volume.
                   \
Whenever a particle goes  through you get a pulse whose height depends on
the particle size.  If you increase your flow velocity, all you do is to
squeeze all of these pulses together in time, but if you allow, say, three
particles in the focal volume at any given instant of time, during transit
time Tt, you will see not just one big peak but three overlapping peaks of
various magnitudes.  So, the wiggles in the lower figures correspond to
particles.  The only requirement is  that the time between two successive
ripples should be longer than the resolution time of the electronics Tr.  In
other words, if the change here is slower than what your electronics can re-
solve, then you can handle a higher  particle concentration.   In order  to
process this, you need to be able to handle multiple pulses.  What you  gain
is a ratio of Tt/TR, which may be a  factor of 10 or so.  This is  very  nice
conceptually, but if you  use this technique without care, for real particle
sizing or sizing in real time, you are in trouble because the height you
measure then is not necessarily the  same as the original height of the pulse.
                                   231

-------
I(t)f
           Laser
                      Sample
                      Volume
,Lens      Aperture
                   (b)
                                         Light/
                                         Block
                                       (a)
     Kt)f
                                                 I     I	1_
                                                                       Small
                                                                  I  /'Particle
                         (c)
             H   HT
                    (d)

                                                 I    I    I    I
                                                      HH
                         (e)
        a.  ExperinientaX arrangement.
        b.  Current pulses for  a modest flow speed.
        c.  Current pulses for  an increased flow speed.
        d.  Viewing several flowing particles in the scattering volume.
        e.  Viewing several flowing particles in the scattering volume
           at  increased speed.
                Figure 8.   Multiple particle sizing.
                                   232

-------
So you have to do some fancy footwork, so to speak.  What you.could do
is to shape your laser profile in such a way that the rise and drop of
the pulse becomes very sharp.  If you can do that, then you are all right.
Let me show this to you with the following viewgraph (Figure 9a).  Let me
emphasize again that we have been working on this only six months so that
we haven't really done any experimental work on this concept.
     Let me assume that I have three particles coming in, and I somehow
can shape them nicely.  This can be done perhaps by aperaturing the laser
profile or sending the laser beam through some optical fibers which, I don't
know for sure at this moment.  The overlapping pulses will now give
you something like this (Figure 9b).  If you can now design an electronic
circuit in such a way that whenever a rise in the current is confronted,
you ask the device to sample the difference between the heights of the
current right after and right before such a rise.  When this goes up again,
you sample the difference between this height and that height and you get
a second pulse=  Whenever the pulse comes down, you ask the device not to
do anything but just to be ready for the subsequent rise of the current.
This we know we can do.  Now, you may want to ask me what are the limita-
tions of this sharp rise.  There are three things that can limit this:
one is how sharp you can make your laser beam profile, the second
is the resolution time of the electronics, and the third is the size of
the particle.  When a particle enters the focal volume, depending on the
diameter of the particle, it takes a finite time to enter the viewing
region and that will also smooth out the sharp rise. , The longest of these
times will determine the rise time in question.  So, the number of multiple
particles you can sample equals the ratio Tt/TR.  I think 5 or 10 is about
right for this ratio.
     The last things that I will discuss for this conference are some possi-
ble ways to increase the range of sizable particles.  Again,  (Figure  10)
this is one of the intensity ratio curves plotted versus alpha, which I
showed in Figure 3 for angles of 10°:5°.  Consider small size particles,
perhaps an alpha of 2, which corresponds to 0.4 microns for helium-neon lasers
and 0.3 for Argon-ion lasers.  In order to sense this portion of the  curve,
                                   233

-------
             K-Tt—I
(a)  Individual Overlapping  PuUes
                                        Time
                                        Time
 (b)   Sum of Overlapping Pulses
                                        Time
 (c )   Retrieved  Pulse   Height
   Figure 9.  Determination of true pulse height.
                   234

-------
              In
                 11
0.99

0.95

0,90
1.01    1.01  x 10
                       "2
1.05   5.13 x 10

1.11  10,5 x 10"2
                "2
Will help small  a.
           A = .6328 u
•v
i.a
0.8
0.6
0.4
0.2





»
2 6 10 14 a
0.4 1.2 2.0 2.6 y
BACKSCATTERING


   5°/10° ratio is

   possible in the

   backward direction.

   This will help large particles,
                                        LASER
                                                             I
                                                            *
                                                            D
                                                           CK-
       Figure 10.  Possible ways to increase the sizeable  particle range.
                                      235

-------
assuming you can avoid noise problems,  you have  to  be  able  to  read  the
difference between two small numbers.   You can help that  by using a log-
arithm amplifier.  Suppors that instead of measuring this ratio  you measure
the logarithm of the inverse ratio.  Taking,  for example, the  ratios
from .9 to .99 which corresponds to  this part of the curve, their inverse
will give you these numbers.  If I take the logarithm after that, you
have a factor now of 10 difference with which to work.   In  other words,
by using logarithmic amplifiers which  are commercially available, you might
be able to expand the intensity-ratio  scale.  Another thing that one can
do is to do something about the large  particle range here.   Remember that.
in order to sense the large particles,  you want  to  be  able  to  work  with
small angles.  In the forward direction, this is very  difficult; 5°
is probably the smallest angle to use^unless  you want  to  detect  the laser
beam.  You might be able to do back  scattering,  however.   You  can shine
your laser beam through a beam splitter and focus it down onto particles.
You put your two photodetectors, for example, at 5° and at  0°.  The back
scattered light will come and hit this  beam splitter and  go this way.  That
will not mix with the forward laser  beam.  That  way you can do a ratio
of 5°:0°.  Then, you can, perhaps, hope to do better with large  particles.
     That is all of the viewgraphs that I have.   Let me summarize by
saying that we have been able, in the  laboratory, to separate  the two over-
lapping pluses and get a particle size  distribution, even though we have
some problems with noise and calibration.  We don't have  a particle genera-
tor like you have.  I hope my presentation will  give you  experts a  chance
to think seriously about light scattering as  a viable technique  for
particle sizing.
                                  236

-------
DISCUSSION

ENSOR:  Just a comment-  Generally, light-scattering is more sensitvie to
particle shape than to refractive index; this may reduce its practicality.
SHE:  I don't know about that.  There was a paper sent to me, after we
published that Applied Optics paper, by Brinkworth of England.   They  show
that backward scattering, really backward, is as good as forward in terms
of its insensitivit.y to the refractive index.  If you think about it, per-
haps it is not too surprising.
ENSOR:  Well, the imaginary refractive index should be more...
SHE:  Perhaps.  I don't know.
ENSOR:  And lidar is used in cloud physics to tell whether you have ice
crystals or snow or^ike that.
SHE:  That's something that one could consider.  Of course, back scattering
has less intensity, but if you are working with large particles5 you  are
all right.  For small particles, the backward and forward intensities are
comparable.
BOLL:  If I remember rightly, Hodkinson did his work based on the assump-
tion of uniform spherical particles.  Flyash particles tend to have a
certain amount of shell structure to them, because as the gas cools down,
you get things condensing out on the surface and while it would seem  to be
a small effect, I was wondering if you had given any thought to how it
would affect those calibration curves.
SHE;  I didn't do any work myself.  Gravatt did look into some such parti-
cles.  Let me say, first of all, in light-scattering we always assume
spherical particles.  Other shapes are quite complicated even though,  in
principle, you can determine different particle shapes.  In practice,
it is not easy.  However, Gravatt did check the particle shape question
in sizing and.found that it is not that sensitive to shapes if you can
tolerate the accuracy of about 10 percent.
                                   237

-------
BOLL:  I'm speaking of refractive index variations  in rate.
SHE:  That's a good question.   That may"or may not  be important.   I  haven't
done any work.  If the particle is much smaller than the wavelength,
certainly it doesn't make much difference.
BOLL:  I was thinking that the PILLS data were more compact and wondering
why.  Maybe that could be the  explanation.
SHE:  Could be.  That's the first time I saw the data also.
ENSOR:  I would like to comment here that Meteorology Research has been
using different approaches in  applied light-scattering.   I  feel  that
applied light scattering is one of the forgotten aspects of engineering.
However, we, in general, have  been using across-the-stack techniques
and inverting the information  into forward light-scattering angles,  and
with proper signal processing  you get more information about the size
distribution.  Our specific objective is to try to  get a better measure-
ment of aerosol mass and aerosol  volume.  I might point out that there
are other techniques available.
SHE:  Yes, well I guess that everybody knows that light-scattering
techniques measure different diameters than those measured by impactors.
ENSOR:  Well, there is a lot of information on light-scattering around.
SHE:  It is difficult to get it all out.
SPARKS:  Thank you.
                                  238

-------
    GENERAL DISCUSSION OF CASCADE IMPACTORS OR ALTERNATIVE DEVICES
         D. B. Harris, Process Measurements Branch, IERL-RTP
HARRIS:  The only alternatives that I want to discuss right now are our
attempts to use the Series Cyclones system to give us particle size
distributions.  You have seen some of the results as far as their field
useability goes in the work that Bill Kuykendal presented earlier this
morning.  I would just like to go into it a little bit to show the cali-
bration end of things and what we're hoping to apply these devices to.
The main reason we looked at using Series Cyclones instead of impactors
is that we were having to supply sufficient material to other people to
do testing for chemical or biological health effects purposes.  When you
get to the point where you have a 10 milligram collection per stage and
one of these guys comes and says to you, "I need a gram to work with,"
you don't look forward to running impactors for the rest of your life to
get enough sample for him.
     The Series Cyclones look like our only alternative.  The device that
I brought in yesterday (Figure 1) was our first attempt at it.  The system
operated at 1 cfm* and had three sizing cuts to give us particles in the
range of about 3 microns maximum size.  Figure 2 is a set of curves of col-
lection efficiencies of these devices.  Their shape is why I made the comment
that it looks like an impactor collection efficiency curve.  The classical
cyclone curves usually have less slope in the straight portion of the curve.
The cut point is not as sharp.  As a matter of fact, some of them look
better than you get on some impactor data.  Data from five cyclones are shown
in Figure 2.  Two of them are units that were designed four years ago for
us by McCrone Associates.  They were arranged in a parallel configuration
as we were attempting to develop a sizing scheme that would allow us to do
isokinetic sampling while running parallel cyclones and the bypass filter,
adjusting the filter,to meet-the isokinetic requirements.  Subsequently,
Southern used two of these designs in developing the Series Cyclones, adding
 See metric conversions listed on page 252.
                                   239

-------
Figure 1.   One ACFM series cyclone showing the three cyclones
           and Gelman back-up filter in the preferred orienta-
           tion.  The nozzles which were developed for isokinetic
           sampling from 10 fps to 100 fps are also shown.
                             240

-------
ro
Figure 2.
                                        PARTICLE DIAMETER , MICROMETERS
                        Collection efficiency versus  particle diameter.
                        p  =  1.35  g/cc, 22 °C, 29.5 "Hg).
(5 cyclones, 1.0 ACFM,
                                                                                                 LEGEND
                                                                                                 •   "NF"
                                                                                                 •   "TIB"
                                                                                                 A   "TZA"
                                                                                                     "CZ"
                                                                                                     "AL"

-------
one in the middle to give us a cut size we needed in that range.   The others
shown in Figure 2 are two cyclones that have been used as precutters for the
Brink impactor.
     The next major step that we took was to try to increase the  flow rate
of the units.  Evan at 1 cfm, when you start talking about the concentrations
that Reed [Cass] and Gil [Sem] were finding out >at Nucla, a good  bit of your
lifetime could go by sampling on the stack.  The first attempt was to come
up with a high-vol scheme.  We decided that we should try to get  it into a
form that we would have a chance to use in the field.  Our effort was aimed
at adapting it to the high volume sampling system that we had for mass
sampling at this low concentration.  That train is commercially available
from the Acurex people.  Again, I prevailed on Joe [McCain] and Wally [Smith]
to come up with some cyclones that would give us three cuts when  operated
under 5 cfm conditions, and they did (Figure 3).  Unfortunately,  I didn't
stipulate in the contract that they had to carry it out on its first field
test.  All the cyclones operated fairly well, but the large cyclone was
                               t
about 3 feet high.  It did cut at 12 microns and we were aiming for 10,
but it didn't fit in the Aerotherm oven.  The basic idea still works very
well.  We got three different size classifications plus we had plenty of
material for the toxicity people.  We had a series of 10 tests, run up and
down the West Coast by TRW Systems, using that scheme.  Those poor boys did
not love to lug this thing up and down on their backs, and I have heard
about it every time that I have seen them since.
     We did get sufficient sample to do some cytotoxicity testing.  The pre-
liminary results have given us enough shocks so that we decided to try to
get a system that would be more useful.  The data indicated that some of
these plants that we have previously assumed to be innocuous in certain size
ranges, usually predominantly in the respirable size range, weren't  too nice
at all.
     The next iteration involved asking the Acurex people to try to  get the
cyclone train into a configuration that we could move around a little bit
better.  Figure 4 shows the scheme that they have come up with.  They have
the whole unit sitting inside the oven.  For people familiar with their
                                   242

-------
                                                          AEROTHERM OVEN
ro
-F=.
co
            STACK
OVEN OR

THERMAL BLANKET
                                                                                                     IMPINGERS
                                                                                GAS METER

                                                                                TC-        FINE ADJ.
                                                                                           BYPASS VALVE
                                                                                        AIRTIGHT

                                                                                        VAC. PUMP
                                                                                                       VACUUM GAGE
                             10/i CYCLONE
                                                                                DRY TEST

                                                                ORIFICE AP      METER

                                                                MAGNEHELIC GAGE
                                                                       COARSE

                                                                       ADJ. VALVE
                                    Figure 3.  SoRI concept of multicyclone train.

-------
Figure 4.   Acurex concept of multicyclone train.
                      244

-------
normal unit, this one is about 3  inches bigger on each side than the standard.
Something you can manage, but still fairly  large.  The main difference is
that they used a stub-cyclone configuration similar to what we've seen with
the Aridersen and the Sierra  impactor,  as opposed to a classically designed
cyclone that has been tapered.  It  has a separate collection box and it has
the capability of using the  pressure drop across one cyclone as the flow
rate control,
     Figure  5 shows the first calibration data we had on this unit.  We
were aiming  at 10, 3, and 1  micron  cuts.  Anything larger than 10 microns
we figured was going to hit  somebody in the head rather than get inside
the body, so we didn't worry about  that too much.  Three microns is the
classical definition running around here (today, anyway) of what is respirable,
and 1 micron was our feeling for  t!he size range which would penetrate into
                                                                        ^,
tbe alveolar area of the bronchial  system.  However, we ended up with pro-
bably the most efficient cyclone  the world  has ever seen.  The little unit
was running  about 0.3 micrometer  for its D^Q.  The reason that this is
extrapolated on the plot is  that  we ran out of capability to generate material
small enough to calibrate it.  So that     curve is extrapolated and the
shape below  1 micron is uncertain.  The other cut points missed a little bit.
We got 5 to  12 microns.  Since we had  this  idea in our head that we wanted
these certain size cuts, we  asked Joe  [McCain] to fiddle around a
little bit and see what variations  he  could make.  The results are shown in
Figure 6.  Basically by changing  the inlet  and outlet tubes, he came up with
the type of  configuration needed.  The main reason I present this is to say
that cyclones seem to  lend themselves  very  well to tuning at a specific flow
rate to whatever size  range  you are interested in.  With the range we have
in Figure 6, something  like  1.5 to  4.5 microns, it does not seem to be out
of the realm of possibility  to think of making a unit with a number of
stage cuts similar to  an impactor.  Impactor data and cyclone data seem  to
agree fairly well.  For the  data  Bill  [Kuykendal] presented this morning,
the cyclone  cuts were  laying right  over the curves that we were getting
with impactors.
                                    245

-------
IVJ
                                                          LEGEND


                                                        Small Cyclone


                                                    A   Middle Cyclone
                                                                                                    50.0
                                                     PARTICLE DIAMETER,
               Figure 5.  Collection efficiency versus particle  size for  the 5 ACFM  Series Cyclone.
                           (4 ACFM, 1.35  g/cc, 29.5"  Hg, 22 °C).

-------
                      100
ro
                                                                                             LEGEND
Inlet
0.82"
0.50"
0.62"
0.62"
0.62"
Outlet
0.63"
0.63"
1.609"
0.63"
1.609"
                                                                                        Vortex Buster

                                                                                        (Small Cyclone)
                        0.4
*i>ARTlCLE DIAMETER,
IOD
                     Figure 6.   Collection efficiency  versus particle size for the 5 ACFM Series Cyclone with
                                modifications  to  middle and small cyclones.  (4 ACFM, 1.35 g/cc, 29.5" Hq,
                                22 °C).

-------
     Figure 6 also graphically presents  the  changes  in  the  size  curves  as
inlet and outlet nozzles are changed.  This  is  both  the middle cyclone  and
the small cyclone.  The curve represented  by the  square shows results of
pushing the small  cyclone up to where  its  D5Q is  on  the order of 1 micron,
actually we've got about 1.2 microns.  The other  3 curves are three  varia-
tions on the middle cyclones where we're trying to get  the  cut point
closer to 3 microns.  The middle of the  three curves does this at about 3.5
microns.  We are presenting this information to show that you can, by working
with either the outlet or inlet tube,  change the  cut capability  of the  cyclone.
     That is basically where we've come  to with the  cyclones.  We'll probably
be using more of them in the field since we  are becoming more involved  in the
source assessment idea, which is to be able  to push  back to the  health  end
of things the effect that particle sizing  might have.
     The last thing that I would like  to present  is  something for everybody
who wants to make a few bucks if they  ever build  it. This  is what we  need
for an ideal particle size instrument  as our needs stand now.  Figure  7 shows
the criteria.  We need in situ measurement and furthermore, we'd like  it
so that at the time that we're doing the actual sizing, the stack doesn't
know we're around.  If you don't have  sample nozzles to mess up  your results,
then your results are not going to get messed up  by  the sample nozzles. At
the same time, after you get done sampling,  you want to collect  this sample.
The size distribution should be kept separated for the  other problems  of
chemical and biological testing.  All  the  data should be  real  time,  because
of all the process upsets that people  have told us about which  cause pro-
blems with impactors.  We need to get  rid  of that problem.   Again, we have
the problems of simultaneous inlet and outlet measurements.  We  might be
able to feed data into a black box so  that we get instantaneous  efficiency
results out of it.  The machine must be  able to handle concentration ranges
similar to the ones we have seen where we've got  perfectly clear stacks or
ones we can walk across. Our size range  needs and our temperatures  (and we're
in the process of moving the upper temperature up as we get into the coal
conversion schemes where we have to face 1600 or  2000°F) are shown.  So you
can have a little flexibility here in  the  upper limits.  All of this,  of
course, should be able to fit into something the  size of an HP 65 so that'
everybody can carry it around with them.  Any questions?

                                   248

-------
                 IN-SITU PARTICLE SIZING

         COLLECTION OF SIZE FRACTIONATED SAMPLE

                    REAL TIME OUTPUT

       SIMULTANEOUS INLET AND OUTLET MEASUREMENTS

PARTICULATE CONCENTRATION RANGE:  0.01 to 25 g/m3
                                  10 to 1010 particles/cc

     AERODYNAMIC SIZE RANGE:  0.01 TO 10 MICROMETERS
              -,
              t
             TEMPERATURE:  AMBIENT TO 450°C

                   PORTABLE OPERATION
          Figure 7.   "Ideal"  instrument features.
                           249

-------
DISCUSSION
         Bruce, when you show a point on your rate efficiency curve,  is
that the result of one run or the average of several  runs?
HARRIS:  It varies.  On a calibration of the cyclone  in the original
report, I think that it was the average of several  runs.  At the point
where we were trying to do the repair job, i.e.,  to alter the cut points,
it was one run because we had problems to fix and we  had one in the field
and we had to get the data out in one or two days.   This work was done for
us by Southern.
	:  From back here, the cyclone looked awfully good.  Could you tell
us what the costs are?
HARRIS:  Well, the first set of data was with the 3 foot high set of
instruments.  I don't know yet what it is going to be by the time we get
all the ramifications that we want done in it, done.   The cyclones themselves
are on the order of $4 or $5K per modification if you just  wanted to do that.
If you wanted to get all the modifications we're having put into the
unit in order to do the source assessment (which means we're having to
insert an organic trapping device capable of trapping organics at a
temperature somewhere between the oven temperature and the temperature
of the impinger where we get the condensible inorganics), we're probably
talking about a total package starting to approach $20,000.
	:  So the expense would keep you from building, let's say, a
six-stage cyclone.
HARRIS:  No, I don't think it would be that much.
MCCAIN:  You can build a six-stage cyclone train  for  about 'the same price
as you can build a six-stage impactor, maybe cheaper.
	:  Okay.  Why wouldn't you do it then?
HARRIS:  Well, for one thing you then expand the volume th.at you have to
heat.  Since this is an extractive system, you have to use an oven or some
heating method.  There are not too many people with 4 foot holes in their
duct.  If your conditions allow you to run at 1 cfm instead of 5 cfm and

                                 250

-------
you get enough dust to work with, then you can get smaller.   Another thing
you could do would be to change to an axial type of configuration and use
the higher flow rates.
QLIN:  We're looking into designing something like this commercially.   At
first cut I would say that the price of 3 cyclone stages followed by a
filter that fits into a 3 inch port (I think that that's an  essential
ingredient) would probably run you about the same price as a standard
Cascade impactor.
	:  Well, the Cascade impactors run over a large price range.  Would
you give me a little better idea?
MCCAIN:  $500 a cyclone.
OLIN:  That sounds about right.  I'd like some feedback on this because we
fooled around with it a little bit in terms of preliminary designs.  I'd
say something that runs from  say 0.2 cfm up to 1 or 2 cfm so that you can
get the outlet of low or high efficiency devices and maybe 3 cuts.  I think
more than that .does become fairly complicated.
HARRIS:  The big problem that we have was the source assessment needs.
We're going after 5 cfm.
	:  I always thought that the problem was that you couldn't get down
on particle size, but you sound like you can get down to 0.3 of a micron.
HARRIS:  At 5 cfm we do.  We  haven't tried to get down below that at 1
cfm.  The one that we had originally set out for 0.5 micron ended up cut-
ting at 0.8.  The problem that we've had so far is that we don't have
any equation which applies to cyclones, especially when you get down to
small sizes.  All we have is  the classical cyclone equation.  It seems that
when you shrink the cyclone down to where you have a significant portion
of the flow stream exposed to the wall, so that wall effects come in, the
classical cyclone equation just goes to pot.  We haven't come up with an
equation that is able to describe what's going on.  It's been a matter of
something like:  we've got this body and this head and we're going to change
the inlet diameter, change the outlet tubes, put in a flow buster to  break
up the action a little bit.

                                   251

-------
QLIN:  I think that the thing that is evident here from the work done by
SRI and EPA is that you are in a different regime.  It's a laminar regime
and the classical  turbulence equations  of Lapel  and others just don't
seem to apply.  Indeed, you can get very small  cutoffs without reentrain-
ment.  So I think  that it is a kind of new area.   The one disadvantage
is that you're not collecting the material on a filter or something, so
you have to brush  the particles out of the cyclone.
HARRIS:  The thing that helps in that respect is  that we're aiming to
collect 0.5 or 1  gram or even greater quantities.
OLIN:  To collect  larger amounts.
RAO:  One thing,  if you plan to use cyclones with  different flow rates,
what calibration do you use?
HARR.I S:  Probably  just use a function of the flow  rate through the cyclone
until we get something better.  Get points at two  different flow rates
and extrapolate.

                           CONVERSION FACTORS
     Although EPA's policy is to use metric units  in its publications,
certain nonmetric  units were used in this paper for convenience.  Readers
more familiar with the metric system are asked to use the below factors to
convert to that system:
                    cfm x 28.3 = liters/min
                    5/9 (F-32) = C
                    ft x 0.3   = m
                    in. x 2.5  = cm
                                   252

-------
     FIELD EXPERIENCE WITH CASCADE IMPACTORS FOR BAGHOUSE EVALUATION
                        Reed Cass, GCA Corporation
SPARKS:  Tt looks as if everyone  is here, and most important of all  our first
speaker is here; we'll get started.  We're going back to impactors now.
The first speaker is Reed Cass of GCA and he is going to talk about how cas-
cade impactors work when you're trying to evaluate a baghouse.

CASS:  Gentlemen, GCA has recently conducted sampling programs on two utility
boilers fitted with fabric filter dust collectors.  In-stack impactors
were used to measure the fractional particulate collection efficiencies of
the baghouses.   I would like to relate some of what we'learned while using
the impactors, focusing on the precision and accuracy of the impactor
measurements and the problems which we encountered.
     The first testing program was conducted at Nucla, Colorado generating
station, which has been previously discussed.  Steam is generated at Nucla
by Springfield stoker-fired, traveling-grade boilers with fly ash reinjection
These boilers are fitted with Wheelaborator-Frye fabric filter dust collectors,
which use reverse flow and shaking for cleaning.
     The second  testing program was conducted at the Sunbury Steam Electric
Station which is located in central Pennsylvania.  The station has a capacity
of approximately 400 megawatts, which are generated by four steam turbines.
Two of the turbine units, rated at 87.5 megawatts each, are supplied by four
anthracite fired Foster-Wheeler boilers.  These boilers burn a pulverized
fuel mixture of  15 to 35 percent  petroleum coke with the remainder made up
of anthracite silt and No. 5 buckwheat anthracite.  Each of these boilers
is served by a Western Precipitation fabric filter dust collector, which
utilizes reverse air flow to clean the bags.
     In the series of 22 tests at the Nucla station, the utilization of
two identical Andersen impactors  sampling simultaneously for 6 hours
afforded an opportunity to examine the precision of the technique.  Figure
1  shows the sampling locations in the stack for the impactors as well  as
                                   253

-------
                                                 IMPACTOR
                                                 SAMPLING
                                                 POINTS
Figure 1.   Cross  section  of  baghouse  outlet  sampling
           location  showing  sampling  points.
                   254

-------
the traverse points for an Aerotherm High Volume Sampler, which was  used
according to Method 5 technique.  The stack cross-section was located more
than 8 diameters from the nearest upstream and downstream disturbances,
providing what should have been a homogeneous effluent for the impactors.
     Table T presents the geometric mean mass concentrations as measured
by each impactor and both impactors.  As can be seen, the geometric  mean
concentrations as measured by each impactor are very close to each other.
However, the average absolute value of the differences .of each pair  of
measurements is about 70 percent of the geometric mean of all measure-
ments.  This means that for any given paired sample, a significant
difference is quite apt to be observed between the two impactors.  If
one does not take the absolute value of the difference, however, the
average difference is only about 18 percent of the goemetric mean of
all of the measurements.  This shows that one impactor does not always
tend to be biased, and that the differences, though large, are probably
random.
     Table 2 presents an analysis of the mass median diameters as deter-
mined by each impactor.  This shows the same trend as the mass concentra-
tions analysis, except that the geometric standard deviations are some-
what smaller.  Further, the average absolute value of the difference of
each paired measurement is only about 40 percent of the average geometric
mean of all the measurementsj and the average difference between each
pair, not taking the absolute value, is only about 5 percent of the
geometric mean.  It would appear, therefore, that substantial, apparently
random differences are quite apt to be observed in the mass median
diameters as determined by paired impactors, but that those differences
are less than those observed for the measured mass concentrations.  The
accuracy of the impactor measurements was determined by comparing the
geometric mean of the impactor concentrations with that of Method 5
outlet measurements.  The Method 5 geometric mean concentration of 0.0025
grains per dry standard cubic foot is only about 66 percent of the impactor
geometric mean concentration.  Despite the poor agreement between the
Method 5 and impactor measurements in terms of mean concentration, no
specific difficulties were encountered and the actual field  use of the
impactors was nearly trouble free.
                                   255

-------
   Table 1.   COMPARISON  OF OUTLET  IMPACTOR  CONCENTRATIONS

Impactor X
Impactor Y
Impactor X + Y
Geometric mean
concentration grains/dscf
0.0041
0.0036
0.0038
SIX - Yf _ 0 OQ26 E(X - Y) =
n U'°°'-b n
Geometric standard
deviation
2.0085
2.298~2
2.1419
-0.0007
Table 2.   COMPARISON OF OUTLET IMPACTOR MASS MEDIAN DIAMETER

Impactor X
Impactor Y
Impactor X + Y
Geometric mean
mass median diameter, ym
8.57
8.13
8.34
Geometric standard
deviation
1.51
1.50
1.50
                 = 3.26
.45
                            256

-------
     At Sunbury, however, several problems with in-stack impactors arose,
which not only severely complicated field use but also adversely affected
the accuracy of the data.  These problems included anomolous weight
gains on the Andersen glass fiber substrates, weight losses of University
of Washington greased impaction inserts, skewed size distributions due to
excessive particulate capture while using an Andersen cyclone precollector
and due to probe losses which amounted to significant portions of the total
sample.
     The anomolous weight gain problem was first indicated while sampling
at Sunbury when it was noticed that the weight gains of the Andersen glass
fiber substrates during the 6 hour test were much greater than would have
been expected based on visual inspection.  The substrates,  which gained
substantial amounts of weight, had almost no signs of particulate on
them.  To test the theory that the Andersen substrates were gaining weight
which was not particulate, a particle-free sample flue gas was drawn through
an impactor for the normal sampling time of 6 hours.  As shown  in Table
3, when the Andersen substrates were weighed, they had all  gained weight,
confirming that the weight gain was not due to part
-------
               Table  3.  ANDERSEN SUBSTRATE WEIGHT GAINS
                         WHEN SAMPLING  FILTERED  FLUE  GAS
Stage
Pref i 1 ter
0
1
2
3
4
5
6
7
F
Weight gain, mg
5.7
4.8
3.5
3.8
2.6
2.0
1.8
1.6
0.1
0.8
Table 4.  ANDERSEN SUBSTRATE WEIGHT GAINS WHEN SAMPLING FILTER FLUE GAS

Stage
Pref i 1 ter
0
1
2
3
4
5
6
7
F
Substrate weight gain, mg
Upper
4.5
6.3
6.4
5.2
4.3
2.1
1.7
1.2
1.0
0.0
Lower

3.3
2.9
1.8
2.0
1.0
0.8
0.5
0.6
0.3
Upper- lower

3.0
3.5
3.4
2.3
1.1
0.9
0.7
0.4
-0.3
lower 1nn
upper x 10°

52
45
35
46
48
47
42
60

                                   258

-------
inserts coated with grease as impaction surfaces.  Initially, the inserts
were coated with Dow Corning Hi-Vacuum grease,  however, when these inserts
were weighed after sampling, it was found that  they had all lost weight.
After some experimentation, inserts coated with polyethyleneglycol and dried
in an oven at 300°F overnight were found to be  satisfactory, showing sub-
stantially decreased weight loss problems.
     In essence, we found substantial weight gains for the glass fiber
substrates, but we were able to compensate for  them at least partially.   We
also found the losses using Dow Corning Hi-Vacuum Grease to be too large
for our purposes, but polyethyleneglycol coatings performed satisfactorily,
although also producing some losses.
ENSOR:  What was your change in weights with the glycol?  Do you have the
        values?
CASS:   No sir, I don't.
        When the size distribution curves for the first 21 Sunbury inlet
Andersen impactor tests with a cyclone precollector were plotted, the size
distribution curves indicated that the precollector was collecting a large
portion of particles which would have normally  been impacted on the upper
impactor stages.  The curves shown in Figure 2  are clean examples of how the
form of the curves is affected by the precollector.  In an attempt to
determine what portion of the particles was being removed from the upper
stages by the precollector, an additional impactor sample was taken through
a gooseneck probe for each of the remaining 10  tests.
     Upon examination of cumulative distribution curves for these runs, no
consistent difference between the size distribution determined by the
Andersen with the precollector and the Andersen with the gooseneck nozzle
could be found.  This is apparent from the size distribution curves pre-
sented in Figure 3.  Also the mean percentage of the mass collected by the
impactor precollector for the tests was approximately 35 percent of the total
mass collected while the mean percentage of the mass collected in the goose-
neck probe for the same series of tests was approximately 33 percent of the
total mass collected.  Thus, it appears that both the cyclone precollector

                                   259

-------
   100
CO
tc
its
l-
fel
5
O
£C
O
K
UJ
H
Ul
S
<

O


u
z

O

O
tc
UJ
10
1.0
    0.1
   i   I    I   '  I  '  I   I   r~TT
                                  6 INLET  RUN A

                                  a INLET  RUN B
           I    I   i  I  i  I  J_i   !
                                I
I    I
     2    5    10   20  30 40    60     80   90  95   98

      PERCENTAGE OF  MASS LESS THAN OR EQUAL  TO STATED SIZE
 Figure 2.  Inlet cumulative particle size distribution for run 17.
                        260

-------
   100
      FT~T
te.
ui
H
UJ
2
o
DC
O
£C

U)

I-

Ul
O

2
o
te
ui
    10
    1,0
   0.1
                                 O  INLET  RUN A


                                 Q  INLET  RUN B

                                    (NO  CYCLONE)


                                 +  INLET  RUN C
           !    I   i  I  i  I   I  i  I   i   lil    I
     2    5    10    20 30  40    60     80  90   95  98

      PERCENTAGE  OF MASS LESS THAN OR  EQUAL  TO STATED SIZE
  Figure 3.   Inlet cumulative particle size distribution for run 23.



                        261

-------
and the gooseneck probe removed approximately the same size fraction and
amount of particles.  This finding is not surprising because the centrifugal
force applied to the partictles in the gooseneck probe is the same type
of force that is applied in the cyclone.   The large percentage of the total
sample which was collected in the probe was also evident in the series of
tests at Nucla, Colorado.   The mean percentage of mass caught in the Nucla
inlet impactor gooseneck probe was approximately 21 percent of the total
sample collected and the mean percentage of mass caught in Nucla outlet
gooseneck probe was approximately 15 percent of the total  mass collected.
    In our previous measurements with impactors using gooseneck probes,
we have assumed that the fraction caught in the probe was  larger than that
impacted on the top stage of the impactor.   From our recent experience,
it has been shown that the gooseneck probe may remove particles which
would normally impact upon the second impactor stage.  Therefore, whenever
possible, straight probes should be used; however, when it is necessary to
use 90° probes, they should be designed to impart the least centrifugal
force to the sample stream possible.  Also, a 90° elbow probe should be
used rather than a gooseneck probe because the particles are exposed to
centrifugal acceleration for a greater duration in the gooseneck probe in
which the particles pass through an arc of approximately 160°.  Whatever
probe is used, it should be checked for losses, which may be substantial
and affect the size distribution curves for the impactor.   Thank you.
                                   262

-------
DISCUSSION

ENSOR:  When you noticed the weight gain on the substrates, did you
ever run two of the substrates through the filtered flue gas to see if
they gained the same amount of weights?
CASS:  Yes, we did.  That is presented in Table 4.
	:  Was that what that table was?  The extra weight was the particulate?
CASS:  No, that was when filtered flue gas was passed through the impactor.
SPARKS:  Thank you.
                                    263

-------
          LOW PRESSURE IMPACTORS  FOR  IN-STACK PARTICLE  SIZING
               Dr.  M.  J.  Pilat, University of Washington
SPARKS:   About everybody is  back  so  we  will  get  sta-rted  again.   Awhile
back we got interested in defining the  performance  of particulate control
devices in the particle size range below which conventional  cascade
impactors work (from say 0.1 to 0.01 micron  diameter).   The  next two
people are going to tell  us  about a  couple of techniques for measuring
fine particles,and then this afternoon  we will have a panel  discussion on
another method.
     Mike Pilat is our first speaker,and Mike has been working on a super-
high -pressure drop cascade impactor  to  push  the  sizing capabilities down
below 0.1 micron diameter, and he is going  to tell  us all about how good
this impactor is.

PILAT:  To provide a proper  perspective for  our  type of  research at the
University of Washington, I  think we should  be classified more as users
of cascade impactors in support of our  research  on  plume opacity, source
emission characterization, and particulate control  equipment development.
     After our initial work  on the Mark I,  II, and  III models of our U of
W Source Test Cascade Impactors,which size  particles in  the  0.3 to 30
micron diameter range (and where  the now commercially available cascade
impactors operate), we have  :focused  our cascade  impactor research on those
particles in the 0.02 to 0.2 micron  diameter range.  This research has
resulted in a cascade impactor with  gas pressure taps located on those
impactor stages with less than 0.2 micron D™ stage cut  diameters, as
shown in Figure 1.  These pressure taps are  to monitor the absolute gas
pressure in order to obtain  the magnitude of the Cunningham Correction
factor,  as shown in equation 1, used in the design  of cascade impactors,
                         D50 = l-	-H                            CD
                                  264

-------
                               •Inlet Nozzle
                                      Stags
                                   Collection Plate
                      To Vacuum Pump
                                           To
                                       Pressure Causa
Figure 1.   Cross  section of  prototype Mark IV U,W,
             Source Test Cascade Impactor.
                        265

-------
assuming 0.145 for the Stokes number and with y the gas  viscosity,  D^
the jet diameter, p  the particle density,  and V.  the gas  velocity  in
the jet.  As a matter of practicality,  the  jet diameter  is limited  to
about no smaller than 0.008 inch diameter,  the gas viscosity is fixed,
the particle density is set, the gas velocity is probably  limited by
sonic flow, but the Cunningham Correction factor is not  limited.  There-
fore, by reducing the absolute gas pressure,  one can get larger magni-
tudes of the Cunningham Correction factor,  as shown in Figure 2.   For
example, at an absolute gas pressure of 100 mm Hg, unit  density,  and
0.02 microns particle diameter, the Cunningham Correction  factor is
larger than 100.  Therefore, in essence, the  Mark  IV model is a
cascade impactor in which we have a high pressure  drop and low gas
absolute pressures on the downstream stages.
     In 1971-1972, we had a graduate student  who performed a laboratory
study on our Mark IV impactor.  A potassium sulfate aerosol  was generated
using a modified Dautreband nebulizer.   Scanning electron  micrographs were
taken of the particles deposited on the particle collection plates.  These
particle images-were counted and sized  using  a Zeiss Particle Analyzer.
When using the scanning electron microscope procedure it is necessary to
be careful with the application of an electrically conductive layer
(gold, palladium, carbon, etc.) such that you neither cover up nor
significantly alter the size of the collected particles.  After the
laboratory study, we constructed a field sampling  model  (Mark IVC)
which operates at about the 0.2 to 0.3  acfm gas flow rate.  The Mark IVC
sizes in the 0.02 to 0.2 micron diameter range and is located downstream
of the Mark III.
     Using the Mark III and IVC impactors,  we measured particle sizes
at a sulfite pulp mill and a pulverized coal-fired boiler.  Based on the
results of these tests we designed a Mark IVD model which is mainly oriented
towards sampling at the outlet of control devices  where  the particle mass
concentration is lower.  The Mark IVD samples at gas flow rates  up  to  1.8
acfm.  The Mark IVD sampling train was  improved to include a control
unit to house the pressure gauges and temperature  controls, a low pressure
drop tubular condenser, and a larger vacuum pump.   A 90-mm diameter filter
                                  266

-------
1000 h
PARTICLE DIAMETER
    (microns)
                          01
                                J	1   1
    10                   100                   1000
         Pressure (millimeters of mercury)
  Figure  2.   Cunningham Correction  Factor as a function
             of absolute gas  pressure.
                       267

-------
holder was used (instead  of the 47-mm diameter holder as is used in the
Mark IVC train) in order  to reduce  the gas pressure drop downstream of
the impactors.   It is  interesting to note that in some source tests we
have been collecting very little particle weight in the filter, indica-
ting that the particles are not bouncing on through the Mark IV even
though we are using quite high gas  velocities in the stage jets.  A
schematic illustration of the U of  W Mark III and IV cascade impactor
sampling train is shown in Figure 3.
     For sampling at the  inlet to control devices, we have been using a
BCURA cyclone, an extended version  of the Mark III which has 11 jet stages
(we call this a Mark V model), and  a Mark IVC.  With the greater number
of jet stages in the Mark V, we are able to divide the particle into a
greater number of size fractions which helps to prevent overloading of the
particle collection plates.
     Our Mark III-IV sampling data  at the outlet of control devices on coal
fired power plants has shown little particle mass in the smaller particle
size ranges,  approximately less than 2 percent by mass less than 0.2 micron
in diameter.
CALVERT:  What kind of particle diameter?  Physical or aerodynamic?
PILAT:  The particle diameter defined by the equation

                               2.61 y D. N1/2
                         n  = I	   J
                          60   VcVo

which is conventionally called the  "effective cut-off diameter" and  using
a particle density of  unity.  Thus  this D5Q is an aerodynamic  cut-off
diameter (diameter of  hypothetical  sphere of unit density  which  is collected
with an efficiency of  50  percent by the impactor stage  in  question,  regard-
less of the actual particle's true  size, shape, or density).
     At any rate, this is the  status we're  in  right  nowj-we're mainly heading
towards that higher flow  rate  system  in order  to  get a  sample downstream in
the same time increment as the  upstream sample.
     I have these handouts and  I also  have  a  paper  that we presented at
Vancouver,and I'll make that available to  those who  are interested,and I think
that that's about all  I care to say unless  you have  a few questions.

                                   268

-------
ro
                          „.._.      MARK 3
                          BCURA      IMPACTOR
                         CYCLONE       /
                          GAS FLOW
               MARK 4
              IMPACTOR

                L
                                                     TO
                                                 CONTROL BOX
                                                              90mm
                                                             FILTER
                                                             HOLDER
                                         THERMOCOUPLE

                                         PRESSURE TAP
                                                                       STACK WALL
                                                                                         CONDENSER -
                                            G1O

                                             ^
                                                                           CONTROL BOX
                                                        DRY GAS
                                                         METER
Figure  3.  U  of  W Mark III  - Mark  IV impactor sampli
                                                                                                   ing  train

-------
DISCUSSION

BQLL:  Do your particle diameters  assume  a  particle density?
PILAT:  Yes, a particle density of unity.
BQLL:  When you reduce the data, what formula do you use to calculate
the Cunningham Correction factor?
PILAT:  We use an equation reported by Davies in 1945 (Proc.  Phys.  Soc.,
Vol. 57, pp 259-270):

                     C = 1 + ^-[1.257 +  0.40 exp (-1.10 D,-n/2A)]
                             U50
where x is the mean free path of the gas  molecules.
BOLL:  The reason I asked is that  I understand that when the  Cunningham
Correction factor gets large and important, you have to pay attention to
how the gas molecules  bounce off the particles, because it makes a  difference
in the slip.  Sometimes they bounce off elastically, and sometimes  they
don't, and that's not important when you  are talking about correction
factors of 1.1 and 1.2; but if you are talking about 100 or 120, then it
may be important.
N.ICHOLS:  But that 100 is because  of reduced pressure.  You still have the
same differential mass between your particle and your molecule, as  if it
were a correction factor of 1.1.  You get the difference because of the
vast change in pressure.
ENSOR:  I think that a point to note here is that the Davies  equation for
C is an empirical fit to measured  data, and the question is whether the equa-
tion is being used for a gas pressure region in which the data were originally
obtained.
BOLL:  But it changes  with what the particles are.  You are not going to
mix water and oil particles.
PILAT:  Perhaps I should discuss some :of the possible problems.  In the
draft impactor guidelines which Bruce Harris handed out, the maximum gas
velocity is in the 70  meter/second range.  Well, we are exceeding that with

                                   270

-------
gas velocities up to 0.9 Mach number.  During our initial laboratory
tests, we did blast-off the deposited particles on some stages of the Mark
IV.  However, we have modified the impactor stage design to correct this
situation.  The proof of these improvements is provided by the scanning
electron micrographs of the particles collected during the field
testing.
     Another possible problem is the use of a Stokes number magnitude of
0.145 for the calculation of the particle D5Q.  Our studies have shown that
this Stokes number at 50 percent particle collection efficiency of a stage
is somewhat dependent on the jet Reynolds number and/or the absolute gas
pressure.  Our laboratory and field calibration work will clarify this
situation.  I hope this is not evading your question.
BOLL:  No, I think it is to the point.  But I do think one has to worry
about whether the Cunningham Correction factor is the same for each
particle material.
RAO:  The effect of material is very little.  It has been pointed out by
Fuchs that we all use oil drop constants for everything.
BOLL:  That is not my understanding.
CALVERT;  Lee Byers knows all about that.  He studied accommodation coefficients
with regard to thermophoresis.  I don't know whether the mystery got cleared
up.  It's been years.
BYERS:  We were working at atmospheric pressure but at high temperature.
OLIN:  Mike, when you drop the pressure, do you have a sonic orifice
between the Mark  III and the Mark IV?
PILAT:  No, we stage down the gas pressure progressively.  It  is someone
else who has the  system that passes the gas through a throttling valve.
OLIN:  Drops the  gas pressure down and the whole impactor  runs  at  a  lower
pressure.
PILAT:  This other system is for sampling the atmospheric  aerosol.
OLIN:  Does that mean that the last stage on your Mark IV  is running choked?

                                  271

-------
PILAT:  No, not knowingly.   We may achieve sonic  flow sometimes  because
we are pushing some test prototypes at higher gas flow rates  than they
were originally designed for,  mainly because we want to sample at the
outlet of control devices which have fairly low particle mass concentra-
tions.  But our design is not  such that we can go through a throttling
setup to lower the gas pressure substantially in  one stage at say Mach
one.  In fact, we are trying to even out the Mach numbers on  the stages
such that they are relatively  constant.
OLIN:  That is a little different approach that you are using compared to
the standard low-pressure impactor.
PILAT:  Well, it is mainly that we are trying to  avoid possible  problems
that may arise at higher gas velocities such as the sonic shock  wave at
Mach one.  It looks like, from the experimental data, the high gas
velocity may not be that important and this is based on the fact that
we are not seeing particles of too great a diameter (larger than
the upstream stage D-IQQ) in tne filter and in the lower Mark IV  stages with
field samples.  These particles are depositing on the proper stage
collection plates and this implies that the particles are not bouncing or
being blown off.
OLIN:  So the Mark IV then is  like your Mark III, but it has smaller holes
and less of them.  Is that it?  So you are just going to higher  and higher
gas velocities?
PILAT:  The main thing that is different on our Mark IV is that  it has
pressure taps on the stages, operates at a lower  absolute gas pressure, and
has a sequence of stage jet diameters and number  of jets per stage that
allows the gas pressure lowering to be achieved without having the gas
velocity go sonic on any one stage.
OLIN:  But the Mark IV operates at a lower pressure because you are dropping
the pressure across succeedingly smaller and smaller orifices.
PILAT:  Yes, now if you were to take the conventional commercially available
cascade impactor and just lower the gas pressure  at the impactor outlet  by
using a large vacuum pump, you can lower the stage particle  D50s somewhat.
Our first prototypes of the Mark IV were longer bodied Mark  III models with
some pressure taps on the lower stages to watch what was going on.

                                   272

-------
CALVERT:  I am not straight on the diameters.  The diameters you reported,
are these the computed physical diameters?
PILAT:  Yes.
CALVERT:  Isn't it the aerodynamic diameter?
PILAT:  Well, it is the same aerodynamic diameter as reported by the other
cascade impactors when calculated using the equation presented earlier.
CALVERT:  What I call aerodynamic diameter is the diameter multiplied by
the square root of the density and square root of the Cunningham
Correction factor.
PILAT:  I am referring to the stage D™.
OLIN:  For an equivalent spherical particle with density of one.
LILIENFELD:  I have two questions,  the first is what was the pressure
at the last stage?
PILAT:  This has varied over a considerable range, from say 3 to 10 inches
of mercury.  This range is dependent upon the Mark IV stage configuration
(number of stages, jet diameter, and number of jet holes per stage) and  upon
the capabilities of the vacuum pump used.
LILIENFELD:  The second question is what kind of substrates are you using?
PILAT:  Excellent question.  We have used flat foils (non coated foils)
and greased substrates;.  In the Mark IV at higher gas temperatures, there
is a  problem with the weight loss of the grease.  We have done quite a bit
of testing with no grease at all.  The scanning electron micrographs of
these samples without grease looked quite good with regard to not having
particles larger than expected on the Mark IV foils.
RAO:  What about wall losses?
PILAT:  I don't have the numbers here, but our test procedure frequently
includes washing all of the parts using an ultrasonic cleaner and then
evaporating the solvent to dryness and then weighing the residue.  We have
some disagreement between our field data and that laboratory data reported
by SRI.  We have not been able to find the nozzle losses which Wally did.   In
fact, we were strongly considering changing our nozzle design, but decided
against it after looking at our field data.

                                   273

-------
SMITH;  One thing that we did not do during  our laboratory tests  is  to
load up the impactors.  That makes a difference.
PILAT:  Yes.
SMITH:  You are talking about milligrams  which are pretty large quantities
of material.
PILAT:  The percentage particle losses appear not to be that great at all.
SMITH:  With regard to coating surfaces and  so forth, I think it  is  hard
to generalize effectively the ultimate jet velocity.  This is definitely
a function of the particle size too.  In the Sierra impactor, for example,
the jet velocity was increasing with decreasing particle size, but the
particle bounce was decreasing at the same time as you went to the smaller
particle sizes.  So who knows what happens when you go to the submicron
particle sizes.
PILAT:  Yes, in fact we have been working on coal-fired power boilers
(research supported by the Electric Power Research Institute) and it
appears that our main particle overloading (and then blow-off) problems
may be in the 4 to 6 micron diameter range,  not in the Mark IV size range.
OLIN:  Well, while we are dwelling on this subject of bounce, maybe this
should be held for discussion, but it seems  to me that as you go  to smaller
and smaller particles, that the bounce effect does get smaller and smaller
because it has to be something to do with the particle momentum which has
the particle diameter cubed in it.  It is particle mass times velocity
so that the effect goes down like the cube of the particle size.   Further-
more, the Van der Waals active forces zoom up for the smaller particles
and thus, there are a lot of things going for you as you go to smaller and
smaller sizes, which eliminates the problems of bounce.  My feeling is that
above 3 microns, you have bounce and below that I'm not sure.
PILAT:  One of our graduate students designed a prototype Mark IV with very
sharp cut-off curves and measured the collection efficiency versus particle
size.  His data showed that the larger particles were very effectively
scoured off the particle collection plates.   This illustrates the fact that
it is possible to design and build a low pressure cascade impactor that
does have the particle bounce and/or scouring problems.
                                   274

-------
	:  Mike, when you were using your  impaetor train in the field, did
you have both the Mark III and Mark  IV  plates greased, or just one?
PILAT:  At times we actually  had both impactors uncoated.  Sometimes we
had grease only on the Mark III collection  plates.  We have not yet
operated a simultaneous  test  with one train using greased foils and the
other train  having uncoated inserts.
	:  Mike, I understand you to say that  you intend to calibrate at the
small particle sizes.  Could  you describe this?
PILAT:  We are working on generating an electrically conductive submicr.on
aerosol so that we will  not have to  coat the particles with gold, graphite,
etc.  We have a gold  hot-wire aerosol generator constructed and we are
running tests with it.   We will use  scanning electron microscopy to size
the particle deposits.
	:  How  do you weigh the  particles  collected with your low pressure
impaetor?
PILAT:  We use a Mettler balance.   It is a  real problem to achieve accurate
weighing with the small  samples.  We have not been  successful in using the
Cahn electrobalance,  either the model C2 or the model 4100, and we tried
quite a bit  on this.
WILLIAMS:  What type  of  weights were you trying to  weigh?
PILAT:  We like, of course, to  get  samples  in say,  the milligram range.
But sometimes we  have low weights down  in the 0.1 to 0.01 milligram
range.  You  can sometimes  have  a visible particle sample that is in the
noise range  of the balance.
SPARKS:  Thank you, Mike.
                                    275

-------
          SUBMICRON PARTICLE SIZING EXPERIENCE ON A SMOKE  STACK
                USING THE ELECTRICAL AEROSOL SIZE ANALYZER
                  Gilmore J. Sem,  Thermo-Systems, Inc.
SPARKS:  Gil Sem from Thermo-Systerns will  now talk about the field use
of the mobility analyzer for obtaining size information down to below a
few tenths of a micron.

SEM:  The items I plan to cover today are:
      1.  A brief review of the design and calibration of the electrical
          areosol size analyzer,
      2.  A report on the successful use of the analyzer to measure the
          penetration of submicron particles through a baghouse filter
          on a coal-fired power plant, and
      3.  A discussion of problems we observed during the field pro-
          gram and suggestions for further work to solve these problems.
WHY SHOULD WE MEASURE SUBMICRON PARTICLES IN STACKS?
     Although particles smaller than 1 ym diameter are only a small fraction
of the total mass emissions of particles from coal-fired power plants,  these
particles are a very important component of the air pollutants emitted
from the stack.  Submicron particles remain airborne in the atmosphere  for
periods of days, traveling long distances from the source.  When inhaled,
these particles penetrate deeply into the human respiratory system, posing
a more serious health hazard than the -larger particles.  The potential
harmfulness of submicron particles, and their penetration through control
equipment, is a function of particle size.  Thus, measurement of the parti-
cle concentration as a function of size is important for characterizing the
performance of control equipment as well as for characterizing the potential
harmfulness of the emissions.
THE ELECTRICAL AEROSOL SIZE ANALYZER
     The electrical analyzer, either in its present form or in the earlier,
much bulkier form,  has been used for over 10 years for the measurement of
ambient urban air,  smog chamber, and laboratory aerosols.   It's only
within the past year, or perhaps, past 6 months, that  it has been  applied
to stack aerosols.   Southern Research Institute, under Wallace Smith and Joe
                                  276

-------
McCain, has used it in several types of stacks, mostly within the last 3
months.  Battelle Institute in Columbus has used it within the past 3
months for coal-fired power plant evaluation.  Monsanto Research Corp.
is preparing to use an analyzer for control equipment evaluation.  Meteor-
ology Research Inc., under David Ensor and Richard Hooper, is using it for
baghouse evaluation on a coal-fired boiler.  I believe one or two other
groups are in the process of applying the analyzer on gas turbine engines.
     Figure 1 shows the analyzer, consisting of 2 modules:  a larger flow
module (35 x 30 x 62 cm LWH) and a smaller control and readout module.  The
entire instrument weighs 20 kg excluding vacuum pump.  The instrument
basically places a known, reproducible electrostatic charge on each parti-
cle; a second stage removes-particles smaller than a known, selectable size
from the aerosol stream; a  third stage measures the charge level carried by
particles passing through the second stage.  The measured charge level of the
third  stage  is proportional to aerosol concentration.  I will not discuss the
theory any further in this talk.
     Aerosol enters the center tube shown in Figure 1 at 4 LPM.  An additional
46 LPM enters the instrument to serve as sheath air.  The two streams do
not need to come from the same source, but should be within several degrees
of the same temperature and neither should be saturated with water vapor.
     An analyzer has been calibrated by exposing it to a series of mono-
disperse aerosols of known sizes.  All analyzers are compared with a cali-
brated standard instrument before shipment.  If they are kept clean by proper
cleaning every few weeks of operation, the calibration remains constant.
However, it is best to compare a field unit with a laboratory-based instru-
ment or with a known test aerosol before any critical program.  If this
precaution is followed, almost any field malfunction will be obvious from
the data.  Incidentally, I strongly recommend that anyone on a field program
bring along a spare charger assembly.  The charger is the most "breakable"
item on the analyzer and can be field repaired only by replacement.
                                   277

-------
Figure 1,   Photograph of the Model  3030 electrical  aerosol
           size analyzer.
                           278

-------
FIELD APPLICATION TO STACKS:   NUCLA POWER PLANT
     Figure 2 shows the power plant near Nucla.   It  is  small,  only  36 Mb!
total; 12 MW in each of 3 boilers.   The coal  is  mined about  6 to 8 miles
away,and its composition is quite variable.
   Figure 2.  Nucla Power Plant of the Colorado-Lite Electric  Association.

     We measured the No. 2 boiler emissions.   We had an excellent set of
sampling ports on the downstream side of the  baghouse.   It was  more than
10 diameters downstream of the last bend at the bottom of the stack.   We
had sufficient space for equipment.
     The effluent from the stack was not visible at any time.  The visibility
as measured by an MRI stack nephelometer was  in the range of  22  km
inside the stack downstream of the baghouse.   This was verified by the
electrical analyzer which measured submicron  particle concentration levels
similar to those measured in clean rural outdoor areas.
     Our equipment, shown at the downstream location in Figure  3, included
a diffusion battery of parallel plate design, a Gardner nuclei  counter,  an
electrical analyzer and (not shown) a diffusion drying system.   The down-
stream sample coming from the stack was heated to about 160°F up to the
entrance of the first diffusion drier.  The aerosol than passed through
about 3 to 4 feet of polyethylene tubing to a  second diffusion drier before
                                        279

-------
Figure 3.   Instrumental  setup downstream of the  baghouse  showing  the  EASA,
           the diffusion battery,  and the Gardner  nuclei  counter.
                                   280

-------
entering the electrical  analyzer and diffusion battery.   The  analyzer  sheath
air was dried before it  entered the analyzer.   Filter  samples were  taken
behind each diffusion battery stage in addition to Gardner data.  On some
runs, the analyzer sampled behind each diffusion battery stage in an attempt
to get one more field comparison of the analyzer and diffusion battery.  I
have no diffusion battery or filter data to show at this time; Ensor and
Hooper will report fully at a later time.
     The equipment, shown at the upstream  location in  Figure  4 without the
diffusion battery, included a dilution system capable  of ratios at  least as
great as 1000X.  The emissions sample, heated to 160   to 220°F, enters a
diffusion drier, then passes through the diluter, and  finally another
diffusion drier before entering the analyzer and/or diffusion battery.
Dilution ratios of about 20-30 X were required to bring  the concentration
within range of the analyzer.
   Figure  4.   Instrumental setup upstream of the baghouse showing the sample
              port,  diluter,2 diffusion driers, the EASA, and the strip chart
              recorder.
                                     281

-------
     Figures 5 and 6 show upstream distributions  measured  on  Nov.  12  and
Figure 7 shows the downstream distribution  measured  on  Nov. 12.  Note that
the concentration scale is 5000X more sensitive  in the  downstream  graph.
Except for the peak at 0.015 ym on the downstream distribution which  I
will discuss later, the shapes of the distributions  are similar.
     Figure 8 shows the upstream distribution and Figure 9 shows the  down-
stream distribution on Nov. 13.  All  5 sets of data  were taken with about
11 MW (near full load) generation on  Boiler No.  2.   Note that Nov. 13
had similar data to Nov. 21 with some rather minor differences,  including
a much lower peak around 0.015 ym on  the downstream  data.   Note  that  most
of the volume, and therefore mass, of the aerosol is contained by  parti-
cles greater then 0.1 ym.
     Figure 10 shows the volume distributions plotted with a  log concentra-
tion scale.  We see that the  six  averaged  sets of data  plotted here,
representing 32 separate size distributions, are quit consistent.
     Figure 11 shows the penetration  of the baghouse for the   three pairs of
data;  two on Nov.  12 and one on Nov.  13.  Note that  penetration is about 0.0001
between 0.1 and 1.0 ym and may be slightly  greater below 0.1  ym.   Although
the measurements may certainly be a factor  of 2  in error,  the baghouse
definitely has high efficiency, even  for particles below 0.1  ym.
     An important point regarding the data  reduction:  all distributions
shown are averages of multiple analyzer measurements.  Since  the concen-    ,
tration of aerosol in the stack varies by typically  + 10 percent from one
10-second period to another, even with good combustion  control;  any
single distribution may not be very accurate.  By averaging a number of
measurements, preferably  five or  more,  much greater accuracy can  be  obtained.
The use of a strip chart recorder allows one to  see  more easily  if the con-
centration level is changing during a run.   Figure  12 shows typical  strip
chart traces of upstream effluent with the  EASA,
PROBLEMS, POSSIBLE SOLUTIONS, AND SUGGESTED FUTURE WORK
     The two downstream distributions shown earlier, especially on the Nov.
12 data, show a very significant peak in the measured volume distribution
between 0.01 and 0.018 ym.  It should be emphasized  that the distributions
shown  are volume distributions, not surface or number distributions.  The

                                   282

-------
   30 r
to
    25
 E
 0
10
O
 a.
 a  20
    '5
 oc


 Ul
 o
UJ
2
 2

 UJ

 I-
 V)
 Q.
    10
NUCLA , COLORADO,  POWER PLANT
UPSTREAM OF  BAGHOUSE
18.37 X  DILUTION

DATA SHOWN UNDILUTED
   NOV. 12, 1975

 14:38-14:44 HRS

AVERAGE OF 4 RUNS
0

.005
        i  i I  i u
.01     .02       JOS     .1

       PARTICLE  DIAMETER,
                                          .2
                                        .5
IX)
 Figure  5.  Upstream particle size distribution, 14:38 - 14:44,
           Nov. 12, 1975.
                               283

-------
    30 r
 o
 i  25
10
 O
 Q.

 Q
 <
 >
    20
     15
 
-------
ro
10
 i
 £•  4-
oc.
t-
UJ
O
 Ul
 cc

 CO
g  0
       NUCLA, COLORADO, POWER PLANT
       DOWNSTREAM OF BAGHOUSE
       NO DILUTION , 2 DIFFUSION  DRIERS
                  NOV. 12, 1975
                 I6:20H6'40HRS
          AVERAGE OF
            5 RUNS
                       AVERAGE OF
                         6 RUNS
   MM
                                        I    II  I  III  I 11
.005    .01     .02       .05      .1      .2
               PARTICLE DIAMETER, Dp,
                                                 .5
                                                       1.0
 Figure 7.  Downstream particle size distribution,  16:20 - 16:40,
           Nov. 12,  1975.
                     285

-------
    40,-
    35
    30-
    25
to
g
x

£
    20
 o
 fj-
 I
 §
 Ul
 Ul

 (O
 Q.
     10-
           NUCLA , COLORADO, POWER PLANT
           UPSTREAM OF BA6HOUSE
           31.45 X DILUTION
           DATA SHOWN UNDILUTED
   NOV. 13 ,1975
 I7:50-I8:I5HRS
AVERAGE OF 12 RUNS
     .005    .01     .02       .05     .1      .2

                    PARTICLE DIAMETER, Dp , /im
                                          .5
1.0
 Figure 8.  Upstream particle size distribution, 17:50 - 18:15,
           Nov.  13, 1975.
                          286

-------
to

 o
 ex
O
<3
  A
Z
O
 LU
 O


 8
 UJ
 UJ
 a:
1    0
     .005
           NUCLA, COLORADO, POWER PLANT
           DOWNSTREAM OF BA6HOUSE
           NO DILUTION , 2 DIFFUSION DRIERS
             .01
   NOV. 13,1975

 2l:38-2l'44 MRS

AVERAGE OF 4 RUNS
.02       .05      .1     .2


 PARTICLE DIAMETER, Dp
.5
Figure 9.  Downstream particle size distribution,  21:38 - 21:44,
           Nov.  13, 1975.
                        287

-------
IU
I05
I04
"e
,0
Vo3
a.
O
o»
> 10^
o
z
g
ONCENTRA'
O
o
UJ
;§ i
o
< J
a.
ni
- O RUN 1 NOV. 12,1975
A RUN 2 NOV. 12,1975
= Q RUN 3 NOV. .3,1975 UPSTREAM
- OF BAGHOUSE
Q
: © @ °
@ A
_ ^ A A
•*
•^
: i
•*
i^
-
^ 8
MM
^H
~ O
E O
E ° o © &
o Q
= DOWNSTREAM
E OF BAGHOUSE
- Q
-
EG
- s
I Q
1 1 I i i i II i i i i i I i I i i
    .01     .02      .05     .1      .2       .5      (JO
             PARTICLE  DIAMETER, Dp
Figure  10.  Composite particle size distribution,  Nov.  12-
           13, 1975.
                     288 ,

-------
     1.0
o
5
QC
^^

8    -05
tu
Z    .02
UJ
•I-
I    .
    .005
x
u
 -  .002
    .001
  .0002
   .0001
 .00005
     .02
                            —O—  RUN  I    NOV. 12,1975
                            —A—  RUN  2    NOV. 12,1975
                                    RUN  3    NOV. 13,1975
                                                         	A--^^r	©
                                                                -o-
      .05           ,1            .2                 .5

        PARTICLE  DIAMETER ,  /tm

Figure 11.   Baghouse  penetration of submicron particles.

                  289
1.0

-------
Figure 12.   EASA strip  chart  recording upstrean of the baghouse,
                                290

-------
overwhelming number of particles in the 0.01 - 0.018 ym size range on the
Nov. 12 data accounts for over 99 percent of the total measured number
concentration.  The only way we could accurately measure this distribution
was to record this channel separately and then increase the chart recorder
sensitivity to get the size distribution of larger particles.  I have rarely
seen an aerosol in this size range which is so monodisperse.
     Figure 13 shows a typical downstream strip chart recording of the
entire size range,and Figure 14 shows the trace for 0.03 ym and larger.
     Particles in the 0.01 - 0.018 ym range are probably produced by
condensation within the sampling system shortly before the aerosol enters
the analyzer.  The aerosol almost certainly is not produced in the analyzer
itself because the nuclei counter also detected the particles at the inlet
to the diffusion battery.  Joe McCain of Southern Research Institute has
also seen this aerosol and suggests that it is an acid mist, possibly
HpSO^  He can usually get rid of the very small particles by sufficient
dilution.  We could not dilute the downstream samples to prevent conden-
sation because the undiluted aerosol concentration was already approaching
the lower detection limit of the analyzer.
     I feel that the data above 0.05 ym  are representative of what  is
in the stack because the volume distribution reaches a very low level around
0.05 ym and then increases, an indication of 2  independent, nonoverlapping
distributions.  We collected a yellow-tinted liquid in a water trap between
the 2 diffusion driers which smelled like acid, but we did not analyze the
liquid.  Since this aerosol is casting some doubt on the validity of the
data below 0.05 ym, some work should be done to identify it and prevent it.
     The possibility that the 0.01 ym droplets are generated in the sampling
system is just one illustration of the need for careful development and
evaluation of the sampling system for submicron aerosols.  Since particles
below 0.1 ym are governed by diffusion, coagulation, and condensation
rather than gravity and inertia; a sampling system optimized for  1-10 ym
particles will not necessarily work well for particles below 0.1 ym.
                                   291

-------
Figure 13.   EASA strip chart recording  downstream of baghouse  showing
            the data for 0.01  -  0.2 um  particles.
 Figure  14.   EASA  strip chart recording downstream of baghouse showing
             the data  for 0.03 - 1.0 ym particles with the 0.01 -
             0.03  ym particles not  included.
                                  292

-------
     The data shown earlier were taken during relatively constant plant opera-
tion with no shaking of the filter bags.  When the boiler was pushed to its
maximum of 13 MW, the upstream concentration fluctuated much more and the
baghouse went into a continuous shake cycle.  The total current measured with
the analyzer on the downstream side showed an increase of greater than 2X
whenever a bag was shaken.  The concentration level did not reach anything
resembling constant conditions for at least 10 minutes after shaking.  Since
a continuous shake cycle calls for one shake every few seconds, no data were
taken during shaking.
     Similarly, when the coal composition fluctuated, so did the submicron
aerosol.  When the power plant was operating at non-ideal conditions one
day, the submicron aerosol concentration fluctuated by 3-4X at the inlet to
the baghouse over a time period of 10  to 20 minutes.  No valid size  distribu-
tion data could be taken during these periods.
     Since these unstable periods occur often in a steam boiler, and indeed
may be more typical than stable conditions on many boilers, it would seem
important to try to characterize the emissions during unstable periods.
Batch sampling, as illustrated in Figure 15 with the SRI sampling system,  .,
can be used to obtain submicron aerosol samples during unstable effluent
conditions.  The submicron instruments can then operate on a stable batch
of aerosol immediately after  the batch sample is taken.  This technique
has been used successfully for measuring other combustion aerosols such as
cigarette smoke, candle flames, propane torches, auto exhaust, and others.
I have constructed a batch sampler shown in Figure 15, which can fit into
the sampling-dilution system  as shown in Figure 16.  It uses a flexible
plastic bag within a rigid-wall housing.  The bag is filled by drawing air
from the chamber between the  bag and the rigid housing.  I have not yet
had a chance to try the batch system.  One word of caution:  do not allow
sunlight to reach the aerosol in a batch sampler or you may have a photo-
chemical smog chamber.
     It is important to note  that some of the cascade impactor data  taken
in stacks does include bag shaking, electrostatic precipitator rapping,
coal composition variations,  changes in operating conditions, etc.   Thus,
data from the submicron near-real-time sensors may not match well with
cascade impactor data in these cases.
                                   293

-------
           FLOWMETER
                          DRY , NEUTRAL
                         DILUTED , AEROSOL
                            SAMPLE
                       Kr -85 NEUTRALIZER
                         (OPTIONAL)
                        DIFFUSION DRIER K&  S
                          (OPTIONAL)   "**!  '"
    CYCLONE
   (OPTIONAL)
    tft
   EFFLUENT
   STREAM
                                                          EASA
                                                                        PUMP
               RECIRCULATINC
                CLEAN , DRY
               DILUTION AIR
!p

1 1
1 <—

-\
ION
/


MANOMETE
4 1
^=i










R


!
i
1

1
1
1
1
1
	 »•
1
1
I
1
I
1_
1
1
PERFORATED FALSE END PLAT
^V— ->>__——-—_— -^
^ ^N
-ITx
PERFORATED SAMPLE \ »
EXHAUST TUBE 	 ^J |
1 1
| 1
Ij
T- MEMBRANE I !
\ 	 s J
\ 	 £>
S~
<- RIGID-WALL CHAMBER
^- BATCH SYSTEM

I . i f
E
P
|
I
1
I
1
1
1
1
1
1
1

	

^
•





»


i
1
— i


                                  FILTER
                                             DRIER
                                                                        PUMP
                                                               BLEED VALVE
Figure 16.   Southern Research Institute sampling-dilution system modified
             with  batch sampling system suggested for  use on  emissions
             streams with  fluctuating  aerosol  concentration and/or size
             distribution.
                                            294

-------
Figure 15.   Photograph  of  a  batch  sampling system.
                              295

-------
     The process variations point out the desirability of measuring up-
and downstream simultaneously.   Since we had only one instrument at Nucla,
we settled for the assumption of constant aerosol.
     We are confident that we obtained good, valid  penetration data from
the Nucla Power Plant using the electrical  analyzer.   However, we feel  that
this is just a beginning.   Much work remains to develop a good sampling,
dilution,and cooling system to assure the delivery  of a valid sample to the
sensor under a wide variety of stack effluent conditions.   After that,  much
work will be needed to evaluate many control devices  on many different
processes as well as to characterize the submicron  aerosol being emitted
into the atmosphere.
     However, to emphasize the point as strongly as possible, the sample
extraction, conditioning,  and delivery system is the  key to accurate sub-
micron sizing of stack effluents with the electrical  analyzer.  The majority
of measurement errors at this time are attributable to the sampling system,
not to the sensor.  The future of this technique for  submicron particle •
sizing in stacks appears bright if the sampling problems are solved and
fully evaluated.
ACKNOWLEDGMENTS
     This work was partially supported by Electric  Power Research Institute,
Robert Carr, Project Officer.  Dr. David Ensor and  Mr. Richard Hooper
invited me to work with them on their major baghouse  evaluation program.
I appreciate the opportunity to work with them and  to present some of their
results in this talk.  The field work was performed at the Nucla, Colorado
power plant of the Colorado-Ute Electric Association.
                                   296

-------
DISCUSSION
                    t
WAGMAN:  I notice that you plot your data out to 1 micron; isn't this
beyond the minimum in the mobility-size relationship, so that you have
some overlap with the smaller size particles?
SEM:  The minimum in the mobility curves occurs at something like 1.5 or
2 microns, and certainly the data are not as accurate up around 1 micron as
they are around a tenth of a micron, for example.  If I were to apply error
bars for the aerosol that was reaching the instrument, I would apply some-
thing on the order of +_ 50 percent by the number for the largest size that
I showed and perhaps more on the order of +_ 15 to 20 percent around a tenth
of a micron.  When you get down to 0.01 micron, I would apply something
like a 50 percent error again.
WAGMAN:  How do you presently define the size that you get from a mobility
analyzer?  In a cascade impactor, for example, we refer to it as inertial
or aerodynamic diameter.  How do you define the size that you get from a
mboility analyzer?
SEM:  The size is not a nice clean characteristic like aerodynamic diameter,
but it is not all that different from it.  The instrument is calibrated by
using monodispersed aerosols,which are essentially spherical,and feeding
them to the instrument in a number of sizes of monodispersed aerosols.  So
  •f
really, it's an equivalent size based on the spherical particle and it's
close to a diameter squared kind of relationship.  So it's not all that
far from an aerodynamic diameter.
WAGMAN:  Isn't there a problem here depending on which material you select as
your calibrating material?  You're tying it into a material of that composi-
tion, in as much as it would be a dependence of charging characteristics
on the composition of the particles that you happen to use?
SEM;  The instrument uses diffusion charging, rather than field charging.
In an electrostatic precipitator in a power plant, composition certainly
does make a difference because you're using primarily field charging there.
The instrument uses diffusion charging,which strictly depends on  the statis-
tical chance of an ion approaching a particle with essentially enough momentum
to stick to it.
                                   t,
                                  297

-------
WAGMAN:   But the question of whether it sticks to it may very well  depend
on composition.
SEN:  Not for ions on particles in this size range.   It's essentially
independent of composition.
WAGMAN:   I have spoken to Earl  Knutson about this matter.  He actually
conducted work some years ago on this.  He said that he did find a  dependence.
SEN:  I  think that you'll find, though, that that dependence was very small.
We're talking about a 5 or 10 percent kind of dependence from perhaps a
metallic particle which was  very conductive to a plastic type particle, for
example.
	:  Gil, you do not have an electric field in the charging region?
SEM:  There is a very small  electric field, about 50 volts across 1 centi-
meter.  That's all.
OLIN:  Two-stage charging  generates them at high voltage and then  moves
them across the gap at low voltage.
LILIENFELD;  What is the space charge field?
SEM:  The NT product is 107>
LILIENFELD:  Do you calculate the space charge field?
SEM:  You can calculate the charge level, and that comes out to something very
near NT = 10 .  But we also measure the current which is flowing across the
aerosol  gap.and so actually we're measuring the ion current which is
passing through the very fixed volume that the aerosol is passing through.
Those two numbers tend to agree with each other within a few percent.
(David Piu of the University of Minnesota has determined that space charge
contributes about 15 percent to the field in the charging region).
LILIENFELD:  Are the 50 to 100 volts per centimenter the result of the space
charge?
                                                          i
SEM:  No, that's strictly an applied voltage to get the ions to move across
the aerosol stream.
                                  298

-------
LILIENFELD:  Well, I think that the space charge could be several  times
higher than that; the field produced by the space charge.
SEM:  You'll have a minor effect due to that field, certainly.   But an NT
of 10  fully predominates on the charging mechanism.  I have several
references which I can refer you to on questions of the design  and theory
and the calibration of the instrument in case you're interested.  (See
Pui, D.Y.M., Ph.D. Thesis, University of Minnesota, Mechanical  Engineering
Dept., 1975).
WAGMAN:  Have you ever used this instrument downstream of an electrostatic
precipitator?  Do you have any problems there?  I wondered because Bill
Kuykendal this morning was talking about our troubles of using  it  downstream
for a particulate which was charged.
SEM:  Joe, can I refer that question to you?  I haven't.
MCCAIN:  We've done a limited amount of work with comparisons of diffusional
results with this—apparently successful.  We use charge neutralizes in
the dilitters.  We don't have one at point of entry into the probe.
	:  Did you use Krypton 85?
MCCAIN:  No, Polonium 210.
KUYKENDAL:  I think that may have been part of our trouble.  We were in  a
time tight experimental procedure and we just didn't have time  to  optimize
the sampling system.  We tried the one charge neutralizer that  we  had
available, a Krypton 85 source.  We were not using dilution.  It didn't  work.
SEM:  I think, also, in a place where you have a powder which has  been re-
dispersed, the submicron particles are not redispersed as well  as  are the
particles above 1 micron.  You could have a very definite difference due to
a small change in relative humidity,and how well you are dispersing the  sub-
micron particles which would definitely give you trouble if you are trying
to size them.
MCCAIN:  Did you (Kuykendal) have a way to knock out relatively large parti-
cles ahead of the mobility analyzer?
                                   299

-------
KUYKENDAL:  Other than in the sampling  system itself,  we  didn't.
SEM:  It is a good idea, if you have predominately large  particles, to knock
out the particles with something,  let's say a 2 micron cutoff device,
either cyclone or a good impactor.   Any other questions?
ENSOR:  Our data are  really  in  preliminary  form; there  are many more  things
we can do.  We have some preliminary chemical anaysis  here  from  impactors,
which indicates it was sulfur-rich  brine on the plates.  There is  a possi-
bility that we may have been below  the  acid dew point  in  the stack itself,
and it may not have been particles  in the stream.   Again, these  are all
preliminary data.
MCCAIN:  Well, we've seen it where  we've definitely introduced it  in the
train.  We made a change of dilution of as  little  as 10 or  15 percent,
and this resulted in a change in concentration with the condensation nuclei
counter by a factor of almost 10 .
ENSOR:  My question is, how do you  really know what was created  in the train?
I mean you're manipulating something in the train  and  suddenly things
change.  That's not conclusive proof.  What I'm saying is,  this  is a poten-
tially serious problem in submicron testing, and it's  not always straight-
forward.
                                 300

-------
                 PANEL DISCUSSION ON DIFFUSION BATTERIES

              Joseph D. McCain, Southern Research Institute
SPARKS:  We'll start the discussion of diffusion batteries and their use
in-stack, their use out in the field to take size distributions somewhere
around .01, .05 to a few tenths of a micron particle diameter.  They fill
in the gap below the impactor region.  The way that we'll do this is to
allow Joe [McCain] of Southern Reserach, Dave [Ensor] of Meteorology
Research, Seymour [Calvert] of APT—  Pedro [Lilienfeld]  or Reed  [Cass] does
somebody from GCA want to talk about theirs?.
LILIENFELD:  Perhaps.
SPARKS:  Perhaps.  Well, we'll let these three, anyway, talk about the things
that they are doing.  Joe [McCain] and Dave [Ensor] both have more of a
classical diffusion battery configuration, parallel plates.   Seymour [Calvert]
has decided that chemical engineers knew something about mass transfer to
cylinders, so he developed a new geometry of diffusion batteries.  If each one
of you could maybe take about 10 minutes to 15 minutes at the most to say
what you're doing, then we'll let Pedro [Lilienfeld] decide if he wants
to say anything, and let anybody else make a comment.  Joe, if you want to
go first.

MCCAIN:  We started in this several years ago, trying to characterize some
fine particle emissions from a pulp mill, and some of the equipment we made
then is still in use today.  It was pretty unconventional then for stack
sampling.  We were probably conservative in our diffusion battery design,
and you'll see the result of that shortly.  Let me run through a little
bit of the history of our field work with some slides.  For those of you who
aren't familiar with a diffusion battery system, the detectors that are
used for diffusional sizing are not capable of coping with the particle
concentrations that are normally encountered in flue gases and extensive
sample dilution is necessary as Gil Sem has already mentioned.  Figure 1
is an earlier version of the slide that Sem used showing the essentials of

                                   301

-------
                                                             Flowmeters
      Cyclone Pump
Neutralizer
                Participate
                Sample tine
                                   Diffvisional Dryer
                                       (Optional)
              St-t—£'  Dilution  ^>
                                         Neutralizer    Pressure
 Recirculated
 Clean Dilution
 Air
         Filter
                                   Bleed
        Figure 1.  Optical  and diffusional sizing system.
                               302

-------
our sample conditioning system which involves a metered flow into  a  dilution
device from which a secondary sample is taken to the various particle  moni-
tors (an optical particle counter, CN counters reading total concentration
of ultrafine particles and diffusion batteries with a condensation nuclei
counter downstream of them.)
     The diffusion battery operates on the principal of removal  of particles
by Brownian motion, which results in the migration of particles to the walls
of the battery.  The diffusion batteries which we have generally used  in our
field work are of the parallel plate type, using one millimeter separation
between the plates.  The residence time for gases flowing through those
batteries ranges from a few seconds upwards to an hour.  They remove par-
ticles on the order of  .01 to 0.3 ym in size.  The diffusional rates of
moderately large particles are quite slow, and consequently the required
residence time in the battery is long, which results in long delay times
in the measurements.  This does not make the method particularly suitable
for obtaining data on a rapidly fluctuating source.  Seymour [Calvert]
will talk, I'm sure, on the screen type which has  much shorter transport
times.  We are trying some screen type diffusion batteries too, but he's
got a lot more experience with them than we do.
     We try to keep all aerosol transport lines as short as possible to
minimize losses in the sampie lines prior to entry into the diluter or
instrumentation.  A charge neutralizer is used in the diluter to remove
electrostatic charges from the particles at that point, but it would be better
to have it in the stack.  We have used Climet and Royco optical particle
counters with pulse height analyzers, General Electric condensation nuclei
counters (which weigh about 150 Ibs) and an Environment-One condensation
nuclei counter that we've had moderate success with (you can't use it as it
arrives from the factory).  It uses pneumatic valves that operate on pressure
differentials of about 2 inches of water, and if you put the inlet or exit of
the counter at a differential pressure to ambient on the order of 2 inches
of water or greater, it'll lock the valves open or shut and it won't function
at all.  We've modified ours to at least partially alleviate this problem.
It has been used, when well treated, with reasonable success in the field.
As a standby device we generally take along a Gardner manually operated CN
counter in case all of the automatic ones fail.  The G.E. CNC device is a nice

                                   303

-------
rugged instrument, but the 150 Ib weight does make it a little big  to carry
around (G.E. is now making a smaller, lighter version).
     Our diffusion batteries consist of a small  one with 12 1-mm wide,
10-cm high, 38 cm long channels and four large ones, each of which  has
96 channels of the same dimensions as those on the small one.  Each of the
large ones weighs about 65 Ib .  Figure 2 illustrates one of the diffusion
batteries and shows its penetration characteristics.
     We could have made them a good deal more compact and somewhat  lighter
but we were very conservative.  We wanted them to work the first time out
and we wanted to be sure that we didn't have any manufacturing problems in
the critical dimensions.  We used a slot geometry with the long axis verti-
cal so that settling wouldn't play a significant role in particle losses.
As sample extraction is necessary with these instruments one of the problems
one immediately encounters is probe losses.  Figure 3 shows a plot  of
theoretical probe losses based on settling losses in long tubes at  one
extreme in size and diffusional losses at the other.  For moderately high
flow rates, on the order of 0.5 or so (10 1/min) even with a 10-ft  probe,
the losses are generally inconsequential over the range 0.01 to 1 micron.
Thus, sample extraction for this size range does not introduce significant
errors.   Electrostatic effects might play an additional role.  We
haven't  really explored that, other than in some limited experiments
in the lab.  In those experiements they did not seem to be particularly
bad over the range from about a half micron down to 0.01 ym  in probes of
the dimensions that we use and at the flow rates that we use.  Electrostatic
effects in diffusion batteries themselves can be very significant and charge
neutralization prior to passing the aerosol through the diffusion battery
is required.  One can get electrostatically induced losses  of 90 percent in
a diffusion battery very easily for some sizes.
     Figure 4 shows the overlap in size distribution on a number basis
obtained using three different methods.  The plot shows the differential
size distribution (dN/d log D) versus size as obtained with  cascade
                                   304

-------
 Figure  2a.   Parallel  plate diffusion  battery.
             or 96 channels 0.1  x  10 x 48 cm.
The batteries have 12
       #70-
                  0.01
                            ARTICLE DIAMETER, M"»
Figure 2b.  Penetration curves for monodisperse aerosols (12 channels,
            0.1 x 10 x 48 cm).
                               305

-------
                                     FLOWRATE
    DIA,
  ,64 CM
  ,64 CM
  .64 CM
 1.6 CM
LENGTH
183 CM
305 CM
183 CM
305 CM
 0
  ,01
,1                     1,0
 PARTICLE  DIAMETER,
                                                       10
Figure 3.  Probe losses due to  settling and diffusion for spherical parti-
          cles having a density of 2.5 gram/cc  under conditions of
          Laminar Flow.
                               306

-------
CO
o
            10'
            I013
         10

          N
Q

O


X
2
•o
             10
              I03
              10
              10
                                                              1—i  i    inn	r   i  i    1
                                           o  o
        ESP POWER OFF
                    o
                                           o   o
                                                               D0
         ESP POWER ON
                    D   a  a  a  a
                o
         o  °  a
   a   D            a
                                  o  IMPACTOR
                                  A  OPTICAL
                                  D  DIFFUSIONAL
                                                                                   A
                                                                                  O
                      '    I  I
                                        -LL
                 I.  I   I.
               0.001
           Figure 4.
0.005   0.01
    0.05   O.I
PARTICLE  DIAMETER,
                                                          0.5
I  I i  M I
  5     10
            dN/d  log D vs. D at inlet  and outlet of electrostatic precipitator installed
            on a  coal-fired power boiler.

-------
impactors starting at about 10 vim and going down to about 0.5 urn,  optical
particle counters in the .3 to 1 urn range and diffusional sizing on down
to  .01 urn.  The overlap and matchup of the size distributions is pretty
good in most cases being generally within about a factor of 2 in the concen-
tration at any particle size as determined by the three independent tech-
niques.
     Most of our use of the optical counters is to monitor concentration
fluctuations resulting from process changes and the like and not for primary
data.  However, we would like to get some idea of how well optical sizes
are correlating with sizes from impactors.  Figure 5 shows a correlation
based on sedimentation data with the optical particle counter obtained
in  the field.  We did sedimentation sizing using diffusion batteries
tilted on their side as dynamic settling chambers (or horizontal elutriators).
From this data we obtained estimates of the particles being counted in the
various optical size channels for which the particle counter was set.  It
appears that we obtained a fair correlation between sizes obtained with the
optical measurements and Stokes diameters.
     Figure 6 compares data obtained with a Thermo-Systerns System Model 3030
Electrical Aerosol Analyzer, an optical particle counter, and the diffusion
battery measurements at a common source.  We've got inlet and outlet data
on  the same plot here.  The inlet data are the solid symbols, the outlet data
the open symbols.  There is not a great deal of difference in the data at the
inlet and that at outlet, which says that the control device was a poor
collector for that size range of particles.  We couldn't really tell much
difference at all between the inlet and outlet on this particular control '
device.  The matchup between the diffusional measurements and the electrical
aerbsol measurements  was quite good on this particular  source.   It has
typically been about that good on most of the sources where we  have obtained
such common measurements.  But that is only a very limited number of places.
and we don't really have a large data base for comparison.  However, it does
look like diffusional arid electrical methods give comparable results.   Limited
experiements in our laboratory also indicated that we got comparable results,
and I think that Sinclair and Liu compared diffusion batteries  and  the Thermo-
Systems device  and got very excellent agreement during  some work at the
University of Minnesota.

                                   308

-------
J.U. U



5.0


e
n Diameter,
o
0
JJ
OJ
HI
•-4
4) n ^
w °-5


0.1
0








• 12 (
+ 100









.1


.-





Channel
Channe


















batt
1 bat


















.ery
-ter:






y


0








dai
f di





/
(


5








ta
ita




/







;









/

















*s

















/







1








/'
Jt
f







.0
























































5


















0








































, 	 1
| 	 »^


1 	 1 	 ,
I '
1
1
I
t







i
%
10.
         Indicated Optical Diameter (equivalent PSL dia.), microns
Figure 5.   Correlation  of optical and  sedimentation diameters.   Data
t-igure a.       .    using fly ash obtained from a coal-fired  boiler.
                              309

-------
        10'
        I06
      o
      I
      I-
      UJ
      a:
      2
      3
         105
        104
         10"
          0.01
                            INLET OUTLET
                             •    O - ELECTRICAL

                             4    0 - DIFFUSIONAL

                             A    A - OPTICAL
•


O
                                         j_
         0.1                 1.0

          PARTICLE  DIAMETER,'urn
                                                                 10
Figure  6.   Inlet and outlet-size distributions as  obtained with
            optical,  diffusional, and electrical  techniques.
                               310

-------
SEM:  I think that what they were doing was using the differential  mobility
analyzer, which uses a similar kind of principle and comparing that to
diffusion batteries.  In fact, they are doing more Work on it.
MCCAIN:  Figure 7 shows the fractional efficiency curve obtained for a
precipitator.  It shows a very characteristic phenomenon with ESP's dip in
collection efficiency in the vicinity of three tenths or a half micron.
     We've frequently encountered a problem that Gil mentioned—the apparent
formation of an aerosol in our sampling system as the result of what appears
to be a vapor phase component of the gas stream passing through the dew
point in the sample conditioner and forming a very small characteristic size
aerosol.  We attribute this to sulfuric acid condensation.  It hasn't been
demonstrated to really be that.  The fact that it appears to be something
that is formed in our sampling system is based on the observation
10 percent change in the dilution of the stack aerosol sample has frequently
lead to a concentration increase as detected by the instruments of a factor
of 1,000 to 10,000 rather than a 10 percent change in concentration.  Dif-
fusion battery measurements on such aerosols confirmed what Gil has shown
that it was quite small.  Diffusion batteries designed to eliminate particles
smaller than about .02 microns would remove essentially all of it.   Apparently,
something had formed in the sample conditioner not in the flue.  If it
existed in the flue then the small change in dilution would have resulted
in small change in indicated concentration by the instruments.  It is the
very sudden concentration increase with slight change in dilution that leads
us to believe that it is a dew point phenomenon (condensation).
	:  What did your temperature tell you about it?
MCCAIN:  We don't know what the temperatures are precisely.  We're taking  a
300-320° flue gas and suddenly mixing that with ambient air, and we don't
know what the detailed temperature structure is in that mixing process.  We
are certainly bringing the sample down very rapidly in temperature, and it's
a question of whether it mixes fast enough to keep the vapor from ever
going over the dew point concentration.  We just don't know what the history
is through the mixing zone.  It is certainly possible for the sample to go
                                   311

-------
CO


*»
e
s
Q.
S
u
o
u.
u.
UJ
z
g
i

8
^^



99.98
99.9
99.8
99.5
99
98

95

90




60

30
— — 1 I
_ A
— &
T" A
. — + 4-
	 A
" O
° c^ °
•^•I^H*
A MEASUREMENT METHOD:
— A CASCADE IMPACTORS
O OPTICAL PARTICLE COUNTERS
4- DIFFUSIONAL
PRECIPITATOR CHARACTERISTICS:
r TEMPERATURE - 335 °C
	 SCA - 85 M2/(M3/sec)
CURRENT DENSITY- 35 nA/CM2
Mi
1 1
0.05 O.I 0.5 1.0 5.0
	
MVW
-__
i ,
-_


	

	

—

*""*
___
—
~ ~
10.
                                                 PARTICLE DIAMETER, urn
            Figure  7.   Measured fractional efficiencies for the San Juan  hot  side electrostatic precipi-
                       tator with the operating parameters as indicated,  installed on a pulverized coal
                       boiler.

-------
through the dew point.  It does not always occur, we've had a lot more
trouble with it occurring on boilers that are burning a high sulfur  coal
than on boilers that are burning a low sulfur coal or sources that do
not have sulfur present, which  is why we  speculate that it may be
sulfuric acid.
                                    313

-------
                 PANEL DISCUSSION ON DIFFUSION  BATTERIES
       THE METEOROLOGY RESEARCH, INC.,  EXTRACTIVE SAMPLING  SYSTEM
                         FOR SUBMICRON  PARTICLES
             Dr. David S. Ensor, Meteorology Research,  Inc.*


SPARKS:  Thank you.  Dave, do you want to go next?

ENSOR:  The MRI extractive sampling system for control  device evaluation
is briefly described.  The diluter is a three-stage unit with adjustable
dilutions up to 2,000:1. The diffusion  battery is 5 units with particle size
cuts from 0.01 to 0.2 microns by diameter.  A modified  cascade impactor is
used as an in-stack precutter.
INTRODUCTION
     The objective of this presentation is to discuss the major points of
the MRI extractive sampling system.  The equipment has  been constructed over
the past year, tested in the laboratory, and used in a  major field test.
EXTRACTIVE SAMPLING EQUIPMENT
     In order to determine the penetration of submicron particles through a
control device, their effective mass is measured at both inlet and outlet
sample locations.  Mass measurements are made only after the sample is extracted
through a large particle knockout device and diluted for compatibility with
particle detection instrumentation.  Figure 1 is an illustration of the normal
sampling train.
     Usually, the sample is extracted at nonisokinetic flow rates due to sys-
tem design and instrumentation specifications.  This should not effect  sample
collection of the submicron particles of interest.
     Sample Train
     Precutter
     For source sampling, a precutter is used to prevent large particle con-
tamination of the fine particle sample  train.  For many purposes, a modified
  Coauthored by Richard G.  Hooper, Meteorology Research, Inc.

                                   314

-------
                      Heat  Traced
en
ILine Diffusional Single Enclosure
1 — _. Drver /
. — 1 	 "1
s. — 1 	 :._— —3 I 	
1
1
Cascade Dilutnon
Impactor • /-\P1 System ,
Precutter \r - '
r~"1 Fi ne Filter I
Coarse
Filter
^\ n
Pump \^S \ U
P2
1
_J


TP1 TP2 TP3
nJ___L--J__
TT 1
I
i
I

!_
i
i
i
1


!
1
1


I
1
R

1 !
Diffusion Batteries
Nos. 1, 2, 3
,Tp/i
i>ingle Enclosure
Diffusion Battery
TP5 No- 4
Single Enclosure
Diffusion Batter'
No. 5
	 TDfi
                                                                                      "Single Enclosure
                       Figure 1.   Flow diagram for submicron particle source testing.

-------
MRI Model 1502 Cascade Impactor can be used if assembled  to  mate with the
environmental  being sampled.   The advantages of using  the modified  impactor
by eliminating collection discs and filter media is the flexibility of deter-
mining the D<-0 separation point by manipulation of stage  jet diameters.  When
assembled properly, the impactor allows samples to be  col letted for approxi-
mately eight hours without plugging.  The inlet of the precutter-impactor
should face downstream of particle flow.
     Dilution System
     Sample dilution is a critical aspect to a good and proper sampling train
and requires experienced engineering design and flexible  experimental capa-
bilities.  Diluting the sample serves a dual purpose:
     •  Matching sample concentrations to particle detection
        capabilities, and
     •  Reducing the dew point of the sample.
     Dry dilution air is created by recirculating air through a bed of CaS(L
desiccant (dew point, -90°F)  and then filtering to prevent contamination.
Dilution is accomplished by a three-stage process of mixing the dry, particle
free dilution air with the sample.  Sample flow is measured by venturi-type
flow meters preceding each dilution stage, while dilution f!6ws are measured
across orifice-type meters.  Temperatures and pressures are also monitored
throughout the flow scheme.  Tubing diameters in the sample path are reasonably
large (0.375 inch diameter) to minimize particle loss due to diffusion and tubing
lengths are short to minimize sample residence time.  Flow control is
accomplished by manipulation of the dilution >.air control  valves.
     Dilution ratios of about 3:1 to 2,000:1 can be obtained by adjustment
of the control valves.
     A diffusional dryer is sometimes used to remove mositure from the
aerosol stream.  The diffusional dryer is a wire screen  tube containing  the
aerosol.  The annular area between the screen and an outer container  is
packed with desiccant.  Effort has been put forth to establish  the linearity
and efficiency of the diluter.  First, calibration curves have  been  established
for both the venturi- and orifice-type meters.  Using dilution  rates  deter-
mined from the flow meter calibrations and comparing to  particle measurements
at the inlet and outlet of the diluter gives an estimation  of  particle loss
and/or growth in the diluter.
                                   316

-------
     Diffusion Battery
     Diffusion is a process by which particles are removed from a gas;
this removal is characterized by the theoretical relationships between par-
ticle diffusivity (D  ) and parameters characteristic of the gas.   Einstein
introduced a theory which applies to particles with characteristic dimensions
the same as or greater than the mean free path of the gas molecules, and
deduced that particle diffusivity may be expressed as follows:

                      Dp  =  CKT/3 TT yd                           (1)
                                                     p
Where                 Dp  =  Particle diffusivity, cm /sec
                      C   =  Cunningham correction factor
                      K   =  Boltzman's constant, ergs/°K
                      T   =  Absblute temperature, °K
                      d   =  Particle diameter, cm
                                                         2
                      y   =  Fluid viscosity, dyne-sec/cm .

     The diffusion battery is a chamber with a large surface area where the
aerosol particles are pushed to the wall with Brownian Diffusion.  The
basic batter can be either a bundle of small diameter tubes or an array of
narrow rectangular ducts.
     In a plane-parallel channel  (with a cross-sectional length much greater
than its width), the  concentration ratio or battery penetration for both
steady state and unsteady state conditions is expressec by De Marcus1.
formula:
n/n  = 0.9149 exp-K885u  + 0.0592

       0.026exp
                    0

                                "151v
                                                    3
Where            n    =  outlet concentration,  No-/cm
                 n    =  inlet concentration,  No-/cnr
                 o
                 y    =  XD
                 D    =  particle diffusivity,  cm /sec
                 X    =  diffusion battery parameter, sec/cm .

                                   317

-------
     The battery parameter  is:

                     Y  _  L4hWN  _  4LWN
                     A  —  	n    ~
                            h*Q     hQ

Where                Q  =  the  volumetric  flow rate,  cm  /sec
                     L  =  length of battery channel, cm
                     W  =  width  of the  channel,  cm
                     N  =  number of channels
                     h  =  height of the channel,  cm.
     The battery parameter  X includes the  physical  dimensions  of the  battery
and the gas flow rate which are manipulated to achieve particle separation.
     An example of the parallel plate diffusion battery  is  illustrated  in
Figure 2.  The spacing and  number of channels are varied to create the
distinguishing parameter characteristic  of each battery. Table 1  lists
the representative parameters used  in a  recent study  conducted by MRI.

                                 Table 1

                 Battery No.                 Battery  Parameter
                    1                              284 sec/cm
                    2                             1065
                    3                             3975
                    4                            23272
                    5                            65274
To obtain fractional penetration  information, total particle counts are taken
before and after each battery with  a condensation nuclei counter (Gardner
Assn., Inc.).  Also, to add credibility to the data recovered by the CN
counter, filter samples are taken before and after each  battery; field
weighings of these filters  provide  an estimation  of particle density
through the battery system.  Figure 3 illustrates the flow scheme and  test
points for any given battery.
     Two instruments are used for particle detection in conjunction with
the diffusion system.

                                    318

-------
Filter
Holder
  Flow
  Meter
         O
        To Vacuum
        Pump
      Figure 2.  Single parallel plate diffusion battery.
                                                              Sample
                                                              In
                                                               To Next
                                                               Battery
                               319

-------
CO
N3
o
                   TP 1
From
Dilution
System
                                                    (Y^Vacuum Pump


                                                         CN  Counter
                                                                                             Flowmeter
                                                                                                    To
                                                                                                    Vacuum
                                                                                                    Pump
                                                                                             Diffusion
                                                                                             'Battery

                                                                                             Filter
                                          Figure  3.   Diffusion battery system.

-------
     1.  A condensation nuclei  (CN) type detector, Gardner Assn.,
         Inc.; and
     2.  A particle mobility analyzer, Model 3030 Thermo-Systems, Inc.
The CN type instrument gives total particle count by referencing the instru-
ment readout with a factory supplied calibration curve.  The mobility analyzer
gives a particle distribution,  and consequently total count, by referencing
a factor supplied calibration factor to the digital electrometer current
values displayed and continuously recorded.  Furthermore, the mobility
analyzer provides an opportunity to continuously observe the effects of
process variations at both inlet and exit sampling sites.  Finally, the
greatest usage of the mobility  analyzer is the quasi-continuous particle
distribution information  provided for the respective sample location.
     The mobility analyzer is portable enough for near-simultaneous measure-
ments at inlet and on sample locations.  Care is exercised in transportation
as well as sample preparation;  the sample is required to be dry and electrically
unbiased.  Sheath air for the instrument is pulled through a bed of CaS04
desiccant to reduce moisture levels in the instrument.  The output is recorded
on a strip chart and subsequently uses factory determined calibration con-
stants.
     The filter samples are used to collect aerosol for microscopic inspection,
chemical analysis, and to determine mass concentration.  Particulate mass
concentration can be used with  number concentration to estimate particle
density.
SUMMARY
     The MRI fine particle extractive sampling system has been briefly reviewed.
The laboratory and field  data obtained with the system were in good agreement
with cascade impactor tests and had good internal consistency between the
various particle sensing  methods.  It is expected that these data will be sub-
ject to future research papers.  Work is continuing in refinement of the equip-
ment and test techniques.
ACKNOWLEDGMENT
     The extractive sampling train and diffusion battery were developed with
Meteorology Research, Inc., internal research funds.
                                    321

-------
                 PANEL DISCUSSION ON DIFFUSION  BATTERIES

           Dr. Seymour Calvert,  Air Pollution Technology,  Inc.
SPARKS:  Seymour?

CALVERT:  I haven't any slides for this.   I  wasn't prepared to talk about
our diffusion batteries so I'll wave my hands around in the air a lot,
maybe draw something on the blackboard if necessary, but the essentials
are as Les Sparks alluded to.   About 2 to 2-1/2 years ago, we got into
the diffusion battery business and, upon hearing about the sad and
strenuous adventures of Joe McCain and other people at Southern Research
at carrying parallel plate batteries around, I decided that there ought
to be a better way and came upon the idea of using screens, a series of
screens, as the diffusional elements.
     Contrary to a popular notion, there is  a theoretical basis for screen
type batteries.  There are some uncertainties in the theory,but essentially
you build it up from basically a description (mathematical model) of dif-
fusional transfer to a single cyliner.  Depending on whose theory you use
there, there is some room for choice.  Then  the next point that comes up is
that the cylinders are not isolated, but they are close together in really
three dimensions by the time you put tqgether a battery, so there are inter-
action effects.  These have to be accounted  for.  We have put together a
theoretical model and have calibrated our screen type diffusion battery
using a parallel plate battery as the analytical instrument and have some
pretty consistent results.  We have used this in the field and have packaged
it in aluminum suitcases, and we use a two-stage series dilution so that we
can go up to a thousand-to-one without requiring very much dilution air.
The aerosols that we have used for calibration purposes are spray-dried
sodium chloride.  We have also used some small polystryene latex particles,
but these proved to be troublesome in experimental  use.
                                   322

-------
     We use condensation nuclei counters: a Gardner as the particle detector
and counter.  We also carry  a  Pollock  type condensation nuclei counter along
in the field and keep it in  the lab or the motel room and calibrate the
Gardner against it every night.  We've found that this is necessary.  We would
like a better condensation nuclei  counter or essentially a particle counter
of some sort.
     I think this just  about does  it.  We are presently initiating a more
comprehensive study  program  of the screen type diffusion battery, and we will
be looking more thoroughly into such questions as transfer to single cylinders,
interaction between cylinders, optimizing wire diameter mesh size,  you
name it.

DISCUSSION

	:  Characterizing  a particle  counter, what do you call a better
particle counter?
CALVERT:  We're using these manual condensation nuclei counters, and they
have some bias, and there are some  peculiarities in the way they count.  You're
able to set it at essentially different expansion ratios, which are supposed
to give you different size ratios  and  they don't really work out the way
that they are supposed  to.   It drifts  from day to day.  You have to keep
the optics clean, carry extra  bulbs,, extra batteries, extra everythings as
we've talked about, a little redundance doesn't hurt.  The Pollack is just
to big and heavy and beastly to carry  around.  So it would be nice to have
a good continuous condensation nuclei  counter, let's say about the size
of the Environment I counter that  SRI  would feel was a good reliable
instrument.  I gather so far that  you're not too happy with it.
MCCAIN:  Only in part.  It's  lightweight and when it works it does  all  right,
but I wouldn't use it as a primary instrument, the only instrument without
something to back it up.
                                   323

-------
CALVERT:  No.  We used to use the G.E. CNC connected with preconditioning
(chemical preconditioning) steps.  Back at Penn State we were using it for
gas analysis purposes and we were flying it , but we had to rent big air-
planes.
BOLL:  I seem to recall that at the Minnesota Symposium there were some good
words said about Professor Picard's continuous CN counter, and I wonder if
you or Joe McCain would like to comment on that.
CALVERT:  I've no experience with it, but I've read some papers about that
and then one by Sinclair using alcohol as the condensing medium, but I've
no personal experience with it.  Sounds like  a great idea;  it would be a
good approach if it worked.
MCCAIN:  Presuming it would work.  We built one several  years ago  with the
same sort of general idea using water vapor.   We didn't achieve enough super-
saturation in ours to reliably detect very small particles.   The calibration
would be the biggest problem to maintain a stable condition.   I don't think
that his is commercially available.
BOLL:  No, it is not commercially available as far as I know.
MCCAIN:  It would be a do-it-yourself thing.   It might be a little expensive
to build and get it to work.
HARRIS:  The main thing is trying to get enough supersaturatibn to get it
down to small things.
MCCAIN:  It takes several hundred percent to  do well on small sizes.  We got
a few percent with ours.
CALVERT:  The one thing that I might say.thinking about it.  The weight
advantage of the screen type battery, size and weight advantage, we can get
say 10 liters per minute sample capacity in something a foot long and about
1 inch diameter; 20 liters per minute in about an inch and a half diameter so we
can handle the same sample, volumetric flow rate as a big parallel plate
battery in a small screen setup.  The basic difference being we don't have to
have laminar flow through the casing.  We can have either laminar or  tur-
bulent or mixed boundary layers on the wires, which is another story.
Essentially, we can have pretty high superficial velocities  through a unit.
This proves to be a big advantage for field use.

                                  324

-------
	:  We found that the dilutions  have  to be extremely critical.  What
do you use to measure flow rates and what sort of dilution do you use,  how
do you dry your air?  That sort of thing, Joe, too?
CALVERT:  Yes, we went to the multistage  dilution system because we didn't
want to, among other things, we didn't want to have to meter a very small
sample flow rate from the stack.  You would like that flow rate to be as big
as possible.  If you want 10 cc  per minute or some huge rate like that or a
liter per minute, then if you want a thousand to one dilution on the other side,
you now need a thousand  liters per minute of filtered air, etc.  It achieves
two things: (i.) It keeps the meter  size  large (I don't Tike the~idea of
using a very small diameter orifice  or Venturi, a restriction-type
meter) and  (2.) It keeps  the amount of dilution air down.  So we use glass
Venturi meters for our sample measurement and for measurement of the sample
for aerosol streams.  For measurement of  dilution air we use rotameters,
the filtered dilution air.
	:  In your glass Venturi, the throat diameter is about what?
CALVERT:   Let's  see.   Something  on  the order of 1/16", 1/16" to 1/8", or so.
MCCAIN:  What we use  are calibrated orifices of various sizes for the sample
stream on  the order of twenty-thousandths of an inch and up, depending on
the flow rate we want.
	:  DO you have any trouble  with edge effects or anything?
MCCAIN:  There were  some  losses  behind  the orifice  that were calibrated out,
and there  is  some  difficulty with metering with single stage dilution.
	:  What  is  the  greatest you've  ever  had to dilute?
MCCAIN:  We've probably run on the order  of a couple thousand to one.  We've
gotten things that,  orifices and the like, that at  least calculate to provide
up to  4,000 to  1 if necessary.   More frequently,  it will range  from  10 to  1
to 500 to  1.
ENSOR:  You can  really affect your dilution by throttling the sample.
MCCAIN;  We throttle the  sample. We pull a sample  at a high flow rate through
a cyclone,  through a majority of the probe.and then split off a small flow
through the metering orifice to  the  diluter.

                                   325

-------
        Do you get dilution that high?
CAUVERT:  We generally, I  think we've run  in  the field  up to about  100 to
1, or there about.  Really, the dilution ratio required depends on  the  limits
of the condensation nuclei counter.   We were, as a matter of fact,  trying
to run the battery with the aerosol  as is  and do sometimes  keep it, so we
can run the aerosol through our diffusion  battery without dilution  and pull
samples out at various points after  so many sets of screens and dilute
those.  Dilute the sample  and then run that through the CNC.   So you can
do it either way and we've run it with filters in after so  many stages
Of screen and also had a bypass filter so  we'd get a check  on  the total
loading below the cut point of the impactor and precutter.
                                  326

-------
                 PANEL DISCUSSION ON DIFFUSION BATTERIES

                    Pedro Lilienfeld, GCA Corporation
SPARKS:  Thank you, Seymour.  Pedro?

LILIENFELD:  The  person who  actually did the work in this area is not here,
it is Doug Cooper,,and so  I'm not entirely up to all of the things that were
done in this area, but I  can relay a few of the conclusions and observations
that were made in the use of both condensation nuclei counters and diffusion
batteries in the  field.
     In one case, the problem of spurious generation of particles was also
observed, that which Gil  Sem reported this morning, measurements with the
mobility analyzer where a large number of small particles was observed pro-
bably without a generation system being in the source itself, but somewhere
between the source and the place of detection.  It appears that that problem
is very much associated with how the dilution is done and where it is done
and at what temperatures  it  is done.  So it all points to a condensation/
evaporation problem somewhere between where the sample is taken before it
reaches the place of measurement.  We seem to have resolved that to a certain
extent by building a diluter that fits right into the stack.  By minimizing
the length of line between the sample extraction point and where one dilutes
it, it seems that the problem of spurious generation is minimized or at least
reduced drastically.
     Doug [Cooper] also worked on the problem of modeling the errors
associated with the cut characteristics of diffusion batteries and comparing
them with impactors.  In  his work, he came to the conclusion that small
uncertainties in  the characteristics of a diffusion battery can lead to very
large errors in the measurement of size distributions using diffusion
batteries, mainly because of the very gentle slope cuts of the diffusion
batteries as compared to  an  impactor.  One has to watch out very much for
such errors in trying not to come,.to any sweeping wrong conclusions when
such measurements are performed.
                                  327

-------
     Some problems with the condensation nuclei  counter are  found  in  the
field.  I'm just going to read some that he mentioned.   The  Environment I
CNC which we have operates by measuring the light obscuration produced  by
the cloud,which results from an adiabatic expansion  of  the water-saturated
sample taken about once per second from a continuous sampling flow of 30 to
60 cubic centimeters per second.  This extinction is assumed to be linearly
related to particle number concentration and independent of  particle  size
for particles larger than 0.001 micrometers in diameter.  Recent work on
CMC's has somewhat tarnished their reputation.  Cabro et al.  have  showed
that the minimum detectable particle size is quite sensitive to relatively
small changes in the adiabatic expansion ratios in the device and  somewhat
sensitive to typical changes in ambient temperature.
     Walter Anjenica found that the supersaturations actually developed in
the CNC were not as high as expected from the adiabatic expansion  ratio,
thus raising the minimum detectable particle size to 0.01 micrometers,  some-
what dependent upon aerosol  material tested.
     Katz and Curshoned reported that even in the simple case of sodium
chloride particles, the minimum detectable particle size was three  times
higher than theoretically predicted.
     In general, it has been reported that nonwetting hydrophobic  particles
are not correctly counted.  As for linearity of response, Lu and Pui, 1974.
reported that the Environment I device, which is the one we were using  in
most of our tests, became nonlinear at concentrations above 50,000 per cubic
centimeter,which is an order of magnitude less than its nominal upper limit
of concentration.  Lu and Pui also found out that the Environment  I device
that they tested was low by a constant factor of 2.5 in comparison
with carefully determined concentrations in this linear arrangement.  The
General Electric model CNC was a factor of 1.4 too high.  However, in many
applications these errors are not significant as long as linearity is main-
tained.
     From all of this one can conclude that the best way to  use the CNC
measurements is  as ratios rather than absolute numbers.  This should  be
limited to the linear response range.  Furthermore, variations in  tempera-
ture should be avoided,and aerosols of different material should  not be
compared directly.  That is all that I have to contribute.
                                   328

-------
SPARKS:   Thank you.  Anybody else have any comments or experience with
diffusion batteries, condensation nuclei counters?  Or anything  else
they want to tdlk about?
     We'd like to thank all of you for coming to the seminar and I
don't know if we've solved any problems but we sure have discussed
them.
                                   329

-------
                      GENERAL DISCUSSION -  THURSDAY
SPARKS:  This afternoon's speakers are really zipping  right  along.   The
next thing that is supposed to be on the agenda is a general discussion  of
everything that has happened so far.   The floor is open  if anybody  has
anything they want to say, if they have found something  that might  solve
somebody's problems that they haven't mentioned or anything.

BOLL:  I'd just like to take about a minute to remind you, I'm sure that
you arl-e all acquainted with the B&W Impactor that was  designed and  the
operating procedure was published in December, 1971.  I'd like to remind you
that I think that has a few useable points that may still be valid.  One
of them is a cap over the inlet nozzle that would prevent anything  from
getting in through the nozzle while the impactor is warming  up.  Another
is a critical flow orifice that was built into the impactor  that allows  you
to start the flow rate up almost instantaneously and shut it down quickly
so that you can sample very little,in 10 seconds if you  like.  Another  is
a sheath that covers the impactor and provides a thermally conductive
path tending to minimize the possibility of thermal gradients from  one  to
the other.  They could get pretty horrible when you're sampling the high
temperature duct and you've got part of the impactor hanging out of the cold
end.  That leaves you with the problem of deciding what is the gas  tempera-
ture through the nozzles that are doing the impacting, and,  therefore,  what
is the cut point?  It does use a plain substrate which neither gains weight
nor loses weight.  Whether or not there is "bounce" is open  to some
debate, apparently.  We have seen "bounce," but we judge the quality of
the data by whether the cone of the impacted particles looks sound or
whether there are any particles out on the edges of the  cup.   So I just
brought up those points for your consideration when drawing  up manuals
and procedures.
                                   330

-------
HARRIS:  I think those are very valid points, especially the problem of
thermal gradients.  Another thing, thinking about using these devices
on pilot scale, we have to remember that these do have some finite volume
to them, and if you try to jam a 3-inch Andersen Impactor into an 8-inch
duct, you'll pretty well disturb the flow stream where you are trying to
get a sample.  The same can be said for most impactors.  Although we say
that you would like to sample in-situ, there are times when you're going
to be stuck with using a probe.  Have to make the best of the situation.
We have, by the .way, and I don't know the company right off-hand, we have
purchased blankets which just wrap around the Brink Impactor for heating
purposes.  Rather than having to wrap the whole thing in heating tape,
double-insulate and all that, you just buy these in any shape you want, and
they fit nice and snugly and tie on.  You can also get one to fit the
Andersen, or something else.  While I think about the heating problems,
if you are going to use some kind of heat, I would like to recommend
that you use your temperature-sensing element for controlling the
temperature as the gas stream flow exiting from the impactor immediately
after the impactor, rather than trying to use either the temperature of the
surface of the unit or try to sense somewhere further on down the stream.
It is amazing the amount of cooling that can happen in a gas flow stream in
5  inches.  We  first started off with just a little nipple, matter of  fact,
just 3  inches  behind  the Brink Impactors, and managed  to barbeque the
heating stage because the controller was always trying to drive the thing
to a hotter temperature.   It  got  up  to  500°-600° and  fried  it,  even  though
we were reading less than 300°F in the control circuits.  So when you get into
these regimes where you're having to heat, high heat, you have a lot of
chance of creating more problems.
SPARKS:  Everybody's happy, I guess.
SMITH:  We don't have a lot of experience with those cyclones collecting gram
quantities.  Do you know from the TRW folks where the  stuff ended up  in
the smaller cyclones?  Did it work its way into the cup?
                                    331

-------
HARRIS:  If I remember right,  we had  some evidence  that the vortex was
getting down into the collecting columns, because a  lot of the material
would be on the side of the collection cup.   But I don't remember any  large
deposition found on the walls  of the  units.   It seems  to get  down into  the
collection cup fairly readily.
SMITH:  That's pretty consistent with what we saw on the calibration
aerosols, but one flaw in  our  calibration procedures,  I guess,  is
that we really didn't load them up with a lot of material.  We  just measured
the collection efficiency  with th& ammonium  flourascein and the  Spectronic
88.
HARRIS:  Well, when TRW took them out in the field,  it was before we  had
them calibrated.  We took  them out, ran them, and tried to figure out  what
we had.  They tried a very crude type of calibration:  the type  where you
threw some dust in the air, sucked it through, then  see where it ended
up, then tried to do some  SEM measurement on it.  In one case,  in that case,
it is fairly heavily loaded and their results indicated that  the last  cyclone
was less than half a micron, and they  were getting down to the region  where
they were having some resolution problems.  When we came up with individual
particle calibrations, it  was 0.3 microns, so there seems to  be some  sort
of correlation involved there, but we haven't done  it  to the point where we
know it was fortuitous or  really true.
CALVERT:  This is a prosaic item but  an important one.  Can  we agree  on
what size the sampling ports ought to be?  The standard is  now 3 inches,
but we're pushing for 4 inches,and when I was just  thinking,  feeling,  that
we were getting pretty porty,  I started getting requests asking to put in
6 inch ports.  I've seen these grow from 2 inches up to 3, which seemed pretty
big at the time but now,with heating  blankets around an impactor and other
goodies, a 3 inch port isn't big enough and a 4 is  just barely big enough
and then if you have to turn the apparatus very much on an  elbow and so
forth, you'd like to have  more.  Yet, I think that it  is important to have
some kind of agreement on  a standard  approach so that when people are
putting in sampling ports, they'll put in big enough ones so we can all  be
happy nationwide.
                                   332

-------
HARRIS:  Arbitrarily, I say put in 6 inch.
CALVERT:  I suppose we ought to think big.
HARRIS:  It is kind of hard to ask them for a manhole, though.
CALVERT:  You wouldn't want them to.
HARRIS:  Then we would have ceiling problems.
CALVERT:  No, you would have technician loss problems.
HARRIS:  But one indication is the plant that we are going to next week is
a fairly new power plant,and they have 6-inch ports throughout their system.
They are also one of the few places that I know of where the sampling crew
that worked for them got involved in designing the sampling ports.  And
they are nice to work with.  If we have a choice, I'd like to see 6-inch
ports.  They are still manageable, but they give you a chance to try to
do some different things rather than try to sneak it in and hope you don't  scrape
the side clean with your nozzle going in.
CALVERT:  With flanges?
HARRIS:  I believe that these happen to be nipples, I think nipples are what
they need.
CALVERT:  You could put a flange on, ok, nipples rather than unions for
coupling thread.  You could put a flange on.
HARRIS:  Right, plus the fact that when you have the nipple there, that's
the inside diameter which is what you are going to be dealing with, so you
have a pretty good idea of what you need where, when its a coupling, and
you have to worry about how far down is your nipple going to be and bring
                                                                  V
it on down.
ENSOR:  With a cyclone sampling large amounts of flash, do you have problems
with the deposit migrating during your tests, grating against themselves
and creating new particles.
HARRIS:  We don't know.  I suspect that after next week, next couple of weeks,
we'll have a pretty good idea of what is going on.  The work that RTI did with
the 1.0 CFM train, they had pretty good samples.  Do you recall any problems
they mentioned in collection, Bill [Kuykendal]?
                                    333

-------
KUYKENDAL:  No, but I think that the collection was  assumed  to  be where  it
was supposed to be.
HARRIS:  I don't think that we ever saw any evidence that there was  a  patch
of material, then it was clear, then another patch of material.
MCCAIN:  Weren't some of those cyclones, on the abrasion question,  I think
that some of those were anodized aluminum,  weren't they?  Did you see  any
evidence of scouring through the anodized layer?
KUYKENDAL:  On the outside?
MCCAIN:  Inside opposite the inlet section, talking  about abrasion.
KUYKENDAL:  (To VanOsdell) You were there.
VANOSDELL:  I never took one apart.  I  wasn't involved while the cyclones
were being run.  I remember that a great deal  of the loading was on the  first
cyclone.
SMITH:  With our calibrations the curve goes up to 100 percent  and  stays as
opposed to the impactors, and you would hope that even for large quantities
they would stay up, but it is not inconceivable that things would  begin  to
be reentrained after you collect a lot  on the walls.
HARRIS:  Looks like we stuck our heads  in the sand with you.
OLIN:  I might mention that the ambient sampling people at EPA fooled around
with the Aerotech tube and ran it at 40 CFM, that was about 6 years ago, and
they thought that this was going to be  the way to sample.  That unit did
break up floculates and give you a higher respirable mass than you normally
should have.  I think the reason there  is it's a larger unit than  is really
pressed in terms of the sheer forces in there were pretty high.  They ran
at a cutoff of 3.5 microns and didn't have anything  above it, for
one thing, to take out real big particles.  That's one thing that should be
looked into.
HARRIS:  Yes, unfortunately, the type of development that we've been  involved
with is that somebody has a problem, so we get something that  works,  then
go back and see if it really does the job.
                                    334

-------
CALVERT:  On your draft guidelines, the second on data reduction, I  think
the cut diameter approach is dealt with unfairly by comparing it with the
deconvolution method on an aerosol with a very small standard deviation
and the cut diameter method becomes better as the standard deviations gets
bigger so if you're talking about a realistic geometric standard deviation
then the cut,diameter method is plenty good.
MCCAIN:  It's even mentioned in that when you run real data with the
deconvolution method it often tends to go to pieces.  It requires large
amounts of negative particles in some sizes on occasion to duplicate what
your masses are on the stages of the impactor. It just doesn't work  on
every case.
ENSOR:  We've had the same experience with the Picknett data reduction
techniques.  It will blow up for unknown reasons.
HARRIS:  Yes, I think that what we're seeing is that, except for the very
rare situation, we've gotten a perfectly adequate particle size description
using D5o's, rather than messing around with the Picknett.
CALVERT:  I think it goes all the way back. The first paper that I remember
on it is one by Mercer where he essentially validated that technique and
indicated that it does depend on the sigma g.
HARRIS:  Really, I think that it was introduced when we had those scalping
cyclones, because some of them had some pretty lousy distributions,  and some
of the original scalpers did tend to have those tipped where they were
extending over three size ranges, so that if you were assuming that it was
all D50, all of a sudden you didn't have enough material to work with down
here in the second stage because it was in the cyclone.  In that case, you
had to do some subtracting in order to get a realistic distribution.  I think
your point is well taken.  Those people that have had a chance to look at
that thing and have some comments, please tell me, I would  appreciate it.
Tomorrow we're going to sit down and incorporate what everybody  has said into
the next edition of this.  Then we'll be going out and will probably be
recommending to our people whondo tests for use that they  use this.  We'd like
to get feedback now.

                                    335

-------
SPARKS:  The one thing that all  of our contracts are working on is that they
will take blanks when they go out into the field.  They will make blank
samples, so that they will at least discover that there are problems with
their substrates.  I think they're all now, the technical  directive for
that was fairly recent, but I think they've all now had some experience
trying to live with taking blanks in the field.  I haven't heard any com-
plaints, so I guess it's working reasonably well.
ENSOR:  It's really no problem,  because you can do blanks  before the
traverses.   You're not worried about isokinetic sampling,  so they require very
little attention.
                                  336

-------
                             SEMINAR ATTENDEES
S. P. Schliesser
Monsanto Research Corp.
Research Triangle Park, N.C.

Dr. David S. Ensor
Meteorology Research, Inc.
464 N. Woodbury Rd.
Altadena, California  91001
(213) 791-1901

Joseph D. McCain
Southern Research Institute
2000 9th Avenue, S.
Birmingham, Alabama  35205

Grady B. Nichols
Southern Research Institute
2000 9th Avenue, S.
Birmingham, Alabama  35205

R. L. Byers
Exxon Research and Engineering
Box 101
Florham Park, N.O. 07932

Pedro Lilienfeld
GCA/Technology Division
Burlington Road
Bedford, Mass.  01730

Colin J. Williams
Cahn Instruments
16207 S. Carmenita Rd.
Cerritos, California  90701
(now with Perkin-Elmer Corp.)

Roger Shigehara
EPA, OAQPS, ESED, EMB
(MD-19)
Research Triangle Park, N.C. 27711
(919) 688-8146, ext. 276
Gilmore J. Sem
Thermo-Systems,  Inc.
2500 N. Cleveland Ave.
St. Paul, Minn.   55113

Wallace B. Smith
Southern Research Institute
2000 9th Avenue, S.
Birmingham, Alabama   35205
(205) 323-6592
Dale R. Blann
Acurex Corporation
485 Clyde Ave.
Mt. View, California
(415) 964-3200
94042
Dr. Mike J. Pilat
Dept. of Civil  Engineering
University of Washington
Seattle, Washington  98195
(206) 543-4789

Reed Cass
GCA/Technology Division
Burlington Road
Bedford, Mass.   01730
(617) 275-9000

R. M. Boll
Babcock & Wilcox
Research Center
P. 0. Box 835
Alliance, Ohio
(216) 821-9110

Roy  Neulicht
EPA, OAQPS, ESED, EMB
(MD-19)
Research Triangle Park, NC 27711
(919) 688-8146
                                    337

-------
Sam R. Turney
Monsanto Research Corp.
Research Triangle Park, N.  C.

Dr. C.Y. She
Physics Department
Colorado State University
Fort Collins, Colorado  80523

Michael C. Osborne
EPA, EMSL, Q.A.B.  (Md-77)
Research Triangle Park, N.C.  27711
(919) 549-8411, ext. 2580

Dale Harmon
IERL-RTP
Research Triangle Park, N.C.  27711

Jerry Cur
Sierra Instruments, Inc.
P. 0. Box 909
Carmel Valley, California  93924

Henry H. S. Yu
Monsanto Research Corporation
1515 Nicholas Rd.
Dayton, Ohio  45407
(513) 268-3411, ext. 259
Dr. J. A. Dorsey
IERL-RTP, PMB (MD-62)
Research Triangle Park, N.C.  27711
(919) 549-8411, ext. 2557

D. B. Harris
IERL-RTP, PMB (MD-62)
Research Triangle Park, N.C.  27711

Dr. Dennis Drehmel
IERL-RTP
Research Triangle Park, N,C.  27711

Ken Knapp
ESRL-RTP
Research Triangle Park, N.C.  27711

Jack Wagman
ESRL-RTP
Research Triangle Park, N.C.  27711
A. Kishan Rao
Midwest Research Institute
425 Vblker Blvd.
Kansas City, Mo.  64110

Geddes H. Ramsey
IERL-RTP (MD-61)
Research Triangle Park, NC

Ken Graves
Andersen 2000, Inc
2000 Sullivan Rd.
College Park, Ca.  30337
(404) 768-1300

John 01 in
Sierra Instruments, Inc.
P. 0. Box 909
Carmel Valley, California  93924
(408) 659-3177

Dr. Seymour Calvert
Air Pollution Technology, Inc.
4901 Morena Blvd.
Suite 402
San Diego, California  92117

Richard Hooper
Meteorology Research Inc.
464 W. Woodbury Rd.
Altadena, California   91001

W. B. Kuykendal
IERL-RTP, PMB  (MD-62)
Research Triangle Park,  N.C.  27711

Dr. L. E. Sparks
IERL-RTP
Research Triangle Park,  N.C.  27711

Jim Abbott
IERL-RTP
Research Triangle Park,  N.C.  27711

Gary  L.  Johnson
IERL-RTP
Research Triangle Park,  N.C.  27711
                                   338

-------
                                TECHNICAL REPORT DATA
                         (Please read Instructions on the reverse before completing)
 . REPORT NO.
 EPA-600/2-77-060
                                                      3. RECIPIENT'S ACCESSION-NO.
. TITLE AND SUBTITLE
Proceedings:  Seminar on In-stack Particle Sizing for
 Particulate Control Device Evaluation
                                                      5. REPORT DATE
                                                       February 1977
                                                      6. PERFORMING ORGANIZATION CODE
 . AUTHOR(S)

Douglas Van Osdell, Compiler
                                                      8. PERFORMING ORGANIZATION REPORT NO.
 . PERFORMING ORGANIZATION NAME AND ADDRESS
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, NC 27709
                                                      10. PROGRAM ELEMENT NO.
                                                      1AB012  ROAPAAS90
                                                      11. CONTRACT/GRANT NO.

                                                       68-02-1398, Task 32
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 COVERED
                                                       Proceedings; 12/75-9/76
                                                      14. SPONSORING AGENCY CODE
                                                        EPA/600/13
15. SUPPLEMENTARY NOTES IERL.RTP project officer for this document is D.B. Harris, Mail
Drop 62,  919/549-8411 Ext 2557.
16. ABSTRACT
          The proceedings document discussions during an EPA/IERL-RTP-sponsored
seminar on In-stack Particle Sizing for Particulate Control Device Evaluation.  The
seminar, organized by lERL-RTP's  Process Measurements Branch, was held at IERL-
RTP in North Carolina on December 3 and 4, 1975.  The first day's discussion was on
the use of cascade impactors, including calibration, field use experience, weighing
techniques, and elemental analysis of impactor samples.  Sizing techniques other than
with cascade impactors were discussed on the second day, with special emphasis on
diffusion batteries. The proceedings consist of edited versions of the seminar spea-
kers' transcripts.  Some were edited into technical paper format;  others remain in
conversational tone.  Visual aids presented by  the speakers are included.
17.
                             KEY WORDS AND DOCUMENT ANALYSIS
a.
                 DESCRIPTORS
                                          b.lDENTIFIERS/OPEN ENDED TERMS
                                                                   c. COSATI Field/Group
Air Pollution
Particles
Measurement
Impactors
Calibrating
Weight Measurement
Sampling
                                           Air Pollution Control
                                           Stationary Sources
                                           Particulate
                                           In-stack
                                           Cascade Impactors
                                           Diffusion Batteries
                                                                   13 B
                                                                   14B
18. DISTRIBUTION STATEMENT

 Unlimited
                                           19. SECURITY CLASS (This Report)'
                                           Unclassified
                                                                   21. NO. OF PAGES
                                                                        343
                                           20. SECURITY CLASS (This page)
                                           Unclassified
                                                                   22. PRICE
EPA Form 2220-1 (9-73)

-------