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)
------- |