EPA-R2-73-160
February 1973
Environmental Protection Technology
Develop an Operational System
for Evaluating and Testing Methods
and Instruments for Determining
the Effects of Fuels and Fuel Additives
on Automobile Emissions
Office of Research and Monitoring
U.S. Environmental Protection Agency
Washington, D.C. 20460
-------
EPA-R2-73-160
Develop an Operational System
for Evaluating and Testing Methods
and Instruments for Determining
the Effects of Fuels and Fuel Additives
on Automobile Emissions
by
H. R. Blosser
Battelle Memorial Institute
Columbus Laboratories
505 King Avenue
Columbus, Ohio 43201 :
Contract No. 68-02-0324
Program Element No. All002
Project Officer: John E. Sigsby
Chemistry and Physics Laboratory
National Environmental Researcl) Center
Research Triangle Park, North Carolina 27711
Prepared for
OFFICE OF RESEARCH AND MONITORING
ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D. C. 20460
February 1973
-------
This report" has been reviewed by the Environmental Protection Agency
and approved for publication. Approval does not signify that the
contents necessarily reflect the views and policies of the Agency,
nor does mention of trade names or commercial products constitute
endorsement or recommendation for use.
11
-------
MANAGEMENT SUMMARY
Contract No. 68-02-0324
"Develop an Operational System for Evaluating and Testing Methods
and Instruments for Determining the Effects of Fuels and Fuel Additives on
Automobile Exhaust Emissions".
Objective of Research
The original objective of this program, given in the title, was
changed to a study of the behavior of glass filters used to collect particu-
late matter from automobile exhaust.
Significance of Research to Date
The weights of particulates collected from auto exhaust emissions
showed unacceptable variations within lots, among filter types, and among
filters from different manufacturers.
Data presented in this report show that the weight variations are
real, and that they are related to automobile exhaust particulate collections
rather than to particulate collections in general. Possible physical and
chemical causes of these variations were studied, but no clear cause or causes
were identified positively. The evidence presented indicates that the causes
may be more subtle than originally suspected, and that their identification
will require extensive statistical analysis of experimental data.
Application of Research Results
Until the causative factors of weight variations are identified,
reproducible data are most likely to be obtained by tight controls over all
known variables, including the exclusive use of glass filters from a specific
manufacturing lot.
Goals For Next Period
No further work is scheduled under the present contract. Specific
suggestions for additional work are given in the "Future Work" section of this
report.
BATTELLE — COLUMBUS
-------
ABSTRACT
Causes of observed weight variations in collected particulates from
automobile exhaust were sought. Chemical and physical properties of unused
glass fiber filters were studied, and some chemical analyses of collected exhaust
particulates were performed. No clear indication of a single cause for the
variations could be discerned. An extensive statistical analysis of data
obtained from additional experiments is suggested as a means of pinpointing
the causes of the weight variations.
BATTELLE — COLUMBUS
-------
TABLE OF CONTENTS
INTRODUCTION 1
OVERALL PROGRESS 2
Literature Survey 2
Experimental Collections ...... 2
Filters Used 3
Automobile Exhaust Particulate Collection 4
Ambient Air Particulate Collection 7
Detailed Data Concerning The Automobile Exhaust and Ambient Air
Particulate Collection Data 20
Study of Filter Characteristics 20
Physical Properties 21
Weight Uniformity 21
Thickness and Density 21
Abrasion Resistance 21
Pressure Differential 26
Humidity Effects . 26
Conclusions Concerning Physical Properties ..... 36
Chemical Properties. ......... 36
Study of the Collected Samples 37
Elemental Analysis <, . . . 39
Hydrocarbon Analysis 39
Thermogravimetric Analysis 44
Conclusions Concerning Analysis of Collected Samples 45
Calibration of Exhaust Dilution Tunnel 46
CONCLUSIONS AND RECOMMENDATIONS . . . 49
FUTURE WORK ..... ..... 49
LIST OF FIGURES
Page
Figure 1. Sampling Port Locations, Looking Toward Automobile ......... 5
Figure 2. Example of Large and Rapid Weight Gain in «L0070 RH Atmosphere. ... 31
Figure 3. Example of Weight Loss in R£)% RH Atmosphere. . 32
Figure 4. Example of Weight Loss in »07o RH Atmosphere. 33
Figure 5. Example of Slow and Small Weight Gain in «LOO% RH Atmosphere .... 34
Figure 6. Weight Loss in ȣ)?<, RH Atmosphere 35
Figure 7. Particle Size Distribution of Aerosol Used to Calibrate Dilution
Tunnel 51
BATTELLE — COLUMBUS
-------
TABLE OF CONTENTS
(Continued)
LIST OF TABLES
1.
2.
3.
4.
5.
6.
7.
8.
9.
Table
Table
Table
Table
Table
Table
Table
Table
Table
Table 10.
Table 11.
Table 12.
Table 13.
Table 14.
Table 15.
Table 16.
Table 17.
Table 18.
Table 19.
Table 20.
Table 21.
Table 22.
Table 23.
Summary of Collections on Glass Filters 6
Auto Exhaust Particulate Collection on Glass Filters, Run 1 9
Auto Exhaust Particulate Collection of Glass Filters, Run 2 10
Auto Exhaust Particulate Collection on Glass Filters, Run 3 11
Auto Exhaust Particulate Collection on Glass Filters, Run 4 12
Auto Exhaust Particulate Collection on Glass Filters, Run 5 13
Ambient Air Collection on Glass Filters, Run 1 14
Ambient Air Collection on Glass Filters, Run 2 „ 15
Ambient Air Collection on Glass Filters, Run 3 .... „ 16
Ambient Air Collection on Glass Filters, Run 4 17
Ambient Air Collection on Glass Filters, Run 5 , 18
Ambient Air Collection on Glass Filters, Run 6 ..... 19
Weight Variation of Glass Filters 22
Thickness Measurements and Calculated Densities of Glass Filters . . 23
Abrasion Loss of Glass Filters ..... 25
Pressure Drop of 47-mm Glass Filters ..... 27
Weight Change of Glass Filter as a Function of Humidity 29
Chemical Analysis of Glass Filters 38
Carbon and Hydrogen Analysis of MSA Glass Filters 40
Hydrocarbon Content of Exhaust Collections on Glass Filters 42
Hydrocarbon; Content of Exhaust Collections on MSA Glass Filters. . . 43
Distribution of Dye Aerosol at Replicate Sampling Sites 48
Size Distribution of Dye Aerosol 50
ii
BATTELLE — COLUMBUS
-------
DEVELOP AN OPERATIONAL SYSTEM FOR EVALUATING AND TESTING METHODS
AND INSTRUMENTS FOR DETERMINING THE EFFECTS OF FUELS AND FUEL ADDITIVES
ON AUTOMOBILE EXHAUST EMISSIONS
by
E. R. Blosser and J. F. Foster
INTRODUCTION
The initial purpose of this research program was to "develop an
operational system for evaluating and collaboratively testing methods proven
applicable to the measurement of the influence of fuels and fuel additives on
vehicle emissions". As the project progressed, various difficulties emerged
that prevented the successful pursuit of the original goal. Therefore, a change
in objective was discussed and agreed upon with the project monitor. The revised
objective, briefly,stated, was to "develop an understanding of the critical
properties and characteristics of such filters which will contribute to the
precision and reproducibility of measurements of the particulate loading of
(2)
automobile emissions".
Four principal sections comprise this report: first, the work (primarily
a literature search) performed under the original scope of work; second, the experi-
mental work to establish the nature and magnitude of the problem; third, the experi-
mental work with different filters in an attempt to discover the reason(s) for varia-
tions; and fourth, recommendations for future work.
(1) Scope of work, negotiated Contract 68-02-0324.
(2) Introduction, Twelfth Monthly Report, Contract 68-02-0324 (May 8, 1972).
BATTELLE — COLUMBUS
-------
2
OVERALL PROGRESS
Literature Survey
One of the initial tasks of this program was to identify fuels and
fuel additives used in automobiles, and the means employed to measure their
effect upon particulate emissions in the exhaust. A literature search was
conducted, using a list of key words likely to identify articles relevant to
the subject. Over 1,000 references were located and partially indexed and
abstracted, and reprints or copies of the more pertinent ones were obtained.
In this mass of literature very little was found dealing directly with fuels,
fuel additives, and resulting automobile exhaust particulate emissions. Speci-
fically, particulate collection schemes were not discussed, nor were compositions
given for fuel additives. Further, it proved not practical for EPA to supply
information on additive compositions, and a few direct inquiries to manufacturers
produced no information. These formulations appear to be held securely as
proprietary information and efforts to analyze fuels for the additives' composi-
tions would have been beyond the scope of this program. For these reasons, with
the agreement of the project monitor, the program was redirected toward an
understanding of particulate filtration of particulate generated from nonleaded
fuel.
Experimental Collections
As redefined, the immediate goal of this program became a study of glass
fiber filters used to collect automobile exhaust particulate matter. Simply stated,
not all filters give the same net weight of exhaust particulate matter even though
the collections are made under identical conditions. The variations are not merely
BATTELLE — COLUMBUS
-------
3
random, nor can they be attributed to the several inherent errors of such collections.
The variations are real in the sense that they can be reproduced, as is shown in the
next sections.
Filters Used. For this study six different types of filters were
obtained. Five of these are glass fiber; one was made of quartz fibers. One glass
filter had an organic binder, not present in the other five. One filter was consider-
ably thicker than the others. All filters were either received as 47-mm-diameter
precut disks or were cut in our laboratory from sheet or roll stock,to 47-mm-diameter
disks.
The complete identification of the filters is as follows.
Whatman GF/A (No Lot No.) Box l^
Box 2
Whatman GF/B (No Lot No.) Box 1^
Box 2
Gelman Type A Batch 8170
Batch 8172
(a)
Gelman Type E Lot 8168 Box 1 '
Box 2
Mine Safety Appliance CT 75428, Lot J 7020
(47-mm-diameter disks cut from 8 x 10-inch sheets)
Pallflex Tissuequartz Style 2500 QAO, QC No. C-14581
(47-mm-diameter disks cut from 65-foot roll)
(a) Used in experiments where samples were taken from only one box. Box 1 and 2
are arbitrary numbers assigned to two boxes of the same lot number.
Whatman: Manufactured by W & R Balston, Ltd., England.
Gelman: Manufactured by Gelman Instrument Company, Ann Arbor, Michigan.
Mine Safety Appliance: Manufactured by Mine Safety Appliances Company, Pittsburgh, Pa«
Tissuequartz: Manufactured by Pallflex Products Corp., Putnam, Connecticut.
BATTELLE — COLUMBUS
-------
Automobile Exhaust Particulate
Collection
The key experiments of this program were the collections made from
diluted automobile exhaust and from ambient air. For these experiments the
automobile and associated equipment were those purchased and used on another
project and were used with the Sponsor's permission. The automobile is a 1970
Ford with a 351-cubic-inch engine and automatic transmission. The automobile
was run on a chassis dynamometer under tape-controlled Los Angeles cycles. Each
run consisted of 8 L.A.-4 cycles, begun after initial warm up; no forced engine
cooling was done between cycles. The fuel was a nonleaded research fuel,
RE 141-B, supplied by the Coordinating Research Council, the co-sponsor of the
project associated with the automobile. The engine exhaust was diluted approxi-
mately 20:1 with filtered ambient air forced into a 2-foot-diameter stainless
steel dilution tunnel maintained at an excess pressure above ambient of 1 inch
of water. Sampling was performed at a position about 30 feet beyond the dilution
point, using 13 collection points in a plane perpendicular to and within the
dilution tunnel. See Figure 1 for the arrangement of the sampling inlets in the
tunnel. The relative humidity (R.H.) at this point was approximately 50 percent
at 95 to 105 F. The flow through each filter was controlled by calibrated critical
flow orifices set for 0.50 cfm. A large mechanical vacuum backing pump maintained the
lines downstream of the orifices at about 22-inches Hg negative pressure (200 Torr).
Before beginning each run, the filters to be used were allowed to equili-
brate for at least 1 hour in the constant temperature and humidity weighing room
(74 F, 54 percent R.H.). After weighing the filters to ± 1 M>g, they were installed
in the holders and the run was begun. About 3 hours of operation, equivalent to about
61 miles, were required to complete the 8 L.A. cycles. Upon completion of each run, the
filters were removed from the holders and equilibrated in the constant temperature-humidity
BATTELLE — COLUMBUS
-------
5
room before reweighing. No evidence of weight drift, ± 1 |0,g, suggested that
equilibration in the weighing room was reasonably complete under these conditions.
The weights of the collected particulate matter from the first run are
given in Table 1. Duplicates of each of six filter types were used (triplicates
of the MSA filter), with agreement between replicates ranging from good to poor.
Among types, the agreement was not satisfactory; indeed, this lack of agreement
is the subject of this study. Two types were below the overall average weight
gain (Whatman GF/A and Tissuequartz); two were "average" (Gelman Type E and MSA);
and two were "high" (Whatman GF/B and Gelman Type A).
FIGURE 1. SAMPLING PORT LOCATIONS, LOOKING TOWARD AUTOMOBILE
(Numbers are the filter locations mentioned in the
text and tables.)
An insufficient number of each type of filter prevented reasonably
confident conclusions to be drawn from this single run. Therefore, the
experiment was repeated to improve the satistical reliability of the data.
For subsequent runs, the MSA filter was arbitrarily chosen as the reference
filter. Three runs were made using the same automobile and tunnel conditions
BATTELLE — COLUMBUS
-------
TABLE 1. SUMMARY OF COLLECTIONS ON GLASS FILTERS
Automobile Exhaust
Run No.
Filter Type (Table No.)
13
H
H
ffl
r
r
m
1
n
0
c
2
0
c
Ul
Whatman GF/A
Ditto
n
Whatman GF/B
Ditto
Gelman Type A
Ditto
Gelman Type E
Ditto
Tissuequartz
Ditto
MSA
Ditto
n
n
n
"
1
3
5
I
3
1
4
1
4
1
5
1
2
3
4
5
(2)
(4)
(6)
(2)
(4)
(2)
(5)
(2)
(5)
(2)
(6)
(2)
(3)
(4)
(5)
(6)
Number p Ratio ,
of Percent Type Run No.
Filters Avg. Dev. a) MSA (Table No.)
2
4
5
2
5
1
4
2
5
1
5
3
12
3
3
3
[ 6.2]
7.4
14.8
[36. ]
12.9
-
10.0
t 0.36]
6.2
-
2.4
[21.2]
12.9
7.4
10.0
3.9
Averages 8.8
0.69
0.56
0.48
1.66
0.90
1.74
0.74
1.06
0.91
0.45
Minus Value
-
-
-
-
-
2
3
4
5
6
1
2
3
4
5
6
(8)
(9)
(10)
(H)
(12)
(7)
(8)
(9)
(10)
(H)
(12)
Ambient Air
Number
of
Filters
10
7
10
10
10
9
3
3
3
3
3
Percent
Avg. Dev.(a)
14.4
2.1
7.1
3.6
12.1
10.0
11.7
3.3
10.7
10.0
1.5
7.9
Ratio (b),
Type
MSA
0.97
0.95
1.08
1.03
1.07
-
-
-
-
-
-
[ ] Excluded from computation of average (see page 20).
(a) Percent average deviation = £|wt - wtl
= x 100
wt
(b) R ti = average weight of particulate collected on given filter type
average weight of particulate collected on MSA filters during same run
-------
7
as described. In each run three MSA filters were randomly located in the
holders, and five each of two other filter types. Weighing was done in the
constant humidity-temperature room with at least 1 hour equilibration. The
detailed results of these runs are given in Tables 2 through 6, and are
summarized in Table 1. More will be said about these runs later in this report,
but first the ambient air collections will be described.
Ambient Air Particulate Collection
Ambient air collections of particulate matter were made using the
same types of filters as were used for the automobile exhaust collections. The
differences in conditions were:
(1) The automobile was not operating
(2) The blower supplying filtered ambient air to the dilution
tunnel was not operating
(3) The sampling was conducted for longer times
(4) The temperature and humidity in the tunnel were ambient.
Apart from these necessary differences, all other conditions were as nearly identical
as possible with the automobile exhaust collections. Thus, variables such as weigh-
ing errors, handling losses, humidity effects, etc. should be directly comparable in
the two different sets of collections.
The details of ambient air collection data are given in Tables 7 through
12, and are summarized, with the automobile exhaust data, in Table !„
One key point stands out in Table 1. This is the observed fact that
(a) different filter types give different collected weights of automobile exhaust
particulate matter, while (b) different filter types give the same weights of collected
ambient air particulate matter. The disparity between the performance of, for example,
Whatman GF/A filters and MSA filters when filtering automobile exhaust and
when filtering ambient air is clear; it makes no difference, relatively,
what reference base is chosen for data normalization. Therefore, it
BATTELLE — COLUMBUS
-------
8
seems evident that one or more of the following conditions exists for automobile
exhaust collections and does not exist, or exists to a remarkably lesser extent,
during ambient air collections.
(A) Assumption: The higher observed normalized automobile exhaust
collections are most nearly correct. Therefore, lower normalized results indicate
losses.
(1) Some exhaust particulates pass through some filters.
(2) Particulates are lost from some filters after collection.
(3) Exhaust gases weaken some filters and thus lead to mechanical
handling losses.
(4) Exhaust gases remove from some filters materials present in
these filters.
(5) Exhaust gases ablate bulk fibers from some filters.
(B) Assumption: The lower observed normalized automobile exhaust
collections are most nearly correct. Therefore, higher normalized values indicate
pickup.
(1) Some nonparticulate component(s) of exhaust are retained by some
filters.
(2) Some component(s) of air are retained by some filters to an enhanced
degree in the presence of exhaust.
BATTELLE — COLUMBUS
-------
TABLE 2. AUTO EXHAUST PARTICULATE COLLECTION ON
GLASS FILTERS, RUN 1
Filter
Filter Type Location
Whatman GF/A, Box 1
Whatman GF/B, Box 1
Gelman Type A,
Batch 8170
Gelman Type E, Box 1
MSA
Tissuequartz
Overall Average
10
6
7
8
4
3
2
5
1
13
12
9
11
Gain Per Average
Filter, VR ue
288. 272.
255.
421. 655.
890.
[1403.] 687.
687.
419. 418.
416.
369. 395.
520.
295.
[ 55.](a) 180.
180.
431.
Ratio (b>,
Type
MSA
0.69
1.66
1.74
1.06
_
0.45
[ ] Excluded from further computation.
(a) Filter torn upon removal from holder.
/u\ p (-• = average weight of particulate collected on given filter type
average weight of particulate collected on MSA filters during
same run
ATTEULE — C O U U IV! B U S
-------
10
TABLE 3. AUTO EXHAUST PARTICULATE COLLECTION ON
GLASS FILTERS, RUN 2
Filter Type
MSA
Ditto
M
it
n
n
ti
it
it
n
n
M
it
it
n
Filter
Location
10
6
7
8
4
3
2
5
1
13
12
9
11
Blank
Blank
Gain Per '"""'(M
Filter, jig Average Avg. Dev. ; Avg. Dev.
347
411
407
501 ;
451 ;
Lost |> 411 53. 12.9
335 |
430
345
544
300
438
422
-16
-23
(a) Average deviation = S | wt - wt
n
,. v _ , . _. average deviation , -._
(b) Percent average deviation = °-^ x 100
wt
BATTELLE — COLUMBUS
-------
11
TABLE 4. AUTO EXHAUST PARTICULATE COLLECTION ON
GLASS FILTERS, RUN 3
Filter Type
MSA
Ditto
"
in
Whatman GF/A
Ditto
ii
it
ti
it
Whatman GF/B
Ditto
"
ii
ii
Filter
Location
2
8
12
Blank
5
9
10
13
11
Blank
1
3
4
7
6
Blank
o . ., Percent
Gain Per , , ^v
Filter, |Jig Average Avg. Dev. Avg. Dev.
1
287
360 > 324 24. 7.4
324 '•
3 J
181 ~
[313] ,
173 > 182 8.5 4.7
175. j
199 '
-33
304 ]
263 i
345 :• 292 38. 12.9
224 \
325 j
-41
Type
MSA
-
0.56
0.90
[ ] Excluded from further calculation.
/ N A J • .-• £ | Wt - Wt |
(a) Average deviation = —' L
... _ . , . . . average deviation nn_
(b) Percent average deviation = °—2 x 100
wt
P • _ average weight of particulate collected on given filter type
(c; a 10 average weight of particulate collected on MSA filters during
same run
BATTELLE — COLUMBUS
-------
12
TABLE 5. AUTO EXHAUST PARTICULATE COLLECTION ON
GLASS FILTERS, RUN 4
Filter
Gain Per
Filter Type Location Filter, jig Average Avg. Dev.
(a)
Percent
Avg. Dev.
(b)
Ratio
Type
MSA
(c),
MSA
Ditto
it
it
Gelman Type A
Ditto
Ill
III
Gelman Type E
Ditto
8
2
12
Blank
5
1
13
7
6
Blank
4
3
11
9
10
Blank
363 1
272 ;;
311 j
-9
279 1
209
220 f
225 \
[752] J
-21
301 }
257
314
279
277 J
-9
315
32.
10.
233
23.
10.
0.74
286
18.
6.2
0.91
[ ] Excluded from further calculation.
Z I wt - wtj
(a) Average deviation =
n
„, _ ... average deviation ,nn
(b) Percent average deviation = a-^ x 100
wt
. .. „ . _
(c) a 10 -
average weight of particulate collected on given filter type
average weight of particulate collected on MSA filters during
same run
BATTELLE — COLUMBUS
-------
13
TABLE 6. AUTO EXHAUST PARTICULATE COLLECTION ON
GLASS FILTERS, RUN 5
Filter Type
MSA
Ditto
it
ii
Tissuequartz
Ditto
ii
it
ii
n
Whatman GF/A
Ditto
ii
ti
it
M
Filter
Location
10
4
13
Blank
9
5
7
6
8
Blank
11
2
3
12
1
Blank
(c\
Percent Ratio v ,
Gain Per ,, (b) Type
Filter, p.g Average Avg. Dev. ' Avg. Dev. MSA
342 1
370 > 349 14. 3.9
336 '
-33 J
-62 1
-58 !
-59 > -59 1.4 2.4 Minus
-60 ; Value
-57 )
-45
205
185
166 "> 167 25. 14.8 0.48
175
106 „.
-3&
£ wt - wt
(a) Average deviation = —'
,,. • . , . . . average deviation 1r._
(b) Percent average deviation = - Q— - - x 100
wt
. . _
(c) a o -
average weight of particulate collected on given filter type
average weight of particulate collected on MSA filters during
same run
AT T E U L E — COLUMBUS
-------
14
TABLE 7. AMBIENT AIR COLLECTION ON GLASS FILTERS, RUN 1
Filter Type
MSA
Ditto
11
it
ii
n
ii
ti
n
ii
it
ii
n
n
M
Filter Location
1
2
3
4
5
6
7
8
9
10
11
12
13
Blank
Blank
Gain Per Filter, (J.R
403
389
Lost
536
459
522
492
537
426
454
Lost
Lost
[766]
0
-10
Average - 468
Average deviation - 47
Percent average deviation - 10
•3
23-hour collection (19.5 m total per filter)
BATTELLE — COLUMBUS
-------
15
TABLE 8. AMBIENT AIR COLLECTION ON GLASS FILTERS, RUN 2
Filter Type Filter Location Gain Per Filter. |J.g
MSA 2 406
Ditto 8 561
" 12 465
" Blank -23
Average - 477
Average deviation - 56
Percent average deviation - 11.7
Whatman GF/A 1 418
Ditto 3 403
11 4 529
11 5 550
11 6 365
" 7 386
" 9 514
11 10 594
11 426
ii 13 449
11 Blank -32
Average - 463
Average deviation - 67
Percent average deviation - 14.4
_ _. Whatman GF/A ,. ....
Rati°> —Til °'97
23-hour collection (= 19.5 m total per filter)
BATTEL.LE — COLUMBUS
-------
16
TABLE 9. AMBIENT AIR COLLECTION ON GLASS FILTERS, RUN 3
Filter Type Filter Location Gain Per Filter. p.g
MSA 1 727
Ditto 6 665
11 7 705
" Blank 5
Average - 699
Average deviation - 23
Percent average deviation - 3.3
Whatman GF/B . 2 654
Ditto 3 636
4 666
" 5 658
" 8 702
" 9 664
11 10 651
" 11 [255]
11 12 [215]
11 13 [227]
11 Blank -38
Average - 662
Average deviation - 14
Percent average deviation - 2.1
Whatman GF/B _
Ratl°> MSA - 0-*>
[ ] Filter removed after 8.5 hours; value excluded from
further calculation.
3
23.5-hour collection (= 20.0 m total per filter)
BATTELLE — COLUMBUS
-------
17
TABLE 10. AMBIENT AIR COLLECTION ON GLASS FILTERS, RUN 4
Filter Type _ Filter Location Gain Per Filter,
MSA
Ditto
ii
it
1
5
9
Blank
565
590
448
-30
Average - 534
Average deviation - 57
Percent average deviation - 10.7
Gelman Type A 2 483
Ditto 3 522
4 593
" 6 568
11 7 621
" 8 640
" 10 597
11 11 637
11 12 537
ii 13 593
" Blank -20
Average - 579
Average deviation - 41
Percent average deviation - 7.1
Gelman Type A , ...
Ratl°» - MSA - l'°*
19-hour collection (= 16.2 m total per filter)
BATTEULE — COLUMBUS
-------
18
TABLE 11. AMBIENT AIR COLLECTION ON GLASS FILTERS, RUN 5
Filter Type Filter Location Gain Per Filter,
MSA 2 806
Ditto 5 830
" 9 810
" Blank -3
Average - 815
Average deviation - 10
Percent average deviation - 1.2
Gelman Type E 1 843
Ditto 3 795
" 4 798
6 815
11 7 830
11 8 856
11 10 873
" 11 872
11 12 799
11 13 891
" Blank 4
Average - 837
Average deviation - 30
Percent average deviation - 3.6
Gelman Type E . ..
MSA 1>03
3
18.25-hour collection (=15.5 m total per filter)
BATTEULE — COLUMBUS
-------
19
TABLE 12. AMBIENT AIR COLLECTION ON GLASS FILTERS, RUN 6
Filter Type Filter Location Gain Per Filter. \JR
MSA 4 1371
Ditto 8 1428
11 11 1392
" Blank 1
Average - 1397
Average deviation - 21
Percent average deviation - 1.5
Tissuequartz 1 1349
Ditto ' 2 1257
" 3 1399
11 5 1671
11 6 1818
" 7 1351
11 9 1886
10 1448
11 12 1300
" 13 1438
" Blank -44
Average - 1492
Average deviation - 180
Percent average deviation - 12.1
_ . . Tissuequartz n _-,
Ratio, rr^* 1.07
3
23.5-hour collection (= 20 m total per filter)
BATTELLE — COLUMBUS
-------
20
Detailed Data Concerning the Automobile Exhaust and Ambient Air
Particulate Collection Data. The detailed data that were summarized in Table 1
are presented in Tables 2 through 6 (automobile exhaust) and Tables 7 through 12
(ambient air exhaust). In each table the filter type, its location within the
dilution tunnel (Figure 1), the measured weight gain or loss after collection,
the average weight change for each type, and the precision statistics are given
where applicable.
All averages are simple arithmetical averages, and all precision
statistics ignore difference signs. Average deviations rather than standard
deviations were used because the number of observations varies from run to run.
The precision data using standard deviations are prejudiced against sets having
fewer observations because the divisor is n-1, rather than n as is used here.
Filter location information is included to demonstrate the interchang-
ability (equivalence) of the 13 locations. Where applicable, the data in Tables 1
and 2 have been corrected for an improper critical flow orifice. "Lost" data refer
either to individual filters that were mechanically ruined upon removal from their
holders, or (in Table 7) to a malfunction in part of the vacuum manifold.
Occasional data are excluded from computations. The cause of these
abnormal values is not known. All precision data from Table 2 are neglected in
further calculations; since there are only one or two observations for all but the
MSA filters, it seemed best to discard all the precision data from that run.
Study of Filter Characteristics
Work reported in the preceding section confirms the observation by
persons at EPA and elsewhere that not all glass fiber filters collect, uniformily
and consistently, the particulates in automobile exhaust. In attempts to discover
the causes of these observed variations the different filter types used in this
study were subjected to various examinations, both physical and chemical.
BATTELLE — COLUMBUS
-------
21
Physical Properties
Physical properties can play an important role in the collection and
retention of particles and in the handling characteristics of the filters. The
properties studied and the results are described in the following sections„
Weight Uniformity. Variations of weight within a given type or lot of
a manufacturer are one indication of uniformity or quality control. From each of
the package units of the filters obtained for this study, random individual filters
were weighed to the nearest 0.1 mg on a Cahn digital balance Model DTL. From 10 to
40 such weighings from each type or batch gave the data shown in Table 13. It is
seen that the standard deviation of the weights of individual filters from a single
package (lot, type, or batch) ranges from 1.13 mg to 3.86 mg per 47-mm-diameter
filter. Repetitive weighings of the same filter, weighed to the nearest 0.01 milli-
gram, gave a percent coefficient variation of 0.01. An interesting point is the
difference in the standard deviations of 2 batches of the Gelman Type A filter.
No other (known) different batches of a given type from a given manufacturer
were used in the program so an "average" batch-to-batch variation cannot be
established.
Thickness and Density. The thicknesses of ten samples of each filter
type were measured using an Ames gage with a 1.25-inch-diameter foot and a pressure
2 2
of 25 g (= 3.16 g/cm = 0.718 oz/in )„ The results are shown in Table 14.
Abrasion Resistance. Any weight loss inadvertently caused by handling
between the tare and the gross weighings will of course give a fictitiously low
value for the collected particulate matter. Two methods were employed to assess
possible losses by handling. The first method, insertion in and removal from the
filter holder (Gelman Model 1235), with the 1-7/8-inch-OD rubber 0-ring removed, was
not a valid means for measuring differences. In this test the seal was formed
between the metal screen and the mating wall, which is normally the internal
support for the 0-ring. The mating ring is free to turn on the filter during assembly.
BATTELLE — COLUMBUS
-------
TABLE 13. WEIGHT VARIATION OF GLASS FILTERS
Gelman
Whatman Type A
Tissue-
MSA quartz
GF/A GF/B Batch Batch Type E From 4 From
0
H
H
in
r
r
m
1
n
0
r
C
2
C
Box 1 Box 2 Box 1 Box 2 8170 8172 Box 1
Weight, milligrams, 94.44 96.54 267.49 267.58 151.43 135.90 138.36
average
Range
Low 92.3 93.9 263.8 262.6 148.3 128.4 134.7
High 96.8 98.6 271.7 271.7 155.5 141.4 141.3
; h
a(a) 1.38 1.39 2.20 2.50 2.35 3.86 1.81
Average a 1.38 2.35 1.88
fM
vw 1.45 0.88 1.55 2.84 1.36
;^-i / ~* \ L.
<*\ rr \/S< W - W>
Box 2 Sheets 65' Roll
137.67 107.48 116.75
134.1 105.3 114.4
14]. 2 108.2 119.3
1.95 1.13 1.61
1.05 1.38
Nl
isi
(b) v = - x 100.
w
-------
TABLE 14. THICKNESS MEASUREMENTS AND CALCULATED DENSITIES OF GLASS FILTERS.
Weight
m
H
H
m
p
p
m
1
n
0
r
C
2
D>
C
(0
Whatman GF/A
WhatmanGF/B
Gelman Type A
Gelman Type E
MSA
Tissuequartz
per 47-mm
diam. filter,
mg
94
267
151
138
107
r\
mg/cm
5.4
15.4
8.7
7.9
6.2
6.7
v(a)
1.45
0.88
1.55
1.36
1.05
1.38
rnilW
15.0
37.8
27.6
31.0
17.0
23.3
Thickness
ram
0.38Q
0.96l
0.702
0.78?
0.432
0.592
v(a)
3.2
1.7
2.5
2.3
3.1
1.8
Density
g/cnv3
0.142
0.16Q
0.124
°'10l
0.143
o.n4
NJ
LO
__...
(a) v = -7- x 100, where a =~\ ^\~-^
(b) Mil = 0.001 inch.
-------
24
With care, no loss (to the nearest 0.1 mg) was observed in most cases for any
filter; sporadically, gross amounts (milligrams) of filter were torn from a
filter, but this event was usually preceded by allowing the mating ring to
rotate during the tightening step. It was concluded that, although this test
is not highly reliable, with reasonable care all but the Tissuequartz lost
either nothing (<0.1 mg), or major amounts could be readily detected visually
and thus were cause for rejecting that specific collection. The Tissuequartz
did seem more prone to tearing even with careful handling.
The second evaluation for handling properties measured abrasion
resistance. Samples of each filter were subjected to a "sieving" motion
for 10-, 20-, and 40-minute periods, using new filters for each period. The
results of this evaluation are presented in Table 15. The MSA and Tissuequartz
filters lost amounts of material proportional to the length of agitation, but
the Whatman and Gelman filter losses did not increase proportionally with
increased time.
(R
The agitation was provided by a Rotap machine used to sieve samples
of ore. The filters were placed, inlet side down, on a 1/8-inch opening sieve.
The motion is horizontal; one side of the sieve moves back and forth in a
straight line while the other side moves back and forth in a large arc perpendi-
cular to the other horizontal motion. Simultaneously the assembly is tapped
from above once per cycle; the frequency is about 120 cycles per minute. These
motions probably ablate the filters far more severely than would any but very
careless handling.
From the abrasion studies it was concluded that (a) normal handling
with reasonable care should result in less than a 0.05-mg loss for any filter
except Whatman GF/B; that type gave a 1-minute average loss of 0.1 mg; (b) losses
1ATTELLE — COLUMBUS
-------
TABLE 15. ABRASION LOSS OF GLASS FILTERS
(Rotap Agitation)
Gelman
Whatman Type A
0
5
H
m
r
p
m
I
0
r
C
2
0
C
(A
GF/A . .
Box 1 Box 2
Run 1 (10 min) 0.5 0.3
Weight loss, mg
Run 2 (20 min) 0.4 0.1
Weight loss, mg
Run 3 (40 min) 0.5 0.0
Weight loss, mg
Average loss /min, mg 0.027 0.012
"-— •**> '"
Average loss/min, mg 0.020
: GF/B Batch Batch Type E
Box 1 Box 2 8170 8172 Box 1 Box 2 MSA
1. 2. 0.1 0.4 0.0 0.0 0.5
1. 2. 0.1 0.3 0.0 0.0 0.7
2. 2. 0.1 0.7 0.1 0.2 1.6
0.070 0.120 0.006 0.024 0.0008 0.002 0.042
V ^ S ^ ^ :'
0.100 0.015 0.001 0.042
Tissue-
Quartz
0.5
0.8
1.9
0.044
0.044
U1
-------
26
from insertion in and removal from the filter holder are either <001 mg or major,
and visual inspection should be made to guard against the major losses. In any
case, the data shown in Tables 7 through 12 suggest that abrasion losses, if they
occur, at least are reproducible within a filter type to about ± 50 jig. That
figure is the average total error of the ambient air experiments (8 percent of
600 p.g), and includes possible errors from other causes in addition to ablative
losses.
Pressure Differential. The pressure drop across the 47-mm filters was
measured at a flow rate of 1.0 cfm, controlled by a calibrated critical flow
orifice. The pressure differential was measured at a tee between the exit side
of the filter and the limiting orifice using an open-end water manometer. The
results, given in Table 16, show that all but the Whatman GF/B have pressure
drops quite close to 18 inches of water; the GF/B is a much heavier and thicker
filter than the others.
Humidity Effects. A significant contribution to the wide spread
among observed weights of collected automobile exhaust particulate matter could
be made by differences in adsorbed water on the filter. If different types of
filters, or even different filters from the same lot, adsorbed varying amounts
of water from the exhaust stream, these differences would be reflected in the
measured weight changes attributed to particulate matter.
To measure the moisture effect two large desiccators were set up;
one with Drierite® in the bottom part, the other with water. Two or more
weighed samples of each filter type were placed on wire screens just above the
Drierite or water and allowed to equilibrate at room temperature (72-74 F). After
a time period ranging from overnight to one week, each filter was removed from
the desiccator and placed on the recording balance pan. Within the enclosed
BATTELLE — COLUMBUS
-------
27
TABLE 16. PRESSURE DROP OF 47-timTa^ GLASS FILTERS
[Inches of water differential/ ' at
1.0 cfm(c)]
4,(« ,(•) v«>
Whatman GF/A, Box 1
Whatman GF/A, Box 2
Whatman GF/B, Box 1
Whatman GF/B, Box 2
Gelman Type A, Batch 8170
Gelraan Type A, Batch 8172
Gelman Type E, Box 1
Gelman Type E, Box 2
MSA
Tissuequartz
16.4
17.6
43.8
44.6
21.7
20.6
18.5
18.5
18.2
15.9
0.55
0.47
1.0
1.1
1.9
0.68
0.28
0.27
0.32
0.49
3.3
2.7
2.3
2.5
8.8
3.3
1.5
1.4
1.7
3.1
(a) Effective diameter = 35 mm = 1-3/8 inch = 9.62 cm2 = 1.49 in.2,
(b) 1 inch of water differential = 1.87 torr (mm Hg).
(c) 1 cfm = 0.0283 ra3/min.
(d) Average of AP determinations on 6 individual filters.
9
, s / £(AP::- AP)
(e) a =
(f) v = ;£ x 100.
AP
BATTELLE — COLUMBUS
-------
28
volume of the balance pan was a small dish containing either Drierite or water.
(Samples from the desiccator containing Drierite were weighed in the balance
with water, and vice versa.) The transfer from the desiccator to the balance
was accomplished as rapidly as possible (2 to 3 seconds), and the readings were
begun immediately, since the balance requires only a few seconds to stabilize.
Either 50 or 200 mg was tared from the balance so that the weighings could be
made on the 100-mg full-scale range with a sensitivity of 0.1 mg. A few
weighings were made with almost the entire filter weight tared out so that the
balance could be set on the 10-mg full-scale range and read to 0.01 mg. The
recorder was also set to higher sensitivity with the zero point depressed.
Thus, it was possible to record very small weight changes, as was done for the
chart in Figure 5. Since it is difficult to anticipate the weight of the filter
closely enough to make all the necessary adjustments before removing the filter
from the desiccator, quite a few experiments were aborted because an incorrect
estimate was made, and the readings,were off-scale.
The data from these experiments are shown in Table 17. The "AW"
columns are the weight change of the filter while on the balance, and are the
gain or loss upon a change in the relative humidity. The "Time to Constancy"
columns are read from the chart recording and are the approximate time at
which the rate of change becomes nearly zero. This estimate is somewhat
subjective, but should be consistent from filter to filter. The "Aw from
50 percent RH" column is the weight difference of the filter as weighed before
placing it in the desiccator, and at the constancy point defined above.
BATTELLE — COLUMBUS
-------
TABLE 17. WEIGHT CHANGE OF GLASS FILTER AS A FUNCTION OF HUMIDITY
(a)
ID
J>
H
-1
m
r
r
m
n
0
r
C
2
o
c
in
Filter Sample
Whatman GF/A (Box 1)
(Box 2)
Average
Whatman GF/B (Box 1)
(Box 2)
Average
Gelman Type A, Batch 8170
Average
Gelman Type A, Batch 8172
Average
Gelman Type E (Box 1)
Average
(Box 2)
.Average
MSA
Average
Tissuequartz
Average
Total AW, mg
0.1
0.5
0.1
0.1
0.2
0.1
0.3
0.0
0.1
0.1
0.1
0.1
0.2
0.3
0.2
0.4
0.3
0.3
0.2
0.0
0.1
0.0
0.5
0.3
0.3
2.2
3.6
4.1
5.3
3.8
Dry to Wet
Time to
Constancy, min
[70]
2
~2
^2
/w2
[15]
~2
--
rt
•^2
•^*2
'"•'2
~2
^2
~2
f\
f\
r\
f~^2.
--
^2.
..
-2
~2
'-•'2
52
46
60
...... 26,
46
AW From
sa50% RH, mg
-0.1
0.0
0.0
-0.2
-0.1
-1.0
-0.8
+0.1
-0.6
-0.6
-0.1
-0.1
-0.1
-0.2
-0.1
-0.3
-0.1
-0.2
0.0
-0.1
-0.1
0.0
-0.3
-0.3
-0.2
-1.5
-1.6
-0.6
-0.4
-1.0
Total AW, mg
10.5
13.6
9.5
7.7
10.3
4.0
21.1
(13)
4.6
4.2
4.4
23.8
23.8
23.8
12.8
10.3
11.5
3.0
1.1
2.0
39.5
29.6
22.2
51.1
30.6
5.4
4.5
4.9
Wet to Dry
Time to
Constancy, min
4
3
6
3
4
6
8
7
2
2
2
10
9
9
6
7
6
1
1
1
13
11
8
23
14
6
8
7
AW From
sa50% RH, mg
-0.8
0.0
+0.2
+0.1
-0.1
+ 1.8
-0.3
+0.7
+0.2
+0.4
+0.3
-4.8
-0.2
-2.5
-0.1
-0.1
-0.1
+0.2
. +0.3.
+0.2
-0.1
+0.1
-0.5
-0.2
-0.2
-0.5
-0.7
-0.6
(a) See text for explanation of terms.
[ ] Not included in average.
-------
30
Two facts are clear from the data. First, the loss of weight
in going from the humid desiccator to the dry balance is far more notice-
able than the reverse (dry to wet). Typical recorder plots of these
changes are shown in Figure 2 through 6. Second, true equilibrium is not
established within the balance enclosure. If true equilibrium were
established, the gain in weight going from dry to wet should equal the
loss in weight going from wet to dry. Only the Tissuequartz filter
approximates this situation. It appears that most filters, whether from
the dry or the wet atmospheres, rather quickly (<30 minutes) arrive at
a weight close (± 0.1 to 0.2 mg) to that at ambient (50 percent) relative
humidity. The additional weight gained in the wet atmosphere apparently
is acquired over a much longer time period than the hour or less that the
filter was on the balance. No measurement of nor control over the actual
humidity in the balance was possible, and it is inferred that the R.H. in
the balance is close to ambient (« 50 percent) when the weighing commences,
because the cover was opened to insert the sample.
In retrospect, the experiment would be more meaningful if weight
changes to ambient R.H. had been sought. The presence of Drierite or water
under the balance cover did not establish equilibrium at high or low R.H.
during the weighing period, but only introduced another (unknown) variable.
In spite of that uncertainty, the differences among the filters should be
valid qualitatively.
Results in Table 17 and Figures 2 and 3 show that the Tissuequartz
equilibrates quite rapidly, either up or down. This can be interpreted to
mean that the material is sensitive to humidity changes, and that Tissue-
quartz filters must be weighed under well controlled humidity conditions.
Other differences shown in Table 17 are the marked variation between the
BATTELLE — COLUMBUS
-------
31
«-rtrN1eD IN U. S. A.
IB
1
-l_.
ri
— U-
."Li
q-:
m
•-T-
4_
:ll'
wLi
^
ii-P-r'
A|:;f r
tobi::
I
;"[
Li
•a
7t:
f-.pL
I
:S
£-Mftt*
liVl
o
h-tt
3
SI
':] _i:
Lj—iT
:-t»t;
fi
S
Him
roi,
o
:::
g
•T££
FIGURE 2. EXAMPLE OF LARGE AND RAPID WEIGHT GAIN
IN «1007o RH ATMOSPHERE
Tissuequartz, 1 week in «07o RH (Drierite),
then put in recording blance having a pan
of water. This figure illustrates the
large and rapid weight gain of a filter
exposed to high humidity.
BATTELLE — COLUMBUS
-------
32
FIGURE 3. EXAMPLE OF WEIGHT LOSS IN «0% RH ATMOSPHERE
Tissuequartz, 1 week in xs 100% RH (over H20),
then put in recording balance having a pan of
Drierite. This figure illustrates a small
and fairly rapid weight loss to equilibrium.
BATTELLE — COLUMBUS
-------
33
•-RINltD Iti O. «.*.
FIGURE 4. EXAMPLE OF WEIGHT LOSS IN ^0% RH ATMOSPHERE
Whatman GF/B, 1 week in «1001 RH (over H20),
then put in recording balance having a pan of
Drierite. This figure illustrates a large
weight loss over a moderate time period.
BATTELLE — COLUMBUS
-------
34
t-RIWIcTI III U. 8.
v-YI i-lQ r-00171 ^HILAuEUPhl*. PA
"
FIGURE 5. EXAMPLE OF SLOW AND SMALL WEIGHT GAIN IN ^100% RH ATMOSPHERE
Whatman GF/A filter, 15 hours in ^0% RH (Drierite), then put
in recording balance having a pan of water. This figure
illustrates the (relatively) small and slow but continuing
weight gain of a dry filter exposed to high humidity.
BATTELLE — COLUMBUS
-------
35
FIGURE 6. WEIGHT LOSS IN «07. RH ATMOSPHERE
MSA, 1 week in «100% RH (over H20),
then put in recording balance having
a pan of Drierite. This figure
illustrates a large and relatively
long weight loss to equilibrium.
BATTELLE — COLUMBUS
-------
36
two Gelman Type A batches where Batch 8170 lost 4.4 mg and Batch 8172 lost 23.8 mg
in the humid to dry changes; the relatively low moisture loss (4.9 mg) of the
Tissuequartz; and the high loss (30.6 mg) of the MSA filter. In all cases, the
excess weight gained at high humidity was lost in less than 15 minutes.
Conclusions Concerning Physical Properties. Within a given type, lot,
or box of filters, the variations of physical properties of individual filter
disks are small, usually from 1 to 3 percent of the property. Random errors
in measurement probably account for some but not all of the variations. In those
instances where two boxes or lots of a specific type were checked, the variations
between group averages for certain properties appeared larger than the standard
deviation for individual measurements (see Table 13, Gelman Type A; Table 16,
Whatman Type GF/A; Table 17, several pairs).
Many large differences among the various filter types are apparent from
the data in Tables 13 through 17. Some differences, such as thickness, are
immediately apparent, even visually. Others, such as humidity response, must be
measured. These data were plotted in many ways in an attempt to discover a
relationship between some physical property and the measured exhaust particulate
collections. No such relationship, consistent among filter types, was found.
It is concluded that physical properties, in and of themselves, do not
account for variations observed in exhaust collections. If such a correlation
does exist, either it lies with a property not measured, or else the measurements
reported here were not sufficiently sensitive or accurate to reveal the correlation.
Chemical Properties
Chemical analyses were made of the filters using optical emission spectro-
graphic, X-ray fluorescence, and combustion techniques. These analyses were made on
the entire filter; additionally, optical emission spectrographic analyses were made
on the portion of the filter extracted by refluxing a mixture of HC1-HNO through the
BATTELUE — COLUMBUS
-------
37
filters for 3 hours (standard EPA procedure). The extraction data are of interest
primarily in specific applications calling for the acid extraction of collected
materials followed by an analysis for specified elements. The direct analysis
is of more general interest, especially for applications that do not remove the
sample from the filter prior to analysis.
The data in Table 18 show that there are no major differences in
metallic (cation) elements among the filters, with the very noticeable exception
that the Tissuequartz, made from SiO? and not glass, is far cleaner except for
Pb and a few other elements. This fact could be a deciding consideration in
choosing a suitable filter, if elemental analyses of the exhaust particulate
matter is desired.
The presence of an organic binder in the Gelman Type E filter is
apparent from the high C and H values.
It has been suggested by EPA that the Zn content correlates with the
collection of automobile exhaust particulates. From the data in Table 18 it is
not possible to establish a definite correlation. Two filters have very low Zn
contents while the other four have much higher amounts averaging about 7.5 mg per
2
47-mm filter (= 270 IJ-g/cm ). However, one of the two filters containing the low
zinc content is Tissuequartz and its properties are definitely unlike those of the
other filters. Thus the MSA, having a low zinc content, perhaps alone should be
compared with the other four filters. On this basis the MSA filter does show a
greater auto emission particulate collection. On the other hand, the Whatman GF/B
and the Gelman Type E both had high Zn contents, yet gave exhaust collections within
10 percent of the MSA filters.
Study of the Collected Samples
If unused filters yield no striking clues concerning their subsequent
behavior as collectors of diluted automobile exhaust, then the analysis of the
BATTELLE — COLUMBUS
-------
TABLE 18. CHEMICAL ANALYSIS OF GLASS FILTERS
(Micrograms of Element Per 47 mm Filter [= 17.34 ctn^j)
Whatman
GF/A. Box 1 (94 me)
Fleme«
H
LI
Be
B
C
V
m
0 Mg
> Al
•« si
H ?
m 8
r Cl
r »
m C«
1 ^
V
n cr
0 Mn
r Fe
C Co
2 Hi
m Cu
0 Zn
C As
«1 Sr
Zr.
Mo
Cd
Sn
Sb
Pb
Bi
Ba
Total percent
by weight of
elements other
than Si
Acid
Extract
<0.01
T<0.3
<0,3
<0.3
6
<0.3
<0.3
T<0.1
Hi
<1
Hi
<0.1
<0.3
<0.3
<2
T<1
<0.3
Direct
83
3
<0.5
900
111
5
5500
200
200Q
Major
30
2
40
2000
2000
10
<1.
1.
5
30
<1.
<1.
<1.
4700
<10
50
10
<1.
<2
<2
<2
10
<1
2800
20.
Whatman
GF/B. Box 1 (267 me)
Acid
Extract
<0.01
K0.3
<0.3
T<0.3
9
<0.3
<0.3
T<0.1
Hi
<1
Hi
<0. 1
<0.3
<0.3
<2
T<1
<0.3
Direct
250
3
<1
2700
209
12
16,000
250
5500
Major
30
5
20
2500
5500
30
<3.
3.
10
80
<3.
<3.
<3.
13,000
<30
130
20
<3
<5
<5
<5
10
<3
8000
17.
Gelman
Type A, Batch 8170
(151 me)
Acid
Extract
<0.01
2.
<0.3
T<0.3
6
<0.3
<0.3
0.1
Hi
)
Acid
Extract
<0.01
T<0.3
<0.3
0.3
3
<0.3
<0.3
T<0.1
<3
<1
0.3
<0.1
<0.3
<0.3
<2
<1
<0.3
Direct
34
2
<0.5
1100
111
4
11,000
750
2000
Major
60
40
20
1100
4500
2. :
<1.
2.
<2
55
<1.
<1.
<1.
<10
<10
20
3
<1
<2
<2
<2
<2
<1
10
19.
Pa 11 f ex
Tissue Quartz (117 tug)
QAO
Acid
Extract
<0.01
T<0.3
<0.3
T<0.3 -
3
<0.3
<0.3
4
<3
<1
T<0.1
<0.1
<0.3
0.4
<2
1
<0.3
Direct
130
<1
<0.5
1.
180
27
20
120
120
Major
8
5
20
10
120
6.
<1.
<1.
<1.
23
<1.
<1.
6.
<10
<10
10
2
<1
<2
<2
<2
25
<1
1
0.7
GO
-------
39
filters after collections might give insight into the causes of variabilities
among filters in collection. However, the analyses described in the following
sections did not satisfactorily explain the glaring differences in Table 1.
Elemental Analysis. The thirteen MSA filters for which data are given
in Table 3, plus two blank MSA filters were analyzed for C and H. Opposite
quarters of each filter were taken as duplicate samples for this analysis using
a Perkin-Elmer Elemental Analyzer with the sample wrapped in platinum gauze and
held in a quartz ladle. The results are shown in Table 19.
The C/H ratios, and the ratios of C to exhaust and H to exhaust were
calculated as shown. If gross amounts of carbon were present as elemental carbon,
or if much water were present, the C/H ratio would be much different than the
approximately 10:1 ratio found in the samples. (Aliphatic hydrocarbons have
a C/H ratio of about 12:2 =6, and aromatic hydrocarbons have a ratio of about
12:1 = 12.) The ratio of 1.3 to 1«4 in the blanks suggests residual water.
No obvious correlation can be seen between high exhaust amounts
(Column 2) and high C or H values. A correlation would be manifested by a
constant ratio in the last two columns, but these ratios are not constant.
Hydrocarbon Analysis. Two sets of filters with exhaust particulate
collections were analyzed for hydrocarbons using a extraction-gas chromatographic
technique. The first set analyzed is described in Table 2, two or three filters
of each type studied in this program. An entire filter was extracted for six hours
in CH_C1_ using a micro Soxhlet extractor. The solvent was removed in a Kuderna-
Danish concentrator, and the residue was diluted to 100 p.1 with CH_C1?. Immediately
2 \il of this solution was injected into the chromatographic column. The column,
4 feet of 1/8-inch stainless steel, was packed with 2.5 percent Dexsil (high
temperature silicone grease) on Supelcoport. The temperature was programmed,
BATTELLE — COLUMBUS
-------
TABLE 19. CARBON AND HYDROGEN ANALYSIS OF MSA GLASS FILTERS
0
H
H
m
r
p
m
1
n
0
r
C
I
a
c
U)
Filter Location
1
2
3
4
5
6
7
8
9
10
11
12
13
Average
Blank
Blank
Average deviation
Percent Ave. dev.
Gain per
filter, u.g
12.9W
filter, p,g(b>
176
173
lost
192
186
137
141
150
163
163
153
lost
199
168
35
38
21
12.6
H per
filter, p,g
-------
41
starting at 50 C, at + 20 C per minute to a maximum of 300 C. Dry nitrogen was
the carrier gas, and the detector was a flame ionization unit. These conditions
were selected, after some experimentation, to give a partially resolved envelope
containing most of the heavier hydrocarbons. The envelope was broad, starting
at about 6 minutes and ending at about 16 minutes, with many partially resolved
peaks superimposed. Comparing the sample chromatograms with standard straight
chain hydrocarbons, the sample elution covered the approximate carbon number
range of C..- through C-g. Most of the aliphatic, aromatic, saturated, unsaturated,
and polynuclear compounds would be represented within the broad envelope; highly
polar compounds such as carboxylic acids would not be detected in this manner.
The analyses were repeated later by evaporating the CH Cl_ solvent,
then rediluting to 100 \il and repeating the injection and elution. The data from
both analyses are given in Table 20.
Rather clearly, the data in Table 20 are not sufficiently precise to
permit conclusions to be drawn. No blanks were run except for the Tissuequartz
filters and this lack further contributes to an overall uncertainty. The
method for determining total hydrocarbons needs further refinement if this aspect
of filter variability is to be explored.
The MSA filters with exhaust particulates, for which data are given in
Tables 3 and 19, were analyzed for total hydrocarbons using the extraction -
chromatography method described above. In this case one quarter of the filter was
extracted; two injections, elutions, and area measurements were made for each filter.
The hydrocarbon data, and pertinent data from Tables 3 and 19, are shown in Table 21.
The important fact in Table 21 is that the hydrocarbon blank is essentially
the same as the average hydrocarbon value found in the filters with exhaust particulates,
(Note that such was not the case for Tissuequartz, Table 20.) Any future work must
IATTELLE — COLUMBUS
-------
42
TABLE 20. HYDROCARBON CONTENT OF EXHAUST COLLECTIONS ON GLASS FILTERS
Filter Type(a)
Whatman GF/A
Whatman GF/A
Whatman GF/B
Whatman GF/B
Gelman Type A
Gelman Type A
Gelman Type E
Gelman Type E
MSA
MSA
MSA
Tissuequartz
Tissuequartz
Blank
Average
Filter
Location^3'
10
6
7
8
4
3
2
5
1
13 •
12
9
11
Blank
Gain per
filter, ng (a>
288
255
421
890
1,403
687
419
416
369
520
295
55
180
--
Hydrocarbon ( '
Run 1
1.91
1.48
3.17
3.82
2.92
2.26
3.45
3.61
2.46
3.35
3.11
3.50
3.39
0.73
Run 2
2.65
1.90
2.98
4.59
1.82
2.50
2.42
3.76
3.60
3.14
2.87
3.93
4.07
0.60
Average
1.98
3.64
2.38
3.31
3.09
3.72
0.66
(a) Data are from Table 7.
(b) Arbitrary units (square inches under elution peak),
BATTELLE — COLUMBUS
-------
43
TABLE 21. HYDROCARBON CONTENT OF EXHAUST COLLECTIONS
ON MSA GLASS FILTERS.
Gain per C per
Filter Location Filter, M.g(a) Filter, M-g (b)
1
2
3
4
5
6
7
8
9
10
11
12
13
Average
Average deviation
Percent Avg. Dev.
Blank
Blank
(a) From Table 3.
(b) From Table 19.
(c) Arbitrary units
345
335
lost
451
430
411
407
501
438
347
422
300
544
411
53
12.9
-16
-23
(square
176
173
--
192
186
137
141
150
163
163
153 :
lost
199
168
21
12.6
35
38
inches under peak).
H per HC per Ratio
Filter, |ig (b) Filter (c> HC/Exhaust(d)
20
20
--
20
19
12
14
13
18
18
14
--
28
17.7
4
26.1
26
26
Average of
1.8
2.8
--
2.0
2.8
2.8
2.4
3.2
3.1
3.3
2.6
--
4.4
2.83
0.47
16.9
3.0
2.4
two determinations,
0.0052
0.0084
--
0.0044
0.0065
0.0068
0.0059
0.0064
0.0071
0.0095
0.0062
--
0.0081
0.0068
0.0011
16.2
--
--
not
corrected for blank.
(d) Exhaust = Gain per filter, Column 2.
BATTELLE — COLUMBUS
-------
44
establish unequivocally the relationship of the blank and the sample. Is the
blank merely a reagent and column blank, or is it truly a filter blank? If it
is a filter blank, then is the extraction-chromatography method really detecting
any hydrocarbons in the sample? Are there, in fact, any hydrocarbons there to
detect? These questions must be answered before any conclusions can be made
concerning the role of hydrocarbons in the observed particulate weight.
If the question of blanks is ignored, and only the relationship between
the measured hydrocarbon peak and the observed weight is examined, a slight
correlation may exist. Certainly the data should be subjected to rigorous
statistical analysis; these data, and those in Tables 19 and 20, were not
considered sufficiently precise or comprehensive to warrant such analyses.
Thermogravimetric Analysis. If part or all the observed variations
in the weight of collected automobile exhaust particulates are caused by adsorbed
or condensed hydrocarbons, which normally would have passed through the filter
as a gas but were retained in varying amounts, then it should be possible to drive
off these extraneous substances by heat. This hypothesis was tested by performing
thermogravimetric analysis (TGA) on three of the MSA filters with exhaust particu-
lates (Tables 3, 19, and 21) and two blank MSA filters„ For the TGA runs, one-
fourth of a 47-mm-diameter filter was taken. This portion was placed in the pan
of the TGA, which had been carefully calibrated and tared in the appropriate weight
range. The chamber was evacuated to the 10 torr range using rotary and oil
diffusion pumps, and the steady state weight of the filter was obtained. Heating
at the rate of 4 C per minute was then begun, using a controlled resistance heater
surrounding the pan and sample leg of the TGA unit. Weight loss was recorded on a
stripchart recorder. The maximum temperature of 490 to 500 C was reached in about
two hours. At the end of the heating period the furnace was removed and the leg
allowed to cool to ambient temperature, at which point the steady state weight was
again observed.
BATTELLE — COLUMBUS
-------
45
The two blank samples gave remarkable weight losses, 0.0890 and 0.0900 mg
from starting weights of 27.2072 and 28.6584 mg, respectively. The blank weight
loss for these MSA filters corresponds to about 0.35 mg per 47-mm filter, nearly
the same as the weight loss found at ambient temperature when MSA filters were
transferred from humid to dry conditions (Table 17). The two types of weight loss
may not be similar, however. At ambient temperature and pressure the loss of water
was essentially complete after about 20 to 25 minutes whereas the loss of weight in
the TGA experiments was still occurring after 2 hours. Thus, the similarity in
weight losses in the humidity-ambient temperature and the TGA experiments probably
is coincidental.
Similar TGA runs were made using quarter-filter segments from filters 9,
10, and 13 (Tables 3, 19, ;and 21). These showed weight losses of 0.540, 0.628,
and 1.160 mg, respectively, calculated on an entire 47-mm-diameter filter. If the
average blank value of 0.358 mg is subtracted from each sample weight loss, the net
losses are 0.182, 0.270, and 0.802 mg, respectively. The first two values are within
reason considering the weights of particulate matter collected, but the value from
filter 13 is about 50 percent more than the total weight of sample.
Thus, the TGA data present a confusing picture. On the one hand the blank
runs and the runs for two samples are plausible. On the other hand the run for
Sample 13 is not logical, and casts doubt on the validity of the technique employed.
Conclusions Concerning Analysis of Collected Samples. The analytical
steps failed to discover the cause of differences observed among collected exhaust
samples. The approach taken in this program was to attempt to find a single component
of the samples to account for the variations. Perhaps this approach is still valid, and
the search should continue. An alternative approach suggests itself, however. Per-
haps the weight variations are not caused by a single event, such as hydrocarbon
BATTELLE — COLUMBUS
-------
46
adsorption, but rather are a manifestation of many unrelated events whose
summation produces the observed variations. If the latter be true, then future
work should stress statistical studies to discover the more subtle relationships
among variables. If such relationships can be identified, then means can be
devised to reduce the dependency of the exhaust collection upon these variables.
Suggestions for additional studies along both lines are presented in the future
work section of this report.
Calibration of Exhaust Dilution
Tunnel
The dilution tunnel used in this program was constructed for a project
studying the fate of automobile exhaust particulate matter in the atmosphere,
sponsored by the Coordinating Research Council and Environmental Protection Agency.
The tunnel is about 2 feet in diameter and about 36 feet long. About 30 feet from
the dilution point are located 13 sampling ports arranged in a vertical and a
horizontal line, 3 ports per quadrant plus 1 in the center. Each port leads to a
fitting outside the tunnel; a filter holder with a critical flow orifice is attached
to each fitting, and the exit sides of the filters are connected via a manifold to
a large mechanical vacuum pump. The location of the inlet ports was shown in -
Figure 1.
It is important to be certain that each of the 13 sampling ports is
equivalent to all the others. If the ports would not sample representative
fractions of the diluted exhaust, i.e., if the exhaust stream preferentially
followed a limited path through the tunnel, then comparisons among filters sampling
the various ports with the assumption that all sampling positions are identical
would not be valid.
BATTEULE — COLUMBUS
-------
47
The uniformity of distribution of the diluted exhaust was checked by
injecting a fluorescent dye aerosol into the exhaust stream before dilution,
catching the dye on the 13 filters, and measuring the quantity by nondispersive
fluorimetry. The uniformity of the critical flow orifices is checked simultaneously
by this experiment. An aerosol can ("spray paint" type) was prepared containing
dye, toluene solvent, and Freon 12 propellant. The dye was injected into the
exhaust stream through a small copper tube whose exit axis was concentric with
the automobile exhaust pipe. With the automobile running on the chassis dynamometer
at a steady 35 mph and 3 hp, the contents of the aerosol can was slowly injected
into the exhaust. Simultaneously with the filter collections, a cascade impactor
collection was made to determine the aerosol particle size distribution.
The uniformity of distribution of the aerosol on the filters is given
in Table 22. It is clear that the variations are minor and that the exhaust is
mixed essentially completely throughout the cross section of the tunnel. The
percent coefficient of variation is less than that found for any other parameter
of the filter materials measured in this study. This fact shows that flow variations
within the tunnel are not a major source of filter-to-filter variations.
One filter holder (location 12) was found to have a 1.0 cfm critical
orifice, approximately twice the 0.5 cfm orifices of the other 12. Data from
this holder in this report have been corrected by a factor of 0.5 when possible.
BATTELLE — COLUMBUS
-------
48
TABLE 22. DISTRIBUTION OF DYE AEROSOL AT REPLICATE
SAMPLING SITES
Filter Location^ '
1
2
3
4
5
6
7
8
9
10
11
12
13
Photometer Reading
Arbitrary Units
80.5
81.2
80.5
78.3
83.0
79.4
85.6
78.0
76.4
85.5
74.0
75.5(b)
82.3
(a) See Figure 1 for location diagram
(b) Corrected _- see text
/ ~ 2
CT =\/—i—^~—<— , where R is the arbitrary photometer
•J n- i
reading (=3.59)
Percent coefficient of variation v = — x 100 (=4.48)
R
BATTELLE — COLUMBUS
-------
49
In those cases when a correction might not be valid, the data were discarded as
"lost". The improper critical flow orifice was replaced with the correct size
before making the collections reported in Tables 4, 5, and 6 and 8 through 12.
Particle size distribution of the aerosol was determined by drawing
a portion of the diluted exhaust through a Battelle cascade impactor backed by
a glass fiber filter. The amount of dye impacted on each stage of the impactor
and caught by the backing filter is given in Table 23, which shows that the
major portion of the aerosol is less than 1 Mm. A plot of the cumulative percent
of each size is shown in Figure 7.
CONCLUSIONS AND RECOMMENDATIONS
Many conclusions may be drawn from the data presented in this report;
some conclusions have been presented at the end of each major section.
Two primary conclusions are apparent. First, real differences exist
and can be measured among the weights of diluted automobile exhaust collected
on different types of filters; similar differences are not apparent when particulate
matter from ambient air is collected. Second, no single factor, or combination of
factors, has been identified as the cause of these variations. Many causes, both
within the filters themselves and also within the collected samples, were examined
as possible causative factors. Although some trends and possible relationships
might be postulated from the data, these cause-effect relationships are tenuous
at best; more data are needed, in order to establish definite links.
Recommendations are presented in the concluding section.
FUTURE WORK
The key to the successful search for the causes of weight variations
in exhaust collections may ultimately lie in statistics. The search for a
single variable responsible for the weight differences, either in the filter or
BATTELLE — COLUMBUS
-------
50
TABLE 23. SIZE DISTRIBUTION OF DYE AEROSOL
Stage Number
Filter
7
6
5
4
3
2 (a)
Effective
Stoke 's Diameter,
Pan
0.25
0.25
0.50
1.0 :
2.0
4.0
8.0
Percent of
Total on Stage
76.8
10.1
2.3
3.1
4.1
2.5
1.1
Cumulative
Percent
76.8
86.9
89.2
92.3
96.4
98.9
100.0
(a) Stage 1 (16 pm) not used in cascade impactor assembly.
BATTELLE — COLUMBUS
-------
51
QQ Q
yy.y
qq
QO
yo
qs
*~ PO
CP
1 op
>. ^
«— tf\
c 70
^ fin
u 60
Q_ co
C)U
-------
52
the sample, did not succeed. While not dismissing the possibility that that
single variable has been overlooked, it seems more probable that many factors
have a cumulative effect. If so then it follows that the number of observations
should be large enough to minimize the impact of wild or false data and to
allow trends to be seen. Of course the cost of acquiring large amounts of data
can escalate beyond reason if the experimental work is not carefully planned and
designed statistically so that the data can be interpreted meaningfully.
The search for variability relationships should be confined to intertype
studies rather than intratype studies. For the initial studies, the variability
among filters of a given type should be treated as a fact, not investigated.
Intratype variability-properly follows the successful identification of variability
among different types of filters.
Translated into specific steps, these broad suggestions lead to the
following program.
(1) Select two to four types of filters for intensive study. The
Whatman GF/A and the MSA types should be included since they contain two distinctly
different concentrations of zinc.
(2) Make repetitive simultaneous exhaust collections using the selected
types, varying controllable parameters one at a time.
(a) Face velocity at filter.
(b) Filter holder configuration.
(c) Dilution of exhaust.
(d) Humidity and temperature at collection point.
(e) Ambient humidity during tare and gross weighings.
It is realized some of these variables are mutually dependent.
(3) Statistically analyze the data and attempt to correlate observed
weights with the variables.
BATTELLE — COLUMBUS
-------
53
(4) If correlations are found, analyze several samples from the
extreme ends of the range, e.g., some from low and some from high face velocities
if that is a correlating variable.
(5) Devise experimental conditions to measure the exit stream from
each filter. This could involve monitors placed in the volumes immediately
following each filter to measure water vapor, hydrocarbons, or any untrapped
particulate species. Alternatively, backup filters and/or impingers could be
used to trap material escaping the first filter.
(6) Experiment with predrying beds as has been done in some
ambient air particulate studies. It is recognized that such beds may alter
the chemical composition of the exhaust and introduce another variable whose
effect will be very difficult to measure.
(7) It may prove desirable to alter the exhaust characteristics, as
by running the engine too rich, to cause exaggerated collection errors. This
approach has the severe disadvantage of yielding high weights of particulates
that may swamp or obscure the smaller but significant variations now observed.
(8) Refine analytical methods to determine specific components.
This task is not incompatible with the hypothesis that more than one cause of
variation is responsible, since each individual cause eventually should be studied.
Because water and hydrocarbons remain the most probable, but unproven, causes of
variations, better means are needed to determine them. Modified extraction -
chromatographic techniques should be developed for hydrocarbons; perhaps it will
become necessary to separate the hydrocarbons into classes, even individual
compounds. For water, a modified TGA technique is envisioned, wherein the amount
of water evolved as a function of temperature is measured, not the weight loss.
BATTELLE — COLUMBUS
-------
54
If these experiments are successful in identifying the causes of
variations among the few types of filters, the next step is to determine if
the same causes also affect other filter types. Paired, statistically valid
experiments should be able to establish whether or not the findings from the
first filter types are applicable to all other types. If as expected the first
results are applicable to other filter types, the variables causing weight
differences among filter types can be summarized as generally applicable, with
statistical limits given for the average effect of each variable.
When that phase is completed, the next task is to define statistical
limits for variations within a given filter type. Many of these data will have
been assembled during the other studies, but should be retabulated to stress
intratype variability.
BATTELUE — COLUMBUS
-------
BIBLIOGRAPHIC DATA
SHEET
1. Report No.
EPA-R2-73-16Q
| 3. Recipient's Accession No.
4. Title and Subtitle
Develop an Operational System for Evaluating and Testing
Methods and Instruments for Determining the Effects of Fuel
and Fuel AdditiVR^ on Aiitnmnhilp F.TTII <;<;i nrm
5. Report Date
.. February 1 973
6.
7. Author(s)
F R B
8. Performing Organization Rcpt.
No.
9. Performing Organization Name and Address
Battelle Memorial Institute
Columbus Laboratories
505 King Avenue
10. Project/Taslc/Work Unit No.
11. Contract/Grant No.
Tnl i
OViin
12. Sponsoring Organization Name and Address
Chemistry and Physics Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
13. Type of Report & Period
Covered
Final Report, 1 year
14.
15. Supplementary Notes
media.
--Initial contract scope of work was modified to cover evaluation procedures forfilter
16. Abstracts
Causes of observed weight variations in collected particulates from automobile
exhaust were sought. Chemical and physical properties of unused glass fiber
filters were studied, and some chemical analyses of collected exhaust
particulates were performed. No clear indication of a single cause for the
variations could be discerned. An extensive statistical analysis of data
obtain'ed from additional experiments is suggested as a means of pinpointing
the causes of the weight variations.
17. Key Words and Document Analysis. 17o. Descriptors
Automobile exhaust
Glass Fiber Filters
Particulate collection
Chemical analysir
Humidity effects
17b. Identifiers/Open-Ended Terms
17c. COSATI Field/Group
18. Availability Statement
19..Security Class (This
Report)
UNCLASSIFIED
20. Security Class (This
Page
UNCLASSIFIED
21. No. of Pages
22. Price
FORM NTIS-35 (REV. 3-72)
USCOMM-DC
------- |