EPA-600/2-78-038
June 1978
Environmental Protection Technology Series
EVALUATING AND OPTIMIZING
ELECTRON MICROSCOPE METHODS
FOR CHARACTERIZING AIRBORNE ASBESTOS
Environmental Sciences Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development. US. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR) .
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL PROTECTION TECH-
NOLOGY series. This series describes research performed to develop and dem-
onstrate instrumentation, equipment, and methodology to repair or prevent en-
vironmental 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.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA 600/2-78-038
June 1978
EVALUATING AND OPTIMIZING ELECTRON MICROSCOPE
METHODS FOR CHARACTERIZING AIRBORNE ASBESTOS
by
A.V. Samudra, F.C. Bock, C.F. Harwood, and J.D. Stockham
IIT Research Institute
Chicago, Illinois 60616
Contract No. 68-02-2251
Project Officer
Jack Wagman
Director, Emissions Measurement and Characterization Division
Environmental Sciences Research Laboratory
Research Triangle Park, North Carolina 27711
ENVIRONMENTAL SCIENCES RESEARCH LABORATORIES
OFFICE OF RESEARCH AND DEVELOPMENT
U. S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, N. C. 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Sciences Research
Laboratory, U. S. Environmental Protection Agency, and approved for publica-
tion. Approval does not signify that the contents necessarily reflect the
views and policies of the U. S. Environmental Protection Agency, nor does
mention of trade names or commercial products constitute endorsement or
recommendation for use.
ii
-------
FOREWARD
The occurrence of asbestos or asbestiform minerals as pollutants in the
ambient air and in supplies of food and drinking water has caused consider-
able concern because occupational exposures to these fibrous materials have
been found to induce mesothelioma of the pleura and peritoneum, as well as
cancer of the lung, esophagus, and stomach, after latent periods of about 20
to 40 years.
Electron microscopy is currently the principal technique used to
identify and characterize asbestos fibers in ambient air and water samples.
Because of the poor sensitivity and specificity of conventional bulk
analytical methods, electron microscopy is also being used for routine
measurement of airborne or waterborne asbestos concentrations. The several
laboratories that perform such analyses generally have reasonable internal
self consistency. However, interlaboratory comparisons have shown that the
results obtained by the separate laboratories are often widely different.
In recognition of this problem, the Environmental Sciences Research
Laboratory, U. S. Environmental Protection Agency, initiated a comprehensive
two-year study (June 1975 - June 1977) through EPA Contract No. 68-02-2251
to evaluate the various electron microscope procedures currently used for the
measurement of airborne asbestos concentrations. The scope of work included
the development of an optimum procedure incorporating the best features of
current methods together with whatever improvements in sample collection,
specimen preparation, and electron microscope examination that seem desirable
for enhancement of accuracy and precision and for reduction of analysis time
and cost.
A manual entitled "Electron Microscope Measurement of Airborne Asbestos
Concentrations — A Provisional Methodology Manual" describing an optimized
method resulting from this study has been published as EPA Report 600/2-78-
178 (August 1977). This final report contains a detailed account of the
investigation and the experimental data supporting the provisional methodology.
Jack Wagman
Project Officer
A. Paul Altshuller
Director
Environmental Sciences
Research Laboratory
Research Triangle Park, N.C.
iii
-------
ABSTRACT
Electron microscopy is currently the principal technique used to identify
and characterize asbestos* fibers** in ambient air and water samples. Varia-
tions in instrument capabilities, operator proficiency, and the myriad of tech-
niques used in microscopy laboratories have resulted in wide data scatter.
Under Contract No. 68-02-2251, a research program was initiated to evaluate the
electron microscope methods and subprocedures currently in use at different
laboratories to measure airborne asbestos fiber concentrations and develop a
composite procedure that would minimize the variability of results.
Other objectives of the program were to provide a handbook describing the
optimized method and to test the ruggedness of the optimized method through
interlaboratory analyses.
A five-phase program of statistically designed experiments was used to
evaluate 19 major independent variables and 50 subprocedures (or variable
levels). The data from transmission and scanning electron microscopy examina-
tion were analyzed by statistical techniques to evaluate the effects of the
independent variables and subprocedures on two major dependent variables,
asbestos fiber number and mass concentrations. Multiple criteria were used to
select the independent variable levels for the optimized procedure.
The optimized method for estimating the concentration of asbestos fibers
in ambient air samples has the following features:
1. Use polycarbonate membrane filters to collect ambient air samples.
2. Coat the polycarbonate filter with a thin layer of carbon to lock-in
the collected fibers.
* Asbestos is used as a collective term for the six minerals: chrysotile,
amosite, crocidolite, and the asbestiform varieties of anthophyllite,
actinolite, and tremolite.
** The term fiber is used for a particle with an aspect ratio of 3:1 or
greater, and with substantially parallel sides.
iv
-------
3. Transfer the collected fibers to 200 mesh electron microscope grids
in a modified Jaffe washer using chloroform to dissolve the filter.
4. Examine the grids in a transmission electron microscope (TEM) at a
screen magnification of 16,OOOX.
5. Identify and characterize each fiber from its morphology and
selected-area electron diffraction (SAED) pattern. Place the fibers
into three categories: chrysotile, amphibole, and non-asbestos.
6. Enumerate the number of asbestos fibers using the field-of-view
method. If mass concentration information is needed, measure the
length and width of each fiber and compute the mass from the fiber
volume and density data.
A manual describing the optimized method was prepared and reviewed by
six independent laboratories. The manual was subsequently published as
Environmental Protection Agency Technology Series Document EPA-600/2-77-178,
"Electron Microscope Measurement of Airborne Asbestos Concentrations - A
Provisional Methodology Manual", August, 1977.
The ruggedness of the optimized method was tested by the six laboratories.
These laboratories used the method to analyze two filters upon which airborne
asbestos fibers had been deposited and were carbon coated to prevent loss of
the fibers. One sample was prepared by sampling pure UICC chrysotile that
was aerosolized into a large chamber. The second sample was collected from
the air inside an industrial plant processing asbestos. These samples were
labeled "Lab" and "Field" sample, respectively.
The interlaboratory studies showed the average precision of chrysotile
fiber concentration estimates, as determined by the ratio of the standard
error of the mean to the mean expressed as a percentage, was about 21% for
both the lab and field samples. The average precision of chrysotile mass
concentration estimates was 22% for the lab sample and 44% for the field
sample when ashing of the filter was used as a subprocedure. When ashing was
not used, the average precision for the field sample was 54%. The lower pre-
cision of the chrysotile mass concentration estimates for the field sample is
attributed to the presence of a few large bundles of fibers. These few
bundles do not affect the number concentration estimates but significantly
influence mass estimates. As a result, it is suggested that fiber bundles
3
greater than 1 ym should be reported separately.
-------
The accuracy of the transmission electron microscope procedure for esti-
mating the mass concentration of chrysotile was determined by comparing the
computed mass from fiber volume and density data with the computed mass calcu-
lated from the magnesium concentration obtained by x-ray fluorescence spectro-
3
metry (XRF). A chrysotile fiber density of 2.6 g/cm was used to compute
fiber mass from size data. A factor of 3.8 times the magnesium concentration
was used to determine chrysotile fiber mass by XRF. For the lab sample, the
mass estimates for chrysotile agreed within 10%. For the field sample, the
mass estimate obtained by electron microscopy was a factor 4.2 less than
that obtained by XRF. The difference is attributed to the presence of
fiber bundles and other sources of magnesium in the field sample.
Testing of the subprocedure that incorporates ashing, resuspension, ultra-
sonification, and refiltering of the lab and field samples gave inconclusive
results. The mean fiber lengths were decreased by the ashing subprocedure and
fiber concentration estimates were significantly increased (4-8 times higher
for ashed sample than the unashed) . The data cannot resolve whether the ap-
parent increase in fiber concentration in the ashed sample results from fiber
breakage or results from less interference in observing and identifying small
fibrils in the diluted ashed sample. It is suggested that the ashing subpro-
cedure should be used only when the direct transfer method is not suitable.
This report is submitted in fulfillment of EPA Contract No. 68-02-2251,
IITRI Project No. C6351, by IIT Research Institute under sponsorship of the
Environmental Protection Agency. This report covers the period July, 1975,
to June, 1977.
vi
-------
CONTENTS
Foreword . . . * iii
Abstract iv
Figures , . ix
Tables xii
Acknowledgments xv
1. Introduction 1
2. Conclusions 3
3. Recommendations 6
4. Scope of the Work 8
Introduction 8
Strategy 9
5. Statistically Designed Experimental Plans 17
Introduction 17
Five phase program 18
6. Experimental Work 28
The preparation of laboratory filters of controlled asbestos
loading in Phase 1 28
Sample preparation for TEM 34
Examining samples in transmission electron microscope 34
Data recording 36
Experimental work in Phase 2: Study of fiber identification
method. 36
Experimental work in Phase 3: Evaluating a TEM and an SEM . . 37
Experimental work in Phase 4: Ashing and sonification .... 40
Experimental work in Phase 5: Study of direct drop method . . 44
7. Results and Discussion of Phase 1 45
Introduction 45
Criteria selected 45
Summary of Phase 1 results 45
Statistical distribution tests 47
Mass concentration estimates 67
8. Results and Discussion of Phases 2, 3, 4 and 5 , . 70
Phase 2 results 70
Results and discussion of Phase 3 87
Phase 4 results 90
Statistical analysis of Phase 5 data 104
9. Provisional Optimized Method and Round-Robin Testing 109
Preparing calibration filters . 109
Final choice of samples for round-robin test 110
Independent estimate of chrysotile mass concentrations .... Ill
vii
-------
CONTENTS (continued)
10. Results and Discussion of Round-Robin Tests 113
Poisson distribuiton tests 113
General procedures 115
Interlaboratory comparisons 121
Graphical representation of results .... 125
Accuracy and precision of estimates 125
Effect of ashing, ultrasonification, and reconstitution .... 137
Conclusions 145
References . 147
Appendices
A. The Experiment Design for Phase 1 152
B. Regression Analysis 157
C. Poisson Distribution Tests 166
D. Optimized Method for Measurement of Airborne Asbestos Concentrations 175
E. Estimating Chrysotile Mass on Air Filters Using Neutron Activation
Technique 177
F. X-Ray Fluorescence Analysis of Standard Samples of Chrysotile . . . 179
viii
-------
FIGURES
Number Page
1 Flow chart of electron microscope procedures for estimating size
distribution and concentration of airborne asbestos 10
2 (a) Top view and size view of aerosol chamber showing location of
apparatus. (b) Flow diagram of aerosol generator 30
3 Graphical presentation of performance equation 9 in Phase 1. Net
contribution to square root of fiber concentration (no. of all
fibers/cm3 of air) 59
4 Graphical presentation of performance equation 9 in Phase 1. Net
contribution of square root of fiber concentration (no. of all
fibers/cm3 of air) 60
5 Graphical presentation of performance equation 10 in Phase 1. Net
contribution to natural log of mass concentration of all fibers,
Vlg/m3 of air 61
6 Graphical presentation of performance equation 10 in Phase 1. Net
contribution to natural log of mass concentration of all fibers,
yg/m3 of air 62
7 Estimated number concentration of chrysotile fibers in the nine
Phase 2 samples (standard classification method), with 95% confidence
intervals 82
8 Estimated mass concentration of chrysotile fibers in Phase 2 (standard
and alternative classification methods) in relation to filter
composition, transfer method, and identification technique, with 90%
confidence intervals 83
9 Estimated geometric mean length of chrysotile fibers in Phase 2
(standard classification method) in relation to filter composition
and transfer method, with 90% confidence intervals . 84
10 Estimated percent of all Phase 2 fibers that were exceptional
(ambiguous or other by the standard classification method) in
relation to transfer method and identification technique, with 90%
confidence intervals 85
11 Graphical presentation of performance equation 6 in Phase 4. Net
contribution to square root of fiber concentration, 106/cm2 of
filter 97
ix
-------
FIGURES (continued)
Number
12 Graphical presentation of performance equation 9 in Phase 4. Net
contribution to mean log of fiber length (ym) 98
13 Graphical presentation of performance equation 1 in Phase 5. Net
contribution to square root of fiber concentration (106 fibers/cm2) . 108
14 90% confidence intervals (inner) and 95% confidence intervals on the
mean fiber number concentration 117
15 95% confidence intervals about the means in laboratory air sample 154
(see Tables 42 and 45) 126
16 95% confidence intervals about the means in field air sample 661 (see
Tables 43 and 46), ashed samples only 127
17 95% confidence intervals about the means in field air sample 661 (see
Tables 44 and 47), unashed samples only 128
x
-------
TABLES
Number Page
1 Procedural Variables of Electron Microscope Method 11
2 Independent Variables Selected for Evaluation in This Study 15
3 Representative Dependent Variables in Phase 1 19
4 Independent Variables, Phase 1 20
5 Phase 1 Experiment Design (Compact Notation) 21
6 Phase 1 Experimental Plant (Long Notation) 22
7 Phase 2 Experiment Design 24
8 Phase 3 Experiment Design 25
9 Phase 4 Experiment Design 26
10 Phase 5 Experiment Design . 27
11 Experimental Scheme for Simulated Air Samples for Phase 1 35
12 Scheme of Electron Microscope Parameters for Fiber Identification
Methods in Phase 2 38
13 Details of the Ashing Parameters Used in Phase 4 42
14 Summary of Phase 1 Data 46
15 Tests for Applicability of the Poisson Distribution to Number of
Fibers Per Field 48
16 Variable Level Frequency Distribution in Two Groups 50
17 Precision in Fiber Count Per Field as a Criterion for Optimizing. . . 53
18 Variable Level Frequency Distribution in Two Group 54
19 Dependent Variables, Phase 1 ..... 57
20 Signs of Coefficients of Independent Variables in Performance
Equations, Phase 1 58
xi
-------
TABLES (continued)
Number Page
21 Optimization of Variable Levels According to Four Different
Criteria 66
22 Estimates of Number and Mass Concentration of All Fibers Per Unit
Volume of Air, Phase 1 68
23 Numbers of Fibers Observed and Classified, Phase 2 Samples. ...... 72
24 Concentrations of Chrysotile Fibers, Phase 2 Samples 73
25 Size Distributions of Chrysotile Fibers, Phase 2 Samples 74
26 Concentrations of All Fibers and of Fibers of "Ambiguous" and
"Other" Categories, Phase 2 Samples 75
27 Phase 2 Regression Equations 78
28 Properties of Phase 2 Regression Equations 79
29 Summary of Phase 3 Data 88
30 Difference in Number of Chrysotile Fibers Counted When Same Grid
Openings are Observed Under SEM and Convensional TEM Mode in
JEOL 100C 91
31 Estimating Chrysotile Asbestos in Phase 4 Samples (Ashing and
Sonification Experiments) 92
32 Size Distribution Characteristics of Chrysotile Fibers in Phase 4
Samples (Ashing and Sonification Experiments) 93
33 Values of Dependent Variables in Phase 4 95
34 Means and Standard Deviations of Dependent Values in Phase 4 96
35 Regression Equations in Phase 4 96
36 Characteristics of Fiber Length in Cumulative Distribution in
Phase 4 100
37 Characteristics of Fiber Width in Cumulative Distribution in
Phase 4 101
38 Values of Dependent Variables in Phase 5 105
39 Fiber Number and Mass Concentration in Phase 5 106
40 Tests for Applicability of the Poisson Distribution to Number of
Fibers Per Field 114
xii
-------
TABLES (continued)
Number Page
41 Mean Values and Lower and Upper Limits of Fiber Concentration
Estimate According to Poisson Distribution 116
42 Summary of Round-Robin Test Results on Air Sample 154 118
43 Summary of Round-Robin Test Results on Field Sample 661 (All Samples
Ashed) 119
44 Summary of Round-Robin Test Results on Field Sample 661 (All Samples
Analyzed Without Ashing) 120
45 95% Confidence Intervals on the Mean Estimates for Individual
Operators in Air Sample 154 (See Table 42) 122
46 95% Confidence Intervals on the Mean Estimates for Individual
Operators in Field Air Sample 661 (See Table 42) (Ashed Samples
Only) 123
47 95% Confidence Intervals on the Mean Estimates for Individual
Operators in Field Air Sample 661 (See Table 44) (Unashed Samples
Only) 124
48 Precision of Fiber Concentration Estimates on Laboratory Air
Sample 154 130
49 Precision of Fiber Concentration Estimates of Field Sample 661 (All
Samples Ashed) 132
50 Precision of Fiber Concentration Estimates of Field Sample 661
(Unashed) . 133
51 Precision of Different Measurements in the Two Samples 134
52 Effect of a Few Large Bundles on Number Concentration and Mass
Concentration of Chrysotile in Field Sample 661 ..... 136
53 Effect of Low Temperature Ashing and Reconstltution of Fiber
Concentration Estimates 138
54 Effect of Low Temperature Ashing and Reconstitution of Mean Fiber
Dimensions 139
55 Length Distribution in Ashed and Unashed Samples, Data from
Operator Number 2 141
56 Length Distribution in Ashed and Unashed Samples, Data from
Operator Number 4 142
xiii
-------
TABLES (continued)
Number Page
57 Length Distribution in Ashed and Unashed Samples, Data from
Operator Number 5 143
58 Length Distribution in Ashed and Unashed Samples, Data from
Operator Number 6 144
xiv
-------
ACKNOWLEDGMENTS
The authors gratefully acknowledge the expert technical support of
several colleagues. Meticulous work in preparing simulated air samples in
our laboratory by Mr. David R. Jones deserves a special mention. The assist-
ance of Mr. Jones and Mr. George Yamate in electron microscopy work is also
acknowledged. Prompt secretarial effort of Miss Elaine Brown and Miss Bonnie
Fitzpatrick is duly acknowledged.
Acknowledgments are also due to Dr. Philip Cook of EPA, Duluth, MN,
Mr. J. M. Long of EPA, Athens, GA, Mr. John Miller of EPA, Research Triangle
Park, NC, Dr. Ralph Zumwalde of NIOSH, Cincinnati, OH, Dr. Edward Peters of
A.D. Little, Cambridge, MA, and Miss Wendy Dicker of Ontario .Ministry of
Environment, Toronto, Canada, for reviewing the methodology manual and for
participating in the round-robin test of air samples.
Finally, the authors thank Dr. Jack Wagman of the Environmental Protection
Agency, Research Triangle Park, NC, for his continued interest and encouragement,
xv
-------
SECTION 1
INTRODUCTION
The association of asbestos microfibers with adverse health effects
prompted various governmental agencies and private industries to consider the
electron microscope for characterizing microfibers in air. The choice of the
electron microscope is ideal because of its ability to detect very fine fibers,
estimate the size and shape, and identify each fiber from morphology and elec-
tron diffraction supplemented, in some instances, by energy-dispersive X-ray
analysis. No other instrument matches the modern analytical electron micro-
scopes in overall capabilities. The information gained on the quantity of
fibers and their characteristics in given localities can be utilized to under-
stand the significance of fiber exposure in terms of health hazard.
Unfortunately, because of the recent and rapid utilization of electron
microscopes for quantifying fiber concentration levels, there is no standard
methodology. In general, many different techniques and procedures have been
used to collect samples, prepare samples for electron microscopy, examine the
samples in the electron microscope, and interpret and evaluate the results.
As a consequence, the various laboratories performing the asbestos character-
ization in air samples have reasonable intralaboratory agreement; but, inter-
laboratory agreement is totally unacceptable. This wide variability in results
of electron microscope studies makes the technique unacceptable in a court-
of-law, and the electron microscope results are generally treated as order-
of-magnitude estimates for broad comparisons only. Since this is a serious
limitation on a powerful tool, it is very important to understand the sources
of this variation and minimize or eliminate the variation by appropriate
optimization.
The U.S. Environmental Protection Agency realized the need for the develop-
ment of an optimum methodology, particularly with respect to asbestos in air
because of its known association with cancer. The present study at IIT Research
Institute, directed at developing an optimum analytical methodology for
1 '
-------
determining asbestos* fibers** in the ambient air, uses statistically designed
experimental techniques for simultaneous evaluation of a large number of inde-
pendent variables and subprocedures. As stated in the scope of work, the inves-
tigation included "the development of an electron microscope procedure incor-
porating the best features of the current methods together with whatever im-
provements in sample collection, specimen preparation, and electron microscope
examination seem desirable for enhancement of accuracy and precision and re-
ducing analysis time and cost." The optimum procedure sought in this study is
one that yields maximum information on asbestos fiber characteristics in the
airborne state (from studying fibers collected on suitable filters) , including
fiber count and size distribution, as well as mass concentration.
* Asbestos is used as a collective term for the six minerals: chrysotile,
amosite, crocidolite, anthophyllite, actinolite, and tremolite.
** The term fiber is used for a particle with an aspect ratio of 3:1 or
greater, and with substantially parallel sides.
-------
SECTION 2
CONCLUSIONS
The major conclusions drawn from the statistically designed experiments
and the interlaboratory investigations are itemized below.
1. The optimized method for characterizing the asbestos levels in
ambient air samples by electron microscopy is summarized as follows:
a. Collect ambient air samples for asbestos analysis on 0.4 ym
pore size polycarbonate membrane filters.
b. Secure the collected fibers to the filter as soon as possible
after collection by vapor depositing a 40 nm thick layer of
carbon on the filter.
c. Transfer the collected fibers to an electron microscope grid
by dissolving the filter in a Jaffe washer using chloroform
as a solvent.
d. Place the electron microscope grid in a transmission electron
microscope (TEM) and observe the' fibers at a magnification of
20,OOOX (screen magnification 16,OOOX). A lower magnification,
about 10,OOOX, is adequate for samples containing predominently
amphibole asbestos fibers or where the aim is to assess the
total mass of the asbestos fibers and the detection of very
small fibers is unimportant.
e. Identify and characterize the observed fibers by their mor-
phology and selected area electron diffraction pattern. Use
energy dispersive X-ray spectrescopy, if available, to aid
in the classification of fibers not classified by SAED. The
ED X-ray technique is particularly useful for characterizing
amphibole asbestos minerals that exhibit indistinguishable
electron diffraction patterns.
f. Enumerate the asbestos fibers using the field-of-view method
for medium and high fiber loading levels and the full-grid
opening method for low loading levels. If mass concentration
data are needed, measure the length and width of each fiber
and compute the mass from the fiber volume and density.
2. The identification of fibers based on morphology and electron dif-
fraction, as proposed by the optimized method, is adequate for clas-
sifying fibers into three categories: chrysotile, amphibole asbestos,
-------
and other minerals. This identification scheme provides the largest
amount of information for the analysis time; it is 2 to 3 times
faster than methods involving X-ray analysis. Thus, the proposed
methodology is cost effective.
3. The subprocedure of ashing, ultrasonification, and refiltration of
the collected material is not recommended despite the lack of
strong evidence to show the subprocedure is detrimental. Ashing
decreases the fiber lengths. This should result in an increase in
the number of fibers; but, the fiber number concentrations of ashed
and unashed samples were statistically equivalent in Phase 4
data. In the interlaboratory tests, however, ashing subprocedure
gave decreased fiber length and increased fiber number concentration,
as compared with the unashed sample. The data could not resolve
whether the apparent increase in fiber number concentration in the
ashed sample resulted from fiber breakage or from reduced inter-
ference in detecting and identifying small fibrils of chrysotile.
It is suggested that ashing be reserved for those instances where
the TEM grids prepared by the optimized method are unsuited for
analysis. These instances include those where fiber loadings are
high and dilution is necessary and where the presence of organic
matter obscures the observation of the fibers.
4. The presence of a few large bundles of fibers strongly influence
mass concentration estimates but has no significant effect on num-
ber concentration estimates. To circumvent the effect of bundles,
it is suggested that bundles greater than 1 ym^ be counted as single
entities. These bundles should be reported separately and, if mass
information is needed, a significant number of bundles should be
counted and sized.
5. The conventional transmission electron microscope is superior to
the scanning electron microscope for detecting and identifying
chrysotile fibrils. The superiority results from the higher re-
solution and the stationary image in the TEM.
6. Samples of airborne chrysotile prepared in the laboratory and an-
alyzed by several laboratories using the optimized procedure
showed good precision and accuracy. The ratio of the spread between
the 95% confidence limits to the mean value was about 0.48 for
chrysotile fiber number concentration and about 0.40 for chrysotile
mass concentration. The mass concentration computed from size and
density data compared favorably with the mass estimate obtained by
X-ray fluorescence.
7. Samples of air collected in a plant handling asbestos materials
give less precision and accuracy than the pure chrysotile sample
prepared in the laboratory. When ashing was used as a subprocedure,
the ratio of the spread between the 95% confidence limits to the
mean value was about 0.49 for the chrysotile fiber concentration
estimates and about 1.57 for the chrysotile mass concentration es-
timates. Without ashing, the corresponding values were 0.62 and
2.34. The mass estimates based on size and density were a
factor of 4.2 less than the mass obtained by X-ray fluorescence.
The lower precision estimate of the field sample is due to the
-------
presence of a few large fiber bundles and to the possible presence
of sources of magnesium other than chrysotile. These factors com-
bine to underestimate the chrysotile mass obtained by electron
microscope and overestimate the chrysotile mass obtained by X-ray
fluorescence.
-------
SECTION 3
RECOMMENDATIONS
The following recommendations are made to further the development and
acceptance of the optimized method for characterizing airborne asbestos by
electron microscope.
1. Sampling and collection methods were only briefly addressed in the
present study. Further study in five areas is recommended. First,
the deployment of polycarbonate filters in the field presents han-
dling problems. Field investigators prefer the cellulose ester
filters. While cellulose ester filter deposits can be transferred
to electron microscope grids in a Jaffe washer using acetone, the
microscopist prefers the clarity of the transferred deposit from
polycarbonate filters. Therefore, a technique for transferring the
deposit from cellulose ester filters to polycarbonate filters needs
development. Secondly, the effect of face velocity on the collec-
tion of fibers needs clarification. Personal samplers, with 1/5 the
face velocity of high-volume samplers, gave higher fiber count
estimates. Thirdly, a method for securing the deposit to the fil-
ter substrate at the collection site needs to be devised. Fourthly,
the rearrangement or loss of fibers on the filter due to handling
and transportation of the sample to the laboratory needs to be
determined. And, fifthly, the collection efficiencies of the poly-
carbonate and cellulose ester filters for small filters need
evaluation.
2. Computer programs for the quantitative identification of asbestos
minerals from energy-dispersive X-ray spectra need to be developed.
At present, the method is only semiquantitative. A quantitative
method should allow obtaining proportions of the various elements
or their oxides and comparing them with standard reference spectra
stored in a computer memory. A search technique should use reit-
erative techniques to narrow the choice among the possible refer-
ence spectra and the unknown by assigning probabilities to the de-
gree of match between the spectra. Some preliminary work along
these lines is reported by Millette and McFarren.[47]
3. The modifications suggested in this report for dealing with fiber
bundles need to be tested. New and improved methodology, such as
stratified analyses, needs to be investigated. These modifications
should improve the reliability of the fiber mass concentration data
computed from fiber size and density estimates.
4. Further work is needed to determine the effect of the total area
-------
scanned, i.e., the number of fibers counted, on the reliability
of the fiber concentration estimate.
5. The present study produced inconclusive results for a few sub-
procedures. More definitive studies are suggested to clarify the
importance of these subprocedures. The effects of low temperature
ashing, diluent and dispersant selection, and ultrasonic treatment
should be isolated.
6. Intralaboratory reproducibility should be determined. Duplicate
samples should be exposed at different times to the entire sequence
of processing steps from collection to TEM examination.
7. Additional round-robin tests are suggested to obtain a complete
picture of interlaboratory variability.
8. Techniques, other than electron microscopy, should be sought to
assess the mass of asbestos minerals present in ambient air
samples.
-------
SECTION 4
SCOPE OF THE WORK
INTRODUCTION
Although asbestos fibers are a definite health hazard [1-27], the effects
of low-dosage, chronic inhalation exposures from natural and occupational en-
vironments have not been defined [23-27]. It is believed that fiber charac-
teristics and size distribution are important parameters in addition to the
amount of asbestos in the inhaled air [21,22]. There are several methods
available for determining the mass of asbestos [27-39]. X-ray diffraction
[31-35], X-ray fluorescence, differential thermal analysis [36], infrared spec-
troscopy [37-39], neutron activation analysis [28], and atomic absorption can-
not distinguish fibrous from non-fibrous minerals and cannot give fiber size
distribution data. Optical and electron microscope methods allow the size dis-
tribution data to be obtained. The main limitations of the optical microscope
methods are the inability to detect fibers smaller than about 0.2 ym diameter.
Electron optical instruments allow much better resolution and facilitate recog-
nition of asbestos from non-asbestos minerals from studying morphological fea-
tures [15-17,30,40-43]. Further authentic identification is possible using
electron diffraction data obtained with a transmission electron microscope
[44-48] and by using X-ray emission spectroscopy data in electron-probe instru-
ments or scanning electron microscopes [35,42-44,47-49]. New analytical elec-
tron microscopes are now available which allow all three types of data (mor-
phological, electron diffraction, and X-ray spectroscopy) on individual fibers
for unique and fullest characterization of particles [47-49]. However, because
of the rapid acceptance and utilization of electron optical instruments, there
is no standard methodology available. Several laboratories perform the analysis
of airborne asbestos fibers and while they claim a reasonable internal consis-
tency, the results obtained by separate laboratories are often widely different.
The purpose of this study was to evaluate the methods and subprocedures cur-
rently in use in different laboratories, and to select and develop a composite
-------
procedure which will minimize the variability of results. The second objective
of the program was to prepare a detailed handbook describing the optimized
method without ambiguity. The third objective was to test this optimized
method by several independent laboratories in a round-robin test and evaluate
the results and to further improve it.
STRATEGY
A strategy to five tasks was planned for evaluating and optimizing a
large number of important procedures, subprocedures, or variables.
Task 1; Literature Search and Survey of EM Procedures
Published Work and Experience of Analytical Laboratories—
A reference library containing over a thousand articles on asbestos-related
topics was compiled. Several prominent investigators were contacted and asked
to supply details of their methods for estimating asbestos. Selected labora-
tories were visited to study first-hand the different aspects of specimen col-
lection, specimen preparation and examination, and evaluation of the data. The
laboratories visited were:
(a) Dow Chemical Laboratory, Midland, MI; Dr. Don Beamon
(b) Battelle Memorial Institute, Columbus, OH; Dr. Heffelfinger
(c) Atomic Energy Research Establishment, Harwell, England; Dr. A. E.
Morgan
(d) Pneumoconiosis Research Institute, Penarth, Wales; Dr. V. Timbrell
(e) Franklin Research Institute, Philadelphia, PA; Dr. A. Pattnaik
(f) Naval Research Laboratory, Washington, DC; Dr. L. S. Birks
(g) Mount Sinai Laboratory, New York, NY; Dr. Arthur Langer
(h) McCrone Associates, Chicago, IL; Dr. I. Stewart
(i) National Institute for Occupational Safety and Health, Cincinnati,
OH; Dr. Ralph Zumwalde
List of Possible Variables—
The electron microscope method involves several steps such as sample col-
lection, sample preparation, sample examination, and interpretation of results.
A generalized scheme for quantitative characterization of asbestos is illustrated
in Figure 1. Table 1 lists the various steps and subprocedures which may be
-------
Sample
Air
Prepare Filter
Sample
Ashing of Filter
Prepare TEM
Grid
EM Examination
of Grid
Fiber Identification
and Count
Evaluate, Interpret,
and Estimate
Concentration
Dilution and
Redeposition
on Filter
Figure 1. Flow chart of electron microscope procedures for estimating size
distribution and concentration of airborne asbestos.
10
-------
Table 1
PROCEDURAL VARIABLES OF ELECTRON MICROSCOPY METHOD
Variable
Variable Label
Levels
3
4
5
6
10
11
12
13
Sampling Variables
Asbestos Source Type and Time
Distance from Source
Sampler Type
Volume Sampled
Filter Type
Pore Size
Locking of Particulates to
Filter
Filter Examination Variables
Location on Filter
Magnification, nominal
Measurement Method
Examiner
Ashing Variables
Portion of Original Filter
Used for Ashing
Ignition Before Ashing
(1) raw fiber
(2) cement industry
(3) plastics industry
(1) point
(2) near point
(3) ambient
(1) personal
(2) hi-vol
(1) small
(2) medium
(3) large
(1) cellulose acetate
(2) polycarbonate
(1) 0.2 ym
(2) 0.4 ym
(3) 0.8 ym
(1) none
(2) carbon coating
(3) gelatinizing
(1) center
(2) mid-radius
(3) periphery
(1) 5,OOOX
(2) 10,OOOX
(3) 20,OOOX
(1) ruler
(2) eyeball using micrograph
(3) eyeball using fluorescent screen
(1) #1
(2) #2
(3) #3
(1) pieshape #1
(2) pieshape #2
(3) pieshape #3
(1) none
(2) yes
11
-------
Table 1 (continued)
Variable
Variable Label
Levels
14 Ashing
15 Duration of Ashing
Suspension and Redeposition
16 Dilution Medium
17 Bonification
18 Duration of Sonification
19 Type of Redeposition Filter
Grid Preparation Variables
20 3 mm Sample Location
21 Deposition Method
22 Type of EM Grid
23
24
25
Mesh Size of the EM Grid
Filter Side During Washing
Grid Examination Variables
Fiber Identification Method
26
Grid Opening
(1) low temperature
(2) high temperature
(1) short (2 hrs)
(2) long (24 hrs)
(1) Toluene
(2) Aerosol O.T. 0.1%
(3) Aerosol O.T. 0.2%
(1) none
(2) low energy
(3) high energy
(1) short
(2) medium
(3) long
(1) cellulose acetate
(2) polycarbonate
(1) center
(2) mid-radius
(3) periphery
(1) cold finger soxhlet extraction
(short duration)
(2) Soxhlet extraction (long duration)
(3) Jaffe-Method
(1) copper
(2) nickel
(1) 200-mesh
(2) 400-mesh
(1) particle side up
(2) particle side down
(1) TEM-Morphology
(2) TEM-Morphology plus diffraction
(3) TEM-Morphology plus chemistry
(4) TEM-Morphology plus diffraction
plus chemistry
(5) SEM-Morphology
(6) SEM-Morphology plus chemistry
(1) center
(2) mid-radius
(3) periphery
12
-------
Table 1 (continued)
Variable
Variable Label
27 Field Selection
28 Magnification
29 Operator
30 Experience of Operator
Levels
(1) random
(2) consecutive
(3) full grid opening
(1) 5,OOOX
(2) 10,OOOX
(3) 20,OOOX
(1) #1
(2) #2
(3) #3
(1) short
(2) long
13
-------
important. The list is by no means complete. One can extend it further.
It is clear that a large number of steps are involved in following any of
the several possible paths. At each step, a multitude of choices is possible.
There is no apriori way of choosing a particular level or a step on a rational
basis. Therefore, one must evaluate these and provide a rational basis for
selection of a step and selection of the proper level in each variable.
Task 2: Selection of Procedures or Subprocedure Variables and Experimental
Plan
The number of possible variables listed in Table 1 is large and evaluating
all of them would mean spreading the experimental effort too thinly over too
large an area. To avoid this and to achieve meaningful estimates, the choice
was narrowed down to 19 as shown in Table 2. In view of this large number of
variables, it was necessary to adopt a multiphase approach, each phase util-
izing a statistically designed experimental plan. This approach yields a maxi-
mum of information with a high degree of statistical significance for a given
experimental effort.
A highly fractionated factorial design was used. Independent variables
were controlled simultaneously according to a predetermined experimental scheme.
Each of the tests (or set of data) represented a unique combination of several
independent variables.
Task 3; Statistical Evaluation of the Variables
The results obtained from Task 2 were evaluated using statistical methods.
The data allowed several dependent variables to be studied and explained their
relationship with the independent variables.
Task 4; Development of an Optimal Procedure
The evaluation from Task 3 was to lead to selecting those independent
variables and their level which gave the least variability of results. One
could formulate a combination of these variables into a composite procedure.
These subprocedural steps are described in detail in the form of a manual.
Task 5: Statistical Evaluation of the Optimal Method
The performance of this optimized method was evaluated in a round-robin
test on the same air samples by independent laboratories.
14
-------
Table 2
INDEPENDENT VARIABLES SELECTED FOR EVALUATION IN THIS STUDY
Variable
Composition of Sample
X~ Concentration of Sample
Xo Sampling Instrument
X^ Filter Type
X5 Pore Size, nominal
Levels
Xy
X15
Filter Orientation
(Particle Side)
2.3 mm Portion
Location
Use of Carbon Coating
Transfer Method
Magnification, nominal
Grid Opening
Choice of Fields
Identification
Ashing
Sonification
(1) #1 100% Chrysotile
(2) #2 60% Chrysotile + 40% Amphibole
(3) #3 70% Chrysotile + 20% Amphibole + 10% Non-
Asbestos Fiber
(1) Low (2) Medium (3) High
(1) Hi-Vol
(2) Personal
(1) Nuclepore (polycarbonate)
(2) Millipore (cellulose acetate)
(1) 0.2 ym
(2) 0.4 urn
(3) 0.8 ym
(1) Down (2) Up
(1) Periphery
(2) Mid-radius
(3) Center
(1) Yes
(2) No
(1) Soxhlet Extraction 1 (short)
(2) Soxhlet Extraction 2 (long)
(3) Jaffe Method
(1) 5,OOOX
(2) 10,OOOX
(3) 20,OOOX
(1) Periphery
(2) Mid-radius
(3) Center
(1) Random
(2) Consecutive
(3) Full Grid Opening
(1) Morphology plus chemistry
(2) Morphology plus diffraction
(3) Morphology plus chemistry plus diffraction
(4) Morphology alone
(1) High Temperature
(2) Low Temperature
(1) Low Energy
(2) Medium Energy
(3) High Energy
15
-------
Table 2 (continued)
Variable
X_- Redeposition Filter
16
%• ^ 2.3 mm Portion of
Redeposition Filter
X Instrument
Ignition Before Ashing
Levels
(1) Millipore (cellulose acetate)
(2) Nuclepore (polycarbonate)
(1) Periphery
(2) Mid-radius
(3) Center
(1) JSM 50A
(2) JEOL 100C
(1) No
(2) Yes
16
-------
SECTION 5
STATISTACALLY DESIGNED EXPERIMENTAL PLANS
INTRODUCTION
There are different ways of investigating systems having many independent
variables.
Two Extreme Approaches
One approach consists of varying only one variable at a time, holding all
others fixed. This is the simplest and most unambiguous way of evaluating each
of the variables by itself. However, this approach does not provide informa-
tion on a variable when other variables are also changed simultaneously. At
the other extreme, to obtain all passible combinations of all independent vari-
ables and their levels may require an astronomically large experimental effort.
For the problem being studied, in which there are 19 independent variables with
19
three possible levels per variable, the number of tests would be 3 , which is
an astronomically large number, and it would clearly be impractical to proceed
with this complete factorial approach.
Fractional Factorial Designs
This approach allows evaluation of a large number of independent variables
with a reasonable experimental effort. It is the best approach for screening
the most important variables from other less important ones. These important
variables can then be studied in greater detail in subsequent small experiments.
A good discussion of the statistical design of experiments may be found in
References 50-54.
In this project, fractional factorial experiment designs of a special class
were utilized for the investigation and optimization of procedural variables
in the electron microscope examination of airborne asbestos. These designs are
characterized as having three levels per factor (which can be reduced to two
where desired) and being highly efficient in the sense that the number of test
combinations is small compared with the number of effects that can be estimated.
17
-------
The use of properly coded values (linear and quadratic) permits orthogonal esti-
mates of the effects of the independent variables to be computed by multiple
regression analysis. These three-level compact orthogonal designs were con-
structed by extension of the method for constructing similar two-level designs
described by Youden [51]. A detailed discussion of the Phase 1 design, as an
illustration of the general class, is given in Appendix A.
From the experimental data on each sample, characterizing quantities such
as fiber number concentration, fiber mass concentration, mean fiber length,
etc., were computed. These measured or estimated quantities are designated as
the dependent variables. Some dependent variables were subjected to suitable
mathematical transformations far the usual purposes of linearization of the
effects of the independent variables, variance stabilization, and normalization
of the distribution of residuals. The transformations applied include the
logarithmic, the square root, and the arcsine square root [55-57]. Table 3
lists a few representative dependent variables considered in Phase 1.
A stepwide least-square multiple regression method was used to construct
the equation relating each chosen dependent variable to the independent vari-
ables [58-60]. A detailed discussion of the procedure and resulting equations
is given in Appendix B.
FIVE PHASE PROGRAM
We decided to study the selected 19 variables in five phases, each phase
using a fractional factorial design.
Phase 1
In this phase, we examined the procedures employed when no ashing and
resuspension were undertaken and fibers were identified by morphology alone.
Twelve procedural variables (see Table 4) were studied in a plan of 27 tests.
The 12 variables comprise five variables of sample collection (X.-X,.), four
variables of sample preparation (X,-X ), and three variables of sample examina-
tion (X-jQ-X-io) *n transraission electron microscope. The compact notation of
the scheme is shown in Table 5. The 12 independent variables are denoted by
^1~^12" T^e numkers *n each row refer to the variable's value (or level code)
for the particular combination. Table 6 shows the same scheme in long notation
for easy recognition of variable level combinations.
18
-------
Table 3
REPRESENTATIVE DEPENDENT VARIABLES IN PHASE 1
Variable Definition
Y Mean Ln (fiber width, ym)
Y_ Mean Ln (fiber length, ym)
Yj. Mean Ln (aspect ratio)
3
Y Mean Ln (fiber volume, (ym) )
3
YQ Square root (estimated number of fibers per cm of air sampled)
3
Y - Ln (estimated mass concentration of fibers in the air, yg/m )
2
Y . Square root (estimated number of chrysotile fibers per cm of
filter)
-9
Y „ Ln (estimated mass concentration of chrysotile fibers, 10 gm per
12 cm2 of filter)
19
-------
Table 4
INDEPENDENT VARIABLES, PHASE 1
Variable
INDEPENDENT VARIABLES OF FILTER LOADING
X-j Composition of Sample in
Aerosol Chamber
XT Concentration of Sample
on Filter
Xg Sampling Instrument
X4 Filter Type
X5 Pore Size, nominal
Levels and Codes
(1) 100% Chrysotile
(2) 60% Chrysotile
+ 40% Amphibole
(3) 70% Chrysotile
-i- 20% Amphibole
+ 10% Non-Asbestos Fiber
(1) Light
(2) Medium
(3) Heavy
(1) High Volume
(2) Personal
(1) Nuclepore
(2) Millipore
(1) 0.2 ym
(2) 0.4 ym
(3) 0.8 ym
XiL=-l XjQ= 1
xa= 1 X,Q= 1
XiL= 0 XiQ=-2
X2L=-1 X2Q= 1
X2L= 0 X2Q=-2
X2L= 1 X2Q= 1
X3Q=-2
X3Q= 1
X,,q=-2
X.,0- 1
X5L=-1 X5Q= 1
X5L= 0 X5Q=-2
X5L= 1 X5Q= 1
INDEPENDENT VARIABLES OF TEM GRID PREPARATION
)<6 Filter Side
Xy 2.3 mm Portion Location
Xg Use of Carbon Coating
Xg Transfer Method
INDEPENDENT VARIABLES OF TEM EXAMINATION
XIQ Magnification, nominal*
Xj} Grid Opening Location
X-J2 Choice of Fields
(1
(2
(1
(2
(3
Particle side down
Particle side up
Periphery
Mid-radius
Center
(1) Yes
(2) No
(1) Soxhlet Extraction 1 (short)
(2) Soxhlet Extraction 2 (long)
(3) Jaffe Method
(1) 5.000X (screen mag. 4.000X)
(2) 10.000X (screen mag. 8.000X)
(3) 20.000X (screen mag. 16.000X)
(1) Periphery
(2) Mid-radius
(3) Center
(1) Random choice of small fields
(2) Small fields, consecutive
(3) Entire grid opening as a field
X6Q= 1
X6Q=-2
X7L=-1 X7Q= 1
X7L= 0 X7Q=-2
X7L= 1 X7Q= 1
X8Q=-2
X8Q= 1
X9L=-1 X9Q= 1
X9L= 1 X9Q= 1
X9L= 0 X9Q=-2
X10L=-1 X10Q= 1
XI - n y n— 9
1 o L— U Jk\ o y~ -C.
X10L= 1 X10Q= 1
XnL=-l XnQ= 1
XuL= 0 XnQ=-2
XuL= 1 XnQ= 1
X12L=-1 X12Q= 1
Xi2L= 1 X12Q= 1
XJ2L= 0 X12Q=-2
*The actual magnification at the fluorescent screen is somewhat smaller than the nominal or
camera magnification, depending upon the design geometry of each transmission electron
microscope.
20
-------
Table 5
PHASE 1 EXPERIMENT DESIGN (COMPACT NOTATION)
Factor (Independent Variable)
Combination
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
4
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
X
^,
1
1
1
2
2
2
3
3
3
1
1
1
2
2
2
3
3
3
1
1
1
2
2
2
2
3
3
^3
2
2
2
1
1
1
2
2
2
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
1
1
1
4
2
2
2
1
1
1
2
2
2
2
2
2
2
2
2
1
1
1
1
1
1
2
2
2
2
2
2
-5
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
4
1
1
2
1
1
2
1
1
2
1
2
1
1
2
1
1
2
1
2
1
1
2
1
1
2
1
1
'^7
3
1
2
1
2
3
2
3
1
3
1
2
1
2
3
2
3
1
3
1
2
1
2
3
2
3
1
X8
O
2
2
1
2
1
2
1
2
2
2
1
2
1
2
2
2
2
1
1
2
2
2
2
1
2
1
2
Xg
1
2
3
3
1
2
2
3
1
2
3
1
1
2
3
3
1
2
3
1
2
2
3
1
1
2
3
X10
2
1
3
2
1
3
2
1
3
3
2
1
3
2
1
3
2
1
1
3
2
1
3
2
1
3
2
%
2
1
3
3
2
1
1
3
2
2
1
3
3
2
1
1
3
2
2
1
3
3
2
1
1
3
2
-12
1
3
2
2
1
3
3
2
1
2
1
3
3
2
1
1
3
2
3
2
1
1
3
2
2
1
3
21
-------
Table 6
PHASE 1 EXPERIMENTAL PLAN (LONG NOTATION)
Samp 1 e
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Independent Variables
Xl
Compo.
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
X2
Loading
Low
Low
Low
Med
Med
Med
High
High
High
Low
Low
Low
Med
Med
Med
High
High
High
Low
Low
Low
Med
Med
Med
High
High
High
X3
Sampler
P
P
P
HV
HV
HV
P
P
P
HV
HV
HV
P
P
P
P
P
P
P
P
P
P
P
P
HV
HV
HV
X4
NP/MP
M
M
M
N
N
N
M
M
M
M
M
M
M
M
M
N
N
N
N
N
N
M
M
M
M
M
M
X5
Pore
Size
0.22
0.45
0.8
0.2
0.4
0.8
0.22
0.45
0.8
0.22
0.45
0.8
0.22
0.45
0.8
0.2
0.4
0.8
0.2
0.4
0.8
0.22
0.45
0.8
0.22
0.45
0.8
X6
Particle
Side
Down/Up
—
Up
_„
-_
Up
__
—
Up
„_
Up
_-
Up
--
-_
Up
Up
--
--
Up
--
Up
—
X7
3 mm
Location
Ctr
Peri
MR
Peri
MR
Ctr
MR
Ctr
Peri
Ctr
Peri
MR
Peri
MR
Ctr
MR
"Ctr
Peri
Ctr
Peri
MR
Peri
MR
Ctr
MR
Ctr
Peri
X8
C-
Coat
—
Yes
Yes
--
Yes
—
—
Yes
--
Yes
--
—
—
Yes
Yes
--
--
-_
--
Yes
-_
Yes
_ _
X9
Filter
Remova 1
Sox 1
Sox 2
J
J
Sox 1
Sox 2
Sox 2
J
Sox 1
Sox 2
J
Sox 1
Sox 1
Sox 2
J
J
Sox 1
Sox 2
J
Sox 1
Sox 2
Sox 2
J
Sox 1
Sox 1
Sox 2
J
X10
Magnif i .
1.000X
10
5
20
10
5
20
10
5
20
20
10
5
20
10
5
20
10
5
5
20
10
5
20
10
5
20
10
Xll
Grid
Opening
Loc.
MR
Peri
Ctr
Ctr
MR
Peri
Peri
Ctr
MR
MR
Peri
Ctr
Ctr
MR
Peri
Peri
Ctr
MR
MR
Peri
Ctr
Ctr
MR
Peri
Peri
Ctr
MR
X12
Choice
of Field
Random
Full Grid •
Consecutive
Consecutive
Random
Full Grid
Full Grid
Consecutive
Random
Consecut ive
Random
Full Grid
Full Grid
Consecutive
Random
Random
Full Grid
Consecut ive
Full Grid
Consecut ive
Random
Random
Ful I Grid
Consecut ive
Consecut ive
Random
Full Grid
Footnotes:
P = Personal
HV = High-volume
Ctr = Center
Peri = Periphery
MR = Mid-radius
Sox 1 - Short C-Coat = Carbon Coat
Sox 2 = Long M - Millipore/Cellulose Acetate
J = Jaffe N = Nuclepore/Polycarbonate
-------
Phase 2
In phase 2, three variables, X^, Xg, and X , were examined in nine tests.
The compact notation of the scheme is shown in Table 7. The independent vari-
ables, their levels, and coding are also listed in Table 7. The coding is
necessary to balance the design and facilitate the regression analysis. Further
explanation may be found in Appendix A.
Phase 3
This phase was intended for comparing two instruments used in the secondary
electron mode. The JEOL JSM 50A, an excellent scanning electron microscope,
and JEOL 100C, the latest scanning-transmission electron microscope, were eval-
uated in SEM mode using either morphology alone or in conjunction with elemental
analysis by X-ray probe. Phase 3 also used nine tests. The scheme is shown in
Table 8.
Phase 4
Phase 4 provided data for examining effects of ashing and redeposition pro-
cedures. We examined four variables, X-,, X., ~, X . , and X , in nine tests.
The compact designation of the scheme is illustrated in Table 9. The variable
names and different levels in each and their coding are also listed in Table 9.
Phase 5
In phase 5, we examined the variables of the direct drop method being fol-
lowed at a few laboratories. The method uses a liquid suspension (either a
water sample or resuspended ash of an air filter). Instead of redepositing the
particles onto another filter (see phase 4), a microdrop is withdrawn and de-
posited directly on a carbon-coated TEM grid and allowed to dry. Two common
ways of drying the drop are keeping the drop-side facing up or keeping the
drop-side facing down. There have been conflicing reports about the uniformity
of deposit on the 3 mm diameter grid. Hence, another variable evaluated in
this phase was the grid opening location on TEM grid (X,-), i.e., peripheral
(level 1), mid-radious (level 2), and central locations (level 3). The scheme
for phase 5 experiment and the orthogonal coding are given in Table 10.
23
-------
Table 7
PHASE 2 EXPERIMENT DESIGN
Sample
Description
Code*
2113
2121
2132
2211
2222
2233
2312
2323
2331
Factors
^
1
1
1
2
2
2
3
3
3
(Independent
J^9-
1
2
3
1
2
3
1
2
3
Variables)
X13-
3
1
2
1
2
3
2
3
1
Independent
Variable
Code
Description.
of Variables
Composition of Sample
Transfer Method
Identification Method
Levels
(1) Pure chrysotile
(2) Chrysotile plus amosite
(3) Chrysotile plus
crocidolite plus
wollastonite
(1) Soxhlet 1
(2) Soxhlet 1 with carbon
coating
(3) Jaffe
(1) Morphology plus
chemistry
(2) Morphology plus
electron diffraction
(3) Morphology plus
chemistry plus
electron diffraction
Codes
XiL=-l
XiL= 0 XiQ=-2
XiL= 1 XiQ= 1
X9L=-1 X9Q= 1
X9L= 0 X9Q=-2
X9L= 1 XsQ= 1
Xi3L=-l X13Q= 1
X13L= 0 X13Q=-2
Xi3L
Xi3Q
*First digit in the sample description code shows the phase number; second, third,
and fourth digits refer to the levels of independent variable used.
24
-------
Table 8
PHASE 3 EXPERIMENT DESIGN
Test
1
2
3
4
5
6
7
8
9
Sample
Description
Code*
3142
3111
3142
3212
3242
3241
3341
3342
3312
Factors
Xl-
1
1
1
2
2
2
3
3
3
(Independent
X
Ljr-
4
1
4
1
4
4
4
4
1
Variables)
^1 8-
iCT^
2
1
2
2
2
1
1
2
2
Independent
Variable
Code
1
Description of Variables
Composition of Sample
Identification Method
X1_ Analytical Electron Microscope
1 o
Level
(1) Pure chrysotile
(2) Chrysotile plus amosite
(3) Chrysotile plus
crocidolite plus
wollastonite
(1) Morphology plus X-ray
analysis
(4) Morphology
(1) JSM-50A
(2) JEOL 100C
*First digit in the sample description code refers to the phase number; the
second, third, and fourth digits refer to the levels of the independent vari-
ables used.
25
-------
Table 9
PHASE 4 EXPERIMENT DESIGN
Sample
Number
1201
1202
1203
1204
1205
1206
1207
1208
1209
Sample
Description
Code*
42211
41112
42213
41221
42222
42123
42121
42222
41223
— t-6-
2
1
2
1
2
2
2
2
1
Factors (Independent
— t9-
2
1
2
2
2
1
1
2
2
Variables)
X14-
1
1
1
2
2
2
2
2
2
JL^
1
2
3
1
2
3
1
2
3
Independent
Variable
Code
X16
19
X
14
X
15
Description of Variables
Filter Type
(both primary and secondary)
Ignition
Ashing
Bonification
Levels
(1) Millipore
(2) Nuclepore
(1) No
(2) Yes
(1) High temperature
(2) Low temperature
(1) Low energy
(2) Medium energy
(3) High energy
Codes
X16Q=-2
Xi6Q= 1
Q=-2
Xi9Q= 1
XmQ=-2
X1SL=-1 X15Q= 1
XisL= 0 Xi5Q=-2
XX5L= 1 XisQ= 1
*First digit of the sample description code refers to the phase number; second,
third, fourth, and fifth digits refer to the levels of the independent variables
used.
26
-------
Table 10
PHASE 5 EXPERIMENT DESIGN
Sample
Number
5101
5102
5103
5104
5105
5106
Sample
Description
Code*
523
522
521
513
512
511
Factors (Independent
X,
2
2
2
1
1
1
Variables)
xn
i±
3
2
1
3
2
1
Independent
Variable
Code
X6
o
X11
J.JL
Description
of Variables
Orientation of drop
during drying on
TEM grid
Radial Location of
Opening on TEM grid
Levels
(1) Drop side up
(2) Drop side down
(1) Center
(2) Mid-radius
(3) Periphery
Codes
X6= 1
X6— 1
V T T V ^_> 1
AH L=— 1 AH (}— 1
XnL= 0 XnQ=-2
XnL= 1 Xn Q= 1
*First digit of the sample description code refers to the phase number;
second and third digits refer to the levels of the independent variables
used.
27
-------
SECTION 6
EXPERIMENTAL WORK
THE PREPARATION OF LABORATORY FILTERS OF CONTROLLED ASBESTOS LOADING IN PHASE 1
Introduction
Phase 1 of the sratistically designed study to evaluate the electron micro-
scope analytical methodology for determining asbestos required that filters be
prepared under controlled conditions to obtain three different asbestos concen-
trations. Both polycarbonate (Nuclepore) and cellulose acetate (Millipore)
filters, with pore sizes of 0.2, 0.4, and 0.8 ym, were used and samples were
collected using both high volume samplers [61] (with 20 cm x 25 cm filters) and
personal samplers [62] (with 3.5 cm diameter filters).
Filters could be prepared in several ways, preferably by simultaneous sam-
pling using different filter types, pore sizes, and samplers. Filters could be
prepared by:
• taking samples close to a natural source
• preparing solutions of known asbestos concentration by ultrasonic
treatment of water and filtering from liquid suspension
• preparing asbestos aerosols and sampling from an aerosol cloud of
calculated concentration
Sampling from a natural asbestos source, for example, an asbestos products
factory, would be the most convenient, but unfortunately, it has the serious
disadvantage that the concentration of the source is not known.
Filter samples could be prepared from liquid suspension of known concen-
tration of asbestos minerals. However, the disadvantage with this method is
that the deposition of fibers from water suspension onto a filter may not be
equivalent to that obtained from an aerosol cloud.
Simultaneous sampling from an aerosol cloud of known concentration appears
the best since it simulates normal operating conditions while allowing some
control of the aerosol concentration.
28
-------
Experimental
The Aerosol Chamber—
A spherical chamber fabricated from welded steel plate with a diameter
of 5.5 m was utilized to obtain an aerosol cloud. The volume of the chamber is
3
86 m . The inside of the chamber is coated with an epoxy-phenolic material
(Plasite 7122) to prevent corrosion. The chamber can be cleaned by a hot
water spray to wash down the walls, and by a high volume extraction system to
purge the chamber through an absolute filter device at the rate of 12 air
changes per hour.
Inside the chamber there is a catwalk as shown in Figure 2a. Three high-
volume samplers and six personal samplers were mounted on the catwalk. The
aerosol cloud entered the chamber from the generator located outside the
chamber. A fan inside the chamber circulated the air to ensure a uniformly
mixed aerosol.
Ultrasonic Treatment to Break Fibers to a Sufficiently Fine Size—
The UICC asbestos minerals have a very coarse particle size, which is
unsuitable for charging in an aerosol cloud. Three ultrasonic devices were
tested to determine their efficiency in breaking up asbestos into fibers under
10 ym in length. They were:
• Ultra-Sonic Industries - System Forty
80 Watts Bath Type
• Polytron Cell Disruptor - PT10
5000 Watts with High Speed Agitator
• Branson Sonifier - W 185C
100 Watts Horn Type
Tests were conducted by weighing out a small quantity of asbestos and
suspending it in distilled water to give an asbestos concentration of about
0.3% by weight. Aerosol OT was added as a dispersing aid at a concentration
of about 0.2%. Ultrasonics were applied,for time periods of 5, 10, 20, and
30 minutes. Using each device, the sample was then diluted to a concentration
of 0.03% with distilled water.
The Branson Sonifier was the only unit found suitable for achieving small
enough fiber lengths in chrysotile asbestos. By varying the time of the ultra-
sonic treatment, the chrysotile asbestos could be reduced to any fiber length
desired. The most satisfactory chrysotile dispersion was produced by giving a
29
-------
Aerosol
Inlet
Access
Flanges
Control
Panel
Air Circulatin
Fan
High-volume
Samplers
Structural
Support
Purge Air Out
Figure 2a: Top view and side view of aerosol chamber
showing location of apparatus.
Test
Aerosol
Particle Charge Neutralizer
Air at
45 psig *
Dryer
Valve
Filter
Pressure
Regulator
HXH
Dilution
Flowmeter
Atomizer
Flowmeter
Atomizer
Impactor
Figure 2b. Flow diagram of aerosol generator
30
-------
45 minute treatment at 100 watts power to 250 mg of asbestos suspended in 150 ml
of water with 2% of Aerosol OT added as a dispersing agent. The results of the
treatment were checked by optical and electron microscopes.
The Branson unit was found to be less effective with amosite asbestos and
fiber glass. A series of hand-grinding experiments were performed using an
agate pestle and mortar. A techniques was developed which les to satisfactory
dispersion of both amosite and fiber glass. It consisted of wet hand grinding
a 100 mg quantity of fiber in a few drops of 1:1 solution of water and Aerosol OT
for 30 minutes.
Aerosol Generation—
The Sierra Instrument Company's Model 133G Fluid Atomization Aerosol Gener-
ator utilizes air-blast atomization and inertial impaction to produce aerosols.
9
It could produce particles at rates of up to 10 particles per second. The
droplet size was variable from 0.03 to 3 ym.
The generator is schematically illustrated in Figure 2b. It consisted of
a dryer, a pressure regulator, an absolute filter, an adjustable valve, two pre-
cision flowmeters, a fluid atomizer, an impactor, and an ionizer.
High pressure air was supplied to the generator at a minimum pressure of
45 psig. The air passed through a chemical dryer and a pressure regulator
which reduced the pressure to 35 psig. The air then flowed through an absolute
filter and was subsequently divided into two fractions: the atomizer air and
the dilution air.
The atomizer air flowed through a flowmeter and a Collison-type atomizer.
As the air passed through the nozzles of the atomizer, it produced a spray of
the suspension directed against a baffle. The spray was then carried by the
air through an impactor where the large droplets were removed, leaving an
aerosol of a narrow size distribution. The remaining droplets then flowed to
a mixing tee located upstream of the ionizer.
After flowing through the filter, the dilution air flowed through a manually
adjusted valve. It then passed through a flowmeter and into the mixing tee.
From the mixing tee, the diluted aerosol flowed into the ionizer where it was
mixed with bipolar ions and the solvent evaporated. The aerosol was then
exhausted through the outlet located on the side of the generator housing.
Care was taken to adjust the fiber concentration to a point where each droplet
31
-------
formed would contain 0 or 1 fiber the vast majority of the times. This precau-
tion is required to minimize agglomeration or clumping of the fibers as the
water evaporated. The ionizer employed a radioactive source (1 milli-curie of
Krypton 85 gas) to neutralize static charge on the particles.
During preliminary runs, contamination of the aerosol chamber by the high-
volume samplers was observed. The requirement that high-volume sampling time
be kept below a total of one hour, coupled with the failure of aerosol genera-
tors producing larger droplets to provide an adequately dispersed aerosol, led
to a modification to the aerosol generator. Provision was made to pump asbes-
tos slurry, whose concentration was adjusted to compensate for the fiber loss
and evaporative water loss, in the atomizer unit. The Sierra Atomizer was
operated for periods of 16 to 80 hours on a continuous basis using this make-up
system.
This method was used since the fiber sizes were small, enough to remain
in suspension indefinitely with minimal air recirculation. Trial and error
tests were necessary to control the desired concentration of fiber loading on
the filter. It was found that the Sierra Fluid Atomization Aerosol Generator
was not capable of delivering the airborne concentrations required in a few
hours.
Experiments to use other aerosol atomizers for obtaining higher concen-
trations in a short time proved futile because of unpredictable dispersion of
fibers. Such air samples would be unsuitable for good electron microscopy work.
The problem remained on how to obtain a reasonably high aerosol concentra-
tion in the chamber while at the same time ensuring the quality of the dispersion.
We settled upon a procedure that assumed a very low decay constant for the con-
centration of the aerosol in the chamber and involved operating the Sierra aero-
solizer for long periods (up to 95 hours) to build up a suitable concentration.
The assumption was deemed reasonable since the gravitational sedimentation of
the ultra-fine particles produced was negligible and the large diameter (18 foot)
chamber gave a low wall effect. Thus, all the aerosol dispersions were finally
made with one instrument, the Sierra Fluid Atomizer.
Details of Experimental Work in Phase 1 Samples Prepared in Chamber
In all, 27 samples were prepared as detailed in Table 6. Each sample was
unique and had to be individually prepared.
32
-------
-' o
The effective area of the 37 mm personal samplers was 6.7 cm versus
2
406.5 cm area of 20 cm x 25 cm filter in high-volume samplers. Also, the
difference in flow rates of the two devices was quite significant. The per-
-------
accounts for the distribution of sampling times seen in Table 11 for the high-
volume samplers.
SAMPLE PREPARATION FOR TEM
The sample preparations involved numerous subprocedures as described
below.
Carbon Coating
Filters needing carbon coating (independent variable Xg, level 1) were
placed in a vacuum evaporator and given a thin (40 nm thick) coating of carbon.
Cutting Out 2.3 mm Diameter Segment
Small discs were cut from each filter according to the variable X7
(level 1 - peripheral location, level 2 - mid-radius, and level 3 - central
location). A standard 2.3 mm punch was used for this purpose.
Particle Transfer Method
Three methods of filter dissolution were used. Variable Xg, level 1 is
the Soxhlet Extraction of short duration, level 2 is the Soxhlet Extraction of
long duration, and level 3 is the Jaffe method [63]. The solvent used depended
on the type of filter; acetone for cellulose acetate and chloroform for poly-
carbonate. Duration for filter dissolution also depended on type of filter.
Short and long durations were four hours and eight hours for the acetone extrac-
tion of cellulose acetate and eight hours to 16 hours for the chloroform extrac-
tion of polycarbonate filters. The Jaffe method used a 24 hour duration.
Filter Topside (Particle Side) Orientation
Variable X, had two possibilities. Levels 1 and 3 refer to particle side
down (i.e., during filter dissolution, the particles should be in direct con-
tact with the carbon-coating of the grid) and level 2 referred to keeping par-
ticle side up (not in direct contact with the carbon-coating of TEM grid).
EXAMINING SAMPLES IN TRANSMISSION ELECTRON MICROSCOPE
The study of samples in transmission electron microscope involves the
following important considerations.
Grid Opening Location
After positioning a grid in the transmission electron microscope, an area
of about 2 mm diameter was available for examination. The first step was to
34
-------
Table 11
EXPERIMENTAL SCHEME FOR SIMULATED AIR SAMPLES FOR PHASE 1
Ui
Run
No.
1
2
3
Sample
No,
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Fiber
Composition3
C
c
C
c
c
c
c
c
c
C J- A
C + A
C + A
C + A
C + A
C + A
C + A
C + A
C + A
C + A + F
C + A + F
C + A + F
C + A + F
C + A + F
C + A + F
C + A + F
C + A + F
C + A + F
Sampler
Typeb
P
P.
P
Hi~vol
Hi-vol
Hi-vol
P
P
P
Hi-vol
Hi-vol
Hi-vol
P
P
P
P
P
P
P
P
P
P
P
P
Hi-vol
Hi-vol
Hi-vol
Filter
Tvpec
M
M
M
N
N
N
M
M
M
M
M
M
M
M
M
N
N
N
N
N
N
M
M
M
M
M
M
Pore
Size, um
0.22
0.45
0.80
0.2
0.4
0.8
0.22
0.45
0.80
0.22
0.45
0.80
0.22
0.45
0.80
0.2
0.4
0.8
0.2
0.4
0.8
0.22
0.45
0.80
0.22
0.45
0.80
Air
Volume
Filtered, I
32
32
32
9116
9200
8400
512
512
512
2080
2030
2040
124
124
124
513
513
513
7.6
7.6
7.6
29.2
29.2
29.2
7920
7701
9167
Sampling
Time , min
16
16
16
14
13
11.2
256
256
256
5.25
4.5
- 3.0
62
62
62
258
258
258
3.8
3.8
3.8
14.6 -
14.6
14.6
20
17
13.5
Expected Mass Concentration
of Fibers on Filter, vig/cm^
Chrysotile
1.385
1.385
1.385
6.503
6.563
5.993
22.161
22.161
22.161
1.484
1.448
1.455
5.367
5.367
5.367
22.204
22.204
22.204
1.406
1.406
1.406
5.404
5.404
5.404
24.159
23.491
27.963
Amphibole
0.987
0.964
0.968
3.572
3.572
3.572
14.777
14.777
14.777
0.401
0.401
0.401
1.543
1.543
1.543
6.897
6.706
7.983
Fiberglass
0.201
0.201
0.201
0.771
0.771
0.771
3.448
3.353
3.992
Anticipated Mass Concentration
of Fibers in Chamber, Ug/m3
Chrysotile
290
Chrysotile Amphibole »-,.
290 193
Chrysotile Amphibole
1240 354
Fiberglass
177
C = chrysotile
C + A = chrysotile + amosite
C + A + F = chrysotile + amosite + fiberglass
P = Personal
M = Millipore
N = Nuclepore
-------
reduce the> magnification to a minimum (which was about 90X on the JEOL 100C
transmission microscope) and to select a grid opening from central, mid-radius,
or peripheral location as called for by the scheme (see Table 6).
Magnificat ion
The required grid opening was brought to the center of the screen and the
electron microscope was adjusted to 5,000, 10,000, or 20,000 nominal magnifi-
cation, again as required by the scheme (see Table 6).
Choice of Fields
In variable X , the field of view could be chosen as a rectangular area
(of the tiltable section) of the fluorescent screen and these fields could be
either selected in a random fashion (level 1) or in a consecutive (adjacent)
fashion (level 2). Alternately, the entire grid opening could be treated as
one single field of view (level 3). For scanning the entire grid, the area was
scanned from the top corner sideways until the grid bar was reached. The
field was displaced upwards slightly and again scanned sideways until the op-i
posite boundary of the grid opening was reached. This was repeated until the
entire grid opening area was scanned. The levels for this variable were ex-
plained in the overall scheme (see Table 6).
DATA RECORDING
f
All particles with an aspect ratio 3:1 and greater and having substantially
parallel sides were considered as fibers. The length and width of each fiber
were estimated in mm by visual comparison with graduated circles on the fluores-
cent screen and all fibers visible in each field of view were counted and se-
quentially numbered. No attempt was made to recognize each type of mineral
fibers; only a computed average density was assumed for all fibers in each
sample. For fibers extending beyond the perimeter of the field of view, the
length within the field of view was estimated and the fiber was treated as a
half fiber for fiber concentration estimation.
EXPERIMENTAL WORK IN PHASE 2: STUDY OF FIBER IDENTIFICATION METHOD
The main objective of this phase was to evaluate the three methods of
fiber identification; (a) morphology in conjunction with X-ray analysis,
(b) morphology in conjunction with electron diffraction, and (c) morphology in
conjunction with both electron diffraction and X-ray analysis.
36
-------
Filter Preparation
Polycarbonate filters of 47 mm diameter and 0.4 ym pore size were used.
Known volumes of standard liquid suspensions of UICC asbestos minerals were fil-
tered through these filters. Filter 1 represented only chrysotile. Filter 2
represented a mixture of chrysotile plus an amphibole (amosite). Filter 3
represented a mixture of chrysotile, an amphibole (crocidolite), and a contam-
inant mineral (wollastonite).
Particle Transfer Method
All methods used chloroform as the solvent for dissolution of the filter.
Method 1 consisted of Soxhlet extraction for eight hours of the polycarbonate
filter (without carbon-coating). Method 2 consisted of Soxhlet extraction for
eight hours of the same polycarbonate filters coated previously with carbon.
Method 3 consisted of first carbon-coating the filter and then Jaffe washing
for 24 hours.
Since these three methods were applied to the same three initial polycar-
bonate filters, this scheme was expected to give a close comparison among the
methods of particle transfer.
Electron Microscopic Examination
EM Parameter Selection—
On the JEOL 100C electron microscope, the various parameters, such as
accelerating voltage, beam spot size, tilt angle, screen magnification, etc.,
could be adjusted to obtain the best performance for achieving specific infor-
mation. Morphological examination and electron diffraction analysis were done
at 0° tilt angle and 100 kv accelerating voltage, whereas the X-ray analysis
was conducted at 40° tilt angle and 40 kv accelerating voltage. The scheme for
different microscope parameters is shown in Table 12.
EXPERIMENTAL WORK IN PHASE 3: EVALUATING A TEM AND AN SEM
The main objective of this phase was to compare two electron microscopes
used in the secondary electron emission mode.
Preparation of Filters
The filters used in this study were 0.4 ym pore size polycarbonate, pre-
pared by filtering liquid suspension. Sample 1 referred to chrysotile alone,
Sample 2 referred to chrysotile plus an amphibole asbestos (amosite), and
37
-------
UJ
00
Table 12
SCHEME OF ELECTRON MICROSCOPE PARAMETERS FOR FIBER IDENTIFICATION METHODS IN PHASE 2
VARIABLE
COMBINATION
CODE
X.
1
1
1
2
2
2
3
3
3
2Lx
i
2
3
1
2
3
1
2
3
X.
3
1
2
1
2
3
2
3
1
Ace.
TT n ,_
Volt
KV
100
100
100
100
100
100
100
100
100
MORPHOLOGY
Beam
bpOt
Size*
1
1
1
1
1
1
1
1
1
•»*• •
Magni-
fication
160,000
160,000
160,000
160,000
160,000
160,000
160,000
160,000
160,000
Tilt
Angle
Degree
0
0
0
0
0
0
0
0
0
ELECTRON DIFFRACTION
Ace.
Volt
KV
100
-
100
_
100
100
100
100
-
Beam
Spot
Size*
1
-
1
_
1
1
1
1
-
Camera
Length
cm
20
-
20
-
20
20
20
20
-
Tilt
Angle
Degree
0
-
0
-
0
0
0
0
-
Ace.
Volt
KV
40
40
-
40
-
40
-
40
40
X-RAY' FLUORESCENCE
Beam
C* 4_
Spot
Size*
3
3
-
3
-
3
-
3
3
»* j
Magni-
fication
44,000
44,000
-
44,000
-
44,000
-
44,000
44,000
Tilt
A«rv1 ft
Angle
Degrees
40
40
-
40
-
40
• -
40
40
* Beam spot size refers to a setting on the JEOL 100C electron microscope. Spot size refers to a large size
and spot size 3 refers to a significantly smaller size beam.
-------
Sample 3, to a mixture of chrysotile, amphibole (crocidolite), and a contam-
inant (wollastonite).
Particle Transfer Method
All filters were prepared by Soxhlet condensation washing for eight hours
using chloroform as a solvent. The carbon-coated TEM grids used were nickel
marker (finder) grids to facilitate examining the same grid opening in the
two instruments.
EM Instrument Parameters
The electron microscope parameters were chosen such that the highest capa-
bility of each instrument was not exceeded. The comparison was done at 10,OOOX,
which represented the highest usable magnification in JSM 50A scanning electron
microscope.
Instrument
Identification
Method
Accelerating
Voltage
KV
Tilt Angle
degrees Magnification
JEOL 100C TEM
JEOL-JSM 50A SEM
Morphology
Morphology +
X-ray
Morphology
Morphology +
X-ray
100
40
40
40
0
35
5
20
10,000
10,000
10,000
10,000
Specific grid openings were examined in the two instruments in succession.
Electron Microscope Examination
Morphological identification in secondary electron imaging was based on the
fiber dimensions rather than the internal, or surface structure, because of the
difficulty in focusing the image. The focusing difficulty was due partly to the
movement of fiber images under the beam because the fibers acquired electrical
charge. In the JEOL 100C instrument, a tilted specimen was more difficult to
focus because of the height differences created by tilting the grid.
39
-------
The sequential image formation, as observed on the CRT screen in the
scanning mode, was strenuous on the eye. Thus, scanning a grid opening while
watching the secondary electron image required a meticulous effort to avoid
double counting. It was found that the number of fibers recognized in a sec-
ondary electron image was smaller than those from a transmission scanning elec-
tron image of the same field of view. However, since the scanning transmission
mode is not available on most common SEM microscopes, only secondary electron
imaging was used for the purposes of comparison between the JSM 50A and JEOL
100C. In the morphological identification method, there were no special dif-
ferences between amphibole fibers and wollastonite and, hence, this classifica-
tion was subjective and was based on the observed chunkier appearance of
wollastonite.
In the identification based on both morphology and X-ray analysis, emphasis
was placed on the X-ray spectrum information. Interpretation was qualitative
and, hence, subjective. The presence of Si and Fe was interpreted to mean
amosite or crocidolite, while the presence of Si and Ca was interpreted as in-
dicating wollastonite, and the presence of Si and Mg was interpreted as chrysotile.
(A rigorous and quantitative analysis should consider the relative proportion of
these elements also as described by Millette [47]). In general, the X-ray count
rate was quite small and, hence, the X-ray peaks were also small. Fibers which
did not given recognizable X-ray peaks were classified as ambiguous.
The sequence of examining samples was random to avoid bias.
EXPERIMENTAL WORK IN PHASE 4: ASHING AND BONIFICATION
Filter Preparation
Twelve membrane filters, eight polycarbonate, 0.2 ym pore size, and four
cellulose acetate, 0.45 ym pore size, were used separately to filter a standard
chrysotile suspension. The nominal amount of chrysotile (0.13 x 10 grams) was
2
deposited by filtering on 37 mm diameter filters (with effective area of 9.6 cm )
Q 2
to achieve a calculated chrysotile concentration of 13.5 x 10 gm/cm . After
filtration, the filters were stored in disposable petri dishes and air-dried in
the clean work bench.
Six of the polycarbonate and three of the cellulose acetate filters were
cut in half for ashing and sonification tests. Samples on polycarbonate mem-
brane and one cellulose acetate membrane were transferred to 200-mesh carbon-
coated copper grids to serve as control standards. Transfer to the TEM grids
40
-------
was accomplished using the Jaffe washer method with chloroform (for polycar-
bonate) or acetone (for cellulose acetate) as solvents.
Ashing and Sonification Procedures
Preignition—
Nine filter samples were rolled and placed with the fiber side facing the
wall in 25 mm diameter pyrex test tubes. Preignition of three filters (two
polycarbonate and one cellulose acetate) was accomplished by moistening the
filters with 95% ethyl alcohol and igniting the filter by heating the pyrex
tube (without an open flame on the filter) prior to completion of ashing.
Two more filters (one polycarbonate and one cellulose acetate) were pre-
pared directly (without the ashing step). Details of the scheme for phase 4
samples are summaried in Table 13.
Ashing—
Three different ashing techniques were used:
1. Three samples, including the preignited cellulose acetate, placed
in separate 25 mm diameter pyrex test tubes and placed in a muffle
furnace at room temperature. The temperature was then slowly raised
to 500°C and held overnight.
2. Two samples in 25 mm diameter pyrex test tubes were ashed using
nascent oxygen generated low-temperature asher (Model 302 LTA sup-
plied by LFE Corporation, Waltham, MA). The asher was operated at
50 watts and ashing continued overnight.
3. The four remaining samples were also ashed in the LTA using a slow
start of 25 watts for the first half-hour, and completed by 2^ hours
at 50 watts. It was observed that the majority of the membrane was
ashed in the first half hour at low power.
Ultrasonic Dispersion—
After the ashing treatments, three different levels of dispersion were
applied. In each case, the glass tube containing the ash was filled with dis-
tilled water containing 1% Aerosol OT as a dispersion aid.
1. Low energy ultrasonic treatment was applied from a normal laboratory
ultrasonic cleaning bath (e.g., Bendix UTL-4B-1). The sample tube
was placed in the neck of a 250 ml water filled conical flask such
that it was held upright. Ultrasonic energy was applied for
15 minutes to disperse the ash.
2. Medium and high energy ultrasonic dispersion was applied from a
Branson Sonifier (Model 200). Fitted with a variable power supply,
41.
-------
Table 13
DETAILS OF THE ASHING PARAMETERS USED IN PHASE 4
Sample
Number
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
Sample
Description
Code*
42211
41112
42213
41221
42222
42123
42121
42222
41223
Initial
Filter
NP
MP
NP
MP
NP
NP
NP
NP
MP
NP
MP
Preignition Ashing Treatment
No HT 500 °C Overnight
Yes HT 500°C Overnight
No HT 500°C Overnight
No LT 25 watts-30 min
50 watts-150 min
No LT 25 watts-30 min
50 watts-150 min
Yes LT 25 watts-30 ijiin
50 watts-150 min
Yes LT Same as above
No LT 25 watts-30 min
50 watts-150 min
No LT 25 watts-30 min
50 watts-150 min
Ultrasonic Treatment
L Bath 15 min
M Branson Unit
35 watts-2 min
H Branson Unit
65 watts-2 min
L Bath 15 min
M Branson Unit
35 watts-2 min
H Branson Unit
65 watts-2 min
L Bath 15 min
M Branson Unit
35 watts-2 min
H Branson Unit
65 watts-2 min
Final
Filter
NP
MP
NP
MP
NP
NP
NP
NP
MP
*First digit of the sample description code refers to the phase number; second, third, fourth, and
fifth digits refer to the levels of independent variables used.
-------
this unit supplied ultrasonic vibrations at 20,000 cps via a microtip
probe which was immersed in the suspension.
3. Medium level energy was applied by using the Branson sonifier (Model
200) (equipped with a microtip probe) at its lowest setting (No. 1)
for a period of two minutes. The energy was measured at 35 x^ratts of
output power. High energy was applied at the highest allowable set-
ting (No. 7) again for a period of two minutes. The output power was
measured at 65 watts.
After the ultrasonics had been applied, the probe was washed and the wash-
ings were collected in the sample tube. Between samples, the probe was cleaned
by operating it three times in distilled water containing Alconox detergent,
then rinsing four times in filtered distilled water.
The sample from each of the dispersion experiments was filtered through a
25 mm filter of the same type and pore size as the starting filter. The filter
was dried and stored in a disposable petri dish in a clean work bench.
Particle Transfer Method—
Grid preparation was accomplished using the Jaffe washer method (without
carbon-coating of filters) with analytical grade chloroform as the solvent for
the Nuclepore membranes and analytical grade acetone for the Millipore membranes.
The apparatus arranged in a clean air bench consisted of a glass petri dish con-
taining a stack of five microscope slides with a strip of Whatman filter paper
laid over the slides. Solvent was gently poured into the dish to bring the
level to the top of the slides.
A 3 mm copper grid, carbon side up, was placed on the Whatman filter. A
3 mm disc cut from the membrane filters could then be gently placed (particle
side down) on top of the grid. Solvent was added dropwise to restore the sol-
vent level as required. The filters took 24-72 hours to completely dissolve.
The dish was covered.
Transmission Electron Microscopy—
The grids resulting from these experiments were studied using a JEM-7
Transmission Electron Microscope (TEM) operated at 100 kv and at a nominal
magnification of 10,OOOX.
Each sample was mounted and examined using TEM. A number of grids suffi-
cient to exceed a count of 100 fibers were examined. Data taken were the number
of grid openings examined and the diameter and length of each fiber observed.
43
-------
Only undamaged grid openings were counted and each grid opening examined was
surveyed completely.
EXPERIMENTAL WORK IN PHASE 5: STUDY OF DIRECT DROP METHOD
The objective of this phase was to study the direct drop method of prepara-
tion of TEM grids from liquid suspensions.
Preliminary experiments had shown a 5 y& droplet to be of appropriate size
for coverage of a standard 200-mesh carbon-coated electron microscope grid, and
a suspension was prepared such that a 5 y& drop would contain sufficient chryso-
2
tile asbestos fibers to give loading equivalent to a 10 nanograms per cm filter
loading. The same base chrysotile stock suspension used in phases 2 and 4 was
used.
The grids were mounted on a 2.5 cm x 7.5 cm glass microscope slide using
double-stick tape. The droplets were applied with a 5 y£ syringe from the
freshly prepared suspension. Four grids were used, two of which were allowed
to dry in an inverted position, and two as deposited. The grids were prepared
in a clean work bench and were covered during the drying process.
Grid openings located near the center, mid-radius, and periphery of the
droplet were examined and fiber counts and size distribution measured using
a JEM-7 TEM microscope at the same conditions used in phase 4.
44
-------
SECTION 7
RESULTS AND DISCUSSION OF PHASE 1
INTRODUCTION
Phase 1 represented the largest body of experimental data in this multi-
phase program. These data were analyzed in a variety of ways to extract rele-
vant information for evaluating 12 variables. It was necessary to decide on
specific criteria to be used in determining which variables were important and
which levels were desirable.
CRITERIA SELECTED
The following four criteria were selected:
1. Conformance to the Poisson distribution.
2. Precision in fiber count per field of view.
3. Number concentration of chrysotile fibers per unit volume of air
sampled.
4. Mass concentration of chrysotile fibers per unit volume of air sampled.
Criteria 1 and 2 refer to fiber frequency distribution characteristics.
Criteria 3 and 4 refer to detection and estimation of number and mass of chryso-
tile fibers in air samples.
SUMMARY OF PHASE 1 RESULTS
Table 14 summarizes the data from Phase 1. The various entries are as
follows:
Column 1 lists the combination code (or sample number)
Colume 2 lists the number of fibers counted
Column 3 lists the number of fields examined
Column 4 lists the mean number of fibers per field of view
Column 5 lists the area of each field of view
45
-------
Table 14
SUMMARY OF PHASE 1 DATA
1
Data Base
and Sample NO.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
41
(Repl icate of 1)
34
(Dupl icate of 4)
44
(Repl icate of 4)
36*
(Dupl icate of 6)
120
(Dupl icate of 20)
121
(Dupl icate of 21)
2
No. of Fibers
Counted
144
211
164
223
201
178
269
237
221
60
203
210
695
139
208
218
218
200
196
24
26
34
227
201
169
205
249
215
200
206
64
33
35
3
No. of Fields
200
12
211
46
26
4
1
4
72
276
108
8
1
200
19
17
2
7
15
280
200
200
4
186
204
140
7
75
39
21
200
220
221
4
Mean No. of
Fibers/Field
0.72
17.55
0.78
4.85
7.72
44.50
269.0
59.3
3.06
0.216
1.88
26.25
695-0
0.695
10.95
12.8
109.0
28.5
13.06
0.086
0.13
0.17
56.8
1.08
0.827
1.462
35.6
2.86
5.12
9.8
0.32
0.15
0.158
5
Area of
each Field
x 10~6 cm2
1.0
72.0
0.25
1.0
4.0
72.0
72.0
4.0
0.25
0.25
1.0
72.0
72.0
1.0
4.0
0.25
72.0
4.0
72.0
0.25
1.0
4.0
72.0
1.0
4.0
0.25
72.0
1.0
1.0
1.0
0.25
0.25
1.0
6
No. of
Fibers/cm^
x 10°
0.72
0.244
3.10
4.85
1.93
0.62
3.74
14.82
12.25
0.862
1.88
0.364
9.67
0.695
2.74
51.2
1.515
7-12
0.181
0.344
0.13
0.042
0.79
1.08
0.207
5.86
0.494
2.86
5.12
9.8
1.28
0.60
0.158
Small fields of view were chosen in a consecutive sequence.
46
-------
Column 6 lists the number of fibers per square cm of the filter
STATISTICAL DISTRIBUTION TESTS
Distribution of Fibers in a Microscopic Field of View
Since, in an electron microscope method, only a very small area of the
sample is examined and an assumption is made that the area examined is repre-
sentative of the entire sample for computing the average fiber concentration
and fiber characteristics, it is important to check whether this assumption is
statistically sound. This is done by comparing the variation of fiber distri-
bution with a Poisson distribution model.
Poisson Distribution Tests
An analysis of the Phase 1 data was performed to determine whether the
variation in the observed numbers of fibers per field in the various samples
was in accordance with the Poisson distribution, and if not what the nature of
the departure was. The Poisson sequence for the expected numbers of fields
containing 0, 1, 2, . . . fibers is
(F)(e~X)(l, A, \2/2l, A3/3I, . . .)
where F is the total number of fields and X is the mean number of fibers per
field. It is considered desirable that the Poisson model hold, since this is
an indication of truly random sampling, and simple methods of establishing con-
fidence intervals for the mean number of fibers per unit volume of air can be
applied.
To investigate this question, 21 statistical tests were made, as summarized
in Table 15. The data for each test consisted of the fiber counts per field
that were recorded during the EM examination of a particular Phase 1 sample or,
in three instances, a duplicate pair of samples. The duplicate pairs of samples
were: 4 and 34; 20 and 120; 21 and 121. A pair consisted of different 2.3 mm
diameter portions of the same filter. It was shown that the two samples of
each pair were in good agreement, and therefore the counts were combined for the
present purpose.
Each set of test data was analyzed by means of computer program POISSON-1
written at IIT Research Institute for the purpose of determining the goodness
of fit of the distribution. The listing of this program is given in Appendix C.
The printouts for two illustrative samples, 1 and 26, and presented in Appendix C.
47
-------
Table 15
TESTS FOR APPLICABILITY OF THE POISSON DISTRIBUTION TO NUMBER OF FIBERS PER FIELD
00
Test
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Samples in
Data Base
1
2
3
4 & 34 f
5
9
10
11
14
15
16
19
20 & 120 f
21 & 121 f
22
24
25
26
36
41 (Repl.of
44 (Repl,of
Size of Field
cm2 x 10-6
1.0
72.0
0.25
1.0
4.0
0.25
0.25
1.0
1.0
4.0
0.25
72.0
0.25
1.0
4.0
1.0
4.0
0.25
0.25
1) 1.0
4) 1.0
No. of
Fields,
F
200
12
211
85
26
72
276
108
200
19
17
15
498
420
200
188
203
140
215
75
21
No, of
Fibers
144
210
164
423
201
221
60
203
139
208
218
196
57
61
34
201
169
205
64
215
206
Mean No,
Fibers per
Field, A
0.72
17.50
0.78
4.98
7.73
3.07
0.22
1.90
0.70
10.95
12.82
13.07
0.11
0.15
0.17
1.07
0.83
1.46
0.30
2.87
9.81
Degrees
of
Freedom
2
1
2
7
4
5
1
4
2
2
3
2
1
1
1
3
2
3
1
5
2
Chi-
Square
30.54
7.69
11.84
21.83
1.45
120.33
16.19
8.81
137.02
2.46
1.95
3.35
26.80
6.72
35.46
7.61
114.44
3.23
10.73
16.75
6.89
Good Fit
to
Probability Poisson
.001>P
.01>P>.001 (*)
.001>P
.01>P>.001 (*)
.9>P>.8 *
.001>P
.001>P
.10>P>.05 *
.001>P
.3>P>.2 *
.7>P>.5 *
.2>P>.1 *
.001>P
P = .01 (*)
.001>P
.10>P>.05 *
.001>P
.5>P>.3 *
R = ,001
,01>P>.001 •(*)
,05>P>.02 (*)
f Combined data from original ana duplicate j-mm 11
* Conform to Poisson.
(*) Borderline conformation to Poisson.
Absence of * signifies poor agreement to Pofsson.
-------
Presented in Table 15 are the results of all 21 goodness-of-fit tests.
The EM field size is specified and the following quantities from the computer
analysis of the data are given: the number of fields (F) , the number of fibers,
the mean number of fibers per field (X), the degrees of freedom for assessing
the goodness of fit of the Poisson distribution, and the total Chi-square value.
The range within which the probability, P, lies is given in the last column,
determined as a function of Chi-square and the degrees of freedom from a stan-
dard table IV in reference 55.
A number of tests revealed good agreement between the observed numbers of
fibers per field and the numbers computed from the Poisson model of random vari-
ation, while other tests revealed poor agreement. The samples in which the fit
was particularly good are 5, 11, 15, 16, 19, '24, and 26. In these seven in-
stances, the probability of a worse fit due purely to accidents of sampling was
less than 1 in 1,000. In the remaining five instances (samples 2, 4 and 34
combined, 21 and 121 combined, 41, and 44), the probability values are greater
than 1 in 1,000 but less than 1 in 20.
For the purpose of analyzing the EM procedural factors in relation to com-
pliance and non-compliance with the Poisson distribution, the samples in Table 15
were assigned to two categories: those with P _< 0.001 and those with
P > 0.001.
The cases conforming definitely to Poisson distribution are denoted by
asterisks and those borderline cases are shown by bracketted asterisks in
Table 15.
Tendency Towards Poisson Distribution as a Criterion for Optimizing Variable
Levels
It is interesting to understand why 12 cases tend to follow Poisson distri-
butions whereas the remaining nine cases do not. The frequency of variable-
levels among the cases in each class can be examined. The variable-levels were
studied for each variable and the frequency distribution of these levels is
summaried in Table 16.
For example, consider the variable X, , the filter-type. Of the 12 cases con-
forming to Poisson distribution, six had Nuclepore filters and the other six had
Millipore filters. Of the nine cases not conforming to Poisson, two had Nucle-
pore filters and seven had Millipore filters. For another example, consider
variable Xq, the method of filter dissolution. Of the 12 cases conforming to
49
-------
Table 16
VARIABLE LEVEL FREQUENCY DISTRIBUTION IN TWO GROUPS
xl
J.
X
z
X
J
X4
*T
x
X6
\f
X7
/
X8
o
X9
*F
X10
JLU
1 1
X12
jL£-
Variable
Composition
Loading
Sampler
Filter
Pore Size
Particle Side
2.3 mm
Location
Carbon Coat
Transfer
Method
Magnification
Grid Opening
Location
Choice of
Field
Level
1
2
3
L
M
H
Hi-Vol
Personal
NP
MP
0.2
0.4
0.8
Down
Up
Peri
MR
Ctr
Yes
No
Sox 1
Sox 2
Jaffe
5
10
20
Peri
MR
Ctr
Random
Consecutive
Full Grid
Opening
Group 1
12 Tests
Conforming
to Poisson
Distribution
5
3
4
5
[5]
2
5
[7]
[6]
6
[5]
4
3
[10]
2
4
3
5
5
[7]
3
3
[6]
4
[6]
2
5
3
4
[7]
3
2
Group 2
Nine Tests
Not
Conforming
4
2
3
4
[3]
[2]
[2]
7
[2]
7
4
[2]
3
[3]
6
3
3
3
[1]
8
4
4
[1]
2
[2]
5
3
1
5
3
6
[o]
Remarks On
Best Choice to
Achieve Maximum
Frequency in
Conforming and
Minimal Frequency
in Nonconforming
Best Choice
Best Choice
Best Choice
Best Choice
Best Choice
[] indicates the highest frequency in Group 1 and the lowest frequency in
Group 2.
50
-------
Poisson distribution, three were prepared by Soxhlet method 1, three by Soxhlet
method 2, and six by Jaffe method. Among the nine cases deviating from Poisson
distribution, four were prepared by Soxhlet method 1, four by Soxhlet method 2,
and only one by Jaffe.
If we how hypothesize that the variable levels should be chosen which are
conducive to Poisson distribution (i.e., maximum frequency among the levels),
then we can make a clear-cut choice in some variables. For example, in vari-
able X,, particle side down is definitely preferable to particle side up.
Similarly, in variable X-, the filter dissolution method, the Jaffe method has
the maximum frequency and hence is conducive to obtaining Poisson distribution
in fibers.
However, it is difficult to make a clear-cut choice in some cases. For
example, in variable X. , the frequency is 6 for Millipore and 6 for Nuclepore.
In order to avoid such indecisive cases, one may look into another group (those
deviating from Poisson's distribution) and select the variable level which is
the least conducive to deviation from Poisson distribution (i.e., select the
least frequency). For example, in variable X,, Nuclepore filter with a low
frequency 2 is preferred to Millipore with frequency of 7*
Thus, a choice of variable level should be such that it corresponds to
the maximum frequency in the group conforming to Poisson distribution and also
to the least frequency in the group deviating from Poisson distribution.
Following such a criterion, variables X_, X, , X&, X-, and X-Q give a
definitive choice in variable level. In variables X~, X,., Xg, and X-2> a com-
promise has to be made. In the remaining cases of variables, X^, X?, and X.^,
the choice is not governed by the variable levels. The best choices in the
levels in the nine out of 12 variables studied are indicated in Table 16.
Most of the choice can be explained rationally. It should be noted that
compliance with Poisson distribution is one of the many rational criteria that
can be used in selecting variable levels. Other criteria, such as least vari-
ability in electron microscopy results, are applied in the next step of statis-
tical analysis.
Precision ift Fiber Counts per Field as a Criterion for Optimizing Independent
Variables /
In a manner similar to that discussed in the previous section, one can also
51
-------
use the precision in fiber counts per field as a criterion for optimizing inde-
pendent variables.
Table 17 lists for each sample (Column 1) the mean (Column 2), the standard
error of the mean (Column 3), and the ratio of standard error to the mean
(Column 4). These statistical quantities are based on each field as a'unit of
analysis. The relative standard error (i.e., R.S.E. = standard error of
mean/mean) has been chosen as a measure of the precision. A value of 0.10 and
less has been arbitrarily chosen to indicate good precision and higher values
to indicate high variability or poor precision. The categories of good and
poor precision are denoted in the remarks column.
It is found that our of a total of 28 cases, 12 have been classified as
having good precision and 16 as having poor precision. Following the same form
of analysis as was explained earlier for compliance with Poisson distribution,
the frequency distributions in the variable levels have been developed.
Selecting the variable levels with the highest frequency in the good precision
group and the least frequency in the poor precision group indicates definite
trends as noted in Table 18.
It is interesting to compare these trends with those according to the
criterion of compliance with Poisson distribution. A comparison between the
two criteria (see Tables 16 and 18) shows that in the majority of cases, the
choice of the best variable level is identical in two criteria. In a few cases,
the best choice in one does not match with the best choice in the other. For
example, in variable X», the Jaffe method appeared the best in the criterion of
compliance with Poisson distribution. However, in the best precision criterion,
Soxhlet method 1 appeared quite comparable with the Jaffe method.
Consideration of Fiber Characteristics
In the discussion so far, we had referred to only the frequency distribu-
tion of fibers. Now we consider the other characteristics of the sample,
namely, the size distribution of length, width, aspect ratio, volume, and mass
of chrysotile fibers. These quantities are termed statistical descriptors.
Statistical Descriptors of the Observed Fibers on a Per-Sample Basis
Included in the Phase 1 data base is a unit record for each of the almost
8,000 fibers observed under the electron microscope. A table was prepared by
computer from the fiber records of each sample separately containing summary
52
-------
Table 17
PRECISION IN FIBER COUNT PER FIELD AS A CRITERION FOR OPTIMIZING
Std. Error
Sample
1
2
3
4 + 34
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20 + 120
21 + 121
22
23
24
25
26
27
36
41
44
Mean
0.72
17.50
0.7773
4.9765
7.7308
44.50
(Not Enough
59.25
3.0694
0.2174
1.8796
26.25
(Not Enough
0.6950
10.9474
12.8235
109.0
28.5714
13.0667
0.1145
0.1452
0.1700
56.75
1.0691
0.8325
1.4643
35.5714
0.2977
2.8667
9.8095
of Mean
0.0787
4.8218
0.0699
0.3036
0.4659
18.3416
Data)**
16.25
0.5029
0.0335
0.1353
4.7650
Data)**
0.2184
1.6463
0.9044
9.0
2.0914
0.9282
0.0176
0.0221
0.0572
11.2129
0.0896
0.1676
0.1108
6.6471
0.0580
0.2556
1.1436
Std. Error/Mean
0.10
0.27
0.08
0.06
0.06.
0.41
0.27
0.16
0.15
0.07
0.18
0.31
0.15
0.07
0.08
0.07
0.07
0.15
0.15
0.33
0.19
0.08
0.20
0.07
0.18
0.19
0.08
0.11
Remarks
R.S.E.*10.1
good
good
good
good
good
good
good
good
good
good
good
good
R.S.E.>0.1
poor
poor
poor
poor
poor
poor
poor
poor
poor
poor
poor
poor
poor
poor
poor
poor
* R.S.E. stands for relative standard error.
** Only one full grid opening was examined and was considered as a unit of
analysis. Therefore, the data are inadequate for estimating the precision
from one field to the next.
53
-------
Table 18
VARIABLE LEVEL FREQUENCY DISTRIBUTION IN TWO GROUPS
xl
1
x2
fm
X.
3
x4
x_
X6
\^
X7
/
X8
X9
X10
i \*
xn
X12
Variable
Composition
Loading
Sampler
Filter
Pore Size
Particle Side
3 mm Location
Carbon Coat
Transfer Method
Magnification
Grid Opening Loc.
Choice of Field
Level
1
2
3
L
M
H
Hi-Vol
Personal
N.P.
M.P.
0.2
0.4
0.8
Down
Up
Peri
MR
Ctr
Yes
No
Soxhlet 1
Soxhlet 2
Jaffe
5,000
10,000
20,000
Peri
MR
Ctr
Random
Consecutive
Full Grid
Group 1
12 Tests Showing
Good Precision
5
4
3
5
3
k
4
8
[8]
4
[5]
4
3
8
4
3
3
[6]
[7]
5
[5]
2
5
3'
[6]
3
3
[53
4
[6]
4
2
Group 2
16 Tests Showing
Poor Precision
6
4
6
5
7
*
5
11
[3]
13
[4]
5
7
10
6
6
5
[5]
[0]
16
[/,]
7
5
6
[4]
6
6
[5]
5
[*»]
6
6
Remarks
Better Choice
Best Choice
Best Choice
Better Choice
Best Choice
(close 2nd best)
Best Choice
Best Choice
Best Choice
54
-------
statistics. A typical printout for sample 1 is shown in Table A-4 of Appendix A.
The total number of fibers that did not extend beyond the EM field were noted.
Values of the following variables were first found on a per-fiber basis within
each sample:
V^ - Fiber width in micrometers (considered as diameter)
V2 - Fiber length in micrometers (that portion of the fiber within the EM
field in the case of a fiber that crossed the boundary of the field)
V» - Aspect ratio, V = V /V
•J -3 »L JL.
2
V, - Fiber volume in cubic micrometers, (ir/4) (V1) (V?)
V-, - Natural logarithm of V
V_- - Natural logarithm of V^
£* -L ib
V-, - Natural logarithm of V,
V, - Natural logarithm of V,
The following statistical descriptors for the designated variables were then
computed from the individual fiber values: the total, the mean, the standard
deviation, the standard error of the mean, the variance, the minimum value, and
the maximum value. The variables for which these quantities are given are:
V, - Over all fibers
Vn - Over all fibers
V?1 - Over fibers lying wholly within their fields
V_, - Over fibers lying wholly within their fields
V,- - Over fibers lying wholly within their fields
The total fiber volume of the sample is required for estimating the mass
concentration of fibers in the atmosphere. A log-normal model of random vari-
ation among fibers is considered appropriate for width (or diameter), length,
aspect ratio, and volume [56].
Statistical Analysis of Phase 1 Fractional Factorial Experiment
For evaluating the effects of independent variables on the statistical
descriptors (or dependent variables), we used a regression analysis technique
[59,60].
55
-------
Dependent Variables—
Certain of the sample descriptors (or measured response) were analyzed in
relation to the 12 controlled factors (or independent variables) of the Phase 1
experiment design by constructing performance equations with the descriptors as
the dependent variables. The dependent variables chosen are listed in Table 19.
These are the dependent variables, or observed responses, of the experiment.
A square root transformation is appropriate in response to Yg because it in-
volves number count [56]. In all other responses, designated Y through Y ,
natural logarithmic transformations are appropriate [57].
Regression Analysis—
Regression equations were constructed to express each dependent variable
in terms of the coded independent variables. The best values of the coefficients
were determined by statistical regression methods using the stepwise regression
program BMD02R from the BMD library of statistical programs [59].
The signs of the coefficients of independent variables in regression equa-
tions are listed in Table 20. A positive sign in this table means that the de-
pendent variable increases in value as one increases the coded value of the
independent variable. A negative sign represents a decreasing trend and an
absence of any sign means that the dependent variable is not significantly
affected. This is easy to visualize for the linear components. For the quad-
ratic components and for a combination of linear and quadratic components, the
effects associated with the different levels of the experimental factors can
be clearly displayed in the form of plots. If the magnitude of the effect was
not statistically significant, it was dropped from further consideration. A
complete treatment of the method is illustrated in Appendix B.
The results of the analyses for Y and Y are given in detail in Tables B-l
through B-4 and summarized in Figures 3 through 6 for a quick comparison. Fiber
count concentration Y_ and fiber mass concentration estimates Y _ are the most
commonly considered responses for quantitative EM work. Other responses, YI
through Yg, are of secondary importance and such detailed analyses of these
are not presented.
Discussion of Main Effects
In this section, we discuss separately the effect of each variable on the
two dependent variables, namely, the fiber count estimate (Y_) and the mass coi
centration estimate (Y .). Rational explanations are offered where possible.
56
-------
Table 19
DEPENDENT VARIABLES, PHASE 1
Variable Definition*
YI Mean Ln (fiber width, micrometers)
Y2 Standard error of Y!
Y3 Mean Ln (fiber length, micrometers)
Y4 Standard error of Y3
Y5 Mean Ln (aspect ratio)
Y6 Standard error of Y5
Y7 Mean Ln (fiber volume, micrometers3)
Y8 Standard error of Y7
Y9 Square root (estimated number of fibers per
cm3 of atmosphere)
Y10 Ln (estimated mass concentration of fibers
in the atmosphere, micrograms per cubic meter)
* The unit of observation for these variables is a combination as
specified in the experiment design. All logarithms are to base e.
57
-------
Table 20
SIGNS OF COEFFICIENTS OF INDEPENDENT VARIABLES IN PERFORMANCE EQUATIONS,
PHASE 1
Dependent Variables^ '
Independent
Variable Code* Y, Y2 Y3 Y,, Y5 Y6 Y7 Y8 Y9 YH
1. X:L Composition + + +
2. XjQ Composition - - - - - +
3. X2L Concentration ____ ____
4. X2Q Concentration +
5. X3Q Sampler type + + + +
6. X^Q Filter type + _ _
7. X5L Pore size
8. X5Q Pore size
9. X6Q Filter side +
10. X7L Location on filter — — — — —
11. X7Q Location on filter
12. X8Q Carbon coating + _ + _ + _ + _
13. X9L Transfer method +' + + + -
14. X9Q Transfer method — + + + —
15. X10L Magnification - + + +
16. X10Q Magnification — — +
17. XnL Grid opening loc.
18. XnQ Grid opening loc. + +
19. X12L Choice of fields
20. X12Q Choice of fields - + + + - + +
^ ' For explanation of dependent variables, please see Table 19.
* Independent variables designated Xx-Xi2 are the same as described in Table 4.
Each of these have linear (L) and quadratic (Q) components. A complete
description of these is given in Appendix A.
58
-------
Ul
o
0
o
o
CU
f>
•rl
Pn /-s
U-i 'H
O ^5
M
C
0
4_t
a)
^
6.0
5.0
4.0
3.0
2.0
1.0
0.0
-1.0
-2.0
-3.0
-4.0
-5.0
ft 0
.
: , 1 '
: l 1 III
-
-
r
it ii i
cu cu co
rH rH CO
•H -H Ct)
CU -M CU -P 0) rH
rH O -U O -U t)D
•H CO -iH CO -H. M '•j
4J >-, CO >> CO CU 2
OMOMO^i H^ »C
COJ36J3S-H 4J 3 >i O O
>-, O •< U •< En J3 -r) > > CO
ij tsJD "T? ^3 I W
.flOOOOO >rl <" ^ *H "
CJ^O-t |Z
X6: Filter Xg: Carbot
Side Coating
(d) (e)
Figure 3. Graphical presentation of performance equation 9 in Phase 1,
Net contribution to square root of fiber concentration (no.
of all fibers/cm3 of air).
-------
ON
O
o
c
o
CJ
•U M—I
o o
0) O
CO w
cr ^
M
4-1
O
CU
a
6.0
5.0
4.0
3.0
2.0
1.0
0.0
-1.0
-2.0
-3.0
-4.0
-5.0
f. o
T
i
1
,__(
4-1
(U
,_ |
*£3
X
0
co
x9
]
!
1 1
CV|
4-1
CU
r-4 O>
fj im
X "4-1
O «
: Transfer
Method
I
I
1 1
x
X 0
0 0
0 0
o
« o
I
-
1
b^
o
o
o
o
CM
X^Q: Magnification
I
1
»^"»
M
0)
,e
a.
•H
(U
PM
Xll =
1
1
(0
3
•H
TJ
CO
1
T3
•H
£
Grid
I
i
)-<
0)
4-1
fi
(U
u
Opening
Location
1 I
1 1
CU
•H
4-1
j3
C O
O 01
*d to
pj d
cd o
prf O
X]^: Choice
Fields
1
1
T3
•H
M
O
(U
M
•H
4_)
c
w
of
(a)
(b)
(c)
Figure 4^ Graphical presentation of performance equation 9
in Phase 1. Net contribution to square root of
'fiber concentration (no. of all fibers/cm^ of air).
-------
CO
co
>-•
4-1
(-1 O
4J -H
C -U
O C(J
CJ S-i
0) 0)
*£
O
o
1.2
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-1.0
-1.2
^j
O
•H
4-1
O
co
oo
vo-st-
4-1 01
O 4J
CO -H
CO
CO
Cfl
H
00
t-l
ooo
r-cN^
X,: Composition
of Sample
-(a)
'O
cy
cu
Bd
Concentration
on Filter
(b)
o
(X
01
s
Z
0)
(-1
o
o.
•H
-H
S
X : Filter
Type
(c)
0
oo
X_: Pore Size
(d)
Figure 5. Graphical presentation of performance equation 10
in Phase 1. Net contribution to natural log of mass
concentration of all fibers, yg/m3 of air.
-------
ON
to
TO • j-.i— — - _ _ , ,
03 4-1
«S o 1.0 -
co " •*•*• T
^4j[ 0.8-
OD
an. 0.6- L
o2 0.4- T
i— i < £ 0.2 T
O PM
•"-• ' 0 0
iH
C iH
° , 3 -
OJ ^0 4J 4J
^ « n ^ * », H
Q, LJ
Xy: 3 mm Portion Xfl: Carbon X : Transfer XIQ: 1
Location Coating Method
(a) (b) (c)
••' 1 l l
it f r •
cu "C
•S i
° 8' a 3 a
§O O C 0 *J
o o cd o C
^ CM p3 o .. w
Magnification X „: Choice of
Fields
(d) (e)
i
i
-
.
Figure 6. Graphical presentation of performance equation 10 in Phase 1. Net
contribution to natural log of mass concentration of all fibers,
of air.
-------
X.^ Composition of Sample; This variable affects the total fiber count
estimate. The composition level 2 (viz, 60% chrysotile plus 40% amosite) gives
a higher value than either the composition 1 or the composition 3. The vari-
able X.^ also affects the mass concentration estimates, the composition 2 shows
higher value than compositions 3 and 1. The probable reason for this is that
the amosite fibers are generally blocky, whereas we assume a cylindrical geome-
try for computing the volume. This assumption of cylindrical shape tends to
over-estimate the volume of amosite fibers.
Xp Concentration on Filter: Light concentration gives the highest relative
value for fiber count estimates and the heavy concentration gives the lowest
relative value. Exactly the same trend is apparent for the mass concentration
estimates. The most probable explanation of this is the possibility of aggre-
gation of fibers in heavy concentration samples, leading to a failure to count
all fibers.
X0 Sampler Type; Personal sampler appears to give a higher value of fiber
j
counts than the high-volume sampler. This variable has no significant effect
on mass concentration estimates. One plausible explanation may be that the
face velocity of particles in a personal sampler is l/5th that in the high-
volume sampler.
X, Filter Type; Millipore filters appear to give slightly higher mass
concentration estimates than the Nuclepore; however, the filter type does not
significantly affect the fiber count estimate.
Xg. Pore Size; Pore size between 0.2 and 0.8 ym does not significantly
M~3 rr ~*H****amm
affect the fiber count estimate; on the other hand, the 0.4 ym filters
appear to yield the highest mass concentration estimates.
Xfi Filter Side on Grid; Keeping the particle side up results in signifi-
cantly lower fiber count estimate, but does not affect the mass concentration
estimate. A probable explanation is that only the very small and fine fibers
tend to be washed away and their mass is not appreciable. Keeping particle
side down is the best method to avoid this type of fiber loss.
X? 2.3 mm Portion Location; This variable has no significant effect,
either on the fiber count estimate or the mass concentration estimate.
63
-------
X0 Carbon Coating: Carbon coating of the filter very definitely gives a
o *
higher value of fiber count estimate and also a higher value of the mass con-
centration estimate. These data give credence to the theory that the carbon
coating locks the fibers in place and prevents their washing away.
XQ Filter Transfer Method: The Jaffe method of filter dissolution gives
the highest value of the fiber count estimate and also the highest value of the
mass concentration estimate. Thus, these data bear out the contention of the
advocates of the Jaffe method that the method is very gentle and slow and, hence,
does not wash away any fibers. The Soxhlet method 2 gives the lowest value of
fiber count estimate, presumably because of fiber loss due to extended duration
of washings.
X-Q Magnification; 20,OOOX magnification gives the highest value of the
fiber count estimate, but the effect is very slight on the mass concentration
estimate. The explanation is that with higher magnification, more fine and
short fibers are visible; however, their volume contribution is quite small.
X, Grid Opening Location; A mid-radius location gives low values of fiber
count estimate, but does not significantly affect the mass concentration estimate.
Presumably, small fibers may migrate from the center of the filter towards the
periphery — if it is not carbon coated, the effect being most noticeable at
the mid-radius.
X. 2 Choice of Fields; Random and consecutively selected fields give similar
values of the fiber count estimate and significantly higher values of the esti-
mate than those for the entire grid opening. The same trend also holds for the
mass concentration estimate. One possible explanation is that the operator may
be unknowingly skipping empty fields (with no fibers), thus introducing a bias.
Another possible explanation is that a full grid opening examined required a
long time for fiber counting and often caused operator fatigue, which could
have resulted in lower fiber count.
Optimum Choice of Variable Levels
It is a reasonable assumption that variable levels which give the highest
values of the fiber count estimates (Y_) and/or the mass concentration esti-
mates (YIQ) are the optimum levels. (There is no reason to suspect external
contamination, which could increase the fiber count or the mass concentration
estimates. If fiber migration occurred, there will be some areas with higher
true concentration but other areas will be lower in concentration.) Thus, high
64
-------
values are associated with efficiency of fiber retention, fiber recognition,
counting, and sizing, and low values are associated with fiber loss, inefficient
technique, etc.
Based on these assumptions, the best choices of variable levels for maxi-
mizing (Yg and YIQ) are summarized in Table 21. These choices are also com-
pared with those based on the earlier chosen criteria of compliance with Poisson
distribution and with the internal precision of fiber counts per field.
Though the optimum choice is somewhat dependent on the criterion chosen,
many very remarkable trends emerge.
The variables X , X,, and X^ (air sampling variables) do not affect re-
sponses Y_ and Y. _. However, considerations of Poisson distribution and better
precision allow choice to be NP over MP as indicated in Table 21. Variable X., ,
is not generally within one's control. Variable X- (concentration on filter)
should be kept low (in practical ambient air monitoring applications this
would not be a problem).
Among the variables of TEM grid preparation, variable X7 (2.3 mm portion
location) appears immaterial. Still, it is recommended to cut 2.3 mm portions
from widely separated locations for duplicate of triplicate grids. Variable XQ
o
offers a clear-cut choice. Carbon-coating of filters is recommended for pre-
serving the particulate integrity and distribution. Variable X, again offers a
clear-cut choice. For transferring particles to carbon substrate, the particle
side of the filter should be kept facing down, i.e., in contact with the carbon
substrate. In variable Xq, the filter transfer method, the Soxhlet method 2
should be eliminated. Indications suggest the superiority of the Jaffe method
and this is reinforced by the fact that it is less susceptible to operator
technique.
Among the variables of TEM examination, variable X^, the grid opening
location of preference, is in the center or the peripheral regions. The vari-
able X]Q, the magnification, appears to give a clear choice of 10,OOOX. However,
if we assume that the criterion of maximizing the fiber counts is more important
than the mass concentration and the other criteria, then the choice is 20,OOOX.
From a practical standpoint, higher resolution and higher magnifications are
important for detecting the very fine fibers [46] which may have a greater
chance of remaining airborne. Therefore, we recommend a magnification of
20,OOOX be used for fiber counting, sizing, and studying morphology. For cases
65
-------
Table 21
OPTIMIZATION OF VARIABLE LEVELS ACCORDING TO FOUR DIFFERENT CRITERIA
Variable
Xi Composition
Xj Loading
Level
1
2
3
L
M
H
Sq. Root of
Est. No. of
Fibers/cm^
Highest
Highest
Ln Poisson- Precision
Est. Mass Distr. in Fiber
Cone. yg/n>3 Compliance count/field Remarks
Highest
Highest A low loading level is
Best certainly preferable.
However, this requires
Xj Sampler
XA Filter Type
Xj Pore Size
X6 Particle Side
Filter Transfer
Magnification
Hi-Vol
Personal
NJ>.
M.P.
0.2
0.4
0.8
Down
Up
Higher
[Higher]
X. 3 mm Portion Loc. Peri
' MR
Ctr
Xo Carbon Coating Yes
No
Soxhlet 1
Soxhlet 2
Jaffe
5,000
10,000
20,000
X.. Grid Opening Loc. Peri
MR
Ctr
Low
Choice of Fields Random [High]
Consecutive High
Full Grid Low
covering several grid
openings for counting
enough number of
fields.
Variable X3 is prob-
ably Insignificant.
Better Better Nuclepore appears a
Higher better choice.
Best Pore size smaller or
Highest equal to 0.4 urn is
preferable.
[Higher] [Better] Keeping particle side
down is definitely
better and must be
adopted for transfer-
ring fibers to carbon
substrates.
Highest Variable "7 is prob-
ably insignificant.
Best Duplicate grids should
be prepared from dif-
ferent locations.
[Higher] [Better] Carbon coating of fil-
ter Is certainly better
and should form a
necessary step in sam-
ple processing.
Best More work needed for
dec i d i ng between
(2nd Best) Soxhlet 1 and Jaffe.
Soxhlet 2 should be
eliminated.
5.000X is too low to
Best give reliable EM esti-
mates. While 10.000X
appears best overall,
20.000X is preferred
when fiber count con-
centration and detec-
tion of small fibers
is more important
than mass concentration
estimate.
Variable xu is prob-
Best ably insignificant.
Grid openings should be
chosen from all loca-
tions with equal
frequency.
[Best] Though random choice
of small fields is
best, in practice, it
is easier to use full
grid opening, which
eliminates -the fibers
crossing the field of
view.
[Higher]
(2nd Highest)
[Highest] [Highest] [Best]
Highest Best
Highest (2nd Highest)
[High]
High
Low
[ ] Best Choice, ( ) 2nd Best Choice
66
-------
where majority of fibers are of amphibole asbestos, a lower magnification
(e.g., 10,OOOX) may be sufficient.
Variable Xj~» the choice of fields, appears to give the random choice of
Small fields as the best in all respects. However, in practice, this can
cause problems in counting fibers longer than the field of view and also fibers
extending beyond the perimeter of the field of view. Also, the operator un-
knowingly may tend to skip empty fields, thereby introducing a bias. These
difficulties can be avoided by using full grid opening as one field. If the
fiber concentration is low enough, there will be no operator fatigue.
MASS CONCENTRATION ESTIMATES
In addition to fiber number concentration, the mass concentration of
asbestos in air may be an important parameter. Table 22 lists the details of
the air sampling parameters, namely, the effective filter area (Column 2), the
2
volume of air filtered in leters (Column 3), the air volume filtered per cm
of the filter (Column 4). It also lists the total area examined in the EM
(Column 5)s the observed fiber counts (Column 6), the estimate of fibers per
ml of air (Column 7), the total volume over all fibers observed (Column 8),
fiber density (weighted average) (Column 9), and estimated mass concentration
3
yg/m in air (Column 10).
Comparison of Observed Mass Concentrations with Those Expected
When we compare the estimated mass concentration values from Table 22
with those from Table 11 as expected from the aerosol generation parameters,
we find there is a substantial difference. The EM estimates are smaller by
a factor of 100-300 in samples 1 to 18 and by a factor of 300-1,000 in samples
19 to 27.
Sample 11 gives a substantially larger value than all the rest. This is
explained by the fact that a few very large fibers were detected in this
3
sample, as listed below. The 203 fibers counted had a total volume of 11.19 urn .
3
The three large fibers listed below account for 7.5 ym of the volume (see p. 69)
The detection of these large fibers indicates that the large fraction of
the total mass is accounted for by a few large fibers in the aerosol chamber.
It is likely that these large fibers have settled by gravity rather than being
drawn onto the filter by the air sampler's suction. Another possible explana-
tion is that the air circulating fan might have acted as an impactor and removed
67
-------
oo
Table 22
ESTIMATES OF NUMBER AND MASS CONCENTRATION OF ALL FIBERS PER UNIT VOLUME OF AIR, PHASE 1
Samples
1
2
3
4 & 34
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20 & 120
21 & 121
22
23
24
25
26
27
Filter
Area,
cm2
6.7
6.7
6.7
406.5
406.5
406.5
6.7
6.7
6.7
406.5
406.5
406.5
6.7
6.7
6.7
6.7
6.7
6.7
6.7
6.7
6.7
6.7
6.7
6.7
406.5
406.5
406.5
Air Vol.
Filtered,
liters
32
32
32
9116
9200
8400
512
512
512
2080
2030
2040
124
124
124
513
513
513
7.6
7.6
7.6
29.2
29.2
29.2
7920
7701
9167
Air Vol.
per Unit
Area, I/cm2
4.78
4.78
4.78
22.43
22.63
20.66
76.42
76.42
76.42
5.12
4.99
5.02
18.51
18.51
18.51
76.57
76.57
76.57
1.13
1.13
1.13
4.36
4.36
4.36
19.48
18.94
22.55
Area
Scanned,
10"6cm2
200
864
53
85
104
288
72
16
18
69
108
576
72
200
76
4.25
144
28
1080
125
421
800
288
186
816
35
504
Obs.
Fiber
Count
144
210
164
423
201
178
269
237
221
60"
203
210
695
140
208
218
218
200
196
57
61
34
227
200
169
205
249
Est. Fibers
per cm3
of Air
151
51
647
222
85
30
49
194
161
170
377
73
521
38
148
670
19.8
93
161
404
128
9.7
181
247
10.6
309
21.9
Obs . Fi ber
Volume,
ym3
0.383
1.726
0.470
1.302
0.422
0.507
2.472
1.386
0.436
0.167
11.194
1.057
2.285
3.844
1.121
0.348
0.751
0.678
1.778
0.093
0.241
0.362
0.492
0.674
0.811
0.587
8.383
Fiber
Density,*
Af
g/cm
2.43
2.43
2.43
2.43
2.43
2.43
2.43
2.43
2.43
2.58
2.58
2.58
2.58
2.58
2.58
2.58
2.58
2.58
2.54
2.54
2.54
2.54
2.54
2.54
> 2.54
2.54
2.54
Est. Mass Con-
centration, All
Fibers, yg/m3
0.974
1.016
4.508
1.660
0.436
0.207
1.092
2.754
0.770
1.220
53.590
0.943
4.424
2.679
2.056
2.759
0.176
0.816
3.700
1.672
1.287
0.264
0.995
2.111
0.130
2.249
1.874
* Fiber density refers to the average density for all mineral fibers considering their weight proportions, in the
mixture used.
-------
UNUSUALLY LARGE FIBERS DETECTED
IN SAMPLE 11
Fiber No.
5
99
110
L
ym
9.4*
5.6*
1.25*
W
ym
0.9
0.56
0.5
Volume
(ym)3
6.1
1.12
0.25
* All of these fibers extended beyond
the field perimeter; hence, these
lengths represent only underestimates
of the true length.
a substantial amount of asbestos from the air in the aerosol chamber. Sample
Number 11 was collected on a high-volume sampler with a large surface area
(406.5 cm2). It is only by chance that their presence within the areas
randomly selected on the grid was noted. This example points out the large
bias introduced by a few large fibers in the estimate of the total volume and,
hence, in the total mass concentration.
69
-------
SECTION 8
RESULTS AND DISCUSSION OF PHASES 2, 3, 4 AND 5
PHASE 2 RESULTS
In the Phase 2 analysis, two types of fiber classification were used. The
"standard classification" method allowed six categories, viz, chrysotile, amosite,
crocidolite, wollastonite, ambiguous, and other. In the "alternative classifi-
cation" method, only four categories were allowed, viz, chrysotile, amosite,
crocidolite, and wollastonite. This is, in the alternative classification
method, the ambiguous and other fibers were assigned to one of the four cate-
gories based on the best operator judgment. For example, if a sample was studied
using morphology, diffraction, and X-ray, the standard classification was based
on all three tests simultaneously. Any fiber not giving a characteristic dif-
fraction pattern or X-ray analysis was classified as ambiguous. In the alter-
native classification method, such fibers were assigned to the four mineral
categories based on other available information. For example, a fiber that did
not give a recognizable diffraction pattern, but gave a recognizable X-ray
analysis, was classified based on X-ray analysis and/or morphology. Similarly,
a fiber that did not give a distinct X-ray analysis was classified based on
electron diffraction and/or morphology. The category "other" includes fibers
for which X-ray and/or diffraction measurements could not be made.
In the standard classification method, ambiguous and other categories
constitute a variable percentage of the fibers. Since the ambiguous and other
categories cannot be assigned any fixed density, their mass concentration is
subject to error. The difficulty is avoided, although not eliminated, in the
alternative classification method, where there are no ambiguous and other
categories. The ambiguous and other fibers combined will be designated
"exceptional".
Phase 2 Data
Specified properties of the population of individual fibers in each of
the nine Phase 2 samples were determined by application of IITRI computer
70
-------
program SIZ1. The results of the data reduction are presented in Tables 23
through 26.
The fiber counts are given in Table 23 for the six fiber categories
allowed under the standard classification method and the four categories al-
lowed under the alternative classification method. Additional items of infor-
mation per sample are: the total fiber count, the area scanned (in units of
-4 2
10 cm ), and the number of grid openings. The area per grid opening was
-4 2
0.72 x 10 cm . The area examined in the case of each of the last two
samples was a portion of a grid opening. Column 14 lists the total time on
TEM for studying each sample.
ft ")
The estimated number concentration (in units of 10 /cm of filter) and
—9 2
mass concentration (units of 10 gm/cm of filter) are given in Table 24 for
fibers classified as chrysotile under the standard and alternative classifi-
cation methods. The 95% confidence limits for the number concentrations were
calculated by the computer program POISSON 2 under the assumption of a Poisson
distribution of the fibers. The listing for POISSON 2 is given in Appendix C.
This table also lists the total TEM time for inspecting 100 fibers in
each, sample, which is a measure of the experimental effort required. It is
quite evident that the experimental effort required is substantially dependent
on the method of analysis. For example, the method based on morphology and
electron diffraction requires the least time and, hence, should be considered
preferable among the three methods.
The size distributions of the individual chrysotile fibers within each
sample as classified by the two methods, are characterized in Table 25. The
properties treated are: fiber length (micrometers), fiber diameter or width
-12
(micrometers), and fiber mass (units of 10 gin). For each quantity, the
geometric mean and the mean and standard deviation of the natural logarithms
(to the base e) are given.
The estimated number concentration of all fibers combined, and of fibers
classified as ambiguous and other under the standard classification method, are
listed by sample in Table 26. Also listed are the percentages of the total
fibers classified as ambiguous, other, and exceptional (ambiguous and other
combined).
71
-------
Table 23
NUMBERS OF FIBERS OBSERVED AND CLASSIFIED, PHASE 2 SAMPLES
Sample
escription
Code*
2113
2121
2132
2211
2222
2233
2312
2323
2331
Combined
Area
Spflnn^r!
10~4 cm2
1.44(2)t
0.72(1)
2.16(3)
0.72(1)
1.44(2)
0.72(1)
0.72(1)
0.34(1)
0.14(1)
8.40(13)
No.
"F^HiayQ
Total
149
114
228
137
287
227
398
188
176
1904
Standard Classification Method
Chrys-
otile
93
64
209
117
221
130
204
56
39
1133
Amo-
site
0
1
0
10
11
4
0
0
0
26
Crocid-
olite
0
0
0
0
0
0
43
38
61
142
Wollas-
tonite
0
0
0
0
0
0
100
3
36
139
Ambig-
uous
50
49
1
10
22
93
45
57
11
338
fit-KoT-
6
0
18
0
33
0
6
34
29
126
Alternative
Classification Method
Chrys-
otile
149
113
228
125
276
223
219
85
58
1476
Amo-
site
0
1
0
12
11
4
0
0
0
28
Crocid-
olite
0
0
0
0
0
0
56
94
72
222
Wollas-
tonite
0
0
0
0
0
0
123
9
46
178
Total
Time
on
TEM
Hrs.
9.0
6.0
4.0
,7.0
3.5
12.0
7.0
10.5
8.0
-•J
N3
* For explanation of the sample description code, see Tables 7 and 12.
t Number in parentheses is number of grid openings.
-------
Table 24
CONCENTRATIONS OF CHRYSOTILE FIBERS, PHASE 2 SAMPLES
Sample
)escription
Code*
2113
2121
2132
2211
2222
2233
2312
2323
2331
2113
2121
2132
2211
2222
2233
2312
2323
2331
Identification
Method Usedt
STJ
M+D+X
M-KiT
M+D
M+X
M+D
M+D+X
M+D
M+D+X
M+X
ALT!
Number Concentration
Estimate, 106/cm2
of Filter**
iNDARD CLASSIFICATION
0.646 (0.521, 0.791)
0.889 (0.684, 1.135)
0.968 (0.841, 1.108)
1.625 (1.344, 1.948)
1.535 (1.339, 1.751)
1.806 (1,508, 2.144)
2.833 (2.458, 3.250)
1.647 (1.244, 2.139)
2.786 (1.981, 3.808)
SRNATIVE CLASSIFICATK
1.035 (0.875, 1.215)
1.569 (1.293, 1.887.)
1.056 (0.923, 1.202)
1.736 (1.445, 2.069)
1.917 (1.697, 2.157)
3.097 (2.704, 3.532)
3.042 (2.652, 3.472)
2.500 (1.997, 3.091)
4.143 (3.145, 5.356)
Mass Concentration
Estimate, 10~9 gm/cm2
of Filters
METHOD
4.781
7.138
9.818
14.731
15.528
10.742
26.008
25.190
35.007
)N METHOD
6.126
10.527
9.959
14.787
16.857
20.586
29.385
37.133
48.340
TEM Time for
Studying 100
Fibers, Hrs.
6.04
5.26
1.75
5.10
1.22
5.29
1.76
5.59
4.54
* For explanation of the sample description code, see Tables 7 and 12.
** Numbers in parentheses are 95% confidence limits based on the Poisson
distribution.
t M+D refers to morphology + electron diffraction,
M+X refers to morphology + X-ray analysis, and ,
M+D+X refers to morphology + electron diffraction + X-ray analysis;
73
-------
Table 25
SIZE DISTRIBUTIONS OF CHRYSOTILE FIBERS, PHASE 2 SAMPLES
Sample
Descriptioi
Code*
Fiber Length,
n Geom.
Mean
Mean
Ln
ym
St. Dev.
Ln
Fiber Width, ym
Geom.
Mean
Mean St. Dev.
Ln Ln
Fiber Mass, 10~12 g
Geom.
Mean
Mean
Ln
St. Dev.
Ln
STANDARD CLASSIFICATION METHOD
2113
2121
2132
2211
2222
2233
2312
2323
2331
0.7931
0.8238
0.8162
0.7287
0.7565
0.7413
0.8747
0.9651
1.0466
-0.2318
-0.1938
-0.2032
-0.3165
-0.2791
-0.2993
-0.1339
-0.0355
0.0456
0.7899
0.6754
0.8270
0.8657
0.7700
0.7387
0.6119
0.5990
0.6468
0.0467
0.0521
0.0531
0.0522
0.0515
0.0483
0.0584
0.0687
0.0644
-3.0634
-2.9549
-2.9349
-2.9532
-2.9670
-3.0297
-2.8409
-2.6786
-2.7423
ALTERNATIVE CLASSIFICATION
2113
2121
2132
2211
2222
2233
2312
2323
2331
0.6780
0.6996
0.7587
0.6737
0.7028
0.7254
0.8467
0.9912
0.8214
-0.3886
-0.3573
-0.2761
-0.3590
-0.3527
-0.3210
-0.1664
-0.0088
-0.1967
0.7968
0.7142
0.8366
0.8929
0.7761
0.8281
0.6463
0.5988
0.7339
0.0438
0.0507
0.0521
0.0503
0.0499
0.0471
0.0594
0.0669
0.0654
-3.1275
-2.9821
-2.9554
-2.9896
-2.9987
-3.0553
-2.8236
-2.7044
-2.7277
0.4070
0.3888
0.3597
0.3744
0.4442
0.3209
0.3036
0.3426
0.2696
METHOD
0.4200
0.3929
0,3600
0.3962
0.4337
0.3547
0.3151
0.3611
0.3296
0.00354
0.00456
0.00471
0.00405
0.00409
0.00354
0.00609
0.00929
0.00887
0.00266
0.00367
0.00420
0.00348
0.00357
0.00329
0.00610
0.00906
0.00717
-5.6557
-5.3896
-5.3589
-5.5090
-5.4991
-5.6448
-5.1017
-4.6789
-4.7250
-5.9297
-5.6076
-5.4730
-5.6602
-5.6362
-5.7178
-5.0998
-4.7036
-4.9381
1.3246
1.1463
1.3381
1.3566
1.3878
1.1215
0.9960
1.0828
0.8529
1.3575
1.1996
1.3548
1.4462
1.3677
1.2635
1.0213
1.0705
1.0018
* First digit of the sample description code refers to the phase number; second, third, and fourth
digits refer to the levels of independent variables used. For further explanation of the sample
description code, see Tables 7 and 12.
-------
Ul
Table 26
CONCENTRATIONS OF ALL FIBERS AND OF FIBERS
OF "AMBIGUOUS" AND "OTHER" CATEGORIES, PHASE 2 SAMPLES
Sample
Description
Code*
2113
2121
2132
2211
2222
2233
2312
2323
2331
All Fibers
Number
Concentration
Estimate,
106/cm2
of Filter
1.035
1.583
1.056
1.903
1.993
3.153
5.528
5.529
12.571
Ambiguous Fibers
Number
Concentration
Estimate,
106/cm2
of Filter
0.347
0.680
0 . 0046
0.139
0.153
1.291
0.625
1.676
0.786
Percent
of
Total
Fibers
33.56
42.98
0.44
7.30
7.66
40.97
11.31
30.32
6.25
Other Fibers
Number
Concentration
Estimate,
106/cm2
of Filter
0.0416
0.0
0.0833
0.0
0.2291
0.0
0.0833
1.000
2.071
Percent
of
Total
Fibers
4.03
0.0
7.89
0.0
11.50
0.0
1.51
18.08
16.48
All
Exceptional
Fibers
Percent
of Total
Fibers
37.59
42.98
8.33
7.30
19.16
40.97
12.81
48.40
22.73
* First digit of the sample description code refers to the phase number; second, third, and fourth
digits refer to the levels of independent variables used. For further explanation, see Tables 7
and 12.
-------
Methods of Data Analysis
Statistical methods applied in the analysis of the Phase 2 data are as
follows.
Confidence limits for number concentrations were calculated by IITRI
program POISSON 2 (Appendix C) under the assumption of a random distribution
of fibers on the grid.
Multiple regression analyses were made on each of the sets of dependent
variables defined below by means of a modified version of the stepwise regres-
sion program BMD02R from the BMD package of statistical programs [59]. The
data input for each regression analysis included the orthogonally coded values
of the independent variables given in Table 7 (X^, X.^, X9L, XgQ, X^L, X^Q)
in addition to the values of the dependent variables. ;
Regression Equations 1
The dependent variables for the regression analyses have the following
symbols and definitions.
Symbol Definition
YI- Square root of the estimated number concentration of
chrysotile fibers in units of millions per car of filter
Y-„ Natural logarithm of the estimated mass concentration of
chrysotile fibers in units of nanograms per cm of filter
Y- Natural logarithm of the geometric mean length of chrysotile
fibers in micrometers
Y- Natural logarithm of the geometric mean width of chrysotile
fibers in micrometers
Y , Natural logarithm of the geometric mean mass of chrysotile
fibers in units of lO"1^ grams
Y-_ Square root of the estimated number concentration of all
fibers in units of millions per cm^ of filter
Y . Arcsine of the square root of the proportion of all fibers
classified as exceptional (ambiguous or other) using the
standard classification method
There are nine values of each dependent variable, i.e., one value per Phase 2
sample.
The number concentration estimates were subjected to the square root
transformation (Y. and Y ). The mass concentration estimates and the geometric
76
-------
mean fiber lengths, widths, and masses were subjected to the logarithmic trans-
formation (Y,2» YO» Y , and Y ). The exceptional fiber percentages were sub-
jected to the arcsine-square-root transormation (Y ). The selected transfor-
mations are often employed in analyzing the effects of independent variables on
three kinds of dependent variables by means of analysis of variance or regres-
sion analysis [56,57].
The 12 regression equations constructed from the Phase 2 data are given
in Table 27. In each equation, only those candidate independent variables
appear that have effects that are significant at the 10% probability level.
The method of equation construction is described in Appendix B.
Some overall properties of the equations are given in Table 28: the number
of residual degrees of freedom, the residual standard deviation, and the degree
2 2
of determination, R . R ranges from 73 to 98% in the group of equations, sig-
nifying that the values of the dependent variables are, in general, strongly
influenced by the independent variables included in the equations.
The first six equations are based on data in which fibers were classified
by the standard method. The next five equations are based on data in which
fibers were classified by the alternative method. The final equation refers to
the number concentration of all types of fibers combined, and hence the method
of fiber classification is not applicable.
Discussion of Phase 2 Results
The results obtained in Phase 2 will be considered in relation to the
three factors that were systematically varied in the experiment design: (1) the
three different filter preparations; (2) the three transfer methods; and (3) the
three techniques of fiber identification, with the further contrast between the
standard and alternative methods of classifying fibers.
The Three Filter Compositions—
In the Phase 2 experiment design (Table 7), it was the intent to vary the
fiber composition in the preparation of the three filters, with essentially pure
chrysotile on the first filter, a mixture of chrysotile plus amosite on the
second filter, and a mixture of chrysotile plus crocidolite plus wollastonite
on the third filter. The results of EM examination confirm that this aim was
achieved (Table 23), the only evidence of contamination being the single amosite
fiber in a sample from the first filter, intended to be pure chrysotile (sample
2121). Based on the counts made by the alternative classification method, about
77
-------
. Table 27
PHASE 2 REGRESSION EQUATIONS
STANDARD METHOD OF FIBER CLASSIFICATION
(1) Y = 1.247 + O.
(2) Y12 = 2.629 + Q.704(X]L) + 0.117(X9D - 0.174(X13L) - 0.066(X13Q)
(3) Y3 = - 0.183 + 0.084(X1D + O.OSS^Q) + 0.038(X9D
(4) Y = - 2.907 + 0.116(X.,L) + O.OSSCX^)
(5) Y14 - - 5.283 + O.SlSCXjL) + 0.134(X1Q)
(6) Y13 = 0.523 - 0.062(X9Q) + 0.103(X13L) + 0
ALTERNATIVE METHOD OF FIBER CLASSIFICATION
(7) YU - 1.458 + 0.344(X1L)
(8) Y12 = 2.876 + 0.735(XjL) + 0.219(X^L)
(9) Y3 = - 0.274 + 0.108CX.JL) + 0.041(X1Q)
(10) YX = - 2.929 + O.ISSCX^) + 0.043(X1Q)
(11) Y14 = - 5.418 + 0.378(X1L) -t- 0.126(3^)
METHOD OF FIBER CLASSIFICATION NOT APPLICABLE
(12) Y15 = 1.791 + 0.824(X1L)
78
-------
Table 28
PROPERTIES OF PHASE 2 REGRESSION EQUATIONS
Equation Method of Fiber
Number Classification
1 Standard
2
3
4
5
6
7 Alternative
8
9
10
11
12 Inapplicable
Dependent
Variable
Yll
Y12
Y3
Yl
Y14
Y13
Yll
Y12
Y3
Yl
Y14
Y
Residual
Degrees of
Freedom
7
4
5
6
6
5
7
6
6
6
6
7
Residual
Standard
Deviation
0.1380
0.1276
0.0428
0.0664
0.1679
0.1002
0.1923
0.1099
0.0705
0.0683
0.1792
0.4487
R2
Percent
82
98
92
80
84
82
73
98
77
84
86
74
79
-------
99.8% of the fibers on the first filter were chrysotile; about 95.9% of the
fibers on the second filter were chrysotile and about 4.1% were amosite; on
the third filter about 47.5% of the fibers were chrysotile, 29.1% were croci-
dolite, and 23.4% were wollastonite.
The number and mass concentrations of chrysotile fibers on the three
filters varied substantially. This is evident from the estimates of these quan-
tities given in Table 24, with concurrence between the standard and alternative
classification methods. (Note that the first group of three samples came from
the first filter, the second group of three from the second filter, and the
third group of three from the third filter.) The pattern is a marked increase
in both the number and mass concentrations of chrysotile fibers in the progres-
sion from the first to the second to the third filter. This trend is also
clearly revealed by equations 1, 2, 7, and 8 (Table 27) in which the linear
variable associated with the filter preparations, XL, appears with positive
coefficients.
The Three Transfer Methods—
The three transfer methods employed in grid preparation were: (1) Soxhlet 1,
(2) Soxhlet 1 with carbon coating, and (3) Jaffe, also with carbon coating
(Table 7). The candidate coded independent variables representing possible
differences in performance of the transfer method are X_L and X.Q. The regres-
sion analysis revealed significant differences in relation to Y.-, i.e., natural
logarithm of chrysotile mass concentration. Variable XqL appears in both equa-
tions 2 and 8 with positive coefficients. The indicated effect is an increase
in the estimated mass concentration of chrysotile fibers as the transfer method
is changed from Soxhlet 1 to Soxlet 1 with carbon coating to Jaffe. The effect
is manifest regardless of the method of fiber classification.
Further effects of transfer method are significant in two of the equations
based on data in which the standard fiber classification method was employed,
i.e., equations 3 and 6 (see Table 27). The dependent variable in equation 3
is Y , the natural logarithm of the geometric mean length of chrysotile fibers.
The independent variable X»L is in the equation with a positive coefficient.
The effect brought out is a trend in the direction of increasing length of
chrysotile fibers in changing from Soxhlet 1 to Soxhlet 1 with carbon coating
to Jaffe. The dependent variable in equation 6 is Y , representing the per-
centage of fibers classified as exceptional under the standard classification
80
-------
method. In this equation, the independent variable XqQ (having coded values of
1, -2, and 1) is present with a negative coefficient. The indicated effect is
that the percentage of all fibers classified as exceptional tends to be higher
when the transfer method is Soxhlet 1 with carbon coating than when the trans-
fer method is either Jaffe or Soxhlet 1 without carbon coating.
The Three Fiber Identification Techniques—
The three techniques employed in identifying fiber types were: (1) mor-
phology plus X-ray fluorescence, (2) morphology plus electron diffraction, and
(3) morphology plus X-ray fluorescence plus electron diffraction (Table 7). The
candidate coded independent variables representing possible differences in per-
formance of the identification techniques are X _L and X _Q. In the regression
analyses, differential performance of the three techniques emerged as statis-
tically significant in two of the equations, 2 and 6 (Table 27). The dependent
variable in equation 2 is Y , natural logarithm of estimated mass concentration
of chrysotile fibers on the filter. Independent variables X ,L and X ,Q are in
the equation with negative coefficients. The pattern of effects is that the
third technique of fiber identification (morphology in conjunction with both
X-ray fluorescence and electron diffraction) tends to result in lower estimates
of chrysotile mass concentration than the first two techniques. Note, however,
that this effect was not significant when all fibers were assigned to the chemi-
cal species on the basis of the available evidence (alternative classification
method).
The other equation in which the performance of the identification tech-
niques differs has the dependent variable Y , which represents the percentage
of fibers classified as exceptional. In this equation, No. 6, both X 3L and
X _Q are included with positive coefficients. The pattern of effects is that
the third identification technique (morphology plus X-ray fluorescence plus
electron diffraction) results in the highest percentage of exceptional (ambig-
uous or other) fibers, the second technique (morphology plus electron diffrac-
tion) results in the lowest percentage of exceptional fibers, while the first
technique (morphology plus X-ray fluorescence) results in an intermediate per-
centage of exceptional fibers.
Selected Plots
Significant findings from the analysis of the Phase 2 data are illustrated
-,
by the confidence-interval plots of Figures 7 through 10. The estimated number
81
-------
0)
4J
4.0
3.5
3.0
2.5
0)
CH
2
2.0
1.5
1.0
0.5 -
I
I
2113 2121 2132 2211 2222 2233 2312 2323 2331
Phase 2 Samples (See Table 7)
Figure 7. Estimated number concentration of chrysotile fibers in the nine
Phase 2 samples (standard classification method), with 95%
confidence intervals.
82
-------
B,
o\
o
C
o
4-1
c
50
"
35
S 30
a
o
y
W or
» ^J
X
O
en
?*,
H
6
20
i c
15
10
Figure 8.
Solid Lines:
Dashed Lines:
Standard Method
Alternative Method
•
Tt
F
T
t
T T
•
I
I
l
o
CO
o co
CO 0
+ + T)
iH 4J (3
•H -H O
4J iH 4-1
,C O O
u »-i a
4-1
4J Cd
r-H O
O O
M
O
O
25
i—i w
•a°
4J 4-1
•H CO
>4-IO
O
>>
60
O f£
CX. X
o +
+ c
c o
E*. O -H
o' •«; o
1-1 o cd1
O"-CU' 'W
43 rH'.H-l
+
rn
r2
o
oo
O Wi -H
iH 4-1 4-1 I
O 0 0 X
J3 0) cfl
Filter
Composition
(a)
Transfer Method X
(b)
13*
Identification
Technique
(c)
Estimated mass concentration of chrysotile fibers in Phase 2 (standard
and alternative classification methods) in relation to filter
composition, transfer method, and identification technique, with 90%
confidence intervals.
83
-------
1.1
1.0
e
4-i
CO
c
• Q)
0.9
•H
d>
rH
•H
o 0.8
cc
I"
0.7
0.6
Tt
4-1
O
CO
01
CD 4-1
4-1 -H
01 T) C
r-l rH O
•H O 4-1
4-1 TJ CD
O -rl CO
CO U rH
•H (U
4-1 4J
O iH
CO CO
r"1 9
t_j C9
^4 £3
^J -
-------
50
40
30
20
10
o
OOO
c
o
CM ,a
4-1 td
0) U
3-S3
O -H O
01
<4-l
4-i
C8
OC
O
OX
^ O
00 C iH
O O 4->
r-l Vi O
O 4-J C8
43 O }-i
O. 01 M-l
Ij _,pj ^4*4
O W -H
>. O
00 C -H
O O W
O
O
,C
O >-i «
0) M-l »S
T-< M-I i
W-HX
X : Transfer Method
y (a)
Identification Technique
(b)
Figure 10. Estimated percent of all Phase 2 fibers that were exceptional
(ambiguous or other by the standard classification method) in
relation to transfer method and identification technique,
with 90% confidence intervals.
85
-------
concentrations of chrysotile fibers in the nine samples (standard classification
method) are shown in Figure 7. The confidence intervals were computed at the
95% probability level, assuming a Polsson distribution of fibers. The increas-
ing number concentration of chrysotile fibers in the successive filter prepara-
tions is clearly evident. Also apparent is substantial variation in the esti-
mates of the number concentration between samples within filters.
In Figures 8^ 9, and 10, the plotted values were computed from the Phase 2
regression equations, showing the effects of the controlled experimental factors.
Each point estimate has an associated 90% confidence interval. First, the
point values and confidence limits of the various dependent variables, i.e.,
Y „, Y_, and Y ,, were computed for specified combinations of values of the inde-
_L^ J J..J
pendent variables in the equations. Then the computed values of the dependent
variables were converted back to customary physical units by inverse transfor-
mations, e.g., the exponential transformation of values of the logarithmic
dependent variables.
Significant differences in estimated mass concentrations of chrysotile are
displayed in Figure 8. Results obtained using the standard method of fiber
classification, derived from equation 2, are denoted by the confidence inter-
vals drawn as solid lines. Results obtained using the alternative (forced)
method of fiber classification, derived from equation 8, are denoted by the
dashed-line confidence intervals. There is a very marked increase in chrysotile
mass concentration in the three successive filter preparations, based on either
the standard or alternative method of fiber classification. The significant
increase in the estimated chrysotile mass concentration as the transfer method
is changed from Soxhlet 1 without carbon deposition to Soxhlet 1 with carbon
deposition, and finally to Jaffe, is also evident regardless of fiber classifi-
cation method. With respect to Identification technique, the significant con-
trast is between the third technique (morphology together with both X-ray
fluorescence and electron diffraction) versus the other two when the standard
classification method is used: under these conditions the requirement of a
consensus of morphological, X-ray, and electron evidence understandably tends
to result in a lower estimate of the chrysotile mass concentration. The con-
trast disappears, however, when the exceptional fibers are assigned to the most
probable chemical species.
Figure 9 illustrates the significant contrasts that are implicit in equa-
tion 3. Geometric mean length of chrysotile fibers was smallest in the second
86
-------
filter preparation, largest in the third. Estimated geometric mean fiber length
increased in step with the three qualitative levels of transfer method, with the
Jaffe method indicated to result in somewhat the longest fibers.
Factors influencing the percent of all fibers classified as ambiguous or
other under the standard classification method, are illustrated in Figure 10,
based on equation 6. Considering the three transfer methods tested, Soxhlet 1
with carbon deposition appears to result in a somewhat higher percentage of
exceptional fibers than the other two methods. The identification technique of
electron diffraction in conjunction with fiber morphology resulted in the lowest
percentage of exceptional fibers, the combined technique of both X-ray fluores-
cence and electron diffraction in conjunction with morphology resulted in the
highest percentage.
Summary
The Phase 2 data support the choice of the Jaffe method of transfer of
fibers to the EM grid and the choice of electron diffraction plus morphology as
the technique for fiber identification.
RESULTS AND DISCUSSION OF PHASE 3
Phase 3 Objectives
Phase 3 evaluated the capabilities of two instruments working in the secon-
dary electron imaging (SEM) mode. Both instruments were capable of analyzing
small particles by means of an X-ray fluorescence probe. The instruments were
the JEOL JSM 50A, a modern high quality instrument designed primarily for SEM
operation; and the JEOL 100C analytical electron microscope. Identical areas
on marker grids were observed using the two microscopes. A pseudo-random
sequence was used in analyzing samples to avoid biases. The data from Phase 3
are summarized in Table 29.
Discussion of Phase 3 Results
The results from Phase 3 were not subjected to statistical analysis because
the information needed from Phase 3 could be obtained by a less rigorous evalua-
tion of the data. Additionally, because of the need for key-punching, computer
runs, and statistical analysis, the total analysis becomes very time consuming.
The results from tests 1, 2, 5, and 6 gave the difference in the values
obtained for fiber counts when the identical areas were observed using the
JEOL 100C and JSM 50A instruments, both in SEM mode. In the two instances where
87
-------
Table 29
SUMMARY OF PHASE 3 DATA
Test
No.
]
2
3
1*
5
6
oo 7
00
8
9
Compos i t ion1
1
1
1
2
2
2
3
3
3
Identification
Method2
M
M+X
M
M+X
M
M
M
M
M+X
Instrument
Used for
SEM
100C
50A
IOOC
100C
IOOC
50A
50A
IOOC
100C3
Total Area
Examined
x lO-W
2.16*
2.16*
2.16
1.44
1 . 44**
] . /,!,**
0.72
0.72***
0.72***
Total No.
of Fibers
104
88
127
50
114
87
42
126
153
Number of Fibers of Each Type of Fiber
Chrysotile Amosite Crocidolite Wollastonitt
104
58
127
23 3
106 8
74 13
31 65
106 12 8
58 53 9
Ambiguous
24
33
Mixture Composition
1. Chrysot ile
2. Chrysotile and Amosite
3. Chrysotile and Crocidolite
and Wollastoni te
2 M = Morphology
X = X-ray Analysis
3 IOOC used in STEM Mode
*,**,*** Identical
Areas Examined
-------
a direct comparison was made, the JEOL 100C gave higher number of fibers; 18%
higher from tests 1 and 2, and 31% higher from tests 5 and 6. A further com-
parison was made between the use of the scanning transmission mode (STEM),
test 9, and the SEM mode, test 8, both measuring fibers ,on identical areas using
the JEOL 100C instrument. The test showed a significant improvement on the
fiber count, 21%, when using the STEM mode. The reason for the increase in the
fiber count was observed under the JSM 50A and JEOL 100C using the SEM mode and
the JEOL 100C using the STEM mode probably results from the respective resolu-
tion capabilities of the instruments. The claimed resolution limits of the
JSM 50A and JEOL 100C in SEM mode and the JEOL 100C in STEM mode are claimed to
be 7 nm, 4 nm, and 2 .nm, respectively. In addition, the STEM mode gives an
image on the fluorescent screen with higher contrast and consequently fibers
are more obvious. One reason for the improved resolution results from the
higher accelerating voltage (100 kv) used with the JEOL 100C as opposed to the
40 kv used with the JSM 50A.
Tests were made to consider the difference in the fiber counts when differ-
ent areas were observed using the same instrument. From tests 1 and 3 using the
JEOL 100C, it can be seen that from different areas (openings) of the same grid
gave results which varied from 104 fibers in test 1, to 127 fibers for test 3,
a difference of 22%.
In combination, the results obtained from using different instruments and
observing different areas (openings) of the same grid, the difference in the
results can be very large indeed. Tests 2 and 3 compared the results obtained
from the JEOL 100C and JSM 50A when different areas (openings) of the same grid
were interrogated; test 2 gave 88 fibers while test 3 gave 127 fibers, a dif-
ference of 45%. Similarly, test 4 gave 50 fibers while test 5 gave 114 fibers,
a difference of 128%. Again, test 7 gave 42 fibers while test 8 gave 126 fibers,
a difference of 200%.
It should be noted that in every case, the results from the JSM 50A were
lower than when the JEOL 100C was used (see Table 29).
A further test was made to compare the results from the JEOL 100C in STEM
mode with those from a different area on the same grid using the SEM mode. The
results from tests 9 and 7, respectively, showed a very large, 264%, increase
in fiber counts being noted when the STEM mode was used.
89
-------
Tests 2, 4, and 9 indicate that when X-ray analysis is used for fiber type
identification, a substantial proportion, 34%, 48%, and 22%, respectively, of
all the fibers cannot be classified. This is because the X-ray yield from the
fiber is too low, or is ambiguous. The results do not indicate any trends in
terms of the instrument used or the fiber type to be identified. It should be
noted, however, that more detailed studies could reveal such trends.
A final test,'which was not part of the original Phase 3 effort, was added
because of its obvious pertinence to the overall objectives of the study. The
test evaluated the performance of the superior SEM instrument, the JEOL 100C,
against the same instrument operating in the conventional TEM mode. Six iden-
tical grid openings were scanned in both SEM and TEM modes; the results are
given in Table 30. Fibers were recognized by morphology alone. It can be seen
that the TEM mode gives consistently higher fiber counts (with an average over
six grid openings) of plus 79%. Stationary image of the conventional TEM mode
is less fatiguing to the eye than a scanning image.
Application of t-test shows that this difference is quite significant
(t value of 2.27 is significant at 5% probability for 10 degrees of freedom).
Conclusions from Phase 3
The conclusions drawn from Phase 3 are as follows:
1. In secondary electron imaging mode (SEM), the higher resolution
JEOL 100C gave consistently higher values than the JSM 50A.
2. The X-ray probe analysis indicated that approximately one-third of
the fibers could not be positively identified even when laboratory
prepared samples were utilized. The JSM 50A X-ray probe was easier
to use than the JEOL 100C in that it gave higher count rates and
could be operated at a lower tilt angle.
3. Higher fiber counts are obtained with higher resolution equipment.
.When compared with the JEOL 100C, SEM, the counts were improved by
21% when switching to the STEM mode and 79% (average of six) in the
conventional TEM mode. Thus, the conventional TEM is the desired
mode for asbestos analysis.
PHASE 4 RESULTS
Phase 4 was designed to evaluate parameters of ashing, ultrasonification,
and reconstitution of samples.
Table 31 summarizes results from Phase 4 for fiber number concentration
and mass concentration. Table 32 summarizes results on dimensions of observed
90
-------
Table 30
DIFFERENCE IN NUMBER OF CHRYSOTILE FIBERS COUNTED
WHEN SAME GRID OPENINGS ARE OBSERVED UNDER SEM
AND CONVENTIONAL TEM MODE IN JEOL 100C
No. No.
1
2
3
4
5
6
AVERAGE VALUES
STD. DEVIATIONS
STD. EFFOR OF
MEAN
SEM1
Fibers
48
18
38
31
48
48
38.5
12.23
4.99
CTEM2
No. Fibers
100
24
62
51
104
72
68.8
30.31
12.37
*
Increase
TEM Over
SEM
108
33
63
65
116
50
79
1 SEM 100 kv, 0° tilt, 10,OOOX (secondary electron
mode).
2 CTEM 100 kv, 0° tilt, 16,OOOX (conventional
transmitted electron image).
91
-------
vo
NJ
Table 31
ESTIMATING CHRYSOTILE ASBESTOS IN PHASE 4 SAMPLES
(ASHING AND BONIFICATION EXPERIMENTS)
Sample
Numbert
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210*
1211*
Area
Examined
10-4 Cm2
3.2
6.4
1.92
1.28
5.12
1.92
1.28
2.56
4.48
2.56
2 . 56
Fiber Concentration
Estimate, 106/cm2
of Filter
0.4531
0.1672
0.6719
0.7891
0.1973
0.5573
0.8594
0.4023
0.2857
0.4922
0.3750
Mass Concentration
Estimate, 10~9 gm/cm2
of Filter
2.287
3.471
6.913
7.903
1.829
7.402
14.43
6.693
3.160
8.712
17.53
t For explanation of tb,e sample number, please see Tables 9 and 13.
* Unashed.
-------
OJ
Table 32
SIZE DISTRIBUTION CHARACTERISTICS OF CHRYSOTILE FIBERS IN PHASE 4 SAMPLES
(ASHING AND BONIFICATION EXPERIMENTS)
Length ym
Sample
Humbert
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210*
1211*
Mean
0
1
0
0
0
0
1
0
1
1
.6341
.0135
.8819
. 9039
.9689
.9661
.9386
.0238
.6818
.2255
.6383
Std.
Dev.
0.3659
2.2693
1.0231
1.1187
0.6788
1.6861
0,7594
0.8998
0.3960
1.2621
2.7305
Mean
Ln (Length)
-0.5832
-0.4734
-0.3766
-0.3483
-0.2458
-0.4686
-0.2776
-0.2653
-0.5319
-0.1019
0.1311
Diameter
Mean
0.0556
0.0552
0.0601
0.0600
0.0588
0.0575
0.0672
0.0663
0.0636
0.0430
0.0572
Std.
Dev.
0.0260
0.0432
0.0391
0.0364
0.0311
0.0358
0.0417
0.0406
0.0411
0.0308
0.0417
pm
Mass fLO"14
Mean
Ln (Dia.)
-2
-3
-2
-2
-2
-3
-2
-2
-2
-3
-3
.9906
.0887
.9847
.9604
.9562
.0003
.8544
.8615
.9089
.2829
.0089
Mean
0.5048
2.076
1.029
1.002
0.9270
1 . 328
1.679
1.663
1.106
1.770
4.676
Std.
Dev.
0.6670
8.394
2.812
2.413
1.382
4.402
4.031
4.438
3.634
12.79
36.26
gm)
Mean
Ln (Mass)
-33.4816
-33.5678
-33.2831
-33.1862
-33.0753
-33.3863
-32.9035
-32.9053
-33.2669
-33.5848
-32.5261
t For explanation of the sample number, please see Tables 9 and 13.
* Unashed.
-------
fibers in Phase 4. Tables 31 and 32 also list results on two initial filters,
one polycarbonate and one cellulose acetate, studied without the ashing and
reconstitution step.
Dependent Variables in Phase 4
The dependent variables in Phase 4 are as follows.
DEPENDENT VARIABLES, PHASE 4
Variable _ Definition _
Y - Square root of estimated number of chrysotile
fibers per square centimeter of filter
Natural log of estimated mass concentration
of chrysotile fib
square centimeter
.«
of chrysotile fibers on filter, nanograms per
Y_ Mean of natural log of chrysotile fiber
lengths (ym)
Regression Analysis
The dependent variable Y was subjected to a square root transformation
and Y - to a log transformation to normalize the distributions. Similarly,
variable Y» was chosen as the mean of the natural log of length of the indivi-
dual fibers. -
Regressions were performed on each of the dependent variables using step-
wise regression program BMD02R from the BMD library of statistical programs.
The data input to the program included coded values of the independent vari-
ables found in Table 9 and the values of dependent variables listed in Table 33.
The mean values and standard deviation of the dependent variables are
listed in Table 34. The resulting regression equations are given in Table 35.
In any given equation, only those independent variables appear that have coef-
fients significant at the 20% probability level.
Each of the equations describes a relationship between a dependent variable
and the various independent variables.
The net effects and their confidence limits are shown graphically in
Figures 11 and 12.
94
-------
Table 33
VALUES OF DEPENDENT VARIABLES IN PHASE 4
Fiber Concentration
Sample
Number*
1201
1202
1203
1204
1205
1206
1207
1208
1209
106 Fibers
2
per cm
0.4531
0.1672
0.6719
0.7891
0.1973
0.5573
0.8594
0.4023
0.2857
Square Root
of Fiber
Concentration
Yn
l-L
0.6731
0.4089
0.8197
0.8833
0.4442
0.7465
0.9270
0.6343
0.4537
Mass
10~8 Gran
per cm
0.2287
0.3471
0.6913
0.7903
0.1829
0.7402
1.443
0.6693
0.3160
Concentration
Natural Log
of Mass
is Concentration
V
LZ.
-1.4753
-1.0581
-0.3692
-0.2353
-1.6988
-0.3008
0.3667
-0.4015
-1.1520
Mean Ln
(length)
YT
j
-0.5832
-0.4734
-0.3966
-0.3483
-0.2458
-0.4686
-0.2776
-0.2653
-0.5319
* For explanation of the sample number, please see Tables 9 and 13.
95
-------
Table 34
MEANS AND STANDARD DEVIATIONS OF DEPENDENT VARIABLES IN PHASE 4
Regression Standard
Equation Variable Mean Deviation
6 Y Square root of chrysotile 0.6662 0.1966
fiber concentration
(million fibers/cm2
filter)
7 Y Natural log of mass -0.7027 0.6748
concentration of chryso-
tile (10~9 gm/cm2 filter)
8 Y Mean log of chrysotile -0.399 0.1229
fiber length (micrometers)
Table 35
REGRESSION EQUATIONS IN PHASE 4
Phase 4
(6) Yn - 0.6619 - 0.0781(X15L) + 0.0852(X Q)
(7) Y12 - - 0.7027
(8) Y3 = - 0.3990 + 0.0478(XUQ) - 0.0354(X15Q)
96
-------
vD
•vj
C X""N
6 H
•H CD
1 1 Ik
4-J 4-J
tfl rH
If l^j
o c
•M Q) M-l
0 O
C C
O O C"N
•HUB
+J 0
3 !-i
fi 0) ^
•H ,Q OJ
M -H O.
Jj f-^.
C ca
O *|i I 1 1
u o
-------
VO
oo
60
O
C
CO / — s
gj E
S— i
j- •-
o
4-1 43
4-1
C 60
o c
•H V
4J h4
42 H
•H
-------
Interpretation of Phase 4 Results
Square Root of Fiber Concentration of Chrysotile—
Figure 11 shows that high energy ultrasonification increases the fiber
concentration as compared with the medium energy ultrasonification. This is
most probably due to the disruption of fibrils into shorter fibrils as
evidenced by the decrease in the mean fiber length (see Figure 12).
It is difficult to rationalize why low energy ultrasonification (which
is conducted in a different sonifying equipment than the medium and high
energy ultrasonification) should give the highest estimate of chrysotile
fiber concentration and also the smallest mean fiber length.
Figure 12 shows that high temperature ashing results in shorter fibrils
as compared with the low temperature ashing, presumably because of the vio-
lent nature of high temperature ashing. This conclusion may be interpreted
to mean that low temperature ashing is preferable to the high temperature
ashing. Also from Table 32, it is clear that mean fiber length is larger for
the unashed samples than that for both low temperature ashing and high tempera-
ture ashing. This suggests that ashing may be shortening the fiber length.
Cumulative Size Distributions for Chrysotile Fibers—
The mean fiber length and width are susceptible to a large variation if a
few large fibers are present in the group of fibers observed. Hence, a cumu-
lative distribution of length and width for all eleven samples are listed in
Tables 36 and 37, respectively.
It is clear that the maximum value of length of fiber in each sample
is different, but minimum length is almost constant. The different percentile
10th, 20th, . . . 90th values show that the length is always lower (regardless
of which ashing and ultrasonification treatment is employed) than that for the
unashed samples. Thus, qualitatively one can conclude that the ashing and
ultrasonification treatments chosen in this study lead to a shortening of the
length of fibers and hence should be used very cautiously.
From Table 37, it appears that width does not show much change or any
consistent trends of alteration.
Number Concentration of Chrysotile Fibers—
In the data, we have nine samples studied with various combinations of
ashing and sonification treatment and two samples studied directly (without
99
-------
o
o
Table 36
CHARACTERISTICS OF FIBER LENGTH IN CUMULATIVE DISTRIBUTION IN PHASE 4
Sample Code
Length (urn) :
Minimum
Maximum
Median
Maximum Length for
(percentile) :
10th*
20th
30th
40th
50th
60th
70th
80th
90th
1201
2,0
20.0
4.129
2.0
2.933
2.99
4.035
4.129
4.865
5.009
5.943
8.171
1202
1.0
175.0
4.115
1.966
2.853
2.993
3.869
4.115
4.968
6.153
8.328
15.00
1203
.2.0
80.0
4.945
2.00
2.987
3.905
4.063
4.945
5.713
6.723
9.486
12.119
1204
2.0
80.0
5.018
2.965
3.091
4.000
4.144
5.018
5.685
6.649
7.962
9.982
1205
2.0
30.0
6.011
3.011
3.153
4.086
4.973
6.011
6.786
7.994
11.779
14.744
1206
1.0
100.0
3.957
2.000
3.036
3.173
3.836
3.957
4.799
5.042
7.329
11.569
1207
2.0
40.0
5.143
2.927
3.887
4.005
4.947
5.143
6.073
8.041
9.753
14.445
1208
2.0
40.0
5.948
2.000
2.948
3.915
4.999
5.948
6.237
8.214
10.011
19.627
1209
1.0
20.0
4.212
2.018
3.001
3.095
4.068
4.212
4.934
5.844
7.800
8.121
1210
1.0
80.0
7.543
2.922
3.924
5.035
5.791
7.548
8.020
10.308
12.248
15.019
1211
2.0
200.0
8.49
3.925
4.827
5.757
8.125
8.492
9.765
11.532
14.042
21.136
* 10th refers to the 10th percentile of the distribution. The numbers in this row refer to the
maximum length in each sample, for the IQth percentile.
-------
Table 37
CHARACTERISTICS OF FIBER WIDTH IN CUMULATIVE DISTRIBUTION IN PHASE 4
Sample Code
Width (ym) :
Minimum
Maximum
Median
Maximum Width for
(percentile) :
10th*
20th
30th
40th
50th
60th
70th
80th
90th
1201
0.2
1.0
0.397
0.200
0.295
0.298
0.302
0.397
0.404
0.501
0.604
0.802
1202
0.1
2.0
0.305
0.197
0.201
0.295
0.300
0.305
0.311
0.409
0.572
0.976
1203
0.1
2.0
0.402
0.197
0.203
0.298
0.306
0.402
0.416
0.491
0.688
0.984
1204
0.1
2.0
0.400
0.199
0.295
0.301
0.307
0.400
0.415
0.503
0.603
0.803
1205
0.2
1.2
0.399
0.200
0.298
0.302
0.306
0.399
0.406
0.500
0.598
0.804
1206
0.100
2.000
0.398
0.199
0.294
0.300
0.306
0.398
0.412
0.482
0.584
0.798
1207
0.200
2.000
0.415
0.200
0.291
0.299
0.405
0.415
0.495
0.589
0.801
0.831
1208
0.200
2.000
0.411
0.200
0.293
0.299
0.404
0.411
0.484
0.582
0.800
0.833
1209
0.200
2.000
0.408
0.200
0.291
0.296
0.302
0.408
0.481
0.581
0.599
0.825
1210
0.2
2.0
0.291
0.2
0.2
0.2
0.2
0.291
0.295
0.300
0.408
0.493
1211
0.2
2.0
0.30
0.2
0.2
0.2
0.29
0.300
0.408
0.488
0.596
0.971
* 10th refers to the 10th percentile of the distribution. The numbers in this row
refer to the maximum length in each sample for the 10th percentile.
-------
ashing). We may compare the chrysotile fiber number concentration in these
two groups.
CHRYSOTILE FIBER NUMBER CONCENTRATION 106/cm2 in
ASHED SAMPLES AND IN UNASKED SAMPLES
Nine Ashed Samples " Two Unashed Samples
Mean Value of Fiber
Number Concentration
Standard Deviation
Standard Error of Mean
Variance
0.4870
0.2510
0.0837
0.00700
0.4336
0.0829
0.0586
0.00343
t . 0.4840 - 0.4336 = c
/O.00700 + 0.00343
This t-value is not statistically significant with 9 degrees of freedom, thus
indicating that the slight increase in chrysotile number concentration in the
ashed samples compared with the unashed samples is unimportant.
It should be noted here that in preparing these filters, carbon-coating was
I
not used, because some filters were cellulose acetate. Thus, the fibers on the
Phase 4 filters were not locked and during the Jaffe wash may have resulted in
a variable loss. This possibility may have made the evaluation of ashing and
Bonification variables difficult.
Although the length distribution data suggest that ashing and ultrasonifi-
cation treatments tend to decrease the length of fibers, the fiber concentra-
tion estimates suggest that the effect of ashing is not significant.
102
-------
MASS CONCENTRATION OF CHRYSOTILE nanogram/cm2 IN
ASHED SAMPLES AND UNASHED SAMPLES
Nine Ashed Samples Two Unashed Samples
Mean of Mass
Concentration,
nanogram/cm^
Standard Deviation
Standard Error of Mean
2
(Standard Error)
6.0098
3.9308
1.3103
1.71680
13.121
6.2353
4.4090
19.43965
fc 13.121 - 6.0098 ..
/I.71680 + 19.43965
This t-value for 9 degrees of freedom is significant at a probability
between 10 and 20%. It indicates that the average mass concentration of chryso-
tile in the ashed samples is lower than that in the unashed samples. This sug-
gests some loss of chrysotile during ashing and reconstitution step.
Clearly, more work is required to establish, quantitatively, the effects of
ashing and sonification treatment. Redeposition filters should be polycarbonate
and these should be coated with evaporated carbon to lock all particulates prior
to Jaffe wash.
One possible explanation for the failure to detect strong alteration in
fiber characteristics due to ashing subprocedure is that the initial stock solu-
tions had been subjected to high energy ultrasonics to break down the fibers to
a small enough stable size. Thus, a subsequent ashing and ultrasonic treatment
had only a marginal effect on the fiber dimensions. One needs a sample that has
not seen prior ultrasonic treatment.
Another possible area for future work is the effect of diluting the sample
(without ashing). This can be done by dissolving the primary filter with
particulate matter in a suitable solvent and then to redeposit it, after appro-
priate dilution, onto a polycarbonate filter.
Phase 4 Conclusions
1. Ashing and ultrasonic treatments should be used only when direct
sample preparation and examination are not possible. These cases
103
-------
include presence of organic matter or high total particle density on
primary filter.
2. Since ashing and reconstitution involves elaborate procedure, much
care is necessary in handling the products.
STATISTICAL ANALYSIS OF PHASE 5 DATA
The Experiment Design
In Phase 5, two independent variables were used: X&, the orientation of
the droplet during drying (face up or down), and X^, the radial location of
the grid opening (position used: center, mid-radius, or periphery). A full
factorial experiment was run, with the design indicated in Table 10. The vari-
ous levels of the independent variables and their codes were also listed in
Table 10. The dependent variables considered were: mean Ln liber length and
square root of fiber concentration. The values of these in the various cases
are given in Table 38, along with their means and standard deviations. These
data were based on the results of computer analysis given in Table 39, similar
to that used in Phases 2 and 4.
Regression Analysis
As in Phases 2 and 4, the dependent variable, Y . , fiber concentration, was
subjected to a square root transformation since it is essentially a count of
fibers. The dependent variable Y was chosen as mean Ln fiber length. The Ln
transformation was used to "smooth out" the large variations usually found in
length measurements and to normalize the distribution.
Regressions were performed on each of the dependent variables using the
stepwise regression program BMD02R from the BMD library of statistical programs
[59]. The data input to the program included the coded variables of the inde-
pendent variables found in Table 10 and the values of the dependent variables
from Table 38.
The resulting regression equations are given below.
REGRESSION EQUATIONS - PHASE 5
0.5024 - 0.11903(X6) + 0.0514(XUQ)
- 0.6025
104
-------
Table 38
VALUES OF DEPENDENT VARIABLES IN PHASE 5
Variables
Sample
Number
5101
5102
5103
5104
5105
5106
Square Root of
Fiber Concentrations
0.6124
0.4815
0.7705
0.4507
0.3176
0.3818
Mean Ln
Fiber Length
-0.6069
-0.5058
-0.6453
-0.8795
-0.5469
-0.4307
Mean
Standard Deviation
0.5024
0.1648
-0.6025
-0.1553
105
-------
Table 39
FIBER NUMBER AND MASS CONCENTRATION IN PHASE 5
Input mass: 10.3 x 10"9 gm /cm2
Fiber Count
(fibers)
Fiber Concentration
(10 fibers per sq. cm.)
Mass Concentration
_Q
(10 gm per sq. cm. )
Length Mean
(pm) Standard Deviation
Mean Ln
Geom Mean
Diameter Mean
urn) Standard Deviation
Mean Ln
Geom Mean
Mass Mean
(10~14 gm) Standard Deviation
Mean Ln
Geom Mean
5101
24
0.3750
1.963
0.7522
0.7375
-0,6069
0.5451
0.0368
0.0208
-3.4201
0.0327
0 . 5285
1.078
-34.3641
0.1191
5102
89
0.2318
1.312
0.7209
0.4562
-0.5058
0.6030
0.0468
0.0341
-3.2123
0.0403
0.5659
1.611
-33.8474
0.1997
5103
76
0.5937
2.116
0.6681
0.4853
-0.6453
0.5243
0.0412
0.0201
-3.2794 .
0.0377
0.3563
0.7563
-34.1212
0.1518
5104
39
0.2031
1.318
0.5295
0.4069
-0.8795
0.4150
0.0549
0.0301
-3.0318
0.5513
0.6489
1.301
-33.8603
0.1971
5105
71
0.1009
0.8827
0.7417
0.6876
-0.5469
0.5787
0.0613
0.0379
-2.9408
0.0528
0.8752
2.227
-33.3455
0.3298
5106
84
0.1458
1.073
0.7792
0.4776
-0.4307
0.6501
0.0555
0.0274
-3.0010
0.0497
0.7357
1.242
-33.3497
0.3284
-------
In any given equation, only those independent variables appear that have coeffi-
cients significant at the 20% probability level. That is, an independent varia-
ble did not enter an equation if it was determined that its coefficient value
could have occurred with probability 20% or greater due to accidents of sampling.
It is worthwhile to note each case which independent variables are in a given
equation and which ones are absent.
These net effects and their confidence limits are shown graphically in
Figure 13.
The results of the attempted regression of Y_, mean Ln fiber length, on
the independent variables showed no significant effect of any factor at the 80%
confidence level. That is, all the variation in mean Ln fiber length would
occur by chance with probability at least 20%.
Conclusions from Phase 5
Phase 5 investigated the effect of placing a drop of liquid containing
fibers in suspension directly onto an EM grid. The results indicated that sur-
face tension effects tended to move the fibers as the drop dried with the result
that an uneven fiber loading was observed on the grid.
In particular, as shown graphically in Figure 13, a grid allowed to dry
with the drop facing down gave higher fiber counts than when dried rlghtside up.
Further, a point on the mid-radius of the grid gave lower values than either the
periphery of the grid or the center of the grid.
For these reasons, the use of the direct drop method is not recommended.
107
-------
O
co
0.18
c
•H 0.16
•u
2 0.14
4-1
§ 0.12
u
o 0.10
n 0.08
o>
3 0.06
mc^ 0.04
o e
w-Ii 0.02
o co
£ « 0.00
CU -H
% *» -0.02
g& -0.04
o~ -0.06
g -0.08
tj -0.10
43
•H -0.12
i j
g -0.14
u
w -0.16
0)
» -0.18
-0.20
-L
< i
-
J_ _L
•V
••»
•B
•V
••
< )
•V
•••
II 1 1 1
-------
SECTION 9
PROVISIONAL OPTIMIZED METHOD AND ROUND-ROBIN TESTING
A provisional method was developed as based on the results of the five-
phase program. The subprocedures were adjusted to achieve the best precision
(see Table 21). Some adjustments were made to accommodate practical considera-
tions. A brief description of the optimized method is given in Appendix D.
Also, procedures or subprocedures which gave inconsistent results or which
altered the initial sample were eliminated from the optimized method. The
final draft of the detailed manual has been issued separately [64].
ROUND-ROBIN TESTS
In order to determine the ruggedness of the provisional optimized method,
a round-robin test was planned. This test was to serve as interlaboratory exper-
ience of the method and provide important feed back to further improve the op-
timized method. Of the several laboratories sought for participation, the fol-
lowing laboratories offered to cooperate.
1. Environmental Protection Agency, Research Triangle Park, NC
2. Environmental Protection Agency, Athens, GA
3. Environmental Protection Agency, Duluth, MN
4. National Institute for Occupational Safety and Health, Cincinnati, OH
5. A.D. Little, Cambridge, MA
6. Ontario Ministry of the Environment, Toronto, Ontario Canada
PREPARING CALIBRATION FILTERS
It would be ideal to prepare filters with known amounts of asbestos, known
size distributions of asbestos fibers, and to use these for round-robin tests.
This would serve as an independent external absolute calibration for evaluating
the absolute accuracy of the electron microscopy method. The concept of a fully
known and characterized sample proved elusive, since none of the methods could
109
-------
completely characterize a sample. Non-microscopic methods would not give size
distributions, parameters and fiber counts. The most one could expect to
achieve was to measure the total mass of asbestos per unit area of the filter.
Even this apparently easy measurement was quite difficult to achieve. Our
own repeated attempts at estimating chrysotile mass on filters of Phase 1, by
atomic absorption measurements, proved futile because of the very small mass
involved.
Being aware of the recent work on neutron activation technique at AERE,
Harwell, England, Dr. Morgan was asked to prepare six polycarbonate filters
with known small mass concentrations of asbestos. The method consisted of com-
pacting UICC asbestos into a disc, irradiating it with an intense flux of neu-
trons in a nuclear reactor, then eroding the disc with an air jet and collecting
the resulting aerosol on 47 mm diameter polycarbonate filters. The actual mass
of the asbestos deposited on the filter was measured by radio-activity measure-
ments of some short-lived isotopes of rare-earth elements present as trace
impurities. With some trial and error, Dr. Morgan succeeded in depositing
three different levels of mass concentration on these filters.
Our own work on two of these calibration filters showed that the filter
dissolution using Jaffe technique required much longer durations. Also, the
samples contained a wide size distribution of fibers and several bundles or
aggregates of fibers. One more serious problem with these samples was that it
was extremely difficult to obtain good electron diffraction patterns even at
100 kv. For these reasons, these calibration samples were discarded from con-
sideration in the round-robin test.
FINAL CHOICE OF SAMPLES FOR ROUND-ROBIN TEST
Following are two air samples we selected for round-robin test.
1. Air sample collected under controlled conditions in our laboratory
using aerosol generations in Phase 1. (High-volume samplers, poly-
carbonate filter, 0.4 Urn pore size, 708 liters/min for 13 minutes.)
Standard UICC chrysotile mineral was used for obtaining the aerosol
cloud.
2. A field air sample collected at the Johns-Manville Plant in Waukegan,
Illinois. (High-volume sampler, polycarbonate filter, 0.4 urn pore
size, 560 liters/min for one hour.)
The provision for two samples was meant to avoid total failure of the round-robin
test, if the field sample posed any unsurmountable problems.
110
-------
INDEPENDENT ESTIMATE OF CHRYSOTILE MASS CONCENTRATIONS
Neutron Activation Analysis^
Segments representing about half the areas of these filters were dispatched
to Dr. Morgan at AERE, Harwell, England, to estimate the chrysotile mass using
ultrasensitive neutron activation technique. Our letter is appended (see
Appendix E). Dr. Morgan's reply explaining the problems, and his inability to
estimate the low mass, is also appended (see Appendix E).
Fortunately, it was possible to obtain estimates of the chrysotile mass
concentration in the filter deposits by X-ray fluorescence spectrometric analysis
for magnesium. Fluorescence intensities above background of the Mg-k line were
measured with a simultaneous multi-wavelength spectrometer (Siemens MRS-3),
adapted for use with thin filter-deposited samples, using procedures described
by Wagman [65]. Values for chrysotile were derived from magnesium concentration
data, on the basis of the chemical formula Mg_Si 0 (OH),. Further detailes are
given in Appendix F.
The chrysotile mass concentration estimates on the two polycarbonate filters
used in the round-robin test, as determined by the X-ray fluorescence method in
Dr. Wagmen's X-ray laboratory, are as follows:
Chrysotile Mass Standard Deviation of Ratio
Concentration Mass Concentration Standard Error/Mean
Air Sample yg/m3 yg/m3 x 100-
Lab Sample 154
Field Sample 661
2.452
57.919
0.096
1.015
1.598
0.715
It is evident that the X-ray fluorescence method gives highly reproducible
results. The mass concentration estimate for the laboratory sample should be
fairly accurate, inasmuch as it consisted of high purity chrysotile. The esti-
mate for the field sample is likely to be too high because some of the magnesium
present is associated with materials other than chrysotile.
The filter segments were carbon-coated at IITRI laboratory tacked to the
bottom of disposable petri dishes and mailed to each of the participating
111
-------
laboratories along with specific instructions and with copies of the provisional'
method.
Dr. Anant Samudra visited the participating laboratories to (1) discuss and
demonstrate the fine aspects of the optimized method, (2) to explain the proper
use of the electron diffraction capability of the transmission electron micro-
scopes, and (3) to obtain criticism and comments on the provisional method. Most
of the electron microscope data were received later in the mail. These data were
reorganized using fortran coding forms and transferred to key-punch cards. The
data consisted of 54 sets and required 9,000 cards. The statistical descriptors
or characterizing parameters were derived for each data set, i.e., for each
separate TEM grid examined.
112
-------
SECTION 10
RESULTS AND DISCUSSION OF ROUND-ROBIN TESTS
The voluminous data from round-robin tests allow several analyses. It is
proposed to first check the Poisson distribution test and then to study the sum-
maries of statistical descriptors or characterizing parameters for the two air
samples.
POISSON DISTRIBUTION TESTS
Goodness of Fit with Poisson Distribution
This test requires the data about the number of fibers observed in a field
of view 0, 1, 2 ... etc., and the corresponding frequency of occurrence. These
data were extracted from the basic electron microscope data of the 54 data sets
mentioned earlier. The minimum data should consist of 40 fields of view and,
when arranged for fiber frequency, should give a minimum of three class intervals.
Following the method outlined in Phase 1 analysis, data from round-robin
tests were analyzed by the Poisson 1 program. The results are summarized in
Table 40. Of the 54 sets of data, 30 sets are in appropriate form for Poisson
distribution tests. Our of the 30 tests, 19 conformed definitely to the Poisson
distribution, seven are borderline cases, and only two tests definitely do not
conform to- the Poisson distribution. In data sets 76 and 77, tests cannot be
applied because they had only two class intervals and, hence, no degree of
freedom.
The finding that the majority of the sets conform to the Poisson distribu-
tion may be interpreted to mean that the differences in sample preparation and
sample contamination have not altered (or rearranged) the initial random settling
of fibers on the Nuclepore filter.
Confidence Intervals on the Mean Number Concentration
The Poisson distribution model allows computing intervals, for any given
degree of confience, on the value of X, the mean number of fibers per field.
113
-------
Table 40
TESTS FOR APPLICABILITY OF THE POISSON DISTRIBUTION TO NUMBER OF FIBERS PER FIELD
Data
Set
No.
1
2
3
4
5
6
7
8
9
18
19
21
22
23
24
25
26
31
41
42
43
45
73
74
75
76
77
86
90
91
Size of
Field
cm2xl(T6
0.18
0.725
0.18
0.235
0.235
0.235
0.235
0.235
0.235
0.309
0.309
0.187
0.187
0.187
0.187
0.187
0.187
0.72
0.19
0.472
0.472
0.18
O.lOSt
O.lOSt
O.lOSt
0.06 t
0.06 t
0.464t
O.lOSt
o.iosf
No. of
Fields
58
28
40
40
62
30
33
43
42
43
37
95
93
89
71
80
113
34
44
21
30
56
102
100
100
100
100
97
100
100
No. of
Fibers
106
106
114
110
108
75
104
105
109
101
78
141
103
106
105
105
104
154
98
101
100
102
27
31
44
19
8
107
59
39
Mean No.
Fibers per
Field, X
1.83
3.79
2.85
2.75
1.74
2.50
3.15
2*44
2.60
2.35
2.11
1.48
1.11
1.19
1.48
1.31
0.92
4.53
2.23
4.81
3.33
1.82
0,26
0.31
0.44
0.19
0.08
1.10
0.59
0.39
Degrees
i of
Freedom
3
4
3
3
3
3
4
4
4
4 .
3
3
2
2
3
3
2
4
4
2
4
3
1
1
1
0
0
2
1
1
Chi
Square
0.61
2.03
4.47-
5.73
10.3
1.35
1.18
8.67
7.77
6.65
8.28
8.59
7,77
1.77
2.09
10.3
25.3
1.39
1.61
0.63
3.33
1.31
6.16
3.40
3.42
0.12
0.40
13.9
4.14
0.30
Probability
0.90>P>0.80
0.80>PXh70
0.30>P>0.20
0.20>P>0.10
0.02>P>0.01
0.80>P>0.70
0.90>P>0.80
0.10>P>0.05
0.20>P>0*10
0.20>P>0.10
0.05>P>0.02
0.05>P>0.02
0.05>P>0.02
0.50>P>0.30
0.70>P>0.50
0.02>P>0.01
0.001>P
0.90>P>0.80
0,90>P>0.80
0.80>P>0.70
0.70>P>0.50
0.80>P>0.70
0.02>P>0.01
0.10>P>0.05
0.10>P>0.05
no test
no test
0.001>P
0.05>P>0.02
0.70>P>0.50
Good Fit
to Poisson
*
*
*
*
(*)
*
A
*
*
*
(*)
(*)
(*)
*
*
(*)
*
*
*
*
*
(*)
*
*
(*)
*
1 Fiber .per Field of
View Represents so much
Fiber Concentration,
106 fibers/m3
245.47
69.94
242.77
188.02
188.02
188.02
188.02
188.02
188.02
142.99
142.99
236.36
236.36
236.36
236.36
236.36
236.36
61*37
232.55
93.61
93.61
245.47
1135.63
1135.63
1135.63
1995.14
1995.14
257.95
1135.63
1135.63
t This takes into account the dilution factor in ashing also.
* Conform to Poisson.
(*) Borderline conformation to Poisson.
-------
In this study we have obtained 90% confidence as well as 95% confidence limits
on the mean X in each set since the size of the field varies from one set to
another, we normalized all these values of A and its confidence interval to give
the number concentrations in standard units (namely, 106 fibers/m3). The nor-
malized values are listed in Table 41 and are shown graphically for the labora-
tory air sample in Figure 14 for a quick comparison.
Precision Measured by Ratio of Standard Error of X to the Mean Value of X
Precision can be measured from the ratio of standard error to the mean to
the mean value of X. These ratios, expressed as a percentage, are also listed
in Table 41. These values appear quite comparable among different sets, i.e.,
for different laboratories and operators (with a few exceptions in the case of
the field air sample) the precision of the fiber count estimate is about 10%,
which is quite good.
GENERAL PROCEDURES
For each of the 54 data sets, statistical descriptors or characterizing
parameters were derived using a special fortran program. These characterizing
parameters are summarized for the laboratory air sample in Table 42 and for the
field air sample studied with ashing in Table 43 and for the field sample studied
without ashing in Table 44.
f\ ^
Column 3 lists - number concentration of all fibers, 10 /m
Columns 4 and 5 list - mean fiber length and mean fiber diameter, urn
-15 3
Column 6 lists - mean fiber volume, 10 cm
-9 3.3
Column 7 lists - volume concentration of all fibers, 10 cm /m
6 3
Column 8 lists - number concentration of chrysotile fibers, 10 /m
Columns 9 and 10 list - mean fiber length and mean fiber diameter, \im
Column 11 lists - mean fiber volume, 10 cm
3
Column 12 lists - chrysotile mass concentration in air, yg/m
In Tables 42 through 44, the numerical values in column 3 are slightly lower
.than the corresponding values listed in Table 41. This is because fibers cros-
sing the perimeter of the field of view have been treated as half fibers in
computing fiber concentration in Tables 42 through 44-
115
-------
Table 41
MEAN VALUES AND LOWER AND UPPER LIMITS OF FIBER CONCENTRATION
ESTIMATE ACCORDING TO POISSON DISTRIBUTION
Data
Set
SAMPLE
1
41
42
43
2
3
4
5
6
7
8
9
21
22
23
24
25
26
31
18
19
Lab
154
RTF
RTF
RTF
RTF
Athens
NIOSH
IITRI
IITRI
IITRI
IITRI
IITRI
IITRI
IITRI
IITRI
IITRI
IITRI
IITRI
IITRI
Duluth
Ontario
Ontario
Number Cone.
of all Fibers
106 Fibers/m3
Mean
448.72
517.89
450.27
312.01
230.72
691.90
517.06
327.53
470.05
592.64
459.15
487.91
350.76
261.90
281.51
349.58
310.10
(217.46)
277.93
355.89
301.43
any
J \J/O
Confidence
Lower
379.50
434.87
379.13
262.58
195.14
588.97
438.65
277.52
384.50
500.32
387.88
413.64
303.73
220.76
238.02
295.45
262.12
(183.65)
242.16
282.84
247.52
Limits
Upper ,
527.27
612.54
531.15
368.46
271.14
808.43
605.80
384.31
569.70
697.37
539.99
572.15
403.47
308.46
330.91
411.04
364.94
(255.99)
317.76
396.23
363.92
95?
jf *J n
Confidence
Lower
367.22
420.45
366.68
253.88
188.87
570.76
424.93
'268.68
369.65
484.15
375.48
"400.67
295.22
213.67
230.45
285.77
253.12
(177.74)
235.77
273.55
238.23
Limits
Upper
542.74
631.14
547.07
379.50
279.06
831.26
623.29
395.41
589.26
718.05
555.79
588.69
413.64
317.43
340.60
423.10
375.58
(263.54)
325.49
408.10
376.21
Std. Error
xlOO
Mean
10.26
9.67
9.48
9.47
10.47
7.06
14.60
12.52
11.14
9.07
12.21
10.18
10.67
11.67
11.06
10.41
12.02
(12.90)
7.75
12.10
13.73
45
SAMPLE
73
74
75
76
77
90
91
86
ADL
661
IITRI
IITRI
IITRI
IITRI
IITRI
IITRI
IITRI
Ontario
447.00
300.94
352.04
499.68
(379.08)
(159.61)
670.02
442.89
(284.52)
376.80
212.36
254.38
382.71
(247.40)
( 79.81)
533.74
332.74
(240.92)
527.03
414.50
474.69
642.76
(556.64)
(287.30)
832.41
578.03
(334.04)
364.52
197.60
239.62
363.40
(227.45)
( 67.83)
509.90
314.57
(233.18)
542.74
437.22
499.68
671.16
(592.56)
(315.23)
864.21
605.29
(343.84)
11.96
23.50
21.84
20.52
(24.47)
(42.37)
14.07
17.82
(15.47)
Numbers in brackets are considered tentative, since the data did not conform to Poisson
distribution or when the test for conformity could not. be applied.
116
-------
5
(U
o
ca
*J
m
«
850
800
750
700
650
600
550
500
450
400
350
300
250
200
150
Sample 154
tll
1414243 2 3 456789 212223242526 31 18 19 45
Data Set Number
Figure 14. 90% confidence intervals (inner) and 95% confidence intervals
on the mean fiber number concentration.
117
-------
Table 42
SUMMARY OF ROUND-ROBIN TEST RESULTS ON AIR SAMPLE 154
oo
......
0) 1 j 2 ; 3
o 1 Number
a ; Concent ration
a Data, 1 of All
a" Set i Fibers,
Code) Filter! 106/m3
[
I 4 154-2 434.8
1, 5 154-8 j 297.2
1 6 154-6 i 429.3
!• 7 154-3 | 547.0
1 8 154-1-A1 432.8
1 9 154-1-B 436.5
2 21 154-8 328.4
2 22 154-4 230.3
2 23 :154-5-A 268.3
2 24 '154-5-B 329.6
2 25 , 154-5-C 289.5
2 26 154-5-D 199.8
3 1 154-8 421.1
3 41 154-8-A 486.2
4 42 154-8-B 439.1
4 43 154-8-C 293.3
5+6 31 154-7 266.2
7 45 :i54-4-A 447.1
8 46A 154-4-B 37. 1
8 46B 154-4-B 55.1
8 47A 154-4 24.23
8 47B 154-4 16.15
; ' (ashed)
9 2-1 154-6A 209.2
9 2-2 154-6-B 215.5
10 3-1 154-3-A 623.5
10- 3-2 154-3-B 570.5
IV 18 154-2-A 310.9
11 19 154-2-B 274.4
11 L.., — ..L
" I
4 L 5
Size Distribution of
Mean
Length,
pm
1.061
0.780
0.838
0,770
0.714
0.894
0.995
1.385
0.943
0.934
1.062 i
0.987 ,
1.236 :
1.151
0.810 ,
0.888
1.384 ~*
0.873
1.352 '
1 .096
.209
.177
.739
.682
.104
.025
.032
.310 ,
Mean
,
6
All Fibers
Mean
Diameter, ! Volume,
pm 10~15 cm3
1
0.067
0.059
0.067
0.057
0.055
0.061
0.067
0.077 ;
0.060
0.070
0.067
0.072
0.042
0.049
0.049
0.046
0.053
0.060 _^
0.053
0.044 ^
0.076
0.061
0.064
0.068
0.044
0.052
0.046
0.040
4.301
2.161
3.472
2.284
2.297
3.117
4.778
7.520
3.320
5.557
4.243
5.966
1.638
2.045
1.945
2.020
3.995 ^
3.595
4.673
2.043
40.33
4.19
6.44
0.002
3.590
3.822
3. 179
5.613
....i i _..J
7
Volume
Concentration
of All
Fibers,
10~s c.m3/m3
1870.5
642.2
1490.5
- 1249.3
994.1
1360.6
1569.1
1729.6
890.7
1831.6
1228.3
1192.0
689.8
994.3
885.4
8
Number
Concentration
of
Chrysotile,
296.1
222.9
369.8
421.6
284.3
268.6
108.2
52.2
103.6
136.5
141.8
92.0
412.6
486.2
439.1
592.5 145.1
1063.5 I 140.8
1607.3
173.4
112.6
977.2
67.7
•
1347.2
1724.4
2238.4
2180.5
988.3
1540.2
355.1*
170.6
180.7
292.4
242.8
126.4
36.7
9
10
Size Distribution of
Mean
Length,
pm
1.316
0.846
0.912
0.874
0.876
1.064
1.318
2.017
1.091
1.161
1.291
1.035
1.251
1.151
0.810
Mean
Diameter,
pm
0.076
0.063
0.069
0.062
0.063
0.072
0.081
0.108
0.074
0.083
0.078
0.091
0.042
0.049
0.049
1.067 0.055
1.569 0.064
0.953 0.060
1.879 n
1.697
1.658
1.794
^ 1.298 ~*
1.602
„ —
•
0.069
0.070
0.068
0.086
0.056
0.038
' 11
Chrysotile
Mean
Vo 1 ume ,
ID'15 cm3
5.769
2.574
3.912
2.742
3.162
4.414
9.103
18.978
5.667
10.096
6.130
8.762
' 1.662
2.045
1.948
3.207
6.174
' 3.890
7.445
7.932
7.210
8.428
3.011
1.718
L
12
Chrysotile
Mass
Concent rat ion
in Air,
Pg/m3
,
4.442
1.492
3.761
3.006
2.391
3.083
2.561
2.570
1.526
3.584
2.260
2.097
1.783
2.585
2.224
1.210
! 2.270
r 3.591
_,
'''
3.303
3.726
5.482
5.320
0.989
0.164
1
* Morphology alone.
-------
Table 43
SUMMARY OF ROUND-ROBIN TEST RESULTS ON FIELD SAMPLE 661 (ALL SAMPLES ASHED)
-------
Table 44
SUMMARY OF ROUND-ROBIN TEST RESULTS ON FIELD SAMPLE 661 (ALL SAMPLES ANALYZED WITHOUT ASHING)
Operator Code
1
1
2
5
5
6
4
4
10
1
Data
Set
Code
71
72
80
83A
83B
84
53
54 1
50
2
Filter
661-4*
661-4*
661-4*
661-7*
661-7*
661-7*
661-1*
661-1*
661-3*
3 rrr ,
i i
Number
Concentration
of All
Fibers,
106/m3
79.7
78.0
21.4
71.4
53.14
53.1
23.6
33.2
43.2
6
J
Size Distribution of All Fibers
Mean
Length,
Um
3.067
.1.490
2.916
1.728
2.197
2.100
3.519
2.284
1.551
Mean
Diameter,
pm
0.167
0.134
0.178
0.138
0.119
0.091
0.280
0.151
0.109
Mean
Volume,
10- 1S cm3
293.0
53.43
124.87
58.07
95.61
31.90
1781.0
82.55
53.05
7
Volume
Concentration
of All
Fibers,
10~9 cm3/m3
23352.0
4167.5
2672.0
4146.2
5080.7
1693.9
42031.6
2740.7
2291.8
8
Number
Concentration
of
Chrysotile,
106/m3
34.0
47.3
9.3
36.53
18.27
23.3
18,6
27.6
11.2
9
10
n
Size Distribution of Chrysotile
Mean
Length,
\im
4.853
1.589
3.719
2.614
4.227
3.064
3.713
3.361
2.25
Mean
Diameter,
jam
0.171
0.131
0.145
0.170
0.163
0.098
0.313
0.151
0.120
Mean
Volume
10-15 cm3
328.1
63.03
90.32
83.54
218.5
48.53
2203.0
81.53
62 .'97
12
Chrysotile
Mass
Concentration
in Air,
Ug/ra3
29.035
7.755
2.193
7.935
10.37
2.933 j
106.7
5*483 __^
1-827
-------
INTERLABORATORY COMPARISONS
A direct and simple statistical method is follwed here for comparing
transmission electron microscope estimates for the two air samples examined in
the round-robin test.
Data from the different laboratories were considered in three types of
comparisons.
1. Data for the laboratory air sample (unashed) as summarized in
Table 42
2. Data for the field air sample with ashing as summarized in Table 43
(ashed only)
3. Data for the field air sample without ashing presented in Table 44
In each case, estimates of the following four parameters can be compared:
6 *?
1. Number concentrations of all fibers, 10 /m
—9 3 3
2. Volume concentrations of all fibers, 10 cm /m
6 3
3. Number concentrations of chrysotile fibers, 10 /m
3
4. Mass concentrations of chrysotile fibers, ugM
For each of these parameters, sample means and standard deviations were
computed for each operator. Confidence intervals about the mean at the 95%
level were calculated and graphed for comparison with the overall mean for each
case. These are listed in Tables 45, 46, and 47. The quantities listed in each
table are as follows:
Operator Code - 1 to 11
n - Number of observations (or grids examined)
t (y, n-1) - t-value at 95% confidence (a = 0.05) and n-1 degrees of
freedom
s - Standard deviation of the observed values
SEm - Standard error of the mean value
t'SEm - Interval for obtaining 95% confidence limits
x - Mean value
Lower - Lower limit at 95% confidence, i.e., x - f(SEm)
Upper - Upper limit at 95% confidence, i.e., x - f(SEm)
121
-------
Table 45
95% CONFIDENCE INTERVALS ON THE MEAN ESTIMATES FOR INDIVIDUAL OPERATORS IN AIR SAMPLE 154 (SEE TABLE 42)
Si
to
"T T"^~
^Operator Code' '• j
i :
|«i '6
E
1
2
,
6
t (|, n-1) 2.571; 2.571
— — — j '— 4-— -.
1
; Concentration] s 79.24
of All
Fibers SEm* 32-35
,(106/ra3) t-SEm 83.2
x 429.6
1 lower 346.4
upper 512.8
j
Volume s '. 421.33
Concentration' _
/IA~ 9 3/ 3\ SEm . j 172.01
V ^- / Hi /
! if SEm i 442.2
! |x .1267.8
J
lower ; 825.6
'upper 1710.0
j i
i
Concentration s 72.36
of Chrysotile s&n 2g ^
f SEm 75.95
x 310.6
lower 234.6
.
upper ' 386.5
i
j
52.42
i 21.40
55.0
I 274.3
219.3
329.3
362.04
147.80
c
380.0
1406.9
: 1026. 9
!l786.9
32.60
; 13.31
34.2
^
i 105.7
7.15
139.9
Chrysotile is 1.030) 0.682
"ass „. isEm 0.420
Concentration
(Mg/m1) in fSEm .1.081
Alr x 3.029
! lower 1 . 948
upper 4.110
1
0.278
0.716
2,433
1.717
3.149
1 1
I
\
2
12.706
46.03
32.55
413.6
, 453.6
40.0
1 867.2
215.31
152.25
i
1934.5
842.0
'Negative
2776.5
52.04
36.80
467.5
449,4
Negative
916.9
0.567
0.401
5.094
2.184
Negative
7.278
2
|
1
1 5&6
i 1
i
12.706,
103.10
72.90
926.3
366.2
Negative
1292.5
185.90
131.45
1670.2
724.0
Negative
2394.2
207.89
147.00
1867.8
292.1
ttegat i ve
2159.9
0.717
0.507
6.442
1.717
legatlve
8,159
/
i --.-
1
'.
f T" "]
; 1
7 j 8 ' 9
I 2-2
^
12.706 12.706
16.76 4.45
H.85 3.15
, 150.6 40.0
: 266.2 447.1 31.2 > 212.4
—
Negative 172.4
' — -- 181.8 252.4
i
197.21 266.72
i — 139.45 188.60
! -- — 1771.8 2396.3
1063.5 1067.3 292.6 1535.8
' J
i — ' — [Negative Negative
— ' — 2064.4 3932.1
j (
7.14
; ! — 5.05
64.1
1
140.8
355.1 — 175.6
' — , — — 111.5
; I
t
- , ,,
• — . 239.7
—
,
'
2,260
—
—
— - 0.299
— ' | ., 0.211
— ! -- ; 2.686
I
i 10
i
! 2
1
': 12.706
11
2
12.706
1
' 37.48 25.81'
26.50
18.25
: 336.7 231.9 ,
597.0 292.6
260.3 60.7
I
933.7 i 524.5
: 40.94
390.25
28.95 275.95
367.8 3506.2 ,
2209.4 i 1264. 2
.1841.6 Negative,
2577.2 !4770.4
! 1
!
: 35.07 63.43
: 24.80 44.85
i 315.1 569.9
267.6 i 81.6 '
Negative Negative;
' '
j 582.7 ' 651.6 ;
'. 1
AU 1
Operators
Combined
26
2.060
143.98
28.24
58.2
332.4
274.2
390.6
526.29
103.21
212.6
1248.3
1035.7
1460.9
131.01
26.74
55.3
230.2
174.9
285.5
f
0.115 0.583
0.081 0.412J
1.033) 5.238
3.591 — I' 3.515J 5.401
• 0.829
6.201
- .1
4,368
0.577
Negative
6,434 5.815
1
1.291
0.264
0.545
2.725
2.180
3.270
J
* SEm stands for standard error of the mean.
-------
Table 46
rO
95% CONFIDENCE INTERVALS ON THE MEAN ESTIMATES FOR INDIVIDUAL OPERATORS IN FIELD AIR SAMPLE 661
(SEE TABLE 43) (Ashed Samples Only)
! 1 f — r — ] •
1 i ; •
Operator Code; | 1
1 ' j '
2
•n j 9 2
t (|, n-l)i 2.306' 12.706
!
; Concentration s | 159.59 11.10
of All ;
Tibell SE°>* 53.20 ! 7.85
. (lOVm3) fSEm j 122.7 i 99.7
x 347.4 84.4
'lower ' 224.7 | Negative
'upper 470.1 184.1
i
Volume 's i 12260. 61 1899.43
Concentration. ; ,„„, „-, ; , .,,, .„
,,n-
-------
Table 47
95% CONFIDENCE INTERVALS ON THE MEAN ESTIMATES FOR INDIVIDUAL OPERATORS IN FIELD AIR
SAMPLE 661 (SEE TABLE 44) (Unashed Samples Only)
Operator Code
Concentration
of All
Fibers
(106/m3)
Volume
Concent rat ion
(10-9 cmVm3)
Concentration
of Chrysotile
(106/m9)
Mass
Concentration
Chrysotile
In Air
(Ug/ta3)
n
t
-------
For some operators, there is only one observation. In such a situation,
only the value is plotted, but no estimate for standard error is obtainable and
the confidence interval is undetermined. These are shown as the point values
with no confidence interval, where it should be noted that this represents a
confidence interval that is indefinite.
GRAPHICAL REPRESENTATION OF RESULTS
These results are graphed in Figures 15 (a, b, c, and d), 16 (a, b, c, and
d), and 17 (a, b, c, and d) for the observed parameters and in cases 1, 2, and
3 previously mentioned.
When the computed lower limit is negative, the confidence interval is
truncated at zero. This suggests that La-transformation should be used to avoid
non-positive values, in better agreement with the physical situation.
The overall mean in each case has been plotted as dashed lines to facilitate
direct comparison with estimates of individual operators. Laboratories and
operators whose confidence intervals overlap the overall averages can be con-
sidered in good agreement. Some operators have narrower confidence intervals
due to a larger number of replications.
The .results for Tables 43 and 44 (field air sample) showed greater varia-
tions than that in laboratory sample, and thus the change in scale for the plots
should be noted.
ACCURACY AND PRECISION OF ESTIMATES
Accuracy of TEM Chrysotile Mass Estimates
As described earlier, it is impossible to obtain complete characterization
of an asbestos sample by some method independent of electron microscopy= At the
most, only, the chrysotile mass estimate may be obtained by an independent method.
- '*V
Of the several approaches tried, only high-precision X-ray fluorescence spec-
trometry appeared useful [65]. The overall chrysotile mass estimates by the
electron microscope method on the two round-robin tests are compared below with
those by X-ray fluorescence analysis for magnesium content.
125
-------
Number
Concentration of
All Fibers (106/m3)
(a)
H-1 M
*• OO O N>
O O O O O
O Q O O O
200
0
5000
§
3^ 4000
ume
§"B ^ 3000
U U .O
C ^
u o 2000
•3 1000
0
Number
Concentration of
Chrysotile (106/m3)
(c)
Ml-1 NJ NJ
W O Ul O U1
O O O O C
OOP O O
0
g • 8.000
•H )->
4J -H
Q> BJ <£
MWB 6-°°°
*J C -H ^N
O 4) T3
>, erT ^ 4.000
W> 0 E
J= O ~-
O 60
m
'- I
1
1 1
i 2 :
4
I- J
- r i .
1 2
" i :
1 2
•
3 4
3 A
3 4
-•
i i 1 i i i
5&6 7 8 9 10 1
_ T I
1 1
i 566 7 8 9 10 1
T i ~~? 1
t 5&6 7 8 9 10 1
n -
i
L All Operators
T
"I
1
1 All Operators
f ,
1 T
1 All Operators
X-ray
Flour escence
T- /
Figure 15. 95% confidence intervals about the means in laboratory air
sample 154 (see Tables 42 and 45).
126
-------
1500
^ 1250
o°o 1000
O r-4
U 1-1 x^
§OOO
O O O
4J -H
te .
: f
. I. 4 .
^
• i ,
| - | { *
12 3456 89 11 All Operators
2
p
! T , I
77 ;
p
10
X— Tinv
Fluorescence
B
* ,
11
All Operators
Figure 16.
95% confidence intervals about the means in field air sample 661
(see Tables 43 and 46), ashed samples only.
127
-------
ro
°^ 100
C3 O
O i-l
VH i-l ^-I
(U 4J s~* _c
.0 ea co nj 75
S I* |i ^^^
3 u 01
z § 3 50
U E*4
e
0 rH
" < 25
0
c
o
•H
2^g
«j e x-s
0 CJ J3
tf \^^
oT 30,000
| S 20,000
=i 10,000
o
Number
mcentration of
rysotile (106/itt3
(c)
S , 8 .
> o o
0 £ °
0
o
H tJ
4J -rl
0) 41 •<
Sd c 60.0
§S"
-------
Electron Microscope Estimate X-ray Fluorescence
Chrysotile Mass Chrysotile Mass
Concentration in Air Concentration in Air
Air Sample yg/m3 yg/m3
Lab Sample 154 2.725 2.452
Field Sample 661 15.396 57.919
(ashed & unashed)
Field Sample 661 13.751 57.919
(ashed samples only)
As expected, the agreement is good for the laboratory air sample 154, but
poor for the field air sample. Direct comparison of results on field sample is
inappropriate because XRF measures all magnesium in Chrysotile, non-chrysotile,
and even non-fibrous minerals present. For the laboratory air sample, where
such interferences are avoided, the EM method is in good agreement with the
XRF method. These estimates from X-ray fluorescence method are also included in
Figures 15d, 16d, and 17d for direct comparison with estimates of individual TEM
operators.
Precision of TEM Estimates in Laboratory Air Sample
One way of comparing precision of individual operators is illustrated in
Table 48. Here, the ratio of the standard error to the mean value is expressed
as a percentage and used as a measure of precision. Smaller value of this ratio
signifies better precision.
From Table 48 the precision appears very good (less than 10%) for operators
1, 2, 3, 9, 10, and 11 for number concentrations of all fibers (see column 5).
The precision for operator 11 is poor (55% and 71%) for number concentrations
and mass concentrations of Chrysotile respectively (see columns 8 and 11). In
chrysotile fiber number concentration estimate, operator 11 is the lowest (,81.55)
and operator 3 the highest (449.4). The mean value of the estimate for all
6 3
operators is 230.0 x 10 fibers/m . In chrysotile mass concentration estimate,
operator 11 is the lowest (0.58) and operator 10 the highest (5.4). The mean
3
value of the estimate for all operators is 2.72 yg/m .
129
-------
Table 48
PRECISION OF FIBER CONCENTRATION ESTIMATES ON LABORATORY AIR SAMPLE 154
1
Operator
Code
1
2
3
4
5&6
7
8
8*
9
10
11
2
No.
of
Tests
6
6
2
2
1
1
1
1
2
2
2
3
4
5
All Fibers
Number Concentration
10 6 Fiber s /m3
Mean
429.6
274.26
453.65
266.20
266.2
447.1
43.1
19.4
212.35
597.0
292.65
Std.
Error
32.35
21.42
32.55
72.90
—
—
—
—
3.15
26.50
18.25
Std. Error ,_0
Mean
7.53
7.81
7.18
19.91
—
—
—
—
1.48
4.44
6.24
6
7
8
Chrysotile Fibers
Number Concentration
106 Flbers/m3
Mean
310.55
105.72
449.4
292.1
140.8
355.1
—
—
175.65
267.6
81.55
Std.
Error
29.54
13.31
36.80
147.00
—
—
—
—
5.05
24.80
44.85
Std. Error ,nn
Mean
9.51
12.59
8.19
50.32
— T
—
2.88
9.27
55,00
9 | 10
11
Chrysotile Fibers
Mass Concentration
yg/*3
Mean
3.029
2.433
2.184
1.717
2.260
3.591
—
—
3.515
5.401
0.577
Std.
Error
0.420
0.278
0.401
0.507
--
—
—
—
0.211
0.081
0.412
Std. .Error ...
Mean Kl°°
13.88
11.44
18.36
29.53
—
—
--
—
6.02
1.50
71.49
12
13
14
Percentage of Fibers
Identified by SAED
Mean
72.59
39.30
99.00
76.33
52.66
—
—
—
82.70
55.19
27.01
Std.
Error
3.489
3.953
1.000
23.670
—
—
—
~
1.140
6.515
13,630
Std. Error .
Mean
4.81
10.06
1.01
31.00
—
—
—
~
1.38
11.81
50.47
* Ashed and reconstituted.
-------
Precision of TEM Estimates in Field Air Sample
Precision as measured by the ratio of standard error to the mean value,
expressed as a percentage, on field sample is summarized in Tables 49 and 50.
The precision for chrysotile fiber number concentration in the ashed
samples varies from 5.69 to 43 (see column 8 in Table 49), and for chrysotile
mass concentration varies from 14.20 to 94.5 (see column 11 in Table 49).
The precision for chrysotile fiber number concentration in unashed samples
varies from 16.36 to 33.32 (see column 8 in Table 50), and for chrysotile mass
concentration varies from 13.33 to 90.24 (see column 11 in Table 50).
Comparison of Precision in Round-Robin Samples
Table 51 lists the average precision value and its standard deviation for
the laboratory sample and the field sample.
It is evident that the mean precision for chrysotile fiber number concen-
tration is almost the same for the laboratory sample and the field sample, ashed
as well as unashed. However, the mean precision for chrysotile mass concentra-
tion is much better for the laboratory sample (21.74) than that for field sample
(44.17 for ashed and 53.81 for unashed). In general, there is no difference in
mean values of precision between ashed and unashed field samples.
Ashed Field Sample - (See Table 49)
In chrysotile fiber number concentration, operator 4 is the lowest (12.07)
and operator 5 is the highest (238.5). The mean for all operators is
6 o
108 x 10 fibers/m . In chrysotile mass concentration, operator 8 is the
lowest (1.01) and operator 6 is the highest (39.21). The mean for all opera-
tors is 13.85 ug/m .
Unashed Field Sample - (See Table 50)
In the chrysotile fiber number concentration estimate, operator 2 is the
lowest (9.3) and operator 1 is the highest (40.65). The mean value of estimate
6 3
for all operators is 25.12 x 10 fibers/m . In the chrysotile mass concentration
estimate, operator 10 is the lowest (1.83) and operator 4 is the highest (56.09).
3
The mean value of the estimate for all operators is 19.36 yg/m .
The spread in estimates is much higher in the ashed field sample than in
the unashed field sample.
131
-------
Table 49
PRECISION OF FIBER CONCENTRATION ESTIMATES ON FIELD SAMPLE 661 (ALL SAMPLES ASHED)
1
Operator
Code
1
2
3
4
5
6
8
9
11
2
No.
of
Tests
9
2
3
1
2
2
3
2
2
3
4
5
All Fibers
Number Concentration
106 Fibers/m3
Mean
347.3
84.4
29.3
16.3
494.7
463.8
153.3
204.4
260.1
Std.
Error
53.20
7.85
10.52
3.03
61.86
83.91
11.06
4.10
0.55
Std. Error
., xiuu
Mean
15.31
9.30
35.86
18.59
20.99
18.09
7.21
2.00
0.21
6
7
8
Chrysotile Fibers
Number Concentration
106 Fibers/m3
Mean
171.8
36.9
25.4
12.0
238.5
167.8
21.4
143.0
69.6
Std.
Error
28.37
4.30
10.94
2.58
53.0
8.85
6.68
8.15
11.75
Std. Error
_. xiuu
Mean
16.52
11.65
43.00
21.37
22.22
5.27
31.11
5.69
16.01
9
10
11
Chrysotile Fibers
Mass Concentration
UR/m3
Mean
13.22
4.39
30.15
13.36
27.75
39.21
1.01
2.75
5.90
Std.
Error
9.28
2.13
4.29
3.48
19.62
37.07
0.16
1.11
1.05
Std. Error
.. xiuu
Mean
70.20
48.51
14.23
26.05
70.70
94.54
15.34
40.36
17.79
12
13
14
Percentage of Fibers
Identified by SAED
Mean
66.094
50.385
95.873
87.57
49.66
37.06
15.77
69.96
28.24
Std.
Error
4.287
0.385
3.327
7.35
2.72
4.80
5.42
2.60
4.28
Std. Error
„ XIUU
Mean
6.48
0.76
3.47
8.39
5.48
12.95
34.38
3.72
18.71
-------
Table 50
PRECISION OF FIBER CONCENTRATION ESTIMATES ON FIELD SAMPLE 661 (UNASHED)
1
Operator
Code
1*
2*
4*
5*
6*
10*
2
No.
of
Tests
2
1
2
2
1
1
3
4
5 '
All Fibers
Number Concentration
106 Fibers/m3
Mean
78.85
21.4
28.4
62.3
53.1
43.2
Std.
Error
0.85
—
4.80
9.13
—
—
Std. Error 1QO
Mean xl°°
1.07
' —
16.90
14.65
—
—
6
7
8
Chrysotile Fibers
Number Concentration
1Q6 Fibers/m3
Mean
40.65
9.3
23.6
27.4
23.3
11.2
Std.
Error
6.65
—
4.50
9.13
—
—
Std. Error
„ • '- XXUU
Mean
16,36
—
19.07
33.32
— r
'
9
10
11
Chrysotile Fibers
Mass Concentration
yg/m3
Mean
18.39
2.75
56.09
9.15
2.93
1.83
Std.
Error
10.64
—
50.61
1,22
—
—
Std. Error
' XJ-UU
Mean
57.85
—
90.24
'13.33
—
—
12
13
14
Percentage of Fibers
Identified by SAED
Mean
70.515
47.58
95.48
45.89
43.75
47.31
Std.
Error
1.365
—
2.521
5.270
—
—
Std. Error
Mean KiUU
1.94
—
2.64
11.48
—
—
OJ
* Unahsed direct transfer.
-------
Table 51
PRECISION OF DIFFERENT MEASUREMENTS IN THE TWO SAMPLES
All Fibers
Concentration
106/m3
Mean Value of
Precision*
Standard Deviation
of Precision
Laboratory Sample
Unashed
7.80
5.78
Field Sample
Ashed Unashed
14.17 10.87
11.00
8.56
Chrysotile
Fibers
106/m3
Chrysotile
Mass
Concent rat ion
yg/m3
i
Mean Value of
Precision**
Standard Deviation
of Precision
Mean Value of
Precision***
Standard Deviation
of Precision
21.11
21.79
21.74
23.71
-
19.20 22.92
12.11 9.11
44.17 53.81
28.95 38.61
* Obtained from mean of the values listed in column 5 of Tables 48, 49, and
50, respectively.
** Obtained from mean of the values listed in column 8 of Tables 48, 49, and
50, respectively.
*** Obtained from mean of the values listed in column 11 of Tables 48, 49, and
50, respectively.
134
-------
Explanation of Large Variation in Field Sample
A large amount of this variation can be attributed to the presence of large
bundles or fiber aggregates occasionally found in this sample. Quantitative
characterization of samples containing fiber bundles is difficult for the fol-
lowing reasons:
1. One cannot readily estimate the volume of an aggregate of fibers. The
current method of assigning some average length and width and assump-
tion of a cylindrical shape results in a gross overestimate of the
volume.
2. Fiber bundles generally have a large volume as compared with majority
of individual fibers. The presence of fiber bundles makes the fiber
distribution bi-modal. A few large bundles can account for a dispro-
portionately large percentage of the total particulate volume and mass.
Table 52 shows that substantially large fractions of total chrysotile mass
can be accounted for by a few chrysotile bundles. The contribution of these
large bundles (i.e., larger than 1 ym3 in volume) to the chrysotile number con-
centration is relatively small.
Characterizing of Fiber Bundles
At present there is no rational method for characterizing fiber bundles.
Grouping fiber bundle entities along with individual fibers leads to problems
as explained above. Elimination of fiber bundles through high energy ultrasonic
treatment should not be undertaken because this may radically alter the initial
sample characteristics. A complete disregard of the bundle entities, big and
small, would result in biasing the data. In such cases, the following modifi-
cations are suggested:
1. Fiber bundles encountered should be reported as bundle entities with
tentative average lengths and widths.
2. An arbitrary cut off, for example, bundles with volumes greater than
a 1.0 ym3, should be used to separate the large bundles from the
other fibers or small bundles.
3. When a few large bundles are encountered during the random scans, we
recommend that after collecting data on 100-200 fibers, the sample be
searched for large fiber bundles only collecting data on these large
bundles (say 20 or 30) by scanning over large areas. A somewhat lower
magnification, say 10,OOOX or 5,OOOX, would be better for this. This
will enable one to obtain more representative distribution of fiber
bundles and their number and volume concentrations in the initial
sample.
135
-------
Table 52
EFFECT OF A FEW LARGE BUNDLES ON NUMBER CONCENTRATION AND
MASS CONCENTRATION OF CHRYSOTILE IN FIELD SAMPLE 661
1
Data
Set
73
74
75
76
77
78
79
69
70
90
91
71*
72*
80*
81
82
83*
84*
86
51-1
51-2
51-3
52
53*
54*
50
48
49
2
Number Cone.
of all Chrys.
106/m3
116.9
158.9
278.2
139.7
49.9
123.7
137.1
41.2
32.6
223.7
318.0
34.0
47.3
9.3
238.5-
167.8
27.4
23.3
58.5
47.3
14.9
14.1
12.1
18.6
27.6
11.2
21.5
143.1
3
Number Cone, of
Chrys . Bundles
Greater than
1 ym3 in Size
106/m3
11.36 •
1.66
3.32
1.65
0.83
1.47
2.35
4
Actual Number
(counted) of
Chrys . Bundles
Greater than
1 ym3
1
2
2
1
4
2
1
8
10
5
Mass Cone.
of all Chrys.
yg/m3
3.029
5.035
10.740
3.411
4.211
2.957
0.746
2.850
8.204
1.760
80.629
29.035
7.755
2.757
22.62
9.91
9.153
2.933
6.949
21.58
33.83
35.01
13.5
106.7
5.483
1.827
1.014
2.749
6
.'
Mass Cone.
of Chrys.
Bundles
' Ug/ffl3
•>.
•';
':
63.8840
24.2986
15.5115
3.5265
15.8360
27.7713
34.3326
8.3418
100.6012
• -
s
* Unashed.
136
-------
4. In the analysis of the particulate data, these large bundles should
be treated separately from the other fibrous particulates and their
number and mass reported separately.
EFFECT OF ASHING, ULTRASONIFICATION, AND RECONSTITUTION
The field sample was studied with the inclusion of an ashing step by nine
operators (see Table 43). The same sample was also studied without low tempera-
ture ashing, ultrasonification, and reconstitution by six operators (see
Table 44).
Effect of Ashing on Number Concentration
One question most commonly asked is whether ashing, ultrasonification, and
reconstitution alter the initial sample. To answer this question, one can com-
pare the number concentration estimates and mean fiber length and width dimen-
sions, in the unashed and ashed samples.
Table 53 lists the data for number concentrations of all fibers in the ashed
samples (column 2), unashed samples (column 3), and the ratio of the concentra-
tion estimates for ashed samples to those for unashed samples. Similar quanti-
ties for number concentration for chrysotile fibers are listed in columns 5, 6,
and 7, and for mass concentration of chrysotile in columns 8, 9, and 10.
From columns 4 and 7 of Table 53, it is clear that the data from operators
1, 2, 5, and 6 distinctly show an appreciable increase in the reported number of
all fibers as well as the chrysotile fibers. This is in contrast to data from
operator 4, who reports a net loss of fibers due to the ashing step.
The increase in fiber number concentrations due to ashing step may have two
possible explanations:
1. Fiber breakage and breaking of bundles into fibrils,
2. A reduced interference by non-fibrous debris in the ashed sample, thus
facilitating unhindered detection of fibers.
If the number of fibers is increased due to breakage, it should decrease the
mean fiber length and mean fiber width in the ashed and reconstituted sample.
However, this would also result if ashing led to reduced interference with the
detection of relatively short and thin fibers.
Effect of Ashing on Mean Length and Mean Diameter of Fibers
The data on mean fiber length for all fibers and for chrysotile fibers are
summarized in Table 54. The ratio of the mean fiber length in the ashed sample
137
-------
Table 53
EFFECT OF LOW TEMPERATURE ASHING AND RECONSTITUTION OF FIBER CONCENTRATION ESTIMATES
OJ
CO
Operator
Code
1
2
3
4
5
6
8
9
10
11
All Fibers
Mean Number Concentration
106 Fibers/m3
Ashed
347.39
84.45
29.33
16.3
494.7
463.8
153.3
204.4
Unashed
78.85
21.4
.
28.4
62.3
53.1
—
—
Ratio
Ashed
Unashed
4.406
3,946
—
0.574
7.941
8.734
—
—
43.2
8
10
Chrysotile Fibers
Mean Number Concentration
10s Fibers/m3
Ratio
Ashed
Chrysotile Fibers
Mean Mass Concentration
yg/m3
Ratio
Ashed
260.6
171.79 40.65 4.226
36.9 9.3 3.968
25.43
12.1 23.6 0..513
238.5 27.4 8.704
167.8 23.3 7.202
21.5
143.1
11.2
58.5
Ashed Unashed Unashed Ashed Unashed Unashed
13.22 18.39 0.719
5.52 2.756 2.006
30.15
13.50 56.09 0.241
22.62 9.15 2.472
9.99 2.93 3.410
1.01
2.75
1.83
6.95
11
Remarks
Does Ashing Increase the
Number of Fibers Counted?
Definitely
Definitely
Not enough data
No, there is a definite loss
Definitely
Definitely
Not enough data
Not enough data
Not enough data
Not enough data
-------
Table 54
EFFECT OF LOW TEMPERATURE ASHING AND RECONSTITUTION OF MEAN FIBER DIMENSIONS
1
Operator
Code
1
2
3
4
5
6
8
9
10
11
2
3
4
5
6
7
All Fibers - Characteristic Dimensions
Mean
Fiber Length
Um
Ashed
1.358
1.534
2.258
2.568
1.075
1.005
1.180
1.510
—
1.248
Unashed
2.279
2.915
~
2.902
1.928
2.100
--'
—
1.551
--
Ratio
Ashed
Unashed
0.60
0.53
—
0.88
0.56
0.48
--
—
—
—
Mean
Fiber Diameter
um
Ashed
0.082
0.106
0.198
0.206
0.079
0.082
0.077
0.055
—
0.071
Unashed
0.051
0.178
—
0.216
0.126
0.091
--
--
0.109
—
Ratio
Ashed
Unashed
0.54
0.60
—
0.95
0.63
0.90
--
__
—
—
8
9
10
11
12
13
Chrysotile Fibers - Characteristic Dimensions
Mean
Fiber Length
Um
Ashed
2.154
1.851
2.483
2.963
1.450
1.474
1.529
1.659
—
2.029
Unashed
3.221
3.719
— "
3.037
3.152
3.064
~
~
2.255
--
Ratio
Ashed
Unashed
0.67
0.50
~
0.98
0.46
0.48
--
—
—
-—
Mean
Fiber Diameter
Mm
Ashed
0.081
0.093
0.217
0.231
0.093
0.097
0.097
0.058
—
0.100
Unashed
0.151
0.145
—
0.232
0.168
0.098
—
—
0.120
— • ••
Ratio
Ashed
Unashed
0.54
0.64
~
1.00
0.55
0.99
—
—
—
-—
14
Does Ashing
Reduce
Observed Mean
Fiber Length?
Definitely
Definitely
Does Ashing
Reduce
Observed Mean
Fiber Diameter?
Definitely
Definitely
No Test Possible
N.S.*
Definitely
Definitely
N.S.
Definitely
N.S.
No Test Possible
No Test Possible
No Test Possible
No Test Possible
CO
VO
* N.S. stands for not significant,
-------
to that in the unashed sample is listed in column 4 for all fibers, and in
column 10 for chrysotile fibers, Similar quantities for mean fiber diameter
are listed in column 7 for all fibers and in column 13 for chrysotile fibers.
Data from operators 1, 2, 5, and 6 support the contention that more fibers
are being generated in ashing step due to fiber breakage. The data from
operator 4 are inconclusive.
The second explanation may also be simultaneously correct, but it is diffi-
cult to verify. Operators 2, 5, and 6 have reported increased chrysotile mass
estimates in ashed samples (see Table 53). This seems to suggest the second
explanation.
The reported low number concentration in ashing step by operator 4 may be
explained in two ways:
1. All fibers are not retained during the ashing and reconstltution.
2. Agglomeration occurs in the ashing step.
If the first explanation was valid, the mass concentration of chrysotile would
be reduced after ashing. Table 53, column 10, does show that there was a sub-
stantial loss (75%) of chrysotile mass concentration.
If the second explanation was valid, the mean fiber dimensions should
increase. Table 54 shows a slight decrease for mean fiber length for all fibers
and practically no increase in mean length and mean diameter of chrysotile
fibers. Hence, we may conclude that this is a real possibility of a true loss
of fibers because of the several transfer steps in the ashing and reconstitution
of samples.
We had used the mean fiber length and mean width for assessing the effect
of ashing and sonification step. An alternative method would be to consider
the entire length distribution. Table 55 lists the length distribution of
fibers for unashed and ashed samples for operator 2. Tables 56, 57, and 58 show
the length distributions reported by operations 4, 5, and 6, respectively.
Frequencies have been expressed as percent frequencies for a direct comparison
between ashed and unashed samples.
In all four cases, it appears that the largest fibers reported in unashed
samples were longer than those for the ashed sample. Also, there were fewer
long fibers in ashed sample as compared with the unashed sample.
140
-------
Table 55
LENGTH DISTRIBUTION IN ASHED AND UNASKED SAMPLES,
DATA FROM OPERATOR NUMBER 2
Ashed
(Data Set 70)
% Frequency
1.85
1.85
1.85
1.85
. 1.85
1.85
3.70
7.41
3.70
1.85
3.70
1.85
7.41
1.85
1.85
1.85
5.55
3.70
11.11
20.37
5.55
1.85
Length, um
15.15
14.54
10.61
7.57
7.27
6.97
6.67
6.36
5.76
5.15
4.85
4.54
3.94
3.64
3.33
3.03
2.75
2.42
2.12
2.06
1.82
1.57
1.51
1.45
•(-Median-*- 1.33
1.21
1.09
0.97
0.91
0.73
0.61
0.48
0.36
Unashed
(Data Set 80)
% Frequency
1.92
1.92
3.84
1.92
1.92
1.92
1.92
5.77
1.92
3.84
7.69
1.92
1.92
7.69
«-MediaiH- 5.77
7.69
7.69
1..92
5.77
1.92
11.51
3.84
5.77
1.92
100.00 100.00
141
-------
Table 56
LENGTH DISTRIBUTION IN ASHED AND UNASHED SAMPLES,
DATA FROM OPERATOR NUMBER 4
Ashed
(Data Set 52)
% Frequency
1.08
1.08
2.17
1.08
2.17
1.08
2.17
1.08
1.08
9.78
6.52
3.26
10.86
1.08
4.35 -^-Median-*
5.43
20.65
5.43
1.08
2.17
9.78
1.08
1.08
3.25
Length , ym
31.0
27.0
19.5
16.0
14.0
11.5
11.0
10.0
9.0
8.0
7.75
7.50
7.00
6.00
5.00
4.75
4.50
4.25
4.00
3.75
3.50
3.00
2.50
2.25
2.17 -Hfedian-*
2.00
1.75
1.50
1.25
1.10
1.00
0.75
0.70
0.65
0,60
0.50
0.45
0.40
0.35
0.30
0.25
Unashed
(Data Set 53)
% Frequency
1.0
1.0
1.0
2.0
1.0
1.0
1.0
1.0
4.0
1.0
2.0
1.0
1.0
1.0
6.0
1.0
7.0
6.0
9.0
1.0
6.0
1.0
9.0
5.0
1.0
8.0
2.0
3.0
3.0
4.0
3.0
3.0
2.0
100.00
100.0
142
-------
Table 57
LENGTH DISTRIBUTION IN ASHED AND UNASKED SAMPLES,
DATA FROM OPERATOR NUMBER 5
Ashed
(Data Set 81)
% Frequency
0.89
0.89
0.89
0.89
0.89
0.89
0.89
0.89
0.89
0.89
2.68
0.89
0.89
3.57
2.68
1.78
2.68
0.89
0.89
7.14
1.78
2.68
8.93 +Med
16.07
16.96
12.50
8.03
100.00
Length, ym
13.0
12.8
8.8
8.5
7.0
4.5
4.2
4.0
3.9
3.7
3.5
3.3
3.1
3.0
2.9
2.8
2.7
2.6
2.5
2.4
2.3
2.2
2.0
1.8
1.7
1.5
1.4 «-Median-»-
1.3
1.2
1.1
1.0
0.9
0.8
lian-* 0.7
0.6
0.5
0.4
0.3
Unashed
(Data Set 83)
% Frequency
1.33
1.33
1.33
1.33
1.33
1.33
1.33
2.67
1.33
1.33
2.67
2,67
5.33
1.33
5.33
5.33
1.33
1.33
5.33
4.00
5.33
1.33
6.67
4.00
5.33
12.00
5.33
5.33
4.00
100.00
143
-------
Table 58
LENGTH DISTRIBUTION IN ASHED AND UNASHED SAMPLES,
DATA FROM OPERATOR NUMBER 6
Ashed
(Data Set 82)
% Frequency
0.95
0.95
0.95
0.95
0.95
0.95
0.95
0.95
0.95
1.90
0.95
0.95
0.95
0.95
0.95
0.95
0.95
2.86
2.86
9.52
7.62
8.57 ^Median-*-
10.47
5.71
19.05
11.43
3.81
0.95
Length, ym
0.30
9.00
6.00
4.50
4.30
4.00
3.10
3.00
2.80
2.50
2.30
2.10
2.00
1.90
1.80
1.70
1.60
1.50
1.40
1.30
1.20
1.10
1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
Unashed
(Data Set 84)
% Frequency
3.12
3.12
9.38
3.12
12.50
3.12
6.25
3.12
+-Median-> 3.12
3.12
9.38
6.25
9.38
6.25
3.12
12.50
3.12
100.00 100.00
144
-------
The tendency for the distribution to shift towards the small length after
ashing is quite evident.
With the limited data from this study, we tentatively conclude that ashing
and sonification can significantly alter the measured sample characteristics.
The exact mechanism of this difference remains obscure.
In this study of ashing and reconstitution, aerosol O.T. solution was used
for resuspending the ash. There is a possibility that aerosol O.T. may have
contributed to some breakage of chrysotile fibers and obscured the effects of
actual ashing and ultrasonic treatment. So, further work is needed to evaluate
the effects of ashing, dilution and reconstitution on the asbestos sample.
A full-scale study is also needed to evaluate the effect of dilution,
without the ashing step. This could be done by dissolving the initial filter
in a suitable solvent and then redepositing the solids onto a new polycarbon-
ate filter after appropriate dilution.
CONCLUSIONS
This round-robin test was carried out with very little opportunity for
most of the participants to become familiar with the provisional procedure.
Much better results should be expected if such a test were repeated after
participants obtained more experience with the procedure.
From the round-robin test, the following conclusions can be reached.
1. It is difficult to determine absolute accuracy of the electron
microscope estimates. The overall mean estimate for chrysotile
mass concentration according to the electron microscope method is
2.72 yg/m3 in the laboratory air sample. This may be compared to a
chrysotile concentration of 2.45 yg/m3 for the same sample as deter-
mined independently by X-ray fluorescence speetrometric analysis of
magnesium. Thus, the EM estimate of chrysotile mass concentration
differs by only 10% with that by the X-ray fluorescence method.
2. In laboratory sample, the ratio of spread between 95% confidence
limits to the mean value was 0.48 for chrysotile fiber concentra-
tion and about 0.40 for chrysotile mass concentration.
3. In the field air sample, studied with ashing, the ratio of the
spread between 95% confidence limits to the mean value was about
0.49 for chrysotile fiber concentration and about 1.57 for chryso-
tile mass concentration. In the same sample, studied without
ashing, the corresponding values are 0.62 for chrysotile fiber
concentration and about 2.34 for chrysotile mass concentration.
145
-------
4. Presence of a few large fibers or fiber bundles strongly influence the
mass concentration estimates and mean values of length, width, and
volume of fibers, but does not significantly affect the number concen-
tration of fibers.
5. The following modifications are suggested to mitigate the adverse
effects of fiber aggregates on sample characterization.
a. Bundles or aggregates of fibers larger than 1 ym3 (as judged
by mean length, diameter* and cylindrical shape assumption)
should be counted as single entities.
b. Bundle entities should be treated separately and reported
separately from other single fibers in the statistical analysis
of fiber characteristics.
c. Representative data on bundles can be collected by scanning the
sample at a lower magnification (e.g., 5,000 X).
6. In any air sample, the precision of the estimates can be improved by
studying at least three or four TEM grids.
146
-------
REFERENCES
1. Thomson, J.G., R.O.C. Kaschula, and R.R. McDonald. Asbestosis as a Modern
Urban Hazard. S. African Med. J., 37: 77, 1963.
2. Thomson, J.G. Asbestos and the Urban Dweller. Ann. N.Y, Acad. Sci.,
132: 196, 1965.
3. Selikoff, I.J., J. Chrug and E.G. Hammond. The Occurrence of Asbestosis
Among Insulation Workers in the United States. Ann. N.Y. Acad. Sci.,
132: 139, 1965.
4. Hourihane, D.O. The Pathology of Mesotheliomata and Analysis of Their
Association With Asbestos Exposure. Thorax, 19: 268-278, 1964.
5. Owen, W.G. Mesothelial Tumors and Exposure to Asbestos Dust. Ann. N.Y.
Acad. Sci., 132: 674, 1965.
6. Newhouse, M..L. and H. Thompson. Mesothelioma of Pleura and Peritoneum
Following Exposure to Asbestos in the London Area. Brit. J. Ind. Med.,
22: 261, 1965.
7. Whipple, H.E. Biological Effects of Asbestos. Ann. N.Y. Acad. Sci.,
132: 766, 1965.
8. Knox, J.F., R. S. Doll and I.D. Hill. Cohort Analysis of Changes in Inci-
dence of Bronchial Carcinoma in a Textile Factory. Ann. N.Y. Acad. Sci.,
132: 526, 1965.
9. Harington, J.S., et. al. Studies of the Mode of Action of Asbestos As a
Carcinogen. S. African Med. J. 41(32): 800-804, 1967.
10. Tabershaw, I.R. Asbestos as an Environmental Hazard. J. Occup. Med.,
16: 32, 1968.
11. McDonald, A.D., et. al. Epidemiology at Primary Malignant Mesothelial
Tumors in Canada. Cancer, 26(4): 914-919, 1970.
12. Bader, M.E., et. al. Pulmonary Function and Radiographic Changes in 598
Workers with Varying Duration of Exposure to Asbestos. Mt. Sinai J. Med.
37(4): 492-500, 1970.
13. Gross, P., et. al. The Pulmonary Response to Fibrous Dusts of Diverse
Compositions. Amer. Ind. Hyg. Assoc. J. 31(2): 25-32, 1970.
14. Reeves, A.L., et. al. Experimental Asbestos Carcinogenesis» Environ. Res.
4(6): 495-511, 1971.
15. Langer, A.M., I.J. Selikoff and A. Sastre. Chrysotile Asbestos in the
Lungs of Persons in New York City. Arch. Environ. Health., 22: 348-361,
1971.
147
-------
16. Selikoff, I.J., et. al. Asbestos Air Pollution. Arch. Environ. Health,
25(1): 1-13, 1972.
17. Selikoff, I.J., W.J. Nicholson and A.M. Langer. Carcinogenicity of Amosite
Asbestos. Arch. Environ. Health, 25(3): 183-188, 1972.
18. Gilson, J.G., Health Hazards of Asbestos. Composites, 3: 59, 1972.
19. Sherrill, R. Asbestos - The Saver of Lives has a Deadly Side. N.Y. Times
Magazine, January 21, 12 pp., 1973.
20. Asbestos Health Question Perplexes Experts. Chem. Eng. News. 51(50): 18,
1973.
21. Timbrell, V. Inhalation and Biological Effects of Asbestos. In: Assess-
ment of Airborne Particles. Ed. T.T. Mercer et. al. Charles C. Thomas
Publ., p. 429, 1972.
22. Stanton, M.F. and C. Wrench. Mechanism of Mesothelioma Induction with
Asbestos and Fiber Glass. J. National Cancer Inst., 48: 797, 1972.
23. Report of the Advisory Committee on Asbestos Cancers to the Director of the
International Agency for Research on Cancer. In: Biological Effects of
Asbestos. International Conference, Lyon, France, International Agency for
Research on Cancer, IARC Sclent..Publ. 8: 341, 1973. i
24. National Emission Standards for Hazardous Air Pollutants - Asbestos, Beryl-
lium, Mercury. Fed. Register. 38(66): pp. 8820-8850, 1973.
25. Wagner, C. Disputes on the Safety of Asbestos. New Scientist, 61(888): 606,
1974.
26. Wagner, J.C. Current Opinions on the Asbestos Cancer Problem. Ann. Occup.
Hyg., 15: 61-64, 1972.
27. Background Documentation onTEvaluation of Occupational Exposure to Airborne
Asbestos. Joint ACGIH-AIHA Aerosol Hazards Evaluation Committee. Amer.
Ind. Hyg. Assoc., 36(2): 91-103, 1975.
28. Keenan, R.G. and J.R. Lynch. Techniques for the Detection, Identification
and Analysis of Fibers. Amer. Ind. Hyg. Assoc. J., 31:587, 1970.
29. Mumpton, F.A. Characterization of Chrysotile Asbestos and Other Members
^ of the Serpentine Group of Minerals. Siemens Review, XLI, 75, 1974.
30. Kramer, J.R., 0. Mudroch and S. Tihor. Asbestos in the Environment.
McMaster University, Report submitted to Research Advisory Board, Inter-
national Joint Commission and Environment Canada, 1974.
31. Richards, A.L. Estimation of Trace Amounts of Chrysotile Asbestos by X-ray
Diffraction. Anal. Chem. 44: 1872, 1972.
32. Birks, L.S. Quantitative Analysis of Airborne Asbestos by X-ray Diffraction.
Naval Res. Labs., Final Report, EPA-650/2-75-004, 1975.
148
-------
33. Crable, J.V., and M.J. Knott. Application of X-ray Diffraction to the
Determination of Chrysotile in Bulk and Settled Dust Samples. Amer. Ind.
Hyg. J., 27: 382-385, 1966
34. Crable, J.V., and M.J. Knott. Quantitative X-ray Diffraction Analysis of
Crocidolite and Amosite in Bulk or Settled Dust Samples. Amer. Ind. Hyg.
J., 27:449-453, 1966.
35. Cook, P.M., J.B. Rubin, C.J. Maggiore, and W.J. Nicholson. X-ray
Diffraction and Electron Beam Analysis of Asbestiform Minerals in Lake
Superior Waters. Proc. Intern. Conf. on Environmental Sensing and
Assessment, Pub. by IEEE, Piscataway, NJ, #$(2): 1-9, 1976.
36. Schelz, J.P. The Detection of Chrysotile Asbestos in Low Levels of Talc
by Differential Thermal Analysis. Termochem. Acta., 8: 197-203, 1974.
37. Taylor, D.G., C.M. Nenadic, and J.V. Crable. Infrared Spectra for Mineral
Identification. Amer. Ind. Hyg. Assoc. J., 31: 100, 1970.
38. Bagioni, R.P. Separation of Chrysotile Asbestos from Minerals that Inter-
fere with Its Infrared Analysis. Environ. Sci. and Tech., 9: 262, 1965.
39. Gadsden, J.A,, J. Parker, and W.L. Smith. Determination of Chrysotile
in Airbone Asbestos by an Infrared Spectrometric Technique. Atmos. Env.,
4: 667, 1970.
40. Heffelfinger, R.E., C.W. Melton, D.L. Kiefer, and W.M. Henry. Development
of a Rapid Survey Method of Sampling and Analysis for Asbestos in Ambient
Air. Battelle Columbus Labs, Ohio, Final Report APTD-0965, EPA Contract
No. CPA-22-69-110, 46 p., 1972.
41. Richards, A.L. Estimation of Submicrogram Quantities of Chrysotile
Asbestos by Electron Microscopy. Anal. Chem., 45: 809, 1973.
42. Beckett, S.T. The Evaluation of Airborne Asbestos Fibers Using a Scanning
Electron Microscope. Ann. Occ. Hyg., 16: 405, 1973.
43. Pattnaik, A., and J.D. Meakin. Development of Scanning Electron Microscopy
for Measurement of Airborne Asbestos Concentrations. Final Report,
Franklin Inst. Res. Lab., Philadelphia, PA, EOA 650/2-75-209, 84 p.,
National Tech. Inf. Service, Springfield, VA, 1975.
44. Ruud, C.O., C.S. Barrett, P.A. Russell, and R.L. Clark. Selected Area
Electron Diffraction and Energy Dispersive X-ray Analysis for the Identi-
fication of Asbestos Fibers, A Comparison. Micron, 7: 115-132, 1976.
45. Mueller, P.K., A.E. Alcoer, R.Y. Stanley, and G.R. Smith. Asbestos Fiber
Atlas. U.S. EPA Publication No. EPA-650/2-75-036, National Tech. Inf.
Service, Springfield, VA, 1975.
46. Gerber, R.M., and R.C. Rossi. Evaluation of Electron Microscopy for Process
Control in the Asbestos Industry. The Aerospace Corporation, Los Angeles,
CA, Final Report, EPA-600/2-77-059, 55 p., National Tech. Inf. Service,
Springfield, VA, 1975.
149
-------
47. Mlllette, J.R., and E.F. McFarren. EDS of Waterborne Asbestos Fibers in
TEM, SEM, and STEM. Proc. of IITRI Symposium on SET/1976/1, 451-460, 1976.
48. Beamon, D.R., and D.M. File. Quantitative Determination of Asbestos Fiber
Concentrations. Anal. Chem., 48(1): 101-110, 1976.
49. Maggiore, C.J., and I.E. Rubin. Optimization of an SEM X-ray Spectrometer
System for the Identification and Characterization of Ultramicroscopic
. Particles. Proc. of IITRI Symposium on SEM/1973, 129-136, 1973.
50. Cochran, W.G., and G.M. Cox. Experimental Designs. Wiley, 1950.
51. Youden, W.J. The Collaborative Test. Journal of the Association of
Official Agricultural Chemists, 46(1): 55-62, 1963.
52. Cramer, H. Mathematical Methods of Statistics. Princeton University
Press, 1946.
53. Fisher, R.A. Statistical Methods for Research Workers. Oliver and Boyd,
10th Ed., 1946.
54. Davies. O.L. (Ed.). The Design and Analysis of Industrial Experiments.
Oliver and Boyd, 1954.
55. Fisher, R.A., and F. Yates. Statistical Tables. Oliver and Boyd, 1948.
56. Bartlett, M.S. The Use of Transformations. Biometrics, 3(1): 39-52, 1947.
57. Herdan, G. Small Particle Statistics. Butterworth & Co., 2nd Ed., 1960.
58. Daniel, C., F.S. Wood. Fitting Equations to Data. Wiley-Interscience,
1971.
59. Dixon, W.J. (Ed.). BMD Biomedical Computer Programs. University of
California Press, 1977.
60. Draper, N.R., and Smith, H. Applied Regression Analysis. Wiley, 1966.
:
61. Tentative Method of Analysis for Suspended Particulate Matter in the Atmos-
phere (High-Volume Method) 11101-01-70T. Methods of Air Sampling and
Analysis, Intersociety Committee, Amer. Public Health Assoc., 1015 18th St.,
N.W., Washington, DC, 1972.
62. Recommended Procedures for Sampling and Counting Asbestos Fibers; Joint
AIHA-ACGIH Aerosol Hazards Evaluation Committee, Amer. Ind. Hygi. Assoc. J.,
36(2): 83-90, 1975.
63. Jaffe, M.A. Handling and Washing Fragile Replicas. Proc. of EMSA, Toronto,,
September 1948. J. Appl. Phys., 19(12): 1189, 1948.
64. Samudra, A.V-, and C.F. Harwood. Electron Microscope Measurement of
Airborne Asbestos Concentrations - A Provisional Methodology Manual. U.S.
Environmental Protection Agency, Publication No. EPA-600/2-77-178, 55 p.,
National Tech. Inf. Service, Springfield, VA, August 1977.
150
-------
65. Wagman, J. Chemical Composition of Atmospheric Aerosol Pollutants by
High Resolution X-ray Fluorescence Spectrometry. In Colloid and Inter-
face Science, Vol. II. M. Kerker (Ed.), Academic Press, 1976.
151
-------
APPENDIX A
THE EXPERIMENT DESIGN FOR PHASE 1
The design of the fractional factorial experiment for studying 12 variables
is shown in compact form in Table A-l, where Xx through X12 represent the 12
variables, or subprocedures, listed in Table A-2. The indices 1, 2, and 3 repre-
sent the variable levels as defined in Table A-2.
In order to evaluate the 12 independent variables, the variable levels were
orthogonally coded as shown in Table A-2. Each three-level variable has one
linear and one quadratic component, denoted by L and Q respectively, while each
two-level factor has one quadratic component. The coding scheme shown in Table A-3
is chosen to satisfy three conditions:
1. The sum of linear components for each variable is zero^ (8 linear com-
ponents) . ;
2. The sum of quadratic components for each variable is zero (12 quadratic
components).
3. The sum of the cross products of each pair of components is zero (190
pairs of components).
This is illustrated by considering the variable XI.
Level
1
2
3
Linear
Comp. Code
-1
1
_0
Total 0
Quadratic
Comp . Code
1
1
^2
Total 0
Cross Product
-1
1
_0
Total 0
Thus, it satisfies all the three conditions specified above. This coding scheme
ensures the orthogonality, i.e., independence ,and non-correlation of the variables
considered, and helps to bring out even relatively small effects of the 12 controlled
factors.
152
-------
Table A-l
PHASE 1 EXPERIMENT DESIGN
Combination
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Factor
*l .,
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
^
1
1
1
2
2
2
3
3
3
1
1
1
2
2
2
3
3
3
1
1
1
2
2
2
2
3
3
A3
o
2
2
2
1
1
1
2
2
2
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
1
1
1
*4
2
2
2
1
1
1
2
2
2
2
2
2
2
2
2
1
1
1
1
1
1
2
2
2
2
2
2
V
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
^6
1 ,
1
2
1
1
2
1
1
2
1
2
1
1
2
1
1
2
1
2
1
1
2
1
1
2
1
1
x7
3
1
2
1
.2
3
2
3
1
3
1
.2
1
2
3
2
3
1
3
1
2
1
2
3
2
3
1
^8
2
2
1
2
1
2
1
2
2
2
1
2
1
2
2
2
2
1
1
2
2
2
2
1
2
1
2
*g
1
2
3
3
1
2
2
3
1
2
3
1
1
2
3
3
1
2
3
1
2
2
3
1
1
2
3
-10
2
1
3
2
1
3
2
1
3
3
2
1
3
2
1
3
2
1
1
3
2
1
3
2
1
3
2
^•11
2
1
3
3 .
2
1
1
3
2
2
1
3
3
2
1
1
3
2
2
1
3
3
2
1
1
3
2
/\4 A
—12
1
3
2
2
1
3
3
2
1
2
1
3
3
2
1
1
3
2
3
2
1
1
3
2
2
1
3
153
-------
Table A-2
INDEPENDENT VARIABLES, PHASE 1
Levels and Codes
Variable
INDEPENDENT VARIABLES OF FILTER LOADING
X-, Composition of Sample in
Aerosol Chamber
(1) 100% Chrysotile
(2) 60% Chrysotile
+ 4035 Amphibole
(3) 70% Chrysotile
+ 20% Amphibole
•*• 10% Non-Asbestos Fiber
XiL=-l
Xjl= 1
X,Q= 1
XiQ= 1
XjL= 0 XjQ=-2
c on Filter (1) Light X2L=-1 X2Q= 1
(2) Medium. X2L= 0 X2Q=-2
(3) Heavy _ X2L= 1 X2Q= 1
X, Sampling Instrument- (1) High Volume •> X3Q=-2
J (2) Personal XjQ= 1
X,, Filter Type (1) Nuclepore X*Q=-2
4 (2) Millipore X,Q= 1
Xc Pore Size, nominal (1
(2
(3]
I 0.2 pm X5L=-1 X5Q= 1
1 0.4 \tm XSL= 0 X5Q=-2
I 0.8 um X5L- 1 X5Q= 1
INDEPENDENT VARIABLES OF TEM GRID PREPARATION
Xs FiHer Side (1]
; . <2<
Particle side down X6Q= 1
Particle side up X6Q=-2
X7 2.3 mm Portion location (1) Periphery X7L=-1 X7Q= 1
' (2) Mid-radius X7L= 0 X7Q=-2
(3) Center X7L= .1 X7Q= 1
XB Use of Carbon Coating (1
8 (2
XQ Transfer Method (1,
•|i!
INDEPENDENT VARIABLES OF TEM EXAMINATION
Yes X8Q=-2
No X«Q= 1
Soxhlet Extraction 1 (short) X9L=-1 X9Q= 1
Soxhlet Extraction^ (long) X9L= 1 X9Q= 1
Jaffe Method X9L= 0 X,Q=-2
Xln Magnification, nominal* (1) 5.000X (screen mag. 4.000X) Xi0L=-l X10Q= 1
u . ' (2) 10.000X (screen mag. 8.000X) X,0L= <3 X,0Q=-2
(3) 20,OOOX (screen mag. 16.000X) Xi0L= 1 X10Q= 1
Xn Grid Opening Location (1
11 (2
(3
X12 Choice of Fields (1
Periphery XML=-1 XnQ= 1
Mid-radius Xul-= 0 XuQ=-2
) Center XnL= 1 XnQ= 1
Random choice of small fields Xi2L=-l Xi2Q= 1
Small fields, consecutive Xt2L= 1 X12Qs 1
Entire grid opening as a field Xi2L= 0 Xi2Q=-2
*The actual magnification at the fluorescent screen is somewhat smaller than the nominal or
camera magnification, depending upon the design geometry of each transmission electron
microscope.
154
-------
Table A-3
VALUES OF CODED INDEPENDENT VARIABLES, PHASE 1 COMBINATIONS
Cn
XI
Comb.
•; l
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
L
-1
-1
-1
-1
-1
-1
-1
-1
-1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
Q
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
-2
-2
-2
-2
-2
-2
-2
-2
-2
X2
L
-1
-1
-1
0
0
0
1
1
1
-1
-1
-1
0
0
0
1
1
1
-1
-1
-1
0
0
0
1
1
1
Q
1
1
1
-2
-2
-2
1
1
1
1
1
1
-2
-2
-2
1
1
1
1
1
1
-2
-2
-2
1
1
1
X3
X
1
1
1
-2
-2
-2
1
1
1
-2
-2
-2
1
1
1
1
1
1
1
1
1
1
1
1
-2
-2
-2
X4--
Q
1
1
1
-2
-2
-2
1
1
1
1
1
1
1
1
1
-2
-2
-2
-2
-2
-2
1
1
1
1
1
1
.,, )
L
-1
0
1
-1
0
1
-1
0
1
-1
0
1
-1
0
1
-1
0
1
1
0
1
-1
0
1
-1
0
1
(5
~±
1
-2
1
1
-2
1
1
-2
1
1
-2
1
1
-2
1
1
-2
1
1
-2
1
1
-2
1
1
-2
1
X6
Q
1
1
-2
1
1
-2
1
1
-2
1
-2
1
1
-2
1
1
-2
1
-2
1
1
-2
1
1
-2
1
1
L
1
-1
0
-1
0
1
0
1
-1
1
-1
0
-1
0
1
0
1
-1
1
-1
0
-1
0
1
0
1
-1
X7
TF
1
1
-2
1
-2
1
-'2
1
1
1
1
-2
1
-2
1
-2
1
1
1
1
-2
1
-2
1
-2
1
1
X8
X
1
1
-2
1
-2
1
-2
1
1
1
-2
1
-2
1
1
1
1
-2
-2
1
1
1
1
-2
1
-2
1
. X9
L
-1
1
0
0
-1
1
1
0
-1
1
0
-1
-1
1
0
0
•-1
1
0
-1
1
1
0
-1
-1
1
0
JL
i
i
-2
-2
1
1
1
-2
1
1
-2
1
1
1
-2
-2
1
1
-2
1
1 .
1
-2
1
1
1
-2
X10
Xll
X1Z
TT T— D: T— P:
0
-1
1
0
-1
1
0
-1
1
1
0
-1
1
0
-1
1
0
-1
-1
1
0
-1
1
0
-1
1
0
-2
1
1
-2
1
1
-2
1
1
1
-2
1
1
-2
1
1
-2
1
1
1
-2
1
1
-2
1
1
-2
0
-1
1
1
0
-1
-1
1
0
0
-1
1
1
0
-1
-1
1
0
0
-1
1
1
0
-1
-1
1
0
-2
1
1
1
-2
1
1
1
-2
-2
1
1
1
-2
1
1
1
-2
-2
1
1
1 •
-2
1
1
1
-2
-1
0
1
1
-1
0
0
1
-1
1
-1
0
0
1
-1
-1
0
1
0
1
-1
-1
0
1
1
-1
0
1
-2
1
1
1
-2
-2
1
1
1
1
-2
-2
1
1
1
-2
1
-2
1
1
1
-2
1
1
1
-2
-------
Table A-4
STATISTICS OF FIBER CHARACTERISTICS BY SAMPLE
SAMPLE 1
Total Mean S.D. S.E. Var. Min. Max.
3
Volume (ym) 0.382780 0.0026582 0.0047790 0.0003982 0.0000228 0.0000614 0.0359526
(over all fibers)
LN Width (Uffi) -433.108917 -3.0077009 0.5166071 0.0430506 0.2668829 -3.6888795 -1.6739764
(over all fibers)
LN Length.(ym) -95.527122 -0.7405204 0.6529610 0.0574900 0.4263580 -2.0794415 0.8109303
(contained fibers)
LN Ashed Ratio 291.630096 2.2606983 0.6369779 0.0560828 0.4057407 1.3862944 4.0943446
(contained fibers)
LN Volume (ym)3 -901.003296 -6.9845219 1.4403129 0.1268125 2.0745010 -9.6987667 -3.3255527
(contained fibers)
The total number of fibers observed = 144
The number of fibers contained within their field =129
-------
APPENDIX B
REGRESSION ANALYSIS
The experimental data resulting from execution of the design presented can
be best analyzed by multiple regression methods, i.e., find values of the regres-
sion coefficients of a model equation that provide the best fit, taking account
of the magnitude of the experimental error. A stepwise fitting procedure will be
applied making use of the computer program BMD02R from the BMD library of statis-
tical programs [59],
The stepwise method of fitting is preferred because it selects only those
candidate terms for inclusion in the regression equation that contribute signif-
icantly to the prediction of values of the dependent variable. The regression
coefficients enter into the equation as multipliers to compute the values of the
dependent variable on the basis of values of the significant independent variables
and covariates.
In the stepwise multiple regression analysis, intermediate response equa-
tions are obtained through the insertion, at each step, of the candidate term
that makes the greatest contribution to the reduction in the residual sum of
squared deviations or, alternatively, the deletion of a variable whose contribu-
tion falls below a specific threshold significance level. This contribution is
measured by the F-value (square of the t value, which is the regression coeffic-
ient divided by its standard error). At each step of computation, the regression
coefficient and the F-value associated with each variable in the equation are
given, together with the potential F-value for each variable not in the equation.
The threshold F-values for inclusion and deletion of terms, set by the analyst,
at the 20% level determine at what point the fitting process terminates. The can-
didate variables need not be all the independent variables and covariates together.
These variables can be separated into blocks to investigate the relationships
among the independent variables and covariates separately as well as in combina-
tion.
In the stepwise fitting procedure it is often found that some (perhaps many)
of the candidate regression coefficients have estimated values that are not sig-
nificantly above the "noise level;" hence, those terms should be excluded from the
regression equation for the sake of simplicity and predictive accuracy. The sum
157
-------
of squares and degrees of freedom associated with these excluded terms can, in
many instances, be pooled to refine the variance estimate.
The final values of the regression coefficients are computed together with
their standard errors and levels of statistical significance (F or t values) .
For each observation, the value of e (difference between observed and predicted
value of the dependent variable) is computed as a check on possible out-liers in
the data.
In the regression equations for each dependent variable, only those terms
were retained having coefficients significant at the 20% probability level. That
is, a term was dropped if it was determined that its coefficient value could
occur by chance 20% or more of the time due merely to accidents of sampling.
Table B-l gives the full information on regression equation 9 for dependent
variable Yq, the square root of the estimated fiber count per ml of air. The
equation contains 14 terms, the other candidates having been dropped because of
their statistical non-significance. The equation is of the form
+ B6(X6Q) -I- B7(XgQ) + Bg(X9L) + B^p) + BIO(XIQL)
+ Bn(X1()Q) + B12(XUQ) + B13(X12Q) [1]
Since 27 tests were run, there remain 13 degrees of freedom (i.e., 27-14) for
estimating the magnitude of the residual variation. The residual standard dev-
2
iation was calculated to be 1.9012. The degree of determination, R , was found
to be 96%, i.e., 96% of the variation in the values of Y_ is accounted for by
the independent variables in the equation. For each term in the equation, the
values found for the regression coefficients, B through B „, are listed along
with their standard errors and variance ratios (which indicate their relative
significance). The least-squares fit for equation Y~ was found to be
Y9 = 12.300 + 0.7810^1.) + 0.489(X1Q) - 1.714(X2D + 0.426(X2Q)
+ 0.811(X3Q) + 1.001(X6Q) - 1.601(XgQ) - 1. 590(^1)
- 1.796(XgQ) + 4.265(X1QL) + 0.821(X1QQ)
+ 0.936(XUQ) + 1.358(X12Q) [2]
158
-------
Table B-l
PERFORMANCE EQUATION NO. 9
Dependent Variable:
Number of Tests:
Number of Terms in Equation:
Residual Degrees of Freedom:
Residual Standard Deviation:
2
Degree of Determination, R :
Independent Variable
Constant term
X0= '
Composition of sample
X.L (Coded -1 , 1, 0)
XJQ. (Coded 1, 1, -2)
Concentration on filter
X,L (Coded -1 , 0, 1)
X^Q (Coded 1, -2, 1)
Sampler type
X3Q. (Coded -2, 1)
Filter side
X6Q (Coded 1 , -2)
Carbon coating
X8Q (Coded -2, 1)
Transfer method
X-L (Coded -1, 1, 0)
X|Q, (Coded 1,1, -2)
Magnification
XinL (Coded -1, 0, 1)
XjgQ (Coded 1, -2, 1)
Grid opening location
X^Q (Coded 1, -2, 1)
Choice of fields
X12Q (Coded 1, 1, -2)
Yg = Square
fibers
27
14
13
1.9012
962
Regression
Coefficient
b
12.300
0.781
0.489
-1.714
0.426
0.811
1.001
-1.601
-1.590
-1.769
4.265
0.821
0.936
1-358
root (estimated
per cm3 of air)
Standard
Error
sb
0.448
0.259
0.448
0.259
0.259
0.259
0.259
0.448
0.259
0.448
0.259
0.259
0.259
no. of
Variance
Ratio
F
3.04
3.57
14.64
2.71
9.82
14.96
38.30
12.59
46.77
90.61
10.07
13.07
27.55
1 l-J.Jl.UII 111] I
+nro I iKrorw
-------
Thus, given particular levels for each of the factors X-^ through X12, a
corresponding estimate for YQ may be found by using the appropriate coded values
for XL through XT_Q defined in Table A-2. For example, for combination 1, the
*. 1Z
estimated value for Y. is found in the following way. Table A-l gives the levels
of the factors X through X used in combination 1, and for each of these, Table
A-2 shows the appropriate coded values of the variable XL through X.-Q- Thus,
for combination ls the coded values of the terms involved in the equation for Y
are given by:
Variable
XIL
xlQ
x2L
X2Q
Coded
Value
-1
1
-1
1
Variable
X3Q
X6Q
X8<>
Coded
Value
1
1
1
Variable
X9L
X9Q
XIOL
Coded
Value
_j
1
0
Variable
X10Q
XUQ
X12Q
Coded
Value
-2
-2
1
so that the estimated value for Yq is found to be
Y9 (combination 1) - 12.30 + 0.781(-1) + 0.489(1)
-1.714C-1) + 0.426(1) + 0.811(1) + 1.001(1)
- 1.601(1) - 1.590(-1) - 1.769(1) + 4.265(0)
+ Q.82K-2) + 0.936(-2) + 1.358(1) = 12.024
Equation 2 can be rewritten, grouping together the terms associated with the
separate factors
12.30 + [0.781(X.L) + 0.489(X,Q]
1 1
+ [-1.714(X2L) '+ 0.426(X2Q)]
+ [0.811(X3Q)] + [1.001(X6Q)J
+ [-1.601(XgQ)] + [-1.590(XgL) - 1.769(X9Q)]
+ [4.265(X1QL) + 0.821(X1QQ)]
+ [0.936(XUQ)] + [1.358(X12Q)]
[3]
160
-------
In this form Yg, is seen to equal its mean value (12.300) plus or minus the net
effect contributed by each factor taken at its specified level. These net effects
are listed in Table A-2 for each level of each factor. For example, at level 2
of factor X1, that is, for a sample composition of 60% chrysotile and 40% amphibole,
since the coded values corresponding to this level are given by X L = +1, X.Q = +1,
the corresponding net contribution factor X, at this level is found to be
I0.78K+1) + 0.489(+1)] - 1.270
The standard error of the net effect is found from the standard error of each
term by the formula
SE = /(STX^)2 + (SnX.)2
L L q q
where X^, X are the coded values for the appropriate .level of the linear and
quadratic components, respectively, and S and S the standard errors of the
linear and quadratic components, respectively. For example, for level 2 of
factor X.
SE = /(I x 0.488)2 + (1 x 0.259)2 = 0.518
In the case that the linear component of some factor is missing, the corresponding
term is dropped from the formula and SE = /(S_X) ; similarly, if the quadractic
component is missing, SE = S_.
The 80% confidence limits for each relative effect were found using student's
t with 13 degrees of freedom, t = 1.350. Thus, for example, for level 2 of XI,
the lower limit of the 80% confidence interval is found to be
1.270 - (1.350 x 0.518) = 0.572
and the upper limit
1.270 + (1.350 x 0.518) = 1.968
That is, with at least 80% certainty, the net effect of choosing level 2 of XI
would be to raise the estimated mean value of Y_ by an amount lying in the 80%
confidence interval between 0.572 and 1.968. The confidence limits for the
various net effects are given in Table B-2 and are presented graphically in
Figures 3 and 4 of the main report.
A similar analysis is presented for Yj_ (the natural log of the estimated
concentration of the fibers in the atmosphere) in Tables B-3 and B-4 and Figures 5
and 6 of the main report. From Table B-3 it is seen that the number of terms in
the regression equation for Y-. was 11, leaving 16 degrees of freedom; the resi-
dual standard deviation was 0.6683, and the degree of determination R = 81%.
161
-------
Table B-2
EFFECTS OF EM PROCEDURAL FACTORS ON SQUARE ROOT OF ESTIMATED NUMBER OF FIBERS
PER CUBIC CENTIMETER OF AIR, FROM EQUATION 9
xl
X2
X3
X6
X8
x9
xto
X11
XT2
Factor
Composition of
Sample
Concentration
on Filter
Sampler Type
Filter Side
Carbon Coating
Transfer Method
Magnification
Grid Opening
Location
Choice of
Fields
Level
100% Chrysotile
60% C + 40% Amphibole
70% C + 20% Amphibole
+ 10% Fiberglass
Light
Med i urn
Heavy
High volume
Personal
Particle side down
Particle side up
Yes
No
Soxhlet 1
Soxhlet 2
Jaffe
5,ooox
10,OOOX
20.000X
Per i phery
Mid-radius
Center
Random, small
Consecutive, small
Entire grid opening
Relative
Effect
-0.292
1-270
-0.978
2.140
-0.852
-1.288
-1.622
0.811
1.001
-2 . 002
3.202
-1 .601
-0.179
-3.359
3.538
-3.444
-1.642
5.086
0.936
-1.872
0.936
1.358
1.358
-2.716
Standard
Error
0.518
0.518
0.518
0.518
0.518
0.518
0.518
0.259
0.259
0.518
0.518
0.259
0.518
0.518
0.518
0.518
0.518
0.518
0.259
0.518
0.259
0.259
0.259
0.518
80% Confidence Limits
Lower
-0.990
0.572
-1.677
1.442
-1.551
-1.987
-2.321
0.461
0.651
-2.701
2.503
-1.951
-0.877
-4.057
2.839
-4.142
-2.341
4.388
0.586
-2.571
0.586
1.008
1.008
-3.415
Upper
0.406
1.968
-0.279
2.838
-0.153
-0.589
-0.923
1.161
1.351
-1.303
3.901
-1.251
0.519
-2.661
4.237
-2.746
-0.943
5.784
1.286
-1.173
1.286
1.708
1.708
-2.017
162
-------
Table B-3
PERFORMANCE EQUATION NO. 10
Dependent Variable:
Number of Tests:
Number of Terms in Equation:
Residual Degrees of Freedom:
Residual Standard Deviation:
2
Degree of Determination, R :
Independent Variable
Constant term
X0= '
Compos i t i on of samp 1 e
X^ (Coded -1, 1, 0)
Concentration on filter
X2L (Coded -1, 0, 1)
Filter type
X^Q (Coded -2, 1)
Pore size
X Q (Coded 1, -2, 1)
3 mm Portion Location
X?L (Coded -1, 0, 1)
Carbon coating
XgQ (Coded -2, 1)
Transfer method
XgQ (Coded 1, 1, -2)
Magnification
XinL (Coded -1, 0, 1)
X "0_ (Coded 1, -2, 1)
Y. = Log (estimated mass concen-
tration of fiber micrograms
per cubic meter)
27
11
16
0.6683
8U
Regression
Coefficient
b
0.286
0.331
-0.466
0.184
-0.121
-0.215
-0.351
-0.450
0.283
-0.159
Standard
Error
sb
0.158
0.158
0.091
0.091
0.158
0.091
0.091
0.158
0.091
Variance
Ratio
F
4.41
8.73
4.11
1.77
1.87
14.93
24.51
3.23
3-04
Choice of fields
X10Q (Coded 1, 1, -2) 0.134 0.091 2.18
163
-------
The various coefficients of the equation are listed along with their standard
errors and variance ratios.
Table B-4 then presents the net effects of the different levels of the
various factors on Y1f. together with their standard errors and 80% confidence
limits. The confidence intervals are presented graphically in Figures 5 and 6
of the main report.
164
-------
Table B-4
EFFECTS OF EM PROCEDURAL FACTORS ON NATURAL LOGARITHM OF ESTIMATED MASS
CONCENTRATION OF FIBERS (MICROGRAMS PER CUBIC METER), FROM EQUATION 10
xl
x2
x4
X5
X7
X8
X9
X10
X12
Factor
Composition of
Sample
Concentration
on Filter
Filter Type
Pore Size
3 mm Position
Location
Carbon Coating
Transfer Method
Magnification
Choice of
Fields
Level
100% Chrysotile
60% C + 40% Amphibole
70$ C + 203 Amphibole
+ 10% Fiberglass
Light
Med i urn
Heavy
Nuclepore
Millipore
0.22 ym
0.^5 ym
0.80 ym
Periphery
Mid-radius
Center
Yes
No
Soxhlet 1
Soxhlet 2
Jaffe
5,OOOX
10,OOOX
20.000X
Random, small
Consecutive, small
Entire grid opening
Relative
Effect
-0.331
+0.331
0
+0.466
0
-0.466
-0.368
+0.184
-0.121
+0.242
-0.121
+0.215
0
-0.215
+0.702
-0.351
-0.450
-0.450
+0.900
-0.442
+0.318
+0.124
+0.134
+0.134
-0.268
Standard
Error
0.158
0.158
0.158
0.158
0.158
0.158
0.182
0.091
0.091
0.182
0.091
0.158
0.158
0.158
0.182
0.091
0.091
0.091
0.182
0.182
0.182
0.182
0.091
0.091
0.182
80% Confidence Limits
Lower
-0.542
0.120
-0.211
0.255
-0.211
-0.677
-0.611
0.062
-0.243
-0.001
-0.243
0.004
-0.211
-0.426
0.459
-0.473
-0.572
-0.572
0.657
-0.685
0.075
-0.119
0.012
0.012
-0.511
Upper
-0.120
0.542
0.211
0.677
0.211
-0.255
-0.125
0.306
0.001
0.485
0.001
0.426
0.211
-0.004
0.945
-Ov229
-0.328
-0.328
1.143
-0.199
0.561
0.367
0.256
0.256
-0.025
165
-------
APPENDIX C
POISSON DISTRIBUTION TESTS
LISTING OF COMPUTER PROGRAM POISSON-1 FOR CHECKING CONFORMITY WITH THE POISSON
DISTRIBUTION
C PROGRAM POISSONi TO COMPARE OBSERVED AND CALCULATED EVENT DISTRIBUTIONS
C AND DETERMINE TMC GOODNESS OF FIT OF THE POISSON MODEL TO THE DATA,
C WRITTEN BY F C BOCK, IIT RESEARCH INSTITUTE,
REAL M,ML
DIMENSION VO(99),FOC99)»Fi(99),F2t99)»T£MPn2>
C—NO J3 THE NUMBER OF PAIRS OF VALUES OF VO AND FO TO BE READ IN
C—Nt
C— »VO(
C— F0(
C--F1
P*-F2
109
til
tl5
US
ur
S THE NUMBER OF CELL FREQUENCIES TO BE CALCULATED* FQR 0 TO N}»1 EVENTS
* 19 A SPECIFIED NUMBER OF EVENTS PER CELL* I*1»,.,«NO
) IS THE OBSERVED NUMBER OF CELLS WITH VOU) EVENTS PER CELL
S THE COMPLETE ARRAY OF OBSERVED CELL FREQUENCIES INCLUDING ZEROS
S THE COMPLETE ARRAY OF CALCULATED CELL FREQUENCIES
READ(5.1J1) NO, Nl, TEMP
PORMAT(2I<|,l2Afc)
«RITE(6*U3) TEMP
8TOP
PO*MATCB
-------
125
127
128
CHI2s99**2/F?(J)
w»lTt(6tl?7) Jl»F1(J)tf-2(J)»D»CHI2
rORMAT(lX»I6»Fll.',0»ri5.4i2F12,RKAT(/lx« 'TOTAL. OBSERVED CELLS »
/1X»'TOTAL CALCULATED CELLS »
/1X>'TOTAL OBSERVED EVENTS *
/IXi'TOtAi. CALCULATED FVENTS »
//tx» 'STATISTICS APPLYING TO NO, EVENTS PER CELL**/
/1X»IHF.AN EVENTS PER CELL «
/1X»'SUM OF SQUARED DEVIATIONS » »F10.4
/IXs'DEliRLES OF FRFEDOM a iF10.4
/Itlt 'VARIANCE s »F10.4
/IXt»STANOARD DEVIATION s iFlO.fl
/1X» 'STANDARD hR«QR OF MtrAN a
*»ITE(6»133)
FORMAT('IDVtRALL CHl-SQUAPE TEST FOR GOODNESS OF
tRlCS {NO. LVEMT8 PER CELL) COMBINED'/' SO THAT NO
2| QUCNCV IS LESS THAN "i.(>'//• RANGE OF»/
3 1X,'MQ. EVCNTS OBSERVED CALCULATED DIFFERENCE'
4 IX,' PER CELL NO. CELLS NQ, CELLS 0*C
52/C'/)
KsO
FIT» WJtTH CATE60
COMPUTED CELL FR
(0»C)**
RF1=0,
TCHI2B9.
00 139 J=lfN1
IF(J.LT.Ni) GOTO 135
F2.r,F-.3.')J GOfO 137
135
137
1375
GOTO
1F(«F2.LT.3.0) r.OfO 139
IF(K.Ltf.l) GOTO 1385
DsRFl 1-^22
CHJ2=D**2/HF2?
167
-------
TCHl2sTCHI2*CH!2
wRlTt(6.138) Jilt J??»RFUtRF22iDieMl£
138 pORMATUX»I4f i-»»I3iF10.0tF15.'li2F12,4)
1585 J11*Jl
J22SJ2
JtaJ
RF1»0.
RF2«0.
IF(J.LT.NUOR.HAKK.EQ.l) GOTO 139
GOTO 1375
15*» CONTINUE
NCPSK-?
w»lTr.(6?l
-------
The computer printouts from program POISSON-1 for two typical cases are
given in Tables C-l and C-2. In the top segment of each printout, the possible
numbers of fibers per field (labeled "events per cell") are listed at the left:
0, 1, 2, 3, etc. The succeeding columns are: the observed numbers of fields
having the specified numbers of fibers in them, the corresponding calculated num-
bers of fields based on an assumed Poisson distribution, and the differences
between the observed and calculated numbers of fields.
The next segment of the computer output provides the following overall items
of information on the sample. (1) The total number of fields observed (F) . (2) Th.
total calculated number of fields assuming that the Poisson distribution applies;
this is made equal to the observed number. (3) The total number of fibers observed
including those crossing the field perimeter as well as those lying entirely within
their field. (4) The total calculated number of fibers; this is made equal to the
observed number. (5) The mean number of fibers per field; this is the sample value
of the Poisson parameter A. (6) The sum of squared deviations around the mean.
(7) The degrees of freedom, one less than the number of fields. (8) The variance.
(9) The standard deviation, i.e., the square root of the variance. (10) The stand-
ard error of the mean. Items (6) through (10) are the usual sample statistics,
computed by treating each field as a unit of observation without reference to the
Poisson distribution.
The final segment of each printout gives the results of the goodness-of-fit
test for the Poisson distribution. The classes defined by the number of fibers
per field are grouped to the extent necessary for each of the calculated numbers
of fields to be no smaller than 3.0 so that the Ghi-square values are not unduly
inflated. For example, in sample 1 the classes after grouping are: 0 fibers per
field, 1 fiber per field, 2 fibers per field, and 3 or more fibers per field.
The smallest calculated number of fields is 7.32 for the last grouped class, and
the corresponding observed number is 17.
The succeeding columns are the observed numbers of fibers in the classes
after grouping the calculated numbers, the differences between the observed and
calculated numbers, i.e. 0-C, and the contributions to Chi-square, i.e. (0-C) /C.
The final items of information are the number of classes after grouping, the
degrees of freedom associated with the total Chi-square value, and the total Chi-
square value itself. The number of degrees of freedom is two less than the number
169
-------
Table C-l
PRINTOUT OF RESULTS FROM POISSON-1 PROGRAM
~f "tCase of poor agreement with Poisson model)
"WO.
PER CELL,
I
2
4
5
OBSERVED
NO, CELLS
7
ft
1
22,
12.
4,
1.
0,
0,
CALCULATED
NO. CELLS
97,3504
70,0923
25.2332
6.0560
1.0901
.1570
,0l«f
,0019
,0002
.0000
DIFFERENCE
0-C
29,0923
•3.2332
2.9099
-.1570
.9981
•0002
.0000
(1) tOTAt-0S»C*VED CEtLf «
<2) TOfAL CALCULATED CELLS •
(3) TOTAL OBSERVED EVENTS «
(4) WAtrtatcotimro EVENT*
100,0000
200,0000
ua,oooo
144,0000
N0§ EVENT8
(5) MEAN EVENTS PER CELL *
(6) SUM OF SQUARED DEVIATIONS •
(7) -&f«RE«-UF FREEDOM • " "~
(8) VARIANCE «
(9) STANDARD DEVIATION *
<10) fTANDAWO ERROR OF
.7200
246,3200
199,0000
1,2378
1,1126
.0787
.y. EVENTS
PER CELL
0- 0
I- 1
2" 2
OBSERVED
NO. CELLS
120*
41.
22,
17,
CALCULATED
NO, CELLS
97.3504
70,0923
25.2332
7,3240
DIFFERENCE
o-c
22,6496
•29,0923
•3,2332
9,6760
5,2696
12,0710
• 4U3
12.7634
NO, CLASSES AFTER GROUPING • 4
DEGREES OF FREEDOM • I
—TflTAl CHI-SOUARE • - 30.5423
-------
Table C-2
PRINTOUT OF RESULTS FROM POISSON-1 PROGRAM
26 (case of good agreement with Poisson Model)
EVENTS
PER CEU
i
2
_-5
4
5
6
7
8
9
10
H
12
OBSERVED
NO, CEILS
38.
42.
32,
16.
11,
0.
0,
It
0.
0.
0.
0.
0.
CALCULATED
NO* CELLS
32.3740
47,4048
14,7071
16,9404
6.2014
U8161
,4432
.0927
,0170
.0026
,0004
.0001
.0000
DIFFERENCE
o»c
5,62*0
-5.4048
-2,7071
••9404
4,7986
•1,8161
.,4432
,9073
.0170
,0026
• 0004
,0001
,9000
(l)TOTAL OBSERVED CELLS «
(2mf*t™C*tmATED CELLS •
(3)TOTAL OBSERVED EVENTS *
(4)TO?AL CALCULATED EVENTS «
140.0000
140*0000
205,0000
205.0000
STATISTICS APPLYING TO NO, EVENTS PER CELL
1.8643
238.8214
139,0000
j,7isi
1,3106
,1108
(5)*f*trtVENtfr ?ER CELL »
(6)SUM OP SQUARED DiVIATSQNS
(7)DESREE3 OF FREEDOM •
(9)STANDARO DEVIATION •
(lO)STANDARD ERROR OF MEAN •
NO. EVENTS
PER CELL
0"
1*
~2'tf
3"
4*
0
1
2
OBSERVED
NO, CELLS
38.
42,
32.
16,
12.
CALCULATED DIFFERENCE
NO, CELLS 0-C
32.3740
47,4048
34.7071
16,9404
8,5736
5,6260
•5,4046
•2,7071
•.9404
3,4264
,9777
,6162
,2112
,0522
1,3693
NO, CLASSES AFTER SROUPINS « 5
DEGREES OF FREEDOM • 3
TOTAL CHX«S8UARE » 3.2266
-------
of classes because two constraints based on the data were imposed in computing the
Poisson sequence, i.e., equality of the observed and computed number of fields and
equality of the observed and computed number of fibers. Or, in other words, the
two parameters F and X were evaluated from the sample data.
Sample 1 is an illustrative case in which there is poor agreement between the
actual data and the Poisson model (P < .001). On the other hand, sample 26 is a
case in which there is good agreement (.5 > P > .3). Inspection of the observed
and calculated numbers of fibers per cell for sample 1 after pooling (bottom seg-
ment of the computer printout) reveals the pattern of departures from the Poisson
frequencies: there is an excess of observed fields with no fibers, and also an
excess of observed fields with three or more fibers, as compared with the calculated
frequencies; these excesses are of course balanced by deficiencies in the observed
fields with one or two fibers in them. This is the general pattern to be expected
if there is a tendency for fibers to aggregate beyond that would occur simply by
chance settling. In all cases in which there was a poor fit of the Poisson distri-
bution the same type of pattern occurred in the departures of the observed frequen-
cies from the calculated.
172
-------
LISTING OF COMPUTER PROGRAM POISSON-2 FOR OBTAINING PERCENT CONFIDENCE LIMITS
ON THE MEAN OF A POISSON VARIABLE
C PROGRAM POISSON2 TO COMPUTE CONFIDENCE LIMITS FOR THE EXPECTED VALUE
C OF A POISSON RANDOM VARIABLE IN A SPACE OF THE SIZE EXAMINED (THE
C SAMPLE SPACE) AND IN THE REFERENCE SPACEt WHICH MAY DIFFER IN SIZE
C PROM THE SAMPLE SPACE,
C WRITTEN BY F C BOCKt IIT RESEARCH INSTITUTE.
C PCCONF IS THE CONFIDENCE LEVEL. I, t,» THE PERCENT PROBABILITY THAT THE
C L-XPEC.TED VALUE OF THt POISSON VARIABLE LIES NITKIN THE COMPUTED
C CONFIDENCE INTLRVAL
c SI*EI is THE NUMBER OF SPATIAL UNITS IN THE SAMPLE SPACE
C SJfE2 IS THE NUMBER OF SPATIAL UNITS IN THE REFERENCE SPACE
C C IS THE OBSERVED NUMBER OF OCCURRENCES (COUNT) IN THE SAMPLE SPACE
C RMEAN IS THE OBSERVED NUMBER OF OCCURRENCES TRANSLATED TO THE REFERENCE SPACE
C LCLi IS THE LOWER CONFIDENCE LIMIT FOR THE EXPECTED VALUE OF C
C UCL1 IS THE UPPER CONFIDENCE LIMIT FOR THE EXPECTED VALUE OF C
C LCL2 IS THE LOWER CONFIDENCE LIMIT FOR THE EXPECTED VALUE OF RMEAN
c ucL2 is THE UPPER CONFIDENCE LIMIT FOR THE EXPECTED VALUE OF RMEAN
REAL LCLt(LCL2
DIMENSION LABELS)
101 FORMAT<» CONFIDENCE LIMITS FOR THE EXPECTED VALUE OF A POISSON VAR
SIABLE IN A SPACE OF THE SIZE EXAMINED (THE SAMPLE SPACE)!/
S I AND ALSO IN THE REFERENCE SPACE*)
WRITE(6»103)
103 FORMATC'U CONFJDENCE'tSX'SIZE OF I i SX'OBSERVED* »4X»LIHITS ON TME«»
S TX*SIIE OF*»3XmFERENCE»«4X*LIMITS ON THE'*/* PROBABILITY* t«X
S *SAMPLC1F»2Xr«FREQUENCYif3X»EXPECTATION OF C' »/IX 'REFERENCE* i2X
S «FREeUENCY*t3X*EXPECTATIOM OF M*/3X *Pf RCENT «,6X »SPACE« »7X»C
S »LOWER ' f1X» UPPER* ?7X* SPACE1* 6X *M i i8X*LOWER* »<|Xt UPPER • »3X
$ »CAS£ DESCRIPTION*/)
HI READ(5fH3) PCCONF»Ct SIZE! tSI?E2. LABEL
FORMAT (4FS.-0»6A6)
iP^PCCqMF.eE.^W^*,) STOP
AUPMA*(100»PCCONF)/200
RNUa2*C
LCll«CHIOFPU»ALPHA»RNll)/2
RNU*2»(C+1)
gCLl«CHIOFP< ALPHA »RNU)/2
FACTOR*SIZE2/SXZEl
LCL2«FACTOR*LCL»
UCL2»FACTOR*UCL1
WRITE{6il2l) PCCONF, SIlEttC»LCLliUCLl»SIfE2rRMFA»*tLCL2»UCL2»LABEL
121 FORMAT<4XF5»2»'lXF8,?,2XF9,3,2X,2(F».3»lX),3XP8.2t3XF9,3»2X,
S 2(F8,3tlX),6A6)
GOTO 111
REAL FUNCTION CHIOFP(P»RNU)
C CHIOFP COMPUTES AN APPROXIMATE VALUE OF CHl-sSQUARE FOR GIVEN PROBABILITY
C P fTNTEGRAL FROM CHISQUARE TO INFINITY) AND DEGREES OF FREEDOM RWU,
C RNU SHOULD NOT BE LESS THAN 30. FROM HANDBOOK OF MATHEMATICAL FUNCTIONS.*
C NBS APPLIED MATH. SERIES 55* 26.4.17
173
-------
CMIQFPs-t.
ircp.LL'.o.oR.p.tit.i j
RETURN
RfAL FUNCTION XCIFP(P)
p COMPUTES AN APPRQXlMATK VALUF OF THE STANDARDIZED NORMAL DEVIATE
C X AS A FUNCTION OF .'fug PROBABILITY P (II IT P Lfc O.S AND P IS THE AREA
C UKPCR THE NORNAL DENSITY CURVE TO THK RIGHT OF X). FROM HANDBOOK OF
c MATHEMATICAL KINC.TIONS* wes APPtito MATH, SERIFS V.M afe.a.as (HASTINGS)
XOFPS999.
ircp.LL'.t'.oR.p.&e.n
PlsP
IP(P.GT.O.b) Pt»1-P
rsSQRT(AtOG(l/lP»*Pl)»
$ +.10I508*T»T*T))
IFfP.GT.O.S) XOFPs-XOFP
RETURN
•XOT
9«3. 93. I. /I/I 1. CHRYSOi 2H3t STAND
95. ott. ,72 1, CHRYSO* 2'2tt STAND
9«J. 209. 2.J6 I. CMRYSCt 2132. STAND
99999
174
-------
APPENDIX D
OPTIMIZED METHOD FOR MEASUREMENT OF AIRBORNE ASBESTOS CONCENTRATIONS
PROVISIONAL METHODOLOGY
Short Form
(1) Collect an air sample on a Nuclepore filter, pore size 0.4 ]am, using
a high-volume or personal samplers.
(2) Coat filter portion with about 40 nm thick carbon film using a vacuum
evaporator..
(3) A 60 or 100 mesh stainless steel mesh is placed on top of a filter
paper stack or form sponge contained in a petri dish. Chloroform is carefully
poured into the petri dish until the level is just touching the stainless steel
mesh. A 2.3 mm diameter (or 1 mm x 2 mm) portion of carbon coated filter is put
particle side down on a 200 mesh carbon coated copper electron microscope grid
and this pair placed on the steel mesh. The 2.3 mm diameter (or 1 mm x 2 mm)
portion is wetted with a 5 yJl drop of chloroform. The Nuclepore will be dis-
solved away in 24 to 48 hours. Chloroform may be added to maintain the level
in the petri dish.
(4) Examine the EM grid under low magnification TEM to determine its suit-
ability for high magnification examination. Ascertain that the loading is suit-
able and is uniform, that a high number of grid openings are intact, and that
the sample is not contaminated.
(5) Systematically scan the EM grid at a magnification of 20,OOOX (screen
magnification 16,OOOX). Record the length and breadth of all particles observed
if they have an aspect ratio of 3:1 or greater and substantially parallel sides.
Observe the morphology of the fiber using the 10X binocular and note whether a
tubular structure characteristic of chrysotile asbestos is present. Switch into
SAED mode and observe the diffraction pattern. Note whether the pattern is
typical of chrysotile or amphibole asbestos, or whether it is ambiguous, or
neither chrysotile or amphibole.
175
-------
(6) Count 100 fibers in several grids squares, or alternatively count all
fibers in at least 10 grid squares. If more than 300 fibers are observed in
one grid square, then a more lightly loaded filter sample should be used. If
no other filter sample can be obtained, the available sample should be trans-
ferred onto a 400 mesh grid. Processing of the sample using ashing and Boni-
fication techniques should be avoided wherever possible.
(7) Fiber number concentration is calculated from the following
/ 3 Fiber Count . Total Filter Area
s/m No. Fields Counted * Area of a Field
1
Volume of Air Sampled
Fiber mass for each type of asbestos is calculated by assuming that the breadth
measurement is a diameter, thus the mass can be calculated from
Mass (ym) = j- length (jam) • [diameter (ym)] • density (g/cm ) 10
The density of chrysotile is assumed to be 2.6 g/cm , amphibole 3.0. The mass
concentration for each type of asbestos is then
Mass Concentration m ,. i « jrxn T^T. c mi. m / \
t i 3\ c -D *.- i Total Mass of All Fibers of That Type (yg)
(yg/m-3) of a Particular = „ /r—K&
Type of Asbestos Volume of Air Sampled (m )
(8) Other parameters characterizing the asbestos fibers are:
(a) Length and width distributions of chrysotile fibers.
(b) Volume distribution of chrysotile fibers.
(c) Fiber concentration of other types of asbestos species.
(d) Relative proportion of chrysotile fibers with respect to total
number of fibers.
176
-------
APPENDIX E
ESTIMATING CHRYSOTILE MASS QN AIR FILTERS USING NEUTRON ACTIVATION TECHNIQUE
m
I IT Research Institute
10West 35 Street. Chicago, Illinois 60616
312/567-4000
March 22, 1977
Dr. Arthur Morgan
Environmental and Medical Sciences Div.
Building 551
AERE Harwell, Oxfordshire
0X11 ORA
England
Dear Dr. Morgan:
Pursuing the telephone conversation of Dr. Harwood with you on
18 March, I have prepared a few Nuclepore membranes for analyzing
chrysotile by neutron activation. The samples are as follows:
Sample No. Area of Membrane
142 4-1/4" x 6"
154 4-1/4" x 4-3/4"
2 8" x 10" (Blank)
4 8" x 10" (Blank)
Transmission electron microscopic study of sample 142 gave us an
estimate of about 0.07 Ug/cm . Since I am providing 4-1/4" x 6" area
of this membrane, I hope there is enough mass of chrysotile for an
accurate measurement by neutron activation. I am also enclosing a
sample of asbestos used for preparing the membrane samples.
Please analyze these samples at your earliest and bill us.
Yours very truly,
Anant V. Samudra
Research Scientist
AVS/eb
encl.
177
-------
n II
ti--«l*—-»
Environmental and Medical
Sciences Division , 3551
AERE Harwell, Oxfordshire ' .
0X11 ORA
Tel: Abingdon (0235) 24141 Ext. ^22
Telegrams: Aten, Abingdon
Telex 83135
Date 31st May 1977
Dr C F Harwood
IIT Research Institute
10 West 35 Street
Chicago
Illinois
USA
Dear Colin
We have attempted to assess the amount of chrysotile on the membrane filter
samples you sent using neutron activation analysis. Unfortunately, however,
the amount of chromium on the blank filters is sufficient to prevent an
accurate determination of fibre at the level required, using the 5°Cr(nY) Cr
reaction. I have discussed the possibility of using infra-red analysis with
people in our analytical group and they do not feel that measurements can be
made at the level required using this technique either.
There will of course be no charge for this work but I should point out that
£600 is still outstanding for our work on the fibre loaded filters.
I understand that Dr Hearsey of our Marketing and Sales Department has written
to you about this.
With best regards
Yours sincerely
A Morgan
17"8
-------
APPENDIX F
X-RAY FLUORESCENCE ANALYSIS OF STANDARD
SAMPLES OF CHRYSOTILE
An independent means for measuring the mass concentration of asbestos in
filter-deposited samples is needed if such samples are to be useful in deter-
mining the accuracy of mass concentration estimates made by electron micro-
scopy.
Air sample 154, used in the inter-laboratory comparisons of the pro-
visional optimal procedure, consisted of high purity chrysotile deposited on
Nuclepore filters in an aerosol chamber. X-ray fluorescence (XRF) analysis of
these deposits for Mg provided a convenient, independent and non-destructive
means for determining the mass concentrations of chrysotile on these filters.
The XRF measurements were carried out at the EPA Environmental Sciences
Research Laboratory (Research Triangle Park, North Carolina) with a simultaneous
multiwavelength spectrometer (Siemens MRS-3) adapted for air pollution samples
using procedures described by Wagman [65]. Fluorescence intensities above back-
ground of the Mg K line were measured using 1000-second counting intervals.
The calibration standard consisted of a vacuum-evaporated film of Mg deposited
2
at a concentration of 47 yg/cm on mylar film. The Mg K sensitivity was 70.73 cps
o
per yg/cm and the minimum detectable limit for a 1000-second count, on the basis
2
of 3a above background, was 1 ng/cm .
Precision analysis of a series of XRF measurements of chrysotile deposits
indicated a relative standard deviation of less than 4 percent. Particle size
and other fluorescence attenuation correction factors were not needed because
EM examination of chrysotile deposits showed that all fibers had diameters under
0.1 ym with only rare instances of fiber overlap. The method of computation is
illustrated as follows.
179
-------
SAMPLE CALCULATION METHOD
Method: Measurement of Mg by XRF using Siemens MRS-3
Chrysotile Concentration = 3.8 x Mg Concentration
9
Magnesium: Vacuum evaporated Mg on Mylar and uniform Mg cone, of 47 yg/cm
Standard
S (Mg sensitivity) = 70.7315 counts/sec/yg/cm2
N (Nuclepore background) = 640 counts/1000 seconds
B
3i/N~
Minimum Detection Limit B
for 1000 sec counting inno
Mg X
= 0.00107 yg/cm2
MEASUREMENTS ON NUCLEPORE FILTER 154
Measurement
1
2
3
4
5
6
Counts /sec -blank
0.980
1.022
1.019
1.076
1.017
1.090
yg Mg/cm
0.0139
0.0144
0.0144
0.0152
0.0144
0.0154
Mean = 0.0146 yg Mg/cm"
2
Standard Deviation = 0.00057 yg Mg/cm
Relative Std. Dev. = 100 x (Std. Dev)/Mean = 3.9%
2
Particle size correction factor, (1 + ab) , is very
small (within Std. Dev.) and hence neglected.
Assuming all Mg is in chrysotile form, and that
Chrysotile Mass Concentration = 3.8 x Mg Concentration
Chrysotile Mass Cone. = 3.8 x (0.0146 + 0.00057) yg/cm2
= 0.0555 + 0.0022 yg/cm2
180
-------
3
Total Volume of Air Sampled = 9.2 m
2
Total Filter Area = 406.5 cm
Air Volume/cm2 of Filter = 0.02263 m3/cm2
Chrysotile Mass Cone. _ 0.0555 ± 0.0022
in Air Sampled '
= (2.452 + 0.096) yg/ra3
The X-ray fluorescence measurements of chrysotile are
summarized in Table F-l.
Table F-l
MEASUREMENT OF CHRYSOTILE MASS
CONCENTRATIONS BY XRF
ANALYSIS FOR MAGNESIUM
Sample No.
154
142 C
154 A
154 B
168 D
Mg Cone.
yg/cm
0.0146
0.0117
0.0111
0.0108
0.0120
Chrysotile Cone.*
/ 2
yg/cm
0.0555
0.0445
0.0422
0.0410
0.0456
ygM3
2.452
1.966
1.865
1.812
2.015
* A chrysotile/Mg factor of 3.8 was used.
The aerosol volume sampled per unit
filter area was 0.02263 m3/cm2.
181
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA 600/2-78-038
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
EVALUATING AND OPTIMIZING ELECTRON MICROSCOPE
METHODS FOR CHARACTERIZING AIRBORNE ASBESTOS
5. REPORT DATE
June 1978
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
A.V. Samudra, F.C. Bock, C.F. Harwood, and J.D. Stockham
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
IIT Research Institute
10 West 35th Street
Ihieago, Illinois 60616
10. PROGRAM ELEMENT NO.
1AD712 BA-14 (FY-77)
11. CONTRACT/GRANT NO.
68-02-2251
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory - RTF, NC
Office of Research and Development
U. S. Environmental Protection Agency
Research Triangle Park, N. C. 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final Report 6/75-6/77
14. SPONSORING AGENCY CODE
EPA/600/09
15. SUPPLEMENTARY NOTES
This report complements EPA Report 600/2-77-178 entitled "Electron Microscope
Measurement of Airborne Asbestos Concentrations — A Provisional Methodology Manual"
16. ABSTRACT
Evaluation of EM methods for measuring airborne asbestos fiber concentrations and size
distributions was carried out by studying a large number of variables and subprocedures
in a five-phase program using elaborate statistically designed experiments. Observa-
tions were analyzed by advanced regression techniques to evaluate the effects of
independent variables and subprocedures. It was shown that the optimized method for
estimating airborne chrysotile should have the following features: (a) collecting an
air sample on Nuclepore filter; (b) coating the Nuclepore filter with carbon; (c)
transferring the particulate deposit to a 200-mesh electron microscope grid using
chloroform in a modified Jaffe-wick washer; (d) examining the grid at about 10,000 x
nagnification (20,000 x for counting very fine fibers); (e) counting fibers using a
(field of view method; and (f) identifying the type of asbestos from morphology and
selected area electron diffraction.
provisional manual of instructions was prepared (EPA Report 600/2-77-178) and six
independent laboratories participated in an interlaboratory test of the proposed method
using two air samples. One of these was prepared at IITRI from pure aerosolized UICC
chrysotile, and the other was an ambient air sample collected by IITRI personnel in a
factory processing asbestos. Intercomparison of the results from the separate labora-
tories yielded some preliminary estimates of the precision and accuracy of the provi-
g-irmal
KEY WORDS AND DOCUMENT ANALYSIS
rAir pollution
'Asbestos
:Serpentine
'Amphiboles
Measurement
DESCRIPTORS
^Electron microscopy
*Electron diffraction
b.iDENTIFIERS/OPEN ENDED TERMS
Chrysotile
COSATl Field/Group
08G
HE
14B
8. DISTRIBUTION STATEMENT
IELEASE TO PUBLIC
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF PAGES
197
20. SECURITY CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2223-1 (Rev. 4-77) PREVIOUS EDITION is OBSOLETE
182
------- |