EPA-600/4-82-021
April 1982
BULK SAMPLE ANALYSIS FOR ASBESTOS CONTENT
EVALUATION OF THE TENTATIVE METHOD
by
E. P. Brantly, Jr.
K. W. Gold
L. E. Myers
0. E. Lentzen
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, NC 27709
Contract No. 68-02-3222
Project Officer: J. J. Breen
Project Officer: M. E. Beard
Office of Pesticides and
Environmental Monitoring Systems
Toxic Substances
U.S. Environmental Protection Agency
Washington, DC 20460
Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
-------
FOREWORD
Measurement and monitoring research efforts are designed to anticipate
potential environmental problems, to support regulatory actions by develop-
ing an in-depth understanding of the nature and processes that impact health
and the ecology, to provide innovative means of monitoring compliance with
regulations, and to evaluate the effectiveness of health and environmental
protection efforts through the monitoring of long-term trends. The Environ-
mental Monitoring Systems Laboratory, Research Triangle Park, North Carolina,
is responsible for: assessing environmental monitoring technology and sys-
tems; implementing Agency-wide quality assurance programs for air pollution
measurement systems; and supplying technical support to other groups in the
Agency including the Office of Air, Noise, and Radiation, the Office of
Pesticides and Toxic Substances, and the Office of Enforcement.
This report describes the results of an evaluation of a tentative method
developed for the measurement of asbestos in bulk insulation materials. The
method is designed to support the Asbestos-in-Schools Program of the Office
of Pesticides and Toxic Substances.
Thomas R. Hauser, Ph.D.
Director
Environmental Monitoring Systems Laboratory
Research Triangle Park, North Carolina
ii
-------
ABSTRACT
The U.S. Environmental Protection Agency Asbestos-in-Schools Program
was established in March 1979 to provide information and technical assistance
to the public for addressing problems presented by asbestos-containing
insulation materials in school buildings. Because there were no existing
standard procedures for the qualitative and quantitative analysis of asbestos
in bulk materials, the Office of Pesticides and Toxic Substances, Washington,
DC, and the Environmental Monitoring Systems Laboratory, Research Triangle
Park, NC, jointly sponsored an effort to produce a practical and objective
analytical protocol.
Draft procedures were written for the analysis of bulk samples by
polarized light microscopy (PLM) and X-ray powder diffraction (XRD) based on
information presented at a conference of interested parties from government,
university, and commercial laboratories. Following review by the conferees,
the Tentative Method for the Determination of Asbestiform Minerals in Bulk
Insulation Samples (March 1980) was submitted to a performance testing
program that involved multiple laboratory analysis of prepared samples with
known asbestos content. This report presents the results of the testing
study and provides observations and preliminary characterization of the
utility and operational parameters of the Tentative Method.
PLM quantitative analysis employs a point counting procedure to estimate
the relative area occupied by asbestos fiber within the microscope fields of
view. This must be compared with the known weight of asbestos in the sample
in order to characterize the accuracy of the method. Data produced by the
point counting procedure are compared with those produced by the typical
quantitation procedures used by some of the participating laboratories.
Accuracy and precision of the point counting procedure are considered in two
contexts: (1) as PLM is currently used, regarding reported data as a direct
estimate of weight percent of asbestos present; and (2) allowing adjustments
f the data to account for bias and variance in the relationship between the
i i i
-------
relative area occupied by asbestos and the known weight percent of asbestos
in the sample. Information is also presented on within-laboratory variance
and the frequency of false negatives and fplse positives.
A very limited amount of data was returned for characterizing the XRD
protocol. Both thin-layer and thick-layer (bulk) techniques were used for
quantitative XRD analysis. Because of the small number of XRD reports, and
the nonequivalence of methods employed, it is not possible to draw any firm
conclusions on the precision and accuracy of the XRD protocol. A general
comparison of bulk and thin-layer techniques with respect to precision,
accuracy, and sensitivity is made.
Comments received from participating laboratories and recommendations
for continued investigation of asbestos bulk sample analysis are presented.
This report is submitted in fulfillment of Contract No. 68-02-3222 by
Research Triangle Institute under the sponsorship of the U.S. Environmental
Protection Agency. This report covers the period March 1, 1980, to Decem-
ber 31, 1980, and work was completed as of April 10, 1981.
iv
-------
CONTENTS
Section Page
Foreword ii
Abstract ill
Figures vii
Tables viii
1 Introduction 1
2 Summary 4
3 Recommendations 9
3.1 Polarized Light Microscopy 9
3.2 X-Ray Powder-Diffraction 11
3.2.1 Identification and Characterization of
Standard Reference Materials 11
3.2.2 Further Development and Evaluation of
Thin-Layer and Bulk Methods of Analysis 12
4 Analytical Methods 14
5 Design of the Evaluation Study and Data Analysis 22
5.1 Study Design 22
5.2 Sample Preparation 22
5.3 Definitions of Laboratory Groups 25
5.4 Statistical Definitions 27
5.5 The relationship Between Area Percent and Weight
Percent Estimates 28
5.6 The Area/Weight Relationship in Point Count Data. ... 29
6 Method Evaluation: Polarized Light Microscopy . . 37
6.1 Bias of the PLM Data 37
6.2 Precision of the PLM Data 40
6.3 Accuracy and Precision, After Data Transformation ... 45
6.4 Estimation of Within-Analyst Variance 50
6.5 False Positives and Negatives for Point Count
Data 52
6.6 Operating Characteristic Curves 54
6.7 General Observations 58
6.8 Conclusions 59
7 Method Evaluation: X-Ray Powder Diffraction 61
8 Comments 70
9 References 73
v
-------
CONTENTS (continued)
Section Page
Appendixes
A Statistical Procedures A-l
B Tentative Method B-l
C Participating Laboratories C-l
D Instructions and Reporting Forms D-l
E Conference Participants E-l
vi
-------
FIGURES
Number Page
1 Examples of X-ray diffraction patterns . 17
2 Group P data, by sample series 30
3a Group P data for samples containing chrysotile 31
3b Group P data for samples containing amosite 32
4a Natural logarithms of Group P data for samples
containing chrysotile 33
4b Natural logarithms of Group P data for samples
containing amosite 34
5 Ninety-percent confidence intervals for means of
Group P data 39
6 Average percent bias of Group P data 42
7 Group standard deviations 43
8a Group P standard deviation vs. asbestos weight percent ... 44
8b Group P standard deviation vs. Group P mean 44
9 Coefficient of variation of Group P data 46
10 Theoretical operating characteristic curves 55
11 Operating characteristic curves for Group P data 56
12 Comparison of bulk XRD analysis by asbestos type 66
13 Comparison of thin-layer XRD analysis by asbestos type ... 67
-------
TABLES
Number Page
1 Sample Composition 23
2 Quantitative Results of PLM Analyses 26
3 Means and Standard Deviations of Reported PLM Results,
by Group 38
4 Average Percent Bias of Group P Data 41
5 Means and Standard Deviations of Predicted Weight, by
Group 48
6 Average Percent Absolute Error: Unadjusted vs.
Transformed PLM Data, by Group 49
7 Analysis of Duplicate Samples, Group P 51
8 Reported XRD Results 62
9 Means and Standard Deviations of Reported XRD Results. ... 63
10 Average Absolute Errors of Reported XRD Results 63
11 Coefficients of Variation of Reported XRD Results 65
vi i i
-------
SECTION 1
INTRODUCTION
Inhalation of asbestos fibers was first recognized and managed as an
occupational health hazard in Great Britain. Epidemiological and experi-
mental studies have since provided extensive evidence of increased risk of
pulmonary fibrotic disease (asbestosis), pleural and peritoneal mesothelioma,
and other carcinomas due to both occupational and nonoccupational exposure
to asbestos fiber.1
The U.S. Environmental Protection Agency (EPA), in cooperation with the
Consumer Product Safety Commission and other Federal agencies, has investi-
gated the potential for hazardous exposures from asbestos-containing products.
The life-cycle of manufactured goods was followed from mining and milling of
asbestos ore through disposal of the used manufactured products. Significant
exposures Were discovered in several use categories, including the applica-
tion of sprayed-on insulations, applicat n of patching plaster, and use of
asbestos filters in food and drug processing.
The U.S. Government has sought to limit exposure to asbestos through a
variety of legislative and regulatory actions.2 Asbestos was listed as a
potential hazardous air pollutant in 1971 (36 FR 5931) and airborne emis-
sions were regulated under the Clean Air Act in 1973 (38 FR 8820). Most
notable to this study was the action taken in April 1973 to ban the spray
application of insulation products containing more than 1 percent asbestos
by weight (38 FR 8820).
In March 1979, EPA established the Asbestos-in-Schools Program to
provide information and technical assistance to the public for addressing
the problems presented by asbestos-containing materials in school buildings
(44 FR 54676). EPA has published several guidance documents3 4 that contain
technical information on the identification and control of potential exposures
to asbestos fibers. The guidance documents and other information are avail-
1
-------
able through toll-free telephone numbers maintained by the Research Triangle
Institute and the EPA Office of Pesticides and Toxic Substances (OPTS)
Industry Assistance Office. Additional assistance is available 1rom EPA-
designated Regional Asbestos Coordinators.
EPA has recommended that polarized light microscopy (PLM) be used for
estimating the asbestos content of bulk samples, to be supplemented with
X-ray powder diffraction (XRD) should add'^onal information on the sample
be required.3 Part I of the guidance document includes guidelines for PLM
and XRD analysis, but notes the lack of standard protocols for either PLM or
XRD. It also notes the absence of any program for qualification of PLM or
XRD laboratories. (Since publication of the document, EPA has established a
voluntary quality assurance program5 for laboratories capable of PLM analysis
of bulk samples. Information on the program may be obtained from RTI.)
Because there were no existing standard procedures for the qualitative
or quantitative analysis of asbestos in bulk materials, the Office of Pesti-
cides and Toxic Substances, Washington, DC, and the Environmental Monitoring
Systems Laboratory, Research T-iangle Park, NC, jointly sponsored an effort
to produce a practical and objective analytical protoco1. The task was
initiated by reviewing existing literature and conferring with recognized
experts in the field. Results of this survey indicated that PLM is the most
appropriate analytical method, to be augmented by XRD when necessary.
In an effort to optimize the application of these techniques to the
specialized task at hand, a conference of knowledgeable and interested
parties from government, university, and commercial laboratories was con-
vened. The symposium "Methods Definition for the Polarized Light Microscope
and X-Ray Diffraction Analysis of Bulk Samples for Asbestos" was held at the
U.S. Bureau of Mines, Avondale Research Center, Avondale, MD, on October 23-24,
1979 (see Appendix E for attendees). Conferees discussed techniques for
sample preparation and qualitative and quantitative analysis by both PLM and
XRD.
Following this "conference, PLM and XRD analytical procedures were
drafted to reflect the best inputs or consensus agreemenffl*"Of the conferees.
Drafts were subsequently circulated twice for review by conference parti-
cipants and other professionals active in the analysis of asbestos. Follow-
ing review, the Tentative Method for the Determination of Asbestiform Mine als
2
-------
in Bulk Insulation Samples (March 1980) was submitted to a performance
testing program that involved multiple laboratory analysis of real-world
samples and prepared samples with known asbestos content. This report
presents the results of the testing program and provides preliminary observa-
tions and characterization of the Tentative Method's utility and operational
parameters. Recommendations for revision of the Tentative Method and for
further investigation of PLM and XRD analysis are also presented.
Revisions pursuant to the recommendations of this report and the comments
received in the evaluation study have been incorporated into the method.
The current revision of the method may be found in the EPA report Interim
Method for the Determination of Asbestos in Bulk Insulation Samples (October
1981).
3
-------
SECTION 2
SUMMARY
An interlaboratory study was conducted to evaluate the accuracy, preci-
sion, and general utility of the Tentative Method for the Determination of
Asbestiform Minerals in Bulk Insulation Samples (March 1980). Twenty-two
commercial and four government laboratories were each supplied with eleven
samples. Eight of the samples were formulated with a known weight of amosite
or chrysotile and a matrix material containing primarily gypsum. Within-
laboratory duplicates, blanks, and "real-world" samples of sprayed insulation
were also included in the materials distributed to laboratories. Four
laboratories (two commercial, two government) chose not to participate in
the study. The twenty-two participating laboratories provided a total of
thirty PLM reports and six XRO reports.
The Tentative Method includes procedures for qualitative and quantita-
tive analysis of bulk samples by polarized light microscopy (PLM) and X-ray
powder diffraction (XRD). Identification of asbestos fibers by PLM requires
the observation of six optical properties: morphology, color and pleochroism,
refractive indices (or dispersion staining colors), birefringence, extinction
characteristics, and sign of elongation. PLM quantitative analysis uses a
point counting procedure to estimate the percent area occupied by asbestos
fibers within the microscope fields of view. The prepared samples distributed
in this study contained a knoym weight percent of asbestos. Because PLM
• analysis produces an estimate of the relative area occupied by asbestos, the
relationship between reported area percent and the known weight percent of
asbestos was investigated.
Reported area percent data are best correlated with the known weight
percent values when regressions are performed in natural logarithmic coordi-
nates, indicating that the relationship between area and weight involves a
power transformation. Analysis of the regression shows that variation in
the relationship is attributable to differences between laboratories and to
differences between asbestos types (chrysotile and amosite).
4
-------
Reported PLM data were divided into three groups based on the quantita-
tion procedure(s) used by the reporting laboratory.
Group P—(Point count) PLM asbestos area percent determinations by
the point count procedure (Interim Method).
Group B—(Both) PLM asbestos area percent determinations by the
laboratories' own methods for laboratories that also
provided data by the point count method.
Group 0—(Other) PLM asbestos area percent determinations by the
laboratories' own methods for laboratories declining to
use the point count method.
Data in Group 0 (other) contributed by different laboratories were not
necessarily produced by the same quantitation procedure. Six laboratories
that contributed data to Group P (point count) also reported results produced
by their own quantitation procedures. This data set is designated Group B
(both). Four of the laboratories produced closer estimates of the true
weight percents by their own method than by point counting, while the other
two laboratories reported closer results by point counting. It is unclear
what, if any, relationship exists between the Group P and Group B data
contributed by any one laboratory. It is possible that an estimate produced
by point counting could have influenced a laboratory's own procedure, or
vice versa.
Considering reported PLM results as direct estimates of the weight per-
cent of asbestos present, it was found that Group 0 (other) is significantly
more biased than Group P (point count). Groups P and B (both) are similarly
biased. Point counting has a greater positive bias on amosite samples than
on chrysotile samples. For a sample containing 10 percent chrysotile by
weight, the average bias (b) of Group P (point count) is 18.5 percent; for
50 percent chrysotile, b = -24.2 percent; for 10 percent amosite, b = 118.5
percent; for 50 percent amosite, b = 12.1 percent.
A regression relating standard deviations and means of reported results,
when performed in natural logarithmic coordinates, did not establish any
difference between Groups P, B, and 0 with respect to precision. The stand-
ard deviation of Group P (point count) is directly related to the mean
reported value, and thus precision may be expressed as the coefficient of
variation (CV). The CV is less than 100 percent on samples containing more
than approximately 6 percent asbestos by area, and less than 50 percent on
5
-------
least 5 percent asbestos by weight, would result in three false negatives
with a probability less than 0.03 and possibly as lev* as 0.001.
Identification of sample components by XRD analysis is accomplished by
comparison of the sample diffraction pattern with standard reference powder
diffraction patterns. Quantitative analysis involves measuring the inte-
grated areas of diagnostic peaks selected from the full XRD scan of a thin-
layer sample. Quantitative analysis must include a correction for matrix
absorption effects and comparison with suitable external standards. XRD
affords information only on crystal lattice structure and not on crystal
morphology. XRD analysis, therefore, cannot distinguish between asbestos
minerals and their non-asbestiform varieties. The presence of fibrous
particles in a sample must be determined by an optical technique such as
PLM.
The six laboratories reporting XRD results were grouped into two general
categories for purposes of data analysis. These categories, thin-layer and
bulk, were defined on the basis of the XRD technique used for quantitative
analysis. Three of the laboratories performed the requested analyses using
some variation of the thin-layer method of quantitation included in the
Tentative Method. The remaining three laboratories used alternative bulk
or thick-layer methods of quantitation. It should be emphasized that within
categories none of the methods used were strictly equivalent. Moreover,
within the thin-layer group, no laboratory followed the Tentative Method
protocol exactly.
Because of the small number of participating laboratories reporting XRD
results, and the nonequivalence of methods employed, it is not possible to
draw any firm conclusions from the reported results about the accuracy and
precision of the XRD method. However, from a general comparison of bulk vs.
thin-layer methodology, the following observations can be made:
1. The bulk method appears to be at least as accurate and precise
as the thin-layer method over the range of samples included
in this study and significantly more accurate for the analysis
of chrysotile; and
2. There is a suggestion that the thin-layer method of analysis
may be more reliable (i.e., more sensitive) than the bulk
method at the 1 percent level of chrysotile in a simple
matrix.
7
-------
Data produced by thin-layer methods of analysis included one false
negative out of three analyses of the 4.9 percent chrysotile sample. The
same laboratory reported chrysotile false positives for all amosite samples
and for the blank sample with reported chrysotile values ranging from <1 to
8 percent. A second laboratory reported one false negative out of three
analyses in the 19.4 percent chrysotile sample.
Data produced by bulk methods of analysis included two false negatives ;
out of three analyses of the 1.2 percent chrysotile sample. One of these
laboratories also reported a false positive amosite in the 4.9 percent
chrysotile sample.
8
-------
SECTION 3
RECOMMENDATIONS
The study presented in this report is a preliminary evaluation designed
to determine the precision and accuracy of the Tentative Method as applied
to carefully prepared samples. It should be emphasized that the samples
analyzed consisted of only two types of asbestos fiber and a single matrix
material. Only one type of asbestos was included in any given sample. One
of the main obstacles to reliable analysis of bulk samples is the varia-
bility of sample composition. Complete characterization of the method
presented herein requires that several issues be addressed, as discussed
below. The highest priority, however, should be assigned to investigations
that will extend the application of the method to a range of real-world
samples involving different fiber types and matrices.
3.1 POLARIZED LIGHT MICROSCOPY
Several aspects of the PLM method require further investigation.
Briefly, future studies should be designed to determine the following:
1. The feasibility of specifying definitive sample preparation
procedures to be used prior to quantitative PLM analysis;
2. The proportion of total variance attributable to individual
procedures of the method, i.e., sample preparation, sub-
sampling, and point counting;
3. The proportion of total variance contributed by within-
laboratory variability;
4. The effect of specific variables within the point counting
procedure, including the number of points to be counted,
magnification used, and the possible bias introduced by the
use of a 25-point reticle instead of a cross-hair reticle;
5. The possibility of introducing a staged point counting process
that would allow fewer counts to be determined on samples
with s high percentage of asbestos;
9
-------
6. The effect of more than one type of asbestos being present in
* a bulk sample;
7. The feasibility of individually calibrating PLM laboratories
with information derived in round robin sample analysis
programs.
The protocol supplied to laboratories in this study contained a provi-
sion for reporting less than 1 percent asbestos in a sample if fewer than 7
of 400 points are scored for asbestos (Appendix B, PLM p. 8). This provi-
sion was based on an approach to the data involving hypothesis testing and
on the assumption that results of repeated analysis of samples with small
amounts of asbestos would fit a Poisson distribution. Sufficient data are
not currently available to support the Poisson assumption for analysis of
"real-world" samples. Additionally, the hypothesis testing approach is not
appropriate to the typical use of laboratory data. It is therefore recom-
mended that the provision be deleted and that the simple arithmetic percent-
age be used for determining asbestos content at all levels.
The confidence interval calculation (Appendix B, PLM p. 9) presently
included in the PLM protocol is misleading. It does provide a good estimate
of reliable bounds for the relative area occupied by asbestos fiber in the
examined fields of view. However, because of other sources of variation
(sampling, subsampling, sample and slide preparation), the confidence inter-
val may not be thought of as reliable bounds for the percent asbestos in the
material from which the sample was taken. It is therefore suggested that
the calculation of the confidence interval be deleted from the method.
Finally, it is apparent from the results of this study that some type
of training would be required to achieve comparable application of the PLM
protocol between laboratories. While point counting is a classical petro-
graphic technique, it is not a standard procedure in the majority of labora-
tories currently analyzing bulk samples for asbestos. Training alternatives
might include regional courses and distribution of split samples analogous
to the NIOSH program for the asbestos air sampling method.
It should also be noted that the PLM method presented, although an im-
provement over subjective techniques, is still a procedure for estimating
the relative area occupied by asbestos fiber and matrix material, and re-
quires an area-to-weight conversion to apply to the Federal standard
(38 FR 8820). Alternative analytical techniques that measure weight percent
-------
directly or that provide an empirically more satisfying relationship to
relative weight of asbestos fiber should be sought and investigated.
3. 2—X-RAY POWDER DIFFRACTION
There are two major areas in the application of XRD techniques to
quantitative analysis of asbestiform minerals in bulk materials that require
further investigation: (1) identification and characterization of standard
reference materials, and (2) further development and evaluation of thin-
layer and bulk methods of analysis.
3.2.1 Identification and Characterization of Standard Reference Materials
The most common concern of laboratories participating in the evaluation
of the XRD protocol was the lack of wel1-characterized, readily available
reference materials. Both NIOSH6 and the Bureau of Mines7 have conducted
rather extensive studies in this area; however, these materials are not
available in large quantities for general use. In addition, the UICC stand-
ards* are net exceptionally pure and have been reported to be in dwindling
supply. Therefore, a thorough, systematic investigation of asbestiform
minerals for use as standard materials should be undertaken. This should
include identification of major sources; determination of availability and
cost; and complete mineralogical characterization and determination of
purity, particle size distributions, and powder diffraction patterns of
materials from these sources.
Since asbestos minerals vary in composition depending on the source and
exhibit different behaviors in grinding, peak positions and/or relative
intensities of XRD patterns may vary from sample to sample. This variabil-
ity is particularly problematic for the amphibole minerals. A quantitative
study to assess the comparability of X-ray response of asbestos minerals
from different sources should be conducted. If possible, observed differ-
ences between different samples of the same asbestos variety should be
correlated with specific sample characteristics (e.g., chemical composition
and particle size).
^Prepared by the International Union Against Cancer. Available from: UICC,
MRC Pneumoconiosis Unit, Llandough Hospital, Penarth, Glamorgan CF6IXW, UK,
and commercial distributors.
11
-------
3.2.2 Further Development and Evaluation of Thin-Layer and Bulk Methods of
Analysis
The need for further development and evaluation of both thin-layer and
bulk methods of XRD analysis is underscored by the following observations:
few laboratories are currently set up to routinely perform the thin-layer
analysis as prescribed; the proposed thin-layer method of quantitation
is considerably more time-consuming and costly than bulk or thick-layer
methods; and for samples analyzed in the methods evaluation study, the
bulk method was at least as accurate and precise as the thin-layer method.
In particular, a comparison of the bulk and thin-layer methods should
be made over a variety of asbestos types and matrix materials, with atten-
tion given to sample preparation requirements, instrument, requirements,
sensitivity, precision, accuracy, and speed and cost of analysis.
For both bulk and thin-layer methods, the following areas of investiga-
tion are proposed:
1. Assessment of sample preparation requirements—The require-
ment to grind the sample and standards to a comparable particle
size of 10 |jm is essential for rigorous quantitative analysis.
It is recognized, however, t.hat this is often time-consuming
and costly and may not be feasible for some samples. In
addition, since .the matrix material itself may alter the
grinding characteristics of the asbestos, the validity of
standards prepared in a manner identical to the sample mater-
ials is questionable. A systematic investigation of the
effects of various grinding and matrix reduction techniques
(e.g., milling, ultrasonication) on the different asbestos
minerals in a variety of "common" matrices should be -onducted,
and changes in relative peak intensities, peak profiles, and
positions monitored as a function of such parameters as
grinding time, temperature, and type of mill.
2. Assessment of preferred orientation effects on quantitative
analysis—This should include evaluation of the dapendence of
preferred orientation effects on sample preparation techniques,
sample particle size, and sample substrate. For bulk methods,
filtration, back-packing, and pelletizing methods of sample
preparation should be evaluated; for thin-layer methods, the
effects of the filter medium and sample particle size should
be investigated. This could be extended to include an assess-
ment of the feasibility of preferentially orienting the
sample fibers prior to analysis to maximize reproducibility,
with evaluation of instrument requirements and applicability
to routine screening programs.
12
-------
3. Assessment of the effect of the use of the step-scanning mode
of analysis on the limits of detection—This should be evalu-
ated for both methods by comparing sensitivities obtained
with and without step-scanning for each asbestos mineral in a
variety of matrices.
4. Assessment of absorption correction requirements and techniques—
The use of an internal standard for absorption correction
should be systematically evaluated. For thin-layer methods
of analysis, the internal correction should be compared with
the proposed method of absorption correction by measurement
of the attenuation of a silver filter substrate peak. Evalu-
ation of the latter method of absorption correction requires
further assessment by XRD of the variability of the silver
content in silver membrane filters both between filters and
between front and back sides of the same filter.
Since XRD offers the possibility of rapid, sensitive, automated analysis
of asbestos at a time when a major increase in screening and monitoring
efforts is projected, it is hoped that the results of such investigations
will allow refinement of the present XRD method to one that is less costly
less time-consuming, and better suited to routine analysis of bulk materials.
13
-------
ACTION 4
ANALYTICAL METHODS
A Tentative Method has been developed for the analysis of asbestos in
bulk insulation materials by polarized l^'ght microscopy (PLM) and X-ray
powder diffraction (XRD) techniques. The Tentative Method is presented in
Appendix B. Procedures for qualitative and quantitative analysis have been
included to address Federal regulations that limit the asbestos fiber content
of sprayed insulation materials to 1 percent by weight (38 FR 8820).
Classical petrographic techniques are specified in the PLM protocol for
identification of asbestos fibers and other components of bulk samples.
SuDsamples of bulk material are prepared by appropriate techniques, immersed
in an oil of kno'..,- refractive index, and ex?mined with both single and
crossed polars. Asbestos fibers are positively identified by the observa-
tion of six optical properties: morphology, color and pleochroism, refrac-
tive indices (or dispersion staining colors), birefringence, extinction
characteristics, and sign of elongation.
There are several deterrents to the reliable quantitation of asbestos
in bulk samples by PLM, includi-ii] variable matrices, the small amount of
sample examined, and variation in the optical properties of the asbestos
minerals. Optical methods measure the relative area occupied by asbestos
fiber and matrix material within the microscope fields of view. At present,
most analysts using optical methods attempt the quantitation of asbestos
either by visual estimation or by comparison of the microscope field of view
with graphics prepared to correspond to area concentrations of 1 percent,
5 percent, 10 percent, etc. Such procedures have been shown in previous
studies8 to be highly variable because of differences among analysts in
training, experience, anct application.
Quantitation is performed ip. fie PLM procedure by a point counting
technique. Point sampling is used in various fields of study to estimate
the relative area within specified boundaries occupied by a particular
14
-------
subject (type of rock, soil, plant, etc.)- Point counting is used in petrog-
raphy to estimate the relative areas of minerals in thin sections of rock.
The technique assumes that particles within the field of view are of equal
thickness and are randomly oriented with respect to the microscope light
path.9 Preliminary testing indicated that despite the violation of these
assumptions by insulation materials, point counting might be applied to the
quantitation of asbestos in bulk samples with less variability than pre-
viously used subjective techniques.
Qualitative analysis of bulk materials by XRD is performed with a
minimum of matrix reduction. Samples are initially scanned over limited
O O
diagnostic peak regions for the serpentine (7.36 A) and amphibole (8.2-8.5 A)
minerals, using standard slow-scanning methods for bulk sample analysis.
All samples that exhibit diffraction peaks in the diagnostic peak regions
for asbestiform minerals are submitted to a full (5°-60° 20; 1° 20/min)
qualitative XRD scan. Typical X-ray powder diffraction patterns for indi-
vidual sample components and for mixed samples are presented in Figure 1.
Sample constituents are identified by comparison of the sample diffraction
pattern with standard reference powder diffraction patterns. When subsequent
quantitation is required, particular note is made of possible interferences.
The proposed thin-layer procedure for quantitation of asbestos in bulk
samples by XRD is a modification of the NIOSH-recommended method for the
analysis of chrysotile in air samples.10 The procedure involves initial
comminution of the bulk material to approximately 10 |jm by cryogenic milling
techniques and deposition of an accurately known amount of the sample on a
silver membrane filter. The mass of asbestos is determined by measuring the
integrated area of the selected diagnostic peak, correcting for matrix
absorption effects, and comparing with suitable external standards. Analyt-
ical problems and limitations of the method are clearly identified in the
protocol. Although there is ample evidence that this method is capable of
measuring microgram quantities of asbestos in relatively simple systems with
reasonable accuracy, precision, and speed,11 its reliability for quanti-
tative analysis of asbestos in bulk samples has not been fully charac-
terized.
It should be emphasized that XRD affords information only on crystal
lattice structure and not on gross crystal morphology. The XRD technique,
15
-------
therefore, cannot distinguish between the asbestos minerals and their non-
asbestiform varieties. This can be demonstrated by comparing the diffrac-
tion patterns for antigorite (nonfibrous serpentine) and a mixed antigorite/
chrysotile sample in Figure 1. Particle morphologies must be determined by
an optical technique such as PLM. It is therefore recommended that XRD be
used only as a corroborative procedure and not as an independent analytical
method.
16
-------
MATRIX MATERIAL
V - VERMICULITE
M - MICA
G - GYPSUM
25°
15°
30°
20°
5°
45°
50°
Figure 1. Examples of X-ray diffraction patterns.
-------
AMOSITE
10.4
28.5
_l 1 i 1 l_
J _l 1 L
26
60° 65° 50c 45° 40° 35° 30° 25° 20° 15" 10°
CHRYSOTILE
45 40° 35° 30° 25°
15° 10° 5'
Figure 1 (continued)
18
-------
SAMPLE SERIES D (19.4% AMOSITE)
A » AMOSITE
B - BASSANITE (CaS04 • H H,0»
M- MICA
V - VErtMICULITE
65*
10*
60*
45*
40*
20°
ir
20
SAMPLE SERIES E (10.4XCHRYSOTILE)
B - BASSANITE
C - CHRYSOT1LE
Q - GYPSUM (C*S04 • 2H20)
M- MICA
V - VERMICULITE v
MS
>(01
30°
25*
45*
20°
15*
10°
50°
55°
-» 2$
Figure 1 (continued)
19
-------
ANTIGORITE
Figure 1 (continue J)
-------
50% CHRYSOTlLi/SO* ANTIGORITi
12.0
Figure 1 (continued)
21
-------
SECTION 5
DESIGN OF THE EVALUATION STUDY AND DATA ANALYSIS
5.1 STUDY DESIGN
An interlaboratory testing program was designed and executed with the
following objectives:
1. Evaluate the between-laboratory precision and accuracy and
within-laboratory variation in applying the Tentative Method;
and
2. Evaluate the error rate of the method relative to the Federal
1 percent weight criterion for asbestos content of sprayed-on
insulation materials (38 FR 8820; April 6, 1973).
Twenty-two commercial and four government laboratories were each sup-
plied with eleven samples. Eight of the samples were targeted at specific
weight percents of asbestos fiber. Two species of asbestos were used,
chrysotile and amosit.e. One matrix material, containing primarily gypsum,
was used in all prepared samples. Target weights were designed to cover a
wide range of asbestos concentrations approximately equally spaced on a loga-
rithmic scale. Blanks (Series F) were provided as controls and for deter-
mining the method's potential for producing false positives. The "real-world"
sample (Series J) was included for comparison of between-laboratory variance.
Duplicates (Series K) were included to estimate the average within-laboratory
variance. Target weights and allowable limits for matrix and asbestos fiber
in each sample series are presented in Table 1.
Samples were assigned to laboratories by use of a permuted random
number series. A list of participating laboratories is included as Appen-
dix C. Cover letters, instruction sheets, and reporting forms are included
as Appendix D.
5.2 SAMPLE PREPARATION
The following procedure was used for preparing each sample. Asbestos
fiber was weighed onto either glassine paper or an aluminum boat, depending
22
-------
TABLE 1. SAMPLE COMPOSITION
Target
Actual
Fiber
Wt. of
Wt. of
Series
wt. %
wt. %
type
asbestos (g)
matrix (g)
C
1
1.2
Chrysotile
0.05 + .005
4.95 + .05
A
4
4.9
Chrysotile
0.20 + .01
4.80 + .05
E
16
19.4
Chrysotile
0.80 + .01
4.20 + .05
I
64
74.5
Chrysotile
3.20 + .01
1.80 + .05
H
2
2.5
Amosite
0.10 + .01
4.90 + .05
G
8
9.8
Amosite
0.40 + .01
4.60 + .05
D
16
19.4
Amosite
0.80 + .01
4.20 + .05
B
32
38.8
Amosite
1.60 + .01
3.40 + .05
F
0
0
None
None
3.0 - 5.0
J
-
50.0*
Chrysotile
-
-
Kt
Varies
Varies
Chrysoti le
•
•
*Mean of reported area percents, Groups P and B.
tSeries K samples were provided as duplicates and included samples from
series C, A, E, and I.
-------
on the amount, and sealed until used. Matrix material was weighed into an
aluminum boat and then sealed in a plastic samplfe bag until used. The
asbestos was transferred from the glassine paper or aluminum boat to a
beaker with deionized water, amended with 0.5 rfIL 1% sodium dodecyl sulfate
(SDS) solution per 100 mL deionized water. The detergent solution was used
to facilitate fiber dispersal and to reduce adherence of matrix particles to
asbestos fibers. To break up large fiber bundles, the suspension was soni-
cated* according to the following schedule:
Series A,G,H - 15 s at 100 W
Series C,D,E - 30 s at 100 W
Series B,I - 60 s at 100 W
/
After sonication, the asbestos suspension was transferred to a 240-mL
container. The dry matrix material and 50 to 200 mL of deionized water were
added. The sample was mixed on a Waring blender for 10 to 15 s at medium
speed. The resulting suspension was filtered in a Mi 11iport- apparatus
(Millipore cellulose ester filter, C.45 pm pore), transferred to an aluminum
boat, and dried overnight at 105° C. The dried sample was sealed in a
plastic sample container.
Quality control measures were instituted at several points in the
sample preparation phase. Analytical balances were prechecked with a set of
weights calibrated against weights traceable to the National Bureau of
Standards. Balances were accurate to within 0.3 mg.
Several packets of asbestos fiber and matrix material were selected for
reweighing before sample preparation. Differences between first and second
weighings were negligible for both asbestos fiber and matrix material and
did not introduce significant variations from target weights.
Control samples of matrix material were treated with the sample prep-
aration steps above. The average solubility of the gypsum matrix was 0.2 g/
100 mL deionized water amended with SDS. Average matrix weight loss due to
drying at 105° C and allowing equilibration at room temperature was 10.8 per-
cent. Corrections for weight loss due to matrix dissolution and drying were
used to determine "Actual weight in Table 1.
*Bronson model W185 with 1/2-in. disruptor horn, conical tip.
24
-------
5.3 DEFINITIONS OF LABORATORY GROUPS
Twenty-six laboratories were asked to perform PLM analyses of the
provided samples. Twelve of the twenty-six laboratories have facilities for
XRD analysis and were asked to also analyze the samples by XRD. Four labora-
tories receiving samples chose not to participate in the study. The twenty-
two participating laboratories returned a total of thirty PLM reports and
six XRD reports. PLM data are summarized in Table 2. XRD data are presented
and discussed in Section 7.
Of the 30 sets of PLM results, 19 were produced by following the point
count method closely enough to be included in the evaluation of the technique.
Three laboratories returned results of analyses by more than one analyst;
such results are treated independently. Data not produced by the point
count method are included in separate groups.
The reported PLM results were classified by quantitation procedure, as
follows. The number of laboratories in each group is in parentheses.
Group P - (Point count) PLM asbestos area percent determinations
by the point count method (n = 19).
Group B - (Both) PLM asbestos area percent determinations by
laboratories' own methods for laboratories that also
provided data by the point count method (n = 6).
Group 0 - (Other) PLM asbestos area percent determinations by
laboratories' own methods for laboratories declining to
use the point count method (n = 5).
Data in Group 0 (other) contributed by different laboratories were not
necessarily produced by the same quantitation procedure. Six laboratories
that, contributed data to Group P (point count) also reported results produced
by their own quantitation procedures. This data set is designated Group B
(both). Four of the laboratories produced closer estimates of the true
weight percents by their own method than by point counting, while the other
two laboratories reported closer results by point counting. It is unclear
what relationship exists between the Group P and Group B data contributed by
any one laboratory. It is possible that an estimate produced by point
counting could have influenced a laboratory's own procedure, r.r vic-^
25
-------
TABLE 2. QUANTITATIVE RESULTS OF PLM ANALYSES
(Percent Asbestos by Area)
Sample series
A
B
C
0
E
F
G
H
I
JC
Q.
Laboratory
Group
Cah
4.9
A
38.8
C
1.2
A
19.4
C
19.4
N
0
CO
cn
A
2.5
C
74.5
C
C
PO
P
6
49
5
46
14
0
22
13
5De
40
54
PE
P
12
87
4
76
16e
0
42
9
93
49
19
PF
P
14
60
9e
36
45
0
62
26
77
53
4
PG
P
3
63
0
48
lle
0
36
17
65
59
35
PHI
P
1
24
0e
17
7
0
5
2
42
50
0
PH2
P
2
26
0
13
10
0
6
2
38e
37
35
PJ
P
9e
69
7
37
38
0
39
18
60
51
12
PK
P
2
65
2
47
14e
0
38
25
55
59
14
PM
P
18. 5e
61
6
50
53
0
27.5
12
72.5
67
18.5
PN
P
2
20
le
29
14
0
9
5
26
9
0
PP1
P
10e
77
11
68
44
0
40
20
84
74
41
PP2
P
2e
31
2
28
19
0
13
4
78
54
6
PR
P
18
61
7
61
33
0
52
29
70e
35
73
PS
P
18
28
17e
18
16
4
15
13
98
70
3
PT
P
4
26
1
14
11
0
9
2
61
PV
P
9
47
4e
34
38
0
10
13
75
53
4
PW1
P
0
49
0
28
10
0
PW2
P
6
43
0
30
11
0
20
6
48
PW3
P
3
44
3
39
8
0
26
9
BSS
B
10
35
15
30
20
0
25
10
95
65
4
BNN
B
1.5
75
0
65
30
0
60
40
55
35
0
BPP
B
6
80
4
53
18
0
28
15
65
35
5
BKK
B
5
60
4
40
18
0
35
30
65
60
13
BJJ
B
3
46
7
16
15
0
10
5
50
63
7
BOO
B
4
50
1.5
50
13
0
18
8
48
38
50
01
0
30
50
15
40
60
0
40
40
85
70
35
02
0
70
80
0
60
70
0
50
10
80
80
80
03
0
15
85
15
40
45
5
55
45
85
65
0
04
0
1
35
1
25
10
0
30
3 .
85
50
1
05
0
8
75
6
42
25
0
25
22
93
60
10
'Asbestos type: C = Chrysotile; A = Amosite; N = None
''fr®-.-. • -:tos by weight.
"Environm* "
^Duplicate samples.
eSample series from which duplicate sample IK) wa« drawn.
26
-------
5.4 STATISTICAL DEFINITIONS
A method is accurate if it tends to give results that are close to the
correct results. A method is precise if it tends to generate values that
are close to ear*i other. More specifically, accuracy and precision in the
case of a single sample with known weight percent asbestos equal to w, where
wx, w2, wn are the results of independent analyses, may be defined by
the equations:
1 n
Mean squared error (MSE) - - I (w. - w)2 (1)
n i=l 1
1
Sample variance* (S2) = - I (w^ - w)2
(2)
where •
w = the true value,
- 1 n
w = - I w^ is the average of the w., and
i = 1, 2, ..., n.
Letting
it follows that
BIAS = AVERAGE ERROR = w - w , (3)
MSE = BIAS2 + VARIANCE . (4)
(See Appendix A, Section A.l.) Thus, all questions relating to the accuracy
and precision of a method at a given level of asbestos content may be expressed
and answered in terms of the average bias and the standard deviation of
reported results at that level.
In accord with Eisenhart,12 it is the opinion of the authors that it
is not possible to adequately express accuracy, or overall correctness, in
^Maximum likelihood estimate; the unbiased estimate (divisor = n-1) was
used for actual computations.
27
-------
terms of a single numeric measure. At least two measures of the quality of
a measurement process are required for its appreciation. It is most natural
to separately consider the systematic and the random components of error.
The systematic component is bias, and the standard deviation of the random
component is often referred to as (im)precision. In the following analysis,
therefore, most questions concerning overall accuracy will be addressed in
terms of bias and precision.
5.5 THE RELATIONSHIP BETWEEN AREA PERCENT AND WEIGHT PERCENT ESTIMATES
As indicated in Section 5.1, one of the objectives of the present study
is to evaluate precision and accuracy in applying the Tentative Method.
Samples were prepared with known weights of asbestos fiber and nonasbestos
matrix. Quantitative analysis by PLM results in the estimation of the
average percent area occupied by asbestos fiber within examined fields of
view. While the area occupied by asbestos within the field of vi°w is ob-
viously dependent on the amount of asbestos present, the estimation of per-
cent area is not a direct measure of the known quantity, percent by weight.
To evaluate the accuracy of point counting, therefore, the relationship be-
tween the reported estimates of area percent and the known values of weight
percent must be investigated.
A microscope field of view is essentially a two-dimensional projection
of a portion of the mounted (three-dimensional) sample. The projected area
of a solid cylinder (fiber) may be expressed in terms of £2, where 2. is some
unit of linear measure, e.g., millimeter. The weight (mass) of the cylinder
is the product of its volume, in terms of £3, and its density, in g/S.3. The
projected area and the weight of the cylinder are therefore related by sbme
power transformation involving H2 and £3. (This is a necessarily simplified
version of the more complicated model anticipated, which involved considera-
tions of relative area, relative volutfie, specific gravity, and geometry of
sample constituents.) By extension, projected area percentages of specific
particles in a uulk sample might also be related to the particle weight
percentages of those particles by a power transformation.13 In anticipation
of this relationship, target weight percents of prepared samples were chosen
to be approximdtaly equally spaced on a logarithmic scale.
28
-------
5.6 THE AREA/WEIGHT RELATIONSHIP IN POINT COUNT DATA
Group P (point count) data are presented in Figure 2. The range of the
values reported for each sample series is apparent. For example, reported
values for percent asbestos vary from 13 percent to 76 percent for samples
containing 19.4 percent amosite by weight, and from 26 percent to 98 percent
for samples containing 74.5 percent chrysotile by weight. Figures 3a and 3b
present Group P (point count) data for chrysotile and amosite separately.
Comparison of the two figures reveals a difference between the ways in which
chrysotile and amosite data are related to the A=W line. This suggests that
the area/weight relationship is different for the two asbestos types. The
same data are presented in natural logarithmic coordinates in Figures 4a and
4b. The increased linearity of the data in natural log coordinates is
consistent with the preliminary assumption that a power function is involved
in the area/weight relationship.
Linear regression in logarithmic coordinates* was used to study the
relation between area percents A as measured by point counting and nominal
weight percents W. A standard equation for the power transformation model
was used.
If A = bWc ,
then In (A) = c In (W) + In (b) . (5)
The relation (5) was fitted to the data of Group P (point count) using
the General Linear Models procedure of the Statistical Analysis System, a
preprogrammed statistical procedures package. The slope (c) and intercept (b)
wer.3 each allowed to vary with laboratory (PD, PE, etc.) as well as with
asbestos type. This is equivalent to fitting individual lines to each
laboratory's data separately for chrysotile and amosite. The results of the
regression are presented in Appendix A, Section A.2. The results o* the
analysis support the following conclusions.
1. There is a significant difference between the area/weight
relationships for the two types of asbestos. (The variation
is best demonstrated in Figures 3a and 3b, plotted in original
coordinates, which show that lines fit to amosite and chryso-
*This and all subsequent uses of logarithms or logarithmic coordinates refer
to natural (base e) logarithms.
29
-------
(C) Chrysotile
(A) Amosite
3
T
70
A
0
A
s
A
\
A
A
60
A
K
?
7
A
K
C
50
I
A
O
A
A
V
AC
A
R
A
A
C
U0
A
A
9
X
AC
H
A
A
m
C
30
~ A
A
A
A
A
A
1 A
A
A
A
20
~ A
A
n
A
1 CA
c
\
c
A
AC
1 A
c
A
A
10
~ CA
c
A
C
1 c
C
1 ACA
c
A
1 CA
c
0
~ AC
c
0
5
11
15 20
25 30 35 UO s'5
ASB2ST0S W5IC5HT PE?C2:r?
SOTS: * OBS HAD HISSIHG TALUK'S 9P OBS HIDDiN
"Series 0 is assigned a weight % = 19;
Series E is assigned a weight % = 20.
Figure 2. Group P data, by sample series.
30
-------
I
110 ~
100
?0
ao
70
60
so
?0
20
10
0
~-
3
¦5 10 1S 20 25 3 0 35 «3 u5 50 55 6 3 65 "'O 75 9 0
HS82STOS WEIGHT ?ESCI3T
SOTS; 2 DBS HAD SISSI'JG VAL3SS 55 OBS
(Letters refer to the laboratory contributing the dsta point,)
Figure 3a. Group P data for samples containing chrysotiie.
31
-------
90 ~
I E
I
I
80 ~
I p
I 2
I
70 * J
I ?
X I K
S I G
9 1?? K
!! 60 ~ r
S I
T I
0 I
S I R
50 ~ a o
\ i a v
R l o
A I S
40 ~ J W
PI k J A =
SI 3 P /
? I V
C I ?
; 30 ~ » S
S| 3 ? S
T I ? « H
20 ~ ? « S
' 3 c
I 0 ? H
10 i s n'//'
I S *
I? P
» » ~ » ~ * —
1 - 10 15 20 25 30 35 40 IS
ASBESTOS 3 SIGHT PirC2"T
son: * ens *(,\~ si?«:»5 »u'j:s J9 03s sio?rs
(Letters refer to the laboratory contributing the data point.)
Figure 3b. Group P data for samples containing amosite.
32
-------
I
5.5 ~
•5.0 ~
I
I s
4.5 ~ B
t i r
0 | s
S | G
\ <1.0* R K
o | D
ii r h
T I J H
H 3.5 ~ B
a l
l *
o I
» 3.0 ~ P
I ' H
* I E
? I T 0
j :.5 ~ 2
A I ? G
|F J H
? I »
2.0 ~ J H
? I
c i a o
e I ?
a 1.5 ~
T I 2 7
I
| V G
1.0 ~
I
I * "
I
0.* ~
I
I
0.1* K
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
1CGAPI7HK OP V SIGHT ?S?C2:IT
M07S: 2a GTS HA!1 II^S-MG 7.M0FS 2S 03S HIODIN
(Letters refer to the laboratory contributing the data point.)
Figure 4a. Natural logarithms of Group P data for samples containing chrysotile.
33
-------
4.5 ~
I
0
s
8
1
*
4
H
a
o
F
4.0
3.5
3.0
2.5
\
a
2 '.o
IV
t
-------
tile samples separately would have different slopes and
intercepts. The effect is present but somewhat less apparent
in Figures 4a and 4b.)
2. There are significant differences between the area/weight
relationships in data submitted by different laboratories.
(The between-laboratory variation is apparent in Figures 4a
and 4b. Laboratories H and N tend to report the lowest
values and laboratory F tends to report high values.)
3. The between-laboratory and asbestos-type effects are not in-
dependent. [All second-order interactions are significant
(see Appendix A, Section A.2). Previous tests had shown that
third-order interactions are not significant.]
The importance of individually calibrating laboratories can also be
seen when considering precision. Precision more than doubles (standard
deviation decreases by more than half) when laboratory effects are incorpo-
rated into the appropriate regression (results not shown). Equal variances
were assumed for all laboratories or groups. The assumption is known to be
false, so this comparison is presented as a descriptive rather than an
inferential statistic.
Operationally, the above conclusions suggest that gains in accuracy of
the Group P data (and, by extension, of future PLM analyses by the point
count method) may be made by individually calibrating laboratories. Further
gains may be made by calibrating separately for chrysotile and amosite.
Similar tests indicate that the same conclusions hold for Group B (both) and
for Group 0 (other), although the dependence on asbestos type is marginal
for Group 0. "Calibration" does not necessarily mean providing calibration
coefficients to all laboratories. It is meant to imply any means by which
laboratory-specific adjustments of quantitative results can be made. Two
ways of accomplishing such an adjustment are a round-robin analysis program
and the development and distribution of "standard" samples! Laboratories
provided with wel1-characterized samples are likely to modify and improve
their techniques until their quantitative results consistently correspond
with reference values.
The conclusions stated above must be taken into account in the data
analysis for evaluation of the Tentative Method. It is of interest to
examine the performance of the method in two contexts: (1) as PLM is cur-
rently used, regarding reported area percent as a direct estimate of weight
35
-------
percent; and (2) allowing adjustments of the data that could reasonably be
expected, such as calibrating for the effects observed above by transforming
area percent estimates to predicted weight percents. For this reason, the
main analysis of the PLM data will be presented in three parts. First,
accuracy (bias) will be examined assuming the area percent results are
estimates of weight percent without any adjustments. Second, the precision
of the PLM methods will also be considered without adjustments of the data.
Finally, accuracy and precision will ac,-iin be considered after transforming
the area percent estimates to predicted weight percent values.
-------
SECTION 6
METHOD EVALUATION: POLARIZED LIGHT MICROSCOPY
6.1 BIAS OF THE PLM DATA
Group means and standard deviations for sample series A through J are
summarized in Table 3. Recall that Group P (point count) is data produced
by point counting, Group 0 (other) is data produced by laboratories' own
methods, and Group B (both) is data produced by laboratories' own methods
for those laboratories that also contributed point count data. Note in
Table 3 that for six of nine cases the mean of the point count group (MP) is
closer to the nominal weight than the mean of Group B (MB). This is not a
significant difference, and it appears that Groups P and B are comparably
biased. Individual t-tests for differences between MP and MB performed for
each sample are not significant. A more powerful test for differences
between biases using regression analysis also supports the conclusion that
Groups P and B are comparably biased (see Appendix A, Section A.3).
Note also in Table 3 that means of Group 0 results are consistently
higher than those of Groups P and B. Sign tests suffice to show that Group 0
is significantly more biased than Groups P and B. Group 0 means were further
from the nominal weight than Group P means on all nine prepared samples; the
probability of a result this extreme occurring by chance is 2(b9) = 0.004
(two-sided test). Group 0 was further from the loaded weight than Group B
on eight of the nine samples; the corresponding probability is 2[(l+9)/29] =
0.04 (two-sided test). Group 0 is the furthest from the loaded weight of
the three PLM groups on eight of the nine samples, with probability 2[l/39 +
9 • 2/39] < 0.002 (two-sided test). The difference in bias between Groups 0
and P is confirmed by regression analysis (see Appendix A, Section A.3).
The 90-percent confidence intervals for Group P (point count) data for
chrysotile and amosite samples are presented in Figures 5a and 5b, respec-
tively. Calculations were performed on log-transformed data and the results
then exponentiated. The procedure used accounts for unequal variance of
37
-------
TABLE 3. MEANS AND STANDARD DEVIATIONS OF REPORTED PLM RESULTS,
BY GROUP (P, B, 0)
(percent asbestos by area)
Series
Type
Weight
%
Means
Standard deviations
MP
MB
MO
SP
SB
SO
C
Chrysotile
1.2
4.2
5.3
7.4
4.5
5.3
7.3
A
Chrysotile
4.9
7.3
4.3
24.8
6.3
2.9
27.5
E
Chrysotile
19.4
21.7
19.0
42.0
14.8
5.9
24.6
I
Chrysotile
74.5
64.3
63.0
85.6
19.6
17.3
4.7
H
Amosite
2.5
12.5
18.0
24.0
8.6
13.9
18.3
G
Amosite
9.8
26.2
29.3
40.0
16.9
17.3
12.7
D
Amosite
19.4
37.8
42.3
41.4
17.7
17.5
12.4
B
Amosite
38.8
48.9
57.7
65.0
19.5
17.4
21.5
F
None
0.0
0.2
0.0
1.0
0.9
0.0
2.2
J
Environmental
-
50.7
49.3
65.0
16.1
14.7
11.2
38
-------
5a. Chrytotilt
5b. Amotita
120
100
I
1
m
I
<
20
Asbtstos Weight P«rc»nt
Figure 5. Ninety percent confidence intervals for means of Group P data.
39
-------
sample series containing different weight percents of asbestos;14 15 the
confidence statement holds simultaneously at all levels of asbestos. The
figures should be interpreted as follows. For a sample containing 20 percent
chrysotile by weight = 20), the mean estimate of relative area by point
counting will be between approximately 16 and 23 percent, with a probability
of 0.90. Similarly, for Wj, = 50, 90 percent of the means of point count
analyses will be between approximately 31 and 44 percent. The same relation-
ships demonstrate a positive bias for Group P analysis of samples containing
amosite. For a sample with = 20, the 90-percent confidence limits are
approximately 27 and 38 percent; for = 50, the limits are approximately
41 and 71 percent.
Using the midpoints of the confidence intervals, the average percent
bias of Group P (point count) analyses was estimated at several weight
percent levels. These are presented in Table 4 and Figure 6. The percent
bias varies with weiqht percent of asbestos similarly for amosite and chryso-
tile samples. Point counting has a greater positive bias on amosite samples
than on chrysotile samples and, in fact, underestimates asbestos content in
samples containing mi re than about 18 percent chrysotile by weight.
6.2 PRECISION OF THE PLM DATA
Precision will be evaluated with an approach based on the standard
deviation of reported results. The standard deviation is the most common
measure of variation or imprecision. If one group or method is systematically
more variable than another, the trend may be evident in a plot such as
Figure 7, in which group standard deviations (SP, SB, SO in Table 3) are
related to nominal weight percentages. Group P sample standard deviations
are larger B on six of nine samples, but are reasonably
*S /
comparable on all except sample E. Group 0 standard deviations exceed those
of Groups P and B on the four samples with less than 5 percent asbestos,
suggesting that Group 0 data are less precise on samples in this range.
The standard deviation of reported PLM data increases as the weight
percent of asbestos increases for all groups; i.e., variance is directly
related to the percent asbestos present. Figures 8a and 8b present the
relationship between standard deviation (SP) and weight percent (W) or mean
reported area percent (MP), respectively, for Group P. Data points conform
40
-------
TABLE 4. AVERAGE PERCENT BIAS OF GROUP P DATA
Asbestos type
Weight percent
W
CI midpoint
X
% Bias = ^ x 100
Chrysotile
10
11.85
18.5
20
19.60
-2.0
50
37.90
-24.2
Amosite
10
21.85
118.5
20
32.35
61.8
50
56.05
12.1
41
-------
100
amosite
chrysotite
.20
100
Asbestos Weight Percent
-20
-40
Figure 6. Average percent bias of Group P data.
42
-------
PLOT CF 3P«WEIGHT SY^CL USoD IS P
PLOT OF SD IGMT SYMPOL USoO IS 9
PLOT OF S0• wC 13HT ;YK90L USED IS 0
30
27
3
0 2»
U
21
IS
P*
15
12
I p
: »B
C 5 IZ 15 2'j 25 30 3? «0 «5 SO 55 SO 65
WEIGHT (%)
NOTi: » CC Wi0 MISSING VALUES
*P and B have the same value.
^Series D is assigned a weight % = 19;
Series E is assigned a weight % =¦ 20.
Figure 7. Group standard deviations.
7* e; 85 i
43
-------
a chrysotile
• amojrta
*
t
I I I 1 __1
0 20 40 ?0 80 100
Asbestos Weijj-iv Percent
Figure 8a. Group P standard deviation vs. asbestos weight percent.
A chrysotile
• amosite
20
16
12
8
4
20
40
60
80
100
Group P Mean
Figure 8b. Group P standard deviation vs. Group P mean.
44
-------
more closely to a single curve in Figure 8b. Regression of In (SP) on
In (W) (R2 = 0.81) ai.-J op in (MP) (R2 = 0.96) suggests that the variance is
more systematically related to the reported area percent asbestos than to
the known weight percent asbestos.
The suggested difference in precision between Group 0 (other) and
Groups P (point count) and B (both) may be due to the larger bias of Group 0
results (Section 6.1). Because variance is directly related to reported
area percent, larger standard deviations may result simplv from the Group 0
tendency to report higher values. Differences between groups with respect
to precision were investigated with regression analysis, allowing information
to be combined from samples loaded at different weight percentages. The
results of the regression (see Appendix A, Section A.4) indicate that, when
differences in bias between groups are accounted for, Groups P, B, and 0 are
not significantly different with respect to precision.
Precision is sometimes expressed as the percent relative standard
deviation or coefficient of variation (CV = 100 SP/MP). CV is related to
Group P means in Figure 9. The CV is less than 100 percent on samples with
more than approximately S percent asbestos by area and less than 50 percent
on samples with more than approximately 32 percent asbestos by area. At a
mean reported value (MP) of 10 percent asbestos, CV = 79 percent; at MP =
20 percent, CV s 61 percent; at MP = 50 percent, CV = 41 percent.
6.3 ACCURACY AND PRECISION, AFTER DATA TRANSFORMATION
As stated earlier, it is of interest to evaluate the accuracy of the
PLM methods after adjusting for the relationship between reported area
percent and the known weight percent of the samples. This will allow not
only a better understanding of what reported PLM data mean, but will also
indicate what improvements might be made in data quality by adjusting PLM
area percent estimates to better represent weight percent. Such an adjust-
ment of the data generated in the present study is possible using parameters
similar to those determined in the regressions discussed in Section 5.5.
This adjustment formally applies only to the samples and laboratories in
this study and would be questionable for other laboratories and other
samples of different composition. Further study would be required to deter-
45
-------
a
1.1
1.0
.9
c
3
s
•8 7
c
0)
3 c
^ chrysotile
• amosite
.3 .
X
J.
20
40 60
Group P Mean
80
100
Figure 9. Coefficient of variation of Group P data.
46
-------
mine if a small family of functions, specific for asbestos and matrix type,
could be used gererally to adjust PLM data to more adequately represent
percent asbestos by weight.
Groups P, B, and 0 area-percent data (A) were adjusted for laboratory
a
and asbestos-type effects to yield predicte-i weight percents (W) for each
individual result. Means and standard deviations of W were then computed
a a
for each group (MP, SP, etc.). These are presented in Table 5. Details of
the transformation may be found in Appendix A, Section A.5.
a
The accuracy of predicted weights (W) was evaluated by determining the
average percent absolute error (ERROR (W)), or the average difference between
W and W. This is calculated as
ni W.. - W-
ERROR (W) = -^ I — — 100 %
ni j=l Wi
where j = 1, . . . , and n. indexes the reported analyses for each group on
+ h '
the i sample. ERROR (W) is then compared to the same quantity computed
for the untreated data,
1 ni Aii"Wi
ERROR (A) = I —— 100% .
a
ERROR (A) and ERROR (W) are tabulated by group in Table 6.
The most obvious and expected result in comparing the average percent
absolute errors of treated and untreated data is the considerable gain in
accuracy (reduction of error) that results from the transformation A -» W.
For example, the average Group P inaccuracy for unadjusted data on samples
containing asbestos is 155 percent. After transformation, the inaccuracy
drops to 31 percent, or only one-fifth of the original.
For Groups P and B the percent error is fairly stable over the five
samples between 2.5 and 20 percent asbestos by weight. The average of
a
ERROR (W) for these five samples is 30 (±5) percent for Group P and 32 (±11)
percent for Group B. For Group 0 (other), the corresponding values are more
variable and tend to indicate less accuracy. The Group 0 average ERROR (W)
for the same five samples is 63 (±35) percent.
47
-------
A
TABLE 5. MEANS AND STANDARD DEVIATIONS OF PREDICTED WEIGHT (W),
BY GROUP (P, B, 0)
Means Standard deviations
Weight
Series
Type
%
MP
MB
MO
SP
SB
SO
C
Chrysotile
1.2
2.1
3.4
1.9
1.1
2.6
0.5
A
Chrysotile
4.9
4.3
3.4
4.7
1.6
1.2
3.4
E
Chrysotile
19.4
20.5
18.1
17.8
9.1
11.5
8.6
T
J
Chrysotile
74.5
72.9
73.0
74.0
3.7
6.0
2.7
H
Amosite
2.5
3.1
2.8
5.5
1.2
0.3
3.3
G
Amosite
9.8
11.1
9.7
13.9
4.9
4.4
7.4
D
Amosite
19.4
21.7
19.0
14.3
6.4
7.3
6.6
B
Amosite
38.8
33.9
36.6
35.3
5.8
5.0
6.1
48
-------
TABLE 6. AVERAGE PERCENT ABSOLUTE ERROR: UNADJUSTED VS. TRANSFORMED PLM DATA,
BY GROUP (P, B, 0)
Group P
Group B
Group 0
Series
Type
Weight
%
Error (A)
Error (W)
Error (A)
Error (W)
Error (A)
Error (W)
C
Chrysotile
1.2
313.9
75.0
370.8
179.6
563.3
62.2
A
Chrysotile
4.9
105.1
28.4
42.5
36.0
438.0
56.7
E
Chrysotile
19.4
60.5
34.0
21.7
44.0
130.0
36.0
I
Chrysotile
74.5
27.2
3.2
24.6 *
4.1
14.9
2.3
H
Amosite
2.5
406.7
23.7
620.0
13.5
860.0
120.5
G
Amosite
9.8
178.9
34.7
199.3
37.6
308.2
68.2
D
Amosite
19.4
106.9
30.9
128.1
29.4
117.9
33.2
B
Amosite
38.8
47.3
14.3
51.9
10.3
71.4
12.8
-------
After variance due to laboratory and asbestos-type effects is removed
by the above transformation, residual variance is reflected in dispersion
A
about the regression line of W on W. In this context, mean squared error
about the regression line is a measure of precision. The analyses were per-
formed in log coordinates since the correlations obtained typically exceeded
R = 0.99. The basic result is that Group P is significantly more precise
than Groups B and 0 after between-laboratory variance is removed. Spherically,
allowing the slope (c) and intercept (d) to vary with laboratory .nd asbsstos
type, the Group P mean squared error (MSE(P) = 0.123) was less than that of
Groups B and 0 (MSE(B) = 0.264, MSE(O) = 0.226). The differences between
Group P and Groups B and 0 are significant at the 0.01 and 0.05 level, res-
pectively, by standard two-sided F tests.
The abeve analysis shows that, if laboratories had access to information
with which they could calibrate theii' results (according to the area-weight
relationship for each laboratory and asbestos type), considerable gains in
accuracy and precision of results could be achieved. The gains would be
greater for laboratories using the point counting quantitation procedure
thajn for laboratories using alternative procedures.
6.4 ESTIMATION OF WITHIN-ANALYST VARIANCE
A duplicate sample was included among the samples sent to each labora-
tory in an effort to collect preliminary data for estimating within-laboratory
variance. Since data supplier; by different analysts from the same laboratory
are being treated independently, what will be estimated is actually within-
analyst variance. More than one implicate sample per analyst would be re-
quired to adequately characterize tnis component of total variance, but a
rough estimate may be gained from the present information.
Samples from the chrysotile series (C, A, E, and I) were reassigned to
sample series K and then distributed. Returned results included analyses of
five duplicate samples from Series C, four from Series A, and three each
from Series E and I, as summarized in Table 7.
On 11 of the 15 reported pairs, the duplicate variance estimate was
less than 25 percent of the total variance estimate for the corresponding
sample series. Each within-sample median variance was less than 25 percent
of the total variance for the sample. It therefore appears that within-
50
-------
TABLE 7. ANALYSIS OF DUPLICATE SAMPLES, GROUP P
Weight
Std. dev.
Median S
Sample
Series
Type
LABa
Results
S
(S2)
(S2)
C
Chrysotile
1.2
PHI
0, 0
0
PV
4, 4
0
PN
1, 0
0.7
0.7
4.5
PF
9, 4
3.5
(0.5)
(20.3)
PS
17, 3
9.9
A
Chrysotile
4.9
PM
18.5, 18.5
0
PJ
9, 12
2.1
2.5
6.3
PP2
2, 6
2.8
(6.3)
(39.7)
PP1
10, 41
21.9
E
Chrysotile
19.4
PK
14, 14
0
PE
16, 19
2.1
2.1
14.8
PG
11, 35
17
(4.4)
(219.0)
I
Chrysotile
74.5
PR
70, 73
2.1
PH2
38, 35
2.1
2.1
19.6
PD
50, 54
2.8
(4.4)
(384.2)
dSee Table 2.
^See Table 4.
-------
analyst variance probably contributes less than 25 percent of the total
variance, which again implies that most of the variance in the point count
data (Group P) is due to between-laboratory differences. This is not sur-
prising and is consistent with the already noted doubling of precision that
results from calibration of individual laboratories (See Section 5.6).
6.5 FALSE POSITIVES AND NEGATIVES FOR POINT COUNT DATA
One of the important characteristics of the point count procedure to be
evaluated is the likelihood of its generating false positives and false
negatives. A false positive occurs when an analyst reports asbestos present
in a sample that does not contain asbestos. A false negative occurs when an
analyst reports no asbestos present in an asbestos-containing sample.
Group P data for the 1.2 percent chrysotile samples (Series C) included
five false negatives out of a total of 19 analyses. Two of the false nega-
tives were reported by analysts in the same laboratory (PHI, PH2).
Group P data for the 4.9 percent chrysotile samples (Series A) contained
one false negative in 19 analyses. The same laboratory reported one of the
false negatives on the 1.2 percent samples. No false negatives were reported
for any samples containing amosite or more than 5 percent chrysotile.
One false positive was reported for the blank samples (Series F) in
Group P. Laboratory S reported 4 percent tremolite-actinolite asbestos
present. Since no other laboratories reported tremolite or actinolite
asbestos in any samples, and Laboratory S reported it only for sample F, the
false positive is probably due to contamination.
The point counting procedure involves counting 400 points on eight
subsamples of the material being analyzed. For the false negatives discussed
above, laboratories scored three or fewer points for asbestos fiber, and
thus less than 1 percent (4/400) asbestos was reported. The Tentative
Method procedure supplied to the laboratories included the statement, "if
seven or fewer of 400 non-empty points are scored for asbestos fiber, report
less than one percent asbestos." This provision was based on regarding data
as a test of the hypothesis that the percent asbestos in a sample is less
than or equal to one. A Poisson distribution of the results of repeated
analysis of low percentage samples was assumed. At present, however, data
are not sufficient to determine whether repeated analysis of "real-world"
52
-------
samples fits the Poisson model. Additionally, it is more appropriate to the
normal use of reported data to regard it as estimating percentage rather
than testing a specific hypothesis. It is therefore recommended that the
simple arithmetic percentage ([points scored for asbestos t total points
counted] x 100) be used at all levels until further data may justify the
Poisson assumption and/or a specific percent criterion is established. The
simple arithmetic percentage was used for the determination of false negatives
in this study.
The one false negative reported for point counting analysis of the
4.9 percent chrysotile sample represents a false negative probability of
0.05 (1/19). This estimate formally applies only to the present study;
analysis of "real-world" samples may be subject to a higher rate. Current
EPA guidance recommends at least three samples be taken from each "sampling
area," defined as "any area, whether contiguous or not . . . which contains
friable material that is homogeneous in texture and appearance.1,16 For this
study, the probability of obtaining three false negatives on the 4.9 percent
chrysotile sample was 0.053, or less than 0.001. Taking 0.10 as a conserva-
tive estimate of the rate of false negatives for samples containing 5 percent
asbestos, the probability of obtaining false negatives on all three samples,
if they each contain at least 5 percent asbestos, is 0.103 or 0.001. If the
false negative rate was even as high as 0.30, then the probability of false
negatives on all three samples would still be only 0.027.
The counting of 400 points provides a good estimate of the area percent
of asbestos within the examined fields of view. The counting of four times
this number (1,600 points) would be required to double the precision of the
estimate. The accuracy of point count data, especially in samples containing
small amounts of asbestos, is strongly dependent on factors other than the
number of points count*'! These include representative sampling of the bulk
material, adequate sampTe preparation, and uniform dispersal of the sample
material on slides. Variation associated with these sources greatly affects
the lower detection limit of the method and is likely to be more responsible
for the occurrence of false negatives than is the actual point counting
procedure. Further study is required (see Section 3.1) to determine what
improvements and standardization can be achieved in these aspects of the
current methodology.
53
-------
6.6 OPERATING CHARACTERISTIC CURVES
The evaluation of false negatives is complicated by the area-weight
relationship as discussed previously. The quantity of interest is not the
quantity that is being directly measured. Therefore, it is relevant to
evaluate the error rates of the point count method after transforming the
data to predicted weights (A -+ W; see Section 6.3). The evaluation will use
a criterion other than false positives and negatives.
Figure 10 presents the operating characteristic (0C) curves for an
idealized error-free method and a hypothetical very accurate method. The
proportion of laboratories reporting less than level X of asbestos in a
given sample is plotted against the known weight percent of asbestos in the
sample. For samples with les: than X percent asbestos, the decision is
correct if and only if a laboratory's result is also below X. It is desir-
able that the 0C curve remain close to 1.0 below the decision criterion X
and drop rapidly to 0 as the threshold X is crossed. This is illustrated by
the curve in Figure 10. For samples containing less than X percent asbestos,
a majority of laboratories report less than X percent present; for samples
containing more than X perc*,it asbestos, a minority of laboratories report
less than X percent present; for samples containing exactly X percent asbes-
tos, one-half of the laboratories report less than X percent present.
Figure 11 presents the 0C curves for the adjusted point count data at
four different criteria levels: Figure 11a, X = 1 percent; lib, X = 2 per-
cent. ~ 5 percent; lid, X = 7 percent.^ The proportion of laboratories
re»-. / , yielding predicted weights (W) less than X is plotted
ag&-V | : v k. ,-n weight percent of asbestos in the sample. Thus, in Figure lib,
more tiui.i 60 percent of the W were less than 2 percent for the 1.2 percent
chrysotile sample (Series C), approximately 40 percent of the W were less
than 2 percant for the 4.9 percent chrysotile sample (Series A), and none of
the W were less than 2 percent for the 9.8 percent amosite sample (Series G).
Figures 11a and lib suggest that laboratories are not able to reliably
distinguish between samples containing <1 percent and >1 percent asbestos or
between samples containing <2 percent and >2 percent asbestos. The perform-
ance of the method improves at X = 5 and is better still at X = 7. Figure lid
suggests that laboratories, after calibration, are able to reliably decide
whether a particular sample contains <7 percent or >7 percent asbestos.
54
-------
1.0
\
\
\
\
\
\
0.5
\
o
\
\
\
An error-free method
— A very accurate method
\
\
Figure 10.
Percent Asbestos by Weight
Theoretical operating characteristic curves.
-------
Proportion of Laboratories
Reporting Less than X Percent
0
0
0
0
to
CO
CJ1
00
<0
is)
-------
Proportion of Laboratories
Reporting Less than X Percent
poo o o o o
m w *. m m m so
Proportion of Laboratories
Reporting Lass than X Percent
ooooppppp.-"
CJI
w
1
GO
O
®
a
#•
*
or
or
<
S
2.
to
t ©
-------
This result applies only to data that have been adjus' Hi fo" v. iance due to
both laboratory differences and to asbestos type. Th . j ju jnt formally
applies only to the samples and laboratories in this i.' j ~nd would be
questionable for other laboratories and other samples of different composi-
tion. Data corrected only for variance due to asbestos type do not show the
improved performance demonstrated for fully calibrated data in Figures 11c
and lid.
6.7 GENERAL OBSERVATIONS
Several general points should be made before concluding this section.
Table 3 shows that between-laboratory variance on saitiple series J (Group P,
SP(J) = 16.1) is comparable to that of similar weight percent series (SP(B) =
19.5, SP(I) = 19.6). This suggests that the formulated samples were not
significantly less variable in composition than the "real-world" samples
distributed and thus that the results of this study, at least with respect
to precision, are reasonable estimates of what would be seen in point count-
ing analyses of samples normally submitted to laboratories.
Several laboratories reported results by the point count method that
were notably more accurate than the Group P average. Laboratories PP2, PS,
PT, PV, and PW2 (Table 2) are among the more experienced of the participating
laboratories in bulk sample analysis. Laboratory PP2, which reported the
results most consistently close to the true weight percent values, is known
to have an internal quality control program involving preparation and analy-
sis of asbestos-containing standard samples.
It was noted earlier that some relationship may exist between Group P
(point count) and Group B (both) data contributed by any one laboratory.
The estimate produced by point counting may have influenced the result of
the laboratory's own method, or vice versa. The problem of interpreting
Group B is further complicated by comparison with Group 0 (other). Group 0
consists of data produced by laboratories that used only their own quantita-
tion procedure. As in Groups P and B, there are differences between labora- •
tories. Lab 04, which reported the "best" results of Group 0, is one of the
more experienced PLM laboratories participating in the study. However, as
was sh-iwn in Section 6.1 Group 0 is significantly more biased than Group B.
The dissimilarity of Groups B and 0 suggests tnat, although both sets of
58
-------
data were produced by the respective laboratories' own quantitation proce-
dures, there is some systematic difference between the procedures used.
6.8 CONCLUSIONS
For the sake of clarity, the definitions of the groups into which the
PLM data were classified are restated below.
Group P--(Point count) PLM asbestos area percent determinations by
the point count method.
Group B--(Both) PLM asbestos area percent determinations by the
laboratories' own methods for laboratories that also provided
data by the point count method.
Group O--(Other) PLM asbestos area percent determinations by the
laboratories' own methods for laboratories declining to use
the point >.ount method.
The following conclusions are indicated by the analysis of PLM data.
A considerable amount of the variation in the data can be
removed by linearly regressing the natural logarithms of area
percent (reported data) on the natural logarithms of weight
percent (known values). This finding is consistent with the
assumption that area and weight are related by a power func-
tion.
There is significant variation in the area/weight relation-
ship because of differences between laboratories, differences
between asbestos types (amosite and chrysotile), and interac-
tions between laboratory and asbestos type.
Groups P and B appear similarly biased. Group 0 results have
a significantly higher bias than Groups P and B.
Group P average bias (b) varies with the type and weight
percent (W) of asbestos in a sample. For samples containing
chrysotile, b = 18.5 percent at W = 10 and b = -24.2 percent
at W = 50. For samples containing amosite, b = 118.5 percent
at W = .10 and b = 12.1 percent at W = 50.
Groups P, B, and 0 are similarly precise when the effects of
bias are removed.
Precision of Group P data may be described by the coefficient
of variation. At a mean reported value (MP) of 10 percent
asbestos, CV = 79 percent; at MP = 20 percent asbestos,
CV = 61 percent; at MP = 50 percent asbestos, CV = 41 percent.
59
-------
Group P is significantly more precise than Groups B and 0 in
terms of residual variance after removing variance due to
laboratory and asbestos-type effects.
Considerable gains in precision and accuracy of PLM data are
possible by individual calibration of laboratories, espe-
cially for laboratories using point counting.
Within-analyst variance probably accounts for less than
25 percent of the total between-laboratory variance in Group P
data.
Several false negatives and a false positive were included in
point count results. The false negatives are more likely due
to variability in sample and slide preparation than to the
counting procedure per se. The false positive was likely due
to sample contamination. The rate of false negatives is such
that the analysis of three samples of a suspect material, if
each contained at least 5 percent asbestos by weight, would
result in three false negatives with a probability less than
0.03 and possibly as low as 0.C01.
The data, after adjustment for between-laboratory and asbestos-
type effects, suggest that laboratories using the point count
method are better able to resolve the difference between
samples containing <7 percent and >7 percent asbestos by
weight than they are able to resolve samples containing
<1 percent from >1 percent.
60
-------
SECTION 7
METHOD EVALUATION: X-RAY POWDER DIFFRACTION
Twelve laboratories received samples for analysis by the proposed XRD
method; five laboratories participated and returned six sets of results.
These data are summarized in Table 8. Three of the data sets (X1-X3) were
produced by some variation of the proposed thin-layer method of quantitation;
the other sets (X4-XG) were produced by alternative bulk or thick-layer
methods of quantitation. The seven laboratories declining to perform the
requested analyses indicated either that the method was too time-consuming
and costly, they lacked adequate facilities and expertise, or they felt the
method .as inadequate. It should be emphasized that none of the "thin-layer"
laboratories followed the Tentative Method protocol exactly; similarly, bulk
methods employed by laboratories reporting X4-X6 were not strictly equivalent.
A notable deviation of the actual methods employed from the Tentative Method
was the failure of all laboratories to use step-scanning analysis for quan-
titation.
Laboratories were instructed to determine and report XRD results inde-
pendently of any information derived from PLM analysis. This is not the
appropriate procedure for typical laboratory analysis of submitted samples.
As has been noted earlier, XRD affords information only on crystal lattice
structure and not on particle morphology. The presence of asbestiform
particles must be determined by an optical procedure such as PLM.
Means and standard deviations of all reported XRD results are shown in
Table 9. Average reported values for XRD are shown for bulk methods,
thin-layer methods, and all methods together. Except for Series G, the
means of the bulk methods are closer to the reference values than those of
the thin-layer methods.
Average absolute errors of reported results for bulk and thin-layer
methods are shown in Table 10. Comparing the average errors with a two-
61
-------
TABLE 8. REPORTED XRD RESULTS
(percent asbestos)
Thin-layer
Bulk
Series
Type
Weight %
XI
X2
X3
X4
X5
X6
C
Chrysotile
1.2
1.0
3.0
5.0
0.0
0.0
3.0
A
Chrysotile
4.9
3.0
7.0
0.0
3.0
9.0
1.0
E
Chrysotile
19.4
4.0 '
0.0
7.0
13.0
31.0
10.0
I
Chrysotile
74.5
55.0
45.0
.
74.0
75.0
H
Amosi te
2.5
2.0
3.0
4.0
2.0
G
Amosi te
9.8
3.0
11.0
17.0
23.0
25.0
D
Amosite
19.4
18.0
37.0
20.0
32.0
20.0
B
Amosi te
38.5
53.0
69.0
51.0
75.0
30.0
F
Amosite
0
.
0.7
0.0
0.0
J
Chrysotile
~50%a
63
35
51
40
K
Chrysotile
(Duplicate
of C,A,E, or I)
42(1)
15.0(E)
35.0(E)
1.0(C)
aArea percent asbestos, mean of Groups P and B.
-------
TABLE 9. MEANS AND STANDARD DEVIATIONS OF REPORTED XRD RESULTS
(percent asbestos)
Thin-layer Buik Pooled
Series
Type
Weight %
M
S
M
S
M
S
C
Chrysoti le
1.2
3.0
2.0
1.0
1.7
2.5
1.8
A
Chrysotile
4.9
3.3
3.5
4.3
4.2
4.3
3.3
E
Chrysotile
19.4
3.7
3.5
18.0
11.4
10.8
10.9
J
Chrysotile
74.5
50.0
7.1
74.5
0.7
62.2
14.7
H
Amosite
2.5
1.5
0.7
3.0
1.4
2.8
1.0
G
Amosite
9.8
7.0
5.7
21.7
4.2
15.8
9.0
D
Amosite
19.4
28.0
12.7
24.0
6.9
25.6
8.3
B
Amosite
38.8
61.0
11.3
52.0
22.5
55.6
17.6
F
None
0
0.2
0
0
0
0
0
TABLE 10. AVERAGE ABSOLUTE ERRORS OF REPORTED XRD RESULTS
(percent asbestos)
Series
Type
Weight %
Average absolute error
Thin-layer (n) Bulk (n)
C
Chrysotile
1.2
1.9
(3)
1.4 (3)
A
Chrysotile
4.9
3.0
(3)
3.3 (3)
E
Chrysotile
19.4
15.7
(3)
9.1 (3)
I
Chrysotile
74.5
24.5
(2)
0.5 (2)
H
Amosite
2.5
0.5
(2)
1.0 (2)
G
Amosi te
9.8
4.0
(2)
11.9 (3)
D
Amosi te
19.4
9.0
(2)
4.6 (3)
B
Amosi te
38.8
22.2
(2)
18.7 (3)
F
None
0
0.7
(1)
0 (2)
n = Number of reported results.
-------
sided t-test, there is no significant difference at the 5-percent level
between laboratories performing the analyses by bulk methods and those
using thin-layer methods, although, as noted above, there is a suggestion
that the bulk methods are more accurate.
Estimates of precision, given by the coefficient of variation (CV),
calculated as the standard deviation divided by the mean reported value, are
shown in Table 11. Comparing CVs with either a sign test or a paired t-test
again showed no significant difference between bulk and thin-layer methods
(two-sided P > 0.4). Considering individual CVs, those for bulk methods are
all less than or equal to those for thin-layer methods, except for Series C
and B, further suggesting that the bulk methods are at least as precise as
the thin-layer methods, as applied by laboratories in this study.
The overall impression from these results, that bulk analysis is at
least as accurate and precise as thin-layer analysis, is further supported
by the results of a more detailed analysis of both bulk and thin-layer
methods by asbestos type.
Linear regression analyses of the reported results for chrysotile
samples and amosite samples for bulk methods gave the following results (see
Figure 12):
1. Reported results and reference values are better correlated
for chrysotile than amosite (i.e., correlation coefficients
are significantly different at the 5 percent level, by a
two-sided t-test);
2. Analysis of chrysotile is significantly more precise than
amosite (i.e., variances about the regression are signifi-
cantly different at the 5 percent level by a two-sided F-test);
and
3. Analysis of chrysotile appears more accurate (chrysotile
slope = 1.00, intercept = -0.55; amosite slope = 1.23, inter-
cept = 3.76), although a two-sided t-test for difference
between the slopes is not significant at the 5 percent level.
This is probably due to the large imprecision in the estimate
of the amosite slope. The results for chrysotile in this
regard are particularly striking, with the regression line
being essentially indistinguishable from the theoretical
y = x line with slope = 1.
In contrast, linear regression analyses of the reported results by
individual asbestos types for thin-layer methods (Figure 13) revealed no
64
-------
TABLE 11. COEFFICIENTS OF VARIATION OF REPORTED XRD RESULTS
Coefficient of variation
Series
Type
Weight %
Thin-layer
Bulk
Pooled
C
Chrysotile
1.2
0.67
1.7
0.72
A
Chrysotile
4.9
1.06
0.98
0.77
E
Chrysotile
19.4
0.95
0.63
1.0
I
Chrysotile
74.5
0.14
0.01
0.24
H
Amosite
2.5
0.47
0.47
0.36
G
Amosite
9.8
0.81
0.19
0.57
D
Amosite
19.4
0.45
0.30
0.32
B
Amosi te
38.8
0.19
0.43
0.32
65
-------
100
90
80
70
60
50
40
3G
20
10
0
•Pi
/
/
/
/
/
/
1.23x + 3.76
y = 1.00x - 0.55
r = .98
sy/x = 5-27
A
C
40
Reference Asbestos Weight %
bars represent ±1 a.
Figure 12. Comparison of bulk XRD analysis by asbestos type.
66
-------
100 r—
1.68X - 5.06
¦a
®
3
o
(A
a
4)
ce
.65x - 1.38
y/x * 5-89
Reference Asbestos Weight %
•Error bars rspresent ;1 o.
Figure 13. Comparison of thin-layer XRD u..alysis by asbestos type.
67
-------
significant differences in correlation or precision between analyses for
chrysotile and amosite; correlation coefficients and standard errors about
the regression for the two asbestos types v/ere not significantly different
at the 5-percent level by a two-sided t-test. The slopes for chrysotile and
amosite were, however, significantly different at the 5-percent level.
Analysis for both asbestos types is biased (chrysotile, negative; amosite,
positive) with slopes of the regression lines significantly different from 1.
A further comparison between bulk and thin-layer methods, by asbestos
type, indicates that for analysis of chrysotile, the bulk methods are signi-
ficantly less biased than thin-layer methods. A two-sided t-test indicates
that the slope of 0.65 for chrysotile analysis by thin-layer methods is
significantly less than that of 1.00 for chrysotile by bulk methods at the 5
percent level. No significant difference in slopes (bias) was observed
between methods for amosite.
Data produced by thin-layer methods of analysis included one false
negative out of three analyses of the 4.9 percent chrysotile sample. The
same laboratory reported chrysotile false positives for all amosite samples
and for the blank sample with reported chrysotile values ranging from <1 to
8 percent. One false negative out of three analyses was also reported at
the 19.4 perce.it chrysotile level.
Data produced by bulk methods of analysis included two false negatives
out of three analyses of the 1.2 percent chrysotile sample. One of these
laboratories also reported a false positive amosite for the 4.9 percent
chrysotile sample.
Four laboratories reported results on duplicate samples (Series K)
included in the samples sent to the laboratories. However, because of the
small number of observations, no firm conclusions can be drawn about the
intralaboratory variance of this method or its relative contribution to the
total variance of the reported resets.
These result? do give evidence that XRD is capable of detecting chryso-
tile at the 1 percent level in a simple matrix and suggest that at this
level a thin-layer method of analysis may be more reliable. Further investi-
gation is required to determine reliable detection limits over a variety of
sample materials for both bulk and thin-layer methods. Although it is
problematic whether such limits could be firmly established given the matrix
68
-------
dependency of the sensitivity of the method and the extreme variability
observed in bulk insulation matrix materials, sensitivity would be expected
to improve if step-scanning analysis were routinely employed.
It should be emphasized that because of the small number of laboratories,
participating in this study and the diversity of methods actually employed,
it is not possible to draw any firm conclusions from the results of these
analyses. However, the following observations can be made:
1. The bulk methods appear to be at least as accurate and precise
as thin-layer methods over the range of samples included in
this study and significantly more accurate for the analysis
of chrysotile; and
2. There is a suggestion that thin-layer methods of analysis may
be more reliable (i.e., more sensitive) than bulk methods at
the 1 percent level of chrysotile in a simple matrix.
Since chrysotile is the most commonly occurring asbestos mineral in
bulk insulation materials, and since most laboratories routinely performing
quantitative analysis of asbestos in insulation samples use a bulk method of
analysis, the first observation suggests that for a wide-scale screening
program, use of bulk methods of XRD analysis, ancillary to PLM, should be
given further consideration. It should be noted, however, that the sugges-
tion that bulk methods are at least as accurate and precise as thin-layer
methods may be due to the fact that the laboratories performing the analyses
by bulk methods were more experienced in this method of analysis than those
using thin-layer methods. The second observation indicates a need to further
evaluate both methods at the 1 percent level of detection. Recommendations
for further development and evaluation of both bulk and thin-layer methods
are detailed in Section 3.2.
69
-------
SECTION 8
COMMENTS
Comments made by participating laboratories in written reports were a
valuable source of information for this technical evaluation and may con-
tribute to future methods refinement. Selected comments under consideration
are discussed below.
Commenting on PLM, analysts most frequently voiced concern over the
additional time and effort required for quantitative analysis by point count-
ing. Estimates of the impact of the PLM quantitative section cite doubled
and tripled analysis times and projected cost increases of 150 to 200 percent
Analysts objected most strongly to the use of point tounvi. or, samples
containing a relatively high percentage of asbestos, e.g., more than 30 per-
cent. It is felt that the objectivity of the quantitative estimation proce-
dure is not as critical at this level as it is in the 1 to 10 percent range.
Additionally, eye fatigue reduces the number of samples an analyst can
analyze per day by point counting. Practice and greater familiarity with
the method may, however, bring analysis time more in line with that of
current procedures. The use of a staged counting process that would allow
the counting of fewer points on high percentage samples could.be investigated.
There were some problems with application of the PLM method. Specific-
ally, some operators were biased toward picking out fibers and found it dif-
ficult to subsample "randomly." Also, teasing the sample apart with forceps
and/or dissecting needles resulted in an uneven distribution of sample
material on the microscope slide. This is in contrast to the experience of
one laboratory that milled all samples before analysis. Milling samples re-
sults in a finer grained material that can be distributed more evenly on
slides. However, mill'ng may disrupt fiber bundles or comminute fibers to
<3 pm in length, thus distorting the results of quantitative analysis or making
fibers more difficuH to identify.
Overlaying particles (i.e., asbestos fibers superimposed on matrix, or
vice versa) were also a problem in the analysis for which no guidance was
70
-------
offered. The method has been modified to require that a point be scored for
both categories when overlays occur.
Several reports commented that the definition of "fiber" used in the
method was confusing and did not require positive identification of the
particles as asbestos. Changes in the text have been made to correct this
di screpancy.
Two errors in the computation of the confidence interval (CI) were dis-
covered and corrected. On pagr 10 of the protocol, the CI should be ±0.035
instead of 0.018. On the PLM reporting form (Appendix D), the square root
symbol was omitted. Most laboratories realized this error and corrected for
it in their reports. It is recommended elswhere in this report that the CI
computation be deleted from the method.
Analysts at one laboratory provided extensive review comments on the
PLM protocol. The reviewers felt primarily that reliable quantitative
analysis is not possible using microscopical techniques without allied
quantitative chemical and physical procedures, which should be separate from
and more stringently specified than sample preparation procedures for quali-
tative PLM analysis. The following points were emphasized in support of
their recommendation.
1. Subsamples of a bulk material taken with forceps are unlikely
to be representative;
2. The method does not contain a description of how subsamples
are to be uniformly dispersed on the microscope slide;
3. Grinding a sample with mortar and pestle, an optional sample
preparation step, may cause separation of amphibole bundles,
which would bias point counting results;
4. Grinding a sample with a Wylie mill, an optional sample
preparation step, may cause the shearing of particles with
fibrous habit (>3:1 aspect ratio) from prismatic particles;
5. Step-by-step descriptions should be given of quantitative
matrix reduction procedures (specifically, low-temperature
ashing, NaOH &nd CH3C00H dissolution, and gravimetric calcu-
lations) with instructions as to when they are to be used on
representative sample types most frequently encountered.
An additional objection raised was that point counting does not provide for
the quantitation of phases that occur as coatings, such as binders and
resi ns.
71
-------
It is recognized that all sample preparation steps prior to point
counting must be performed quantitatively for analytical results to be
meaningful. However, absolute standardization of such procedures has not
been thought feasible for bulk samples because of the extreme range of
sample composition encountered. Several other reviewers with considerable
analytical experience have stated that for the majority of samples no matrix
reduction should be performed. Sample preparation procedures were therefore
included in the method as optional steps to be used at the discretion of the
analyst. Further systematic investigation would be required to determine if
stricter guidelines could be successfully applied.
The quantitative XRD procedure in the Tentative Method is more time
consuming and costly than alternative bulk techniques. Results of the study
further indicate that at this time very few laboratories are set up to
perform the thin-layer analysis as prescribed. Several comments were received
concerning obstacles to the generalized use of quantitative XRD methodology.
Acquisition of appropriate asbestos standards is expected to be a major
problem. UICC standards are not exceptionally pure and are reported to be
in increasingly short supply. It is essential that reliable sources of
standard materials be identified and that the purity of the standards be
known or determine'' before accurate calibration curves can be obtained.
The problem obtaining comparability between standard and sample
asDestos materials is cf critical importance, with no easy or straightfor-
ward solution immediately presenting itself.
The main drawback to the thin-layer procedure as applied to bulk samples
is the sample preparation step. The requirement to grind the sample to pass
a lO-pm sieve is not only time consuming and costly but may not be feasible
for all samples. Furthermore, because the matrix material itself may alter
the asbestos grinding characteristics, the validity of standards prepared in
a manner similar to sample materials is questionable.
Two typographical errors were detected in the XRD protocol and have
been corrected. Specifically, in Section 7.2.3.9 (line 6), "0.1 mg," the
total sample weight deposited on the silver membrane filter for analysis,
should read "ca. 1 mg." In Table 2 (line 4), the powder diffraction file
number for nonfibrous "amosite" should read 17-745 instead of 17-795.
Participating laboratories have been notified of both changes to the protocol.
72
-------
SECTION 9
REFERENCES
1. Charles Poole and Harry Teitelbaum, Support ripcument, Asbestos Contain-
ing Materials in Schools: Health Effects and Magnitude of Exposure,
EPA-560/12-80-003, U.S. Environmental Protection Agency, Washington,
DC, October 1980, 116 pp.
2. Joseph J. Breen and Elizabeth F. Bryan, Federal Register Citations Per-
taining to the Regulation of Asbestos, EPA-560/6-79-007, U.S. Environ-
mental Protection Agency, Washington, DC, April 1979, 40 pp.
3. Environmental Protection Agency, Asbestos-Containing Materials in School
Buildings: A Guidance Document, Parts 1 and 2, EPA/OTS #C00090, March
1979.
4. D. Lucas, T. Hartwell, and A. V. Rao, Asbestos-Containing Materials in
School Buildings: Guidance for Asbestos Analytical Programs, EPA 560/13-
80-017A, U.S. Environmental Protection Agency, December 1980, 96 pp.
5. E. P. Brantly, Jr., and D. E. Lentzen, Asbestos-Containing Materials in
School Buildings: Bulk Sample Analysis Quality Assurance Program, EPA
560/13-80-23, August 1980, 27 pp.
6. J. L. Graf, P. K. Ase, and R. G. Draftz, Preparation and Characterization
of Analytical Reference Minerals, DHEW (NI0SH) Publication No. 79-139.
June 1979.
7. W. J. Campbell, C. W. Huggins, and A. G. Wylie, Chemical and Physical
Charactorization of Amosite, Chrysotile, Crocidolite, and Nonfibrous
Tremolita for National Institute of Environmental Health Sciences
Oral Ingestion Studies, U.S. Bureau of Mines Report of Investigation
RI8452, 1980.
8. A. V. Rao et al., Analysis of Battelle Bulk Asbestos Duplicate Samples,
Research Triangle Institute (Project 43U-1864-04), February IS, 1980,
4 pp.
9. F. Chayes, Petrographic Modal Analysis: An Elementary Statistical Ap-
praisal, John Wiley & Sons, New York, NY, 1956, 133 pp.
10. NIOSH Manual of Analytical Methods, Volume 5, U.S. Department of Health,
Education, and Welfare, August 1979, pp. 309-1 to 309-9.
73
-------
11. M. Taylor, Methods for quantitative determination of asbestos and
quartz in bulk samples using X-ray diffraction, The Analyst, 103(1231):
1009-1020, 1978.
12. Churchill Eisenhart, Realistic evaluation of the precision and accuracy
of instrument calibration systems, Journal of Research of the National
Bureau of Standards (C), 67C(2):Apri1-June 1963.
13. E. M. Chamot and C. W. Mason, Handbook of Chemical Microscopy, Volume
One, John Wiley, New York, NY, 3rd ed., 1958.
14. G. J. Lieberman, R. G. Miller, Jr., and M. A. Hamilton, Unlimited
simultaneous discrimination intervals in reqression, Biometrika,
54:133-145, 1967.
15. J. S. Garden, D. G. Mitchell, and W. N. Mills, Nonconstant variance
regression techniques for calibration-curve based analysis, Analytical
Chemistry, 52:2310-2315, 1980.
16. Federal Register Citation, 45 FR 61987, Wednesday, September 17, 1980,
Part VI, Environmental Protection Agency, Friable Asbestos Containing
Materials in Schools; Proposed Identification and Notification.
-------
APPENDIX A
STATISTICAL PROCEDURES
A-l
-------
APPENDIX A
STATISTICAL PROCEDURES
A. 1
1 n
Mean squared error (MSE) = - 1 (w^ - w)"
1 n
Sample variance (S2) = - X Cw. - w)2
n i=l 1
where w = true value,
- ] n
w = - I w. is the average of the , and
i = 1, 2, ... i
If Bias = Average Error ¦•= w - w;
MSE = i I (w. - w)
i=I
1 n - -
= - I (w^ - w + w - w)2
1 n - - 1 n - 1 n -
= - I (w - w)2 + 2(w - w) - I (w - w) -!¦ - I (w - w)
i=l i=l i=l
1 n
= - I (Wi. - w)2 + n(w - w)2
MSE = Variance + Bias2 .
A.2 '*
The importance of accounting for asbestos and laboratory effects can be
seen from the results in Table A-l. The variable LA was regressed on LW,
A-2
-------
TABLE A-l. LABORATORY AND ASBESTOS TYPE EFFECTS REGRESSION
RENSFAI, LISEAR 30DELS P80CSDUR2
DSDSND£*'T TAFIABt?: IA
SO0KC?
OF
SO?) OF SQUARES
SEAN SQ0AR2
f VALUE
(1005 L
52
126U.Q0720J82
2u.
3251385U
197.97
EFPOi?
76
9.33«298U6
0.
1 22B7235
PP > F
UNCOPS'CTSD TOTAL
121
127a.2«5S0?23
0.0001
R-sqrMfi::
C. ?,
S'D DLT
LA aEAN
0.«O2«i"?C
11.5207
0.35053152
2.
96539U18
soufc0.
rt?
71 ?F, I SS
F ULUS
PE > F
A5B
LAB
Vi
L»«tsn
L»"T.AP
A3 B* LAB
2
16
1
1
16
16
1139.72«!»779a
27.1002776«
a3.j<»5«Q99Q
0.71*92507
5. P Of # 101 u
3.5 1370013
«637.9«
13. 78
6?5.57
5.33
2.95
<1.33
0.0001
0.0001
0.0001
0.01*1
0.0008
0.0001
save* bp
USB t
LAB IS
La l
LV*1S~1 1
IWLA6 16
ArB'tAn 16
?YP° I? 33
F UL'JS
PH > F
5.53191251
BU.-'a
0.0001
19. 0 U69-5663
9. 18
0.0001
21.66«fl«2P6
176.35
0.0001
0.*=0092383
7.3 3
0. 0001'
5.87623781
2.99
0.000 '!
9.5137 0013
«.33
0.0001
A-3
-------
with both slope and intercept allowed to vary with asbestos type and labora-
tory. The third order interaction term LW*ASB*LAB was not significant
(P>.5) and the regression was rerun without this term. All the remaining
terms are highly significant.
A.3
To test for differences in bias between Groups P, B, and 0, weighted
GLM regressions were performed in logarithmic coordinates allowing slopes
and intercepts to vary with group and with asbestos type. The weights for
these regressions were obtained from preliminary regressions of standard
deviations on means of logarithms of reported results. The relations did
not change significantly with Group (P, B, 0), so the same line was used for
all three groups. The results of the full regression are shown in Table A-2.
Allowing all effects and interactions among GROUP, ASB and LW, three of the
interaction terms were insignificant (P > 0.5). The regression was rerun
with these interactions omitted. Results are in Table A-3, and show Groups P
and B to be similarly biased (P = 0.34, not significant) while Group 0 is
more biased than Group P (P << 0.01).
A.4
The regression related the logarithm of group standard deviations (LS)
to the logarithm of group means (LM) using the model
log s = a log x(l - x) + b + error .
The function x(l - x) was employed because it produced a higher overall
correlation (R2 = 0.98) than the regression using x (R2 = 0.93). The regres-
sion parameters a and b did not vary significantly with asbestos type for
any of the groups (P, B, 0). By this analysis, Groups P, B, and 0 are not
significantly different with respect to precision (P > 0.3 for all pairwise
comparisons of slopes and intercepts).
A.5
The following steps were used in transforming reported area percent
data to predicted weights (A -» W). All regressions were performed in two
ways: (1) unweighted; and (2) weighted for differences in variance between
sample series. No discrepancies between weighted and unweighted regressions
A-4
-------
TABLE A-2. GROUP EFFECTS REGRESSION, ALL INTERACTIONS
a .iodels ??ocf3'IR;
DEPEJ'D'-'l? V^IABIS; l\
»rI.:MT: v:nv
seuFcr
DF
£ OH OF e90 AfiSS
!5EAM SQUAK 3
HOC L
12
1 1 °9 2.9 0°6 5728
099.'4091381 1
s:fq?
2 15
''"'O.O 423 78 54
1. 25601106
UNCOBP^CTFD TCTAI.
">27
1 22fi 2. 9520 J5B 3
°-SQI»P?
C. V.
3TD 0£V
LA 82A H
0. 9770f5
4311
1.12071899
2-99402753
SOIfCH
DF
TYPE r SS
F VAL'J* PP > F
A? D
2
1 1 '46 5. Q 7°Qij7 39
a5K«.'»« 0.0001
TKV.'P
2
21.77609024
9.67 0.300?
LW
1
M1.-i(»««52<32
H1.63 0.0001
A.i8*G?C'JP
2
2. u«258002
O.OQ 0.3739
1^*8 53
1
9.26°18d89
7.3"? 0. 007 1
L-»¦; FC'J?
2
1.? 1450236
0.52 0.5933
L1» N ? -1» "¦" o rp
2
0.2023u557
0.18 0.9226
F V&LUS
795.70
PS > F
0.0001
snr cz
3"
TY?* IV SS
F V.U'JS
PP > P
1
I3.* 237b 106
1 a. 'S
0.0002
if o" r
j
5.982P'K1 )
2.3 8
0.0949
L<
1
,rO.=, 769 31 J5;
200.30
0.0001
2
0.->8701153
0. 27
0.7610
L '» * 1 5 3
1
7.9c«77Q71
6. ~>U
0.0125
1 »*T !-0')n
2
1.2 5«u SCO
0. 50
0.6076
s°« t ou^
2
0. 2023-455'=•7
0. 0 8
0.«226
A-5
-------
TABLE A-3. GROUP EFFECTS REGRESSION, SIGNIFICANT INTERACTIONS
GENERAL LINEA ? JIODEI.S PtfOCE">URE
VA9I\B! E: L\
WEIGHT: VINV
sousce
10DEL
2 r-F C 9
UNCORRECTED TOTAL
DP SUN OF EQUATES
6 11988.°3065277
221 274.02133306
227 12262.95203583
MEAN SQOAEE
1°9fl. 15 51087 9
1.23991576
P VALUfc
1611.52
PR > P
0.0001
r-souape
0. Q 7 7 6 5 c.
C. V.
?7. 1 Q12
STD DEV
1. 1 1 351505
LA .iEAN
2. 00 40 2 75 3
SOURCE
A3?
35 0'JP
L'J
LW«\SD
DF TYPE I SS
2 11U65.97908739
2 21.77f0902'4
1 -19 1. 8 868C2H2
1 9.28859232
F VALUE
4 ^73.6°
8. 70
396.71
7. U 9
PR •> F
0. 0001
0.0002
C.0001
0.0067
SOUF.CE
A~B
•JIPO'JP
LW
L « • A S B
DF
TYPE IV SS
22. 7057 530
21.46203795
2^3.32102770
9.2885^232
F VALUE
13. 31
9.66
2°3.02
7. no
PR > P
0.0001
0.0002
3. 000 1
U.0067
? \F \ 'A r"rc
\ r a
3 POL' P
LW
LWAS?
A ,10 SI
\*OSI
•~!tr YS
isrir. a:e
1. Q'4C°? 2 36
0. 7?683 223
0 . 0 u 7 7 2 1 °2
0.41351216
0.50000000
0 . 7u <55 1
-0 . ?1f,e 4 0 a
0.30000000
T FOR HO:
?AF MET rF = 0
9. 47
5. 16
0.'1C
4. 16
18. 18
-2 .
> |T|
0 .0 0 0 1
0.0131
J. 34 '4 7
0.0001
0.0001
0.0067
STD EPFU5 OF
ESTIJIA-E
0.19U04^97
0.15450008
0.092f3Q2?
0.09539073
0. 041 2267i»
0.07550152
Bias comparison, B and 0 vs. P.
A-6
-------
i/ere found with respect to the significance of various effects. Attention
i/ill be focused only on the results of unweighted regressions.
For each group (P, B, 0), the following sequence of regressions was
jerformed.
1. Regress LW (log of nominal weight percent) on LA (log of
reported result) for each combination of laboratory and
asbestos type. That is, estimate the slopes and intercepts
of LW = c(asb,1ab) • LA + d(asb,lab). The functional notation
stresses the dependence of slopes and intercepts on asbestos
type and laboratory, as discussed in Section 5.5.
2. Using the parameters estimated in 1, adjust the reported data
A to
W* = edAc .
3. Regress W on W*, with forced zero intercept, according to
W = rW* .
4. Using the parameters estimated in 1 and 3 (c and b = red),
transform W* to predicted weights W.
W = bAW*c .
The predicted weight W corresponding to each reported result
is thereby obtained.
The reason for step 3 is that the regressions in Step 1 tend to balance
out errors in sign and magnitude. However, if equal errors, opposite in
sign, are exponentiated, then an imbalance occurs. The regression in 3 is
designed to compensate for this, and can be achieved with a hand calculator
by entering every (W*, W) pair a second time as (-W*, -W).
A-7
-------
APPENDIX B
TENTATIVE METHOD
March 198.0
Revisions have been made to the Tentative Method pursuant to the con-
clusion? of this study. Appendix B should not be used by laboratories as a
reference or ana1- Seal protocol. Current editions of the Interim Method
for the Determination of Asbestos in Bulk Insulation Samples are availubv
from Gene Brantly, Research Triangle Institute, 800-334-8571.
B-l
-------
TENTATIVE METHOD FOR THE DETERMINATION OF ASBESTIFORM
MINERALS IN BULK INSULATION SAMPLES
BY POLARIZED LIGHT MICROSCOPY AND X-RAY POWDER DIFFRACTION
A tentative method is carefully drafted from available source
information. This method is still under investigation and
therefore is subject to revision.
Environmental Monitoring Systems Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
March 1980
-------
POLARIZED LIGHT MICROSCOPY
1. Principle and Application
1.1 Bulk samples of building materials taken for asbestos identification
are first examined for homogeneity and preliminary fiber identification at low
magnification. When discrete layers are identified, each is treated as a
separate material and should receive individual characterization. Positive
identification of suspect fibers is made by analysis of subsamples with the
polarized light microscope. Asbestos quantitation is performed by a point-
counting procedure.
1.2 This method is applicable to all bulk samples submitted for identi-
fication and quantitation of asbestos components.
2. Range
The range of the analysis is dependent on the amount of material exam-
ined. Quantities of asbestos in a building material sample will be subject to
wide variation in reported results because of sampling variation in an inhomo-
geneous matrix. Quantities below 1 percent are reported as The upper
detection limit is 100 percent. There is no measure of sensitivity presently
available.
3. Interferences
Fibrous organic and inorganic constituents may pose a challenge to
identification, separation, and quantitation of the asbestiform mineral content.
Spray-on binder materials may coat fibers to impart color and obscure optically
determined parameters to the extent of masking the fiber identity. Fine
particles of other materials may also adhere to fibers to an extent sufficient
to cause confusion in identification.
4. Precision and Accuracy
Adequate data for accuracy and precision measurements art not currently
available.
5. Apparatus
5.1. Analysi s
5.1.1. A low-power binocular microscope, preferably stereoscopic,
is used to examine the bulk insulation sampie as received.
-------
b.1.1.1. Microscope: binocular, 10-45X.
5.1.1.2. Light Source: incandescent or fluorescent.
5.1.1.3. Forceps, Dissecting Needles, and Probes
5.1.1.4. Glassine Paper or Clean Glass Plate
5.1.2. Sample preparation apparatus requirements will depend upon
the insulation sample type under consideration. Various physical ana/or
chemical means must be employed for an adequate sample assessnent.
5.1.2.1. Ventilated Hood or negative pressure glove box.
5.1.2.2.
Microscope Slides
5.1.2.3.
Covers 1ips
5.1.2.4.
Disposable gloves.
5.1.2.5.
Mortar and Pestle: agate or porcelain
5.1.2.6.
Wylie Mill (optional)
5.1.2.7.
High-Speed Blender (optional)
5.1.2.8. 100-mL Beakers and Assorted Glassware (optional)
5.1.2.9. Centrifuge (optional)
5.1.3. Compound microscope requirements: A polarized light micro-
scope complete with polarizer, analyzer, port for wave retardation plate, 360°
graduated rotating stage, substage condenser, lamp, and lamp iris.
5.1.3.1. Polarized Light Microscope: described above.
5.1.3.2. Objective Lenses: 10X, 25X, and 45X or near equivalent.
5.1.3.3. Dispersion Staining Objective Lens (optional)
5.1.3.4. Ocular Lens: 10X minimum.
5.1.3.5. Eyepiece Reticle: cross hair or 25 point (Available
from Preiser Scientific and other microscope dis-
tributors)
5.1.3.6. Michel-Levy Interference Color Chart
5.1.3.7. Red I Retardation Plate (First-order compensator)
5.1.3.8. Abbe Refractometer (optional)
6. Reagents
6.1. Sample Preparation
6.1.1. Distilled Water
6.1.2. 0.5 N H?SQ4: ACS reagent grade (optional)
6.1.3. 0.5 N HC1: ACS reagent grade (optional)
6.1.4. Sodium metaphosphate (NaP03)6 (optional)
- 2
J. j
-------
6.2. Analytical Reagents
6.2.1. Refractive Index Liquids: 1 490-1.570, 1.590-1.720 in
0.002- or 0.004-step increments.
6.2.2. Refractive Index Liquids for Dispersion Staining: high-
dispersion series, 1.550, 1.605, 1.630.
6.2.3. UICC Asbestos Reference Sample Set: Available from: UICC
MRC Pneumoconiosis Unit, Llandough Hospital, Penarth,
Glamorgan CF6 1XW, UK.
6.2.4. Tremolite-asbestos (source to be determined)
6.2.5. Actinolite-asbestos (source to be determined)
7. Procedures
NOTE: Exposure to airborne asbestos fibe .. is a health hazard. Bulk
samples submitted for analysis are usually friable and may release fibers
during handling or matrix reduction steps. All sample and slide preparation
should be carried out in a ventilated hood or glove box with continuous airflow
(negative pressure). Handling of samples without these precautions may result
in exposure of the.analyst and contamination of samples by airborne fibers.
The level of airborne fibers should be monitored in accordance with NI0SH
Analytical Method #P&CAM 239: Asbestos Fibers in Air (see DHEW/NIO'SH publica-
tion no. 79-127, February 1979).
Refractive index liquids typically contain several toxic compounds,
including brominated naphthalene, brominated and iodinated ring compounds, and
hydrogenated terphenyls. (disposable gloves should be worn by the analyst to
avoid prolonged skin contact with these materials. Prepared slides and waste
bulk material should be disposed of in accordance with proper procedures for
toxic substances. Asbestiform materials should be double-bagged, labelled,
and buried in appropriate landfill or burial sites.
7.1 Samplinq: Samples for analysis of asbestos content shall be taken
in the manner prescribed in the guidance document Asbestos-Containing Mate-
rials in School Buildings, EPA #C00090, part 1. If there are any questions
about the representative nature of the sample, another sample should be request-
ed before proceeding with the analysis.
7.2 Analysi s
7.2.1. Gross Examination: Bulk samples of building materials taken
for the identification and quantitation of asbestos are first examined for
homogeneity at low magnification with the aid of a stereomicroscope. The
3
-------
core saapl? is carefully removed froa the saepling canister onto a glassine
transfer paper or clean glass plate. If possible, note is aade of the top and
bottoa orientation. When discrete strata are identified, each is treated as a
separate Mterial so that fibers are identified and quantitated in that layer
only.
7.2.2. Saapie Preparation: Bulk Mterials submitted fo«* asbestos
analysis involve a wide variety cf aatrix materials. Representative subsaaples
say not be rsadily obtainable by siaple aeans in heterogeneous aaterials, &nd
various steps nay be required to alleviate the difficulties encountered. In
aost cases, however, the best preparation is aade by using fine-pointed forceps
to saople at several places froa the bulk material. Forcep saaples are ineersed
in a refractive index liquid on a aicroscope slide, teased apart, covered with
a cover glass, and observed with the polarized light aicroscope.
Alternatively, attempts nay be aade to hoaogenize the saaple or eliminate
interferences before further characterization. The selection of procedure 's
dependent upon the saaples encountered and personal preference. The following
are presented as possible aethods.
A aortar and pestli can soaetiaes be used in size reduction of soft or
loosely bound aaterials though this aay cause satting of so cue saaples. Such
saaples aay be reduced in a Wylie aill. Apparatus should be clean and extreae
care exercised to avoid cross-contaaination of saaples. Periodic checks of
the particle sizes should be aade during the grinding operation so as to
preserve any fiber bundles present in an identifiable fora.
Treataent aay occasionally be reoyired to eliainate interferences. For
ceaentitious aaterials, dissolution of the calcareous substances aay be effected
with warm dilute sulfuric acid (0.5 N at 65° C) [8]. Calcite may be dissolved
with want! dilute hydrochloric acid (9). Wash twice with distilled water,
boing careful not to lose the particulates during decanting steps. Centrifug-
ation of the suspension will prevent significant fiber less. Prolonged acid
contact with the saaple aay alter the optical characteristics of cNrysotile
fibers and should be avoided.
Coatings and binding aaterials adhering to fiber surfaces nay be removed
by treatment with sodiua metaphosphate (10). A4d 10 at of 10 g/L sodiun
netaphosphate to a snail (0.1 to 0.5 nL) sample of bulk material in a 15-tjL
-------
glass centrifuge tube. For approximately 15 seconds each, stir the aixture on
a vortex sixer, place in an ultrtjtnic bath and then shake by hand. Repeat
the series. Boil the contents of the tube over a bunsen flame for about 20
seconds. Collect the dispersed solids by centrifugation at 1000 rp* for 5
¦inutes. Wash the saaple three times by suspending in 10 aL distilled witter
and recentrifuging. After washing, resuspend the pellet in 5 aL distilled
water, place a drop of the suspension on a microscope slide, and dry the slide
at 110° C.
In saaples with a large portion of cellulosic or other organic fibers, it
aay be useful to ash part of the saaple and view the residue. Ashing should
be perforaed at temperatures below 550° C. It should not be performed in an
open flaae or high-temperature oven because dehydration of the asbestos minerals
results in changes of refractive index and other key parameters, and possible
artifact formation. Ashing and acid treatment of saaples should not be used
as standard procedure. In order to aonitor possible changes in fioer character-
istics, the material should be viewed microscopically before and after any
saaple preparation procedure. Use of these procedures on saaples to be used
for quantitation requires a correction for percent weight loss.
7.2.3. Fiber Identification: Positive identification of asbestos
requires the determination of the following optical properties.
Morphology
Color and pleochroism
Refractive indices
Bi refringence
Extinction characteristics
Sign of elongation
Table 1 lists the above properties for coanercial asbestos fibers. Figure \
presents a flow diagraa of the exaaination procedure. Natu-al variations in
the conditions under which deposits of asbestiforn minerals are formed will
occasionally produce exceptions to the published values and differences fron
the UICC Standards. The sign of elongation is determined by use of the Red I
retardation plate and crossed polars. Refractive indices nay be determined by
either the Becke line test or dispersion staining. Becke lines are eaxinized
by viewing with a nearly closed substage diaphrags.
5
\
-------
lit IIM.I I Oil
«iigl« (d«
Chrytol1 If
Cnxidulllt
Wdvy flUsrv MUr bund1*1
h«*« -.pUycd tnlt «od "Mnk»"
At{«Kt K^lu typically 'SO. I
Colsrltt; . nonplaochroli
^tr«lqtit, riyjld Mb*ri
A»p«ct ratio lr»U«tly >10:1
irnmlth, r.onpl*ocfirotc
OfMvnr lucluvlont My b«
^•rry»i>l
$*r4tgtil fltxrrt, Ittt rigid
Ibin Mailt* thick Uteri
«nd Imnillti cnuw.il>, blu* to
|Hir;l«-blw In color. PI*o-
chrelc Blrrfrlngvnc* it ion
41, G 004
1 49i l 660 I MM M.-T
1 bib-1 t.% I.6SV ¦ lit
I 6i4-3 101 1 668-1 111
Oil
0?/
0»S
AnttupiiylllU-
itlwtlet
Ircnilll*-
KIInolIt*
avbe»to*
Vtriijlil Iltort traS.«clcul«r I S%-1 bW
c lc«v*Q» IrtgMnlt S60
or priiMllc cie«v«9« Iraq-
¦tnli Slngl* crylt4l* pr*-
<*oaln.«t« Atpact ratio <10.1
Colorlcvt to pile green
161S 1 6/6
I 6^^¦I 688
02?
0??
tn« retorinc* i
Sua rtfrrimc '
''liber*. tub|t(ltd to Icil liy My be brown I ih
^ llwri dtlliieil IuvIih) atficd ratio • I I
ct at litf t Ion foi lr«^e»ti £o^»o-*ll* lllwrt iktliiguUli it 0"
• Weo Inlo'Utliin in Idtif* I lut been revived for current edition* of the PIN mMumI
o-?o«
-------
Polarised light microscop? analysis: For each type of material identified by examination of sample at low nugniticoiiun.
Mount spacially dispersed sample in 1.550 Rl liquid. (If using dispersion staining, mount in 1.550 HD.) View at 100X
with bath plane polarued light and cruised polars.*
Fibaa
pineal
Fihan
4
(mm aMttwari ifafet a 1MX m*4MX
I
FANiptMil
fHwi m iw:r«pic (dii^ptu M il
at itigs iiKUm uauad
paten)
^tuiUi film IkMi:
flkm an mtinuop* (n&M
•¦tiaetteB et §9" laurafc oi
lUga (tMittJ
FAmatNl
I
IiMiMmcMykH.
Rapart • IX ntniK.
prnnt.
Mnwtd w&d:
S 28 m» awtoint iumttm.
Rl typtt&l • t£)
8-388 urn 4iana
-------
Inexperienced operators may find that the dispersion staining technique
is sore easily learned, and should consult references 3 and 6 for guidance.
Central stop dispersion staining colors are presented in Table 2. Available
high-disp«rsion (HO) liquids should be used.
7.2.4. Quantitation of Asbestos Content: Asbestos quantitation is
performed by a standard point-counting procedure. An ocuiar reticle is used
to visually superimpose a pdint or points on the microscope field of view.
Record the number of points positioned directly above each kind of particle or
fiber of interest. Score only points directly over asbestos fibers or non-
asbestos matrix material. Oo not score empty points for the closest particle.
This provides a determination of the areal percent asbestos. Reliable conver-
sion of areal percent to percent of dry weight is not currently feasible
unless the specific gravities and relative volumes of the materials are known.
For the purpose of this method, "fibers" are defined as having an aspect
ratio greater than 3:1 and substantially parallel borders.*
A total of 400 points superimposed on either asbestos fibers or non-
asbestos matrix material must be counted over at least eight different prepara-
tions of representative subssap'es. Take eight fine-pointed forcep samples
and mount each separately with the appropriate refractive index liquid. Count
50 nonempty points on each preparation, using either
1. A coss-hair reticle and mechanical point-counting stage; or
2. A reticle with 25 points and counting at least 2 randomly selected
fields.
For samples with fixtures of isotropic and anisotropic materials present,
viewing the sample with slightly uncrossed polars or the addition of a Red I
retardation plate to the polarized light path will allow simultaneous discrim-
ination of both fiber types. Quantitation should be performed at the lowest
magnification of t^e polarised light microscope which can effectively distin-
guish the sanple components. Confirmation of the quantitation result by a
second analyst on soat percentage of analyzed sassples should fte us«d as standard
quality control procedure.
If seven or fewer of 400 non-e«pty points are scored for asbestos fiber,
report <1 percent asbestos." For all other results, the percent asbestos is
calculated as follows:
"and being positively identified as asbestos,
onitted in final draft.
„ 0
-------
p -- a/n
% asbestos = 100 p
where
a = nuaber of asbestos counts,
n = nus&er of noneopty points counted (400).
The value reported should be rounded to the nearest percent.
If a > 10, 95 percent corsfidenre intervals can be constructed about p by
the equation
p ~ 1.96 x ^ ea
where q = 1-p.
Example:
a = 60 points
p = 60/400 = 0.15
q = 1 - 0.15 = 0.85
p ~ 1.96 0.15 ~ 0.018"
Report: 15 t 2%
8. References
1. Rohl, Arthur N., Arthur M. Langer, and Ann G. Wylie. "Mineral Character-
ization of Asbestos-Containing Spr&y Finishes." Appendix H. Asbestos
Containing Materials in School Buildinqs: A Guidance Document. t-'art I.
POTTTia)ff5557T573".
2. Caapbell, W. J., R. L. Blake, L. L. Brown, E. E. Cather, and J. J. Sjoberg.
Selected Silicate Minerals and Their Asbestifona Varieties: Kineraloqi-
cal Definitions and Identification-Characterization. U. S. Bureau of Mines
Information Circular 8751, 1977.
3. McCrone, Walter C. "Evaluation of Asbestos in Insulation," Arerican Labora-
tory. December, 1979.
4. Asbestos-Containinq Materials in School Buildinqs: A Guidance Oocuaent.
^a/6Ys *cooo$o, isW. ——
5. Kerr, Paul F. Optical Mineralogy, 4th ed. McGraw Hill, New York, 1977.
6. McCrone, w. C.. L. B. McCrone. and J. G. Oelly. Polarized Light Microscopy.
Ann Arbor Science Publishers, Ann Arbor, 1978.
7. Hogg, R., and A. Craig. Introduction to Mathematical Statistics. Third
Edition. MacMillan, 1973.
8. Personal cow unication. Walter C. McCrone.
•correct result is 0.15 t 0.035.
9
f'l I J
-------
9. Persona) coMirnication, Martin Rutstein.
10. Personal coaaunication, Ja^es S. Webber.
11. Personal covaunication, Yachain Yaffe.
9. 91bIioqraphy
NcCrcs«, Walter C. New Asbestos Particle Atlas. Ann Arbor Science
Publishers, Ann Arbor, June 1980.
10
-------
Acknowledgment
This «ethod is jased in part on contributions and review coments of
participants in the *yaposiun "Methods definition for the polarized light
aicroscope and x-ray diffraction analysis of bulk saaples for asbestos," U. S.
Bureau of Nines, Avondale Research Center, Avondale, Maryland, October 23-24,
1979.
11
•)s /"s
-------
X-RAY DIFFRACTION
1. Principle and Applicability
1.1 The theory of X-ray diffraction (Xftfl) 1s we 1 l~docmerited (1, 2].
Any solid, crystalline Material will diffract an 1mp\ingent beam of parallel.
Monochromatic X-rays whenever Bragg1 s Law,
A = 2d sin 6,
1s satisfied for a particular set of planes 1n the crystal lattice,
where
o
A » the X-ray wavelength, A;
d * the Interplanar spaclngs of the set of reflecting lattice planes,
o
A; and
0 3 tfte angle of Incidence between the X-ray beaa and the reflecting
lattice planes.
By appropriate orientation of a sample relative to the Incident X-ray beam,
an X-ray diffraction pattern can be generated that. In uost cases, will be
uniquely characteristic of both the chemical composition and structure of
the crystalline phases present.
Unlike optical Methods of analysis, however, XRO cannot determine
crystal Morphology. Therefore, In asbestos analysis, XRO does not distin-
guish between fibrous and nonflbrous fores of the serpentine end amphlbole
olnorals (Table 1). However, when used In conjunction with optical Methods
such ^s polarized light Microscopy (PLM), XRO techn1qu«s can provide a reli-
able analytical Method for the Identification and characterization of asbestl-
fora Minerals In bulk materials.
Bulk Material staples are Initially analyzed by PLM for Identification
and quantitation of asbestos. Subsequent analysis by XRO proceeds 1n two
stages.
For qualitative analysis, saaples are Initially scanned over limited
O
diagnostic peak regions for the serpentine (7.36 A) and amphlbole (8.3-
O
8.5 A) Minerals (Table 2). Standard slow-scanning Methods for bulk sasplo
analysis May be used for Materials shown by PLM to contain Major amounts of
asbestos (>5-10 percent). Oetectlon of elnor or trace amornts of asbestos nay
1
-------
TABLE 1. THE ASBESTOS MINERALS ANO THEIR
HONASBESTIFORM ANALOGS
Asbestlfom Nonasbestifor*
SERPENTINE
Chrysotlle Antigor1t«, lizardite
AMPHIBOLE
Anthophy11i te asbestos Anthophy1111«
CuMtngtonite-grunerite asbastos Cuani ngtoni te-grun«H te
("Aaoslte")
Croc1dolite
Treaolite asbestos
Act1nolite asbestos
2
-------
N
\
TABLE 2. PRINCIPAL LATTICE SPACINGS OF ASBESTIFORM
_ MINERALS
Minerals
Principal d-spacing* (A)
and relative intensities
JCPOS
Powder diffraction file (3]
Number
Chrysotile
"Aaosite"
7.37too
3.6570
4.5750
21-543
7.36|oo
3.66,o
2 *S«s
25-645
7.10ioo
2.33ao
3.5570
22-1162 (theoretical)
8.33|oo
3.O670
2.79670
17-795*(nonf i brous)
8.22ioo
3.060,4
3.2570
27-1170 (UICC)
Aivthophyl 11 te
3.05xoo
3.06100
3.24«0
8.3370
8.269s
3.23(0
9-455
16-401 (synthetic)
Actinollte
2.72|oo
2.54)oo
3.40,o
25-157
Crocidol ite
8.35too
3.10ss
2.720*5
27-1415 (UICC)
Treaolite
8.3&ioo
2.706)00
13)oo
3.12)oo
3* 14ttS
2.706eo
2.705«o
8.43«0
8.44«0
13-437bh
20-1310° (synthetic)
23-666 (synthetic
alxture with
richter
-------
require special sample preparation and step-scanning analysis of selected
diagnostic peaks. All samples that exhibit diffraction peaks in the diagnostic
peak regions for asbestlfora Minerals are submitted to a full (5°-60* 26;
1° 26/ain) qualitative X'AD scan and their diffraction patterns compared with
standard reference powder diffraction patterns to verify Initial peak assign-
aents, and to Identify possible aatrlx interferences when subsequent quantita-
tive analysis will be perfora&d.
Accurate quantitation of asbestos in bulk samples by XRO is critically
dependent on particle si^e distribution, crystallite size, preferred orienta-
tion and aatrlx absorption effects, and coaparabillty of standard reference
and sanple Materials. The most intense diffraction peak t^at has been shown
to be free froa interference by prior PLM or qualitative XRO analysis is
selected for quantitative determination of each asbestlfora mineral.
A "thin-layer" aethod of analysis [4, 5] is recooaended in which, subse-
quent to coaalnution of the bulk aatcrlal to <10 pa by suitable cryogenic
Billing techniques, an accurately known aaount of the saaple is deposited on a
silver aeabrane filter, and the aass of asbestifora material is determined by
aeasuring the integrated area of the selected diffraction prak using a step-
scanning aod«, correcting for aatrlx absorption effects, and comparing with
suitable calibration standards.
An alternative "thick-layer" or buU aethod (6] nay be used for semiquanti-
tative analysis.
1.2 This aethod is applicable as a confirmatory method for identifica-
tion and quantitation of asbestos in bulk aaterial samples that have undergone
prior analysis by PIM.
2. Range and Sensitivity
2.1. The range of the aethod has not been determined.
2.2. The sensitivity of ttie aethod has not been determined. It will
be variable and dependent upon many factors, including matrix effects
(absorption and interferences), diagnostic reflections selected, and their
relative intensities.
3. Limitations
3.1. Interferences: The use of XRD for idantification and quantitation
of asbesti form minerals in bulk samples aay to a severely Halted by tin
4
/ • / /
-------
presence of other interfering Materials in the sample. For naturally occurring
¦aterials the commonly associated asbestos-related aineral interferences can
usually b* anticipated, however, for fabricated aaterials the nature of the
interferences may vary greatly (Table 3) and present sore serious problems in
identification and quantitation.
The interference problem is further aggravated by the variability of the
silicate Mineral powder diffraction patterns associated with alterations in
the crystal lattice arising fro* differences in isomorphous substitution and
degree of crystal 1inity. This variability often makes unambiguous identifica-
tion of the asbestos Minerals by comparison with standard reference diffraction
patterns difficult. The amphiboles, for example, exhibit a wide variety of
very similar chemical compositions, with the result b«ing that their XRD
patterns are characterized by having major (110) reflections of the moncclinic
amphiboles and (210) reflections of the orthorhoobic anthophyllite separated by
less than 0.2 A (3],
Common interferences are listed below.
3.1.1. The serpentine and aaphibole minerals occur naturally in
both fibrous and nonfibrous forms (Table I). X-ray diffraction techniques,
however, cannot distinguish between these two varieties. Therefore, >n the
absence of confirmatory PlM data, the identification of asbestos by XRO
methods is not definifjvo. In addition, the presence of nonasbestiform
serpentines and aaphibole* vn a sawple will pose severe interference problems
in the quantitative analysis of their jsbestifero analogs, unless special
sample preparation techniques and instruments on are used (9).
O O
3.1.2. Chlorite has major peaks at 7.19 A and 3.58 A that interfere
o o
with both the prinary (7.36 A) and secondary (3.66 A) peaks for chrysotile;
resolution of the priHary peak, to givf good quantitative results eay be possible
when a step-scanning mode of operation is employed.
O
3.1.3. Halloysite has a peak at 3.63 A that interferes with the
O
s*conaary (3.66 A) peak for chrysotile.
O
3.1.4. Kaolinite has a major peak at 7.15 A that nay interfere with
O
the primary peak of chrysctile at 7 36 A when present at concentrations of >10
O
percent. However, the secondary chrysotile peak at 3.66 A may be used for
quantitation.
5
/, A
-------
TABLE 3. COWtON CONSTITUENTS IN INSULATION MO
WALL MATERIALS (?/
A. Insulation aaterials
Chrysotile
"Aaosite"
Crocidolite
•Rock wool
•Slag wool
•Fibir glass
Gypsuc (CaS04 • H20)
Veraiculite (aicas)
*Perlite
Clays (kaolin)
•Wood pulp
*Paper fibers (talc, clay, carbonate fillers)
Calcfus silicates (synthetic)
Opaques (chroaite, aagnetite Inclusions in serpentine)
Heaatite (inclusions In "aaosite")
Magnesite
•Oiatoaaeeou* earth
8. Spray finishes or paints
Bassanlte
Carbonate ainerals (cal rite, doloalte, vaterlte)
Talc
Treanlite
Anthophyl11te
Serpentine (including chrysotile)
Aaosite
Crocidolite
•Mineral wool
•Rock wool
•Slag wool
•Fiber glass
Clays (kaolin)
Micas
Chlorite
Gypsum (CaS04 • H20)
Quartz
•Organic binders and thickeners
Hydroftrfgnesite
Wo 11astonite
Opaques (chroalte, Mgnetite inclusions in serpent*ne)
Heaatite (inclusions in "aaosite")
•Acorphous aaterials—contribute only to overall scattered radiation and
increased X-ray background.
6
-------
3.1.5. Gypsy has a aajor peak at 7.5 A that overlaps the 7.33 A
peak of chrvsoti1e when present as a aajor saaple constituent.
3.1.6. Cellulose has a broad peak that partially overlaps the
O
secondary (3.66 A) ehrysotile peak [6].
3.1.7. Overlap of both the primary and secondary peaks of crocidollte
O O 0 0 ——————
(8.3S A, 3.10 A) and aaosKe (8.33 A, 3.06 A) presents serious Interference
probleas when these Minerals occur In the presence of one another.
3.1.8. Carbonates say also Interfere with quantitative analysis.
CaC03 has a peak at 3.035 k that overlaps the secondary peaks of crocldolIte
O O "
(3.10 A) and awoslte (3.06 A) when present 1n concentrations of >S percent.
(Reooval of carbonates with a dilute acid wash 1s possible; however, if present,
ehrysotile aay be partially dissolved by this treataent (10].)
3.1.9. Interference between similarly spaced strong reflections of
talc and anthophylllte will significantly reduce the sensitivity of the
aethod far anthophylllte In the presence of talc. The anthophylllte peak at
1 O
8.9 A 1s ofttn usktd by a strong talc ptak at 9.3 A. Similarly, talc peaks
o o o o
at 3.12 A, 4.53 A, and 4.56 A interfere with dnthophyl1ite peaks at 3.OS A
O O
and 4.50 A. For quantitation, the 8.26 A of anthophylllte Bust be used.
O
3.1.10. A aajor talc peak at 3.12 A also interferes with a primary
treaolite peak at this saae position and '<1th secondary peaks of crocidollte
and S52ll£2- 1" th* presence of talc, the 8.38 A treaolite peak should be
used for quantitation.
o o
3.1.11. Overlap of peaks at 8 26 A and 8.38 A for anthophyllite and
treaolite. Interference when these ainerals are analyzed in the presence of
one another; however, adequate resolution aay be attained in the step-scanning
node of operation.
3.2. Matrix Effects
3.2.1. If a Cu X-ray source is used, tne presence of iron at high
concentrations in a saaple will result in significant X-ray fluorescence,
leading to loss of peak intensity along with increased background intensity
and an overall decrease ia sensitivity. This situation nay be corrected by
choosing an X-ray source other than Cu; however, this is often accorspaniad
7
l> • <-
-------
both by loss of intensity and by decreased resolution of closely spaced reflec-
tions. Alternatively, use of a diffracted beaa aonochroaator will reduce
background fluorescent radiation, enabling weaker diffraction peaks to be
detected.
3.2.2. X-ray absorption by the saaple aatrix will result in overall
attenuation of the diffracted beaa and aay seriously Interfere vith quantitative
analysis. Absorption effects say be ainiaized by using sufficiently "thin"
saaples for analysis (4, 11, 12). However, unless absorption effects are
known to be the saae for both saaples and standards, appropriate corrections
should be aade by referencing diagnostic peak areas to an interrtl standard
(6) or filter substrate (Ag) peak (4, S).
3.3. Particle Size Oependence: Because the Intensity of diffracted X-
radlation is particle-size dependent. It is essential for accurate quantitative
analysis that both saaple and standard reference aaterials have siailar
particle size distributions. The optiaua parMcle size range for quantitative
Analysis of asbestos by XRD is 1 to 10 pa (13). Comparability of saaple and
standard reference aaterlal particle size distributions should be verified by
optical aicroscopy (or other suitable aethod) prior to analysis.
3.4. Preferred Orientation Effects: Preferred orientation of asbesti-
fora ainerals during saaple preparation often poses a serious problea In
quantitative analysis by XRO. A nuatoer of techniques have been developed for
reducing preferred orientation effects In "thick layer" saaples [6, 13).
However, for "thin" saaples on aesbrane filters, the preferred orientation
effects see* to be both reproducible and favorable to enhancement of the
principal diagnostic reflections of asbestos Minerals, actually Increasing the
overall sensitivity of the sethod [11, IS). (Further investigation Into
preferred orientation effects in both thin ayer and bulk saaples and the
utility of a saaple spinner in afniaizing these effects 1s required.)
3.5. Lack of Suitably Characterized Standard Materials: The problea of
obtaining and characterizing suitable reference aaterials for asbestos analysis
is clearly recognized. NIOSH has recently directed a sajor research effort
toward preparation and characterization of analytical reference aaterials,
Including asbestos standards (14); however, these are not available In large
quantities for routine analysis.
8
I , I
-------
In addition, th« problea of ensuring comparability of standard reference
and saapie aaterials, particularly regarding crystallite size, particle size
distribution, and degree of crystal 11n1tys has yet to be adequately addressed.
For c*aaple, Langer at al. {16] have observed that In Insulating aatrlces,
chrysotlle tends to break open Into bundles aore frequently than aaphlboles.
This results In a 1Ine-broadening effect with a resultant decrease In sensi-
tivity. Unless this effect Is the saae for both standard and saapie aaterials,
the aaount of chrysotlle In the saapie will be underestimated by XRD analysis.
To ainialze this problea, it is essential that standardized matrix reduction
procedures be used for both saapie and standard aaterlals.
4. Precision and Accuracy
4.1. Precision of the aethod has not been determined.
4.2. Accuracy of the aethod has not been deteralned.
5. Apparatus
5.1. Saapie preparation apparatus requirements will depend upon the
saapie type under consideration and the kind of XRD analysis to be performed.
5.1.1 Mortar and Pestle: Agate or porcelain.
5.1.2. Saapie Hill: SPEX, Inc., freezer alii, or equivalent.
5.1.3. Bulk Saapie Holders
5.1.4. Silver Heabrane Filters: 25-ea dlaaeter, 0.45-vjh pore size.
Selas Corp. of Aaerlca, Flotronlcs Oiv., 1957 Pioneer Road, Huntington Valley,
PA 19006.
5.1.5. Microscope Slides
5.1.6. Vacuua Filtration Apparatus: Gelaan Mo. 1107 or equivalent,
and siae-ara vacuua flash.
5.1.7. Mlcrobalance
M.S. Ultrasonic Bath or Probe: Model W140, Ultrasonics. Inc.,
operated at a power density of approxieately 0.1 W/nt, or equivalent.
5.1.9. Volumetric Flasks- 1*1 volume.
9
-------
5.1.10. Assorted Pipettes
5.1.11. Pipette Bulb
5.112. Honserrated Forceps
5.1.13. Polyethylene Wash Bottle
5.1.14. Pyrex Beakers: 50-aL woli
5.1.15. Oeslccator
5.1.16. Filter Storage Cassettes
5.1.17. Magnetic Stirring Plate and Bars
5.1.18. Porcelain Crucibles
5.1.19. Muffle Furnace
5.2. X-Ray Olffractlon Unit, equipped with:
5.2.1. Constant Potential Generator; Voltage and aa Stabilizers
5.2.2. Autoaated Plffractoaeter with Step-Scanning Mode
5.2.3. Copper Target X-Ray Tube: High Intensity, fine focus,
preferably.
5.2.4. X-Ray Pulse Height Selector
5.2.5. X-Ray Detector (with high voltage power supply): Scintilla-
tion or proportional counter.
5.2.6. FocusIttfl Graphite Crystal Monochroaator; or Nickel Filter
(If Cu source is used, and iron fluorescence is not a serious problea).
5.2.7. Data Output Accessorles:
5.2.7.1. Strip Chart Recorder
5.2.7.2. Decade Scaler/Tlaer
5.2.7.3. Oigltal Printer
5.2.8. Sample Spinner (optional).
5.2.9. Instruaent Calibration Reference Sptclaen: o-quart? refer-
ence crystal (Arkansas quartz standard, Phillips) or equivalent.
6. Reagents
6.1. Standard Reference Materials: The reference eMorlals listed below
10
-------
are Intended to serve as a guide. Every attempt should be aade to acquire
pur* reference Materials that are comparable to saaple MVirials being ana'yz*4.
6.1.1. Chrvsotlle: UICC Canadian, or NIEHS PUstlbest. (UICC
reference Materials available fro®: UICC, MRC Pneumoconiosis Unit, Llandough
Hospital, Penarth, Glamorgan, CF61XW, UK).
6.1.2. Crocldollte: UICC; NIEHS (Or. Jack Noore), Research Triangle
Park, NC.
6.1.3. Aaosite; UICC; NIEHS (Or. Jack Moore), Research Triangle
Park, NC.
6.1.4. Anthophylllte: UICC
6.1.5. Troaollte Asbestos: Wards Natural Science Establlstwent,
Rochester, N.Y.; Cyprus Research Standard, Cyprus Research, 2435 Military
Ave., Los Angeles, California 90064 (washed with dilute NCI to reaove sull
acount of calclte inpurlty); India treoollte, Rajasthan State, India.
6.1.6. Actlnollte Asbe«tos
6.2. Adhesive: Tape, petrgteua Jelly, etc. (for attaching silver aea-
brane filters to holders).
6.3. Surfactant: 1 percent aerosol OT aqueous solution or equivalent.
6.4. Isopropanol: ACS Reagent Grade.
7. Procedure
7.1. SawplInq: Saaples for analysis of asbestos content shall be
collected as specified in EPA Guidance Oocunent 9C0090, Asbestos-Containing
Materials in School Buildings (7).
7.2. Analysis: All saaples shall be analysed initially for asbestos
content by PLM. XRO shall be used as an auxiliary wvthod when a second,
independent analysis is requested.
Note: Asbestos is a toxic substance. All handling of dry aateriaU
should be performed in an operating fu»e hood.
7.2.1 Saaple Preparation: The Method of $«sple preparation required
for HRO analysis will depend on: (1) the condition of the sa*s>le received
(sample size, homogeneity, particle size distribution, and overall ccr^osHion.
II
i
-------
as determined by PLM); and (2) the type of XRO analysis to be performed
(qualitative, quantitative; thin layer or bulk).
Bulk saterials are usually received as inhooogeneous Mixtures of coaplex
coaposition with very large partic'* size distributions. Preparation of a
homogeneous, representative saaplc froa asbestos-containing Materials is
particularly difficult because the fibrous nature of the asbestos Minerals
inhibits Mechanical aixing and stirring, and because »'11ing procedures aay
cause adverse lattice alterations.
Coop1etc nethods of saaple preparation are detailed in the appropriate
analytical sections. A discussion of specific Matrix reduction procedures is
given below.
7.2.1.1. Hilling: Mechanical Billing of asbestos Materials has
been shown to decrease fiber crystallinity, with a resultant decrease in
diffraction intensity of the speciMen; the degree of lattice alteration Is
related to the duration and type of Milling process. Therefore, all Milling
tiMes should be kept to a Minimal.
For qualitative analysis, particle size is not. in general, of critical
importance, and initial characterization of the Material with a ainiaua of
Matrix reduction is often desirable to document the coaposition of the staple
as received. Bulk saaples of very large particle size (>2-3 an) should be
cooainuted to <100 pa by careful grinding of all or a substantial portion of
the original Material in a aortar and pestle or other suitable aill (e.g., a
aicrohaaaer Mill or equivalent). When using a aortar and pestle for grinding,
the saaple should be aoistened with ethanol, or soae other suitable wetting
agent, to ainiaize exposures.
For accurate, reproducible Quantitative analysis, the particle size of
both sanple and standard Materials should be reduced to 1 to 10 pa (Section 3.3).
Ory ball Milling at liquid nitrogen teaperatures (e.g.. Spex Freezer Mill, or
equivalent) for a aaxiaua tiae of -10 ain should be used to cbtain satis-
factory particle size distributions while protecting the integrity of the
crystal lattice (4). Bulk tMplet of very large particle size aay require
grinding in two stages for full Matrix reduction to <10 m* (6.14).
Final particle size distributions should always be verified by
optical Microscopy or other suitable Method.
12
-------
7.2.1.2 Low Teaperature ashing: For Materials shown by PLN to
contain large Mounts of cellclosic or other organic aaterials, it aay be
desirable to ash the samples prior to analysis to reduce background interference
(see Section 7.2.2 of the PLN Method).
7.2.1.3. Reaoval of Carftonate Interferences: Because of the
interference caused by soae carbonates in the detection of asbestifonn ninerats
by XRO (Section 3.1.9). it aay be necessary to reaove these interferenttf by a
siaple acid leaching procedure prior to analysis (Section 7.2.2 of the PLN
Method).
7.2.1.4. AH saaples should be exaoined aicroscoplcally before
and after each aatrix reduction step to aonitor changes In saaple particle
size, coaposition, and crystal Unity, and to ensure saaple representativeness
and hoo&gev>eity for analysis.
7.2.2. Qualitative Analysis
7.2.2.1. Initial Screening of Bulk Materia)
The bulk aaterial received aay be either a "total" saaple or a
"single layer" saaple. In either instance, initial qualitative analysis
should be performed on a representative, hoaogeneous portion of U* saaple
with a ainiaua of saaple treataent.
1. Grind and aix the saaple for 5 to 10 ainutes with a aortar ami
pestle (or equivalent aethod, see Section 7.2.1.1.) to a final
particle size of <100 pa.
2. Pack the saaple into a standard bulk saaple holder. Care
should be taken to ensure that a representative portion of the
allied saaple is selected for analysis. Particular attention
should be paid to avoid possible size segregation of the saaple.
(Note: Use of a back-packing aathod of bulk saaple preparation
aay reduce preferred orientation effect*.)
3. Mount the saaple on the diffractoaeter and scan over the dlag-
o
nestle p«ak regions for the serpentine (7.36 A) and aaphibole
(6.3-0.5 A) ainerals (see Table 2). The X-ray diffraction
equipment should b* optiaized for intensity. A slow scanning
speed of 1° 20/ain is recoaaended for adequate resolution. Us*
of a saaple spinner is optional.
13
/> I (s*
-------
4. Subalt ail snples that exhibit diffraction peaks In the diag-
nostic regions for asbestiforci alnereis to a full qualitative
XRO scan (5°-60° 20; 1° 20/aln) to verify initi-.l peak assign-
ments and to Identify potential aatrix Interferences when
subsequent quantitative analysis Is to be performed.
5. Coapare the saap!e XRO pattern with standard reference powder
diffraction patterns (I.e., JCPOS PDF data (3) or those of
other well-characterlied refersnce materials). Principal
lattice spaclngs of asbestlfora Minerals are given In Table 2;
coaeon constituents of bulk insulation and wall materials are
listed in Table 3.
7.2.2.2. Oetectloin of Minor or Trace Constituents: Routine screening
of bulk Materials by XJtO aay fail to detect saall concentrations (<5 percent)
of asbestos. The Halts of (Detection will, in general, be iaproved if aatrix
absorption effects are ainiaized, and if the staple particle site is reduced
to the ootiaal 1 to 10 mo range, provided that the crystal lauice is not
degraded in the ail ling process. Therefore, in those instances where confiraa-
tion of the presence of an anbestifora aineral at very low levels is required,
or where a negative result (Section 7.2.2.1) is in conflict with previous PlM
results, it aay be desirable to prepare the saaple as for quantitative analysis
(Section 7.2.3) and step-scaiv over appropriate 20 ranges of selected diagnostic
peaks (Table 2). (Accurate transfer of the saaple to the silver aeabrane
filter is not necessary unless subsequent quantitative analysis is to tie
perforaed).
7.2.2.3. Identification of OiscreU Saaple Phases
In soae instances, confirmatory identification of discrete
. saaple phases (i.e., bundles f fibers) by XRO aay be rrcctisary. The following
procedure is recoaaendad.
1. It necessary, reduce staple particle size to
-------
3. Analyze according to the procedure described in Section 7.2.2.1.
7.2.3. Quantitative Analysis: The proposed aethod for quantitation of
asbestos in bulk sanples is s Modification of the NIOSH-recoaeended thin-layer
aethod for chrysotile in air (5]. (The thick-layer aethod of M. Taylor a^y be
used for semiquantitative analysis (6). However, this requires the addition of
an Internal standard, use of a specially fabricated saaple press, and relatively
large aaounts of standard reference aaterials. Additional research is required
to evaluate the coaparabi11ty of thin- and thick-layer aathods for quantitative
asbesto; analysis.)
7.2.3.1. Mil 1 and size all or a substantial representative
portion of the saaple as outlined in Section 7.2.1.1.
7.2.3.2. Ory in a 100° C oven for 2 hr; cool in a desiccator.
7.2.3.3. Weigh accurately to the nearest 0.01 ag.
7.2.3.4. Saaples shown by PlU to contain large aaounts of
cellulosic or other organic aaterials, and/or carbonates, should be subaitted
to appropriate aatrlx reduction procedures described in Sections 7.2.1.2 and
7.2.1.3. After ashing and/or acid treataent, repeat the drying and weighing
procedures described above, and deteraine the percent weight loss, L.
7.2.3.5. Quantitatively transfer an accurately weighed aaount
(50-100 ag) of the saaple to a l-l voluMtric flask with approxlaately 200 al
isopropanol to which 3 to 4 drops of surfactant have been added.
7.2.3.6. Ultrasonicate for 10 ain at a power density of approx-
iaately 0.1 W/al, to disperse the saaple aaterial.
7.2.3.7. 01 lute to voluae with isopropanol.
7.2.3.8. Place flask on aagnetic stirring plate. Stir.
7.2.3.9. Place a silver aeabrane filter on the filtration
apparatus, apply a vacuua, and attach the reservoir. Release the vacuuu and
add several silliliters of isopropanol to the reservoir. Vfgsjvously hand
shake the asbestos suspension and {mediately withdraw an aliquot froa the
center of ?he suspension so that total saaple weight, Wy, on the filter will
be approxieat«ly 0.1 ag.* Ho not adjust the volume in the pipet by <»xpe151ng
part of the suspension; if aore than th* desired aliquot is withdrawn, discard
•correct anount is ca. 1.0 «c>.
15
/. ^
-------
the aliquot and res use the procedure with a clean pipet. Transfer the aliquot
to the reservoir.. Filter rapidly under vacuum. Oo not wash reservoir walls.
Leav? filter apparatus under vacuum until dry. Resove reservoir, release
vacuus, and reaove fiHer with forceps.
7.2.3.10. Attach the filter to a flat holder with suitable
adhesive and p'iace on the diffractoaeter. Use of a sample spinner is optional.
7.2.3.11. For each asbestos aineral to be quantitated select
a reflection (or reflections) that has toon shown to be free froa interferences
by prior PlM or qualitative XRO analysis and that can be used unaabiguously as
an index of the mount of act«rial present in the saople (see Table 2).
7.1.3.12. Analyze the selected diagnostic reflections) b>
step scanning in tncreaents of 0.02° 26 for an appropriate fixed tie® and
fntegrciing the counts. (A fixed count scan aay be used alternatively;
however, the aethod chosen should be used consistently for all samples and
standards.) An appropriate scanning interval should be selected for each
peak, and background corrections aade. For a fixed tiae scan, aeasure tha
background on each side of the peak for one-half the peak-scanning tiae. The
net intensity, I., is the difference between the peak integrated count and the
total background count.
7.2.3.13. Oeteraine the not count, , of the filter 2.36 A
silver peak following the procedure in Section 7.2.3.12. Reaov* the filter
froa the holder, reverse it, and reattach ft to the holder. Oeteraine the net
count for the unattenuated silver peak, Scan tiaes aay be lets for
aeasure»ent of silv«r peaks than for saaple peaks; however, they should be
constant throughout the analysis.
7.2.3.14. Noraalize all raw, net intensities (to correct for
instruaent instabilities) by referencing thea to an external standard (e.g.,
e
the 3.34 A peak of an o-quartz refer*nee crystal). After each unknown is
o
scanned, deterafne the net count, Ir, of the reference speciaen following the
procedure in Section 7.2.3.12. The normalized intensities are determined by
O
dividing the peak intensities by Ir:
I - I 1°
t = *5 I = .nd to -
a Y Ag | • ana Ag °
r r r
16
b '/
-------
8. Calibration
8.1. Preparation of Asbestos Standards
6.1.1. Wf, 11 and size standard asbestos Materials according to the
procedure outlined In Section 7.2.1.1. It Is essential that equivalent,
standardized Matrix reduction and sizing techniques be used for both standard
and saaple Materials.
8.1.2. Dry In a 100° C oven for 2 hr; cool 1n a desiccator.
8.1.3. Prepare two suspensions of each standard in isopropanol by
weighing approxiaately 10 and SO «g of the dry Material to the nearest 0.01 ag.
Quantitatively transfer each to a 1-L volumtric flask with approxlaately
200 al isopropanol to which a few drops of surfactant have been added.
6.1.4. Ultrasonicate (at 5 W power, or equivalent) for 10 win to
disperse the asbestos Material.
8.1.5. 01 lute to voluMe with Isopropanol.
8.1.6. Place flask on Magnetic stirring plate. Stir.
8.1.7. Prepare, in triplicate, a series of at least five standard
filters to cover the desired analytical range, using appropriate aliquots of
the 10 and SO ag/L suspensions.
Mount a filter on the filtration apparatus. Place a few Milliliters
of isopropanol in the reservoir. Vigorously hand shake the asbestos suspen-
sion and iMMediately withdraw an aliquot froa the center of the suspension.
Oo not adjust the voluae in th$ pipet by expelling part of the suspension; If
aore than tn« desired aliquot is withdrawn, discard the aliquot and resuae the
procedure with .i clean pipet. Transfer the ati^sn to reservoir. Koep
the tip of the pipet near the surface of the isopropanol. Filter rapidly
under vacuus. Oo not wash down the sides of the reservoir. Leave the vacuum
on for a tine sufficient to dry the filter. Release the vacuum and remove the
fiIter with forceps.
8.1.8. Ntjunt the filter on a Tlat holder. Perfors step scans on
selected diagnostic reflections of the standards and reference specimen using
the procedures outlined
-------
8.1.9. Determine f.he normalized Intensity for each peak measured,
A
I®, as outlined in Section 7.2.3.14.
9. Calculations
9.1 For tact) asbestos reference aaterial, calculate the exact weight,
deposited on each standard filter froa the concentrations of the standard
suspensions and aliquot volumes. Record the weight, w, of each standard.
A
Prepare a calibration curve by regressing I® on w. Poor reproducibility (±15
percent RSO) at any given level indicates problems in the saaple preparation
technique, and a need for new standards. The data should fit a straight line
equation.
9.2. Oeteralne the slope, a, of the calibration curve in counts/micro-
A
graa. The Intercept, b, of the 11n« with the I£ axis should be approximately
0. A large negative Intercept indicates an error In determining the back-
ground. This may arise from Incorrectly measuring the baseline or froa
interference by another phase at the angle of background measurement. A large
positive intercept indicates an error in determining the baseline or that an
{¦purity is included in the aeasured peak.
A
9.3. Using the normalized 'ntenslty, 1^ , for the attenuated silver peak
of a saaple, and the corresponding noraallzed intensity froa the unattenu&ted
silver peak, 1^, of the saaple filter, calculate the transaittanc«, T, for
each saaple as follows (17, 18):
T «
1°
1 Ag
9.4. Determine the correction factor, f(T), for s.«ch sample wsrdi*?#
to th« formula:
f(T) = T
l-i
where
R
sin ©
a
18
-------
= angular position of the aeasured silver peak (froa Bragg's Law), and
6 = angular position of the diagnostic asbestos peak.
9.5. Calculate the weight, in aicrogr«as, of the asbestos aaterial
analyzed for in each staple, using the appropriate calibration data and absorp-
tion corrections:
ia " b
a ~ - f(T)
9.6 Calculate the percent coaposition, P , of each asbestos aineral
analyzed for in the parent aaterial, froa the total saaple weight, W^, on
the filter:
W. (1-.01L)
Pa " x 100
where
P - percent asbestos aineral in parent aa'^rial;
W = aass of asbestos aineral on filter, in mQ;
Wy = total saaple weight on filter, in pg;
L = percent weight loss cf parent aaterial on ashing
and/or acid treatoent (Section 7.2.3.1).
19
-------
10. References
1. Klug, H. P., and L. E. Alexander. 1979. X-ray 01ffraci1on Procedures
2nd ed. New York: John WiUy and Sons.
2. Azaroff, L. V., and M. J. Buerger. 1958. The Powder Method of X-ray
Crystallography. New York: McGraw-Hill.
3. JCPOS-International Center for Diffraction Data Powder Olffraction
File. Swarthmore, PA.
4. Lange, B. A., and J. C. Haartz. 1979. Determination of Microgram
Quantities of Asbestos by X-ray Olffraction: Chrysotlle In Thin Oust
Layers of Matrix Material. Anal. Chea. .51(4):520-525.
5. NIOSH Manual of Analytical Methods, Vol use 5, U.S. Oept. HEW, August 1979,
pp. 309-1 to 309-9.
6. Taylor, M. 1978. Methods for the Quantitative Determination of
Asbestos and Quartz in Bulk Samples Using X-ray Olffraction. The
Analyst. 103(1231): 1009-10?C.
7. Asbestos-Containing Materials in School Buildings: A Guidance Document.
EPA/0TS #C00090, 1979.
8. Krause, J. B., and W. H. Ashton. Mlsldentificatlon of Asbestos In
Talc, pp. 339-353. Proceedings of Workshop on Asbestos: Definitions
and Measurement Methods, (NBS Special Publication 506). 1977. (Issued
1978). Edited by C. C. Gravatt, P. 0. LaFlour, and K. F. Helnrich.
National Measurement Laboratory, National bureau of Standards, Washington,
0. C.
9. Birks, L., M. Fatemi, J. V. Gilfrich, and E. T. Johnson. 1975.
Quantitative Analysis of Airborne Asbestos by X-ray Oiffraction.
Naval Research Laboratory Report 7879. Naval Research Laboratory,
Washington, O.C.
10. Stanley, H. 0. The Oetection and Identification of Asbestos and
Asbestlform Minerals in Talc, pp. 325-337. Proceedings of Workshop on
Asbestos: Definitions and Measurement Methods, (NBS Special Publication
506). 1977. (Issued 1978). Edited by C. C. Gravatt, .P. 0. LaFleur,
and K. F. Helnrich. National Measurement Laboratory, National Bureau
of Standards, Washington, D. C.
11. Rlckards, A. L. 1972. Estimation of Trace Amounts of Chrysotile
Asbestos by X-ray 0'ffractlon. Anal. Chtm. 44(11):1872-3.
12. Cook, P. M., P. L. Smith, and 0. G. Wilson. A«phibole Fiber Concentra-
tion. Determination for a Series of Community Air Samples: Use of X-
Ray Diffraction to Supplement Electron Microscope Analysis. Electron
Microscopy and X-Ray Applications to Environmental and Occupation
Health Analysis. 1977. Edited by P. A. Russell and A. E. Hutchtngs.
Ann Arbor Science Publications. Ann Arbor, Michigan.
20
I *, •,
-------
13. Rohl, A. N., and A. M. Langer. 1929. Identification And Quantitation
of Asbestos in Talc. Environ. HeaHh Perspectives. 9:95-109.
14. Graf, J. L., P. K. Ase, and R. G. Oraftz. Preparation and Characteriza-
tion of Analytical Reference Minerals. OHEW (NIOSH) Publication
Mo. 79-139. June 1979.
15. Haartz, J. C., B. A. Lange, R. G. Oraftz, and R. F. Scholl. Selection
and Characterization of Fibrous and Nonfibrous Aaphlboles for Analytical
Methods Development, pp. 295-312. Proceedings of Workshop on Asbestos:
Definitions and Measurement Methods, (MBS Special Publication 506). 1977.
(Issued 1978). Edited by C. C. Gravatt, P. 0. LaFleur, and K. F. Heinrich.
National Measurement Laboratory, National Bureau of Standards, Washington,
O.C.
16. Langer, A. M. Environmental Sciences Laboratory. Mount Sinai School
of Medicine of the City University of New York. New York, New York.
Private coamunication.
17. Leroux, J. 1969. Staub-Reinhalt Luft, 29:26 (English).
18. Leroux, J. A., B. C. Oavey, and A. Paillard. 1973. Am. Ird. Hyg. Assoc.
J. , 34:409.
Ack'iowledgaent
This method Is based In part on contributions review comments of
participants in the syaposlum "Methods definition for the polarized light
Microscope and *-ray diffraction analysis of bulk saat)les for asbestos,"
U.S. Oureau of Mines, Avondale Research Center, Avondale, Maryland, October
23-24, 1979.
21
-------
APPENOIX C
participating laboratories
C-l
£k
-------
Mr. Reg Jordan
Northrop Services, Inc.
Post Office Box 12312
Research Triangle Park, North Carolina 27709
Nr. R. J. Kuryvial
Petrography Consultant
31720 Hilltop Road
Golden, Colarado 80401
Nr. James Schirripa
Industrial Hygienics, Inc.
755 New York Avenue
Huntington, New York 11743
Nr. Douglas Allenson
Herron Testing laboratories
5405 lichaaf Road
Cleveland, Ohio 44131
Nr. Stuart Salot
Certified Testing Laboratories. Inc.
2905 East Century Boulevard
South Gate. California 90280
Nr. Jams ttebber
Environmental Health Center
Division of Laboratories and Research
New York Oepartmer* of Health
Tower building, Eapire State Plaza
Albany, New York 12201
Nr. Bernard Erlin
Erlfn, His* Associates
811 Skokie Boulevard
Morthbrootr, Illinois 60062
Nr. Rolwrt Reinhart
LFE Corporation
Environmental Analysis Lab Oivision
2030 Wright Avenue
Richmond. California 94804
u. i... ia . i .
n>. nun my 9 me
Maryland Mineral Analysis laboratory
Post Office 80* V
CoMege Park. Maryland 20740
Pr. Robert N. Ginsberg
teoscience Consultants, Inc.
Post Office Bo* 341366
Coral Gables. Florida 33134
C-?
Nr. Jerome Krause
Colorado School of Nines
Mineralogy Projects Exploration
I Nining Oivision
Post Office Box 112
Golden, Colorado 80401
Nr. Ronald 0"aft*
I IT Research Institute
10 West 35th Street
Chicago, Illinois 60616
Nr. Jia Snarr
Utah Biomedical Test Laboratory
520 Wakara Way
Salt Lake City. Utah 84108
Nr. Nark Palenik
Walter McCrone Associates, Inc
Electron Optics Group
2820 S. Michigan Avenue
Chicago, Illinois 60616
Nr. Gary C. Allen
Sunbelt Associates, Inc.
6961 Hayo Road
New Orleans. Louisiana 70126
Nr. Martin Rutstein
State University of New York
Department of Geological Sciences
Col lege at New Palt;
Mew Pall/. New »ork 12562
Nr. N. A. 8eg
American Nicroscopy Laboratory
0. 3410 12th Ave. East
Tuscaloosa. Alabama 35405
Mr. Charles Muggins
U.S. Bureau of Mines
4900 LaSalle Road
Avondale. Maryland 20782
Or. Eric Steel
U.S. National Bureau of Standards
Building 222, Roofl A121
Washington. O.C. 20234
Hr. Gay lord Atkinson
Midwest Research Institute
425 Volkor Blvd.
Kansas City. Missouri 64110
-------
Hr. Mike Boucher
MJH Associates
1334$ Foliage Avenue
Apple Valley, Minnesota 55124
Hr. G«n« Brant Iy
Research Triangle Institute
P.O. So* 12194
Research Triangle Park,
North Carolina 27709
C-3
-------
APPENDIX 0
INSTRUCTIONS A*0 REPORTPiG FOSMS
0-5
-------
R F 3 tA 9 C H TRIANGLE INSTITUTE
R\
' « t f o r ' ¦ c • « o • i I < » «
March 28, 3980
Mr. Carl Mellon
Battelle ColtiMbus Laboictory
SOS Kmjt Ave.
Coluobus, Ohio 43201
Dear Mr. Melton:
Your liboritory hit indicated 4 willingness to test the tentative EPA
Methods (or identification jno quant 1 tat ion of asbestos in bulk Materials.
Enclosed please find a third draft of the proposed Methods and a selection of
sa»ples chosen to characterize their accuracy and precision. These Methods
¦ust be followed carefully for your results to provide a Meaningful reflection
of the potential of the Methods. Vote in detail any deviation in your appli-
cation of the Method. Should you *0 desire, coapare the results obtained by
this Method with those obtained using your laboratory's standari procedures.
Ve will welco«e any coments on the procedures. Testing mist be performed and
reported (received) no later than April 21, 1980, to be useful in this draft's
evaluation. EPA has authorized a standard payment of $20 per saaple for the
polarized light Microscope results and $40 per saaple for X-ray diffraction
results received by that date.
Should there be any question in this Matter or should you find your
laboratory unable to respond as requested, please contact me or Gene Brantly
immediately at 1 -800-J34-8S71, ext. 6745.
Thank you for your continued interest in the Methods developMent progrj*.
Sincerely yours,
0. F.. Lentzen, Ph.D.
Environmental Scientist
DF.l/lb
-------
TENTATIVE y£?HCO FCP THE OETCRMIKATICN OF ASP£ST!FCPr HINEPALS
!fl BJir. SAM>LES PT POLAR!ZED L.'GhT MICROSCOPY
Y<» 1idat ion Study
Laboratory: PT! Sanple t.
Analyst:
Analytical method
Gross sample appearance-
SubsarpImg, na'.ri* reduction or sample preparation steps
FIBfR ICCKTiriCATIC*
Are fibers present? YES NO
For ail fiber types, corplete the following
TYPE 1 TYPE 2
Extinction characteristics:
Sign of elongation:
Refractive indices:
Fiber rorphology. color: "* *
Fiber identity: " ~
cuaktitatiow
Magnification used:
•lunber of ron-e*pty points counted (n) ¦ ACO
Points counted for asbestos fibers (a) »
Relative area occupied by asbestos (p»a/n) ¦
q • 1-p "
95°. confidence intervals (p + 1.96 po/n) •
Reported percent asbestos: ___
cownts
°evie*fed by
-------
INSTRUCTIONS
ANALYSIS OF BULK MATERIALS BY X-RAY POWER DIFFRACTION
1. Validation of the Quantitative XRO aethod Is an Important objective of
this study. We recognize, however, that lack of suitable reference
aaterials may present a problea to soae laboratories. For the purposes
of this study, plaase proceed with the quantitation using the standard
Materials available and report all raw data, along with the final results,
as requested on the report fora.
It should be eaphaj1zed that this study 1s Intended to validate the
XRO aethod Independent of the PLM aethod. If prior analysis by PLM has
been done, 3o not let these results Influence the outcoae of the XRO
analysis.
2. Complete one data fora for each saaple analysed. If aultipie aeasureoents
are aade on a single sample, report Individual results on separate data
form; do not average results.
3. All deviations froa the proposed aetnod should be noted where appropriate.
4. Details of aatrlx reduction stops should Include equipnent specifications
(e.g., type of grinding aill) wtd the length of tiae for each procedure
where appropriate.
5. All calibration curves, dlffractograas, and Intensity data Must be
included, as requested, for accurate assessaent of reported results.
6. Please Include any coaaents you aay have om this aethod. In particular,
coapare with yoar standard XRO atthod, note specific problea areas, and
detail recoaaendatlons for iaproveaent. If you so iesire, analyze the
saaples by your standard XRO rethod (report results on the data fonts
provided, noting deviation froa proposed aethod), and conpare your results
with those obtained by the proposed aethou.
-------
I*fUY POWCfH OIFfMCTIM AMLTSIS MPO«T
laboratory _____________________________
I. SMVIE I0CNTIFICAT10N
Staple lafcel or $ at tecetveJ:
ItteriUry • MtlgnM:
Cross SmqU *de«iriAC«:
II. ElPEIIKMTAl
HatrU tort^ctlon or Saaole Preparation:
Particle S1Hng: Nttitetf
final Particle S1*e Distribution (range, Md) na ______
Mounting Media:
Typ«(s) of AAalytiS: frgntltatWe ___SestQwanl1tat1v« Qw«11
SmoU SplMier: yet no
[«Ura|l MUrm* SU"<«f« (t»»* toitrf, rtf«r««KI ptlk N):
Calibration Standard Material(») (asbestos type, source, eiatrl* reduction,
final particle site):
III. KSUlTS
Please Attach copies of:
1. Calibration curves. including regression equations. and
2. All dlffractoqrees. appropriately labeled Mv>it be accurately i«oe*ed (29). with
exact scanning Intervals note*.
Asbestos Present: ires *o Other Materials Present and * ______
ConftraeJ >y PlM: ye» He
r'
Asbestos Identification
Major Peaks (20)
Olagnosttc peak (26)
Integrated peak area, i,
(cos) *
teference peak area, I
(CPS)
Absorption correction,
f(T)
Sanple weight, M.. (yg)
X ut loss en asMng or
acid wash. I
Mass of asbestos on
filter. va, (ug)
t Asbestos. Pa
SsOwated error
IXBlA !aii
Major Peaks unaccounted for.
Analyst:
Cements:
0-5
-------
APPENDIX E
CONFERENCE PARTICIPANTS
E-l
-------
PARTICIPANTS IN RESEARCH TRIANCLE INSTITUTE SPONSORED
SYMPOSIUM ON 'METHODS DEFINITION FOR THE POLARIZED LIGHT MICROSCOPE
AND X-RAY DIFFRACTION ANALYSIS OF SULK SAMPLES FOP ASBESTOS"
HELD AT THE U.S. BUREAU OF MINES, AVONDALE RESEARCH
CENTER, AVONDALE. MARYLAND
OCTOBER 23-24. 1979
Dr. Gary Allen
1'nlversity of New Orleans
Dept. of Earth Sclonces
New Orleans, U 70122
< 5CWi) 283-0325
Dr. Charles Anderson
Athens Environnencal Research Lib.
U.S. EPA
College Station Road
Athens, Georgia 30605
(404) 346-1183
Michael Beard
U.S. EPA MD-77
Environmental Monitoring System
Laboratory
Research Triangle Park, N.C. 27711
(919) 541-2623
Janes Bergen
Kta Shedd
Robert Vlrta
U.S. Bureau of Hlnes
4900 LaSalle Road
Avoodale, Maryland 20782
(301) 436-7541
Michael Boucher
U.S. Bureau of Mlncn
P.O. 1660
Twin Cities, Minnesota 55111
(612) 725-4614
Dr. Jos^pn Srecn
U.S EPA OTS-TS '93
E/jSt Tover Waters Uc Mall
'.01 M Street S. V.
Washington, D. C. 20460
(202) 755-8060
Larry Broun
I'. S. Bureau of Mines
P. 0. Box 70
Albany, Oregon 97321
(503) 967-5843
FTS 420-5043
Dr. Willlas J. Caapbell
U.S. Bureau of Mines
4900 LaSail* Road
Avondale, Maryland 20782
(301) 436-7501
Ulllard Dixon
U. S. Dept. of Labor 0SHA
390 Uabara Way
Salt Lake City, Utah 84108
(.101) 524-5366
Lurry Dorsty
U.S. EPA OTS-TS 793
East Tower Waterside Mall
401 M Street S. W.
Washington, D. C. 20460
(202) 755-8060
Mr. Ronald Draft*
IITRI
10 West 35th Street
Chicago, Illinois 60616
(312) 567-4291
Dr. Janet Haartz
N10SH
Division of Engineering and
Physical Science
4o76 Colunbia Parkway
Cincinnati, Ohio 45226
(513) 684-4323
Z-2
-------
PARTICIPANTS (Continued)
Charles Hugglns
U.S. Bureau of Mines
4900 LaSalle Road
Avondale, Maryland 20782
(301) 436-7565
Lois Jacob
U.S. EPA
Office of Enforcement (EN 362)
401 m Street S.U.
Washington, D.C. 20460
(202) 755-9404
flruce A. Lange
U.R. Crace & Coapany
62 Whltleaore Avenue
Cambridge, Massachusetts 02140
(617) 876-1400, ext. 155
Or. Stephen L. Law
U.S. Bureau of Mines
4900 LaSalle Road
Avondale, Maryland 20782
(301) 436-7542
Dr. Donald Lentsen
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, N.C. 27711
1-800-334-8571, ext. 6745
Donna Lucas
Rose«-ch Triangle Institute
Research Triangle Park, N. C. 2771 •
(919) 541-6633
Dr. Ualtei C. McCrone
2820 South Michigan Avenue
Chicago, Illinois 60616
(312) 842-7100
Carl ftclton
Battelle Colunbus
505 King Avctiue
Colunbus, Ohio 43201
(614) 424-5305
Eric Sta*l
National Bureau of Standards
Bldg. 222, R*. A121
Washington, D.C. 20234
(301) 921-2875
Cindy Stroup
U.S. EPA OTS-TS 793
East Tower Waterside Mall
401 M Street S.W.
Washington, D.C. 20460
(202) 755-8060
Steve Wllllaas
Research Triangle Institute
Research Triangle Park, SiC. 27711
(919) 541-6146
John Wilson
Office of Toxic Substances (TS 794
U.S. EPA
401 M Street S.W.
Washington, D.C. 20460
Dr. Ann Wylle
University of Maryland
Departaent of Geology
College Park, Maryland 20742
(301) 454-3548
Yasl Yaffe
Lt'C Envlronoental
2030 Wright Avenue
Richaond, California 94P04
(415) 235-2633
E-3
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO. 2.
EPA-600/4-82-021 ORD Report
3
4. TITLE AND SUBTITLE
BULK SAMPLE ANALYSIS FOR ASBESTOS CONTENT: EVALUATION
OF THE TENTATIVE METHOD
B. REPORT DATE
March 1982 -
6. PERFORMING ORGANIZATION CODE
43U-1808-50
7. AUTHOR(S)
E.P. Brantly, Jr., K.W. Gold, L.E. Myers, and
D.E. Lentzen
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, NC 27709
10. PROGRAM FLEMENT NO.
11. CONTRACT/GRANT NO.
68-02-3222
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Monitoring Systems Laboratory
U.S. EPA, Research Triangle Park, NC 27711
Office of Pesticides and Toxic Substances
U.S. EPA, Washington, D.C. 20460
13. TYPE OF REPORT AND PERIOD COVERED
Finalt March-December 1980
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The U.S. Environmental Protection Agency Asbestos-in-Schools Program was estab-
lished in March, 1979 to provide information and technical assistance to the public
for addressing problems presented by asbestos-containing insulation materials in
school buildings. Because there were no existing standard procedures for the quali-
tative and quantitative analysis of asbestos in bulk materials, the Office of Pesti-
cides and Toxic Substances, Washington, DC, and the Environmental Monitoring Systems
Laboratory, Research Triangle Park, NC, jointly sponsored an effort to produce a
practical and objective analytical protocol. Draft procedures were written for the
analysis of bulk samples by polarized li^ht microscopy (PLM) and X-Ray powder dif-
fraction (XRD). Following review, the Tentative Method for the Determination of
Asbestiform Minerals in Bulk Insulation Samples (March, 1980) was submitted to a
performance testing program that involved multiple laboratory analysis of prepared
samples with known asbestos content. This report presents the results of the testing
study and provides observations and preliminary characterization of the utility and
operational parameters of the Tentative Method.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b. IDFNTIF IE RS/OPEN ENDED TERMS
e. COSATl Field/Group
18. DISTRIBUTION STATEMENT
Release to P 'blic
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
135
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Irorm 2220-1 (Rev. 4-77) previous edition is obsolete
t
\
------- |