Tobacco Documents Online

CS National Bureau of Standards Special Publication 506. Proceedings of the Workshop on Asbestos: Definitions and Measurement Methods held at NBS, Gaithersburg, MD, July 18-20, 1977. (Issued November 1978) IDENTIFICATION AND COUNTING OF MINERAL FRAGMENTS R. J. Lee, J. 5. Lally, and R. M. Fisher U. S. Steel Corp., Research Laboratory Monroeville, Pennsylvania 15146 Abstract Positive identification of submicrometer-diameter mineral fragments, especially amphiboles, requires both chemical and crystallographic analysis. At present, only electron optical methods can be used for this purpose, and considerable care must be taken to ensure that (1) the x-ray spectra and diffraction patterns pertain only to the particle in question (that is, spatial-resolution limitations must be recognized); (2) x-ray data are compared with well characterized reference standards; (3) overlapping chemical composition and/or similar crystal structures of mineral series are recognized; (4) crystal fragments are tilted into zone-axis orientation before recording the electron-diffraction pattern; and (5) appropriate statistical criteria are used to evaluate the significance of the results. The required procedures are time consuming (and costly), but less rigorous methods are subject to considerable uncertainty, which limits the validity of the data and its usefulness in any assessment of biological effects. Adoption of a definition of mineral fibers based on an aspect ratio of 10:1 and parallel edges would eliminate most non- asbestos mineral fragments from consideration, and reduce the analytical problems to more manageable proportions. Analysis of the face orientations of amosite fibers (commercial amphibole asbestos) and grunerite fragments (nonasbestiform amphibole) reveals pronounced distinctions which originate in their different crystal growth or cleavage characteristics. Key Words: Amosite; amphibole; asbestos; electron diffraction; fibrous; grunerite; mineral identification; non-fibrous. Introduction The organization in 1977 of a workshop devoted to identifying points of agreement and disagreement on definitions and measurement methods for asbestos was a most welcome and logical initiative on the part of the National Bureau of Standards (NBS) and the Occupa- tional Safety and Health Administration (OSHA). Changes in meaning of the term "asbestos fiber," which have occurred with the advent of concern about very fine particles (only observable in the electron microscope), have been discussed and deplored by Tibor Zoltai [1,2]1. Such loosening of the definition of asbestos results in the inclusion of varieties of sheet silicates, chain silicates, and even non-silicates. Malcolm Ross [3] has pointed out that serpentine, amphibole, clay, mica, chlorite, and alumina-silicates are prime exampies of widely occurring minerals that could be erroneously classified with "asbestos." As a further consequence, the term "emission sources" becomes broadened to include extended, naturally occurring geological formations. - 'Figures in brackets indicate the literature references at the end of this paper. Three digit bracketed numbers, e.g., [113] refer to reciprocol space vectors. 387

The use of vague terminology coupled with limited biological research data has blurred the distinction between scientific fact and speculation regarding the health hazard resulting from exposure to low concentrations of silicate dust particles. The correct identification of micrometer-size mineral particles and the accurate measurement of their concentration in air or water samples is not easy. Unfortunately, reported observations of mineral varieties in samples collected at locations where these minerals should not occur, coupled with differences of many orders of magnitude in the particulate concentrations reported by various laboratories, confuse those with the political or administrative responsibility for reacting to public concern about environmental quality. Hopefully, publication and distribution of the proceedings of the NBS/OSHA workshop will help repair the damage to the credibility of analysts and their procedures that public controversy has engendered. United States Steel has expended a great deal of effort in developing reliable methods for identifying and counting particles. The details have been published elsewhere [4,5]. The purpose of the present report is to emphasize the precautions that must be taken to avoid errors and illustrate, the results that can be obtained if appropriate methods are used. Fiber characteristics pertinent to the definition of "asbestos" are also discussed at the end of this report. Particle Identification It is necessary to use electron microscopy, either scanning (SEM) or transmission (TEM), to examine particles with dimensions of a few micrometers or less. Transmission electron microscopes can be adjusted to obtain electron-diffraction (ED) patterns from single particles which provide cyrstallographic information. Scanning electron microscopes are usually equipped with facilities for x-ray emission spectroscopy which provide elemental information about single particles (if properly dispersed). Some hybrid instruments combine SEM, TEM, ED, and x-ray functions. Characteristic Diffraction Patterns. The term "characteristic pattern" has come into common use in connection with the identification of particles of minerals by transmission electron microscopy as if such characteristics (diagnostic patterns) could be ascribed to each mineral. As a general rule this is incorrect. Many minerals have very similar dif- fraction patterns, as illustrated in figures 1 and 2, and it is impossible to distinguish between them by visual inspection. If they are measured and interpreted (indexed as in figures 1 and 2), they can be used as references for other identical (not similar) pat- Chr.p;ntle Figure 1. Selected area electron diffraction patterns of mjneral particles which all show a d-spacing of about 5 A. 388

CS Aclinolite Figure 2. Indexed electron diffraction patterns of representative amphiboles and non-amphiboles showing essentially similar appearance. terns. These are good patterns obtained by carefully tilting into a"zone axis" orienta- tion. If this is not done, the patterns are diffuse, irregular, and useless. Because of its tubular shape and helical structure, the diffraction pattern of chrysotile is not strongly dependent on orientation and is recognized more readily. Another factor affecting electron-diffraction analysis, as illustrated in Figure 2 for actinolite, is that every mineral has several different "zone axis" patterns, which depend on the orientation of the particle with respect to the electron beam. This is also illustrated in figure 3 for the amphibole mineral grunerite. However, the presence of extensive twinning, as in figure 3 for grunerite, or fine-scale exsolution in the pyroxenes (as in augite in figure 2), can give rise to closely spaced spots in a diffraction pattern. This may lead the unsuspecting microscopist to conclude that he is observing a large d-spacing when, in fact, the pattern contains "extra" spots due to reciprocal lattice spikes, multiple zone-axis orientations, or satellite spots. 0 The 5.3 A d-spacing of the c-axis is frequently taken as definitive of the amphiboles. yet, of the eight diffraction patterns shown in figures 1 and 2, six have d- spacings ti5 A, although only two of these are from amphibole particles. In general, as discussed by ,Zoltai [1], many minerals fragment into acicular particles and have d-spacings ry5 lf Because of differences in fragmentation characteristics, the broad faces of the two mineral varieties, amosite and grunerite, occur along different crystallographic planes, so the patterns are usually different. bN ~ 389 W ~ 0 ~ ~. 0 0

01b Grunerite (3jp1 Figure 3. Indexed electron diffraction patterns of different orientations of grunerite achieved by tilting specimens in the electron microscope. The characteristic growth habit of amosite asbestos gives rise to a fiber with the largest face on (100) planes, which will then lie flat on electron-microscope grids [6]. As a result, the nearest reciprocal lattice section will contain b*x[103]*, as pointed out by Nord [7]. A typical pattern from amosite, shown in figure 4(a), contains b*x[113]*, the basic vectors for zone axis perpendicular to the [100]* direction in reciprocal space. In contrast, a cleavage fragment should lie near a (110) face, and the expected zone-axis diffraction patterns nearly perpendicular to [110]* would not contain b*. While the cleavage fragments are more randomly oriented, the diffraction pa-ttern in figure 4(b), [021]*x[221]*, and the pattern containing the twin spots [130]*x[1l1]* in figure 3, are both within about 15° of [110]*. Thus, the predominant face of the particle can be determined if the pattern is indexed on the assumption that the fiber axis c in real space, [103]* in reciprocal space, is perpendicular to the beam. 390

ir © ~-- .~ . ~ , ~ ~~ ~ ~ CS Amosite f Figure 4. Electron micrographs and diffraction patterns of amosite and grunerite particles which indicate that the asbestos and non-asbestos varieties can be distinguished if the diffraction patterns are interpreted properly. Preliminary results of a study of the typical orientations of a Penge amosite, and a grunerite from Presque Isle, Michigan, are shown in table 1. In this comparison, 75 percent of the amosite particles had b* vectors within 100 of the normal to the electron beam, thus indicating they were lying on (100). Only 18 percent of the grunerite particles had this orientation. The remaining 75 percent were more than 30° away from [100]* on the [010]* side of the [110]* direction in reciprocal space. The cleavage fragments are not as tightly clustered about [110]* as the amosite particles are about [100]*. These data suggest that it may be possible to distinguish amosite fibers from grunerite cleavage fragments by looking for b* reflections. Very recent SEM observations of crystal-growth habits of amosite and grunerite have revealed that these distinctions between fiber formation by "delamination" and fragment formation by cleavage are also evident in bulk samples, figure 5. Table 1. Face orientations of Amosite and Grunerite fibers. Face Amosite Grunerite (100) 17 (74%) 7 (010) 1 2 (110)a 5 30 (77%) a Cleavage plane. 391

a b 20,0001 Figure 5. Cross-sections of grunerite and amosite looking down the c-axis. (a) Exsolution lameliae, cleavage traces and 100 parting are all evident in grunerite. (b) The delamination of thin lamellae produced by fine-scale twinning (ti50 A) along (100) is shown. This leads to the development of amosite particles with large 100 faces. X-ray Analysis. As mentioned, the development of energy-dispersive spectroscopy as an adjunct to scanning or transmission microscopy has made it possible' to obtain information about the elemental composition of small particles. It is tempting to assume that the x-ray spectra are diagnostic, but in many cases this is not true. Unless precautions are taken, near-by particles can contribute to the x-ray spectra. Absorption and fluorescence effects depend on particle size and orientation, so their x-ray spectra may differ from those of reference standards. These effects are not great; but for many minerals, differences in composition are also not large, (as shown in figure 6 with x-ray spectra from amphiboles, serpentines, and non-amphiboles). ® ® N1gAISi Ca Fe ® © ® MyS, Fe Figure 6. X-ray spectra of representative amphiboles, serpentines, and non-amphiboles showing essentially similar appearance illustrating difficulty of identifying silicate minerals by non-quantitative x-ray spectroscopy. 392 I

C?) Improved Identification Technigues. These difficulties with identification, although formidable, have been overcome with effort and better understanding of electron-diffraction and x-ray spectroscopy. Key features of the method are (1) tilting of particle into one or two zone axis orientations; (2) obtaining an x-ray spectrum from the same particle; (3) comparison of diffraction and x-ray data by computer with a complete library of all minerals that could be present. Figure 7(a,b) [4,5] shows block diagrams for the program DIFFPAT, which simulates zone-axis diffraction patterns for any crystal with known unit- cell parameters and space-group symmetry, and SEARCH, which compares measured parameters obtained from zone-axis diffraction patterns with the simulated patterns generated by DIFFPAT. The minerals included in the computer library are listed in table 2. Prior consideration of chemical information eliminates many of these minerals and reduces computer time considerably. Table 2. Silicate minerals used in programs "DIFFPAT" and "SEARCH".a PYROXENES AMPHIBOLES SERPENTINES TALCS CHLORITES Augite Actinolite Amosite Minnesotaite Clinochlore Diopside Anthophyllite Antigorite Pyrophyllite Penninite Enstatite Arfvedsonite Berthierite Talc Prochl ori te Hedenbergite Barkevikite Chamosite Talc-Chlorite Hyperstene Cumningtonite Chrysotile Jadeite Eckermanite Cronstedtite MICAS Johannsenite Edenite Greenalite CLAYS Pigeonite Ferrogedrite Lizardite Biotite Spodumene Gedrite Cl intonite Apophyllite Glaucophane Glauconite I11ite Grunerite KAOLINS Lepidollte Montmorillonite PYROXINOIDS Hastingsite Margarite Prehnite Holmquistite Dickite Muscovite Vermiculite Pectolite Hornblende Donbassite Paragonite Rhodonite Kaesutite Halloysite Phlogopite Wollastonite Katophorite Kaolinite Stilpnomelane Pargasite Nacrite Xanthophyllite Richterite Zinnwaldite Riebeckite Tremolite Tschermakite a Obtained from Ref. 1. 393

SYAT - CALCUILATEO SII. S1=... OCNLC - C4CULATFD 0'SF.LCINGS. ORROENS EY SIII. ANO SEPARATES INTO ALLOWAELE Ntq FOREICOEN VECTORS FICES A ZONE AxR FINOS TNE LARGEST ALLOWED Pp.ECING EELONGING TO IT - THEN CNECES FUR OOOELE OIFFRKnON • FINALLY CqMRFS THE WOLF SETNEEN THE VECTONS TO SIWILATF THE DIFIRACTNy/ NLTFERN Figure 7. (a) Summary flow chart for computer program DIFFPAT which calculates reciprocal lattice diffraction patterns. Enter D , ARaM Dau From DIFFPAT Figure 7. (b) Block diagram of the program SEARCH which compares measured parameters obtained from zone-axis diffraction patterns with simulated patterns generated by DIFFPAT. 394

(CS In many cases, a well characterized x-ray spectrum and one good low-order diffraction pattern are sufficient for a positive identification. If the x-ray spectra are not available, or if there is more than one mineral matching the selected-area diffraction pattern, two patterns must be measured and indexed for positive identification. These two approaches to positive identification are illustrated in figure 8. These methods, although very reliable, are very time consuming. A much simpler method based on grouping partitles into "classes" is described in the subsequent section, "More Rapid Electron Microscope Methods." Figure 8. Schematic illustration of particle-identification criteria. Sample Preparation and Particle Counting Even when all necessary precautions are taken with particle identification, a number of factors can affect their apparent concentration. Very large variability between laboratories has been reported [8], in part due to problems with sample preparation methods, particulate losses, contamination, casual identifications, subjective definition of 3:1 aspect ratio, and inherent statistical limitations. Particularly large variations can occur for conditions of relatively high dust loadings. In this case the air sample volume is very small and the corresponding scale factor will be very large. The particle density on the electron-microscope specimen is still likely to be rather high, with the result that observation and positive identification of mineral fragments will be much more difficult than for low-particle-density samples of ambient air. Improved Sample-Preparation Procedure. The block diagram of our well proven procedure [4] is shown in figure 9. The first step in sample preparation is to ash the filter and its contents. This low-temperature ashing is an important aspect of the sample preparation since it eliminates organic debris. Next, the sample is agitated ultrasonically to break up agglomerates. If friable material is present, this may result in some modification of the original size distribution. The alternatives to this procedure are "direct" examination or "rubout" procedures. In the first case, the sample may contain extensive debris; and in the second case, modification of the size distribution may occur. 395 0 ~ w ~ 0 ~ r ~ a

I Pump Carbon Coating 1-11-w1 Filter Punch 3 mm Ottk Spec & Cont F41'" Low Temp Ashing DisscHe Filtar ryi Ultrasonic Vibration TEM Grid N411--~ Filtering Spec & Control Gold Coat Figure 9. Block diagram of steps in sample preparation procedure. The material is now collected on a second filter paper by suspending the sample in distilled water and pulling the water through a vacuum aspirator. Dilution may be used to control the amount of material deposited on the filter. In addition, the particle distri- bution will be much more uniform than that on the original filter. This second filter is carbon-coated. The carbon film serves to entrain the particles and provide a support material that is transparent to the electron beam. A 3-mm disk is punched from the laboratory filter; the filter material is then dissolved in acetone, leaving the carbon film and entrained particulates. The film Is picked up on the lettered grid, which permits determination of the exact position of a particle. This technique is rapid, and does not differ in any substantive manner from any method in which a carbon film is deposited on the filter to entrain the particulates. Finally, a thin film of gold is deposited on the replica, which serves as a calibra- tion material for the diffraction work; failure to include this calibration precludes measurement of SAD patterns with sufficient accuracy to allow differentiation between the complex crystal structures of the minerals under consideration. Particle Losses. In any sample preparation technique, particulate losses may be an important source of error and some method must be developed for quantifying any loss. To check on this, the same carbon-coated filter material was examined before and after dissolution. As illustrated in figure 10, the losses are negligible, confined only to those particles in the body of the filter that are not entrained in the carbon. Since filters with 0.2-micron-diameter pores are used, these losses are minimal. Other labora- tories report losses as high as 40 to 80 percent of the particulate matter deposited on the filter. A-P.vh.r.L^, nrl Fi~tr• B-Panccles in Carbon Film Figure 10. Comparison of samples before and after dissolving filter. (a) Particles on filter after carbon coating. 0 (b) Identical particles entrained in carbon film after dissolution of the filter material. w 0 ~ 396 ~ J