Tobacco Documents Online

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) AN OVERVIEW OF ELECTRON MICROSCOPY METHODS Clayton 0. Ruud Denver Research Institute University of Denver Denver, Colorado 80208 Abstract According to a recent National Academy of Sciences Report, animal deposition model studies have shown the fiber size has some effect upon the toxicity of mineral microfibers, the long thin ones appearing to be most active [1]1. However, the extrapolation of these results to the relative carcinogenicity in humans must be tempered by the consideration that an experimental animal model has not been established. Moreover, the size range to be considered long, thin microfibers is not clearly defined, that is to say, the shortest length may be on the order of one micrometer or ten micrometers. For this and other reasons most scientists in the field consider that it is necessary to obtain data on length and width, as well as on concentration and species of mineral fiber fragments in the environment. Due to these considerations, microscopy methods are necessary for mineral fiber analysis, and because of the small size of the particles, electron microscopy is necessary. This paper will describe the methods and techniques of electron microscopy which are most generally applied. These are the transmission electron microscope-selected area electron diffraction (TEM-SAED) and the scanning electron microscope-energy dispesive x-ray spectroscopy (SEM-EDXS) methods. The advantages and disadvantages of these two techniques will be discussed, including their relative proficiency in detecting sub-micrometer fiber fragments. Their ability to identify the species of mineral, sample preparation techniques, statistical considerations and the cost of analysis will also be reviewed. The application of various techniques and methods based upon the TEM-SAED or SEM-EDXS systems will be discussed, including situations where one or the other is the optimum method. The advantages of combined systems, scanning transmission electron microscopy with SAED and EDXS, will be discussed. Also new approaches of combination and computer controlled methods using both TEN and SEM will be described. In conclusion, the state of the art will be discussed in terms of general considerations necessary for the selection of an electron microscopy technique for mineral fiber analysis. Key Words: Amphibole asbestos; asbestos; chrysotile; electron diffraction; energy-dispersive x-ray spectroscopy; mineral microfibers; scanning electron microscopy; selected-area electron diffraction; transmission electron microscopy. 1Figures in brackets Indicate the literature references at the end of this paper. 221

Background Collection of mineral particles for identification and counting is usually done by filtering the medium, air or water, through cellulose ester membrane (Millipore) or perforated polycarbonate (Nuclepore) filters, thereby concentrating them through deposition on the filter's surface. The effective minimum particle collection size is always less than one half a micrometer. The optical microscope is used extensively for counting mineral fibers collected from occupational environments, but it is generally agreed that this is a matter of expedience and not due to adequacy. By far the greatest number of asbestos mineral fibers found in the environment, including occupational environments, are below the resolving power of the optical microscope. Since neither epidemiology nor animal studies on the relative toxicity of mineral microfiber have shown conclusively that those less than 0.5 pm in diameter or width are innocuous, it has been considered prudent to count, size, and identify all particles with an aspect ratio greater than 3 to 1 which are tens of micrometers in length and shorter. Although the long thin fibers seem to be more active in animal deposition- model studies, the shortest active fiber length has not been established [1]. Also, because a number of mineral and man-made microfibers are suspected of producing varying degrees of adverse health effects, the identification or classification of a mineral fiber as to species is important. Until the effect of size, morphology, species and other properties of microfiber can be related to toxicity, it will be necessary for the analyst to characterize the distribution of a number of these parameters from environmental samples. Electron Microscopy The group of analytical instruments which provides more of what are considered the important parameters mentioned above is that of the electron microscopes. Both. transmission and scanning electron microscopy have been used extensively for mineral fiber identification, sizing and counting, and both types of instruments and their related characterization techniques have their place. The transmission electron microscope (TEM) with selected area electron diffraction (SAEO) Is considered the most widely applicable instrument, although it has some disadvantages which will be discussed. This technique requires that the image forming electrons travel through the sample and therefore the sample matrix must be transparent to the high kinetic energy (usually about 100 KeV) electrons. SAED also requires that the electrons travel through the matrix as well as some part of the microfiber to be iden- tified. SAED is used to characterize the crystal structure of the particle of interest and is valuable for the identification of the type or class of fiber, e.g., serpentine asbestos, amphibole asbestos, non-crystalline or non-asbestos. Scanning electron microscopy (SEM) can be compared with reflected light microscopy. However, images are formed electro-optically, usually by secondary electrons produced by a focused electron beam in the sample. The technique usually employed for species identification is energy dispersive x-ray spectroscopy (EDXS) which determines the energy of x-rays emitted from the sample. This emitted x-ray energy spectrum is caused by the electron beam interaction with the sample and can be used to qualitatively and semi- quantitatively identify the elemental content of a microfiber. A third type of instrument which combines the advantages of both the TEM and SEM is the scanning transmission electron microscope (STEM). This instrument has been used by a number of laboratories, most of which have procured it specifically for asbestos microfiber counting and identification. Essentially it is a transmission electron microscope equipped with scanning and focusing coils so that a focused beam of electrons can be scanned over the sample or pinpointed in a particular area. The most general mode of application is to obtain a shadow image as with the TEM, then perform SAED and/or EDXS as desired. The focused beam should produce a brighter SAED pattern for particle identification than in the TEM, and if an elemental analysis is desired this may be obtained from the same particle without transferring the specimen to another electron beam instrument. 222

There is another type of electron microscope which has been used only sparsely for asbestos mineral fiber analysis. This instrument is an SEM with an electron detector below- the specimen for transmission imaging. This allows a transmitted electron image to be formed and the instrument might be called a transmission scanning electron microscope (TSEM). Application of this technique will be discussed in a subsequent section. Needless to say, combinations of SEM and TEM instruments have been and are being used for microfiber analysis also. Applications There are four important considerations in the selection of an electron microscopy method for the counting and characterization of microfibers. These are: observability, specificity, sample preparation and analysis cost. Observability Observability is concerned with the sharpness and contrast of the microfiber image against the matrix. This controls the relative ability of the microscopist to find microfibers, measure them, and characterize their morphology. Flinckinger and Standridge [2] compared fiber counts with SEM and TEM from water samples and concluded that for small fibers TEN gave much higher counts, about an order of magnitude or greater. Ruud et al. [3] showed the relative clarity of SEM and TEM images illustrating the superior contrast of the latter (see figure 1). The highly magnified shadowgraph obtained in transmission electron microscopy is for the most part an accurate representation of the length and width or diameter of the fiber. Chrysotile fibers are usually circular bundles of fibrils or round single fibrils. Often the fibrils can be distinguished in a TEM image by the fact that they are tubular and the hollow center can be seen in the electron microscope image [3]. While this tubular appearance is characteristic of chrysotile, it is not always present so that if a fiber does not appear to be hollow this does not rule Nt out as chrysotile. Amorphous material can be attached to the surface and fill the tubes, thereby giving the appearance, as far as density is concerned, that the fiber is solid [4]. At any rate it is well to have an identification method in addition to morphology for chrysotile and it is imperative for the amphibole minerals since non-asbestos material can appear in the electron microscope to be fibrous, i.e., they may have a 3:1 length-to- width ratio. Also, many chain silicate non-asbestos minerals fracture in the same general way as the asbestos minerals so that morphology does not lead to a reliable identification. See figure 2 from Ruud et al. [3]. The most effective additional identification method is selected area electron diffraction, which will be discussed subsequently. 223

Figure 1. Comparison of SEM and TEM image clarity for a microfiber fonn an environmental sample. Top is SEM image and bottom is TEM image. The marks in the upper left corner of each micrograph are 1 micrometer apart. 224

Figure 2. A TEM micrograph of the mineral wollastonite. The superior image contrast of small microfibers and the clarity of internal voids in the TEM can be understood when the mechanism of image production and resolution of the two types of instrumentation is compared. The TEM relies upon the electron opacity of the microfibers which depends upon the thickness but which is invariably several times higher than that of the specimen substrate. SEM relies upon the production of secondary electrons for imaging and the relative difference of their efficiency of production between microfibers and substrate is often rather slight. In spite of these considerations, a recent report issued by the EPA [5] judged the two techniques as equal with respect to fiber counting. However, the sample type and analytical procedure covered In that report were very specific and not what may be generally expected or applied in environmental samples. The sample source was a laboratory prepared and dispersed Canadian chrysotile. The TEM sample preparation was one which is se}dom if ever used in TEM preparation because it is complicated and prone to fiber loss and contamination. This report therefore cannot be used as justification for the general use of SEM for microfiber sizing and counting. The electron microscope magnification used to locate and measure microfibers is an important concern and generally varies from 4000 to 20,000 times. It should be obvious that the lower the magnification used to find microfibers consistent with sharp contrast, the higher the likelihood of missing very fine ones. At 10,000X a 0.1 micrometer fiber would appear to be 1 mm wide and at 4000X it would be 0.4 mm wide. On the other hand, the lower the magnification used to search for microfibers the larger the area of electron microscope specimen observed, thereby improving counting statistics for a given amount of analysis time. Specificity Specificity is concerned with the identification of a microfiber species. In the SEM, clues as to the elemental content may be obtained by EDXS, and these can sometimes be used to identify the microfiber. With the TEM, SAED is usually employed for speciation. SAED produces a pattern which is indicative of the crystal structure of a microfiber. This crystal structure can then be related to the type or species of fiber. Usually only classification is possible, but in the case of chrysotile asbestos it is usually readily identified by SAED. 225 2063105020

The basis for SEM-EDXS is that electron beam microchemical analysis may sometimes be used to distinguish particles of various minerals (6,7,8]. The most common method presently in use is the energy dispersive x-ray system (EDXS) attached to an SEM. X-ray wavelength dispersive analyzers and the conventional electron microprobe have been used; however, their routine application is negligible in asbestos microfiber analysis because the high electron beam currents required may damage the specimen and the microanalysis procedure is relatively time-consuming. Semi-quantitative electron beam x-ray microchemical analysis in the electron microscope is based on the fact that a beam of high energy electrons incident upon a particle generates x-rays with energies that are characteristic of the elements present in that particle. Only those elements heavier than sodium (atomic number 11) can be practically detected. An EDXS detector placed in the electron microscope sample chamber close to the specimen converts the energy of the x-ray photons to voltage pulses which are amplified, digitized and stored in a multichannel analyzer or a minicomputer. In the EDXS identification of microfibers, ambiguities can arise from x-rays produced by adjacent or adhering particles, from instrumental uncertainties in determining the exact chemical composition of a particle [9], or from the fact that a given mineral can exist over a wide range of compositions (10]. Rs much as a 10 percent variation in the element x-ray intensity can be expected from any one mineral sample [71 or even a single microfiber (11]. To further confuse the matter we have observed many mineral particles that are often associated with asbestos materials which show a 3:1 length-to-width ratio and give EDXS spectra that cannot be distinguished from the asbestos types. Figure 3 shows an example of SEM-EDXS data from an anthophyllite microfiber and a lizardite cleavage fragment with a greater than 3:1 aspect ratio. Anthophyllite is an amphibole asbestos mineral and lizardite is a non-asbestos polymorph of chrysotile. However, the EDXS spectrum from the two are indistinguishable. A number of examples of this type of possible misidentification of mineral microfiber appear in Ruud et al. [3]. In spite of the above considerations, a number of researchers have surmised that each of the asbestos minerals can give x-ray spectra that usually are characteristic enough, when combined with fiber morphology, to allow their mineral identification [6,7,12]. Visual observation of the semi-quantitative fiber x-ray spectrum is the usual method of fiber identification; however, three-component diagrams have been used after subtracting the continuous background from the semi-quantitative x-ray spectrum for further extrapolation of the data [6]. For these analyses, matrix corrections are rarely used. Typically, iron, magnesium, and silicon are plotted on the three component diagram and compositional boundaries for the asbestos minerals established. In addition to the major shortcomings mentioned in the previous paragraph, this added refinement suffers from its failure to use all compositional data obtained such as presence or absence of sodium, calcium, aluminum. and manganese which might aid in identification [6]. As has already been discussed, observation of proper elemental intensities by energy- dispersive x-ray analysis is generally not sufficient for positive identification of fibers. For example, chrysotile, anthophyllite, and fibrous talc, which have similar elemental compositions, may be difficult to differentiate (3,6]. These considerations make the sole use of SEM-EDXS unreliable in its general appli- cation to the identification of fibers and microfibers. There are specific cases where the source of the sample is well characterized and the absence of particles of nearly similar chemical composition has been confirmed that it may be useful. Considering the uncertainties in SEM application to the identification of micro- fibers, it is understandable that transmission electron microscopy coupled with selected area electron diffraction has been selected by many researchers as the most viable method for identifying and counting asbestos fibers [1]. Although this method has some disadvantages, the overriding advantage is that usually it is specific with respect to the identification of chrysotile or amphibole microfibers and it permits accurate size measurement of particles even when that size is on the order of fractions of micrometers in diameter. 226

Figure 3. SEM-EDXS spectra from anthophyllite asbestos (top) and lizardite (bottom) samples. ' Selected area electron diffraction can be readily accomplished on a modern transmission electron microscope and a pattern observed in about 10 seconds and recorded usually in less than two minutes. However it usually requires an experienced microscopist and some fine manipulation of the specimen in the SAED mode for production of a clear pattern. The two-dimensional SAED pattern of diffraction spots has the advantage, in the case of some asbestos microfibers, that it contains certain outstanding characteristics that can be recognized at a glance. This is particularly true for the more common type of asbestos, the serpentine mineral chrysotile [4,13], figure 4. 227 2063105022

I Figure 4. Chrysotile asbestos TEM micrograph (top) and SAED pattern (bottom). The SAED pattern of a single chrysotile fiber or fibril is analogous to a rotating or oscillating crystal x-ray diffraction pattern in which the long dimension of the fiber tends to lie parallel or nearly parallel to the supporting membrane and therefore is perpendicular to the incident beam corresponding to the axis of rotation being normal to the beam in the usual type of rotating crystal x-ray exposure. This analogy is also 228 ~

cv, (CS partially true for amphibole fibers. In x-ray patterns the spots are arranged in lines, universally called "layer lines," with the spacing between the lines dependent upon the periodicity of the crystal structure in the direction of the axis of rotation (see, for example, Barrett and Massalski [14]). The analogous layer lines in SAED are also very prominent and their spacing reveals the crystal periodicity in the direction of the fiber axis. From a quick view of the layer line spacing one cannot distinguish between chrysotile, tremolite, and amosite which all have layer line spacings corresponding to a periodicity of approximately 0.53 nm, but this group of materials can often be distinguished from some others of interest, for example wollastonite, lizardite, antigorite, albite, hedenbergite, or diopside [3]. Fortunately there is no need for a detailed study of the pattern in order to positively Identify chrysotile. The chrysotile diffraction pattern has very prominent streaks on layer lines other than the central one, and some streaking also may be seen on the central one [13]. Some spots of normal sharpness also occur; these are on the central layer line and alternate ones (2nd, 4th, etc.). The streaks are seen on the pattern in figure 4 and can also be seen on the fluorescent screen of the electron microscope. The geometry of the pattern is known for orthochrysotile, clinochrysotile, parachrysotile and mixed ortho plus clino varieties [15], and the origin of the streaks is now well understood as resulting from disorder in the stacking of the prominent layers in the crystal (the hydroxyl, magnesium oxygen-hydroxyl, silicon and oxygen layers). The series of researches beginning with Warren in 1941 and extending through many studies by Whittaker in 1956, have shown that the layered structure is curved cylindrically around the axis of the fiber, the axis with 0.53 nm periodicity in clino and ortho varieties. This is called the c axis in some of the papers [16], but is called the a axis on others [15]. There is x-ray evidence [16] that the layers are wrapped in a helical cylindrical manner and this is confirmed by electron microscopic views of the cross-section of the chrysotile tubes by Yada [17]. This curvature of the structure accounts for the preqence of the prominent layer lines, which are perpendicular to the length direction of• the fiber. Amphibole minerals exist in both asbestiform and massive varieties. Numerous names have been given to varieties of the amphibole groups, and the many different types of atoms substituted in the different members of the groups [18] add to the natural difficulties of identifying them. It is not surprising that the Joint Conmittee on Powder Diffraction Standards x-ray powder data file contains many cards of diffraction patterns differing from each other by small amounts. SAED patterns prepared in this laboratory of known samples of the amphibole asbes- tiform minerals tremolite, crocidolite and amosite have prominent rows of spots which resemble the layer lines of rotating crystal x-ray patterns and which we will also call 'layer lines.' There are especially closely spaced spots on each of these layer lines, far more closely spaced than they are in the rows of spots from the minerals hedenbergite, albite or wollastonite, for example [3]. We have rarely observed any non-asbestos material exhibiting the characteristic layer line spacing and spot patterns within the layer lines displayed by asbestos mineral fibers. However, this author has recently been informed that pyroxenes have been observed to produce asbestos-like SAED patterns. Although chrysotile is usually readily distinguished from the asbestiform varieties of amphibole (crocidolite, amosite2, anthophyllite, tremolite and actinolite), it is not easy to distinguish one variety of these amphiboles from another because the spacing of prominent rows of spots in these are the same, and the differences occur only in the arrangement of spots along the rows. However, an experienced microscopist can learn to distinguish on sight a pattern usually characteristic of an asbestos fiber from the patterns of most non-asbestos minerals commonly associated with them. Crystalline materials that exist in the form of thin plates also produce SAED patterns with many spots, but these in general are arranged in a two-dimensional array in which there are not such prominent layer lines in a single direction. ZAmosite - a discredited term. N 229 ~ r+ 0 w a N A

As mentioned above, SAED is used extensively as the major criterion for the identification of mineral microfibers [1,2,3]. However, it should be mentioned that the method is empirical and has not been rigorously tested. The possibility exists that some species of non-asbestos mineral fibers or microfibers may produce a high incidence of SAED patterns characteristic of chrysotile or the asbestos amphiboles. An example of this, which has been mentioned, is pyroxenes. Transmission electron microscopes and STEM equipped with an energy dispersive x-ray detector are available which allow simultaneous observation of morphology, crystal structure and elemental composition. These microscope systems have been used to study fibers of known asbestos origin as well as environmental and material samples [12,19]. It would be highly advantageous if a thorough crystallographic examination of the SAED pattern could be performed in the few seconds in which patterns are now cursorily examined. This is technologically possible, but requires the building of a TEM or STEM with a television camera in place of the fluorescent screen coupled to a computer programmed to index and classify the pattern with respect to standard or calculated patterns. These facilities are extremely expensive and few laboratories will be so equipped in the near future. However, studies of the patterns with respect to mineral type, cleavage and fiber orientation are needed. Sample Preparation As previously mentioned, samples for electron microscopy analysis of microfibers are generally collected on cellulose ester membrane (Millipore) or perforated polycarbonate (Nuclepore) filter media [5,19]. For analysis by the SEM the latter medium, due to its smooth surface, is preferred. SEM preparation is usually done by coating the surface directly with an electrical conducting material, e.g. gold, silver, carbon or silicon monoxide [5]. More complicated methods have been used for SEM preparation of samples collected on Millipore [9]. These filters with their rough surface are not generally suitable for direct coating for SEM because small fibers may be masked by protrusions of the surface. In TEM, STEM, and TSEM analysis, the matrix must be nearly electron transparent to electrons of about 100 KeV energy. This requires that the filtrate (particles) be mounted upon electron microscope grids with very thin, on the order of 100 Angstroms, carbon or metallic substrates and the filter material dissolved away. Several dissolution techniques are used, including the Jaffe wick and condensation washing. Generally these techniques are relatively simple and maintain the original particle size distribution and relative particle location. Some investigators have reported particle losses as high as 60 percent with the condensation washing technique compared with less than 10 percent with the Jaffe wick method [20]. Coating the filter and filtrate with a conductive layer prior to dissolution has been proposed as a technique to minimize particle loss [19,21]. Also, careful control of the condensation washer can reduce filtrate loss to much less than 10 percent. Most laboratories apply a second carbon, metallic or silicon monoxide coating to the filtrate after filter dissolution to reduce the probability of particle loss. The choice of conduction coating is varied; however, many laboratories have been considering fine grained metallic coatings because of superior contrast and the fact that a reference pattern is provided on the SAEg patterns. The general preparation technique discussed in the previous paragraph is known as the "direct transfer" method. A variety of more complicated techniques include the direct transfer procedure as the last few steps. This includes ashing of the sample which is required when a considerable amount of organic material is collected with the inorganic microfibers or sometimes is used as a preliminary step to redistributing the filtrate for a more uniform or more suitable concentration. Dissolution of the collection filter substrate and subsequent refiltering has also been used. Needless to say, whether TEM or STEM is performed, the particulates must be distributed as uniformly as possible on the filter sample. This is a vital consideration in the statistics of analysis which will be covered by another author in this publication. Ashing can be performed in a low temperature oxygen plasma device or at high temperatures in a muffle or tube furnace. There are pros and cons to all redistribution procedures which must be considered by the analyst; however, it is always highly desirable to process control specimens, i.e., 230

blanks, when preparing samples for fiber counting and analysis. These blanks confirm a clean preparation environment or bear witness to laboratory contamination. Another preparation technique which has been used off and on is the so-called "rub- out" technique. This was used early in the electron microscopy analysis of microfibers and has been applied by the Mount Sinai group [22]. High particle losses and the destruction of the true particle size distribution to produce only a mass concentration are cited as disadvantages with this technique. Other techniques have also been cited as viable, including that in a recent EPA report [5]. However, most have been discarded in favor of the direct transfer method alone or preceded by ashing only when necessary. The added specimen handling necessary for transmission electron analysis has often been cited as a serious disadvantage to TEM, STEM and TSEM analysis. However, experienced laboratories have developed preparation routines and techniques which make particle losses, contamination and labor time negligible. The usual amount of time lag in preparation of a transmitted electron sample is about four hours. Analysis Cost The amount of electron microscope time necessary for an analysis is the major consideration affecting cost, and is dependent upon many factors, not the least of which is the sample from which the specimen was produced. The size distribution, particle loading and uniformity of distribution are just three of these. If a very limited amount of microscope time requires that the analyst use only a low magnification, e.g., 4000X, then the small microfibers may be missed. Computer image analysis has been used by a few laboratories [9] and can be applied directly on an electronic image as produced in the SEM, STEM and TSEM or on photomicrographs produced by an electron microscope. Direct computer image analysis is also possible with suitably modified imaging devices mounted into TEM's. This technique can greatly reduce the amount of microscope time requireil for microfiber searches but is prone to certain errors, especially where high concentrations of microfibers and other particles are present. The application of microfiber identification techniques affects the microscope time as well as imaging. The TEM image is essentially instantaneous, whereas an SEM image must be acquired with time and takes several seconds to form. Furthermore, on a typical SEM the time for one EDXS analysis is 100 or more seconds. As a conse- quence most analysts working with SEM and STEM only obtain analyses from selected microfibers, not all of those found. SAED usually requires 10 to 30 seconds to form an image suitable for recognition by the microscopist and is usually performed on all microfibers found. Recording of this image is done selectively on a few microfibers and usually requires 100 to 200 seconds. The beam focusing feature available on all STEM and some TEM can reduce the recording time by producing a brighter SAED pattern. Technique Development A number of laboratories are evaluating the various electron microscopy techniques used in the analysis of microfibers. That this is necessary is evident from the wide discrepancy in results produced on similar samples by different laboratories and/or microscopists (23]. No two laboratories perform sample preparation or microfiber analysis exactly the same and some are markedly different. However, over the past three years a number of laboratories have markedly improved their analytical reliability in spite of the overwhelming statistical uncertainties. This author is aware of some new approaches to the identification, counting and measurement of microfibers. United States Steel Research Laboratories are applying a specially equipped TSEM with an EDXS detector located for a very high x-ray take-off angle, higher than possible in a standard unit. This sytem is computer controlled using criteria from the transmitted electron image data at 1g,000X magnification processed through an image analyzer to locate microfibers. The geometry of the system and the sample and x-ray detector distance (less than 1 cm) are such that a very adequate EOXS spectrum can be accumulated an order of magnitude faster than with standard electron microscopes, SEM or STEM. After a statistically significant number 231 2063105026

of microfibers are found and EDXS data obtained from each, they are classified with respect to aspect ratio and EOXS spectrum. The specimen is then transferred to a TEM, a 1200 KeV instrument in this case, where some microfibers in each EDXS clas- sification are selected and an SAED pattern obtained for identification. It is recog- nized that a 1200 KeV TEM is not' readily available; however, the SAED could be performed on most TEN instruments with 80 to 100 KeV. The advantage in the transmitted image over that usually produced in an SEM is greater visibility of particles, as has previously been stated. Moreover, the tech- nique has a great advantage over those presently applied from the standpoint that a large number of microfibers are analyzed at least through classification and this is a tremendous statistical advantage. Conclusion In conclusion there are a few points that should be made. 1. The transmitted electron image is generally accepted as being superior for counting and measuring microfibers as compared with a secondary or backscatter electron image. 2. Selected area electron diffraction is generally accepted as the best criterion for the identification of asbestos mineral microfibers, although a few non-asbestos minerals may be mistaken for asbestos. 3. The statistical consideration affecting electron microscopy of microfibers is a source of considerable error and new techniques are being and must be developed to relieve these problems. 4. There are a few specific situations where the SEM can be applied to the counting of microfibers, especially where the source and species mixture are well characterized. 5. Although the TEN-SAED method of asbestos mineral microfiber counting and identifica- tioh is not absolute, it is the best compromise of accuracy and cost available. The author would like to thank C. S. Barrett and J. M. Dement for their contribution to this paper. References [1] National Academy of Sciences, "Drinking Water and Health," Part I, 1977. [2] Flinckinger, J. and Standridge, J., Identification of fibrous material in two public water supplies, Env. aci. and Tech. 10, No. 10, 1028-1032 (Oct. 1976). [3] Ruud, C. 0., Barrett, C. S., Russell, P. A., and Clark, R. L., Selected area electron diffraction and energy dispersed x-ray analysis for the identification of asbestos fibers, a comparison, Micron 7, 115-132 (1976). [4] Mumpton, F. A., Characterization of chrysotile asbestos and other members of the serpentine group .inerals, Siemens Review XLI, 7th Special Issue, 75-84 (1974). [5] Gerber, R. M. and Rossi, R. C., Evaluation of electron microscopy for process control in the asbestos industry, EPA-600/2-77-059 (Feb. 1977). [6] Rubin, I. B. and Maggiore, C. J., Elemental analysis of asbestos fibers by means of electron probe techniques, Env. Health Perspect. 9, 81-84 (1974). 232

[7] Ferrell, Jr., R. E., Paulson, C. G., and Walker, C. W. , Evaluation of an SEM-ES method for identification of chrysotile, In Proc. 8th Annual SEM Symposium, Johari, 0. and Corvin, E. (eds.). IIT Research InstStute, Chicago, I11., 537-546 (1975). [8] Langer, A. M. , Rubin, I., and Selikoff, I. J., Electron microprobe analysis of asbestos bodies, Histochem. Cytochem. J. 20, 735-740 (1975). [9] Pattnaik, A. and Maakin, J. D., Development of scanning electron microscopy for measurement of airborne asbestos concentration, U.S.E.P.A., Office of R and 0 Pub. No. EPA 650/2-75-029. Research Triangle Park, N.C., (1975). [10] Speil, S. and Leineweber, J. P., Asbestos minerals in modern technology, Env. Res. 2, 166-208 (1969). [11] Pooley, F. D., The identification of asbestos dust with an electron microscope microprobe analyzer, Norelco Reporter, 23, No_2, 5-9 (Oct. 1976). [12] Dement, J. M. , Zumwalde, R. D., and Wallingford, K. M. , Asbestos fiber exposures in a hard rock gold mine, In Proc. N.Y. Acad. of Scci. Conf. Occup. Carcinogenesis. (1975). [13] Clark, R. L. and Ruud, C. 0., Transmission electron microscopy standards for asbestos, Micron 5, 83-88 (1974). [14] Barrett, C. S. and Massalski, T. R., Structure of Metals, (McGraw-Hill, New York, 654, 1966). [15] Zussman, J., Brindley, G. W., and Comer, J. J., Electron diffraction studies of serpentine minerals, Amer. Mineralogist 42, 133-153 (1957). [16] Whittaker, E. J. W. , The structure of chrysotile, Acta Cryst. 27, 659-664 (1956). r [17] Yada, K., A study of microstructures of chrysotile asbestos by high resolution microscopy, Acta Cr st. 9, 855-857 (1971). [18] Deer, W. A., Howie, R. A., and Zussman, J., Rock Forming Minerals, (Wiley, New York, Vol. 3, p. 270, 1963). [19] Cook, P. M., Rubin, J. B., Maggiore, C. J., and Nicholson, W. J., X-ray diffraction and electron beam analysis of asbestiform minerals in Lake Superior waters. In Trans. Inst. Electrical Electronic Eng. (in press) (1974). [20] Beaman, D. R. and File, D. M., The quantitative determination of asbestos fiber concentrations. The Dow Chemical Company, unpublished report (1975). [21] Oriz, L. W. and Loom, B. L., Transfer technique for electron microscopy of membrane filter samples, Am. Ind. ~jg. Assoc. J., 423-425 (1974). [22] Rohl, A. N., Langer, A. M., and Selikoff, I. J., Environmental asbestos pollution related to use of quarried serpentine rock, Science, 196, 1319-1322 (17 June 1977). [23] Brown, A. L., Taylor, W. F., and Carter, R. E., The reliability of measures of amphibole fiber concentrations in water, Env. Res. 12, 150-160 (1976). N ~ Y O 233 0 ~

Discussion J. LEINEWEBER: I would like to make one comment with regard to Clay Ruud's remark about using the central channel of the chrysotile fiber for identification. This is good a reasonable percentage of the time, but you can run into chrysotile fibers such that this channel is not very visible and may be pretty well filled up with the non-cystalline material. So, it cannot be used as positive identification. C. RUUD: I know. R. FISHER: I want to get clarification whether you advocate visual identification from the diffraction patterns and visual counts in contrast to recording micrographs. It seems to me desirable to have your data in a form that others can confirm, look at your diffraction patterns, look at your counts, and not rely on visual observations that are just recorded in a pad or notebook. RUUD: I hear what you say, and I would like to record every pattern or every micro- graph that is projected on the screen, but I can't afford to do this; my sponsor won't stand for it. So, what we do is, we record typical SAD patterns we see in particular samples or sets of samples. When we see something different than that, something unusual, strange, we record it. I agree that it would be nice to have everything recorded for posterity, but it takes too much time. FISHER: Well, at this stage it is essential to have records that can be accepted by others, I am afraid. I agree the costs are high, and people will have to pay them, but I think that any data that are not recorded for confirmation and detailed examination are going to be challenged in all kinds of situations. RUUD: As I say, we record typical ones; we save the samples and since the samples are on finder grids any grid can be found and the data confirmed. J. ZUSSMAN: I'd like to make three comments concerning Dr. Ruud's paper. One concerning electron diffraction patterns. I think he has very much underplayed the varia- tions and variability one can get in electron diffraction patterns, depending very much upon the orientation of the grain, the way it lays on the stage. If you look at these patterns carefully, you see enormous numbers of different effects; I would have made this comment anyway; I make it still much more strongly now, having heard a lot of judgments are made perhaps without even taking photographs. From looking down on the screen you can certainly not see the subtle variations which are nevertheless important, produced by orientation effects. Secondly, you mention the scanning electron microscope as being best for chemical analytical purposes. I don't think it is capable of an accuracy that can be obtained by the transmission electron microscope with suitable attachments, or STEM, which brings me to the third point. You showed that lizardite and anthophyllite were not distinguishable from their x-ray fluorescence spectra, and this is surprising. The magnesium-to-silicon ratio for lizardite is 1.5 to I in atomic ratio, the other ratio is 0.9 to 1, and I think there is a detectable difference. The reason why we may not pick up this difference is that your crystal has the wroog kind of thickness so that the crude ratios of peak height are not indicative of concentration. The crystal has to be of a suitable thickness for this to be so. RUUD: Regarding the last comment, we can rotate the fiber or change the position in the microscope, and get different ratios, and, as someone pointed out yesterday, just be going along the fiber you may get different ratios. So, that's one reason why I do not have too much confidence in energy-dispersive x-ray spectroscopy. The first comment had to do with selected-area diffraction and the variability of patterns. We do not study them that carefully. We do not try to distinguish between the various amphiboles, amphibole asbestos materials. We do not have the time to study individual patterns that carefully. We looked at the possibilities of trying to get good d-spacings from them; it seems like it is a good possibility if we could connect the computer into a vidicon or a camera tube in the bottom of a TEM or STEM and put it directly into a computer; I think that would be great. But, so far I know of only one microscope equipped that way, and it is not used for asbestos analysis. 234