Tobacco Documents Online

1.0 0.9 0.8 0.7 0.6 0.5 0.4 • • • • • ChrYSOtile Mg/Si Ferroactinolite Ce/Si F• 0.1 Grunerite M Si • n 0.3 0.2 0.1 Ok 0.9 • 0'8 Amosite Fe/Si 0.7 • • • • • 0.3 0.6 • x Hombiende AI/Si 02 •• 0.5 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.B~ Figure 3. Elemental intensities ratioed to the Si intensity as a function of mineral fiber diameter. The scales for chrysotile, grunerite, and amosite are on the left and on the right for ferroactinolite and hornblende. / x \ 0 exp ' plm pm rl csc y I Ix \,J Ix \, \ S /i \ ~/ rl y C~/ rl eXp (- Plm pm r2 csc ~/ \10 / x \Ix / CIx / 2 exp ` pl pm r2 csc ~~ Si r m Si r2 exp '- pl pm r2 csc \ m (5) where p/p Im is the mass absorption coefficient for x or Si radiation by the mineral, pm is the mineral density, W is the x-ray take-off angle and (IX /ISi°)r /(Ix°/ISi°)r is the ratio of the generated intensities which is independent of r. The intensity is assumed to be generated at the center of the fiber. Rearranging yields Im ~ a pm cscV(r2-rl) (plm plmi) . 2 259 (6) N 0 a w F+ 0 ~ 0 ~ A

This expression provides a satisfactory fit (± 10 percent) to the experimental data in figure 3 except in the case of contamination at small fiber diameters [1]. Equation 6 illustrates that it is the difference between the mass absorption coefficients that deter- mines the magnitude of the absorption effect. When p/p'mineral >>'p/plmineral' a decrease in Iw/ISi occurs with decreasing size because the relative increase in emission will be greater for the element with the larger absorption coefficient. Thus, in grunerite there is a greater relative increase in Si emission (N/pl9~unerite = 1455) than in Fe emission (1'/plgrunerite - 65) and a subsequent 25 percent decrease in I(Fe)/I(Si) as the diameter decreases from 1.5 to 0.15 pm. When N/plmineral << N/p jmineral' Ix/ISi increases with decreasing size because the relative increase in emission is greater for x than for Si. Thus in grunerite, where N/plMg = 3460 and p/pISi = 1455, there is a greater grunerite grunerite relative increase In Mg emission and a subsequent 50 percent increase in I(Mg)/I(SI) as the size decreases from 1.5 to 0.15 pm. The easiest way of correcting for such effects is to use calibration curves of the type shown in figure 3. Combining eqs. (1) and (5) shows that (SA8)til(SA8)tZ = R2/Rl where t is the film thickness (r = t/2). In the case of a very thin film or fiber, taking the limit in eq. (6) as t approaches zero gives: AB not-so-thin , t PB u_ A In A8(thi--~ ~ -pfilm cscy ~ ( plfilm - plfilm ) (7) which is in accord with the expression published recently by Goldstein et al. (23]. The SCu Si' S5n Si and 5Cr Si values used to calculate the Cu-Sn-Cr values were corrected for absorption using SA8 (not-so-thin) values from eq. (7), and in all cases the relative error in concentration decreased as shown in Table 3. Figure 4 can be used as a guide to determine when an absorption correction is advisable. When the absorption coefficient difference for a given particle radius or film thickness is above the line, the absorption correction will be greater than 10 percent and should be taken into account. Many of the amphibole fibers with diameters of 0.2 pm and over require absorption corrections (1]. 260

10,000 5000 500 100 0 I 100 200 pfXmtflim/2 in µg/cm2 300 Figure 4. A(pIp) = u/p Ifilmne u/P Ifilmne (pt)film = film mass thickness. When the value of o(u/p) for a particular film thickness is below the line, the absorption correction will be less than 10 percent. The absorption correction will exceed 10 percent for values above the lines. The values shown for amosite and crocidolite indicate that the absorption correction is significant for relatively thin fibers. Instrumentally Induced Contamination Superimposed on the absorption effects just described is the sample contamination which occurs when the hydrocarbons from the vacuum pump fluids are decomposed by the electron beam and deposited on the sample surface [10]. The deposited thickness can, in time, represent an appreciable portion of the total sample thickness. The magnitude of the problem depends upon; 1) the cleanliness of the vacuum system; 2) the electron beam current density; 3) the duration of the analysis; and, 4) the difference in absorption by carbon for the x-ray lines of interest. The magnitude of the latter effect can be estimated from the following expression: with In {Ix/I5i)Woithoutnation = pc tc csc>y pICiK - plC (Ix/ISi)contamination 261 0.2 µm diameter amosite (Mg/Si) **-0.4 pm diameter crocidolite (Fe/Si) I (8)

where pC is the density of carbon and tC is the thickness of the carbon deposit in cm. Figure 5 shows the observed variation of IMg/ISi in chrysotile with time for different current densities. The analysis of small (300-400A) chrysotile fibers often requires a small electron beam (higher current density) and a longer analysis time (>5 minutes) to generate credible counting 'statistics. Even though p/pl~~ - N/pI~g is 800, the rapid decrease in IMg/ISi can only be partially accounted for by contamination implying other electron beam induced effects. When the difference in absorption coefficients Is small, contamination is not a serious problem as indicated in figure 5 for the Cu-Cr-Sn film. µ CuKa CrKa SnLa MgKa SiKa p o 5 14 51 1170 360 1.0 018 0.8 0.4 0. I 10 Chrysotlle Fibers 3.7 Km Beam a Chrysotile Fibril 0.6 pm Beam Cu-Sn-Cr Thin Film 0.7 pm Beam i I 20 30 Time In Minutes 1 40 50 Figure 5. Elemental intensity ratios as a function of the duration of electron bombardment in an ATEM operated at 80 keV. IM9JISi and ICr/ISn are plotted for chrysotile asbestos fibers and a Cu-Sn-Cr thin film respectively. The beam diameter for each analysis is indicated on the curves. The mass absorption coefficients for the indicated radiation by carbon are also shown. Optimu. Conditions for Analysis In thin films, theory predicts [24] that the peak-to-background ratio should vary approximately as In U with E , increasing rapidly at low U and then more slowly, where U is the over-voltage ratio, °acceleration potential/excitation potential. This is not always observed experimentally as shown in Table 4. The failure to increase continuously with voltage is, in part, due to the background contribution from extraneous radiation which varies from instrument to instrument. The superiority of the STEM (vs. TEM) configura- tion is indicated in Table 4 where the two STEM instruments have their best peak-to- background ratios at the highest voltage. Unfortunately, fiber or particle counting in the STEM mode is not practical [2]. When column modifications are completed, the optimum operating conditions should be experimentally determined for each instrument. Note that low voltage operation will promote absorption and backscatter effects and reduce the effectiveness of SAEO on thicker fibers. 262

Table 4. Experimental determinations of the acceleration potential providing the maximum peak-to-background ratios in the ATEM. E in keV for maximum Investigator Instrument and mode X-ray line opeak to background This report EM 300-TEM CuK 60 This report EM 300-TEM SnL 40 Russ[39] EM 300-TEM FeK 50 Joy & Maher[25] JEOL 100B-STEM M9K 100 Mizuhira[29] JEOL 100C-TEM Na-CiK 20-40 Ga11e et al.[19] Cameca-TEM A1K, Au 20 Geiss & Kyser[27] EM 301-STEM Fe and CuK 100 While there are some mineralogical ambiguities that cannot be resolved by EDS, a well-designed ATEM with the appropriate column modifications used in conjunction with good analytical procedure can provide distinctive mineral spectra that are of great utility in fiber identification. ' Selected Area Electron Diffraction Vastly differing claims have been published as to the utility of SAED in the identifi- cation of mineral fibers: Ampian [28] finds that positive identification using SAED is only forthcoming from carefully indexed patterns yielding accurate lattice parameters. Ross [29] found SAED patterns of asbestos minerals difficult to obtain and interpret and that 200 keV was required to have distinct patterns. Beaman and File [1] reported that only about 10 percent of the chrysotile fibrils examined in a standard gave distinct patterns (40 percent were crystalline). Biles and Emerson [30] reported that most chrysotile fibers in beer did not give identifiable patterns. Samudra [31] reported that 99 percent of the chrysotile fibers in the size range of 200-1200 A provided good patterns. Much of this variation can be accounted for. A distinctive SAED pattern for chrysotile: 1) has a characteristic layer line spacing; 2) is streaked in alternate layer lines; and 3) shows some characteristic reflections, e.g., those in the second row from center are often quite distinctive. We classify as positive only those fibers exhibiting all of these characteristics. Fibers showing only the correct layer line spacing as determined visually on the fluorescent screen are clas- sified as ambiguous; the streaking or characteristic reflections are not sufficiently distinctive to permit positive identification. Patterns without systematic reflections or distinctive layer lines are classified as unknown and the sum of positive, ambiguous, and unknown is termed crystalline. The percentage of fibers in each category has been deter- mined as a function of fiber size using different instruments, standards, and sample preparation methods. Droplets of 10 pL volume, prepared from the dispersion of a high purity chrysotile standard [32] in water, were placed on carbon-coated formvar films on TEM grids. The samples were examined at 00 tilt in a Philips EM300 at 80 keV and a JEOL 100B at 60 and 100 keV. Fiber searching was carried out in the selected area mode with the diffraction aperture in position and focused to minimize the time lapse between finding a fiber and obtaining a SAED pattern. The aperture size at the specimen level was 1-2 pm, the camera length was minimized, and the SAED patterns were focused with the diffraction and objective lens controls. 263 2063105058

Figure 6 shows that less than 15 percent of the individual chrysotile fibrils (300- 400 A in diameter) provide positive SAED patterns. A significantly larger portion (20-50 percent) do exhibit the correct layer line spacing (positive + ambiguous) as observed on the fluorescent screen. For the fraction of positive fibers to exceed 50 percent, the fibers must contain over 3 fibrils. 100 80 60 40 20 0 I 2 3 4 5 No. of Chrysotile Fibrils in Fiber e 7 Figure 6. The percentage of chrysotile fibers in a standard providing the indicated quality of the SAEO pattern is shown to depend upon the number of fibriis in the chrysotile fiber. The results obtained on two different instruments are plotted along with previously reported results [1j. All samples were prepared using 10 uLl water droplets containing suspended chrysotile. The results obtained in instrument B were similar at 60 and 100 keV. The lower two curves in figure 6 compare the present results with earlier work [1]. The differences are due to the present use of slightly more stringent requirements for positive identification and possibly to the use of different standards (Wards in reference 1 vs. Union Carbide). Figure 7 illustrates that the percentage of fibers providing diffraction patterns in every category is lower when using samples prepared by the Jaffe extraction of carbon-coated Nuclepore as compared to water droplets. This is presumably due to the carbon coating and/or the presence of some residual Nuclepore. Note that the positive fiber category is not significantly affected by sample preparation. 264

as 80 20 0 100 2 Positive + Ambiguous Positive In Water Drop After Jaffe Extraction Of Nuclepore 3 4 5 No. of Chrysotile Fibrils In Fiber a Figure 7. The percentage of chrysotile fibers providing the indicated SAEg pattern quality is shown to depend, to some extent, on the method of sample preparation. The results for 10 pL water droplets are compared with those obtained after Jaffe extraction of a Nuclepore filter in chloroform. All samples were examined in instrument A. 7 The primary reasons for the differing claims are the use of different criterion for classifying a pattern as positive and differences in the fibril content of the fibers being examined. A rigorous definition of positive SAED is needed if identification errors are to be avoided and interlaboratory agreement achieved. Figure 6 shows that over 70 percent of the fibers containing. three fibrils show the correct layer lines spacing (positive + ambiguous category). Most published SAED patterns are not from single fibrils as indicated by the presence of partial rings and diffraction spot smearing or multiplicity [28,33]. To a lesser extent, the reported variation is due to differences in: 1) standard source and treatment; 2) sample preparation methods; 3) instrumental capabilities; 4) operator judgment; and 5) diffraction technique. In the river, tap water, and lake samples we have studied, the chrysotile has consisted predominantly of fibers with 3 or less associated fibrils with single fibrils appearing most frequently. The fibers in 50 percent NaOH produced from chlorine cells using chrysotile asbestos diaphragms are predominantly fibrils and 80 percent have lengths less than 2 pm and 95 percent have lengths less than 5 Ns. Identification based on morphology or SAED alone in these cases has not been particularly reliable because less than 20 percent of the chrysotile fibers had a tubular appearance and only 5-30 percent gave positive SAED patterns. Those fibers identified as chrysotile had EDS spectra and fibril diameters characteristic of chrysotile. 265 2063105060

In counting fibers with the ATEM, searching with the diffraction aperture in place is not practical because the field diameter is decreased from about 7 pm to 1 pm. When counting in the TEM mode, the fiber is subjected to more electron beam bombardment before a diffraction pattern can be obtained. When searching with the diffraction aperture in position, the SAED patterns from chrysotile fibers containing three or less fibrils generally fade within 30 seconds to such an extent as to be unidentifiable. This electron beam induced change is due to dehydroxylization [28] and carbon contamination. Reliability of the Method If a sufficient number (typically 60-100) of fibers are analyzed [1,2], the method will generally provide concentrations that are accurate within a factor of two. The reproducibility is considered to be represented by the coefficient of variation or 100o/mean fiber concentration. Inter-laboratory reproducibility between two different Dow laboratories measuring chrysotile in 50 percent NaOH, which is a relatively clean sample, has recently been better than 20 percent (see Table 5). This is reasonably good performance for the small, amount of material being detected as shown in Table 5. The idgntification of an 1000A long chrysotile fibril corresponds to the detection of 3 x 10 18 grams of material [24]. The results will not be this good for a series of labora- tories using a variety of sample preparation techniques and differing criteria for fiber identification. Table 5. Experimentally measured asbestos concentrations. Sample Concentration in millions of fibers per liter Mass of asbestos in parts per billion by weiqht Midland, MI Tap Watera 0.6 0.001 Waste Water Effluenta 10-400 0.2-10 50% NaOHa 50-5000 0.5-40 Duluth Tap Waterb 25 25 50% NaOHa sample 1 Dow Lab A Dow Lab B 380 380 50% NaOH sample 2 380 300 50% NaOH sample 3 530 520 50% NaOH sample 4 1900 1500 a Chrysotile b Amphibole In order to achieve good reproducibility, we adhere to the following: 1. Use a sample preparation method with proven low fiber loss such as the extraction of carbon-coated Nuclepore [2,5,6] or apply a fiber loss correction to each sample [1,2]. 2. Count only samples that have a uniform distribution of solids on the TEM grid, i.e., the fibers per unit area should not fluctuate widely [1,2]. 266

C2 3. Count until a sufficient number of fibers (generally 60-100) have been detected so that number of fibers per unit area does not change significantly with additional counting [1,2]. 4. Use a sample volume that provides a particulate density with minimum inters ferences from non-fibrous solids. 5. Modify the TEM column to reduce electron scattering and secondary fluorescence. 6. Subtract off-fiber EDS spectra from fiber spectra. 7. Correct for absorption, when present, using standards or relative sensitivity factors. 8. Minimize contamination rates, when possible, by the use of low current density and short analysis times. 9. Experimentally determine the optimum acceleration potential which often differs for EDS and SAED performance, necessitating a compromise. 10. Use a reasonable and consistent scheme for classifying fibers. The authors wish to thank L. Sturkey and W. A. Knox of The Dow Chemical 6mpany, Walnut Creek, California, for helpful discussions concerning selected area electron dif- fraction and R. H. Geiss and D. F. Kyser of IMB, San Jose, California, for their komments and critical review of the manuscript. The assistance of D. J. Peterson of Dow Chemical of Canada Ltd. and E. B. Bradford of Dow, Midland, in performing the experimental measure- ments is also gratefully acknowledged. References [1] Beaman, D. R. and File, D. M., Anal. Chem. 48, 101 (1976), also in Proceedings Microbeam Analysis Society igth AnnualZ'onference, paper 31 (1975). [2] Beaman, D. R. and Walker,. H. J. , in FDA Symposium on Electron Microscopy of Microfibers (Aug. 1976) in press. [3] Ortiz, L. W. and Isom, B. L. , in 32nd Annual Proceedings of EMSA 554 (1974). [4] Zumwalde, R., In FDA symposium on Electron Microscopy of Microfibers (Aug. 1976) in press. [5] Cook, P. M., Rubin, I. B., Maggiore, C. J., and Nicholson, W. J., in Proceedings of International Conference on Environmental Sensing and Assessment Section 34-1 I.E.E.E. Las Vegas (1976). [6] Anderson, C. H. and Long, J. M., Preliminary Interium Procedure for Fibrous Asbestos, U.S. Environmental Protection Agency, Athens, GA (1977). [7] Benefield, D., The Dow Chemical Company, Freeport, Texas, private communication (1977). [8] Millette, J. R., in FDA Symposium on Electron Microscopy of Microfibers (Aug. 1976) in press. [9] Stewart, I., in FDA Symposium on Electron Microscopy of Microfibers (Aug. 1976) in press. 267 2063105062

[10] Beaman, 0. R. and Isasi, J. A., Electron Beam Microanalysis, 5TP506, American Society for Testing and Materials, Philadelp ia 9722T [11] Philibert, J. and Tixier, R., in Physical As ects of Electron Microscopy and Microbeam Analysis; Seigel, B.M. and Beaman, D.R., ds John Wiley and Sons New York, 333 1975). [12] Bolon, R. B. and McConnell, M. D., in Scanning Electron Microscopy/IITR[/SEM/76, Part 1 (1976). [13] Russ, J. C., in Scanning Electron Microscopy/IITRI/SEM/77 1 335 (1977). [14] Joy, 0. C. and Maher, 0. M., in Scanning Electron Microscopy IITRI/SEM/77 1 325 (1977). [15] Zaluzec, N. J. and Fraser, H. L., in Proceedings Microbeam Analysis Society, 11th Annual Conference, paper 14 (1976). (16] Packwood, R. H., Laufer, E. E. , and Roberts, W. N., in Proceedings Microbeam Analysis Society, 12th Annual Conference, paper 115 (1977). [17] Geiss, R. H. and Huang, T. C., X-ray Spectrometry 4 196 (1975). [18] Kyser, D. F. and Geiss, R. H., in Proceedings Microbeam Analysis Society, 12th Annual Conference (1977) paper 110; also private communication with R.H. Geiss (1977). [19] Ouncumb, P., J. de Microscopie 7, 581 (1965). (20] Cliff, G. and Lorimer, G. W., J. Microscopy 103, 203 (1975). [21) Russ, J. C., in Proceedings Microbeam Analysis Society, 8th Annual Conference (1973) paper 30; also in Edax Editor, 5, 11 (1975); also J. Submicr. Cvtol., 6, 55 (1974). [22] Sprys, J. W. and Short, M. A., in Proceedings Microbeam Analysis Society, 11th Annual Conference (1976) paper 9; also private communication with Sprys, J.W. (1977). [23] Goldstein, J. I., Costley, J. L., Lorimer, G. W., and Reed, 5. J. B., in Scanning Electron Microscopy IITRI/SEM/77 1, 315 (1977). [24] Beaman, 0. R., in Modern Techniques for the Detection and Measurement of Environmental Pollutants, 10th Rochester International Conference on Environmental Toxicity (May 1977) in press. [25] Reed, S. J. B. and Ware, N. G., X-ray Spectrometry 3, 149 (1974). [26] Rao, P. and Lifshin, E., in Proceedings Microbeam Analysis Society, 12th Annual Conference, paper 118 (1977). [27] Beaman, D. R. and Solosky, L. F., in Proceedings Microbeam Analysis Society, 9th Annual Conference, paper 26 (1974). [28] Ampian, S. G., in FDA Symposium on Electron Microscopy of Microfibers (Aug. 1976) in press. (29] Ross, M., in FDA Symposium on Electron Microscopy of Microfibers (Aug. 1976) in press. [30] Biles, B. and Emerson, T. R. , Nature, 219, 93 (1968). [31] Samudra, A. V., in Scanning Electron Microscopy IITRI/SEM/77 1(1977). 268