0 Bewertungen0% fanden dieses Dokument nützlich (0 Abstimmungen)
15 Ansichten7 Seiten
X-ray Microanalysis of Small Precipitates in an iron base alloy is described. The influence of precipitate size, position and composition is evaluated. Optimum experimental conditions for microanalysis of these precipitates are established.
Originalbeschreibung:
Originaltitel
MTA,83-STEM-EDS Xray Microan of Small Prec in Fe Bas Th Foils
X-ray Microanalysis of Small Precipitates in an iron base alloy is described. The influence of precipitate size, position and composition is evaluated. Optimum experimental conditions for microanalysis of these precipitates are established.
X-ray Microanalysis of Small Precipitates in an iron base alloy is described. The influence of precipitate size, position and composition is evaluated. Optimum experimental conditions for microanalysis of these precipitates are established.
P. DOIG and P. E. J. FLEWlTT The experimental conditions for detection of segregated elements in the X-ray spectrum recorded from small precipitates _< 100 nm diameter contained within a thin foil of an iron base alloy using STEM- EDS X-ray microanalysis are described. The influence of precipitate size, position and composition, foil thickness, electron accelerating voltage, and X-ray emission intensity on its detectability are evaluated. Optimum experimental conditions for microanalysis of these precipitates are established and the limitations of current techniques discussed. I. INTRODUCTION SMALL precipitates, -<100 nm diameter, are important microstructural features in controlling the mechanical and chemical properties of many commercial ferrous and non- ferrous alloys. Interpretation of such properties requires a description of their morphology, distribution, and crys- tallography, as well as their chemical composition. The most widely available technique for undertaking micro- analysis of these small precipitates is that of scanning transmission electron microscopy combined with energy dispersive X-ray spectroscopy (STEM-EDS) of thin foil samples. This technique, however, suffers practical limi- tations in that small precipitates with diameter less than the foil thickness must be analyzed simultaneously with the surrounding matrix. Thus, it is desirable, whenever pos- sible, to examine either very thin regions of foil or utilize techniques for extracting the precipitates from the matrix. In practice, it is not possible to prepare foils -< 50 nm thickness such that precipitates with diameter less than this value will always be contained within the foil. The alternative of examining precipitates which have been extracted from the matrix may introduce spurious artifacts into the microanal- ysis and prejudice any correlation with the microstructure. Since the X-ray spectrum recorded from small precipi- tates contained within the foil volume is derived from both the precipitate and matrix, the relative proportions of the total signal from each will depend on the foil thickness, the precipitate size, and its depth within the foil together with the local volume electron flux intensity distribution within the sampling electron probe. Intuitively, use of smaller incident sampling electron probes with higher inci- dent energies and thinner foils will result in a larger propor- tion of the recorded X-ray quanta being derived from the precipitate and therefore optimize the microanalysis. These conditions, however, serve to reduce the total number of emitted X-ray quanta thereby decreasing the statistical confidence in the recorded data. These two competing effects are nonlinearly related in that the relative benefits of increasing the proportion of X-rays emitted from the precipitate are not directly complemented by an equivalent decrease in the statistical confidence of the data. Thus, there exists some compromise between these competing P. DOIG and P. E. J. FLEWI TT are bot h Resear ch Offi cers wi t h Cent ral Electricity Generat i ng Board, Sout h East ern Regi on, Scientific Servi ces Depart ment , Gravesend, Kent , Uni t ed Ki ngdom. Manuscri pt submi t t ed November 18, 1982. effects that results in a maximum sensitivity for any micro- analysis which may be achieved by adjustment of the experi- mental variables. The purpose of this paper is to examine the influence of the experimentally adjustable variables of incident electron probe size, total electron flux, electron accelerating voltage, foil thickness, and precipitate position within the foil on the sensitivity of microanalysis of small precipitates contained within thin foils. II. THEORETICAL ANALYSIS OF THE X-RAY EMISSION FROM A SMALL PRECIPITATE In this work we assume spherical precipitates with a uni- form concentration, Co, of a particular element and diameter d positioned at a depth t ' in a foil of thickness t~ shown schematically in Figure 1. The measured characteristic X-ray intensity from a given microstructural feature, I ' , is a convolution of the volume electron flux, l ( v ) , and the solute composition, C ( v ) , distributions within the total sampled region of foil: I ' = K I v I ( V ) C ( v ) d v [1] where K is a constant describing the efficiency of X-ray generation, emission, and detection for the particular ele- ment of interest. An analytical expression which approxi- mates the electron flux distribution within the foil, I (v), has been given by the authors: 1'2 I ( r , t) = Ie[Tr(2o "2 + / 3 / 3 ) ] - 1 . exp[_rZ/ (2cr 2 + /3t3)] [2] for a total incident electron flux of Ie where I ( r , t ) is the electron flux at a distance r from the center of the probe at a depth t in the foil, o, is a measure of the incident electron probe size such that the probe diameter, B (FWHM) -~ 2.35~r, and/3 is a parameter defining the electron scattering characteristic of the foil material given by: /3 = ( 4 Z / E o ) 2 " p / a . 500 [3] Here Z, A, and p are the atomic number, the atomic weight, and density of the foil material and E0 is the electron accelerating voltage in eV;/3 is then given in units of nm -1. The assumption of uniform precipitate composition C (v) = Co in Eq. [1], and a matrix composition = 0 reduces this convolution to a simple integration of I (v) over the volume of the precipitate. Thus, integration of Eq. [2] over the METALLURGICAL TRANSACTIONS A VOLUME 14A, OCTOBER 1983 - - 1943 (a)
O u . - d. ~ 1 t' (b) Co 0 (c) Fig. 1- - ( a) Transmission electron micrograph showing a typical small second phase precipitate, P, contained in a thin foil sample. (b) Schematic diagram showing the position of the included precipitate within the thin foil. (c) Schematic composition profile across a small precipitate. precipitate volume gives the total electron flux within the precipitate and thereby from Eq. [1] its contribution to the total X-ray signal. The accuracy of this X-ray measurement is given by a statistical consideration of the finite number of detected X-ray quanta. The significance of the measurement is re- lated to the counting statistics in the characteristic X-ray energy peak, Np, compared with those in the background of the spectrum Nb (Figure 2). Confidence in any measured peak intensity (or integrated counts) is dependent on both the peak and background such that for a large number of counts where Gaussian statistics may be assumed (N - 20): Peak intensity = Np "+" (Np) I/2 [4] Background intensity = Nb + (Nb) ~/2 [5] Detection of a peak requires its magnitude to be greater than the error in its estimation such that: Up >-- (Up + Ub) ''2 + (Ub) 1/2 [6] 1944--VOLUME 14A, OCTOBER 1983 I--- Z I - - I I [ / / 1 / \ i ~ Np [ N b I ENERGY Fig. 2- - Schematic diagram defining the number of X-ray counts recorded in the characteristic peak, Np, and in the background, No. For any given electron source the total incident electron flux, le, is related to the incident probe size, B, by: 3'4'5 l e oc B 8/3 [7] Thus, the total emitted X-ray intensity, I ' , increases with beam size. The background X-ray intensity in a recorded spectrum is proportional to the electron probe current, Ie, the foil thickness, tl, and the total recording time, T, such that: 6 N b = K' BS/ 3hT [ 8] where K' is a constant for the particular X-ray energy and electron illumination system which characterizes the "effective" electron intensity for X-ray generation and detection. The X-ray intensity, Np, from the precipitate is given by the volume integral: Up = K'RB8/3T JvI(V) dv [9] where R is the characteristic X-ray peak-to-background ratio (Np/Nb) in the spectrum obtained from a thin foil sample of unit composition, and V is the volume distribution of the precipitate. Here, the integral term represents that fraction of the foil volume which contributes to the charac- teristic X-ray peak. The inequality in Eq. [6] defines the condition for de- tecting a characteristic X-ray energy peak from the pre- cipitate in a recorded X-ray spectrum. Nb and Np may be calculated from Eqs. [8] and [9] for given values of foil thickness tl and material /3, spectrum recording time T, electron probe size B, "effective" electron intensity K' , pre- cipitate composition Co, depth t ' , and diameter d. Thus, from Eqs. [6], [8], and [9], we may establish conditions for detecting the segregated element in the X-ray spectrum from a precipitate of given composition contained within a thin foil using STEM-EDS X-ray microanalysis. III. RESULTS OF THE THEORETI CAL ANALYSIS For a given precipitate geometry, the volume integral component of Eq. [9] decreases with increasing incident METALLURGICAL TRANSACTIONS A low Io w fd v
Np/ N b Detected ! low N b d ~ d Fig. 3- - Sc he ma t i c diagram showing the range of electron probe size, B, required for the X-ray detection of a small precipitate of diameter, d, contained within a thin foil. d* is the mi ni mum precipitate diameter whi ch may be detected in the recorded X-ray spectrum requiring a critical electron probe size, B*. O (a) 0 3 0 10 1 log d \ electron probe size, i.e., a smaller proportion of the total electron flux interacts with precipitate. Conversely, the total electron current increases rapidly (Eq. [7]) such that the resulting total electron flux which interacts with the precipitate will tend to increase. Therefore, from these two competing effects it is possible to define a range of electron probe diameter over which the inequality in Eq. [6] may be satisfied such that the intensity of characteristic X-rays emit- ted from the element segregated to the precipitate may be detected in the recorded spectrum. Clearly, for very small precipitates the inequality will never be satisfied in that the volume integral term (Eq. [9]) tends to zero and the segre- gation cannot be detected. This overall behavior is shown schematically in Figure 3 where the electron probe size range for detection is given as a function of the precipitate diameter. For small electron probes the overall X-ray count rate is low such that statistical noise in the recorded spec- trum precludes detection whereas for large probes the vol- ume integral term decreases such that the measured Np/Nb is reduced to a level below that for which the increasing statis- tical confidence in the data is able to compensate. The form of the curve defining the bound for detectability displays a minimum at a precipitate diameter d*. This diameter is a measure of the overall sensitivity for microanalysis. Calculated curves defining the limits of probe size B for detection of a segregation to a precipitate with diameter d are shown in Figure 4 as a function of (a) the product METALLURGICAL TRANSACTIONS A 2 t ~ O 1 0 0 1 2 (b) l og d Fig. 4- - Ca l c ul a t e d plots showing the range of electron probe size, B, for detection of a precipitate wi t h di amet er d in the X-ray spectrum as a function of: (a) product Ro = RCo = 10, 1, 0.1. (b) ' Effect i ve' electron flux K' T = 10 -I, 10 2, 10 3, 1 0 - - 4 nm-ll/3. VOLUME 14A, OCTOBER 1983-- 1945 o (c) 0 (e) n m m 0 I I I I i I i I I | 1 log d I I I I I I I I ] 1 log d ,00 100 0 0 _ 2 0 0 I I I l I I I I I I I ! o 3 t 1 0 1 2 ( d ) l og d Fig. 4--Calculated plots showing the range of electron probe size, B, for detection of a precipitate with diameter d in the X-ray spectrum as a function of: (c) foil thickness h = 1 0 0 , 2 0 0 , 300 nm. (d) Precipitate depth t ' = top, center, and bottom of foil. (e) Electron accelerating voltage E0 = 100, 200, 300 kV for standard conditions given in the text. Ro = RCo, (b) "ef f ect i ve" el ect ron flux, K' T, (c) foil thick- ness, tl, (d) precipitate dept h, t ' , and (e) el ect ron accel er- ating vol t age, E0. These cur ves are eval uat ed for t ypi cal standard condi t i ons whi ch relate t o the operat i on of a t ypi cal 100 kV STEM i nst rument fitted wi t h an ener gy dispersive X- r ay spect romet er: (i) i ron base f o i l s - - d e f i n i n g / 3 in Eq. [3] (ii) foil t hi ckness = 200 nm (iii) el ect ron accel erat i ng vol t age = 100 kV (iv) "ef f ect i ve" el ect ron flux, K' T = 0. 01 nm -11/3 (v) an assumed pr oduct R0 = RCo = 1 (vi) with t he preci pi t at e dept h = t l / 2, i. e. , at t he cent er of the foil. Al l curves have the f or m shown schemat i cal l y in Fi gure 3 wi t h a mi ni mum in t he precipitate di amet er, d*, at a critical val ue of el ect ron pr obe size, B*. Fi gure 4(a) shows t he i nfl uence of precipitate composi t i on. The val ue of d* de- creases by a fact or of - 2 . 5 for a t enfol d i ncrease in peak- t o- backgr ound rat i o, R0, wi t h an associ at ed but smal l er decrease in el ect ron pr obe size, B*. B* tends to a l ower limit with i ncreasi ng R0 whi ch for t he above st andard con- ditions is - 2 0 nm. Fi gure 4(b) shows t he effect of changes in the "ef f ect i ve" el ect ron flux, K' T, whi ch i ncl udes bot h t he r ecor di ng t i me, 1 9 4 6 - - V O L U M E I 4 A , O C T O B E R 1 9 8 3 M E T A L L U R G I C A L T R A N S A C T I O N S A T, and the intensity parameter K' . This latter term is de- pendent on the electron source brightness, the geometry and design of the electron optical illumination system and the efficiency of X-ray generation, emission and detection for the element of interest. The particular value of K'T for a microanalysis experiment is reflected in the background i nt ensi t y val ue Nb (Eq. [8], Fi gure 2). The results in Figure 4(b) show that increasing K'T by a factor of ten decreases d* by about 30 pct with, as before, a decrease in B* t oward a l ower l i mi t i ng val ue of - 2 0 nm f or K'T = 1 0 - 3 nm -H/3. The influence of the experimental variables of foil thick- ness, tl, precipitate depth, t ' , and electron accelerating volt- age, E0, are shown in Figures 4(c) to (e). Figure 4(c) shows that increasing the foil thickness, tl, over the range 100 to 300 nm increases d* nonlinearly by a decreasing amount from 12 to 17 nm only with a corresponding change in B*. The minimum value of B* for a 100 nm thickness foil is 15 nm. Figure 4(d) shows the effect of precipitate depth, t ' , for three positions within the foil: the top, center, and bot- tom. Here, d* increases with depth into the foil as a result of the increasing effect of the electron scattering within the foil reducing the proportion of the total electron flux inter- acting with the precipitate. For these conditions d* increases from 12 to 18 nm with an associated change in B* from 15 to 60 nm. Changes in the electron accelerating voltage, E0, influ- ence the electron scattering behavior described by /3 in Eqs. [2] and [3], and also the parameters K' and R0. Consid- ering firstly the influence on/ 3, increasing the electron ac- celerating vol t age from 100 to 200 kV reduces d* by - 1 0 pct and is again accompanied by a corresponding change in the associated electron probe diameter from 20 to 12 nm (Figure 4(e)). This improvement in detectability with increasing electron accelerating voltage is evaluated for a constant "effective" electron flux, K'T. For a constant true electron probe current, however, since the X-ray ionization cross section, Q, and therefore K' is proportional to the factor ln(E~/Eo)/(E~Eo) where Ec is the characteristic X-ray energy, 6'7 increasing the accelerating voltage from 100 to 200 kV decreases the "effect i ve" electron flux by - 4 0 pct. Complementing this decrease, however, is the associated change in peak-to-background ratio, R, for this increase in electron accelerating voltage. Theoretically, 6,8 an im- provement of up to 40 pct may be achieved, 8 but in practice this is difficult to realize because of increasing contributions to the background from spurious X-rays within the micro- scope and detector. Thus, for a constant current in the elec- tron probe, increasing the electron accelerating voltage from 100 to 200 kV has almost no effect on the value of d*. For a constant electron source geometry, the electron probe cur- rent increases linearly with electron accelerating voltage. 3,4,9 However, for instruments with variable electron acceler- ating voltage, the illumination system geometry is not nec- essarily constant and this simple linear relationship will rarely apply. Nevertheless, the results of Figure 4(b) show that even a twofold increase in the "effective" electron flux will have little effect on d*. The influence of the "effective" electron flux term, K'T, on the minimum diameter, d*, is more clearly illustrated in Figure 5. Here, d* is plotted as a function of the term K'T over the range 10 4 to 10 -1 nm 11/3 for R0 values of 0. 1, 1, and 10, and the other conditions as defi ned above for Figure 4. In addition, results are shown for R0 = 10 and foil thicknesses of 100 and 300 nm. The data for 200 nm thick foils are described by the empirical equation: log(d*) = 0.75 - 0.18 log (K' T) - 0. 37 log ( Ro) [10] Thus, a tenfold increase in the "effect i ve" electron flux decreases d* by a factor of - 1 . 5 whereas a similar increase in R0 decreases it by a factor of - 2 . 3 . In practice R0 is limited primarily by the resolution of energy dispersive spectrometers. The "effect i ve" electron flux term, K' T, how- ever, may be increased most significantly by using higher brightness electron sources such as LaB6 or field emission tips where improvements of up to X100 in electron current density may be achieved 4 for smaller probe sizes. Any in- crease in K'T by extending the spectrum acquisition time, T, is limited by the mechanical and electrical stability of the microscope. For these small precipitate sizes, - 1 0 nm, a realistic period for microanalysis is - 1 0 0 seconds. IV. DI SCUSSI ON The above theoretical consideration of STEM-EDS X-ray microanalysis of small precipitates illustrates the compro- mise between reducing spatial resolution and improving counting statistics associated with increasing the elec- tron probe size. For given experimental conditions of foil thickness tl, precipitate di amet er (d), position ( t ' ) and segregated element concentration (Co oc R0), electron accel- erating voltage (E0), effective flux intensity in the probe (K' ), and recordi ng time (T) there exists an opt i mum incident electron probe diameter (B*) for detection of the particular element of interest within the X-ray spectrum recorded from that precipitate. For iron base foils, electron scattering within the foil, characterized by the term /3 in Eqs. [2] and [3], results in optimum electron probe di- ameters in the range 10 to 60 nm for the standard conditions considered in the present work. For precipitates located at the center of a foil of thickness >- 100 nm the optimum probe size is >-20 nm regardless of the "effective" electron flux parameter K'T. This beam diameter is larger than usually obtained in current STEM instruments when operating in the 1.o 0 , i i i I i I 1 [ i i i I I I i l l I i I i i i i 1 [ i - 4 - 3 - 2 - I 1o9 K'T Fi g. 5 - - Ca l c u l a t e d pl ot s o f mi n i mu m pr e c i pi t a t e d i a me t e r , d*, wh i c h ma y be de t e c t e d i n t he X- r a y s p e c t r u m as a f u n c t i o n o f t he e l e c t r o n f l ux, K' T ( nm it/3) f o r R0 = 0, 1 a n d 1, t~ = 2 0 0 n m a n d Ro = 10 wi t h t~ = 100, 200, a n d 3 0 0 nm. C METALLURGICAL TRANSACTIONS A VOLUME 14A, OCTOBER 1 9 8 3 - 1 9 4 7 normal imaging mode but may be achieved by overriding the first condenser lens current controls or selection of aper- ture size. It is noteworthy that the mi ni mum segregated precipitate diameter in these cases is smaller than the elec- tron probe size which is required for their detection in the X-ray microanalysis spectrum. The "effect i ve" electron in- tensity paramet er K' is proportional to the electron source brightness. Our JEOL JEM 100CX STEM combi ned with a Link Model 860 EDS microanalysis system has a value for K' of - 5 x 10 -5 nm -H/3 s -j for iron K~ X-radiation using a conventional thermal hairpin tungsten filament at 100 kV (evaluated using Eq. [8] and the definition for Nb shown schematically in Figure 2). Thus, for a foil thickness of 200 nm, X- r ay recordi ng t i me of 200 seconds (K' T = 10-2nm-ll/3), and a 10 nm diameter electron probe we ob- tain a background count Nb = 1000 counts, an integrated peakNp = 4 x 104 counts (R = 40 for FeK~) with an over- all X-ray count rate for the total spectrum of 1000 s -~. The maxi mum count rate which can be accommodat ed by such an EDS system is - 2 0 0 0 s -~ for a processing time of 40/ zs. Thus, the maxi mum beam diameter which can be efficiently utilized for microanalysis of iron base foils of this thickness is =13 nm. The calculated mi ni mum precipitate diameter detected in the X-ray spectrum vs incident electron probe size curves for the standard conditions of Figure 4 and K' T = 5 x 10-~,5 x 10-2,5 x 10-3, and5 x 10 - 4 nm -11/3 are shown in Figure 6. For K' T = 5 x 10 - 3 nm -H/3 char- acteristic of our JEOL system, we find a mi ni mum diameter of 19 nm when limited by the X-ray count rate that is ac- cepted by the X-ray spectrometer. For higher values of K' T associ at ed wi t h br i ght er el ect r on sour ces, i . e. , fi xed T = 100 seconds, the X-ray count rate limitation further reduces the maxi mum electron probe size which may be used for X-ray excitation. These limits are superimposed as broken lines on the curves of Figure 6 for count rate limits of 2000 and 10,000 s ~. Clearly, for higher K' T values the finite X-ray count rate significantly increases the mi ni mum diameter, d*, f r om the t heoret i cal l y calculated values. Typically, a value of K' T = 5 x 10 -~ nm -H/3, which is 100 times greater than that attainable from the present JEOL tungsten thermal filament, represents an upper limit which might be achieved with field emission illumination systems. This would reduce the mi ni mum diameter to 15 nm com- pared with a theoretical d* of 7 nm and the 19 nm using the conventional tungsten filament system. Increasing the maxi- mum acceptable X-ray count rate to 10,000 s -1, however, decreases d* to 11 nm for this value of K'T. Thus, any attempt to i mprove the sensitivity of X-ray microanalysis by increasing K' through i mprovement s in electron flux in- tensity within the probe must be accompanied by a corre- sponding i mprovement in the maxi mum X-ray count rate which may be accommodat ed by the EDS system. Although the results described in this work have been developed for iron base foils, the overall principles apply to other alloy systems and the conclusions concerning the interaction between precipitate diameter, electron probe size, "effective" electron intensity, and maxi mum X-ray count rate apply even to low atomic number elements such as aluminum base and K' T values up to 100 times that achieved using conventional thermal tungsten filaments and accelerating voltages of up to 300 kV. In the present work 130 O 1 0 0 1 2 log d Fig. 6 - - Ca l c u l a t e d pl ot s of t he el ect ron pr obe size r ange requi red for X-ray detection of precipitate cont ai ned wi t hi n a t hi n foil base foil as a f unct i onof pr eci pi t at e di amet er f or val ues of K' T = 5 10 ' , 5 10 2, 5 x 10 3, and 5 x 10 -4 nm - " / 3 usi ng t he st andard condi t i ons for Fig. 3. Super i mposed on t hese cur ves are br oken lines s howi ng the ma xi mum pr obe size whi ch ma y be used whe n l i mi t ed by the X- r ay det ect or count rate of 2000 and 10,000 s ~. we have used a simple analytical approxi mat i on to describe the electron flux distribution within the total vol ume of the thin foil, Eq. [2], and a specific analytical formulation for the total electron flux as a function of electron probe size has been assumed (Eq. [7]). Regardless of the absolute accu- racy of these descriptions, the overall conclusions of the theoretical analysis depend only on an increase in electron probe current which is greater than a linear function of probe size. V. CONCLUSI ONS The principal conclusions which may be derived from the results of the present work are: 1. For precipitates which are contained within a thin foil there exists an opt i mum electron probe size for their detection in their X-ray spect rum obtained using STEM- EDS X-ray spectroscopy. 2. For precipitates contained within iron base foils of thick- ness ->100 nm the opt i mum el ect ron pr obe size for their detection in the X-ray spectrum is greater than 10 nm diameter. 1948--VOLUME 14A, OCTOBER 1983 METALLURGICAL TRANSACTIONS A 3. For EDS systems, where the saturation X-ray count rate is -<10,000 s -l , the optimum electron probe size for detecting segregations in the X-ray spectrum of small precipitates in iron base foils produces sufficient counts to saturate the spectrometer such that in practice a smaller probe size must be used. Thus, experimental conditions are optimized by using the largest electron probe size which just saturates the EDS to produce an acceptable maximum X-ray count rate. ACKNOWLEDGMENT This paper is published by permission of the Director General, Central Electricity Board, South Eastern Region. REFERENCES 1. P. Doig, D. Lonsdale, and P. E. J. Flewitt: Proceedings of "Quantitative Microanalysis with High Spatial Resolution", 1981, Manchester, Inst. Met. Book 277, p. 41. 2. R Doig, D. Lonsdale, and P. E. J. Flewitt: Scripta Met., 1982, vol. 16, p. 201. 3. D. C. Joy: Proceedings 6th Annual SEM Symposium, IITRI, 1973, Chicago, IL, O. Johari, ed. , p. 743. 4. D. C. Joy: "Advances in Analysis of Microstructural Features by Elec- tron Beam Techniques", 1974, Met. Soc. Inst. Phys. London, p. 20. 5. D. M. Poole and P. M. Martin: Met. Rev., 1969, vol. 14, no. 133, p. 61. 6. P. J. Goodhew and D. Chescoe: Micron, 1980, vol. 11, p. 153. 7. M. Green and V. E. Cosslett: Proc. Phys. Soc., 1961, vol. 78, p. 1206. 8. N. J. Zaluzec: Proceedings of "Ninth International Congress on Electron Mi crocopy", J . M. Sturges, ed. , Microscopical Society of Canada, Toronto, 1978, pp. 548-49. 9. J . I . Goldstein, D. E. Newbury, P. Echlin, D. C. Joy, C. Fiori, and E. Lifshin: Scanning Electron Microscopy and X-ray Microanalysis, PLenum Press, 1981, pp. 202-31. METALLURGICAL TRANSACTIONS A VOLUME 14A, OCTOBER 1983-- 1949