Electron Probe Microanalysis of Submicron Alloy Films

EPMA has been applied to electron beam evaporated and r.f. sputtered layers of about 200 tig/cm. This includes analysis of trapped argon and oxygen ranging from 0.2 to 10 wt%. For primary electron energies of 5 to 15 keV depth range of X-ray emission may be predicted with an accuracy of ± 10 nm. Errors of quantitative analysis are discussed for various correction models and should not exceed 3.5 %. I INTRODUCTION Electron probe microanalysis (EPMA) may be performed at a sufficiently low primary electron energy (Eg) for which the depth range of X-ray emission is reduced to about 0.2 to 0.5 um. This letter considers some practical aspects of EPMA under conditions of thin film analysis: Experimental depth ranges compared to the theoretical predictions, evaluation of analytical sensitivities, accuracy provided by various models of matrix correction, and simultaneous quantitative analysis of incorporated argon and oxygen. II DEPTH RANGE OF X-RAY EMISSION Experimental depth ranges have been studied for FeLa, AuMa, FeKa, and AuLa X-ray lines covering the range of critical excitation energies (Ec). from 0.7 keV to 12 keV. Pure element films of 0.05 to 0.5 nm were deposited on carbon by electron beam evaporation. The exact mass deposition was determined by chemical analysis. X-ray intensities normalized to a thick film standard were measured as a function of primary electron energy. The extrapolation of this curve to the saturation level at low EQ yields the primary electron energy for which the film thickness represents the depth range of X-ray emission. In Figure 1 experimental depth ranges are compared to those obtained by the range relation of Castaing [1]. The influence of mass absorption was calculated by the absorption factor of Philibert. For AuLa and FeKa this model gives an accurate representation of the measured depth ranges. For the X-ray lines with E c < 5 keV, Figure 1 shows larger deviations from the experimental data which may be explained by the strong influence of Xray absorption. Alternatively, theoretical depth ranges have been derived from depth distribution functions 0 (pz) calculated by the formula of Parobek and Brown [2]. Figure 1 shows an accurate fit to the experimental data for FeLa and AuMa, but significant deviations for Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:19842145 C2-622 JOURNAL DE PHYSIQUE Castaing's formula --i'arobek,Brown $,pz, 0 experimental' , electron energy I keV 1 Fig. 1 Experimental and calculated depth range of X-ray emission for the element targets of Fe and Au as a function of primary electron energy. lL' Ib io io h 40 $0 $0 do atomic number Fig. 2 Overvoltage and counting rate (crystal spectrometers) of some elements from C to Bi corresponding to a depth range of 0.3 pm. Data refer to a concentration of 30 wt% in a matrix of Fe. FeXa and AuLa (Ec > 5 keV). Within the limits pointed out, both models enable one to predict depth ranges of 0.2 to 0.5 pm with an error of about + 10 nm correspondinq to the error estimated for the determination of experimental depth ranges. A comparable depth range for X-ray lines separated by a large difference of the critical excitation energy may be obtained by measurements at two electron energies. I11 ANALYTICAL SENSITIVITY Analysis of thin films at low overvoltages (EO/Ec<2) or by low energy X-rays combined with low electron energies (EO<10 keV) conflicts with the requirements of analytical sensitivity. This concerns the ability to distinguish between two compositions that are nearly equal. The Xray intensity of 17 elements from carbon to bismuth was calculated by ZAF correction for the electron energy representing a linear depth range of 0.3 pm in a matrix of 70 wt% of iron (Fig. 2)\ This was based on measurements of the pure element or a well defined cbmpound by crystal spectrometers. A suitable choice of the X-ray line results in at least 1000 cps (beam current 100 nA) corresponding to 3 measurements, each at a counting time of 30 seconds, to obtain a sensitivity of 0.8%. This should be acceptable even for a large series of samples considering the stability of the applied computer-controlled instrument (CAMEBAX) . IV ACCURACY OF QUANTITATIVE ANALYSIS Atomic number correction may be a significant limitation to the accuracy of quantitative thin film analysis. This was studied for An-Cu alloys (NBS SRM 482) simulating a large difference of atomic number between specimen and pure element standard (AZ = 40). Conventional ZAF correction was performed by a commercial on-line program [ 3 ] using backscattering factors of Duncumb combined with the Bethe formula for the calculation of stopping power factors. Analytical errors of the applied ZAF model show a tendency towards overcorrection with decreasing EO (Fig. 3). Alternatively, the atomic number correction was performed by modified expressions of Love and Scott [ 4 ] for the backscatter factor R and the stopping power factor S. Figure 3 indicates improved accuracy for AuMa at electron energies of 5 to 10 keV. CorrecA u M a \ \ C u K u 10 .\ 20 wt % Au 20wt% Cu