Characterization of SiGe Bulk Compositional Standards with Electron Probe Microanalysis R. B. Marinenko*, J. T. Armstrong*, S. Turner*, E. B. Steel*, and F. A. Stevie** ^National Institute of Standards and Technology (NIST), MS 8371, Gaithersburg, MD 20899 **Analytical Instrumentation Facility, Room 318, EGRC Box 7531,1010 Main Campus Drive, North Carolina State University, Raleigh, NC27695-7531 Abstract. Bulk SiGe wafers cut from single-crystal boules were evaluated with the electron probe microanalyzer (EPMA) for micro- and macroheterogeneity for use as primary standards for future characterization of SiGe thin films on Si that are needed by the microelectronics industry as reference standards. Specimens with nominal compositions of 14 at. %, 6.5 at. %, and 3.5 at. % Ge were rigorously tested with wavelength dispersive spectrometers (WDS) using multiple point, multiple sample, and duplicate data acquisitions. The SiGe 14 is a good bulk reference material for evaluation of SiGe thin films. INTRODUCTION Silicon germanium technology is having a profound impact on the wireless and computer industries due to the development of electronic components that are smaller, faster, less noisy and require less power than conventional silicon technology [1]. This trend is expected to continue since the fabrication cost of silicon germanium devices is nearly as inexpensive as silicon devices, while little retooling is required. Characterization of the SiGe thin films used in the new devices has therefore become necessary. NIST is working with the semiconductor industry to use electron probe microanalysis (EPMA), Auger analysis, analytical electron microscopy (AEM), and secondary ion mass spectrometry (SIMS) to determine the composition, extent of heterogeneity, and thickness of industry-produced SiGe thin films on Si. These films will become available to laboratories for characterization as part of a NIST "Interactive Materials" reference program designed to share materials and analytical results among microanalysis laboratories. From commercial sources we obtained several Si wafers coated with SiGe thin films of different thicknesses. An abundance of samples could be cut from any one wafer for distribution to the industrial community for use as reference standards after characterization of each wafer for Si and Ge compositions and evaluation of the extent of compositional and thickness heterogeneity. This paper discusses the first step in the development of this project - the evaluation of SiGe bulk specimens for use as primary standards for the microanalysis of these SiGe thin films on Si. EXPERIMENTAL SiGe specimens cut from single-crystal boules normal to the growth axis were obtained from Virginia Semiconductor. The nominal compositions were 3.5 at. % Ge in Si, 6.5 at. % Ge in Si, and 14 at. % Ge in Si. The specimens were mounted on brass disks with conductive carbon tape without carbon coating for testing with the electron probe microanalyzer (EPMA) using five wavelength dispersive spectrometers (WDS). EPMA uses a high energy, focused electron beam (up to 40 keV) to excite characteristic x rays in a specimen. The sample excitation volume is a sphere with a diameter of approximately 1 |im. It is therefore ideal for determining the extent of heterogeneity on the micro scale (point-to-point variability) and on the macro scale (sample-to-sample variability). For the certification of NIST Standard Reference Materials (SRMs) for microanalysis, this laboratory has developed analytical and statistical procedures for the testing the micro- and macroheterogeneity of flat, polished materials [2]. RESULTS AND DISCUSSION Initial Heterogeneity Testing The purpose of the initial heterogeneity testing of these specimens was to determine if any specimens CP683, Characterization and Metrology for VLSI Technology: 2003 International Conference, edited by D. G. Seiler, A. C. Diebold, T. J. Shaffner, R. McDonald, S. Zollner, R. P. Khosla, and E. M. Secula 2003 American Institute of Physics 0-7354-0152-7/03/$20.00 238 showed appreciable microheterogeneity. Line traverses and random sampling procedures were used. On each specimen two 40 point traverses normal to one another were run at two randomly selected locations with steps of 2 (im or 5 Jim between points. Such traverses are routinely run during heterogeneity testing to determine if distinct phases across short distances (5 jim to 100 (im) within specimens are present. Table 1 lists analysis parameters, average counts/20 s observed per traverse, and the la uncertainty observed for each sample for the traverse with the largest uncertainty. These data suggest that the SiGel4 specimen is the least heterogeneous while the SiGe3.5 is the most heterogeneous on the |im scale. Plots of x-ray counts vs. point sequence also confirmed that these three specimens exhibit different levels of heterogeneity. Random sampling tests on each specimen were run to evaluate the overall specimen heterogeneity. Each specimen was divided into rectangular sectors, or samples, excluding the extreme specimen edges. Four or five randomly selected points were sampled within each sector and x-ray counts were acquired three times on the same point for an acquisition time of 20 s or 40 s. Data from the five fixed spectrometers were acquired simultaneously and corrections were made for current drift (usually less than 1% relative). Data were processed with a nested design analysis of variance procedure used previously at NIST [2,3,4]. From the repeated readings taken from the randomly selected points within each sample (or sector), and from the known mass fraction of each element as well as the background counts on each spectrometer, variance components were calculated for between points (a 2), for between samples (a 2), and for the experimental uncertainty (a 2). The overall variance (o 2) in Ew W mass fraction due to the heterogeneity contribution was calculated for each group of analyzed samples or sectors as, = U (T 2 + Urr 2 + Sw Pw (1) the sum of the between specimens, between points, and experimental variances. Table 2 lists the expanded uncertainties (3a, where a = V aw2 ) in % mass fraction determined from data for each specimen - they are 1 % to 2 % relative for Si and Ge in SiGel4 and for Si in SiGe6.5, but for Ge in the SiGe6.5 they are 2 % to 3.5 % relative. For the SiGe3.5 these uncertainties are 4.5 % relative for Si and as much as 15 % relative for Ge. The consistently higher uncertainty for Ge is due to the much lower counting statistics observed for Ge compared to Si, due in part to the lower concentration of this element in the specimens and the lower spectrometer efficiency for Ge. SiGe6.5 was is two large pieces, designated "top" and "bottom." Data taken from the top portion are shown in Fig 1 where the total x-ray counts acquired in 20 s are plotted vs. point number. Data from the three repeated readings taken from four randomly selected locations in each of the four sectors, totalled 48 points. The error bars are the expanded uncertainties (±3a) determined from Poisson counting statistics, where a is equal to the square root of the average number of counts/20 s. This specimen definitely shows heterogeneities between sectors as did the statistics in Table 2. A similar plot (not shown here) of data taken from the SiGel4 sample showed considerably less heterogeneity, while for SiGe3.5 there is considerably greater heterogeneity, as expected from Table 2. For this reason, the SiGe3.5 was excluded from subsequent heterogeneity tests. Since this was an initial test, the extent of heterogeneity of SiGe6.5 was not sufficient at this point to exclude it from further testing. Having five spectrometers with only two elements to analyze allows the acquisition of duplicate data on more than one spectrometer and/or the acquisition of data from two different x-ray lines of the same element as can be seen in Table 2 and Fig. 1. This type of information helps to sort out instrumental problems such as spurious noise signals that occasionally occur over the duration of an experiment from true compositional changes in the specimen. Throughout these analyses we used plots of x- Table 1. Uncertainties (la) for 40-Step Traverses Specimen Steps Element/Line/Crystal/Spectrometer (urn) SiKocTAPl GeKccLiF2 GeKccLiF3 SiKoPET4 SiGel4 2 Cts/20s o(rel %) 1862735 0.25% 2784659 0.51% 5 SiGe3.5 Cts/20s 3169570 a(rel %) 1.39% 40 pts/traverse, 20 kV, 80 nA, 20 s counting times SiGe6.5 5 Cts/20s o(rel %) GeLocTAPS 123457 134272 201235 444393 0.38% 62545 0.60% 43240 4.19% 0.40% 68594 0.67% 47606 4.09% 0.30% 109363 0.47% 74876 4.34% 0.31% 635956 0.40% 710021 1.09% 239 Final Heterogeneity Testing ray counts for a given element vs. acquisition sequence or plots of Si x-ray counts vs. Ge x-ray counts to visualize the consistency of the data, but only a few are presented here. At the voltage and current used in these initial tests, the count rate for SiKa on TAP1 was too high, swamping the counter with too many photons. Occasional spurious signals were also observed that could not be attributed to concentration changes. For subsequent tests the voltage and current were lowered to 15 kV and 60 nA, respectively, in hopes of reducing erratic signals. The SiGel4 and SiGe6.5 wafers were cleaved into smaller specimens, approximately 0.5 cm x 0.5 cm each. Five specimens from each were tested to determine if they were sufficiently similar for use as reference standards. Three were intended for the bulk quantitative analysis with ICP-OES (Inductively Coupled PlasmaOprtical Emission Spectroscopy) while two would be retained as primary reference standards. Table2. Extent of Specimen Heterogeneity (Initial Test) Dimension # Sectors, Expanded Uncertainty (3 a ) due to # Points, Specimen Heterogeneity, % mass fraction Specimen Sector Size # Repeats SiK SiK GeK GeK GeL (mm) Time/pt TAP1 LiF2 LiF3 TAP4 PETS SiGel4 0.21 0.16 0.39 0.33 20 x 11.4 12,4,3 0.31 5x3.8 20s SiGe6.5 0.84 15x8.3 1.28 0.75 4,4,3 0.51 1.11 Bottom 7.5x4.2 40s SiGe6.5 17.2 x 10.2 0.63 0.67 0.58 4,4,3 0.38 0.97 Top 7.9x5.1 20s SiGe3.5 35.4 dia 2.95 3.20 4,5,3 1.13 3.05 0.80 20s SKTAP1 . $!€ 8 W $$::;,.::•;.•:•; , X 4S SO Figure L Plots of x-ray counts as acquired and in sequence with the electron micropiobe with five fixed WD spectrometers. The.data were from three repeated from each Dffour randomly selected points in four different sectors of the top half of specimen SiOe6,5, 240 The Si and Ge data were acquired separately at different currents to get better counting statistics and spectrometer efficiency. The SiKa line was analyzed using both the TAP 1 and TAP5 crystals with the current lowered to 10 nA at 15 kV to avoid the high count rates that flooded the TAP1 counter in the initial heterogeneity test discussed above. Ge was tested in a second run using the TAP1 and TAPS crystals for the GeLa line, and the LiF2 and LiF3 crystals for the GeKoc line. A15 keV excitation potential and current of 60 nA was used for the Ge test. The instrument was attended constantly during the acquisition to be sure that the focus was maintained between specimens. Plots of all data points taken in sequence from the five specimens of each of the SiGel4 and SiGe6.5 samples and for the five repeated 7-point analyses of the Si and Ge wafers are shown in Fig. 2 (counts vs. point no). Corrections were made to the counts data for current changes during the experiment. The bars correspond to the ± 3 a uncertainty based on the average number of counts from Poisson counting statistics. The GeKoc plots are not included in Fig. 2 since they duplicate the GeLa plots at lower count rates. As expected from previous observations, the data from the SiGel4 specimens show consistency among points and among specimens and is similar to, though not quite as consistent as, the data taken from the Si and Ge wafers. The SiGe6.5 data shows less consistency among points and among specimens with points falling outside of the ± 3a uncertainty bars. Table 3 lists the expanded uncertainties, 3a, for the set of data taken at each of the SiKa and GeLa lines from each of the Si and Ge wafers, and the SiGel4 and SiGe6.5 samples. The value of a was calculated from the contributions to the heterogeneity cited in equation (1) above. These data confirm that the SiGel4 is a good material for use as a bulk primary reference standard. The heterogeneity of this specimen is only about 0.5 % mass fraction more than that observed for the pure Si and Ge wafers. The SiGe6.5 is obviously more heterogeneous, particularly for Ge where the expanded heterogeneity contribution is almost 5 % mass fraction. Data for the wafers do not include an uncertainty for SiKaTAPl because the between specimen variance was less than the between point variance, resulting in a negative number when the latter was subtracted from the former, terminating the calculations. Also, the value for GeLa on TAP1 is larger than for GeLa on TAP5 because a single outlier reading from one data point was not excluded from the calculations. Quantitative EPMA K-value data (ratio of x-ray counts from unknown to that of a standard) were also taken for Si and Ge from each of the five SiGel4 and SiGe6.5 specimens using the Si and Ge wafers as reference standards. The same randomly selected seven or eight points used to collect the x-ray counts data above from the SiGe specimens were used for the k-value data. The Si and Ge wafers were the reference standards for the determination of the k-ratios. Plots of the K-value data showed the same trends within and between specimens observed for the counts data in Fig. 2. These analyses provided numerous datapoints for quantification since a k-value is the first estimate of composition in the EPMA quantification process. Special efforts are needed to accurately determine the composition of Si-Ge alloys by quantitative electron microprobe analysis. The same experimental data, when processed through the various electron microprobe correction procedures currently in use by different laboratories can yield concentrations that vary by over 20 % relative. There is considerable uncertainty in the mass absorption coefficients for this system, particularly for SiKa x-rays by Ge. Using a multiple-keV correction evaluation procedure developed by our laboratory [5], we determined the best combination of mass absorption coefficients and correction procedures that correctly predicted the change in measured k-values with electron beam accelerating potential for the various SiGe alloys we analyzed. Using this correction, we determined that our best estimate of the composition of the nominal SiGe6.5 alloy is 6.7 at. %±0.1 at. % Ge and the composition of the nominal SiGe 14 alloy is 14.4 at. % +0.1 at.% Ge. These results will be augmented with bulk composition quantition using ICP-OES. CONCLUSION The SiGe 14 is represents a good microanalysis reference standard due to the low heterogeneity of this material, with the maximum expanded uncertainty due to heterogeneity no greater than 1.5% relative mass fraction for Ge and considerably less for Si. The SiGe6.5 is not as good as the SiGe 14 material with an expanded uncertainty due to Ge heterogeneity of about 5 % relative mass fraction. The SiGe 14 material will be used to set up a set of secondary standards of SiGe films on Si that will be useful for industry and become part of our Interactive Materials program. 241 REFERENCES [1] Subbanna, S., Myerson, B., Barnum, J.C., OConnell, T., and St. Onge.,S., Future Fab Internatl, Issue 11 (June 29, 2001). [2] Marinenko, R. and Leigh, S., "Heterogeneity Evaluation of Research Materials for Microanalysis Standards Certification," Micros. MicroanaL, Cambridge Press, New York (in review). [3] ISO Guide 35, Certification of reference material — General and statistical principles,2nd ed., Internal. Org. for Standardization, Switzerland (1989). [4] Neter, J., Kutner, M.H., Nachtsheim, C. J., and Wasserman, W., Applied Linear Statistical Models, 4th ed., Irwin/ McGrawHill, Chicago (1996) pp. 1121-47. [5] Armstrong, J. T., Marinenko, R. B., and Davis, J. M., "A Simple Method for Determining Optimum Corrections for High-Accuracy EPMA in Difficult Chemical Systems" in Micros. Microanal. ed. by Voelkl, E., et. al., Cambridge Press, 8, Suppl. 2, New York (2002,) pp. 438-9. Table 3. Uncertainty due to Specimen Heterogeneity Expanded Uncertainty (3 o) in % Mass Fraction Sample SiGel4 SiGe6.5 Si&Ge Wafers SiKcc TAP1 1.22 1.75 n/a SiKcc TAPS 1.29 1.55 0.94 GeLa TAP1 1.38 4.67 1.36 GeLa TAP5 1.55 4.64 0.60 Figure 2. Plots of x-ray count data simultaneously taken with the electron microprobe with two different TAP crystals, spectrometers 1 and 5, are compared for specimens SiGel4 (top), SiGe6.5 (middle), and the Si and Ge wafers (bottom). For the SiGe specimens 7 or 8 points were each read twice from five different samples. For the Si and Ge wafers, 7 points were each read twice at the beginning, in the middle, and at the end of the data acquisition time. All data are plotted in acquisition sequence. Vertical ticks at bottom of plots denote sample changes for the SiGe specimens, but for the Si and Ge wafers, they designate time breaks between repeated data acquisitions on the same 7 data points. 242
© Copyright 2025 Paperzz