J.T. Armstrong, R.B. Marinenko (CSTL, Gaithersburg), J.G. Pellegrino (EEEL, Gaithersburg), and K.A. Bertness (EEEL, Boulder)

Objective: Develop analytical and correction procedures that will enable industrial users to perform routine electron microprobe analyses of the elemental compositions of AlXGa1-XAs thin layers on substrates like GaAs with a relative accuracy of better than 1%.

Problem: AlXGa1-XAs is a semiconductor system with very important properties for a variety of commercially-significant microelectronic and optoelectronic applications. To manufacture working devices that utilize these properties, the composition of the AlXGa1-XAs phase must be known with a precision of better than one percent relative and an accuracy of about one percent relative (e.g., X = 0.200 ± 0.002). Moreover, phases of economic interest encompass well over half of the compositional range in the AlAs-GaAs system. Electron probe microanalysis (EPMA) is one of the most commonly employed analytical procedures for characterization of these materials. However, the accuracy of EPMA in this system is restricted due to the atypically high degrees of characteristic fluorescence and x-ray absorption of some of the x-ray lines used in analysis, as well as the relatively large compositional differences between the analyzed phases and the typically available standards. Different laboratories analyzing the same phases in this system with a variety of commonly-employed commercial analysis programs, obtain compositions that differ from each other and differ from the correct compositions by as much as 20 to 30% relative (instead of the 2-3% relative accuracy often obtainable in EPMA). As a result, many laboratories use empirical corrections or calibration curves relative to internal laboratory reference compounds of uncertified (unknown) composition. Because of this, any inter-laboratory comparison of the relation of electrical or optical properties to composition is subject to serious error.

Approach: We have been working to characterize by EPMA a series of new standard reference materials (SRMs) in the system AlXGa1-XAs. Films were synthesized by molecular beam epitaxy with five different known compositions (with 'X' ranging from 0.1 to 0.63) by careful monitoring of the time evolution of the film properties during epitaxial growth. The compositions were determined by in-situ optical reflectance spectroscopy and ex-situ reflection high-energy electron diffraction (RHEED). The films were then analyzed by high precision EPMA at multiple accelerating potentials. The homogeneity of the films were confirmed and multiple sets of analytical data (including analysis of numerous primary standards) were collected to yield data sets with precisions better than 0.5% relative. We devised a new analysis scheme involving a combination of x-ray lines not typically employed to minimize the uncertainties and artifacts produced by secondary x-ray fluorescence. We determined that there were not enough internally consistent measurements of mass absorption coefficients (MACs) for Al, Ga, and As on which to base the various parameterizations of MACs currently used in commercial correction programs. We added a set of NIST-produced theoretical MACs that appear to be more self-consistent. The data was processed using all of the commonly-employed correction algorithms and the various sets of MACs. We determined the internal self-consistency of the measured ratios of intensities of samples to standards by the a-factor (calibration curve) method and determined a set of superior empirical a-factors for analyses in this system.

Results and Future Plans: The results obtained by processing the same data set with the various combinations of MACs and correction algorithms varied by over 30%. However, when the NIST MACs and the NIST/Caltech "CITZAF" correction algorithms were used, the calculated compositions of all five films agreed with the reflectance/RHEED data within 2% relative (better than 1% relative for four of the five samples). The same degree of agreement is obtained using the best-fit empirical a-factors. This suggests that the level of accuracy required by industry can be obtained using either of these two procedures. Further testing of the accuracy of these procedures is planned as well as a study of III-V semiconductor compounds in the more complex system (Al, Ga, In) (As, Sb, P).

Web Contact micro@nist.gov