 |
Fig. 1 Wafer reflectance (black) at 925 nm as a function of time during the growth of a layer of Al0.20Ga0.80As at 595 °C. Also shown are the results of an optimized fit (gray) and fit residuals. |
K. A. Bertness,** J. T. Armstrong, R. B. Marinenko, L. H. Robins,
A. J. Paul, J. G. Pellegrino, P. M. Amirtharaj, D. Chandler-Horowitz,
National Institute of Standards and Technology, **Mailstop 815.04, 325 Broadway, Boulder,
CO 80303
* Contribution of the National Institute of Standards and Technology and not subject to
copyright.
Accurate in situ measurements of composition are important for increasing the efficiency of epitaxial crystal growth processes. In situ measurements allow course-corrections in the operating parameters of crystal growth systems when used either as pre-run calibrations or for adjustments throughout actual device runs. This article illustrates another application— where in situ measurements are used to increase the amount of information available for determining composition with high absolute accuracy. The ultimate goal of this project is to produce standard reference materials (SRMs) of certified alloy composition to mole-fraction uncertainty of 0.002 for a range of important III-V alloys. Our approach is to apply several techniques of composition determination to a large number of specimens, identify and reduce sources of uncertainty in those techniques, and thereby bracket the composition determination on an absolute scale.
We describe here efforts to determine the composition of AlGaAs layers using in situ optical reflectance spectroscopy (ORS) data. Reflection high energy electron diffraction (RHEED) oscillations are used to independently determine the composition of the AlGaAs layers. The results are compared with ex situ measurements of photoluminescence (PL), photoreflectance (PR), x-ray rocking curves, and electron microprobe analysis (EMPA). Agreement among these techniques is presently converging to within 3 % for most specimens, and we expect that further improvements are possible by reducing source of error.
ORS data was acquired by reflecting monochromatic light at near-normal incidence from a semiconductor wafer mounted in the growth chamber of a molecular beam epitaxy (MBE) system. The optical access window was heated to minimize deposits, particularly arsenic. Additional details have been published elsewhere.[1] Before each growth run, the AlAs, GaAs, and AlGaAs growth rates were measured separately using the monolayer oscillations in the intensity of the zero-order beam in the RHEED pattern. The Al mole fraction was calculated by dividing the AlAs growth rate by the sum of the AlAs and GaAs growth rates, resulting in an estimated uncertainty of 0.004. ORS does not measure the composition directly; however, the growth rate measurements from RHEED and ORS are based on independent physical processes. Over a large sample set, agreement between RHEED and ORS growth rates indicates that the composition determined by RHEED has persisted throughout the run. The index of refraction derived from ORS can also be used to derive composition, although our present data set is insufficient for accurate determinations in this manner.
|
Fig. 1 Wafer reflectance (black) at 925 nm as a function of time during the growth of a layer of Al0.20Ga0.80As at 595 °C. Also shown are the results of an optimized fit (gray) and fit residuals. |
Typical ORS data during growth of a thick Al0.20Ga0.80As layer are shown in Fig. 1, along with a theoretical fit to the data using algorithms based on the virtual-substrate formalism.[2, 3] The variables optimized by the fit are the index of refraction and absorption coefficient of the layer, the growth rate, the effective index and absorption coefficient of the underlying layers, and an overall reflectance scaling factor.
 |
Fig. 2. Index of refraction at 925 nm and 585 to 600 °C, and ORS growth rate deviation from RHEED oscillation growth rate for a series of layers with different Al mole fractions. |
Figure 2 summarizes two important layer properties derived for a number of growths with different Al compositions. The growth rate determined from ORS fits was within 3 % of that from RHEED for most runs. Additional information is needed to determine whether the deviations result from variations in growth conditions or errors in the measurement techniques. There is a fairly large variation in the index derived for the set of runs with Al mole fraction near 0.20. Some variation originates with the difficulty the modeling encounters in distinguishing among changes in reflectance scaling, growth rate, and index of refraction. Deviations can also be introduced into the RHEED data because of flux transients and drift over time. RHEED is also acquired without substrate rotation, and therefore more subject to spatial variations in the flux. Finally, this set also comprises runs with different doping levels, doping types and growth temperatures (from 585 °C to 600 °C), all of which can alter the index.
 |
Fig 3. Comparison of in situ and ex situ composition determinations. |
In figure 3, we compare the in situ composition measurements with the conventional ex situ measurement techniques of PL, PR, x-ray diffraction rocking curves, and EMPA. More specimens are needed to improve the statistical significance of the data, but in this preliminary set, we see that the in situ measurements agree with each other and with the EMPA data well. The EMPA data were analyzed using procedures developed specifically for this project.[4] Some sources of error that we have begun to investigate further are the effects of sample temperature and doping variations in PL and PR-based composition measurements. The data also show that x-ray rocking curves lose accuracy when the substrate peak and epilayer peak begin to overlap significantly at low mole fraction. The overlap complicates the curve-fitting used to determine peak separation, and also makes the conversion of strain to composition more sensitive to small offsets introduced by doping, low-level impurities, or substrate misorientation.
In conclusion, we have found that in situ techniques agree with ex situ conventional methods of determining alloy composition in AlGaAs to within 3% for most of the specimens studied to date. We expect that the uncertainties can be reduced to the point where we can certify composition to 1% to 2% uncertainty, depending on the mole fraction.
1. K. A. Bertness, R. K. Hickernell and S. P. Hays, Mat. Res. Soc. Symp. Proc. (MRS, Pittsburgh, 1999), vol. 570, p. 157.
2. W. G. Breiland and K. P. Killeen, J. Appl. Phys. 78, 6726 (1995).
3. D. E. Aspnes, J. Opt. Soc. Am. A 10, 974 (1993).
4. J. T. Armstrong, Microscopy and Microanalysis 4 (suppl. 2), 226 (1998).