Using optical reflectance to measure GaN nucleation layer decomposition kinetics

GaN nucleation layers (NLs) that are used in the two-step GaN growth process on sapphire partially decompose during the ramp from low to high temperature, T. In this paper, we show how the extent of this decomposition can be measured during each growth run using optical reflectance. Using reflectance, the NL decomposition rates were observed to increase with increasing P and T, and decreasing NH3 flow. Also, the NL decomposition rate did not change as the NL growth T was varied, but did increase as the NL thickness increased. The NL decomposition has an activation energy, E-A, of 2.7 +/- 0.2 eV for all annealing conditions studied, while the pre-exponential factor, A(0), depends strongly on the annealing conditions and initial NL thickness. Using the measured NL decomposition kinetics, an analytical model is used to fit the reflectance waveforms which also includes the T dependence of indices of refraction. The model was applied to several sets of reflectance waveforms where the annealing conditions were intentionally varied and the values of A(0) measured and compared. In addition to directly fitting the reflectance waveforms, we introduce another way to quantify decomposition rates measured under constant or varying T. This approach requires a constant E-A for the reaction kinetics throughout the entire T range of interest, which is the case for the GaN NL decomposition. For this approach, time is scaled by an integrated form of the Arrhenius exponential (i.e. exp(-E-A/k(B)T)). We have called this transformation of the time scale the "kinetic advancement," because it allows kinetic processes at constant or varying T to be compared directly. If the reflectance waveform or NL thickness is plotted vs. the kinetic advancement, a straight line results with the slope equal to A(0). The usefulness of both reflectance waveform fitting procedures is discussed, specifically how the reflectance waveform, can be compared and quantified during every growth run in order to verify NL evolution repeatability. (c) 2005 Elsevier B.V. All rights reserved.