Chloride Vapor Phase Epitaxy (ChVPE) of GaN using GaCl and ammonia as the group III and V precursors, respectively, is a promising technique for growth of thick and quasi-bulk GaN epilayers. ChVPE combines high crystal growth rates with a low defectiveness, which makes it very attractive for industrial production of GaN substrates for III-nitride heterostructures [1,2]. At once, the technique is now employed for growth of the (Al/Ga/In)N device heterostructures themselves [3,4].

In spite of the technique advantages, wide-ranging commercial use of ChVPE is currently hampered by some technological problems related, in particular, to non-uniform supply of the precursors to the wafer. Due to high reactivity of the precursors, there is a trend for growth of parasitic polycrystalline GaN deposit on the reactor walls and occasionally for gas-phase nucleation of the reaction by-products. In order to avoid these harmful effects, the initial reagents are normally supplied separately into reactors and are intermixed only next above the wafer. This, in tern, may result in poor intermixing of the precursors and eventually in thickness and quality non-uniformities of the epilayers. Thus, the flow dynamics and species transport in reactors for GaN ChVPE become key factors affecting the process efficiency.

To simulate and optimize GaN ChVPE in reactors of realistic geometry, we have developed a predictive model of the process, coupling a detailed 3D description of the species transport and quasi-thermodynamic description of the surface kinetics. The transport part of the model accounts for the mass, momentum, heat, and radiation transfer while the surface kinetic part is based on the idea that adsorption and desorption represent the limiting stages of surface kinetics. The model has been carefully verified by comparison of the computed and experimental dependencies of the growth rate on the process parameters.

Examples of the model verification are presented in Figs. 1 and 2. Fig. 1 illustrates, in particular, good agreement of the computed and experimental data on the GaN growth rate vs. temperature in the atmosphere of He or H2 (data of Seifert et al [5]). Agreement of the computed and experimental data on the growth rate as a function of temperature and species flow rates are shown in Fig. 2 (data of Ilegems et al [6]). It is seen from the figures that computations quantitatively reproduce the experimental data except for an ambiguous dependence of the growth rate on the H2 flow rate (Fig. 2d). The latter effect seems to be related to a considerable scatter of the N2 sticking probability on GaN and to possible convective flow instability. The computations reproduce, in particular, the linear and non-linear dependencies of the growth rate on the GaCl and NH3 flow rates, respectively, and attribute them to a considerable contribution of the desorbed species fluxes.

Fig.1. Computed dependencies of the GaN growth rate on temperature
in the atmosphere of He (red) and H2 (blue)
compared to the data of Seifert et al [5].

Fig.2. Computed dependencies of the GaN growth rate on temperature (a)
and on the HCl (b), NH3 (c), and H2 (d) flow rates
compared to the data of Ilegems et al [6].

After verification, the model has been applied to simulation, analysis,and optimization of GaN ChVPE in reactors of different geometry. The results of computations for a horizontal single-wafer rotating-disk reactor are shown in Fig. 3. Here, the computed distributions of the species molar fractions over the rotating disk and upper susceptor surface are presented. The distributions are considerably non-uniform, which results in non-uniform “instantaneous” distributions of the growth rate and V/III ratio over the wafer (see Fig. 4). It is of interest that in spite of this non-uniformity, the corresponding angle-averaged radial distributions prove rather uniform (see Fig. 5), which favors the epilayer quality.

Fig.3. Distributions of the GaCl (a), NH3 (b), H2 (c), and HCl (d)
over the upper surfaces of the rotating disk and susceptor.

Fig.4. “Instantaneous” distributions of the V/III ratio and GaN growth rate
over the rotating disk.

Fig.5. Angle-averaged radial distributions of the V/III ratio and GaN growth rate.

References

1. R.J. Molnar, W. Gotz, L.T. Romano, and N.M. Johnson, J. Cryst. Growth 178, 147 (1997).
2. R. Cadoret and A. Trassoudaine, J. Phys.: Condens. Matter 13, 6893 (2001).
3. N. Takahashi, R. Matsumoto, A. Koukitu, and H. Seki. J. Cryst. Growth 189/190, 37 (1998).
4. Yu.V. Melnik, A.E. Nikolaev, S.I. Stepanov, A.S. Zubrilov, I.P. Nikitina, K.V. Vassilevski, D.V. Tsvetkov, A.I. Babanin, Yu.G. Musikhin, V.V. Tretyakov, and V.A. Dmitriev, in MRS Symposium Proceedings 482, 245 (1998).
5. W. Seifert, G. Fitzl, and E. Butter, J. Cryst. Growth 52, 257 (1981).
6. S.Yu. Karpov, A.S. Segal, D.V. Zimina, S.A. Smirnov, A.P. Sid’ko, A.V. Kondratyev, Yu.N. Makarov, D. Martin, V. Wagner, and M. Ilegems, to be published in MRS Symposium Proceedings 743 (2003).

Our publications on GaN ChVPE

1. S.Yu. Karpov, D.V. Zimina, Yu.N. Makarov, B. Beaumont, G. Nataf, P. Gibart, M. Heuken, H. Jurgensen, and A. Krishnan, Phys. Stat. Sol. (a) 176, 439 (1999).
2. S.Yu. Karpov, A.S. Segal, D.V. Zimina, S.A. Smirnov, A.P. Sid’ko, A.V. Kondratyev, Yu.N. Makarov, D. Martin, V. Wagner, and M. Ilegems, to be published in MRS Symposium Proceedings 743 (2003).