Atomic Design of Polarity of GaN Films Grown on SiC(0001)

Commun. Theor. Phys. (Beijing, China) 41 (2004) pp. 609 613 c International Academic Publishers Vol. 41, No. 4, April 15, 2004 Atomic Design of Polarity of GaN Films Grown on SiC(0001) DAI Xian-Qi, 1,2, WU Hua-Sheng, 2 XU Shi-Hong, 2 XIE Mao-Hai, 2 and S.Y. Tong 3 1 Physics Department, Henan Normal University, Xinxiang 453002, China 2 Physics Department, University of Hong Kong, Hong Kong, China 3 Department of Physics and Materials Science, City University of Hong Kong, Hong Kong, China (Received July 8, 2003) Abstract Ab initio total energy calculations are used to determine the interface structure of GaN films grown on 6H-SiC(0001) with different substrate reconstructions. The results indicate that GaN films grown on bare SiC(0001) are of the Ga-polarity, while GaN films grown on SiC(0001) with Si adlayer are of the N-polarity if there is no N-Si interchange at the interface. With the interchange, the GaN films are of the Ga-polarity. PACS numbers: 68.35.Ct, 61.72.Nn, 68.55.-a, 81.15.Aa Key words: polarity, total energy calculation, adlayer, interface The lattice polarity of III-nitride epitaxial films grown on different substrates is an important topic because the polarity exerts great influence on the structural, optical, and electrical properties of the film. [1] Specific properties influenced by polarity include surface morphology, chemical behavior, luminescence spectra, the formation of the metal/gan interface, and AlGaN/GaN heterostructures. [2] Most devices are built on the polar basal plane of wurtzite GaN, and the characteristics of these devices depend on whether the GaN film has the Ga- or N-polarity. [3] GaN films of Ga-polarity are found to exhibit different properties from those of N- polarity. It has been widely believed that Ga polarity films are obtained by metalorganic chemical-vapor deposition (MOCVD) grown on saphire. Recent MBE-grown GaN films on SiC, however, show different polarities depending on the substrate condition. While the film has an N-polarity on the 6H-SiC(000 1) substrate, the polarity behavior ( ) on 6H-SiC(0001) is more complicated. On the 3 3 R30 0 SiC(0001) substrate, GaN films are found to exhibit the Ga polarity. [4 6] On the 3 3 SiC(0001) substrate, however, the polarity of the GaN films is less certain. It is the objective of this work to investigate the polarity behavior by the first-principles total energy calculations. We determine the total energies of a number of possible interface structures of GaN on 6H-SiC(0001). From these calculations, the polarity as well as the most favored structure on different substrates is determined. The total energy calculations are performed using the Vienna Abinitio Simulation Package (VASP), which uses local density functional theory and the first-principles pseudopotential methods. [7] This package has been successfully applied to III-nitride systems. [8] In our calculations, the Ga 3d orbitals are included in the valence band, the interaction of which with the atomic cores is treated by non-normconserving ab initio Vanderbilt pseudopotentials. [9] The total energy is minimized by optimizing atomic positions using a conjugate-gradient method. To model GaN films grown on different 6H- SiC(0001) substrates, a supercell scheme is used. The supercell contains eight bi-layers of SiC and several layers of GaN, depending on the model. The dangling bonds of C atoms on the backside of the supercell are saturated by hydrogen atoms. Substrate structural parameters used in the calculation are obtained from optimization of bulk 6H-SiC(0001). The atoms in the GaN layers and the top bilayer of SiC are allowed to relax along the c-axis. Approximately 10 Å of vacuum between the GaN layer and hydrogen atoms is included to prevent interaction between surfaces. Figure 1 shows the models in the clean 6H-SiC system. In panel (a), a Ga-polarity film is shown with N atoms bonded at on-top sites above the surface Si atoms. Panel (b) shows an N-polarity film with Ga atoms bonded at on-top sites. The calculations show that the total energy of model (a) is 1.06 ev lower than that of model (b). This is a significant amount of energy difference and the result indicates that the Ga-polarity is definitely preferred on the clean substrate. This result is consistent with the earlier experimental [4,5] and theoretical findings. [10] The optimized structural parameters are as follows. For model (a), the N-Si bond length is 1.75 Å, and the vertical separation between the Ga and N layers is 0.85 Å. These values agree well with the experimentally determined interplanar separation, which ranges between 2.4 to 2.65 Å in the vicinity of the interface. [11] The Si atoms beneath the N layer are at near bulk positions. The top Ga atoms The project supported by Hong Kong Research Grants Council under Grant Nos. N CityU 7281/00P, CityU 7396/00P, HKU 7095/01P, and HKU 7096/01P Correspondence author, E-mail: xqdai@henannu.edu.cn

610 DAI Xian-Qi, WU Hua-Sheng, XU Shi-Hong, XIE Mao-Hai, and S.Y. Tong Vol. 41 in Fig. 1(a) are found to occupy the B registry, which has an energy 0.01 ev lower than if they were to occupy the C registry. Fig. 1 Structural models for the interface of GaN layers on the ( 3 3 ) R30 0 SiC(0001) substrate. The models are shown in the (11 20) projection. (a) Ga-polarity with formation of N-Si bonds; (b) N-polarity with formation of Ga-Si bonds at the interface. Fig. 2 Structural models for the interface of GaN layers on the 3 3 SiC(0001) substrates. The models are shown in the (11 20) projection. (a) N-polarity with formation of N-Si bonds; (b) Ga-polarity with formation of Ga-Si bonds at the interface; (c) An additional layer of N and Ga atoms on top of model (a). Figure 2 shows models for the Si-adlayer SiC(0001) substrate. In panel (a), an N polarity model is shown, where N atoms are bonded at 3-fold sites on the top Si substrate layer. Panel (b) shows a Ga-polarity model, where Ga atoms are bonded to the top Si layer instead. The calculations show that the total energy for model (a) is significantly lower, by 6.00 ev. This indicates that the GaN film strongly prefers to form the N-polarity on the Si-adlayer SiC(0001) substrate. This result is in agreement with recent experimental findings. [12] The extra Si layer on the surface inverts a Ga-polar face to an N-polar face, just like the role played by an Mg layer in the MBE epitaxial growth of wurtzite GaN. [13] The optimized structural parameters determined by the calculations are: for model (a), the Si-Si bond length is 2.35 Å, which is the same as in bulk Si. The Si-N bond length is 1.85 Å and the vertical separation between the N and Si layers is 0.58 Å. The Ga-N bond length is 2.26 Å, which is considerably

No. 4 Atomic Design of Polarity of GaN Films Grown on SiC(0001) 611 longer than the typical Ga-N bond (1.90 Å) in bulk GaN. The Ga-N atoms are found to prefer the B registry, which is lower in energy by approximately 0.01 ev than the C registry. It is interesting to examine the energy differences of E 1 = 1.06 ev and E 2 = 6.01 ev respectively for the models shown in Figs. 1 and 2. The electronegativity of N is much higher than that of either Ga or Si, so the N-Si bond in model (a) of both figures has a much lower energy than the Ga-Si bond in model (b) of both figures. Furthermore, while there is only one N-Si bond vs. one Ga-Si bond in the models of Fig. 1, there are 3 N-Si bonds vs. 3 Ga-Si bonds in the models of Fig. 2. Thus, the energy difference E 2 is expected to be much larger than E 1. The fact that E 2 is greater than 3E 1 shows that while nearest neighbor bond energies provide the correct trend, further neighbor bond energies are important in determining the total energy of the system. Figure 2(c) shows the site registry when an additional bilayer is added on the N-polarity model of Fig. 2(a). The second bilayer is found to prefer the A-registry, which is 0.08 ev lower in energy than the C-registry. Additional bilayers occupy successively the B and A registries, forming a BABA wurtzite film with the N-polarity. These results indicate that if the GaN film is grown by adsorption of N atoms on Si dangling bonds at the interface, then the film will have the N-polarity as shown in Fig. 2(c). Fig. 3 Four possible models for the interface of GaN layers on the 3 3 SiC(0001) substrate with N-Si interchange at the interface. The models are shown in the (11 20) projection. The 6H-SiC(0001) surface consists of a hexagonal Bravais lattice with a close packed stacking (ABCACB ) along the c-axis. For the ( 3 3 ) R30 0 SiC(0001) reconstruction, there are 1/3 monolayer (ML) of Si adatoms on symmetry-allowed T 4 sites of Si-terminated SiC. For the 3 3 SiC(0001) reconstruction, there is an Si tetramer (Si trimer + Si adatom) with an extra Si adlayer on Siterminated SiC. [14,15] Because the 1/3 ML of Si adatoms in the ( 3 3 ) R30 0 structure and the Si-tetramers in the 3 3 structure are easily removed during the MBE growth

612 DAI Xian-Qi, WU Hua-Sheng, XU Shi-Hong, XIE Mao-Hai, and S.Y. Tong Vol. 41 process, we assume that these atoms are not present in the models for the interface. It might indicate that the GaN grown on 3 3 substrate is of N-polarity, and the GaN grown on ( 3 3 ) R30 0 substrate is of Ga-polarity. There is, however, the possibility that the N atoms may replace the top layer Si, resulting in a different structure. In Fig. 3, we show four such possible models. In Figs. 3(a) and 3(b), the vertical Ga-Si pair occupy the B registry, with either Si or Ga bonded respectively to the N atoms. In panels (c) and (d), the vertical Ga-Si pair occupy the C registry, with Si or Ga bonded respectively to the N atoms. The calculations show that the total energy is the lowest for the model shown in Fig. 3(c), which is 0.35 ev lower than that of model 3(d) and 0.05 ev lower than the model shown in Fig. 3(a). The models 3(b) and 3(d) differ in energy by less than 0.01 ev. More important is the fact that the model in Fig. 3(c) is lower in energy by 0.07 ev than that in Fig. 2(a). These results merit two observations. (i) If there is sufficient energy for the N atom to break the surface Si-Si bond, then the most energetically favored model is that shown in Fig. 3(c). (ii) The fact that the total energy is much higher when Si atoms are placed at the topmost layer as in Figs. 3(b) and 3(d) indicates that Si atoms are not likely to segregate as a surfacant the way that Mg atoms do in GaN films. This may be the reason why one rarely finds any trace of Si atoms at the surface of GaN films growth on SiC substrates. Fig. 4 Adding a layer of N atoms to the model shown in Fig. 3(c). (a) The N atoms adsorb at 3 fold-sites on top of Ga atoms; (b) The N and Ga surface atoms interchange in positions. The N-Si interchange model of Fig. 3(c) has the lowest energy. The follow-up question is: What is the polarity of the GaN film if such an interchange indeed takes places? In Figs. 4(a) and 4(b), we consider such two models: (a) The direct adsorption of an additional layer of N atoms at either the A or B registry on top of Ga atoms in Fig. 3(c), resulting in an N-polarity film, and (b) The additional N layer interchanges with the Ga layer, with the Ga atoms occupying either the A or B registry, resulting in a Ga-polarity film. The total energy results indicate that registry A is preferred in both cases. The energy at registry A is 0.07 ev lower than that at registry B for Fig. 4(a) and 0.02 ev lower for Fig. 4(b). Comparing the models shown in Figs. 4(a) and 4(b), the model in Fig. 4(b) has a lower energy by a large amount, i.e. 0.41 ev. This means that the GaN film will have the Ga-polarity, if the N-Si interchange does take place at the interface. In conclusion, first principles total energy calculations of the initial configurations of GaN layers grown on 6H- SiC(0001) with different reconstructions show that the polarity of the GaN film depends on the structure of the substrate. The results indicate that GaN films grown on clean substrate is always of the Ga-polarity. On the other hand, GaN films grown on Si-adlayer substrate has the N-polarity if there is no N-Si interchange at the interface. This may be the likely scenario for growth at low temperatures. In the case that the N-Si interchange does take place at the interface, the calculations show that the GaN film will have the Ga-polarity. This case may correspond to growth at higher temperatures. These results indicate that polarity engineering of GaN films may be achieved through either substrate reconstruction or temperature variation.

No. 4 Atomic Design of Polarity of GaN Films Grown on SiC(0001) 613 References [1] A. Gassmann, T. Suski, N. Newman, C. Kisielowski, E. Jones, E.R. Weber, Z. Liliental-Weber, M.D. Rubin, H.I. Helava, I. Grzegory, M. Bockowski, J. Jun, and S. Porowski, J. Appl. Phys. 80 (1996) 2195. [2] J.F. Schetzina, Mater. Res. Soc. Symp. Proc. 395 (1996) 123. [3] Z. Liliental-Weber, J. Washburn, K. Parula, and J. Baranowski, Microsc. Microanal. 3 (1997) 436. [4] R.F. Davis, T.W. Weeks, Jr., M.D. Bremser, S. Tanaka, R.S. Kern, Z. Sitar, K.S. Aliey, W.G. Perry, and C. Wang, Mater. Res. Soc. Symp. Proc. 395 (1996) 3. [5] S.M. Seutter, M.H. Xie, W.K. Zhu, L.X. Zheng, H.S. Wu, and S.Y. Tong, Surf. Sci. 445 (2000) L71. [6] M.H. Xie, S.M. Seutter, W.K. Zhu, L.X. Zheng, H.S. Wu, and S.Y. Tong, Phys. Rev. Lett. 82 (1999) 2749. [7] G. Kresse and J. Hafner, Phys. Rev. B47 (1993) 558; G. Kresse and J. Furthmuller, Comput. Mat. Sci. 6 (1996) 15; Phys. Rev. B54 (1996) 11169. [8] U. Grobner, J. Furthmuller, and F. Bechstedt, Appl. Phys. Lett. 74 (1999) 3851. [9] D. Vanderbilt, Phys. Rev. B41 (1990) 7892. [10] R.B. Capaz, H. Lim, and J.D. Joannopoulos, Phys. Rev. B51 (1995) 17755. [11] J.N. Stirman, F.A. Ponce, A. Pavlovska, I.S.T. Tsong, and D.J. Smith, Appl. Phys. Lett. 76 (2000) 822. [12] B. Yang, A. Trampert, B. Jenichen, O. Brandt, and K.H. Ploog, Appl. Phys. Lett. 73 (1998) 3869. [13] V. Ramachandran, R.M. Feenstra, W.L. Sarney, L. Salamanca-Riba, J.E. Northrup, L.T. Romano, and D.W. Greve, Appl. Phys. Lett. 75 (1999) 808. [14] K. Reuter, J. Bernhardt, H. Welder, J. Schardt, U. Stark, and K. Heinz, Phys. Rev. Lett. 79 (1997) 4818. [15] U. Stark, J. Schardt, J. Bernhardt, M. Franke, K. Reuter, H. Welder, K. Heinz, J. Furthmuller, P. Kackell, and F. Bechstedt, Phys. Rev. Lett. 80 (1998) 758.