Thin Solid Films 443 (2003) 9–13 GaN growth on single-crystal diamond substrates by metalorganic chemical vapour deposition and hydride vapour deposition P.R. Hageman*, J.J. Schermer, P.K. Larsen Department of Experimental Solid State Physics III, Mathematics and Computing Science, University of Nijmegen, Toernooiveld 1, Nijmegen 6525 ED, The Netherlands Received 26 November 2002; received in revised form 28 May 2003; accepted 7 June 2003 Abstract In this study a thick hexagonal GaN layer has been grown on a (110) single crystalline diamond substrate utilising two different deposition techniques. Using an AlN nucleation layer, metal–organic chemical vapour deposition (MOCVD) has been used to deposit an initial GaN layer on a (110) single crystal diamond substrate. The layer consists of closely packed GaN grains with a thickness of approximately 2.5 mm and with different orientations with respect to the substrate. Low temperature photoluminescence indicates a poor optical quality of the layer due to poor structural properties andyor a high incorporation of impurities. This layer was used as a template in a hydride vapour phase epitaxy (HVPE) growth experiment. As a result of this, the GaN grain size has increased enormously and the layer consists of large, hexagonal shaped pillars with a diameter of approximately 50 mm and a height of more than 100 mm protruding from a polycrystalline background having a more uniform thickness. PL spectra of this film show a strongly increased intensity of the exciton related emissions when compared to the MOCVD deposited film. X-Ray diffraction analyses revealed that the dominant orientation of the GaN crystallites perpendicular to the substrate changed from w001x for the thin MOCVD film to w112x for the HVPE layer. 2003 Elsevier B.V. All rights reserved. PACS: 61.82 Bg; 68.55 Jk; 78.30 Fs; 78.55 Cr Keywords: Chemical vapor deposition; Diamond; Epitaxy; Nitrides 1. Introduction diamond crystal lattice to that of silicon insinuates the possibility of depositing GaN on diamond. This is The development of nitride based technology for both strengthened by some recent reports about AlN growth optical and electronic applications has showed the last on diamond w2,3x. decade an enormous progress. Besides further improve- It is clear that lattice constants and thermal expansion ment of the material quality, specifically for electronic coefficients of substrate and film should match as close devices, addressing the thermal management of the as possible for hetero-epitaxial deposition to be success- substrate is required for performance improvement. ful. In Table 1 the lattice constants of GaN, sapphire, Si Although use of alternative substrates may be a solution and diamond are given together with their thermal for this, one has to bear in mind that the development expansion coefficients w4x. From this table it follows of device quality material on these substrates may be a that the lattice mismatch of GaN on diamond is 11.8%. difficult and time-consuming task. With this in mind, Although this is a considerable value, GaN of device growth of GaN on diamond substrates, especially type quality can be deposited on sapphire substrates using an IIa diamond w1x, may be attractive because it has the appropriate nucleation layer, in spite of the lattice highest thermal conductivity of all materials ()2000 mismatch of nearly 16.1% w5,6x between sapphire and Wym K at room temperature). The resemblance of the GaN. The use of an appropriate nucleation layer could *Corresponding author. Tel.: q31-24-3653158; fax: q31-24- overcome the problems associated with the lattice mis- 3652620. match between GaN and diamond. Nevertheless, to the E-mail address: email@example.com (P.R. Hageman). best of our knowledge, the growth of GaN on diamond 0040-6090/03/$ - see front matter 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0040-6090(03)00906-4 10 P.R. Hageman et al. / Thin Solid Films 443 (2003) 9–13 Table 1 graphic orientation of the deposits was determined from ˚ Lattice constant (A) and thermal expansion coefficient (=10y6 u –2u curves as obtained using a Bruker D8 Discovery Ky1) of GaN (a-axis), sapphire, diamond and Si w4x ˚ X-ray diffractometer with a Cu target (ls1.54060 A) Substrate GaN Sapphire Si Diamond and a 4-bounce monochromator Ge (022). The phase (a-axis) purity of the material was assessed by photolumin- ˚ Lattice constant (A) 3.189 4.758 5.420 3.567 escence (PL) measurements performed at 4 K with a Thermal expansion 5.59 7.5 4.70 4.38 HeCd laser (325 nm) as excitation source with power coefficient (=10y6 Ky1) densities up to I0s50 Wycm2 and incident at approxi- mately 308 to the normal of the sample surface. The PL emission was dispersed by a 0.6-m monochromator and has not been reported. However, diamond deposition on detected by a cooled GaAs photomultiplier. The spectral a GaN substrate was found to result in the growth of a resolution was 0.4 meV in the region from 3.2 to 3.55 limited number of individual grains only. This was eV. The set-up was calibrated by using a mercury lamp. argued to be a consequence of the poor wetting of diamond on GaN w4,7x. Depending on the growth 3. Results and discussions conditions a similar behaviour may be expected for the growth of GaN on diamond. MOCVD growth on the sapphire substrate resulted in In this paper we report on the growth of GaN on a a smooth 2.2 mm thick (001) GaN layer as described (110) single crystal diamond substrate with MOCVD in previous work w11x. In contrast with the deposition using an AlN nucleation layer w7x and subsequent of GaN on sapphire or silicon, where the substrate enlargement of the layer thickness by HVPE. The results determines the orientation of the epilayer, the GaN layer are compared with those of GaN layers, which were on the diamond substrate consists of differently oriented, grown on sapphire substrates. closely packed GaN grains. In Fig. 1 a SEM picture is shown of this approximately 2.5 mm thick continuous 2. Experimental procedure polycrystalline layer, which indicates a high nucleation density as was intended by the use of the AlN nucleation In the growth experiments type IIa (110) natural layer. The different orientations of the grains in the layer diamond and (001) sapphire crystals were used as are confirmed by X-ray diffraction measurements. The substrates. The diamond substrates were polished to X-ray u –2u curve of this layer, as shown in Fig. 3a, within 38 of the exact orientation and have a surface shows that that the GaN layer is hexagonal and that the roughness less than 20 nm. Immediately before growth use of a (110) diamond substrate has not resulted in the the diamond substrates were cleaned in boiling H2SO4 growth of a (110) GaN layer. In fact, the spectrum in with sodium nitrate, followed by an etch in HCly Fig. 3a shows a very strong w002x GaN peak, indicating H2SO4 (3:1) w8x. The nitride layers were grown in a that a relatively large fraction of the grains is aligned horizontal, low-pressure metal–organic chemical vapor with the w001x orientation perpendicular to the substrate. deposition (MOCVD) reactor using trimethylgallium This is eighter imposed by the (110) substrate, or the (TMG), trimethylaluminium (TMA) and ammonia result of evolutionary growth and selection of (001) (NH3) as precursors. A 10 nm thick AlN nucleation layer is deposited in a H2 carrier gas stream at 50 mbar at 850 8C using a TMA and a NH3 partial pressure of 4.9=10y6 bar and 1.0=10y1 bar, respectively w9x. After deposition of the nucleation layer, the temperature was raised to 1170 8C and GaN was grown for 1 h. Next the MOCVD pre-grown samples were used as templates for the growth of thick GaN layers using hydride vapour phase epitaxy (HVPE). The GaCl growth species were in-situ synthesised by passing a 30 sccm flow of pure HCl over liquid gallium (99.9999%) at 890 8C in a home-built reactor w10x. HVPE deposition took place for 1 h at a temperature of 910 8C. For the MOCVD growth of GaN on sapphire the procedure as described in w11x was used. After MOCVD as well as after additional HVPE, the surface morphology of the layers was examined using optical differential interference microscopy (DICM) and Fig. 1. SEM picture of a GaN layer deposited on a (110) single crystal scanning electron microscopy (SEM). The crystallo- diamond substrate using MOCVD. P.R. Hageman et al. / Thin Solid Films 443 (2003) 9–13 11 than 100 mm (see Fig. 2b) protrude from a more homogeneous background layer showing grains with a diameter of approximately 20 mm. The hexagonal struc- ture of the GaN is reflected in the appearance of these pillars. The X-ray u –2u curve measurements, as shown in Fig. 3b, indicate that the preferential alignment of the GaN crystallites with respect to the substrate is changed from (001) to (112) and has increased considerably. Also, it is an indication of the preservation of the hexagonal orientation. In GaN the w112x orientation is at an angle of 17.18 with the w001x orientation of the hexagonal top face of the pillar-shaped structures, which correlates well with the fact that the pillars are not perpendicular to the substrate (see Fig. 2). From this morphology it can be deduced that during HVPE the conditions are such that the highest growth rate is in the w112x direction. As a result those grains in the polycrystalline GaN template which are aligned with their w112x orientation perpendicular to the substrate start to protrude from their immediate surroundings. Under the high deposition rate gas phase conditions of HVPE the extended grains encounter a higher concen- tration of the Ga growth precursor. This further enlarges the growth rate of these particular grains while the differently oriented grains in the background more and more suffer from a GaCl depleted gas phase. Recently, a similar growth mechanism was reported to yield N001M Fig. 2. SEM pictures of a GaN layer deposited using HVPE on a template consisting of a (110) single crystal diamond substrate cov- ered by a MOCVD GaN film. (a) overview and (b) detailed image of the hexagonal, columnar structures protruding the more uniform background. GaN from initially randomly oriented GaN nuclei w12x. Since there is no obvious direct 1:1 relation between the diamond (110) surface and the GaN (001) surface, the latter mechanism is more likely. Besides reflections from GaN, the spectrum shows the reflections of w220x dia- mond originating from the single crystal (110) diamond substrate below the relatively thin GaN layer. The approximately 10 nm thick AlN nucleation layer is too thin to show up in the spectrum. As for the growth of GaN on Si(111) substrates w9x, the AlN nucleation layer should be further optimised with respect to its layer thickness and growth temperature to improve the GaN material quality. After HVPE growth, the thickness of the uniform GaN layer deposited on the MOCVD grown GaN layer on a sapphire substrate has increased to 127 mm. As shown in Fig. 2 the roughness of the polycrystalline GaN layer on the diamond substrate has increased Fig. 3. X-ray diffraction spectra (u-2u curves) of (a) the MOCVD enormously. Large, hexagonal shaped pillars with a GaN layer layer grown on a (110) diamond substrate and (b) after diameter of approximately 50 mm and a height of more enlargement of the GaN layer thickness using HVPE. 12 P.R. Hageman et al. / Thin Solid Films 443 (2003) 9–13 sapphire w14x. The most likely candidate for the impurity is carbon from the diamond substrate that incorporates in the MOCVD grown GaN film. Carbon acts as a deep acceptor and quenches the band edge emissions when present in GaN w15x. After HVPE growth the sample exhibits a very intense and detailed PL spectrum (see Fig. 4b). For the HVPE layer on a diamond substrate the intensity of the spec- trum is approximately 20 times higher than that of the initial MOCVD layer, thereby implying a considerable reduction of the amount of non-recombinative radiation centres andyor a large improvement of the structural quality. The peak at the high-energy side of the spectrum (at 3.472 eV) is attributed to excitons bound to a neutral donor (D0BE). This position confirms that the structure of the GaN layer is hexagonal and not cubic. In the latter case the D0BE peak is expected at a position of Fig. 4. Low temperature photoluminescence spectra of (a) the MOCVD GaN layers grown on (110) diamond and (001) sapphire 3.26 eV due to the reduced bandgap of cubic GaN substrates and (b) after enlargement of the GaN layer thickness using compared to hexagonal GaN w16x. HVPE. The position of the D0BE peak at 3.472 eV is almost identical with that found for a 400 mm thick HVPE textured diamond layers as a result of a gas phase grown, free-standing GaN layer w10x. This reveals that gradient of the growth precursor concentration w13x. in contrast to the HVPE layer grown on sapphire (D0BE For GaN growth on Si substrates, epilayers in excess at 3.449 eV), the layer on the diamond substrate is of 0.4 mm thickness were found to suffer from severe almost stress free. This agrees well with the absence of cracking due to the large tensile strain in the GaN layer cracks in the layer indicating that the thermal stresses formed when cooling down from growth temperature to are effectively accommodated at grain boundaries of the room temperature w9x. A similar result was obtained in polycrystalline layer. In the DAP recombination region the present work for the GaN layer as obtained by between 3.05–3.27 eV the zero-phonon peak is found HVPE on the sapphire substrate. For GaN growth on at 3.262 eV with a strongly reduced intensity compared diamond the difference in thermal expansion coefficients to the D0BE signal. At the low energy side two longi- is even larger than for GaN on Si or sapphire (see Table tudinal-optical (LO) phonon replicas are found at, 1). However, only a few cracks are observed in the respectively, 92 meV and 184 meV from the DAP peak, HVPE grown layer grown on the diamond substrate (see being one and two times the LO-phonon energy of GaN. Fig. 2a). As will be shown below, the PL data indicate The presence of these phonon replicas indicates a nearly that the HVPE GaN layer is almost stress-free. The fact perfect crystallinity of the individual crystallites. The that the GaN forms no closed layer and its crystallites are almost ‘free-standing’ ensures that the thermally difference in PL signature between the MOCVD and induced strain is effectively absorbed at the grain bound- HVPE grown GaN layer on the diamond substrate is aries of a polycrystalline GaN layer resulting in an explained by the fact that, during HVPE growth the almost stress-free layer. incorporation of carbon originating from the substrate, In Fig. 4 the PL spectra of the MOCVD grown GaN is eliminated by the initial GaN layer. layers on the diamond and the sapphire substrate are The broad band between 2.0 and 2.4 eV in the PL given. In contrast to the signal obtained from the spectra is known as ‘yellow luminescence’, and was reference sample on sapphire, the PL signal from the commonly detected regardless of the growth technique. MOCVD GaN layer on a diamond substrate has a very It is generally believed that this yellow luminescence low intensity (see Fig. 4a). This implies that the con- involves electronic states associated with intrinsic centration of non-radiative recombination centres is defects in the material, such as vacancies andyor inter- high. The spectrum reveals no peaks in the excitonic stitials w17,18x and is due to the radiative recombination region. In fact, the only features observed lay in the from a shallow donor to a deep localised acceptor state region where signal is expected from donor-acceptor- w19x. In diamond a similar broad band between 2.4 and pair (DAP) recombination. These characteristics point 3.2 eV referred to as ‘band A luminescence’, was argued to a bad structural quality of the material or to a high to be a result of donor–acceptor pair recombination w20x concentration of incorporated donors and acceptors, and later was demonstrated to be dislocation related which is not found for the MOCVD grown sample on w21x. P.R. Hageman et al. / Thin Solid Films 443 (2003) 9–13 13 4. Conclusions w3x M. Ishihara, T. Manabe, T. Kumagai, T. Nakamura, S. Fujiwara, Y. Ebata, S. Shikata, H. Nakahata, A. Hachigo, Y. Koga, Jpn. J. Appl. Phys. 40 (2001) 5065. In this study the deposition of GaN on diamond w4x M. Oba, T. Sugino, Diamond Relat. Mater. 10 (2001) 1343. substrates was demonstrated for the first time. As a w5x H. Amano, N. Sawaki, I. Akasaki, Y. Toyoda, Appl. Phys. nucleation layer for the MOCVD experiments we have Lett. 48 (1986) 353. used a low temperature AlN layer. The resulting w6x Akasaki, H. Amano, Y. Koide, K. Hiramatsu, N. Sawaki, J. MOCVD GaN layer is polycrystalline and hexagonal Cryst. Growth 98 (1989) 209. with, as a result of carbon incorporation originating w7x M. Oba, T. Sugino, Jpn. J. Appl. Phys. 39 (2000) L1. from the substrate, a very poor optical quality as w8x W.J.P. van Enckevort, G. Janssen, W. Vollenberg, M. Chermin, determined by photoluminescence. Additional enlarge- L.J. Giling, Surf. Coat. Technol. 47 (1991) 39. w9x P.R. Hageman, S. Haffouz, V. Kirilyuk, A. Grzegorczyk, P.K. ment of the thickness of the GaN film using HVPE Larsen, Phys. Stat. Sol. A 188 (2001) 523–526. preserved the hexagonal structure and resulted in layer w10x V. Kirilyuk, P.R. Hageman, P.C.M. Christianen, M. Zielinski, of good optical quality, i.e. with a strongly reduced P.K. Larsen, Appl. Phys. Lett. 79 (2001) 4109. intensity of the donor-acceptor pair recombination relat- w11x F.K. de Theije, A.R.A. Zauner, P.R. Hageman, W.J.P. Enckevort, ed peaks in the PL spectrum. X-Ray diffraction analyses P.K. Larsen, J. Cryst. Growth 197 (1999) 37. revealed that the dominant orientation of the GaN w12x A. van der Drift, Philips Res. Rep. 22 (1967) 267. crystallites perpendicular to the substrate changed from w13x J.J. Schermer, F.K. de Theije, W.A.L.M. Elst, J. Cryst. Growth w001x for the thin MOCVD film to w112x for the HVPE 243 (2002) 302. layer. w14x S. Haffouz, V. Kirilyuk, P.R. Hageman, L. Macht, J.L. Weyher, P.K. Larsen, Appl. Phys. Lett. 79 (2001) 2390. w15x H. Tang, J.B. Webb, J.A. Bardwell, S. Raymond, Appl. Phys. Acknowledgments Lett. 78 (2001) 757. w16x D.J. As, L. Lischka, Phys. Stat. Sol. (A) 176 (1999) 475. The authors would like to acknowledge M.P. Moret w17x J. Neugebauer, C.G. Van der Walle, Appl. Phys. Lett. 69 and P. Mulder for the SEM pictures and XRD measure- (1996) 503. ments and V. Kirilyuk, L. Macht and T. Ochalski for w18x P. Boguslawski, E. Briggs, J. Bernholc, Phys. Rev. B51 (1995) performing the PL measurements. 17255. w19x P. Perlin, T. Suski, H. Teisseyre, M. Leszczynski, I. Grzeory, References J. Jun, S. Porowski, P. Boguslawski, J. Bernholc, J.C. Chervin, A. Polian, T.D. Moustakas, Phys. Rev. Lett. 75 (1995) 296. w1x J.E. Field (Ed.), Properties of Natural and Synthetic Diamond, w20x P.J. Dean, Phys. Rev. (1965) A588. Academic Press, London, 1992. w21x H. Kawarada, Y. Yokata, Y. Mori, K. Nishimura, A. Hiraki, J. w2x H. Gong, X. Jiang, J. Cryst. Growth 235 (2002) 333. Appl. Phys. 67 (1990) 983.