Single-Crystal Aluminum Nitride Substrate ...

Mat. Res. Soc. Symp. Proc. Vol. 680E © 2001 Materials Research Society

Single-Crystal Aluminum Nitride Substrate Preparation from Bulk Crystals J. Carlos Rojo, Leo J. Schowalter*, and Kenneth Morgan Crystal IS, Inc., Latham, NY 12110 Doru I. Florescu and Fred H. Pollak, Brooklyn College, Brooklyn, NY 11210 Balaji Raghothamachar and Michael Dudley SUNY, Stony Brook, NY 11794 ABSTRACT Large (15mm diameter) single-crystal AlN boules have been prepared using sublimationrecondensation growth. X-ray topography shows that the dislocation density averages less than 103 cm2 in some of the substrates but also that the dislocations are not uniformly distributed. Also, strain due to the differential expansion with the crucible walls seems to cause severe cracking in the periphery of the crystal and high-strain regions. Thermal analysis using the Scanning Thermal Microscopy (SThM) reveals a thermal conductivity of 3.4 ± 0.2 W/K-cm, which is the largest value ever reported for AlN. INTRODUCTION Aluminum nitride (AlN) has received attention as a candidate for III-nitride epitaxy applications due to its close lattice match, minimal differential thermal expansion compared to GaN, and high thermal conductivity. There is interest in AlN substrates as a competitive substrate for heteroepitaxial growth of GaN until commercial bulk GaN substrates become available. In addition, AlN is a more desirable substrate than GaN for device structures that require Al-rich nitride epitaxial layers such as solar-blind UV detectors, UV light sources and high power microwave devices. Among the properties that the ideal substrates for III-nitride device fabrication should posses, thermal conductivity plays a role of paramount importance. The next generation of high-power, high-temperature electronic devices will demand a high thermal conductivity substrate capable of dissipating the large amounts of heat generated at the active layers to prevent rapid aging and catastrophic failure. Slack et al. [1] estimated the thermal conductivity of AlN to be 3.19 W/cmK and a more recent study [2] claims the AlN thermal conductivity value to be even higher, approximately 5.9 W/cm-K. In either case however, the value of the thermal conductivity strongly depends on the concentration of oxygen impurities in AlN crystals. Oxygen is able to substitute nitrogen in AlN up to very high concentrations. This substitutional defect has associated an aluminum vacancy, which can be a very efficient center for phonon scattering [3]. Therefore, minimizing oxygen impurities during the growth process is crucial to produce high thermal conductivity AlN substrates. *

on sabbatical from the Physics, Applied Physics and Astronomy Dept., Rensselaer Polytechnic Institute, Troy, NY 12180 E2.1.1

Out of all the techniques used to grow bulk AlN, the sublimation-recondensation technique has been recognized as the most promising method to produce large AlN boules. While the sublimation-recondensation technique, first developed by Slack and McNelly [4], has been known for several decades, not much progress has been achieved in scaling up the process. Even though high quality AlN crystals with full-width-at-half-maximum (FWHM) of less than 40 arcsec have been reported before [5], those crystals were mainly platelets or needles and, therefore, too small to allow device fabrication. To the best of our knowledge, we report here the largest bulk AlN crystal grain grown by the sublimation-recondensation technique with a structural quality that can be compared to that of the best AlN platelets gown under much lower growth rates.

CRYSTAL GROWTH OF ALN SUBSTRATES We have grown several AlN boules by the sublimation-recondensation method using dense AlN ceramic with an oxygen content ranging from 200 to 500ppm of oxygen by weight as starting material. In this method [4], a thermal gradient drives the sublimation of the polycrystalline AlN starting material and posterior recondensation at the colder part of the crucible as a single crystal. A relative movement between the thermal gradient and the crucible sets the driving rate, which for our different growth runs was between 0.65 and 0.9mm/h. If the growth condition, including temperature gradient and gas pressure, are adequate, the growth rate will be equal to the driving rate. However, under inadequate conditions for a given driving rate, platelets or polycrystalline material may result. Several wafers were cut from 15mm diameter AlN boules with diameters ranging from 10 to 15mm and thickness ranging from 750 to 1,000 µm before polishing. As shown in Fig. 1a, some of these wafers consist of several grains (labeled as 1, 2, and 3 in Fig. 1a) with single crystal appearance occupying most of the wafer and separated by what appears to be grain boundaries. The perpendicular axis to the major grain forms approximately 45o with the c-axis. The rest of the wafer (~1/3) seems to be a polycrystalline region (labeled as 4 in Fig. 1a). The evolution of grains 1-3, shown in Fig. 1, indicates that the single crystal grains get larger as the boule grows, while the surface area occupied by the polycrystalline region decreases. This behavior is very advantageous to growing large, self-seeded crystals since it guaranties that, under adequate nucleation and growth conditions, one or very few good single crystal grains will eventually prevail over polycrystalline regions. Fig 1b shows a 10mm-diameter, 850 µm-thick single crystal wafer obtained from a different boule than that shown in Fig 1a. This wafer exhibits no polycrystalline regions nor large angle grain boundaries. As a main difference with the substrate in Fig.1a, the periphery of the substrate in Fig.1b exhibits multiple cracks. Using back reflection diffraction, the surface orientation of the 1cm diameter wafer crystal shown in Fig.1b was found to be ~6-8 degrees away from the (10 1 1) plane. The different grain structures observed in the two wafers described above are most likely caused by the difficulty to reproduce the nucleation conditions early on during the process. We have observed that the reproducibility of the growth results can be substantially improved by accurately controlling the operational conditions in the system, especially, the thermal gradients and the starting material quality.

E2.1.2

a

b

Fig. 1. Picture of large substrates cut from AlN boules (grids are in mm). The substrate on the left (a) exhibits three main single crystal grains (1,2, and 3) and a polycrystalline region (4). The substrate on the right (b) does not present polycrystalline regions, but displays severe cracking area close to the periphery.

STRUCTURAL CHARACTERIZATION The microstructural characterization of Crystal IS wafers has been carried out using Synchrotron White Beam X-ray Topography (SWBXT) at the Dept. of Materials Science & Engineering, SUNY at Stony Brook, NY. This technique enables an extensive and detailed characterization of structural defects such as dislocations, stacking faults, twins, grain/subgrain boundaries, precipitates, etc. The diffraction pattern has been recorded using the area-filling white X-ray beam at beamline X-19C at the National Synchrotron Light Source. An enlargement of the diffraction pattern (see Fig. 2a) recorded for the substrate shown in Fig 1a shows that the crystal, with a thickness of approximately 800 µm, is divided into three main subgrains (agreeing with the optical observation) each with a high degree of perfection. A close up of the lower central region of the largest grain, shown in Fig. 2b, reveals groups of dislocations with a local density of around 103-104 cm-2. Regions in between those indicated by G have dislocation densities that are lower by at least two orders of magnitude. The original nucleation source for the dislocations, from which the multiplication occurs, appears to be closely associated with the region of high deformation observed in the topograph. The deformation is inhomogeneous and quite severe in a few locations, such as at D in Fig. 2b. The dislocation configurations observed in Fig. 2b are clearly the result of plastic deformation processes [6,7] which must occur away from the growth interface and, therefore, after the growth process. Whether they are formed at growth temperature or during the cooling process is unclear. However, the largest stresses are expected during the post-growth cooling, most probably due to the thermal expansion mismatch between the crucible and the grown boule. This is supported by the location of the high deformation areas on the crystal periphery.

E2.1.3

a

b

Fig. 2. a) Enlargement from a region of the transmission diffraction pattern showing the main three subgrains in the sample as labeled in Fig.1a, and b) close up of the central region of the larger grain (labeled as number 1). Transmission diffraction patterns for the substrate shown in Fig. 1b were recorded after the plane (10 1 1) was oriented perpendicular to the x-ray beam direction. The absence of a scanning apparatus precluded obtaining a topograph of the entire wafer in one recording. Hence, two topographs were recorded for each reflection. However, the large amount of strain in the sample and the corresponding distortion of the x-ray topograph on the film prevented a mosaic from being created from these separate topographs to produce a single picture. The wafer can be separated in several regions as shown in Fig. 3a. The entire wafer is a single crystal but two long cracks were observed. Therefore, topograph images are split into three parts. The topograph from the upper half of the wafer contained regions 1 and 2 while the topograph from the lower half is chiefly constituted of region 3 and small parts of region 2. The left side of the entire wafer is characterized by relatively lower strains than the right side. Figure 3b shows the 1 103 reflection from the top and bottom halves of the wafer. Overall, the wafers are characterized by large strains because of which the topographic images are heavily distorted. Strain on the right side of the wafer is much higher than that on the left side. In the low strain regions, subgrain boundaries (misorientations varying from 10-30 seconds of arc) can be discerned by orientation contrast in the peripheral regions. Fig. 4 shows a high magnification topograph from the peripheral region of the bottom half of 1 103 reflection (Fig. 3b). The contrast from this region indicates a dense cellular network of dislocations (with density around a few times 105cm-2). Qualitatively, defect density in the periphery appears to be much higher (>106cm-2) than that in the interior sections of the strain-free regions. The density of dislocations find in some large areas of the AlN substrates analyzed here is much better than that typically reported for quasi-bulk GaN grown on sacrificial sapphire by hydride vapor phase epitaxy (HVPE) (>106 cm-2) [8]. Dislocation densities in the range of 102 to 105 cm-2 have been reported for bulk SiC [9,10]. However, epitaxial layers of AlN or GaN grown on SiC typically have higher densities with 106 cm-2 being typical [11,12].

E2.1.4

a

b

Fig. 3. a) Optical picture of the substrate shown in Fig. 1 depicting the different strain regions revealed by the topography and the three different regions produced by the two major cracks, and b) transmission topograph (g = 1 103) showing those three different regions produced by the two cracks. Subgrain boundaries (SB) can be observed in the periphery of the wafer.

From the analysis of the topograph obtained from both substrates it is clear that low-density of dislocation large bulk AlN single crystal can be grown by the sublimation recondensation technique at growth rates that largely exceeded any other previously reported results on AlN crystal growth. Plastic deformation and severe cracking at the periphery of the crystal, very likely resulting from the differential expansion of the crucible and the crystal, remains an issue at the present time.

Fig 4. High magnification of the bottom half of the transmission topograph (g = 1 103)in Fig.3b showing a dense cellular network of dislocations (C).

E2.1.5

THERMAL CONDUCTIVITY ANALYSIS We have studied the thermal conductivity on several AlN substrates using the Scanning Thermal Microscopy (SThM) technique. SThM does not require thick samples, is a non-destructive technique, and has a micron range resolution. These characteristics confer this technique an enormous potential to measure the thermal conductivity of thin substrates and reveal the relationship between local density of defects and thermal conductivity. A detailed description of this technique can be found in [13]. The value of the thermal conductivity in substrates as those shown in Fig. 1 revealed a thermal conductivity ranging from 3.3 to 3.4 W/cm-K, being the maximum value obtained 3.4 ± 0.2 W/cm-K. This value is the largest ever reported for AlN substrates using this or any other technique (see [13]), and it is much larger than the 2.85 W/cm-K obtained by Slack in bulk crystals grown using the sublimation-recondensation technique. According to the latest results obtained using the SThM method, the best thermal conductivity value reported for LEO films and free standing thick HVPE GaN is 2.1 W/cm-K. A thermal conductivity of 3.9 W/cm-K has been measured on SiC substrates [14] using the SThM method. These results indicate that, in terms of thermal conductivity, our single crystal AlN substrates are clearly superior and comparable to the available GaN and SiC substrates, respectively. CONCLUSIONS Using the self-seeded sublimation-recondensation technique we have been successful in growing up to 15mm diameter AlN boules containing single crystal grains of more than 80mm2 and several mm in length using driving rates between 0.65mm/h and 0.9mm/h. The structural characterization of single crystals obtained using that growth technique reveals that an average density of dislocations in the order of 103-104 cm-2 can be achieved using this growth technique. Using the SThM technique we have measured the thermal conductivity of those substrates, 3.4 ± 0.2 W/cm-K, which is the largest reported for AlN. We have also identified the differential thermal expansion between the crucible and the crystal to be the main factor causing severe plastic deformation and cracking. Eliminating the presence of highly strained regions and cracking, controlling better the nucleation step, and reducing the point defect density are all currently major issues of concern to the production of high-quality AlN substrates. Radial scale up will also require controlling the larger radial thermal gradients. However, based on the present status of our technology development, we do not anticipate any major physical obstacle to produce 50mm diameter, or even larger, cost-efficient AlN substrates using the sublimation-recondensation growth technique.

E2.1.6

ACKNOWLEDGMENTS This work has been partially supported by ONR, BMDO, and AFRL. The authors would gratefully like to acknowledge the support of Dr. C.E.C. Wood (ONR), Dr. C. Litton (AFRL) and T.E. Wille (AFRL). The authors also acknowledge helpful discussions with Prof. G.A. Slack. REFERENCES [1] G.A. Slack, R.A. Tanzilli, R. Pohl, J.W. Vandersande, J. Phys. Chem. Solids 48, 641 (1987) [2] A. Witek, Diamond and Related Materials 7, 962 (1998) [3] G.A. Slack, J. Phys. Chem. Solids 34, 321 (1973) [4] G.A. Slack, T.F. McNelly, J. Crystal Growth 34, 263 (1976) [5] M. Tanaka, s. Nakahata, K. Sogabe, H. Nakata, M. Tobioka, J. Appl. Phys. 36, L1062 (1997). [6] V. Audurier, J.L. Demenet, J. Rabier, Philos. Mag. A, 77, 825 (1998). [7] M. Azzaz, J.P. Michel, V. Feregotto, A. George, Mater. Sci. Eng. B, B71, 30 (2000) [8] R.P. Vaudo, G.R. Brandes, J.S. Flynn, X. Xu, M.F. Chriss, C.S. Christos, D.M. Keogh, F.D. Tamweber. Proceedings of International Workshop on Nitride Semiconductors. Nagoya, Japan; 24-27 Sept. 2000; 15-18 p.1002. [9] E.V. Kalinina, A.S. Zubrilov, N.I. Kuznetsov, I.P. Nikitina, A.S. Tregubova, M.P. Shcheglov, V.Y. Bratus. Proc. of the ICSCRM ’99; Research Triangle Park, NC, USA; 10-15 Oct., 1999; 338-342, 497 (2000). [10] S.F. Avramenko, M.Y. Valakh, V.S. Kiselev, M.Y. Skorokhod. Metallofizika i Noveishie Tekhnologii, 22, 33 (2000) [11] C.D. Lee, V. Ramachandran, A. Sagar, R.M. Feenstra, D.W. Greve, W.L. Sarney, L. Salamanca-Riba, D.C. Look, B. Song, W.J. Choyke, R.P. Devaty. J. Electron. Mater 30, 162 (2001) [12] T.S. Zheleva, N. Ok-Hyun, W.M. Ashmawi, J.D. Griffin, R.F. Davis. J. Crystal Growth, 222, 706 (2001) [13] D.I. Florescu, V.M. Asnin, F.H. Pollak, Compound Semiconductor 7(2) March, 62 (2001). [14] D.I. Florescu, F.H. Pollak, Wide Bandgap Electronics, MRS Proc. Vol. 680E, ed: T. E. Kazior, P. Parikh, C. Nguyen, E. T. Yu (2001) (accepted)

E2.1.7