Mater. Res. Soc. Symp. Proc. Vol. 831 © 2005 Materials Research Society

E3.23.1

Initial Stages of Growth of Gallium Nitride via Iodine Vapor Phase Epitaxy WJ Mecouch*,1, BJ Rodriguez2, ZJ Reitmeier1, J-S Park1, RF Davis1, Z Sitar1 Department of Materials Science and Engineering, North Carolina State University, Raleigh, NC 27695 2 Department of Physics, North Carolina State University, Raleigh, NC 27695 1

Abstract Thin layers of GaN have been deposited on 1µm thick MOVPE GaN(0001) thin film substrates using a novel vertical iodine vapor phase epitaxy system. The system features three concentric flow zones that separate the reactant gasses until they reach the substrate. Hydrogen flows through the innermost zone to deliver iodine vapor from an external bubbler to the molten Ga maintained at ~1050°C and GaI to the substrate; high-purity ammonia flows through the outermost zone; nitrogen flows through the middle zone to prevent reaction between the growth species at the GaI nozzle. GaN growth was found to be a function of time, with decreasing concentration of iodine the likely cause of a decrease in growth rate at longer growth times. The step-and-terrace microstructure of the MOVPE seeds was replaced with a smooth morphology film after the shortest growth experiment. Star-shaped features with hexagonal symmetry grew on the surface with increasing growth time. These features became the tops of hexagonal pyramids; these pyramids grew competitively and dominated the final growth surface. The surface of the films grown for the longest period contained a step-and-terrace microstructure; however, the density of steps of was lower than that on the surface of underlying MOVPE substrate. Introduction Hydride vapor phase epitaxy (HVPE) has long been used for growing thick layers of GaN having a reduced density of threading dislocations [1] relative to the densities of these defects within thin films of this material grown via MOVPE and MBE. The most common HVPE process route involves the reaction of flowing HCl with the surface of molten Ga to form GaCl vapor which is transported to the substrate and which reacts with flowing NH3 to grow a GaN layer. Sufficiently thick layers can be separated from substrates having a different chemistry via, e.g. laser liftoff [2], and used as stand-alone substrates for the growth of GaN-based material device layers [3]. The reduction in the defect density realized in both the bulk-like GaN substrates and the subsequently grown layers has been associated with the improved performance of blue [4], near-ultraviolet [4] and ultraviolet light emitting diodes [5]. A less chemically aggressive alternative to HCl-based HVPE has been developed [6,7] which uses iodine as a substitute for HCl and is referred to as iodine vapor phase growth (IVPG). A horizontal reactor was employed in the process route described in Refs. [6,7]. Growth rates to 32 µm/hr [7] were reported. A unique vertical, up-flow reactor is being used in the present research to investigate the initial stages of IVPG of GaN. The experimental approach and the results and summary of this study are presented in the following sections. Experimental Procedures The vertical reactor shown schematically in Fig. 1 and described in detail in Ref. 8 consists of three flow zones. I2 is transported within the inner zone to the Ga source, where it reacts to form GaI vapor, which is transported to the GaN substrate. The shield zone contains flowing UHP N2 that separates the GaI and the NH3 until they reach the seed. The outer zone *

Corresponding author: e-mail address: [email protected]

E3.23.2

(2) Figure 1. Schematic of the reaction tube cross section. The reaction tube is shown on the left side, and the centerline of the reactor as a dashed line on the right. A silica nozzle, shown in grey, sits above the gallium source. Flow in the inner zone begins with (1) the I2 / H2 mixture. After reaction with gallium, the GaI / I2 / H2 mixture flows (2) to the seed. The shield zone (3) contains shield N2. The outer zone (4) contains ammonia and outer N2.

(4) (3)

Ga (1)

contains NH3 as the nitrogen source for GaN, as well as flowing N2 that is referred to as the outer N2. The reactor also features a solid I2 source that is heated independently from the reaction tube. The temperature and pressure of this source are continually monitored and controlled to control the concentration of I2 being transported within the vapor to the reaction tube. Two important changes have been made to the system described in Ref. 8: 1) H2 to carry the iodine within the inner zone and 2) rotation of the seed holder. Hydrogen was incorporated as the sole carrier gas to getter oxygen within the iodine source or elsewhere in the system and thereby improve the purity of the as-grown films. The effectiveness of this change has not been investigated. Growth uniformity across the seeds was improved using seed rotation, as shown in the scanning electron micrographs presented in Figs. 2 and 3. The GaN films were grown at 1050°C and 200 torr total pressure. The flow rates of ammonia, carrier H2, shield N2 and outer N2 were 1.00 slm, 0.25 slm, 0.62 slm, and 1.25 slm, respectively. The substrates used in this work were 1µm thick GaN films grown via metal-organic vapor phase epitaxy (MOVPE) on 6H-SiC(0001) wafers containing a 100 nm thick AlN buffer layer. The final growth surface of these substrates contained a step-and-terrace microstructure [9]. The initial stages and the mode of IVPG growth of GaN were investigated by employing both a reduced rate relative to that used in the research described in Ref. 8 and short periods of time. The growth rate was reduced by lowering the concentration of iodine in the source ampoule to 5%. This concentration was controlled by adjusting the pressure in the source ampoule to obtain the desired vapor concentration at the measured source temperature. The pressure within the source was controlled via a needle valve. The flow rate through the source was controlled with diaphragm valves. The growth rate for each run was calculated from the values of the weight change of the substrate, the period of the run, the density of GaN (6.1 g/cm3) and the growth area of each seed. Growths within the periods of 30 seconds, 1, 2, 5, 10, and 60 minutes were investigated. Growth time was defined as starting when the iodine source was opened to the reactor. For runs longer than 1 minute, the source flow was initiated and followed by an increase in the pressure in the source from 200 torr to the run value between 400 and 600 torr depending on the source temperature. The time to achieve a stable pressure in the source was typically less than 30 seconds. For the runs having periods of 30 and 60 seconds, the pressure was allowed to build in the source to near the run value before starting the iodine flow to the reactor. The needle valve was also preset in these runs to maintain the source pressure after the flow of iodine was initiated. Filling the iodine source before starting the flow of this reactant reduced the strong transients in the concentration of iodine during the short period runs. However, small

E3.23.3

Figure 2. Cross sectional SEM images taken from a sample grown without rotation. Calculated thickness was 47 µm, but the SEM images show thicknesses of a) 35 µm, b) 50 µm, and c) 75 µm.

Figure 3. Cross sectional SEM images taken from a sample grown with rotation. Calculated thickness was 40 µm, which compares well with the SEM images, which show thicknesses of a) 42, b) 44 and c) 44 µm measured in the SEM images.

adjustments to the needle valve setting were required to pinpoint the source pressure. A fresh substrate was used for each deposition. During heating and cooling within the temperature range of 500°C-1050°C, the sample was held under a flowing gas mixture of 1 slm NH3 and 1.25 slm of N2. The latter gas was divided among the outer (0.5 slm), shield (0.5 slm) and carrier (0.25 slm) zones. The gallium vapor pressure is negligible even at 1050°C; thus, there should have been no significant Ga transport or GaN growth during these periods. The microstructure of the IVPG GaN layers were characterized using a Nikon optical microscope (OM) and atomic force microscopy (AFM) using a TM Microscopes AutoProbe CPResearch scanning probe microscope in contact mode with a silicon nitride cantilever (0.1 N/m). Scans were acquired over 2 µm x 2 µm and 20 µm x 20 µm regions. Results and Discussion Growth for one hour was conducted initially to compare the growth rate of GaN achieved using the conditions noted above with those achieved using the conditions given in Ref. [8]. The growth rate obtained in the present study was 7.5 µm/hr. The reduced iodine concentration compared to previous work [8] led to a reduced growth rate at otherwise similar growth conditions. This data point fits well with the iodine concentration vs growth rate trend established in Ref. [8]. The calculated thicknesses, based on the weight change of the sample, and the growth rate measured in each run in the present research are presented in Table 1. The drop in the growth rate as a function of growth time is likely due to differences in the iodine concentration within the N2 flowing from this source. When the source is isolated prior to a run, the iodine has time to reach the high equilibrium concentration in the vapor that will be initially carried in the H2. If the evaporation of iodine during growth is limited such that equilibrium is not achieved, the concentration of iodine in the vapor phase will continually decrease throughout a run, because the iodine can never evaporate sufficiently fast to reach the equilibrium concentration. Since the iodine controls gallium transport, a continued decrease in the iodine concentration leads to an overall decrease in the growth rate of the GaN. As the initially fast growth rate is averaged over longer periods of slow growth, the calculated growth rate decreases.

E3.23.4

Sample A B C D E F

Growth time 30 sec 1 min 2 min 5 min 10 min 60 min

Table 1 Thickness (µm) 0.3 0.8 1.1 1.5 2.5 7.5

Growth rate (µm/hr) 42 48 32 17.5 15 7.5

The five sequences of OM and AFM micrographs contained in Fig. 4 show the evolution of the surface microstructure of the GaN films as a function of growth time. The OM and AFM images were not necessarily obtained from the same spot in the films. Images were obtained which are representative of the films that were grown in each time interval. After 30 sec. an essentially complete film has formed that contains the random array of small black spots shown in Figure 4 a) that are uncoalesced areas within the film. The surface microstructure of the coalesced areas is smooth, as shown in Figs. 4 a), b) and c). Fig. 4 c) shows that the step-and-terrace microstructure of the MOVPE substrate [9] is covered within the first 30 seconds of IVPG. The initial microstructure of the I2-based GaN film appears to have larger undulations than the MOVPE seed [9], as well as valleys with rough edges, such as the one extending vertically through the film shown in Fig. 4 c). Star-shaped features with six lobes appeared in some areas of the film during the next 30 sec. of deposition, as shown in Fig. 4 d). The AFM image, Fig. 4 e), was taken from a region between these features and shows evidence of the initial formation of clusters in the microstructure, but no definitive star-shaped features. The latter features appear in greater density in Fig. 4 f) in concert with hexagonally-shaped islands that are clearly defined in Fig. 4 g). The 1 min. and 2 min. depositions were conducted under different iodine concentrations; however, the presence of the star-shaped features in Figs. 4 d) and f) and the appearance of island facets and well-defined islands in Figs 4 e) and g) indicate that the same mechanism(s) of growth was operable whether a low concentration of iodine was used, as in the growths for 30 sec and 1 min, or a high concentration was used, as in the runs longer than 1 minute. The star features and the islands grow larger during the next 4 minutes of deposition, as shown in Figs. 4 h) and i), respectively. The latter have average diameters to 15 µm and are beginning to grow preferentially to form the pyramidal features observed in Figs. 4 j) and k) after an additional 5 minutes of growth. The latter micrograph shows the tops of several pyramids within one image; a boundary is also observed around the central star-shaped feature. These pyramids are similar to the subgrain features observed by Etzkorn and Clarke [10] in a 22µm thick GaN film grown on sapphire via HVPE. They attributed the structure to slight misorientations between the growing grains. A step-and-terrace microstructure is shown in Fig. 4 l); however, the density of steps is significantly reduced relative to that present on the surface of MOVPE GaN layers [9] similar to the one used in this work. This shows that homoepitaxial growth via IVPG does not mimic the step and terrace features of the substrate as closely as heteroepitaxial growth via MOVPE [9] or HVPE [11]. The dislocation densities indicated by the pits (black dots) in Figs 4 c) and l) are similar, although the pits are more evident in Fig. 4 l). The densities are similar even though the samples were from different locations within the GaN substrate film. The micrograph in Fig. 4 c) was

E3.23.5

Fig. 4 By row: a), b), and c): 30 sec. of I2 flow during growth, d) and e): 1 min, f) and g): 2 min, h) and i): 5 min, and j), k), and l): 10 min. Left column: Optical micrographs, black scale bar represents 50 µm. Middle column: AFM scans of 20 µm x 20 µm areas. Vertical scales are b) 16 nm, e) 14 nm, g) 20 nm, i) 30 nm, and k) 16 nm. Right column: AFM scans of 2 µm x 2 µm areas. Scans b) and c) are in the same area, but k) and l) are different areas of the sample grown for 10 minutes. Vertical scales are c) 3.6 nm and l) 12 nm.

acquired from near the center of the original 2” SiC wafer and shows a pit density of ~1.4 x 107 cm-2; the micrograph in Fig 4 l) was taken from the near edge of the original SiC wafer and

E3.23.6

shows a pit density of ~2 x 107 cm-2. The pits may be more obvious in Fig. 4 l) because the film is thicker, and the pits have grown larger in the lateral dimensions. It is also important to note that Figs. 4 k) and 4 l) were not obtained from the same areas on the sample, as a change in the microscope stage was made to obtain the smaller area scans. Valcheva et al. [12] investigated the relationship between nanopipes and growth mode in HVPE GaN. They determined that a growth mode dominated by hillocks around nanopipes that propagated into the growing film pertained when substrates with smooth surfaces were used. By contrast, a rough surface on a substrate provided more nucleation sites, which promoted stepflow growth. As previously implied, the surfaces of GaN substrates used in the present research possessed both low values of RMS roughness (1.3 nm from a 10 µm x 10 µm area of a 1 µm thick film [9]) and a step-and-terrace microstructure. These former characteristic allowed the initial growth of an IVPG film having a smooth morphology which mirrored the larger undulations of the seed layer, as shown in Figs. 4 b) and e). The latter feature and the dislocations within the terraces provided a very large number of sites for the nucleation and growth of islands and pyramids. Ostwald ripening occurred with growth time resulting in the enlargement of selected islands and pyramids at the expense of others, as shown in Figs. 4 g), i), and k). The pyramids show macroscopic features of having grown via a spiral growth mechanism in films deposited at rates that exceed those used herein. Island overgrowth has been observed by Bendersky et al. [13] during the deposition of GaN via HVPE directly on a bare SiC substrate. Summary The initial stages of I2-based vapor phase homoepitaxial growth of GaN films have been investigated using optical microscopy and AFM. The step-and-terrace nature of the MOVPE seed layer was covered by a surface that mimicked only the larger undulations of the seed. Starshaped features with six-fold symmetry gradually formed on the surface with increasing growth time. These features increased in diameter, and evolved with time to become the tops of hexagonal pyramids that coalesced and became the dominant surface features. Step-flow was the initial growth mode; however, exaggerated growth of selected pyramids and the spiral growth of hillocks around dislocations with screw or mixed character caused these features to dominate the final surface. Small area AFM scans showed the step density in the IVPG films to be much lower than the original MOVPE seed layer. 1 W. Götz, L.T. Romano, B.S. Krusor, N.M. Johnson, R.J. Molnar, Appl. Phys. Lett. 69, 242 (1996). 2 T. Paskova, V. Darakchieva, P. Paskov, U. Sodervall, B. Monemar, J. Cryst. Growth 246, 207 (2002). 3 N.G. Weimann, M.J. Manfra, J.W.P. Hsu, K. Baldwin, L.N. Pfeiffer, W. West, S.N.G. Chu, D.V. Lang, R.J. Molnar, Compound Semiconductors 2002 Inst. Phys. Conf. Ser. 174, 223 (2003). 4 X.A. Cao, S.F. LeBoeuf, M.P. D’Evelyn, S.D. Arthur, J. Kretchmer, C.H. Yan, Z.H. Yang, Appl. Phys. Lett. 84, 4313, (2004). 5 A. Yasan, R. McClintock, K. Mayes, S.R. Darvish, H. Zhang, P. Kung, M. Razeghi, S.K. Lee, J.Y. Han, Appl. Phys. Lett. 81, 2151 (2002). 6 M. Suscavage, L Bouthillette, D. Bliss, S-Q. Wang, C. Sung, phys. stat. sol. (a) 188, 477 (2001). 7 V. Tassev, D. Bliss, M. Suscavage, Q.S. Paduano, S-Q. Wang, L. Bouthillette, J. Cryst. Growth 235, 140 (2002). 8 W.J. Mecouch, Z.J. Reitmeier, J-S. Park, R.F. Davis, Z. Sitar, phys. stat. sol. (c) In press. 9 A.M. Roskowski, P.Q. Miraglia, E.A. Preble, S. Einfeldt, R.F. Davis, J. Cryst. Growth 241, 141 (2002). 10 E.V. Etzkorn, D.R. Clarke, J. Appl. Phys. 89, 1025 (2001). 11 Y. Golan, X.H. Wu, J.S. Speck, R.P Vaudo, V.M. Phanse, Appl. Phys. Lett. 73, 3090 (1996). 12 E. Valcheva, T. Paskova, B. Monemar, J. Cryst. Growth 255, 19 (2003). 13 L. Bendersky, D. Tsvetkov, Y. Melnik, J. Appl. Phys. 94, 1676 (2003).