Halogen-free vapor phase epitaxy for high-rate growth of GaN bulk crystals

Here, we propose a halogen-free vapor phase epitaxy (HF-VPE) technique to grow bulk GaN single crystals. This technique employs the simplest reaction for GaN synthesis (reaction of Ga vapor with NH3) and can potentially achieve a high growth rate, a prolonged growth duration, a high crystal quality, and a low cost. The analyses of thick HF-VPE-GaN layers grown under optimized growth conditions revealed that high-quality crystals, both in terms of dislocation density and impurity concentration, are obtained at high growth rates of over 100 µm/h.

G aN-based optoelectronic devices such as lightemitting and laser diodes have been successfully developed and applied to solid-state lighting and high-density optical memory. 1) The next target application for GaN, which has excellent electronic properties such as a high electrical breakdown field and a high electron mobility, is power devices. [2][3][4] One of the requirements for highperformance and highly reliable GaN-based power devices is large high-quality native GaN single-crystal wafers produced at a low cost; therefore, the development of crystal growth techniques is important to obtain bulk GaN crystals. However, the equilibrium N 2 pressure over congruent molten GaN is expected to be >6 GPa; 5) therefore, pulling up a GaN crystal from molten GaN is unrealistic. Alternative techniques, such as high-pressure nitrogen solution growth, 6) Na-flux growth, 7,8) ammonothermal growth, 9,10) and hydride vapor phase epitaxy [11][12][13] (HVPE), have been proposed. The former three techniques have the advantage of producing high-quality crystals, although the growth rates are low (≤10 µm=h); therefore, these techniques would be appropriate for preparing high-quality seed crystals.
In contrast, the HVPE technique has an advantage of high growth rate (∼100 µm=h), which has contributed to the production of self-standing GaN wafers, especially as substrates for blue-violet laser diodes. Self-standing GaN wafers have been produced by a two-step process of thick GaN layer growth (300-1000 µm) on foreign substrates and subsequent substrate removal. The self-standing wafer production process is a compromise as a result of solid ash (NH 4 Cl) generation due to the reaction of the HVPE growth process [GaCl(g) + 2NH 3 → GaN + NH 4 Cl(s) + H 2 ], which results in the closing of the exhaust with the ash, thereby hampering the long-duration process required to grow large bulk crystals.
To address the ash issue in the HVPE process, other vapor phase growth techniques without chlorine and chloride (halogen and halide) as reactants, such as sublimation growth, 14) oxide-source vapor phase epitaxy, 15) and hightemperature vapor phase epitaxy, have been explored. 16) These techniques have the advantage of ash-free growth, but still suffer from the incorporation of impurities and the long-term instability of the growth rate, which causes quality and productivity issues.
To simultaneously achieve ash-free growth, low impurity incorporation, and the long-term stability of the growth rate, we propose halogen-free vapor phase epitaxy (HF-VPE) as a significantly improved technique that utilizes the simplest reaction of Ga(g) + NH 3 → GaN + 3=2H 2 to grow GaN at high rates for prolonged durations, and employs the most stable TaC-coated graphite crucibles=susceptors to prevent impurity incorporation. [17][18][19] Here, we examined the dependence of the HF-VPE growth rate on the growth conditions; the obtained results revealed that the high temperature and low pressure in the Ga source crucible are appropriate growth conditions for high-rate growth, confirming that the physical mechanism of Ga supply in HF-VPE is the simplest (evaporation of Ga) without parasitic reactions. Furthermore, we discuss the growth rate limit in terms of the evaporation efficiency and demonstrate HF-VPE growth at a high rate with low-concentration impurity incorporation that results in high-quality thick GaN layers. Figure 1 shows a schematic diagram of the vertical HF-VPE growth setup, which consists of growth [seed holder and seed substrate (sapphire or GaN=sapphire template)] and source (molten Ga source, Ga crucible to hold the source, and inner=outer crucibles) zones. All the crucible and seed holder components are made of TaC-coated graphite. 19,20) The crucible and quartz-glass components are contained in a vacuum chamber (quartz-glass tube). A vertically movable helical radio-frequency (RF) coil used to heat the growth setup is placed outside the vacuum chamber. The temperatures at the seed holder and Ga crucible are monitored by two-color pyrometers placed at the top and bottom viewports of the chamber, which feedback into both the RF output power and the vertical position of the RF coil to independently control the temperatures of the Ga crucible and seed holder.
There are three process-gas flow channels in the HF-VPE growth setup: carrier-N 2 , sheath-N 2 , and NH 3 diluted with N 2 . The carrier-N 2 in the innermost channel, the spacing between the Ga crucible and the inner crucible, transports Ga-related gas-phase species generated from the molten Ga source to the outer crucible. The carrier-N 2 with Ga-related gas-phase species is mixed with sheath-N 2 in the outer crucible, and then the mixture flows out from the crucible through a multihole source-gas outlet. Outside the outer crucible, all gases (N 2 , Ga-related gas, and NH 3 ) are mixed and reach the substrate surface, which drives GaN crystal growth.
The main differences in the growth setup between the proposed HF-VPE and similar VPE techniques 16,21) are the addition of the sheath-N 2 flow channel and the full adoption of TaC-coated components, which contributes to the following several advantages: 1) prevention of the direct reaction of molten Ga with NH 3 , which could cause the boiling and=or creeping of the molten Ga and result in the incorporation of Ga droplets into the grown crystals; 2) independent control of the Ga feed rate k Ga and Garelated-species partial pressure to suppress the formation of GaN polycrystals at the source-gas outlet; and 3) significant suppression of impurity incorporation that causes surface morphology degradation and nanopipe defect formation. 17) Furthermore, the absence of NH 4 Cl ash generation means that exhaust closing does not occur, which enables the growth period to be extended. Thus, HF-VPE is potentially more favorable for the growth of high-quality bulk GaN crystals than other VPE techniques.
A number of short-duration preliminary growth runs (10-20 min) were conducted to determine the dependences of the Ga feed rate k Ga and the GaN growth rate R gr on the growth conditions (crucible temperature, pressure, and carrier-N 2 flow rate). k Ga and R gr were evaluated on the basis of the weight changes of the Ga source crucibles and sapphire seed substrates before and after the growth runs. After the optimization of the growth conditions for both the crystal quality and the growth rate, high-quality thick GaN layers were deposited on a metalorganic vapor phase epitaxy (MOCVD)-GaN=sapphire template to verify the crystal quality by scanning electron microscopy (SEM) observation, X-ray rocking curve (XRC) measurement, transmission electron microscopy (TEM) observation, and secondary ion mass spectrometry (SIMS) analysis.
Figures 2(a)-2(c) show the dependences of k Ga and R gr on the Ga crucible temperature T, the crucible-interior pressure p c (calibrated from the background pressure through temperature-dependent viscous gas flow and resultant differential pressure consideration), and the carrier-N 2 flow rate Q carrier . The dependences of k Ga and R gr on the growth conditions were very similar. The R gr data were subsequently replotted with respect to k Ga , as shown in Fig. 2(d), which clearly revealed an almost perfect linear relationship, i.e., the ratecontrolling factor for GaN growth is solely the Ga supply. Hereafter, k Ga with respect to the individual growth conditions is discussed to address the mechanism of Ga supply and appropriate growth conditions.
The temperature dependence of k Ga was well fitted by the Antoine equation, which is an empirical equation used to precisely fit the temperature dependence of the saturated vapor pressure of any substance; therefore, the mechanism of Ga supply is plausibly evaporation. To confirm this mechanism, the temperature dependence of k Ga was drawn as an Arrhenius plot (Fig. S1 in the online supplementary data at http://stacks.iop.org/APEX/10/045504/mmedia). The plot was fitted by the Arrhenius equation (regression formula shown in Fig. S1) to obtain the activation energy for the Ga feed, E feed a . The experimentally obtained E feed a (206 ± 20 kJ=mol) was comparable to the heat of vaporization of Ga 22) (see inset of Fig. S1 in the online supplementary data at http://stacks.iop.org/APEX/10/045504/mmedia); therefore, the mechanism of Ga supply must be evaporation (see the online supplementary data at http://stacks.iop.org/APEX/10/ 045504/mmedia for details), and k Ga should be strongly dependent on the saturated vapor pressure.
To obtain insight into the further enhancement of k Ga , we evaluated the evaporation efficiency evap Ga , which is defined by the experimentally derived partial Ga pressure in the carrier-N 2 flow p Ga over the saturated Ga vapor pressure p sat Ga , where p Ga and p sat Ga are calculated as where V mol gas is the molar volume of an ideal gas (24.465 L=mol), M Ga is the atomic weight of Ga (69.723 g=mol), and T is the Ga crucible temperature (in Kelvin). Equation (2) was derived by fitting the saturated Ga vapor pressure data 22) with the Antoine equation. The dependences of p sat Ga , p Ga , and evap Ga on T are shown in Fig. S2(a) in the online supplementary data at http://stacks.iop.org/APEX/10/045504/mmedia. Although p Sat Ga and p Ga had similar dependences, p Ga was always smaller than p sat Ga ( evap Ga ¼ 2432%), and the deviation increased with T. Accordingly, evap Ga gradually decreased with increasing T (increasing k Ga ). The decrease in evap Ga with increasing k Ga was also observed with respect to p c and Q carrier [Figs. S2(b) and S2(c) in the online supplementary data at http:// stacks.iop.org/APEX/10/045504/mmedia]. If the cause of this decrease in evap Ga is revealed, then additional measures to realize ideal evap Ga (∼100%) could be obtained. To elucidate the cause of this decrease in evap Ga , the dependences of k Ga on p c and Q carrier are discussed in detail. The dependences of k Ga on p c and Q carrier were well fitted by power law equations, as shown in Figs. 2(b) and 2(c), the scaling factors of which were approximately −0.69 and +0.27, respectively. The scaling factor for the regression formula for the pressure dependence (≈ −0.7) is quite close to −1, which corresponds to that for the pressure dependence of the gas diffusion coefficient. 23) This coincidence strongly indicates that the diffusion of Ga vapor governs the Ga evaporation rate. In addition, the scaling factor for the Q carrier dependence (≈ +0.3) is not far from +0.5, which corresponds to that for the dependence of the molecular flux through a stagnant boundary layer (with a thickness inversely proportional to the square root of the gas flow velocity). 24) Thus, diffusion through the boundary layer (4-9 mm thick under the present growth conditions) on top of the molten Ga source surface could be the major factor limiting k Ga and evap Ga . A longer path (or time) for carrier-N 2 flow along the molten Ga source surface to attenuate the locally saturated Ga vapor (in the boundary layer) through mutual diffusion between Ga vapor and carrier-N 2 , i.e., increase in molten Ga source surface area, could be a solution. An effective measure for increasing the surface area of the molten Ga source is under investigation and the results will be reported soon.
Both k Ga and R gr were significantly enhanced by increasing the temperature and decreasing the pressure, while they were moderately enhanced by increasing Q carrier . Furthermore, if the ideal evap Ga is achieved, then R gr would reach ∼500 µm=h because R gr is proportional to k Ga . However, the present R gr of 100-200 µm=h is superior or at least equivalent to those achieved by other vapor phase growth techniques, and sufficient for growing bulk-like GaN epilayers and verifying the crystal quality. Thus, the growth conditions were reoptimized both in terms of the growth rate and the quality of the resultant GaN crystals, and thick and thin HF-VPE-GaN layers on MOCVD-GaN=sapphire templates were grown to assess the quality of the epilayers.   implies good crystal quality. The visual appearance [inset photo in Fig. 3(a)] revealed the almost mirrorlike surface of the HF-VPE-GaN growth front, where some large pits were observed. These large pits emerged in the layer probably owing to particle incorporation during growth; therefore, a more intensive optimization of the growth conditions will be required to achieve the crystal quality required for power device applications. The cross-sectional image [ Fig. 3(b)] also shows the featureless fractured surface of the HF-VPE-GaN layer, where the growth thickness was determined to be 55 µm, indicating a growth rate of 117 µm=h.
XRC measurements were also conducted to quantitatively evaluate the crystal quality of the thick HF-VPE-GaN layer (see the online supplementary data at http://stacks.iop.org/ APEX/10/045504/mmedia for XRC measurement results). The full widths at half maximum (FWHMs) for the (0002) and (11 22) reflections of an XRC ω-scan obtained from the HF-VPE-GaN layer were 233 and 213 arcsec, respectively. These FWHMs were better than those obtained for the MOCVD-GaN template seed (255 and 273 arcsec, respectively), indicating that the crystal quality of HF-VPE GaN improved with increasing thickness. This improved quality is attributed to a reduction in dislocation density through pair annihilation (as reported for thick HVPE-grown crystals). 25) Thus, the HF-VPE technique is demonstrated as capable of growing high-quality GaN at high growth rates of over 100 µm=h.
To confirm the quality of initial growth (with or without an increase in dislocation density due to flaws under the initial growth conditions) and the propagation behavior of dislocations (with or without unintended stress and faceting during growth), cross-sectional TEM observations in the interface region between the MOCVD-GaN seed (with a 25 nm show that the majority of defects were mixed and edge threading dislocations, which is typical of GaN crystals. Figures 3(c) and 3(d) clearly indicate no prominent increase in dislocation density during initial growth, and no drift or flexion in the propagation behavior of dislocations, confirming that the growth conditions employed for HF-VPE growth were roughly optimized throughout the growth process.
Finally, the SIMS analysis revealed that the HF-VPE-GaN layers contain very low concentrations of impurities. The typical concentrations of Si, H, C, and O impurities were ∼1 × 10 17 (unintentionally doped), <3 × 10 16 (background level), <8 × 10 15 (background level), and <6 × 10 15 cm −3 (background level), respectively. The concentrations of impurities (H, C, and O) that could generate unintentional carriers (except for Si) are much lower than those for MOCVD, HVPE, and other techniques; 21) therefore, the HF-VPE technique could also be suitable for growing high-purity thick drift layers for vertical GaN power devices.
In conclusion, an alternative HF-VPE technique was proposed for producing bulk GaN single crystals at high growth rates. The investigation of the dependences of R gr and k Ga on the major growth parameters revealed that the mechanism of Ga supply is simple evaporation, and high growth rates of 100-200 µm=h were achieved by increasing the Ga crucible temperature and decreasing the pressure. Furthermore, the increase in the surface area of the molten Ga source was predicted to further enhance the Ga evaporation efficiency and GaN growth rate. The analyses of optimized HF-VPE-GaN layers confirmed that the crystal quality in terms of morphology, dislocation density, and purity is quite promising for obtaining high-quality GaN bulk crystals and for forming thick drift layers for vertical GaN power devices.