Introduction

GaN is of current interest for the fabrication of blue light emitting diodes [Reference Nakamura, Senoh and Mukai1], lasers [Reference Nakamura, Senoh, Nagahama, Iwasa, Yamada, Matsushita, Kiyoku and Sugimoto2] and for high power electronic devices [Reference Mohammad, Salvador and Morkoc3]. It has been shown that the GaN quality plays a strong role in the device performance [Reference Nakamura, Senoh and Mukai1-Reference Mohammad, Salvador and Morkoc3]. Typically, GaN is grown at high temperature (> 1000 °C) using MOVPE with N to Ga ratios larger than 1000 on sapphire or SiC substrates using a thin buffer layer. The high temperature is necessary for efficient catalytic dissociation of NH₃ and the large V/III ratio is needed to offset the N loss from the growing film [Reference Koleske, Wickenden, Henry, DeSisto and Gorman4]. The growth temperature is larger (i.e. 100-500 °C) than the threshold temperature for GaN decomposition, and it is not currently understood to what extent GaN decomposes during growth at MOVPE pressures.

Previously, we suggested that decomposition of the GaN film during growth enhances GaN ordering, by eliminating more weakly incorporated Ga and N atoms [Reference Koleske, Wickenden, Henry, DeSisto and Gorman4]. At equilibrium the growth rate is zero. For a positive growth rate close to equilibrium, the incorporation rate of atoms into the growing lattice is slightly larger than the decomposition rate [Reference Koleske, Wickenden, Henry, DeSisto and Gorman4-Reference Heckingbottom, Chang and Ploog6]. Recently, we showed how the GaN decomposition rate is enhanced in flowing H₂ for pressures greater than 100 torr [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman7]. This result was explained by assuming chemical dissociation of H₂ on the GaN surface which then increases Ga surface mobility and enhances N₂ desorption. This paper reports a more detailed study of GaN decomposition in both H₂ and N₂ as a function of pressure and temperature, where the role of H₂ on the enhanced GaN decomposition rate is clarified.

Experimental Details

Details of the GaN growth are discussed elsewhere [Reference Fatemi, Wickenden, Koleske, Twigg, Freitas, Henry and Gorman8]. The GaN films were grown at 76 torr using a close-spaced showerhead reactor design. This same reactor was also used to study the GaN decomposition. The growth process resulted in specular GaN growth over the 2” sapphire wafer, with excellent thickness uniformity. Temperature reproducibility of the susceptor was a major concern for the decomposition study. The temperature was calibrated by observing the melting point of 0.005” diameter Au wire and correlating it to a thermocouple in close proximity to with the backside of the susceptor. Temperature measurement in the reactor was found to be reproducible to within 5 °C after 8 months of use.

For the decomposition study, pieces of the GaN on sapphire were cleaved and weighed to within 0.1 mg using an analytical balance [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman7]. Repeated weighing of the GaN on sapphire pieces were reproducible to within 0.1 mg. The pieces were reintroduced into the reactor and heated under varying conditions using either 6 SLM flow of H₂ or 3 SLM of N₂. Each piece was ramped at 25 °C per minute to the annealing temperature, which ranged from 800 to 1130 °C. After annealing for a set time and cooling, each piece was re-weighed in air to determine the mass loss. If Ga droplets were observed, they were removed by etching in dilute HNO₃ and rinsing with DI water. (On some samples which were not etched, the Ga droplets were found to be very stable in air even up to several months, suggesting minimal oxidation of the liquid Ga droplets.) Each piece was then weighed again to determine the weight of liquid Ga. Finally, the piece was annealed at 1080 °C until the remaining GaN was decomposed, leaving only the initial bare sapphire surface. The bare sapphire weight was used to calculate the sapphire area in order to convert the measured weights to kinetic rates (atoms/cm²) [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman7].

Results

After annealing GaN in the absence of NH₃, the most notable change in the GaN surface morphology is the appearance of Ga droplets as shown in Fig. 1. For Fig. 1, the GaN surface was annealed at a temperature of 811 °C for 20 minutes in H₂ at a pressure of 150 torr. The Ga droplets are observable as the lighter regions in the phase contrast image (Fig. 1(a)) and as the darker regions in the transmission image (Fig 1(b)). Because the GaN decomposition rate is larger than the Ga desorption rate, the liquid Ga droplets accumulate on the surface and coalesce into larger droplets, similar to the liquid droplet growth mechanism developed by Family and Meakin [Reference Family and Meakin9]. In flowing H₂ Ga droplets were observed for pressures greater than 22 torr [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman7] for anneals at 992 °C. Compared to the Ga droplets in flowing H₂, the Ga droplets in flowing N₂ were barely discernable even at the highest magnification of 1000x. The droplet size increased as both the anneal temperature and the pressure were increased in H₂ and in N₂ for temperature greater than 1000 °C.

Figure 1. (a) Phase contrast and (b) transmission Normarski images of GaN surface heated for 20 minutes in H₂ at a pressure of 150 torr at a temperature of 811 °C. The bar on image (a) indicates a length of 20 μm.

For temperatures ranging from 800 - 1000 °C, the H₂ pressure had a strong influence on both the quantity of Ga and the Ga droplet size. This is shown in Fig. 2, where the kinetic rates for Ga accumulation (i.e. liquid Ga on the surface), GaN decomposition, and Ga desorption are plotted as a function of pressure. Figs. 2(a) through 2(c) show the rates at anneal temperatures of 992 °C, 902 °C, and 811 °C, respectively. It is clear from Fig. 2(a) that the GaN decomposition rate (filled circles) and the surface Ga accumulation rate (open squares) increase as the pressure is increased. The Ga desorption rate (filled diamonds) changes slightly as a function of pressure, peaking near 76 torr in Fig. 2(a), 100 torr in Fig. 2(b), and 120 torr in Fig. 2(c). The increase in the Ga desorption rate at these pressures is due to a maximum in the Ga surface area to volume ratio which then decreases as the droplets coalesce. Clearly, as shown in Fig 2(b) and 2(c), the buildup of Ga on the surface directly coincides with the increase in the GaN decomposition rate.

Figure 2. Plot of the GaN decomposition rate in H2 (solid circles) at various pressures at (a) 992 °C, (b) 902 °C, and (c) 811 °C. The Ga desorption rate (solid diamonds) and the rate of Ga accumulation (open squares) on the surface during the anneal are also plotted.

Arrhenius plots of the GaN decomposition and Ga desorption rates in both H₂ and N₂ are plotted in Figs. 3 and 4, respectively. Because the GaN decomposition rates depend strongly on the H₂ pressure, the decomposition rates in H₂ (solid circles, solid line) are plotted only for 76 torr. From this data, the fit yields a pre-exponential of (6.3±0.4)x10³⁰ cm⁻²s⁻¹ and an activation energy, E_A, of 2.96 ± 0.06 eV. This is in close agreement with previous measurements of GaN decomposition in vacuum, where pre-exponentials of 4x10²⁹ cm⁻²s⁻¹ [Reference Munir and Searcy10] and 5x10²⁸ cm⁻²s⁻¹ [Reference Groh, Gerey, Bartha and Pankove11] and an E_A of 3.1 eV [Reference Munir and Searcy10, Reference Groh, Gerey, Bartha and Pankove11] were measured. For comparison, values of the pre-exponentials and E_A for the GaN decomposition and Ga desorption kinetics are listed in Table 1. Arrhenius plots of the GaN decomposition at higher H₂ pressures gave slopes that were smaller than the slope measured at 76 torr, giving significantly smaller E_A at higher pressures. These values for E_A were not explicitly calculated because it is not clear if the GaN decomposition at these higher pressures is governed by a single, simple chemical mechanism. The GaN decomposition rates were also measured in N₂ at pressures of 76 (open diamonds) and 150 torr (open squares). At fixed temperature in N₂, the GaN decomposition rate was approximately constant for pressures up to 400 torr. In N₂, an exponential fit gives a larger pre-exponential of (1.2±0.1)×10³² cm⁻²s⁻¹ and an substantially larger E_A of 3.62 ± 0.04 eV compared to the kinetic parameters measured in H₂. The GaN decomposition kinetic parameters measured in N₂ are closer to the mass spectroscopy work of Ambacher and coworkers [Reference Ambacher, Brandt, Dimitrov, Metzger, Stutzmann, Fischer, Miehr, Bergmaier and Dollinger12] as shown in Table I.

Figure 3. Arrhenius plot of the GaN decomposition rate vs. the reciprocal temperature under H₂ (filled circles, solid line) and N₂ (open diamonds and squares, dashed line). The pre-exponent and activation energy from the fit are listed in Table 1.

Figure 4. Arrhenius plot of the Ga desorption rate vs. the reciprocal temperature under H₂ (filled and open circles and squares, solid line) and N₂ (filled and open diamonds, dashed line). The pre-exponent and activation energy from the fit are listed in Table 1.

GaN decomposition was also studied as a function of time at a fixed pressure of 150 torr and temperature of 811 °C. Images of the surface after (a) 3, (b) 10, (c) 20, and (d) 80 minutes are shown in Fig. 5. As shown in Fig. 5, the average Ga droplet size increases as the surface is annealed as described by the model of Family and Meakin [Reference Family and Meakin9]. After an initial incubation time, the GaN decomposition rate, Ga surface accumulation rate, and the Ga desorption rate were relatively constant in time suggesting zeroth order kinetics. Also, for one experiment the GaN surface was predosed with trimethylgallium for 10 minutes at 600 °C prior to the high temperature anneal in H₂. In agreement with the work of Pisch and Schmid-Fetzer who observed enhanced GaN decomposition on Ga predosed surfaces [Reference Pisch and Schmid-Fetzer13], a 34% increase in the GaN decomposition rate was observed on the Ga predosed surface. While the increase in the GaN decomposition rate by Ga predosing the surface is significant, it is not as large as the increase observed in Fig. 2 when the H₂ pressure is increased from low (< 76 torr) to high pressure (> 150 torr). From this work it is apparent that the GaN decomposition rates depend strongly on the ambient gas (i.e. H₂ or N₂), the pressure, and the condition of the initial surface.

Figure 5: The growth in the Ga droplet size is shown for GaN surfaces anneal at 811 °C for (a) 3, (b) 10, (c) 20, and (d) 80 minutes in H2 at a pressure of 150 torr.

Contrary to the large range in the measured kinetic parameters for GaN decomposition, the measured E_A for Ga desorption were found to be both independent of pressure and ambient gas composition. The data shown in Fig. 4 for H₂ were measured at 40 (solid circles), 76 (open circles), 150 (solid squares), and 250 (open squares) torr. From a fit to all the data a pre-exponential of (6.6±0.5)×10²⁹ cm⁻²s⁻¹ and an E_A of 2.74±0.06 eV were measured [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman7]. In N₂, the desorption rates were measured at 76 (solid diamonds) and 150 (open diamonds) torr, yielding a pre-exponential of (5.3±0.4)×10²⁸ cm⁻²s⁻¹ and an E_A of 2.69±0.08 eV. Both measurements of the Ga desorption E_A are in excellent agreement with the value of 2.8 eV for Ga desorption from liquid Ga [Reference Honig and Kramer14], and 2.69 eV for Ga desorption from GaN in vacuum [Reference Brandt, Yang and Ploog15]. The preexponential measured in H₂ is 12.5 larger than in N₂, which is close to the mass ratio of N₂ to H₂ of 14. The mechanism for Ga desorption suppression in N₂ and implications for GaN growth will be discussed in the next section.

Discussion And Conclusions

GaN thermal decomposition has been extensively studied in vacuum [Reference Munir and Searcy10-Reference Ambacher, Brandt, Dimitrov, Metzger, Stutzmann, Fischer, Miehr, Bergmaier and Dollinger12, Reference Johnson, Parsons and Crew16-Reference Bolgar, Gordienko, Ryklis, Fesenko and Samsonov19]. From these previous studies activation energies, E_A, of 3.1 eV [Reference Munir and Searcy10, Reference Groh, Gerey, Bartha and Pankove11] and 3.93 eV [Reference Ambacher, Brandt, Dimitrov, Metzger, Stutzmann, Fischer, Miehr, Bergmaier and Dollinger12] were measured as listed in Table 1. In this study, we have shown how the measured kinetic parameters vary depending on the pressure and ambient gas. For example, in N₂ a larger E_A (3.62 eV) is measured compared to a lower E_A (2.96 eV) measured in H₂ at 76 torr. The E_A are even lower in H₂ when the pressure is increased above 100 torr. These differences in the E_A reflect a change in the GaN decomposition mechanism when GaN is heated in H₂ vs. heating in N₂.

The decomposition rate enhancement at high pressure coincides with an increase in liquid Ga coverage. Recently, Pisch and Schmid-Fetzer showed that liquid Ga can catalyze GaN decomposition for temperatures as low as 720 °C [Reference Pisch and Schmid-Fetzer13]. Although Ga droplet formation coincides with an increase in the GaN decomposition rate, it is not known if the liquid Ga accumulation is the cause or a result of the increased GaN decomposition rate. In Fig. 2, we show that the GaN decomposition rate is enhanced at higher pressure in H₂. However, no enhancement is observed in N₂. This implies that the GaN surface is chemically altered in H₂ and that N is preferentially removed from the lattice while the Ga desorption rate remains relatively constant.

To preferentially remove N from the lattice, the H₂ must first dissociate and adsorb on the surface. Ga metal is known to dissociate H₂ at high temperatures to form Ga hydrides [Reference Remy20]. The hydrogenated N and Ga species have the potential to be both more mobile and also more volatile. More mobile hydrogenated Ga species have been proposed to explain the increase in the Ga diffusion length when H₂ or atomic H is used in the MBE growth of GaAs [Reference Morishita, Nomura, Goto and Katayama21]. In addition, Okamoto and coworkers have recently showed a suppression of 3D growth morphology when atomic H is used during MBE growth of GaN [Reference Okamoto, Hashiguchi, Okada and Kawabe22], implying an increase in the surface mobility for the hydrogenated Ga atoms. Increasing the Ga diffusion length would more rapidly uncover new areas of the GaN surface for N₂ desorption, which is 10-10⁴ times faster than the Ga desorption rate for the temperature range of 800-1100 °C [Reference Koleske, Wickenden, Henry, DeSisto and Gorman4, Reference Brandt, Yang and Ploog15]. Furthermore, the surface H can form more volatile NH_x species. Both an increased rate of NH_x desorption and an increased GaH mobility, will lead to the formation and growth of Ga droplets as shown in Figs. 1 and 5. When N₂ is substituted for H₂, the E_A for decomposition is larger because the chemical pathways for forming hydrogenated N and Ga species are absent. In N₂, the kinetic barrier to GaN decomposition is larger and may be limited by the formation and desorption rate of N₂.

Compared to the GaN decomposition kinetics, Ga desorption is simpler. The E_A listed in Table 1 all are in the range 2.69-2.8 eV, implying that the Ga desorption mechanism is similar under varying pressures and ambient gas flows. The agreement in the E_A for Ga desorption from liquid Ga [Reference Honig and Kramer14] and the similarity of the measured E_A values in Table 1, suggest that the Ga atoms desorb from a Ga rich surface. As shown in Fig. 4, the measured pre-exponential factor in H₂ is 12.5 times the measured pre-exponential factor in N₂, which is close to the mass difference of 14 between N₂ and H₂. With the N₂ and H₂ molecules initially near room temperature (21 °C), their impact with the hot surface (near 800-1050 °C) would likely result in a general heat removal from the surface. The heat transfer between the N₂ and the hot surface is more efficient than between H₂ and the surface, because of its larger mass (more impulsive collision) and slower speed (mean speed of N₂ is 3.7 times slower than H₂) its collision time with the surface is longer. As a result of the increased collision time, heat transfer from the hotter surface Ga atoms to the cooler N₂ molecules is more efficient, resulting in a reduced population density of the Ga surface vibrations which lead to Ga desorption. The net effect of this is a reduction in attempt frequency (i.e. preexponential factor) for Ga desorption in N₂ compared to H₂.

This study of GaN decomposition has several consequences for the growth of GaN. We have found that the material quality is substantially improved when GaN growth is conducted above 100 torr. When the GaN epitaxial layer is grown above 100 torr, we find a near doubling of the electronic mobility (µ > 500 cm²/Vs for intentionally Si doped films with n = 2-3x10¹⁷ cm⁻³) compared to films growth at 76 torr [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman7]. For films grown above 100 torr, the GaN grain size increased from < 1 µm to 2-5 µm, which may be directly responsible for the increased mobility [Reference Hersee, Ramer and Malloy23]. Other groups using close-spaced or high speed rotating disk reactors have also reported improved electric properties when the pressure of the GaN growth is greater than 100 torr [Reference McDermott, Pittman, Gertner, Krueger, Kisielowski, Lilienthal-Weber and Weber24, Reference Han, Ng, Biefeld, Crawford and Follstaedt25]. Larger grains and narrower x-ray rocking curve widths have been reported for the growth of unnucleated GaN on sapphire in H₂ compared to N₂ [Reference Kistenmacher, Wickenden, Hawley and Leavitt26], suggesting that the enhanced decomposition in H₂ aids in the breakup of smaller grains. Changes in the nucleation layer evolution during the ramp from low to high temperature have also been observed as a function of H₂ pressure [Reference Han, Ng, Biefeld, Crawford and Follstaedt25, Reference Ramer, Zheng, Kranenberg, Banas and Hersee27]. Enhanced GaN decomposition in H₂ may also increase the size and aid in the coalescence of the low temperature nucleation layers as shown by Han et al. [Reference Han, Ng, Biefeld, Crawford and Follstaedt25]. If insufficient NH₃ is supplied, the GaN nucleation layer may be entirely decomposed [Reference Kobayashi, Akasaka and Kobayashi28], especially if the nucleation layer is annealed under higher H₂ pressure.

This study illustrates the significant differences between reduced and atmospheric pressure MOVPE GaN growth and the influence of carrier gas chemistry. We have demonstrated the direct influence of H₂pressure on the GaN decomposition rate. A more thorough understanding of the GaN decomposition mechanism at the higher growth pressures currently used by many groups may serve to clarify the mechanisms contributing to GaN growth.