Introduction

Metallorganic vapor phase epitaxy (MOVPE) is currently being used to grow GaN for the fabrication of blue light emitting diodes [Reference Nakamura, Senoh, Iwasa, Nagahama, Yamada and Mukai1], lasers [Reference Nakamura, Senoh, Nagahama, Iwasa, Matushita and Mukai2] and for high power electronic devices [Reference Mohammad, Salvador and Morkoc3]. For MOVPE growth, NH₃ is typically used as the N source and high temperatures (> 1000 °C) are required to efficiently dissociate (i.e. 40-50 %) the NH₃ [Reference Liu and Stevenson4], because of the large N-H bond strength [Reference Weast5]. As a result, MOVPE growth temperatures are 100-500 degrees Celsius larger than the threshold temperature for GaN decomposition in vacuum [Reference Groh, Gerey, Bartha and Pankove6,Reference Grandjean, Massies, Semond, Karpov and Talalaev7] and in H₂ and N₂ [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman8,Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman9]. The high rate of N₂ desorption is compensated by using large flows of NH₃ [Reference Koleske, Wickenden, Henry, DeSisto and Gorman10], however the extent of GaN decomposition that occurs during growth has not been measured.

The recent studies of Grandjean et al. [Reference Grandjean, Massies, Semond, Karpov and Talalaev7] and Rebey et al. [Reference Rebey, Boufaden and Jani11] have shown dramatic decreases in the GaN decomposition rate when small NH₃ flows are dosed onto GaN surface. For example, Grandjean et al. measure a GaN decomposition rate of 5 Å/s at 875 °C in vacuum, while under an NH₃ flux of 1.7×10¹⁷ cm⁻² s⁻¹, the decomposition rate drops to 0.03 Å/s [Reference Grandjean, Massies, Semond, Karpov and Talalaev7]. To explain the decrease in the GaN decomposition rate in NH₃, a site-blocking model has been proposed where the adsorbed NH₃ blocks sites necessary for N₂ formation and desorption [Reference Grandjean, Massies, Semond, Karpov and Talalaev7]. A similar site blocking mechanism has also been proposed to explain reduced GaN growth when the NH₃ flux is increased [Reference Briot, Clur and Aulombard12]. In this paper we suggest that H also blocks sites on the GaN surface and H surface coverage effects must be considered in order to properly describe the GaN decomposition and growth kinetics.

Experimental Details

Details of the GaN growth [Reference Wickenden, Koleske, Henry, Gorman, Culbertson and Twigg13] and decomposition [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman8,Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman9] are discussed elsewhere. The GaN films used in this study were grown and decomposed in a close-spaced showerhead MOVPE reactor. The growth and decomposition rates were determined from weight loss using an analytical balance [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman8]. The GaN films were grown at 1030 °C using 32 µmoles of trimethylgallium (TMGa), 2 SLM NH₃ and 4 SLM of H₂ at pressures ranging from 40 to 300 torr. GaN decomposition was studied under similar flow conditions as the growth. The measured weights were converted to growth and decomposition rates per surface area (i.e. cm⁻²s⁻¹) following Ref. 10. Expressed this way, a rate of 1.14×10¹⁵ cms⁻² s⁻¹ corresponds to a thickness of 1 µm per hour. Temperature was calibrated by observing the melting point of 0.005” diameter Au wire on sapphire and correlating it to a thermocouple in direct contact with the susceptor underside [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman8]. After 2 years of use the set point temperature needed to melt the Au wire was reproducible to within 10 °C.

Results

The change in the GaN decomposition rate, k_GaN, at a temperature of 992 °C as NH₃ is substituted for the H₂ flow is shown in Fig. 1. In Fig. 1, k_GaN is plotted as the NH₃ flow is increased from 0 to 2 SLM for pressures of 40, 76, and 150 torr. The total flow rate was kept constant at 6 SLM with the balance being H₂. Note that k_GaN decreases from ≈ 1×10¹⁶ cm⁻²s⁻¹ to ≈ 1×10¹⁴ cm⁻²s⁻¹ as the NH₃ flow increases. At 150 torr, the minimum k_GaN occurs at a flow of 0.75 SLM of NH₃. At 76 torr, the k_GaN minimum occurs between 1.25 and 2.0 SLM of NH₃. At 40 torr no minimum in k_GaN is observed. At 150 torr and a flow of 2 SLM of NH₃, k_GaN is ≈ 4×10¹⁴ cm⁻²s^−1.

Fig 1. Plot of the GaN decomposition rate as a function of the NH₃ flow rate at three different pressures. For this plot the GaN was heated to a temperature of 992 °C using H₂ and NH₃ for a total flow of 6 SLM.

In Fig. 2, the data from Fig. 1 are replotted as a function of the NH₃ gas density, which depends on pressure. In Fig. 2, it appears that the k_GaN measured at different pressures have a common minimum at a NH₃ density of ≈ 1×10¹⁹ cm⁻³. This NH₃ density is the same order of magnitude as the calculated N desorption rate from GaN, which should be 9×10¹⁹ cm⁻²s⁻¹ at 992 °C [Reference Koleske, Wickenden, Henry, DeSisto and Gorman10]. Two different dependencies of k_GaN on the NH₃ density are evident in Fig. 2. At lower NH₃ density, the k_GaN drops steeply as the NH₃ density increases from 3×10¹⁷ cm⁻³ to a value of near 1×10¹⁹ cm⁻³. For NH₃ densities greater than 1×10¹⁹ cm⁻³, k_GaN increases. This differs from the behavior observed by Grandjean et al., where k_GaN only decreased for increasing NH₃ flow [Reference Grandjean, Massies, Semond, Karpov and Talalaev7].

Fig. 2. Plot of the GaN decomposition rate measured as a function of the NH₃ density at 992 °C. The filled circles (red) were measured at 150 torr, the open squares (blue) were measured at 76 torr, and the filled diamonds (green) were measured at 40 torr. The lines are fits to the data using the expression k_GaN = a + b[NH₃] + c[NH₃] ^x . For the fits the values of a and b are the same, while the value of x is fixed from −1.0 to −3.0 and c is varied for the best fit.

To determine the dependence of k_GaN on the NH₃ density, [NH₃], separate fits were calculated for [NH₃] both less than and greater than 1×10¹⁹ cm⁻³. For [NH₃] < 1×10¹⁹ cm⁻³, a functional form of k_GaN = c[NH₃] ^x was used and the data were fit by varying c, keeping x constant. For [NH₃] > 1×10¹⁹ cm⁻³, a linear functional form, k_GaN = a + b[NH₃], fit the data well. The series of lines shown in Fig. 2 are a combination of the two fits (i.e. k_GaN = a + b[NH₃] + c[NH₃] ^x ). For the combined fits, 5 curves were calculated for 5 values of x ranging from −1.0 to −3.0, keeping the linear fit constant. Clearly, the data are best fit using with x = −1.5 to −2.0.

Similar to NH₃, the GaN decomposition rate in N₂ is lower when compared to the rates measured in H₂ [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman9], however, in mixed N₂ and H₂ flows the rate is substantially larger than in mixed NH₃and H₂ flows. This is shown in Fig. 3, where k_GaN is plotted vs. the N₂ fraction of the total flow (i.e. [N₂] + [H₂]). For these measurements, the GaN films were annealed at 992 °C at pressures of 76 and 150 torr. In Fig. 3, k_GaN at 150 torr is reduced from 1.6×10¹⁶ cm⁻²s⁻¹ in pure H₂ to 3.5×10¹⁴ cm⁻²s⁻¹ in pure N₂ (factor of 45). For a 1:1 mixture of N₂ and H₂ at 150 torr, k_GaN decreases slightly to 8×10¹⁵ cm⁻²s⁻¹, which is a factor of 2 compared to k_GaN in pure H₂. This is a significantly smaller decrease when compared to the decrease in a 1:2 mixture of NH₃ and H₂ (factor of 120). Note that k_GaN in pure N₂ and in mixed H₂ and NH₃ flows can be similar. For example in pure N₂, k_GaN is 3.5×10¹⁴ cm⁻²s⁻¹, while in mixed H₂ and NH₃ at 150 torr, k_GaN is 4×10¹⁴ cm⁻²s⁻¹, as shown in Fig. 1. The solid and dashed lines in Fig. 3 are cubic fits to the k_GaN vs. N₂ fraction. The cubic dependence is a result of the expected dependence of surface H coverage (i.e. [H]³) for ammonia formation via the reaction 3H + N → NH₃. Previously, Thurmond and Logan also demonstrated NH₃ formation when GaN is heated in H₂ by titration of the basic exhaust gas [Reference Thurmond and Logan14].

Fig. 3. Plot of the GaN decomposition rate vs. the ratio of the N₂ concentration to the total (i.e. N₂ + H₂) gas concentration. The GaN decomposition rate is shown at total pressures of 76 and 150 torr. The solid and dashed lines are cubic fits to the data.

In Fig. 4, the GaN decomposition and growth rates are plotted vs. pressure. In Fig. 4(a), k_GaN is plotted for GaN films annealed at 992 °C in H2 [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman8]. Also in Fig. 4(b), the GaN growth rate at 1030 °C is plotted for conditions where 2 SLM NH₃, 4 SLM H₂, and 32 µmoles of TMGa were used. Finally, in Fig. 4(c) the GaN decomposition rate is plotted using the same conditions as (b) except no TMGa was used and hence decomposition was observed. Note that the decrease in the GaN growth rate as the pressure increases in Fig. 4(b) coincides with an increase in the GaN decomposition rate in Fig. 4(c). Also, the k_GaN shown in Figs. 4(a) and 4(c) have a similar shape as the pressure increases and these curves are nearly identical if the k_GaN in Fig 4(c) are multiplied by 30. This similarity in shape implies that surface H plays a similar role in the GaN decomposition for both pure H₂ and mixed NH₃ and H₂ gas environments.

Fig. 4. Pressure dependence of the a) GaN decomposition rate in 6 SLM H₂ measured at T = 992 °C, b) GaN growth rate using, 32 µm TMGa, 2 SLM NH₃ and 4 SLM H₂ at T = 1030 °C, and c) GaN

Discussion and Conclusions

From the data presented in Fig. 1, the GaN decomposition rate is greater than 1×10¹⁴ cm⁻²s⁻¹ (i.e. ≈ 1/10 μm/hour) even in mixed NH₃ and H₂ flows. This is important for GaN growth because it suggests that some level of decomposition occurs during decomposition rate using 2 SLM NH₃ and 4 SLM H₂ at T = 1030 °C. The only difference between b) and c) is the use of TMGa in b).

growth as previously speculated [Reference Koleske, Wickenden, Henry, DeSisto and Gorman10]. Currently, we are growing GaN at a pressure of 130 torr and a temperature of 1030 °C [ Reference Wickenden, Koleske, Henry, Gorman, Culbertson and Twigg13]. Under these growth conditions, the rates for growth and decomposition are 1.2×10¹⁵ cm⁻²s⁻¹ and 4×10¹⁴ cm⁻²s⁻¹ respectively as shown in Fig. 4. If the growth rate equals the incorporation rate minus the decomposition rate [Reference Koleske, Wickenden, Henry, DeSisto and Gorman10], this means that the incorporation rate is ≈ 4 times decomposition rate under these growth conditions.

The decrease in k_GaN as the NH₃ density increases is due to NH₃ adsorption, which blocks sites needed for GaN decomposition. As shown in Fig. 2, the decrease in k_GaN depends on the −1.5 to −2.0 power of the NH₃ density. Since Ga desorption from GaN has been shown to be independent of H₂ pressure [Reference Koleske, Wickenden, Henry, Twigg, Culbertson and Gorman8], reactions which remove N from the lattice probably influence the GaN decomposition rate more. This is clearly observed in the cubic dependence of k_GaN in Fig. 3 where NH₃ formation is favored at higher pressure. For N₂ formation and desorption one or both of the N atoms diffuse across the surface until they combine to form N₂. If open surface sites are necessary for N diffusion, blocking of these sites by NH₃ or H would decrease the hopping rate and as a consequence the N₂ formation rate would be decreased. If two (one) N must migrate for N₂ formation, the N₂ desorption kinetics would be second (first) order in the number of open surface sites. As the NH₃ density on the surface increases, the decrease in the GaN decomposition rate should be between first and second order, i.e. k_GaN α [NH₃]⁻¹ or k_GaN α [NH₃]⁻², depending on the details of N₂ formation and desorption. From Fig. 2, it is clear that the decrease in k_GaN vs. NH₃ density is closer to second order (power of −2) than first order.

At higher NH₃ densities (> 1×10¹⁹ cm⁻³), the GaN decomposition rate increases linearly. This may be due to a decrease in the NH₃ site blocking suppression of N₂ desorption or a general increase in the H surface coverage. The increased H coverage could block sites necessary for NH₃ adsorption. Surface H has also been shown to aid in NH₃ adsorption and dissociation on GaN [Reference Bartram and Creighton15] and on Al [Reference Kim, Bermudez and Russell16]. In addition, large H coverage can favor NH₃ reformation and desorption by combining with adsorbed NH_x species as suggested by Fig. 3. In contrast to NH₃, site blocking with H should lead to an increase in the decomposition rate.

Several groups have observed decreases in growth rate when H₂ is used in place of N₂ [Reference Ambacher, Brandt, Dimitrov, Metzger, Stutzmann, Fischer, Miehr, Bergmaier and Dollinger17, Reference Hashimoto, Amano, Sawaki and Akasaki18], when the growth pressure is increased [Reference Khan, Skogman, Schulze and Gershenzon19], and when higher NH₃ fluxes are used for growth [Reference Briot, Clur and Aulombard20]. In Fig. 4(b), the GaN growth rate decreases as the growth pressure increases. It is clear from Fig. 4(c) that the reason the growth rate decreases is because the increased GaN decomposition at higher pressures. However, to fully explain the reduction in the growth rate, the full effect of gas phase depletion of the TMGa also needs to be considered.

GaN grown in H₂, where GaN decomposition is enhanced compared to N₂, appears to have better crystalline order compared to GaN growth in N₂. Kistenmacher et al have shown that the FWHM of the GaN films grown in H₂ had narrower x-ray rocking curve linewidths and were better aligned compared (i.e. smaller mosaic dispersion) to GaN films grown in only N₂ [Reference Kistenmacher, Wickenden, Hawley and Leavitt21]. Better alignment is also observed in laterally overgrown GaN when H₂ is used instead of N₂ [Reference Tadatomo, Ohuchi, Okagawa, Itoh, Miyake and Hiramatsu22]. Schön and coworkers find smoother morphologies and better electrical properties when growth is conducted in H₂ compared to N₂ [Reference Schön, Schineller, Heuken and Beccard23]. Better electrical properties are observed for GaN grown at higher pressures where GaN decomposition is enhanced [Reference Wickenden, Koleske, Henry, Gorman, Culbertson and Twigg13, Reference Bell, Smith, McDermott, Gertner, Pittman, Pierson and Sullivan24, Reference Ng, Han, Biefeld and Weckwerth25]. Recently, we have observed a near doubling of electron mobility in films grown at 150 torr compared to 76 torr, keeping all other growth parameters the same [Reference Wickenden, Koleske, Henry, Gorman, Culbertson and Twigg13]. In this study, growth at higher pressure led to increased GaN grain size in the films [Reference Wickenden, Koleske, Henry, Gorman, Culbertson and Twigg13], suggesting that the increased GaN decomposition at higher pressure plays a significant role in determining the grain size of the GaN film.