Electrolysis: a surprising method for scaling GaN crystal growth?

Native substrates are the ideal platform for GaN lasers and high-power LEDs. But manufacturing this material is tricky, to say the least. If standard melt-based techniques are employed, bulk growth of high-quality single-crystal GaN and InN demands incredibly high melting points and nitrogen gas vapor pressures. For GaN, nitrogen s vapor pressure at its equilibrium melting point exceeds 6 GPa and for InN it is even higher.

This melt-type approach also suffers from very slow growth rates that ultimately restrict the size of the boule that can be produced. Even the most successful exponents of this technique, the Unipress group at the Institute of High Pressure Research Center in Warsaw, Poland, can only produce crystals up to 25 mm in size.

At Sandia National Laboratories in Albuquerque, NM, we have started to investigate an alternative process for producing bulk GaN – electrochemical solution growth. Funded by the Department of Energy s solid-state lighting program, this approach promises to produce bulk GaN from a molten salt. If the technique is successful, it will offer a fast and simple method for bulk growth of all forms of metal nitride, including InN, AlN and GaN. This is because growth from ionic precursors in solution can eliminate the high overpressures that are essential for conventional techniques, while simultaneously addressing the issue of slow kinetics.

One obstacle facing any solvent-based approach is the difficulty associated with dissolving GaN. The most successful approach to date is ammonothermal synthesis, which involves a liquid ammonia solvent and mineralizers, such as sodium-based materials, which are added to increase GaN solubility. The awkward conditions required for this form of growth make the technique difficult to scale, so a batch approach is often adopted, with numerous seeds added to each growth run. The requirement for forming liquid ammonia at the typical growth temperatures of 500–800 °C is particularly irksome because this requires pressures of 4000–5000 atm. However, it is worth noting that significant progress has been made through refinements to the growth apparatus. Fumio Kawamura and co-workers from Osaka University, Japan, have managed to produce 2 inch material by this method.

The traditional method for making other refractory materials, such as SiC and oxides of magnesium, also tend to require precursors to be dissolved at high temperatures. With this approach, the solid crystal forms as the solution cools. High temperatures are needed at the start, so that as the crystal grows and the precursor concentration falls, lower temperatures can be introduced to maintain the supersaturated solution and ultimately ensure continual growth. But there is an inherent drawback – it is not possible to use the ideal crystal growth temperature throughout the process.

A key advantage of our approach is that it avoids all of these problems and forms precursors in situ by electrochemical means. GaN is difficult to dissolve in molten chloride salts, but its ionic precursors are highly soluble and these can be continuously formed in solution by oxidizing gallium metal and reducing nitrogen gas. This means that the molten chloride salt acts as both a solvent and an electrolyte.

A key part of our process is a nitrogen reduction reaction to form negatively charged nitrogen ions. This reaction has previously been studied by Takuya Goto and Yasukio Ito from Kyoto University, Japan. They have demonstrated that it is possible to continuously reduce nitrogen gas to N3– ions in a molten chloride salt at 450 °C. However, they had some very different applications in mind from GaN growth. Nuclear fuel reprocessing and hardness coatings were their main research themes, and they used the reduction reaction to electroplate nitride films on the surface of several transition, lanthanide and actinide metals.

If conversion of the nitrogen gas to ions is very slow, it could restrict the GaN growth rate. But it is possible to produce nitride ion concentrations of 2.9 mol% or greater in LiCl at 650 °C, according to the research by Wataru Utsumi and co-workers from the Synchrotron Radiation Research Center at Japan s Atomic Energy Research Institute. This means that if the creation of nitrogen ions is the growth-limiting process, then the GaN crystal s growth rate would still exceed 1 mm/h under diffusion-limited conditions. Such a high growth rate could be great news for our community – it could ultimately lead to cheaper substrates for low-cost, high-throughput commodity applications such as LEDs for solid-state lighting.

Early steps
We are aiming to use this type of reaction to develop a process for making large, high-quality bulk nitride crystals. Our preliminary experiments have focused on the reaction of nitride ions with gallium in molten salt. We used electrolysis to create GaN at the interface of a molten gallium pool (figure 1), with Li3N providing the nitride ion source.

Under atmospheric pressure and an inert gas environment, we dissolved Li3N in a fused LiCl–KCl eutectic solution at 450 °C. A current of 5 mA was swept for two hours to oxidize the gallium in the presence of nitride ions. The melt was then cooled to room temperature before the salt was rinsed away with deionized water and the gray insoluble GaN submillimeter-sized particles were filtered.

Scanning electron microscopy and X-ray diffraction (figure 2) revealed GaN with the required and expected wurtzite crystal structure. No photoluminescence was observed under optical excitation at 266 nm, but this was not surprising. We believe that this experimental configuration will produce lithium reduction at the cathode and gallium oxidation at the anode. Since lithium is a liquid at the growth temperature, and it is suspended above the gallium pool in this experimental configuration, it is possible that this metal could have dripped into the pool and contaminated the GaN. Or, the crystals might simply be defective due to the very low growth temperature (450–500 °C) that is used in this unoptimized experiment.

A light gray powder was also produced during our first attempt at making GaN from the reduction of nitrogen gas. The growth of larger crystals was probably hindered by the bubbling of nitrogen gas near the surface of the molten gallium pool in the small quartz test-tube reactor, which continually disturbed the growth surface.

However, even though this particular electroplating process is unsuitable for growing high-quality, thick, insulating crystals, the formation of GaN by this relatively simple method is very encouraging. This is because the ability to reduce nitrogen to nitride ions, which can be present in large concentrations in solution, could provide the key pathway to controlled bulk growth of large-diameter GaN.

Our next-generation reactor, which should enable scaling of our GaN growth (figure 3), involves the formation of N3– ions from nitrogen at the cathode, which is balanced by the production of Ga3+ ions from gallium at the anode. In a motionless solution this would cause gallium and GaN plating at the electrodes. However, this situation is avoided by a rotating seed crystal that is held just below the surface of the melt. This scenario is analogous to both the MOCVD vertical rotating disk reactors used for nitride growth and the rotating disk electrode, which has been used to study electrochemical reactions and diffusion rates of species in solution.

We plan to use our spinning seed crystal to control the fluid dynamics and draw the ionic precursors to the seed surface. These ions can then diffuse across the boundary layer and react on the seed s surface. Continuing the analogy with the rotating-disk MOCVD reactor, the electrodes substitute for the mass-flow controllers and the molten salt takes the place of the carrier gas.

This process promises stable growth because the ionic precursors can be introduced into solution at precisely the same rate that they are being consumed by the growing crystal. This process takes place at or near atmospheric pressure, so there are no obvious barriers that could prevent scaling of the crystal dimensions in the horizontal plane.

It should be possible to pull a boule from the surface of this melt because the growth region is close to the solution s surface. The precursors can also be replenished while they are being consumed, so their concentration should not waiver. This means that steady-state conditions are maintained throughout the process and the growth surface can be set to a constant, optimum growth temperature. If a thermal gradient is required, this can be introduced and maintained throughout the growth. In fact, we cannot see any barrier to prevent boule growth continuing for as long as is practically desired.

The process that we are proposing would have a reaction rate controlled by the nitride ion, which is encouraging from a reaction kinetics perspective. Our GaN electroplating experiments showed that GaN can grow at a rate of 1 mm in two hours under non-ideal conditions. This suggests that the reaction will proceed at a fast enough rate to yield high growth rates, but slow enough to prevent small crystals from forming in solution.

One potential pitfall for our process is the reaction of the gallium and nitrogen ions outside the boundary layer, rather than on the seed crystal. This reaction is a distinct possibility with the fluid flow patterns we expect to have in our crucible, but we believe that this scenario can be avoided by inserting a divider between the electrodes. This would isolate the gallium and nitrogen ions until they reach the boundary layer. Under such a scheme, separate potentiostats or power supplies would be used to control each reaction.

We are currently characterizing all of the electrochemical reactions – in particular the nitrogen reduction – as a function of temperature and several other relevant criteria. This will be followed by attempts to demonstrate GaN growth on a seed at the desired location in the melt. If this is successful, the last milestone will be to replicate this growth process on a high-quality template, such as a HVPE-grown seed crystal. This will require optimization of temperature, which will most likely have to be increased from 450 °C.

Under our current program, which is being funded by the National Energy Technology Laboratory, we will ignore the role of impurities present in the molten salt, and thermal management. Both of these factors could have a detrimental impact on growth quality. Nevertheless, the experiments that we have planned will determine whether our technique has the potential to deliver affordable high-quality, large-area GaN for our community.

Our technology has tremendous potential, and although it is still in its infancy, we are already collaborating with a small company in Silicon Valley to develop manufacturing methods in conjunction with the research at Sandia. This will ensure that our technology will be compatible with mass production processes.

Further reading
T Goto et al. 1998 Electrochimica Acta 43 3379.
H Tsujimura et al. 2004 J. Electrochem. Soc. 151 D67.
T Okabe et al. 2001 J. Electrochem. Soc. 148 E219.