Bulk GaN: Ammonothermal trumps HVPE

CURRENT FEATURES

Bulk GaN: Ammonothermal trumps HVPE

Mar 01, 2010

Today’s GaN substrates are manufactured by a HVPE process that requires high temperatures and substantial reactor maintenance. Ammonothermal growth can address both these issues, while producing material with far fewer dislocations in a more efficient manner, says Ammono’s Robert Dwilinski, Roman Doradzinski and Marcin Zajac.

GaN-based devices have an important role to play in the portfolio of energy-saving technologies making an ever-increasing impact on our world. LEDs based on this wide bandgap material can deliver incredibly efficient light emission, and nitride transistors are promising devices for efficient, high-power output at high frequencies.

Although nitride based devices are already capable of delivering an impressive performance, they are held back by the limited availability and high cost of a native substrate. Devices are usually built on sapphire, silicon and SiC, and heteroepitaxial growth of nitrides on these platforms leads to a high density of defects in these films. These result from a difference in lattice parameters and thermal expansion coefficients between the materials. What is needed by the industry is a bulk GaN substrate that can drive improvements in the efficiency of optoelectronic and high-temperature electronic devices, but does not cost the earth.

Developing a GaN substrate is very challenging, because it is impossible to use standard methods to grow a boule. The Czochralski and Bridgman techniques employed for GaAs manufacture are not applicable, because GaN decomposes into gallium and nitrogen gas.

It is possible to grow high-quality GaN crystals by combining high temperatures (of about 1500 °C) with extremely high nitrogen pressures (of the order of 15 kbar). But such high pressures prevent the use of large growth chambers. In addition, crystal seeds cannot be used, which imposes serious limitations on crystal size.

Other techniques have also been developed, including HVPE, a growth technology used for most of today’s freestanding GaN substrate production. Although material produced by this technique is undeniably a commercial success - it has provided the bedrock for 405 nm lasers deployed in Blu-ray players - it suffers from a high dislocation density that stems from the use of non-native seeds. Even after the seed has separated, the freestanding HVPE GaN is still highly stressed and bowed.

Better quality material can be produced by a sodium-flux technique, which involves the growth of GaN crystals in a vessel containing a gallium-sodium mixed metal melt and pressurized nitrogen gas. However, there are still many problems to solve, such as poor growth of seeds, heterogeneity, mosaicity and poor scalability.

Ammonothermal growth

At Ammono, a company which is based in Warsaw, Poland, we are pursuing a more promising technique involving convection-driven transport of an ammonia solution, followed by crystallization of GaN on native seed crystals. Advantages of our approach include growth in equilibrium conditions, growth in a closed system and scalability.

Our success in manufacturing GaN stems from long-term experience in ammonothermal crystallization of this material. Efforts in this direction can be traced back to the early 1990s, when two of us (Robert Dwilinski and Roman Doradzinski), plus Leszek Sierzputowski and Jerzy Garczynski built the first ammonothermal set-up for GaN synethesis at the Institute of Experimental Physics, at the University of Warsaw. The first breakthrough was the growth of micro-crystalline GaN powder by a chemical reaction between gallium and ammonia. Alkali-metal amides, such as LiNH2 or KNH2, were added into the reaction zone to play the role of mineralizers, highly increasing the reactivity of the solution.

Growth at 550 °C and 5 kbar produced GaN crystals in the form of a wurtzite-type microcrystalline powder with grains up to 5 μm in size. This material revealed highly intensive photoluminescence with very sharp peaks of near band-edge emission accompanied with a weaker parasitic yellow band. These promising results sparked the formation of our company in 1999, renting at the very intial stage several labs owned by Polish government institutes. At this time we also started to collaborate with Nichia Corporation, Japan.

We have learnt how to produce relatively large pieces of GaN by taking advantage of the chemical transport of ammonia solution in a temperature gradient. In 2003 we started selling GaN substrates, although these were not available on the open market at that time. Three years later we transferred to our own facility, an incredibly beneficial move that allowed us to design a laboratory and production and office facilities tailored to our specific needs. Two years ago we made a further investment, installing large-diameter autoclaves capable of simultaneous growth of many GaN crystals. The workforce has also increased, and today we have 50 highly trained staff.

Thanks to refinement in our ammonothermal technology, we can now manufacture high-quality, bulk c-plane GaN seeds up to 2-inches in diameter with perfect crystalline quality. Recent additions to our product portfolio include non-polar m-plane, a-plane and semi-polar GaN substrates. These provide a platform for fabricating blue and green lasers and LEDs that are free from the strong internal electric fields hampering optoelectronic devices grown on conventional, polar surfaces.

Fig.3. m-plane non-polar GaN can be made by slicing material from a piece of 12 mm-thick, oneinch GaN. The non-polar substrate is 11 mm by 22 mm in size

The ammonothermal technique that we have adapted is an analogue of the hydrothermal technique used for commercial mass production of α-quartz. GaN-containing feedstock is dissolved in one zone of the high-pressure autoclave, before being driven by a temperature gradient to a crystallization zone. Here, GaN crystallizes on native seeds thanks to the supersaturated solution (Fig. 1).

Typical temperatures and pressures are 0.2- 0.5 GPa and 500 °C - 600 °C, respectively. Mineralizers are added to enhance the solubility of GaN in ammonia. Our growth technology is actually an ammonobasic version of the ammonothermal technique, with pure alkali metal or alkali metal amides such as LiNH2, NaNH2 or KNH2 used to introduce NH2 - ions into the solution. At the start of the previous decade we realized an unusual, but beneficial feature of this particular approach - the solubility of the solution decreases with increasing temperatures. The consequence is that soluble GaN can be transported from a low-temperature dissolution zone offering high solubility to a higher-temperature crystallization zone with lower solubility. To realize an efficient re-crystallization process with this approach, the high-temperature, seed-containing zone has to be placed below the low-temperature zone containing feedstock (see Fig. 1).

Ammonothermal growth has several strengths: it enables growth of high-diameter, truly bulk seeds with perfect crystalline quality; it is highly controllable and reproducible at process temperatures of just 500-600°C and pressures of 0.2-0.5 GPa; and it is perfectly scalable with the size of the autoclaves. The dimensions of the autoclave are the only limit to the size of the crystal, and it is possible to grow hundreds of them in one run.

Further advantages are the conversion of almost all the feedstock material into the final product, thanks to the use of a closed system; relatively easy reactor maintenance; and growth that can be continued up to any thickness, because the crystal quality does not deteriorate as the process time is increased. The latter benefit holds the key to the growth of quality, non-polar substrates of any size.

In comparison, growth of GaN by HVPE is hampered by the use of non-native seeds, far higher growth temperatures of 1100°C, and the use of an open reactor rather than a closed system. The later weakness means that only a small fraction of the raw materials are converted into the product.

For example, just 5-15 percent of the flowing GaCl3 incorporates gallium into the GaN crystal, which equates to wasting at least 85 percent of gallium. If HVPE reactors are used to grow GaN, then there is also the need to etch away parasitically nucleated crystals and regularly exchange elements, costly tasks that take time and impair productivity.

Material benefits

We manufacture 1-inch and 1.5-inch diameter, c-plane orientated, polished GaN substrates that have been sliced and round-shaped from a larger crystal. Our ammonothermal method allows scaling of substrate sizes, leading to production of 2-inch seed monocrystals (see Fig.2). We now plan to ramp the production and shipment of 2-inch GaN substrates to a high volume after building up a sufficient stock of seeds.

By slicing thick, 1-inch diameter GaN crystals, we have created non-polar substrates with a surface area of several square centimeters (see Figure 3). Semi-polar substrates, such as the (2021) orientation that has been used to make a green laser, can also be realized by this approach. We have set ourselves the target of producing the first 1-inch non-polar wafer in the near future.

Hall effect experiments and contactless methods have verified that it is possible to control the electrical properties of the substrates with appropriate doping to realize n-type, p-type and semi-insulating material. Tuning the electrical properties of our material will enable it to find application in both optoelectronic devices requiring a highly conductive platform, and HEMTs that must be grown on a semi-insulating substrate.

The exceptional crystallinity of our bulk material is revealed by X-ray rocking curve measurements that show a peak with a full width at half maximum (FWHM) of just 16 arcsec (Fig. 4), measured for the (0002) crystallographic plane (c-plane substrates), (1010) plane (non-polar substrates), and (2021) plane (semi-polar substrates (Fig.4). Incredibly low values have also been obtained for appropriate asymmetric planes.

One of the downsides of GaN that has been produced by HVPE is its high internal stress, which leads to an unwanted curvature of the crystal planes. We have measured this bending by studying the systematic shift in the maxima of diffraction peaks on the Ω-axis (see Fig.4). This effect reflects the systematic change of the inclination of the (0002) plane when moving along the measurement line. The X-ray data has been used to calculate a radius of curvature for HVPE-grown GaN of 2-12 m, which is at least three orders of magnitude smaller than that for ammonothermal GaN (see Fig. 4). In other words, the crystalline quailty of our GaN is extremely flat, indicating that there is very little built-in stress in our material. The high degree of crystallinity of this material allows it to be used as a seed for the growth of additional pieces of bulk GaN without any loss in the crystal quality of the product.

Ammono GaN substrates ready to dispatch

Although the dislocation density in HVPE-grown GaN continues to improve, typical values are still of the order of 106 cm-2. This density is far higher than that for ammonothermal GaN - after etching the material in potassium hydroxide, hexagonal pits were observed by microscopy with a density of just 5 x 103 cm-2 (see Figure 5).

The combination of mechanical and chemical-mechanical polishing has created epi-ready surfaces that have provided a base for homo-epitaxial growth of thin films with excellent properties. Optical and X-ray measurements indicate that it is possible to grow strain-free homoepitaxial layers with excellent quality on our polar and non-polar substrates. Photoluminescence is dominated by an intensive, perfectly resolved excitonic structure that is uniform across the entire sample surface. The width of the bound exciton peak is just 0.3 meV.

Reflectance spectra reveal the truly non-polar character of m-plane oriented GaN (see Figure 6), and X-ray and microscopic measurements show that the resulting epitaxial layer has a high crystal quality with very few dislocations. The FWHM of the X-ray diffraction peak is just 22-25 arcsec, and the threading dislocation density is less than 5 x 104 cm-2.

Although the ammonothermal growth rate is much lower than that for HVPE, its perfect scalability makes it by far the most promising method for high-volume manufacture of bulk GaN, partly because it is possible to produce hundreds of crystals in one run. Our next goals are further development and up-scaling of this method, plus the realization of lower operating costs by further automatization of the process.

If we can execute on this front, we will deliver lower-cost, higher-quality substrates than the HVPE-produced material on the market today. And that should ultimately lead to a hike in the performance of commercial, energysaving, high-power optoelectronic and electronic devices.