AlN: can it become a universal substrate for III-nitrides?

III-nitrides are undoubtedly a remarkable family of semiconductors. Unlike their semiconductor cousins, high-quality films of this material can be grown on almost any substrate - sapphire, silicon and even glass.

However, despite their propensity for forming single crystals on practically anything, no III-nitrides occur naturally. Consequently, there has been continuing interest in manufacturing a native, single crystal substrate for these nitrides since the first demonstration of epitaxial growth of GaN by HVPE in 1969 by Paul Maruska and co-workers at RCA Sarnoff. Over the intervening decades nitride devices have kicked on to demonstrate unexpectedly good performance characteristics despite high dislocation densities. Blue LEDs, for example, can realize excellent reliability and power at dislocation densities of 5 x 109/cm2, a value that would kill light emission in their GaAs and InP longer wavelength cousins. However, in general, semiconductor devices have shown highest performance on low-defect native substrates, so there is no reason to suspect that the III-nitrides, even with their unique growth habit, will be any different.

There are three options for native substrates: AlN, GaN and InN. To date, no one has produced bulk crystal InN; GaN has been grown from solution at high pressure; and AlN has been produced using relatively straightforward physical vapor transport (sublimation). Another strength of AlN is that it is the most promising universal substrate for epitaxy of a wide variety of nitride devices, including LEDs, lasers, RF and surface acoustic wave (SAW) devices. Of all the III-nitrides, it has the largest bandgap; highest thermal conductivity, breakdown electric field and SAW velocity; and the smallest a-lattice parameter. What’s more, deposition of high-quality epilayers on this platform is relatively easy, because all AlInGaN compositions are in compression when grown on AlN, thus minimizing cracking probability.

However, although the PVT process used to grow true bulk AlN crystals was first identified by Glen Slack and co-workers at GE Research Lab more than 35 years ago, it has proven extraordinarily difficult to grow large diameter, low-defect crystals, even with well established experience in SiC growth. Initially, there was little interest in scaling crystal dimensions of AlN, but this has now changed thanks to the dramatic success of the nitrides.

Today, a handful of companies have reported success in developing at least small, high-quality AlN substrates. These include the US firms Crystal IS, Hexatech and Fairfield Semiconductor; the Japanese materials specialist Sumitomo Electric Industries; the German outfit CrystAlN; and ourselves, Nitride Crystals, which has bases in both the US and Russia. To our knowledge, we, along with Fairchild, are the only companies shipping AlN substrates on a commercial basis. We focus on sales of round substrates, while Fairchild ships 10 mm squares. Of the other players, CrystAlN has recently entered the market, and Crystal IS and Hexatech have reportedly established internal production of AlGaN devices such as deep UV LEDs. Perhaps the fact that both Crystal IS and Hexatech have decided to focus their AlN wafer manufacturing toward their own device products is the clearest indication of the potential importance of that material.

Our approach to operating in the AlN substrate and device market is based on this belief: The AlN substrate and device markets will never be significant unless major device players and substrate manufacturers adopt the technology. Therefore, we place no restrictions on how our customers use our AlN substrates. Production of our substrates has been governed by two major factors: nitride epitaxy on sapphire has a dislocation density of 5 x 107 – 5 x109/cm2, yet LED performance is excellent; and commercial manufacturers of optical devices need 2-inch wafers as a minimum size. Taking these factors into account, we have set ourselves the goal of scaling our wafer capability to 2-inch diameter as quickly as possible while maintaining quality that is “good enough.” What is our take on “good enough”? That the density of substrate dislocations is three-to-six orders of magnitude lower than that of epitaxy on sapphire. Figure 1 AlGaN epitaxy on AlN

We have developed production technology for delivering a range of 15 mm diameter AlN crystals and epi-ready substrates, which can be shipped with the aluminum face epi-ready polished, and the nitrogen-face polished or fine lapped, with US flats. All variants have excellent crystallinity, low dislocation density, high UV transparency and high resistivity. This is borne out by X-ray diffraction maps that show that the single peaks of the best substrates have a full-width half maximum less than 100 arcsec in both asymmetric and symmetric scans. These substrates provide a foundation for growing high-quality AlGaN layers (see Figure 1). We have also recently shown that it is possible to produce larger crystals - we have achieved fully mono-crystalline 2-inch diameter AlN by using low defect SiC seeds that we also grow by PVT.

Scaling to 2-inch

Early in our AlN development we realized that AlN can be successfully seeded by PVT on SiC. The growing AlNand- SiC seed forms a solid solution. Near the interface the concentration of silicon and carbon in AlN can be higher than 5 percent, giving rise to conductive AlN; however, the concentration of these two elements falls rapidly with distance from the interface, where resistivity is far higher. Second generation (grown on AlN seeds) AlN has resistivity in excess of 5 x 1011 Ohm-cm. We have not studied the effects of silicon concentrations of more than 5 percent in AlN, but on the basis of the seed growths we speculate that it will be possible to make conductive AlN (or perhaps more correctly AlSiN?).

Figure 2 Thick, micropipe-free 60 mm diameter 6HSiC seed

A micropipe-free, low dislocation seed is required only for the initial growth of a thick AlN layer. Figure 2 shows the initial micropipe-free 6H-SiC seed. Once this thin AlN layer is separated from the SiC, it is then employed for the growth of the AlN bulk crystal. Figure 3 shows the AlN layer separated from the SiC seed and attached to the crucible lid. The pattern of cracks that heal during bulk growth attests to the high quality of the AlN seed layer.

Figure 3 AlN layer separated from SiC seed and attached to crucible lid

Typically, as shown in Figure 3, the seeds used to produce 2-inch AlN have a diameter that is much larger. This removes edge striations and defects that occur during growth. The high quality of this material is revealed by the lack of features in crossed-polarizer images (see figure 4). Before the AlN crystals are ground to 2-inch diameter, they are used as seeds for fabricating AlN material of this size.

Figure 4 A crossed polarizer photo of a typical wafer. The edge striations have not been fully removed by diameter grinding

AlN sublimation issues

Our outfit, just like the team at CrystAlN, has extensive expertise in the growth of bulk single crystal SiC. However, despite this background, scaling up the growth process to 2-inch has been surprisingly difficult. Unlike SiC, AlN dissociates congruently into aluminum and nitrogen gas; however, aluminum vapor at high temperature is extremely reactive and forms lowertemperature eutectics with many materials that would be otherwise inert. To deal with these issues, we use TaC crucibles, which is a process that Hexatech and CrystAlN have also reportedly adopted. These crucibles are employed for the growth of the initial thick layer of AlN in a graphite system. A combination of TaC and tungsten crucibles, along with tungsten reactors, is then used for growth of bulk crystals. The historical evolution of our AlN crystals from 15 mm diameter to 2-inch diameter is illustrated in Figure 5. In an intermediate phase, the single crystal center was surrounded by a polycrystal ring as we expanded the micropipe-free SiC seeds. Recently, however, we have progressed to the production of fully mono-crystalline material.

Figure 5 Evolution of AlN crystals and substrates from 15 mm to 2-inch diameter

There are wide variations in the reports of optimum growth conditions for AlN. Hexatech reports that they achieve very high quality crystal growth only on the nitrogen-face and only at temperature of 2600 °C. On the other hand, we are able to grow “good enough” AlN crystals on the aluminum-face or nitrogen-face at relatively low temperatures, ranging from 1950 °C to 2050 °C, using a near atmospheric pressure N2.

We utilize the VirtualReactor-AlN code from Soft-Impact to model growth conditions in detail. Figure 6 shows an example of a VR simulation of crystal growth.

Figure 6 VirtualReactor simulation of crystal growth

Making substrates

For AlN substrate production, crystals are separated from seed holders and the head cropped off, before being ground to a 2-inch diameter and oriented by X-ray. Both major and minor US-convention flats are ground into the crystal. The round crystal is mounted to a slicing fixture and the mis-orientation angle of the substrates is fixed with X-rays, before a multi-wire diamond saw slices the crystal into wafers. The sliced wafers are lapped as needed and then subjected to chemical mechanical polishing (CMP). Finally, the wafer is cleaned and packaged in a nitrogen atmosphere.

We have developed our own CMP final polishing for the production of seed substrates; however, final polishing of the AlN substrates can be made by Novasic (France) on request. Novasic has established a unique, industryrecognized polishing capability for hard materials such as sapphire, SiC and AlN.

Characterization data, such X-ray diffraction and atomic force microscopy (AFM) measurements, is typically collected on the wafers after cleaning. A good assessment of the wafer production process is revealed by the quality of the epitaxy. Mapping of the surface topography with an AFM reveals the absence of scratches in the epi-layer, attesting to the quality of the CMP.

Figure 7 AlN characterization

The dislocation density of the majority of crystals is in the range 1 x 104 to 1 x 105/cm-2 (see Figure 7). This meets our criterion of “good enough.” The ultimate test of the substrate quality is epitaxy growth and device performance. UVA LEDs grown on AlN substrates can show a four-fold efficiency improvement over those grown on sapphire; however, to date, deep UV LEDs made by Crystal IS on their AlN substrates have not shown remarkably improved efficiency. This may be due to factors such as impurities rather than crystal quality.

For deep UV LEDs the transparency of the AlN substrate is particularly important. Due to its 6.02 eV bandgap AlN should be transparent down to nearly 200 nm. However, early substrates, even though optically almost colorless, had a rather sharp transmission cut off near 300 nm. Positron annihilation spectroscopy indicates that the main defect in AlN crystals may be an aluminum vacancy complexed with oxygen. This defect has an absorption feature in the blue that gives AlN crystals their usual orange/yellow color. But, it does not normally affect the UV absorption. Other impurities are likely to be responsible for the low energy cut off. This is confirmed, at least for our growth, since increased pre-purification of the AlN source powder leads to a shorter cut-off. Indeed with the cleanest source materials the cut-off is typically just 205 nm.

AlN Applications

AlN substrates have already been used for a wide variety of devices. Excellent performance of RF structures grown on this platform has been reported; however, these devices are not expected to become commercially important until 4-inch diameter AlN becomes available. Using the SiC seeding process AlN could be scaled to 4- inch diameter in a rather straightforward manner. Crystal IS has reported deep UV LEDs on AlN and shown that thick, fully strained pseudomorphic layers of AlGaN can be grown on this foundation. Meanwhile, SAW devices have been made by a variety of groups on AlN. This gives great promise for AlN as a universal substrate for III-nitrides.

The quest to obtain a 2-inch diameter AlN substrate has been pursued for many years. Our efforts have yielded material of this size with a dislocation density that is threeto- six orders of magnitude lower that that associated with epitaxy on sapphire. The many users of this substrate have obtained excellent epitaxial results, which are realized without the need for special buffer layers. While AlN forms an unavoidable oxide layer, simple procedures can remove any residual oxide.

Today the pricing of AlN is similar to that of SiC when it was first commercialized. But AlN prices will drop as sales increase. Substrates made from both these wide bandgap materials should have similar production costs in years to come. Since AlN is UV transparent, it may be usable in laser lift-off, although to date no publication of such results has been made. This could reduce the cost of AlN to that of repolishing, making it competitive with sapphire, GaAs and GaP, despite its greater initial cost.

Further reading
J. R. Grandusky et al, Applied Physics Express 3 072103 (2010)
G. A. Slack et al. Mat. Res. Soc. Symp. Proc. 798 Y10.74.1
US Patents 6,547,877 and 7,056,383
F. Tuomisto et al. J. Cryst. Growth 310 3998 (2008)