What is the best substrate for GaN?

In much the same way that GaN-on-sapphire is the clear-cut leader for high-brightness LED manufacture, GaN-on-SiC is today s winning ticket for wireless/RF products based on the nitrides. GaN-on-silicon is yet to make its mark in either of these markets, but it is still seen as the most promising material combination for achieving the holy grail of high device performance from a chip grown on a low-cost large-diameter platform.

In the case of LEDs, it s difficult to envisage the displacement of GaN-on-sapphire with GaN-on-silicon at the high end of the market in the near future, unless there s a technological breakthrough. If the recent progress in the development of free-standing GaN is maintained and migrated onto larger-diameter material, who would argue against a switch to high-power LED production on this native platform.

The GaN wireless/RF market, like the LED sector, is currently debating the merits of different types of substrate, and silicon is being considered once again as a low-cost platform that can energize and catalyze the market. However, defense is the only real market today, and in this sector performance is the main driver for success.

The focus on performance makes GaN-on-SiC the only contender, because it is the only combination that can fulfill market demands. While GaN-on-silicon continues to improve, its gain and thermal performance are not good enough for most defense applications, and it is difficult to see this type of transistor undergoing a performance hike that will enable it to compete with GaN-on-SiC at the highest level.

Of course, all is not perfect in the GaN-on-SiC world – issues relating to substrate quality have dogged this technology for many years. But over the last 18 months these concerns have been swept away through improvements in the uniformity and reproducibility of material quality. Today, the greatest challenge for SiC is cost, with 3 inch semi-insulating (SI) SiC costing 20 times that of a silicon equivalent.

Several vendors are working towards a reduction in material cost by migrating to larger substrates. US substrate suppliers Cree and II-VI are leading the way by producing very-high-quality 100 mm material that should have a lower cost per unit area than its 3 inch sibling by 2010, and become even more affordable further down the line. So does this mean that the price of SiC will eventually reach parity with that of silicon? Probably not, but does it really need to?

The inferior performance of GaN-on-silicon is also preventing it from enjoying success in many commercial RF applications, particularly radar. However, is it possible that GaN-on-silicon researchers can exploit a cost advantage over GaN-on-SiC and catapult this device s performance to the state of the art, before GaN-on-SiC prices fall substantially?

While it s impossible to know what lies ahead, it is clear that the longer that it takes for GaN-on-silicon to make incremental performance improvements, the more likely it is that GaN-on-SiC becomes a stronger competitor in terms of cost, due to the inevitable gains that come from economies of scale and progression up the learning curve. The market analyst Strategy Analytics forecasts that four-fifths of the GaN wireless/RF market will use SiC by 2012, and it wouldn t be surprising to see this materials technology grabbing all of that market.

Where GaN-on-silicon may play a major role is in wireless/RF products for emerging (not yet imagined) applications, requiring moderate performance at low cost, which could appear as the market blossoms. This scenario is not unusual – there are numerous examples of lower specification products with good economic credentials that fill, or even create, a gap in markets that were previously dominated with expensive, over-specified products.

A further opportunity for GaN-on-silicon, which offers massive potential, is power electronics. Here, success requires the overthrow of the incumbent technology, silicon, while seeing off the threat from another emerging wide-bandgap technology – SiC. While silicon majors in low cost and SiC offers the promise of high performance, GaN-on-silicon has the attributes to excel on both of these fronts.

Although cost is the primary motivator in the power electronic market, alternative technologies that only offer a reduction in expenditure are not guaranteed to enjoy success, because there is a financial penalty associated with the switch from one type of device to another. Consequently, new technologies have to offer a tangible, compelling value proposition in the form of significantly improved functionality. GaN-on-silicon is potentially up to this task, thanks to its combination of improved efficiency, reduced size, greater reliability and long-term potential for integration with silicon transistors.

The GaN power electronic market is still at an embryonic stage, but when sales do take off they have the opportunity to grab a share of a market that has been valued at $3.5 billion by Yole Développement. Investment is certainly not lacking in this class of device – it is just a question of whether progress is rapid enough to fend off competition from SiC or a new, advanced silicon technology.

IQE is continuing to develop GaN-on-silicon, GaN-on-SiC and GaN-on-sapphire technologies, with the aim of tailoring production and development activities towards providing the most appropriate technology for the relevant market.   

View pdf of article