SiC readies itself to save the world
Al Gore and the International Panel on Climate Change were named the winners of the Nobel Peace Prize while I was on the way to Otsu, Japan, to attend the 2007 International Conference on Silicon Carbide and Related Materials. This fact was not lost on the participants in the plenary session of the ICSCRM conference, affectionately termed "ice cream". On the first morning, representatives of SiCED and Toyota mentioned this piece of news in the same breath as the potential efficiency benefits of replacing silicon with SiC. When describing the energy savings arising from using SiC devices as industrial inverters, SiCED s Dethard Peters commented: "The world is waiting for solutions to global warming and we have silicon carbide in our hands."
Addressing the crowd at the conference banquet, the governor of Shiga prefecture, of which Otsu is the capital, picked up the theme. With a background as an academic ecologist and having jurisdiction over Lake Biwa, the world heritage site onto which the conference backed, she was proud to be associated with a potentially "environmentally friendly" technology.
Such a reputation will hardly do the SiC community any harm, but perhaps it is a little early to be trying to get associated with Gore et al. "We thought we had enough to do getting defect levels down and making devices, and now we have to save the world?" was one anonymously reported response to the green tinge that shaded the conference.
Back in the plenary session, Kimori Hamada of Toyota emphasized that his company anticipated using SiC in its hybrid cars at some point in the 2010s. However, his requirements for SiC demand high performance at a cost to match silicon.
SiCED s Peters almost immediately delivered on that requirement, asserting in his plenary talk that the Schottky diodes that it has helped Infineon to develop are economically competitive with silicon today. Enjoying a privileged position in German industry (owned by Siemens and collaborating with Infineon), SiCED is looking to develop JFETs and MOSFETs rapidly to join Infineon s Schottky diodes.
Regardless of these commercial promises, many of the conference participants were nervous about the industry s prospects, not least the representatives of the growing number of SiC wafer suppliers. While Cree remained the most-mentioned provider in the talks, SiXON – a well reputed Japanese wafer supplier – announced its closure. In SiXON s place, two new SiC firms made their debut: Crysband of South Korea and TanKeBlue of China.
Cree – keeping ahead of the pack
In the remaining plenary talk, Cree s Al Burk went over the history of SiC growth, leading up to his firm s current leadership in the substrate market with its 4 inch (100 mm in Cree s terminology) zero-micropipe wafers. Burk focused on the predecessors who made Cree s current position possible, before rushing through the design details of the 8 × 100 mm wafer warm-wall planetary SiC-VPE reactors in use at the company. Burk s colleague Joe Sumakeris explained the elegant way in which Cree can improve the thickness and doping uniformity of its 100 mm wafers, using horizontal flow rather than planetary reactors. Introducing additional propane to the chamber through a secondary carbon source channel helps to reduce the thickening and increased doping seen at the wafer edges. Adding supplementary hydrogen stops graphite precipitation in the reactor and the undesirable defects that this causes in the growing wafers.
Having nailed down zero-micropipe growth on the 4 inch scale, Cree is now lowering "1c-screw" dislocation density. Robert Leonard presented detailed wafer mapping to show that process developments at Cree have more than halved the number of 1c-screw dislocations. Standard processes gave 1c densities of 850 cm–2, but Leonard showed improved median densities of 325 cm–2 that went as low as 175 cm–2 on 3 inch 4H SiC wafers. Cutting 1c-screw dislocation density lowers the leakage current of devices grown on SiC substrates, according to Leonard.
A good chunk of research presented at ICSCRM was focused on developing ways to improve crystal quality that avoid Cree s strong intellectual property position. For example, Cree has patented the KOH etching of wafers to reduce the density of basal plane dislocations (BPDs). These cause stacking plane faults that degrade the performance of certain SiC devices when a current is applied, so an alternative method coming out of the US Naval Research Laboratory in Washington attracted a lot of interest. In a single step of what he called an interrupted growth method, Robert Stahlbush was able to convert 30% of a wafer s BPDs to less problematic threading etch dislocations (TEDs). This can then be repeated to diminish BPD levels further. While remaining wary of releasing precise details of the process, Stahlbush explained that the discovery was made while studying BPD evolution during epitaxial growth. During these studies, some BPDs would turn to TEDs at the beginning of each epitaxial run.
How quickly can SiC grow?
With end users like Toyota putting price-for-performance at the top of their SiC wish list, the High Growth Rate session drew a lot of attention. The key to the achievement of high growth rates is suppressing nucleation in the gases used in the reactors. The session split in two, between European researchers who use different types of chlorine-based chemicals to suppress nucleation, and Japanese researchers who focus on gas control.
In chlorine chemistry, US start-up Caracal and Linköping University attained 4H SiC growth rates of higher than 100 µm/h by using methyltrichlorosilane as a growth precursor. The same collaboration achieved on-axis crystal growth that produced 100% of the 4H SiC polytype at 20 µm/h. On-axis growth is difficult because it can lead to a SiC crystal composed of a mixture of polytypes, but it is desirable because it will eliminate the BPD problem.
By contrast, researchers from Japan s Center of Research for the Electrical Power Industry attained a 135 µm/h growth rate using a more conventional gas mixture. The key to stopping nucleation in hydrogen/silane/propane is growing your crystal at 40 Torr pressure, according to Masahiko Ito. A similar low-pressure method at 20 Torr, with a greater focus on reactor design, scored a growth rate of 140 µm/h for Yuuki Ishida of Japan s Advanced Institute for Science and Technology.
Toyota was not the only Japanese conglomerate present. Among the device-related papers, big hitters like Toshiba, Panasonic, Mitsubishi, Hitachi, Nissan and representatives from Japanese electrical power companies stepped up to show their form in the SiC arena. Amid a plethora of devices, a strong body of MOSFET research came through to show SiCED and its collaborators that they will not have the SiC MOSFET field to themselves. Many groups focused their attention on the problem of the trapping of electrons between SiC and any potential gate oxide without producing commercial devices.
While we wait for the full range of SiC devices to work through to commercialization, the next to market could result from military contracts. Victor Veliadis of Northrop Grumman presented a 0.1 cm2 vertical JFET that has attained breakdown voltages of more than 1980 V for the US Army. David Spry from NASA showed how the SiC VJFETs that he is developing for sensing applications can operate stably at over 500 °C for 2000 h, with tests still ongoing.
So, although there was much in Otsu to be positive about, it is clear that the SiC community faces a myriad of challenges before it can meet its full potential. In the meantime there are a lot of people performing work with a limited amount of cash from sales to end consumers flowing into the industry.
Specialist power electronics consultancy Yole Développement has estimated that SiC device sales in the power sector came to $15 million for 2006, split evenly between Infineon and Cree.
On the Thursday of the conference week, Cree announced that its total sales for the preceding quarter had hit $113 million – $8.5 million of this revenue was derived from materials, and power devices brought in just $4 million. Just as Cree is a big fish in the small SiC pond, so the kind of applications that were discussed at ICSCRM are currently of secondary financial importance at Cree.
The upside to this is that industry giants continue to line up behind SiC, lending weight to Yole s claims that the annual device market will be worth $800 million by 2015. After Cree s results were announced, its Otsu contingent gathered for a celebration, confident that they will be around to reap the benefits of the industry s expansion. Many other companies may be equally confident, but SiXON has showed how the competition between SiC wafer makers is getting tougher. Perhaps, in such a tightly balanced business, the motivation for each individual company to make breakthroughs and stay ahead of its competitors will help to kick off the SiC device growth ramp sooner than anticipated.
