Silicon carbide moves into the spotlight

The late January release of Cree’s new CMF20120D MOSFET, which the company claims as the “first fully qualified commercial silicon carbide power MOSFET,” marks yet another announcement in a string of new silicon carbide (SiC)-based power products coming from Japanese companies.

For example, Rohm announced the industry first mass production of a SiC double-diffusion MOSFET (or DMOSFET) right before Christmas. Then, in January, Mitsubishi Electric announced that it had developed a power conditioner mounted with SiC power semiconductor.

Certainly, SiC has been receiving accolades as the next-generation power semiconductor material of choice touting its numerous benefits including its low thermal expansion, lower density, lower power loss, higher power density, and higher voltage resistance than standard silicon wafers.

A recent EDN interview with John Palmour, co-founder of Cree, further illustrates some of the mounting support for SiC as the next generation technology of choice in power ICs. In fact, Palmour sees SiC as a challenger that “will emerge from and transcend the niche power market to rival silicon in high-power devices.” But his views are hardly surprising, given the long history of research and funding Cree has put into developing SiC technology.

SiC comes with numerous challenges that will prevent it from becoming ubiquitous, at least in the short term. Firstly, gallium nitride (GaN) has also emerged as a major contender for adoption in new RF and switching power devices. What’s more, although advances are being made, manufacturing of SiC is still particularly challenging due to its brittle characteristics, high hardness, and poor machinability. Fractures can often result when trying to machine hard brittle materials like SiC due to its low fracture toughness.

What’s more, while SiC has a higher electron mobility than silicon, GaN’s electron mobility is even above SiC, which means that GaN should ultimately be the best choice for very high frequencies. GaN also has almost an order of magnitude better material limited performance potential compared to SiC.

But, with all that being said, in the end it is the performance per cost ratio that drives market adoption, particularly in consumer markets, and is, in many ways, just as important as performance. GaN on SiC wafers cost about 20% more on average than their SiC on SiC counterparts. Additionally, silicon substrates are readily available in larger wafer diameters than GaN, which further drives down wafer fabrication costs. Putting together performance and cost, the price per performance of GaN is much more than double that of SiC.

Still, it’s important to remember that in situations where cost is not the primary motivator, such as RF-frequency applications above the S-Band (ex: in defense systems like those required for electronic warfare applications), GaN on SiC will remain the preferred choice thanks to its superb RF performance. More RF GaN on Si HEMTs are also being developed. Supporters of these devices are primarily targeting applications like wireless base stations (including 3G, 4G, and WiMAX), which are currently dominated by silicon LDMOS devices.

There certainly remains some grey areas where either option may succeed over the other (such as in RF applications in the low MHz to S-Band range of frequencies), but these will probably be decided by several factors including declining costs, the emergence of new applications, and the performance requirements of these applications. While it is true that silicon carbide will continue to remain a quickly growing technology for power products, particularly in MOSFET design, both SiC and GaN will most likely settle into their respective niches, with each playing their own important role.