Phil Neudeck, Electronics Engineer, NASA’s John Glenn Research Center, Cleveland, OH

Neudeck: There are two main failure mechanisms. The first is what I would call semiconductor physics. In the case of silicon, heat will actually turn a semiconductor into a conductor. That’s just physics. The temperature at which that happens depends on a fundamental material property called the “bandgap” of the semiconductor. Silicon has a bandgap of about 1 volt and silicon carbide has a bandgap of about 3 volts. The difference in that bandgap actually translates into a big difference in terms of the temperature at which a semiconductor turns into a conductor. Once a semiconductor becomes a conductor, you really can’t do the things you need to do with it in terms of making transistors and electronic circuitry. So that’s a physics mechanism.

There’s also a second mechanism with respect to chemical degradation. Actually, the mechanism we’ve been fighting with silicon carbide is that silicon carbide has always been a very good semiconductor, even as high as 600°C glowing red-hot. We’ve always known that you can build a transistor in silicon carbide and it will work fine, but the issue is the chemical degradation of the interfaces and the junctions that you need to actually make a transistor. In the case of silicon, you don’t really have that much of a chance, because silicon is much more reactive than silicon carbide. The chemical reactions and diffusions take place and the silicon device will degrade very rapidly.

So, there are two things that prevent silicon from operating at high temperatures, whereas with silicon carbide it’s still a semiconductor at high temperatures and we’ve figured out how to control the interfaces to where they can last a good long time at high temperatures. That’s sort of the fundamental discovery, if you will, in silicon carbide – how to control those chemical reactions.

NTB: So that is really the key? Are there any other differences between a conventional IC and one of these high-temperature silicon carbide ICs?

Neudeck: Well, the silicon carbide tends to be more expensive, so you really would only use silicon carbide where you have to use silicon carbide, such as in an environment like this. Silicon is very well commercialized; the chips in these computers are becoming pretty low-price, high-volume commodities. Silicon carbide is not at that stage where it’s cheap or much of a commodity. It’s definitely much tougher to work with than silicon; that’s definitely a difference.

In the end, we’re not as far along in the technology maturity process as silicon. For instance, in these integrated circuits that we’re talking about for our 500° C demonstration, it’s only a couple of transistors and three resistors interconnected together. I’ve got some other circuits in the lab that haven’t run as long as the one we’re talking about in the press release. They’re at about the four- or five-transistor level. Obviously, the number of transistors in a silicon chip in your computer is on the order of about a million transistors.

The silicon people have had decades to refine their technology, make the individual transistors smaller and faster. That process has gone on in silicon for decades. With silicon carbide, certainly we can follow the path that silicon took in terms of shrinking things and making more complex circuitry. I think we’ll be able to make more complex circuitry in silicon carbide more rapidly because we’ve learned from how silicon did it. But we still have that path to go down.