When Jayant Baliga invented the insulated gate bipolar transistor in the 1970s—something for which he won the 2014 IEEE Medal of Honor—it went from prototype to commercial product in about a year and a half. His quest to make silicon carbide a key material for power electronics has taken decades longer. But thanks to some recent innovations, he’s almost there. Two are new devices, impossible to make in silicon, and the third is a foundry process that will finally make SiC devices cost-competitive with silicon.
Silicon carbide is such desirable stuff because it has a much wider bandgap than silicon. That is, it takes more energy—about 3 times as much—for its electrons to go from being bound to their atoms to flowing as current than it does in silicon. So, devices made from SiC can be smaller and yet handle higher voltages more efficiently. Silicon power electronics tend to lose 10 percent of their power, while silicon carbide loses just 7 percent. And the higher frequency operations means the passive components attached to them—inductors and capacitors—can be smaller.
The problem has always been silicon carbide’s price relative to silicon. Right now it’s about five times as expensive. “Our goal is to get it down to 1.5 times the cost of silicon devices,” says Baliga, Distinguished University Professor at North Carolina State University. Key to that is a project called Power America Institute. “Power America was formed to create a foundry in [the] U.S. to accelerate the production” of devices made from wide-bandgap materials such as silicon carbide. The project chose X-Fab in Lubbock, Tex., which has a line dedicated to silicon carbide production.
The economic argument goes like this: The more wafers of silicon carbide X-Fab processes, the less each will cost, because the cost of the processing equipment will be spread over more wafers. “Hopefully that will begin the ‘virtuous cycle’: lower cost will lead to higher use; higher use leads to greater production volume; greater production volume further reduces cost, and so on. And consumers are getting a better, more energy-efficient product,” says Baliga.
But filling the silicon-carbide line was a problem. They needed new companies to develop silicon carbide devices to increase the throughput at X-Fab, but those companies faced the steep cost of developing their own manufacturing processes. That’s where Baliga stepped in. He and colleagues invented a manufacturing process called PRESiCE that companies can license from North Carolina State University and use at X-Fab. Last month at the International Conference on Silicon Carbide and Related Materials Baliga presented data showing good yield rates for PRESiCE on X-Fab’s silicon carbide line. One maker of silicon power electronics has already licensed the process for use at X-Fab, and two others are in talks, he says.
Another advantage of X-Fab is that its manufacturing line uses 6-inch-wide (15-centimeter) silicon carbide wafers instead of the 4-inch (10-cm) wafers used by other manufacturers. With more than double the area, each wafer can pack in many more devices, driving down the cost per device even further.
PRESiCE was originally geared to make silicon-carbide MOSFETs and diodes, standard components of power electronics. But Baliga’s team recently invented two new devices—things that work in silicon carbide but not silicon—and proved they could be made using the PRESiCE process to sweeten the deal.
One is called a JBSFET (junction barrier Schottky diode integrated metal oxide semiconductor field effect transistor). Many high-volume power electronics applications, such as inverters in solar panels and industrial motor drivers, require a MOSFET paired with a diode, explains Baliga. The JBSFET combines them in one device to save 40 percent on silicon carbide area and cut the amount of chip packaging.
The device is more than just integrating the transistor and the diode on the same chip, he says. “We’ve merged the diode into the MOSFET cell structure so they share the current path,” he says.
He expects JBSFET products to appear on the market in 2018.
The second new device is called the BiDFET, for bidirectional field effect transistor. It’s a four-terminal device that’s particularly useful in making a kind of AC to AC converter called a matrix converter. “The advantage of a matrix converter… is that it allows AC to be converted to AC directly” without a DC step that requires a bulky, potentially unreliable capacitor, he says.
According to IHS Markit, the market for silicon carbide discrete power device and modules is worth about US $270 million out of a total market of $16 billion. “But it’s growing very fast,” says Kevin Anderson senior analyst for power semiconductors at IHS. He sets the compound annual growth rate at 31.4 percent per year through 2025.
Power America’s plan to reduce the cost of silicon carbide devices by increasing the volume at the X-Fab Foundry “makes sense,” says Anderson. “One of the drivers of the cost is sheer volume. There’re a number of factors, but anything that drives volume would help in the market.” However, he calls Baliga’s 1.5-times-silicon price goal “aggressive.” Currently a third to half the cost of a silicon-carbide power device comes from the cost of the material itself compared to a mere 5–7 percent for a silicon device. “The very large volume difference between silicon and silicon carbide is a big factor in this difference, hence additional volume helps drive the economy of scale for silicon carbide.However, we do not foresee silicon carbide wafers ever approaching the price of silicon wafers.”
There are other reasons to be hopeful. A silicon carbide foundry “also drives innovation,” says Anderson. “In a foundry people with good ideas can get to work on them without having to do process development, which can be expensive.”
Samuel K. Moore is the senior editor at IEEE Spectrum in charge of semiconductors coverage. An IEEE member, he has a bachelor's degree in biomedical engineering from Brown University and a master's degree in journalism from New York University.