IGBT Inventor Crusades Wide Bandgap Semiconductors
As Professor Jay Baliga, North Carolina State University, steps up to lead the $140 million 'Next Generation Power Electronics National Manufacturing Innovation Institute', Compound Semiconductor asks about his past, plans and relentless pursuit of the power semiconductor.
You invented the silicon IGBT, have pioneered wide bandgap semiconductor devices, and you've just received the IEEE Medal of Honor, where did this all start?
I invented the IGBT in 1980, the same year that I derived an equation that, for the first time, related resistance in power devices to material properties. I called it Baliga's figure of merit. So when I discovered this, I also started looking for alternative semiconductor materials to silicon, and the first promising material I found was gallium arsenide. Working at GE at the time, we put together a group and developed the first wideband gap power device based on GaAs in 1985. We commercialised this and these devices are still available across the industry today.
At any point did you feel torn between the two competing technologies; silicon and wide bandgap semiconductors?
It was actually a pretty tough time for me as I was trying to develop the IGBT as quickly as possible and then I had to do the GaAs programme as well. But I had always seen wide band gap semiconductors as the future, although I had not appreciated how long it would take for wide bandgap semiconductor devices to become commercial. With the IGBT I commercialised it in a remarkable ten months, which was why it became so widely used in so many diverse applications from air cooling systems to drives and lighting. Thirty years has been a very long time, but wide bandgap semiconductors are now cannibalising on the IGBT, and this is how I always believed it would happen.
The first wideband gap semiconductor you worked on was GaAs, what came next?
My equation predicted that with SiC, which has an even wider bandgap, I could get a 1000 fold improvement [in performance]. I'd predicted a 13.7 times improvement with GaAs, so this was another two orders of magnitude. When I found this one thousand fold opportunity, I started to look at how to make devices. But this was the early 1980s, and there was no SiC technology at the time. So we went to see Professor Robert Davis in the materials science and engineering department at North Carolina State University. He developed some growth processes for SiC wafers, and his students later spun off the company, Cree. Cree started to produce SiC wafers and this, as well as my interest in collaborating with Dr Davis, brought me to North Carolina from GE.
So what power devices followed?
In 1991, I set up a power semiconductor research centre at the university. This was, and is, an industrial consortium and with the funding the came from that consortium we made the first high performance SiC electronic devices. In 1992, we announced a very high performance Schottky rectifier and by 1995 we had developed a very high performance power MOSFET. I do think these devices helped the industry appreciate this technology; these systems had proven the theory, so many programs to develop more devices followed. By 2000 to 2005 we started to see a lot of products, particularly Schottky rectifiers and junction barrier Schottky (JBS) rectifiers based on an idea I proposed for silicon in the 1980s. These rectifiers were the first products and then we had power MOSFETS. Today, Infineon, Cree, Rohm and a whole bunch of companies now manufacture these SiC products.
It's no secret that the cost of SiC devices is hampering adoption; do you see a solution?
Yes, the problem for the industry right now is the cost of these products. It is much much greater than silicon, making the devices too expensive for mainstream applications, despite the advantages. In the 1980s, silicon carbide was thought of as a material for, say, high temperature, radiation-hard applications but my belief has always been that the material can be used in mainstream power electronics and this is what I have been championing since then. But to make that happen you have to have low enough costs to be competitive with silicon, this is the goal of the new manufacturing innovation hub at North Carolina State University.
How will the new hub achieve cheap silicon carbide devices?
We will bring together the materials suppliers of SiC as well as GaN, and align the materials with the foundry that will make the chips. We will then develop high volume manufacturing technology and also bring in companies that package these devices so they can operate at higher temperatures than silicon.
The entire supply chain is involved; how do you organise the research and intellectual property issues for example?
These details are being worked out and the US Department of Energy is involved in the process. We plan to collaborate very strongly with the [hub partners]. My focus is on SiC devices and processing so I will work with the foundry that is going to build SiC chips. We will share whatever IP comes out of this as part of the consortium.
I'm used to developing IP, I have 120 patents.
Will the hub focus more on SiC devices than GaN devices?
No, GaN is now coming along very strongly. We have 600V devices and these are getting very competitive. GaN can be grown on silicon so the cost structure is very different to SiC. We will have to see how this pans out, but I would say we have equal emphasis on SiC and GaN.
What devices can industry expect when the project finishes in five years?
For GaN, 600V seems to be the sweet spot and for SiC I have been saying all the way from 600V to 5000V for MOSFETS. Essentially, industry needs two devices; the rectifier and the switch. Both will be developed and the power rating will be driven by the end users, that's why we have them in the consortium; they will provide the incentive to the device manufacturers to create the products that are needed. This helps as when manufacturers create products without that interaction, it's like throwing a device over a fence and hoping someone will use it. But an end user will say, we just really need this device for the next electric vehicle for example. I had this benefit at GE when I was developing the IGBT; I knew exactly what the applications were so I could design my device to meet the needs and get it commercialised very quickly. This consortium will enable that. We want to bring as many applications as we can to this consortium, to proliferate the technology, expand volume and drive the costs down. This is the goal of the hub, to bring all of this together.
That's quite a goal?
Yes it is, but it is a goal that will make my 30 year old vision come true.