Taking GaN to the Next Level of Scalability

//php echo do_shortcode(‘[responsivevoice_button voice=”US English Male” buttontext=”Listen to Post”]’) ?>

Among the challenges of the wide-bandgap (WBG) industry is how to achieve the manufacturing scale that enables it to be cost-competitive. Gallium nitride (GaN), in particular, enables greater conversion efficiency and performance at higher switching frequencies and power densities when compared with conventional silicon devices.

On a visit to Silicon Valley in 2022, EE Times met with Cem Basceri, the co-founder, president and CEO of Qromis, in Santa Clara, Calif. Basceri talked about some of the challenges of the WBG industry and how he had focused on commercializing his years of materials science research to improve the manufacturability of GaN. He talks about the Qromis substrate technology (QST) in “Engineered Substrate Scales GaN Technology.”

In April, EE Times caught up with Basceri  to talk more about the wider industry challenges and dig deeper into how he thinks Qromis is facilitating some of the change needed that he referred to last summer.

Cem Basceri with an 8-inch GaN on QST power device wafer at Qromis’ Santa Clara, Calif., office. (Source: Qromis)

Cem, what do you see as the big challenge in this the WBG industry?

The biggest challenge is manufacturability and scale, which has been a major obstacle to date. The fundamental issue is foundry economics. If the materials just stay in an IDM (integrated device manufacturer) or in boutique fabs or the non-semi spec environments, together with huge variations for RF on different substrates, or having different substrates for power, plus dealing with variations like thick substrates, thin substrates, small substrates and so on, that would never be accepted by foundries and won’t scale.

Everything ultimately comes down to cost or scale, and the foundry. Our foundation is basically a fabless approach and to create a WBG foundry ecosystem. Without the foundries, it’s very difficult to scale any material or any solutions. With that underlying approach, we created the QST technology, which is semi-spec and can run in any foundry today. It’s also a platform where you can make power, RF and scale on each application—high power to low power and so on.

The QST engineered substrate from Qromis. (Source: Qromis)

If you look at the WBG industry today, it’s all about materials. The substrate is so critical and the epitaxy. Native substrates are expensive and challenging in terms of scale and diameters, for many years. Fundamentally, it all comes down to substrate. Finding or creating a substrate that would address everything from cost to scale to performance has been the major challenge in the industry.

That forced innovation for us. Right now, the QST structure is used for GaN, but it can be applied to different interfaces and different WBG, maybe over the next decade that will be realized.

The other aspect is, especially for this type of substrate, which is very boutique, you need an entirely different infrastructure, and you need substrate manufacturing. With QST, the way we designed it is that any fab, any semiconductor fab can make the QST substrate in its existing infrastructure, in any silicon semiconductor fab. An example is Vanguard International Semiconductor—it’s a silicon foundry and in CMOS.

When we delivered this technology to Vanguard and showed them how to make the substrate, they took only three months to make this substrate. That brings a clear advantage for GaN right now. You can go to any foundry today, same spec, CMOS fab compatible.

Cost wise, I think there is a perception today that because it’s a new material, everybody thinks it costs thousands of dollars, but it doesn’t. There is always of course the question, how about GaN on silicon versus GaN on QST? Let’s say that if the price difference is, once it’s made internal, for example, if it’s $100 or $200 difference, if you get 4,000 or 5,000 dies on each substrate, that’s a fraction of a cent. It increases a little bit, but you enjoy the yield, there’s no breakage, and you can go to high voltage on different applications and use cost effective CMOS fabs.

So that’s the underlying problem. The rest—making devices, finding applications and so on—can all be done, but everything hinges on the substrate.

Is that what you’re enabling, allowing these fabs to license this capability, and technology?

That’s right. And if they don’t want to make it internal, for that we supply substrates. Vanguard makes those for us exclusively, and we also gave a license to Shin-Etsu—they are also making substrate products.

You mentioned when we met last summer that Shin-Etsu was doubling their capacity with this. Tell me, how are they doing?

Yes, they are even increasing more than that, especially for the 2024-2025 timeframe based on demand. Customers have different choices—they can either buy from Shin-Etsu or from us, or they can get a license and make it internal.

What’s your progress so far in pursuing the fabless business model path you’ve taken?

The showcase for us was we ran thousands of wafers at Micron to validate this internally. Then the showcase was to bring it to a foundry and see how it works. That was the big major milestone, enabling Vanguard from ground zero—they didn’t have any GaN experience or GaN device. In a couple of years, they made the substrate, they made the epi, and then they just released the device. That, in my opinion, is a great story. That’s from ground zero—a foundry made a GaN device. They are the only 8-inch GaN foundry service in the world right now.

Closeup of 8-inch GaN on QST 650-V power IC device wafer. (Source: Qromis)

When was the ground zero you referred to?

That started in the 2016-2017 timeframe. They started with substrate, and then epi, and then device qualification, reliability and all that. Then the second milestone was to bring another pure play materials company Shin-Etsu.

One of the things you said, is that there is an unfulfilled promise in the III-V industry? Is it the fact that it isn’t able to scale?

If you look at GaN or silicon carbide (SiC), if you look at them just as semiconductors, they’re great materials. But so far, either you make devices or solutions on a platform that’s extremely expensive. Or you try to make it on non-native substrates that aren’t as suitable, like GaN on silicon, and you cramp the device with stress, with dislocations, and lots of issues and breakage. So, you don’t get the full promise of either material. Because of cost reasons, and because of the unavailability of the native substrates. So, that was the unfulfilled promise.

Then the second one in my opinion—compared to other semiconductor businesses like memory or logic or such—unfortunately, we don’t yet see a systematic approach like a real semiconductor approach where you do a systematic study, look at different platforms and then address all your road-map items with confidence.  Hopefully, that will be coming up soon.

What is the big thing that needs to happen in the industry that will make that happen—that systematic approach?

One is investments, and there is still some resistance to new materials in the industry. To me, it’s an interesting contrast. For example, when I worked in memory, we would scan every material possible on the periodic table or every different platform across XYZ axes. You would just do a systematic study to make decisions. But unfortunately, most of the industry right now doesn’t have that bandwidth. They still stay in that comfort zone, and I understand it requires investments or changes. I think that will be another next big breakthrough in my opinion.

You mean investment in the substrate?

In everything: substrates, research and development, and real manufacturing lines, where development takes place in the semi fabs, not just in boutique fabs. But it’s coming, step by step.

Where do you see yourself going as a fabless materials company and your business model?

So far, our model is working very well. A good analogy would be like the Soitec business—except that we don’t have a fab yet, and I’m not sure if we should. As we discussed previously, there are so many good players in the world that we can enable them, work in partnership. We brainstormed quite a bit with Mike [Noonen] in the beginning, and we looked at the map and saw that there are so many players, so let’s work with everybody. Our role is therefore to enable them: we enable academics, we enable materials companies, foundries and IDMs. At the same time, of course, we do our R&D, and we have another two decades of different solutions around QST.

You have focused on GaN. What about silicon carbide?

SiC is a fascinating material, I worked on it for many years, and I’m very certain it will coexist together with GaN for different applications. The way that we designed QST, if you look at the cross section of the substrate, the top surface is silicon—really thin. So, that’s the growth surface. The reason we picked silicon is because silicon is available and to grow GaN on silicon. That interface, that silicon, can be changed to any other material. It could be SiC, very thin layers, cost effective and it’s only a 0.2- or 0.3-micron thin layer. So, it could be a GaN interface, it could be SiC. Then you can apply it to different applications.

But you as Qromis are not planning that right now?

Not us. But we cover those in our portfolios and all such IP. But we leave it to the industry. So, maybe Shin-Etsu could make that interface or someone else. We are open to that.

Finally, you talked about 1200 V EV being critical, what did you mean by that?

So, it’s the 1200 V segment, which is EVs definitely, as well as industrial applications or solar. The reason for 1200 V is because QST can deliver. One good reason is because you don’t want to leave that segment on the table. We can even go higher, like 2,000 V, 2,400 V or beyond. The EV is very critical because an EV, in my opinion, is one of the applications taking full advantage of GaN. The high switch, it reduces the cost, bill of materials, reduces system size and weight, and also increases the performance. So, that’s important for EVs, and a huge market. So, of course, in the EV also, it’s not just one application—there’s 650 V, 1200 V, inverter and so on.

What are the other significant industry challenges that we haven’t covered?

I think we covered most of it. For these materials to go beyond today, with more applications, more adoption and all that, scale is very important, and also the foundry economics, plus fast cycles: industry can’t wait for 6-inch to 8-inch transition for two decades. That’s not going to work. So, I’m hoping we can go from 6-inch to 8-inch to 12-inch within six months or 12 months, depending on the fabs or the needs.

Scroll to Top