SemiWiki.com - Podcast EP362: An atomic layer in compute. A discussion of what can be built as graphene photonics scales with Cedric Huyghebaert

Episode Date: August 21, 2026

Daniel is joined by Dr. Cedric Huyghebaert, CTO of Black Semiconductor. Cedric is one of the world’s leading experts in the development and integration of 2D materials, particularly graphene, into industrial semiconductor manufacturing. With a PhD in Physics from KU Leuven and over 25 years at imec, Cedric built and led… Read More

Hello, my name is Daniel Nenny, founder of semi-wiki, the Open Forum for Semiconductor professionals. Welcome to the Semiconductor Insiders podcast series. My guest today is Dr. Cedric Acrobat, CTO of Black Semiconductor. Cedric is one of the world's leading experts in the development and integration of 2D materials, particularly graphene, into industrial semiconductor manufacturing. So we has a PhD from physics at KULUBN and over 25 years at IMEC. Cedric built and led the scientific programs that bridged graphene from lab to waferscale industrial process.

He joined Black Semiconductor in April of 2022, bringing with him a body of institutional knowledge that no other commercial organization possesses. Welcome to the podcast, Cedric. Great. It's a pleasure to meet you. You are very well known. But first, let me ask you, what brought you to Black Semiconductor? Well, I had the chance that I make to lead the nanomaterial structure research group. And the idea of that group was always to build functionality out of nanostructured materials.

2D materials were an important part of that portfolio. And so in 2010, I was for the first time exposed to graphene. And we found interesting value proposition in these two. 2D materials and one of the biggest challenge was was in integrating that I was part of a lot of European programs around 2D materials and it's in that framework that I met Daniel Scholl the the founder of black semiconductor and we always had a lot of respect for each other's research Daniel finished his PhD at a MO Research Institute and founded black semiconductor and at some

point in time, he asked me, could you imagine joining a startup? And I think he asked me at exactly the right moment in my life. I was established that I could have stayed until my retirement, but at that moment in time, closing the circle from bringing a material in a research stage in the lab, doing the first path-finding steps in a production fab, and finally now trying to make the first a scaled product out of that was very exciting and that's the reason I joined black. So Cedric, why did you name the company black semiconductor? Well, it's actually not so complicated. Graphene is coming from graphite. It's a first, it's a layer that you peel off from a graphite part and graphite is black and we decided to

call the company black semiconductor because of that. It is a similar. It is a similar to a similar name to remember and I think it covers relatively well what we try to do. So in that sense we evaluated a lot of more names that were way too complicated and at the end the simplest version was the best version so that's why we are called black semiconductor. I agree completely great name. So let's get into it. What made graphene the material you decided to build a company around? I mean what What did you see that others may have missed? Well, when we started graphene research, everybody was excited about graphene being the replacement for silicon.

And people were hoping to build the fastest possible transistors. Very soon it was clear that graphene has no band gap and because of that you cannot switch it off. So transistors was not the application field for graphene. graphene had remarkable optical properties and we see a gigantic opportunity in combining these optical properties with the integratability of graphene and we think we can solve a problem for the industry and that's why we created black semiconductor. Oh great. So what are the three hardest process engineering problems in integrating graphene onto a 300 millimeter wafer?

Yeah, that's a good question. Graphene is a complicated material. You cannot go to a foundry and ask them, hey, can you build a graphene-based device for us? Because no foundry has graphene in the portfolio. The first big problem is graphene is a monolayer material, so there's just one atomic tick. And you need to go to high temperature to create crystalline material. So we need to grow graphene. This is the first challenge, is how do you know, you know, to go to high temperature to create crystalline material. So we need to grow graphene. This is the first challenge is how do you grow and how do you make single crystalline wafer level graphene? And then because of the high temperature requirements very often and the epitaxial requirements, you need to do that on a different template. So this is one of the first challenges that we had to address. Another challenge is how do you build a dielectric stack? How do you encapsulate graphene? Graphene is a self-passivated materials, so it has only chemical bonds in the X, Y direction, and it's self-passivated in the Z direction. So when you start to grow something on graphene, there is no

dackling bond there to help you in seeding or in nucleating the start of a growth. And this is a very complicated problem that needs to be addressed in a very careful way. And we have solutions for that. And then the last problem that is really critical in graphene is how do you build contacts that are scalable at what the industry is used to? In labs, we're building contacts in a completely different way. And actually, the industry has abundant a way to build these contacts since more than 30 years because yield-wise, they could not reach acceptable numbers there. And we have developed a proprietary process to make contacts, damaskin contacts, to go. graphene and this is a key development in order to be able to scale graphene to thousands or

millions of devices at wafer level. So these are I think the really core developments and the nuts that needs to be cracked in order to make graphene from a fancy material to something that starts to be a technology. Okay but what makes you confident you can solve these? I mean very complex challenges? Well, we are building on more than 20 years of experience of a lap insights. And of course, these lab insights are approached in a very different way compared to where you would approach in standard silicon integration. So what we have done is we have, based on that understanding and based on the understanding

of what Tier 1 tool vendors are offering, we went in discussion with tool vendors and try to match that understanding to processes and yeah, tool designs that are available on the market. So the very interesting thing was that a lot of tool vendors were quite interested to join that journey with us because it's a very different approach that is required, but nonetheless they recognize that the way their tools are designed offers opportunities to develop this novel technology. And this was a very interesting journey together with the tool vendors because they opened up their portfolio of most advanced tools in order to allow us to be successful with the graphene integration. So I think semiconductor industry is always an industry, an ecosystem industry. I don't think you can solve the problems of graphene by your own, but together with the help of these very seasoned industries, giants, we can really make big step forwards.

You need to understand that graphene is three Engstrom-Tick. So this is, I think, the most advanced Z direction. This is the most scaled layer. And you really need to have processes that are only touching that layer in a very gentle way. Otherwise, you are, you're breaking up that material and basically it will not perform. So finding the tools that are just like be able to control. these three angstrom take layers. This was the one of the big challenges. Now we are building that FAP and what we know that we have to demos and to demonstrators,

we have a first version of our technology, but there is the big work that needs to be done in stabilizing the technology. And of course this is a work of iterating again and again and stabilizing all the processes and the interaction between all the processes. We are very excited that we can start our facility by the end of this year and that we can gradually go into that exercise in stabilizing and calming down the processes in such a way that the yield and the stability of our processes will go up and we will see the performance being validated. I think in general we needed to build our FAP because in research institute and they're doing fantastic work. But you can, for such a novel material,

you can count on two to three learning cycles per year. And actually to stabilize and to accelerate the adoption of such a technology, you need to do at least 20 to 30 learning cycles to really come down and having the yield numbers that are required. And I think this is why we are building this facility here together with the T1 tool vendors in such a way that we can demonstrate that stability of that processing.

Amazing. You know, you're building a process that has never been built before on equipment that has really never been configured in this way for a material. No production line has handled at this scale. What does it mean to write the process manual that introduces 2.5D of material manufacturing capability the industry will eventually, you know, learn on and maybe even standardize on? So writing a process manual for 2D materials starts by understanding what is fundamentally different for 2D materials compared to what you have for silicon. And there is one thing that is fundamentally different is that most of the materials until now were starting from bulk properties of material. And 2D materials actually bulk is very limited. So the interface is defining the performance of your material. So when you are writing an integration scheme for 2D materials, you're not looking to the material per se,

but you need to look exactly at the way how you build that material in a material stack and all what is around matters. And this is one of the big insights that changes the way you need to address the integration needs. needs. And this is also something that we see when we build up this startup here, is that when we hire people from the industry, they need to change their minds in adopting this novel way of thinking about the materials. And I think we see that for 2D materials, but it's also something that you see in other parts of the industry. More and more materials are thinned down. What happens at the interface is basically becoming more and more important because you cannot push the effects in the bulk anymore.

And I think 2D materials is the ultimate way of that question that is more and more visible in the industry. And that's why it's so exciting. I think we are dealing really with the future integration problems of the semiconductor industry. Why can integrated graphene photonics be added to any chip? And why does that matter for the customers designing systems today? Well, the first thing is we are growing our graphene and then we are transferring it on the waiver. This means that we do not need to have access to high temperatures to make the crystalline material there. So we can really integrate graphene at low temperatures and this makes that it's compatible within the back in the flight. So you can really build that graphene photonics layer. on every chip because the temperature budget that you have is very low. So this is quite unique because if you compare that to competitors like silicon photonics or where you need to make junctions and so on, so you need to have crystalline material, you need to do implantations, you need to do aniles.

This is not compatible with CMOS that has already been processed. So in that sense, our integration method allows to do that seamlessly on top or in the back end of line, we can build in optical components. And this is quite interesting because it allows your chip to be bilingual, I call it. So it talks in the electronic domain, but it can also talks in the optical domain.

And this is quite interesting if you understand the problems of interconnect or the limited reach that you can have with copper. I think everybody understands that copper cannot provide the density and the reach that is required for advanced compute systems. This is now accepted by the industry, that photonics will come.

And that's why you see all these efforts for co-package optics and scale out and scale in is moving towards photonics. Our way of integrating that photonics without having to do complex packaging, to use complex advanced packaging solutions like hybrid bonding or chiplets that needs to be combined with photonic ICs and so on, allows to have a more intimate integration and cheaper integration and higher yielding integration of the photonics inside the circuit. On the longer term, this allows a complete new way of looking to problems because today, the compute is in the center of the chip and you always need as if you want to bring data outside your chip you always need to hop towards the shoreline where the serializes light pumps your data outside and then you need to decilize again and then you can you you go to your next chip if you can integrate your photonics on every system or in every system together with the electronics

You can really optimize where you collect the data and making a direct optical link to where you need to deliver the data. So you can have an IP block that is linked to another place and so on. And you can basically you can intensify the possibilities and you can create novel ways of, well, novel architectures. And we want to enable the others to rethink computing with our technology because some of these boundaries that are typically there with the other. there with the other technologies are becoming available or the boundaries are kind of removed by our technology and of course this this this is a long-term effort because i i understand that the industry will first have to see the results and and gradually um before they can adopt that but this is how we see the added value of black is that that we really allow the others to rethink computing and redesigning

systems in a completely different way. Great. What other unique capabilities does graphene photonics offer that other technologies cannot deliver? And more importantly, what industry will benefit first? Well, I think the unique capabilities that graphene offer is that intimate integration together with deep into systems. The other thing that graphene offers also is because we can reduce the the distances and so on. And graphene is also a very efficient way of converting light.

So at the power consumption that you will have with links based on graphene, these are very low. And the performance also, the physics tells us that the speed, there is basically very limited reason why the physics would limit the speed of the devices. We are not fully there yet because at this moment in time, the graphene is not a perfect material, but there is a lot of ways we can improve that material over time. This will improve the device performance. The other thing that is quite unique about graphene is that we have a way to modulate

and the photo detector are in the same material. If you compare that to a lot of other materials, they have all an extremely good modulate. or they have a good photo detector, but it's very rare that you have the modulator and the photo detector based on the same material. And this is quite interesting for the cost of ownership of building systems with graphene, because the photo detector and the modulator are just in one layer of material. Finally, what is also very interesting about graphene photonics is that it is inherently insensitive to temperature variations, or there is a very small dependence on the graphene photonics.

temperature variations it doesn't affect the modulation efficiency and it doesn't affect the the photo detector efficiency so you can put the system and you don't need to correct for temperature variations and this could become a very big asset especially when you combine it with CMOS that is known to give spikes of heat at uncontrolled moments and uncontrolled places And because of that, I think this makes graphene quite unique and useful to combine with CEMOS. Right.

So last question, Cedric. If graphene photonics scales, what changes? You know, not just for a black semiconductor, but how will the industry build computing systems 10 years from now? Yeah. So on the short term, I think what changes is that we can reduce the chip-to-chip communication bottleneck, which is already a problem in AI clusters, and this is also where we see the opportunity.

The other thing on the short term where it can bring solutions for the industry is that because of the relatively lean integration scheme, we can most probably bring photonics to novel markets because it will become affordable to have photonics where today other solutions are too expensive. So this is most probably a way to create novel markets. On the longer term, I think the big advantage is it will create completely different compute architectures. And I see opportunities in the way that the compute is happening, but also in combinations where you have optical compute combined with CMOS. If you look today to the problems of optical neuromorphic computing, one of the best. big problems is the periphery. It's how to control that whole optical neural network.

And that is bringing a lot of additional power requirements and complexity. The fact that we can co-integrate the photonics on the CMOS and really bring these things close together, we can most probably solve a lot of bottlenecks. And I think our system will be become now we are focusing on the interconnect space, but I think on the longer term, the system will become hybrid, where you optimize the compute in the, in the photonic space or in the or in the CMO space, there where it's the most efficient. And these are novel solutions that are difficult to conceive with other technologies. Great. Excellent discussion.

Cedric, it's a pleasure meeting you. We want to have you back again for sure. And next time you're out, a conference, it would be pleasure to meet. you, maybe SPIEE next year. Thank you, Daniel. That concludes our podcast. Thank you all for listening and have a great day.