Matt Martin, Managing Director of Oxford Instruments Nanoscience is interviewed by Yuval Boger. Matt discusses his company’s focus on dilution refrigerators for quantum computing, the differentiation between different refrigerators, modular designs, and scalability considerations. We also cover service, maintenance aspects, recent developments, and much more.
Full Transcript
Yuval Boger: Hello, Matt, and thank you for joining me today.
Matt Martin: Thank you, Yuval
Yuval: So, who are you and what do you do?
Matt: My role is managing director of Oxford Instruments Nanoscience, which I’ve been doing for six months. Prior to that, I spent four and a half years as the director of engineering. I’ve been looking at our portfolio and basically rebuilding our roadmap and our collaborations and that’s been really exciting.
Yuval: And what does Oxford Instruments do?
Matt: Oxford Instruments was the first spin-out of Oxford University, focused on superconducting magnets and cold dilution units. This was in 1959, so we’ve been around for just over 60 years. Effectively, in terms of the quantum space, we are probably the most historic company doing dilution refrigerators and very large magnets. Our history has entirely been providing research equipment into the materials measurement and materials characterization markets.
Yuval: I like that. So, quantum will be an overnight revolution, 60 years in the making.
Matt: Pretty much!
Yuval: So educate me if you can. Obviously, dilution refrigerators are used with superconducting qubits. Now, I work for a company that makes neutral atoms. We operate at room temperature. But let’s say I was in the market for a dilution refrigerator. What sets one vendor apart from the other? What are the important characteristics of a dilution refrigerator?
Matt: That very much depends on what you want. Ultimately if I distill it down, in superconducting qubits, what does a dilution refrigerator do? And the answer is it provides a cold environment. As long as it provides that cold environment reliably, repeatedly, and allows it to stay cold for years at a time, it’s probably meeting the minimum requirement. But there are then many priorities and preferences that would influence your decision on which solution to use.
Yuval: What sets your company apart?
Matt: We have made the coldest continuously operating dilution refrigerator. We have systems that can comfortably go down to four millikelvins and stay there. We’ve got some of the lowest vibration systems in the marketplace for the research space. And for us, when we developed the latest version of our Proteox range we looked closely at wiring reliability and modularity.
We decided the best solution was to do our wiring off-system in what we call a secondary insert. With the secondary insert, you can qualify it, and move it between a small fridge, a medium fridge, and a very large quantum scaling fridge like the Proteox QX. It’s always the same. It can always do the same thing. Even if I’ve qualified the experiment small, then I can, with confidence, take it all the way through to large.
We looked at modularity and tried to understand the quantum scaling market when we designed the product. And the feedback we’re getting back from the customers is wow, somebody’s actually done modular properly. And that’s really positive to get that sort of reaction.
Yuval: How well does it scale if I went from 100 superconducting qubits to 10,000 in a couple of years? How well does it scale in terms of, whether is it a single fridge, is it multiple fridges? How does power consumption scale? Anything you can share with me about scaling would be wonderful.
Matt: That’s a really good question. And there’s probably a few things to pull out. One is about cooling power. One is about wiring scaling. One is about the total capacity of the fridge. So I’ll probably go back to my modular answer to that. If we take a small fridge, we have a bunch of components that make that fridge work very well for us. There’s a pulse tube, a type of cooler, there is the dilution unit itself, there’s the wiring solution, and there’s the gas gap which we use to control the temperature between the top and the bottom of the fridge and cool it down. And in the smaller version of our fridge, the Proteox MX, all of these components exist. When we go to the larger or the mid-scale version, the same things exist but in a multiple of two. The form factor is still round.
Now when we went and looked properly at scaling, anybody who’s tried tessellating anything finds out that round things don’t tessellate very well. In developing the larger Proteox QX dilution refrigerator we asked the simple question, if I needed to put lots of fridges together, what’s the best shape that tessellates well and easily? The answer is people are very used to laying out squares like chess boards. And we asked how do we make sure that all the connectivity between all of these common elements go together? We’ll put lots of squares in a line, we’ll have access front and back, and then all the connectivity faces will be side and side. We can probably make the line as long as anybody wants. We can go from a few hundred qubits in a single secondary insert to two, three, four hundred qubits on two secondary inserts to perhaps a thousand or more qubits on many secondary inserts of which I can put multiples in modules of QX. We’ve built a very modular and linear model.
Effectively QX for us is a block, it’s a square, I can put four or six of these secondary inserts in it and I can put as many QXs together as I like to create an environment for one thousand or two thousand qubits. We answered the wiring question by using secondary inserts, so the next challenge is scaling the cooling power and maintaining environmental efficiency. There are different ways to do this with different cooling methodologies, which becomes a balance between power and cost. Depending on our customers, that answer is different.
Without breaking any of the confidences of the conversations we’ve had, groups of customers have each answered that question differently. That’s a long answer but it’s quite a complex question. Going back to my original point, what makes your fridge different? One aspect is the way that we can scale and the consistency of the components we use and the fact that we genuinely have a modular system which means that the limit ultimately comes down to budget.
Yuval: If I spoke to some of your customers and I know some of these relationships are confidential, would they also concur that cooling is not the major hurdle to scaling the number of qubits?
Matt: Cooling is a hurdle and depending on what they want to do, cooling will matter. The thing that lots of our customers will talk about is wiring. Anybody who’s really trying to scale is coming up with their own wiring solutions. There’s also a whole bunch of work going into cryo-CMOS: Cryo-CMOS refers to CMOS (complementary metal-oxide semiconductor) technology that operates at cryogenic temperatures. Sticking cryo-CMOS in the fridge poses its own cooling requirements. It’s a different story again about where you need to take the heat out in a fridge.
In quantum, when you go from the brute force approach that people are talking about at the moment, to a scaling fridge that doesn’t use brute force or doesn’t use flexible wiring, there’s a whole bunch of new cooling challenges to be resolved.
In designing a fridge for customers, the question is always what matters most to you? Because depending on what other programs a customer has, sometimes they want maximum space under the mixing chamber because they have a lot of wiring that needs to be resolved. Other customers are asking to give us the biggest space and the most cooling power because they have a lot of parametric amplifiers or cryo-CMOS that they want to cool. And other people have said, we just need space. Just give us the biggest thing you can possibly give us. What we need to do is do the experiments because we need to prove that we will solve our issue with error correction by just having more and more of it and then proving that a big system will work. Again, I don’t want to break any confidence, but every single one of those questions has been answered differently by any number of customers in the last couple of years.
When we launched Proteox QX at APS last year, we launched a concept. We didn’t launch a product. Everybody’s taken the concept and gone, ‘can I have one that’s green? Can I have one that’s bigger? Can I have one that’s taller? Can I have one that’s a specific requirement for our robot?’ And that’s been really enjoyable getting to properly understand what people want and provide solutions.
Purely and simply, I had a group of engineers that have been doing fridges for years, and those fridges have been going into research and they’ve enjoyed making the lowest vibration. They’ve enjoyed making good cooling powers. And they’ve enjoyed creating better automation software. But ultimately it wasn’t a completely different challenge. Whereas going to scaling, it was a very different concept that the engineers had to work through and the scientists had to work through. That was a challenge and people have really enjoyed those conversations.
Yuval: I’m curious, are dilution fridges a significantly growing part of your business? Is it just a small thing? And you have a lot of other non-quantum things that are really keeping the lights on?
Matt: Quantum is such an attractive market. There is so much investment going in. Everybody says you’ve got to put all your efforts into quantum because otherwise you’ll miss out. And that’s true. There is such an opportunity here. But fundamentally, we’re doing work to support superconducting qubits.
Which modality of qubits will win? Will the NV (nitrogen-vacancy) guys win? Will the photonics guys win? Will the superconducting guys win? The answer is nobody really knows today. We know where the bulk of the investment is going. But just because that’s where the money’s gone doesn’t guarantee that that’s the space that will become successful. I mentioned at the beginning that we’re a company that provides environments and measurement capability.
As much as we’re focusing on quantum, the core historical measurement part of our business, the customers doing electrical transport measurement, thermal, any number of different variations of those things is still something that we are completely focusing on. Quantum is only one of the two pillars that are part of what we do.
We’ve got products like our Teslatron or our Spectromag that are used in research universities throughout the world to do general materials characterization, and that’s not a market we’re moving away from. If anything, we’re trying to reinforce that market for the simple reason that we don’t know which quantum modality will win, and you don’t back all of your money on one horse if you’re spread betting.
Very much we’re looking to underpin our historical dominance in magnet technology. We’re doing some great collaborations with people like the MagLab in Florida and with European project ISABEL, to try and drive that future magnets technology. We’re moving from Low-Temperature Superconductors to High-Temperature Superconductors, as much as we’re doing all this work in trying to scale quantum with other customers. This is what makes it such an exciting time for us and Oxford Instruments at the moment.
We’ve got so much to do, and we had to grow the team dramatically in the last couple of years to do it. I’m trying to find people who understand large-scale systems. We’re trying to find more and more research application scientists, and PhD physicists to drive our measurement market. We’ve almost got two camps forming inside the business about the way we need to operate. One is an industrial company, and one is an absolute materials research instrument company, and that’s been great trying to manage that transition.
Yuval: Quantum is changing very rapidly, so I’m curious, what do you know today that you didn’t know, say nine months ago?
Matt: We know that various companies have got great roadmaps, and being able to work with them and see the directions they’re going in and how that’s affecting our product has really changed some of our thinking. We know that the way people are trying to scale is different than we thought it was, which we certainly didn’t know nine months ago. We know that as part of trying to understand what those challenges are, we found out that we can contribute more to those roadmaps than we thought we could.
I’ll give you an example. How do you install a system that’s got to weigh many tons, and will have maybe dozens of racks of ancillary equipment? It became a question that wasn’t just about scaling science. It became a question about how we fitted our fridges with customers like OQC to data centers. What are the constraints of the data center? When we talked about scaling and how you physically fit a fridge that is four tons, and then you put two, three, or four of them together and say, well, how do you get that into a building and how do you manage the vibration and how do you manage the installation? What is the service engineer commitment? We didn’t know that nine months ago. When was the first time that anybody put a quantum computer in a data center? It all happened in the last 12 months. And now we’ve done four of them.
We’ve really learned about how effectively the superconducting qubit is translating into application requirements. But for us, we are not making the computer. We are with the support team. We’re the pit crew that says we can make your car, and your environment work. And the constraints of all of those people looking at how they deploy have taught us lots of different things.
Yuval: If there’s such a thing as a typical project, how long does it take from here’s a purchase order to now you want me to sign off on the delivery?
Matt: We’ve done a lot of work in our factory to optimize lead time. For our standard product, if you bought a standard fridge from us today and you dropped a PO and it fitted our standard configuration, we could ship that to you in under six months. If you came to us and said, we’ve got a real bespoke requirement here, depending on what you’re asking for, that could take anywhere between nine months, a year, or be a two-year program.
We’ve got a great history of doing world-leading equipment, both in dilution refrigerators or in magnet systems with dilution fridges in them. People using our equipment have won Nobel prizes, and some of those are of course complicated, bespoke projects. We enjoy one as much as we enjoy the other, but it really depends on your requirements.
What we’re trying to do is move away from doing lots and lots of bespoke project at the moment, purely and simply, because that’s how we’ve liberated the capacity to focus on the quantum scale. As a business, you can’t focus on everything. We’ve made some really, really clear decisions on looking after the general measurement market and focusing on quantum and quantum scaling. They’re the two key pillars that we’re working on at the present time.
Yuval: If I run maintenance for a significant data center, now you deliver a dilution fridge to me. Do you need to train me in things that I don’t know how to do today? Do you leave a person behind with the fridge to help with the maintenance? How does it work from that perspective?
Matt: With industrial scaling, we’ve done a lot of work developing service solutions for people. Most of our customers want the environment including a service contract that’s part of the solution. That makes their problems go away. Our service team really enjoyed going, right, how is that different from what we provide today? Our Plasma team is used to doing service models. So it’s the great advantage of being part of the Oxford Instruments group. We walked over to the Plasma guys and asked how do you deliver 24×7 support to a plasma tool? I’ll borrow that. And that’s exactly what we’ve done. We borrowed knowledge from across our business and said, we’ll do that for a fridge. And that’s been great to be able to provide those services.
Yuval: As we get to the end of our conversation, I wanted to ask you a hypothetical. If you could have dinner with one of the quantum greats or maybe the refrigerator greats, who would that person be, dead or alive?
Matt: How do you define the greats? I’ll answer it personally. When I first started in quantum, my background was semiconductors. So different physics. When I first came across this industry, I went, what do I need to learn about quantum? I went onto YouTube and spent a good period of time watching a lot of Andrea Morello’s videos on introducing quantum. I’ve never met the gentleman, but I would really love to spend dinner with him and say, thank you very much for that entire video series that you did. And can we talk in person? I don’t know where you put him in the pantheon of quantum physicists, but for me personally, I found his videos characterful, entertaining, and insightful. I’d love to have dinner with him.
Yuval: Wonderful. Matt, thank you so much for joining me today.
Matt: Thank you. Pleasure.