Why Silicon Spin Qubits Can Ride Existing Chip Fabs

Summary

Andrew Dzurak, founder and CEO of Diraq, a company developing silicon-based quantum computers built with existing chip fabrication technology, is interviewed by Yuval Boger. Andrew and Yuval talk about the merits of their technology, the reason many companies have taken alternative approaches, when silicon dot computers will be available, a hypothetical dinner with Paul Dirac, and much more.

The full transcript is on the Quantum Computing Report site here

Transcript

Yuval Boger: Hello, Andrew, and thank you for joining me today.

Andrew Dzurak: Hi, Yuval. It's a pleasure to be with you.

Yuval: So who are you and what do you do?

Andrew: My name's Andrew Dzurak, and I'm the CEO and founder of Diraq, which is a full-stack quantum computing company. We're headquartered in Sydney, Australia, but with very strong links all around the world.

I'm also a professor at the University of New South Wales, which is how I got into the business. I've been working in the field of quantum computing, specifically silicon-based quantum computing, for over 20 years and have been looking to commercialize this technology for a long time.

We've had a few preparations for commercial activity, and finally, in May this year, we launched the company with some support from private equity investors.

Yuval: What took you so long, if I may ask? I mean, you were working on this for 20 years. Was it a particular technical milestone that you were waiting to achieve, or maybe IP licensing from the university, or anything of that nature?

What caused you to start the company just a few months ago?

Andrew: Mm-hmm. So it was primarily related to issues around IP and so on. We had developed the core technology for our CMOS-based approach around 2014, 2015. We did some crucial demonstrations and published some work in Nature, followed up by quite a lot of other results.

And then it was really a matter of negotiating with the university and other parties in Australia in order to effectively get the company investable, which meant we needed to have a clear IP ownership structure. And so a lot of the negotiations were to get the IP essentially out of the university and into our company, which is where it is now.

And we also have a clear stream of IP ownership moving forward. But during that period, we were able to essentially tick off a lot of the technical milestones that we needed to do anyway. Fortunately, my group's been very well funded by both the Australian government and US government investment for 20 years.

And so we had a lot of resources from research grants, which allowed us to achieve a lot even while we were in that period waiting for commercial investment.

Yuval: So what are silicon-based computers, or spin qubits, or silicon dots? I saw so many names related to that.

Andrew: Yeah. So spin qubits — there are different flavors, but the type that we are exploring in Diraq is based on single electrons or a small group of electrons.

And we use the quantum mechanical property known as the spin of an electron, which is related to a tiny magnetic moment that the electron has. As well as its charge, it also has a magnetic moment, which is quantum mechanically related to a phenomenon known as spin. You can think about it as a ball of charge spinning on an axis, and it can either spin clockwise or anti-clockwise, and therefore its magnetic moment points either north or south when put in an external magnetic field.

And so we use that to encode the zeros and ones, our qubit states. That's the general thing for all sorts of spin qubits. In our particular implementation, we hold those electrons inside tiny transistor-like devices. In fact, they are actually just modified transistors of the type that you would have billions of on a chip in your mobile phone or in your laptop.

And so one of the big breakthroughs that we made around 2014, 2015 was when we showed that it was possible to modify a standard silicon transistor on a chip and operate it as a spin qubit, and read out its spin and control its spin with very high accuracy, or fidelity.

Yuval: So it follows that a big advantage of this approach is the scalability, right?

You would think that just like one could create billions of transistors on a regular computer chip, one would one day be able to create a very large number of qubits this way. Would that be an accurate statement?

Andrew: That's a very accurate statement. That's essentially the major commercial and strategic advantage that our technology has.

And just to give you a bit of background, I mentioned I've been in the research business for twenty years. Initially, I was working on a different type of spin qubit in silicon, one that was confined by a single atom. And I worked on that for a long time, and still have some very talented colleagues working on that technology in Sydney and elsewhere.

But I was concerned about the ability to manufacture it at scale. And so it was around 2007 that I first showed in our group that it was possible to operate a modified transistor at the single-electron level. And so it's the scalability that is the big advantage, and I recognized that early.

And then, once we patented and demonstrated the high-fidelity qubits on that technology, we've really been spending all our time designing ways to ensure that this on-chip scalability can work successfully.

Yuval: Is that similar technology to what Intel is working on?

Andrew: It's very similar technology to what Intel is working on.

In fact, almost identical. I think it's fair to say that our breakthroughs at UNSW inspired a number of groups and a number of entities around the world, and certainly I know that our work inspired the interest of Intel to get involved in that. And they've been making some very impressive demonstrations in their foundry.

They demonstrated a high-fidelity spin qubit in the technology we designed around a year or so ago. And that's very, very encouraging for the manufacturability. I was always convinced that it was possible to produce these in large-scale manufacturing facilities, and it appears to be the case.

And indeed, it's very encouraging for us as we are working with our fabrication partners around the world to develop large-scale chips based on this technology.

Yuval: So you're looking at major chip companies more as partners than potential competitors?

Andrew: That's right. For commercial reasons, confidentiality reasons, I can't say any specifics about particular partners that we're working with.

But we do have partners who are significant players in this area, and we see them as partners moving forward. So our company is very much focused on the design, the assembly, the operation, and the implementation of quantum computing based on our technology. We won't be doing the chip manufacture internally in our company.

We require the manufacturing capabilities of what's called a tier-one chip manufacturing foundry. Those foundries cost many billions of dollars to create, and so we'll be working in partnership with them. Just as many companies, like Apple and so on, go out to chip manufacturers to make the chips for their products, that's exactly the approach we'll be adopting with Diraq.

Yuval: What does connectivity look like between the qubits? Is it nearest neighbor? What kind of operations can you do, or how far does one qubit need to be from another to implement a two-qubit gate?

Andrew: Right. So the natural two-qubit gate or connectivity is nearest neighbor. It's implemented via what's known as the electron-electron spin-spin exchange interaction, or Heisenberg exchange.

And so for that you need a small amount of electron wave function overlap, to get very technical, and that's related to the size of the transistor or quantum dot that we use to confine the electron wave function. And typically, in our devices, we need center-to-center distances between our quantum dots of somewhere between 50 and 100 nanometers.

Okay? So anything in that range, it's possible to perform two-qubit gates. Now, in addition to that, a year ago we published in Nature Communications an experimental demonstration of the ability to move an electron and its spin between quantum dots coherently, while maintaining quantum coherence.

And what that means is that we can do shuttling of qubits through an architecture. You can pick up an electron qubit from one site, move it around an array of quantum dots, and take it to another site. So it allows beyond-nearest-neighbor connectivity, and that can be very, very helpful in providing a bit of extra flexibility in architectures and in error correction.

Yuval: How soon would you anticipate that you would have, or someone would have, a chip that works, even if it only had dozens or hundreds of qubits?

Andrew: So we've actually, on our website, if people want to take a look, diraq.com, spelled out our technical roadmap that we've worked up together with our partners.

And we see that well within a decade, we will have CMOS chips with — we've publicly stated — 256 qubits by the end of our second phase, which, without going into specific timescales, is, let's say, somewhere in the range of five to six years. But in fact there's the opportunity to go to much larger numbers than that.

So while we're happy to very publicly state we're confident of reaching 256 in five to six years, we're seeing the potential to go much larger than that, and to some extent that's going to be dependent on the investment that we have in the company and what we're able to do with our fabrication partners.

This is really just a matter of having the financial resources to be able to undertake the manufacture. And there's some process development that needs to be done together with our partners. This is primarily related to just slightly modifying the integration arrangements for the wiring.

Because for the qubits, it's a little bit different wiring than is used for standard transistors. It's just the layout of the addressing of the gates. And so it's really just a matter of investment with our partners. So we think that we can get to many thousands well within a decade.

And ultimately, many hundreds of thousands, potentially a million, by the end of the decade.

Yuval: From a software perspective, does it look like just a general gate-based quantum computer? I'm sure there's low-level control that's specific to your technology.

But in a few years, once the chip is available, if I had a Qiskit program, would I expect to be able to run it on your computer as well?

Andrew: Yes, you would. We're designing "standard," in quotation marks, gate-based quantum computing. We have some slight modifications of error correction schemes, but they're largely surface code-based, if you're familiar with that.

We are looking, both internally and with partners, to develop some aspects of custom software ourselves, but we also want to make sure that we have appropriate compatibility with other systems as well. But the general principle, gate-based computing, that's what we're working on.

Yuval: You've been publishing in very high-end journals, Nature and others, for many years, so the work has not been kept a secret.

Why do you think other people are using neutral atoms, or ions, or superconducting, or other technologies if the advantages of silicon-based are so clear in terms of scalability, in terms of operating temperature, and others?

Andrew: Yeah, that's a really great question. And the answer, I think — well, I'm pretty confident — is that a lot of those alternative technologies are based on systems that were already demonstrated some time ago.

And where the entry level to performing experiments is much lower than it is in silicon qubits. So, for example, the technology related to superconducting electronics had been developed by IBM in the 1980s. And superconducting quantum interference devices, which share a lot of commonalities with spin qubits, had been developed in the late '80s and early '90s.

And so it was a relatively straightforward step to start making qubits based on superconducting circuits. And furthermore, the actual size of those devices is very large. A typical superconducting qubit is of order one hundred microns across, right? In contrast, a silicon qubit is less than one hundred nanometers.

So we're talking about a scale of between one hundred and one thousand in length, which corresponds to a scale of between ten thousand and one million in surface area. So it's essentially much easier to make devices like that. Similarly with ion traps; in that case, there'd been some great work on atomic clocks.

Some of the first control of quantum states was done in those quantum systems. People like David Wineland and Serge Haroche, Nobel Prize winners who'd worked on single-atom control. So there was a lot of capability already in place experimentally, and it was therefore — I don't want to use the word straightforward, but it was a more straightforward path in a university or research institute laboratory to do that.

In contrast, silicon qubits, because they're so small, require very advanced nanofabrication. And usually, in order to get reliable devices, this is only done in tier-one semiconductor chip foundries, which, as I mentioned earlier, have capital investments of billions of dollars. Now, there are a handful of laboratories at universities and government-funded research institutes that have capabilities in this area.

An example is our facility at the University of New South Wales. We had an excellent facility known as the Australian National Fabrication Facility. And so we were able to get into the zone of doing demonstrations. But it's important to note that it's right on the edge of what you can do in a government-funded, and certainly a university, lab.

While you can do demonstration devices, and that's what our group has been doing over the past 10 to 15 years, it's very, very difficult to get yield to go to large numbers of devices where every gate, every transistor works. So that was the long answer. The short answer is the cost of entry into silicon qubits is very high.

You need a lot of resources. You need very advanced fabrication, and that's not generally available to most research groups. So that's why those other technologies were the ones that have been explored in the most detail and have made the most advances so far. My view, and the view of a lot of colleagues and people who are looking at this — and I think smart investors see this — is that once the decision is made to make the investments that are required in silicon and in chip foundries, then when you get that rapid scaling, you'll basically be able to shoot past those numbers of qubits.

Yuval: Could you tell me a little bit about the company?

How many people are you right now? What kind of funding, if you can disclose? I think you announced a chairman recently.

Andrew: That's right. So at the moment, the company is around twenty-five engineers, with a very small business team so far. We were founded with just over twenty million Australian dollars of investment from our partners, Allectus Capital, who are the major shareholders in the company.

And we are at the moment in the process of additional capital raising, and we're looking to double the size of the technical team to around fifty in just the next year or so with that additional investment. As you mentioned, we've just appointed a new chairman of the board, Bill Jeffrey.

Bill was formerly the CEO and president of SRI International, formerly the Stanford Research Institute. And before that, he was the CEO and president of HRL Laboratories, formerly Hughes Research Labs, a major defense and technology contractor for the US government.

And in fact, HRL Labs has been a collaboration partner of our group for a long time. And so Bill brings a wealth of experience in the semiconductor area, in government and defense investment, and in the venture capital space in the Bay Area. Bill's based in the Bay Area. Our relationship with the US is very important, as it is with all of our other partners around the world, in the UK and Europe, et cetera, and Japan.

So we see Diraq, although headquartered in Sydney, very much as a global enterprise. So Bill's involvement is really, really crucial to help cement the strong links that we already have with the US.

Yuval: Excellent. And as we get close to the end of our conversation today, I'm wondering, a little bit off topic, if you could have dinner with anyone in the quantum field, dead or alive, who would that be?

Andrew: Well, that's an easy one for me to answer, isn't it? Because it's Paul Dirac. Interestingly, though, it might be a stilted conversation, because Dirac was famously a man of few words. But when he did say those words, they were amazingly profound. So it might be a tough dinner to have, but it would be an amazing experience, and I couldn't really pick anyone else.

But if I can have a runner-up: in contrast, a person who wasn't a man of few words was Richard Feynman, and I think that for laughs and for entertainment, as well as for genius, he'd run a close second.

Yuval: Excellent. How can people get in touch with you to learn more about your work?

Andrew: Well, just go to our website, diraq.com, D-I-R-A-Q dot com, and you can see about our technology, but there's also a contact page there that you can reach us on, and we'd be happy to come back to you.

Yuval: Andrew, thank you very much for joining me today.

Andrew: It's been an absolute pleasure, Yuval.