Podcast | Quantum Error Correction on a Single Qubit— with Nord Quantique

Julien Camirand Lemyre: You’re correcting errors on a single-qubit system to build an architecture that leverages for quantum error correction at scale, and you require fewer resources when you scale the system. It’s a way to improve performance and reduce the need for large-scale architectures.

Konstantinos Karagiannis: Error correction typically involves taking a lot of physical qubits and using them to create one logical qubit. Ratios vary by modality and approach, so getting a single fault-tolerant qubit may take seven to a thousand physical ones. What if there was a way to correct most of the errors that appear on each qubit instead? Scaling up would certainly be a lot easier. Find out the details behind one such new approach in this episode of The Post-Quantum World. I’m your host, Konstantinos Karagiannis. I lead Quantum Computing Services at Protiviti, where we’re helping companies prepare for the benefits and threats of this exploding field. I hope you’ll join each episode as we explore the technology and business impacts of this post-quantum era.

Our guest today is the president, CTO and cofounder of Nord Quantique. Julien Camirand Lemyre, welcome to the show.

Julien Camirand Lemyre: Thank you for having me. Pleased to meet you.

Konstantinos Karagiannis: Today, we’re going to primarily be discussing error correction, which is hugely important, of course, in advancing quantum computing. Your company has a unique approach about that that we’ll get to in a moment. But first, for those who might be new to the topic — maybe it’s their first episode — how is error correction typically handled in quantum computing?

Julien Camirand Lemyre: First, why do we think quantum error correction is important? Something we’ve seen from the quantum industry in the past year is enormous progress in terms of technology development and what platforms are accessible right now to end users as well. But one challenge that remains is, it’s still hard to control qubit to a level we would like to control.

These qubits are still making errors. Right now, hardware platforms, whatever the modality, are somehow limited to an error every 1,000 operation or so, or 10–3 error rates — it depends on how you like to count things. The way people have envisioned solving this issue is the same way you usually approach this problem in classical information or classical computing — basically, by building redundancy inside the process you’re doing, or inside the chip, in the case of quantum computers. The idea, simply, is, if you have a qubit that’s faulty, you can add more of them and call this a logical qubit. You operate them in such a way that these together would form a logical qubit that will then be able to carry a quantum computation, and you run an algorithm that enables you to correct these errors.

Now, the challenge is that with the typical error rates we get on this hardware, we will require quite a bit of resources, quite a bit of qubits, to be able to do quantum error correction — we currently envisioning having the need for roughly 1,000 to 10,000 physical qubits per logical qubit that encodes the quantum information. This is the way we usually envision building quantum error correction: by building these arrays of qubits, on top of which we run these quantum error-correction algorithms to correct errors as they occur within the quantum computation.

Konstantinos Karagiannis: I like to think about classical error correction as like an iceberg: What you see, that’s the qubit, you think, and then underneath, there are all these other qubits helping to keep it floating.

Now, Nord Quantique is doing something a little different, and that might have some big impact on scaling. Let’s talk about that. Error correction in a single qubit built-in — how does that work?

Julien Camirand Lemyre: In a single physical system that we call a logical qubit, the idea is the same. I was talking about redundancy — you also refer to the iceberg. It’s always the same trick. We’re trying to add redundancy or add ways to gain knowledge about the quantum information embedded in a quantum system. We do this by using microwave resonators or microwave cavities, in which we can store many microwave photons. These photons, we’re able to control within our superconducting qubit platform to form a single logical qubit — a qubit on top of which we can perform quantum error correction.

That we are able to do this comes from the fact that we have built-in redundancy rising from the many photons inside the cavity or the quantum objects. The idea is the same. Now, the impact is twofold. One way, this provides a way to correct errors on single physical qubits. This is something we find interesting in the perspective of reducing the error rates on physical qubits, which is not possible with more traditional qubits.

On the other hand, it also opens up ways to think about scaling that are a bit different, because while we do this, we also have access to something unique to this platform. We have access to the information of whether a qubit was faulty and have some level of confidence of whether a qubit, in particular, is faulty or not. This can be helpful when you build an architecture with quantum error correction embedded for the whole platform together to lower the overhead in terms of scaling. The more information you have in that case, the more quantum error correction you can perform on that architecture.

Konstantinos Karagiannis: How different is this modality from other types of photonic systems?

Julien Camirand Lemyre: It is very different. Our platform is a superconducting platform, so we do use microwave photons. This is natural for any superconducting platform. The way you would drive a traditional transmon qubit, which has been the building block in superconducting qubits for more than a decade now, is by driving these systems with some photons. Also, resonators are part of the toolbox of these systems.

Now, in a photonic platform, if you compare it to something like other companies would do using photons of light instead, it’s very different. First, our modes are not propagating modes. We don’t have photons that are propagating through space. They’re localised in quantum objects. That we are using the superconducting toolbox provides us a way to build nonlinearity within the system. This is helpful because we have a lot of control in the system we do, and this is why we are also able to produce, more easily, these quantum states that can be error corrected. This is the main difference between optics and superconducting — the level of nonlinearity you can build in this system and how localised these can be in space.

Konstantinos Karagiannis: Thanks for clarifying that, because anytime people hear “photons,” they get a little confused, and I just want to make sure listeners understood that. I got to see images of this device, and it’s definitely unique and memorable. In this phase, we still need some of that.

Everyone’s familiar with the chandelier refrigeration look, but when it gets right down to what’s going on the chip level, it becomes a little more muddy. You’ve got this thing — is it like an acorn or something like that? It looks almost like a hexagonal nut or something you turn.

Julien Camirand Lemyre: Exactly. The image you’re referring to is a superconducting cavity. It’s the first prototype we’ve been using in that perspective. We like these cavities because they produce photons with a very long lifetime for superconducting circuits. Photons can live for up to a millisecond — a bit more, sometimes — in these cavities, which gives us a lot of time to play with them, in some sense.

The exact shape you’re referring to is for tiling. When we are thinking about scaling, the system is also tiling them together. The shapes come from that direction. They’re cavities, there’s a hole in it. This hole defines a mode together with a pose that is within the mode. It’s just like a coaxial cable that has been cut on one end and grounded on the other end, which produces a resonant mode in which photons of that frequency can live.

Konstantinos Karagiannis: I was going to ask about scaling. Is that designed so you can cram them together in, like, a honeycomb grid?

Julien Camirand Lemyre: We have different designs we’re exploring as well. That’s the first step we’re taking with this platform.

Konstantinos Karagiannis: People can visit your site from the show notes and see if they want to get a visual sense of what this looks like.

Julien Camirand Lemyre: Not only do we have the pictures, but we also have a bunch of blog posts that explain, what is this technology? What are bosonic codes? Where are the microphoton stores? These kinds of questions as well.

Konstantinos Karagiannis: Are there performance numbers available to quantify how effective this approach is right now?

Julien Camirand Lemyre: When we created Nord Quantique, the challenge was to build a system that can have this disadvantage of being error-corrected by themselves. Our core focus in the early years was to build a system, of course, but also to operate it and demonstrate that quantum error correction is effective on these systems. The way we’ve measured it is by using the system as a memory. We’re measuring the lifetime: How much time can we preserve the quantum information of the collective photon states that live inside the system, this cavity, if we do error correction or if we don't do it? What we’ve demonstrated is that by doing error correction, we can extend the lifetime of this system by a factor of 24%.

This is the first step. Of course, we’re trying to crank up that number. We have a lot of efforts across the stack, from software algorithm to hardware development, to push these numbers up. But this is the first demonstration, and this is important in one perspective, because quantum error correction is extremely challenging. I was referring to the need for redundancy before, and we’re seeing more and more of these experiments trying to achieve this with arrays of qubits. People are starting to succeed with these arrays. But we do need a lot of qubits in these architectures. Google did an experiment with roughly 50 qubits, and QuEra, with 200 atoms, to get to the point where you see effects of quantum error correction. In our case, it’s on a single system, and this is what we like about it.

Konstantinos Karagiannis: When you have these humming along, is it the idea to have them be able to be reset quickly? Is that something you’re trying to focus on for running?

Julien Camirand Lemyre: We do reset some elements in our system. Any quantum error-correction platform needs to have some forms of reset.

The idea is twofold: Coming back to what I said about quantum error correction at the beginning, on the one hand, it is a way to improve performance and reduce error rates, in some sense, in these systems. The other way is to use the information that is emerging from the fact that you’re correcting errors on single-qubit system to build an architecture that leverages for quantum error correction at scale and require fewer resources when you scale the system. It’s a way to improve performance and reduce the need for large-scale architectures.

Konstantinos Karagiannis: When you’re working on an approach like this, I imagine if any of these qubits fail or introduce errors earlier than expected or something, that is going to have to trigger some kind of alert. It sounds like the whole process itself is very involved. Is this a full-stack approach, then, to trying to get these qubits to behave?

Julien Camirand Lemyre: Exactly. This involves a lot of efforts we have been focused on at Nord Quantique, and this goes all the way down to hardware. Improving hardware performance is important, and not only for us but also for the old industry and the old superconducting qubit community as well. But in our case, algorithm development is also very important. We have been testing different ways of performing quantum error correction on these systems, falling into some pitfalls, but also finding things that succeed in terms of improving performances. To do this, we do have a stack that covers up everything from controlling the hardware itself to simulating these systems accurately and dealing with quantum error correction. It’s a development that goes across what we like to call real hardware — hardware and hardware control, but also the middleware stack that involves quantum correction

Konstantinos Karagiannis: That stuff will live with this system and evolve with it in the future, and then you’re going to ultimately have something that the public can interface with on the other end. But these layers will be there from now. On the topic of more of a software approach, I was reading about bosonic polyplus. Can you talk about that?

Julien Camirand Lemyre: One challenge of migrating from these arrays of qubits to having a system with a lot of photons is that when we’re simulating these systems, it’s quite involved. Every time we’re trying to simulate a single one of these bosonic modes, we require quite a lot of classical computing power to be able to simulate these things. This is a challenge. If we’re trying to simulate the whole system completely and get all the physics that is out of there, we are limited to a handful of modes we can deal with. That means if we are trying to build an architecture and trying to see how, for example, quantum error correction, which was the point in this bosonic polyplus paper — if we’re trying to see how many of these qubits will interact — we need approximations.

We need ways to deal with qubits inside an error-correction architecture, and, more specifically, with the types of qubits we’re playing with, which is the bosonic codes we’re working with. Now, the challenge is that in the quantum error-correction community, people have found ways to describe a system that can be efficiently described for the purpose of quantum error correction. Usually, when you simulate, one of the most notable codes, for example, is the surface code, so we have ways to simulate this efficiently so we can simulate the system for quantum-error correction purposes using hundreds or thousands of qubits. This is possible.

Now, here, we’re adding a bit of complexity because our systems have an internal degree of freedoms that are important when we’re also thinking about quantum error correction. We want to know, is this qubit faulty, or not, which is something that usually, in quantum refraction code, the way we simulate it, we don’t necessarily care about. This paper was about developing this method where we are not tracing out the complexity of the system, but reducing it to some qubit parts that are there, but also keeping some information about, was there leakage to other levels, and were there errors occurring on these qubits later? And this we integrate then into the concatenate code, which is the architecture surrounding using many of these bosonic qubits together. This is the purpose of this code, and maybe it links to your previous question.

As we come up with new technology like this, there are also other challenges that come from across the stack. One of these challenges is designing the system and getting an accurate design to simulation and basically polypluses away in that direction — making sure that when we’re simulating this quantum error correction architecture, we also take into account the noise, or at least, in good approximation, the noise that is occurring in the system and how they interact.

Konstantinos Karagiannis: To get to error correction, sometimes you hear about error suppression, error mitigation. Is there anything new there that’s being done in your hardware? Is there something that might even warrant its own name?

Julien Camirand Lemyre: The ideas are not necessarily new, in the sense that these are error corrections. The essence is, with the bosonic codes, we are building these qubits with bosons. What is the exact algorithm we’re running on this system to do error correction? The innovation, in our case, comes in the paper — not the bosonic polyplus paper we have out, but the one on the system itself. We show autonomous quantum error correction, which means we don't have a need for feedback. We have a system that autonomously corrects errors on the system as they occur. This is something we’ve put up together — this new block inside the sequence, how we design it and how we make it work inside the hardware. But this is pretty specific in our case — nothing that deserves a new name, but the ensemble is new in itself.

Konstantinos Karagiannis: That creates this wonderful qubit. How soon do we see putting two of these together and actually having gate operations in the real world?

Julien Camirand Lemyre: We’re getting there. This is why we exist. We need to have these systems not only performing quantum regression at memory but also, at some point, being integrated inside an architecture. We presented some early work at the March meeting last year, this big physics conference where a lot of the science is happening from the quantum computing field as well. We already showed a demonstration of the building blocks required to get to gates. For the results, you’ll have to wait a few months, but it’s coming. This is something we are working on right now and tying into the next step for the company, which is scaling these systems in the end.

Konstantinos Karagiannis: So the goal is to, over the summer or something, hopefully, get a couple of gates interacting. But then, if you want to take a gamble and think farther out, what kind of timeline would you predict for, let’s say, 100 of these logical qubits being available?

Julien Camirand Lemyre: Right now, we are working toward the smaller-scale system, and this will rapidly bring us to the level of 10 logical qubits. This is the next big challenge we are up to. We’re targeting 2028 for having something that is reaching up to the numbers you’re describing — to 100 qubits. This is, of course, very challenging, but we believe, with the superconductive platform, something we like about this is that there’s a lot of knowledge. There’s a big industry also working on scaling around this, and this is something we can leverage to get to that point as well.

Konstantinos Karagiannis: When we hit that point, if we have 100 logical qubits, even though it doesn’t automatically guarantee some kind of advantage, there’s that feeling that it does, because classical computers cannot simulate that many perfect qubits. We just can’t do it. We choke somewhere in the 50s. What do you see happening then? We’re going to be on the cusp of some kind of advantage. What problems or use cases are you most excited to see come to life with such a frontier you might reach?

Julien Camirand Lemyre: There are a lot of problems that excite us. Given the size of the company, we’re focused on a few of them. We like to partner with other people, and we’re working with a company on developing some new materials, and they’re interested in learning how these systems can get to that point where they can have advantage for material and chemistry designs. This is something that is exciting to us as well, and we believe that with roughly 100 of these logical qubits, we can solve them — materials science, material designs, and trying to simulate and design new chemicals. We will have some news around this sooner than later.

Konstantinos Karagiannis: There are some partnerships already in place where companies are so interested in solving these problems that they’re going to start working with you now, and that’s a great belief factor they have in you. That’s got to feel good.

Julien Camirand Lemyre: We are working toward logical qubits, but we also have these bosonic systems that are interesting in themselves, and on the roadmap, what opportunity lies ahead in terms of these material simulations and material developments is something we are still trying to keep ourselves aware of as we grow the company.

Konstantinos Karagiannis: That’s great. All this is going on in an exciting area. A lot of our listeners might not know about it. I’d love to talk about it. But can you describe Sherbrooke, where you are? It’s an Innovation Zone. What has that meant personally to you and the creation of this company?

Julien Camirand Lemyre: It is actually named the Innovation Zone — that is the actual name. Sherbrooke was recently designated by the provincial government. We have two government layers in Canada: We have Quebec and Canada. The provincial one has designated Sherbrooke as the quantum city, if you’d like. It means that there’s a lot of investment, and, more specifically, it’s a center where quantum is happening in Quebec and within Canada. It’s more and more recognised that Sherbrooke is one of the leading centers in Canada, with worldwide recognition also in terms of quantum development.

For Nord Quantique, more specifically, what it means is, Sherbrooke has this long history of working closely between academia and industry. A lot of the research, for example, will be done in close collaboration between industry and academia. This ties in to the microelectronic sectors. Sherbrooke is also the tipping point of what we call the northeast microelectronic corridor, which runs all the way down to Albany and ends up here in the Sherbrooke and Vermont ecosystem. This means there’s also a lot of supporting infrastructure that has been built alongside this northeast microelectronic corridor.

It turns out that throughout the years, this evolution also created a lot of opportunity for new technology emergence and new technology development, including quantum. For Nord Quantique, we turned that into an opportunity trying to leverage this ecosystem. We’re working closely, for example, with the Institute Quantique, a leading research center. It’s also where Nord Quantique was born, so we’re a spin-off of that research center and now have close collaboration with researchers there and other development projects.

We’re also using labs there. We have been renting infrastructure since the beginning of our operations. We are renting a superconducting fab, where we are fabricating and designing our chips and looking for the next steps into how we can leverage this infrastructure and how we can build the next stages into scaling these systems, bringing these superconducting circuits to scale. This is on the tech-development side.

What’s exciting also about the hub is, now it’s a real hub. There are other quantum computing companies in the ecosystem here. It’s a good cradle for interactions with providers, with developers, with government entities. That’s all gathering into Sherbrooke. It’s pretty exciting right now and bringing momentum in the activities in the region.

Konstantinos Karagiannis: That’s great. It has a lot of parallels to what’s going on in the U.S. We have a couple of quantum hubs here. We have Chicago at the Chicago Quantum Exchange, and we have Colorado and everything going on in that region. It is great to see this kind of model happening around the world because we all have a lot to accomplish here.

You’ve been there a long time. Before you started this company in 2020, you were there eight years before that, living in that ecosystem already.

Julien Camirand Lemyre: I was born in this ecosystem, but raised alongside it, I would say, as I was doing my studies here. We’re fortunate enough to see this growing, and being an active part of it made us aware of how to leverage it. You’re right to say that there’s a parallel to what’s happening in the U.S. but also in the different hubs in Europe. This is needed for the quantum industry. We need these poles where companies can develop together and bring momentum for end users to get traction and build the industry early.

Konstantinos Karagiannis: I have two more questions on this topic: One, with all these other companies that are also going to be growing in the area, do you already find you’re creating synergies with them? Are there folks who are starting to solve one of the problems you had and you’re going to team up with them to take care of them? Is that kind of thing already occurring?

Julien Camirand Lemyre: The short answer is yes, and the long answer is yes, and across the stack. There are a lot of different developers. Some people are only working on some software parts. It turns out that is a solution to what we are working with. When integrating and testing these solutions with these people, we are trying to bring our needs to these providers and say, “This is what we need to develop our technology and codevelop these products that are helpful for the whole industry and with end users.” The zone here is great to get traction, to get attention for quantum and to build synergy on how and when quantum will be accessible, and starting to do some of these pilot projects with these companies. Yes, we’re already seeing the benefits of the zone, but we’re also seeing that this is still increasing. Momentum is still rising. This is what we see in all ecosystems right now. There’s a lot of momentum — the industry is growing.

Konstantinos Karagiannis: Having been in this experience of being there, what advice do you have for those looking to get started in QIS? Every once in a while, people reach out to me. They want to know what to do, how to get started. Everyone has a different background; everyone’s come to it in a different way. Do you have any insights you’d like to share for those trying to get a start?

Julien Camirand Lemyre: Quantum is really changing. The industry is very different from when I started, as I mentioned earlier. At that time, the first D-Wave computers were out, and we were curious to see what these machines would do. Now, there are plenty of different modalities running with different advantages and disadvantages. What we’re seeing now in the industry is that there are a lot of needs from a broad set of expertise — experiences that are needed, from engineering to quantum science development to more business kinds of things as well. There are a lot of opportunities. The field is evolving, and it’s important to stay alert and understand that you can be part of this and you can be evolving alongside the field as well.

Konstantinos Karagiannis: Now, we’ve got a sense of the timeline and what’s coming for these machines. Do you see any real end user applications you’d also like to see — leaving the partner space to building materials, things like that? Do you envision, if you put your futurist hat on, anything folks might not be thinking of that they might be able to accomplish in trying to access a machine like this in the future?

Julien Camirand Lemyre: It’s hard to predict. But we’re excited, and other people are excited as well. We’ve been having a lot of interaction with Sumitomo Corporation, in Japan, around what can be done in the energy sector as well, and where to work with that. There are a lot of unknowns about these machines, and the more the technology progresses, the more we lower rates, the more qubits we have, and we’ll start seeing more and more traction and development based on these problems and the industry-relevant applications as well.

Konstantinos Karagiannis: Yeah, energy is often overlooked. People don't bring it up often enough.

Thanks a lot. I wanted to have you on about this exciting technology that just had to be mentioned here. It’s always good to see new attempts at revitalising these modalities. Thanks for sharing all your knowledge on this.

Julien Camirand Lemyre: Thanks for having me on the show. It was a great pleasure to discuss it with you.

Konstantinos Karagiannis: Now, it’s time for Coherence, the quantum executive summary, where I take a moment to highlight some of the business impacts we discussed today in case things got too nerdy at times. Let’s recap.

Typical quantum error correction can involve many physical qubits and some type of array to result in one logical, or fault-tolerant, qubit. The sometimes-large ratios of physical to logical qubits cause serious scaling issues, and that’s the reason you might have heard we could need millions of qubits one day. Nord Quantique is focused on building a better qubit of sorts: Error correction is handled in each qubit, allowing for easier scaling. It’s much easier to build 100 qubits and be able to use most of them rather than build 100,000.

Nord Quantique’s semiconductor qubit has microwave resonators or cavities that can store many more microwave photons. This setup allows autonomous correction of errors caused by phase and bit flips using both sonic codes, and yields 14% fewer errors on a qubit. Even the hexagon shape of the qubit module has scaling in mind. Imagine a beehive of them in the future. It’s still early days for the company’s approach, but Nord Quantique hopes to have a multiqubit system this year and a 100-logical-qubit system by 2028. Customers are already partnering to take a crack at materials-science use cases.

That does it for this episode. Thanks to Julien Camirand Lemyre for joining to discuss Nord Quantique, and thank you for listening. If you enjoyed the show, please subscribe to Protiviti’s The Post-Quantum World, and leave a review to help others find us. Be sure to follow me on all socials @KonstantHacker. You’ll find links there to what we’re doing in Quantum Computing Services at Protiviti. You can also DM me questions or suggestions for what you’d like to hear on the show. For more information on our quantum services, check out Protiviti.com, or follow Protiviti Tech on Twitter and LinkedIn. Until next time, be kind, and stay quantum-curious.