Transcript | Topological Quantum Computing – with Anita Ramanan from Microsoft Azure Quantum

An 85-year-old idea may hold the key to nearly error-free qubits. The Majorana fermion has taken on almost mythical status, but Microsoft recently solved a critical technical hurdle to creating topological qubits with the particle. How close are we to quantum computers with this technology? Find out more about topological quantum computing 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 technical program manager lead for optimization at Azure Quantum, Anita Ramanan. Welcome to the show.

 Hi, Konstantinos. Thank you for inviting me. It’s great to be here.

I’m glad to have you here. Microsoft put out a paper on ArXiv this summer that’s causing a bit of a stir — at least to me — and, like most papers, it has a very short title — just kidding. It’s “Indium Arsenide Aluminum Hybrid Devices Passing the Topological Gap Protocol,” so it just rolls off the tongue.

We love the catchphrase.

Yes, but that topological part, that’s huge, and something we should be excited about, so I’m glad you’re here. But before we dive into all that, could you tell our listeners about what you do at Microsoft?

As you mentioned, I’m a technical program manager lead for optimization in the Azure Quantum team. Up until quite recently, I was a software engineer, so I used to focus on building optimization software projects to tackle some of our customers’ biggest optimization problems. I completed my degree in physics — atomic and particle physics and physical chemistry — and during my time in uni discovered computer gaming and started taking up World of Warcraft and writing game worlds and decided, “You know what? Maybe I can make some differences, some impact in the world, with tech, and see how that’s applied to healthcare,” at the time instead of going into kind of physics academia.

I wound up as a software engineer and overheard at a Christmas party, someone mentioned quantum computing, so my ears pricked, and then I inserted myself into the conversation. Long story short, I wound up interacting with the quantum computing team and eventually moving out here to work for these wonderful folks. I have been focusing on solving these optimization problems — from transportation and optimization for logistics through to healthcare applications like optimizing how long it takes to do an MRI scan, for example, so there’s an incredible variety of quantum-inspired optimization projects, as well as quantum computing projects, that being in this field enables.

Yes. We’ve done some of the same things. I just coauthored a paper on optimization, but that’s a topic for a different episode, I believe, so we’ll get back to this one. Let’s ease in. This is a hard topic, so we’ll just take it one piece at a time.

This all started way before anyone thought about quantum computing. It was 1937. Ettore Majorana — who was a fascinating fellow, if anyone wants to look him up — he came up with a fermionic wave equation named for him, and he predicted there should be fermion particles that are their own antiparticle, so it all sounds very sci-fi and interesting. How did these fermions become a Holy Grail for quantum computing?

I’m going to take this a little bit further — well, not back in time, but back in terms of technology. You’ve most likely heard of the various different content computing platforms that various different organizations, institutions, etc., are developing today. These are what are known as noisy intermediate scale quantum devices, or NISQ devices. These typically rely on encoding quantum information in the individual quantum states of quantum particles: anything from ions, for example — excited electronic states of ions — through to photons, electrons, things like that.

Now, this is incredible, because these particles are readily available. We can manipulate them relatively easily using technology that exists today, and we can start building these machines. And you must have seen in the news that they’re getting bigger, they’re getting more fault-tolerant, and we’re seeing some exciting new applications and ideas coming through.

At Microsoft, we decided that we would take a slightly different approach. We saw this paper from 1937, and Michael Freedman and a few of our wonderful scientists thought, “Is there a way that we can leverage topological properties to encode quantum information rather than using the individual quantum properties of these single electrons or ions, or what have you?” Now, the idea behind that is that by encoding information in a topological property, it should be more resilient to noise, but what does that mean? What does “the topological property” mean in the first place?

What you do is, you get these nanowires, and this is the indium arsenide aluminum hybrid, and you cool them down to nearly 0 Kelvin — as cold as you can get it, so we’re talking millikelvin. At these superlow temperatures, the electrons in this material start to behave weirdly — you can almost think of it like they’re lining up in a little chain of electrons, and you can do a little mathematical trick that puts a half-electron cap on each end of these chains.

Now, instead of encoding information in the states of these individual electrons, we braid them, these chains. You think of it like if you got a bunch of strings and braided them together or tied a knot in them and then you shook it around, that perturbation doesn’t affect the actual information that you’ve encoded in that braid pattern or that knot anywhere near as much as, say, if we’re encoding that in the state of an individual string. The idea behind a topological qubit is that by using this braiding, by encoding information in topological properties, we can leapfrog a lot of the problems that are the reason that we don’t have fully scaled quantum computers today — namely, decoherence. The quantum particles are incredibly susceptible to noise from the environment, and if we can do something to lower the amount of error correction that we need to deal with that, it can only be a good thing.

That’s where Microsoft is going. It’s new physics. It’s exciting. The March paper was an incredible milestone. What they achieved was, our research teams managed to create and control these Majorana zero modes in the lab for the first time, which is a huge milestone. It has taken 85 years from the initial conception of this idea to realize that in the lab, and it’s exciting. This is the fundamental building block that we can now take forward and start building these topological qubits out of.

Yes. That is one of the best explanations I’ve heard.

Thank you.

 It is terrific, and hopefully, your team are all still OK, because a year after he published that paper, he disappeared in weird circumstances, so hopefully, all that’s gone now, right? We’re all back to normalcy now.

Yes, we’re all good.

You’re all good, yes. Yes, seriously, the more you look into his life, the stranger it is.

I feel like that’s true for most of quantum.

Yes. We’ve achieved this topological gap, and it’s statistically noticeable because it’s 30 microelectronvolts, and noise is somewhere around 10, so there’s the proof. We’ve got that gap.

How do you anticipate this moving through development now? Now that we’ve achieved this stage, what would be next in trying to turn this into an actual usable qubit?

That’s the million-dollar question, isn’t it?

Yes, definitely.

We’ve got this. We’ve got the topological gap. It has been reviewed, or we’ve proven that it exists, which is a milestone in and of itself. The next thing that we need to do is, now that we have these Majorana zero modes, we need to turn them into qubits. Once we can build a qubit, we need to build another few qubits. And from those, then we’ll need to prove entanglement, and then we can start to get into building this scalable quantum machine.

We have a long road ahead of us. We’ve achieved this significant milestone, but that’s just the first step. This now unlocks us and enables us to start building these qubits. Of course, we’re not working in isolation here. This isn’t the new space race. We’re collaborating with others in the field. We’re focusing on building out a platform in Azure Quantum that enables users to access, of course, our own first-party hardware when it’s available, but also our partners’ hardware, so right now, you can access quantum hardware from a variety of different partners — for example, ionQ, Rigetti, Pasqal, and they’re up and coming — but the idea is to provide a platform that allows the user as much choice as possible and meets them where they are.

We’re focused not just on building this hardware platform, which is, of course, the key technology that underlies the whole thing, but also in building the full stack on top of that. That includes cryogenic control systems and software. It includes things like the Quantum Intermediate Representation. It includes Q# and the QDK and Azure Quantum as a wrapper around all of these services for our customers. We need innovation at all levels of that stack to make this a reality and to bring us into this fault-tolerant future.

 Are you currently working with all the hardware vendors that are accessed through Azure Quantum to do even more optimization down the stack, as you say?

Yes. We’re working with our partners to understand the limitations of current hardware, to understand how we can design and optimize algorithms to run on the minimum set of hardware possible. Of course, the idea is that we want to be able to identify these applications, but quantum computing is unfortunately not a silver bullet that will solve all of our problems in life. We need to identify those specific scenarios where we think that we can have a big impact.

For example, gaining the ability to accurately predict the outcome of chemical processes, or the manufacture of drugs or chemicals or materials, and scaling to high accuracy, or large problem sizes, in those domains is often inaccessible for a classical machine. We need these quantum machines to help us solve these problems, and that’s not something that anyone can do in isolation. And it has been incredible seeing progress being made and this competitive push toward building these scalable systems.

When you’re trying to consider the current landscape of machines that are out there now and where this future device would land, a lot of questions come to mind for me. For one, let’s talk briefly about error correction. Right now, we’ve got some pretty wild and wide-ranging estimates on how many qubits are needed to create a logical qubit. IBM has said it could be as bad as 1,000:1. IonQ has said it could be as good as 16:1. Did you have any ideas on what we might see with topological? Could it be just 1:1? Could it be that each one is a logical qubit, and is it that optimistic?

That is a brilliant question. I would love to be that optimistic. Honestly, it’s in very early stages. We’re figuring that out. I’m not privy to the specific numbers, but the idea behind the topological qubit is, you would need fewer. We’re hoping orders of magnitude, but it’s very early in this process.

Is there a set experiment that comes next that will explore the next steps and turning one of these into a qubit?

The research team has a road map that’s defined very specifically. We have a lot of next steps planned out, as you might imagine. Unfortunately, that is not something that we share publicly, but rest assured, there are plans in place, and we are pursuing this with everything we’ve got. This is something that we’re heavily invested in, and we are champing at the bit to get this out there in front of our customers.

It will make quite a splash, obviously. So there are no real guesses, then, as to a time frame for getting some gated qubits?

That’s something that we do have internally, but again, we’re not sharing publicly at this time.

If you were to think about the types of applications that you would like to see take advantage of universal gate-based, fault-tolerant quantum computing, what would be your top three for that approach?

Oh gosh, that’s a great question. Nitrogenase is one of the ones that I’m excited about. The production of ammonia as an industrial process is something that takes an incredible amount of energy. A measurable percentage of the global carbon emissions every year are caused by this process, and it’s incredible because it has enabled us as humanity to expand far further than we would have been able to. It has enabled the industrial agriculture movement. But at the same time, if we could make that process more efficient, require less energy, we can have a huge impact on things like reducing carbon emissions.

There is an enzyme, nitrogenase, which is naturally occurring, but the active site of this enzyme involves a lot of transition metals, and when you try to model transition metals in the way they interact, the problem is that their electrons are so close together that you can’t really ignore the quantum interactions that they have between them and between the molecules that they’re interacting with. We can’t use a classical machine to do much useful modeling of these things at all, so the idea is that when we are able to have a quantum computer running at scale, we’ll be able to much more effectively model those materials and hopefully come out with processes that will enable us to industrially produce this enzyme so that we can produce ammonia at room temperature and pressure, which would be groundbreaking.

We’ve also done a bunch of research at Microsoft into designing catalysts for carbon capture — some incredible chemistry work that a few people in our team did to design algorithms and optimize them to run on the minimum set of hardware that we can get this to run on. Of course, it’s still out of reach for the kind of computers we have today. But hopefully, in the nearish future, we should be able to see some gains from that — those kinds of applications, these problems that are intractable on classical machines, but they’re going to take probably millions of qubits, realistically, to tackle these fully industrial-scale problems, and that’s the future we’re working toward.

 I’m always amazed when I hear about the actual pure-science stuff that Microsoft is doing.

You don’t think about it.

Exactly. Most people don’t think about it. They’re, like, “Microsoft—they’re obviously making an application that’s going to be downloadable tomorrow, and that’s what they do,” and it’s not the case. I’ve seen demos of things like server farms that go underwater to run on the vibrations of the ocean.

Yes, Project Nautilus. That was great.

Yes — Microsoft is doing some cool things here.

Yes. We’re heavily invested in research. We’re invested in the hardware, the software, you name it. We’re going for the full stack here.

How about in Azure Quantum now? We’ve got these simpler use cases that you could play around with and start to tweak. How is it going with building reusable code? Is there a push being made to have these modules and algorithms that are ready to pull and use for newer developers out there who want to experiment?

This is part of our mission at Microsoft, in particular, when we’re coming toward looking at quantum computing — to bridge that gap between software developers, software engineers, and researchers in quantum information theory and quantum physicists who all want to use this tech, and there are various benefits that all of these groups have. And helping them all to work better together is pretty important for our acceleration in the future.

We’ve focused on developing things like Q# and the Quantum Development Kit. They include a lot of libraries to help with these kinds of things. They implement a lot of the basic functionalities that you would hope for as a software developer, and we’re also busy building out libraries to support this kind of development, so if you want to do the quantum Fourier transform, you don’t need to recode the QFT every time you want to do it. You can just call the QFT module and it does it for you. There’s an implementation there. Of course, if you want to implement that yourself, you can do that, but we’re trying to make this as easy as possible for developers to come in.

Things like our Azure Portal Notebooks experience is one example of how we’re trying to make life a bit easier for people who are new, who are trying this out, who just want to spin up an Azure Quantum workspace, run some experiments in a notebook, and just get started quickly and see what they can achieve with the hardware that’s available today. We’ve added credits, so if you sign up to Azure Quantum, you get $500 worth of hardware credits, which you can use against any of our hardware providers that are available. It gives you a chance to play around, to experiment and to see what’s possible with the technology as it stands.

 Do you envision people will be coming in and then just thinking, “I have this use case,” and they just will type it in and be presented with a couple of options for, let’s say, doing something like fraud detection or some kind of machine learning approach?

That would be incredible. We’re a little bit far away from having the plug-and-play approach, but I would love to see a scenario or a future where that is the reality. At the minute, what we’re more looking at is, we have a wonderful samples repository on GitHub. There are many different samples of varying complexity on there, and we are working to make sure that those are integrated smoothly into this Notebooks experience so you can take those samples, modify them, add them together, run them and use that as a starting plate for your own exploration.

Great. Before anyone wants to use their credits to actually run shots on, let’s say, an ionQ machine or something, they’re probably going to want to use simulation first just to get the code right. Do you offer any kind of noise simulation so that that way, they can get used to the way that the results will be? Because let’s face it, you don’t get idealized results on hardware yet.

 I wish. Yes, that would be amazing. Yes, we do. We offer a noisy simulator. I can’t remember if our third-party simulators also offer noisy ones, but we’re working on one for sure. Also, the enabling, I believe there are some Python packages that you can use to add noise as well, and we’re working to enable those as well.

Have you personally had a chance to use the very newest machines yet that are coming online, like Pasqal, for example?

I ran a sample on one of the prereleased ones, but I won’t say which, because I’m not sure I’m supposed to, but, yes, I have, and everything was running smoothly as far as my tests went. I love the inside track — we get a few weeks’ notice before this stuff hits the public, so we can experiment with it inside and internally, and it’s always exciting to see.

You never know what kind of like hiccups could be introduced — silly little things. It’s good to give them a little run first. You can’t say which one, but would you say that you find this newer machine to be bringing newer capabilities as far as better gate depth and things like that, deeper circuits?

I can’t comment specifically on that, but any time that we add a new option for hardware for our users is a good thing and increases diversity. Of course, each different system has its own nuances, its own strengths, its own weaknesses, and the ability, without changing your code, to instantly switch to run between one or the other, between simulation, between real hardware, is an incredible tool for both research as well as for software engineers.

That’s great. One last thing I’d like to know is, are you doing any work right now in making it smoother to have classical and quantum work together, like workloads that will intelligently pass off pieces to quantum machines, a hybrid approach?

Yes, so hybrid is the word of the day, I guess.

Yes, definitely.

This is something that we are absolutely invested in and are exploring actively. We are looking to make that easier for users, so keep your eyes open for announcements in the future, but that’s all I can say at the minute.

I was hoping there’s some hybrid thing coming that would be fun to play with, because that is the goal of having these platforms: You have your workload and just add quantum like you’d add a GPU or whatever else to your workload in the past.

Exactly. It’s this point solution. It’s not going to run your entire workload. That would be a waste of everyone’s money, just like coding an entire application on a GPU would be a complete waste of everyone’s time and money. But identifying those use cases where we can apply quantum computers to their best effect where the classical machine just can’t deal with it and then passing back over to the classical machine, because let’s be honest, we’ve been optimizing them for many years. They’re pretty good. You can run all sorts of awesome things on them, and we might as well make use of those strengths as well as the ones from the quantum computer.

Yes, because I don’t see classical computers going anywhere ever. They’re always going to work in tandem. Quantum will always be a specialized device, just like a GPU — they’ll always have things they’re better at. It’ll be worth constantly keeping an eye. Obviously, you’re one of the big three, so it’s not like we’re going to forget that Azure Quantum is there.

We’re not going anywhere. We’re invested in this, and we’re excited for the future.

I can’t wait to hear that topological qubits are running gates. That will be a very exciting day for me, so I’m looking forward to that.

Me too.

Thank you so much for your time. I appreciate this.

Thank you. Same — it has been a pleasure.

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.

In 1937, Ettore Majorana wrote a paper hypothesizing a fermion that could be its own antiparticle. Fast-forward to 2004, and Microsoft Station Q looked at that paper and decided that the Majorana fermion could provide a new way to accomplish quantum computing. A year ago, Microsoft’s research culminated in a paper describing a topological-gap protocol that would prove if a device met the criteria for this new approach.

It took 18 years, but this year, the team made a breakthrough. They cooled nanowires to just millikelvin above absolute zero. Electrons began to line up in chains. The team braided these chains and encoded information in the resulting topological structure. Quantum information encoded this way is less susceptible to interference or decoherence than information encoded in a single string of particle.

The paper just published this summer demonstrated that Majorana zero modes and the topological gap have been proven. The measured 30-microelectronvolt topological gap has tripled the noise level found in the experimental setup, and that quantifies the actual stability of the phase. Now, it’s time for future work to apply this achievement to building qubits, entangling them and then creating gates.

Microsoft continues expanding partnerships with quantum hardware vendors to provide access to multiple quantum computers via Azure Quantum. The team is also constantly improving the full stack used to interact with the machines. However, a major goal is still to build a topological system that the company hopes could have orders of magnitude fewer errors than disk error machines.

Microsoft is concerned with finding better practical uses for quantum computers. Anita is excited about some use cases you might not hear about too often. One is improving nitrogenase production to help reduce carbon emissions. A similar-goal project is helping solve carbon capture. For those who want to experiment with Azure Quantum on their own, there is a $500 credit available now for running real hardware. The site is constantly improving its access to libraries and reusable bits of code as well.

That does it for this episode. Thanks to Anita Ramanan for joining to discuss topological quantum computing and Microsoft Azure Quantum, 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 Twitter and Instagram @KonstantHacker. You’ll find links there to what we’re doing in Quantum Competing 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 ProtivitiTech on Twitter and LinkedIn.

Until next time, be kind, and stay quantum curious.