Microsoft has been putting money into the science of quantum computing for quite a few years now, funding some of the basic research that could let us produce computers based on the fact that electrons can be in many states at once (which is called ‘superposition’).
In the computers we use today, electrons flow through transistor ‘gates’ inside a processor that are either open or closed – one or zero in binary – but what we care about is the gate, not the electron. With a qubit (short for quantum bit), it’s the electrons themselves that store the information, and that’s one and zero at once, and everything in between.
Wire together 300 qubits in a quantum computer and they could store more information than there are atoms in the universe. Plus ‘entanglement’ between the qubits means one operation in a quantum computer does the same amount of actual computing as many normal operations, so programs run far more quickly.
But there are some big problems in the way of progress when it comes to making these tiny computing devices. Namely how to keep control of the physics of exotic matter well enough to make a reliable qubit that doesn’t lose the result of its computation before you can retrieve it, not to mention wiring more than a few of them together into a full system, and keeping it very cold while you run it.
Most people tackling quantum computing are looking at superconducting qubits, but Microsoft is taking a completely different approach.
“Imagine laying out thousands of tops in a gym and getting them all spinning at once, in a complicated configuration – with some of them going clockwise and some going counter-clockwise,” Microsoft research chief Peter Lee told TechRadar. “For superconducting qubits, we have the engineering knowledge to do that, over time. It’s a remarkable engineering achievement that the world knows how to do that. But it’s not stable; within tens of microseconds it falls apart.” And the more qubits you add, the worse the problem gets.
Instead, Microsoft is betting on topological qubits, which don’t use the properties of the electron – which can be changed by the slightest interaction with what’s around it, like the electrical field of any nearby electronics – but rather, the order in which some very exotic particles called Majorana fermions or anyons, change position.
“Under conditions of magnetic flux,” explains Lee, “they orbit in certain patterns, and if you imagine them dragging a thread behind them as they orbit around one another, they braid that thread into a pattern – and the pattern of the braiding encodes the quantum computing. There could be a lot of noise and wiggling in the orbit, it might be an ugly orbit – but it’s definitely orbiting in a clear pattern.”
Superconducting qubits still hold a lot of potential, says Lee, but topological qubits could leapfrog them. “Superconducting qubits are to quantum computing what vacuum tubes are to digital computers – but topological qubits are to quantum computing what transistors are to digital computing. When we were building computers out of vacuum tubes they were very useful – you could compute a better missile trajectory – but it was never going to be a scalable technology.”
From theory to engineering
Anyons were conceived of by an Italian physicist, Ettore Majorana, in the 1930s, then a physics professor called Alexei Kitaev “stumbled over them decades later and realised if they existed, you could use them to do quantum computing,” says Lee.
Another math prodigy, Michael Freedman, had been thinking about knot theory as a way of making a quantum computer; when he joined Microsoft Research in 1997 he teamed up with Kitaev. By 2004, Freedman was convinced that it would be possible to build topological qubits – as long as anyons were actually real.
He went to Craig Mundie, who was running Microsoft’s research strategy – Mundie didn’t just hire some more theoretical physicists, he funded six academic experimental physics labs around the world to run experiments to prove, one way or the other, if they exist.
A lab in the Netherlands found something with the right properties in 2011, then another in Holland in 2012. “Suddenly, there it was, and you could start dreaming about the possibility of building qubits with this topological protection,” enthuses Lee.
The fringe theory had become mainstream. In 2014, the mathematicians and theoretical physicists of Station Q – the quantum research lab Microsoft set up in Santa Barbara in 2006 – got some new colleagues. Microsoft’s Quantum Architectures and Computation group is run by Burton Smith – co-founder of Cray, the original supercomputing company – and Doug Carmean, the architect behind key Intel chips like the Pentium 4 and the Nehalem Core i5 and i7 (the first chips to use high-K metal gates), so they have plenty of experience working right at the limit of what you can do with silicon.
And now they’ve been joined by Todd Holmdahl. Microsoft’s new corporate vice president for quantum has been responsible for launching products like Xbox and Kinect, and he hired key Microsoft hardware people like Alex Kipman, Kudo Tsunoda and Steven Bathiche, so his new job is a sign that Microsoft thinks the possibility can become reality.
“I don’t think there’s one defining moment but we’re at this inflection point between science and engineering,” Holmdahl told TechRadar, giving three reasons why he thinks it’s time to move from research to engineering, even though many of the problems are far from solved.
“We have a really good line of line sight to controlling a topological qubit – we’ve been emphasising those over the last ten years and more because we believe they will have a longer coherence time, they will be immune to noise and they will allow us to scale much faster. We believe we’re getting very close to controlling that.”
“The second big thing is that we work with a number of growers that grow the materials for us and build the devices, and we’re finding that we’re able to do this much more quickly than in the past.” That means Microsoft can try out new ideas for making and connecting qubits more quickly, too. “As we go through the engineering process, it’s a huge advantage to be able to get through the prototyping phase in a rapid way.”
Microsoft is already able to grow nanowires for qubits ‘relatively quickly’ but it’s moving on to 2D electron gas processes which could mean that in the future, manufacturing qubits could be done at the same speed as making transistors. That means prototypes in a timeframe of weeks “and then several months for production time,” Holmdahl suggests. “Ultimately, I believe something like that will be the solution that allows us to scale the fastest in the future.” Again, that’s Microsoft taking a different direction from most of the other teams working on quantum computing.
“The third area is that the design and architecture that we are building looks to be very scalable and we have created a roadmap for the future of what our quantum computer might look like. Not only do we have an idea of what one qubit looks like, but what hundreds and thousands of them look like – how they’ll be laid out, how we’ll be able to control them, how we’ll be able to have them talk to each other.”
Quantum computing needs more than just qubits; it needs a whole computer, and a programming model for it. “The physics, the qubits, are just one part of that,” Holmdahl points out. “One of the beauties of working at Microsoft is that we have access to the other things that are very important – the classical computer we will need to control the qubits, we have access to the software that's going to run on that classical computer, we have access to the applications that are going to do the amazing things that are out there in quantum computing, these intractable problems. We're working on all of those.”
“We’re more than doubling our resources that we’re investing in this. We’re adding more research resources, we’re adding additional engineering resources to help them out. We’re adding a lot more people, we’re adding dollars for equipment and we’re adding focus.
“I've done this a couple of times, with Kinect and HoloLens, and I believe we’re putting in the right amount of resources that give us the best chance of getting something in the right time. I always bet big on technology that we will be able to figure out a lot of the issues that we need to figure out to make it widely available.”
So, what are those aforementioned intractable problems? “It would be pretty disappointing, after all this genius, pain and heartache has gone into quantum computing, if the only thing we can do with it is break codes,” notes Peter Lee. “But it’s very early days. When we first created computers with vacuum tubes and valves, the big thing we could conceive of doing was calculating better, more accurate missile trajectories.”
One big area of research is quantum chemistry. “We’re looking at how we can make fertiliser better, how to do activation of hydrogen and oxygen better, we’re looking at how we can help with the environment, what are the things we might do with higher temperature superconductors. All these things are possibilities, in addition to AI and machine learning things,” Holmdahl believes. “This technology has a real opportunity to make an impact on our lives in a profound way.”
Thanks to the quantum simulation software development kit Microsoft has come up with, Lee says that they already know that a problem like calculating the energy minimum of ferrodoxin – “if we can do that we have a basic building block for a next generation chemical catalyst,” he adds – would take about nine hours to run on a general purpose quantum computing machine (instead of billions of years on a conventional supercomputer).
And when might we get that quantum computer? Holmdahl has some dates in mind, though he won’t share them and he warns that quantum computing is still an unknown scientific frontier. He says that “some of this stuff is going to take time and there are a bunch of things that may be unsuspected as we go forward.”
Peter Lee says he keeps changing his mind on how long a quantum computer might take to arrive. “Some days I think it could be tomorrow, some days I think I won't live long enough to see it; it’s so unpredictable.”
Holmdahl agrees: “There are days which are like ‘wow, we’re right there on the threshold’, and there are days like ‘wow, it might be a long time away!’ But more and more, I feel like we are right there on the threshold.”
- Looking beyond silicon to discover the processors of the future