Microsoft’s first step to scalable quantum computing

There’s a race going on. Not the one to build artificial general intelligence, whatever that may be, but to deliver the first at-scale quantum computer that’s able to solve a class of massive problems that would take far too long using even the most powerful supercomputer.

Building a quantum computer requires devices that can work at temperatures close to absolute zero, where it’s possible to set, observe, and measure the quantum states of subatomic particles and use those measurements as part of a complex analog computer. More importantly, we need a way to ensure that there are no errors in those measurements and that the probabilities associated with them aren’t affected by how we measure them or by fluctuations in temperature and charge. Then we need tools to build and run the complex quantum circuits used to model problems, going from code to physics and back again, from our high-temperature world to one colder than interstellar space.

It’s enough to make rocket scientists think they have an easy job.

Azure and quantum computing

Microsoft has long been committed to building a quantum computer as part of Azure. It has set up a long-running research project as part of its Station Q initiative, designing and building a new type of quantum computer based on what it calls topological qubits that promise to be a lot more stable than current approaches. The Azure Quantum team has been developing a set of tools and languages designed to program these devices, compiling code into quantum circuits and providing simulators where anyone can experiment with writing quantum code.

Microsoft

Outside of its own work, Microsoft has been sponsoring fundamental research into quantum computing technologies and working with universities around the world. One of the key areas of research has been into Majorana fermions, a theoretical particle that’s eminently suitable for twisting into stable topological qubits.

Like much scientific research, the journey has not been straightforward or simple. The equipment needed to both make and detect Majorana particles is new and complex, as no one has done this work before. The path is perhaps best described as difficult, with possible breakthroughs turning into dead ends.

Majorana 1: A quantum processing unit

This week’s news is a lot more positive, with peer-reviewed papers and a design for a first-generation quantum processor, the Majorana 1. Designed to scale to a million qubits, the first iteration of what Microsoft is calling a topological core has eight qubits, which can help prove many of the requirements for quantum processing.

Microsoft has published a paper in the journal Nature that details how it constructed its first topological superconducting qubits and, more importantly, that these are made up of a pair of trapped Majorana particles with a quantum state that can be measured. The process of measuring its state is complex; the topological structure of a Majorana qubit makes it stable, but it also makes it harder to read that state. Part of the process of developing the Majorana 1 has been developing a new microwave-based technique for getting highly accurate information about a particle.

One advantage to this new technique is that it reduces the complexity of the circuitry around the qubits in a quantum processor. The same technique is used to send digital control signals, taking advantage of the stability of the Majorana qubits. The techniques used with other qubit types require fine-tuned microwave signals to make changes and require complex analog control circuitry, with a much higher risk of induced errors.

Microsoft

Inside Majorana 1’s qubits

The design of the quantum processor is fascinating and at the same time extremely complex. Building the structure needed to hold the topological qubit, a material Microsoft calls a topoconductor (as it’s a superconducting environment for a topological quantum node), requires near atom-level precision. Each layer on the indium arsenide substrate is sprayed on “atom by atom,” as Microsoft says. The slightest error means you won’t get the necessary properties to create a qubit.

Microsoft

And while the basic engineering is in place, the more complex the quantum processor, the harder it will be to deliver that level of accuracy. There’s a certain irony in the fact that to get more scale, you need a scaled-up quantum computer to design those quantum structures. It’s much like conventional processor design, with higher-density transistor layouts needing significant amounts of compute to design, build, and construct.

Having a repeatable technique for building these novel materials is key, as it ensures that Microsoft can turn experimental structures into computing hardware. Getting things right required a lot of simulation for the indium arsenide topoconductor structures and the essential aluminum superconductor.

Building qubits with topoconductors

The topoconductor differs from standard superconductors in the way it hosts qubits. Like standard superconductors, the topoconductor provides a zero-resistance path for Cooper pairs of electrons. Unlike standard superconductors, the topoconductor allows an unpaired electron to exist. This is how the topoconductor sets its quantum state and how that state is measured. With an unpaired electron, the topoconductor has a charge (and different superposition states will have different charges). At the same time, there’s no way to trap that electron, as it’s everywhere in the topoconductor at the same time.

The single qubit building block is called a tetron. It has two superconducting topological wires that host four Majorana Zero Modes (MZM) at each end. The wires are linked by a thinner superconductor. Together with the wire, these four MZMs control and store the state of the qubit. The topological wire is coupled to a pair of quantum dots, which have a different charge depending on the state of the qubit.

This charge is what the microwave detectors measure. The process is extremely accurate, with an extremely low probability of error. However, the extremely low probability of error still isn’t low enough, and there needs to be better error correction for Microsoft to deliver a working quantum computer. This is why the device road map goes up to multi-tetron devices that provide only a handful of qubits. Despite this, Majorana 1-based devices will need only one-tenth of the error-correction hardware of comparable quantum computers.

Those microwave-based measurements are the basis for programming and using the Majorana 1’s tetron-based qubits, as they allow us to connect and disconnect the quantum dots from the tetrons. Instead of requiring complex analog controls to change quantum states, this new approach allows Microsoft to program tetrons digitally, simplifying the connections between the quantum devices and the digital computers that link them to our programs and data. The more complex the tetron structure, the more states can be superimposed in a single wire, letting you use the topoconductors to deliver many different types of quantum information processing.

From a handful of qubits to millions

Microsoft is using this quantum processor as the first building block in a road map that’s intended to end with a scalable quantum computer. There’s a logic in the path to much more complex quantum processors able to support many qubits on a single device. Theoretically it’s a move from the experimental physics that led to the Majorana 1 to an engineering-led approach that first delivers a single qubit, then two, then an array of four-by-two tetrons acting as a two-qubit device with error correction, to much more complex devices with arrays of tetrons delivering multiple error-corrected qubits.

Although this is significant progress, it’s not yet a fully scalable quantum computer. That’s still some time away, as it needs the development of multi-qubit devices based on this first eight-qubit design. But it is a big step that promises a much less complex quantum computer than expected, allowing us to design applications that can help solve much bigger problems than simply finding the prime factors of large numbers. We’ll be able to use quantum computers to simulate complex protein folding or design generic catalysts for recycling plastics or capturing atmospheric carbon.

Having a quantum processor is a start. What’s needed now is a way to go from quantum-ready languages like Q# to code running on this new hardware. Azure Quantum’s simulators are helping developers write those first quantum programs while we wait for the essential quantum compiler that takes our code and builds it into quantum circuits running on the Majorana 1’s braided quantum topoconductors.

There’s a lot of ambition here but still a long way to go. A million-qubit design is already on the drawing boards in Majorana 1 devices that are the same size as a digital watch. Even so, building on Majorana 1’s new techniques and architecture, Microsoft is planning to deliver a full-scale quantum computer in years rather than decades.