Last month quantum computing start-up Quantum Motion opened what it says is the UK’s largest independent quantum laboratory. The Islington lab, which represents a multi-million-pound investment for the University College London (UCL) spin-out, is home to specialist equipment for its scientists and engineers to use. This includes dilution refrigerators, which allow quantum technology to be developed at a temperature close to absolute zero, or -278 degrees Celsius, some 100 times colder than outer space.
“Islington is officially now the coolest part of London,” quipped James Palles-Dimmock, the company’s chief operating officer, at the time. “We’re working with technology that is colder than deep space and pushing the boundaries of our knowledge to turn quantum theory into reality.”
Keeping quantum chips cold is key to ensuring they work accurately and fulfil their promise to outperform classical computers for certain tasks. But as the technology matures and develops, doing this in a sustainable and scalable fashion may prove a challenge. With several types of qubit technology – the building blocks on which quantum computers operate – in development, the one which solves the cooling puzzle most effectively may gain a significant advantage in the race for commercialisation.
Why do quantum computers need to be so cold?
Qubits are the way information is represented in quantum form within a quantum computer. So where a classical computer, which runs on bits, represents data as a one or a zero, quantum data can simultaneously be a one and a zero. In theory, this means a quantum computer can process information much faster and more efficiently than a classical machine.
The technology remains at an early stage, and in November IBM announced it had developed what it claims is the most powerful processor yet, the 127-qubit Eagle. According to Big Blue’s quantum roadmap, it expects to reach “quantum advantage” – the point where quantum machines outperform traditional computers on certain tasks – within two years.
To achieve accurate processing, quantum computers need to operate at extremely low temperatures. This is because the heat generated by the surrounding equipment can interfere with the qubits, says Harrison Ball, quantum engineer at UK quantum computer developer Universal Quantum.
“When we talk about the temperature of a material, what we’re really referring to is the motion of the constituent particles, the atoms,” says Ball. “The colder the temperature, the less motion of those atoms, which means there are contributing less variation in their environment.
The obsession of quantum engineers and physicists over the last few years has been attempting to make the most pristine qubits possible.
Harrison Ball, Universal Quantum
“The obsession of quantum engineers and physicists over the last few years has been attempting to make the most pristine qubits possible, and the way in which you do that is try and produce an environment for the qubit where it interacts with absolutely nothing. That’s why, broadly speaking, colder is better.”
Universal Quantum is developing its quantum machine using trapped ions, or individually charged atoms, as its qubits. This is one of a number of methods for generating and controlling qubits which are in development, and John Morton, professor of nanoelectronics at UCL and co-founder of Quantum Motion, says each of them has its own reasons for needing to operate at a low temperature. Superconducting quantum computers have dominated early deployments.
“The superconducting qubit approach that Google and IBM are following needs low temperatures so they don’t accidentally create cubit errors,” Professor Morton says. “Ion traps use low temperatures because they need to create an incredibly good vacuum in which to operate. In the photonics approach, photons travel around quite happily at room temperature, but if you want to detect the types of photons that are being used you often need superconducting detectors, which work better at extremely low temperatures.”
Quantum computing’s carbon footprint: is it sustainable?
While the enormous carbon footprint of classical computing, particularly when it comes to the emissions of the rapidly increasing number of cloud data centres around the world, is well known, quantum computing promises a more sustainable alternative, despite the ultra-low temperatures that are required.
Professor Morton explains that the new Quantum Motion lab is housed in a standard commercial unit. “Our power requirement is not very different to that of a typical office,” he says.
While energy requirements will increase as quantum machines become more powerful, they are still likely to remain more efficient than their classical counterparts. “In general we anticipate workloads where we’ll have quantum advantage to be more efficient than the classical route,” says Jean-Francois Bobier, partner and director at Boston Consulting Group.
The key factor in this is speed. “Cooling down one of these fridges to a fraction of a degree above absolute zero takes about 10-15 kilowatts,” says Professor Morton. “But with that quantum chip, you can do things that would take vast computing resources to achieve. These machines are not designed to replace a desktop computer, which can use less than a kilowatt of energy a day. They are a replacement for something that consumes much more.”
Google demonstrated this in 2019 with Sycamore, its 53-qubit supercomputer, which it benchmarked against IBM’s Summit, which at the time was the world’s most powerful classical supercomputer. Sycamore was able to complete a random number problem in three minutes 20 seconds. Summit took two and a half days to solve the same problem. This increased speed meant the power consumed by Sycamore to achieve this milestone was orders of magnitude lower – 30 kilowatts compared to the 25 megawatts required by Summit.
Though the nascent quantum computing industry is focused on the ‘fidelity’ (meaning quality and reliability) of qubits, Bobier says this does not need to be at the expense of energy efficiency. “Given all the advantages of quantum computing, exact computation is the priority over energy efficiency – right now fidelity is the key bottleneck,” he says. “We might possibly find a new way to control qubits that is both exact and consumes a lot of energy, but right now we don’t see that, even with superconducting qubits which require dilution fridges. The ratio relative to the calculation speed-up should remain massively in favour of quantum computing.”
Quantum computing’s cooling puzzle
But quantum computing’s cooling requirements bring with them practical challenges.
IBM’s roadmap anticipates that it will release a 433 qubit quantum chip this year, with a 1,000 qubit version to follow. This number will need to grow exponentially to realise the full benefits of quantum computing, Professor Morton says.
“The 100 qubit chip IBM released recently is about 2.5cm square,” he says. “So if you ask yourself what that chip will look like if you have one million qubits, which is likely to be the amount you need to establish a fault-tolerant architecture, then you’re looking at chip which is 2.5m square. The kind of cooling technology required to go to that sort of size hasn’t been worked out, and certainly, if you’re working in superconducting qubits one of the things you’ll need to think about is how to scale the cooling system. It’s definitely one of the challenges.”
IBM’s solution to this is to build its own enormous fridge. The company is currently constructing what it says will be the world’s largest dilution refrigerator. Code-named Goldeneye, it will have a licence to chill a quantum computer of up to one million qubits, and measure some 3m tall by 1.8m wide. The project was announced in 2020 and construction is due to be completed next year. Once operational it will take between 5-14 days to reach the temperature required for a large quantum computer to operate.
Such a sizeable investment may not be practical for companies without IBM’s resources, but other techniques are being investigated. Quantum computing start-up IonQ, for example, is building quantum computers on the Ion Trap architecture, and cools its qubits by using a laser to cool the individual atoms which are required to be in a quantum state, a process known as laser doppler cooling.
Professor Morton says that whoever comes up with the best cooling solution could have a significant advantage as commercial applications for quantum computers start to emerge. “At the moment there are three or four different architectures which are being most actively investigated,” he says. “I think it’s certainly possible that the practicalities of cooling may well influence which qubit technology ends up winning.”
Matthew Gooding is news editor for Tech Monitor.