Phasecraft: the pioneer in quantum software

The Bristol and London company, which has recently opened an office in Washington DC, is a world leader in software for quantum processors. Its mission is to prove the power of quantum chips.

Forward: features are independent pieces written for Mewburn Ellis discussing and celebrating the best of innovation and exploration from the scientific and entrepreneurial worlds.

‘We are writing quantum algorithms for the imperfect quantum computers of today, to accelerate practical quantum advantage from decades away to years away,’ says Cat Mora, director of research operations at Phasecraft, one of the pioneers of the quantum revolution.

It’s an odd situation. Quantum chips are almost here. IBM, Quantum Motion, Infleqtion, D-Wave, and others, have prototypes with a few hundred to a few thousand qubits, able to run code. But they are unstable and ‘noisy’ in the industry parlance – prone to errors which require correction. The large-scale, high-performance fault-tolerant quantum chips that large swathes of the industry will require for many complex tasks are still three to five years away.

Which is why Phasecraft is so unusual. It’s one of the first quantum software firms: so early, the hardware isn’t ready yet. Founded in 2019 by academics from University College London and the University of Bristol, Phasecraft employs 30 researchers who are building the applications that aim to show the world what quantum chips can do, well before they reach the scale and size required for fault-tolerant computation.

The obvious question arises: how and why does Phasecraft create software if the chips aren’t ready?

‘They will exist soon!’ says Mora. ‘We are making sure useful stuff can be done on quantum computers as soon as possible. Quantum computers have been hypothetically thought about for 40 years. In the 1990s we had Shor’s algorithm and Grover’s algorithm, but then the focus switched to how to build the machines. There was an assumption we’d need billions of qubits to do anything useful. Our approach is to get our hands dirty and see what problems we can solve with as little quantum resource as possible.’

What problems is Phasecraft working on?

‘We work with a whole spectrum of companies,’ says Mora. ‘There are materials discovery companies, who want to improve industrial products; pharmaceutical companies researching molecules; and there are companies such as the National Grid working on optimisation problems. Classical computers don’t give good answers. So we try to build an algorithm which can solve them, and minimise the resources required so that it can run on a quantum computer in three or five years, without having to wait for the billion qubit fault-tolerant processor.’

At this juncture, a chipmaker might object. IBM launched the Condor quantum chip in December 2023, with 1,121 superconducting qubits – a big advance on its 433-qubit chip, and the 127-qubit chip in 2021. In January 2024, D-Wave unveiled the 1,200+ Advantage2. Xanadu has the Borealis chip. These exist and can run code.

Mora elaborates: ‘These quantum computers exist. The problem is the vulnerability to noise and their stability. In a classical computer error correction is built in. You can just run your program for as long as you need and not worry. In these quantum processors, you have a limited number of operations before things go pear-shaped.’

So the hardware is not able to run anything meaty. ‘We need thousands of operations to run our algorithms, and currently no chip can support what we need yet.’ Phasecraft’s work currently focuses on testing its algorithms on today’s current crop of quantum computers, including Google, IBM and Rigetti. The lessons Phasecraft gets from testing its algorithms are also shared with the hardware companies to help refine the technology.

No hype

For a truly groundbreaking firm, Phasecraft’s staff are reluctant to brag about their work.

Also present in our interview is Joel Klassen, lead quantum scientist at Phasecraft. He received his PhD from the University of Guelph and the Institute for Quantum Computing, and has an academic interest in fermionic encodings and fermionic quantum information, Hamiltonian complexity theory and the philosophical implications of quantum theory.

Klassen is admirably frank about the uncertain prospects of quantum computing.

‘It is not definitively clear whether the devices that exist today, or might exist in the future, can do something useful. There have been experiments demonstrating quantum supremacy [in which quantum computers outperform classical machines]. Both Quantinuum and Google have achieved this, so we know these devices can do things beyond classical devices. The challenge is doing something useful. The demonstrations are solving problems that aren’t useful to anyone.”

This is therefore the mission of Phasecraft: ‘We want to find useful applications that leverage near-term devices. It’s really important to emphasise that it’s not definitely known one way or another whether something useful can be done. But the huge cost reductions that we’ve seen from our novel algorithms give us confidence that we will get there!’

It is worth noting, there are differing opinions in the industry – including those who feel certain quantum chips are going to deliver social change on an epic scale. These include William Hurley, founder of quantum consultancy Strangeworks, who told Mewburn Ellis: “The quantum revolution is often compared to the industrial revolution, and for good reason.”

To those who believe Klassen is being pessimistic, he retorts: “I disagree that it’s pessimistic. It is actually an optimistic thing from the perspective of a scientist. Because it means there’s something to discover. If we knew nothing could be done, that would be pessimistic.”

On the brink of reality

Clearly the team at Phasecraft find this uncertainty immensely motivating. Led by co-founders Professor Toby Cubitt, CTO and Chief Science Officer, Professor Ashley Montanaro, CEO, and Professor John Morton – who also founded Quantum Motion, profiled by Mewburn Ellis here – Phasecraft is the only company in the world to have access to the globe’s most advanced quantum computers, built by the leaders of the industry such as IBM, Rigetti, QuERA, Google Quantum AI, and others, allowing Phasecraft to test and refine its algorithms on state-of-the-art quantum hardware.

The research output of Phasecraft is impressive. Recent work includes developing the world’s most complex quantum ground-state simulation of the Fermi-Hubbard model in materials science with Google. There are commercial partners too: Phasecraft is working with BT to develop quantum algorithms and software for solving optimisation challenges, such as how best to configure telecoms networks.

In materials discovery, Phasecraft partners with Oxford PV, which makes perovskite solar cells. In 2020, Oxford PV set a world record for power conversion of solar radiation at 29.5%. Phasecraft hopes to develop new materials to beat this record, using quantum algorithms to model how various molecules would perform.

Mora says: ‘The problem with researching solar cells in the real world is you need to build them in a lab and test them – and do that for each variation. It’s expensive. Doing it on a classical computer isn’t very effective. What we are trying to do is build a simulation which can run through hundreds of materials types and combinations and calculate their efficiency.’

All of which sounds pretty useful! Here, Klassen spells out what could happen if Phasecraft and the chip makers are able to deliver the sort of results that optimists like William Hurley predict.

‘Civilisations are defined by their materials. Traditionally, we’ve stumbled across things. When humans discovered bronze we had the Bronze Age. What we want to do is enter an age when we are able to find the best materials for all sorts of purposes. That has profound implications.’

Phasecraft publishes a fascinating Materials Database listing the elements and molecules it could, in theory, simulate. The database shows the current circuit depth and number of qubits needed for simulation, alongside the improvement offered by Phasecraft’s approach. For example, to simulate copper oxide the traditional method needs 1,350 qubits and a circuit depth (meaning the number of layers of two-qubit gates) of 2.907×10⁹. In layman’s terms, a vast amount of quantum hardware. Phasecraft’s superior algorithms can reduce that to 3,964 at a depth of only 71,709. So more qubits – not far off current hardware capabilities. Yes, the qubit number goes up a little with Phasecraft’s model, but this is more than compensated for by the dramatically lower gate count. ‘The qubit number is against us in this example,’ comments Klassen, ‘but if you tell me I can drop the gate count by orders of magnitude, and will only require twice as many qubits, I will happily take that bargain.’

For materials scientists, the ability to simulate interactions at the atomic level would be profound. And the hardware cited in the Phasecraft Materials Database is within reach in the short to medium term.

Quantum specialists will also enjoy the white papers and research published with impressive regularity by Phasecraft’s R&D team. As befits the quantum realm, these are not readily translatable into everyday prose, featuring deep dives into the ‘perturbation of an eigenstate of the system’, ‘Accelerating variational quantum Monte Carlo using the variational quantum eigensolver’ and technical insights into encoding for hardware.

There are accessible insights. Glenn Jones, principal quantum scientist at Phasecraft, for example, published this reflection on the contribution quantum computing could make to enhance Google DeepMind’s Graph Networks for Materials Exploration (GNoME) deep learning tool, which discovered 2.2 million new crystals via neural networks and other radical new methods.

‘Some of the algorithms developed by Phasecraft in recent months have been addressing this specific challenge and could provide improved data and predictions,’ says Jones. ‘These algorithms are still beyond the capabilities of existing quantum hardware, but with the current expected rates of hardware development, within a few years they will be able to enhance tools such as GNoME, enabling a further and potentially revolutionary advance in the discovery of new functional materials.’

The prospect of quantum hardware accelerating the work of Google DeepMind in this and parallel fields is mouthwatering.

Software tailored to each chip

Code written for quantum chips is so new, it is intriguing to hear how it’s done. ‘The languages we use are the ones you typically encounter in regular programming, such as Python,’ says Klassen. ‘The thing is you are not only using a language, you are using a conceptual framework. Designing quantum algorithms involves a lot of sophisticated mathematics and reasoning about what your program is doing.’

It is also the case that the best code must be tailored to each chip. Quantum processors come in an astonishing variety of flavours. Infleqtion’s Hilbert uses caesium atoms trapped in a laser grid in quantum states that have a large spatial ‘reach’, allowing entanglement between qubits across the grid. Quantum Motion focuses on traditional silicon chips cooled to -190°C using a distillation fridge. Silicon Quantum Computing in Australia is building chips with phosphorus atoms embedded in silicon – a novel approach.

‘I think it’s incredibly exciting that there are so many varieties,’ says Klassen. ‘And it’s promising. If there was only one design, that might limit us. We are in an exciting time where there’s a ton of options, and Phasecraft is in the privileged position of being able to access so many varieties of device. Each impacts the way you design algorithms.’

Superconducting qubits versus neutral trapped ions, for example, offer different clock rates. Superconducting qubits offer higher computational iteration, but at the price of higher noise. Trapped ions are slower, but the results are more robust. ‘The algorithm needs to factor that in,’ notes Klassen. Some approaches may split the task into chunks, each running on the chip best suited for the sub-task.

Klassen has advice for students hoping to enter the field: ‘In order to learn how to programme a quantum computer, I would recommend studying linear algebra. Followed closely by group theory and representation theory. Honestly the more maths you know, the better. Just keep learning maths.’

Our door is open

So can businesses simply ring up Phasecraft and start developing quantum solutions for their industrial problem? Mora says she would be delighted to receive enquiries. ‘We wouldn’t turn anyone away!’ Klassen suggests pharmaceutical companies, materials research firms and other tech companies are the best partners. ‘We have the feeling that if something isn’t likely to happen in the next five or six years, we are unlikely to entertain it. But we are working on optimisation problems with energy and telecoms companies, and with the Wellcome Leap Q4Bio [quantum for biology] programme, so we do interact quite widely.’

Mora adds: ‘It’s worth reaching out.’

It is clear Phasecraft is one of the most extraordinary enterprises on planet Earth. It is playing a major role in guiding quantum chips of unimaginable power from the realms of theory to a commercial reality. It is at the forefront of establishing what, if anything, quantum computers will be used for.

‘A common myth being debunked is that quantum computers can solve every problem because of some kind of massive parallelism,’ says Klassen. ‘That is simply not true. Quantum computers are machines that will solve very specific problems tailored to the power of quantum computers.’

But for a select number of challenges, quantum computers have the potential to deliver astonishing results. Klassen, so cautious up to now, spells it out: ‘When a quantum computer is able to deliver, it will yield exponential improvements in runtime for important problems. People always under-estimate the power of exponential improvements. It’s not just a bigger number. You are getting past a brick wall. It’s something really hard to wrap your head around.’

Transforming computing

Andrew Fearnside, Senior Associate and Patent Attorney at Mewburn Ellis, comments:

"Quantum computation has moved from a theoretical exercise to a practical one. Limited, small-scale quantum computers have become increasingly available and Phasecraft has risen to the practical challenges of usefully programming these emerging devices. Developing quantum algorithms that run on big, future quantum computers is essential to the development of this field of technology, of course, but it’s also so important to explore and develop practical, real-world advantages on the noisy intermediate-scale quantum (NISQ) devices of today. That’s why it’s so exciting to see the great commercial progress that’s being made by Phasecraft on this.”

Written by Charles Orton-Jones