Quantum simulations

Quantum technology promises to spark a revolution in computing as significant as teh integrated circuit in the 1950s and 60s. Evrim Yazgin helps us master the basics of quantum computing today, to imagine where quantum simulators will take us tomorrow.

Integrated circuits form the basis of modern ‘classical’ computing. There can be hundreds of these microchips in a laptop or personal computer. Their size has meant that now mobile phones have computing power thousands of times faster than the most powerful supercomputers built in the 1980s.

Since the 1990s, supercomputers have come into their own. The most powerful supercomputer in the world, Frontier based in the US, has a million times more computing power than top-tier gaming PCs.
But these devices are still based on the classical technology of integrated circuits and are therefore limited in their capabilities.

Quantum computers promise to be able to process calculations thousands, even millions of times faster than modern computers.

We’re not there yet, though.

Quantum does not compute

Quantum computers have been in development for decades.

These devices use the principles of quantum mechanics – which produce bizarre and seemingly magical effects – for machines which can do things that aren’t possible using modern classical computers.

Classical computers use bits – zeroes and ones – to encode information as binary signals in transistors in their integrated circuits.

Quantum computers use quantum bits – called qubits. These can encode information as zeroes, ones or a mix of zero and one thanks to the quantum phenomenon known as superposition.

Particles in a superposition of states aren’t defined by single values for their physical properties. Instead, these physical properties are expressed as probabilities.

Engineers can make use of superpositions to store multidimensional computing data in qubits of much greater complexity than an ordinary classical bit.

Extending the effects of superposition over multiple systems – or atoms – leads to quantum entanglement. This phenomenon, which Einstein famously described as “spooky action at a distance”, can be used to link qubits together leading to lip-licking prospects such as unhackable encryption.

But the same quantum mechanics which gives quantum computers their great potential means they are extremely difficult to actually produce.

Gavin Brennen, a professor at Macquarie University in Sydney and director at the Macquarie Centre for Quantum Engineering, helps explain why this is the case.

“The problem with anything quantum is, if we look at the world around us, it’s not very quantum,” Brennen says. “We don’t find ourselves walking through walls. We don’t find objects in superpositions. When things become larger and warmer, and they tend to act less quantum.”

“If you’re working with a single electron, that’s going to act very quantum,” he adds.

“Or maybe a single atom – which is a collection of electrons, protons and neutrons – that can act pretty quantum. But when you try to get thousands of those things to act quantum, it’s very hard.”

Brennen says this is due to an effect called “decoherence”.

“It’s noise. The more things you have to control, the more things they can interact with. And the things they interact with get information. Your quantum system leaks information and when it does that it loses the properties that make it quantum,” Brennen explains.

Brennen says “you have to play a lot of games” including cooling the system down and removing coupling between the quantum device and the rest of the world to try and reduce decoherence.

The physicist is speaking to me online from Helsinki, Finland where he is attending a conference called Quantum Resource Estimation at which researchers are discussing such problems.

“It’s pretty interesting,” Brennen says. “It’s about trying to find ways to make quantum computers more efficient for solving problems, like ways to make error correction work better and tricks to make algorithms faster.”
Quantum error correction is one way of tackling decoherence. The idea is to develop algorithms using more qubits to increase redundancies in the quantum system and reduce signal loss. It’s a bit like having more backups.

Eventually, such methods will lead to quantum devices being able to do things that today’s computers can’t.

Quantum see, quantum do

Among the people most excited to see quantum computers are quantum chemists.

“The UN has declared next year 2025 as the International Year of Quantum Science and Technology,” says Amir Karton, a professor at the University of New England in New South Wales.

“It essentially marks 100 years since the Schrödinger equation.”

Karton explains that this fundamental equation, developed by Erwin Schrödinger in 1925, describes the quantum mechanics of different systems. Solve the equation for a system and you can understand its properties.

“We’ve been able to be able to solve the Schrödinger equations for very small molecules, or very small systems with a small number of electrons in the last 100 years,” Karton explains. “For example, solving the equation for the hydrogen molecule was done in the 1920s.”

Systems with more electrons and other subatomic particles require solving more complex sets of equations. Karton says that solving anything with more than a handful of electrons wasn’t possible until supercomputers came around in the 1990s.

“What we’ve been able to do over the last 5 or 10 years is to model real chemical systems – molecules and materials,” Karton says. “That enables us to design better drugs, better catalysts, better materials for various applications without the need to go into the lab.”

Karton says, for example, that quantum chemists may need to test hundreds or thousands of catalysts to see which ones are most effective. Doing this in a lab is not feasible. Having a quantum machine to simulate this would speed the process up.

“We can calculate a catalytic enhancement of all these potential catalysts and have really good insights of what’s going to work. Then we would collaborate with experimental groups to then test the proposed catalyst.”

For quantum chemists like Karton, quantum computers able to solve the Schrödinger equation for even more complex molecules and materials would be a huge boon. The ultimate question is how can you build a device which is powerful enough to simulate the complex quantum mechanics of molecules?

Decoherence means that such useful machines are still a while away.

But there may be a type of quantum simulator which could give Karton and his ilk something to look forward to in the nearer term.

“The field divides it into 2 types of simulators,” says Brennen. “There’s digital and analogue.”

The digital quantum simulator – loosely, the quantum computers in development – attempts to use algorithms and gates (logical operations) to simulate the quantum mechanics of particles in complex systems.

On the other hand, Brennen explains, “analogue quantum simulators try to mimic the interactions of a system you’d like to understand by designing those interactions into a quantum device that you can control.”

“You tune up, fix the positions of some qubits, and make them interact. Turn on some fields and just let it go. There’s no sense of doing discrete sets of gates with error correction happening in between. You just try to get the thing to mimic what you’re trying to simulate as best as you can, and then let it go and do some measurements.”

Brennen likens this “analogue” quantum simulation to experiments and classical computing.

“If you want to simulate drag on an aircraft, you can do it in an analogue way where you stick a model in a wind tunnel. You see the effect from the small scale and argue that the properties scale up,” Brennen says.

The digital version involves “running some massive computational model of the effect of air pressure on the structure of an aircraft using complicated circuit-based simulation, probably using lots of GPUs.”

Analogue quantum simulators require fewer qubits than digital quantum computers and, therefore, should be easier to produce.

“I’m not a quantum physicist,” admits Karton, “but I think quantum simulators are going to be more successful in the near future.”

Simulators in practice

Dr Joris Keizer is a researcher at the University of New South Wales (UNSW) and the company Silicon Quantum Computing (SQC).

He’s kind enough to speak to me while having breakfast at a hotel in the Netherlands while he is on long-service leave.

“There’s a set of problems, mostly in quantum chemistry and material research, that really are quantum systems that we’re trying to understand,” Keizer says. “These are important things for drugs development and so on.”

Keizer reiterates that: “If you want to simulate a system like that with a universal quantum computer – one with gates – it will take a full-blown quantum computer to do that. So, the idea with quantum simulation is: why don’t we build a quantum system that we actually can engineer, that mimics the quantum properties of that system that you’re trying to simulate?

“It’s basically a quantuam system simulating a quantum system – but it’s a quantum system that we can engineer and that we have control over.”

It comes down to copying the mathematical representations of molecules onto other quantum systems.

“Let’s say you want to simulate a molecule,” Keizer says. “That molecule is described by a Hamiltonian. That’s getting quite technical, but a Hamiltonian just is a mathematical function that describes a quantum system. The whole molecule can be described with a Hamiltonian.

That function gets really complex if the molecule grows larger. It gets to the point where you can’t solve that function anymore. What you can do, though, is map that Hamiltonian to another quantum system that you engineer. You can use that to build your own quantum system that actually is exactly the same, mathematically at least, to the quantum system that you’re interested in.”

Sounds straightforward enough?

But many of the problems ailing “full-blown quantum computer” research also plague quantum simulators.

One problem is isolating the quantum simulator system, so it is not affected by the external environment. Another is using materials which are cheap to build and run.

In dealing with the second of these problems, why look any further than that mainstay of computing technologies: silicon.

Keizer was part of a team led by fellow UNSW researcher and SQC founder Professor Michelle Simmons which, in 2022, released a paper detailing the first coherent quantum simulator produced using silicon.

“Analogue quantum simulation is part of a product line at the moment, and we have demonstrated that we actually can do this stuff,” Keizer says. “We’ve published 2 articles. One is about simulating a very small molecule in 2021, and one is about simulating a much larger system.”

“My role in this has been developing the technique to basically engineer the quantum system that we use to simulate the other quantum system. And that is done by placing atoms, at the atomic scale, in silicon. My role in that is developing the technique to place these atoms where we want to.”

The molecule the team simulated in 2021 was a relatively simple 10-atom molecule, “but that’s about at the absolute limit what a classical computer can simulate”, Keizer says.

“If you extend this to 12 or 14 atoms – and there’s nothing stopping us from doing that – then you come into the realm where classical computers can’t give you the answer anymore. And that’s where it becomes interesting.”

Brennen adds that: “In fact, you could, you could talk about a lot of the devices that are existing now, which are just very prototype quantum computers, as kind of quantum simulators.”

“For example, Google is developing superconducting quantum chips for their quantum computers. But along the way, it turns out that they’ve simulated something called a topological phase. This is a special property of what you can get when you have a lattice of quantum spins, and their lattice of spins is a lattice of qubits.”

This indicates one drawback of quantum simulators. The systems they solve tend to be very specialised. A quantum computer, however, could be programmed to universally solve any number of interesting quantum problems.

Quantum checks and balances – and bets

How long until we see quantum simulators doing things that supercomputers can’t? It’s difficult to say.
“The problem is: how do we know that the thing it did is, is correct?” asks Brennen. “Because it might do something that’s too hard for a supercomputer to mimic, but the answer it gave might be wrong. That’s something that the field has been struggling with.”

Karton stresses that quantum simulators will build upon the advances of quantum mechanics and computing that have come in the decades before now.

“In this primary stage, the only way that these new technologies can be developed is through the availability of reliable data from quantum mechanics done on supercomputers,” Karton says.
“If we didn’t have this data to benchmark and assess the performance of these new technologies, there would be no way to develop these new technologies, because you wouldn’t know if they were doing the right thing.”

“We used experiments as our yardstick. We are, of course, talking about very accurate experiments, but we use those experiments to benchmark and improve. We saw in quantum mechanics which theories work and which of these theories are not good enough. That’s how we went through that developmental stage. That same process will be true for these new technologies.”

“I agree that quantum simulators are a more near-term application for these quantum devices, and people have learned some physics from running these analogue simulators,” adds Brennen. “A lot of what they’ve learned is the capabilities and also the deficiencies of some of their systems. Sometimes you don’t know what your system is good at until you actually try it out on a problem that you can test.

“People have looked at making analogue quantum simulators behave like a system they understand. If it does, great, your simulator passed that test. And if it doesn’t, then you learn something new about the physics of your simulator. That guides the new developments.”

How long will that process take? It’s still a guessing game.

“I suspect that we will see a simulator within I would say 5 years that’s doing something that we simply can’t track in a reasonable amount of time,” says Brennen. “And by reasonable, I mean, like, a lifetime on a supercomputer.

“Now whether it solves an interesting problem is another problem. It might not. It might be simulating a very contrived type of thing that’s just there to show we did something that you can’t do on a supercomputer.

“But for digital quantum computers solving interesting problems, that’s going to take longer, because they really are going to have to use error correction all along the way. I’m still optimistic. I think we’re going to see something useful within a decade, by the early 2030s.”

“I think the technique exists,” says Keizer. “It’s just a matter of pushing a little bit further, a little bit more investment. But this could potentially be something within a couple of years, where we could actually commercialise this and have clients actually accessing our systems” to run quantum simulations.

It feels like we have reached a critical stage in the development of quantum simulators which can explore quantum problems that, up until now, have been virtually impossible to solve on current technologies.
The researchers working on these projects are optimistic. And they understand the impact that such a development will make. We could be about to witness the opening of the floodgates.

“It could be one of the greatest achievements of humanity to get a quantum computer,” says Brennen.

And the floodgates will likely be burst open by a quantum simulator.