‘Magic’ states empower error-resistant quantum computing

Quantum computers have performed the full slate of tricks needed for robust computation. And it all came down to a little magic.

In a pair of papers submitted June 17 at arXiv.org, researchers generated conditions called “magic states,” crucial components of quantum computations. And those magic states were high-quality enough to allow the computers to resist errors, one of the biggest bugaboos of quantum computers.

Quantum computing has the potential to allow calculations that are not possible with classical computers, by taking advantage of the physics that governs tiny scales. The computers are based on quantum bits, or qubits, that are analogous to the bits in a conventional computer. But qubits are finicky, meaning that errors accumulate during calculations, threatening to hold the computers back. So scientists are developing computers that correct errors as they happen, known as fault-tolerant quantum computers.

In the new studies, the researchers demonstrated a full set of error-resistant operations — a prerequisite for a fully functional quantum computer. These operations manipulate qubits during calculations, such as flipping their values or linking them through the quantum phenomenon of entanglement. While some operations can be implemented directly, others are more difficult. They require a workaround that involves generating special configurations known as magic states. Performing those more difficult operations in a fully fault-tolerant way — outperforming operations done without error correction — had eluded researchers until now.

“They’re demonstrating basically the final missing piece in the full fault-tolerant and scalable quantum computing architecture,” says physicist Boris Blinov of the University of Washington in Seattle.

The researchers used two quantum computers made by the company Quantinuum. The machines use electrically charged atoms, or ions, as qubits.

Here’s how the magic happens. Before a calculation begins, qubits are put into special quantum states, that is, the magic states. When one of the easy class of operations is performed on a magic state, it’s equivalent to one of the operations that can’t be performed directly.

Magic states are crucial for quantum computers to gain the upper hand over classical ones. In some sense, magic states are “the keystone that give quantum computers their power,” says quantum physicist Dave Hayes of Quantinuum in Broomfield, Colo., a coauthor of both studies.

Normally, the production of magic states is a shaky affair, with only some qubits actually ending up in the desired states. The two studies took different methods to resolve this issue. In one study, the researchers produced the magic states and then used a technique for detecting errors to check how successful the magic state generation was, throwing out the bad attempts to keep only the good. “If we did do it properly, we end up with a better magic state,” says mathematician Shival Dasu of Quantinuum, a coauthor of that study. “And if we didn’t do it properly, we retry until we do it correctly.”

In the other study, the researchers switched back and forth between two different error-correction techniques. Different techniques allow different operations to be performed easily. The researchers switched from a technique suited to preparing magic states and another which was apt for performing the operations.

The methods resulted in magic states that were high-quality enough to perform error-resistant operations. And the methods are efficient: The number of qubits required to create the magic states is smaller than many previous methods. Dasu’s method, for example, used eight qubits to create magic states, and in the future, the team anticipates using 40 qubits to make even better magic states. Initial ideas for making robust magic states suggested hundreds of thousands of qubits might be required, Dasu says.

That efficiency is important for scientists’ goal of building large-scale quantum computers, which could tackle problems that are otherwise unsolvable. Fewer qubits will need to be tied up in magic states. “The number of qubits that you would eventually need to unlock these applications comes down quite significantly,” says physicist Sebastian Weidt, CEO of the quantum computing company Universal Quantum in Haywards Heath, England. “The way they’re doing it is new, and no one’s done it with this quality before.”