Why it might be impossible to build a practical quantum computer

Physics

After “quantum supremacy”, the next step is scaling up and mastering the errors that dog qubits. But some researchers reckon the noise might always to be too high for useful quantum computers

By Michael Brooks

25 August 2021

IT IS 40 years since physicist Richard Feynman pointed out that quantum systems should be able to carry out an entirely new form of computation that outperforms even the most powerful conventional computers. “Feynman argued that quantum computing should offer an exponential speed-up for many classical computations,” says at the University of Auckland in New Zealand. And with a slew of breakthroughs, quantum computers look like they might now be hitting the big time. Perhaps.

Because they have properties that just don’t exist in the classical world, quantum entities such as atoms, photons, electrons and the like have access to a different set of routines for information processing if used to make quantum bits, or qubits – a potentially much more powerful set.

Part of that is down to quantum superposition, which means a qubit can be used to represent a complex combination of the 0 and 1 binary states used in normal computing. That doesn’t mean it is 0 and 1 at the same time. A better way to put it is that might turn out to be 0 or 1.

Quantum algorithms use a process called “interference” to skew these undefined properties and bias the interactions of multiple qubits in a way that increases the likelihood they will arrive at a final state that contains a solution to the problem they are trying to solve.

That’s where entanglement comes into the mix. The spooky connections between qubits it generates somehow allow for a pattern of interference where the paths leading to each wrong answer destroy one another and cancel out, while the paths leading to the right answer are reinforced.

The power has long been proven. In 2019, Google’s quantum computing team announced it had achieved “quantum supremacy” – when a quantum processor can do things that a classical computer can’t. Its 54-qubit Sycamore processor took just 3 minutes and 20 seconds to solve a problem that would take 10,000 years to crack on the world’s most powerful classical computer, the researchers said.

Which isn’t to say that Google’s quantum computer, or any that has reached quantum supremacy since, is close to doing anything useful. The problem Google cracked was highly esoteric. In May, Isaac Chuang at Massachusetts Institute of Technology, one of the world’s leading authorities on quantum computing, spelled out the current state of the technology in stark terms: “Quantum computing today is actually, from a practical standpoint, quite useless, other than for generating publicity.”

Trial and error

That brings us to the long journey ahead to a practical machine. The inconvenient truth is that, in quantum computing, size matters. Data-holding qubits must maintain their delicate quantum states for a long time, and not succumb to environmental influences such as heat and vibration that can cause them to decohere, creating errors in the computation.

This is a problem that can only be overcome by scaling up. Current estimates suggest that in large, programmable quantum computers, most qubits – perhaps as many as 5 in 6 – will be doing error correction, not computation. That means we are going to need as many as a million qubits before we can do anything truly useful. Keeping so many qubits sufficiently cold or maintaining all their quantum states long enough to do a computation is a monumental engineering challenge.

It could take decades to get there, but the big players are at least making steps in the right direction. IBM is aiming to build a 1121-qubit machine by 2023, and the company has envisaged a colossal helium-cooled refrigerator to contain it. Others, including Winfried Hensinger at the University of Sussex, UK, want to avoid the complications involved with cooling: they are scaling up operations with trapped ion qubits that shuttle around a large circuit to perform computations. Still others are performing computations by sending photon qubits around a silicon nitride chip that can be manufactured at scale using processes already proven in the semiconductor industry.

So do we have a “yes”? Not so fast. Gil Kalai, a mathematician at Israel’s Hebrew University of Jerusalem, has argued that the base level of noise in a quantum computer will always be too high, no matter how many qubits are available. “My analysis asserts that quality error correction won’t be possible,” he says.

Sabrina Maniscalco at the University of Helsinki in Finland is similarly sceptical. “Finding a remedy to the effect of noise induced by the environment is not just, in my opinion, a technological issue, but more of a conceptual and foundational one,” she says. “I would say that I am hopeful, rather than confident.”