Google presents new quantum processor "Willow"

Contents

Five years after the release of its Sycamore chip, the Google Quantum AI team is now presenting its next-generation quantum processor: Willow is its name, the English word for "willow", and it is equipped with 105 superconducting qubits.

The team claims to have demonstrated two scientific achievements with Willow: Firstly, the qubits should make it possible for the first time to effectively correct computational errors that occur during calculations. This was announced by the researchers led by Hartmut Neven, founder and head of Google Quantum AI, in a paper published on Monday Publikation in the renowned journal "Nature". Secondly, the chip is said to have solved a task in five minutes that would take the best supercomputer 10 quadrillion (10²⁵) years. The team announced this in a press release.

Effective quantum error correction

Their first breakthrough, the effective correction of errors, was published by Google researchers led by Julian Kelly bereits in August on the Preprint-Server arXiv.org, but without disclosing the name of their new chip. After a successful peer review process, the results have now appeared in Fachmagazin Nature.

Due to the unstable nature of qubits, quantum computers are still very error-prone today. Calculation results are noisy if several calculation steps are carried out in succession – the results become unusable. Recognizing these errors and correcting them during the calculation is the aim of quantum error correction. Quantum error correction is a major challenge and is considered by many experts to be a necessary step in the development of useful, powerful quantum computers.

The Google researchers implemented a so-called surface code. Instead of storing the quantum information in a single qubit, it is distributed across several so-called physical qubits. Together, these form a logical qubit, i.e. a computing unit. This means that the necessary information is retained even if a single physical qubit makes an error.

Theoretically, the more physical qubits are used for error correction, the lower the error rate of the quantum computer should be. In practice, however, this was not previously possible because the physical qubits were too error-prone. Willow's qubits maintain superpositions five times longer than their predecessor Sycamore. For the first time, they are now stable enough to perform effective error correction. In combination, the error rate of the logical qubit is lower than that of the physical qubits. Errors are suppressed exponentially: if the number of physical qubits is doubled, the error rate of the logical qubit is halved.

With their result, the researchers are laying the foundation for error-tolerant quantum computing. "As the first system below the critical threshold, this is the most convincing prototype for a scalable logical qubit built to date," writes Neven. "It is a clear sign that useful, very large quantum computers can indeed be built."

Faster than the best supercomputer

The second breakthrough that the researchers are demonstrating with Willow is its superiority over classical computers. The "Frontier" supercomputer at Oak Ridge National Laboratory in the USA is said to take 10 quadrillion years to complete a task that Willow solved in less than five minutes. At the time of the test, it was the most powerful supercomputer in the world. In mid-November, "El Capitan" from the Lawrence Livermore National Laboratory knocked it out of first place.

The first version of Willow's predecessor, the Sycamore chip with 53 qubits, was the first to solve a task that exceeds the computing capacity of classic supercomputers, the Google researchers say. They called this ability "quantum supremacy". The chip was said to have solved a task in 200 seconds that would have taken a supercomputer 10,000 years. The news received a lot of media attention, but just a few days later the competitor IBM announced that a supercomputer could solve this task in just 2.5 days using the right methods. Although this would still make the quantum computer faster, the quantum computer is not essential for solving the problem.

Last August, the team renewed the claim with an updated version of the Sycamore chip with 67 superconducting qubits. The researchers estimated that it would take a supercomputer 10 trillion years to perform the same calculation as Sycamore. The researchers are now moving away from the term "quantum supremacy" and are only talking about calculations "beyond classical", i.e. beyond what is classically possible.

The task that the quantum computer – is supposed to solve, then as now –, is called "Random Circuit Sampling" (RCS): it is a quantum algorithm that generates a series of random numbers. The researchers execute a series of randomly selected quantum gates (i.e. quantum physical arithmetic operations) to bring the qubits into a complicated configuration that is difficult for classical computers to simulate.

"I am not aware of any practical use for random circuit sampling," says Sabine Wölk from the DLR Institute of Quantum Technologies. This is a recurring point of criticism of the Google researchers' benchmarking method. However, this is not the goal, explains Neven during a presentation of the new chip. "RCS is not useful for applications. It is a benchmark for comparing two quantum processors or a quantum and a classical processor." Only if a quantum computer performs beyond the classical regime in RCS can one expect advantages in useful applications. However, when asked whether Willow is already suitable for useful applications, the researchers responded evasively. Willow could possibly enable some applications in chemistry and physics, but we will have to wait and see.

What's next?

In addition to the RCS, the company has so far only carried out simulations that are relevant for scientific research, but are not beyond what is classically possible. "Our goal is to do both at the same time," writes Neven, "to move into the realm of algorithms that are beyond the reach of classical computers and that are useful for real, commercially relevant problems."

The company presents its longer-term plans in a roadmap. With Willow, it would have taken an important step towards the third of six milestones. The third milestone, a long-lived, logical qubit, is to be reached in one or two years. The team was unable to say when milestone 6, a large, error-corrected quantum computer with millions of qubits, could be expected.

In the long term, the company expects quantum computers to have a significant advantage in various fields of application. These include the development of new drugs, the design of new batteries for electric cars, advances in fusion research and the exploration of alternative energy sources.

(spa)