In the near future, fault-tolerant quantum computers with millions of qubits will revolutionize science and industry. However, the challenge lies in ensuring that these building blocks operate with sufficient accuracy to avoid errors. Currently, the error rates of qubits are around 1 in 10,000 to 1 in 1,000. To tackle industrially relevant problems, we need to reduce this range from 1 in 109 to 1 in 106. The question arises, how can we accomplish this monumental task?
Despite the difficulty of squeezing such high levels of performance from Google’s current physical qubits, the Google Quantum AI team has developed a roadmap that has guided its research in recent years. The team’s approach gradually enhances the capabilities of the company toward the goal of achieving a fault-tolerant quantum computer.
Then came exciting news in the field of quantum computing! In a recent publication in Nature titled “Suppressing Quantum Errors by Scaling a Surface Code Logical Qubit,” Google shared its latest breakthrough in achieving the second milestone in its roadmap. The Google Quantum AI team has successfully demonstrated a prototype of a logical qubit, which represents the fundamental unit of an error-corrected quantum computer. Through these experimental results, the research team has shown that the performance of this prototype approaches the range required for scalable, fault-tolerant quantum computing.
From Physical Qubits to Logical Qubits
Quantum Error Correction (QEC) marks a significant departure from current quantum computing techniques, in which each physical qubit serves as a computational unit. Instead, QEC offers a path to achieve lower error rates by combining multiple good qubits into a single, superior logical qubit. By encoding information across multiple physical qubits, the resulting logical qubit becomes more robust and capable of performing large-scale quantum algorithms. Moreover, as more physical qubits are used to construct a logical qubit, its performance, and reliability can be further enhanced.
Despite the promise of QEC, there is a major hurdle to overcome: if the errors introduced by each additional physical qubit outweigh the benefits of error correction, QEC will not be effective. Until now, this has been a significant barrier to progress in the field of quantum computing, given the high error rates of physical qubits.
“It is really a full system problem and you will be limited by the weakest link. We need high-quality, coherent qubits (a lot of them), good connectivity, good yield, and demonstrate low-error operations — that is quite a task. After our first milestone, we came up with this roadmap to strategically figure out how to tackle them (issues). We communicate this roadmap more broadly so that others in the industry know how to align their development roadmap,” according to Dr. Julian Kelly, Director of Quantum Hardware at Google Quantum AI, adding that what makes their work harder is the fact that there are so many new technologies now that need to work together at the same time.
To address this issue, the Google Quantum AI team utilized a specific error-correcting code known as a surface code and demonstrated for the first time that increasing the size of the code can decrease the error rate of logical qubits. This achievement was made possible by carefully addressing a multitude of error sources scaled from 17 to 49 physical qubits. The findings are a testament to the potential of producing the logical qubits necessary to realize large-scale, error-corrected quantum computing, with sufficient care and attention to detail.
QEC with Surface Codes
An error-correcting code provides protection to information by duplicating it across several physical qubits. For example, in classical communication, a simple error-correcting code involves sending three bits instead of one, where if one bit is erroneously flipped, the intended message can still be understood by taking a majority vote of all received bits. The same principle applies to quantum computing, where the information is encoded across multiple physical qubits to create a single logical qubit that is more resilient to errors. By increasing the size of the code, the code becomes more tolerant of individual errors, thus enabling scalable fault-tolerant quantum computing.
A surface code is a practical quantum implementation that corrects errors caused by bit and phase flips. It arranges two types of qubits on a checkerboard: data qubits and measure qubits. The logical qubit is made up of data qubits while measure qubits are used for stabilizer measurements. Stabilizer measurements detect errors without revealing the value of the individual data qubits. By tiling two types of stabilizer measurements in a checkerboard pattern, the surface code protects the logical data from bit- and phase-flips. Correlations in the stabilizer measurements identify which error(s) occurred and where.
The main challenge in implementing QEC using the surface code is that the physical qubits used in the code are susceptible to errors, which increase as the number of qubits in the code increases. In order to achieve high protection from QEC, the benefits of the added qubits must outweigh the additional opportunities for errors. However, for this to occur, the physical qubits must have errors below the fault-tolerant threshold, which for the surface code is quite low. This threshold has been difficult to achieve experimentally until recently.
The Path Forward
The Google Quantum AI team is focused on achieving low errors in quantum computing for practical applications. Their target is a logical error rate of 1 in 106 or lower per cycle of QEC. They anticipate gradually increasing Λ, a parameter that measures the performance of their logical qubits, as they improve their physical qubits. A code distance of 17 and a value of Λ=4 would yield a logical error rate below their target. The team has an experimental technique to probe error rates this low with today’s hardware using both two-dimensional surface codes and one-dimensional repetition codes. While the repetition codes can only correct one type of error, they can achieve error rates near 1 in 107 by controlling for new kinds of error mechanisms that are not yet observable with surface codes.
“Our team has a single objective: to make quantum computing a useful technology. To achieve this goal, we strive to build a machine with a million physical qubits. But for those qubits to be useful, each qubit needs to be capable of participating in a large number of algorithmic steps. Quantum error correction technology is a necessary right of passage that maturing quantum computing technology has to go through,” said Dr. Hartmut Neven, Engineering VP at Google and the founder and manager of the Quantum Artificial Intelligence Lab.
The milestone of demonstrating a quantum computer outperforming a classical computer has been followed by three years of focused work by the entire Google Quantum AI team, culminating in the recent breakthrough. The team will continue to use the target error rates to measure their progress toward building fault-tolerant quantum computers. By achieving further improvements, they aim to enter the fault-tolerant regime where logical errors can be exponentially suppressed, enabling the first useful error-corrected quantum applications. Meanwhile, they continue to explore various applications of quantum computing in fields such as condensed matter physics, chemistry, machine learning, and materials science.