Quantum computers aren鈥檛 currently reliable enough for mainstream use, in part because the error rates of their calculations are too high. That could soon change, because for the first time, a quantum computer has demonstrated an error-correction strategy that fixes more errors than it creates, which may provide a practical way to scale up to a machine capable of carrying out genuinely useful computations.
Ordinary computers store data as either a 0 or 1, but errors can cause the bit to 鈥渇lip鈥� to the wrong value, which is why error-correction is a standard feature of modern processors. In quantum computing, the problem is more complex because each quantum bit, or qubit, exists in a mixed state of 0 and 1, and any attempt to measure them directly destroys the data.
Several research teams are working on the problem of quantum error correction but there is a long way still to go. Google announced in July that its Sycamore processor was able to detect and fix computational errors, but the additional hardware needed to do that introduced more errors than it was able to fix.
Advertisement
at the Joint Quantum Institute (JQI) in Maryland and his colleagues have now passed this crucial threshold. The team was able to set the state of a logical qubit 鈥� in this case a group of 13 qubits clustered together to more reliably hold a single piece of data 鈥� and then measure it again 99.4 per cent of the time, despite relying on six individual operations that have only 98.9 per cent reliability. Without error correction, the reliability would be expected to slip down to 93.6 per cent after all six operations.
Unlike the groups at Google and the University of Science and Technology of China (USTC), which have made big strides in recent months with superconducting qubits, the JQI use trapped-ion qubits. Their machine uses up to 32 individual charged atoms that are manipulated with lasers.
The inherently higher stability of trapped ions allowed the team to use a more efficient error-correction strategy called a Bacon-Shor code, which superconducting qubits aren鈥檛 currently high enough quality to use.
Monroe, who is also founder of quantum computing firm IonQ, which , says that error correction is the key to creating practical computers, not simply making more and more qubits. Anyone creating dozens of qubits while having a high error rate is 鈥渟pinning their wheels鈥�, he says, claiming that trapped-ion technology is on a steep upwards slope with only engineering hurdles ahead of it while superconducting qubits are on a flat trajectory with large scientific breakthroughs needed to progress.
Despite this, the only claims of quantum supremacy so far have both included superconducting qubits, and the number of qubits used in them has been rising steadily over the last year.
Monroe concedes, however, that his team was only able to demonstrate error-correction on a single logical qubit, and that the next challenge is to scale up to two or more. 鈥淲e need to think higher now,鈥� he says.
at Imperial College London agrees that the trapped-ion approach does have some advantages over the superconducting plan being followed by Google and USTC. Ions in a trapped-ion computer are physically identical, whereas superconducting qubits can vary, he says. 鈥淲ith superconducting qubits there鈥檚 a lot of surface noise. With each qubit you have to do a lot of tuning to make it as identical as you can to another, whereas nature gives you identical trapped ions.鈥�
Nature
