A new demonstration of error-corrected qubits shows substantial reductions in errors as logical qubits grow in size. Running complex quantum algorithms will require logical qubits, in which collections of hardware qubits are linked so that errors can be detected and corrected. Generally, dedicating more hardware qubits to each logical qubit improves the chances of identifying errors, but because hardware qubits are themselves error-prone, adding more of them also raises the probability that errors occur. If the hardware error rate is too high, growing the logical qubit only makes matters worse.
Several companies have now demonstrated hardware with error rates low enough that logical qubits sit beyond the breakeven point, where adding qubits improves performance. The problem is that the improvement has been modest. Reaching the error rates needed for complex algorithms could require an impractical number of hardware qubits per logical qubit. On Monday, Quantinuum announced that its current generation of hardware can scale well beyond the breakeven point, with each step up in logical qubit size improving some performance measures by a factor of over 10.
The results rely on error correction codes, whose qubits hold data and perform measurements that produce "syndrome data" indicating whether, where and how an error has occurred. Codes are judged by how many hardware qubits they need, how many logical qubits they host, and their distance, a rough measure of how many simultaneous errors a logical qubit can withstand. To test scaling, researchers measure the error rate at increasing distances, such as three, five and seven. If the hardware is past breakeven, the error rate drops with each step, and the size of that drop, the scaling factor, indicates how far beyond breakeven the system operates.
The scaling factor matters greatly in practice. A factor of 10 per step would require only a few increases in distance to cut errors substantially, while a factor near two would demand many more steps. Most prior demonstrations have shown drops closer to two, which could mean thousands of hardware qubits per logical qubit. Quantinuum builds trapped ion quantum computers, in which each qubit is stored in an ion.
Source: Ars Technica · Summarized by HeadlinesBriefing