Abstract:
The Google Quantum AI research team demonstrated "magic state cultivation" for the first time on the Willow superconducting quantum processor, and reported that the error rate of the cultivated quantum state was about 40 times lower than that of the uncultivated state, and the logic fidelity reached 99.99%. The research, published in the journal Science, shows that the solution, which had previously remained largely theoretical, can now be run on real quantum hardware.
The magic state is a special quantum state required to implement universal quantum computing and can be used to perform non-Clifford operations called T-gates. Existing error correction architectures usually consume a large number of qubits and computing resources to prepare high-quality magic states; some estimates suggest that related preparations may occupy up to about 95% of the capabilities of a quantum computer. The cultivation method first injects noisy quantum states into a small area in the chip, and then checks for errors and filters out unqualified results through a series of coordinated measurements to obtain purer logical states.
The experiments in the paper reduced the probability of logical errors from about 0.44% without cultivation to about 0.01% to 0.02%, which is consistent with the statement of "reduced by about 40 times"; however, the experimental results retained after screening accounted for about 8%, indicating that this error improvement is still accompanied by a low success rate, and cannot be simply understood to mean that the resources required for the entire quantum computer have been reduced by 40 times. The team also demonstrated the conversion and transplantation of the color code logical state with a distance of 3 to a surface code with a distance of 5, verifying that the cultivated state can be transferred between different error correction codes. Willow's entire chip has 105 physical qubits, but these results are demonstrations of specific experimental circuits and are not large-scale, practical fault-tolerant programs that have been run on the chip.
The researchers also found in experiments that device temperature fluctuations and electronic device drift can interfere with traditional measurement benchmarks, so they improved hardware control and designed new benchmark testing methods. Google stated that it still needs to verify the cultivation of magic state on a larger scale and solve multiple engineering challenges during the expansion process. The results at this stage prove that the technical route can be implemented on real hardware, providing new progress in lowering the resource threshold of fault-tolerant quantum computing, but there is still a distance from practical large-scale quantum computers.
Comments