Google Cuts Quantum 'Magic State' Errors 40-Fold on Its Willow Chip in a Science Paper

Some of the most expensive ingredients in a future quantum computer are not qubits but special quantum states called magic states. Google Quantum AI says it has found a cheaper way to make them. In a paper published in Science, the team reports a roughly 40-fold reduction in errors and logical T-states at 99.99 percent fidelity on its Willow processor, according to the company's blog post on 15 October.

Magic states are what let an error-corrected quantum machine perform the operations that take it beyond classical computing. Each one can be used only once, so they are scarce, and producing them to a high standard is resource-intensive.

What is magic state cultivation?

The older approach, distillation, was projected by some estimates to eat up as much as 95 percent of a quantum computer's capacity. Cultivation works more like a filter. It prepares a rough state in a small patch of qubits, runs checks, discards any attempt that shows a flaw, and grows the survivors, as The Quantum Insider described it. The team also showed it could move states from color codes into surface codes, a first step it called a technical challenge.

What did Google actually measure?

The headline numbers apply to state preparation, not to the processor as a whole. The main experiment reached an error rate of roughly one in 10,000 after about four hours of reinforcement-learning calibration. Willow has 105 superconducting qubits; the distance-5 experiment that grafted states into a larger code used 54 of them.