quantum error correction

Classical computers correct errors by keeping copies and comparing them. Quantum computers cannot do that. Reading a qubit to see what it holds destroys the superposition it was holding, and quantum states cannot simply be copied, so a different approach is needed.

The approach that works is to encode a smaller number of logical qubits onto a larger number of physical qubits, then repeatedly measure what are called parity checks. These measurements are carefully chosen: they reveal whether an error has occurred and roughly where, without ever revealing the information being protected. Software called a decoder turns that stream of clues into a correction, and the cycle repeats throughout the computation.

The price is hardware. In the Nature demonstration of a low-overhead fault-tolerant quantum memory, 288 physical qubits preserved 12 logical qubits, against nearly 3,000 physical qubits that the widely used surface code would have needed for the same performance. Bringing that overhead down, while keeping the correction faster than the errors, is the central problem the field is working on.