
The maximum error rate below which quantum error correction improves performance.The error threshold is a critical parameter in quantum error correction (QEC). It is the maximum physical error rate that a quantum error-correcting code can tolerate while still allowing the logical error rate to decrease as the code size increases. Below this threshold, adding more qubits and more error-correction cycles improves the reliability of the logical qubit; above it, the correction process introduces more errors than it fixes, making the logical qubit worse. The threshold is a key benchmark for building a fault-tolerant quantum computer.
The concept was formalized in the mid-1990s by Peter Shor and Andrew Steane, who showed that fault-tolerant computation is possible if error rates are below a threshold. The threshold theorem was further developed by Dorit Aharonov and Michael Ben-Or, and later refined for specific codes.
Quantum error correction works by encoding a logical qubit into many physical qubits and measuring syndromes (parity checks) to detect errors without disturbing the quantum state. If the physical error rate per gate or measurement is below a certain value, the probability of an uncorrectable error (e.g., a chain of errors that spans the code) decreases exponentially with the code distance. The threshold is determined by the specific code and the noise model; for the surface code, a leading candidate, the threshold is around 1% per gate for depolarizing noise. Below this, the code can be scaled up to achieve arbitrarily low logical error rates.
The error threshold guides experimental efforts in building quantum computers: it sets the target for gate fidelities and measurement accuracies that must be achieved before QEC becomes viable. Current superconducting and trapped-ion qubits are approaching or surpassing the thresholds for some codes.
The surface code's threshold of about 1% means that if every quantum gate has a 99% chance of working correctly, you can still build a reliable quantum computer — but if it drops to 98%, no amount of scaling will help.















