Gate Fidelity
Gate fidelity measures how close a real operation is to its ideal; it is the headline metric for qubit quality and the gate to error correction.
How good was the gate
Fidelity quantifies the overlap between the state (or process) a gate actually produced and the one it should have produced, on a scale from 0 to 1. A fidelity of 0.999 means the gate is right 99.9 percent of the time, or equivalently that it introduces an error with probability about 0.001 per application. Errors compound, so small per-gate errors still limit circuit depth.
Why the decimals matter
In a circuit with N gates, the chance that all succeed falls roughly as (fidelity)^N. At 99 percent fidelity, about 100 gates already accumulate a large error; at 99.99 percent, thousands are possible. This exponential sensitivity is why hardware teams chase each additional nine of fidelity so hard.
Single versus two-qubit
- Single-qubit gates: routinely above 99.9 percent across modalities
- Two-qubit gates: the bottleneck, typically 99 to 99.9 percent in the best devices
- Readout: often the noisiest step, from about 95 to over 99 percent
Measuring it
Fidelity is estimated by randomized benchmarking (which isolates gate error from state-preparation and measurement error), gate-set tomography (which reconstructs the full gate behavior), and cross-entropy methods. Reported single numbers hide structure: coherent errors, leakage, and crosstalk each degrade fidelity differently and call for different fixes.
The threshold
Fault-tolerant quantum computing requires physical error rates below a code-dependent threshold, often quoted near 1 percent for the surface code. Several modalities now operate below that threshold for their best gates, which is why the field has moved from proving single good gates to building the many-qubit systems where error correction can run.