Analog process fidelity

Analog process fidelity is a whole-device fidelity benchmark for analog quantum simulators: machines that evolve continuously under a native Hamiltonian and cannot run gate-based protocols like randomized benchmarking or cross-entropy benchmarking (XEB). Introduced by Mark, Choi, Shaw, Endres & Choi in 2022 (published as PRL 131, 110601), it lets ergodic quench dynamics play the role that random circuits play in XEB. “Analog process fidelity” is the label the QCMet metrics collection assigned it (metric M10.1); the source literature calls it fidelity estimation via ergodic quantum dynamics, so search under that name too.

How it works

The device quenches a simple initial state under its native, ergodic Hamiltonian, then samples roughly 1,000 bitstrings. The same evolution is simulated classically to get each sampled bitstring’s ideal probability. Under chaotic dynamics, outcome probabilities rescaled by their infinite-time averages follow Porter–Thomas statistics (the same universal fluctuations random circuits produce), so an XEB-style estimator applied to the rescaled probabilities yields the many-body fidelity F_d between the device’s state and the ideal one. Percent-level precision needs only about 10^3 samples, independent of system size.

Strengths and limitations

The protocol needs no gates, ancillas, or hardware changes (just native dynamics and bitstring readout), and it applies to a wide class of analog simulators, not only the neutral-atom platforms QCMet groups it with. The same data supports Hamiltonian learning and noise calibration, demonstrated on Rydberg arrays in Choi et al., Nature 613, 468 (2023). Its core assumption is also its main limit: dynamics must be ergodic enough to produce Porter–Thomas statistics, so the estimator can fail for weakly or non-ergodic evolution and for correlated non-local errors. It also requires classical simulation of the target dynamics, though Shaw et al. showed that extrapolating approximate simulations extends it beyond exactly simulable sizes.

Notable results

Shaw et al. benchmarked highly entangled states on a 60-atom Rydberg simulator (Nature 628, 71 (2024)). The February 2025 QCMet standardization collection adopted it as metric M10.1. For a holistic score built on analog simulators, see the Many-Body Quantum Score.