An Energy Estimation Benchmark for Quantum Computers

TL;DR

This paper introduces a standardized energy estimation benchmark for quantum computers, applied to IBM's device, revealing unexpected fluctuations and limitations of current error mitigation techniques.

Contribution

It proposes a new energy estimation benchmark inspired by quantum chemistry and applies it to characterize real quantum hardware performance.

Findings

Benchmark results show poor correlation with reported gate and readout errors.

Two-hour oscillations in benchmark outcomes were observed over time.

Measurement error mitigation techniques did not resolve observed fluctuations.

Abstract

Certifying the performance of quantum computers requires standardized tests. We propose a simple energy estimation benchmark that is motivated from quantum chemistry. With this benchmark we statistically characterize the noisy outcome of the IBM Quantum System One in Ehningen, Germany. We find that the benchmark results hardly correlate with the gate errors and readout errors reported for the device. In a time-resolved analysis, we monitor the device over several hours and find two-hour oscillations of the benchmark results, as well as outliers, which we cannot explain from the reported device status. We then show that the implemented measurement error mitigation techniques cannot resolve these oscillations, which suggests that deviations from the theoretical benchmark outcome statistics do not stem solely from the measurement noise of the device.