Fig. 1: Quantifying measurement-induced crosstalk in a trapped-ion quantum computer using QIRB.

Quantum instrument randomized benchmarking (QIRB) uses the mid-circuit-measurement-containing (MCM-containing) random circuits shown in (a, b) to measure the average error rate (r_Ω) of MCM-containing circuit layers. QIRB retains the simplicity and elegance of RB methods for measuring gate errors, as it estimates r_Ω as the decay rate of the average circuit success rates (\({\overline{F}}_{d}\)) versus circuit depth, as shown in the examples of (c). We used QIRB to study MCMs in Quantinuum’s H1-1 system, depicted in (g), which arranges 20 ¹⁷¹Yb⁺ ions into auxiliary ("A'') and gate ("G'') zones. Our initial experiments on H1-1 applied micromotion hiding³⁵ only to unmeasured ions in gate zones during MCMs, as ions in auxiliary zones are distant from measured ions. We observed higher error rates in six-qubit QIRB than predicted [compare left with center bars in (d, e)]. We added micromotion hiding for all unmeasured ions, reran QIRB, and observed that the six-qubit error rate was approximately halved [right bars in (d, e)] and was now consistent with the predictions of Quantinuum’s emulator. Results from bright-state depumping experiments (f) independently confirmed the existence of long-range, MCM-induced crosstalk and its mitigation by extra micromotion hiding. These experiments quantify MCM-induced crosstalk errors by measuring the rate at which unmeasured ions, prepared in the \(\left\vert 1\right\rangle\) state, leak out of the computational space as gate-zone ions are measured. All error bars represent bootstrapped 1σ deviations from the sample mean.