Quantum computing progress depends on addressing noise challenges that hinder quantum advantage. Accurate assessment of gate performance is critical, particularly for large-scale operations involving multiple qubits. While individual gate fidelities provide useful metrics, they fail to capture vital information about correlations between gates operating simultaneously.
This research introduces an innovative approach to quantum gate benchmarking using the Character-Average Benchmarking (CAB) protocol. This method’s key advantage is its ability to evaluate fidelity in shallow-depth circuits, making it effective even with higher gate error rates. This characteristic allows for benchmarking substantially larger quantum gates than previously possible.
The experiments successfully benchmarked quantum gates at an unprecedented scale—up to 52 qubits. For a 44-qubit parallel CZ gate, researchers achieved a fidelity of 63.09% ± 0.23%, representing significant progress in large-scale quantum operation performance.
Two important gate types were examined: fully connected gates (with two layers of CZ gates and two layers of local Clifford gates) and parallel CZ gates. The fully connected gates, essential for variational quantum algorithms and quantum simulations, were benchmarked up to 46 qubits with a fidelity of 17.42% ± 0.45%. The parallel CZ gates, crucial for generating multi-party entanglement and preparing graph states, maintained a relatively stable average fidelity of approximately 98% per local CZ gate regardless of total qubit count.
Beyond measuring fidelity, the research introduced metrics to quantify correlations between local gates, providing insights into crosstalk errors without requiring additional experiments. These correlation measurements revealed consistently positive but weak interactions in the 44-qubit configuration, while the 52-qubit system exhibited some negative correlations.
The experimental results aligned with a composite noise model incorporating both local depolarizing noise and ZZ-coupling effects. This model demonstrated that when three gates interact, improving one gate’s fidelity can negatively impact others—an important insight for quantum circuit optimization.
By optimizing global fidelity rather than individual gate fidelities, researchers enhanced a 6-qubit parallel CZ gate from 87.65% to 92.04% fidelity while simultaneously reducing gate correlation from 3.53% to 3.22%. This confirms that global gate optimization strategies yield superior results by accounting for the complex interactions between simultaneously operating quantum gates.
npj Quantum Information, Published online: 26 February 2025; doi:10.1038/s41534-025-00983-5
