Trial-space co-design for quantum simulation on noisy hardware

Abstract

Quantum simulations on noisy hardware are often designed as if the machine were ideal, with device limitations considered only after the basis and measurement plan have been chosen. We instead use information about the physical problem and the device at an earlier stage, to decide which states and evolution times should form a compact simulation basis. Because the target problem remains unchanged, the method adapts only the representation used to solve it. Tests on many-body and molecular models show that this approach can reveal useful directions that ordinary time evolution misses, while also exposing an important limitation: some of these directions can become too similar to estimate reliably. Realistic device data further show that hardware response can change the preferred states and evolution times before execution. Across seven acquisition days, a four-qubit experiment shows a median 54\% reduction in projected-energy error from problem-informed time choices relative to uniform sampling under matched hardware conditions. These results provide a proof of principle for simulation-basis design as a practical interface between quantum algorithms and the hardware that executes them.

Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.