10th Users' Conference of IT4Innovations

Hardware drift and the accuracy of molecular properties from quantum computers

5 Nov 2026, 17:43

1m

atrium (IT4Innovations)

Poster Materials Science (e.g. Computational Chemistry, Materials Engineering, Nanomaterials) Poster Session

Speaker

Przemyslaw Grenda (Charles University)

Description

Quantum chemistry on quantum hardware requires estimating expectation values from repeated circuit measurement. Recent hardware results suggest this choice is less innocent than it looks: beyond a certain point, additional shots stop improving accuracy, plausibly due to hardware-induced drift of parameters.¹ Existing adaptive stopping strategies monitor the raw output distribution and assume the device is stationary during a run, an assumption that has not been tested on real hardware.²

Using a few small molecules as a benchmark, we measure how drift propagates into the electronic energy and into atomic forces. Both are obtained from the same measured reduced density matrices, but the former is a direct expectation value, whilst the latter requires further classical processing of those matrices, so the two are expected to respond differently to drift. Building on this, we develop a stopping criterion that terminates measurement once further shots no longer improve accuracy and evaluate it against fixed-budget and distribution-based alternatives on identical recorded shot streams. All measurements will come from the VLQ quantum computer based in Ostrava.

_{1. Olarte Hernandez, R. O.; Ziems, K. M.; Kjellgren, E.; Kongsted, J.; Coriani, S.; Sauer, S. P. A. Analytical Nuclear Gradients and Hessians on Quantum Hardware via Orbital-Optimized VQE with Error Mitigation. arXiv:2608.08758, 2026.}

_{2. Bisicchia, G.; Bocci, A.; Pimentel, E.; Brogi, A. StableShots: Online Shot Stopping for Quantum Circuit Execution. arXiv:2606.22170, 2026.}