arXiv:2609.24794
Observable-targeted variational quantum simulation of Hamiltonian dynamics
Leonardo Zambrano, Luciano Pereira, Antonio Acín
Abstract
Standard variational quantum simulation seeks to reproduce the evolution of the full quantum state, although many applications require only the expectation values of a few observables. We study a variational method for pure-state Hamiltonian dynamics that updates circuit parameters to reproduce the evolution of selected expectation values. An exact error identity guides the choice of observables, motivating a construction based on repeated commutators of the target with the Hamiltonian. For Pauli observables and Pauli-rotation circuits, the update can be estimated without ancillary qubits or controlled operations for overlap estimation. Across six- and ten-qubit spin, fermionic, and molecular benchmarks, the observable-targeted update extends the median time within the target-error tolerance compared with standard variational quantum simulation with equal shot budgets. These results show that directing the variational update toward the target observable can extend accurate simulation without increasing the measurement cost per time step.
AI Summary
Standard variational quantum simulation (VQS) approximates full-state dynamics even when only a few expectation values are needed. This paper studies updates that instead minimize a least-squares mismatch between selected observables’ variational and Hamiltonian time derivatives. An exact Duhamel error identity expresses the final target error as the accumulated residual paired with the backward-evolved target. This motivates observable dictionaries generated by nested commutators with the Hamiltonian. For Pauli dictionaries and single-parameter Pauli rotations, coefficients require only Pauli measurements on original and parameter-shifted circuits, without ancillas or controlled overlap operations. Classical simulations compare this update with regularized McLachlan VQS across six spin, Hubbard, and molecular benchmarks, using six- and ten-qubit, two-layer circuits. Both methods receive equal shot budgets per time step, with $10^6$–$10^9$ shots per velocity evaluation and 100 measurement realizations. Accuracy is measured by the first crossing of a target-error tolerance, generally $10^{-3}$. Six-qubit median reachable-time ratios are $1.06$–$4.22$ where both medians are observed; ten-qubit reported ratios reach $11.9$, with additional censored lower bounds. The results include sampling noise but not device noise; target expectations are evaluated exactly. Small dictionary residuals alone do not guarantee small target errors without sufficient target-orbit coverage. Dictionary growth can be costly: ten-qubit LiH requires 2,627 observables and roughly $5\times10^5$ measurement settings per update evaluation. Performance also depends on ansatz flexibility and regularization; one nonmonotonic shot-budget trend reflects error cancellation. Related work uses the same least-squares observable-matching formulation.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.