📝 SummarXiv

Abstract

The Path integral Quantum Control (PiQC) algorithm was recently introduced by Villanueva et al. (2025) as a new approach for computing optimal controls in open and closed quantum systems. Originally proposed for pulse-based quantum control, PiQC estimates optimal controls through global averages over quantum trajectories. In this work, we adapt the PiQC algorithm to optimize parametrized quantum circuits by showing that the quantum circuit can be randomized using a continuous dynamics governed by a stochastic Schr\"odinger equation that is compatible with the path integral control framework. In this adaptation, the circuit parameters become the controls to be optimized within PiQC. We refer to this instance of PiQC as the Gate-based PiQC (GB-PiQC) algorithm. We apply GB-PiQC for ground state preparation of electronic structure problems. We benchmark the gate-based and pulse-based versions of PiQC against the Variational Quantum Eigensolver (VQE), which is optimized using the common Simultaneous Perturbation Stochastic Approximation (SPSA) optimizer, on a set of standard molecular Hamiltonians: H2, LiH, BeH2, and H4, mapped to 2-, 4-, 6-, and 6-qubit systems, respectively. For each molecule, the benchmark is implemented at different bond distances, after performing a hyperparameter tuning of each algorithm at a fixed bond distance near the equilibrium geometry. We find that both PiQC algorithms exhibit greater robustness than SPSA to variations in the target Hamiltonian induced by changes in molecular bond distances. Furthermore, PiQC algorithms also achieve superior performance compared to SPSA in most instances, particularly at stretched bond lengths, where the Hartree-Fock solution becomes less accurate and its error grows relative to equilibrium.