Speakers
Description
Variational quantum eigensolvers require ansatzes that are expressive, trainable, and implementable on constrained quantum hardware. Quantum architecture search can discover compact circuits, but the learned topology is obtained only after an additional search process. We present a deterministic Givens-exchange ansatz for molecular VQE that removes this outer architecture-search loop while retaining a chemically interpretable exchange prior. The circuit initializes the computational-basis state with the lowest diagonal Hamiltonian expectation and applies three local $R_Y$ layers interleaved with two ordered all-pair Givens-exchange blocks. Optimization uses only Hamiltonian expectation values; exact diagonalization is reserved for post-optimization errors and fidelities. Across six fixed seeds, every run on coefficient-verified LiH-6 and H$_2$O-8 Hamiltonians, and on a BeH$_2$-6 public-specification candidate, satisfies chemical accuracy, with mean errors of $0.000124$, $0.128558$, and $0.002152~mHa$, respectively. On LiH-6 and H$_2$O-8 these six-seed means are lower than the published point errors of the compared quantum-architecture-search methods. The method is not claimed to be a hardware-resource optimum: Givens macros and CNOT counts are reported at distinct levels, and hardware realization requires native-gate compilation, routing, finite-shot estimation, and noisy reoptimization. The contribution is an accurate, reproducible, and search-free reference architecture for molecular VQE and quantum-architecture-search benchmarking.
| Topic | Quantum variational circuits and their applications |
|---|---|
| Additional Topic | Hybrid quantum-classical approaches for learning and optimization |