arXiv:2610.21782
Quantum-compatible Riemannian conjugate gradient for ground-state preparation
Andres Arcia-Lopez, Marek Gluza, Yudai Suzuki, Cyrus Mostajeran
Abstract
Riemannian optimization provides a natural geometric framework for designing quantum algorithms for ground-state preparation, most notably through quantum imaginary-time evolution. However, existing geometry-aware approaches primarily employ steepest-descent updates, leaving more advanced methods from classical Riemannian optimization largely unexplored. In this work, we introduce quantum conjugate gradient descent (QCGD), a quantum-compatible implementation of Riemannian conjugate gradient that combines the steepest descent direction with information retained from previous search directions. In our numerical experiments, the proposed approach achieves faster convergence per iteration than the baseline double-bracket quantum imaginary-time evolution (DB-QITE) algorithm, and we quantify the associated circuit cost with gate counts produced by Qrisp. This improvement, however, comes at the cost of increased circuit depth for each update step, revealing a trade-off between convergence rate and implementation complexity. Nevertheless, since the circuit depth for those algorithms grows exponentially with the number of iterations, the proposed method may offer practical advantages, particularly in the early fault-tolerant regime.
AI Summary
Can memory-accelerated Riemannian optimization reduce the circuit cost of quantum ground-state preparation? This paper introduces quantum conjugate gradient descent (QCGD), extending double-bracket quantum imaginary-time evolution (DB-QITE) beyond steepest descent. It combines the current energy gradient with previous search directions, then constructs quantum-compatible updates using group-commutator and Trotter approximations, Hamiltonian evolutions, and reflections about previously prepared states. Including zero memory weight recovers DB-QITE at the current state, ensuring an optimized candidate is no worse than that fallback—not guaranteeing dominance over an independently evolving DB-QITE trajectory. State-vector simulations benchmark three updates on open antiferromagnetic Heisenberg chains of 10–20 qubits, starting from singlet products or a one-layer Hamiltonian-variational warm start. QCGD achieves higher final fidelity in every tested case. At 20 qubits, fidelities increase from 0.522 to 0.669 for singlet initialization and from 0.857 to 0.890 for the warm start. For warm-started 12-, 16-, and 20-qubit systems, two QCGD updates match or exceed three DB-QITE updates using approximately 36% of the compiled gates and 40–41% of the depth. These savings concern selected preparation circuits, excluding parameter-search overhead: QCGD evaluates 400 candidates per iteration versus 20 for DB-QITE. Its leading-order primitive count grows as $O((2+\sqrt{5})^k)$, versus $O(3^k)$ for DB-QITE. Stated limitations include sensitivity to parameter-search range, restriction to one Hamiltonian family, two initializations and three updates, and the absence of QCGD-specific cooling and ground-state-fidelity guarantees.
Central claim may not hold: The rescaled memory phases in the final QCGD circuit do not consistently implement the stated Riemannian conjugate-gradient recursion across iterations, weakening the central algorithmic identification despite the reported numerical improvements. Eq. modified-coefficients incorrectly gives theta_k = 1 even though its definition requires theta_k = sqrt(mu_k), which is the value used in the subsequent circuit.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.