arXiv:2610.39129
Reverse Quantum Annealing for Ground-State Energy Calculation in Molecule
Pradyot Pritam Sahoo
Abstract
We present the results of our reverse quantum annealing computations of the ground-state energies of alkaline-earth monofluoride molecules. By partitioning the Dirac--Coulomb electronic Hamiltonian into sub-QUBOs and refining forward-annealing solutions through optimized reverse-annealing schedules, residual configuration errors are systematically minimized. The electronic ground-state wave functions and energies are obtained using the Quantum Annealing Eigensolver algorithm. This work provides a pathway for achieving sub-milliHartree chemical accuracy and advancing molecular electronic structure calculations in the quantum annealing paradigm.
AI Summary
The paper asks whether reverse quantum annealing can refine forward-annealing solutions in a Quantum Annealing Eigensolver (QAE) for relativistic molecular ground-state energies. It studies BeF, MgF, CaF, SrF, and BaF in $(3e,8o)$ and $(7e,14o)$ active spaces using Dirac–Coulomb Hamiltonians. QAE encodes determinant coefficients with 10 binary variables each and partitions the resulting QUBO into hardware-sized blocks. Forward-annealing solutions initialize reverse annealing on a D-Wave Advantage 4.1 processor, testing twelve pause-type and V-shaped schedules with 1,000 reads per schedule. Exact full configuration interaction diagonalization in the same determinant basis provides the reference. Selecting the best reverse schedule for each system reduces residual forward-annealing errors by $82.3\%$–$98.1\%$. Final errors range from $0.305$ to $6.966$ mHa; two of ten systems reach sub-milliHartree accuracy: BeF in $(3e,8o)$ and SrF in $(7e,14o)$. A 20-microsecond V-shaped schedule reversing to $s'=0.40$ gives the largest mean energy lowering in both active spaces, although individual molecules favor different schedules. The paper reports strong schedule and molecule dependence, occasional energy increases in the smaller active space, and reduced mean relative energy lowering as the active space grows: $0.409\%$ versus $0.071\%$, averaged across molecules and schedules. It identifies finite connectivity, embedding and chain effects, stochastic sampling, and separately optimized sub-QUBOs as practical limitations. The results concern refinement against forward annealing within this fixed encoding and decomposition workflow.
Central claim may not hold: The protocol does not explain how reverse annealing optimizes inter-block couplings after concatenation when hardware limitations required partitioning, and decoding or evaluating the full Hamiltonian cannot itself perform that optimization. The stated algorithm omits both normalization of decoded coefficients and a Rayleigh quotient, so its displayed energy expression does not establish the variational energies needed to interpret the FCI benchmarks. The decrease in percentage energy lowering with active-space size is energy-origin dependent and does not establish reduced absolute recovery; the benchmark's mean best-per-system absolute lowering increases from about 21.130 to 26.698 mHa.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 10 pages, 7 figures, comments are welcome