arXiv:2610.03654
A Basis-Aware Approach to Quantum Sampling of the Fermi-Hubbard Ladder
Adam Leicester, Zixu Wang, Jack Mandell, Yangyang Xu, Osama Muhammad Raisuddin, Haimeng Zhang, Jian Shi
Abstract
Sample-based quantum diagonalization can reduce the classical cost of many-body calculations by restricting diagonalization to configurations identified through quantum sampling, but its efficiency depends on how compactly the target state is represented in the sampled basis. We investigate this dependence for sample-based Krylov quantum diagonalization of the half-filled Fermi-Hubbard ladder using physically motivated one-body orbital rotations. Across position, momentum, bonding/antibonding, and leg-mode representations, the most compressible basis varies systematically with interaction strength and hopping anisotropy, with momentum sampling favored in the weakly interacting regime and geometry- or interaction-adapted representations becoming advantageous as the dominant energy scale changes. Near intermediate interaction and isotropic hopping, however, even the most favorable representation remains poorly compressible. Statevector SKQD calculations further show that basis choice strongly affects sampling efficiency. Hardware calculations on systems containing up to 120 qubits retain qualitative signatures of this basis dependence while revealing increasing particle-number leakage and reconstruction sensitivity with system size. These results show that scalable sample-based diagonalization requires co-design of the sampled representation, quantum circuit, and classical reconstruction procedure.
AI Summary
Sample-based Krylov quantum diagonalization (SKQD) samples time-evolved quantum states to select determinants for classical subspace diagonalization. This paper asks how orbital representation affects ground-state compressibility and sampling efficiency for the half-filled, two-leg Fermi–Hubbard ladder. It compares site, kinetic-eigenmode (“momentum”), rung bonding/antibonding, and leg-mode bases using exact diagonalization through 14 sites, statevector sampling, and hardware experiments through 60 sites (120 qubits), with DMRG energy references. The most compact basis depends on the dominant energy scale: momentum orbitals favor weak interactions, bonding/antibonding orbitals strong rung hopping, leg modes dominant leg hopping, and site orbitals stronger repulsion. Near isotropic hopping and intermediate coupling, none is highly compact. At $U/t=5$, retaining 99% of ground-state weight requires approximately 3.5% of the fixed-sector space for 12 sites; a small-system fit gives $k_{0.99}\sim(0.435/N)3.106^N$. This is empirical scaling, not an asymptotic bound. For a 10-site statevector calculation at $U/t=1$, momentum sampling achieves 0.47% relative energy error using 1.4% of the sector, versus 9.5% error using 33% for site sampling. Hardware momentum-basis reconstructions at $U/t=1$ reach relative energy errors of approximately 0.29% and 1.2% for 20 and 60 sites. Only 9.3% and 2.3% of raw samples retain the target particle numbers, respectively. Performance deteriorates at stronger coupling. Configuration recovery enforces particle number but need not recover important determinants or hopping connectivity. Stated limitations include reference-state bias toward momentum sampling, noise and reconstruction sensitivity, limited scaling data, and the absence of systematic orbital or reference-state optimization.
Error found: The 60-site position-basis hardware samples have a target-sector population and mean particle-number distance essentially identical to uniformly random bitstrings, so their concentration around half filling does not by itself demonstrate retained information about the intended sector.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.