Utility-scale Experimental Quantum Computation with Hardware Efficient Ansätze and Calibrated Hamiltonian

Variational quantum algorithms often adapt the ansätze (ansatz) to the target Hamiltonian, leading to deep or hardware-incompatible circuits. In this work we demonstrate an opposite approach: choosing a hardware-efficient CNOT-ladder ansatz and calibrating the Hamiltonian to compensate for its expressibility to solve ground-state energy problem of 100+ site lattice on real quantum computer. This ansatz-calibrated Hamiltonian framework enables two key advantages: (1) it allows restructuring the algorithm into smaller sub-problems which can independently solved off-line classically, leaving quantum computer tasked with recombining partial solutions into full state-vector with fewer parameters to optimize and (2) it leverages circuit-specific error statistics to de-noise observable readouts with nominal noise-mitigation overhead. We simulate the spin-1/2 antiferromagnetic Heisenberg model over a 103-site flat Kagome lattice using IBM’s Heron r1 and r2 processors. The obtained per-site ground-state energy of − 0.4172 J , which approaches the benchmark − 0.4386 J after open boundary correction and approximate error mitigation. Our experiments reveal a peculiar distribution of bond energies when the ansatz is executed on the quantum processor, in contrast to its classical simulation, which merits further investigation.