We present a quantum-native framework for auxiliary-field quantum Monte Carlo (AFQMC), viewed as imaginary-time projection for ground-state calculation through an ensemble of one-body propagators driven by stochastic auxiliary fields. Starting from a Feynman-Kac representation, each trajectory is converted into a sequence of piecewise-constant one-body generators using stochastic Magnus expansions up to second order. The resulting nonunitary one-body slices are implemented either by unitary dilation with a small ancilla register or by an equivalent system-only linear-combination-of-unitaries (LCU) form. The dilation formulation gives a coherent fault-tolerant segmented projector algorithm with linear scale-over-gap dependence, while the LCU formulation gives a near-term sampled estimator whose circuits are fermionic Gaussian and avoid the explicit dilation register. The two modes have different resource assumptions and different manifestations of the AFQMC sign and phase problem: estimator variance in the sampled near-term mode and normalization or postselection-amplitude cost in the coherent mode. Benchmarks on the Hubbard model verify the dilation and stochastic Magnus errors classically, test segment chaining, and demonstrate proof-of-principle multistep executions on IBM quantum hardware.

Close

PrevNext

View figure in article

• Quantum algorithms & computation
• Quantum computation
• Quantum chemistry methods