Resource-Free Quantum Hamiltonian Learning Below the Standard Quantum Limit

Abstract: Accurate and resource-efficient estimation of quantum Hamiltonians is crucial for developing practical quantum technologies, yet current methods typically demand entanglement resources or dynamical control. Here, we demonstrate a method that surpasses the standard quantum limit without requiring entanglement resources, coherent measurements, or dynamical control. Our method relies on trajectory-based Hamiltonian learning, in which we apply local, randomized pre-processing to probe states and apply the maximum-likelihood estimation to optimally scheduled Pauli measurements. Analytically, we establish the emergence of a transient Heisenberg-limited regime for short-time probes for our procedure. Furthermore, we outline how to estimate all Hamiltonian parameters in parallel using ensembles of probe states, removing the need for parameter isolation and structural priors. Finally, we supplement our findings with a numerical study, learning multiple disordered, anisotropic Heisenberg models for a 1D chain of spin-1/2 particles, featuring local transverse fields with both nearest- and next-nearest-neighbour interactions, as well as a gapless XXZ Hamiltonian. Our numerics show that our method needs only one shot per Pauli measurement, making it well-suited for experimental scenarios. The code for our method is available online and open-source.