Efficient Hamiltonian reconstruction from equilibrium states

We describe a novel algorithm that recovers the Hamiltonian and temperature of a quantum state in thermal equilibrium using a restricted set of measurements. The algorithm works by imposing a semidefinite constraint based on free energy minimization, which we interpret in terms of thermodynamic stability. We benchmark the algorithm on the problem of learning a nearest-neighbour Hamiltonian on a 100-qubit spin chain.