Bang-bang preparation of quantum many-body ground states in two dimensions: optimization of the algorithm with a two-dimensional tensor network

A bang-bang (BB) algorithm prepares the ground state of a two-dimensional (2D) quantum many-body Hamiltonian $H=H_1+H_2$ by evolving an initial product state alternating between $H_1$ and $H_2$. We use the neighborhood tensor update to simulate the BB evolution with an infinite pair-entangled projected state (iPEPS). The alternating sequence is optimized with the final energy as a cost function. The energy is calculated with the tangent space methods for the sake of their stability. The method is benchmarked in the 2D transverse field quantum Ising model near its quantum critical point against a ground state obtained by variational optimization of the iPEPS. The optimal BB sequence differs non-perturbatively from a sequence simulating quantum annealing or adiabatic preparation (AP) of the ground state. The optimal BB energy converges with the number of bangs much faster than the optimal AP energy.