Optimal Lower Bounds for Hamiltonian Simulation
This work suggests that for many physical systems, gate count must scale polynomially in $1/\epsilon$, contrary to the complexity suggested by counting coherent oracle queries such as those in the block-encoding model.