Local-in-time error in variational quantum dynamics

The McLachlan "minimum-distance" principle for optimizing approximate solutions of the time-dependent Schrodinger equation is revisited, with a focus on the local-in-time error accompanying the variational solutions. Simple, exact expressions are provided for this error, which are then evaluated in illustrative cases, notably the widely used mean-field approach and the adiabatic quantum molecular dynamics. These findings pave the way for the rigorous development of adaptive schemes that re-size on-the-fly the underlying variational manifold and thus optimize the overall computational cost of a quantum dynamical simulation.