Speedup for quantum optimal control from automatic differentiation based on graphics processing units

摘要

We implement a quantum optimal control algorithm based on automaticdifferentiation and harness the acceleration afforded by graphics processingunits (GPUs). Automatic differentiation allows us to specify advancedoptimization criteria and incorporate them in the optimization process withease. We show that the use of GPUs can speed up calculations by more than anorder of magnitude. Our strategy facilitates efficient numerical simulations onaffordable desktop computers, and exploration of a host of optimizationconstraints and system parameters relevant to real-life experiments. Wedemonstrate optimization of quantum evolution based on fine-grained evaluationof performance at each intermediate time step, thus enabling more intricatecontrol on the evolution path, suppression of departures from the truncatedmodel subspace, as well as minimization of the physical time needed to performhigh-fidelity state preparation and unitary gates.