Optimal control of molecular states in a learning loop with a parameterization in frequency and time domain

Abstract

Learning loops, consisting of femtosecond pulse shaper, feedback signal and evolutionary algorithm, can be used to optimize coherent control processes in complex molecular systems. Using different parameterizations for the algorithm-shaper-interface not only in the frequency but also in the time domain, where the genes of the algorithm directly represent the applied electric field, the optimization process can be dramatically improved and gives direct insight about the physical mechanism responsible for the achieved control. With the prototype system K 2 combined with fs-DFWM we demonstrate control over the vibrational dynamics through parameterizations of linear chirp, impulsive-stimulated-Raman-scattering, and phase-locked double pulses.