The application of evolutionary multi-criteria optimization to dynamic molecular alignment

Abstract

This study introduces the multi-criteria approach to the optimization of dynamic molecular alignment by shaped femtosecond laser pulses, which has been considered so far only as a single-criterion problem. The paper applies advanced Pareto front approximation algorithms to this challenging real-world, high-dimensional, and computationally expensive problem, working with low-dimensional parameterizations of the electric field. Standard approaches (NSGA-II) and their metamodel-assisted extensions based on Kriging, are applied to this optimization task and compared among each other. The study confirms the conflicting nature of the objectives. Interesting features of the problem domain, such as the geometry of the Pareto front are revealed. Furthermore, metamodel-assistance, in particular pre-screening with the Kriging-based expected improvement criterion, proves to be a valuable ingredient for improving the numerical results.