Rostock 2019 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
A: Fachverband Atomphysik
A 7: Interaction with strong or short laser pulses
A 7.7: Vortrag
Montag, 11. März 2019, 15:45–16:00, S HS 3 Physik
Gaussian-process optimization of photoionization dynamics of many-electron systems — •Yi-Jen Chen1, Ulf Saalmann1, Robin Santra2, and Jan M. Rost1 — 1Max Planck Institute for the Physics of Complex Systems, Dresden, Germany — 2Center for Free-Electron Laser Science, Hamburg, Germany
Advances in light-source technology have opened up the possibility to manipulate the properties of light and, thus, the possibility to dynamically steer a quantum system in a desired manner. Since the simulation and measurement of light-driven quantum dynamics are costly, for a given control task, the optimal set of laser parameters has to be found as quickly as possible.
In this talk, we present a Bayesian active learning algorithm based on Gaussian-process regression to tackle such optimal control problems. This algorithm converts the search for the optimum of an expensive target function into that of a relatively cheap cost function. As the derivatives of the cost function permit an analytical form, its optimum can be located efficiently using gradient-based optimization schemes.
We apply this algorithm to maximize the degree of coherence of a photoion created by attosecond photoionization. The reduced density matrix of the ionic wave packet, obtained here by ab-initio solution of the N-electron Schrödinger equation, may be probed experimentally using ultrafast spectroscopy. In addition to numerical efficiency, we discuss how to extract physical insights and to select an experimentally accessible maximum by exploiting the global and local information of the target-function landscape.