Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

K: Fachverband Kurzzeit- und angewandte Laserphysik

K 4: Poster (joint session K/Q)

K 4.9: Poster

Dienstag, 3. März 2026, 17:00–19:00, Philo 1. OG

Neural network reconstruction of hamiltonians and transition dipole couplings from transient absorption spectra — •Ronald Cardenas, Ulf Saalmann, and Jan-Michael Rost — MPI-PKS, Dresden, Germany

Transient absorption spectroscopy (TAS) provides insight into ultrafast electronic dynamics, yet the resulting spectra are often challenging to interpret with conventional tools. In this work, we develop a Convolutional Neural Network (CNN) that reconstructs effective Hamiltonian matrices directly from TAS data. In addition to recovering effective energy levels, the model also predicts transition dipole couplings, which determine electronic coherences and state interactions under external fields. These couplings are essential to determine the optical response for linear and nonlinear optical processes. As they cannot be measured directly, they are typically obtained from ab initio calculations. Such calculations can become demanding for larger systems or situations involving many excited states. The CNN approach provides an alternative route to estimating these couplings using only spectroscopic input. The reconstructed Hamilton matrices enable the calculation of dynamical properties such as time-dependent dipole moment and polarization response. They can also be used for simulations of open-system Lindblad dynamics, coherent control schemes, nonlinear spectroscopy, and strong-field ionization models. Overall, our approach links experimental TAS data to the theoretical parameters needed to model ultrafast light-matter interactions, offering a flexible framework for complex molecular systems.

Keywords: Hamiltonian reconstruction; CNN; Transition dipole couplings