Erlangen 2022 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MO: Fachverband Molekülphysik
MO 6: Theory
MO 6.3: Vortrag
Dienstag, 15. März 2022, 11:00–11:15, MO-H7
Spectral learning for (ro-)vibrational calculations of weakly-bound molecules — •Yahya Saleh1,2, Jannik Eggers1,2, Vishnu Sanjay6, Andrey Yachmenev1,3, Armin Iske2, and Jochen Küpper1,3,4,5 — 1Center for Free-Electron Laser Science, Deutsches Elektronen-Synchrotron DESY, Hamburg, Germany — 2Department of Mathematics, Universität Hamburg, Hamburg, Germany — 3Center for Ultrafast Imaging, Universität Hamburg, Hamburg, Germany — 4Department of Physics, Universität Hamburg, Hamburg, Germany — 5Department of Chemistry, Universität Hamburg, Hamburg, Germany — 6Gran Sasso Science Institute
Weakly-bound complexes of organic molecules with water play diverse roles in various fields ranging from biology to astrochemistry. Planning experiments requires accurate quantum mechanical calculations of (ro-)vibrational energies up to dissociation, which is a challenging task for these systems. Standard predictions for these problems represent the wavefunctions as a linear combination of some fixed basis set. The quality of the predictions deteriorate for highly-excited states. Moreover, the computational costs scale poorly with the dimension of the problem.
We present a nonlinear variational framework to simultaneously compute multiple eigenstates of quantum systems using neural networks. The proposed framework is shown to model excited states more accurately and is believed to scale better with the size of the system. We also present numerical analysis' results and convergence guarantees of the proposed approach.