Center for Nonlinear Studies

Tuesday, January 10, 20232:00 PM - 3:00 PMCNLS Conference Room (TA-3, Bldg 1690)
Colloquium
Machine learning motivated approach applied to discover dynamical phase transitions and quantum resonances
George P. TsironisSchool of Engineering and Applied Sciences, Harvard University and Department of Physics, University of Crete, Greece
In this work we investigate dynamical phase transitions and resonances in non-linear systems described by the Discrete Non-Linear SchrÃ¶dinger (DNLS) equation. We generalize the physics-informed machine learning (PIML) method proposed in Refs [1, 2] that successfully finds the parameters for the targeted energy transfer (TET) [1] of an electron (or exciton) to a target state and the parameters for the self-trapping (ST) [2] transition in a nonlinear dimer for the case of the quantized DNLS [3]. We define a data-free loss function that measures the stability of the system. For given values of the transition matrix element V, and of the nonlinearity parameter Ï‡, of the system, we minimize the loss function and find resonant transfer as a function of boson number N. We discuss the efficiency of the transfer as a function of N.

[1] G. D. Barmparis and G. P. Tsironis, â€œDiscovering nonlinear resonances through physics-informed machine learning,â€ J. Opt. Soc. Am. B, 38, C120-C126 (2021).

[2] G. P. Tsironis, G. D. Barmparis, D. K. Campbell, â€œDynamical symmetry breaking through AI: The dimer self-trapping transitionâ€, Int. J. Mod. Phys. B, 2240001, (2021).

[3] I. Andronis et al., Quantum targeted energy transfer through machine learning tools, arXiv:2212.00556

Host: Avadh Saxena