SAMOP 2023 – wissenschaftliches Programm

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

MO: Fachverband Molekülphysik

MO 6: Poster I

MO 6.18: Poster

Dienstag, 7. März 2023, 16:30–19:00, Empore Lichthof

Classification of noisy spectra using machine learning — •Aritra Mishra and Alexander Eisfeld — Max Planck Institute for the Physics of Complex Systems, Dresden, Germany

A general problem in quantum mechanics is to obtain information of the eigenstates from the experimental measured data which consists inherent noises. In particular, we consider molecular aggregates, where information about excitonic eigenstates is vitally important to understand their optical and transport properties [1,2]. It has been shown that it is possible to reconstruct the underlying delocalised aggregate eigenfunctions from near-field spectra using convolution neural networks  [3].

In this work, we also use a convolution neural network but ask a question related to the eigenstate based classification of the spectra in the presence of noise. Given that each eigenstate correspond to a distinct spectrum, we can assign a class to each of the eigenstate. We add a random noise to these spectra and build a network that can classify the spectra into these classes, in the presence of the noise. We find that the network is also able to classify the spectra of different noise strengths along with the one it has been trained for.
[1] X. Gao and A. Eisfeld, J. Phys. Chem. Lett. 9, 6003 (2018)
[2] S. Nayak, F. Zheng and A. Eisfeld, J. Chem. Phys. 155, 134701 (2021)
[3] F. Zheng, X. Gao and A. Eisfeld, Phys. Rev. Lett. 123, 163202 (2019)