Litcius/Paper detail

Structural Deformation Controls Charge Losses in MAPbI<sub>3</sub>: Unsupervised Machine Learning of Nonadiabatic Molecular Dynamics

Guoqing Zhou, Weibin Chu, Oleg V. Prezhdo

2020ACS Energy Letters83 citationsDOI

Abstract

The rapid increase in perovskite solar cell efficiencies has motivated massive experimental and theoretical efforts aimed at understanding and enhancing the performance. We apply machine learning to nonadiabatic molecular dynamics simulation of nonradiative charge recombination in MAPbI3 and discover that the I–I–I angle is the key structural parameter governing nonadiabatic electron–phonon coupling and the bandgap. Surprisingly, the structure of MAPbI3 is much more important that the motions of MAPbI3, even though the coupling depends explicitly on nuclear velocity. Also surprisingly, rotational and center-of-mass motions of MA influence charge recombination, even though MA does not contribute to electron or hole wave functions. The findings rationalize the unusual temperature dependence of carrier lifetimes in halide perovskites and emphasize inorganic lattice deformation and MA rotation during polaron formation. By detecting nontrivial correlations within complex data and providing accurate quantitative measures, machine learning surpasses traditional analyses and suggests that perovskite performance can be controlled by chemical changes that alter perovskite geometric structure.

Topics & Concepts

PolaronPerovskite (structure)PhononCharge carrierChemical physicsCoupling (piping)ElectronMolecular dynamicsCharge (physics)Lattice (music)HalideChemistryPhysicsMolecular physicsCondensed matter physicsMaterials scienceComputational chemistryCrystallographyQuantum mechanicsAcousticsMetallurgyInorganic chemistryPerovskite Materials and ApplicationsSolid-state spectroscopy and crystallographyChalcogenide Semiconductor Thin Films
Structural Deformation Controls Charge Losses in MAPbI<sub>3</sub>: Unsupervised Machine Learning of Nonadiabatic Molecular Dynamics | Litcius