Machine Learning for Charge and Exciton Dynamics
Quantum chemistry calculations for OSCs are computationally expensive, especially when modeling large-scale systems with time-dependent properties. Machine learning (ML) techniques allow for efficient approximation of electronic properties, making it possible to simulate charge and exciton transport with high accuracy.
Neural network-based Hamiltonians provide a data-driven approach to predicting site energies and electronic couplings, significantly reducing computational costs. These ML models are trained on density functional tight binding (DFTB) calculations, enabling them to generalize charge dynamics across various molecular configurations [5, 6 , 7]. The integration of ML models with nonadiabatic molecular dynamics simulations enables the study of charge and exciton transport over extended time scales. The ability to predict transport parameters dynamically makes ML-based simulations a powerful tool in designing new OSC materials.
Exciton transport in OSCs is also a fundamental process that governs the efficiency
of light-harvesting devices such as solar cells and OLEDs. Unlike free charge carriers, excitons are bound electron-hole pairs that move via resonant energy transfer mechanisms. The Frenkel Hamiltonian is commonly used to model exciton transport in molecular solids [8]. This Hamiltonian describes excitonic interactions at the molecular scale,incorporating key parameters such as site energies and exciton coupling strengths. Nonadiabatic molecular dynamics simulations provide a realistic description of exciton motion, accounting for effects such as thermal disorder and molecular vibrations. By integrating Neural Network-based Frenkel Hamiltonians, we achieve a computationally efficient method for predicting exciton transport properties, enabling the rapid screening of potential materials for optoelectronic applications [7].
Bridging Theory, Experiment, and Device Performance
Bridging theoretical predictions with experimental measurements is crucial for validating computational models and advancing material design. The methodologies developed in this research offer a systematic framework for studying charge and exciton transport across multiple length and time scales. By calculating electronic coupling and examining the interplay between molecular structure and charge/exciton transfer properties, these approaches provide deeper insights into transport mechanisms and their impact on material performance [9, 10, 11, 12, 13, 14, 15].
Using a combination of quantum chemistry, molecular dynamics, and machine learning, we optimize OSC materials for high mobility and efficient exciton transport. The agreement between theoretical and experimental charge mobility measurements underscores the predictive power of our modeling approaches.
References
[1] Weiwei Xie, Daniel Holub, Tom´aˇs Kubaˇr, and Marcus Elstner. Performance of mixed
quantum-classical approaches on modeling the crossover from hopping to bandlike
charge transport in organic semiconductors. Journal of Chemical Theory and Com-
putation, 16(4):2071–2084, 2020.
[2] Sara Roosta, Farhad Ghalami, Marcus Elstner, and Weiwei Xie. Efficient surface
hopping approach for modeling charge transport in organic semiconductors. Journal
of Chemical Theory and Computation, 18(3):1264–1274, 2022.
[3] Alexander Heck, Julian J Kranz, Tomas Kubar, and Marcus Elstner. Multi-scale
approach to non-adiabatic charge transport in high-mobility organic semiconductors.
Journal of chemical theory and computation, 11(11):5068–5082, 2015.
[4] Tomas Kubar and Marcus Elstner. Coarse-grained time-dependent density functional
simulation of charge transfer in complex systems: application to hole transfer in dna.
The Journal of Physical Chemistry B, 114(34):11221–11240, 2010.
[5] Philipp M Dohmen, Mila Kr¨amer, Patrick Reiser, Pascal Friederich, Marcus Elst-
ner, and Weiwei Xie. Modeling charge transport in organic semiconductors using
5
neural network based hamiltonians and forces. Journal of Chemical Theory and
Computation, 19(13):3825–3838, 2023.
[6] Mila Kr¨amer, Philipp M Dohmen, Weiwei Xie, Daniel Holub, Anders S Christensen,
and Marcus Elstner. Charge and exciton transfer simulations using machine-learned
hamiltonians. Journal of chemical theory and computation, 16(7):4061–4070, 2020.
[7] Farhad Ghalami, Philipp M Dohmen, Mila Kr¨amer, Marcus Elstner, and Weiwei Xie.
Nonadiabatic simulation of exciton dynamics in organic semiconductors using neural
network-based frenkel hamiltonian and gradients. Journal of Chemical Theory and
Computation, 2024.
[8] Julian J Kranz and Marcus Elstner. Simulation of singlet exciton diffusion in bulk
organic materials. Journal of chemical theory and computation, 12(9):4209–4221,
2016.
[9] Niklas J Herrmann, Nadine von Coelln, Robin M Teichgreber, Sebastian H¨ofener,
Christian Huck, Farhad Ghalami, Simon Settele, Manuel Hertzog, Marcus Elstner,
Petra Tegeder, et al. Optical and electronic properties of different thin-film poly-
morphs of pdif-cn2 controlled by zone-casting conditions. Journal of Materials Chem-
istry C, 2023.
[10] Lisa Roß, Julius Reitemeier, Farhad Ghalami, Wen Shan Zhang, J¨urgen H Gross,
Frank Rominger, Sven M Elbert, Rasmus Schr¨oder, Marcus Elstner, and Michael
Mastalerz. Two dimensional triptycene end-capping and its influence on the self-
assembly of quinoxalinophenanthrophenazines. Chinese Journal of Chemistry, 2023.
[11] Lucas Ueberricke, Ioana Ciubotaru, Farhad Ghalami, Felix Mildner, Frank
Rominger, Marcus Elstner, and Michael Mastalerz. Di-and tetracyano-substituted
pyrene-fused pyrazaacenes: Aggregation in the solid state. Chemistry–A European
Journal, 26(50):11634–11642, 2020.
[12] Lucas Ueberricke, Julia Schwarz, Farhad Ghalami, Maik Matthiesen, Frank
Rominger, Sven M Elbert, Jana Zaumseil, Marcus Elstner, and Michael Mastalerz.
Triptycene end-capped benzothienobenzothiophene and naphthothienobenzothio-
phene. Chemistry–A European Journal, 26(55):12596–12605, 2020.
[13] Lucas Ueberricke, Farhad Ghalami, Wen-Shan Zhang, Vaishnavi Rao, Frank
Rominger, Rasmus R Schr¨oder, Marcus Elstner, and Michael Mastalerz. Isostruc-
tural charge-transfer cocrystals based on triptycene end-capped quinoxalinophenan-
throphenazine. Crystal Growth & Design, 21(2):1329–1341, 2021.
[14] Lucas Ueberricke, Dennis Mizioch, Farhad Ghalami, Felix Mildner, Frank
Rominger, Thomas Oeser, Marcus Elstner, and Michael Mastalerz. Quinox-
alinophenanthrophenazine based cruciforms. European Journal of Organic Chem-
istry, 2021(34):4816–4823, 2021.
[15] Ritesh Haldar, Mariana Kozlowska, Michael Ganschow, Samrat Ghosh, Marius
Jakoby, Hongye Chen, Farhad Ghalami, Weiwei Xie, Shahriar Heidrich, Yusuke
Tsutsui, et al. Interplay of structural dynamics and electronic effects in an en-
gineered assembly of pentacene in a metal–organic framework. Chemical science,
12(12):4477–4483, 2021.