Charge and Exciton Transfer Simulations Using Machine-Learned Hamiltonians.

Area of Science:

• Computational chemistry
• Materials science
• Organic electronics

Background:

• Understanding charge and exciton transfer in organic semiconductors is crucial for developing advanced electronic devices.
• Current simulation methods can be computationally intensive.

Purpose of the Study:

• To develop an efficient multiscale model for predicting charge-transfer mobilities and exciton diffusion constants.
• To apply machine learning to simulate charge and exciton propagation in organic semiconductors.

Main Methods:

• Kernel ridge regression models were trained to predict electronic and excitonic couplings.
• Semiempirical density functional tight binding (DFTB) data was used as reference.
• Nonadiabatic molecular dynamics and Marcus-based Monte Carlo approaches were employed.

Main Results:

• Machine learning models achieved high accuracy in predicting couplings from DFTB data.
• Simulations reproduced hole mobilities in anthracene within 8.5% of DFTB and 34% of experimental values.
• Exciton transfer simulation costs were reduced by an order of magnitude.

Conclusions:

• Machine learning offers an efficient and accurate approach for simulating charge and exciton dynamics in organic semiconductors.
• This multiscale model significantly reduces computational expense, enabling faster material discovery and design.