Machine-Learning Guided Quantum Chemical and Molecular Dynamics Calculations to Design Novel Hole-Conducting Organic...

Area of Science:

• Materials Science
• Computational Chemistry
• Organic Electronics

Background:

• High charge mobility in organic semiconductors is crucial for advanced applications like printed electronics and solar cells.
• Conventional materials like fullerenes and thiophenes have limitations, driving the search for novel molecular structures.
• Discovering new high-mobility materials requires efficient screening and predictive methods.

Purpose of the Study:

• To develop a predictive model for identifying organic molecules with enhanced charge mobility.
• To utilize machine learning (ML) and computational calculations to guide the discovery of new semiconductor materials.
• To achieve extrapolative discovery of novel high-mobility organic semiconductors.

Main Methods:

• Combined density functional theory (DFT) and molecular dynamics (MD) calculations were employed.
• A machine learning (ML) model was trained on 32 calculated hole mobilities of various organic molecules.
• Sequential learning (active learning) was implemented to iteratively refine the ML model and select compounds for DFT/MD analysis.

Main Results:

• A ML model was constructed using initial data, with a maximum calculated hole mobility of 10^-1.96 cm²/(V s).
• Through 60 cycles of sequential learning and 165 DFT/MD calculations, a fused thioacene molecule was identified.
• This novel molecule exhibited a calculated hole mobility of 10^-1.86 cm²/(V s), surpassing the initial maximum.

Conclusions:

• The integrated approach of ML and DFT/MD calculations enables efficient discovery of high-performance organic semiconductors.
• Sequential learning facilitates extrapolative discovery, identifying materials with properties beyond the initial training dataset.
• The identified fused thioacene represents a promising candidate for next-generation organic electronic devices.