Machine learning-assisted high-throughput screening of transparent organic light-emitting diode anode materials

Area of Science:

• Materials Science
• Organic Electronics
• Computational Chemistry

Background:

• Optimizing work functions of 2D nanomaterials is key for high-efficiency Organic Light-Emitting Diodes (OLEDs).
• Traditional material discovery is slow and inefficient, relying on empirical methods.

Purpose of the Study:

• To accelerate the discovery of transparent OLED anode materials using a target-driven design framework.
• To combine high-throughput virtual screening and interpretable machine learning (ML) for efficient material prediction.

Main Methods:

• Developed a CatBoost ML regression model to predict work functions of 2D nanomaterials (MAE of 0.20 eV).
• Utilized SHapley Additive exPlanations (SHAP) for global and local model interpretation.
• Performed multi-condition screening and density functional theory (DFT) calculations.

Main Results:

• Identified space group as a primary determinant of work function for many 2D nanomaterials.
• Discovered specific space groups (e.g., Pmn2_1, P6̄m2) correlate with high work functions, while others (e.g., P4/mmm, P1̄) correlate with low work functions.
• Identified a promising 2D nanomaterial (PS) with high conductivity (>10^6 S m^-1), transparency (>90%), and favorable work function (>5 eV).

Conclusions:

• The developed ML framework offers a cost-effective approach for designing high-performance 2D nanomaterials.
• The identified 2D nanomaterial (PS) shows potential as a superior alternative to indium tin oxide (ITO) for transparent OLED anodes.
• This study provides novel insights into the work function mechanisms of 2D nanomaterials.