Computational Discovery of TTF Molecules with Deep Generative Models.

Area of Science:

• Materials Science
• Computational Chemistry
• Artificial Intelligence

Background:

• Blue organic light-emitting diodes (OLEDs) require efficient triplet-triplet fusion (TTF) for improved performance.
• Discovering novel materials with specific energy transfer properties is challenging and time-consuming.

Purpose of the Study:

• To develop a computational workflow for discovering novel materials for TTF in blue OLEDs.
• To leverage generative deep learning models and quantum chemical calculations for accelerated material discovery.

Main Methods:

• A workflow combining quantum chemical calculations and deep neural network-based generative models was developed.
• Graph convolution neural networks were trained to predict excitation energies and filter molecules.
• A modified Junction Tree Variational Autoencoder (JT-VAE) was identified as the optimal generative model.

Main Results:

• Generative machine learning models identified promising chemical spaces for TTF materials.
• The workflow effectively filters molecules by predicting energy level alignment, reducing computational cost.
• Several generative models were computationally evaluated, with JT-VAE showing superior performance.

Conclusions:

• The proposed computational approach aids in the computer-aided design of materials for efficient energy transfer and exciton fusion.
• This methodology is crucial for developing next-generation OLED materials with enhanced properties.
• The workflow accelerates the discovery of molecules suitable for triplet harvesting in OLED devices.