AI-Wei Xu-Abstract

OLED materials have become crucial in the displayindustry. We have adopted a graph representation of organic molecules and developeda property prediction model for OLED materials. This graph-based AI predictionis more than 100,000 times faster than quantum chemistry calculations whilemaintaining enough accuracy. Combining million-scale combinatorial enumerationof molecular fragments, we have achieved high-throughput screening of OLEDmaterials and scaled up traditional materials search by several orders ofmagnitude. Through AI-based high-throughput analysis, we have discovered a functionalgroup with a specific connectivity pattern that efficiently modulates the molecularorbital energy levels of blue-emitting materials. Inspired by the recent progressin generative language models, we have performed language modeling for OLED materialsbased on their string representation and explored alternative solutions forgenerating candidate materials. We have built a platform for integrating our AI-enabled discovery processes for OLED materials and expect it to facilitate thediscovery of a wider range of organic semiconductor materials in the future.