Machine-learning Based Prediction of Thermal Limits for Conjugated Organic Materials

Concerns about stability remain a key impediment to further penetration of organic materials into electronics and photonics. While organic light-emitting diodes (OLEDs) have emerged as a mass-market technology, use of organic materials in other active components (transistors, electro-optic modulators, etc.) has lagged despite potential advantages in size, weight, power, cost, and/or performance (SWAP-cp). Such concerns have been particularly acute for integrating organics with conventional semiconductors. Development of software tools capable of inferring thermal stability based on computational and experimental datasets for existing materials would substantially de-risk development and implementation of organic semiconductors. This Phase I STTR project would combine NREL expertise in machine learning related to decomposition of organic materials development of organic semiconductor with NLM expertise in organic semiconductor materials for demanding thermal applications and materials analysis. NLM proposes developing a software tool known as SCANN-DT and based on graph neural networks (GNNs) that can parse a conjugated organic molecule and predict its decomposition temperature and temperature at which functionality is lost. The SCANN-DT model would be offered as a commercial SaaS package to R&D teams working in organic electronics in addition to providing indirect benefits via internal use by NLM to produce robust materials for next-generation HPC systems.