Machine-learning guided discovery of ultralow-threshold organic gain materials towards electrically pumped lasing

Abstract

Developing organic gain molecules with an ultralow amplified spontaneous emission (ASE) threshold has been considered a key condition for realizing electrically pumped organic lasers (EPOLs) over the past 30 years. However, the lasing threshold remains an order of magnitude higher than 0.01 μJ cm⁻² (corresponding to a theoretical threshold current density of 0.1 kA cm⁻²) by the traditional research paradigm, i.e., the experimental trial-and-error method. Here, an efficient approach is demonstrated to discover ultralow-threshold organic gain materials via machine learning. New molecular structures with a minimum ASE threshold of approximately 0.011 μJ cm⁻² (equivalent to 0.1 kA cm⁻²) are found by analyzing 92880 untested organic gain molecules. These molecules with an ultralow ASE threshold feature an A–π–A structure, where A is 10-(4-methylphenyl)acridin-9-one, and the π bridge comprises a three-phenyl conjugated unit with various substituents that modulate charge distribution and polarity, thereby lowering the ASE threshold. Utilizing these molecules as the gain medium is anticipated to reduce the lasing threshold to below 0.01 μJ cm⁻². Our work provides a new paradigm for discovering laser molecules and further advances the realization of EPOLs.