Predicting Molecular Laser Properties from First-Principles Using Machine Learning-Based Nuclear Ensemble Approach...

The accurate prediction of absorption and emission spectra of molecular compounds using quantum mechanical (QM) methods is essential for understanding and designing laser materials, especially when experimental data are limited or inaccessible or when the syntheses are time- and resource-consuming. In this work, a numerical framework is developed to simulate the laser properties of molecular compounds from first principles. The methodology integrates QM calculations in combination with thermal sampling and a Gaussian Mixture Model-based Nuclear Ensemble Approach (GMM-NEA) for spectra reconstruction, and a spectrally resolved spatiotemporal laser simulation model. The GMM-NEA method is extended to include spontaneous and stimulated emission processes, enabling the generation of spectroscopic input for laser modeling. The framework is validated using two boron hydrides, anti-B₁₈H₂₂ and Et₄-anti-B₁₈H₁₈, which exhibit contrasting laser behaviors, while possessing very similar absorption and emission properties. High-level multireference multiconfigurational QM calculations (CASSCF/MS-CASPT2) are employed, and the results show excellent agreement with experimental data. The present framework identifies excited-state absorption as the primary factor responsible for the absence of lasing in Et₄-anti-B₁₈H₁₈. This approach represents not only a potentially predictive in-silico screening of candidate laser compounds but also offers deeper physical insight into the behavior of novel laser materials.