Machine Learning-Aided Screening and Design Rule Discovery for LWIR-Transparent Optical Materials.

Area of Science:

• Materials Science
• Computational Chemistry
• Optics

Background:

• Advancing thermal imaging and sensing requires low-cost, high-performance materials transparent in the long-wavelength infrared (LWIR) spectrum.
• Current LWIR optics often use expensive inorganic materials, hindering widespread application.

Purpose of the Study:

• To develop a machine learning (ML)-driven framework for discovering novel hydrocarbon monomers for LWIR-transparent polymers.
• To establish structure-property relationships for enhanced infrared transparency.

Main Methods:

• Utilized a communicative message-passing neural network (CMPNN) to predict infrared spectra of hydrocarbons.
• Analyzed spectral broadening effects (γ = 4-10) to guide material selection.
• Developed regression models for direct prediction of LWIR window transparency (wT) from ML-derived fingerprints.

Main Results:

• The ML framework demonstrated high predictive accuracy and generalizability across diverse datasets.
• Identified key structure-property relationships, revealing descriptors and substructures linked to high LWIR transparency.
• ML-prioritized motifs aligned with experimentally validated comonomers and proposed new, plausible scaffolds.

Conclusions:

• The ML approach offers a scalable, cost-effective alternative to density functional theory (DFT) for screening IR-transparent materials.
• Provides chemically grounded design insights for next-generation LWIR-transparent polymers.
• Facilitates the formulation of design rules for developing advanced infrared materials.