The vibrational response of polymer networks under load provides a sensitive probe of molecular deformation and enables non-destructive diagnostics. Machine-learned force fields reproduce these spectroscopic fingerprints with quantum-level fidelity in realistic epoxy thermosets. Using MACE-OFF23 molecular dynamics, the experimentally observed redshifts of para-phenylene stretching modes under tensile load are captured, in contrast to predictions from the harmonic OPLS-AA model. The shifts correlate with molecular elongation and alignment, consistent with Badger’s rule, thereby establishing a direct link between vibrational features and local stress. Infrared intensities are predicted through a symmetry-adapted dipole moment model trained on representative epoxy fragments, enabling quantitative validation of strain-dependent responses. The combined approach yields chemically accurate and computationally efficient predictions of vibrational spectra under deformation. These results identify vibrational fingerprints as predictive markers of mechanical state in polymer networks and outline a spectroscopic route to stress mapping and structural-health diagnostics in advanced materials.
Konrad et al. (Mon,) studied this question.