Progress in material science is driven by design and discovery of increasingly complex structures and, e.g., in multi-component/high-entropy alloys the properties depend sensitively on the atomic ordering, but is challenging to determine from experiments. Raman spectroscopy is widely used, powerful, and nondestructive tool for materials analysis that could provide this information. Interpretation of the results would greatly benefit from simulations, but the computational cost for simulating Raman spectra of these complex systems is prohibitively large. In this project, machine-learning based models for simulating Raman spectra of complex materials are developed and benchmarked. From the comparison with experimental spectra of two recently discovered material families, 2D high-entropy alloys and photoferroelectric perovskites, we aim to unravel the atomic ordering and other structural data, their effect on relevant material properties, and thus accelerate the material optimization.

2023

2027