Work

RamanGPT maps bidirectionally between crystal structures and Raman spectra of inorganic materials

Year: 2026
Type: article
Abstract: Raman spectroscopy is one of the most accessible vibrational probes in materials laboratories, but its forward problem (structure to spectrum) is bottlenecked by the cost of density functional perturbation theory, and its inverse problem (spectrum to structure) typically relies on retrieval against curated references. We introduce RamanGPT, a deep-learning framework that addresses both directions for crystalline inorganic materials. The forward model, an Atomistic Line Graph Neural Network (ALIGNN), is trained on the 5,099-material Computational Raman Database and predicts 200-bin spectra over 50–1000 $$\hbox {cm}^{-1}$$ with 53.8% of the 509 held-out materials reaching an active-region cosine similarity of at least 0.354, capturing qualitative features of the target spectrum. The model also shows qualitative agreement with an experimental measurement of metallic 1T $$\hbox {VSe}_{2}$$ , a system absent from the training set. The inverse model fine-tunes DiffractGPT, a large language model already trained for diffraction-to-structure generation, via Quantized Low-Rank Adaptation on Raman-plus-formula prompts, recovering lattice parameters with mean absolute errors of 1.13–2.14 Å and reduced-formula consistency of 92.1% on 508 held-out materials, compared to 17.1% for a formula-conditioned retrieval baseline given identical inputs. A cosine-similarity matcher and an inverse $$\rightarrow$$ relax $$\rightarrow$$ forward consistency loop are deployed at https://atomgpt.org/raman .
Language: English
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7163384766
FWCI:
Cites: 60
Cited by:
Related to: 10