Electronic Density versus Geometry for Machine-Learned Molecular Absorption Spectra
Abstract
Molecular optical absorption spectroscopy provides a direct probe of electronic structure and is widely used for molecular identification, interpretation of photophysical behaviour, and planning of spectroscopy experiments. Calculating the absorption spectra using first-principle excited-state methods, however, is computationally demanding, at least compared to ground-state calculations, which limits their routine application across large molecular sets. Machine-learning (ML) surrogates can reduce this cost and allow rapid spectral prediction. However, their performance depends strongly on how molecular information is represented. Here, we compare using the ground-state electron density versus the molecular geometry as inputs to a ML model for predicting absorption spectra, for a training set of 6874 molecules selected from the QM7 dataset. For each of these molecules, the density was calculated using density functional theory (DFT) and the absorption spectrum was calculated using linear-response (LR) time-dependent DFT (TDDFT). Utilizing the ground-state density as the input to the ML model is motivated by the Hohenberg-Kohn and Runge-Gross theorems, and the fact that the ground-state density encodes information about bonding, charge localisation, and electronic delocalisation. Hence, it may be a more judicious starting point for the ML model compared to the geometry, as it effectively decouples the chemistry of the ground-state. The question we test is whether the benefits of using the density outweigh the (notprohibitive) penalty of requiring an additional single-point DFT calculation for the density. We find that the density-based convolutional neural network achieves a validation correlation of 0.9926, compared with 0.9795 for the best geometry-based graph model, reducing the residual decorrelation, by approximately 64%.
Related
- Disentangling Electronic and Lattice Contributions to Transient Absorption in Metal Halide Perovskites: A First-Principles Study of CH3NH3PbBr3.
- Electronic states near surfaces and interfaces of β-Ga2O3 and κ-Ga2O3 epilayers investigated by surface photovoltage spectroscopy, photoconductivity and optical absorption
- High-Resolution X-ray Absorption Spectroscopy of Soluble Methane Monooxygenase: Room Temperature and Cryogenic Insights into Active Site Electronic Structure Modulation.
- The Evolution of Electronic Structure in Conjugated Molecules Revealed by Mixed-Reference Spin-Flip Time-Dependent Density Functional Theory and Density-Based Analyses: II. Polyacenes
- Excited-state Properties Beyond the Excitation Energy from Orbital-Optimized Density Functional Calculations II: Absorption Spectra
BibTeX
@misc{dhanpal2026electronic,
title = {{Electronic Density versus Geometry for Machine-Learned Molecular Absorption Spectra}},
author = {Siddharth Dhanpal and Peter Elliott and Paolo Emilio Trevisanutto and Alin M. Elena and Gilberto Teobaldi},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.03444},
eprint = {2610.03444},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.03444},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Electronic Density versus Geometry for Machine-Learned Molecular Absorption Spectra
AU - Siddharth Dhanpal
AU - Peter Elliott
AU - Paolo Emilio Trevisanutto
AU - Alin M. Elena
AU - Gilberto Teobaldi
PY - 2026
DA - 2026/10/02
PB - arXiv
DO - 10.48550/arxiv.2610.03444
UR - https://arxiv.org/abs/2610.03444
AB - Molecular optical absorption spectroscopy provides a direct probe of electronic structure and is widely used for molecular identification, interpretation of photophysical behaviour, and planning of spectroscopy experiments. Calculating the absorption spectra using first-principle excited-state methods, however, is computationally demanding, at least compared to ground-state calculations, which limits their routine application across large molecular sets. Machine-learning (ML) surrogates can reduce this cost and allow rapid spectral prediction. However, their performance depends strongly on how molecular information is represented. Here, we compare using the ground-state electron density versus the molecular geometry as inputs to a ML model for predicting absorption spectra, for a training set of 6874 molecules selected from the QM7 dataset. For each of these molecules, the density was calculated using density functional theory (DFT) and the absorption spectrum was calculated using linear-response (LR) time-dependent DFT (TDDFT). Utilizing the ground-state density as the input to the ML model is motivated by the Hohenberg-Kohn and Runge-Gross theorems, and the fact that the ground-state density encodes information about bonding, charge localisation, and electronic delocalisation. Hence, it may be a more judicious starting point for the ML model compared to the geometry, as it effectively decouples the chemistry of the ground-state. The question we test is whether the benefits of using the density outweigh the (notprohibitive) penalty of requiring an additional single-point DFT calculation for the density. We find that the density-based convolutional neural network achieves a validation correlation of 0.9926, compared with 0.9795 for the best geometry-based graph model, reducing the residual decorrelation, by approximately 64%.
ER -