Electronic spectra from TDDFT and machine learning in chemical space.
Raghunathan Ramakrishnan1, Mia Hartmann2, Enrico Tapavicza2
1Institute of Physical Chemistry and National Center for Computational Design and Discovery of Novel Materials, Department of Chemistry, University of Basel, Klingelbergstrasse 80, CH-4056 Basel, Switzerland.
Machine learning models improve time-dependent density functional theory (TD-DFT) predictions for electronic spectra. Training on coupled-cluster (CC2) data accurately reproduces excitation energies for organic molecules.
Area of Science:
- Computational chemistry
- Quantum chemistry
- Spectroscopy
Background:
- Time-dependent density functional theory (TD-DFT) offers efficient computation for electronic spectra.
- TD-DFT predictions often suffer from significant inaccuracies.
- Accurate electronic spectra are crucial for understanding molecular properties and reactions.
Purpose of the Study:
- To enhance the accuracy of TD-DFT electronic spectra predictions using machine learning.
- To develop a data-driven approach for high-throughput spectral predictions.
- To assess the performance of machine learning models across a large chemical space.
Main Methods:
- Machine learning models were trained on deviations between TD-DFT and reference coupled-cluster (CC2) spectra.
- The models utilized data from over 20,000 synthetically feasible small organic molecules.
- Focus was on low-lying singlet-singlet vertical electronic spectra.
Main Results:
- Prediction errors decreased monotonically with increasing training set size.
- With 10,000 training molecules, CC2 excitation energies were reproduced within ±0.1 eV for other molecules.
- Chromophore analysis indicated potential for even higher prediction accuracy.
Conclusions:
- Machine learning effectively corrects TD-DFT spectral prediction errors.
- Data-driven approaches show great promise for accurate, high-throughput spectral calculations.
- Challenges remain in modeling high-lying spectra and transition intensities.
More Related Videos
Related Concept Videos
UV–Vis Spectroscopy: Molecular Electronic Transitions
Molecular Spectroscopy: Absorption and Emission
π Electron Effects on Chemical Shift: Overview
¹H NMR: Interpreting Distorted and Overlapping Signals
As Δν decreases and the signals move closer, the doublets appear increasingly distorted. The intensities of the inner lines increase at the cost of those of the outer lines as the signals are...
Mass Spectrum: Interpretation
Atomic Spectroscopy: Absorption, Emission, and Fluorescence


