Predicting Two-Photon Absorption Spectra of Octupolar Molecules: A Deep-Learning Approach Based Exclusively on...

Area of Science:

• Computational Chemistry
• Materials Science
• Machine Learning

Background:

• Octupolar molecules are crucial for advanced applications due to their strong two-photon absorption (TPA) properties.
• Predicting TPA spectra typically requires computationally intensive quantum chemical calculations.

Purpose of the Study:

• To develop a rapid and accurate deep learning method for forecasting TPA spectra of octupolar molecules.
• To overcome the computational cost associated with traditional quantum chemical simulations.

Main Methods:

• Utilized a graph neural network (GNN) model.
• Input molecular geometric structure as the initial descriptor.
• Predicted maximum two-photon transition wavelength and cross-section.

Main Results:

• Achieved a mean absolute percentage error of less than 4% compared to time-dependent density-functional theory (TD-DFT) calculations.
• The deep learning model accurately reproduced experimental observations.
• Demonstrated a speed increase of nearly 100,000 times compared to conventional quantum calculations.

Conclusions:

• The developed deep learning approach offers an efficient and cost-effective tool for simulating TPA properties.
• This method shows significant potential for high-throughput screening of novel TPA materials.