Predicting Free Energies of Exfoliation and Solvation for Graphitic Carbon Nitrides Using Machine Learning

Area of Science:

• Materials Science
• Nanotechnology
• Computational Chemistry

Background:

• Graphitic carbon nitride (g-C3N4) is a metal-free, visible-light-responsive material crucial for solar energy conversion and thin-film transistors.
• Liquid-phase exfoliation (LPE) is a key method for synthesizing 2D g-C3N4 nanosheets, but solvent selection is critical.
• Accurate prediction of exfoliation (ΔGexf) and solvation (ΔGsol) free energies is vital for efficient LPE, but traditional methods are limited.

Purpose of the Study:

• To develop accurate machine learning (ML) models for predicting ΔGexf and ΔGsol of g-C3N4 in various solvents.
• To accelerate the identification of optimal solvents for the liquid-phase exfoliation of g-C3N4.
• To provide insights into solvent properties influencing g-C3N4 exfoliation and solvation.

Main Methods:

• Generated a database of ΔGexf and ΔGsol values for 49 solvents using molecular dynamics (MD) simulations.
• Included solvent descriptors such as density, surface tension, and dielectric constant in the dataset.
• Compared six ML algorithms, with the extra tree regressor showing the best performance for predicting free energies.

Main Results:

• Developed a highly accurate ML model for predicting g-C3N4 exfoliation and solvation free energies.
• Identified key solvent descriptors that influence the LPE process.
• Experimental validation confirmed the model's predictions through dispersibility tests.

Conclusions:

• Machine learning offers an efficient alternative to MD simulations for predicting solvent performance in g-C3N4 LPE.
• The developed ML models and identified descriptors can guide solvent selection strategies for scalable g-C3N4 nanosheet synthesis.
• This approach facilitates the advancement of g-C3N4 applications in solar energy and electronics.