Predicting the Surface Tension of Deep Eutectic Solvents Using Artificial Neural Networks

Area of Science:

• Green Chemistry
• Computational Chemistry
• Materials Science

Background:

• Deep eutectic solvents (DESs) are gaining research interest as sustainable alternatives.
• Predicting DES physicochemical properties is crucial for industrial applications.
• Surface tension is a key property for many DES applications.

Purpose of the Study:

• To develop a generalized artificial neural network (ANN) model for predicting DES surface tension.
• To utilize molecular parameters as input for the ANN model.
• To establish a computational method for efficient DES property prediction.

Main Methods:

• A comprehensive database of 1571 data points from 133 DES mixtures was compiled.
• An artificial neural network (ANN) model was trained using molecular parameters from the conductor-like screening model.
• The optimal ANN architecture (9-15-15-1) with two hidden layers was determined.

Main Results:

• The ANN model achieved high prediction accuracy for DES surface tension.
• Training and testing yielded R² values of 0.986 and 0.977, respectively.
• An average absolute relative deviation of 2.20% was observed, demonstrating excellent predictive performance.

Conclusions:

• The developed ANN model provides an accurate and efficient method for predicting DES surface tension.
• The model's reliance on molecular parameters reduces the need for extensive experimental work.
• This initiative promotes the development of robust predictive models for DES properties, saving time and resources.