Related Experiment Video
Updated: Jun 25, 2025
06:15
Preparation of Binary and Ternary Deep Eutectic Systems
Published on: October 31, 2019
11.9K
Assessing Viscosity in Sustainable Deep Eutectic Solvents and Cosolvent Mixtures: An Artificial Neural Network-Based
Luan Vittor Tavares Duarte de Alencar1,2, Sabrina Belén Rodríguez-Reartes1,3,4, Frederico Wanderley Tavares2,5
1Department of Chemical Engineering, ETSEQ, Universitat Rovira i Virgili, Avinguda Països Catalans 26, 43007 Tarragona, Spain.
ACS Sustainable Chemistry & Engineering
|May 31, 2024
Summary
An artificial neural network (ANN) accurately predicts the viscosity of deep eutectic solvents (DESs) and their mixtures. This model uses molecular descriptors and shows excellent extrapolation capabilities for designing new green solvents.
Area of Science:
- Green Chemistry and Sustainable Solvents
- Computational Chemistry and Molecular Modeling
Background:
- Deep eutectic solvents (DESs) are emerging as environmentally friendly alternatives in chemical processes.
- Accurate prediction of DES physicochemical properties, particularly viscosity, is essential for their application but challenging.
- Current design methods for tailored DESs are resource-intensive, necessitating advanced predictive models.
Purpose of the Study:
- To develop an accurate artificial neural network (ANN) model for predicting the viscosity of DESs and their mixtures.
- To utilize molecular parameters derived from σ-profiles computed via COSMO-SAC for model input.
- To assess the model's predictive accuracy, extrapolation capacity, and applicability domain.
Main Methods:
- An artificial neural network (ANN) model was trained using a comprehensive dataset of 1891 experimental viscosity measurements for 48 choline chloride-based DESs and their mixtures.
- Molecular descriptors were computed using the conductor-like screening model for the real solvent segment activity coefficient (COSMO-SAC).
- The optimal ANN architecture (9-19-16-1) was determined to describe the logarithmic viscosity.
Main Results:
- The ANN model achieved a high accuracy with an overall average absolute relative deviation of 1.6031% for viscosity prediction.
- The model demonstrated significant extrapolation capabilities, accurately predicting viscosity for systems with untrained solvents like ethanol and 2,3-butanediol.
- The developed ANN model covers an extensive applicability domain, validating its robustness across a wide range of DES compositions and properties.
Conclusions:
- The study successfully established a robust and highly accurate ANN model for predicting DES viscosity using molecular descriptors.
- The model's strong extrapolation capacity facilitates the design and application of novel deep eutectic solvents.
- This work represents a significant advancement in developing open-source, accurate predictive tools for green solvent engineering.
