Machine learning-driven prediction of deep eutectic solvents’ heat capacity for sustainable process design

Aug 5, 2026

Heat capacity, a key physical property for chemical processes, remains insufficiently explored for Deep Eutectic Solvents (DESs) despite their promise as greener alternatives to conventional solvents. The study builds machine learning models to predict DES heat capacity and determine which structural features most strongly control it. A dataset of 530 DESs with experimental heat-capacity values is represented using quantum-chemical COSMO-RS descriptors, then fitted with kNN, RF, MLP, SVM and multiple linear regression. Results show the MLP achieves very low AARD on training and test sets, and SHAP analysis reveals the dominant structural factors.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Introduction

Deep eutectic solvents and the need for heat-capacity data

Heat-transfer context in solar energy and sustainable fluids

DES preparation and tunable physicochemical properties

Reported DES applications and remaining research gaps

Machine learning approach and model comparison

Dataset and experimental heat-capacity values

COSMO-RS-based descriptors

Algorithms and hyperparameter optimisation

Performance metrics (AARD) and results

Interpretability of structural drivers

SHAP structural factor analysis

Journal of Molecular Liquids 418 (2025) 126707 Contents lists available at ScienceDirect Journal of Molecular Liquids journal [homepage:](homepage: www.elsevier.com/locate/molliq)[ www.elsevier.com/locate/molliq](homepage: www.elsevier.com/locate/molliq) | Machine learning-driven prediction of deep eutectic solvents’ heat capacity for sustainable process design
Amit Kumar Halder a,b , Reza Haghbakhsh c , Elisabete S.C. Ferreira a , Ana Rita C. Duarte c, M. Nat´alia D.S. Cordeiro a,*
a LAQV, REQUIMTE/Department of Chemistry and Biochemistry, Faculty of Sciences, University of Porto, 4169-007 Porto, Portugal
b Dr. B. C. Roy College of Pharmacy and Allied Health Sciences, Dr. Meghnad Saha Sarani, Bidhannagar, Durgapur 713212, West Bengal, India c LAQV, REQUIMTE/Department of Chemistry, Faculty of Sciences and Technology, Nova University of Lisbon, 2829-516 Caparica, Portugal | | | | | --- | --- | --- | --- | | A R T I C L E I N F O | | A B S T R A C T | | | Keywords:
Deep eutectic solvents COSMO-RS
Machine learning
Heat capacity prediction Thermodynamic modelling | | Heat capacity, a crucial physical property for chemical processes, is often understudied in Deep Eutectic Solvents (DESs), which in turn are promising green alternatives to environmentally hazardous conventional solvents. This work addresses this gap by developing a machine learning model to predict DES heat capacity and identify key structural features influencing it. We employed a dataset of 530 DESs with corresponding experimental heat capacity values. Quantum-chemical COSMO-RS-based descriptors, capturing detailed information about DES structures, were calculated for each data point. Various machine learning algorithms, namely k-Nearest Neighbours (kNN), Random Forests (RF), Neural Network Multilayer Perceptron (MLP), and Support Vector Machines (SVM) were explored alongside a linear model (Multiple Linear Regression, MLR). Hyperparameter optimisation ensured all models were fine-tuned for optimal performance. The most successful model, based on the MLP technique, achieved remarkably low Average Absolute Relative Deviation (AARD) values of 0.500 % and 3.999 % for the training and test sets, respectively. This signifies a significant improvement in prediction accuracy compared to traditional methods. Furthermore, by applying a SHapley Additive exPlanations (SHAP) analysis, we identified the most crucial structural factors within DES components that govern their heat capacity. This comprehensive investigation offers valuable insights that can pave the way for an efficient design of novel DESs in the future. | | 1. Introduction Heat capacity is a critical property in various chemical and industrial processes, and its optimal value varies depending on the specific application. In processes such as extraction and separation, it is essential that the heat capacity of the solvent be as low as possible to reduce the system’s energy consumption. However, that is not the case in energyrelated applications. Solar energy, for example, stands out as one of the most naturally available renewable sources, with electricity being produced through solar photovoltaic panels or heat via solar thermal systems. In the conventional solar thermal approach, toxic gases are emitted, contributing to global warming. In contrast, an alternative method harnesses solar radiation to generate vapour, using heat transfer or thermal fluids as a medium to store energy [1,2]. While typical thermal fluids have limited heat storage capacity and low thermal stability, molten salts suffer from high freezing points, low heat capacities, and corrosive characteristics. Addressing these challenges has been the focus of extensive research aimed at replacing environmentally harmful compounds with sustainable solvents. Ionic liquids (ILs) initially emerged as a greener alternative due to their non-volatile, chemical and thermal stabilities, and tuneable features. However,

FAQs

Why is heat capacity important for DES-based chemical and industrial processes?

Heat capacity strongly affects energy requirements in operations such as extraction and separation, where lower values can reduce system energy consumption. DESs are investigated as sustainable solvents, making reliable heat-capacity data important.

How is the DES heat-capacity prediction model built in the study?

The work uses a dataset of 530 DESs with experimental heat-capacity values. Each DES is encoded using quantum-chemical COSMO-RS-based descriptors, and multiple machine learning algorithms are trained and tuned.

Which machine learning model performs best, and how is its accuracy assessed?

The multilayer perceptron (MLP) model is reported as the most successful. Accuracy is evaluated using Average Absolute Relative Deviation (AARD) on training and test sets, showing much lower errors than traditional approaches.

How does the study identify the structural features that control DES heat capacity?

A SHAP (SHapley Additive exPlanations) analysis is applied to interpret the trained models. This identifies the key structural factors within DES components that govern heat capacity.

Aug 5, 2026

Upload by Seraphina