Insights into modeling refractive index of ionic liquids using chemical structure-based machine learning methods.

Area of Science:

• Materials Science
• Computational Chemistry
• Chemical Engineering

Background:

• Ionic liquids (ILs) are gaining prominence for their diverse applications and eco-friendly properties.
• Accurate refractive index prediction is crucial for ILs quality control and characterization.

Purpose of the Study:

• To predict the refractive indices of pure ionic liquids using machine learning.
• To identify key factors influencing refractive index variations in ILs.

Main Methods:

• Developed six chemical structure-based machine learning models: XGBoost, LightGBM, CatBoost, CNN, Ada-DT, and Ada-SVM.
• Utilized a large dataset of 6098 data points from 483 ILs, incorporating chemical substructures, temperature, and wavelength as inputs.
• Introduced wavelength as a novel input parameter for machine learning-based refractive index prediction.

Main Results:

• The CatBoost model achieved the highest accuracy, with an R² of 0.9973 and an average absolute percent relative error (AAPRE) of 0.0545.
• Compared to existing literature, the developed models offer advantages in dataset size and predictive precision.
• Identified the -F substructure as the most influential factor on IL refractive index, and observed that refractive index increases with alkyl chain length in imidazolium-based ILs.

Conclusions:

• Chemical structure-based machine learning models offer a powerful and accurate approach for predicting IL refractive indices.
• These methods provide comprehensive insights into the factors governing IL properties.
• The study highlights the potential of ML for advancing IL research and development.