Machine learning-driven generation and screening of potential ionic liquids for cellulose dissolution.

Area of Science:

• Materials Science
• Computational Chemistry
• Chemical Engineering

Background:

• Cellulose processing is hindered by its poor solubility in conventional solvents.
• Ionic liquids show promise for dissolving cellulose but require extensive experimental screening.
• Developing new ionic liquids for cellulose dissolution is a time-consuming, trial-and-error process.

Purpose of the Study:

• To create an efficient computational workflow for identifying novel ionic liquids with high cellulose solubility.
• To generate a large virtual library of potential ionic liquid candidates.
• To accelerate the discovery and development of ionic liquids for cellulose applications.

Main Methods:

• Utilized Monte Carlo tree search and recurrent neural networks to generate a virtual library of billions of ionic liquid candidates.
• Employed machine learning models pre-trained to predict cellulose solubility and melting point.
• Validated promising candidates using the Conductor-like Screening Model for Real Solvents (COSMO-RS).

Main Results:

• Generated a vast virtual molecular library of de novo ionic liquid candidates.
• Screened candidates using predictive machine learning models for solubility and melting point.
• Identified promising ionic liquid candidates through computational screening and validation.

Conclusions:

• The developed workflow significantly accelerates the discovery of ionic liquids for cellulose dissolution.
• The virtual molecular library provides a valuable resource for future theoretical and experimental research.
• This computational approach offers an efficient alternative to traditional experimental methods for developing novel ionic liquids.