From measurement to prediction: accelerating the search for thermally stable ionic liquid

For more than two decades, ionic liquids have been regarded as highly promising solvents for demanding high-temperature applications. Yet one critical challenge has consistently limited their industrial deployment: identifying which ionic liquids can withstand months of operation at temperatures above 200 °C without the need for lengthy and costly experimental campaigns.

At IMIL 2026 – Iberoamerican Meeting on Ionic Liquids (6–10 July 2026) in Salvador, Brazil researchers from CICECO – University of Aveiro (UAv) and the Laboratory of Artificial Chemical Intelligence (LIAC) at EPFL, working on the ILIMITED project, presented two complementary advances that address this challenge from both the experimental and predictive perspectives.

On the experimental side, the team introduced a streamlined thermogravimetric analysis (TGA) protocol capable of recovering the long-term thermal stability metric T₀.₀₁/10h using only two TGA experiments, compared with the five to seven measurements traditionally required. This reduces instrument occupancy to approximately three hours per compound, while maintaining the accuracy needed to reliably rank candidate ionic liquids by thermal stability.

Complementing this work, researchers developed a multi-task graph neural network that predicts long-term thermal stability directly from molecular structure. The model simultaneously learns both short-term degradation behavior (Tonset) and long-term stability (T₀.₀₁/10h), leveraging the physical relationship between these two regimes. This approach is particularly valuable given the limited availability of high-quality long-term stability data, with fewer than 100 reliable measurements currently reported worldwide.

Together, these advances demonstrate how the integration of faster experimental methods and machine learning can help overcome one of the key barriers to deploying ionic liquids in high-temperature industrial processes.

Congratulations to João Fabian, Murilo L. Alcantara, Jan Rittig, Filipe Sosa, Philippe Schwaller and João A. P. Coutinho on their joint effort and valuable contribution!