Chemical space is vast,and the space of physical properties derived from it even more so.Despite over a century of physical chemistry measurements,we still lack large,well-organised data sets of key physical propertie...Chemical space is vast,and the space of physical properties derived from it even more so.Despite over a century of physical chemistry measurements,we still lack large,well-organised data sets of key physical properties such as the interfacial tension of surfactant solutions.Access to such data sets is essential to developing the next generation of predictive models for physicochemical properties using modern approaches such as artificial intelligence and machine learning(ML).Thus,we require experimental methods which can efficiently provide large,reproducible and informative data sets.In this work,we developed a robotic,automated pendant drop module for the efficient characterisation of the interfacial properties of surfactant-containing solutions.Provided with surfactant stock solutions,the module measures a surface tension isotherm,from which multiple important physical properties can be determined,including the critical micelle concentration and the maximum surface excess concentration.The module handles critical experimental challenges and decisions such as choosing an appropriate drop volume,detecting failed measurements and selecting concentration points to measure.To obtain a maximally informative dataset,the platform leverages an active learning algorithm combining Bayesian inference and mutual information to dynamically design experiments in response to collected data.We validate the platform by characterising a range of surfactants and demonstrate the effectiveness of its capabilities by mapping the surface tension of binary surfactant mixtures.This module lays the foundation for efficiently and autonomously generating informative datasets of the interfacial properties of surfactant formulations,paving the way to the next generation of ML models for the prediction of formulation properties.展开更多
基金funding from the National Growth Fund project“Big Chemistry”(1420578)funded by the Ministry of Education,Culture and ScienceThis project was also supported by the European Union and the Swiss State Secretariat for Education,Research and Innovation(SERI)under contract numbers 22.00017 and 22.00034(Horizon Europe Research and Innovation Project CORENET).
文摘Chemical space is vast,and the space of physical properties derived from it even more so.Despite over a century of physical chemistry measurements,we still lack large,well-organised data sets of key physical properties such as the interfacial tension of surfactant solutions.Access to such data sets is essential to developing the next generation of predictive models for physicochemical properties using modern approaches such as artificial intelligence and machine learning(ML).Thus,we require experimental methods which can efficiently provide large,reproducible and informative data sets.In this work,we developed a robotic,automated pendant drop module for the efficient characterisation of the interfacial properties of surfactant-containing solutions.Provided with surfactant stock solutions,the module measures a surface tension isotherm,from which multiple important physical properties can be determined,including the critical micelle concentration and the maximum surface excess concentration.The module handles critical experimental challenges and decisions such as choosing an appropriate drop volume,detecting failed measurements and selecting concentration points to measure.To obtain a maximally informative dataset,the platform leverages an active learning algorithm combining Bayesian inference and mutual information to dynamically design experiments in response to collected data.We validate the platform by characterising a range of surfactants and demonstrate the effectiveness of its capabilities by mapping the surface tension of binary surfactant mixtures.This module lays the foundation for efficiently and autonomously generating informative datasets of the interfacial properties of surfactant formulations,paving the way to the next generation of ML models for the prediction of formulation properties.
