The ML-MULTIMEM project (Machine Learning-aided Multiscale Modelling Framework for Polymer Membranes) is multidisciplinary synergy between the SKEL The AI lab, Institute of Informatics and Telecommunications and the Institute of Nanoscience and Nanotechnology (INN), NCSR Demokritos that kicked off on Monday 15 November 2021. This endeavour aims at applying Artificial Intelligence (ΑΙ) tools in computer simulations of materials for carbon capture applications. The project is supported by a Marie Skłodowska-Curie Postdoctoral Fellowship that targets early-stage career researchers to promote mobility across Europe and foster the acquisition of new skills and the transfer of knowledge across institutions.
Dr Vangelis Karkaletsis, Director of IIT and Dr George Giannakopoulos Researcher at SKEL | The AI lab along with Dr Niki Vergadou of the Molecular Thermodynamics and Modelling of Materials Laboratory, INN, have joined forces to cooperate. Dr Eleonora Ricci is the postdoctoral researcher awarded the Marie Skłodowska-Curie Postdoctoral Fellowship.
“What impact do you expect this collaboration to have?” was the question set to this multidisciplinary research team.
Dr Vangelis Karkaletsis, Director of IIT and Dr George Giannakopoulos, Researcher, SKEL |The AI Lab, IIT
Collaboration across different scientific domains has two major contributions: foundations and perspectives.
First, it sets the foundation for a common vocabulary and understanding of problems and solutions across the domains. It also highlights the limitations of methods and implies a way forward, oftentimes beyond the individual scientific traditions of each domain. Such a foundation allows knowledge sharing with other teams and acts as a catalyst for the launch of new efforts by other teams within our research center. It also allows the generation of an ecosystem, which can reach beyond the organization itself. We have seen this in action through a co-organized workshop we led in 2020, named “AI in Natural Sciences and Technology”, which drew international interest.
The second contribution is a set of novel perspectives, where scientists can combine expert knowledge from their domain with data-driven (and mostly domain-agnostic) approaches. This essentially means discovering new phenomena through a different viewpoint, but also validating the findings through established know-how. It may also mean the opposite: starting from an established view and knowledge of problems and methods, and leading the next steps through data-aware-analysis and retrospection. In any case, complementarity is wealth in this case.
A third, scientifically secondary but humanely primary side-effect, is the fact that new teams are formed, new friends are made, cultures are intertwined, and the roots of collaboration grow deeper across institutes, organisations, and countries. This is no small achievement and I am very happy to be part of it.
Dr Niki Vergadou, MTMML – INN
The development of modern novel technologies requires the design of new materials with controlled properties in order to address the numerous environmental, economic and societal challenges around the globe. In this direction, a fundamental understanding of the diverse interactions and microscopic mechanisms that give rise to the non-trivial behavior in complex materials is necessary. Molecular simulation methods are particularly efficient and reliable for the study of the underlying molecular mechanisms, the design of new materials and the prediction of materials properties.
Within the framework of ML-MULTIMEM project, hierarchical machine learning-assisted multiscale simulation schemes will be developed as a powerful advancement to the existing modeling strategies. These simulation frameworks will enable a deeper understanding of structure-property-processing relations in macromolecular materials and polymer-based membranes and facilitate a successful and more efficient molecular design of high-performance materials.
Developing reliable multi-scale modeling methods is per se a very difficult task, and, furthermore, invoking artificial intelligence methods preserving the correct underlying physics is a challenging goal. However, the synergy between molecular simulation and modelling with machine learning expertise provides a unique opportunity for the extension of the applicability and generalization of these novel techniques on a systematic basis, which will facilitate and enable their use to a wide range of complex chemical systems utilized in a plethora of applications worldwide. Via the present Marie Skłodowska-Curie Postdoctoral Fellowship and the collaboration with Dr Eleonora Ricci, who is an exceptional young researcher with a very solid scientific and technical background, we will have the chance to form this interdisciplinary team and realize this great vision.