Abstract: I will discuss the structure-property relationships in ordered and disordered soft materials using a wide range of computational methods varying from molecular dynamics simulation, enhanced sampling, and machine learning (ML). In recent years, there have been rapid developments and a wide range of applications of ML in different fields, but it is still quite limited in soft matter research. The plausible reasons could be the lack of interpretability, or missing connections to physical/experimentally measurable properties. I will address the problem of the physical/chemical interpretability of the results of ML methods and their application in understanding structure-property relationships in soft matter systems to design novel materials. First, I'll talk about our new data-driven approach, which utilizes the high-resolution details accessible through molecular dynamics simulation and considers the structural information of individual polymer chains. The method identifies one of the material properties such as the glass transition temperature of polymer melts of semiflexible chains. By combining principal component analysis (PCA) and clustering, we identify glass transition temperature at the asymptotic limit even from relatively short-time trajectories [1]. The method can be applied to a wide range of systems with microscopic/atomistic information. More recently we applied this methodology to all-atom acrylic polymer systems [2]. In the end, I'll talk about the polymorphism [3] in polymer crystals. Overall, I will discuss a general framework of how the data-driven approaches could be applied to identify meaningful states in complex systems.

[1] A. Banerjee, H. Hsu, K. Kremer, O. Kukharenko, ACS Macro Lett. (2023)
[2] A. Banerjee*, A. Iscen, K. Kremer, and O. Kukharenko J. Chem. Phys. (2023)
[3] A. Banerjee, T Bereau, J F Rudzinski (to be submitted in JCTC)

Meeting ID: 995 3988 7607
Passcode: 159509