Abstract:  
Block polymers serve as an ideal model for probing self-assembly behaviors in soft matter, given their simplicity in thermodynamic representation through self-consistent field theory (SCFT). In this talk, we will explore the application of SCFT in guiding the design of polymer systems, as well as the potential to leverage deep learning techniques to better understand complex phase formation. While the traditional SCFT method excels in making predictions based on known phases, it faces limitations when tasked with discovering novel phases. This issue stems from the inherent need for prior knowledge of the final structure in SCFT simulations. To address this shortcoming, we utilize a deep convolutional generative adversarial network (GAN), trained on SCFT density fields of known phases. The GAN is then deployed to generate new input fields for subsequent SCFT calculations, thereby providing a pathway to uncovering novel phases in block polymers and other types of soft matter. Furthermore, we have developed a convolutional neural network (CNN) classifier using SCFT solutions. This well-generalized model enables accurate phase identification in molecular dynamic simulations for small molecules, which enhances both the efficiency and interpretability in the simulation workflow. Overall, this talk will showcase the value of integrating deep learning with physics-based simulations to enrich our understanding in self-assembled soft matter systems.

Bio
Pengyu Chen is a Ph.D. candidate in materials science at the Department of Chemical Engineering and Materials Science, University of Minnesota. He received his B.S. degree in Chemistry from Shanghai Jiao Tong University in 2019. His thesis research focused on the investigation of zeolite structures through high-resolution transmission electron microscopy and electron diffraction tomography, conducted under the guidance of Prof. Osamu Terasaki at ShanghaiTech University. Subsequently, he became part of Prof. Kevin Dorfman's group at the University of Minnesota, where his focus shifted to computational simulations of self-assembled block polymers. His current work primarily involves understanding the formation of network phases using polymer self-consistent field theory and machine learning techniques.
 

In-Person Location:  Bldg. 440 A105/A106 
Zoom Link:   https://argonne.zoomgov.com/j/1600746525