Dealing with Architecture-Controlled Bottlebrush Copolymers System using Graph Convolutional Network
발표자
()
초록
내용
In this work, we studied how to deal with architecture-controlled bottlebrush copolymers in solution using Graph Convolutional Network (GCN) based on Dissipative Particle Dynamics (DPD) simulation dataset. Architecture of bottlebrush copolymers was encoded by graphs including connectivity, side chain length, bead types, and interaction parameter of DPD simulation. First, single bottlebrush copolymer chain properties such as radius of gyration, volume of chain and asphericity were predicted using GCN with over 95% accuracy. Second, we used these single chain properties to predict the multi-chain self-assembly behavior in solution using multinomial classification models to learn morphologies of bottlebrush copolymer in selective solvents. With this model, we reproduced and generated the phase diagram of self-assembled morphologies of bottlebrush copolymers having various architectures. Moreover, we are trying to draw the latent space from the graphs to generate new polymer graphs of target morphology.