AI-driven Automated Construction of Block Copolymer Phase Diagrams

Abstract

The self-assembly of block copolymers serves as an effective approach for fabricating various periodic ordered nanostructures. By employing self-consistent field theory (SCFT) to calculate the phase diagrams of block copolymers, one can accurately predict their self-assembly behaviors, thus providing guidance for the fabrication of various novel structures. However, SCFT is highly sensitive to initial conditions because it finds the free energy minima through an iterative process. Consequently, constructing phase diagrams using SCFT typically requires predefined candidate structures based on the experience of researchers. Such experience-dependent strategies often miss some structures and thus result in inaccurate phase diagrams. Recently, artificial intelligence (AI) techniques have demonstrated significant potential across diverse fields of science and technology. By leveraging AI methods, it is possible to reduce reliance on human experience, thereby constructing more robust and reliable phase diagrams. In this work, we demonstrate how to combine AI with SCFT to automatically search for self-assembled structures of block copolymers and construct phase diagrams. Our aim is to realize automatic construction of block copolymer phase diagrams while minimizing reliance on human prior knowledge.