Data‐Driven Prediction of Block Copolymer Morphology Using Coarse‐Grained Modeling and Machine Learning

ABSTRACT Block copolymers (BCPs) spontaneously self‐assemble into diverse ordered nano‐ and microstructures that enable applications in nanolithography, drug delivery, and flexible electronics. Predicting these morphologies remains challenging because the self‐assembly process depends on multiple molecular and thermodynamic parameters, including chain architecture, interaction strength, and processing conditions. In this work, we develop an integrated computational framework combining coarse‐grained molecular dynamics (CG‐MD) simulations, automated image‐based microstructure analysis, and supervised machine learning (ML) to predict quantitative morphological descriptors of BCP systems. A large dataset of simulation‐generated morphologies is analyzed using an automated pipeline to extract physically meaningful descriptors, including domain spacing, interface length, thickness, and structural periodicity. ML regression models trained on simulation dataset accurately predict key morphology metrics across a broad parameter space. Feature attribution analysis further shows that the learned descriptor–parameter relationships recover physically expected trends, with interaction strength and chain length emerging as dominant contributors across several global and local morphology descriptors. This framework establishes a scalable strategy for linking simulation‐derived microstructures with data‐driven prediction, enabling accelerated high‐throughput design and interpretation of BCP self‐assembly.