Bayesian optimization-guided closed-loop morphology optimization of block copolymer blends via synchrotron X-ray scattering

The utility of block copolymers (BCPs) in next-generation nanofabrication and optoelectronics is strictly governed by the long-range ordering and precise periodicity of their microphase-separated domains. However, navigating their multi-dimensional processing space to maximize this structural fidelity can be difficult. Conventional optimization relies on Edisonian trial-and-error, which is resource-intensive, slow, and often fails to identify the global optimum. Here, we address this challenge using a closed-loop morphology-optimization workflow that integrates synchrotron grazing-incidence small-angle X-ray scattering (GISAXS), automated scattering data analysis, and Gaussian process-driven Bayesian optimization (GP-BO). Applied to the swelling-induced assembly of poly(styrene-b-2-vinylpyridine) (S2VP) blends, the BO-guided closed-loop workflow explores the two-dimensional processing space defined by polymer blending ratio and solvent uptake. By jointly optimizing domain spacing and peak sharpness as a measure of structural ordering, the workflow identified target-satisfying conditions using approximately 30 experiments, more than two orders of magnitude fewer than required for uniform coverage at the stated resolution. Overall, we developed a constraint-aware acquisition strategy tailored to morphology optimization, in which domain spacing is treated as a feasibility constraint while structural ordering is optimized within the feasible region. This task-specific formulation differs from applying a standard GP-BO workflow and enables efficient navigation of a coupled landscape containing both monotonic and highly nonlinear responses.