Data-driven discovery and stability of bulk morphologies in A ( BA ′)3 miktoarm star copolymers

The identification of novel mesophases is an important research theme in both simple and complex block copolymer systems. While self-consistent field theory (SCFT) is a standard tool for finding ordered structures, its predictions of the free energy local minima depend heavily on the initial seed fields and the initial cell parameters, complicating efforts to perform an exhaustive search. In order to reduce the high dimensionality of the space of initial conditions, we apply principal component analysis to a selection of phases as a generative model in tandem with SCFT to efficiently search for novel structures in the A(BA')3 star block copolymer system. Application of this technique resulted in several novel morphologies, including a peanut-shaped micelle structure that has not previously been reported. This peanut micelle phase, which we name the peanut structure, appears to be a sphere/cylinder intermediate phase with an appreciable region of metastability. We optimized the stability of this phase by evaluating a surrogate function trained using kernel ridge regression, which is both computationally inexpensive and generalizable to large formulation spaces. Our methodology provides a powerful tool for predicting novel ordered mesophases of block-copolymer systems in a symmetry-agnostic manner.