Coarse-Grained Artificial Intelligence for Design of Brush Networks

We outline a design strategy based on synergistic combination of theoretical and artificial intelligence tools for encoding mechanical properties of brush networks with three architectural parameters: degrees of polymerization (DP) of network strands, n_x, side chains n_sc, and backbone spacers between side chains, n_g. Implementing a two-layer feedforward artificial neural network (ANN), we take advantage of the coarse-grained representation of chemistry specific characteristics defined by monomer projection length l and excluded volume v, Kuhn length b of bare backbone and side chains, and architecture [n_sc,φ=n_g/(n_g+n_sc), n_x] of brush networks and their equilibrium mechanical properties described by the structural shear modulus G and firmness parameter β. In our approach, a five-dimensional input vector [b/l,v/l³,φ,n_scl/b,n_xl/b], corresponding to a coarse-grained representation of network, is mapped into two-dimensional vector [Gv/β, β] representing network mechanical properties. ANN was trained on data sets for brush networks of poly (n-butyl acrylate), polyisobutylene and poly(dimethyl siloxane) strands and used for synthesis of networks with the nearly identical stress-elongation curves made with different monomers or strand architectures.