Optimizing Pair Potentials for Efficient Crystallization Kinetics

We present an inverse-design algorithm to optimize an isotropic pair potential for kinetically efficient and thermodynamically stable self-assembly of a target crystal structure. In general, there is often a trade-off between kinetically and thermodynamically optimal pair potentials. Nonetheless, we show that applying a Fisher Information Metric-guided optimization algorithm "on the fly" - that is, concurrently with a simulated self-assembly process - identifies pair potentials that balance these two competing objectives. We specifically focus on assembling a honeycomb lattice, the two-dimensional analog of the diamond lattice, which is challenging to assemble due to the existence of competing crystal polymorphs at certain densities. Importantly, our approach requires no prior knowledge of the competing polymorphs and identifies optimal pair potentials on a time scale comparable to that of the kinetically optimized self-assembly process. Our results also suggest that time-dependent pair potentials proposed by our optimization approach can further enhance the self-assembly kinetics and avoid kinetic traps associated with competing polymorphs.