Machine learned force fields typically require manual construction of training sets consisting of thousands of first principles calculations, which can result in low training efficiency and unpredictable errors when applied to structures not represented in the training set of the model. This severely limits the practical application of these models in systems with dynamics governed by important rare events, such as chemical reactions and diffusion. We present an adaptive Bayesian inference method for automating the training of interpretable, low-dimensional, and multi-element interatomic force fields using structures drawn on the fly from molecular dynamics simulations. Within an active learning framework, the internal uncertainty of a Gaussian process regression model is used to decide whether to accept the model prediction or to perform a first principles calculation to augment the training set of the model. The method is applied to a range of single- and multi-element systems and shown to achieve a favorable balance of accuracy and computational efficiency, while requiring a minimal amount of ab initio training data. We provide a fully open-source implementation of our method, as well as a procedure to map trained models to computationally efficient tabulated force fields.
Strong anion–cation interaction in lithium-salt/ionic liquid electrolytes leads to ionic association that decreases the Li transference number, even causing it to be negative. We show that these interactions can be greatly reduced by adding cyclic ethylene oxide molecules, and we quantitatively examine the effect using rigorous multispecies concentrated solution theory coupled with molecular dynamics simulations. The added molecules, primarily lithium ionophore V also known as 12-crown-4, have high affinity to lithium, therefore disrupting the lithium cation–anion coupling, resulting in a significantly improved transference number. First, we investigate the lithium–anion spatial correlation by studying their clusters and show that the 12-crown-4 ether allows the formation of previously nonexisting positively charged lithium-containing complexes. We then prove that the chelators actively compete with the anion to coordinate lithium ions by showing that the persistence-over-time of a given anion coordination cage decreases when ionophore molecules are added to the system. Last, we report an increase in the lithium transference number for a variety of chemistries as a function of added 12-crown-4 (and another ionophore, 18-crown-6) molecules, and even positive values can be reached. Our results provide a foundation for new design and optimization strategies to reverse the sign of and increase the transference number in highly correlated concentrated electrolytes.
Neural network force field (NNFF) is a method for performing regression on atomic structure–force relationships, bypassing the expensive quantum mechanics calculations that prevent the execution of long ab initio quality molecular dynamics (MD) simulations. However, most NNFF methods for complex multi-element atomic systems indirectly predict atomic force vectors by exploiting just atomic structure rotation-invariant features and network-feature spatial derivatives, which are computationally expensive. Here, we show a staggered NNFF architecture that exploits both rotation-invariant and -covariant features to directly predict atomic force vectors without using spatial derivatives, and we demonstrate 2.2× NNFF–MD acceleration over a state-of-the-art C++ engine using a Python engine. This fast architecture enables us to develop NNFF for complex ternary- and quaternary-element extended systems composed of long polymer chains, amorphous oxide and surface chemical reactions. The rotation-invariant–covariant architecture described here can also directly predict complex covariant vector outputs from local environments, in other domains beyond computational material science.