Force fields determination for molecular dynamics through Chebyshev polynomials quadrature points

Atomistic simulations are a powerful tool to complement experiments, and their accuracy and reliability are ultimately determined by the adopted interaction scheme, which will also dictate the computational expense and hence the limit on the attainable simulation size and timescales. Nowadays, atomistic modeling offers a spectrum of possibilities ranging from a full quantum-mechanical description of the system, but limited to a couple hundreds of atoms, to coarse-grained approaches capable of simulating biomolecules of billions of particles. However, the field of extreme conditions still reprents an open challenge because a fully reactive picture is needed to provide an accurate description of the dynamically changing potential energy surface of the system, but the space and time scales involved impose a serious limit for ab initio simulations. A force matched pairwise reactive interatomic model based on Chebyshev polynomials (ChIMES) has recently shown to retain the accuracy of DFT calculations while guaranteeing the reactive nature of the system in a number of cases. Indeed, force matching techniques need a lot of data, possibly at multiple thermodynamics conditions, in order to yield accurate predictions. In this work, we show how using a cluster approach based on the Chebyshev polynomials quadrature points helps in quickly stabilizing the fitting process, potentially avoiding the problem of highly correlated data and overfitting issues. Furthermore, we show a proof of concept application of how the "Chebyshev quadrature points cluster" can be used as a stand-alone force field, dramatically reducing the number of data required for the determination of the interatomic interactions.