How Short Is Too Short for the Interactions of a Water Potential? Exploring the Parameter Space of a Coarse-Grained Water Model Using Uncertainty Quantification

Abstract

Coarse-grained models are becoming increasingly popular due to their ability to access time and length scales that are prohibitively expensive with atomistic models. However, as a result of decreasing the degrees of freedom, coarse-grained models often have diminished accuracy, representability, and transferability compared with their finer grained counterparts. Uncertainty quantification (UQ) can help alleviate this challenge by providing an efficient and accurate method to evaluate the effect of model parameters on the properties of the system. This method is useful in finding parameter sets that fit the model to several experimental properties simultaneously. In this work we use UQ as a tool for the evaluation and optimization of a coarse-grained model. We efficiently sample the five-dimensional parameter space of the coarse-grained monatomic water (mW) model to determine what parameter sets best reproduce experimental thermodynamic, structural and dynamical properties of water. Generalized polynomial chaos (gPC) was used to reconstruct the analytical surfaces of density, enthalpy of vaporization, radial and angular distribution functions, and diffusivity of liquid water as a function of the input parameters. With these surfaces, we evaluated the sensitivity of these properties to perturbations of the model input parameters and the accuracy and representability of the coarse-grained models. In particular, we investigated what is the optimum length scale of the water−water interactions needed to reproduce the properties of liquid water with a monatomic model with two-and three-body interactions. We found that there is an optimum cutoff length of 4.3 Å, barely longer than the size of the first neighbor shell in water. As cutoffs deviate from this optimum value, the ability of the model to simultaneously reproduce the structure and thermodynamics is severely diminished.

Estimation of viscous dissipation in nanodroplet impact and spreading

“…The mW model has been proven to accurately reproduce the properties of energetics, density, surface tension of liquid-vapor interface, and short-range structure (hydrogen bond) of liquid water. 27 The cut-off distance used in our simulation is 4.3 Å, which has been proven by Jacobson et al 30 to be the best setting to accurately reproduce the characteristics of real water. Meanwhile, the computational efficiency of the mW model is much higher than that of the other models.…”

Bayesian selection for coarse-grained models of liquid water

“…Furthermore, one should avoid artificially constructed scoring functions that could be biased but rather perform model selection based on rigorous mathematical foundation. In this respect, the Bayesian statistical framework can serve as a powerful tool which has become a popular technique to refine, guide and critically assess the MD models [40][41][42][43][44][45].…”

Uncertainty Quantification at the Molecular–Continuum Model Interface

“…Further research on the reliability of MD predictions involved the analysis of parametric uncertainty on the flux of ions through silica nanopores [22,23]. Jacobson et al [24] applied generalised polynomial chaos to assess the accuracy of a family of coarse-grained water models. A recent study by Salloum et al [25] focused on a two-way coupling of uncertainty across scales with the use of surrogate models for computational efficiency.…”