Molecular dynamics study of water anomalies: a comparison of OPC3 and TIP4P/ models

Abstract Context Classical water models are commonly parameterized to reproduce selected experimental properties, with the dielectric constant often treated as a primary target. However, improving dielectric response within rigid, non-polarizable frameworks can redistribute errors and affect the simultaneous description of structural, thermodynamic, and dynamic properties. In this work, we investigate this trade-off by comparing two widely used rigid water models with distinct parameterization strategies: the four-site TIP4P/ ε model, designed to enhance dielectric properties relative to TIP4P/2005, and the three-site OPC3 model, optimized for overall thermodynamic and dynamic performance. By analyzing structural correlations, density and diffusion anomalies, and excess entropy, we show that both models reproduce the qualitative hierarchy of water-like anomalies, while exhibiting systematic quantitative differences linked to their molecular architecture. TIP4P/ ε provides a more accurate description of short-range structure and density-related anomalies, whereas OPC3 displays softer intermediate-range correlations with comparable dynamic behavior. These results demonstrate that anomaly-based analyses offer a sensitive framework for assessing the redistribution of accuracy induced by different force-field parameterization choices and confirm the continued relevance of rigid water models for large-scale simulations of liquid water. Methods Molecular dynamics simulations were performed using the LAMMPS package. Liquid water was modeled using the rigid, non-polarizable OPC3 and TIP4P/ ε force fields, combining Lennard–Jones and Coulombic interactions. Simulations were carried out over broad ranges of temperature and density using Nosé–Hoover thermostat and barostat schemes. Structural properties were analyzed through radial distribution functions, thermodynamic anomalies were assessed from density–temperature relations, and molecular mobility was quantified via diffusion coefficients. Excess entropy was estimated from pair correlation functions to rationalize the coupling between structure and dynamics.