A physically guided hierarchical parameterization of a rigid three-site water model based on the density anomaly and the dipole moment

Rigid non-polarizable three-point (3P) water models remain essential tools for molecular simulations, but their parameterization continues to be a key challenge. State-of-the-art designs, such as OPC3, often fail to reproduce temperature-dependent properties like the thermal expansion coefficient, a limitation that stems from complex multi-objective optimizations which obscure underlying physical principles. Here, we introduce an alternative paradigm: a hierarchical, physically-guided parameterization that uses the isobaric density anomaly as its primary constraint. This strategy restricts the search for parameters to physically meaningful directions defined by the density anomaly, making the optimization process more transparent. The resulting models achieve competitive overall accuracy while showing superior performance in reproducing the temperature of maximum density and the thermal expansion coefficient, thus resolving persistent deficiencies of existing rigid 3P models. We conclude that this physically-guided parameterization improves the balance between computational feasibility and physical accuracy, offering a promising pathway for the rational design of next-generation water models.