Abstract
The use of an effective intermolecular potential often involves a compromise between accurate, complex functional forms and more tractable simple representations. To study this choice in detail, we systematically derive coarse-grained isotropic pair potentials that accurately reproduce the oxygen-oxygen radial distribution function of the TIP4P-Ew water model at state points over density ranges from 0.88 to 1.30 g/cm³ and temperature ranges from 235 to 310 K. Although by construction these effective potentials correctly represent the isothermal compressibility of TIP4P-Ew water, they do not accurately resolve other thermodynamic properties such as the virial pressure, the internal energy, or thermodynamic anomalies. Because at a given state point the pair potential that reproduces the pair structure is unique, we have therefore explicitly demonstrated that it is impossible to simultaneously represent the pair structure and several key equilibrium thermodynamic properties of water with state-point dependent radially symmetric pair potentials. We argue that such representability problems are related to, but different from, more widely acknowledged transferability problems and discuss in detail the implications this has for the modeling of water and other liquids by coarse-grained potentials. Nevertheless, regardless of thermodynamic inconsistencies, the state-point dependent effective potentials for water do generate structural and dynamical anomalies.
Introduction
In molecular dynamics, the pursuit of computational efficiency often necessitates a reduction in system complexity. Coarse-graining (CG) is a common strategy where multiple atoms are mapped onto a single "bead" or interaction site. For water, the most common CG approach involves representing the entire molecule as a single isotropic (spherical) particle. While this significantly accelerates simulations, it raises fundamental questions about what physical information is lost during the simplification process.
Most research in coarse-graining focuses on transferability—the ability of a model developed at one state point (e.g., room temperature) to function at another. However, this study addresses the more foundational issue of representability: whether a simplified model can accurately capture all relevant physical properties of the reference system even at a single, fixed state point.
Methodology
The researchers utilized the TIP4P-Ew water model as a high-fidelity reference. They mapped a comprehensive grid of state points, covering densities from 0.88 to 1.30 g/cm³ and temperatures from 235 K (supercooled) to 310 K.
To derive the coarse-grained potentials, the team employed Structure Inversion via Empirical Potential Structural Refinement (EPSR). This numerical procedure reverse-engineers a potential that perfectly reproduces the oxygen-oxygen radial distribution function (RDF) of the reference model. A critical theoretical pillar of this work is Henderson’s Uniqueness Theorem, which states that for a given structure, there exists only one unique pair potential that can produce it. This theorem implies that if a structure-matching potential fails to predict pressure or energy, no other radially symmetric pair potential can succeed.
The Representability Trap
The study reveals a fundamental "no free lunch" scenario in molecular modeling. By forcing the CG model to match the two-body structure (the RDF) of water, the resulting potential becomes fixed. The authors found that these unique potentials consistently failed to represent other thermodynamic properties:
- Thermodynamic Inconsistency: The CG models showed significant deviations in internal energy and virial pressure. In some cases, the virial pressure exhibited unphysical trends, such as decreasing as density increased.
- Structural Deficiencies: While the two-body RDF was matched perfectly, the three-body correlations were not. The CG models displayed a peak in bond angles at 60°, a hallmark of simple liquid packing that contradicts the tetrahedral coordination characteristic of real water.
- Energy Mismatch: The simplified spherical beads cannot account for the directional energy contributions of hydrogen bonding, leading to a failure in capturing the system's total internal energy.
Successes in Anomalous Behavior
Despite the quantitative failures in pressure and energy, the coarse-grained models demonstrated a surprising ability to capture water's qualitative anomalies. Specifically, the models reproduced the dynamic anomalies of water, such as the non-monotonic behavior of the self-diffusion coefficient. In water, diffusion can increase upon compression within certain density ranges; the CG models successfully mirrored these minima and maxima, suggesting that the structural information encoded in the RDF is sufficient to trigger these complex dynamical signatures.
Discussion and Future Directions
The "representability problem" identified here suggests that isotropic pair potentials are fundamentally limited. The loss of angular, directional information inherent in hydrogen bonding cannot be fully compensated for by a simple spherical potential if one aims to match both structure and thermodynamics.
To overcome these hurdles, the authors suggest several paths forward:
- Density-Dependent Terms: Incorporating a "one-body" potential term that depends on density might allow researchers to correct pressure and energy without disturbing the structural fit.
- Advanced Fitting: Utilizing Bayesian ensemble approaches or adaptive resolution schemes to find an optimal compromise between competing physical properties.
- Non-Isotropic Models: Moving toward models that account for molecular orientation, though this increases computational cost.
In conclusion, while coarse-graining remains a powerful tool for extending the reach of molecular simulations, researchers must remain cognizant that a model perfectly tuned to structure may be fundamentally incapable of representing the underlying thermodynamics of complex liquids like water.
Summary updated