Wang–Landau approach to the simulation of water clusters

Junqi Yin, D. P. LandauMolecular 2018 · DOI

TL;DR

This paper applies the Wang-Landau Monte Carlo method to simulate water clusters using various empirical water potentials. It demonstrates that while some classical models can reasonably reproduce structural and thermodynamic properties of water clusters, they often fall short of ab initio calculations, highlighting the ongoing challenge of accurately modeling water.

What Problem Does This Paper Address?

Water, a seemingly simple molecule, exhibits remarkably complex behavior, especially in its condensed phases and cluster forms. Accurately modeling water's structural and thermodynamic properties is a long-standing challenge in computational chemistry and physics. No single classical or quantum potential can universally describe all aspects of water systems. This difficulty is amplified when simulating water clusters, where the vast number of possible configurations makes it hard to find the true ground state and accurately predict thermodynamic behavior. The paper aims to evaluate how well different empirical water potentials perform in describing these properties for water clusters, using a robust simulation technique.

Methods

The core of this study is the application of the Wang-Landau (WL) Monte Carlo method. This is a sophisticated, iterative, flat-histogram sampling technique designed to overcome the limitations of traditional methods like Metropolis Monte Carlo, which can get trapped in local energy minima. The WL method systematically explores the energy landscape of a system by performing a random walk in energy space, effectively mapping out the density of states (DOS). From the DOS, all thermodynamic properties can, in principle, be calculated.

For simulating water clusters, the authors employed the NPT ensemble (constant number of particles, pressure, and temperature). To prevent cluster evaporation, a pressure constraint was used, which they found to be equivalent to a fixed box size constraint. The WL update process involves several types of moves: local translational and rotational moves for molecules, volume changes to maintain the pressure constraint, and molecular dynamics-based "global" updates. The simulation is divided into two stages: an initial stage to estimate the DOS and identify low-energy configurations, followed by a production stage where the DOS is fixed to collect thermodynamic data.

Datasets

Not discussed in the paper. The paper focuses on simulations using established and a newly proposed empirical water potentials, rather than analyzing experimental datasets.

Key Insights and Ideas

  • Wang-Landau as a powerful tool: The WL method proves effective for exploring the complex energy landscape of water clusters, allowing for a more thorough sampling of configurations and a better estimation of the ground state compared to simpler methods.
  • Classical potentials show promise but limitations: Several classical water models, particularly TIP4P, GCP, and the new E3B3, can reproduce structural properties of water clusters that are in reasonable agreement with ab initio calculations, especially for smaller clusters.
  • Low-temperature behavior reveals differences: While models tend to converge at room temperature (where they are typically parameterized), significant differences emerge in their low-temperature structural and thermodynamic behavior, indicating varying degrees of accuracy in capturing subtle interactions.
  • E3B3's close relation to TIP4P/2005: The explicit 3-body potential (E3B3) shows very similar results to the TIP4P/2005 potential, suggesting that the 3-body interactions may not drastically alter the thermodynamic behavior of this specific model compared to its 2-body counterpart.
  • Hydrogen bonding is a key differentiator: The pair energy distribution analysis highlights differences in how potentials describe hydrogen bonding, with GCP showing a narrower low-energy band, suggesting a more well-defined hydrogen bond.

Key Results

  • The study evaluated six empirical water potentials: SPC/E, TIP3P, TIP4P, TIP4P/2005, GCP, and E3B3.
  • Ground state structures for clusters up to N=15 were compared with ab initio results. TIP4P, GCP, and E3B3 showed good agreement for cluster sizes up to 15.
  • Radial distribution functions (O-O, O-H, H-H) were analyzed at various temperatures (50K, 150K, 300K). At room temperature, most models showed similar distributions, but at lower temperatures, significant differences were observed, with SPC/E and TIP3P diverging more from TIP4P, GCP, and E3B3.
  • Specific heat curves were computed to study melting behavior. TIP4P, GCP, and E3B3 exhibited similar transition curves, while SPC/E and TIP3P showed distinct behaviors.
  • The computational cost of the models was assessed, with GCP being the most expensive due to its polarizable nature.

Limitations and Future Work

  • Agreement with ab initio is partial: While some classical models show good agreement with ab initio calculations for certain properties, they are not universally consistent, indicating that a perfect classical model for water remains elusive.
  • Parameterization focus: The paper notes that these potentials are designed to reproduce macroscopic properties of water at ambient conditions, which might explain why their low-temperature or cluster-specific behavior can differ significantly from more fundamental quantum calculations.
  • Future work could explore:
    • Further investigation into the differences in structural populations at low temperatures.
  • Extending the simulations to larger cluster sizes or different ensembles.
  • Comparing these results with more advanced quantum mechanical methods for a more definitive benchmark.
  • Exploring the impact of induced dipoles and flexible bonds on cluster properties.

Summary updated

Abstract

The Wang–Landau Monte Carlo method [Wang FG, Landau DP. Efficient, multiple-range random walk algorithm to calculate the density of states. Phys Rev Lett. 2001;86:2050-2053; Determining the density of states for classical statistical models: a random walk algorithm to produce a flat histogram. Phys Rev E. 2001;64:056101] has been established as an effective computational approach for studying complex systems. In this paper, we review the properties of 6 popular empirical water potentials, including a recently proposed explicit 3-body potential, obtained by applying the Wang–Landau method to the simulation of water clusters. Comparing to numerous classical and first principle calculations, we show that several classical water models can provide a consistent description of the structural and thermodynamical properties of water clusters partially in agreement of ab initio results, although they were designed to reproduce macroscopic properties of water at ambient conditions.
BibTeX
@inproceedings{yin2018wanglandau,
  author = {Junqi Yin and D. P. Landau},
  title = {{Wang–Landau approach to the simulation of water clusters}},
  booktitle = {Molecular},
  year = {2018},
  doi = {10.1080/08927022.2018.1506119},
}
Figure