Abstract
One of the many peculiar properties of water is the pronounced deviation of the proton momentum distribution from Maxwell-Boltzmann behavior. This deviation from the classical limit is a manifestation of the quantum mechanical nature of protons. Its extent, which can be probed directly by deep inelastic neutron scattering experiments, gives important insight on the potential of mean force felt by H atoms. The determination of the full distribution of particle momenta, however, is a real tour de force for both experiments and theory, which has led to unresolved discrepancies between the two. In this study, we present comprehensive, fully converged momentum distributions for water at several thermodynamic state points, focusing on the components that cannot be described in terms of a scalar contribution to the quantum kinetic energy, and providing a benchmark that can serve as a reference for future simulations and experiments. In doing so, we also introduce a number of technical developments that simplify and accelerate greatly the calculation of momentum distributions by means of atomistic simulations.
Introduction
Water exhibits numerous anomalous properties that stem from its complex hydrogen-bonding network. Among these is the behavior of its protons, which significantly deviate from classical Maxwell-Boltzmann statistics. This quantum mechanical nature, characterized by zero-point energy and tunneling, is reflected in the proton momentum distribution. Understanding this distribution is critical for mapping the potential of mean force experienced by hydrogen atoms in different phases of water.
Historically, obtaining converged theoretical data and precise experimental measurements via Deep Inelastic Neutron Scattering (DINS) has been challenging. Experiments have occasionally reported "anomalous" non-monotonic shifts in kinetic energy during phase transitions, while Path Integral Molecular Dynamics (PIMD) simulations—the gold standard for such calculations—are notoriously computationally expensive. This work addresses these challenges by introducing a more efficient computational framework and establishing a definitive theoretical benchmark.
Methodology: The Scaled Gradient Estimator
The researchers utilized "Open" Path Integral Molecular Dynamics, where protons are modeled as open-ended polymer chains. To overcome the slow convergence of standard estimators, the study introduces a novel Scaled Gradient (SG) estimator. This virial-like estimator targets the derivative of the end-to-end distribution of the proton path, providing a significantly more stable signal.
As shown in the figure above, the SG estimator reduces statistical error by approximately fivefold compared to standard methods. This efficiency gain allows simulations to reach high precision (statistical error ~2 meV) in just 10 picoseconds of data, whereas previous methods required nanoseconds. To further accelerate the study, the authors employed machine learning, using neural networks trained on high-level density functional theory (B3LYP+D3) and many-body potentials (MBPOL), achieving an overall 20x speedup in wall-clock time.
Results and Discussion: Model Consensus and the Experimental Gap
The study evaluated three distinct water models: q-TIP4P/f, NN-B3LYP+D3, and MBPOL across various states, including supercooled water, hexagonal ice (Ice Ih), and liquid water at room temperature.
A key finding is the remarkable consensus among the high-end models. Despite different underlying physics, all three models produced nearly identical momentum distributions, suggesting a robust theoretical understanding of water's potential energy surface. Furthermore, the simulations showed a simple, monotonic increase in proton kinetic energy with temperature, finding no evidence for the "anomalies" reported in some experimental literature.
However, a persistent 7% discrepancy remains between theory and experiment for room-temperature liquid water. While models and experiments align well for ice and supercooled water, the models consistently predict higher kinetic energy (~158 meV) than current experimental measurements (~146 meV) at 300 K. This gap suggests either subtle physics missing from current models or, more likely, remaining calibration challenges in room-temperature DINS experiments.
Conclusion
By introducing the Scaled Gradient estimator and leveraging machine-learned potentials, this work provides a converged reference for the proton momentum distribution in water. The stability of directional anisotropy across temperature ranges and the consensus between diverse models highlight the maturity of current theoretical frameworks. Future efforts should focus on Ice Ih as a primary benchmark system; its single-crystal nature allows for precise directional anisotropy measurements, which will be essential for resolving the remaining discrepancies between theory and experiment.
Summary updated