Abstract:Computational studies of liquid water and its phase transition into vapor have traditionally been performed using classical water models. Here we utilize the Deep Potential methodology -- a machine learning approach -- to study this ubiquitous phase transition, starting from the phase diagram in the liquid-vapor coexistence regime. The machine learning model is trained on ab initio energies and forces based on the SCAN density functional which has been previously shown to reproduce solid phases and other properties of water. Here, we compute the surface tension, saturation pressure and enthalpy of vaporization for a range of temperatures spanning from 300 to 600 K, and evaluate the Deep Potential model performance against experimental results and the semi-empirical TIP4P/2005 classical model. Moreover, by employing the seeding technique, we evaluate the free energy barrier and nucleation rate at negative pressures for the isotherm of 296.4 K. We find that the nucleation rates obtained from the Deep Potential model deviate from those computed for the TIP4P/2005 water model, due to an underestimation in the surface tension from the Deep Potential model. From analysis of the seeding simulations, we also evaluate the Tolman length for the Deep Potential water model, which is (0.091 $\pm$ 0.008) nm at 296.4 K. Lastly, we identify that water molecules display a preferential orientation in the liquid-vapor interface, in which H atoms tend to point towards the vapor phase to maximize the enthalpic gain of interfacial molecules. We find that this behaviour is more pronounced for planar interfaces than for the curved interfaces in bubbles. This work represents the first application of Deep Potential models to the study of liquid-vapor coexistence and water cavitation.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to study the liquid - gas equilibrium and cavitation rate of water using the Deep Potential model. Specifically, the author uses a machine - learning - based method to simulate the phase - change process of water from liquid to gas, especially the cavitation phenomenon under negative - pressure conditions. Traditional computational studies usually rely on classical water models, while this paper adopts a new method - the Deep Potential model, which is trained based on the SCAN functional in density functional theory (DFT) to describe the behavior of water molecules more accurately. ### Main research contents: 1. **Liquid - gas equilibrium properties**: The surface tension, saturation pressure and vaporization enthalpy at different temperatures (from 300 to 600 K) were calculated and compared with experimental results and the classical TIP4P/2005 model. 2. **Cavitation rate**: Through the seeding technique, the free - energy barrier and nucleation rate were evaluated under isothermal conditions of 296.4 K. 3. **Interface properties**: The orientation distribution of water molecules at the liquid - gas interface was analyzed, and it was found that hydrogen atoms tend to point to the gas phase to maximize the enthalpy gain of interface molecules. ### Key findings: - **Surface tension**: The surface tension calculated by the Deep Potential model is lower than the experimental value and the results of the TIP4P/2005 model, which may be due to the model underestimating the surface tension. - **Nucleation rate**: The nucleation rate predicted by the Deep Potential model is significantly higher than that of the TIP4P/2005 model, mainly because the former has a lower surface tension. - **Tolman length**: At 296.4 K, the Tolman length of the Deep Potential model is (0.091 ± 0.008) nm. - **Molecular orientation**: Water molecules show a preferential orientation at the liquid - gas interface, and hydrogen atoms tend to point to the gas phase. This behavior is more obvious on a planar interface than on the curved surface of a bubble. ### Research significance: This work applies the Deep Potential model to the study of liquid - gas coexistence and cavitation of water for the first time, demonstrating the potential of machine - learning methods in simulating complex phase - change processes. By comparing with classical models and experimental data, the effectiveness and limitations of the Deep Potential model are verified, providing a direction for further improvement of the model.

A Deep Potential model for liquid-vapor equilibrium and cavitation rates of water

Phase Diagram of a Deep Potential Water Model

The Phase Diagram of a Deep Potential Water Model

Toward High-level Machine Learning Potential for Water Based on Quantum Fragmentation and Neural Networks

Vapor-liquid equilibrium of water with the MB-pol many-body potential

Ab Initio Validation on the Connection Between Atomistic and Hydrodynamic Description to Unravel the Ion Dynamics of Warm Dense Matter

Bridging the Gap: Unraveling the Role of Nano-Gas Nuclei in the Non-Equilibrium Water-Vapor Phase Transition

Modeling Liquid Water by Climbing Up Jacob's Ladder in Density Functional Theory Facilitated by Using Deep Neural Network Potentials

A deep potential model with long-range electrostatic interactions

Thermodynamics of Water and Ice from a Fast and Scalable First-Principles Neuroevolution Potential

Signatures of a liquid–liquid transition in an ab initio deep neural network model for water

Machine learning potentials for complex aqueous systems made simple

Mesoscopic Approach for Nanoscale Liquid-Vapor Interfacial Statics and Dynamics

Many-body interactions and deep neural network potentials for water

EMD SIMULATION FOR POTENTIAL AND FORCE AT LIQUID-VAPOR INTERFACE SYSTEM COMPOSED OF WATER MOLECULES

A fresh look at the vibrational and thermodynamic properties of liquids within the soft potential model

Water at Positive and Negative Pressures

Liquid-Vapor Coexistence and Spontaneous Evaporation at Atmospheric Pressure of Common Rigid Three-Point Water Models in Molecular Simulations

Transferable Water Potentials Using Equivariant Neural Networks

Phase behavior of metastable water from large-scale simulations of a quantitative accurate model: The liquid-liquid critical point

An Equation of State For The TIP4P/2005 Model of Water Including Negative Pressures