Stability and Metastability of Liquid Water in a Machine-Learned Coarse-Grained Model with Short-Range Interactions

Abstract

Coarse-grained water models are ∼100 times more efficient than all-atom models, enabling simulations of supercooled water and crystallization. The machine-learned monatomic model ML-BOP reproduces the experimental equation of state (EOS) and ice− liquid thermodynamics at 0.1 MPa on par with the all-atom TIP4P/2005 and TIP4P/Ice models. These all-atom models were parametrized using high-pressure experimental data and are either accurate for water's EOS (TIP4P/2005) or ice−liquid equilibrium (TIP4P/Ice). ML-BOP was parametrized from temperature-dependent ice and liquid experimental densities and melting data at 0.1 MPa; its only pressure training is from compression of TIP4P/2005 ice at 0 K. Here we investigate whether ML-BOP replicates the experimental EOS and ice−water thermodynamics along all pressures of ice I. We find that ML-BOP reproduces the temperature, enthalpy, entropy, and volume of melting of hexagonal ice up to 400 MPa and the EOS of water along the melting line with an accuracy that rivals that of both TIP4P/2005 and TIP4P/Ice. We interpret that the accuracy of ML-BOP originates from its ability to capture the shift between compact and open local structures to changes in pressure and temperature. ML-BOP reproduces the sharpening of the tetrahedral peak of the pair distribution function of water upon supercooling, and its pressure dependence. We characterize the region of metastability of liquid ML-BOP with respect to crystallization and cavitation. The accessibility of ice crystallization to simulations of ML-BOP, together with its accurate representation of the thermodynamics of water, makes it promising for investigating the interplay between anomalies, glass transition, and crystallization under conditions challenging to access through experiments.

Kinetics and Mechanisms of Pressure-Induced Ice Amorphization and Polyamorphic Transitions in a Machine-Learned Coarse-Grained Water Model

“…For reference, the fastest cooling rate at which simulation cells with 8000 ML-BOP particles spontaneously nucleate and grow ice I in isobaric cooling simulations is 0.2 K ns −1 . 76 The green arrows in Figure 1a overlay these isobaric trajectories to the phase diagram of ML-BOP determined in ref 76. We use the CHILL+ algorithm 82 to assess the presence of ice in the simulation cells, confirming that there is not more than 5% cubic and hexagonal ice in the vitrified configurations at 78 K obtained from these quenching simulations (Figure S1).…”

“…As ML-BOP reproduces quite faithfully the melting line of ice I 76 (Figure 10), we interpret that a large contribution to this discrepancy should be attributed to the extremely fast compression rates of the simulations, that are at least 9 orders of magnitude faster than in experiments. The effect of the fast compression can be seen in the persistence of the gap between the equilibrium melting line T m determined by two-phase coexistence 76 and the one obtained by compression even at temperatures well above T g (Figure 10). We find that not only the amorphization pressure but also the mechanism of PIA depends on the mono-or polycrystallinity of the starting ice sample.…”

“…77 Details of the functional form and the parameters of ML-BOP model are discussed elsewhere. 57,76 We also perform simulations with the monatomic water model mW, 49 based on the Stillinger−Weber potential. 75 The monatomic nature of the mW and ML-BOP models and extremely short-range of the interactions (0.42 nm for mW and 0.355 nm for ML-BOP) makes them ∼100 times faster than all-atom rigid-body water models such as TIP4P/2005.…”

“…We obtain polycrystalline ice through cooling simulations of ML-BOP liquid at 0.1 MPa at a rate of 0.2 K ns −1 as presented elsewhere. 76 This results in stacking disordered polycrystalline ice I that is then energy minimized and used for compression simulations at temperatures ranging from 80 to 260 K to estimate the p Ice→LDA (T) line for polycrystalline ice. The compression rates we explore are 100 and 10 MPa ns −1 .…”

“…mW reproduces the experimental structure of LDA 49 and has a high-density amorphous glass 73 but does not have a LLPT. 62,74 However, mW is based on the Stillinger−Weber 75 form, that lacks the functional flexibility to quantitatively represent the response of the structure of liquid water to pressure and temperature, 61,76 resulting in an underestimation of molar volume of ice and compressibility of the liquid 61 that shifts the melting line and polyamorphic transition toward higher pressures than observed in the experiments. 61,73 Different from the Stillinger−Weber potential, the Tersoff functional form 77 embeds the three-body angular terms into an effective bond order parameter that multiplies the attractive term of the two-body potential.…”