A molecular dynamics study of water nucleation using the TIP4P/2005 model

Homogeneous water vapor condensation with a deep neural network potential model

Molecular-level nucleation has not been clearly understood due to the complexity of multi-body potentials and the stochastic, rare nature of the process. This work utilizes molecular dynamics (MD) simulations, incorporating a first-principles-based deep neural network (DNN) potential model, to investigate homogeneous water vapor condensation. The nucleation rates and critical nucleus sizes predicted by the DNN model are compared against commonly used semi-empirical models, namely extended simple point charge (SPC/E), TIP4P, and OPC, in addition to classical nucleation theory (CNT). The nucleation rates from the DNN model are comparable with those from the OPC model yet surpass the rates from the SPC/E and TIP4P models, a discrepancy that could mainly arise from the overestimated bulk free energy by SPC/E and TIP4P. The surface free energy predicted by CNT is lower than that in MD simulations, while its bulk free energy is higher than that in MD simulations, irrespective of the potential model used. Further analysis of cluster properties with the DNN model unveils pronounced variations of O-H bond length and H-O-H bond angle, along with averaged bond lengths and angles that are enlarged during embryonic cluster formation. Properties such as cluster surface free energy and liquid-to-vapor density transition profiles exhibit significant deviations from CNT assumptions.

Abnormal condensation of water vapour at ambient temperature

The homogeneous condensation of water vapor at ambient temperature is studied using molecular dynamics simulation. We reveal that there is a droplet size at the nanoscale where water droplets can be stabilized in the condensation process. Our simulations show that the growth of water droplets is dominated by collision and coagulation between small water droplets after nucleation. This process is found to be accompanied by exceptionally fast evaporation such that droplet growth is balanced by evaporation when water droplets grow to a critical size, approximately 12.5 Å in radius, reaching a stable size distribution. The extremely high evaporation rate is attributed to the curvature dependence of surface tension. Surface tension shows a significant decrease with decreasing droplet size below 20 Å, which causes the total free energy of nanoscaled water droplets to rise after collision and coagulation. Consequently, water droplets have to shrink via fast evaporation. The curvature dependence of surface tension is related to the dielectric ordering of water molecules near the surface of water droplets. Owing to fast evaporation, secondary condensation occurs, and many small water clusters form, ultimately exhibiting a bimodal distribution of water-droplet size.

Variation of Surface Propensity of Halides with Droplet Size and Temperature: The Planar Interface Limit

The radial number density profiles of halide and alkali ions in aqueous clusters with equimolar radius ≲1.4 nm, which correspond to ≲255 H2O molecules, have been extensively studied by computations. However, the surface abundance of Cl-, Br-, and I- relative to the bulk interior in these smaller clusters may not be representative of the larger systems. Indeed, here we show that the larger the cluster is, the lower the relative surface abundance of chaotropic halides is. In droplets with an equimolar radius of ≈2.45 nm, which corresponds to ≈2000 H2O molecules, the polarizable halides show a clear number density maximum in the droplet's bulk-like interior. A similar pattern is observed in simulations of the aqueous planar interface with halide salts at room temperature. At elevated temperature the surface propensity of Cl- decreases gradually, while that of I- is partially preserved. The change in the chaotropic halide location at higher temperatures than the room temperature may considerably affect photochemical reactivity in atmospheric aerosols, vapor-liquid nucleation and growth mechanisms, and salt crystallization via solvent evaporation. We argue that the commonly used approach of nullifying parameters in a force field in order to find the factors that determine the ion location does not provide transferable insight into other force fields.

Properties of water and argon clusters developed in supersonic expansions

Using adiabatic molecular dynamics coupled with the fluid dynamics equations, we model nucleation in an expanding beam of water vapor and argon on a microsecond scale. The size distribution of clusters, their temperature, and pickup cross sections in dependence on velocity are investigated and compared to the geometric cross sections and the experiment. The clusters are warmer than the expanding gas because of the time scale of relaxation processes. We also suggest that their translational and rotational kinetic energies are modified due to evaporative cooling. The pickup cross sections determined for the final clusters using molecules of the same kind increase with decreasing velocity, still obeying the (a+bN1/3)2 law.

Nucleation parameters of SPC/E and TIP4P/2005 water vapor measured in NPT molecular dynamics simulations

Nucleation rates for droplet formation in water vapor are measured in molecular dynamics (MD) simulations of SPC/E and TIP4P/2005 water by monitoring individual nucleation events. The nucleation process is simulated in the NPT ensemble to evaluate the steady-state nucleation rate in accordance with the assumptions of classical nucleation theory (CNT). Nucleation rates measured between 300 and 425 K for the SPC/E model, and between 325 and 475 K for the TIP4P/2005 model, agree with the CNT predictions roughly within the standard deviation of the MD measurements of the nucleation rates.

Conical Shape Fluctuations Determine the Rate of Ion Evaporation and the Emitted Cluster Size Distribution from Multicharged Droplets

The ion evaporation mechanism (IEM) is perceived to be a major pathway for disintegration of multi-ion charged droplets found in atmospheric and sprayed aerosols. However, the precise mechanism of IEM and the effect of the nature of the ions in the emitted cluster size distribution have not yet been established despite its broad use in mass spectrometry and atmospheric chemistry over the past half century. Here, we present a systematic study of the emitted ion cluster distribution in relation to their spatial distribution in the parent droplet using atomistic modeling. It is found that in the parent droplet, multiple kosmotropic and weakly polarizable chaotropic ions (Cs⁺) are buried deeper within the droplet than polarizable chaotropic ions (Cl^-, I^-). This differentiation in the ion location is only captured by a polarizable model. It is demonstrated that the emitted cluster size distribution is determined by dynamic conical deformations and not by the equilibrium ion depth within the parent droplet as the IEM models assume. Critical factors that determine the cluster size distribution such as the charge sign asymmetry that have not been considered in models and in experiments are presented. We argue that the existing IEM analytical models do not establish a clear difference between IEM and Rayleigh fission. We propose a shift in the existing view for IEM from the equilibrium properties of the parent droplet to the chemistry in the conical shape fluctuations that serve as the centers for ion emission. Consequently, chemistry in the conical fluctuations may also be a key element to explain charge states of macromolecules in mass spectrometry and may have potential applications in catalysis due to the electric field in the conical region.

Comparison of molecular dynamics simulations of water with neutron and X-ray scattering experiments

The atomistic structure and dynamics obtained from molecular dynamics (MD) simulations with the example of TIP3P (rigid and flexible) and TIP4P/2005 (rigid) water is compared to neutron and X-ray scattering data at ambient conditions. Neutron and X-ray diffractograms are calculated from the simulations for four isotopic substitutions as well as the incoherent intermediate scattering function for neutrons. The resulting curves are compared to each other and to published experimental data. Differences between simulated and measured intermediate scattering functions are quantified by fitting an analytic model to the computed values. The sensitivity of the scattering curves to the parameters of the MD simulations is demonstrated on the example of two parameters, bond length and angle.

Molecular Dynamics of Heterogeneous Systems on GPUs and Their Application to Nucleation in Gas Expanding to a Vacuum

Expansion of water vapor through a small orifice to a vacuum produces liquid or frozen clusters which in the experiment serve as model particles for atmospheric aerosols. Yet, there are controversies about the shape of these clusters, suggesting that the nucleation process is not fully understood. Such questions can be answered by molecular dynamics simulations; however, they require microsecond-scale runs with thousands of molecules and accurate energy conservation. The available highly parallel codes typically utilize domain decomposition and are inefficient for heterogeneous systems as clusters in a dilute gas. In this work, we present an implementation of molecular dynamics on graphics processing units based on the Verlet list and apply it to several systems for which experimental data are available. We reproduce sufficiently sized clusters but not the experimentally observed clusters of irregular shape.

Thermodynamics and kinetics of crystallization in deeply supercooled Stillinger-Weber silicon

We study the kinetics of crystallization in deeply supercooled liquid silicon employing computer simulations and the Stillinger-Weber three-body potential. The free energy barriers to crystallization are computed using umbrella sampling Monte Carlo simulations and from unconstrained molecular dynamics simulations using a mean first passage time formulation. We focus on state points that have been described in earlier work [S. Sastry and C. A. Angell, Nat. Mater. 2, 739 (2003)] as straddling a liquid-liquid phase transition (LLPT) between two metastable liquid states. It was argued subsequently [Ricci et al., Mol. Phys. 117, 3254 (2019)] that the apparent transition is due to the loss of metastability of the liquid state with respect to the crystalline state. The presence of a barrier to crystallization for these state points is therefore of importance to ascertain, which we investigate, with due attention to ambiguities that may arise from the choice of order parameters. We find a well-defined free energy barrier to crystallization and demonstrate that both umbrella sampling and mean first passage time methods yield results that agree quantitatively. Our results thus provide strong evidence against the possibility that the liquids at state points close to the reported LLPT exhibit slow, spontaneous crystallization, but they do not address the existence of a LLPT (or lack thereof). We also compute the free energy barriers to crystallization at other state points over a broad range of temperatures and pressures and discuss the effect of changes in the microscopic structure of the metastable liquid on the free energy barrier heights.

Formation free energy of an i-mer at spinodal

In statistical mechanics, the formation free energy of an i-mer can be understood as the Gibbs free energy change in a system consisting of pure monomers after and prior to the formation of the i-mer. For molecules interacting via Lennard-Jones potential, we have computed the formation free energy of a Stillinger i-mer [F. H. Stillinger, J. Chem. Phys. 38, 1486 (1963)] and a ten Wolde-Frenkel (tWF) [P. R. ten Wolde and D. Frenkel, J. Chem. Phys. 109, 9901 (1998)] i-mer at spinodal at reduced temperatures from 0.7 to 1.2. It turns out that the size of a critical Stillinger i-mer remains finite and its formation free energy is on the order of k_BT, and the size of a critical tWF i-mer remains finite and its formation free energy is even higher. This can be explained by Binder's theory [K. Binder, Phys. Rev. A 29, 341 (1984)] that for a system, when approaching spinodal, if the Ginzburg criterion is not satisfied, a gradual transition will take place from nucleation to spinodal decomposition, where the free-energy barrier height is on the order of k_BT.

On the transferability of ion parameters to the TIP4P/2005 water model using molecular dynamics simulations

Countless molecular dynamics studies have relied on available ion and water force field parameters to model aqueous electrolyte solutions. The TIP4P/2005 model has proven itself to be among the best rigid water force fields, whereas many of the most successful ion parameters were optimized in combination with SPC/E, TIP3P, or TIP4P/Ew water. Many researchers have combined these ions with TIP4P/2005, hoping to leverage the strengths of both parameter sets. To assess if this widely used approach is justified and to provide a guide in selecting ion parameters, we investigated the transferability of various commonly used monovalent and multivalent ion parameters to the TIP4P/2005 water model. The transferability is evaluated in terms of ion hydration free energy, hydration radius, coordination number, and self-diffusion coefficient at infinite dilution. For selected ion parameters, we also investigated density, ion pairing, chemical potential, and mean ionic activity coefficients at finite concentrations. We found that not all ions are equally transferable to TIP4P/2005 without compromising their performance. In particular, ions optimized for TIP3P water were found to be poorly transferable to TIP4P/2005, whereas ions optimized for TIP4P/Ew water provided nearly perfect transferability. The latter ions also showed good overall agreement with experimental values. The one exception is that no combination of ion parameters and water model considered here was found to accurately reproduce experimental self-diffusion coefficients. Additionally, we found that cations optimized for SPC/E and TIP3P water displayed consistent underpredictions in the hydration free energy, whereas anions consistently overpredicted the hydration free energy.

One-dimensional water nanowires induced by electric fields

Self-aggregation of water vapour molecules under external electric fields is systemically investigated by using molecular dynamics simulations. It is found that small water clusters aggregate into one-dimensional water nanowires along the electric field direction. The electric field strength plays a crucial role in tuning the nanowire structure. Under relatively weak electric fields such as E = 0.1 V Å^-1, square and pentagonal prism-like structures are preferred; when intermediate strength electric fields are applied (E = 1.0 V Å^-1), water nanowires featuring a disordered mixture of four-, five- and six-membered rings are formed; and an open ordered structure which is reminiscent of two-dimensional (2D) ice is observed when the field strength becomes very high (E > 3.0 V Å^-1). Bond parameter analysis based on density-functional theory calculations shows that the electric field affects anisotropically the conformation of water molecules as well as the hydrogen-bond properties. Along the electric field, the H-O bond is stretched and the hydrogen bond shrinks with field strength in contrast to the changes perpendicular to the electric field. As a result, the hydrogen bonding is enhanced along the electric field. Under very high electric fields, the anisotropic hydrogen-bond network opens up via breaking of the bonds perpendicular to the electric field and ultimately relaxes into a loose quasi-2D ordered network.

On the time required to freeze water

By using the seeding technique the nucleation rate for the formation of ice at room pressure will be estimated for the TIP4P/ICE model using longer runs and a smaller grid of temperatures than in the previous work. The growth rate of ice will be determined for TIP4P/ICE and for the mW model of water. Although TIP4P/ICE and mW have a similar melting point and melting enthalpy, they differ significantly in the dynamics of freezing. The nucleation rate of mW is lower than that of TIP4P/ICE due to its higher interfacial free energy. Experimental results for the nucleation rate of ice are between the predictions of these two models when obtained from the seeding technique, although closer to the predictions of TIP4P/ICE. The growth rate of ice for the mW model is four orders of magnitude larger than for TIP4P/ICE. Avrami's expression is used to estimate the crystallization time from the values of the nucleation and growth rates. For mW the minimum in the crystallization time is found at approximately 85 K below the melting point and its value is of about a few ns, in agreement with the results obtained from brute force simulations by Moore and Molinero. For the TIP4P/ICE the minimum is found at about 55 K below the melting point, but its value is about ten microseconds. This value is compatible with the minimum cooling rate required to avoid the formation of ice and obtaining a glass phase. The crossover from the nucleation controlled crystallization to the growth controlled crystallization will be discussed for systems of finite size. This crossover could explain the apparent discrepancy between the values of J obtained by different experimental groups for temperatures below 230 K and should be considered as an alternative hypothesis to the two previously suggested: internal pressure and/or surface freezing effects. A maximum in the compressibility was found for the TIP4P/ICE model in supercooled water. The relaxation time is much smaller than the crystallization time at the temperature at which this maximum occurs, so this maximum is a real thermodynamic feature of the model. At the temperature of minimum crystallization time, the crystallization time is larger than the relaxation time by just two orders of magnitude.

Water nucleation in helium, methane, and argon: A molecular dynamics study

Nucleation of highly supersaturated water vapor in helium, methane, and argon carrier gases at 350 K was investigated using molecular dynamics simulations. Nucleation rates obtained from the mean first passage time (MFPT) method are typically one order of magnitude lower than those from the Yasuoka and Matsumoto method, which can be attributed to the overestimation of the critical cluster size in the MFPT method. It was found that faster nucleation will occur in carrier gases that have better thermalization properties such that latent heat is removed more efficiently. These thermalization properties are shown to be strongly dependent on the molecular mass and Lennard-Jones (LJ) parameters. By varying the molecular mass, for unaltered LJ parameters, it was found that a heavier carrier gas removes less heat although it has a higher collision rate with water than a lighter carrier. Thus, it was shown that a clear distinction between water vapor-carrier gas collisions and water cluster-carrier gas collisions is indispensable for understanding the effect of collision rates on thermalization. It was also found that higher concentration of carrier gas leads to higher nucleation rate. The nucleation rates increased by a factor of 1.3 for a doubled concentration and by almost a factor of two for a tripled concentration.

Crystal Nucleation in Liquids: Open Questions and Future Challenges in Molecular Dynamics Simulations

The nucleation of crystals in liquids is one of nature's most ubiquitous phenomena, playing an important role in areas such as climate change and the production of drugs. As the early stages of nucleation involve exceedingly small time and length scales, atomistic computer simulations can provide unique insights into the microscopic aspects of crystallization. In this review, we take stock of the numerous molecular dynamics simulations that, in the past few decades, have unraveled crucial aspects of crystal nucleation in liquids. We put into context the theoretical framework of classical nucleation theory and the state-of-the-art computational methods by reviewing simulations of such processes as ice nucleation and the crystallization of molecules in solutions. We shall see that molecular dynamics simulations have provided key insights into diverse nucleation scenarios, ranging from colloidal particles to natural gas hydrates, and that, as a result, the general applicability of classical nucleation theory has been repeatedly called into question. We have attempted to identify the most pressing open questions in the field. We believe that, by improving (i) existing interatomic potentials and (ii) currently available enhanced sampling methods, the community can move toward accurate investigations of realistic systems of practical interest, thus bringing simulations a step closer to experiments.

Effects of ensembles on methane hydrate nucleation kinetics

By performing molecular dynamics simulations to form a hydrate with a methane nano-bubble in liquid water at 250 K and 50 MPa, we report how different ensembles, such as the NPT, NVT, and NVE ensembles, affect the nucleation kinetics of the methane hydrate. The nucleation trajectories are monitored using the face-saturated incomplete cage analysis (FSICA) and the mutually coordinated guest (MCG) order parameter (OP). The nucleation rate and the critical nucleus are obtained using the mean first-passage time (MFPT) method based on the FS cages and the MCG-1 OPs, respectively. The fitting results of MFPT show that hydrate nucleation and growth are coupled together, consistent with the cage adsorption hypothesis which emphasizes that the cage adsorption of methane is a mechanism for both hydrate nucleation and growth. For the three different ensembles, the hydrate nucleation rate is quantitatively ordered as follows: NPT > NVT > NVE, while the sequence of hydrate crystallinity is exactly reversed. However, the largest size of the critical nucleus appears in the NVT ensemble, rather than in the NVE ensemble. These results are helpful for choosing a suitable ensemble when to study hydrate formation via computer simulations, and emphasize the importance of the order degree of the critical nucleus.

Effect of surface free energies on the heterogeneous nucleation of water droplet: a molecular dynamics simulation approach

Heterogeneous nucleation of water droplet on surfaces with different solid-liquid interaction intensities is investigated by molecular dynamics simulation. The interaction potentials between surface atoms and vapor molecules are adjusted to obtain various surface free energies, and the nucleation process and wetting state of nuclei on surfaces are investigated. The results indicate that near-constant contact angles are already established for nano-scale nuclei on various surfaces, with the contact angle decreasing with solid-liquid interaction intensities linearly. Meanwhile, noticeable fluctuation of vapor-liquid interfaces can be observed for the nuclei that deposited on surfaces, which is caused by the asymmetric forces from vapor molecules. The formation and growth rate of nuclei are increasing with the solid-liquid interaction intensities. For low energy surface, the attraction of surface atoms to water molecules is comparably weak, and the pre-existing clusters can depart from the surface and enter into the bulk vapor phase. The distribution of clusters within the bulk vapor phase becomes competitive as compared with that absorbed on surface. For moderate energy surfaces, heterogeneous nucleation predominates and the formation of clusters within bulk vapor phase is suppressed. The effect of high energy particles that embedded in low energy surface is also discussed under the same simulation system. The nucleation preferably initiates on the high energy particles, and the clusters that formed on the heterogeneous particles are trapped around their original positions instead of migrating around as that observed on smooth surfaces. This feature makes it possible for the heterogeneous particles to act as fixed nucleation sites, and simulation results also suggest that the number of nuclei increases monotonously with the number of high energy particles. The growth of nuclei on high energy particles can be divided into three sub-stages, beginning with the formation of a wet-spot, increase of contact angle with near-constant contact line, and finally growth with constant contact angle. The growth rate of nuclei also increases with the size of high energy particles.

Limit of metastability for liquid and vapor phases of water

We report the limits of superheating of water and supercooling of vapor from Monte Carlo simulations using microscopic models with configurational enthalpy as the order parameter. The superheating limit is well reproduced. The vapor is predicted to undergo spinodal decomposition at a temperature of Tspvap=46±10 °C (0 °C≪Tspvap≪100 °C) under 1 atm. The water-water network begins to form at the supercooling limit of the vapor. Three-dimensional water-water and cavity-cavity unbroken networks are interwoven at critically superheated liquid water; if either network breaks, the metastable state changes to liquid or vapor.

Size-Dependent Surface Free Energy and Tolman-Corrected Droplet Nucleation of TIP4P/2005 Water

Classical nucleation theory is notoriously inaccurate when using the macroscopic surface free energy for a planar interface. We examine the size dependence of the surface free energy for TIP4P/2005 water nanodroplets (radii ranging from 0.7 to 1.6 nm) at 300 K with the mitosis method, that is, by reversibly splitting the droplets into two subclusters. We calculate the Tolman length to be -0.56 ± 0.09 Å, which indicates that the surface free energy of water droplets that we investigated is 5-11 mJ/m(2) greater than the planar surface free energy. We incorporate the computed Tolman length into a modified classical nucleation theory (δ-CNT) and obtain modified expressions for the critical nucleus size and barrier height. δ-CNT leads to excellent agreement with independently measured nucleation kinetics.

Molecular dynamics study of carbon nanotube as a potential dual-functional inhibitor of HIV-1 integrase

HIV-1 integrase (IN) plays an important role in integrating viral DNA into human genome, which has been considered as the drug target for anti-AIDS therapy. The appearance of drug-resistance mutants urgently requires novel inhibitors that act on non-active site of HIV-1 IN. Nanoparticles have such unique geometrical and chemical properties, which inspires us that nanoparticles like nanotubes may serve as better HIV-1 IN inhibitors than the conventional inhibitors. To test this hypothesis, we performed molecular dynamics (MD) simulation to study the binding of a carbon nanotube (CNT) to a full-length HIV-1 IN. The results showed that the CNT could stably bind to the C-terminal domain (CTD) of HIV-1 IN. The CNT also induced a domain-shift which disrupted the binding channel for viral DNA. Further MD simulation showed that a HIV-1 IN inhibitor, 5ClTEP was successfully sealed inside the uncapped CNT. These results indicate that the CNT may serve as a potential dual-functional HIV-1 IN inhibitor, not only inducing conformation change as an allosteric inhibitor but also carrying small-molecular inhibitors as a drug delivery system.