Using machine learning with atomistic surface and local water features to predict heterogeneous ice nucleation.

Heterogeneous ice nucleation (HIN) has applications in climate science, nanotechnology, and cryopreservation. Ice nucleation on the earth's surface or in the atmosphere usually occurs heterogeneously involving foreign substrates, known as ice nucleating particles (INPs). Experiments identify good INPs but lack sufficient microscopic resolution to answer the basic question: What makes a good INP? We employ molecular dynamics (MD) simulations in combination with machine learning (ML) to address this question. Often, the large amount of computational cost required to cross the nucleation barrier and observe HIN in MD simulations is a practical limitation. We use information obtained from short MD simulations of atomistic surface and water models to predict the likelihood of HIN. We consider 153 atomistic substrates with some surfaces differing in elemental composition and others only in terms of lattice parameters, surface morphology, or surface charges. A range of water features near the surface (local) are extracted from short MD simulations over a time interval (≤300 ns) where ice nucleation has not initiated. Three ML classification models, Random Forest (RF), support vector machine, and Gaussian process classification are considered, and the accuracies achieved by all three approaches lie within their statistical uncertainties. Including local water features is essential for accurate prediction. The accuracy of our best RF classification model obtained including both surface and local water features is 0.89 ± 0.05. A similar accuracy can be achieved including only local water features, suggesting that the important surface properties are largely captured by the local water features. Some important features identified by ML analysis are local icelike structures, water density and polarization profiles perpendicular to the surface, and the two-dimensional lattice match to ice. We expect that this work, with its strong focus on realistic surface models, will serve as a guide to the identification or design of substrates that can promote or discourage ice nucleation.

Influence of pH on Ice Nucleation by Kaolinite: Experiments and Molecular Simulations.

In mixed-phase or ice clouds, ice can be formed through heterogeneous nucleation. A major type of ice-nucleating particle (INP) in the atmosphere are mineral dust particles. For mixed-phase clouds, the pH of water droplets can vary widely and influence ice nucleation by altering the surface of some INPs, including mineral dust. Kaolinite is a commonly occurring clay mineral, and laboratory experiments, as well as molecular dynamics (MD) simulations, have demonstrated its ice-nucleating efficiency at neutral pH. We examine the influence of pH on the ice-nucleating efficiency of kaolinite, in the immersion freezing mode, through both droplet freezing experiments and MD simulations. Droplet freezing experiments using KGa-1b kaolinite samples are reported under both acidic (HNO3 solutions) and basic (NaOH solutions) conditions, covering the measured pH range 0.18-13.26. These experiments show that the ice-nucleating efficiency of kaolinite is not significantly influenced by the presence of acid but is reduced in extremely basic conditions. We report MD simulations aimed at gaining a microscopic understanding of the pH dependence of ice nucleation by kaolinite. The Al(001), Si(001), and three edge surfaces of kaolinite are considered, but ice nucleation was observed only for the Al(001) surface. The hydroxy groups exposed on the Al(001) surface can be deprotonated in a basic solution or dual-protonated in an acidic solution, which can influence ice nucleation efficiency. The protonation state of the Al(001) surface for a particular pH can be estimated using previously measured pKa values. We find that the monoprotonated Al(001) surface expected to be stable at near-neutral pH is the most effective ice-nucleating surface. In MD simulations, the ice nucleation efficiency persists for dual-protonation but decreases significantly with increasing deprotonation, qualitatively consistent with the experimental observations. Taken together, our experimental and MD results for a wide range of pH values support the suggestion that the Al(001) surface may be important for ice nucleation by kaolinite. Additionally, the deprotonation of hydroxy groups on INP surfaces can have a significant effect on their ice-nucleating ability.

Binary salt structure classification with convolutional neural networks: Application to crystal nucleation and melting point calculations.

Convolutional neural networks are constructed and validated for the crystal structure classification of simple binary salts such as the alkali halides. The inputs of the neural network classifiers are the local bond orientational order parameters of Steinhardt, Nelson, and Ronchetti [Phys. Rev. B 28, 784 (1983)], which are derived solely from the relative positions of atoms surrounding a central reference atom. This choice of input gives classifiers that are invariant to density, increasing their transferability. The neural networks are trained and validated on millions of data points generated from a large set of molecular dynamics (MD) simulations of model alkali halides in nine bulk phases (liquid, rock salt, wurtzite, CsCl, 5-5, sphalerite, NiAs, AntiNiAs, and ß-BeO) across a range of temperatures. One-dimensional time convolution is employed to filter out short-lived structural fluctuations. The trained neural networks perform extremely well, with accuracy up to 99.99% on a balanced validation dataset constructed from millions of labeled bulk phase structures. A typical analysis using the neural networks, including neighbor list generation, order parameter calculation, and class inference, is computationally inexpensive compared to MD simulations. As a demonstration of their accuracy and utility, the neural network classifiers are employed to follow the nucleation and crystal growth of two model alkali halide systems, crystallizing into distinct structures from the melt. We further demonstrate the classifiers by implementing them in automated MD melting point calculations. Melting points for model alkali halides using the most commonly employed rigid-ion interaction potentials are reported and discussed.

Analysis of the relative stability of lithium halide crystal structures: Density functional theory and classical models.

All lithium halides exist in the rock salt crystal structure under ambient conditions. In contrast, common lithium halide classical force fields more often predict wurtzite as the stable structure. This failure of classical models severely limits their range of application in molecular simulations of crystal nucleation and growth. Employing high accuracy density functional theory (DFT) together with classical models, we examine the relative stability of seven candidate crystal structures for lithium halides. We give a detailed examination of the influence of DFT inputs, including the exchange-correlation functional, basis set, and dispersion correction. We show that a high-accuracy basis set, along with an accurate description of dispersion, is necessary to ensure prediction of the correct rock salt structure, with lattice energies in good agreement with the experiment. We also find excellent agreement between the DFT-calculated rock salt lattice parameters and experiment when using the TMTPSS-rVV10 exchange-correlation functional and a large basis set. Detailed analysis shows that dispersion interactions play a key role in the stability of rock salt over closely competing structures. Hartree-Fock calculations, where dispersion interactions are absent, predict the rock salt structure only for LiF, while LiCl, LiBr, and LiI are more stable as wurtzite crystals, consistent with radius ratio rules. Anion-anion second shell dispersion interactions overcome the radius ratio rules to tip the structural balance to rock salt. We show that classical models can be made qualitatively correct in their structural predictions by simply scaling up the pairwise additive dispersion terms, indicating a pathway toward better lithium halide force fields.

Simulations of water structure and the possibility of ice nucleation on selected crystal planes of K-feldspar.

Molecular dynamics simulations are employed to investigate the structure of supercooled water (230 K) in contact with the (001), (010), and (100) surfaces of potassium feldspar (K-feldspar) in the microcline phase. Experimentally, K-feldspar and other feldspar minerals are known to be good ice-nucleating agents, which play a significant role in atmospheric science. Therefore, a principal purpose of this work is to evaluate the possibility that the K-feldspar surfaces considered could serve as likely sites for ice nucleation. The (001) and (010) surfaces were selected for study because they are perfect cleavage planes of feldspar, with (001) also being an easy cleavage plane. The (100) surface is considered because some experiments have suggested that it is involved in ice nucleation. Feldspar is modeled with the widely used CLAYFF force field, and the TIP4P/Ice model is employed for water. We do not observe ice nucleation on any of the K-feldspar surfaces considered; moreover, the density profiles and the structure of water near these surfaces do not exhibit any particularly icelike features. Our simulations indicate that these surfaces of K-feldspar are likely not responsible for its excellent ice nucleating ability. This suggests that one must look elsewhere, possibly at water-induced surface rearrangements or some other "defect" structure, for an explanation of ice nucleation by K-feldspar.

Structural behavior of aqueous t-butanol solutions from large-scale molecular dynamics simulations.

Large-scale molecular dynamics simulations are reported for aqueous t-butanol (TBA) solutions. The CHARMM generalized force field (CGenFF) for TBA is combined with the TIP4P/2005 model for water. Unlike many other common TBA models, the CGenFF model is miscible with water in all proportions at 300 K. The main purpose of this work is to investigate the existence and nature of a microheterogeneous structure in aqueous TBA solutions. Our simulations of large systems (128 000 and 256 000 particles) at TBA mole fractions of 0.06 and 0.1 clearly reveal the existence of long-range correlations (>10 nm) that show significant variations on long time scales (∼50 ns). We associate these long-range slowly varying correlations with the existence of supramolecular domainlike structures that consist of TBA-rich and water-rich regions. This structure is always present but continually changing in time, giving rise to long-range slowly varying pair correlation functions. We find that this behavior appears to have little influence on the single particle dynamics; the diffusion coefficients of both TBA and water molecules lie in the usual liquid state regime, and mean square displacements provide no indication of anomalous diffusion. Using our large system simulations, we are able to reliably calculate small angle x-ray scattering and small angle neutron scattering spectra, except at a very low wave vector, and the results agree well with recent experiments. However, this paper shows that simulation of the relatively simple TBA/water system remains challenging. This is particularly true if one wishes to obtain properties such as Kirkwood-Buff factors, or scattering functions at a low wave vector, which strongly depend on the long-range behavior of the pair correlations.

The influence of ion hydration on nucleation and growth of LiF crystals in aqueous solution.

Molecular dynamics (MD) simulations are employed to investigate crystal nucleation and growth in oversaturated aqueous LiF solutions. Results obtained for a range of temperatures provide evidence that the rate of crystal growth is determined by a substantial energy barrier (∼49 kJ mol-1) related to the loss of water from the ion hydration shells. Employing direct MD simulations, we do not observe spontaneous nucleation of LiF crystals at 300 K, but nucleation is easily observable in NVT simulations at 500 K. This contrasts with the NaCl case, where crystal nucleation is directly observed in similar simulations at 300 K. Based on these observations, together with a detailed analysis of ion clustering in metastable LiF solutions, we argue that the ion dehydration barrier also plays a key role in crystal nucleation. The hydration of the relatively small Li+ and F- ions strongly influences the probability of forming large, crystal-like ion clusters, which are a necessary precursor to nucleation. This important factor is not accounted for in classical nucleation theory.

A molecular dynamics investigation of the influence of water structure on ion conduction through a carbon nanotube.

Molecular dynamics simulations are employed to investigate pressure-driven water and ion transport through a (9,9) carbon nanotube (CNT). We consider NaCl solutions modeled with both the TIP3P and TIP4P/2005 water models. Concentrations range from 0.25 to 2.8 mol l-1 and temperatures from 260 to 320 K are considered. We discuss the influences on flow rates of continuum hydrodynamic considerations and molecular structural effects. We show that the flow rate of water, sodium, and chloride ions through the CNT is strongly model dependent, consistent with earlier simulations of pure water conduction. To remove the effects of different water flow rates, and clearly expose the influence of other factors on ion flow, we calculate ion transport efficiencies. Ion transport efficiencies are much smaller for TIP4P/2005 solutions than for those using the TIP3P model. Particularly at lower temperatures, the ion transport efficiencies for the TIP4P/2005 model are small, despite the fact that the nanotube conducts water at a significant rate. We trace the origin of small ion transport efficiencies to the presence of ring-like water structures within the CNT. Such structures occur commonly for the TIP4P/2005 model, but less frequently for TIP3P. The water structure acts to reduce ion "solvation" within the CNT, posing an additional barrier to ion entry and transport. Our results demonstrate that increasing the water structure within the CNT by decreasing the temperature strongly inhibits ion conduction, while still permitting significant water transport.

Crystal structures of model lithium halides in bulk phase and in clusters.

We employ lattice energy calculations and molecular dynamics simulations to compare the stability of wurtzite and rock salt crystal structures of four lithium halides (LiF, LiCl, LiBr, and LiI) modeled using the Tosi-Fumi and Joung-Cheatham potentials, which are models frequently used in simulation studies. Both infinite crystals and finite clusters are considered. For the Tosi-Fumi model, we find that all four salts prefer the wurtzite structure both at 0 K and at finite temperatures, in disagreement with experiments, where rock salt is the stable structure and wurtzite exists as a metastable state. For Joung-Cheatham potentials, rock salt is more stable for LiF and LiCl, but the wurtzite structure is preferred by LiBr and LiI. It is clear that the available lithium halide force fields need improvement to bring them into better accord with the experiment. Finite-size clusters that are more stable as rock salt in the bulk phase tend to solidify as small rock salt crystals. However, small clusters of salts that prefer the wurtzite structure as bulk crystals tend to form structures that have hexagonal motifs, but are not finite-size wurtzite crystals. We show that small wurtzite structures are unstable due to the presence of a dipole and rearrange into more stable, size-dependent structures. We also show that entropic contributions can act in favor of the wurtzite structure at higher temperatures. The possible relevance of our results for simulation studies of crystal nucleation from melts and/or aqueous solutions is discussed.

Comparison of simulation and experimental results for a model aqueous tert-butanol solution.

Molecular dynamics simulations are used to investigate the behavior of aqueous tert-butanol (TBA) solutions for a range of temperatures, using the CHARMM generalized force field (CGenFF) to model TBA and the TIP4P/2005 or TIP4P-Ew water model. Simulation results for the density, isothermal compressibility, constant pressure heat capacity, and self-diffusion coefficients are in good accord with experimental measurements. Agreement with the experiment is particularly good at low TBA concentration, where experiments have revealed anomalies in a number of thermodynamic properties. Importantly, the CGenFF model does not exhibit liquid-liquid demixing at temperatures between 290 and 320 K (for systems of 32 000 molecules), in contrast with the situation for several other common TBA models [R. Gupta and G. N. Patey, J. Chem. Phys. 137, 034509 (2012)]. However, whereas real water and TBA are miscible at all temperatures where the liquid is stable, we observe some evidence of demixing at 340 K and above. To evaluate the structural properties at low concentrations, we compare with both neutron scattering and recent spectroscopic measurements. This reveals that while the CGenFF model is a definite improvement over other models that have been considered, the TBA molecules still exhibit a tendency to associate at low concentrations that is somewhat stronger than that indicated by experiments. Finally, we discuss the range and decay times of the long-range correlations, providing an indication of the system size and simulation times that are necessary in order to obtain reliable results for certain properties.

Simulated conduction rates of water through a (6,6) carbon nanotube strongly depend on bulk properties of the model employed.

We investigate pressure driven flow rates of water through a (6,6) carbon nanotube (CNT) for the TIP3P, SPC/E, and TIP4P/2005 water models. The flow rates are shown to be strongly model dependent, differing by factors that range from ∼6 to ∼2 as the temperature varies from 260 to 320 K, with TIP3P showing the fastest flow and TIP4P/2005 the slowest. For the (6,6) CNT, the size constraint allows only single-file conduction for all three water models. Hence, unlike the situation for the larger [(8,8) and (9,9)] CNTs considered in our earlier work [L. Liu and G. N. Patey, J. Chem. Phys. 141, 18C518 (2014)], the different flow rates cannot be attributed to different model-dependent water structures within the nanotubes. By carefully examining activation energies, we trace the origin of the model discrepancies for the (6,6) CNT to differing rates of entry into the nanotube, and these in turn are related to differing bulk mobilities of the water models. Over the temperature range considered, the self-diffusion coefficients of the TIP3P model are much larger than those of TIP4P/2005 and those of real water. Additionally, we show that the entry rates are approximately inversely proportional to the shear viscosity of the bulk liquid, in agreement with the prediction of continuum hydrodynamics. For purposes of comparison, we also consider the larger (9,9) CNT. In the (9,9) case, the flow rates for the TIP3P model still appear to be mainly controlled by the entry rates. However, for the SPC/E and TIP4P/2005 models, entry is no longer the rate determining step for flow. For these models, the activation energies controlling flow are considerably larger than the energetic barriers to entry, due in all likelihood to the ring-like water clusters that form within the larger nanotube.

Fluctuations and local ice structure in model supercooled water.

Large-scale simulations (up to 32,000 molecules) are used to analyze local structures and fluctuations for the TIP4P/2005 and TIP5P water models, under deeply supercooled conditions, near previously proposed liquid-liquid critical points. Bulk freezing does not occur in our simulations, but correlations between molecules with local ice-like structure (ice-like molecules) are strong and long ranged (∼4 nm), exceeding the shortest dimension of smaller simulation cells at the lowest temperatures considered. Correlations between ice-like molecules decay slowly at low temperature, on the order of a hundred nanoseconds. Local ice-like structure is strongly correlated with highly tetrahedral liquid structure at all times, both structures contribute to density fluctuations, and to the associated anomalous scattering. For the TIP4P/2005 and TIP5P models, we show that the apparent spontaneous liquid-liquid phase separations, recently reported [T. Yagasaki, M. Matsumoto, and H. Tanaka, Phys. Rev. E 89, 020301 (2014)] for small rectangular simulation cells below the proposed critical points, exhibit strong system size dependence and do not occur at all in the largest systems we consider. Furthermore, in the smaller rectangular systems where layers of different densities do occur, we find that the appearance of a region of low density is always accompanied simultaneously by an excess of local ice density, with no separation in time. Our results suggest that the density differences observed in direct simulations for the two models considered here are likely due to long-range correlations between ice-like molecules and do not provide strong evidence of liquid-liquid phase separation.

Melting point trends and solid phase behaviors of model salts with ion size asymmetry and distributed cation charge.

The melting point trends of model salts composed of coarse grain ions are examined using NPT molecular dynamics simulations. The model salts incorporate ion size asymmetry and distributed cation charge, which are two common features in ionic liquids. A series of single-phase and two-phase simulations are done at set temperatures with 50 K intervals for each salt, and the normal melting point is estimated within 50 K. The melting point trends are then established relative to a charge-centered, size symmetric salt with a normal melting point between 1250 K and 1300 K. We consider two sets of size asymmetric salts with size ratios up to 3:1; the melting point trends are different in each set. The lowest melting point we find is between 450 K and 500 K, which is a reduction of over 60% from the charge-centered, size symmetric case. In both sets, we find diversity in the solid phase structures. For all size ratios with small cation charge displacements, the salts crystallize with orientationally disordered cations. When the partial cation charge is far enough off-center in salts with ion size ratios near 1:1, the salts can become trapped in glassy states and have underlying crystal structures that are orientationally ordered. At ion size ratios near 3:1, the salts with large cation charge displacements show premelting transitions at temperatures as low as 300 K. After the premelting transition, these salts exist either as fast ion conductors, where the smaller anions move through a face centered cubic (fcc) cation lattice, or as plastic crystals, where ion pairs rotate on a fcc lattice.

Simulations of water transport through carbon nanotubes: how different water models influence the conduction rate.

The conduction rate of water through (8,8) and (9,9) carbon nanotubes at 300 K and a pressure difference of 220 MPa is investigated using molecular dynamics simulations. The TIP3P, SPC/E, and TIP4P/2005 water models are considered. The pressure-driven flow rate is found to be strongly model dependent for both nanotubes. The fastest model (TIP3P) has a flow rate that is approximately five times faster than the slowest (TIP4P/2005). It is shown that the flow rate is significantly influenced by the structure taken on by the water molecules confined in the nanotube channels. The slower models, TIP4P/2005 and SPC/E, tend to favor stacked ring arrangements, with the molecules of a ring moving together through the nanotube, in what we term a "cluster-by-cluster" conduction mode. Confined TIP3P water has a much weaker tendency to form ring structures, and those that do form are fragile and break apart under flow conditions. This creates a much faster "diffusive" conduction mode where the water molecules mainly move through the tube as individual particles, rather than as components of a larger cluster. Our results demonstrate that water models developed to describe the properties of bulk water can behave very differently in confined situations.

How distributed charge reduces the melting points of model ionic salts.

Using coarse grain model ions and NPT molecular dynamics simulations, we determine melting point trends across a set of salts beginning with a system closely akin to the restricted primitive model. Redistributing the cation charge in salts with size-symmetric, monovalent, spherical ions can reduce the melting temperature by up to 50% compared to the charge-centered case. Displacing the charge from the ion center reduces the enthalpy of the liquid more than that of the solid resulting in a lower melting point. Upon cooling from the liquid phase, the model salts considered either crystallize as orientationally-disordered CsCl solids, or become trapped in glassy states, depending on the amount and extent of the cation charge redistribution. For the latter case, we find an orientationally-ordered crystal structure with space group 111 (P(42m)) underlying the glassy states. The structural and dynamical properties of both the solid and liquid phases of the model salts are discussed in some detail.

Structure and aggregation in model tetramethylurea solutions.

The structure of model aqueous tetramethylurea (TMU) solutions is investigated employing large-scale (32,000, 64,000 particles) molecular dynamics simulations. Results are reported for TMU mole fractions, X(t), ranging from infinite dilution up to 0.07, and for two temperatures, 300 and 330 K. Two existing force fields for TMU-water solutions are considered. These are the GROMOS 53A6 united-atom TMU model combined with SPC/E water [TMU(GROMOS-UA)/W(SPC/E)], and the more frequently employed AMBER03 all-atom force field for TMU combined with the TIP3P water model [TMU(AMBER-AA)/W(TIP3P)]. It is shown that TMU has a tendency towards aggregation for both models considered, but the tendency is significantly stronger for the [TMU(AMBER-AA)/W(TIP3P)] force field. For this model signs of aggregation are detected at X(t) = 0.005, aggregation is a well established feature of the solution at X(t) = 0.02, and the aggregates increase further in size with increasing concentration. This is in agreement with at least some experimental studies, which report signals of aggregation in the low concentration regime. The TMU aggregates exhibit little structure and are simply loosely ordered, TMU-rich regions of solution. The [TMU(GROMOS-UA)/W(SPC/E)] model shows strong signs of aggregation only at higher concentrations (X(t) ≳ 0.04), and the aggregates appear more loosely ordered, and less well-defined than those occurring in the [TMU(AMBER-AA)/W(TIP3P)] system. For both models, TMU aggregation increases when the temperature is increased from 300 to 330 K, consistent with an underlying entropy driven, hydrophobic interaction mechanism. At X(t) = 0.07, the extra-molecular correlation length expected for microheterogeneous solutions has become comparable with the size of the simulation cell for both models considered, indicating that even the systems simulated here are sufficiently large only at low concentrations.

Understanding electrofreezing in water simulations.

Molecular dynamics simulations are used to investigate why external electric fields promote the freezing of liquid water models. It is shown that the melting point of water at a pressure of 1 bar increases significantly when water is polarized by a uniform field. Fields of 1 V/nm and 2 V/nm increase the melting point by 24 K and 44 K, respectively. The increased melting point is mainly due to the favorable interaction of near perfectly polarized cubic ice with the applied field. For a fixed temperature, we demonstrate that the size of the critical ice nucleus decreases with field strength, mostly because the melting point, and hence the true degree of supercooling, is increasing with field. On simulation timescales, ice nucleation is observed at ∼40 K below the field-dependent melting point, independent of the particular value of the field applied. Indeed, we find that even quite highly polarized liquid water retains the characteristic local structures, and the related anomalous properties of water. Our results are obviously relevant to the mechanism of heterogeneous ice nucleation by local surface fields. Local fields will effectively increase the degree of supercooling of locally polarized liquid, decreasing the size of the critical nucleus in the region influenced by the field, hence facilitating ice nucleation.

The Brucella suis IbpA heat-shock chaperone is not required for virulence or for expression of the VirB type IV secretion system VirB8 protein.

UNLABELLED: Brucella suis, facultative intracellular bacterial pathogen of mammals, and Agrobacterium tumefaciens, a plant pathogen, both use a VirB type IV secretion system (T4SS) to translocate effector molecules into host cells. HspL, an α-crystalline-type small heat-shock protein, acts as a chaperone for the Agrobacterium VirB8 protein, an essential component of the VirB system. An Agrobacterium mutant lacking hspL is attenuated due to a misfunctional T4SS. We have investigated whether IbpA (BRA0051), the Brucella HspL homologue, plays a similar role. Unlike HspL, IbpA does not interact with VirB8, and an IbpA mutant shows full virulence and no defect in VirB expression. These data show that the Brucella α-crystalline-type small heat-shock protein IbpA is not required for Brucella virulence. SIGNIFICANCE AND IMPACT OF STUDY: Many bacteria use type IV secretion systems (T4SS), multi-protein machines, to translocate DNA and protein substrates across their envelope. Understanding how T4SS function is important as they play major roles in the spread of plasmids carrying antibiotic resistance and in pathogenicity. In the plant pathogen Agrobacterium tumefaciens, HspL, an α-crystalline-type small heat-shock protein, acts as a chaperone for the essential type IV secretion system component VirB8. Here, we show that this is not the case for all T4SS; in the zoonotic pathogen Brucella suis, IbpA, the protein most related to HspL, does not play this role.

An analysis of fluctuations in supercooled TIP4P/2005 water.

Large-scale, long-time molecular dynamics simulations are used to investigate fluctuations in the TIP4P/2005 water model in the supercooled region (240-190 K). Particular attention is focused in the vicinity of a previously reported liquid-liquid critical point [J. L. F. Abascal and C. Vega, J. Chem. Phys. 133, 234502 (2010)]. Water is viewed as an equimolar binary mixture with "species" defined based on a local tetrahedral order parameter. A Bhatia-Thornton fluctuation analysis is used to show that species concentration fluctuations couple to density fluctuations and completely account for the anomalous increase in the structure factor at small wave number observed under supercooled conditions. Although we find that both concentration and density fluctuations increase with decreasing temperature along the proposed critical isochore, we cannot confirm the existence of a liquid-liquid critical point. Our simulations suggest that the parameters previously reported are not a true liquid-liquid critical point and we find no evidence of two-phase coexistence in its vicinity. It is shown that very long simulations (on the order of 8 µs for 500 molecules) are necessary to obtain well converged density distributions for deeply supercooled water and this is especially important if one is seeking direct evidence of a two-phase region.

Ice nucleation by electric surface fields of varying range and geometry.

Molecular dynamics simulations are employed to show that electric field bands acting only over a portion of a surface can function as effective ice nuclei. Field bands of different geometry (rectangular, triangular, and semicircular cross sectional areas are considered) all nucleate ice, provided that the band is sufficiently large. Rectangular bands are very efficient if the width and thickness are ≳0.35 nm, and ≳0.15 nm, respectively, and the necessary dimensions are comparable for other geometries. From these simulations we also learn more about the ice nucleation and growth process. Careful analysis of different systems reveals that ice strongly prefers to grow at (111) planes of cubic ice. This agrees with an earlier theoretical deduction based on considerations of water-ice interfacial energies. We find that ice nucleated by field bands usually grows as a mixture of cubic and hexagonal ice, consistent with other simulations of ice growth, and with experiment. This contrasts with simulations carried out with nucleating fields that span the entire surface area, where cubic ice dominates, and hexagonal layers are very rarely observed. We argue that this discrepancy is a simulation artifact related to finite sample size and periodic boundary conditions.