A Molecular Mechanism of Ice Nucleation on Model AgI Surfaces - The

Sep 25, 2014 - Coatings to prevent frost. Ricardo Losada , Stefan Holberg , Jeanette M. D. Bennedsen , Klavs Kamuk , Finn Nielsen. Journal of Coatings...
1 downloads 8 Views 5MB Size
Article pubs.acs.org/JPCB

A Molecular Mechanism of Ice Nucleation on Model AgI Surfaces Stephen A. Zielke, Allan K. Bertram, and Grenfell N. Patey* Department of Chemistry, University of British Columbia, Vancouver, British Columbia, Canada V6T 1Z1 S Supporting Information *

ABSTRACT: Heterogeneous ice nucleation at solid surfaces is important in many physical systems including the Earth’s atmosphere. AgI is one of the best ice nucleating agents known; however, why AgI is such an effective ice nucleus is unclear. Using molecular dynamics simulations, we show that a good lattice match between ice and a AgI surface is insufficient to predict the ice nucleation ability of the surface. Seven faces modeled to represent surfaces of both β-AgI and γ-AgI, each having a good lattice match with hexagonal and/or cubic ice, are considered, but ice nucleation is observed for only three. Our model simulations clearly show that the detailed atomistic structure of the surface is of crucial importance for ice nucleation. For example, when AgI is cleaved along certain crystal planes two faces result, one with silver ions and the other with iodide ions exposed as the outermost layer. Both faces have identical lattice matches with ice, but in our simulations ice nucleation occurred only at silver exposed surfaces. Moreover, although hexagonal ice is often the only polymorph of ice considered in discussions of heterogeneous ice nucleation, cubic ice was frequently observed in our simulations. We demonstrate that one possible mechanism of ice nucleation by AgI consists of particular AgI surfaces imposing a structure in the adjacent water layer that closely resembles a layer that exists in bulk ice (hexagonal or cubic). Ice nucleates at these surfaces and grows almost layer-by-layer into the bulk.

I. INTRODUCTION Ice can nucleate in two ways: homogeneously, without a solid surface, or heterogeneously, on a solid surface.1 Ice does not nucleate in the atmosphere until −38 °C unless in the presence of ice nuclei.1−3 Heterogeneous nucleation can occur at much warmer temperatures, and is the mechanism of ice nucleation in a large fraction of ice containing clouds.1,2 Heterogeneous ice nucleation also initiates a large fraction of precipitation in the atmosphere,4 and can influence climate.5 Particles that are known to be good heterogeneous ice nuclei include mineral dusts, certain biological particles, and silver iodide.1,3 However, why some particles function as good ice nuclei while others do not is not understood. The absence of a good molecular-level understanding of heterogeneous ice nucleation means that researchers cannot predict, a priori, the ice nucleation ability of an unstudied substrate. This absence of microscopic understanding can lead to uncertainties when modeling ice nucleation in the atmosphere.6,7 Silver iodide is one of the best heterogeneous ice nucleating substances known, inducing ice formation at temperatures as high as −3 °C.1 Silver iodide is sometimes used as a cloud seeding agent.8 Early researchers often assumed that the good ice nucleation activity of silver iodide was due to the small lattice parameter mismatch between hexagonal ice and silver iodide surfaces.1,9 Subsequent work, however, has called into question the assumption that a small lattice mismatch is indicative of ice nucleation ability.10,11 We do note that ice nucleation at carbon surfaces has been observed12,13 in simulation studies employing the mW water model.14 However, the authors explicitly remark12 that the effect is not due to templating of the ice structure by the carbon surface. Ice © XXXX American Chemical Society

nucleation influenced by a kaolinite surface has also been reported,15 but attributed to the influence of order imposed by the surface in the adjacent water layers, evident in water density profiles, rather than to any particular “match” between ice and the underlying surface. Additionally, in the kaolinite case the simulation results strongly depended on system size, and ice nucleation could not be achieved for larger systems. An alternative explanation of heterogeneous ice nucleation involves active sites, with some property that leads to ice nucleation. At present, the importance of lattice match between hexagonal ice and silver iodide for ice nucleation remains unclear. To improve our microscopic understanding of heterogeneous ice nucleation on silver iodide surfaces, we have carried out detailed molecular dynamics (MD) simulations of ice nucleation on different faces of a rigid lattice model based on silver iodide. The insights gained from this investigation may be applicable to other highly effective ice nuclei such as certain aliphatic alcohol monolayers and proteins from the bacteria Pseudomonas syringae. The microscopic knowledge and understanding obtained from our simulations should prove useful for designing and engineering surfaces that either inhibit or enhance ice formation, and in predicting a priori the ice nucleation ability of a particular substrate. This is of importance in atmospheric science, as well as in many other areas of current research. Special Issue: Branka M. Ladanyi Festschrift Received: August 25, 2014 Revised: September 24, 2014

A

dx.doi.org/10.1021/jp508601s | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

II. MODELS AND SIMULATION METHOD To ensure that our results are not highly dependent on the selection of a particular model, two different water models are considered in this work. These are the six-site (melting point 289 ± 3 K)16,17 and TIP4P/Ice (melting point 270 ± 3 K)18,19 models. Both are rigid water models consisting of sites that interact with other water molecules and with the silver iodide lattice through Lennard-Jones (LJ) and/or Coulombic terms. Since both models gave similar results, only those for the sixsite model are discussed in the main text. Results for TIP4P/Ice are given in the Supporting Information. Unless otherwise stated, all simulation results discussed below are for a supercooling of 20 K, with respect to the melting temperature of the water model employed. Silver iodide is modeled as a rigid lattice of LJ sites with partial positive and negative charges located on the silver and iodide ions, respectively. The LJ parameters (ϵAgO = 2.289 kJ/ mol, σAgO = 0.317 nm, ϵIO = 2.602 kJ/mol, σIO = 0.334 nm, and for six-site water ϵAgH = 0.920 kJ/mol, σAgH = 0.195 nm, ϵIH = 1.046 kJ/mol, σIH = 0.212 nm) and partial charges (±0.6e) were taken from earlier work of Hale and Kiefer.20 We omit the polarization terms included in the force field proposed by Hale and Kiefer because the water models we employ have been parametrized to describe the properties of bulk water without explicitly including molecular polarizability. Additionally, fully including nonadditive polarization terms is not computationally feasible for the large systems and long simulation times needed to investigate ice nucleation and growth. As discussed below (sections 1 and 2), we did extensively test the sensitivity of our results to the electrostatics of the AgI−water interaction by varying the magnitude of the AgI partial charges from zero to 1e, which revealed surprisingly little sensitivity to the actual charges employed. Specifically, the silver exposed β-AgI(0001) surface nucleated ice with charges on AgI ranging in magnitude from zero to 0.8e, while the sliver exposed γ-AgI(001) face nucleated ice with charges from 0.2e to 0.6e, suggesting that our main observations are not strongly dependent on details of the electrostatic contribution to the force field. The geometry of the simulation cell (Supporting Information, Figure S15) is analogous to that employed in previous work.21,22 Two mirrored AgI slabs are used such that any unphysical, long-range electric fields expected due to lattice truncation cancel exactly and do not polarize the water molecules distant from the surface (Figures S16 and S17). The AgI slabs are located 5 nm apart, with 4610−5300 water molecules placed between them, depending on the face under consideration and the associated XY dimensions (>5.5 nm) of the cell (Supporting Information, Table S1). The equilibrium liquid densities, ρ0, as estimated from the middle region of the water density profiles, ranged between between 0.928 and 0.966 g/mL (Table S1). The silver iodide slabs employed in the simulations consisted of four AgI layers with the positions of the silver and iodide ions based on experimental structures.23−25 Simulations were also performed with slabs consisting of two and eight AgI layers, and no significant differences were noted. All MD simulations were carried out under NVT conditions employing the GROMACS 4.5.526 software package with a Nosé−Hoover thermostat,27,28 and particle mesh Ewald electrostatics.29 Bond lengths and angles were constrained with the LINCS algorithm.30 A time step of 2 fs was used in all simulations. Systems were equilibrated at 300 K for 1 ns before

being cooled to 20 K below the model melting point over an additional 1 ns. Simulations were then run at the supercooled temperature for times ranging between 27 ns to more than 0.5 μs (Supporting Information, Tables S2−S5), depending on the system.

III. RESULTS Seven different silver iodide surfaces were considered (Table 1). The different surfaces were generated by cleaving either Table 1. Summary of the AgI Faces Considered, the Associated Mismatches with Ih and Ic, and the Polymorph(s) of Ice Formed mismatch (%) AgI polymorph

AgI surface

Ih

Ic

ice formed

β

Ag exposed (0001) I exposed (0001) Ag+I exposed (101̅0) Ag exposed (111) I exposed (111) Ag exposed (001) I exposed (001)

1.55 1.55 2.03 1.77 1.77 41.61 41.61

2.00 2.00 13.45 2.22 2.22 2.04 2.04

Ih+Ic none none Ih+Ic none Ic none

γ

hexagonal AgI (referred to as β-silver iodide) or face centered cubic (FCC) AgI (referred to as γ-silver iodide). Both forms of AgI can exist under atmospheric conditions. If we cleave a βAgI(0001) plane, a γ-AgI(111) plane, or a γ-AgI(001) plane, two faces are produced, one with sliver ions exposed and one with iodide ions exposed. We refer to these two faces as silver exposed and iodide exposed, respectively. Cleaving a βAgI(101̅0) plane produces two identical faces, so this face does not have silver or iodide exposed versions. We refer to this surface as the silver+iodide exposed β-AgI(1010̅ ) face. All surfaces considered have a small mismatch with hexagonal ice and/or cubic ice (Table 1). The lattice mismatch can be quantified with the equation1 aAgI − a ice 100 δ (%) = a ice (1) where δ is the lattice mismatch, and aice and aAgI are the lattice constants of ice and the silver iodide surfaces, respectively. Although previous researchers often only considered the lattice mismatch between hexagonal ice and AgI surfaces,1,9,31,32 here we consider the lattice mismatch with both hexagonal ice and cubic ice, for reasons that will become clear below. Hexagonal ice, Ih, is the stable form of ice under atmospheric conditions and cubic ice, Ic, is a metastable form of ice. Both phases of ice can be constructed by stacking bilayers of water formed of hexagonal rings in the chair conformation. The phases differ in how the bilayers are stacked. Ih has an ABAB··· stacking resulting in a hexagonal unit cell, whereas, Ic has an ABCABC··· stacking giving a FCC lattice.33,34 The crystal structures are such that the Ih(0001) and Ic(111) faces are almost identical and can connect with little lattice strain.35 Even though all AgI surfaces considered have a small lattice mismatch with Ih and/or Ic, ice formation was only observed for three of the surfaces (Table 1, last column). For the three surfaces that nucleated ice, ice nucleated, and grew nearly as soon as the liquid was supercooled by 20 K (Supporting Information, Figure S1). B

dx.doi.org/10.1021/jp508601s | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

nucleate ice (Supporting Information, Table S4). The highest temperature observed for ice nucleation was 281 K for the sixsite model and 271 K for TIP4P/Ice. For the six-site model, 281 K corresponds to a supercooling of ∼8 K, and 271 K is within the error bars of the reported normal melting point of TIP4P/Ice. The small supercooling required shows that the silver exposed β-AgI(0001) face is a very effective ice nucleus, and is in good qualitative agreement with experimental observations of ice nucleation by silver iodide, that show ice nucleation at just 3 K below melting.1 Snapshots for the silver exposed γ-AgI(111) surface are shown in Supporting Information, Figure S2. The results for this face are very similar to those for the silver exposed βAgI(0001) surface shown in Figure 1A−C. Both Ih(0001) and Ic(111) planes grow parallel to the silver iodide surface. Figure 1D−F shows snapshots of ice growth on the silver exposed γ-AgI(001) surface. For this face, ice Ic is exclusively formed with the Ic(001) plane growing parallel to the surface. Only ice Ic is produced in these simulations because only Ic has a face that can properly match this particular face of γ-AgI (Table 1). Important conclusions can be reached from Table 1, Figure 1, and Supporting Information, Figure S2. First, although hexagonal ice is often the only polymorph of ice considered when discussing heterogeneous ice nucleation on silver iodide surfaces,1,9,31 cubic ice frequently was observed in the simulations, and cubic ice nucleated exclusively on one of the surfaces of silver iodide. Second, contrary to the earlier suggestion, the lattice mismatch between ice and a surface of silver iodide is insufficient to predict the ice nucleation ability of the surface. For example, some silver iodide surfaces that had a small lattice mismatch with Ih and/or Ic did not nucleate ice. A similar conclusion has been reached for other types of surfaces including clays.15,21,22,32,39−41 On the other hand, ice nucleation is highly effective on some AgI surfaces that have a small lattice mismatch with Ic and/or Ih. We conclude that a small lattice mismatch is a useful prerequisite for ice nucleation on AgI surfaces, but other requirements at the atomistic structure level are needed to make an efficient ice nucleus. We note that very recently Fraux and Doye42 have reported simulations of ice nucleation by a model β-AgI(0001) surface. Their surface and water models differed somewhat from those employed here. Of particular note, AgI was represented by a nonrigid lattice model, and the TIP4P/2005 water potential43 was used. Interestingly, the results obtained for their model are very similar to our findings for the β-AgI(0001) surface. Specifically, ice nucleation readily occurred for the silver exposed face, but was not observed for the iodide exposed case. It is worth remarking that the results summarized in Table 1 and those of Fraux and Doye42 do not mean that iodide exposed faces can never nucleate ice, perhaps at a larger supercooling or on longer time scales. However, they do demonstrate that the silver exposed faces are at least much more effective ice nucleating agents. A. Why Do Some Model AgI Surfaces Nucleate Ice and Others Not? 1. Comparison of Simulations for Ag Exposed and I Exposed β-AgI(0001). To better understand why some AgI surfaces nucleate ice and others do not, even though they all have a small lattice mismatch with at least one ice polymorph, we compare the microscopic details of the water−surface interaction for the silver-exposed β-AgI(0001) and the iodide exposed β-AgI(0001) faces (Figure 2). Both faces have identical lattice mismatches but only the silver-

Shown in Figure 1 are snapshots from simulations for two surfaces that successfully nucleated ice at a supercooling of 20

Figure 1. Snapshots of ice growth at different times. Panels A−C are for the silver exposed β-AgI(0001) face at 0, 11, and 36 ns, respectively. Panels D−F are for the silver exposed γ-AgI(001) face at 0, 3, and 6 ns, respectively. Only the first layer of each AgI slab is shown. Silver and iodide ions are silver and green, respectively. Oxygen atoms of liquid water, Ic, and Ih are red, yellow, and blue, respectively, and all hydrogen atoms are black. Note that panels D−F have been rotated to show the (110) plane of Ic, and this is why the water density can appear nonuniform in those snapshots.

K. Snapshots for the third surface that successfully nucleated ice under the same conditions are shown in the Supporting Information (Figure S2). The CHILL algorithm36 was used to determine whether a particular water molecule was part of the liquid, Ih, or Ic phase. Note that this algorithm does not recognize water molecules at the silver iodide surface as ice, so they retain the liquid color code; however, these water molecules form part of the initial ice layer after nucleation has occurred. Figure 1A−C shows snapshots for the silver exposed βAgI(0001) face. For this face both Ih(0001) and Ic(111) planes grow parallel to the silver iodide surface. As remarked above, the Ih(0001) and Ic(111) faces are almost identical and can connect with little lattice strain.35 This stacking disorder has been noted in previous works.37,38 In Figure 1A−C, at ∼11 ns the ice layers growing from both silver iodide surfaces meet, the upper crystal then continues to grow at the expense of the lower one, in what appears to be an Ostwald ripening process. In the simulation, the ice growth eventually stops with some liquid water remaining in the simulation cell. This is because as the ice grows the liquid is compressed until no further ice growth is possible. We conducted several simulations to identify the highest temperature that the silver exposed β-AgI(0001) face would C

dx.doi.org/10.1021/jp508601s | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

Figure 3. Normalized density profiles of oxygen atoms (red) and hydrogen atoms (black) of water as functions of distance from βAgI(0001) surfaces. Panels A and B are for ice (supercooled by 20 K) and liquid water (supercooled by 4 K), respectively, near the silver exposed face. Panel C is for liquid water (supercooled by 20 K) near the iodide exposed face. On the distance scale, zero (vertical green line) coincides with the position of the center of the outermost ion of the face. ρ0 is the average density of oxygen or hydrogen atoms far from the surface. Note the shifted scale in panels A and B.

3A), the oxygen profile shows a repeating doublet indicative of the formation of bilayers with hexagonal rings in the chair conformation. The hydrogen density profile for ice consists of repeating triplets, the central peak is due to hydrogen atoms that are hydrogen-bonded with water molecules in the same bilayer. The less intense peaks on either side indicate hydrogen bonding to ice layers above and below the bilayer, or in the case of the initial peak, to the silver iodide surface. Liquid simulations (Figure 3B) show that near the surface the structure of water is very similar to that of ice. It is apparent that the silver exposed β-AgI(0001) surface imposes an ice-like structure in the adjacent water layer. This ice-like surface layer then allows the next layer of ice to form and so on as ice continues to grow from the surface. In all likelihood this is why this particular AgI face is such an effective ice nucleus. The iodide exposed face of β-AgI(0001) did not nucleate ice. The density profiles shown for this face are for the liquid state (Figure 3C), and are to be compared with those for liquid near the silver exposed face (Figure 3B). It can be seen that the water structure near the iodide exposed face differs significantly from the silver exposed case. On the iodide exposed face the density profiles for both the oxygen and hydrogen atoms exhibit single rather than split peaks. The hydrogen peak is nearly coincident with that of oxygen indicating that the hydrogen and oxygen atoms are essentially coplanar. The water density profiles for the iodide exposed surface suggest a single planar layer of water molecules rather than a bilayer, or, in other words, the hexagonal rings at the surface (Figure 2B) are planar rather than in chair conformations. A similar density profile, with a large single surface peak, was also observed in an earlier simulation of water on this iodide exposed face.49 The failure of the iodide exposed face to order water into an ice-like bilayer likely undermines its ability to nucleate ice. The results discussed above were all obtained with charges of ±0.6e on the silver and iodide sites; we shall refer to this case as the “normally” charged surface. However, as noted above, simulations were performed for a range of charges, and these give additional insight into how the silver exposed face organizes water into chair configurations. For the silver exposed β-AgI(0001) face, ice nucleation was observed for charges ranging from zero to ±0.8e, but not for ±1.0e (Supporting Information, Table S5). This behavior can be understood by examining the charge dependence of the density profiles of

Figure 2. Snapshots of liquid water on β-AgI(0001) surfaces. Panel A is a top view of water (supercooled by 4 K) on the silver exposed face, and panel C is a side view of a single hexagonal ring on this surface. Panels B and D are corresponding views of water (supercooled by 20 K) on the iodide exposed face. Panel E is a side view of a chair structure from ice. Silver ions, iodide ions, oxygen atoms, and hydrogen atoms are silver, green, red, and black, respectively. The yellow lines highlight the hexagonal arrangement of the water on the surface.

exposed face nucleates ice. On both the silver and iodide exposed β-AgI(0001) faces, water forms hexagonal rings (Figure 2A and B). This finding is consistent with previous studies that have shown that the β-AgI(0001) surface is effective at organizing water molecules on its surface into hexagonal rings.20,44−52 For the silver exposed surface (Figure 2A,C), the oxygen atoms of the water molecules have a chair conformation, which resembles the chair conformation in the ice bilayer of Ih(0001) and Ic(111) (compare panels C and E in Figure 2). It is apparently this striking conformational similarity that makes this surface an excellent ice nucleus. The chair conformation is caused by water molecules occupying two different sites at the surface: directly above iodide ions lower in the AgI bilayer, or over interstitial sites between both ions. Water molecules occupying interstitial sites sit closer to the surface than water molecules over iodide ions and thus form the chair conformation. In contrast, for the iodide exposed surface (Figure 2B,D) the hexagonal rings form a more coplanar structure. On the iodide exposed face, all water molecules sit over equivalent lattice sites, such that all oxygen atoms are largely coplanar. This planar layer of hexagonal rings does not match any surface of ice, which is likely why we observe no ice nucleation on this face. The difference between these two faces can also be seen in the normalized density profiles of oxygen and hydrogen, as functions of distance from the surface (Figure 3). For the silver exposed face, density profiles are given for both ice (Figure 3A) and liquid water (Figure 3B) near the surface. For ice (Figure D

dx.doi.org/10.1021/jp508601s | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

liquid water near the silver exposed β-AgI(0001) face; profiles for charges of zero and ±1.0e are shown in Figure 6. We note that the profiles for the zero charge case (Figure 6A) resemble those of liquid water near the normally charged face (Figure 3B). Thus, the ordering of water at this surface appears to be mainly due to the LJ interactions, which are all that remain in the zero charge limit. On the other hand, for charges of ±1.0e, the density profiles (Figure 6B) show significant deviations from those of the normally charged surface. In particular, the high sharp peaks suggest that water molecules at this surface are more tightly held, with the oxygen atoms occupying a more planar arrangement. This distortion of the surface layer inhibits ice nucleation. The γ-AgI(111) faces are very similar to the corresponding faces of β-AgI(0001) and order water in an near identical manner (Supporting Information, Figures S3 and S4). Thus, the discussion above for β-AgI(0001) applies equally to the γAgI(111) case. 2. Comparison of Simulations for Ag Exposed and I Exposed γ-AgI(001). It is also interesting to compare the microscopic details of the water−surface interactions for the silver exposed face of γ-AgI(001) with the iodide exposed face of γ-AgI(001). Both surfaces have identical lattice mismatches to Ic, but only the silver exposed face led to ice nucleation. Configurational snapshots of the initial water layer on the silver and iodide exposed faces of γ-AgI(001) are shown in Figure 4A,B. These snapshots are for liquid water in contact with the

surface. We note that for both faces the oxygen atoms occupy positions consistent with a cubic ice lattice. However, the molecular orientations are different; most notably, for the silver exposed surface, protons of the water molecules are clearly directed inward toward the surface, similar to the structure of water in an Ic lattice (Figure 4E). This fine detail appears to be important for the nucleation of ice. In contrast, for the iodide exposed surface, all of the hydrogen atoms are directed outward away from the surface, which does not match any surface of ice. This mismatch in the fine structure is likely why we observe no ice nucleation on this face. The differences between the two faces can also be seen in the density profiles shown in Figure 5. For the silver exposed face,

Figure 5. Normalized density profiles of oxygen atoms (red) and hydrogen atoms (black) of water as functions of distance from γAgI(001)) surfaces. Panels A and B are for ice (supercooled by 20 K) and liquid water (6 K above the melting point), respectively, near the silver exposed face. Panel C is for liquid water (supercooled by 20 K) near the iodide exposed face. On the distance scale, zero (vertical green line) coincides with the position of the center of the outermost ion of the face. ρ0 is the average density of oxygen or hydrogen atoms far from the surface. Note the shifted scale in panels A and B.

density profiles are given for both ice (Figure 5A) and liquid water (Figure 5B). For ice (Figure 5A), the oxygen profile presents a series of evenly spaced peaks that are from the layers of the cubic lattice of Ic. The peaks in the hydrogen atom profile do not coincide with those of oxygen, indicating that the hydrogen atoms are bonding to water molecules in other layers or to the silver iodide surface, but not to water molecules in the same layer. Profiles for liquid water near the silver exposed face (Figure 5B) clearly show that an ice-like structure is imprinted on the water at the surface, just as it is for the silver exposed βAgI(0001) case, allowing the nucleation and growth of subsequent ice layers. Density profiles for the iodide exposed face do not match the ice structure as well as the silver exposed case (Figure 5C). This is particularly true of the hydrogen atom profile, which clearly shows the absence of protons closer to the surface than oxygen atoms. Thus, just as for β-AgI(0001), the surface-induced organization of the initial layer or two of water molecules is crucial for ice nucleation. Some simulations with different surface charges were also done for the silver exposed γ-AgI(001) face (Supporting Information, Table S5). Ice nucleation was observed for charges ranging from ±0.2e to ±0.6e, but not for zero and ±0.8e. Density profiles obtained for zero charge and ±0.8e are plotted in Figure 7. Both profiles differ considerably from the ±0.6e result for this surface (Figure 5B). At zero charge (Figure 7A), the water molecules appear to have moved away from the surface to some extent, indicating that in this case the Coulombic interactions do significantly influence how water

Figure 4. Snapshots of liquid water on γ-AgI(001) surfaces. Panel A is a top view of water (11 K above the melting point) on the silver exposed face, and panel C is a tilted view of a single FCC-like arrangement on this surface. Panels B and D are corresponding views of water (supercooled by 20 K) on the iodide exposed face. Panel E is a tilted view of a FCC arrangement in Ic. Silver ions, iodide ions, oxygen atoms, and hydrogen atoms are silver, green, red, and black, respectively. The yellow lines highlight the FCC-like arrangement of the water on the surface. E

dx.doi.org/10.1021/jp508601s | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

the bulk. This mechanism does not rule out defects being important for ice nucleation by silver iodide, but does indicate that defects are not necessary in order to explain why silver iodide particles can be highly effective ice nuclei. However, as noted by Fraux and Doye,42 one must be cautious in extrapolating from model simulations to real AgI surfaces. All AgI faces for which we observed ice nucleation have surface dipole moments, and hence some surface reconstruction is expected.53 Clearly, such effects, as well as any other possible influences of thermal motion, are not included in our rigid lattice model of AgI. Finally, we speculate that a templating mechanism similar to that observed for our model AgI surfaces might apply for other effective ice nuclei, such as aliphatic alcohol monolayers, and proteins from the bacteria Pseudomonas syringae.

Figure 6. Normalized density profiles of oxygen atoms (red) and hydrogen atoms (black) of liquid water as functions of distance from the Ag exposed β-AgI(0001) surface. Panel A is the zero charge case (water supercooled by 4 K). Panel B is for AgI carrying charges of ±1.0e (water supercooled by 20 K). The green vertical line at zero coincides with the position of the center of the outermost ions, and ρ0 is the average density of oxygen or hydrogen atoms far from the surface. Note the shifted scale for panel A.



ASSOCIATED CONTENT

S Supporting Information *

Results for the TIP4P/Ice water model, polarization profiles, additional figures and tables referred to in the main text, and four videos of simulation trajectories. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



Figure 7. Normalized density profiles of oxygen atoms (red) and hydrogen atoms (black) of liquid water (supercooled by 20 K) as functions of distance from the Ag-exposed γ-AgI(001) surface. Panel A is the zero charge case. Panel B is for AgI carrying charges of ±0.8e. The green vertical line at zero coincides with the position of the center of the outermost ions, and ρ0 is the average density of oxygen or hydrogen atoms far from the surface. Note the shifted scale for panel A.

ACKNOWLEDGMENTS The financial support of the Natural Science and Engineering Research Council of Canada is gratefully acknowledged. This research has been enabled by the use of WestGrid and Compute/Calcul Canada computing resources, which are funded in part by the Canada Foundation for Innovation, Alberta Innovation and Science, BC Advanced Education, and the participating research institutions. WestGrid and Compute/ Calcul Canada equipment is provided by IBM, Hewlett-Packard and SGI.

is organized at the surface. Conversely, the profiles for ±0.8e (Figure 7B) show that water is more tightly held by the higher charge. Additionally, at the surface the water molecules are now oriented such that both protons are directed inward toward iodide ions (Supporting Information, Figure S14), which likely inhibits ice nucleation. 3. Simulations for the Ag+I Exposed β-AgI(101̅0) Surface. As discussed above, when β-AgI(101̅0) is cleaved only one type of surface is formed and we refer to this as Ag+I exposed, since both silver and iodide ions are in the plane of the surface. Despite the good lattice match to Ih, ice nucleation was not observed for this particular silver iodide face. Configurational snapshots show that the first layer of water on this face does not possess any significant ice-like order (Supporting Information, Figure S5) and density profiles do not show a strong ice-like pattern (Figure S6). Thus, it appears that the atomistic details of this surface do not favor ice nucleation and growth.



REFERENCES

(1) Pruppacher, H. R.; Klett, J. D. Microphysics of Clouds and Precipitation; Springer: New York, 2010. (2) Hegg, D. A.; Baker, M. B. Nucleation in the Atmosphere. Rep. Prog. Phys. 2012, 72, 056801−1−21. (3) Murray, B. J.; O’Sullivan, D.; Atkinson, J. D.; Webb, M. E. Ice Nucleation by Particles Immersed in Supercooled Cloud Droplets. Chem. Soc. Rev. 2012, 41, 6519−6554. (4) Lau, K. M.; Wu, H. T. Warm Rain Processes over Tropical Oceans and Climate Implications. Geophys. Res. Lett. 2003, 30, 2290− 2295. (5) DeMott, P. J.; Prenni, A. J.; Liu, X.; Kreidenweis, S. M.; Petters, M. D.; Twohy, C. H.; Eidhammer, T.; Rogers, D. C. Predicting Global Atmospheric Ice Nuclei Distributions and Their Impacts on Climate. Proc. Natl. Acad. Sci. U.S.A. 2010, 107, 11217−11222. (6) Bartels-Rausch, T. Ten Things We Need To Know about Ice and Snow. Nature 2013, 494, 27−29. (7) Ervens, B.; Feingold, G. On the Representation of Immersion and Condensation Freezing in Cloud Models Using Different Nucleation Schemes. Atmos. Chem. Phys. 2012, 12, 5807−5826. (8) Bruninjes, R. T. A Review of Cloud Seeding Experiments To Enhance Precipitation and Some New Prospects. Bulletin of the American Meteorological Society 1999, 80, 805−820.

IV. CONCLUSION On the basis of our model simulations and the detailed analysis discussed above, a possible mechanism for ice nucleation by silver iodide consists of some silver iodide surfaces imposing a structure in the adjacent water layer that closely resembles a layer that exists in bulk ice (either Ih or Ic). Ice nucleates at these surfaces and grows outward forming successive layers into F

dx.doi.org/10.1021/jp508601s | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

(9) Vonnegut, B. The Nucleation of Ice Formation by Silver Iodide. J. Appl. Phys. 1947, 18, 593−595. (10) Zettlemoyer, A. C.; Tcheurekdjian, N.; Chessick, J. J. Surface Properties of Silver Iodide. Nature 1961, 192, 653. (11) Finnegan, W. G.; Chai, S. K. A New Hypothesis for the Mechanism of Ice Nucleation on Wetted AgI and AgI-AgCl Particulate Aerosols. J. Atmos. Sci. 2003, 60, 1723−1731. (12) Lupi, L.; Hudait, A.; Molinero, V. Heterogeneous Nucleation of Ice on Carbon Surfaces. J. Am. Chem. Soc. 2014, 136, 3156−3164. (13) Lupi, L.; Molinero, V. Does Hydrophilicity of Carbon Particles Improve Their Ice Nucleation Ability? J. Phys. Chem. A 2014, DOI: 10.1021/jp4118375. (14) Molinero, V.; Moore, E. B. Water Modeled as an Intermediate Element Between Carbon and Silicon. J. Phys. Chem. B 2009, 113, 4008−4016. (15) Cox, S. J.; Raza, Z.; Kathmann, S. M.; Slater, B.; Michaelides, A. The Microscopic Features of Heterogeneous Ice Nucleation May Affect the Macroscopic Morphology of Atmospheric Ice Crystals. Faraday Discuss. 2013, 167, 389−403. (16) Nada, H.; van der Earden, J. P. J. M. An Intermolecular Potential Model for the Simulation of Ice and Water near the Melting Point: A Six-Site Model of H2O. J. Chem. Phys. 2003, 118, 7401−7413. (17) Abascal, J. L. F.; Fernández, R. G.; Vega, C.; Carignano, M. A. The Melting Temperature of the Six Site Potential Model of Water. J. Chem. Phys. 2006, 125, 166101−1−2. (18) Abascal, J. L. F.; Sanz, E.; Fernández, R. G.; Vega, C. A Potential Model for the Study of Ices and Amorphous Water: TIP4P/Ice. J. Chem. Phys. 2005, 122, 234511−1−9. (19) Fernández, R. G.; Abascal, J. L. F.; Vega, C. The Melting Point of Ice Ih for Common Water Models Calculated from Direct Coexistence of the Solid−Liquid Interface. J. Chem. Phys. 2006, 124, 144506−1−11. (20) Hale, B. N.; Kiefer, J. Studies of H2O On β-AgI Surfaces: An Effective Pair Potential Model. J. Chem. Phys. 1980, 73, 923−933. (21) Croteau, T.; Bertram, A. K.; Patey, G. N. Adsorption and Structure of Water on Kaolinite Surfaces: Possible Insight into Ice Nucleation from Grand Canonical Monte Carlo Calculations. J. Phys. Chem. A Lett. 2008, 112, 10708−10712. (22) Croteau, T.; Bertram, A. K.; Patey, G. N. Simulation of Water Adsorption on Kaolinite Under Atmospheric Conditions. J. Phys. Chem. 2009, 113, 7826−7833. (23) Burley, G. Structure of Hexagonal Silver Iodide. J. Chem. Phys. 1963, 38, 2807−2812. (24) Hull, S.; Keen, D. A. Pressure-Induced Phase Transitions in AgCl, AgBr, and AgI. Phys. Rev. B 1999, 59, 750−761. (25) Gražulis, S.; Daškevic̆, A.; Merkys, A.; Chateigner, D.; Lutterotti, L.; Quiró, M.; Serebryanaya, N. R.; Moeck, P.; Downs, R. T.; Le Bail, A.; et al. Crystallography Open Database (COD)An Open-Access Collection of Crystal Structures and Platform for World-Wide Collaboration. Nucleic Acids Res. 2012, 40, D420−D427. (26) Pronk, S.; Páll, S.; Schulz, R.; Larsson, P.; Bjelkmar, P.; Apostolov, R.; Shirts, M. R.; Smith, J. C.; Kasson, P. M.; van der Spoel, D.; et al. GROMACS 4.5: A High-Throughput and Highly Parallel Open Source Molecular Simulation Toolkit. Bioinformatics 2013, 29, 845−854. (27) Nosé, S. A Unified Formulation of the Constant Temperature Molecular-Dynamics Methods. J. Chem. Phys. 1984, 81, 511−519. (28) Hoover, W. G. Canonical Dynamics: Equilibrium Phase-Space Distributions. Phys. Rev. A 1985, 31, 1695−1697. (29) Darden, T.; York, D.; Pedersen, L. Particle Mesh Ewald: An Nlog(N) Method of Ewald Sums in Large Systems. J. Chem. Phys. 1993, 98, 10089−10092. (30) Hess, B.; Bekker, H.; Berendsen, H. J. C.; Fraaije, J. G. E. M. LINCS: A Linear Constraint Solver for Molecular Simulations. J. Comput. Chem. 1997, 18, 1463−1472. (31) Thangaraj, K.; Palanisamy, M.; Gobinathan, R.; Ramasamy, P. Dependence of Ice Nucleating Ability on Misfit. J. Mater. Sci. Lett. 1986, 5, 326−328.

(32) Cox, S. J.; Kathmann, S. M.; Purton, J. A.; Gillan, M. J.; Michaelides, A. Non-Hexagonal Ice at Hexagonal Surfaces: The Role of Lattice Mismatch. Phys. Chem. Chem. Phys. 2012, 14, 7944−7949. (33) Carignano, M. A. Formation of Stacking Faults During Ice Growth on Hexagonal and Cubic Substrates. J. Phys. Chem. C Lett. 2007, 111, 501−504. (34) Malenkov, G. Liquid Water and Ices: Understanding the Structure and Physical Properties. J. Phys.: Condens. Matter 2009, 21, 283101−1−35. (35) Johari, G. P. On the Coexistence of Cubic and Hexagonal Ice between 160 and 240 K. Philos. Mag. B 1998, 78, 375−383. (36) Moore, E. B.; de la Llave, E.; Welke, K.; Scherlis, D. A.; Molinero, V. Freezing, Melting and Structure of Ice in a Hydrophilic Nanopore. Phys. Chem. Chem. Phys. 2010, 12, 4124−4134. (37) Malkin, T. L.; Murray, B. J.; Brukhno, A. V.; Anwar, J.; Salzmann, C. G. Structure of Ice Crystallized from Supercooled Water. Proc. Natl. Acad. Sci. U.S.A. 2012, 109, 1041−1045. (38) Kuhs, W. F.; Sippel, C.; Falenty, A.; Hansen, T. C. Extent and Relevance of Stacking Disorder in “Ice Ic”. Proc. Natl. Acad. Sci. U.S.A. 2012, 109, 21259−21264. (39) Nutt, D. R.; Stone, A. J. Adsorption of Water on the BaF2 (111) Surface. J. Chem. Phys. 2002, 117, 800−807. (40) Nutt, D. R.; Stone, A. J. Ice Nucleation on a Model Hexagonal Surface. Langmuir 2004, 20, 8715−8720. (41) Hu, X. L.; Michaelides, A. Ice Formation on Kaolinite: Lattice Match or Amphoterism? Surf. Sci. 2007, 601, 5378−5381. (42) Fraux, G.; Doye, P. K. Heterogeneous Ice Nucleation on SilverIodide-like Surfaces, 2014, arXiv:1407.3528. (43) Abascal, J. L. F.; Vega, C. A General Purpose Model for the Condensed Phases of Water: TIP4P/2005. J. Chem. Phys. 2005, 123, 234505. (44) Fukuta, N.; Paik, Y. Water Adsorption and Ice Nucleation on Silver Iodide Surfaces. J. Appl. Phys. 1973, 44, 1092−1100. (45) Volkov, V. B.; Zhogolev, D. A.; Bakhanova, R. A.; Kiselev, V. J. A Quantum-Chemical Study of the Interaction of the Water Molecule with Silver Iodide Surface Clusters. Chem. Phys. Lett. 1976, 43, 45−48. (46) Hale, B. N.; Kiefer, J.; Terrazas, S.; Ward, R. C. Theoretical Studies of Water Adsorbed on Silver Iodide. J. Phys. Chem. 1980, 84, 1473−1479. (47) Ward, R. C.; Holdman, J. M.; Hale, B. N. Monte Carlo Studies of Water Monolayer Clusters on Substrates: Hexagonal AgI. J. Chem. Phys. 1982, 77, 3198−3202. (48) Ward, R. C.; Hale, B. N.; Terrazas, S. A Study of the Critical Cluster Size for Water Monolayer Clusters on a Model AgI Basal Substrate. J. Chem. Phys. 1983, 78, 420−423. (49) Taylor, J. H.; Hale, B. N. Monte Carlo Simulations of Water− Ice Layers on a Model Silver Iodide Substrate: A Comparison with Bulk Ice Systems. Phys. Rev. B 1993, 47, 9732−9741. (50) Shevkunov, S. V. Adsorption of Water Vapor On the AgI Surface: A Computer Experiment. Russ. J. Gen. Chem. 2005, 75, 1632− 1642. (51) Shevkunov, S. V. Layer-by-Layer Adsorption of Water Molecules on the Surface of a Silver Iodide Crystal. Russ. J. Phys. Chem. 2006, 80, 769−775. (52) Shevkunov, S. V. Clusterization of Water Molecules on Crystalline β-AgI Surface: Computer Experiment. Colloid J. 2006, 68, 632−643. (53) Tasker, P. W. The Stability of Ionic Crystal Surfaces. J. Phys. C: Solid State Phys. 1979, 12, 4977−4984.

G

dx.doi.org/10.1021/jp508601s | J. Phys. Chem. B XXXX, XXX, XXX−XXX