Article pubs.acs.org/JPCC
Adsorptions of HOCl on Ice Surface: Effects of Long-Range Electrostatics, Surface Heterogeneity, and Hydrogen Disorders of Ice Crystal Mahbubul Alam Shoaib and Cheol Ho Choi* Department of Chemistry and Green-Nano Materials Research Center, College of Natural Sciences, Kyungpook National University, Taegu 702-701, South Korea ABSTRACT: The adsorptions of HOCl on ice surface were studied with the help of QM/EFP scheme. HOCl binding energies were predicted to be −12.2 to −17.8 kcal/mol (BSSE corrected values) depending on surface absorption sites. These values are much larger than previous quantum mechanical estimations but are rather in good agreements with experiments and MD study. Structurally, various new surface binding configurations of HOCl, including an unusual penta coordination, were found indicating diverse reacting environments of ice surface. In general, it was found that the ice surface itself as well as HOCl adsorptions are strongly affected by long-range electrostatics, surface heterogeneity and hydrogen disorders of bulk ice, revealing the unique characteristics of ice surface as a reacting environment. As a way of modeling ice surface, we demonstrated that the hybrid QM/EFP scheme is very effective.
■
INTRODUCTION Chemical reactions on ice surface have become one of the most important research subjects recently. The interest in the interactions of gases with ice surface largely prompted by the discovery of the Antarctic ozone hole1 and the recognition that atmospheric ice particles2 can play a dominant role in determining the chemical composition of the atmosphere. Heterogeneous reactions on ice particles in polar stratospheric clouds (PSCs) are involved in the ozone hole over the Antarctic. HOCl plays an important role in this event,1,2 which undergoes photodissociation by sunlight to produce atomic chlorine that catalyzes ozone destruction/depletion. It has been also recognized that the abundance of some important molecules in space, including H2, cannot be explained by pure gas phase formation and that surface reactions on dust grains in space are crucial to explain the production of these molecules.3 Ice surfaces offer a unique reaction environment that is different from those formed by liquid water, gas, or even bulk ice.4 At the temperature of ice, the rate of thermal reaction is expected to be much slower. In addition, the rate of reagent diffusion is reduced by several orders of magnitude. With such conditions, the occurrence of a chemical reaction with an appreciable speed has been considered to be unlikely. However, recent studies show that reactions occur even at substantially low temperatures at the surface of ice.4 Theoretical studies of chemical reactions on ice surface based on ab initio methods predict in general large reaction barriers implying that the surface reactions are not facile, contradicting experimental results.5 So far, a majority of ice surface cluster models usually comprises no more than four water molecules6 (see Figure 1a). To include long-range effects, methods based on periodic boundary conditions7 as well as QM/MM8 have © 2012 American Chemical Society
been also utilized recently. In order to properly model chemical reactions on ice surface, various aspects of the surface should be carefully investigated. Most importantly, long-range electrostatics by bulk ice should be studied. The electrostatic and polarization interactions play a vital role in various condensed phases containing polar and polarizable molecules. Polarization effects of water are particularly large since its molecular dipole increases from 1.855 D for an isolated molecule to 2.6−3.2 D in condensed state.9 Since the electrostatic interactions are directional, it has been generally recognized that they are critical in the orderings of water clusters. Such long-range electrostatic interactions are expected to be important in the formation of ice crystal, where the directional orderings of the tetrahedral network are predominant. It is not unreasonable that the same electrostatic intereactions can also affect the chemical absorptions on the ice surface. In addition, hydrogen disorders of ice crystal should be taken into account. The structure of popular Ih ice is arranged on a hexagonal lattice. Each oxygen atom has four nearest neighbors at the corners of a regular tetrahedron. The hydrogen atoms are covalently bonded to the nearest oxygen to form H2O molecules, and these molecules are linked to one another by hydrogen bonds, each molecule offering its hydrogens to two other molecules and accepting hydrogen bonds from another two. As a result, there is no long-range order in the orientations of the H2O molecules or of the hydrogen bonds. These disorders are summarized by the ice rules. This fact makes the Received: November 30, 2011 Revised: January 12, 2012 Published: January 17, 2012 3694
dx.doi.org/10.1021/jp211538v | J. Phys. Chem. C 2012, 116, 3694−3701
The Journal of Physical Chemistry C
Article
Figure 1. Various ice models of (a) oxygen dangling bond (ODB) with 4 water clusters, (b) hydrogen dangling bond (HDB) with 4 water clusters, (c) 10 water clusters, (d) 13 water clusters, (e) 26 water clusters, (f) 74 water clusters, (g) 219 water clusters, and (h) 484 water clusters.
surface structures and chemical adsorptions of HOCl using our cluster models. To observe the physical and chemical processes that occurred during adsorption, it is important to know the accurate binding energy of HOCl on ice surface.
use of periodic boundary conditions on the ice crystal difficult. Kuo et al.10 also found static distortion even in the oxygen positions from their crystallographic positions. In short, disorders in the orientations of H2O molecules exist within the framework of directional orderings of the tetrahedral network of ice. Therefore, a delicate yet complex and collective electrostatic potential exists in ice crystal and plays a vital role in the formation of its three-dimensional ice structures. Because of this, theoretical approaches based on large clusters rather than the methods utilizing periodic boundary condition are preferred. In addition to electrostatic interactions and hydrogen disorders of ice, surface heterogeneity, in which an oxygen dangling bond (ODB) (Figure 1a) as well as a hydrogen dangling bond (HDB) (Figure 1b) are statistically distributed, needs to be properly modeled. Such microscopic heterogeneity of ice surface is a direct consequence of hydrogen disorders of ice crystal. It can provide diverse reacting environments including multiple interactions. Full geometry relaxation at least near the adsorption sites should be also allowed. This aspect is especially important since it significantly affects direct interfacial interactions between adsorbed molecules and ice surface, which are generally considered as the major contribution to the absorptions. In this article, the effects discussed above were systematically investigated. In the section below, we first describe our cluster models. After that, we present and analyze our results of ice
■
ICE SURFACE MODELS
Since detailed information of ice structure can be found in previous reviews,11 a brief description of ice surface structure shall be given here. The bulk-truncated unreconstructed (0001) surface of hexagonal crystalline (Ih) ice corresponds to the (111) surface of cubic (Ic) ice. Both phases consist of bilayers of tetrahedrally bonded H2O molecules connected by hydrogen bonds; the oxygen planes are spaced by 2.75 Å between bilayers and 0.92 Å within bilayers. The bilayers have the ...ABABAB... stacking sequence in Ih and the ...ABCABC... stacking sequence in ice Ic. Each oxygen is surrounded by four other hydrogenbonded oxygens 2.75 Å away. H is not centered between two O atoms in the hydrogen bond but stays off-center by about 0.38 Å in a random fashion. The (0001) surface can also be either full-bilayer or half-bilayer terminations. In the former, each outermost O atom in the upper layer is hydrogen-bonded to three neighbors of the lower layer. In the case of half-bilayer termination, the upper layer of full-bilayer is missing. Fullbilayer termination is energetically favored over half bilayer due to a larger number of hydrogen bonds. Energy difference between the two surfaces is about 30 kJ/mol per unit cell.12 3695
dx.doi.org/10.1021/jp211538v | J. Phys. Chem. C 2012, 116, 3694−3701
The Journal of Physical Chemistry C
Article
Thus, the full bilayer termination is mainly considered in our model. When the (0001) surface is terminated as a full bilayer, the water molecules in the outer layer have one dangling bond (either an oxygen dangling bond, ODB, or a hydrogen dangling bond, HDB) pointing outward as shown in Figure 1a and b, respectively. These two dangling bond sites can be considered as the primary reactive places due to their outermost positions on ice surface. Their statistical distributions on ice surface represent the microscopic heterogeneity of ice surface, which can provide various reacting environments including multiple chemical interactions. Therefore, it is expected that models that only consider the ODB site cannot describe diverse surface environments. Reference Ice Crystal. In order to design ice surface models that can satisfy the requirements as listed in the introduction, a systematic strategy of building models is proposed. We first generated coordinates of a reference ice crystal of 36 hexagons by 15 hexagons and 5 full-bilayers deep. We ensured that each layer of 36 by 15 will have a (near to) zero dipole moment by randomizing hydrogen bonding distributions. As a result, the entire ice model of 5 layers will also have zero dipole moment. However, one should be aware of the fact that there are so many different ways to arrange hydrogens in an ice structure and the ice rules do not determine a unique structure. However, it is expected that imposing a low or zero dipole moment criteria may significantly reduce the number of possibilities. The initial coordinates of waters in the reference crystal were obtained on the condition that oxygen−oxygen distances, O−H distances, and O−O−O angles are experimentally found values of 2.75 Å, 1.00 Å, and 109.3°, respectively. Cluster Size Effects and QM/EFP Model. Using the reference ice crystal, various ice models were systematically designed as follows. By putting the ODB or HDB site in the center, clusters with 10, 13, 26, 74, 219, and 484 waters were prepared as shown in Figure 1c−h, respectively. Since full quantum calculations on our large clusters are not practical, combinations of pure quantum mechanical methods and the effective fragment potential (EFP)13 method were adopted. Since detailed descriptions of EFP can be found elsewhere,13 the relevant aspects of EFP shall be briefly mentioned here. The EFP is a quantum mechanical polarizable force field with implicit charge transfer and exchange repulsion corrections. The method has been shown to reproduce the correct structure of liquid water14 and has been successfully applied for understanding the solvent-induced shifts in electronic spectra of uracil in water.15 Therefore, EFP method contains the necessary terms in the study of electrostatic and polarization effects of ice surface. The EFP as parametrized with RHF theory was used throughout the current calculations. The particular combination of QM and EFP scheme in this article is a three-layer model (QM relaxed region/QM fixed region/EFP water), where the coordinates in the QM relaxed region are fully relaxed during geometry optimization, while those of QM fixed and EFP regions are frozen (see Figure 2). The basic strategy behind our three-layer scheme is the surface water(s) that is interacting with adsorbate, and its first neighbor waters should be in QM relaxed regions. The quantum atoms in QM fixed region serve as a buffer between QM relaxed and EFP regions. This model includes both the geometric relaxation effect due to the adsorptions and the long-range electrostatic interactions of ice crystals at the same time. On the basis of our three layer strategy, (4/6/0), (4/6/3), (4/6/16), (4/6/64),
Figure 2. Diagram of relaxed QM/fixed QM/EFP scheme. The geometries of the relaxed QM region are fully optimized, while those of the fixed QM and EFP regions are also fixed during optimization.
(4/6/209), (4/6/474), (4/15/465), and (10/9/465) models were designed using the ice models of Figure 1c−h. For example, there are 10 relaxed QM waters, 9 fixed QM waters, and 465 EFP waters in the (10/9/465) model. In order to study the geometric relaxation effects on the adsorption site, three different sizes of geometrically relaxed quantum regions of (4/6/X), (4/15/X), and (10/9/X) were also designed. (X is the number of EFP waters.) The three corresponding QM regions are presented in Figure 3, where the QM relaxed and QM fixed waters are indicated with ball and stick and thick stick representations, respectively. Hydrogen Disorder of Ice Crystal. In order to take the hydrogen disorder effects of ice crystal into account, we prepared three ice models with different hydrogen disorders. They shall be denoted as SA(10/9/465), SB(10/9/465), and SC(10/9/465). The corresponding QM regions of these three models are presented in Figure 4a−c, respectively. As a result of different hydrogen disorders in ice crystal, the surface dangling bonds of these three models are also different with each other representing the heterogeneity of ice surface. Surface Heterogeneity of Hexagon (H) and Tripod (T) Adsorption Sites. In order to consistently describe the reactive sites of ice surfaces due to the heterogeneity, surface sites were categorized as hexagon (H) and tripod (T) sites using triangles, which connects the upper layer waters of the full-bilayer. The hexagon site does not have lower layer water in the middle of triangle, while tripod has one lower layer water. It should be noted that the lower waters of bilayer have always saturated coordination. Therefore, the upper waters are assumed to be reactive sites. These hexagon and tripod sites are further categorized as HX and TX (X = 0, 1, 2, 3) where X represents the number of HDB (hydrogen dangling bond) at the vertices of the triangle. For example, H3 indicates the hexagon site with three HDB at the upper layer waters. We prepared model SA, SB, and SC in such a way that they can represent all possible reactive sites of HX and TX (X = 0, 1, 2, 3).
■
COMPUTATIONAL DETAILS For the calculations of QM regions, MP2 theory was used in combination with 6-31G(d), 6-311G(d,p), and 6-311++G(d,p) 3696
dx.doi.org/10.1021/jp211538v | J. Phys. Chem. C 2012, 116, 3694−3701
The Journal of Physical Chemistry C
Article
Figure 3. Relaxed and fixed QM waters of (a) (4/6/X), (b) (4/15/X), and (c) (10/9/X) models. The numbers in parentheses indicate the numbers of waters. The X represents the number of EFP waters. The relaxed and fixed QM waters are depicted by ball and stick and thick stick representations.
for oxygen numberings), the fully optimized O−O distances were averaged, and the results were plotted in Figure 5, in which the corresponding distances of SA(10/9/465) model are also presented. The averaged O−O distances of SA(4/6/0) model are nearly 2.84 Å regardless of basis sets. A previous theoretical study using the constraint of 4 water clusters at MP2/6-311++G(d,p) reported a slightly longer distance of 2.905 Å.6a It can be seen that the additional 6 waters of SA(4/6/ 0) model impose geometric restrictions to center O−O distances. The averaged O−O distances are significantly reduced, when SA(4/6/3) and SA(4/6/16) models were used. This large reduction of average O−O distance can only be explained by long-range electrostatics. After that, they tend to approach the limiting value of 2.7 Å as the model size increases. The results with 6-311G(d,p) and 6-311++G(d,p) are similar with each other indicating that further increase of basis set would not change the limiting value. In short, it is clearly demonstrated that the surface structures as measured by O−O distance are strongly affected by bulk ice through electrostatics. Electrostatic Effects on the HOCl Adsorption. Binding Energy. There have been many experimental and theoretical studies of HOCl adsorptions on ice. Early theoretical studies adopting four water clusters of the oxygen dangling bond (ODB) site of Figure 1a revealed that the OH bond of HOCl binds the center surface oxygen through an interfacial hydrogen bond in a symmetric tilted-down and anti-
basis sets. All of the computations were done without imposing symmetry unless otherwise specified. The GAMESS (general atomic and molecular electronic structure system)16 program was used for all of the computations. It is noted that in order to perform three layer optimizations of relaxed QM, fixed QM, and fixed EFP models, the GAMESS was modified since the current GAMESS version does not yield correct Hessian and does not allow for the fixing of geometry during saddle point runs. The new modification will be added to the next version of GAMESS. Basis set superposition error (BSSE) corrections on selected models were performed with the counterpoise method.17
■
RESULTS AND DISCUSSIONS Effects of Electrostatics on Ice Surface Structure. The distance between oxygen atoms along each bond of Ih ice crystal is about 2.75 Å.11 The angle between bonds in the crystal lattice is very close to the tetrahedral angle of 109.5°. This tetrahedral bonding angle of the water molecule essentially accounts for the unusually low density of crystal lattice. Therefore, one of the most important geometric parameters of Ih ice is O−O distances, which determine the volume and density of ice. Geometries of the relaxed QM regions’ four waters of our SA(4/6/X) models where X = 0, 3, 16, 64, 209, 474, and SA(4/15/465) can be fully optimized within our model clusters. Since there are three O−O distances of O4−O1, O4−O10 and O4−O7 in the central four waters (see Figure 3a 3697
dx.doi.org/10.1021/jp211538v | J. Phys. Chem. C 2012, 116, 3694−3701
The Journal of Physical Chemistry C
Article
Figure 6. Two possible orientations of HOCl binding configurations of previous studies. (a) Tilted-down. (b) Anti-like.
Table 1. Previous Studies on Binding Energies of HOCl with Water Clusters (in kcal/mol) theoretical binding energies
Figure 4. Various surface models of (a) SA(10/9/465), (b) SB(10/9/ 465), and (c) SC(10/9/465). (The EFP waters are not presented for clarity.) The triangles connect the upper layer waters of the bilayer. Surface sites are categorized as hexagon (H) and tripod (T) sites on the basis of the presence of lower layer water in the middle of the triangle. These hexagon and tripod sites are further categorized as HX and TX (X = 0, 1, 2, 3) where X represents the number of HDB (hydrogen dangling bond) at the vertices of triangle. For example, H3 indicates the hexagon site with three HDB at the upper layer waters.
HOCl−H2Ob
MP4/6-311++G(3df,3pd)//MP2/6311++G(d,p) MP2/6-311++G(d,p)
HOCl·(H2O)4
MP2/6-31+G(d) MP2/6-311++G(d,p) BLYP/6-311++G(d,p) MD simulation experimental binding energies
HOCl·(H2O)360
−5.9 (syn),−5.6 (anti) −6.4 (syn),−6.3 (anti) −10.1 to −11.2c −11.5(−8.8a)d −10.4 to −12.5e −14.3f
−10.5 ± 2,g −14 ± 2h a
Basis set superposition error (BSSE) corrected value. bFrom ref 6a. From ref 6b. dFrom ref 6c. eFrom ref 6d. fFrom ref 6e. gFrom ref 6f. h From ref 6g. c
−11.5 kcal/mol using four water clusters with MP2/6-311+ +G(d,p). After BSSE (basis set superposition error) correction, it becomes −8.8 kcal/mol. With the same four water clusters, Brown and Doren6d predicted it to be −10.4 to −12.5 kcal/mol with BLYP/6-31++G(d,p). Only Geiger et al.6c reported a BSSE corrected value of −8.8 kcal/mol, which shall be considered as a reference value of the four water cluster model for the rest of our article. It appears that the theoretical binding energies by the one water cluster model are smaller than those with four water clusters by 5 kcal/mol, indicating that the binding energies are strongly affected by cluster size. The BSSE corrected value with the four water cluster model is close to Abbatt and Molina’s6f experiments but smaller than Hanson and Ravishankara6g by about 5 kcal/mol. In order to study the model dependencies of HOCl binding energies with larger clusters, we have performed the same calculations with various ice models. In our calculations, the binding structures were studied with T2 absorption site (tripod site with two hydrogen dangling bond) of SA model as shown in Figure 4a. This particular absorption site is modeled with various clusters of SA(4/6/X) (X = 0,3,16,64,209,474), SA(4/15/465), and SA(10/9/465). MP2 with three different basis sets were used. Full quantum calculations at the MP2/6-31G(d) level of theory were also performed on the moderate size clusters of (0/13/0), (0/26/ 0), and (0/74/0) models, in order to examine the validity of our QM/EFP scheme. The resulted energies and structures are presented in Figures 7 and 8, respectively. To compare with earlier results, we also calculated the HOCl binding energies
Figure 5. Averaged O−O distances between central four waters of various cluster models as a function of model size.
like orientations as shown in Figure 6. The HOCl binding energies of earlier studies are presented in Table 1. Experimentally, Abbatt and Molina6f showed the adsorption enthalpy of −10.5 ± 2 kcal/mol, while Hanson and Ravishankara6g reported −14 ± 2 kcal/mol. Theoretically, a molecular dynamics study by Kroes and Clary6e yielded binding energy of 14.3 kcal/mol. With the help of ab initio theories, Dibble and Francisco6a reported −5.6 to −5.9 kcal/mol using one water cluster with MP4 theory. Geiger et al.6c also reported 3698
dx.doi.org/10.1021/jp211538v | J. Phys. Chem. C 2012, 116, 3694−3701
The Journal of Physical Chemistry C
Article
Figure 7. HOCl binding energies as a function of model clusters. The X marks in the figure represent the full quantum single point energy calculations on (0/13/0), (0/26/0), and (0/74/0) with MP2/631G(d).
with the four water model (0/4/0) at MP2/6-311++G(d,p) as shown in Figure 7. The geometries of our (0/4/0) models were taken from the optimized geometries of (4/6/0) model. It should be noted that full geometry optimizations could not be done in earlier four water cluster calculations. Therefore, our results with (0/4/0) model cannot be exactly identical to earlier values. In any case, our binding energies of HOCl adsorption with (0/4/0) model agree well with earlier values as obtained with 4 water clsuters. Figure 7 shows that the HOCl binding energies on ice surface increase monotonically up to (4/6/209) model and then approach to the limiting value of −20.0 kcal/mol, if we follow the MP2/6-311++G(d,p) curve. After the basis set superposition error (BSSE) correction, this limiting value becomes −17.8 kcal/mol, which is about 2 kcal/ mol higher than the experimental upper bound of −14 ± 2 kcal/mol. As compared to the previous reference theoretical value of −8.8 kcal/mol,6a our limiting value is significantly larger. The additional stabilizations of (4/6/0) as compared to the four water cluster model can be attributed to the secondary hydrogen bond between HOCl oxygen and the surface hydrogen (O31−H21) as shown in Figure 8a. However, the monotonic increase of binding energy between (4/6/0) to (4/ 6/209) models can only be explained by long-range electrostatic interactions. Further increase of model size up to (10/9/465) has relatively little effects. Since the model (4/6/209) has a radius of 8 Å from the adsorption site, the electrostatic interactions are effective up to 8 Å from the reacting site. We also performed full quantum single point energy calculations at MP2/6-31G(d) level of theory on the (0/13/0), (0/26/0), and (0/74/0) models. The three full quantum calculations closely reproduce our corresponding energy values of the QM/QM/ EFP scheme at the same basis sets, proving the validity of our combined models. In summary, Figure 7 clearly indicates the existence of long-range electrostatic attractions by bulk ice, which significantly increase the binding energies of HOCl adsorptions. The values as obtained with 6-31G(d) and 6311G(d,p) yielded larger binding energies as compared to those with 6-311++G(d,p), indicating that the use of large basis sets is necessary.
Figure 8. Optimized HOCl binding structures with various SA−T2 model clusters of (a) (4/6/0), (b) (4/6/3), (c) (4/6/16), (d) (4/6/ 64), (e) (4/6/209), (f) (4/6/474), (g) (4/15/465), and (h) (10/9/ 465). The atoms marked with green, red, and white colors are Cl, O, and H, respectively.
Binding Structure. Figure 8 shows the optimized adsorption geometries of HOCl with various ice surface models. As discussed, from Figure 8a, (4/6/0) model shows both a primary hydrogen bond of H33−O4 and a secondary hydrogen bond of O31−H21, which has been missing in earlier four water models. As the model size increases, the primary hydrogen bond length decreases from 1.78 Å to the limiting value of 1.71 Å. The corresponding hydrogen bond lengths of previous theoretical studies are collected in Table 2. The hydrogen bond lengths as calculated with one water and four water models in combination with MP2/6-311++G(d,p) were 1.81 and 1.78 Å, respectively. These are consistent with our values of the smallest model in Figure 8a. However, they are 0.1 and 0.07 Å 3699
dx.doi.org/10.1021/jp211538v | J. Phys. Chem. C 2012, 116, 3694−3701
The Journal of Physical Chemistry C
Article
Table 2. Previous Theoretical Predictions of Hydrogen Bond Lengths (in Å) model
method
distance of H bond
HOCl−H2Oa
MP2/6-31G(d) MP2/6-311G(d,p) MP2/6-311++G(d,p) MP2/6-31+G(d) MP2/6-311G(d,p) MP2/6-311++G(d,p) BLYP/6-311++G(d,p)
1.806 (syn),1.803(anti) 1.733 (syn),1.768(anti) 1.809 (syn),1.807(anti) 1.82 to 1.89b 1.745c 1.781c 1.78(syn)d,1.79(anti)d
HOCl·(H2O)4
a
From ref 6a. bFrom ref 6b. cFrom ref 6c. dFrom ref 6d.
longer than our best value with (10/9/465) model (Figure 8h). The secondary hydrogen bond between the HOCl oxygen O31 and the surface hydrogen H21 appears in (4/6/0) model with 2.17 Å bond length in Figure 8a. However, as the model size increases up to (4/6/474) of Figure 8f, O31 of HOCl is forming a hydrogen bond with H15, implying that the secondary hydrogen bonding is sensitive to the models. Furthermore, when we increase quantum regions such as Figure 8g,h, the binding configuration of HOCl becomes parallel to the surface. Especially in the case of the largest quantum region of (10/9/ 465), the secondary hydrogen bond length becomes 2.12 Å, indicating that additional geometric relaxation in quantum region improves the secondary hydrogen bonding interactions. Overall, both the size of clusters and the size of quantum regions have large effects on the binding structures of HOCl. Effects of Surface Heterogeneity on the HOCl Adsorption. As seen in the previous section, the long-range electrostatics significantly increase the adsorption energies and changes the absorption structures, requiring the use of sufficiently large clusters. For the rest of our studies, therefore, the (10/9/465) clusters were mainly used. Previous theoretical works mainly considered HOCl adsorptions on the oxygen dangling bond (ODB). However, due to the surface heterogeneity, not only ODB but also HDB (hydrogen dangling bond) sites exist on the surface. In order to examine these effects, as well as hydrogen disorders of ice crystal and long-range interactions, three types of ice surface (SA, SB, and SC) were designed as explained in Ice Surface Models section. The resulted surface binding structures of HOCl on these along with binding energies are presented in Figure 9. It is noted that both BSSE corrected and uncorrected (in parentheses) binding energies as obtained with MP2/6-311++G(d,p) are presented. Figure 9a shows a HOCl adsorption structure on H1 site, which represents the hexagon site with one hydrogen (H17) dangling bond. On this site, HOCl is adsorbed through a hydrogen bond (O4−H33) and a secondary hydrogen bond of O31−H17 due to the hydrogen dangling bond (H17). The final binding configuration of HOCl is parallel to the surface. In the case of the H0 site (Figure 9b) where there is no HDB, HOCl is only adsorbed by a single hydrogen bond of O4−H33. As a result, HOCl forms perpendicular binding configuration, and its binding energy is smaller than that in Figure 9a by 3.7 kcal/mol. Although that in Figure 9b is less stable than that in Figure 9a, it is still 3.4 kcal/mol more stable than the four water cluster value of 8.8 kcal/mol.6a With two HDB as shown in Figure 9c, again HOCl binds the surface through two hydrogen bonds of O10−H33 and O31−H5 with parallel binding configuration. As a result, its binding energy is slightly higher than H1 (Figure 9a). Becuase of the three HDB, the H3 site does not have surface dangling oxygen to which HOCl hydrogen can bind. Instead,
Figure 9. Absorption structures and binding energies of HOCl on (a) SA−H1, (b) SA−H0, (c) SC−H2, (d) SC−H3, (e) SB−T1, (f) SC−T2 ,and (g) SC−T3 surface absorption sites. The (10/9/465) model with MP2/6-311++G(d,p) were used for all of the calculations. The BSSE corrected and uncorrected energies (in parentheses) are reported in kcal/mol. The atoms marked with green, red, and white colors are Cl, O, and H, respectively.
according to our calculations, the HOCl hydrogen is attached to the surface oxygen (O16−H33) with secondary hydrogen bond of O31−H6 as shown in Figure 9d. The formation of primary hydrogen bonding of O16−H33 yielded a penta coordinated oxygen site (O16) on ice surface. In general, 3700
dx.doi.org/10.1021/jp211538v | J. Phys. Chem. C 2012, 116, 3694−3701
The Journal of Physical Chemistry C
Article
Notes
HOCl binds more strongly to the H1 and H2 sites due to the secondary hydrogen bondings. However, it can also bind to H0 and H3 with slightly less binding energies. The adsorptions on the three tripod sites of T1, T2, and T3 were also studied. The primary (O4−H33) and secondary (O31− H20) hydrogen bonds are seen in the T1 site (Figure 9e). Similar hydrogen bonds of O13−H33 and O31−H21 are also seen in the T2 site of Figure 9f. It should be noted that the T2 site of Figure 8h is on the SA surface model, while that of Figure 9f is on the SC surface model. It appears even though they are the same T2, the binding energies are also dependent on the surface model used. As in the case of H2, HOCl can be more strongly bound to the T2 site. In the case of T3, there is no oxygen dangling bond to which the HOCl hydrogen can bind, but HOCl hydrogen H33 binds to the hydrogen dangling bond site (O13) with secondary hydrogen bond of O31−H5 creating a penta coordination. As a result, two of the existing hydrogen bonds (O13−H2 and O13−H12) become longer. According to our calculations, HOCl does not bind to the T0 site. The overall binding energies are between −12.2 and −17.8 kcal/mol (including Figure 8h) depending on the local binding sites. These values are 3.4−9.0 kcal/mol higher than the previous reference binding energy of 8.8 kcal/mol.6a However, our values are in good agreements with experimental data6g and MD study.6f It was seen that the surface binding configurations also depend on surface heterogeneity.
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS C.H.C. is indebted to Professor Heon Kang and Professor Rob Lahaye for inspiring discussions and for providing us reference ice coordinates. This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MEST) (No. 2011-0001213 and No. 20110005032).
■
■
CONCLUSIONS The effects of long-range electrostatics, hydrogen disorders of ice crystal, and heterogeneity of ice surface on chemical adsorptions of HOCl were studied with the help of proposed QM/EFP scheme. Our largest ice model consists of 19 quantum waters and 465 EFP waters. Various models were prepared to account for long-range effects and hydrogen disorder. In addition, various adsorption sites representing the surface heterogeneity were also systemically prepared. We have found that ice surface structures are strongly restricted by electrostatics of bulk ice. The same effects significantly increase the binding energies of HOCl. Our predicted HOCl binding energies are −12.2 to −17.8 kcal/mol depending on the particular surface absorption sites. These values are 3.4−9.0 kcal/mol higher than the previous reference binding energy as obtained with four water cluster models. However, they are in good agreement with experimental data6g and MD study.6f Structurally, the primary hydrogen bond lengths are reduced to 1.70−1.76 Å in our cluster calculations. Adsorption sites with two neighboring hydrogen dangling bonds (HDB) yielded larger HOCl binding energies, indicating the importance of secondary hydrogen bondings. We also found that a penta coordinated surface oxygen can be possible, when there is no oxygen dangling bond to which HOCl can be bound. In short, on the basis of HOCl binding energies, it can be concluded that ice surface is providing quite reactive environments to small gas molecules. Our study also indicates that proper inclusions of long-range electrostatics and surface heterogeneity are critical in the modeling of ice surface. As a way of introducing these terms, our hybrid QM/EFP turned out to be useful.
■
REFERENCES
(1) Farman, J. C.; Gardiner, B. G.; Shanklin, J. D. Nature 1985, 315, 207−210. (2) (a) Solomon, S.; Garcia, R. R.; Rowland, F. S.; Wuebbles, D. J. Nature 1986, 321, 755−758. (b) Molina, M. J.; Tso, T.-L.; Molina, L. T.; Wang, F. C.-Y. Science 1987, 238, 1253−1257. (3) Watanabe, N.; Kouchi, A. Prog. Surf. Sci. 2008, 83, 439−489. (4) Kang, H. Acc. Chem. Res. 2005, 38, 893−900. (5) (a) Chou, A.; Li, Z.; Tao, F.-M. J. Phys. Chem. A 1999, 103, 7848−7855. (b) Kim, Y. K.; Kim, S. K.; Kim, J. H.; Kang, H. J. Phys. Chem. C 2009, 113, 16863−16865. (c) Park, S.-C.; Moon, E.-S.; Kang, H. Phys. Chem. Chem. Phys. 2010, 12, 12000−12011. (6) (a) Dibble, T. S.; Francisco, J. S. J. Phys. Chem. 1995, 99, 1919− 1922. (b) Zhou, Y. F.; Liu, C. B. J. Phys. Chem. Solids 1999, 60, 2001− 2004. (c) Geiger, F. M.; Hicks, J. M.; de Dios, A. C. J. Phys. Chem. A 1998, 102, 1514−1522. (d) Brown, A. R.; Doren, D. J. J. Phys. Chem. B 1997, 101, 6308−6312. (e) Kroes, G.-J.; Clary, D. C. J. Phys. Chem. 1992, 96, 7079−7088. (f) Abbatt, J. P. D.; Molina, M. J. Geophys. Res. Lett. 1992, 19, 461−464. (g) Hanson, D. R.; Ravishankara, A. R. J. Phys. Chem. 1992, 96, 2682−2691. (7) (a) Cwiklik, L.; Devlin, J. P.; Buch, V. J. Phys. Chem. A 2009, 113, 7482−7490. (b) Labat, F.; Pouchan, C. Phys. Chem. Chem. Phys. 2009, 11, 5833−5842. (c) Rodziewicz, P.; Rutkowski, K. S.; Meyer, B. Phys. Chem. Chem. Phys. 2011, 13, 14101−14109. (8) Somnitz, H. Phys. Chem. Chem. Phys. 2009, 11, 1033−1042. (9) (a) Coulson, C. A.; Eisenberg, D. Proc. R. Soc. London, Ser. A 1966, 291, 445−453. (b) Chen, B.; Ivanov, I.; Klein, M. L.; Parrinello, M. Phys. Rev. Lett. 2003, 91, 215503/1−215503/4. (10) Kuo, J. L.; Klein, M. L.; Kuhs, W. F. J. Chem. Phys. 2005, 123, 134505/1−134505/6. (11) (a) Hobbs, P. V. Ice Physics; Charendon Press: Oxford, U.K., 1974. (b) Petrenko, V. F.; Whitworth, R. W. Physics of Ice; Oxford University Press: Oxford, U.K., 1999. (12) Materer, N.; Starke, U.; Barbieri, A.; VanHove, M. A.; Somorjai, G. A.; Kroes, G. J.; Minot, C. Surf. Sci. 1997, 381, 190−210. (13) (a) Jensen, J. H.; Day, P. N.; Gordon, M. S.; Basch, H.; Cohen, D.; Garmer, D. R.; Krauss, M.; Stevens, W. J. Effective Fragment Method for Modeling Intermolecular Hydrogen-Bonding Effects on Quantum Mechanical Calculations. In Modeling the Hydrogen Bond; Smith, D. A., Ed.; ACS Symposium Series 569; American Chemical Society: Washington DC, 1994; p 139. (b) Day, P. N.; Jensen, J. H.; Gordon, M. S.; Webb, S. P.; Stevens, W. J.; Krauss, M.; Garmer, D.; Basch, H.; Cohen, D. J. Chem. Phys. 1996, 105, 1968−1986. (c) Chen, W.; Gordon, M. S. J. Chem. Phys. 1996, 105, 11081−11090. (14) Netzloff, H. M.; Gordon, M. S. J. Chem. Phys. 2004, 121, 2711− 2714. (15) DeFusco, A.; Ivanic, J.; Schmidt, M. W.; Gordon, M. S. J. Phys. Chem. A 2011, 115, 4574−4582. (16) (a) Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert, S. T.; Gordon, M. S.; Jensen, J. H.; Koseki, S.; Matsunaga, N.; Nguyen, K. A.; Su, S.; Windus, T. L.; Dupuis, M.; Montgomery, J. A. J. Comput. Chem. 1993, 14, 1347−1363. (b) Fletcher, G. D.; Schmidt, M. W.; Gordon, M. S. Adv. Chem. Phys. 1999, 110, 267−294. (17) Boys, S. F.; Bernardi, F. Mol. Phys. 1970, 19, 553−566.
AUTHOR INFORMATION
Corresponding Author
*Tel: +82-53-950-5332. Fax: +82-53-950-6330. E-mail: cchoi@ knu.ac.kr. 3701
dx.doi.org/10.1021/jp211538v | J. Phys. Chem. C 2012, 116, 3694−3701