JOURNAL OF CHEMICAL PHYSICS VOLUME 113, NUMBER 13 1 OCTOBER 2000 Structure and vibrational spectra of H¿„H2O…8: Is the excess proton in a symmetrical hydrogen bond? Cristian V. Ciobanu Department of Physics, Ohio State University, 174 West 18th Avenue, Columbus, Ohio 43210 Lars Ojamäe Physical Chemistry, Arrhenius Laboratory, Stockholm University, S-106 91 Stockholm, Sweden Isaiah Shavitt and Sherwin J. Singera) Department of Chemistry, Ohio State University, 100 West 18th Avenue, Columbus, Ohio 43210 共Received 15 May 2000; accepted 23 June 2000兲 The energetics, structure, and vibrational spectra of a wide variety of H⫹共H2O兲8 structures are calculated using density functional theory and second-order Møller–Plesset ab initio methods. In these isomers of H⫹共H2O兲8 the local environment of the excess proton sometimes resembles a ⫹ symmetric H5O⫹ 2 structure and sometimes H3O , but many structures are intermediate between these two limits. We introduce a quantitative measure of the degree to which the excess proton ⫹ resembles H5O⫹ 2 or H3O . Other bond lengths and, perhaps most useful, the position of certain vibrational bands track this measure of the symmetry in the local structure surrounding the excess proton. The general trend is for the most compact structures to have the lowest energy. However, adding zero-point energy counteracts this trend, making prediction of the most stable isomer impossible at this time. At elevated temperatures corresponding to recent experiments and atmospheric conditions 共150–200 K兲, calculated Gibbs free energies clearly favor the least compact structures, in agreement with recent thermal simulations 关Singer, McDonald, and Ojamäe, J. Chem. Phys. 112, 710 共2000兲兴. © 2000 American Institute of Physics. 关S0021-9606共00兲30735-8兴 I. INTRODUCTION contains the excess proton bound symmetrically between two waters as in H5O⫹ 2 . In this work we consider a wider range of H⫹共H2O兲8 isomers and find no instances in which the minimum of the potential energy surface has the proton located precisely halfway between two water molecules. Even where imposed symmetry constraints force the proton into a symmetrical bond, we find the symmetric structure not to be a true local minimum. Nevertheless, we find some cases where the excess proton is rather close to a symmetrical arrangement. The H⫹共H2O兲8 cluster was also one of those previously chosen by some of us for thermal simulations17 using the OSS2 empirical potential18 for H⫹共H2O兲n . We generated many stable isomers of H⫹共H2O兲8 and, in this work, had hopes of establishing which were likely to be observed in low temperature beam experiments. We now know that many of these isomers lie so close in energy that even high level ab initio calculations do not have sufficient accuracy to authoritatively identify the global minimum. The study of protonated water clusters is further motivated by their abundance in the upper atmosphere19 and noctilucent clouds,20–22 and their presence in interstellar clouds.23 Bondybey and co-workers24,25 have studied the reactivity of HCl with protonated water clusters, a model for the early steps of ozone depletion reactions. Achatz et al.26 reported measurements of the rates of reaction of acetone and acetaldehyde with protonated water clusters. The protonated water clusters of interest in current laboratory experiments or in the atmosphere are not particularly cold. Schindler et al.24 estimated the temperature of their clusters Compared to their neutral counterparts, relatively little is known about protonated water clusters, H⫹共H2O兲n . After debate spanning many years 共reviewed by Ratcliffe and Irish1,2兲, experiment3 and theory4–6 agree that protonated water dimer, H5O⫹ 2 , is a symmetric structure in which the shared proton lies midway between two oxygen atoms. Ab initio calculations indicate that the excess proton in H⫹共H2O兲3 and H⫹共H2O兲4 is closely associated with one of the water molecules in a hydronium unit,5,7,8 although in H⫹共H2O兲4 there is an isomer only 3.5 kcal mol above the 5,7 global minimum containing a symmetric, H5O⫹ 2 -like bond. ⫹ ⫹ Similar H3O - and H5O2 -like isomers have been noted for H⫹共H2O兲6. 7,9 Wei and Salahub7 find the energy difference between the two forms of H⫹共H2O兲6 to be much smaller than for H⫹共H2O兲4, while Jiang et al.10 find the symmetric form to be markedly lower in energy. Beyond these small clusters, little is known conclusively, either experimentally or theoretically, about the structure of protonated water clusters. In contrast, the structure of neutral water clusters (H2O) n is well established through n⫽10.11–15 We have chosen to investigate H⫹共H2O兲8 in detail with an empirical potential and ab initio techniques for several reasons. From recent studies involving both vibrational spectroscopy and ab initio calculations, Chang et al.16 and Jiang et al.10 suggest that, at 170 K, a fraction of their cluster population of H⫹共H2O兲6 and, tentatively, also H⫹共H2O兲8 a兲 Electronic mail: singer@mps.ohio-state.edu 0021-9606/2000/113(13)/5321/10/$17.00 5321 © 2000 American Institute of Physics Downloaded 10 Jun 2002 to 128.148.49.77. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp 5322 J. Chem. Phys., Vol. 113, No. 13, 1 October 2000 Ciobanu et al. FIG. 2. Double-ring structures of H⫹共H2O兲8 observed in basin-hopping Monte Carlo 共Ref. 27兲 simulations using the OSS2 model 共Ref. 18兲. Structures a, b, and d are actually obtained from DFT/B3LYP structural optimizations, but at this level of depiction the OSS2 and DFT/B3LYP structures are indistinguishable. FIG. 1. Linear, branched, and single-ring structures of H⫹共H2O兲8 observed in basin-hopping Monte Carlo 共Ref. 27兲 simulations using the OSS2 model 共Ref. 18兲. Structures a, c, e, and f are actually obtained from DFT/B3LYP structural optimizations, but at this level of depiction the OSS2 and DFT/ B3LYP structures are indistinguishable. to be near 140 K, while Jiang et al.10 reported their clusters at 170 K. The temperature of polar stratospheric cloud particles has been measured19 to be about 190 K. Therefore, our previously reported thermal simulations17 are relevant to interpretation of these experiments. The present ab initio investigation characterizes the local potential minima which are visited by clusters near 200 K, and anticipates much needed experimental investigations of cold H⫹共H2O兲n clusters. Our method for generating starting structures using the OSS2 empirical potential18 is described in Sec. II. Ab initio methods and results are given in Sec. III. In Sec. IV we focus on a family of cluster isomers, the double five-membered rings, to illustrate the issues that arise as we search for local minima and attempt to characterize the local structure surrounding the excess proton. Having generated data on a relatively large number of H⫹共H2O兲8 isomers, in Sec. V we extract correlations between the local structure of the excess proton, whether it is H3O⫹- or H5O⫹ 2 -like, and the vibrational spectrum. We indeed find a vibrational signature that distinguishes between the H3O⫹- and H5O⫹ 2 -like environments. Finally, we discuss the relation of our work to previous theoretical and experimental studies. without side chains 共Fig. 1d兲, single rings with side chains 共Figs. 1c, 1e, 1f兲, double rings with and without side chains 共Fig. 2兲, triple rings 共Figs. 3a–3c兲, and quadruple ring 共Fig. 3d兲 structures. Within each topological class located in the basin-hopping simulations, there were generally many representatives that differed in the directionality of the hydrogen bonds and in the two orientations that two-coordinate waters can assume and still maintain tetrahedral bonding directions when they simultaneously accept and donate one H bond. 共The latter will be illustrated in Sec. IV.兲 All of the local minima of the OSS2 potential energy surface obtained from the basin-hopping simulations remained local minima at den- II. ISOMERS OF H¿„H2O…8 FROM THE OSS2 MODEL Our principle source of input structures for ab initio study of H⫹共H2O兲8 was a basin-hopping Monte Carlo simulation17 using the OSS2 potential18 for H⫹共H2O兲n , n ⫽8,16. The basin-hopping procedure27 was carried out at 400 K to help overcome the potential barriers separating local minima of H⫹共H2O兲8. It was continued until 1678 different minima were collected. Representative local minima include linear 共Fig. 1a兲, tree-like 共Fig. 1b兲, a single ring FIG. 3. Multiple-ring structures of H⫹共H2O兲8 observed in basin-hopping Monte Carlo 共Ref. 27兲 simulations using the OSS2 model 共Ref. 18兲 共a–d兲, and two cubic structures 共e, f兲 which are local minima of the OSS2 potential but are not observed in the basin-hopping simulations. Structures a, b, e, and f are actually obtained from DFT/B3LYP structural optimizations, but at this level of depiction the OSS2 and DFT/B3LYP structures are indistinguishable. Downloaded 10 Jun 2002 to 128.148.49.77. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp J. Chem. Phys., Vol. 113, No. 13, 1 October 2000 sity functional theory using the Lee–Yang–Parr correlation28 and Becke’s three-parameter exchange29–32 functionals 共DFT/B3LYP兲 and second-order Møller–Plesset33–36 共MP2兲 ab initio levels of theory. In fact, the structures are so similar that OSS2 and ab initio geometries are indistinguishable in visualizations like Figs. 1–3. The basin-hopping procedure allowed us to cast a wide net for locally stable structures but, like any numerical method, it does not guarantee a complete enumeration of local minima on the potential energy surface. In fact, cubic H⫹共H2O兲8 clusters 共Fig. 3, e, f兲 were never located in the basin-hopping search, although they are indeed local minima of the OSS2 potential.37 From an earlier study37 of (H2O) n and H⫹共H2O兲n using graph theoretical methods to enumerate H-bond topologies for a given structure, we know that there are precisely 11 possible H-bond arrangements for cubic H⫹共H2O兲8 with a hydronium ion at one vertex, and four possible H-bond topologies with the proton symmetrically placed along the bond between two waters. These results give the possible topologies, but not all may exist as local minima of a potential energy surface. Using the OSS2 model, we were able to find six 共out of a possible 11兲 actual minima with a hydronium at a vertex and two 共out of a possible four兲 actual minima with a symmetrical, H5O⫹ 2 -like bond. The lowest energy representative, according to the OSS2 model, from each of these two groups was selected for study with ab initio methods in this work. The cubic structure with an H5O⫹ 2 -like bond proved to be the one example of a local minimum of the OSS2 surface that did not optimize to a corresponding local minimum with ab initio methods. Instead the excess proton migrated to one side to form one of the eleven possible H3O⫹-containing water cubes. Incidentally, systematic graph theoretical methods can enumerate all the H-bonding topologies possible for each of the structural types found in our basin-hopping search, or actually all the topologies possible for H⫹共H2O兲8, but given the magnitude and complexity of analyzing all possible structures, these methods were not employed in this work. III. MP2 AND DFTÕB3LYP STUDIES We study the protonated water clusters at two different levels of theory, DFT/B3LYP28–32 and MP2. The geometry optimizations were performed with the ab initio packages, 38 NWCHEM version 3.2 for MP2 共with the largest gradient at convergence set to 0.0008 E h /a 0 兲 and GAUSSIAN9839 for DFT/B3LYP 共largest gradient 0.000 45 E h /a 0 兲. Geometry optimization was rather slow for the MP2 calculations, the number of steps taken being in the range of 100–200. The DFT calculations were less demanding and the geometry optimization was faster, so we could afford smaller tolerances for the energy gradient. Erratic geometrical optimization steps were observed for GAUSSIAN98 with symmetryconstrained DFT/B3LYP optimization unless we used a finer integration grid than the default choice. The starting structures for the geometry optimizations were local minima of the OSS2 energy surface. All geometry optimizations in this study, both B3LYP and MP2, were carried out with the aug-cc-pVDZ* basis Structure of H⫹(H2O)8 5323 previously used in calculations on water clusters by Tsai and Jordan.40,41 This basis is a modified form of the aug-ccpVDZ basis set of Dunning and co-workers.42,43 The modification consists of the deletion of the diffuse p functions on the hydrogen atoms, leaving a diffuse s function on each hydrogen and diffuse s, p, d functions on oxygen. Tsai and Jordan justified the deletion of the diffuse p functions on the hydrogen atoms by the observation that it had no significant effect on the binding energy of the water dimer. In conventional notation the primitive and contracted components of the basis can be described as (10s5 p2d)→ 关 4s3p2d 兴 on oxygen and (5s1 p)→ 关 3s1 p 兴 on hydrogen. In an attempt for more economical calculations, we also tested the smaller ccpvDZ set 共without diffuse functions兲 in the DFT calculations. We found that the geometries and the energy ordering of the structures were greatly affected by using the smaller basis, and our basis of choice remains aug-cc-pVDZ*. Only Cartesian Gaussians, with six components for each d function, were available in the computer program38 used to perform all MP2 calculations. This choice resulted in a total basis size of 302 contracted functions for H⫹共H2O兲8. The B3LYP calculations, performed with a different package,39 used spherical Gaussians, with five d components, resulting in a basis size of 286 contracted functions. Diffuse basis functions, at least on nonhydrogen atoms, were found to be important in calculations on hydrogenbonded systems by Del Bene and Shavitt.44,45 In particular, including diffuse functions was found to reduce the effects of basis-set superposition errors 共BSSEs兲. Correcting for BSSEs in calculations on the H⫹共H2O兲8 cluster, particularly during geometry optimization, would have been rather difficult, and therefore it was considered important to reduce such errors at the outset. Previous work on hydrogen-bonded systems45 共see also a detailed analysis by Xantheas46兲 showed that much of the BSSE can be eliminated with the use of the aug⬘-cc-pVTZ basis, which is a modified form of the aug-ccpVTZ basis of Dunning and co-workers.42,43 The modification in this case consists of omitting all diffuse functions on the hydrogen atoms. This basis was found to give binding energies close to those obtained with bigger basis sets.45 Single-point MP2 energy calculations were carried out with the aug⬘-cc-pVTZ basis for many of the H⫹共H2O兲8 structures at the optimized geometries obtained with the aug-cc-pVDZ* basis in order to obtain more reliable relative energies. The aug⬘-cc-pVTZ basis can be described in the conventional notation as (11s6 p3d2 f )→ 关 5s4 p3d2 f 兴 on oxygen and (5s2 p1d)→ 关 3s2 p1d 兴 on hydrogen, and using Cartesian Gaussians, with six d components and ten f components, adds up to 695 contracted functions for H⫹共H2O兲8. Table I shows the MP2 energies of a number of clusters calculated using the two basis sets, measured with respect to the cubelike structure 共f兲 of Fig. 3. The relative energies differ very little in the two cases 共less than 10%兲. We also find that the dissociation energies 共the last two columns of Table I兲 are in even better agreement, their values differing by less than 1% when we change the basis set. We conclude that our calculations are not affected by unacceptable basis set superposition errors. Downloaded 10 Jun 2002 to 128.148.49.77. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp 5324 Ciobanu et al. J. Chem. Phys., Vol. 113, No. 13, 1 October 2000 TABLE I. MP2 energies (E h ) relative to the energy of the protonated cubelike structure 共3e兲 and dissociation energies 共E h , last two columns兲 calculated using the double-zeta and triple-zeta basis sets. See the text for a discussion of these basis sets. ⌬E 0 ⌬E 0 E diss E diss Structure aug-cc-pVDZ* aug⬘-cc-pVTZ aug-cc-pVDZ* aug⬘-cc-pVTZ 3b 2b 1e 1f 2d 6b 1c 1a 0.003 163 99 0.004 795 85 0.005 051 23 0.005 982 16 0.006 017 02 0.006 309 11 0.007 285 19 0.020 679 43 0.002 836 70 0.004 604 22 0.004 869 25 0.005 722 65 0.005 651 89 0.006 210 61 0.007 034 34 0.019 682 15 0.207 602 25 0.205 970 40 0.205 715 02 0.204 784 09 0.204 749 22 0.204 457 12 0.203 485 82 0.190 086 81 0.207 932 77 0.206 165 26 0.205 900 23 0.205 046 82 0.205 117 58 0.204 558 86 0.203 735 14 0.191 087 33 A. Energetics We performed geometry optimization on a diverse array of H⫹共H2O兲8 structures, ranging from the floppy linear isomer 共1a兲 to rigid cubes like 共3e, 3f兲 共Figs. 1–3兲. The data are presented in Tables II–V. The general trends are captured in Fig. 4. The structures in Fig. 4 are arranged from left to right roughly from most floppy to most compact. Actually, they are ordered according to increasing Gibbs free energy at 190 K, which also places them almost perfectly in order of increasing zero-point energy. The general trend is for the most compact of the H⫹共H2O兲8 structures to have the lowest energy, as depicted by the E 0 values in Fig. 4, but not as distinctly favored as their neutral counterparts.11–15 As a result, when zero-point energy is added the result shows no clear trend and all the structures fall within about 2000 cm⫺1 TABLE III. Dissociation energy and structural data for H⫹共H2O兲8 calculated at the DFT/B3LYP level. Structures are ordered going from H5O⫹ 2 to H3O⫹ character. The designations in the first column refer to the figures in which the structures are shown. Structures followed by a point group symbol were constrained to have that symmetry during optimization. The two energy columns give the dissociation free energy at 190 K 共including ZPE兲 and the dissociation energy 共no ZPE included兲 to 7H2O⫹H3O⫹. The first three entries have at least one imaginary frequency and optimize to a structure with a slightly assymetrical H5O⫹ 2 unit upon release of symmetry constraints. The parameter f is defined in Eq. 共1兲. ⌬G diss 共kcal/mol兲 ⌬E diss(E h ) R OO⬘ 共Å兲 f 1f (C s ) ⫺46.3361 0.204 859 51 2.4055 0.0000 ⫹ ⫺ 6兵⫺ 兩 ⫹ 其 (C 2 ) ⫺49.0051 0.205 331 19 2.4106 0.0000 ⫺ ⫺ 6兵⫺ 兩 ⫺ 其 (C 2 ) ⫺49.5099 0.205 242 25 2.4111 0.0000 ⫹ ⫺ 6兵⫺ 兩 ⫹其 ⫺48.4701 0.205 329 28 2.4106 0.0365 1a 1f 6b 3f 3b 3a 3e 1e 1c ⫺43.0456 ⫺46.1049 ⫺50.0313 ⫺54.6953 ⫺52.3350 ⫺51.4745 ⫺55.1680 ⫺47.0298 ⫺46.8874 0.194 807 84 0.206 278 16 0.205 560 05 0.206 577 26 0.208 158 42 0.201 887 71 0.206 217 22 0.208 011 72 0.205 306 26 2.4154 2.4210 2.4484 2.4866 2.4720 2.5138 2.5236 2.5195 2.5229 0.0619 0.0705 0.1130 0.1413 0.1437 0.1688 0.1737 0.1796 0.1841 Structure of each other. These energies are too close to each other to make a clear prediction of the global minimum energy isomer of H⫹共H2O兲8. According to the estimated Gibbs free energy at either 170 or 190 K 共Table II兲, the more open structures are clearly favored at elevated temperature. The temperatures 170 and TABLE II. B3LYP 共aug-cc-pVDZ兲* energies of various H⫹共H2O兲8 geometries, reported in ascending order of total free energy at 190 K 共second column兲. Columns 3–6 represent free energy at 170 K, electronic energy, zero-point energy 共ZPE兲, and the sum of electronic and ZPE, respectively. Free energies are given in kcal/mol. The designations in the first column refer to the figures in which the structures are shown. Structures followed by a point group symbol were constrained to have that symmetry during optimization. Such a constraint would result in a perfectly symmetric H5O⫹ 2 unit, but in each case we found such constrained structures to be saddle ⫹ ⫺ ⫹ ⫺ points. The fact that structure 6 兵 ⫺ 兩 ⫹ 其 (C 2 ) lies lower in energy than 6 兵 ⫺ 兩 ⫹ 其 is apparently an artifact of numerical grid effects within GAUSSIAN98 共Ref. 39兲. When the grid parameter is increased from default (75 ⫻302) to high (100⫻24⫻48) the energy of the symmetry-constrained structure is higher than the unconstrained structure, ⫺612.024 010 697 vs ⫺612.024 014 162 E h . G 共190 K兲 G 共170 K兲 E 0 (E h ) ZPE (E h ) E 0 ⫹ZPE 1a 1f 1f (C s ) 1c 1e 92.5526 95.6119 95.8431 96.3944 96.5368 95.5808 98.4450 98.5283 99.2398 99.3657 ⫺612.013 517 349 ⫺612.024 987 669 ⫺612.023 569 021 ⫺612.024 015 774 ⫺612.026 721 227 0.202 838 0.205 105 0.204 073 0.207 138 0.207 093 ⫺611.810 679 349 ⫺611.819 882 669 ⫺611.819 496 021 ⫺611.816 877 774 ⫺611.819 628 227 ⫹ ⫺ 6兵⫺ 兩 ⫹其 97.9771 100.5954 ⫺612.024 038 787 0.206 887 ⫺611.817 151 787 ⫹ ⫺ 6兵⫺ 兩 ⫹ 其 (C 2 ) 98.5121 101.0247 ⫺612.024 040 702 0.206 557 ⫺611.817 483 702 ⫺ ⫺ 6兵⫺ 兩 ⫺其 99.0169 101.4261 ⫺612.023 951 762 0.206 232 ⫺611.817 719 762 99.5383 100.9815 101.8420 104.2023 104.6750 102.0656 103.4190 104.2019 106.4027 106.8418 ⫺612.024 269 555 ⫺612.020 597 225 ⫺612.026 867 931 ⫺612.025 286 769 ⫺612.024 926 731 0.208 724 0.210 198 0.210 810 0.213 057 0.213 410 ⫺611.815 545 555 ⫺611.810 399 225 ⫺611.816 057 931 ⫺611.812 229 769 ⫺611.811 516 731 Structure 6b 3a 3b 3f 3e Downloaded 10 Jun 2002 to 128.148.49.77. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp Structure of H⫹(H2O)8 J. Chem. Phys., Vol. 113, No. 13, 1 October 2000 5325 TABLE IV. Dissociation energies to 7H2O⫹H3O⫹ of a family of double five-ring H⫹共H2O兲8 clusters calculated using the OSS2 potential model. The structures are shown in Fig. 6. Structure ⌬E diss(E h ) ⫹ ⫺ 兩 ⫹其 兵⫺ 0.195 153 ⫺ ⫺ 兩 ⫹其 兵⫺ 0.194 997 ⫹ ⫺ 兩 ⫺其 兵⫹ 0.194 818 ⫺ ⫺ 兩 ⫺其 兵⫺ 0.194 798 ⫹ ⫹ 兩 ⫺其 兵⫺ 0.194 637 ⫺ ⫹ 兩 ⫹其 兵⫹ 0.194 555 190 K were chosen because they correspond to conditions reported for recent experiments10 and for polar stratospheric clouds,19 respectively. The free energies clearly favor the less compact structures at elevated temperature, and provide some confirmation of the break-up of ring-containing structures in this temperature range, as was predicted in thermal simulations using the OSS2 potential.17 However, our reported free energies should only be taken as a rough guide. The free energies reported in Table II are based on a harmonic model for internal motion within the cluster. Thermal simulations17 indicate that very large-amplitude motion, including barrier crossings between local minima of the potential energy surface, is prevalent in this temperature range. Therefore, free energies calculated from a harmonic model have to be used with caution in selecting the most probable structure at elevated temperature. DFT/B3LYP, MP2, and OSS2 are in rough agreement with respect to the trends in energy values (E 0 ). The major discrepancy between the various models occurs in the energy of the most compact structures. MP2 calculations favor the protonated cubes 共Figs. 3e, 3f兲 as the lowest energy structure, while DFT/B3LYP and OSS217 both favor the somewhat more open three-ring structure 3b. We do not report zeropoint energies at the MP2 level since the size of our basis sets made frequency calculations at this level impractical. TABLE V. Energies and structural data for H⫹共H2O兲8 calculated at the MP2 ⫹ level. Structures are ordered going from H5O⫹ 2 to H3O character. The designations in the first column refer to the figures in which the structures are shown. The two energy columns give the energy and the dissociation energy 共no ZPE included兲 to 7H2O⫹H3O⫹. The parameter f is defined in Eq. 共1兲. Structure E 0 (E h ) ⌬E diss(E h ) R OO⬘ 共Å兲 f ⫹ ⫺ 6兵⫺ 兩 ⫹其 ⫺610.575 261 5 0.206 000 571 2.422 52 0.069 412 1f 1a 6b 3f 3b 3a 3e 1e 2d 1c 2b ⫺610.574 045 0 ⫺610.559 347 7 ⫺610.573 718 0 ⫺610.579 995 9 ⫺610.576 863 1 ⫺610.572 379 9 ⫺610.580 027 1 ⫺610.574 975 9 ⫺610 574 975 9 ⫺610 572 746 7 ⫺610 575 231 3 0.204 784 089 0.190 086 807 0.204 457 123 0.210 734 953 0.207 602 248 0.203 119 028 0.210 766 237 0.205 715 018 0.204 749 216 0.203 485 817 0.205 970 396 2.425 21 2.427 70 2.456 56 2.487 16 2.482 94 2.516 14 2.526 72 2.521 36 2.527 16 2.532 00 2.535 00 0.083 857 0.087 312 0.126 399 0.144 927 0.156 992 0.172 954 0.177 666 0.180 728 0.185 331 0.191 245 0.193 930 FIG. 4. Energy trends among isomers of H⫹共H2O兲8. The designations along the horizontal axis refer to the figures in which the structures are shown. The three solid curves depict the energy E 0 of various structures using DFT/ B3LYP 共䊊兲, MP2 共䉭兲, and the OSS2 potential 共䊐兲. The zero point energy 共dot-dash兲, E 0 ⫹ZPE 共long dash兲, and Gibbs free energy at 190 K 共short dash兲 are all calculated by DFT/B3LYP. All energies are referred to the corresponding energy of structure la to facilitate comparison. The general trend is for structures to become more compact going from left to right. E 0 generally decreases in this direction, but, counteracting this trend, zero-point energy increases and entropy decreases. B. Is the excess proton in a symmetrical H bond? There is continued speculation as to whether the excess proton is located symmetrically or nonsymmetrically between two oxygen atoms for the larger proton hydrate clusters and, by extension, for the excess proton in bulk. In reality there is a continuum of possible structures between the hydronium and H5O⫹ 2 limits. We identified the hydronium or the H5O⫹ 2 unit within our structures by monitoring bond length. The degree to which the configuration of the excess proton within our set of H⫹共H2O兲8 clusters resembles H3O⫹ or H5O⫹ 2 is quantified by defining a simple measure of symmetry of the excess proton’s environment 共Fig. 5兲. By inspection of bond lengths, all the hydrogen atoms in our H⫹共H2O兲8 clusters can be associated with a single oxygen, forming either water or hydronium units, or if a hydrogen is equally close to two oxygen neighbors, a perfectly symmetric H5O⫹ 2 unit. When a hydronium unit can be identified, let O⬘ designate the closest oxygen atom to the oxygen of the hydronium, so that O–H¯O⬘ is the shortest hydrogen bond formed with the hydronium, and let r OH and r OH⬘ be the oxygen–hydrogen bond distances within that hydrogen bond, as shown in Fig. 5. Note that by definition r OH⭐r O⬘ H⬘ , with the equality applicable in the symmetrical limit. Then a dimensionless measure of the symmetry of the excess proton’s environment is f⬅ r O⬘ H⫺r OH , r OH⬘ ⫹r OH 共1兲 which smoothly interpolates from values near 0.2 in the H3O⫹ limit to zero as the excess proton’s environment approaches the symmetrical, H5O⫹ 2 -type limit. The length R OO⬘ of the shortest hydrogen bond associated with the hydronium unit correlates quite well with f, our Downloaded 10 Jun 2002 to 128.148.49.77. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp 5326 Ciobanu et al. J. Chem. Phys., Vol. 113, No. 13, 1 October 2000 FIG. 5. The hydrogen bond to the excess proton contracts as the excess proton changes from H3O⫹- to H5 O⫹ 2 -like in character. The parameter f measures the asymmetry of this bond, and vanishes in the H5 O⫹ 2 -like limit. R OO ⬘ is the length of the shortest H bond to the hydronium unit. Closed and open points are from DFT/B3LYP and MP2 calculations, respectively. The nearly overlapping points at f ⫽0 are from symmetry-constrained optimizations, and correspond to saddle-point structures. The smooth curve is a quadratic fit to the data R OO⬘ (Å)⫽2.405⫹3.595f 2 . Agmon 共Ref. 47兲 has used bond-energy–bond-order 共BEBO兲 relations 共Refs. 48–50兲 to correlate oxygen–hydrogen bond lengths within a hydrogen bond with the overall hydrogen bond length. The dashed line shows the prediction using Agmon’s empirical parameters for water clusters and assuming that the hydrogen bonds depart uniformly from linearity by ⬃11°, as discussed in the text. measure of H3O⫹/H5O⫹ 2 character, as shown in Fig. 5. The relationship is approximately fit by R OO⬘ (Å)⫽2.405 ⫹3.595f 2 , for both MP2 and DFT/B3LYP data. Also apparent from Fig. 5 is that the H⫹共H2O兲8 isomers we have chosen to study yield a distribution of f values, more or less dense ⫹ between the H5O⫹ 2 and H3O limits. We have seen in previously reported thermal simulations17 that above about 180 K the H⫹共H2O兲8 cluster is found almost exclusively in hydronium-like configurations. Here we find there is no shortage of local minima of either H3O⫹-, H5O⫹ 2 -like, or intermediate character. However, as mentioned previously, we have not discovered a perfectly symmetrical H5O⫹ 2 -like local minimum of H⫹共H2O兲8. It is possible to provide some interpretation of the trend shown in Fig. 5. Agmon47 has used bond-energy–bond-order 共BEBO兲 relations48–50 to correlate oxygen–hydrogen bond lengths within a hydrogen bond with the overall hydrogen bond length. In our terminology, the BEBO relations become n⫽exp关 ⫺ 共 r OH⫺r s 兲 /a 兴 , n ⬘ ⫽exp关 ⫺ 共 r O⬘ H⫺r s 兲 /a 兴 , n⫹n ⬘ ⫽1, 共2兲 where n and n ⬘ are the bond orders of the colavent and H-bonded oxygen–hydrogen bonds, whose sum is con- served. Agmon found that r s ⫽0.956 Å and a⫽0.385 Å fit data from water clusters.47 Relations 共2兲 are sufficient to predict R OO⬘ as a function of f provided we assume a linear hydrogen bond (r OH⫹r O⬘ H⫽R OO⬘ ). Compared with our calculations, this gives the trend in R OO⬘ vs f quite well, although the values of R OO⬘ are about 8.5% too large. Scaling the BEBO prediction of R OO⬘ uniformly by 0.9815, that is, assuming that the hydrogen bonds uniformly depart from linearity by ⬃11°, produces the best agreement with the ab initio data and is shown by the dashed line in Fig. 5. While the hydrogen bond involving the excess proton does depart from linearity by this magnitude in some of the isomers, like 3b and 3e, the departure from linearity is usually not this great. The necessity for the scaling factor more accurately reflects that the BEBO parameters should be slightly altered for the clusters studied in this work. Since bond lengths clearly indicate an H3O⫹ or H5O⫹ 2 unit as special within our optimized structures, we speak of these units as containing the ‘‘excess proton.’’ This language is flawed because it suggests that the excess charge resides on one or several hydrogens of the H3O⫹ or H5O⫹ 2 unit. In actuality the excess charge, as estimated by Mulliken charges,51,52 is dispersed over the entire H3O⫹ or H5O⫹ 2 unit and beyond, as illustrated by the following discussion. In our density functional calculations, the H3O⫹ typically contains excess charge close to 0.78e (e⫽ 兩 electron charge兩 ), with the rest of the excess charge 共from Mulliken population analysis兲 spread over neighboring water molecules. Water molecules which are far from the excess charge have Mulliken populations of roughly ⫹0.22e and ⫺0.44e on hydrogen and oxygen, respectively. We estimate how the 0.78e excess charge is distributed among the hydrogen and oxygen atoms of the hydronium unit by comparing their Mulliken populations with those of waters far from an excess charge, and find that the excess charge is rather evenly distributed among the atoms of the hydronium unit. From ⬃0.2e to ⬃0.25e resides on the oxygen and the remainder is located on the three hydrogens of the H3O⫹ unit. For example, the Mulliken charges associated with the H3O⫹ unit of the protonated cube structure, Fig. 3e, are ⫺0.26e on the oxygen and ⫹0.28e, ⫹0.37e, and ⫹0.37e for the hydrogens. The charge on the hydrogens is especially variable, and it is not uncommon for the charge on hydrogens far from the excess proton to exceed the charge on one of the hydrogens of the H3O⫹ unit. Turning now to the symmetrical limit, one typically finds 0.85e on the H5O⫹ 2 unit with ⬃0.1e on each of the oxygens, ⬃0.20e on the central hydrogen, and the remainder on the other four hydrogens of the H5O⫹ 2 unit. With the caveat that the location of what we call the ‘‘excess proton’’ does not indicate a localization of the excess charge on a single atom, and for lack of better terminology and measure of excess charge,52 we shall continue to use this language. The principle advantage of singling out the hydrogen in the shortest H bond to the hydronium is that it affords us a definition that smoothly connects the hydronium and H5O⫹ 2 -type limit. Downloaded 10 Jun 2002 to 128.148.49.77. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp Structure of H⫹(H2O)8 J. Chem. Phys., Vol. 113, No. 13, 1 October 2000 5327 ⫺ ⫺ 兩 ⫺ 其 , according to whether the dangling hydrogens point in 兵⫺ FIG. 6. A small sampling of the many local minima of the OSS2 potential with the topology of two edge-sharing five-membered rings. The top six structures have the same overall hydrogen bond topology, and only differ in the orientation of the dangling hydrogen on the the four single-acceptor/ single-donor waters. The notation describes this orientation, with a plus or minus sign indicating the position of the dangling hydrogen with respect to the plane of the rings and the horizontal line separating the two rings. The OSS2 potential has local minima for a wide variety of other edge-sharing double five-rings, and the two structures at the bottom are of possible interest either because of near C 2 symmetry 共a兲 or because an H5 O⫹ 2 -type bond appears to be forming in a position different from the edge shared by the two five-rings 共b兲. IV. DOUBLE FIVE-RING H¿„H2O…8 A family of double five-ring isomers 共Fig. 6兲 illustrates the subtleties of locating local minima and characterizing the symmetry of the excess proton’s environment. A double five-ring structure of this type for H⫹共H2O兲8 was, to our knowledge, first identified by Cheng.9 This structure was also examined recently by Jiang et al.,10 and identified as their prime candidate for the isomer of H⫹共H2O兲8 observed in their beam experiments. Actually, the structures depicted by Cheng and by Jiang et al. are not the same, and a wide variety of local minima based on the simple double five-ring structure are found, some with a clear hydronium-like or H5O⫹ 2 -like excess proton and most with intermediate character. In some structures the excess charge can be found in the H bond shared between the two rings, and in others on one of the two rings. In optimizations using both the OSS2 potential and the ab initio calculations, the double five-ring tends to pucker. We define the ⫹z direction as the direction perpendicular to the shared H bond which makes the puckered rings concave upward. We first examine a family of double five-ring structures in which the end waters furthest from the shared H bond are double acceptors 共the top six structures of Fig. 6兲. The four water molecules which are neither end waters nor part of the shared H bond are single acceptors/single donors. If one follows the simple notion that water molecules accept H bonds to one of two lone pairs tetrahedrally disposed around the oxygen, as is borne out in ab initio and experimental data,53 then the dangling hydrogens of the four single-acceptor/single-donor waters will point either above or below the five-ring, depending on which lone pair accepts the H bond. We label the seven possible arrangements within ⫹ ⫹ ⫹ ⫹ ⫹ ⫺ ⫹ ⫺ ⫹ ⫹ ⫺ ⫺ 兩 ⫹其 , 兵 ⫺ 兩 ⫹其 , 兵 ⫹ 兩 ⫺其 , 兵 ⫺ 兩 ⫹其 , 兵 ⫺ 兩 ⫺其 , 兵 ⫹ 兩 ⫺其 , this family as 兵 ⫹ the ⫹z or ⫺z direction. The horizontal line separating the two rows corresponds to the shared edge of the two fiverings, and the plus signs and minus signs give the orientation of the dangling hydrogens with the ⫹z direction pointing from the page to the reader. Also note that structures like ⫹ ⫺ ⫺ ⫹ 兩 ⫹ 其 and 兵 ⫹ 兩 ⫺ 其 are enantiomers. 兵⫺ Six of the possible seven structures listed above were located as minima of the OSS2 potential energy surface, the one omission being the structure with all plus signs. The energies of these structures under the OSS2 potential model are given in Table IV. The structures are quite close in energy. Associating an energy cost with dangling hydrogens in close proximity accounts for the trend in energy, just as we found for other clusters.37 Dangling hydrogens that are adjacent, either across a five-ring or found on the same side of two five-rings, prefer to be on opposite sides of the rings, ⫹ ⫺ 兩 ⫹ 其 is most stable. Dangling hydrogens on explaining why 兵 ⫺ the same side of the five-rings are best tolerated on the convex side of the structure than the concave side, since dangling hydrogens are closer together on the concave side. This fact may explain why having all dangling hydrogens on the ⫹ ⫹ 兩 ⫹ 其 produces an unstable structure rather concave side in 兵 ⫹ than a local minimum on the OSS2 surface. Three of the ⫹ ⫺ ⫺ ⫺ ⫹ ⫺ 兩 ⫹其 , 兵 ⫺ 兩 ⫺ 其 , and 兵 ⫹ 兩 ⫺ 其 , because of their located structures, 兵 ⫺ near C 2 , C 2 v , and C s symmetry, respectively, are candidates for having the excess proton in a perfectly symmetrical environment within the shared H bond. If any of the local minima turned out to have precisely these point group symmetries, then the central H5O⫹ 2 -like bond would have been perfectly symmetrical. However, the optimized OSS2 potential structures broke symmetry in each case, and hence the H5O⫹ 2 unit departs slightly from perfect symmetry. We attempted to optimize the first two such structures under C 2 symmetry using DFT/B3LYP, and in both cases symmetryconstrained optimization led to structures that had imaginary vibrational frequencies and were not true local minima. We have included the symmetry-constrained minima as data points in Fig. 5 to discover trends in H-bond lengths and vibrational spectra versus f. Structure a in Fig. 6 illustrates that there are other double five-ring structures that are candidates for a perfectly symmetrical H5O⫹ 2 -like bond besides the family considered to this point. However, like all other structures, perfect symmetry, C 2 in this case, is broken in structure a within the OSS2 model. Structure b in Fig. 6 illustrates that an H5O⫹ 2 -like excess proton may be found in various locations, not just the shared edge of the double five-ring. V. VIBRATIONAL SPECTRA Our data set of H⫹共H2O兲8 local minima, spanning a range of structures, allows us to explore whether there is a spectroscopic signature of the symmetry of the excess proton’s environment in the vibrational spectrum. The answer is that there is indeed such a signature in the vibrational modes below 2800 cm⫺1. In Fig. 7 the vibrational spectra of 12 isomers of H⫹共H2O兲8, obtained from DFT/B3LYP calculations in the Downloaded 10 Jun 2002 to 128.148.49.77. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp 5328 J. Chem. Phys., Vol. 113, No. 13, 1 October 2000 Ciobanu et al. FIG. 7. Infrared spectra of H⫹共H2O兲8 clusters, arranged from top to bottom according to decreasing symmetry of the excess proton’s environment as measured by the parameter f 关Eq. 共1兲兴. The spectra are obtained from DFT/B3LYP calculations in the harmonic approximation. The label where the corresponding structure can be located is given toward the left-hand side of each spectrum. The mode with imaginary frequency in the top spectrum is placed on the negative frequency axis. The intense vibrations below 2800 cm⫺1 move to higher frequency with increasing f, while the vibrations above 2800 cm⫺1 are relatively insensitive to the symmetry of the excess proton’s environment. This trend is quantified in Fig. 8. harmonic approximation, are arranged in order of increasing f, that is, in order from most symmetrical to least symmetrical excess proton. The top spectrum of Fig. 7 is that of the ⫹ ⫺ 兩 ⫹ 其 in Fig. 6兲 optimized under double five-ring 共labeled 兵 ⫺ the constraint of C 2 symmetry. The excess proton in this structure is exactly equidistant between two oxygens. The intense infrared active modes in the 690–1000 cm⫺1 range are true proton transfer modes: The proton shuffles back and forth between the two oxygens, its motion strongly coupled to bends and wags of the outer hydrogens of the H5O⫹ 2 unit. These modes are least likely to be harmonic in nature, since the effective potential for the central proton is known to be rather flat.5,54 The strong IR-active modes above 2800 cm⫺1 involve very little central proton motion. They are principally stretching of the outer OH bonds of the H5O⫹ 2 unit. As mentioned earlier, the double five-ring designated as ⫹ ⫺ 兩 ⫹ 其 is not stable in C 2 symmetry. The symmetry breaking 兵⫺ can be viewed as a consequence of the effective potential for the central excess proton exhibiting a double-well character with a maximum at the midpoint. The structure, energy, and distribution of vibrational frequencies in the symmetry con⫹ ⫺ 兩 ⫹ 其 共the two top strained and unconstrained versions of 兵 ⫺ spectra in Fig. 7兲 are quite similar, so we suppose that the double-well character is not very pronounced. As one proceeds to structures with increasing values of f 共toward spectra at the bottom of Fig. 7兲 either the double-well character becomes increasingly pronounced, and the excess proton is gradually localized on one side of the double well, or the symmetric well shifts to one much deeper toward one side. The cluster geometry determines which scenario applies. What was a proton transfer mode for the near-symmetric cluster turns, with increasing f, to localized vibrations of an H3O⫹ unit 共with tunnel splittings if the effective double-well is symmetric兲. The trend from symmetric to nonsymmetric excess pro⫹ ton, from an H5O⫹ 2 - to H3O -like configuration, is clearly reflected in the IR-active modes below 2800 cm⫺1. As the motion of the excess proton evolves from proton transfer to H3O⫹ stretch in character, the locus of intensely IR-active modes shifts to higher frequency and merges with what were, at lower f, stretches of the outer OH bonds of an H5O⫹ 2 unit. This is to be expected, since in the H3O⫹-like limit there is no distinction between a central proton and outer hydrogens. This trend is quantified in Fig. 8. The vibrational modes were separated into groups above and below 2800 cm⫺1 and the intensity-weighted average frequency Downloaded 10 Jun 2002 to 128.148.49.77. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp Structure of H⫹(H2O)8 J. Chem. Phys., Vol. 113, No. 13, 1 October 2000 FIG. 8. The intensity-weighted average vibrational frequency of modes below and above 2800 cm⫺1 共closed and open symbols, respectively兲 are plotted as a function of f, the measure of the symmetry of the excess proton’s environment. The solid lines are linear least-squares fits to the data. The average frequency of the modes below 2800 cm⫺1 increases by about 1000 cm⫺1 in the range of structures studied, while the modes above 2800 cm⫺1 have nearly constant average frequency. The linear least-squares fit to the data below 2800 cm⫺1 is 具 共cm⫺1 ) 典 ⫽778⫹5366f . 具典⫽ 兺 i I i i 兺i Ii 共3兲 of the two groups separately calculated. 共In this equation, I i is the IR intensity of vibrational mode with frequency i .兲 The modes above 2800 cm⫺1 have nearly unchanging average frequency, while those below 2800 cm⫺1 increase by ⫹ about 1000 cm⫺1 going from the H5O⫹ 2 to the H3O limit. The average frequency of the latter group is roughly fit by 具 (cm⫺1) 典 ⫽778⫹5366f . The degree of symmetry in the environment of the excess proton can therefore be extracted from vibrational spectra below ⬃2800 cm⫺1. 5329 tween two oxygens in two equivalent directions. Hence the proton transfer coordinate is effectively a symmetric double well. If the zero-point energy along this coordinate is in excess of the barrier height separating the two wells, similar to a dynamic Jahn–Teller effect, then the ground state wave function will have significant density in the symmetrical region. This effect was observed by Ojamäe et al. in reduced5 dimenional model calculations on the H5O⫹ 2 ion. At a slightly stretched oxygen–oxygen separation of 2.6 Å the proton transfer potential has a small barrier, yet the vibrational wave function has a single maximum. Chang and co-workers55 have pointed out another factor which may tend to drive the excess proton away from a nonsymmetrical potential energy minimum and possibly toward a more symmetric average position. Viewing the proton transfer coordinate in an adiabatic picture, they calculated the zero-point energy of several protonated dimethyl ether– water 关 H⫹关共CH3兲2O兴共H2O兲n 兴 clusters as a function of the proton transfer coordinate. They found that, in some cases, the zero-point energy changes significantly along the coordinate, producing a significant shift in the average position of the excess proton. The results presented here and the complementary results of thermal simulations using the OSS2 model17 indicate that care must be taken in the interpretation of vibrational spectra of H⫹共H2O兲8 in the currently accessible temperature range.10 The data of Fig. 7 show that there is no clear signa⫺1 ture of an H5O⫹ 2 -like excess proton above 2800 cm , unlike ⫺1 the marker at ⬃3180 cm that Jiang et al. could exploit for H⫹共H2O兲6. In Fig. 7 one finds vibrational peaks in this fre⫹ quency range in both the top 共H5O⫹ 2 limit兲 and bottom 共H3O limit兲 of the figure. The clear signature of symmetrical or nonsymmetrical excess proton is instead found at lower frequency. VI. DISCUSSION ACKNOWLEDGMENTS In recent work9,10,37 there has perhaps been too much emphasis on categorizing the excess proton in H⫹共H2O兲n as either H3O⫹- or H5O⫹ 2 -like. This article shows that, at least for H⫹共H2O兲8, a continuum of structures between these two limits is possible, and where one draws the boundary between the two types is somewhat arbitrary. At the level of theory we employed we did not find a single truly symmetri⫹ cal H5O⫹ 2 -like excess proton in H 共H2O兲8, although there were examples in which the deviation from such symmetry was small. The wide variety of cluster isomers we had studied has enabled us to identify qualitative trends which will aid in the interpretation of experiments: The degree of asymmetry of the excess proton’s environment is correlated with the bond distance to the nearest water molecule, and, most important, there is a clear upward shift in the vibrations be⫹ low 2800 cm⫺1 from the H5O⫹ 2 to the H3O limit. Even though we never found the minimum of the potential energy surface where the excess proton is precisely halfway between two water molecules, the possibility remains that zero-point vibrational effects may cause some isomers of H⫹共H2O兲8 to be effectively symmetric.55 In isomers like ⫹ ⫺ 兩 ⫹ 其 the excess proton can depart from the midpoint be兵⫺ The calculations reported here were made possible by resource grants from the EMSL Molecular Science Computing Facility at Pacific Northwest National Laboratory, the Swedish National Supercomputer Center 共NSC兲, and the Ohio Supercomputer Center. The authors wish to thank Robert Harrison and David Bernholdt for useful discussions and technical assistance. This research was partially supported by the Swedish Natural Science Research Council 共NFR兲. We are grateful to Huan-Cheng Chang for bringing Ref. 55 to our attention, and for alerting us to the possibility that zeropoint vibrational effects may cause some isomers to be effectively symmetric. 1 C. I. Ratcliffe and D. E. Irish, in Water Science Reviews, edited by F. Franks 共Cambridge University Press, Cambridge, 1986兲, Vol. 2, p. 149. 2 C. I. Ratcliffe and D. E. Irish, in Water Science Reviews, edited by F. Franks 共Cambridge University Press, Cambridge, 1988兲, Vol. 3, p. 1. 3 M. Okumura, L. I. Yeh, J. D. Myers, and Y. T. Lee, J. Chem. Phys. 85, 2328 共1986兲; L. I. Yeh, M. Okumura, J. D. Myers, J. M. Price, and Y. T. Lee, ibid. 91, 7319 共1989兲; M. Okumura, L. I. Yeh, J. D. Myers, and Y. T. Lee, J. Phys. Chem. 94, 3416 共1990兲; L. I. Yeh, Y. T. Lee, and J. T. Hougen, J. Mol. Spectrosc. 164, 473 共1994兲. 4 Y. Xie, R. B. Remington, and H. F. Schaefer III, J. Chem. Phys. 101, 4878 共1994兲. Downloaded 10 Jun 2002 to 128.148.49.77. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp 5330 5 Ciobanu et al. J. Chem. Phys., Vol. 113, No. 13, 1 October 2000 L. Ojamäe, I. Shavitt, and S. J. Singer, Int. J. Quantum Chem., Quantum Chem. Symp. 29, 657 共1995兲. 6 E. F. Valeev and H. F. Schaefer III, J. Chem. Phys. 108, 7197 共1998兲. 7 D. Wei and D. R. Salahub, J. Chem. Phys. 101, 7633 共1994兲. 8 D. Wei and D. R. Salahub, J. Chem. Phys. 106, 6086 共1997兲. 9 H.-P. Cheng, J. Phys. Chem. A 102, 6201 共1998兲. 10 J.-C. Jiang, Y.-S. Wang, H.-C. Chang, S. H. Lin, Y. T. Lee, G. NiednerSchatteburg, and H.-C. Chang, J. Am. Chem. Soc. 122, 1398 共2000兲. 11 K. Liu, M. G. Brown, C. Carter, R. J. Saykally, J. K. Gregory, and D. C. Clary, Nature 共London兲 381, 501 共1996兲. 12 K. Liu, J. D. Cruzan, and R. J. Saykally, Science 271, 929 共1996兲. 13 C. J. Gruenloh, J. R. Carney, C. A. Arrington, T. S. Zwier, S. Y. Fredericks, and K. D. Jordan, Science 276, 1678 共1997兲. 14 U. Buck, I. Ettischer, M. Melzer, V. Buch, and J. Sadlej, Phys. Rev. Lett. 80, 2578 共1998兲. 15 J. Sadlej, V. Buch, J. K. Kazimirski, and U. Buck, J. Phys. Chem. A 103, 4933 共1999兲. 16 H.-C. Chang, J.-C. Jiang, H.-C. Chang, Y.-S. Wang, S. H. Lin, and Y. T. Lee, J. Chin. Chem. Soc. 共Taipei兲 46, 427 共1999兲. 17 S. J. Singer, S. McDonald, and L. Ojamäe, J. Chem. Phys. 112, 710 共2000兲. 18 L. Ojamäe, I. Shavitt, and S. J. Singer, J. Chem. Phys. 109, 5547 共1998兲. 19 J. Schreiner, C. Voigt, A. Kohlmann, F. Arnold, K. Mauersberger, and N. Larsen, Science 283, 968 共1999兲. 20 L. G. Björn et al., J. Geophys. Res. D 90, 7985 共1985兲. 21 E. Kopp, P. Eberhardt, U. Herrmann, and L. G. Björn, J. Geophys. Res. D 90, 13041 共1985兲. 22 T. Sugiyama, J. Geophys. Res. A 99, 3915 共1994兲. 23 W. W. Duley, Astrophys. J. 471, L57 共1996兲. 24 T. Schindler, C. Berg, G. Niedner-Schatteburg, and V. E. Bondybey, Chem. Phys. Lett. 229, 57 共1994兲. 25 C. Berg, U. Achatz, M. Beyer, S. Joos, G. Albert, T. Schindler, G. Niedner-Schatteburg, and V. E. Bondybey, Int. J. Mass Spectrom. Ion Processes 167, 723 共1997兲. 26 U. Achatz, S. Joos, C. Berg, T. Schindler, M. Beyer, G. Albert, G. Niedner-Schatteburg, and V. E. Bondybey, J. Am. Chem. Soc. 120, 1876 共1998兲. 27 Z. Li and H. A. Scheraga, Proc. Natl. Acad. Sci. U.S.A. 84, 6611 共1987兲. 28 C. Lee, W. Yang, and R. G. Parr, Phys. Rev. B 37, 785 共1988兲. 29 A. D. Becke, J. Chem. Phys. 96, 2155 共1992兲. 30 A. D. Becke, J. Chem. Phys. 97, 9173 共1992兲. 31 A. D. Becke, J. Chem. Phys. 98, 1372 共1993兲. 32 A. D. Becke, J. Chem. Phys. 98, 5648 共1993兲. 33 R. J. Bartlett and D. M. Silver, Int. J. Quantum Chem., Symp. 8, 271 共1974兲. 34 R. J. Bartlett and D. M. Silver, Int. J. Quantum Chem., Symp. 9, 183 共1975兲. 35 J. S. Binkley and J. A. Pople, Int. J. Quantum Chem. 9, 229 共1975兲. 36 J. A. Pople, J. S. Binkley, and R. Seeger, Int. J. Quantum Chem., Symp. 10, 1 共1976兲. 37 S. McDonald, L. Ojamäe, and S. J. Singer, J. Phys. Chem. A 102, 2824 共1998兲. 38 J. Anchell, E. Apra, D. Bernholdt, P. Borowski, T. Clark, D. Clerc, H. Dachsel, M. Deegan, M. Dupuis, K. Dyall, G. Fann, H. Fruchtl, M. Gutowski, R. Harrison, A. Hess, J. Jaffe, R. Kendall, R. Kobayashi, R. Kutteh, Z. Lin, R. Littlefield, X. Long, B. Meng, J. Nichols, J. Nieplocha, A. Rendall, M. Stave, T. Straatsma, H. Taylor, G. Thomas, K. Wolinski, and A. Wong, NWCHEM, A Computational Chemistry Package for Parallel Computers, Version 3.2.1 共Pacific Northwest National Laboratory, Richland, WA 1998兲. 39 M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, V. G. Zakrzewski, J. A. Montgomery, Jr., R. E. Stratmann, J. C. Burant, S. Dapprich, J. M. Millam, A. D. Daniels, K. N. Kudin, M. C. Strain, O. Farkas, J. Tomasi, V. Barone, M. Cossi, R. Cammi, B. Mennucci, C. Pomelli, C. Adamo, S. Clifford, J. Ochterski, G. A. Petersson, P. Y. Ayala, Q. Cui, K. Morokuma, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. Cioslowski, J. V. Ortiz, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. Gomperts, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, C. Gonzalez, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, J. L. Andres, C. Gonzalez, M. Head-Gordon, E. S. Replogle, and J. A. Pople, GAUSSIAN98 共Revision A.6兲 共Gaussian, Inc., Pittsburgh, PA, 1998兲. 40 C. J. Tsai and K. D. Jordan, Chem. Phys. Lett. 213, 181 共1993兲. 41 C. J. Tsai and K. D. Jordan, J. Phys. Chem. 97, 5208 共1993兲. 42 T. H. Dunning, Jr., J. Chem. Phys. 90, 1007 共1989兲. 43 R. A. Kendall, T. H. Dunning, Jr., and R. J. Harrison, J. Chem. Phys. 96, 6796 共1992兲. 44 J. E. Del Bene and I. Shavitt, J. Mol. Struct.: THEOCHEM 307, 27 共1994兲. 45 J. E. Del Bene and I. Shavitt, in Molecular Interactions, edited by S. Scheiner 共Wiley, Chichester, 1997兲, p. 157. 46 S. S. Xantheas, J. Chem. Phys. 100, 7523 共1994兲. 47 N. Agmon, J. Mol. Liq. 73,74, 513 共1997兲. 48 H. S. Johnston, Gas Phase Reaction Rate Theory 共Ronald, New York, 1960兲. 49 L. Pauling, J. Am. Chem. Soc. 69, 542 共1947兲. 50 L. Pauling, The Nature of the Chemical Bond, 3rd ed. 共Cornell University Press, Ithaca, NY, 1960兲. 51 R. S. Mulliken, J. Chem. Phys. 23, 1833 共1955兲. 52 Reviews of the deficiencies of Mulliken charges and improved schemes for assigning atomic charges within molecules can be found in the following articles and references cited therein: S. M. Bachrach and A. Streitwieser, J. Comput. Chem. 10, 514 共1989兲; G. Tasi, I. Kiricsi, and H. Förster, ibid. 13, 371 共1992兲; P. S. Bagus, A. Clotet, D. Curulla, F. Illas, and J. M. Ricart, J. Mol. Catal. A: Chem. 119, 3 共1997兲; P. S. Bagus, F. Illas, C. Sousa, and G. Pacchioni, in Electronic Properties of Solids Using Cluster Methods, edited by T. A. Kaplan and S. D. Mahanti 共Plenum, New York, 1995兲. 53 M. D. Tissandier, S. J. Singer, and J. Coe, J. Phys. Chem. A 104, 752 共2000兲. 54 M. V. Vener and J. Sauer, Chem. Phys. Lett. 312, 591 共1999兲. 55 H.-C. Chang, J.-C. Jiang, I. Hahndorf, S. H. Lin, Y. T. Lee, and H.-C. Chang, J. Am. Chem. Soc. 121, 4443 共1999兲. Downloaded 10 Jun 2002 to 128.148.49.77. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp