Toward a Universal Water Model: First Principles Simulations from the

Toward a Universal Water Model: First Principles Simulations from the Dimer to the Liquid Phase. Volodymyr ... Citation data is made available by part...
0 downloads 9 Views 1MB Size
Letter pubs.acs.org/JPCL

Toward a Universal Water Model: First Principles Simulations from the Dimer to the Liquid Phase Volodymyr Babin,‡ Gregory R. Medders,‡ and Francesco Paesani* Department of Chemistry and Biochemistry, University of California, San Diego, La Jolla, California 92093, United States S Supporting Information *

ABSTRACT: A full-dimensional model of water, HBB2-pol, derived entirely from “firstprinciples”, is introduced and employed in computer simulations ranging from the dimer to the liquid. HBB2-pol provides excellent agreement with the measured second and third virial coefficients and, by construction, reproduces the dimer vibration−rotation−tunneling spectrum. The model also predicts the relative energy differences between isomers of small water clusters within the accuracy of highly correlated electronic structure methods. Importantly, when combined with simulation methods that explicitly include zero-point energy and quantum thermal motion, HBB2-pol accurately describes both structural and dynamical properties of the liquid phase.

SECTION: Molecular Structure, Quantum Chemistry, and General Theory

G

freedom, quantum dynamics methods based on basis set expansions of the nuclear wave function are well suited to the study of small complexes,14 while simulation approaches based on the path-integral formalism allow the fully quantummechanical modeling of water in condensed phases.15 In the end, these two components must be combined in an efficient computational scheme that enables the calculation of statistically converged quantities. Unfortunately, the computational cost associated with CCSD(T) makes these calculations prohibitively expensive even for small water clusters. To overcome this computational barrier while still providing an ab initio representation of the molecular interactions, less accurate but more efficient DFT approaches have been widely applied to water simulations. However, the choice of the most appropriate DFT model for water remains the subject of ongoing research.16,17 Alternatively, the PES for a system of N water molecules can be expressed through the many-body expansion of interaction energies, consisting of 1-body, 2-body, ..., N-body terms.18 It has been shown that this expansion converges rapidly for water, such that it is sufficient to take into account only the first few terms.19−23 Since the low-order interactions can be accurately calculated using CCSD(T), the many-body expansion effectively enables the representation of the multidimensional PES at the CCSD(T) level of theory.23 This strategy has been rigorously followed in the development of the CC-pol24 and WHBB25 models. In CC-pol, which describes the water

iven the central role that water plays in nature, it is not surprising that many studies have attempted to develop a microscopic picture of its unique properties. In particular, since the advent of computer simulations, myriad molecular models based on both force fields (including different degrees of empiricism) and ab initio approaches have been proposed. Force field-based models range from coarse-grained representations with no atomistic details (e.g, see ref 1), to classical parametrizations in terms of point charges and rigid bonds (see ref 2 for a recent review), to yet more sophisticated models that account for molecular flexibility (e.g., ref 3), electronic polarization (e.g., ref 4), and charge transfer.5 On the other hand, due to the associated computational cost, purely ab initio models so far have been limited to the use of density functional theory (DFT) (e.g., ref 6). However, despite much recent progress, none of the existing models is capable of correctly describing the properties of water from isolated molecules and small clusters in the gas phase up to the liquid and solid phases. Due to the broad spectrum of (often conflicting) predictions derived from these models, the microscopic behavior of water under different conditions and in different environments remains the subject of continuing debate.7−12 Independently of the number of molecules, the “firstprinciples” modeling of water within the Born−Oppenheimer approximation builds upon two components: a faithful representation of the electronic potential energy surface (PES) and the proper treatment of the nuclear motion at the quantum-mechanical level. In principle, the multidimensional PES can be accurately approximated at the coupled cluster level of theory including single, double, and perturbative triple excitations, CCSD(T),13 which currently represents the gold standard in quantum chemistry. For the nuclear degrees of © XXXX American Chemical Society

Received: November 2, 2012 Accepted: December 4, 2012

3765

dx.doi.org/10.1021/jz3017733 | J. Phys. Chem. Lett. 2012, 3, 3765−3769

The Journal of Physical Chemistry Letters

Letter

Figure 1. 2-body (a) and 3-body (b) interactions along with the second (c) and third (d) virial coefficients. Experimental data36,37 for second and third virial coefficients are shown as black squares and the HBB2-pol values as light blue circles.

exchange-repulsion and charge transfer, which have been shown to make a significant contribution to the 3-body interaction.31,32 The 3-body interaction of HBB2-pol is unique in that it is currently the first that has been fitted to CCSD(T) data. Importantly, the inclusion of induction interactions in the 3body term enables the use of lower-degree polynomials than previously reported,25 resulting in a sizable decrease in the computational cost associated with the many-body expansion. Specifically, the explicit 3-body component takes the form of a third-degree polynomial in the exponential of the interatomic distances, which is invariant with respect to the permutations of equivalent atoms. This short-range correction was found to be indispensable to overcome the limitations of current polarizable force fields (e.g., ref 33), which attempt to describe all nonpairwise additive interactions through point polarizable dipoles. These findings suggest that the neglect of accurate short-range 3-body interactions invariably leads to an incorrect description of the liquid structure (see Figure S3 in the Supporting Information). The ability of HBB2-pol to accurately reproduce the molecular interactions of water is established by comparison to high-level electronic structure calculations and experimental measurements reported in the literature. Figure 1a,b show the correlation plots of the HBB2-pol and CCSD(T) 2-body and 3body interaction energies, respectively. For this analysis, approximately 1400 dimer and 500 trimer configurations were extracted from classical MD simulations of the hexamer, ice Ih, and liquid water. CCSD(T) energies were calculated with the aug-cc-pVTZ basis set34 and corrected for the basis set superposition error through the counterpoise method.35 Nearly perfect agreement is found over the entire range of energies, indicating that HBB2-pol accurately reproduces the energetics of both favorably interacting and distorted dimers and trimers. An additional measure of the overall accuracy of the 2-body and 3-body interactions is provided by the second and third virial coefficients, which directly probe dimer and trimer

molecules as rigid, the 2-body term was derived from CCSD(T) data, while Hartree−Fock calculations were used to fit the 3-body term. WHBB, which allows for molecular flexibility, was parametrized using CCSD(T) and MP2 reference data for the 2-body and 3-body interaction terms, respectively. Both models include higher-body interactions through point polarizable dipoles. CC-pol accurately reproduces the vibration−rotation tunneling (VRT) spectrum of the water dimer and has been used in classical molecular dynamics (MD) simulations of liquid water.24 However, these simulations explicitly neglect nuclear quantum effects such as zero-point energy and quantum thermal motion, which have been shown to be important.15 A new version of CC-pol with flexible monomers has recently been applied to study the water dimer.26 WHBB also reproduces the VRT spectrum of the dimer with high accuracy.27 Due to the computational cost associated with the high dimensionality of the polynomials used to represent the 3-body interactions, quantum simulations with the WHBB model so far have been limited to small clusters.28 Here we report on structural and dynamical properties of water calculated with the newly developed full-dimensional model, HBB2-pol, derived entirely from first principles. Similarly to CC-pol and WHBB, HBB2-pol is built upon the many-body expansion of the molecular interactions. The 1body term associated with intramolecular distortion is described by the spectroscopically accurate PES developed by Partridge and Schwenke.29 The 2-body interaction at shortrange is represented by the HBB2 potential,30 which smoothly transitions between 5.5 Å and 7.5 Å into the sum of electrostatic and dispersion interactions, reproducing the correct asymptotic behavior. The induction contributions to nonpairwise additive interactions are taken into account through Thole-type point polarizable dipoles on all atomic sites using the TTM4-F scheme.33 In addition, an explicit 3body component is introduced to account for short-range 3766

dx.doi.org/10.1021/jz3017733 | J. Phys. Chem. Lett. 2012, 3, 3765−3769

The Journal of Physical Chemistry Letters

Letter

interactions, respectively. The HBB2-pol results are shown in Figure 1c,d along with the corresponding experimental data.36,37 In the calculation of the second virial coefficient, nuclear quantum effects were explicitly included through the path-integral formalism (see Supporting Information for details). Since experimental data for the third virial coefficient are only available at relatively high temperatures where quantum effects are less important, these calculations were carried out at the classical level with each water monomer held fixed at the ground-state vibrationally averaged geometry. For both virial coefficients, the calculated values are in close agreement with the corresponding experimental data over the entire ranges of temperature, providing further support of the accuracy of the 2-body and 3-body interactions of HBB2-pol. The energetics of small water clusters with more than three molecules allows for a quantitative assessment of the ability of HBB2-pol to correctly describe N-body interactions with N > 3 and represents a stringent test for the overall many-body expansion. For this purpose, the relative energies of the lowlying isomers of the water tetramer, pentamer, and hexamer are compared in Figure 2 with the corresponding ab initio values reported in the literature.38,39 Water clusters in this size range are among the largest ones for which highly correlated electronic structure calculations are still feasible. In all cases, HBB2-pol predicts the correct energy ordering of the isomers, with the energy differences being within the intrinsic accuracy of the ab initio reference data.23 The comparisons shown in Figures 1 and 2 demonstrate that the degree of accuracy of HBB2-pol is comparable to that of state-of-the-art quantum chemistry methods. While these highly correlated calculations are restricted to small clusters, quantum simulations with HBB2-pol enable a microscopically detailed characterization of water in condensed phases. To this end, HBB2-pol was used in fully quantum MD simulations of liquid water at ambient conditions (T = 298.15 K and ρ = 0.997 g cm−3). Specifically, path-integral molecular dynamics (PIMD) and centroid molecular dynamics (CMD) simulations were performed to determine structural and dynamical properties of the liquid phase, respectively.15 Both PIMD and CMD are based on Feynman’s formulation of statistical mechanics in terms of path-integrals and have been shown to accurately describe nuclear quantum effects in condensed phases of water.15 All simulations were carried out with a system consisting of 256 molecules in a periodic cubic box. The oxygen−oxygen (O−O), oxygen−hydrogen (O−H), and hydrogen−hydrogen (H−H) radial distribution functions (RDFs) calculated from an 80 ps PIMD simulation are compared in Figure 3 with two sets of experimental data.40,41 The PIMD simulations correctly predict a lower first peak in the O−O RDF, as determined by the most recent experiments. This feature has been proven difficult to reproduce by current force field-based and ab initio models. Some differences between the PIMD results and the experimental data exist for the second peak of the O−H RDF. However, both position and shape of this peak describing the spatial correlation between O and H atoms directly involved in hydrogen bonds are difficult to be experimentally determined as demonstrated by the appreciable differences between the two sets of experimental data (see also Figure S2 in the Supporting Information for a comparison between the calculated and measured differential neutron scattering cross sections). The diffusion coefficient (D) and orientational relaxation time (τ2) calculated at the quantum-mechanical level by averaging over six CMD

Figure 2. Relative energies of the low-lying isomers of the water tetramer (a), pentamer (b), and hexamer (c) computed with the HBB2-pol model (light blue circles) and state-of-the-art quantum chemistry methods (black squares).38,39 A perfect agreement with the results of refs 38 and 39 would not be expected because HBB2-pol was not fitted to ab initio data extrapolated to the complete basis set limit.

trajectories of 10 ps each are listed in Table 1. For both quantities, the CMD results are in quantitative agreement with the corresponding experimental values, providing further evidence of the accuracy of the HBB2-pol model. The average structure of the first hydration shell was determined by labeling the water molecules according to the number of donating hydrogen bonds as nondonor, single-donor, and double-donor. Using the geometric criterion proposed in ref 7, the PIMD simulations predict that ∼57% of the molecules are double donors and ∼38% are single donors. Each molecule is involved, on average, in ∼3 hydrogen bonds. In summary, the full-dimensional HBB2-pol model, based entirely on first principles, has been introduced and employed in calculations of the water properties from the dimer to the liquid phase. Being derived from state-of-the-art quantum chemistry calculations, HBB2-pol represents a major step 3767

dx.doi.org/10.1021/jz3017733 | J. Phys. Chem. Lett. 2012, 3, 3765−3769

The Journal of Physical Chemistry Letters



Letter

ASSOCIATED CONTENT

S Supporting Information *

Details of the HBB2-pol model are provided in the Supporting Information. Figures demonstrating the effects of nuclear quantization and molecular flexibility on the second virial coefficient, the impact of accurate three-body interactions on the O−O RDF, and the accuracy of HBB2-pol with respect to the experimental differential neutron scattering cross-section per atom are also presented in the Supporting Information. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Author Contributions ‡

These authors contributed equally. The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This research was supported by the National Science Foundation through Grant CHE-1038028 (CCI Center for Aerosol Impacts on Climate and the Environment). This work used the Extreme Science and Engineering Discovery Environment (XSEDE), which is supported by National Science Foundation Grant Number OCI-1053575 (allocation TGCHE110009).

■ Figure 3. Oxygen−oxygen, O−O (a), oxygen−hydrogen, O−H (b), and hydrogen−hydrogen, H−H (c) radial distribution functions. Experimental data from refs 40 and 41 are shown as dashed and solid black lines, respectively. The quantum RDFs calculated with the HBB2-pol model are shown as light blue lines.

Table 1. Comparison between the Experimental and Calculated Diffusion Coefficient (D) and Orientational Relaxation Time (τ2) of Liquid Water at Ambient Conditions experiment

simulation

0.23a 2.5b

0.23 ± 0.5 2.5 ± 0.2

D (A2 ps−1) τ2 (ps) a

REFERENCES

(1) Molinero, V.; Moore, E. B. Water Modeled as an Intermediate Element between Carbon and Silicon. J. Phys. Chem. B 2009, 113, 4008−4016. (2) Vega, C.; Abascal, J. L. F. Simulating Water with Rigid Nonpolarizable Models: A General Perspective. Phys. Chem. Chem. Phys. 2011, 13, 19663−19688. (3) Dang, L. X.; Pettitt, B. M. Simple Model Intramolecular Potentials for Water. J. Phys. Chem. 1987, 91, 3349. (4) Fanourgakis, G. S.; Xantheas, S. S. Development of Transferable Interaction Potentials for Water. V. Extension of the Flexible, Polarizable, Thole-Type Model Potential (TTM3-F, v. 3.0) to Describe the Vibrational Spectra of Water Clusters and Liquid Water. J. Chem. Phys. 2008, 128, 074506. (5) Lee, A. J.; Rick, S. W. The Effects of Charge Transfer on the Properties of Liquid Water. J. Chem. Phys. 2011, 134, 184507. (6) Laasonen, K.; Sprik, M.; Parrinello, M.; Car, R. Ab-Initio Liquid Water. J. Chem. Phys. 1993, 99, 9080−9089. (7) Wernet, P.; Nordlund, D.; Bergmann, U.; Cavalleri, M.; Odelius, M.; Ogasawara, H.; Näslund, L. Å.; Hirsh, T. K.; Ojamäe, L.; Glatzel, P.; et al. The Structure of the First Coordination Shell in Liquid Water. Science 2004, 304, 995−999. (8) Clark, G. N. I.; Cappa, C. D.; Smith, J. D.; Saykally, R. J.; HeadGordon, T. The Structure of Ambient Water. Mol. Phys. 2010, 108, 1415−1433. (9) Pieniazek, P. A.; Tainter, C. J.; Skinner, J. L. Surface of Liquid Water: Three-Body Interactions and Vibrational Sum-Frequency Spectroscopy. J. Am. Chem. Soc. 2011, 133, 10360−10363. (10) Nihonyanagi, S.; Ishiyama, T.; Lee, T.; Yamaguchi, S.; Bonn, M.; Morita, A.; Tahara, T. Unified Molecular View of the Air/Water Interface Based on Experimental and Theoretical χ(2) Spectra of an Isotopically Diluted Water Surface. J. Am. Chem. Soc. 2011, 133, 16875−16880.

Reference 43. bReference 44.

toward the long-sought “universal model” capable of describing the behavior of water under different conditions and in different environments.42 In the future, simulation studies with the HBB2-pol model will focus on the characterization of the pressure−temperature properties of water in an attempt to resolve some of the current controversies regarding structural, thermodynamic, and dynamical properties of bulk, interfacial, and supercooled water.7−12 3768

dx.doi.org/10.1021/jz3017733 | J. Phys. Chem. Lett. 2012, 3, 3765−3769

The Journal of Physical Chemistry Letters

Letter

(11) Kumar, P.; Franzese, G.; Stanley, H. E. Dynamics and Thermodynamics of Water. J. Phys.: Condens. Matter 2008, 20, 244114. (12) Limmer, D. T.; Chandler, D. The Putative Liquid−Liquid Transition Is a Liquid−Solid Transition in Atomistic Models of Water. J. Chem. Phys. 2011, 135, 134503. (13) Raghavachari, K.; Trucks, G. W.; Pople, J. A.; Headgordon, M. A 5th-Order Perturbation Comparison of Electron Correlation Theories. Chem. Phys. Lett. 1989, 157, 479−483. (14) Meyer, H. D.; Manthe, U.; Cederbaum, L. S. The Multiconfigurational Time-Dependent Hartree Approach. Chem. Phys. Lett. 1990, 165, 73−78. (15) Paesani, F.; Voth, G. A. The Properties of Water: Insights from Quantum Simulations. J. Phys. Chem. B 2009, 113, 5702−5719. (16) Zhang, C.; Wu, J.; Galli, G.; Gygi, F. Structural and Vibrational Properties of Liquid Water from van der Waals Density Functionals. J. Chem. Theory Comput. 2011, 7, 3054−3061. (17) Lin, I. C.; Seitsonen, A. P.; Tavernelli, I.; Rothlisberger, U. Structure and Dynamics of Liquid Water from Ab Initio Molecular Dynamics - Comparison of BLYP, PBE, and revPBE Density Functionals with and without van der Waals Corrections. J. Chem. Theory Comput. 2012, 8, 3902−3910. (18) Hankins, D.; Moskowitz, J. W.; Stillinger, F. H. Water Molecule Interactions. J. Chem. Phys. 1970, 53, 4544−4554. (19) Xantheas, S. S. Ab Initio Studies of Cyclic Water Clusters (H2O)n, n = 1−6. II. Analysis of Many-Body Interactions. J. Chem. Phys. 1994, 100, 7523−7534. (20) Cui, J.; Liu, H. B.; Jordan, K. D. Theoretical Characterization of the (H2O)21 Cluster: Application of an n-Body Decomposition Procedure. J. Phys. Chem. B 2006, 110, 18872−18878. (21) Xantheas, S. S. Cooperativity and Hydrogen Bonding Network in Water Clusters. Chem. Phys. 2000, 258, 225−231. (22) Hermann, A.; Krawczyk, R. P.; Lein, M.; Schwerdtfeger, P.; Hamilton, I. P.; Stewart, J. J. P. Convergence of the Many-Body Expansion of Interaction Potentials: From van der Waals to Covalent and Metallic Systems. Phys. Rev. A 2007, 76. (23) Góra, U.; Podeszwa, R.; Cencek, W.; Szalewicz, K. Interaction Energies of Large Clusters from Many-Body Expansion. J. Chem. Phys. 2011, 135, 224102. (24) Bukowski, R.; Szalewicz, K.; Groenenboom, G. C.; van der Avoird, A. Predictions of the Properties of Water from First Principles. Science 2007, 315, 1249−1252. (25) Wang, Y.; Huang, X.; Shepler, B. C.; Braams, B. J.; Bowman, J. M. Flexible, Ab Initio Potential, and Dipole Moment Surfaces for Water. I. Tests and Applications for Clusters up to the 22-mer. J. Chem. Phys. 2011, 134, 094509. (26) Leforestier, C.; Szalewicz, K.; van der Avoird, A. Spectra of Water Dimer from a New Ab Initio Potential with Flexible Monomers. J. Chem. Phys. 2012, 137, 014305. (27) Leforestier, C. Infrared Shifts of the Water Dimer from the Fully Flexible Ab Initio HBB2 Potential. Philos. Trans. R. Soc., A 2012, 370, 2675−2690. (28) Wang, Y. M.; Babin, V.; Bowman, J. M.; Paesani, F. The Water Hexamer: Cage, Prism, or Both. Full Dimensional Quantum Simulations Say Both. J. Am. Chem. Soc. 2012, 134, 11116−11119. (29) Partridge, H.; Schwenke, D. W. The Determination of an Accurate Isotope Dependent Potential Energy Surface for Water from Extensive Ab Initio Calculations and Experimental Data. J. Chem. Phys. 1997, 106, 4618−4639. (30) Shank, A.; Wang, Y.; Kaledin, A.; Braams, B. J.; Bowman, J. M. Accurate Ab Initio and “Hybrid” Potential Energy Surfaces, Intramolecular Vibrational Energies, and Classical IR Spectrum of the Water Dimer. J. Chem. Phys. 2009, 130, 144314−144311. (31) Chen, W.; Gordon, M. S. Energy Decomposition Analyses for Many-Body Interaction and Applications to Water Complexes. J. Phys. Chem. 1996, 100, 14316−14328. (32) Mas, E. M.; Bukowski, R.; Szalewicz, K. Ab Initio Three-Body Interactions for Water. I. Potential and Structure of Water Trimer. J. Chem. Phys. 2003, 118, 4386−4403.

(33) Burnham, C. J.; Anick, D. J.; Mankoo, P. K.; Reiter, G. F. The Vibrational Proton Potential in Bulk Liquid Water and Ice. J. Chem. Phys. 2008, 128, 154519. (34) Dunning, T. H. Gaussian-Basis Sets for Use in Correlated Molecular Calculations 0.1. the Atoms Boron through Neon and Hydrogen. J. Chem. Phys. 1989, 90, 1007−1023. (35) Boys, S. F.; Bernardi, F. The Calculation of Small Molecular Interactions by the Differences of Separate Total Energies. Some Procedures with Reduced Errors. Mol. Phys. 1970, 19, 553. (36) Harvey, A. H.; Lemmon, E. W. Correlation for the Second Virial Coefficient of Water. J. Phys. Chem. Ref. Data 2004, 33, 369−376. (37) Kell, G. S.; McLaurin, G. E.; Whalley, E. PVT Properties of Water VII. Vapour Densities of Light and Heavy Water from 150 to 500 °C. Proc. R. Soc. London, Ser. A 1989, 425, 49−71. (38) Temelso, B.; Archer, K. A.; Shields, G. C. Benchmark Structures and Binding Energies of Small Water Clusters with Anharmonicity Corrections. J. Phys. Chem. A 2011, 115, 12034−12046. (39) Bates, D. M.; Tschumper, G. S. CCSD(T) Complete Basis Set Limit Relative Energies for Low-Lying Water Hexamer Structures. J. Chem. Phys. A 2009, 113, 3555−3559. (40) Soper, A. K. The Radial Distribution Functions of Water and Ice from 220 to 673 K and at Pressures up to 400 MPa. Chem. Phys. 2000, 258, 121−137. (41) Soper, A. K.; Benmore, C. J. Quantum Differences between Heavy and Light Water. Phys. Rev. Lett. 2008, 101, 065502. (42) Keutsch, F. N.; Saykally, R. J. Water Clusters: Untangling the Mysteries of the Liquid, One Molecule at a Time. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 10533−10540. (43) Krynicki, K.; Green, C. D.; Sawyer, D. W. Pressure and Temperature-Dependence of Self-Diffusion in Water. Faraday Discuss. 1978, 66, 199−208. (44) Rezus, Y. L. A.; Bakker, H. J. On the Orientational Relaxation of HDO in Liquid Water. J. Chem. Phys. 2005, 123, 114502.

3769

dx.doi.org/10.1021/jz3017733 | J. Phys. Chem. Lett. 2012, 3, 3765−3769