Achieving fast convergence of ab initio free energy perturbation calculations with the adaptive force-matching method

This paper studies the possibility of improving the convergence of ab initio free energy perturbation (FEP) calculations by developing customized force fields with the adaptive force-matching (AFM) method. The ab initio FEP method relies on a molecular mechanics (MM) potential to sample configuration space. If the Boltzmann weight of the MM sampling is close to that of the ab initio method, the efficiency of ab initio FEP will be optimal. The difference in the Boltzmann weights can be quantified by the relative energy difference distribution (REDD). The force field developed through AFM significantly improves the REDD when compared with standard MM models, thus improving the convergence of the ab initio FEP calculation. The static dielectric constant ε_s of ice-Ih was studied with PW-91 through ab initio FEP. With a customized force field developed through AFM, we were able to converge ε_s to 80 ± 4 with 3,600 configurations. A similar ab initio FEP calculation with the TIP4P model would require 220 times more configurations to achieve the same accuracy. Our study indicates that the PW-91 functional underestimates ice-Ih ε_s by about 20%.     

Calibration of force-field dependency in free energy landscapes of peptide conformations by quantum chemical calculations

The free energy landscapes of peptide conformations were calibrated by ab initio quantum chemical calculations, after the enhanced conformational diversity search using the multicanonical molecular dynamics simulations. Three different potentials of mean force for an isolated dipeptide were individually obtained by the multicanonical molecular dynamics simulations using the conventional force fields, AMBER parm94, AMBER parm96, and CHARMm22. Each potential of mean force was then calibrated based upon the umbrella sampling algorithm from the adiabatic energy map that was calculated separately by the ab initio molecular orbital method, and all of the calibrated potentials of mean force coincided well. The calibration method was also applied to the simulations of a peptide dimer in explicit water models, and it was shown that the calibrated free energy landscapes did not depend on the force field used in the classical simulations, as far as the conformational space was sampled well. The current calibration method fuses the classical free energy calculation with the quantum chemical calculation, and it should generally make simulations for biomolecular systems much more reliable when combining with enhanced conformational sampling.     

Static dielectric constants and molecular dipole distributions of liquid water and ice-Ih investigated by the PAW-PBE exchange-correlation functional

The static dielectric constants, ε(s), of ice-Ih and liquid water were calculated using density functional theory with the Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional and the projector-augmented-wave (PAW) approach. Proton disordered ice configurations and uncorrelated liquid configurations were sampled with the electrostatic switching method using force fields specially designed to facilitate the ab initio free energy perturbation calculations. Our results indicate that PAW-PBE underestimates the ε(s) of both ice-Ih and liquid water but predicts the ratio of ice and water ε(s) in good agreement with experimental measurements. PAW-PBE gives average water dipole moments of 2.50 D in ice-Ih and 2.48 D in the liquid. Our results show that the fixed-charge water models developed by adaptive force matching can reproduce the PAW-PBE dipole moments with an error of approximately 5%. The ice and liquid models created in this work have polarizabilities of 1.32 ?(3) and 1.30 ?(3), respectively, along the HOH bisector direction.     

Tests of Monte Carlo perturbation theory for the free energy of liquid copper

Monte Carlo perturbation theory, in which terms in the thermodynamic perturbation series are evaluated by Monte Carlo averaging, has potentially large advantages in efficiency for calculating free energies of liquids from ab initio potential surfaces. In order to test the accuracy of perturbation theory for liquid metals, a series of calculations has been done on liquid copper, modeled by an embedded atom potential. A simple 1/r(12) pair potential is used as the reference system. The free energy is calculated to third order in perturbation theory, and the results are compared to an exact formula. It is found that for optimal reference potential parameters, second order perturbation theory is essentially exact. Second and third order theories give accurate results for significantly nonoptimal reference parameters. The relation between perturbation theory and reweighting is discussed, and an approximate formula is derived that shows an exponential dependence of the efficiency of reweighting on the second order free energy correction. Finally, techniques for application to ab initio potentials are discussed. It is shown that with samples of 100 configurations, it is possible to obtain accuracy and precision at the level of approximately 1 meV/atom.     

The electronic structure of free water clusters probed by Auger electron spectroscopy

(H2O)(N) clusters generated in a supersonic expansion source with N approximately 1000 were core ionized by synchrotron radiation, giving rise to core-level photoelectron and Auger electron spectra (AES), free from charging effects. The AES is interpreted as being intermediate between the molecular and solid water spectra showing broadened bands as well as a significant shoulder at high kinetic energy. Qualitative considerations as well as ab initio calculations explain this shoulder to be due to delocalized final states in which the two valence holes are mostly located at different water molecules. The ab initio calculations show that valence hole configurations with both valence holes at the core-ionized water molecule are admixed to these final states and give rise to their intensity in the AES. Density-functional investigations of model systems for the doubly ionized final states--the water dimer and a 20-molecule water cluster--were performed to analyze the localization of the two valence holes in the electronic ground states. Whereas these holes are preferentially located at the same water molecule in the dimer, they are delocalized in the cluster showing a preference of the holes for surface molecules. The calculated double-ionization potential of the cluster (22.1 eV) is in reasonable agreement with the low-energy limit of the delocalized hole shoulder in the AES.     

Comparison of charge models for fixed-charge force fields: small-molecule hydration free energies in explicit solvent

In molecular simulations with fixed-charge force fields, the choice of partial atomic charges influences numerous computed physical properties, including binding free energies. Many molecular mechanics force fields specify how nonbonded parameters should be determined, but various choices are often available for how these charges are to be determined for arbitrary small molecules. Here, we compute hydration free energies for a set of 44 small, neutral molecules in two different explicit water models (TIP3P and TIP4P-Ew) to examine the influence of charge model on agreement with experiment. Using the AMBER GAFF force field for nonbonded parameters, we test several different methods for obtaining partial atomic charges, including two fast methods exploiting semiempirical quantum calculations and methods deriving charges from the electrostatic potentials computed with several different levels of ab initio quantum calculations with and without a continuum reaction field treatment of solvent. We find that the best charge sets give a root-mean-square error from experiment of roughly 1 kcal/mol. Surprisingly, agreement with experimental hydration free energies does not increase substantially with increasing level of quantum theory, even when the quantum calculations are performed with a reaction field treatment to better model the aqueous phase. We also find that the semiempirical AM1-BCC method for computing charges works almost as well as any of the more computationally expensive ab initio methods and that the root-mean-square error reported here is similar to that for implicit solvent models reported in the literature. Further, we find that the discrepancy with experimental hydration free energies grows substantially with the polarity of the compound, as does its variation across theory levels.     

An efficient method for computing excess free energy of liquid

We present a new theoretical method for efficient calculation of free energy of liquid. This interaction entropy method allows one to compute entropy and free energy of liquid from standard single step MD (molecular dynamics) simulation directly in liquid state without the need to perform MD simulations at many intermediate states as required in thermodynamic integration or free energy perturbation methods. In this new approach, one only needs to evaluate the interaction energy of a single (fixed) liquid molecule with the rest of liquid molecules as a function of time from a standard MD simulation of liquid and the fluctuation of distribution of this interaction energy is then used to calculate the interaction entropy of the liquid. Explicit theoretical derivation of this interaction entropy approach is provided and numerical calculations for the benchmark liquid water system were carried out using three different water models. Numerical analysis of the result was performed and comparison of the computational result with experimental data and other theoretical results were provided. Excellent agreement of calculated free energies with the experimental data using TIP4P model is obtained for liquid water.     

Magnitude and orientation dependence of intermolecular interaction of perfluoropropane dimer studied by high-level ab initio calculations: comparison with propane dimer

Intermolecular interaction energies of 12 orientations of C(3)F(8) dimers were calculated with electron correlation correction by the second-order M?ller-Plesset perturbation method. The antiparallel C(2h) dimer has the largest interaction energy (-1.45 kcal/mol). Electron correlation correction increases the attraction considerably. Electrostatic energy is not large. Dispersion is mainly responsible for the attraction. Orientation dependence of the interaction energy of the C(3)F(8) dimer is substantially smaller than that of the C(3)H(8) dimer. The calculated interaction energy of the C(3)F(8) dimer at the potential minimum is 78% of that of the C(3)H(8) dimer (-1.85 kcal/mol), whereas the interaction energies of the CF(4) and C(2)F(6) dimers are larger than those of the CH(4) and C(2)H(6) dimers. The intermolecular separation in the C(3)F(8) dimer at the potential minimum is substantially larger than that in the C(3)H(8) dimer. The larger intermolecular separation due to the steric repulsion between fluorine atoms is the cause of the smaller interaction energy of the C(3)F(8) dimer at the potential minimum. The calculated intermolecular interaction energy potentials of the C(3)F(8) dimers using an all atom model OPLS-AA (OPLS all atom model) force field and a united atom model force field were compared with the ab initio calculations. Although the two force fields well reproduces the experimental vapor and liquid properties of perfluoroalkenes, the comparison shows that the united atom model underestimates the potential depth and orientation dependence of the interaction energy. The potentials obtained by the OPLS-AA force field are close to those obtained by the ab initio calculations.     

A molecular dynamics study of small water clusters comparing two flexible models for water

We report results from a molecular dynamics study of small water clusters, (H₂O) _n=2,3,4,6,8, comparing the recent polarizable, dissociable (PD) model of Halley et al. with the central force Stillinger-Rahman (SR) model. Dynamics calculations in the microcanonical ensemble quantified short-time averaged temperatures and RMS bond length fluctuations, which, in combination with simulated thermal quenching, identified structures and structural and phase changes of the clusters. Both models generally pointed to similar global minimum energy configurations, but local minima configurations, relative energies, and RMS bond length fluctuations varied more significantly. The PD model, with its incorporation of many-body effects, more accurately reproduced the structural features of clusters predicted from ab initio calculations, although it over-estimated the binding energies. Solid-like, pre-melting, and liquid-like states were identified based upon RMS bond length fluctuations for the O-O, O-H and H-H pairs. The SR and PD models yielded very similar trends in melting temperature as a function of cluster size. This suggests that general trends in the melting phenomenon do not depend strongly upon details of the models. The melting temperatures for the dimer and tetramer closely resembled the bulk melting temperature, while those for other sizes were considerably lower.     

Calculation of the O-H stretching vibrational overtone spectrum of the water dimer

The O-H stretching vibrational overtone spectrum of the water dimer has been calculated with the dimer modeled as two individually vibrating monomer units. Vibrational term values and absorption intensities have been obtained variationally with a computed dipole moment surface and an internal coordinate Hamiltonian, which consists of exact kinetic energy operators within the Born-Oppenheimer approximation of the monomer units. Three-dimensional ab initio potential energy and dipole moment surfaces have been calculated using the internal coordinates of the monomer units using the coupled cluster method including single, double, and perturbative triple excitations [CCSD(T)] with the augmented correlation consistent valence triple zeta basis set (aug-cc-pVTZ). The augmented correlation consistent valence quadruple zeta basis set (aug-cc-pVQZ), counterpoise correction, basis set extrapolation to the complete basis set limit, relativistic corrections, and core and valence electron correlations effects have been included in one-dimensional potential energy surface cuts. The aim is both to investigate the level of ab initio and vibrational calculations necessary to produce accurate results when compared with experiment and to aid the detection of the water dimer under atmospheric conditions.     

Ab initio theoretical calculations of the electronic excitation energies of small water clusters

A direct ab initio molecular dynamics method has been applied to a water monomer and water clusters (H(2)O)(n) (n = 1-3) to elucidate the effects of zero-point energy (ZPE) vibration on the absorption spectra of water clusters. Static ab initio calculations without ZPE showed that the first electronic transitions of (H(2)O)(n), (1)B(1)←(1)A(1), are blue-shifted as a function of cluster size (n): 7.38 eV (n = 1), 7.58 eV (n = 2) and 8.01 eV (n = 3). The inclusion of the ZPE vibration strongly affects the excitation energies of a water dimer, and a long red-tail appears in the range of 6.42-6.90 eV due to the structural flexibility of a water dimer. The ultraviolet photodissociation of water clusters and water ice surfaces is relevant to these results.     

Critical examination of the supermolecule density functional theory calculations of intermolecular interactions

The results of calculations employing twelve different combinations of exchange and correlation functionals are compared with results of ab initio calculations for two different configurations of the water dimer and three different configurations of the thymine-adenine complex. None of the density functional theory (DFT) treatments could properly reproduce the results of coupled-cluster calculations for all configurations examined. The DFT approaches perform well when the interaction energy is dominated by the electrostatic component and the dispersion energy is less important. Two mechanisms that compensate for the missing dispersion component were identified. The first one is the decrease of the magnitude of the intermolecular exchange-repulsion and the second one is the increase of the magnitude of the attractive deformation energy. For some functionals both effects are observed together, but for some other ones only the second effect occurs. The three correlation functionals that were examined were found to make only very small contributions to the deformation energy. The examination of angular and distance dependence of the interactions shows that the currently available DFT approaches are not suitable for developing intermolecular potential energy surfaces. They could however be used to find global minima on potential energy surfaces governed by intermolecular electrostatic interactions. Additional single point ab initio calculations are recommended as the means of validating optimized structures.     

A polarizable force field for water using an artificial neural network

A force field for liquid water including polarization effects has been constructed using an artificial neural network (ANN). It is essential to include a many-body polarization effect explicitly into a potential energy function in order to treat liquid water which is dense and highly polar. The new potential energy function is a combination of empirical and nonempirical potentials. The TIP4P model was used for the empirical part of the potential. For the nonempirical part, an ANN with a back-propagation of error algorithm (BPNN) was introduced to reproduce the complicated many-body interaction energy surface from ab initio quantum mechanical calculations. BPNN, described in terms of a matrix, provides enough flexibility to describe the complex potential energy surface (PES). The structural and thermodynamic properties, calculated by isobaric-isothermal (constant-NPT) Monte Carlo simulations with the new polarizable force field for water, are compatible with experimental results. Thus, the simulation establishes the validity of using our estimated PES with a polarization effect for accurate predictions of liquid state properties. Applications of this approach are simple and systematic so that it can easily be applied to the development of other force fields besides the water-water system.     

An implemented potential of non-rigid water molecules for molecular dynamics simulations

The central force model for a water molecule is corrected by adding a three-body term. The present potential fits both an accurate ab initio potential energy surface and the fundamental vibrational frequencies of gas-phase water. The three-body terms allow us to reproduce the gas-phase IR spectrum by molecular dynamics simulations. Some problems connected with MD simulations of IR spectra are discussed.     

Ab initio potential and dipole moment surfaces for water. II. Local-monomer calculations of the infrared spectra of water clusters

We employ recent flexible ab initio potential energy and dipole surfaces [Y. Wang, X. Huang, B. C. Shepler, B. J. Braams, and J. M. Bowman, J. Chem. Phys. 134, 094509 (2011)] to the calculation of IR spectra of the intramolecular modes of water clusters. We use a quantum approach that begins with a partitioned normal-mode analysis of perturbed monomers, and then obtains solutions of the corresponding Schro?dinger equations for the fully coupled intramolecular modes of each perturbed monomer. For water clusters, these modes are the two stretches and the bend. This approach is tested against benchmark calculations for the water dimer and trimer and then applied to the water clusters (H(2)O)(n) for n = 6-10 and n = 20. Comparisons of the spectra are made with previous ab initio harmonic and empirical potential calculations and available experiments.     

Polarizable interaction potential for water from coupled cluster calculations. I. Analysis of dimer potential energy surface

A six-dimensional interaction potential for the water dimer has been fitted to ab initio interaction energies computed at 2510 dimer configurations. These energies were obtained by combining the supermolecular second-order energies extrapolated to the complete basis set limit from up to quadruple-zeta quality basis sets with the contribution from the coupled-cluster method including single, double, and noniterative triple excitations computed in a triple-zeta quality basis set. All basis sets were augmented by diffuse functions and supplemented by midbond functions. The energies have been fitted using an analytic form with the induction component represented by a polarizable term, making the potential directly transferable to clusters and the bulk phase. Geometries and energies of stationary points on the potential surface agree well with the results of high-level ab initio geometry optimizations.     

Effective force field for liquid hydrogen fluoride from ab initio molecular dynamics simulation using the force-matching method

A recently developed force-matching method for obtaining effective force fields for condensed matter systems from ab initio molecular dynamics (MD) simulations has been applied to fit a simple nonpolarizable two-site pairwise force field for liquid hydrogen fluoride. The ab initio MD in this case was a Car-Parrinello (CP) MD simulation of 64 HF molecules at nearly ambient conditions within the Becke-Lee-Yang-Parr approximation to the electronic density functional theory. The force-matching procedure included a fit of short-ranged nonbonded forces, bonded forces, and atomic partial charges. The performance of the force-match potential was examined for the gas-phase dimer and for the liquid phase at various temperatures. The model was able to reproduce correctly the bent structure and energetics of the gas-phase dimer, while the results for the structural properties, self-diffusion, vibrational spectra, density, and thermodynamic properties of liquid HF were compared to both experiment and the CP MD simulation. The force-matching model performs well in reproducing nearly all of the liquid properties as well as the aggregation behavior at different temperatures. The model is computationally cheap and compares favorably to many more computationally expensive potential energy functions for liquid HF.     

Simulating vapor-liquid nucleation of water: A combined histogram-reweighting and aggregation-volume-bias Monte Carlo investigation for fixed-charge and polarizable models

The method of histogram-reweighting was integrated with a recently developed approach using aggregation-volume-bias Monte Carlo and self-adaptive umbrella sampling to develop the AVUS-HR algorithm that allows for exceedingly efficient calculations of nucleation properties over a wide range of thermodynamic conditions. Simulations were carried out for water using both fixed-charge and polarizable force fields belonging to the TIP4P family (namely, TIP4P, TIP4P-FQ, TIP4P-pol2, and TIP4P-pol3) to investigate the nucleation of water over a temperature range from 200 to 300 K and the concentration of water clusters in the atmosphere at elevations up to 15 km. It was found that the nucleation free energy barriers and atmospheric concentrations are extremely sensitive to the force field, albeit all of the models investigated in this study support the following general conclusions: (i) ice nucleation is not present under the thermodynamic conditions and cluster-size range investigated here, i.e., the critical nuclei possess liquidlike structures, and (ii) the atmospheric concentrations of water clusters under homogeneous conditions are very low with the mole fraction of hexamers being about 10(-10), a number probably too low to influence the solar radiation balance. Compared to the experimental data, the TIP4P-pol3 model yields the most accurate nucleation results, consistent with its excellent performance for the second virial coefficient and the minimum cluster energies.