J. Phys. Chem. 1995, 99, 16460-16467
A Polarizable Intermolecular Potential Function for Simulation of Liquid Alcohols Jiali Gao,* Dariush Habibollazadeh, and Lei Shao Department of Chemistry, State University of New York at Buffalo, Buffalo, New York 14260 Received: June 6, 1995; In Final Form: August 15, 1995@
A polarizable intermolecular potential function (PIPF) for simulation of liquid alcohols has been developed. This represents the first systematic study of a class of organic liquids using such potential functions. The PIPF includes a pairwise additive component, consisting of the familiar Lennard-Jones and Coulombic form, and a nonadditive polarization term. The empirical parameters were optimized through a series of statistical Monte Carlo simulations of liquid methanol, ethanol, 1-propanol, 2-propanol, and 2-methyl-2-propano1, which cover the functionalities of all simple alkanols. The computed heats of vaporization and densities for these liquids using the final parameters are within 1% and 3% of experimental values, respectively. The polarization effects were found to be significant in all liquids, comprising 10-20% of the total energy of the liquids or over 20% of the electrostatic component. A unique feature in the present parameter optimization is to make use of computed polarization energies and induced dipole moments from separate Monte Carlo simulations employing a combined quantum mechanical and molecular mechanical (QM/MM) approach. In the latter calculations, one alcohol monomer is treated quantum mechanically by the AM1 theory, which is embedded in the liquid of the same alcohol represented by the empirical OPLS potential. Our PIPF results are in accord with the combined QM/MM simulations. In addition, structural features including hydrogen bonding interactions and radial distribution functions were examined and were found to be in good agreement with previous computational results.
Introduction
A major difficulty associated with these calculations is a significant increase in computationaltime and memory required to evaluate the induced dipole moments and potential energies. Typically, these requirements scale as #,where N is the number of polarizable sites in the system. Another difficulty in the development of PIPF is due to a lack of quantitative knowledge of the polarization effects. In particular, it was not clear what extent the polarization energy contributes to the total intermolecular interaction energy in solution. As a result, only a limited number of computer simulations of liquids and molecules of biological interest have been reported in the literature. In the past few years, the situation has changed dramatically thanks to advances in computer speed and developments of combined quantum mechanical and molecular mechanical (QM/MM) methods.’~~The solvent polarization effects can be determined a priori from such combined QMMM simulations and can be used to calibrate the empirical parameters in the PIPF potential.8 It is time to go beyond the simple point charge, pairwise potentials that have dominated simulations of liquids and
Computer simulations have become a powerful tool for obtaining detailed insights into the structure and dynamics of liquids and biopolymers.’ Undoubtedly, a critical component in these studies is intermolecular potential functions describing interactions between molecules in the condensed phase system. It is the accuracy of these potential functions that ultimately determine the success of fluid simulations.’ The past two decades saw a tremendous effort in the development of reliable force fields for organic and biomolecular systems. As a result, a number of well-tested force fields are However, a common feature in these force fields is the assumption that the intermolecular potential energy of a condensed phase system is pairwise additive, which can be determined by interactions between pairs of monomers in the liquid. Consequently, the many-body polarization effects are not explicitly treated, but instead, they are incorporated implicitly into the parameters of the pair potentials in an average sense in the parametrization process. This gives rise to the notion of effective pair protein^.^ potential^.'-^ Although such effective potentials can provide In this paper, we report the results of a simulation study of valuable insights and understanding of the dynamics of proteins liquid alcohols using a PIPF potential. Liquid alcohol represents and liquids and will continue to be widely used because of the an important class of compounds for its hydrogen bonding and computational efficiency, neglect of specific polarization effects amphiphilic characteristics and use as organic solvents. The is a severe limitation. It prevents accurate investigations of processes involving changes in molecular environment, such study also provides parameters for amino acids side chains as molecular recognition, protein-DNA interactions, and protein bearing hydroxyl groups. Our goal in this project is to generate allosteric effects.’ The incorporation of explicit many-body a complete set of PIPF functions for simulation of proteins in polarization terms represents a great challenge and a major step aqueous solution. This investigation represents a major imtowards development of more accurate molecular mechanics provement in the treatment of intermolecular interactions using force fields. empirical force field for molecular modeling and simulations of organic and biological systems. Liquid alcohols, especially However, progress in the development of polarizable intermethanol and ethanol, have been the subjects of numerous molecular potential functions (PIPF) for condensed phase simulations has been s ~ o w . ~In- fact, ~ the only liquid that has studies using pairwise potential The only previous been thoroughly studied using a polarizable potential is ~ a t e r . ~ . ~calculation of liquid methanol employing a polarizable potential is that by Caldwell and Kollman,@which was reported while this paper was in preparation. In the following, we fist describe Abstract published in Advance ACS Abstracts, October 1, 1995. @
0022-36541992099-16460$09.0010
0 1995 American Chemical Society
A PIPF for Simulation of Alcohols
J. Phys. Chem., Vol. 99, No. 44, 1995 16461
TABLE 1: Standard Geometrical Parameters for Alcohold bond angles, degrees bond lengths, A 0-H C-H
c-0 c-c
0.945 1.090 1.430 1.530
C-0-H 0-C-H
c-c-0 c-c-c C-C-H H-C-H
108.5 109.5 108.0 112.0 109.5 109.5
Reference 11. details of the computational methods and parametrization. Then, results and discussion for liquid alcohols, including methanol, ethanol, 1-propanol, 2-propanol, and 2-methyl-2-propano1,will be presented. The paper is concluded with future perspectives.
molecule containing atom i. This makes the calculation of the molecular polarizability inaccurate using Applequist or Tholes methods since induced dipoles do not interact with each other within the same molecule.'8a,b Interestingly, both methods can yield anisotropy in the computed molecular polarization tensor, eventhough isotropic atomic polarizabilities are used.'8as18b Nevertheless, the present approximation appears to be reasonable considering the small size of the alcohol molecules investigated in this study. The intramolecular fields will be severely damped if they are included. A compromised treatment is to include intramolecular electric fields in eq 5 from atoms separated by three or more chemical bonds. This approach has been used in some previous computati~ns.~g~~ In eq 5, ai is the polarizability of atom i, which is assumed to be isotropic for simplicity in the present model. The dipole tensor is given as follows
Computational Details
A. Intermolecular Potential. The molecules are represented by interaction sites located on each nucleus. All atoms are represented explicitly in the present polarizable intermolecular potential function (PIPF). Standard bond lengths and bond angles derived from microwave structures, which have been summarized and used in the OPLS (optimized potential for liquid simulations) potentials for liquid are adopted in this study (Table 1). In the present study, these geometrical parameters are kept fixed throughout the liquid simulation, although internal rotations of all dihedral angles are varied in the Monte Carlo sampling. The total energy of the pure liquid system is given below, which consists of a pairwise term and a nonadditive polarization term: 'tot
+ 'pol
= 'pair
(1)
In eq 1, the pairwise potential is enumerated over pairs of monomers in the liquid (eq 2). N
a
