You are on page 1of 7

THE JOURNAL OF CHEMICAL PHYSICS 132, 104505 2010

Computer simulations of aqua metal ions for accurate reproduction of hydration free energies and structures
Xin Li,1,2 Yaoquan Tu,3 He Tian,2 and Hans gren1,a
1

Department of Theoretical Chemistry, School of Biotechnology, Royal Institute of Technology, SE-106 91 Stockholm, Sweden 2 Laboratory for Advanced Materials and Institute of Fine Chemicals, East China University of Science and Technology, Shanghai 200237, Peoples Republic of China 3 School of Science and Technology, rebro University, 701 82 rebro, Sweden

Received 16 September 2009; accepted 11 February 2010; published online 10 March 2010 Metal ions play essential roles in biological processes and have attracted much attention in both experimental and theoretical elds. By using the molecular dynamics simulation technology, we here present a tting-rening procedure for deriving Lennard-Jones parameters of aqua metal ions toward the ultimate goal of accurately reproducing the experimentally observed hydration free energies and structures. The polarizable SWM4-DP water model proposed by Lamoureux et al. J. Chem. Phys. 119, 5185 2003 is used to properly describe the polarization effects of water molecules that interact with the ions. The Lennard-Jones parameters of the metal ions are rst obtained by tting the quantum mechanical potential energies of the hexahydrated complex and are subsequently rened through comparison between the calculated and experimentally measured hydration free energies and structures. In general, the derived Lennard-Jones parameters for the metal ions are found to reproduce hydration free energies accurately and to predict hydration structures that are in good agreement with experimental observations. Dynamical properties are also well reproduced by the derived Lennard-Jones parameters. 2010 American Institute of Physics. doi:10.1063/1.3352567
I. INTRODUCTION

Ions, commonly found in nature, have signicant effects on the physicochemical and biochemical properties of solutions, and ionic solvation thermodynamics has therefore become a very active eld in both experiment and theory.13 Metal ions have attracted special attention since they play important roles in various ways in living systems. For example, Na+ and K+ ions are essential in biochemical processes in the human body, while some heavy metal ions such as Cd2+ and Hg2+ are highly toxic to organisms. During the past decades, the interaction between metal ions and water molecules have been extensively studied, especially by molecular dynamics MD simulations, which can handle a large number of interacting particles. MD simulations serve as a very useful tool to understand the microscopic processes from a theoretical point of view and offer advantages in the calculation of solvation free energies over the extrathermodynamic assumptions used in experimental measurements.4 So far, plenty of MD simulations have been reported for solvated metal ions by using different techniques, from classical interactive potentials to quantum mechanical QM calculations.2,5

Author to whom correspondence should be addressed. Electronic mail: agren@theochem.kth.se.

The advantage of classical MD simulations using force elds is that they are much less time-consuming than highlevel calculations such as QM/molecular mechanical MM methods6,7 and CarParrinello MD8 simulations and offer an easier way to calculate the solvation free energies through free energy perturbation FEP or thermodynamics integration TI algorithms. However, it is essential to use sufciently accurate interactive potential functions to make the simulations realistic and meaningful. Conventional MD simulations use pair potentials or effective pair potentials, which cannot fully take into account the many-body effects between the particles. Specically, in the solvation process of metal ions, the solvent molecules in the rst solvation shell are highly polarized by the net charge on the ion, and such effects cannot be neglected. To solve this problem, some polarizable water models have been proposed918 to account for the polarization effect, i.e., the most important manybody effect. Among these models, the SWM4-DP water model proposed by Lamoureux et al.18 has been proven very useful in the simulations of solvated metal ions.19 In this paper, we present a tting-rening parameterization procedure for MD simulations of metal ions in aqueous solution, based on the SWM4-DP water model. An initial guess of the Lennard-Jones LJ parameters is based on highlevel QM calculations of hexahydrated metal ions, and the parameters are rened to reproduce experimental hydration structures and free energies. The capability of the parameters in reproducing dynamical properties is also evaluated.
2010 American Institute of Physics

0021-9606/2010/132 10 /104505/7/$30.00

132, 104505-1

104505-2

Li et al.

J. Chem. Phys. 132, 104505 2010

II. COMPUTATIONAL DETAILS A. Parameterization protocol

The SWM4-DP polarizable water model18 is used throughout this work. For each metal ion, the LJ parameters for the metal-hydrogen interactions are set to zero, and the LJ parameters for the metal-oxygen interactions are determined through the following steps. Step 1. By using GAUSSIAN 03 program package,20 the octahedral M H2O n+ M for metal complex is opti6 mized at the MP2/def2-TZVPP level, and the six metaloxygen distances are then varied to generate the potential energy curve around the energetically minimized structure at the same level of theory. Here the potential energy denotes the interaction energy between the metal ion and the six coordinating water molecules, i.e., EQM = EQM M H2O n+ EQM Mn+ 6EQM H2O . In 6 the QM calculations the geometries of the water molecules are kept the same as that in the SWM4-DP water model. Step 2. The LJ parameters for modeling the metaloxygen van der Waals vdW interactions, MO and MO, are derived by tting the MM potential energies to the calculated QM potential energy curve in step 1. Here the MM potential energy consists of three terms, the polarization term represented by the harmonic potential of the Drude particle in the SWM4-DP water model, the electrostatic term represented by Coulomb interactions between point charges, and the vdW term denoted by LJ interactions,18 EMM =
i

FIG. 1. Thermodynamic cycle for hydration free energy calculations.

shown in Fig. 2, which provides convenience in the subsequent renements of the LJ parameters. Step 3. Based on the linear relationship between EMM and Ghyd obtained in step 2, the LJ parameters are rened to reproduce the EMM values that correspond to the experimental Ghyd reported by Marcus.3 It takes two or three rounds of renements for each metal ion to nd the appropriate LJ parameters that reproduce the experimental hydration free energies. Step 4. The LJ parameter obtained from step 3 are further rened by comparing the calculated metal-water distances and coordination numbers in the rst hydration shell with experimental observations. Here the experimental data from Ohtaki2 and some other papers1,2124 are referenced. For some ions such as Cd2+, Hg2+, Ca2+ and K+, the LJ parameters from step 3 give too large metal-water distances and coordination numbers. We then shift the minimum of the MM potential curves of these ions toward smaller metal-water distances than that from QM calculations and derive new LJ parameters that reproduce experimental metal-water distances and coordination numbers while maintaining the correct hydration free energies. Several more rounds of renements are needed in this step, and in general less than ten rounds are necessary for each ion during all of the four steps of parametrization. Finally dynamical properties such as self-diffusion coefcient and mean residence time MRT are also calculated to test these derived LJ parameters for metal ions.
0

1 2 kDrOD + 2 4
i MO

s MO

q Mq s + rMs i
12

q sq s
j s,s 6

rss

rMO
OO 12

MO

rMO
OO 6

+
i j

OO

rOO

rOO

where the summations over i and j run over the six coordinating water molecules and s and s represent all interacting sites in the SWM4-DP water model.18 To obtain the LJ parameters, the MM potential is tted to reproduce the energy minimum of the QM potential curve, i.e., MO and MO are determined through two equations. The MM energy is equal to the QM energy at the QM optimized metal-water distance, and the derivative of the MM energy is equal to zero at the same metal-water distance. Such directly derived LJ parameters are then used in MD simulations and FEP calculations using the thermodynamic cycle shown in Fig. 1 to test their capabilities in reproducing experimental data. The results show that these parameters give more or less good structures of the rst hydration shell; however they systematically underestimate the hydration free energies. Also it is found that a linear relationship exists between the MM energies of the hexahydrated complexes and the calculated hydration free energies, as

Hydration free energy (kJ/mol)

-1000

-2000

-3000

-4000

-5000 -3500 -3000 -2500 -2000 -1500 -1000

EMM (kJ/mol)

-500

FIG. 2. Linear relationship between the MM potential energy of the hexahydrated complex M H2O n+ and the corresponding calculated hydration free 6 energy. Black squares show the hydration free energies and MM potential energies calculated from the initial LJ parameters that are directly derived from QM potential energies, while red circles represent the calculated energies from the nal rened LJ parameters.

104505-3

Computer simulations of aqua metal ions

J. Chem. Phys. 132, 104505 2010

B. Simulation details

MD simulations are carried out by the GROMACS program version 4.0.5, double precision .25,26 For each metal ion, the simulation system consists of a metal ion as solute and 1000 SWM4-DP water molecules as solvent in a rhombic dodecahedron unit cell. An isothermal-isobaric NPT ensemble is used in the simulations with P = 1 atm and T = 298 K, maintained by using the ParrinelloRahman barostat27,28 and the NosHoover thermostat,29,30 respectively. All of the simulations are carried out with a time step of 1 fs, and the cutoff radius for the Coulomb and vdW short-range interactions is set to 10 . During the simulations, periodical boundary conditions are applied, and the long-range electrostatic interactions between the atoms are calculated by the particle mesh Ewald PME approach.31,32 The long-range vdW interactions are also taken into account by introducing the dispersion correction and neglecting the repulsion term, as implemented in GROMACS program package.26 Water molecules are kept rigid by the SETTLE algorithm.33 The system is rst subject to an energy minimization and subsequently to a 1 ns simulation to reach equilibrium, after which production simulations of 1 ns as well as free energy calculations are carried out. Dynamical properties such as self-diffusion coefcient and MRT are calculated from the saved trajectories.
1. FEP calculations

bation energies are saved every 2 fs. The transformation of the parameter is performed forward and backward, and the charging free energy is averaged to minimize hysteresis errors. In the FEP calculation of the solvation free energy of the uncharged metal ion, Gcav, the = 0 state does not correspond to anything in water, while the = 1 state corresponds to the uncharged metal ion in water. Here, Gcav = GFEP none M0 . The LJ interactions between the metal ion and the water molecules are not linearly interpolated since such calculations involve the creation of the uncharged metal ion and a singularity problem arises when is close to 0. Therefore soft-core potentials are used to avoid the singularity problem,35,36 with soft-core parameter equal to 0.5 and soft-core power equal to 1 as suggested by Shirts.37 The coupling parameter is also varied from 0 to 1 forward and then back to 0 backward with = 0.05, and the value of Gcav is obtained as an average of free energy integrals.
2. Calculation of dynamical properties

The self-diffusion coefcients D and MRT are calculated to verify the performance of the derived LJ parameters in reproducing dynamical properties of solvated metal ions. In this work we use the Einstein relation to determine the self-diffusion coefcients from the mean square displacement of the ion, 6Dion = lim
t

A thermodynamic cycle is used to calculate the hydration free energy, as shown in Fig. 1. The hydration free energy is therefore divided into two contributions, the charging free energy Gelec and the energy to create an uncharged metal ion in water, Gcav.34 The latter is equal to the solvation free energy of the uncharged metal ion. The charging free energy Gelec is calculated through the FEP approach by introducing a parameter varying from 0 to 1 to couple the two states, i.e., the uncharged and the charged states of the metal ion. During the FEP calculation of the charging process, the Coulomb interactions are linearly interpolated, and the Gibbs free energy difference between the uncharged and the charged states is obtained by integrating the derivative dG / d over ,
1

d R t R 0 dt

To calculate the MRT of water molecules in the rst hydration shell of metal ions, we follow the methodology of Impey et al.38 by simply tting to a function nion t , which decays exponentially with respect to the ion, i.e., nion t nion 0 exp t / ion . The function nion t is dened as nion t =
j

P j t0,t0 + t;t

t0 ,

GFEP =
0

dG = d

d
0

H
NPT;

where the summation runs over all water molecules and the angle brackets denote averaging over all t0 times. P j t0 , t0 + t ; t takes the value of 1 if the water molecule j is in the rst hydration shell of the metal ion at time t0 and t0 + t and does not leave the region for any continuous period longer than t . In all other cases P j t0 , t0 + t ; t takes the value of 0. In our calculations the parameter t is set to 0.2 ps.
III. RESULTS AND DISCUSSION A. LJ parameters for ion-water interactions

where the derivative of the Hamiltonian H with respect to can be calculated from the -dependent potential functions.26 It should be noted that the -dependent total potential energy is not pairwise decomposable due to the polarization effect of the Drude particle in the SWM4-DP water model.19 In this work, the charging free energy is computed as an integral over discrete windows with varying from 0 to 1 and = 0.05. Note that the = 0 state corresponds to the uncharged state and the = 1 state corresponds to the charged state. Here, Gelec = GFEP M0 Mn+ . The simulations are performed sequentially, and the nal conguration of each window is used as the starting conguration of the next one. At each window, the system is subject to 10 ps equilibration and 50 ps production simulations, during which the pertur-

The nally rened LJ parameters for metal-oxygen interactions are listed in Table I. As described in Sec. II, we start from the ab initio potential energy curve of an octahedral M H2O n+ complex and rene the LJ parameters accord6 ing to experimental hydration free energy and structure of the rst hydration shell. The selection of the M H2O n+ 6 model is motivated by the wide representation of the hexahydrated structure in ion-water complexes. This model has taken into account most of the polarization effect and can give a good initial guess for LJ parameters, saving much effort in the tting-rening procedure and therefore making

104505-4

Li et al.
0

J. Chem. Phys. 132, 104505 2010

dGQQ/d (kJ mol [] )

TABLE I. Rened LJ parameters for metal-oxygen vdW interactions in ion-water complex. is in angstrom unit and in kJ/mol. Ion Al3+ Zn2+ Mg2+ Cd2+ Hg2+ Ca2+ Sn2+ Na+ K+
MO MO

-1 -1

-2000
K(I) Na(I) Sn(II) Ca(II) Hg(II) Cd(II) Mg(II) Zn(II) Al(III)

1.9497 2.0965 2.3027 2.2773 2.2772 2.5985 2.5947 2.6314 2.9659

22.1162 15.9457 3.6165 16.1686 19.0876 5.3589 8.4011 1.7019 1.5514

-4000

-6000

-8000 0 50 0.2 0.4 0.6 0.8 1

dGLJ/d (kJ mol [] )

-1

0 -50 -100 -150 -200 0 0.2 0.4

the parameterization strategy very efcient. Moreover, a linear relationship is found between the calculated hydration free energy and the MM potential energy of the hexahydrated complex shown in Fig. 2 , providing convenience in the parameterization procedure.
B. Hydration free energies

-1

The hydration free energies of the studied metal ions, obtained as the sum of corresponding electrostatic part Gelec and vdW part Gcav, are listed in Table II along with experimental data reported by Marcus.3 It can be seen that the experimental hydration free energies are well reproduced by the rened LJ parameters. The electrostatic and vdW parts of the hydration free energy are obtained as integrations of the corresponding energy derivatives over the coupling parameter . The energy derivatives, dGelec / d and dGcav / d , are shown in Figure 3 as functions of . In general, the dominant contribution to the total hydration free energy originates in the electrostatic part, and this part is highly dependent on the net charge of the metal ion. Nevertheless, the vdW part provides nonnegligible contribution. Several factors may affect the accuracy of the free energy integration. First of all, the values of should be carefully selected to ensure adequate sampling. If the value of the energy derivative varies sharply, more windows should be used to give an accurate estimation of the integration. In Fig. 3, we can see that the electrostatic part of the energy derivative is almost linear, and the use of equally spaced windows is therefore expected to work well. The vdW part is nonlinear
TABLE II. Hydration free energies in kJ/mol for metal ions along with experimental values reported by Marcus Ref. 3 . Ion Al3+ Zn2+ Mg2+ Cd2+ Hg2+ Ca2+ Sn2+ Na+ K+ Gelec 4413.7 1861.7 1815.9 1637.8 1621.9 1460.9 1425.0 361.0 295.3 Gcav 109.3 90.2 14.9 112.9 138.0 38.5 69.4 4.8 1.8 Ghyd 4523.0 1951.9 1830.8 1750.7 1759.9 1499.4 1494.4 365.8 297.1 Expt. 4525 1955 1830 1755 1760 1505 1490 365 295

0.6

0.8

FIG. 3. The energy derivatives dGelec / d and dGcav / d coupling parameter .

as functions of the

due to the use of soft-core potentials; however, the variation in the energy derivative is small in magnitude. Therefore the use of the 21 uniformly selected values can give reliable integrations. The simulation time used in the FEP calculations can also affect the reliability of the calculated free energies since long simulation times can signicantly reduce the statistical error. At each , the error is estimated through the improved block averaging method for correlated data, which is proposed by Hess39 and implemented in the g_analyze tool of the GROMACS program package.26 The error estimations are shown as error bars in Fig. 3. It can be seen that the errors are very small comparing with the values of the energy derivatives. For each metal ion, the errors for dGelec / d and dGcav / d are integrated over and then summed to give the error estimation of the total hydration free energy. The largest estimated error of Ghyd is 9.3 kJ/mol for Al3+ and the smallest is 3.4 kJ/mol for K+ . Therefore the use of 10 ps equilibration+ 50 ps production for each is adequate to give an accurate integration of the hydration free energy. Another important issue in free energy calculations is the treatment of long-range electrostatic interactions. The PME method, which can reliably take into account the long-range behavior of Coulomb interactions, is known to introduce artifacts when applied to systems with net charges.4044 Nevertheless, the results reported by Bogusz et al.43 show that the free energies calculated from the standard Ewald summation method is size-insensitive. In this work we also checked the size-dependence of Gelec of the Al3+ ion in aqueous solution. The Al3+ ion is simulated in three rhombic dodecahedron boxes containing 500, 1000, and 2000 water molecules, respectively, under 1 atm and 298 K. The calculated

104505-5

Computer simulations of aqua metal ions


35 30 25
Al(III) Zn(II) Mg(II) Cd(II) Hg(II) Ca(II) Sn(II) Na(I) K(I)

J. Chem. Phys. 132, 104505 2010

C. Hydration structures

g(r)

20 15 10 5 0 1 10 8 1.5 2 2.5 3 3.5 4

The radial distribution functions RDFs of metal ions are shown in Fig. 4. The maximal g r value goes up with respect to the increase in net charge, indicating that the coordination strength is highly dependent on the electrostatic interactions. Detailed results are listed in Table III. Overall, the hydration structures predicted by the rened LJ parameters are in good agreement with experimental data.
1. Al3+

4.5

6 4 2 0 1 1.5 2 2.5 3 3.5 4 4.5 5

The highly charged Al3+ ion has the strongest interaction with the surrounding water molecules. The hydration number is calculated to be 6 by our parameters, in accordance with the experimental value proposed by Ohtaki.2 Under such strong electrostatic interactions, the octahedral structure of the rst hydration shell is very stable. The AlO distance in the rst hydration shell is calculated to be 1.90 , while the experimentally measured value is 1.871.90 .2 The number of water molecules in the second hydration shell is predicted to be 13.2, which is also comparable with the experimental value of 12.2
2. Zn2+ and Mg2+

Coord. Number

rM-O (Angstrom)
FIG. 4. Calculated RDFs and corresponding hydration numbers of metal ions in aqueous solutions.

Gelec values are 4416.3, 4413.0, and 4414.1 kJ/mol, respectively, showing negligible size-dependence. Grosseld et al. also suggested that the PME-induced artifacts would be small given the relatively large system size and the high dielectric constant of the solvent.45 Based on the sizeindependent nature of the Gelec term, the calculated hydration free energy is considered to be reliable, and we can safely expand this conclusion to other metal ions with less net charge than Al3+. Moreover, the cutoff radius of shortranged Coulomb interactions has little inuence on the calculated free energies data not shown . Therefore we conclude that the PME method combined with 10 Coulomb cutoff can give reliable hydration free energies of charged metal ions.

The Zn2+ and Mg2+ ions also have the regular octahedral hydration structures. The hydration numbers of these two ions are both predicted to be 6, in agreement with experimental data.2 Their hydration structures are very similar, except that the rst RDF peak of Zn2+ is sharper than Mg2+ see Fig. 4 , indicating a slightly tighter bonding between the Zn2+ ion and the surrounding water molecules. The metaloxygen distances of Zn2+ and Mg2+ are both calculated to be 2.09 and 2.08 , respectively, comparable to experimental data summarized by Ohtaki 2.082.17 for Zn2+ and 2.00 2.15 for Mg2+ .2 For these two ions, the calculated numbers of water molecules in the second hydration shell are slightly larger than the experimental value of 12.2
3. Cd2+ and Hg2+

The Cd2+ and Hg2+ ions are less extensively studied by MD simulations than their zinc and magnesium counterparts probably due to their large number of electrons. For Cd2+, our LJ parameters give a hydration number of 6.2, indicating

TABLE III. Calculated structural parameters of metal ions in aqueous solution. rmax,1 and rmax,2 in angstrom denote the positions of the rst and second RDF peaks, respectively. Nc,1 and Nc,2 denote the number of water molecules in the rst and second hydration shells, respectively. Available experimental data are shown in parentheses. For Sn2+ the results from two QM MD simulations are listed. Ion Al3+ Zn2+ Mg2+ Cd2+ Hg2+ Ca2+ Sn2+ Na+ K+ 1.90 2.09 2.08 2.27 2.32 2.48 2.55 2.44 2.77 rmax,1 1.871.90 2.002.15 2.082.17 2.30 2.32 2.46 2.39,2.53 2.402.50 2.602.95 6.0 6.0 6.0 6.2 7.0 7.7 8.0 5.9 7.2 Nc,1 6.0 6.0 6.0 6.07.0 7.0 8.0 7.8,8.0 4.08.0 4.08.0 4.14 4.32 4.42 4.46 4.53 4.64 4.75 4.43 4.83 rmax,2 3.994.15 3.954.26 4.104.28 4.30 4.58 4.60,4.90 13.2 15.8 14.5 19.6 17.1 18.4 20.7 17.3 20.8 Nc,2 12.0 12.0 12.0 Ref. 2 2 2 21 22 23 46 and 47 2 1, 2, and 24

22.0,23.6

104505-6

Li et al.

J. Chem. Phys. 132, 104505 2010 TABLE IV. Calculated dynamical properties: self-diffusion coefcient Dion, in 105 cm2 s1 and MRT ion, in picosecond . Experimental estimations of self-diffusion coefcients are shown in parentheses Ref. 55 . Ion Al Zn2+ Mg2+ Cd2+ Hg2+ Ca2+ Sn2+ Na+ K+
a 3+

a variable rst hydration shell of the Cd2+ ion. This is in accordance with the results of DAngelo et al. and Chillemi et al.21,48 The CdO distance in the rst hydration shell is calculated to be 2.27 , close to the experimental value from extended x-ray-absorption ne structure EXAFS measurements 2.30 .21 The MD simulations performed by Chillemi et al. suggest shorter CdO distances 2.22 and 2.25 and larger hydration numbers 6.3 and 6.8 21 by using nonpolarizable SPC/E and TIP5P water models. For Hg2+, the calculated hydration number is 7.0, in accordance with the heptacoordinated result of Chillemi et al.22,49,50 The HgO distance in the rst hydration shell is obtained as 2.32 , in agreement with experimental data from EXAFS and x-ray appearance near-edge structure measurements 2.32 .22 An earlier QM/MM study by Kritayakornupong et al.51 gives a HgO distance of 2.42 and hydration number of 6.2.
4. Ca2+ and Sn2+

Dion 0.427 0.546 0.468 0.603 0.606 0.570 0.602 1.093 1.742 0.541 0.703 0.706 0.720 0.850 0.792 1.334 1.957

ion

n.o.a n.o. n.o. 748.9 333.2 142.4 381.7 14.2 5.0

n.o. = No observable ligand exchange during simulations.

Ca2+ and Sn2+ are larger than the above discussed ions and have the largest hydration numbers among these investigated ions. For Ca2+, the calculated hydration number is 7.7, in good agreement with the octahydrated structure determined from x-ray scattering measurements.23 The calculated CaO distance, 2.48 , is also comparable to that determined in EXAFS experiments 2.46 .23 For Sn2+, the hydration number and SnO distance in the rst hydration shell are calculated to be 8.0 and 2.55 , respectively, in accordance with the result obtained from QM/MM study by Hofer et al. hydration number of 8 and SnO distance of 2.53 .46 A recent QM charge eld study reported by Lim et al. suggests a hydration number of 7.8 and mean SnO distance of 2.39 in the rst hydration shell.47
5. Na+ and K+

As monovalent ions, Na+ and K+ interact much weaker with the surrounding water molecules than the other ions studied in this work, their rst hydration shells are therefore less stable. For the Na+ ion, the hydration number is predicted to be 5.9, comparable to the results obtained from diffraction measurements 4.08.0 .2 Some recent MD simulations also give similar hydration numbers 5.26.0 .19,45,5254 The NaO distance in the rst hydration shell is calculated to be 2.44 , in accordance with both experimental data 2.402.50 2 and simulated results 2.332.49 .19,45,5254 For the K+ ion, the calculated hydration number is 7.2, in agreement with the result from diffraction measurement 7.5 .24 Hydration numbers obtained from other MD simulations are between 6.75 and 8.3.19,45,53,54 The KO distance in the rst hydration shell is calculated to be 2.77 , in good agreement with the experimental value summarized by Marcus 2.80 .1 Other MD simulations predict similar values 2.742.81 .19,45,53,54
D. Dynamical properties

self-diffusion coefcients show qualitative agreement with experimental observations, although the values are systematically underestimated.55 As pointed out by Impey et al.,38 the calculated self-diffusion coefcients suffer large uncertainties due to the lack of statistics since there is only one ion in the simulation box. Therefore the performance of our derived parameters is quite acceptable in qualitatively reproducing the experimental self-diffusion coefcients. MRT is an important property that characterizes ligand exchange processes of the solvated metal ion. Experimentally measured MRTs range from 109 to 109 s, where only the nanosecond time-scale is accessible in MD simulations. In our calculations, no ligand exchange is observed for the Al3+, Zn2+, and Mg2+ ions, in accordance with the experimental observation that the MRTs of these three ions are much larger than the nanosecond time-scale.56 For the Cd2+, Hg2+, Ca2+, and Sn2+ ions, the calculated MRTs are 748.9, 333.2, 142.4, and 381.7 ps, respectively, corresponding well with the experimental estimation of 108 1010 s.46,56 The MRTs of Na+ and K+, 14.2 and 9.0 ps, are much smaller than experimental estimations.56 However, they are close to those calculated by Impey et al. 9.9 and 4.8 ps, respectively .38 A recent QM MD study also gives similar values for the MRTs of aqueous Na+ and K+ ions.54 Since Na+ and K+ have weak interactions with the surrounding water molecules, it is convincing that they have much smaller MRT values comparing with their divalent and trivalent counterparts.
IV. CONCLUSION

To further test the availability of the derived parameters, we calculated the self-diffusion coefcient and MRT for each metal ion; the results are listed in Table IV. Our calculated

In this paper, the Lennard-Jones parameters for modeling metal-oxygen vdW interactions in aqueous solutions are derived through a tting-rening procedure. First the potential energy curve of the hexahydrated metal ion is calculated at the MP2/def2-TZVPP level. The initial guess of LJ parameters is then obtained by tting the MM potential energy curve to the QM one. The LJ parameters are subsequently subject to renement by comparison between calculated and experimental hydration free energies and structures. In the computations, the FEP methodology is used to accurately calculate the hydration free energies, and MD simulations are performed to give structural information of the solvated metal ions. The nally derived LJ parameters are found to well reproduce experimental hydration free energies and

104505-7

Computer simulations of aqua metal ions


25

J. Chem. Phys. 132, 104505 2010 B. Hess, C. Kutzner, D. van der Spoel, and E. Lindahl, J. Chem. Theory Comput. 4, 435 2008 . 26 D. van der Spoel, E. Lindahl, B. Hess, A. R. van Buuren, E. Apol, P. J. Meulenhoff, D. P. Tieleman, A. L. T. M. Sijbers, K. A. Feenstra, R. van Drunen, and H. J. C. Berendsen, GROMACS, User Manual version 4.0, www.gromacs.org 2005 . 27 M. Parrinello and A. Rahman, J. Appl. Phys. 52, 7182 1981 . 28 S. Nos and M. L. Klein, Mol. Phys. 50, 1055 1983 . 29 S. Nos, Mol. Phys. 52, 255 1984 . 30 W. G. Hoover, Phys. Rev. A 31, 1695 1985 . 31 T. Darden, D. York, and L. Pedersen, J. Chem. Phys. 98, 10089 1993 . 32 U. Essmann, L. Perera, M. L. Berkowitz, T. Darden, H. Lee, and L. G. Pedersen, J. Chem. Phys. 103, 8577 1995 . 33 S. Miyamoto and P. A. Kollman, J. Comput. Chem. 13, 952 1992 . 34 A. Warshel, M. Kato, and A. V. Pisliakov, J. Chem. Theory Comput. 3, 2034 2007 . 35 T. C. Beutler, A. E. Mark, R. C. van Schaik, P. R. Gerber, and W. F. van Gunsteren, Chem. Phys. Lett. 222, 529 1994 . 36 H. Liu, A. E. Mark, and W. F. van Gunsteren, J. Phys. Chem. 100, 9485 1996 . 37 M. R. Shirts and V. S. Pande, J. Chem. Phys. 122, 134508 2005 . 38 R. W. Impey, P. A. Madden, and I. R. McDonald, J. Phys. Chem. 87, 5071 1983 . 39 B. Hess, J. Chem. Phys. 116, 209 2002 . 40 F. Figueirido, G. S. Del Buono, and R. M. Levy, J. Chem. Phys. 103, 6133 1995 . 41 G. Hummer, L. R. Pratt, and A. E. Garca, J. Phys. Chem. 100, 1206 1996 . 42 G. Hummer, L. R. Pratt, and A. E. Garca, J. Chem. Phys. 107, 9275 1997 . 43 S. Bogusz, T. E. Cheatham III, and B. R. Brooks, J. Chem. Phys. 108, 7070 1998 . 44 P. H. Hnenberger and J. A. McCammon, Biophys. Chem. 78, 69 1999 . 45 A. Grosseld, P. Ren, and J. W. Ponder, J. Am. Chem. Soc. 125, 15671 2003 . 46 T. S. Hofer, A. B. Pribil, B. R. Randolf, and B. M. Rode, J. Am. Chem. Soc. 127, 14231 2005 . 47 L. H. V. Lim, T. S. Hofer, A. B. Pribil, and B. M. Rode, J. Phys. Chem. B 113, 4372 2009 . 48 P. DAngelo, V. Migliorati, G. Mancini, and G. Chillemi, J. Phys. Chem. A 112, 11833 2008 . 49 P. DAngelo, V. Migliorati, G. Mancini, V. Barone, and G. Chillemi, J. Chem. Phys. 128, 084502 2008 . 50 G. Mancini, N. Sanna, V. Barone, V. Migliorati, P. DAngelo, and G. Chillemi, J. Phys. Chem. B 112, 4694 2008 . 51 C. Kritayakornupong and B. M. Rode, J. Chem. Phys. 118, 5065 2003 . 52 J. A. White, E. Schewegler, G. Galli, and F. Gygi, J. Chem. Phys. 113, 4668 2000 . 53 M. Carrillo-Tripp, H. Saint-Martin, and I. Ortega-Blake, J. Chem. Phys. 118, 7062 2003 . 54 S. S. Azam, T. S. Hofer, B. R. Randolf, and B. M. Rode, J. Phys. Chem. A 113, 1827 2009 . 55 D. R. Lide, CRC Handbook of Physics and Chemistry, 84th ed. CRC Press, Boca Raton, FL, 2003 . 56 L. Helm and A. E. Merbach, Coord. Chem. Rev. 187, 151 1999 .

structures of the aqua metal ions. Dynamical properties such as self-diffusion coefcients and MRTs are also qualitatively reproduced by the derived parameters. The nine metal ions included in our study, Al3+, Zn2+, Mg2+, Cd2+, Hg2+, Ca2+, Sn2+, Na+, and K+, differ in charge and radius, and their hydration free energies and structures are also distinct from each other. Nevertheless, the tting-rening procedure successfully gave the optimal set of LJ parameters for each ion, proving universal validity of the parameterization strategy.
ACKNOWLEDGMENTS

This work was supported by a grant from the Swedish Infrastructure Committee SNIC for the project Multiphysics Modeling of Molecular Materials under Grant No. SNIC 025/08-4.
Y. Marcus, Chem. Rev. Washington, D.C. 88, 1475 1988 . H. Ohtaki, Chem. Rev. Washington, D.C. 93, 1157 1993 . Y. Marcus, Biophys. Chem. 51, 111 1994 . 4 Y. Marcus, J. Chem. Soc., Faraday Trans. 82, 233 1986 . 5 B. M. Rode, C. F. Schwenka, and A. Tongraar, J. Mol. Liq. 110, 105 2004 . 6 A. Warshel and M. Levitt, J. Mol. Biol. 103, 227 1976 . 7 J. Gao, Acc. Chem. Res. 29, 298 1996 . 8 R. Car and M. Parinello, Phys. Rev. Lett. 55, 2471 1985 . 9 L. X. Dang, J. Chem. Phys. 97, 2659 1992 . 10 S. W. Rick, S. J. Stuart, and B. J. Berne, J. Chem. Phys. 101, 6141 1994 . 11 J. Brodholt, M. Sampoli, and R. Vallauri, Mol. Phys. 86, 149 1995 . 12 I. M. Svishchev, P. G. Kusalik, J. Wang, and R. J. Boyd, J. Chem. Phys. 105, 4742 1996 . 13 L. X. Dang and T.-M. Chang, J. Chem. Phys. 106, 8149 1997 . 14 H. Saint-Martin, J. Hernndez-Cobos, M. I. Bernal-Uruchurtu, I. OrtegaBlake, and H. J. C. Berendesen, J. Chem. Phys. 113, 10899 2000 . 15 P. J. van Maaren and D. van der Spoel, J. Phys. Chem. B 105, 2618 2001 . 16 H. A. Stern, F. Rittner, B. J. Berne, and R. A. Friesner, J. Chem. Phys. 115, 2237 2001 . 17 H. Yu, T. Hansson, and W. F. van Gunsteren, J. Chem. Phys. 118, 221 2003 . 18 G. Lamoureux, A. D. Mackerell, Jr., and B. Roux, J. Chem. Phys. 119, 5185 2003 . 19 G. Lamoureux and B. Roux, J. Phys. Chem. B 110, 3308 2006 . 20 M. J. Frisch, G. W. Trucks, H. B. Schlegel et al., GAUSSIAN 03, Revision D.01, Gaussian, Inc., Wallingford, CT, 2004. 21 G. Chillemi, V. Barone, P. DAngelo, G. Mancini, I. Persson, and N. Sanna, J. Phys. Chem. B 109, 9186 2005 . 22 G. Chillemi, G. Mancini, N. Sanna, V. Barone, S. D. Longa, M. Benfatto, N. V. Pavel, and P. DAngelo, J. Am. Chem. Soc. 129, 5430 2007 . 23 F. Jalilehvand, D. Spngberg, P. Lindqvist-Reis, K. Hermansson, I. Persson, and M. Sandstrm, J. Am. Chem. Soc. 123, 431 2001 . 24 G. W. Neilson and N. Skipper, Chem. Phys. Lett. 114, 35 1985 .
2 3 1

You might also like