You are on page 1of 12

Journal of Molecular Structure: THEOCHEM 806 (2007) 23–34

www.elsevier.com/locate/theochem

Structural and thermodynamic properties of liquid ethylene


carbonate and propylene carbonate by Monte Carlo Simulations
Luciene Borges Silva, Luiz Carlos Gomide Freitas *

Departamento de Quı´mica, Universidade Federal de São Carlos, Caixa Postal 676, CEP 13565-905, São Carlos, Sao Paulo, Brazil

Received 9 June 2006; received in revised form 3 October 2006; accepted 11 October 2006
Available online 20 October 2006

Abstract

Monte Carlo simulations have been used to investigate molecular association in pure liquid ethylene carbonate (EC) and propylene
carbonate (PC). Standard force fields have been developed in order to reproduce accurately experimental pure liquid properties. Inter-
molecular interactions were described by combining standard OPLS Lennard–Jones parameters and partial atomic charges from fitting
to the MP2/6-31G* electrostatic potential surface (EPS) with the CHELPG procedure. It was found that the inclusion of MP2 electron
correlation on calculations of the partial atomic charges is required in order to achieve a proper description of the experimental liquid
properties. The resultant force fields yielded average errors of 1–2% in computed densities and heats of vaporization. A thorough char-
acterization of the liquid structures was performed with radial distribution functions, coordination numbers, energy distributions and
dipole–dipole correlations. The electrostatic interactions in condensed phase lead the neighboring molecules to a preferential head to tail
alignment of the dipoles. In contrast, the most energetically favored configuration for the EC and PC dimers in the gas phase exhibited
antiparallel dipoles. The results of simulations with the partial charges set to zero revealed that the electrostatic term plays a determining
role in structuring the liquids. Basis set dependence of EPS was investigated comparing the CHELPG MP2/6-31G* charges and dipole
moments with the ones obtained at the MP2/{6-311G**, 6-31++G**, 6-311++G** and 6-311++G(3df,3pd)} level. Polarization effects
due to intermolecular interaction were also investigate and a good reproduction of the experimental density and heat of vaporization
(1.5% and 2% of deviation, respectively) was accomplished for liquid EC using CHELPG charges obtained for the most stable EC dimer
configuration (antiparallel) at the MP2/6-31++G** level of theory.
 2006 Elsevier B.V. All rights reserved.

Keywords: Monte Carlo simulation; CHELPG charges; Liquid structure; Ethylene carbonate; Propylene carbonate

1. Introduction solvents [3], solutions melting below room temperature


and having a wide range of dielectric constants may be
Cyclic carbonates such as ethylene carbonate (EC) and obtained. PC is a versatile solvent, as it has an extensive
propylene carbonate (PC) are dipolar aprotic solvents liquid range and dissolves a large variety of organic and
which find a broad range of applications as both solubiliz- inorganic substances [4]. Solutions of lithium salts and these
ing and reactional environments. These solvents are notable cyclic carbonates associated with other organic solvents
for their physical properties including high dielectric con- have been the electrolytes of choice for lithium batteries
stants, high boiling points and high dipole moments (Table [5]. Recently, the use of PC in the degreasing, paint strip-
1). EC is mostly used in mixtures with other liquids because ping and cosmetic industries [6–8] has increased significant-
of its high freezing point. Since it has high miscibility with ly due to its biodegrability [9] and low toxicity [10].
water [2] as well as with a large variety of non-aqueous In spite of the great interest in EC and PC in many
fields, the nature and extension of their molecular associa-
tion in the condensed phase remain unclear and has been
*
Corresponding author. Tel.: +55 16 3351 8060; fax: +55 16 3351 8063. the topic of much discussion in the literature. Several stud-
E-mail address: gomide@dq.ufscar.br (L.C.G. Freitas). ies focusing the dielectric properties of cyclic carbonates

0166-1280/$ - see front matter  2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.theochem.2006.10.014
24 L.B. Silva, L.C.G. Freitas / Journal of Molecular Structure: THEOCHEM 806 (2007) 23–34

Table 1 lo simulation technique. For this purpose force fields have


Physical properties of EC and PC [1] been developed in order to reproduce accurate results for
Property EC PC available experimental pure liquid properties. The wealth
Dielectric constant 89.78 (40 C) 64.92 (25 C) of information obtained from the simulations has been
Dipole moments (D) 4.87 4.94 used to characterize the liquid structures, which includes
Boiling point (C) 248 242 radial distribution functions, coordination numbers, ener-
Melting point (C) 36 -49
gy distributions and dipole–dipole correlations. Moreover,
a series of EC and PC dimer configurations in the gas phase
has been evaluated using ab initio calculations in order to
have been reported [11–13]. They consistently suggested the provide further insights into the association properties of
possibility of molecular association in EC and PC. Dielec- the cyclic carbonates. The unusually large experimental
tric measures complemented by IR and NMR data for PC gas phase dipole moments of EC and PC suggest that
have been reported by Simeral and Amey [14]. Their results strong dipole–dipole interactions can be expected in the
indicated that this solvent behaves as a typical polar liquid liquid phase. This issue has also been addressed through
with strong dipole–dipole interactions but nonspecific asso- the results of simulations carried out with the partial charg-
ciation. The absence of specific interactions in EC and PC es set to zero.
was also reported by Payne and Theodorou [13]. IR and
Raman spectroscopy studies by Fini et al. [15] have shown 2. Computational methods
that dipolar forces induce short range orientation in EC
and PC. Results of a Raman study of EC performed by 2.1. Ab initio calculations
Schindler et al. [16] have been interpreted by assuming per-
manent dipole–dipole interactions. Based on FTIR and Several ab initio calculations in the gas phase were per-
Raman data, Klassen et al. [17] have suggested strong formed to both develop the force field and evaluate the
intermolecular association in pure EC liquid. In contrast, interaction in EC and PC dimers. The geometrical param-
theoretical studies of solutions of lithium perchlorate in eters required to describe the EC and PC monomers in the
EC and PC by Li and Balbuena [18] showed little molecu- simulations were computed from MP2/6-31G* ab initio cal-
lar association in EC. Subsequently, a more detailed theo- culations. Experimental studies have shown that the EC
retical study of the association in EC and PC was reported molecule has a planar conformation in the liquid state
by Wang and Balbuena [19]. In their work, ab initio and [27]. A planar ring conformation has also been reported
density functional theory (DFT) calculations performed for PC in the liquid phase [28]. For consistency with the
for dimers showed that the cyclic carbonate molecules are experimental data, the gas phase geometries of the EC
associated mainly through CH–O interactions. and PC monomers were optimized constraining the ring
Computer simulation methods such as Monte Carlo and to a planar conformation. The partial atomic charge distri-
Molecular Dynamics are suitable tools for investigations of butions which were needed to describe the electrostatic
liquid properties from an atomic level [20]. These method- potential in the simulations were computed using the
ologies provide a direct link between the microscopic CHELPG procedure [25]. In order to evaluate the effect
details of the system and its macroscopic properties. A of electron correlation on charge distributions, the EPS
key component in simulations is the quality of the force partial charges were computed from both HF and MP2
fields which are used to describe the molecular interactions densities. The basis set dependence on the CHELPG charg-
[21–23]. Force fields for liquid simulations are usually opti- es for the EC and PC monomers was evaluated by perform-
mized to reproduce experimental pure liquid properties, ing the calculations with the 6-31G*, 6-31++G**, 6-
specially densities and heats of vaporization. Typically, 311G**, 6-311++G** and 6-311++G(3df,3pd) basis set.
intermolecular interactions in force fields are represented Gas phase heat capacities for each molecule were calculat-
by an effective pair potential consisting of Lennard–Jones ed using computed HF/6-31G* vibrational frequencies
and Coulomb terms [20–23]. Standard Lennard–Jones scaled by a factor of 0.91 [29]. Geometry optimizations
parameter are available for all common functional groups using ab initio HF/6-31G*//HF/6-31G* calculations were
found in organic liquid [23]. However, the assignment of performed for a variety of EC and PC dimer configura-
the partial charges used to describe the coulombic potential tions. During the optimizations, the monomer structures
is critical in force field development. An approach that has were kept fixed at their optimized MP2/6-31G* forms.
been pursued is to fit the charges to reproduce electronic All the ab initio calculations were carried out with the
properties, such as dipole moments and electrostatic poten- GAUSSIAN-94 program [30].
tials [24]. The most common alternative in liquid simula-
tions has focused on partial charges from fitting to the 2.2. Force field
electrostatic potential surface (EPS) using ab initio calcula-
tions [25,26]. Consistently with the usual procedure in liquid simula-
The present work presents an investigation of molecular tions [21–23], the EC and PC monomers were described
association in pure liquid EC and PC using the Monte Car- as a set of interaction sites distributed along of the mole-
L.B. Silva, L.C.G. Freitas / Journal of Molecular Structure: THEOCHEM 806 (2007) 23–34 25

cule. In this work, all the atoms are explicitly considered


and the sites are located on the atomic nuclei. Each site
is represented by a set of Lennard–Jones and Coulomb
parameters. The intermolecular interaction energy between
two monomers is then represented in the standard Len-
nard–Jones plus Coulomb form Eq. (1)
XX 12 6
E¼ fqi qj e2 =rij þ 4eij ½ðrij =rij Þ  ðrij =rij Þ g: ð1Þ
i i<j

Standard Lennard–Jones parameters were transferred from


the force field developed for liquid esters [31]. The number
of different parameters was kept to a minimum, such that C
and H atoms were represented by the same r and e values
for all of the CHn groups. Standard geometrical combining
rules are used such that rij = (riirjj)1/2 and eij = (eiiejj)1/2.
The corresponding Lennard–Jones values are listed in
Table 2. Throughout this work, the labels used to define
the EC and PC monomers are: O for the carbonyl oxygen, Fig. 1. Structure of PC monomer and the corresponding labeling of the
C for the carbonyl carbon, OS for the ester oxygen, and Cn atoms used for the EC and PC monomers.
and Hn for the carbon and hydrogen atoms of the CHn
groups as shown in Fig. 1.
The partial atomic charges were obtained from ab initio out in the NPT ensemble at pressure of 1 atm. In order
calculations with the CHELPG method as described in the to compare densities and heats of vaporization with exper-
latter section. The intramolecular interactions were evalu- imental data, simulations were run at temperatures of
ated by harmonic terms for bond stretching and angle 40 C and 150 C for EC and 20 C and 150 C for PC.
bending, and a Fourier series for torsional angle [23]. The Additionally, simulations were also run at 50 C for EC
C–OS bond stretching and both C–OS–CT and O–C–OS and at 25 C for PC for comparison with experimental val-
angle bending parameters were adopted from the work of ues of heat capacities at constant pressure, coefficients of
Charifson et al. [32]. The HC–CT–OS angle bending was thermal expansion (a) and isothermal compressibilities
taken from the OPLS parameters for carbohydrates, and (j). Simulations at higher temperatures for EC were desir-
the OS–C–OS angle value was approached by the OPLS able due to its melting point of 36 C. The simulations for
OS–CO–OS angle bending parameter reported for acetals EC at 150 C and for PC at 20 C and 150 C were per-
[33]. In this study, the cyclic carbonates were constrained formed with both HF and MP2 charge sets. The monomers
to be planar and therefore the ring conformations were were flexible, allowing for bond stretching, angle bending
not altered during the simulations. Thus, only bond lengths and dihedral angle changes for the methyl group in the
and bond angles were varied, with exception of the torsion- PC. In order to compute the heats of vaporization, it was
al motion for the methyl group in propylene carbonate. If required to perform Monte Carlo simulations for a single
distortion from planarity of the ring is desirable, the devel- molecule. The gas phase simulations were performed with
opment of the appropriate torsional parameters is required. 3 · 106 configurations of equilibration, followed by
The torsional parameters for the methyl group in propyl- 3 · 106 configurations of averaging. For the liquid simula-
ene carbonate were taken from related types of interaction tions, periodic boundary conditions were employed with
in the OPLS force field [23]. reference cubic cells containing 267 molecules. For each
liquid, the initial solvent cell was created through the suc-
cessive replication of a monomer coordinates. New config-
2.3. Liquid simulations urations of the systems were generated by random selection
of a molecule followed by random translations, rotations
Monte Carlo simulations using standard procedures [20] and torsional movements. Attempts to change the volume
including Metropolis importance sampling were carried of the systems were made at every 1000 configurations by
scaling the center of mass coordinate of all molecules.
Table 2 The ranges of the four types of motion were adjusted to
Lennard–Jones parameters for simulations of liquid EC and PC achieve an acceptance rate of 40–50% for new configura-
Site r (Å) e (kcal mol1) tions generated. The equilibrated cells ranged from 31 to
O 2.96 0.210
35 Å on an edge. Each simulation consisted of an equilibra-
C 3.75 0.105 tion phase with 1 · 107 configurations in which conver-
OS 3.00 0.170 gence for energies and densities was monitored.
C 3.50 0.066 Averaging for the reported properties were done over an
H 2.42 0.015 additional 1 · 107 configurations. Statistical uncertainties
26 L.B. Silva, L.C.G. Freitas / Journal of Molecular Structure: THEOCHEM 806 (2007) 23–34

were calculated during the averaging stage from the fluctu- Table 4
ations in separate averages over batches of 1 · 105 configu- Charges for the EC and PC monomers calculated with 6-31G* basis set at
the HF and MP2 levels and the corresponding dipole moments
rations. The intermolecular nonbonded interactions were
evaluated within a spherical cutoff radius of 12 Å based Ethylene carbonate Propylene carbonate
on the distance between the ring centers. A standard cor- HF MP2 HF MP2
rection for the Lennard–Jones interactions neglected O 0.610 0.512 0.610 0.510
beyond the cutoff was included in the simulations. Long- C 1.000 0.850 1.020 0.857
range electrostatic interactions were evaluated with the OSa 0.490 0.429 0.505 0.442
OSb 0.547 0.484
reaction field approach [34]. The simulations were per- C1 0.438 0.401
formed with the DIADORIM program, a FORTRAN C2 0.235 0.210 0.213 0.166
code developed in our laboratory [35]. C3 0.356 0.362
H1 0.002 0.002
H2 0.030 0.025 0.030 0.030
3. Results and discussions H3 0.095 0.104
Dipole moment/D 6.24 5.38 6.37 5.71
3.1. Molecular structures and charges a
OS atom attached to the C2 atom.
b
OS atom attached to the C1 atom.
Table 3 reports the gas phase molecular geometries of
EC and PC from optimizations using the ab initio MP2/
6-31G* calculations. monomer, the modulus of the partial charges obtained with
The computed charges for the two cyclic carbonates the HF density are greater than the ones calculated with
from the HF and MP2 densities are compared in Table 4. the MP2 density for both carbonates. This trend is consis-
The 6-31G* EPS charges were used to evaluate the effect tent with the investigation of electron correlation effects on
of electron correlation on charge distributions. Charges atomic charge reported by Carpenter et al. [36]. It was
for the H’s in the methyl group of the PC monomer were found that MP2 correlation correction reduces charge sep-
averaged to obtain a common value in order to avoid prob- arations and such effect is particularly sizable for unsatu-
lems during the simulation due to the rotation of this rated molecules.
group. The most pronounced discrepancy observed The resultant dipole moments of the corresponding HF
between the two sets occurs on the charges of the carbonate and MP2 charge distributions on the cyclic carbonate mol-
group. As a trend, excepting for the methyl group in the PC ecules are also shown in Table 4. The computed charges
with the HF densities yield dipole moments ca. 22% above
of the gas phase experimental data (Table 1) for both cyclic
Table 3 carbonates. The inclusion of MP2 correlation correction
Optimized geometries for the EC and PC monomers using ab initio MP2/
leads the cyclic carbonates to become less polarized with
6-31G* calculations
dipole moments which are 10% and 13% greater than the
EC PC
gas phase experimental measures for EC and PC, respec-
Bond length (Å) tively. Enhancements of 10–20% in the experimental gas
C–O 1.201 1.202
phase dipole moments are desirable to compensate the lack
C–OS 1.365 1.364
C1–OS 1.437 of the polarization effects on condensed phase simulations
C2–OS 1.433 1.437 with fixed charge models [37]. In this context, the MP2
C2–C2 1.539 charges are found to be more appropriate for the cyclic car-
C1–C2 1.541 bonates simulations.
C1–C3 1.513
In order to evaluate the basis set dependence on the
Cn–H 1.092 1.093
CHELPG charges of the EC and PC monomers using basis
Bending angle () set larger than 6-31G*, the charge calculations were
OS–C–OS 110.6 110.6
C–OS–C2 110.5 110.2
performed with the 6-31++G**, 6-311G**, 6-311++G**
C2–C2–OS 104.2 and 6-311++G(3df,3pd) basis set at the MP2 level. The
C–OS–C1 110.9 results are shown in Table 5 along with the resultant dipole
C2–C1–OS 103.6 moments and the corresponding enhancements of the
C3–C2–OS 109.4 experimental measures. All basis set yielded dipole
C1–C2–OS 104.7
H–C2–OS 108.4 108.4
moments which are 7.6–16.9% higher than the experimen-
H–C1–C2 111.3 tal measure and therefore provided the desirable enlarge-
H–C3–C2 110.2 ment in the dipole moments for non-polarizable force
Dihedral angle () fields [37]. Extending the basis set from 6-31G* to
H–C2–OS–C 120.7 120.6 6-311G**, the dipole moment of the EC monomer was
H–C1–OS–C 117.7 not affected while a decrease of 7.2% was obtained for
H–C3–C1–OS 59.0 the PC monomer. The results for the 6-31++G** and
C3–C1–OS–C 122.1
6-311++G** basis set revealed that the inclusion of diffuse
L.B. Silva, L.C.G. Freitas / Journal of Molecular Structure: THEOCHEM 806 (2007) 23–34 27

Table 5
Charges for the EC and PC monomers calculated at the MP2 level using several basis set, the resultant dipole moments and corresponding deviations from
the experimental dipole moment, and the charges for the EC dimmer
6-31++G** 6-311G** 6-311++G** 6-311++G(3df,3pd)
EC monomer
O 0.558 0.522 0.556 0.533
C 0.927 0.878 0.930 0.843
OS 0.461 0.436 0.457 0.399
C2 0.236 0.212 0.230 0.176
H2 0.020 0.023 0.020 0.034
Dipole moment/D 5.69 5.32 5.57 5.46
Deviation/% 16.9 9.3 14.4 12.2
EC dimer
O 0.564
C 0.918
OS 0.436
C2 0.205
H2 0.027
PC monomer
O 0.509 0.476 0.502 0.454
C 0.946 0.878 0.939 0.866
OSa 0.499 0.460 0.490 0.441
C2 0.240 0.194 0.225 0.175
C1 0.345 0.349 0.342 0.313
OSb 0.547 0.507 0.542 0.525
C3 0.330 0.365 0.359 0.375
H1 0.030 0.032 0.029 0.041
H2 0.012 0.008 0.014 0.023
H3 0.100 0.113 0.110 0.118
Dipole moment/ D 5.73 5.32 5.61 5.56
Deviation/% 15.9 7.6 13.6 12.5
a
OS atom attached to the C2 atom.
b
OS atom attached to the C1 atom.

functions leads to the highest enhancements in the dipole


moments for both EC and PC monomers. The largest basis
set 6-311++G(3df,3pd) yielded dipole moments 12% high-
er than the experimental measure for EC and PC which is
only 1.4% and 2.7%, respectively, different from the one
calculated with the 6-31G* basis set. Table 5 also reports
the computed charge distribution for the most stable ab ini-
tio EC dimer configuration using the 6-31++G** basis set
at the MP2 level. As will be shown in the next section, this
configuration corresponds to an antiparallel dipole geome-
try. Attributing the atomic dimer charge distribution to a
monomer, a dipole moment of 5.75 D is obtained, which
is 18% higher than the experimental dipole for an isolated
monomer.

3.2. EC and PC dimers

Intermolecular geometry optimizations of the EC and


PC dimers were performed at the HF/6-31G*//HF/6-
31G* level. For both cyclic carbonates, all the located
Fig. 2. EC dimer configurations optimized with HF/6-31G* basis set. (a)
dimer configurations were found to optimize dipole–dipole
Lowest energy configuration. (b) Local minimum configuration.
interactions. Figs. 2 and 3 show the global minima and an
example of a local minimum configuration which were O–H2 and O–O intermolecular distances. Structures
obtained in the optimization calculations. Table 6 collects labeled (a) correspond to the global minimum configura-
the corresponding calculated interaction energies and the tion for both EC (Fig. 2) and PC (Fig. 3) dimers, which
28 L.B. Silva, L.C.G. Freitas / Journal of Molecular Structure: THEOCHEM 806 (2007) 23–34

arrangement. For PC, the local minima exhibit stacked


configurations with antiparallel dipoles, whose differences
are found in the methyl group position. During the optimi-
zations of the PC dimers, all the located configurations
exhibited antiparallel dipoles. The EC and PC dimer con-
figurations obtained in this work were also found by Wang
and Balbuena in their optimization using DFT calculations
[19]. However, the intermolecular O–H2 distances obtained
with the HF/6-31G* optimizations were consistently larger
than those computed with the DFT calculations at the
B3LYP level.
As shown in Table 6, the stacked antiparallel dimer of
EC is substantially more favorable than the head to tail
conformation. The local minimum of the PC dimer is only
within 0.4 kcal mol1 less bounded than the global mini-
mum conformation. This pattern of stability in the EC
and PC dimer configurations agrees with the results of
Wang and Balbuena [19]. The interaction energies obtained
from the ab initio calculations and from the force field with
both HF/6-31G* and MP2/6-31G* EPS charges are also
compared in Table 6. The greater polarization of the
monomers with HF/6-31G* charges leads the complexes
to a more favorable interaction by ca. 20% than those with
MP2/6-31G* charges. The ab initio HF/6-31G* interaction
energies are in very close agreement with the values obtained
from the force field with MP2/6-31G* EPS charges.
The effect of the basis set and inclusion of electronic cor-
relation on the intermolecular interaction energies was also
evaluated. A single point calculation at the MP2/6-31G*
level was performed on the optimized HF/6-31G* antipar-
allel EC dimer configuration. The resultant interaction
energy revealed that the inclusion of electronic correlation
Fig. 3. EC dimer configurations optimized with HF-6-31G* basis set. (a) at the MP2 level leads the dimer to become less bounded.
Lowest energy configuration. (b) Local minimum configuration. The interaction energy calculated with the 6-31++G**
basis set at the MP2 level on the same EC dimer configura-
presented a stacked conformation with antiparallel dipoles. tion was very similar to the one obtained with the 6-31G*
The local minimum configuration for the EC dimer has the basis set. The interaction energies of the EC dimer obtained
monomers positioned with the dipoles in a head to tail by Wang and Balbuena [19] with the DFT calculations

Table 6
Interaction energies (kcal mol1) for the EC and PC dimers using the force field with HF/6-31G* and MP2/6-31G* EPS charges and the ab initio
calculations, and the intermolecular O–H2 and O–O distances, R, obtained in the HF/6-31G* optimizations
EC PC
I II I II
Force field
HF/6-31G* charges 9.44 7.16 10.24 10.12
MP2/6-31G* charges 7.93 5.42 8.64 8.45
Ab initio
HF/6-31G* 7.98 5.75 8.73 8.56
MP2/6-31G*a 6.83
MP2/6-31++G**a 6.76
B3LYPb 5.93
B3PW91b 4.62
R (O–H2) / Å 2.631(2.466c) 2.964(2.764c) 2.758 2.834(2.466c)
R (O–O) / Å 3.976 6.756 3.894 4.043
a
The calculations included the basis set superposition error (BSSE) corrections using the counterpoise method of Boys and Bernardi [38].
b
From reference [19].
c
Value from reference [19] with B3LYP calculations.
L.B. Silva, L.C.G. Freitas / Journal of Molecular Structure: THEOCHEM 806 (2007) 23–34 29

Table 7 including the cutoff corrections. Heats of vaporization,


Heats of vaporization, DHvap, for the liquid EC and PC for the DHvap, are related to the E(liquid) value through the rela-
simulations performed with the HF and MP2 charges using the 6-31G*
and 6-311++G(3df,3d) basis set for the monomers and the MP2/6-
tionship given in Eq. (2). E(gas) is the average internal
31++G** charges for the dimer energy from a Monte Carlo simulation of a gas phase mol-
T (C) DHvap (kcal mol1)
ecule. RT replaces the PV term in the enthalpy for the ideal
gas, and is considered negligible for the liquid
Calculated Experimental
Ethylene carbonate DH vap ¼ EðgasÞ þ RT  DEðliquidÞ: ð2Þ
HF/6-31G* 150 16.26 ± 0.01 13.50 [39]
MP2/6-31G* 150 13.57 ± 0.01
The calculated heats of vaporization for the liquid cyclic
MP2/6-311++G(3df,3pd) 150 12.53 ± 0.01 carbonates from simulations with both HF and MP2 set
MP2/6-31++G** (dimer) 150 13.23 ± 0.01 charges are compared with the experimental measures in
Propylene carbonate Table 7. The average statistical uncertainties in the com-
HF/6-31G* 20 17.65 ± 0.03 15.60 [40] puted data are ±0.02 kcal mol1. Simulations for both liq-
MP2/6-31G* 20 15.57 ± 0.03 uids performed with HF/6-31G* charges yield heats of
HF/6-31G* 150 15.59 ± 0.07 13.20 [39] vaporization within 12–16% greater than the experimental
MP2/6-31G* 150 13.61 ± 0.06
value, which is substantially above the accepted 1–3% aver-
Table 8 age error for computed properties in liquid simulations
Computed and experimental densities for EC and PC from simulations [23]. In contrast, heats of vaporization computed with the
using the EPS charges MP2/6-31G* charges agree well with experiment deviating
T (C) Density (g cm3) by an average of 1–3% for both liquids. The force field with
Ethylene carbonate MP2 charges also provides a proper temperature depen-
HF/6-31G* 150 1.240 ± 0.001 (1.209a) dence of heats of vaporization for PC, since the simulations
MP2/6-31G* 150 1.192 ± 0.001 reproduced the experimental values at both 20 C and
HF/6-31G* 40 1.347 ± 0.001 (1.3208 [11]) 150 C.
MP2/6-31G* 40 1.307 ± 0.001
In order to investigate the effect of the electrostatic
MP2/6-311++G(3df,3pd) 40 1.300 ± 0.001
MP2/6-311++G**(dimer) 40 1.301 ± 0.001 potential on the thermodynamic results, additional simula-
tions for the EC liquid were performed using different basis
Propylene carbonate
HF/6-31G* 150 1.131 ± 0.001 (1.077 [41])
set. As shown in Table 7, the simulation carried out with
MP2/6-31G* 150 1.085 ± 0.001 the 6-311++G(3df,3pd) basis set at the MP2 level resulted
HF/6-31G* 20 1.241 ± 0.001 (1.2057 [42]) in a heat of vaporization 7.7% smaller than the experimen-
MP2/6-31G* 20 1.228 ± 0.001 tal value. The heat of vaporization for the simulation using
a
Interpolated from data given in Reference [16]. the CHELPG charges computed for the dimer with MP2/
6-31++G** basis set deviated only 2% from the experi-
Table 9 mental measure.
Computed heat capacities, expansibilities (a) and compressibilities (j) for
liquid EC and PC using the MP2/6-31G* EPS charges
Densities are estimated from the average volumes in the
Monte Carlo simulations and are shown in Table 8. The
ECa PCb
1 1
average statistical uncertainties for the computed data are
Cp(gas) (cal mol deg ) 15.64 20.50 ±0.001 g cm3. Density values computed with MP2/6-
Cp(liquid) (cal mol1 deg1)
Calculated 32.37 ± 1.20 35.44 ± 1.10
31G* charges at 40 C for EC and at 20 C for PC agree
Experimental 31.97 [42] 38.77 [43] well with the experimental measures deviating by an aver-
a (105 deg1) age error of 1%. Simulations performed with HF/6-31G*
Calculated 93.41 ± 8.36 64.56 ± 6.08 charges at these temperatures reproduced densities within
Experimental 73.70 [44] 95.80 [44] 1.9–5% higher than the experimental data. The densities
j (106 atm1)
Calculated 35.43 ± 2.71 24.12 ± 1.82
of the EC liquid at 40 C from the simulations with the
Experimental  59.00 [45] MP2/6-311++G(3df,3pd) charges and the MP2/6-
a
Heat capacities calculated at 50 C, a and j calculated at 40 C.
31++G** dimer charge distribution were very similar to
b
Properties calculated at 25 C. the density obtained with the MP2/6-31G* charges.
The remaining thermodynamic properties calculated
were smaller than the values computed at the MP2 level in using formulae from fluctuation theory [20] in the simula-
this work. tions with the MP2/6-31G* set charges are reported in
Table 9. The heat capacities at constant pressure, Cp(li-
3.3. Liquid thermodynamics quid), are calculated from the fluctuations in the total inter-
molecular energy, Cp(inter). For comparison with
The computed thermodynamic properties from the experimental data Cp(inter) needs to be augmented by a
liquid simulations are summarized in Tables 7–9. The total unimolecular term, Cp(gas), which includes the contribu-
potential energy of the liquid, E(liquid), consists of the tion from the translational, rotational and vibrational
intramolecular and intermolecular energy contributions motions of the monomers. Cp(gas) is approached by the
30 L.B. Silva, L.C.G. Freitas / Journal of Molecular Structure: THEOCHEM 806 (2007) 23–34

5 tions in the volume and enthalpy. It is known, however,


that j and a converge slowly and have large uncertainties,
while the energies and volume converge relatively rapid
[29]. The computed heat capacities for EC and PC are in
4 C1 - C1
reasonable agreement with the experimental data. The sta-
tistical uncertainties for both liquid cyclic carbonates are
±1.30 cal mol1 deg1. It should be noted that Cp(gas) less
C2 - C2 R make a substantial contribution to Cp(liquid) values. No
3
experimental data of j for EC was found in the literature
for comparison. The statistical uncertainties of a and j
g(r)

for EC are ±9.0 · 105 deg1 and ±3.0 · 106 atm1 and
C-C
2 for PC are ±5.0 · 105 deg1 and ±1.4 · 106 atm1.

3.4. Radial distribution functions


O -O
1 The structures of the liquids are reflected in radial distri-
bution functions (rdfs), which give the probability of occur-
rence of atoms of type y at a distance r from atoms of type
x, normalized for the bulk density of the liquid [20]. The
0 computed rdfs for the cyclic carbonates with the force field
2 4 6 8 10
using MP2/6-31G* charges are given in Figs. 4–7. Rdfs
r/Å
from the simulations with the MP2/6-311++G(3df,3pd)
Fig. 4. Computed O–O, C–C, C2–C2 and C1–C1 radial distribution charges and the MP2/6-31++G** dimer charge distribu-
functions for liquid EC (segmented line) and PC (solid line) from tion are not shown since they are almost identical to the
simulations at 150 C using MP2/6-31G* EPS charges. Successive curves
are offset 1.0 unit along the y-axis.
rdfs from the simulation using the MP2/6-31G* charges.
All data shown from here apply to simulations performed
at 150 C in order to provide a structural characterization
5 at the same temperature for both liquids.
Overall, the rdfs are characterized by broad peaks, typ-
ical of dipolar liquids [31]. There is a strong similarity in
O - H3 the rdfs of the two cyclic carbonates, though a little broad-
4
ening and lowering of the peaks for PC reflect a slightly
reduction in its structure comparing to EC. In addition,
the peak positions in all of the PC rdfs present a small shift
3 to larger r, which is consistent with the slightly lower den-
O - H1 sity of the PC liquid as shown in Table 8. Fig. 4 illustrates
the O–O, C–C, C2–C2 and C1–C1 rdfs for the two cyclic
g(r)

carbonates. The remaining OS–OS and C3–C3 rdfs are rel-


2
atively featureless and are not presented here. The Hn–Hn
rdfs provide few additional insights and are also not
shown. For both liquids the O–O rdfs exhibit a broad band
O - H2
1 that precedes a peak centered near 6.5 Å. The C–C rdfs
have two peaks with maxima at 4.3 Å and at 6.0 Å. Inte-
gration of the C–C and O–O rdfs out to 5.1 Å predicts
coordination numbers of approximately 3.0 and 2.0 for
0 EC and PC, respectively. Thus, an average of 3.0 and 2.0
2 4 6 8
short O–O contacts in the liquid EC and PC seems to be
r/Å
distinct interactions among the arrangements of the nearest
Fig. 5. Computed O–H2, O–H1, O–H3 radial distribution functions for neighbors of a given molecule. The C2–C2 and C1–C1 rdfs
liquid EC (segmented line) and PC (solid line) from simulations at 150 C exhibit broad bands that peaks near 5.4 Å for EC and
using MP2/6-31G* EPS charges. Successive curves are offset 1.5 units
along the y-axis.
5.7 Å for PC. Integration of the C–C and C2–C2 rdfs for
EC out to 7.1 Å and integration of the C–C, C2–C2 and
C1–C1 rdfs for PC out to 9.0 Å yields ca. 13.0 neighbors
heat capacity of a gas phase monomer less R, to remove the for both liquids.
gas phase PV contribution [29]. The isothermal compress- The most pronounced correlations in both liquids are
ibility, j, is computed from the fluctuations in the volume, those involving the carbonyl oxygen atom and the Cn and
and the coefficient of thermal expansion, a, from fluctua- Hn atoms. The O–Hn are the shortest contacts with peaks
L.B. Silva, L.C.G. Freitas / Journal of Molecular Structure: THEOCHEM 806 (2007) 23–34 31

5 configuration correspond quite closely with the maxima of


the strongest peaks in the rdfs in Figs. 4–6.
The O–Cn rdfs are exhibited in Fig. 6. For PC, a similar
O - C3 profile characterize the O–C2 and O–C3 rdfs, featuring
4
pronounced peaks with maxima at ca. 3.5 Å. In contrast,
the O–C1 rdf presents a distinct profile, which defines a
large plateau, indicating that the C1 atom is less accessible
3 to interactions with the carbonyl oxygen atoms than the C2
O - C1 and C3 atoms. For both cyclic carbonates, integrating the
peaks in the O–Cn rdfs up to their minima yields consis-
g(r)

tently a number of approximately 6.0 neighbors in each


2 case.
To evaluate the role of the electrostatic interactions on
the liquid structures, simulations for the cyclic carbonates
O - C2
at 150 C were performed with the partial charges set to
1
zero. The O–O, C–C and O–C2 rdfs in Fig. 7 illustrate
the differences in the structures of the liquids from the sim-
ulations with and without any electrostatic interactions.
0 Due to the similarity in the structural results between EC
2 4 6 8 10
and PC liquids, only the rdfs from simulations for EC
r/Å
are presented. The removal of the electrostatic interactions
Fig. 6. Computed O–C2, O–C1 and O–C3 radial distribution functions causes a remarkable reduction of the strongest peaks of the
for liquid EC (segmented line) and PC (solid line) from simulations at rdfs computed with the full potential; the peak at 6.5 Å in
150 C using MP2/6-31G* EPS charges. Successive curves are offset 1.5 the O–O rdf, the second peak at 6.0 Å in the C–C rdf and
unit along the y-axis.
the strong peak at 3.5 Å in the O–C2 disappeared almost
completely. Thus, the orientational preference for head to
tail alignment of the dipoles is lost for simulations with
4 non polar models. In contrast, there is a substantial
O - C2 increase of the band at 4.0 Å in the O–O rdfs and in the
first peak at 4.3 Å in the C–C rdfs, indicating a maximiza-
tion of favorable interactions at short distances. The close
3 approach of the O–O and C–C contacts is clearly unfavor-
C -C able to electrostatic interactions.
g(r)

2 3.5. Energy distributions

Energy pair distributions in Fig. 8 show the distribu-


O-O tions of interaction energies between monomers. The ordi-
1
4

3 Ethylene Carbonate
0
2 4 6 8 10 Propylene Carbonate
r/Å q=0
# molecules

2
Fig. 7. Computed O–O, C–C, O–C and O–C2 radial distribution
functions for liquid EC from simulations at 150 C with (a) MP2/6-
31G* EPS charges (segmented line); (b) charges set to zero (solid line).
Successive curves are offset 1.5 units along the y-axis. 1

0
at ca. 2.7 Å according to rdfs in Fig. 5. These Coulombically -10 -8 -6 -4 -2 0 2 4
-1
favorable interactions along with the large separations Interaction energy / kcal mol
found in the O–O contacts are consistent with a description
Fig. 8. Computed energy pair distributions for the liquids from simula-
of a head to tail alignment of the dipoles in the liquids. This tions at 150 C using the MP2/6-31G* EPS charges (ethylene carbonate
picture is reinforced by examining the configuration (b) of and propylene carbonate) and with the charges set to zero (ethylene
the EC dimer in Fig. 2. The interatomic distances in that carbonate).
32 L.B. Silva, L.C.G. Freitas / Journal of Molecular Structure: THEOCHEM 806 (2007) 23–34

nate gives the average number of molecules that are bound for antiparallel orientation of the neighboring dipoles at
to a monomer with the dimerization energy shown in the O–O distances shorter than ca. 3.0 Å. According to the rdfs
abscissa. The bimodal profiles of the distributions with in Fig. 4, only a small population of O–O contacts shorter
the full potential are typical of liquids associated by strong than 3 Å is found in both cyclic carbonates. For EC, the
dipolar interactions [31]. The shoulders at low energy antiparallel orientation of the dipoles is an indication that
reflect favorable interactions with near neighbors, while the liquid phase retains some order which is found in the
the large spikes in high energies represent the weak interac- solid state [45]. In the crystal, the EC molecules are
tions with the distant molecules in the bulk. For both liq- arranged in planes with the molecules in contiguous planes
uids, the distributions reveal that the most attractive pointing with their dipoles in antiparallel directions.
interactions are about -8.0 kcal mol1 and the strongest The tendency of antiparallel alignment of the neighbor-
repulsions are about +3.0 kcal mol1. The contrast with ing dipoles is lost at O–O distances greater than ca. 3.0 Å
the distribution for q = 0 clearly reveals the influence of and a preference for parallel dipoles is emerged through
the electrostatic interactions on the intermolecular energet- two broad peaks centered near 3.5 Å and 6.3 Å. The loca-
ics. The inclusion of electrostatic interactions causes a tion of these maxima are nearly coincident with the posi-
broadening of the distributions at both extremities, indicat- tion of the band and the peak in the O–O rdfs in Fig. 4.
ing that the optimization of some favorable interactions is This indicates that the special interactions at short ranges
compensated by the increase of some unfavorable and the interactions which give rises to the peak at 6.5 Å
interactions. in the O–O rdfs are predominantly characterized by paral-
The interaction energies of the antiparallel dimer config- lel dipoles. The tendency towards parallel orientation of
urations in Figs. 2 and 3 for both cyclic carbonates range the dipoles supports the description of a head to tail
from 7.9 to 8.6 kcal mol1 (Table 6). These energy val- arrangement of the neighboring molecules in the liquids,
ues can be compared to the low dimerization energy limits which was drawn from the analyses of the rdfs in Figs.
of the distributions in Fig. 8. The distributions reveal the 4–7. Moreover, the O–O distances of the antiparallel dimer
presence of a negligible population of dimers with such configurations in Figs. 2 and 3 are within 4.8–4.9 Å. Com-
low energies in condensed phase. Therefore, although anti- parison with the corresponding distances in the dipole–di-
parallel dimer configurations are strongly favored energet- pole correlations reveals that in the range of such
ically in the gas phase, they are only modestly populated in interatomic distances there is no evidence of antiparallel
the liquids. It is worth noting that the similarity in the ener- alignment of the dipoles in the liquids.
getic environment and structural properties of the liquid The present results are consistent with values of the
EC and PC is reflected in the very close heats of vaporiza- Kirkwood correlation factor (g) [46], obtained by some
tion of these liquids in Table 7. authors [13–15]. This factor is calculated from experimen-
tal dielectric measurements and gives a description of the
short range effects which tend to orient a molecule with
3.6. Dipole correlation functions respect to its neighbors. For systems in which intermolecu-
lar interactions orient neighboring dipoles in a parallel
Insights into the molecular orientation in the liquid EC fashion, g is greater than unity. In contrast, for an antipar-
and PC are given by the dipole correlation functions in allel configuration of dipoles, g is less than unity. Experi-
Fig. 9. The cosine of the angle between the dipole vectors mental results for EC at 40 C yielded g equal to 1.20
for all pairs of monomers was averaged as a function of [13]. For PC, it was found values of 1.18 [14] and 1.01
the O–O atomic distance. The results reveal a tendency [13] for g at 25 C. Based on IR and Raman spectroscopy
data, Fini et al. [15] have shown that in the presence of
dipolar forces, some dipole positions give a positive and
0,2 others a negative contribution to the g value, thus leading
this factor to a value close to unity.
0,0
4. Conclusions
cos θ

-0,2
Ethylene Carbonate Monte Carlo simulations have been used to investigate
-0,4
Propylene Carbonate molecular association in the pure liquid EC and PC. For
this purpose, a force field for simulations of liquid cyclic
-0,6 carbonates has been developed. An approach of combining
standard OPLS Lennard–Jones parameters and charges
0 2 4 6 8 10 from fitting to the EPS using the CHELPG procedure
O-O distance / Å has been used to describe the intermolecular interactions.
Fig. 9. Computed dipole–dipole correlations for liquid ethylene carbonate It was found that the inclusion of electron correlation in
and propylene carbonate from simulations at 150 C using the MP2/6- the charge calculations is required to achieve a proper
31G* EPS charges. reproduction of the experimental pure liquid properties.
L.B. Silva, L.C.G. Freitas / Journal of Molecular Structure: THEOCHEM 806 (2007) 23–34 33

The resultant force fields with MP2/6-31G* charges repro- static interactions are fundamental in structuring the
duced densities and heats of vaporization within 1–3% of liquid EC and PC.
the experimental data for both cyclic carbonates. Addition-
ally, a good description of interaction energies for isolated Acknowledgements
dimers from ab initio HF/6-31G* calculations was
obtained. Simulations performed with HF/6-31G* charges This work was partially supported by Fundação de
yielded heats of vaporization within 12–16% higher than Amparo à Pesquisa do Estado de São Paulo-FAPESP under
the experimental value, which is substantially above the Proc. 96/3936-0. L.B.S gratefully acknowledges FAPESP
accepted 1–3% average error for computed properties in for a graduate research fellowship. The authors are grateful
liquid simulations [23]. to Dr. João M. M. Cordeiro for fruitful discussions.
Besides the inclusion of electronic correlation on the
charge set calculations, two additional issues concerning
the electrostatic potential were addressed in this work. References
First, it was evaluated the basis set dependence on the
[1] W.H. Lee, Cyclic carbonates, in: J.J. Logowski (Ed.), The Chemistry
CHELPG charge calculations using the 6-31++G**, of Nonaqueous Solvents, vol. 4, Academic Press, New York, 1976,
6-311G**, 6-311++G** and 6-311++G(3df,3pd) basis Chapter 6.
set at the MP2 level. It was found that the charges calculat- [2] A. D’aprano, Gazz. Chim. Ital. 104 (1974) 91.
ed with basis set larger than 6-31G* yield enhancements in [3] (a) M.L. Lakhanpal, S.C. Ahuja, H.G. Mandal, Indian J. Chem. 16
(1978) 928;
the dipole moments which are considered adequate to pro-
(b) A.K. Srivastava, R.A. Samant, S.D. Patankar, J. Chem. Eng.
vide the required polarization in force fields with fixed Data 41 (1996) 431.
charges [37]. However, the simulation performed with the [4] (a) C.V. Krishnan, H.L. Friedman, J. Phys. Chem. 73 (1969) 1572;
MP2/6-311++G(3df,3pd) charge set produced a heat of (b) M. Salomon, J. Phys. Chem. 74 (1970) 2519;
vaporization 7.7% smaller than the experimental measure. (c) H.L. Yeager, J.D. Fedyk, R.J. Parker, J. Phys. Chem. 77 (1973)
2407.
The second issue evaluated the parametrization of the elec-
[5] M. Wakihara, O. Yamamoto, Lithium Ion Batteries – Fundamentals
trostatic potential using a charge set calculated for a dimer. and Performance, Kodansha Ltd., Tokyo, 1998.
The atomic charge distribution obtained for the most [6] US Pat., 5 449 474, 1995.
stable EC dimer configuration (antiparallel) using the [7] US Pat., 6 169 061, 2001.
MP2/6-31++G** basis set yielded for the monomer a [8] US Pat., 5 993 787, 1999.
[9] K. Beyer, W. Bergfeld, W. Berndt, W. Carlton, D. Hoffmann, A.
dipole moment 18% higher than the experimental measure.
Schroeter, R. Srank, J. Am. Coll. Toxicol. 6 (1987) 23.
A good reproduction of the experimental density and heat [10] (a) R. Papciak, V. Mallory, Acute Toxic. Data 1 (1990) 15;
of vaporization (1.5% and 2% of deviation, respectively) (b) C. Ursin, C. Hansen, J. Van Dyk, P. Jensen, I. Christensen, J.
was reached with the dimer charge distribution. Thus, the Ebbehoej, Am. Ind. Hyg. Assoc. 56 (1995) 651.
charge set computed with a basis set larger than 6-31G* [11] R. Kempa, W.H. Lee, J. Chem. Soc. (1958) 1936.
[12] R.P. Seward, E.C. Vieira, J. Phys. Chem. 62 (1958) 127.
did not provided a good reproduction of the experimental
[13] R. Payne, I.E. Theodorou, J. Phys. Chem. 76 (1972) 2892.
heat of vaporization. On the other hand, the heat of vapor- [14] L. Simeral, R.L. Amey, J. Phys. Chem. 74 (1970) 1443.
ization calculated with the dimer charge distribution is [15] G. Fini, P. Mirone, B. Fortunato, J. Chem. Soc. Faraday Trans. 2 69
comparable to the one computed with the monomer 6- (1973) 1243.
31G* basis set. These results support that the 6-31G* [16] W. Schindler, T.W. Zerda, J. Jonas, J. Chem. Phys. 81 (1984)
4306.
EPS charges provide the proper amount of polarization
[17] B. Klassen, R. Aroca, M. Nazri, G.A. Nazri, J. Phys. Chem. B 102
that is required in the fixed charge force fields for the cyclic (1998) 4795.
carbonates in this work. [18] T. Li, P.B. Balbuena, J. Electrochem. Soc. 146 (1999) 3613.
The simulations provided a thorough characterization [19] Y. Wang, P.B. Balbuena, J. Phys. Chem. A 105 (2001) 9972.
of the molecular structures of the pure liquid EC and PC. [20] M.P. Allen, D.J. Tildesley, Computer Simulations of Liquids,
Clarendon Press, Oxford, 1987.
A general similarity characterizes the structural features
[21] W.F. van Gunsteren, H.J.C. Berendsen, Angew. Chem. Int. Ed. Engl.
of both liquids. For both liquids, the results revealed a 29 (1990) 992.
coordination sphere of about 13 monomers around each [22] W.D. Cornell, P. Cieplak, C.I. Bayly, I.R. Gould, K.M. Merz Jr.,
monomer. It was found that the electrostatic interactions D.M. Ferguson, D.C. Spellmeyer, T. Fox, J.W. Caldwell, P.A.
lead the neighboring molecules to a preferential head to tail Kollman, J. Am. Chem. Soc. 117 (1995) 5179.
[23] W.L. Jorgensen, D.S. Maxwell, J. Tirado-Rives, J. Am. Chem. Soc.
alignment of the dipoles. However, the results also suggest
118 (1996) 11225.
the presence of a small population of nearest neighbors [24] M.M. Francl, L.E. Chirlian, Rev. Comput. Chem. 14 (2000) 1.
with antiparallel dipole orientation to the central mono- [25] C.M. Breneman, K.B. Wiberg, J. Comp. Chem. 11 (1990)
mer. This feature is consistent with the dipole alignment 361.
found in EC in the solid state. Although the antiparallel [26] (a) H.A. Carlson, T.B. Nguyen, M. Orozco, W.L. Jorgensen, J.
Comp. Chem. 14 (1993) 1240;
orientation of the dipoles is not predominant in the liquid
(b) N.A. McDonald, W.L. Jorgensen, J. Phys. Chem. B 102 (1998)
EC and PC, this arrangement corresponds to the most 8049;
energetically favorable in the gas phase. Simulations per- (c) V.E. Barlette, F.L.L. Garbujo, L.C.G. Freitas, Mol. Eng. 7 (1997)
formed with non polar molecules showed that the electro- 439;
34 L.B. Silva, L.C.G. Freitas / Journal of Molecular Structure: THEOCHEM 806 (2007) 23–34

(d) P. Belletato, L.C.G. Freitas, E.P.G. Areas, P.S. Santos, Phys. [34] M. Neumann, J. Chem. Phys. 82 (1985) 5663.
Chem. Chem. Phys. 1 (1999) 4769. [35] L.C.G. Freitas, Programa DIADORIM v. 3.0, Universidade Federal
[27] C.L. Angell, Trans. Faraday Soc. 52 (1956) 1178. de São Carlos.
[28] H. Finegold, J. Phys. Chem. 72 (1968) 3244. [36] J.E. Carpenter, M.P. McGrath, W.J. Hehre, J. Am. Chem. Soc. 111
[29] W.L. Jorgensen, J.D. Madura, C. Swenson, J. Am. Chem. Soc. 106 (1989) 6154.
(1984) 6638. [37] K.M. Merz Jr., J. Comp. Chem. 13 (1992) 749.
[30] M.J. Frisch, G.W. Trucks, H.B. Schlegel, P.M.W. Gill, B.G. Johnson, [38] S.F. Boys, F. Bernardi, Mol. Phys. 112 (1970) 553.
M.A. Robb, J.R. Cheeseman, T. Keith, T.G.A. Petersson, J.A. [39] C.S. Hong, R. Wakslak, H. Finston, V. Fried, J. Chem. Eng. Data 27
Montgomery, K. Raghavachari, M.A. Al-Laham, V.G. Zakrzewski, (1982) 146.
J.V. Ortiz, J.B. Foresman, C.Y. Peng, P.A. Ayala, M.W. Wong, J.L. [40] E.E. Walker, J. Appl. Chem. 2 (1952) 470.
Andres, E.S. Replogle, R. Gomperts, R.L. Martin, D.J. Fox, J.S. [41] F. Murrieta-Guevara, A.T. Rodrigues, J. Chem. Eng. Data 29 (1984)
Binkley, D.J. Defrees, J. Baker, J.P. Stewart, M. Head-Gordon, C. 204.
Gonzalez, J.A. Pople, Gaussian 94 (Revision A.1), Gaussian, Inc., [42] J.W. Peppel, Ind. Eng. Chem. 50 (1958) 767.
Pittsburgh, PA, 1995. [43] E. Wilhelm, E. Jimenez, G. Roux-Desgranges, J.P. Grolier, J. Sol.
[31] J.M. Briggs, T. Matsui, W.L. Jorgensen, J. Comp. Chem. 11 (1990) 958. Chem. 20 (1991) 17.
[32] P.S. Charifson, R.G. Hiskey, L.G. Pedersen, J. Comp. Chem. 11 [44] Y. Marcus, Ion Solvation, John Wiley & Sons Ltd., New York,
(1990) 1181. 1985.
[33] W. Damm, A. Frontera, J. Tirado-Rives, W.L. Jorgensen, J. Comp. [45] C.J. Brown, Acta Cryst. 7 (1954) 92.
Chem. 18 (1997) 1997. [46] J.G. Kirkwood, J. Chem. Phys. 7 (1939) 911.

You might also like