Article

pubs.acs.org/JPCB

The General AMBER Force Field (GAFF) Can Accurately Predict

Thermodynamic and Transport Properties of Many Ionic Liquids
K. G. Sprenger, Vance W. Jaeger, and Jim Pfaendtner*
Department of Chemical Engineering, University of Washington, Seattle, Washington 98105, United States

ABSTRACT: We have applied molecular dynamics to

calculate thermodynamic and transport properties of a set of
19 room-temperature ionic liquids. Since accurately simulating
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

the thermophysical properties of solvents strongly depends

upon the force field of choice, we tested the accuracy of the
general AMBER force field, without refinement, for the case of
ionic liquids. Electrostatic point charges were developed using
Downloaded via KANAZAWA UNIV on May 14, 2024 at 06:19:44 (UTC).

ab initio calculations and a charge scaling factor of 0.8 to more

accurately predict dynamic properties. The density, heat
capacity, molar enthalpy of vaporization, self-diffusivity, and
shear viscosity of the ionic liquids were computed and
compared to experimentally available data, and good agreement across a wide range of cation and anion types was observed.
Results show that, for a wide range of ionic liquids, the general AMBER force field, with no tuning of parameters, can reproduce a
variety of thermodynamic and transport properties with similar accuracy to that of other published, often IL-specific, force fields.

■ INTRODUCTION
Ionic liquids (ILs) are a class of solvents consisting of organic
values. However, most of these force fields have been
developed for only a small subset of ILs. Inspired by prior
salts that are liquid below 100 °C. Many ILs possess successes, we sought to perform a comprehensive evaluation of
thermodynamic and electrochemical properties that make a common force field (GAFF) in order to determine how well
them optimal candidates for a large range of applications.1,2 standard molecular simulation force fields perform for a new
For example, ILs tend to be thermally stable, noncombustible, class of solvents (i.e., ILs) for which the parameters were not
and nonvolatile, making them more environmentally favorable optimized. Though GAFF has been widely used in IL
compared to common organic solvents and thus increasingly simulations for years,7−10 it is precisely this widespread usage
used in green processes such as the separation of organic that motivated our investigation, because a systematic study of
solvents from ILs by supercritical CO2. Cellulosic and algal the accuracy of GAFF across a wide spectrum of ILs has not, to
biomass can be solvated and preprocessed by ILs for biofuel our knowledge, yet been performed. We note that, while only
production. ILs have also been used to selectively absorb gases MD simulations were performed in this study, the force fields
and later release them in temperature dependent processes. generated using this method are not specific to MD but can
Because of the favorable electrochemical window, ILs have also be used in MC simulations.
been used as a replacement for water in certain batteries. Classical force fields consist of two sets of termsbonded
Understanding these interactions at the molecular level can and nonbonded. Bonded parameters consist of bonds, angles,
help researchers optimize the species of IL used for a given and dihedrals and are used to calculate energy potentials that
device or process, and it can also guide the discovery and include force or spring constants; nonbonded parameters
synthesis of new ILs. include Lennard-Jones parameters and electrostatic point
Molecular simulations, in particular Monte Carlo (MC) or charges and are used to calculate 12-6 and Coulomb potentials,
molecular dynamics (MD) simulations, provide a convenient respectively. The sum of the bonded and nonbonded energies
route to predict thermophysical properties of ILs. The accuracy gives the overall potential energy of the system. For the ILs we
of these predictions relies on the accuracy of the force field
will be testing in this paper, GAFF contains all of the needed
used to describe the intra- and intermolecular (bonded and
parameters for calculation of the bonded and Lennard-Jones
nonbonded) interactions of the solvent molecules. Several force
fields have been optimized to reproduce experimental data for terms. This leaves the electrostatic point charges to be
small organic molecules, both synthetic and biological. Among determined. Since the many interesting properties of ILs arise
these force fields are OPLS,3 CHARMM,4 UFF,5 AMBER, and as a consequence of strong electrostatic interactions, which
its companion the general AMBER force field6 (GAFF). Since drive molecular scale fluctuations, accurately modeling the
many ionic liquids are similar in structure to some of the
molecules used to parametrize these popular force fields, they Received: January 22, 2015
have commonly been used as templates for the creation of IL Revised: April 7, 2015
force fields, both in functional form and actual parameter Published: April 8, 2015

© 2015 American Chemical Society 5882 DOI: 10.1021/acs.jpcb.5b00689

J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

electrostatic landscape of each cation and anion is of primary force field for ILs, the resulting densities, molar enthalpies of
importance. vaporization, and self-diffusion coefficients are compared to
The transport properties of ILs are not described well by results from other published force fields, including a modified
classical force fields when full +e/−e charges are assigned to the force field based on the OPLS-AA/AMBER framework and the
cations and anions.11 The electronic point charges are CHARMM 22 force field.
commonly derived from quantum calculations in a vacuum. It Representations of the IL cations and anions are shown in
has been shown, both computationally and experimentally, that Figure 1. As was noted earlier, many recent studies have
this description of a cation or anion in a vacuum with a full
charge does not accurately describe the liquid system.11−13 In a
liquid, neighboring molecules can screen charge, and the
electron density of a molecule can be affected by nearby
molecules of the opposite charge. Because of this, previous
researchers have suggested that classical force fields should
assume that the charge of the cation and anion is actually some
fraction of +e/−e.14 Maginn and co-workers have suggested and
compared several methods to scale the ions’ charges, and they
have found that scaling is necessary to predict the shear
viscosity or self-diffusivity of an IL using MD.11 To overcome
the shortcomings of calculating the electronic structure in
vacuo, ab initio MD can be used with crystal structures of
solidified ILs in order to scale the charges and very accurately
reproduce the properties of liquid phase ILs. A more naive ̈
approach is to simply scale the charges by multiplying by a
factor of 0.7−0.9,8,15−18 as this is near what is typically found in
the ab initio methods. While this approach sacrifices some
accuracy, it is far more amenable to large-scale deployment of
molecular simulation for predicting IL properties. Figure 1. Ionic liquid ions. R = ethyl (Et or E), butyl (B), pentyl (P),
Ultimately, if IL force fields can be constructed quickly and hexyl (H), and octyl (O).
easily without refinement, screening for the optimal ILs for
certain applications could be accomplished with greater
efficiency, and complex simulations involving ILs could be focused on combining water and ILs to look at protein or
performed with greater ease. For example, simulating sugars, enzyme structure and/or stability; thus, additional simulations
proteins, and enzymes in ILs and IL/water mixtures is were performed on pure boxes of water using the TIP3P and
becoming an increasingly relevant area of research, supported SPC/E water models. This was done for two reasons: first, we
by many recently published papers by us and others.19−25 An surmised that the accuracy of TIP3P and SPC/E, given their
even more interesting possibility is that we will be able to widespread use, would be a useful standard with which to
predict the thermophysical properties of yet undiscovered ILs compare the accuracy of GAFF in predicting IL properties, and
or of complex interactions between solutes and ILs. These second, we wished to provide this comparison as a starting
latter simulations could help to screen a large library of ILs and point for future work that could assess the ability of GAFF to
direct chemists in their efforts to discover useful cation−anion predict IL/water mixture properties.
pairs. Other protocols have been proposed for predicting
electrostatic point charges and properties of ILs.11 These
protocols require more rigorous calculations to develop the
■ METHODS
GAFF. To represent the ILs, GAFF6 was the chosen force
force fields. The approach we adopt herein, even though it field, the functional form of which is shown in eq 1. Despite the
could result in a modest loss of accuracy, can be implemented reputation and widespread use of AMBER as a force field to
by a researcher with a basic understanding of standard MD study proteins and nucleic acids, AMBER lacks the parameters
using widely available, and free, programs. Specifically, in this necessary to study many organic molecules. Thus, GAFF was
paper, we propose a simple and fast protocol for accurately designed to improve upon this aspect of AMBER and permit
predicting the thermophysical properties of ILs using GAFF. the detailed study of organic molecules such as those present in
An extensive series of molecular dynamics simulations has ILs in conjunction with studies of biomolecules. In a similar
been conducted featuring 19 combinations of ionic liquids fashion to AMBER, bond stretching and angle bending terms in
under room conditions (298.15 K and 1 bar), including GAFF are evaluated with a simple harmonic functional form
alkylammonium [RAM], 1-alkyl-3-methylimidazolium and the dihedral term is evaluated with a cosine function.
[RMIM], 1-alkylpyridinium [RPyr], and 1-alkyl-3-methylpyr- Coulombic and 12-6 Lennard-Jones terms are used to calculate
idinium [RMPyr] cations and formate [HCO2], acetate [Ac], the electrostatic and van der Waals nonbonded interactions,
alkylsulfate [RSO 4 ], bis[(trifluoromethyl)sulfonyl]imide respectively, considering all pairs of atoms (i and j) in different
[Tf2N], trifluoromethanesulfonate [Otf], and chloride [Cl] molecules or in the same molecule but separated by a minimum
anions. Calculations of IL density, heat capacity, molar enthalpy of three bonds. One−four interactions, or nonbonded
of vaporization, self-diffusivity, and shear viscosity have been interactions separated by three bonds, are reduced by a scale
computed for the ILs and compared to experimental data. Self- factor, in order to achieve identical parameters for intra- and
diffusivity calculations have also been performed at high intermolecular interactions. Parameters include the force
temperatures to see the influence on the accuracy of GAFF’s constants kr, kθ, and Vn, the equilibrium structural parameters
predictions. To further evaluate the effectiveness of the GAFF req and θeq, the partial atomic charges qi and qj, and the
5883 DOI: 10.1021/acs.jpcb.5b00689
J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

Figure 2. Simulated vs experimental IL properties: (A) density (g/cm3), squares = TIP3P water, triangles = SPC/E water; (B) heat capacity (J/mol-
K); (C) molar enthalpy of vaporization (kJ/mol); (D) shear viscosity (Pa·s). The x and y axes show experimental and simulated results, respectively.
Brackets show results for TIP3P and SPC/E water with experimental results listed first.

Lennard-Jones well depth, ε, and radii, Aij and Bij; n is input files for each ion. Atomic charges were scaled by
multiplicity; and λ is a phase angle for torsional angle multiplying each by a factor of 0.8. These methods are very
parameters. similar to some of those employed by Maginn and co-
workers.8,31 This work aims to determine how universal this
Vtot = ∑ kr(r − req)2 + ∑ kθ(θ − θeq)2 approach is for ILs.
bonds angles Ionic Liquid Simulations. Cubic boxes of 125 nm3 in size
Vn and containing 200−1000 ion pairs were generated using
+ ∑ [1 + cos(nϕ − λ)] Packmol32 at close to experimental liquid densities. IL boxes
dihedrals 2
were simulated with periodic boundary conditions and used the
⎡ A Bij qiqj ⎤ GROMACS33,34 4.6.5 package and GAFF force field. Under 1.4
+ ∑⎢ ij
− + ⎥
nm, electrostatics were calculated explicitly; above 1.4 nm,
i<j
⎢⎣ R ij12 R ij 6 εR ij ⎥⎦ (1) particle-mesh Ewald (PME) summations were used to calculate
Quantum Mechanics Calculations. Individual quantum long-range electrostatic interactions. van der Waals interactions
mechanics (QM) calculations were performed on all ions, i.e., were shifted to 0 at 1.2 nm, in order to eliminate artifacts. All
[RAM] (R = ethyl (E), butyl (B), and pentyl (P)), [RMIM] (R bonds between hydrogen and heavy atoms were frozen at
= E, B, hexyl (H), and octyl (O)), [RPyr] (R = P and H), and equilibrium values.
[RMPyr] (R = E) cations and [HCO2], [Ac], [RSO4] (R = IL ions were placed randomly in the box to carry out the
methyl (Me) and ethyl (Et)), [Tf2N], [Otf], and [Cl] anions. liquid-phase simulations. Following energy minimization, each
Geometry optimization and energy calculations were per- of the 19 ILs was simulated in the NVT ensemble for 10 ns, at
formed at the Hartree−Fock (HF) level of theory using the 6- 500 K, to achieve thorough mixing of the ions and to eliminate
31G(d)//6-31G(d) basis set and the electronic structure any artifacts in the initial configurations, thus establishing
program Gaussian.26 All force constants and equilibrium and coordinates independent of the Packmol input. Constant
Lennard-Jones parameters were taken directly from the GAFF temperature was maintained with a global stochastic thermo-
force field. Antechamber27 was used to assign electrostatic point stat.35
charges by the RESP method.28 After the determination of From these 10 ns NVT simulations, unique starting
partial charges, tleap29 and ACPYPE30 were used to generate structures for subsequent simulations were taken from points
5884 DOI: 10.1021/acs.jpcb.5b00689
J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

along the simulation trajectories at 2, 4, 6, 8, and 10 ns. For Table 1. Experimental and Simulated IL Densities (g/cm3)a
each IL, the five unique starting structures then underwent 5 ns
IL experiment simulation error (%)
of pressure equilibration in the NPT ensemble, under room
conditions (298.15 K and 1 bar), using the Berendsen [BAM][HCO2] 0.96840 0.992 2.4
barostat.36 The boxes then underwent a further 10 ns of [BMIM][Ac] 1.05341−43 1.010 −4.1
pressure equilibration in the NPT ensemble at the same [BMIM][MeSO4] 1.20944−48 1.189 −1.6
conditions, using a time step of 1 fs and the Parrinello− [BMIM][Tf2N] 1.43649−53 1.416 −1.3
Rahman barostat37 to accurately capture the correct ensemble [EAM][HCO2] 1.03940 1.085 4.4
fluctuations for calculations of the densities and heat capacities. [EMIM][Ac] 1.09954−56 1.060 −3.6
The reported density and heat capacity values represent the [EMIM][EtSO4] 1.23857−60 1.216 −1.7
values averaged across the five simulations. [EMIM][Otf] 1.38646,54,61,62 1.367 −1.4
One of the five simulation boxes for each IL was used to [EMIM][Tf2N] 1.51951,59,63 1.501 −1.2
initiate triplicate nonequilibrium MD NPT simulations, 1 ns in [EMPyr][EtSO4] 1.22164−66 1.171 −4.1
length, wherein a cosine acceleration factor, or shear rate, of [EMPyr][Tf2N] 1.48967 1.446 −2.9
0.03 nm/ps2 was applied to each particle. This periodic [HMIM][Cl] 1.03668−71 1.007 −2.8
perturbation method was used to calculate the shear viscosity of [HMIM][Tf2N] 1.36755,72−75 1.351 −1.2
the ILs.38 NPT simulations of length 10 or 20 ns depending on [HPyr][Tf2N] 1.39176,77 1.343 −3.4
the IL (see the Results and Discussion) were run to calculate [OMIM][Cl] 1.01045,68−70,78,79 0.974 −3.5
the self-diffusivity of the ions of four ILs. [OMIM][Otf] 1.15780,81 1.163 0.5
To calculate molar enthalpies of vaporization, gas-phase [OMIM][Tf2N] 1.29273,82,83 1.298 0.5
simulations were conducted in the NVT ensemble on a single [PAM][HCO2] 0.95040 0.962 1.3
cation/anion pair in a small box without periodic boundary [PPyr][Tf2N] 1.42184 1.373 −3.4
conditions, at 298.15 K. The gas-phase simulations were run for
100 ns and the resulting trajectories divided into thirds to mean absolute error 2.4
obtain three “trials” of gas-phase enthalpies. Three trials of max absolute error 4.4
liquid-phase enthalpies were calculated from the NPT water model experimental simulation error (%)
production runs described earlier. TIP3P 0.99739 0.973 (0.98039) −2.4

■
SPC/E 0.99739 0.988 (0.99439) −0.9
a
RESULTS AND DISCUSSION Values in parentheses represent simulated densities from Vega and
An overall picture of the results for IL density, heat capacity, Abascal.39
molar enthalpy of vaporization, and shear viscosity is shown via
parity plots in Figure 2; results will later be discussed in detail. the parity line represent two of the three [OMIM]-based ILs.
Density. Using the equilibrium volume of the simulation Notably, these two ILs have the largest experimental standard
cell, we computed the density of the ILs using the standard deviations of the 19 simulated ILs (see Figure 2A); however,
definition. Computed and experimental densities of the 19 ILs they also represent the ILs with the smallest deviation in
and TIP3P and SPC/E water are reported in Table 1. When simulated density from the reported experimental values (0.5%
multiple experimental data points were available, an average of for both). The largest deviation from experimental values
the values was taken to get the single reported experimental occurs for [EAM][HCO2] (4.4%), followed closely by
value in Table 1. [EMPyr][EtSO4] and [BMIM][Ac] (4.1% for both ILs).
A parity plot showing overall trends in the data across the full The trend of underestimating IL density is not unique to the
spectrum of simulated densities is shown in Figure 2A. GAFF force field. Simulated densities for 33 [RMIM]-based
Uncertainty in the experimental data, when possible to assess, ILs, computed using a refined OPLS-AA force field,85 are
is drawn in the plot with error bars representing the standard compared against results of this work and to experimental
deviation of all experimental or simulation data. We note that results in the parity plot shown in Figure 3.
some data points are based upon a single experimental source The majority of the points, as with GAFF, lie below the
and thus no error bars are available. Simulated error bars are parity line. In contrast to GAFF, however, increasing the chain
too small to be seen due to the good reproducibility of the length of the cation generally leads to larger deviations from
simulations. In general, GAFF appears to slightly underestimate
the densities of most of the 19 ILs; however, the same is true of
TIP3P water with excellent agreement seen between SPC/E
and experiment.
For the five points that lie above the parity line in Figure 2A,
i.e., overestimate the density, three of the points correspond to
the three ILs with [RAM] cations. In contrast to most other
experimental densities, we note that only a single source could
be found in the literature for the experimental densities of the
[RAM]-based ILs. Interestingly, the simulated density error of
these specific ILs appears to decrease with increasing chain
length, from 4.4% (C2) to 2.4% (C4) to 1.3% (C5), and the
same trend is also mildly seen for the [Tf2N]-based ILs, from
1.2% (C2) to 0.5% (C8). This is not surprising given the Figure 3. Simulated vs experimental IL densities for this work (black
extensive use of hydrocarbon-containing compounds in the squares) and the work of Sambasivarao and Acevedo85 using a refined
development of GAFF. The other two points that lie just above IL-specific OPLS-AA force field (gray circles).

5885 DOI: 10.1021/acs.jpcb.5b00689

J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

experiment; since the OPLS-AA force field was parametrized 14 ILs. Incidentally, it can be pointed out that the four most
for types of molecules such as carbohydrates, rather than underestimated heat capacities, all in the upper range of Figure
hydrocarbons, this makes sense.86 For estimating densities, 2B, represent ILs with fluorinated anions, specifically [Tf2N].
these results clearly show that GAFF, with no tuned This underestimation appears to lessen quite drastically with
parameters, provides results of comparable accuracy to an IL- increasing chain length, from 10.3% (C2) to 1.4% (C8). As was
specific force field. noted earlier in regard to density, this result is not unusual in
Heat Capacity. Constant pressure heat capacities for the light of the fact that GAFF was parametrized with a significant
ILs were calculated using the classical statistical mechanical fraction of hydrocarbon-containing compounds and increasing
definition the alkyl chain length has the net result of reducing the impact
⎛ ⟨(∂H )2 ⟩ ⎞ ⎛ ⟨H2⟩ − ⟨H ⟩2 ⎞ of inaccuracies introduced from the fluorinated components.
Cp = ⎜ ⎟ = ⎜ ⎟ In reality, the quantity that we calculated in this work
2
⎝ nkBT ⎠ P ⎝ nkBT 2 ⎠P (2) represents only the excess, or residual, portion of the heat
capacity, which accounts for the intermolecular interactions in
where H and T are the enthalpy and temperature of the system, the condensed phase.100 Conversely, the other portion of the
respectively, kB is Boltzmann’s constant, and n is the total heat capacity is an ideal gas contribution that takes into account
number of particles. intramolecular interactions.100 These two contributions to the
Simulated heat capacities for the 19 ILs and corresponding heat capacity are typically represented as shown in eq 3.
experimental values (averaged across the specified experimental
sources) for 14 of the 19 ILs are reported in Table 2; results for ⟨H ⟩ = ⟨H ig⟩ + ⟨H res⟩
= ⟨Φint + K + NkBT ⟩ + ⟨Φnb + PV − NkBT ⟩ (3)
Table 2. Experimental and Simulated IL Heat Capacities (J/
mol-K)a Typically, ideal gas portions of heat capacities can be found
through experiments, but as this type of experimental data is
IL experiment simulation error (%)
currently not available for ILs, the values are calculated from a
[BAM][HCO2] N/A 258 N/A frequency analysis of optimized cation and anion structures
[BMIM][Ac] 38387 391 2.1 from ab initio MD simulations.101 Since GAFF appears to give
[BMIM][MeSO4] 41646 416 0.0 good agreement with experimental heat capacities across a
[BMIM][Tf2N] 56653,88,89 532 −6.1 broad range of ILs using the quick screening method we are
[EAM][HCO2] N/A 185 N/A proposing here, we conclude that the residual heat capacity is
[EMIM][Ac] 32254 320 −0.7 the dominant term and thus deemed it unnecessary to compute
[EMIM][EtSO4] 38746,90,91 382 −1.4 the ideal gas portions of the heat capacities.
[EMIM][Otf] 36354,92 350 −3.6 Molar Enthalpy of Vaporization. Experiments taking
[EMIM][Tf2N] 50963,93,94 456 −10.3 place over the past decade or so have served to refute the idea
[EMPyr][EtSO4] 38995 409 5.2 of ILs as nonvolatile liquids, and on the contrary have proved
[EMPyr][Tf2N] 49196 480 −2.3 that to some extent ILs can be distilled.102 Because of this
[HMIM][Cl] 38197 369 −3.3 revelation, there has been a push to understand the nature of
[HMIM][Tf2N] 62891,95,98,99 591 −5.9 ILs in the vapor phase, and thus to be able to accurately model
[HPyr][Tf2N] 61295 586 −4.2 and even predict molar enthalpies of vaporization of ILs is of
[OMIM][Cl] 44597 447 0.4 growing importance.
[OMIM][Otf] N/A 572 N/A
Previous studies have shown that ILs most likely transition
[OMIM][Tf2N] 67288,95 663 −1.4
into the gas phase via single, neutral ion pairs (i.e., one cation
[PAM][HCO2] N/A 289 N/A
and one anion).100 Thus, as was described in the Methods
[PPyr][Tf2N] N/A 561 N/A
section, the IL molar enthalpies of vaporization were calculated
by simulating a single cation/anion pair in the gas phase. The
mean absolute error 3.6
equation for calculating the molar enthalpy of vaporization is
max absolute error 10.3
shown below
water model experiment simulation error (%)
TIP3P 7539 83 (7839) 9.9 Δhvap(T , P) = ⟨hgas(T )⟩ − ⟨hliquid(T , P)⟩
SPC/E 7539 89 (8739) 18.7
a
Values in parentheses represent simulated heat capacities from Vega = [⟨ugas(T )⟩ + RT ]
and Abascal.39 − [⟨uliquid(T , P)⟩ + Pv]
≈ [⟨ugas(T )⟩ + RT ] − [⟨uliquid(T , P)⟩]
TIP3P and SPC/E water are also reported in Table 2. These
results are also graphically displayed in Figure 2B, which shows = Δu vap(T , P) + RT
the simulated and experimental error of each data point (note
that some points have no experimental error bar due to a single (4)
source and that, for some points, simulation errors are too small where Chevron brackets represent ensemble averaged molar
to be seen). Error bars represent the standard deviation of all quantities. The gas is assumed to be ideal, and Pv contributions
experimental or simulation data. (i.e., pressure−volume work, where P is the pressure and v is
Similar to density, GAFF slightly underestimates the heat the molar volume) in the liquid can be considered negligible
capacities of the majority of the ILs. It appears, however, that when compared to the internal energies and thus can be
GAFF more accurately estimates heat capacities that lie within neglected. Additionally, the liquid internal energies (uliquid(T,
the lower half of the experimental range of heat capacities of the P)) were divided by the total number of ion pairs in the system,
5886 DOI: 10.1021/acs.jpcb.5b00689
J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

since the desired final quantity was the enthalpy of vaporization

(Δhvap) for a single ion pair.
Due to limited experimental data, experimental molar
enthalpies of vaporization and results from this work for only
8 of the 19 ILs are reported in Table 3, as well as in a parity plot

Table 3. Experimental and Simulated IL Molar Enthalpies of

Vaporization (kJ/mol)a
IL experiment simulation error (%)
[EMIM][Tf2N] 135105,106 117 (159,103 146104) −13.0
[BMIM][Tf2N] 135105,106 122 (174,103 151104) −9.5
[HMIM][Tf2N] 139105,106 128 (184,103 157104) −8.0
[OMIM][Tf2N] 150105,106 134 (201,103 162104) −10.1
[PPyr][Tf2N] 134107 125 −6.7
[HPyr][Tf2N] 137107 128 −6.3 Figure 4. Simulated vs experimental IL molar enthalpies of
[EMIM][EtSO4] 164105 143 −12.9 vaporization for [RMIM][Tf2N] ILs (R = E, B, H, and O). Simulation
[OMIM][Otf] 151105 141 −6.7 results are also shown from the work of Santos et al.,103 using a refined
OPLS-AA force field, and Kelkar and Maginn,104 using the CHARMM
mean absolute error 9.1 22 force field.
max absolute error 13.0
water model experiment simulation error (%)
TIP3P 4439
38 (4239) −12.9
are due to additional van der Waals interactions brought about
SPC/E 4439 45 (4939) 2.4
by the increasing chain lengths.103,104 Given the significantly
a increased difficulty of accurately predicting this property over
Values in parentheses represent simulated molar enthalpies of
others such as density or heat capacity, in conjunction with the
vaporization from Santos et al.,103 Kelkar and Maginn,104 and Vega
and Abascal.39 use of an unrefined force field, the errors yielded by GAFF in
predicting the molar enthalpies of vaporization of these
particular ILs are not unreasonable; however, it is recom-
in Figure 2C; simulation error bars are too small to be seen, and mended that for the property of molar enthalpy of vaporization,
multiple experimental sources could only be found for the four as well as for self-diffusivity as it turns out (see next subsection),
[Tf2N]-based ILs; thus, some points have no experimental one should use this method to merely draw qualitative trends
error bars. Error bars represent the standard deviation of all rather than as a means to obtain accurate predictions.
experimental or simulation data. For the [Tf2N]-based ILs in Self-Diffusivity. Though the ability to predict thermody-
Table 3, the reported experimental value is an average of those namic properties with accuracy is a good first assessment of a
from the specified sources. Also reported in Table 3 is data for force field, it does not necessarily ensure that transport
TIP3P and SPC/E water, as well as molar enthalpies of properties will also be well predicted. Thus, in assessing the
vaporization of the [RMIM][Tf2N] ILs from two other accuracy of a force field, it is often very important to calculate
simulation groups103,104 for comparison to the results of this common dynamic properties, such as self-diffusivity and shear
work. viscosity, to complement calculations of static and equilibrium
Our results show that GAFF is able to calculate molar properties.
enthalpies of vaporization of the eight ILs with fair agreement The Einstein relation shown in eq 5 was used to compute the
to experiment, with the largest percent error occurring for self-diffusivity of the cations and anions through calculation of
[EMIM][Tf2N] (−13.0%; note this does not exceed the error the mean square displacement, or MSD, of the ion centers of
of the TIP3P water model). Additionally, at least in the case of mass
the four [RMIM][Tf2N] ILs, GAFF predicts the molar
N
enthalpies of vaporization with similar or better accuracy than 1 d 2
the computed values of other published studies that used D= lim ⟨∑ [ r (⃗ t ) − r (0)]
⃗ ⟩
6 t →∞ dt i = 1 (5)
parameters from a force field based on the AMBER/OPLS-AA
framework extended to [RMIM][Tf2N] ILs,103 and the where the bracketed term is the MSD over time and the factor
CHARMM 22 force field,104 illustrated in Figure 4. of 1/6 arises for a three-dimensional system.
For the six [Tf2N]-based ILs in Table 3, GAFF captures the As with the molar enthalpy of vaporization, limited
well-documented trend, both experimentally and computation- experimental data led to our decision to calculate the self-
ally, of increasing vaporization molar enthalpy with increasing diffusivity of only the four [RMIM][Tf2N] ILs. Computed and
cation carbon chain length. For the four [Tf2N]-based ILs, our experimental self-diffusivities for the cation and anion of each of
results, in agreement with those of Kelkar and Maginn,104 show the ILs are reported in Table 4, along with self-diffusivities from
around a 5−6 kJ/mol increase in the molar enthalpy of the work of Köddermann and Paschek108 for later comparison
vaporization for each two-carbon increase in cation chain and TIP3P and SPC/E water data. All production simulations
length. This trend also generally agrees with the experimental were run in triplicate for 10 ns except where later noted, and
results, though virtually no increase in the molar enthalpy of fitting for calculation of the self-diffusion coefficients was done
vaporization is seen with an increase in carbon chain length over the range 1−9 ns. For ease of visualizing the trends, the
from C2 to C4, experimentally, and a larger increase (10 kJ/ data is also graphically displayed in Figures 5 and 6; error bars
mol) accompanies a jump from a length of C6 to C8. Both past represent the standard deviation of all experimental or
simulations and experiments have proposed that these increases simulation data, and simulation error bars are generally too
5887 DOI: 10.1021/acs.jpcb.5b00689
J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

Table 4. Experimental and Simulated IL Self-Diffusivities (10−11 m2/s)a

IL cation, exp109 cation, sim error (%) anion, exp109 anion, sim error (%)
[EMIM][Tf2N] 4.98 2.29 −53.9 3.10 1.11 −64.1
5.88b 6.00b,108 +2.0 3.70b 2.30b,108 −37.9
[BMIM][Tf2N] 2.76 1.13 −51.8 2.20 0.83 −62.2
3.37b 3.00b,108 −11.0 2.69b 2.20b,108 −18.4
[HMIM][Tf2N] 1.69 0.87 −48.7 1.54 0.61 −60.6
2.11b 1.30b,108 −38.5 1.93b 1.20b,108 −38.0
[OMIM][Tf2N] 1.17 0.41 −65.1 1.17 0.31 −73.4
1.48b 0.50b,108 −66.3 1.48b 0.48b,108 −67.5

mean absolute errorc 54.9 65.1

29.5 40.4
max absolute errorc 65.1 73.4
66.3 67.5
water model exp sim error (%)
TIP3P 23039 586 (54939) 154.8
SPC/E 23039 264 (25439) 14.7
a
Simulated self-diffusivities from Köddermann and Paschek108 and Vega and Abascal39 are included for comparison. bEvaluated at 303 K and 1 bar.
c
Evaluated first for the data from this work and then for the data from Köddermann and Paschek.108

Figure 5. Simulated vs experimental IL cation and anion self-diffusivities for [RMIM][Tf2N] ILs (R = E, B, H, and O): (A) simulation results of this
work compared to experimental values109 at 298.15 K and 1 bar; (B) simulation results from Köddermann and Paschek108 compared to experimental
values109 at 303 K and 1 bar.

Figure 6. Simulated vs experimental109 cation to anion self-diffusivity ratios for [RMIM][Tf2N] ILs (R = E, B, H, and O). Simulation results from
Köddermann and Paschek108 are included for comparison.

small to be seen, whereas points lack experimental error bars The results show that GAFF is able to accurately capture the
due to the availability of only a single experimental source. trend of decreasing cation/anion self-diffusivity with increasing
5888 DOI: 10.1021/acs.jpcb.5b00689
J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

cation chain length, and even to some extent the relatively large as a way to obtain meaningful qualitative trends of self-
nonlinear trend in self-diffusivity of the cation in moving from a diffusivity, or for the purposes mentioned in the Conclusions
cation chain length of C2 to C4. The trend of decreasing cation/ such as a starting point for force field refinement.
anion self-diffusivity with increasing cation chain length has When considering the quantitative and qualitative accuracy
been observed by others and can be explained by an increased of the results and trends of IL self-diffusivity, respectively, it is
steric hindrance brought on by longer cation chain lengths that important to consider the ability of each IL to reach diffusive
serve to slow down the diffusion of the cations through the behavior over the time scale of the simulations. In a review
IL.101 GAFF is also able to capture the interesting trend of the article on computing IL properties with simulation, Maginn
anion self-diffusivity tracking that of the cation, a trend that has suggests computing the time-dependent quantity beta, or β(t),
been observed by Cadena et al.101 In other words, our results as a way to measure whether a system is in the diffusive
show that the self-diffusivity of the [Tf2N] anion is highest regime.100 The equation to calculate β(t) is shown below
when paired with the fastest cation, [EMIM], and that it is
lowest when paired with the slowest cation, [OMIM]; as others d log(Δr 2)
β (t ) =
have noted, this suggests that cations and anions diffuse d log(t ) (6)
through the liquid in pairs or clusters, rather than on their
own.101 Our results also show that, for the four IL cation chain where t is the simulation time and Δr2 is the MSD of the ion
lengths tested, the self-diffusivity of the cation at 298.15 K is centers of mass, similar to eq 5.
always higher than that of the anion, though studies show this is The system can be said to have reached diffusive behavior
not the case for all ILs.101 From a quantitative perspective, when β(t) has reached a value of 1; values below 1 thus indicate
deviations in cation/anion self-diffusivity from experiment subdiffusive behavior.100 The triplicate TIP3P water simulations
range from 49 to 73%. There is consistently a greater deviation all reached β(t) = 1 after only 1 ns, while the triplicate
from experiment of the simulated anion self-diffusivities than of simulations of [RMIM][Tf2N] (R = E, B, and H) had all
the simulated cation self-diffusivities, with the largest percent reached diffusive behavior after 5 ns. Neither simulation of
error occurring for [Tf2N] in [OMIM][Tf2N]. [OMIM][Tf2N], however, reached diffusive behavior over the
Using a force field based on the OPLS-AA/AMBER 20 ns; after leveling off at around 17 ns, the maximum β(t)
framework and specifically modified to reproduce densities value reached for either simulation was 0.6. Thus, for ILs with
and self-diffusion coefficients of ions of [RMIM][Tf2N] ILs, long-chained ions, such as [OMIM][Tf2N], it would be
molecular dynamics simulations were carried out by Ködder- advisible to extend the simulation time, or increase the
mann and Paschek108 to calculate self-diffusion coefficients at simulation temperature, to have more confidence in the
303 K and 1 bar. A similar method to that used in this work was calculated ion self-diffusivities. Indeed, of the four [RMIM]
employed to perform the actual diffusion calculations. As Table cations, the greatest percent error relative to experiment is seen
4 shows, the IL-refined force field results are much more for [OMIM] (65%), and as was noted earlier, of the four
accurate than the GAFF results at the lower cation chain [Tf2N] anions, [Tf2N] in [OMIM][Tf2N] has the greatest
lengths of two and four carbons. As the cation chain length deviation (73%).
increases to eight, however, the error gap between the refined Shear Viscosity. Past simulations of IL shear viscosity have
OPLS and GAFF force fields becomes nearly negligible, easily roughly fallen into two categories: (1) equilibrium MD
visible in Figure 5. Overall, there is about a 25% increase in the simulations that focus on the intrinsic pressure and momentum
mean absolute error for the GAFF cation and anion data sets fluctuations of the system and (2) nonequilibrium MD
over the IL-refined OPLS data sets. Another interesting and simulations that monitor the linear response of the system
informative way to compare the simulation data against each after it is driven away from equilibrium. Nearly all previous
other and to experiment is by plotting the ratio of the cation to attempts using equilibrium methods have led to over-
anion self-diffusivity as a function of cation carbon chain length. estimations of IL shear viscosities by at least an order of
This is shown in Figure 6 below, where gray markers represent magnitude, while calculations using nonequilibrium methods
the experimental data sets at 298.15 and 303 K (the ratios are have tended to agree better with experiment and generally
the same at these two temperatures) and black markers require significantly shorter simulation times for calculations.100
represent the two simulation data sets at 298.15 and 303 K. As For example, in a study performed by Hess, the shear viscosity
Figure 6 shows, the ratios of cation to anion self-diffusivity of a Lennard-Jones fluid using the SPC and SPC/E water
generated with the IL-refined OPLS force field are, in every models was calculated using two equilibrium and two
case except for [EMIM][Tf2N], closer to experimental ratios nonequilibrium methods; of the four methods, the non-
than results generated with GAFF, though GAFF captures the equilibrium, periodic perturbation method (PPM) provided
trend of decreasing cation to anion self-diffusivity ratios with the best results.112 Accordingly, we adopted the PPM method,
increasing cation chain lengths very well. first developed by Gosling et al.,38 to assess GAFF’s ability to
Considering that the OPLS force field is specifically refined predict IL shear viscosities.
for this family of ILs and for the property of self-diffusion and In using PPM to calculate shear viscosity, an external force is
that other groups using this same force field without the applied to the system, which induces a velocity field in the
property refinement saw diffusion results off by at least an order liquid that is described by the transverse portion of the Navier−
of magnitude from experimental values,110,111 it is surprising Stokes equation, shown below
that simply using an unrefined force field such as GAFF with
scaled atomic charges yields such relatively accurate results. ∂ux(z , t ) ∂ 2u (z , t )
ρ = ρax(z , t ) + η x 2
However, as was suggested earlier in regard to the molar ∂t ∂z (7)
enthalpy of vaporization, it is perhaps inadvisible to use this
method for the purpose of obtaining accurate predictions of where ax(z, t) and ux(z, t) are the external acceleration and
self-diffusion coefficients. Rather, we suggest using this method velocity field, respectively. An acceleration can be added to each
5889 DOI: 10.1021/acs.jpcb.5b00689
J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

particle at each MD step, according to the acceleration profile Table 5. Experimental and Simulated IL Shear Viscosities
shown below (Pa·s) Using a Shear Rate of 0.03 nm/ps2 a
2π IL experimental simulation error (%)
ax(z , t ) = a0 cos(kz) k=
lz (8) [BAM][HCO2] 0.07040 0.091 29.7
[BMIM][Ac] 0.34941,42,75,95,114 0.055 −84.2
where k is the wavenumber and lz is the box height in the z- [BMIM][MeSO4] 0.094115 0.082 −12.8
dimension. The acceleration profile is assumed to take this [BMIM][Tf2N] 0.05052,83,116 0.045 −10.6
shape because it satisfies the conditions that the acceleration [EAM][HCO2] 0.03240 0.036 11.3
and velocity profiles be both periodic (since the simulations [EMIM][Ac] 0.13542,54,117 0.027 −80.4
feature periodic systems) and smooth (so local shear rates are [EMIM][EtSO4] 0.09860,75,118−120 0.073 −26.1
small).112 This assumption, together with the assumption that [EMIM][Otf] 0.04254,121 0.041 −2.4
the initial velocity ux(z, t = 0) is zero, leads to the following [EMIM][Tf2N] 0.03383,122,123 0.036 9.3
solution to the velocity profile [EMPyr][EtSO4] 0.15664,95 0.053 −66.2
[EMPyr][Tf2N] N/A N/A N/A
ux(z , t ) = a0τ(1 − e−t / τ ) cos(kz) = u0 cos(kz) (9)
[HMIM][Cl] 18.1068 0.380 −98.0
where τ, the system relaxation time, is given by eq 10. [HMIM][Tf2N] 0.07272,75,83,124,125 0.080 10.2
[HPyr][Tf2N] 0.08776,77,95 0.059 −31.7
ρ
τ= 2 [OMIM][Cl] 20.9368,119,126 0.419 −98.0
ηk (10) [OMIM][Otf] N/A N/A N/A
At sufficiently long times, e −t/τ
goes to zero and the amplitude [OMIM][Tf2N] 0.09183,127,128 0.077 −15.3
of the steady-state velocity profile, u0, is equal to a0τ. By fitting [PAM][HCO2] 0.07840 0.093 19.1
this steady-state velocity profile to a cos(kz) form, one can then [PPyr][Tf2N] 0.07277 0.050 −30.2
measure u0 and use eq 11 to calculate the IL shear viscosity.
mean absolute error 17.4,b 37.4c
a ρ 31.7,b 98.0c
η= 0 − 2 max absolute error
u0 k (11) water model experimental simulation error (%)
TIP3P 9.0 × 10−4 39 3.1 × 10−4 (3.2 × 10−4 39) −65.1
As was described earlier in the Methods section, 1 ns NPT
SPC/E 9.0 × 10−4 39 7.0 × 10−4 (7.3 × 10−4 39) −22.2
simulations were run in triplicate to calculate the shear a
viscosity. Simulations were run at shear rates of 0.01, 0.03, Values in parentheses represent simulated shear viscosities from Vega
0.05, and 1.0 nm/ps2, and though the shear viscosity calculated and Abascal.39 Bolded ILs are not included in Figure 2D, which is
restricted to a range that highlights good fits. bExcludes bolded ILs.
using the smallest perturbation should theoretically be nearest c
Includes bolded ILs.
to the true shear viscosity at the hydrodynamic limit (i.e., k =
0),113 the best overall agreement with experiments was
achieved with a shear rate of 0.03 nm/ps2. Thus, Table 5
shows the computed and experimental shear viscosities for 17 a few similarities can be found among these five most deviant
of the 19 ILs (experimental data at 298.15 K was unavailable for IL values. Recurring anions include [Cl] and [Ac]. In the case
[EMPyr][Tf2N] and [OMIM][Otf]), as well as for TIP3P and of the [Cl]-based ILs, we do not recommend using GAFF to
SPC/E water, at a shear rate of 0.03 nm/ps2. Reported calculate transport properties. ILs are generally considered to
experimental values represent averages of the values from the be highly viscous fluids, and [Cl]-based ILs have significantly
specified experimental sources. The data is also displayed in a higher shear viscosities than most ILs, as Table 5 shows. While
parity plot in Figure 2D, though the range is restricted to GAFF does correctly predict that the 2 [Cl]-based ILs have the
highlight good fits (note that some points have no experimental highest viscosities of the 17 ILs at the temperature used in this
error bars due to a single source and that, for some points, study (298.15 K), the simulation dynamics of the [RMIM][Cl]
simulation error bars are too small to be seen). Error bars ILs were visibly so slow as to imply that a glass-like state may
shown in Figure 2D represent the standard deviation of all have been reached. This may reflect that these particular
experimental or simulation data. systems did not come close to reaching the hydrodynamic limit,
As Table 5 shows for the sets of [Tf2N]- and [HCO2]-based where the shear viscosity is a number rather than a function,
ILs, GAFF is able to reproduce the experimental trend of due to too small of a box length (and thus too large of a
increasing shear viscosity with increasing cation chain length, wavenumber). Indeed, a study by Hu and Margulis on the shear
even despite the significantly large errors that were seen among viscosity of [HMIM][Cl], using OPLS-AA force field
the triplicate simulations of [BAM][HCO2] and [PAM]- parameters, showed that, with extremely large box sizes and
[HCO2] (see Figure 2D). Additionally, as Figure 2D shows, higher temperatures, much better agreement with experiment
below an experimental shear viscosity of approximately 0.1 Pa·s, could be achieved.113
GAFF is able to predict IL shear viscosity with good To test whether GAFF conforms to this same trend that
quantitative accuracy; over this range of experimental values, higher temperatures lead to more accurate IL shear viscosity
which accounts for just over 70% of the total data set, errors estimates, additional simulations were performed at temper-
range from as little as 2% to a maximum of 34%. Above an atures above 298.15 K for which experimental data was
experimental shear viscosity of 0.1 Pa·s (bolded ILs shown in available, for three of the five ILs with the most poorly
Table 5), GAFF predicts shear viscosities with deviations of predicted shear viscosities (see bolded ILs in Table 5), namely,
66−98%; it is important to note that these deviations, however, [EMIM][Ac], [EMPyr][EtSO4], and [HMIM][Cl]. Results are
are comparable to the 65% deviation in the simulated shear shown in Table 6 below, where reported simulated values are
viscosity of TIP3P water compared to experiment. Nonetheless, averages of three simulations run in triplicate, and reported
5890 DOI: 10.1021/acs.jpcb.5b00689
J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

Table 6. Experimental and Simulated IL Shear Viscosities (Pa·s) Using a Shear Rate of 0.03 nm/ps2 and Temperatures above
Room Condition
IL temperature (K) experiment simulation error (%)
[EMIM][Ac] 298.15 0.13542,54,117 0.027 −80.4
353.15 0.01654,129 0.006 −59.5
393.15 0.005129 0.003 −36.9
[EMPyr][EtSO4] 298.15 0.15664,95 0.053 −66.2
323.15 0.04864,95 0.022 −54.5
343.15 0.02464,95 0.012 −51.8
[HMIM][Cl] 298.15 18.1068 0.380 −98.0
353.15 0.23870,129 0.063 −73.4
393.15 0.043129 0.019 −55.4

experimental values are averages of the values of the cited After determining that GAFF, with no refinement, is suitable
literature sources. for predicting thermodynamic and transport properties of ILs
Results show that, in line with the findings of Hu and with reasonable accuracy, we suggest that the protocol we
Margulis,113 as the simulation temperature is increased above describe in this paper might be of interest to other researchers
room temperature, the error in the calculated shear viscosities in the following situations:
decreases for each of the three ILs, by over 40% in the cases of 1. when IL properties are not known experimentally
[EMIM][Ac] and [HMIM][Cl] from 298.15 to 393.15 K.
2. to get a good starting point for force field refinement
Opportunities for Improvement. Unfortunately, GAFF
does not have force field parameters for some of the most 3. to screen known ILs for unknown properties, quickly,
common constituent ions in ILs, such as AlCl4, BF4, and PF6. In easily, and efficiently
such cases, the free tool paramfit can be used to derive 4. to guide in the discovery of new ILs
unknown force field parameters from ab initio calculations. Because there are many additional applications that feature

■ CONCLUSIONS
Densities, heat capacities, molar enthalpies of vaporization, self-
systems of water/IL mixtures, future research will focus on
GAFF’s ability to predict thermodynamic and transport
properties of water/IL mixtures, as well as mixtures of multiple
different ILs.

■
diffusivities, and shear viscosities of 19 ionic liquids have been
calculated using the general AMBER force field and compared
to experiment. Results show that GAFF is able to reproduce AUTHOR INFORMATION
these properties with good accuracy compared to experiment Corresponding Author
and with similar accuracy compared to other published force *E-mail: [email protected].
fields. No refinement to GAFF was required for such accuracy. Notes
Specifically, deviations in simulated IL densities from The authors declare no competing financial interest.

■
experimental values were ca. 1−4%, heat capacity deviations
ranged from 0 to 10%, and deviations in molar enthalpies of
ACKNOWLEDGMENTS
vaporization from experiment were ca. 6−13%. In general,
across the range of ILs tested, GAFF tended to underestimate The authors would like to acknowledge the NSF award CBET-
the densities, heat capacities, and molar enthalpies of 1150596, as well as the Hyak supercomputer system at the
vaporization. Significantly larger errors occurred for calculations University of Washington, for support of this research.
of IL transport properties; cation and anion self-diffusivity
values were all underestimated by GAFF and deviated from
experiment in the range 50−70%. GAFF was, however, able to
■ REFERENCES
(1) Keskin, S.; Kayrak-Talay, D.; Akman, U.; Hortaçsu, Ö . A Review
capture important trends in the data, such as decreasing self- of Ionic Liquids Towards Supercritical Fluid Applications. J. Supercrit.
diffusivity with increasing cation chain length, as well as anion Fluids 2007, 43, 150−180.
self-diffusivities tracking those of the cations. Deviations in IL (2) Mora-Pale, M.; Meli, L.; Doherty, T. V.; Linhardt, R. J.; Dordick,
shear viscosity from experimental values varied widely, from as J. S. Room Temperature Ionic Liquids as Emerging Solvents for the
Pretreatment of Lignocellulosic Biomass. Biotechnol. Bioeng. 2011, 108,
little as 2% to as much as 98%. The largest errors occurred for 1229−1245.
ILs with [Cl], [RSO4], and [Ac] anions, and may indicate that (3) Jorgensen, W. L.; Maxwell, D. S.; TiradoRives, J. Development
those simulations were unable to reach the hydrodynamic limit and Testing of the OPLS All-Atom Force Field on Conformational
and thus that larger box sizes and higher temperatures could be Energetics and Properties of Organic Liquids. J. Am. Chem. Soc. 1996,
used to achieve better agreement with experiment. This theory 118, 11225−11236.
was tested for three ILs, and results showed errors in the (4) Vanommeslaeghe, K.; Hatcher, E.; Acharya, C.; Kundu, S.;
simulated shear viscosities indeed decreased with increased Zhong, S.; Shim, J.; Darian, E.; Guvench, O.; Lopes, P.; Vorobyov, I.;
temperatures. Excluding the most deviant ILs from the data set et al. CHARMM General Force Field: A Force Field for Drug-Like
Molecules Compatible with the CHARMM All-Atom Additive
at 298.15 K, the range of viscosity error was found to be much Biological Force Fields. J. Comput. Chem. 2010, 31, 671−690.
more reasonable, with a max absolute error of ∼30%. (5) Rappe, A. K.; Casewit, C. J.; Colwell, K. S.; Goddard, W. A.; Skiff,
Considering the difficulty of calculating dynamic properties of W. M. Uff, a Full Periodic-Table Force-Field for Molecular Mechanics
ionic liquids with simulation,100 GAFF does surprisingly well at and Molecular-Dynamics Simulations. J. Am. Chem. Soc. 1992, 114,
predicting these properties and capturing important trends. 10024−10035.

5891 DOI: 10.1021/acs.jpcb.5b00689

J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

(6) Wang, J. W.; M, R.; Caldwell, J. W.; Kollman, P. A.; Case, D. A. (26) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.;
Development and Testing of a General Amber Force Field. J. Comput. Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci,
Chem. 2004, 25, 1157−1174. B.; Petersson, G. A.; et al. Gaussian 09; Gaussian, Inc.: Wallingford,
(7) Liu, H.; Sale, K. L.; Simmons, B. A.; Singh, S. Molecular CT, 2009.
Dynamics Study of Polysaccharides in Binary Solvent Mixtures of an (27) Wang, J.; Wang, W.; Kollman, P. A.; Case, D. A. Automatic
Ionic Liquid and Water. J. Phys. Chem. B 2011, 115, 10251−10258. Atom Type and Bond Type Perception in Molecular Mechanical
(8) Liu, H.; Maginn, E. A Molecular Dynamics Investigation of the Calculations. J. Mol. Graphics Modell. 2006, 25.
Structural and Dynamic Properties of the Ionic Liquid 1-N-Butyl-3- (28) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Kollman, P. A.
Methylimidazolium Bis(Trifluoromethanesulfonyl)Imide. J. Chem. Application of RESP Charges to Calculate Conformational Energies,
Phys. 2011, 135, 124507. Hydrogen-Bond Energies, and Free-Energies of Solvation. J. Am.
(9) Jiang, W.; Yan, T.; Wang, Y.; Voth, G. A. Molecular Dynamics Chem. Soc. 1993, 115, 9620−9631.
Simulation of the Energetic Room-Temperature Ionic Liquid, 1- (29) Heaps, C. G. Long-Range Energy Alternatives Planning (LEAP)
Hydroxyethyl-4-Amino-1,2,4-Triazolium Nitrate (HEATN). J. Phys. System [Software Version 2014.0.1.20]. Stockholm Environment
Chem. B 2008, 112, 3121−3131. Institute:: Somerville, MA, 2012.
(10) Liu, H.; Sale, K. L.; Holmes, B. M.; Simmons, B. A.; Singh, S. (30) Sousa da Silva, A. W.; Vranken, W. F. ACPYPE - AnteChamber
Understanding the Interactions of Cellulose with Ionic Liquids: A PYthon Parser interacE. BMC Res. Notes 2012, 5, 367.
Molecular Dynamics Study. J. Phys. Chem. B 2010, 114, 4293−4301. (31) Tenney, C. M.; Massel, M.; Mayes, J. M.; Sen, M.; Brennecke, J.
(11) Zhang, Y.; Maginn, E. J. A Simple AIMD Approach to Derive F.; Maginn, E. J. A Computational and Experimental Study of the Heat
Atomic Charges for Condensed Phase Simulation of Ionic Liquids. J. Transfer Properties of Nine Different Ionic Liquids. J. Chem. Eng. Data
Phys. Chem. B 2012, 116, 10036−10048. 2014, 59, 391−399.
(12) Men, S.; Lovelock, K. R. J.; Licence, P. X-Ray Photoelectron (32) Martínez, L.; Andrade, R.; Birgin, E. G.; Martínez, J. M.
Spectroscopy of Pyrrolidinium-Based Ionic Liquids: Cation-Anion PACKMOL: A Package for Building Initial Configurations for
Interactions and a Comparison to Imidazolium-Based Analogues. Phys. Molecular Dynamics Simulations. J. Comput. Chem. 2009, 30, 2157−
Chem. Chem. Phys. 2011, 13, 15244−15255. 2164.
(13) Hurisso, B. B.; Lovelock, K. R. J.; Licence, P. Amino Acid-Based (33) Berendsen, H. J. C.; van der Spoel, D.; van Drunen, R.
Ionic Liquids: Using XPS to Probe the Electronic Environment Via GROMACS: A Message-Passing Parallel Molecular Dynamics
Binding Energies. Phys. Chem. Chem. Phys. 2011, 13, 17737−17748. Implementation. Comput. Phys. Commun. 1995, 91, 43−56.
(14) Youngs, T. G. A.; Hardacre, C. Application of Static Charge (34) Lindahl, E.; Hess, B.; van der Spoel, D. GROMACS 3.0: A
Transfer within an Ionic-Liquid Force Field and Its Effect on Structure Package for Molecular Simulation and Trajectory Analysis. J. Mol.
and Dynamics. ChemPhysChem 2008, 9, 1548−1558. Model. 2001, 7, 306−317.
(15) Bhargava, B. L.; Balasubramanian, S. Refined Potential Model (35) Bussi, G.; Donadio, D.; Parrinello, M. Canonical Sampling
through Velocity Rescaling. J. Chem. Phys. 2007, 126, 014101.
for Atomistic Simulations of Ionic Liquid [Bmim][PF6]. J. Chem. Phys.
(36) Berendsen, H. J. C.; Postma, J. P. M.; van Gunsteren, W. F.;
2007, 127, 114510.
DiNola, A.; Haak, J. R. Molecular Dynamics with Coupling to an
(16) Sieffert, N.; Wipff, G. The [BMI][Tf2N] Ionic Liquid/Water
External Bath. J. Chem. Phys. 1984, 81, 3684.
Binary System: A Molecular Dynamics Study of Phase Separation and
(37) Parrinello, M.; Rahman, A. Polymorphic Transitions in Single
of the Liquid-Liquid Interface. J. Phys. Chem. B 2006, 110, 13076−85.
Crystals: A New Molecular Dynamics Method. J. Appl. Phys. 1981, 52,
(17) Zhao, W.; Eslami, H.; Cavalcanti Welchy, L.; Müller-Plathe, F. A
7182−7190.
Refined All-Atom Model for the Ionic Liquid 1-N-Butyl 3-
(38) Gosling, E. M.; McDonald, I. R.; Singer, K. On the Calculation
Methylimidazolium Bis(Trifluoromethylsulfonyl)Imide [Bmim]- by Molecular Dynamics of the Shear Viscosity of a Simple Fluid. Mol.
[Tf2N]. Z. Phys. Chem. 2007, 221, 1647. Phys. 1973, 26, 1475−1484.
(18) Youngs, T. G.; Hardacre, C. Application of Static Charge (39) Vega, C.; Abascal, J. L. F. Simulating Water with Rigid Non-
Transfer within an Ionic-Liquid Force Field and Its Effect on Structure Polarizable Models: A General Perspective. Phys. Chem. Chem. Phys.
and Dynamics. ChemPhysChem 2008, 9, 1548−58. 2011, 13, 19663−19688.
(19) Jaeger, V. W.; Pfaendtner, J. Structure, Dynamics, and Activity of (40) Greaves, T. L.; Weerawardena, A.; Fong, C.; Krodkiewska, I.;
Xylanase Solvated in Binary Mixtures of Ionic Liquid and Water. ACS Drummond, C. J. Protic Ionic Liquids: Solvents with Tunable Phase
Chem. Biol. 2013, 8, 1179−1186. Behavior and Physicochemical Properties. J. Phys. Chem. B 2006, 110,
(20) Jarin, Z.; Pfaendtner, J. Ionic Liquids Can Selectively Change 22479−22487.
the Conformational Free-Energy Landscape of Sugar Rings. J. Chem. (41) Almeida, H. F. D.; Passos, H.; Lopes-da-Silva, J. A.; Fernandes,
Theory Comput. 2014, 10, 507−510. A. M.; Freire, M. G.; Coutinho, J. A. P. Thermophysical Properties of
(21) Burney, P. R.; Pfaendtner, J. Structural and Dynamic Features of Five Acetate-Based Ionic Liquids. J. Chem. Eng. Data 2012, 57, 3005−
Candida Rugosa Lipase 1 in Water, Octane, Toluene, and Ionic 3013.
Liquids BMIM-PF6 and BMIM-NO3. J. Phys. Chem. B 2013, 117, (42) Araújo, J. M. M.; Pereiro, A. B.; Alves, F.; Marrucho, I. M.;
2662−2670. Rebelo, L. P. N. Nucleic Acid Bases in 1-Alkyl-3-Methylimidazolium
(22) Nordwald, E. M.; Kaar, J. L. Mediating Electrostatic Binding of Acetate Ionic Liquids: A Thermophysical and Ionic Conductivity
1-Butyl-3-Methylimidazolium Chloride to Enzyme Surfaces Improves Analysis. J. Chem. Thermodyn. 2013, 57, 1−8.
Conformational Stability. J. Phys. Chem. B 2013, 117, 8977−8986. (43) Shiflett, M. B.; Harmer, M. A.; Junk, C. P.; Yokozeki, A.
(23) Fiebig, O. C.; Mancini, E.; Caputo, G.; Vaden, T. D. Solubility and Diffusivity of Difluoromethane in Room-Temperature
Quantitative Evaluation of Myoglobin Unfolding in the Presence of Ionic Liquids. J. Chem. Eng. Data 2006, 51, 483−495.
Guanidinium Hydrochloride and Ionic Liquids in Solution. J. Phys. (44) Pereiro, A. B.; Verdía, P.; Tojo, E.; Rodríguez, A. Physical
Chem. B 2013, 118, 406−412. Properties of 1-Butyl-3-Methylimidazolium Methyl Sulfate as a
(24) Nordwald, E. M.; Armstrong, G. S.; Kaar, J. L. NMR-Guided Function of Temperature. J. Chem. Eng. Data 2007, 52, 377−380.
Rational Engineering of an Ionic-Liquid-Tolerant Lipase. ACS Catal. (45) Singh, T.; Kumar, A. Temperature Dependence of Physical
2014, 4, 4057−4064. Properties of Imidazolium Based Ionic Liquids: Internal Pressure and
(25) Yu, C.-Y.; Wei, P.; Li, X.-F.; Zong, M.-H.; Lou, W.-Y. Using Molar Refraction. J. Solution Chem. 2009, 38, 1043−1053.
Ionic Liquid in a Biphasic System to Improve Asymmetric Hydrolysis (46) García-Miaja, G.; Troncoso, J.; Romaní, L. Excess Properties for
of Styrene Oxide Catalyzed by Cross-Linked Enzyme Aggregates Binary Systems Ionic Liquid + Ethanol: Experimental Results and
(CLEAs) of Mung Bean Epoxide Hydrolases. Ind. Eng. Chem. Res. Theoretical Description Using the ERAS Model. Fluid Phase Equilib.
2014, 53, 7923−7930. 2008, 274, 59−67.

5892 DOI: 10.1021/acs.jpcb.5b00689

J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

(47) Domańska, U.; Pobudkowska, A.; Wiśniewska, A. Solubility and Gas−Liquid Chromatography at T = (313.15, 323.15, and 333.15) K.
Excess Molar Properties of 1,3-Dimethylimidazolium Methylsulfate, or J. Chem. Thermodyn. 2010, 42, 78−83.
1-Butyl-3-Methylimidazolium Methylsulfate, or 1-Butyl-3-Methylimi- (63) Dzida, M.; Chorążewski, M.; Geppert-Rybczyńska, M.; Zorębski,
dazolium Octylsulfate Ionic Liquids with N-Alkanes and Alcohols: E.; Zorębski, M.; Ż arska, M.; Czech, B. Speed of Sound and Adiabatic
Analysis in Terms of the PFP and FBT Models. J. Solution Chem. 2006, Compressibility of 1-Ethyl-3-Methylimidazolium Bis-
35, 311−334. (Trifluoromethylsulfonyl)Imide under Pressures up to 100 MPa. J.
(48) Iglesias-Otero, M. A.; Troncoso, J.; Carballo, E.; Romaní, L. Chem. Eng. Data 2013, 58, 1571−1576.
Density and Refractive Index in Mixtures of Ionic Liquids and Organic (64) González, B.; Calvar, N.; Gómez, E.; Macedo, E. A.; Domínguez,
Solvents: Correlations and Predictions. J. Chem. Thermodyn. 2008, 40, Á . Synthesis and Physical Properties of 1-Ethyl 3-Methylpyridinium
949−956. Ethylsulfate and Its Binary Mixtures with Ethanol and Water at Several
(49) Harris, K. R.; Kanakubo, M.; Woolf, L. A. Temperature and Temperatures. J. Chem. Eng. Data 2008, 53, 1824−1828.
Pressure Dependence of the Viscosity of the Ionic Liquids 1-Hexyl-3- (65) Calvar, N.; Gómez, E.; González, B.; Domínguez, Á .
Methylimidazolium Hexafluorophosphate and 1-Butyl-3-Methylimida- Experimental Determination, Correlation, and Prediction of Physical
zolium Bis(Trifluoromethylsulfonyl)Imide. J. Chem. Eng. Data 2007, Properties of the Ternary Mixtures Ethanol and 1-Propanol + Water +
52, 1080−1085. 1-Ethyl-3-Methylpyridinium Ethylsulfate at 298.15 K. J. Chem. Eng.
(50) Nieto de Castro, C. A.; Langa, E.; Morais, A. L.; Lopes, M. L. Data 2009, 54, 2229−2234.
M.; Lourenço, M. J. V.; Santos, F. J. V.; Santos, M. S. C. S.; Lopes, J. N. (66) González, E. J.; Calvar, N.; González, B.; Domínguez, Á . (Liquid
C.; Veiga, H. I. M.; Macatrão, M.; et al. Studies on the Density, Heat +Liquid) Equilibria for Ternary Mixtures of (Alkane+Benzene
Capacity, Surface Tension and Infinite Dilution Diffusion with the +[EMpy] [ESO4]) at Several Temperatures and Atmospheric
Ionic Liquids [C4mim][NTf2], [C4mim][dca], [C2mim][EtOSO3] Pressure. J. Chem. Thermodyn. 2009, 41, 1215−1221.
and [Aliquat][dca]. Fluid Phase Equilib. 2010, 294, 157−179. (67) González, B.; Corderí, S.; Santamaría, A. G. Application of 1-
(51) Krummen, M.; Wasserscheid, P.; Gmehling, J. Measurement of Alkyl-3-Methylpyridinium Bis(Trifluoromethylsulfonyl)Imide Ionic
Activity Coefficients at Infinite Dilution in Ionic Liquids Using the Liquids for the Ethanol Removal from Its Mixtures with Alkanes. J.
Dilutor Technique. J. Chem. Eng. Data 2002, 47, 1411−1417. Chem. Thermodyn. 2013, 60, 9−14.
(52) Vranes, M.; Dozic, S.; Djeric, V.; Gadzuric, S. Physicochemical (68) Gómez, E.; González, B.; Domínguez, Á .; Tojo, E.; Tojo, J.
Characterization of 1-Butyl-3-Methylimidazolium and 1-Butyl-1- Dynamic Viscosities of a Series of 1-Alkyl-3-Methylimidazolium
Methylpyrrolidinium Bis(Trifluoromethylsulfonyl)Imide. J. Chem. Chloride Ionic Liquids and Their Binary Mixtures with Water at
Eng. Data 2012, 57, 1072−1077. Several Temperatures. J. Chem. Eng. Data 2006, 51, 696−701.
(53) Troncoso, J.; Cerdeiriña, C. A.; Sanmamed, Y. A.; Romaní, L.; (69) He, R.-H.; Long, B.-W.; Lu, Y.-Z.; Meng, H.; Li, C.-X. Solubility
Rebelo, L. P. N. Thermodynamic Properties of Imidazolium-Based of Hydrogen Chloride in Three 1-Alkyl-3-Methylimidazolium
Ionic Liquids: Densities, Heat Capacities, and Enthalpies of Fusion of Chloride Ionic Liquids in the Pressure Range (0 to 100) kPa and
[Bmim][PF6] and [Bmim][NTf2]. J. Chem. Eng. Data 2006, 51, Temperature Range (298.15 to 363.15) K. J. Chem. Eng. Data 2012,
1856−1859. 57, 2936−2941.
(54) Freire, M. G.; Teles, A. R. R.; Rocha, M. A. A.; Schröder, B.; (70) Kenneth, R. S.; Annegret, S.; María-José, T. Viscosity and
Neves, C. M. S. S.; Carvalho, P. J.; Evtuguin, D. V.; Santos, L. M. N. B. Density of 1-Alkyl-3-Methylimidazolium Ionic Liquids. Clean Solvents;
F.; Coutinho, J. A. P. Thermophysical Characterization of Ionic American Chemical Society: Washington, DC, 2002.
Liquids Able to Dissolve Biomass. J. Chem. Eng. Data 2011, 56, 4813− (71) Huddleston, J. G.; Visser, A. E.; Reichert, W. M.; Willauer, H.
4822. D.; Broker, G. A.; Rogers, R. D. Characterization and Comparison of
(55) Fröba, A. P.; Rausch, M. H.; Krzeminski, K.; Assenbaum, D.; Hydrophilic and Hydrophobic Room Temperature Ionic Liquids
Wasserscheid, P.; Leipertz, A. Thermal Conductivity of Ionic Liquids: Incorporating the Imidazolium Cation. Green Chem. 2001, 3, 156−
Measurement and Prediction. Int. J. Thermophys. 2010, 31, 2059− 164.
2077. (72) Widegren, J. A.; Magee, J. W. Density, Viscosity, Speed of
(56) Rosenboom, J.-G.; Afzal, W.; Prausnitz, J. M. Solubilities of Sound, and Electrolytic Conductivity for the Ionic Liquid 1-Hexyl-3-
Some Organic Solutes in 1-Ethyl-3-Methylimidazolium Acetate. Methylimidazolium Bis(Trifluoromethylsulfonyl)Imide and Its Mix-
Chromatographic Measurements and Predictions from COSMO-RS. tures with Water. J. Chem. Eng. Data 2007, 52, 2331−2338.
J. Chem. Thermodyn. 2012, 47, 320−327. (73) Kato, R.; Gmehling, J. Systems with Ionic Liquids: Measure-
(57) Schmidt, H.; Stephan, M.; Safarov, J.; Kul, I.; Nocke, J.; ment of VLE and Γ∞ Data and Prediction of Their Thermodynamic
Abdulagatov, I. M.; Hassel, E. Experimental Study of the Density and Behavior Using Original UNIFAC, Mod. UNIFAC(Do) and
Viscosity of 1-Ethyl-3-Methylimidazolium Ethyl Sulfate. J. Chem. COSMO-RS(Ol). J. Chem. Thermodyn. 2005, 37, 603−619.
Thermodyn. 2012, 47, 68−75. (74) Muhammad, A.; Abdul Mutalib, M. I.; Wilfred, C. D.;
(58) Gaciño, F. M.; Regueira, T.; Lugo, L.; Comuñas, M. J. P.; Murugesan, T.; Shafeeq, A. Thermophysical Properties of 1-Hexyl-3-
Fernández, J. Influence of Molecular Structure on Densities and Methyl Imidazolium Based Ionic Liquids with Tetrafluoroborate,
Viscosities of Several Ionic Liquids. J. Chem. Eng. Data 2011, 56, Hexafluorophosphate and Bis(Trifluoromethylsulfonyl)Imide Anions.
4984−4999. J. Chem. Thermodyn. 2008, 40, 1433−1438.
(59) Seki, S.; Tsuzuki, S.; Hayamizu, K.; Umebayashi, Y.; Serizawa, (75) McHale, G.; Hardacre, C.; Ge, R.; Doy, N.; Allen, R. W. K.;
N.; Takei, K.; Miyashiro, H. Comprehensive Refractive Index Property MacInnes, J. M.; Bown, M. R.; Newton, M. I. Density−Viscosity
for Room-Temperature Ionic Liquids. J. Chem. Eng. Data 2012, 57, Product of Small-Volume Ionic Liquid Samples Using Quartz Crystal
2211−2216. Impedance Analysis. Anal. Chem. 2008, 80, 5806−5811.
(60) Gómez, E.; González, B.; Calvar, N.; Tojo, E.; Domínguez, Á . (76) Oliveira, F. S.; Freire, M. G.; Carvalho, P. J.; Coutinho, J. o. A.
Physical Properties of Pure 1-Ethyl-3-Methylimidazolium Ethylsulfate P.; Lopes, J. N. C.; Rebelo, L. s. P. N.; Marrucho, I. M. Structural and
and Its Binary Mixtures with Ethanol and Water at Several Positional Isomerism Influence in the Physical Properties of
Temperatures. J. Chem. Eng. Data 2006, 51, 2096−2102. Pyridinium NTf2-Based Ionic Liquids: Pure and Water-Saturated
(61) Vercher, E.; Llopis, F. J.; González-Alfaro, V.; Miguel, P. J.; Mixtures. J. Chem. Eng. Data 2010, 55, 4514−4520.
Martínez-Andreu, A. Refractive Indices and Deviations in Refractive (77) Liu, Q.-S.; Yang, M.; Li, P.-P.; Sun, S.-S.; Welz-Biermann, U.;
Indices of Trifluoromethanesulfonate-Based Ionic Liquids in Water. J. Tan, Z.-C.; Zhang, Q.-G. Physicochemical Properties of Ionic Liquids
Chem. Eng. Data 2011, 56, 4499−4504. [C3py][NTf2] and [C6py][NTf2]. J. Chem. Eng. Data 2011, 56, 4094−
(62) Olivier, E.; Letcher, T. M.; Naidoo, P.; Ramjugernath, D. 4101.
Activity Coefficients at Infinite Dilution of Organic Solutes in the Ionic (78) Arce, A.; Rodríguez, O.; Soto, A. Experimental Determination of
Liquid 1-Ethyl-3-Methylimidazolium Trifluoromethanesulfonate Using Liquid−Liquid Equilibrium Using Ionic Liquids: Tert-Amyl Ethyl

5893 DOI: 10.1021/acs.jpcb.5b00689

J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

Ether + Ethanol + 1-Octyl-3-Methylimidazolium Chloride System at Matter? Structure−Property Relationships in Thermochemistry of

298.15 K. J. Chem. Eng. Data 2004, 49, 514−517. Ionic Liquids. Thermochim. Acta 2013, 562, 84−95.
(79) David, W.; Letcher, T. M.; Ramjugernath, D.; David Raal, J. (98) Blokhin, A. V.; Paulechka, Y. U.; Kabo, G. J. Thermodynamic
Activity Coefficients of Hydrocarbon Solutes at Infinite Dilution in the Properties of [C6mim][NTf2] in the Condensed State. J. Chem. Eng.
Ionic Liquid, 1-Methyl-3-Octyl-Imidazolium Chloride from Gas− Data 2006, 51, 1377−1388.
Liquid Chromatography. J. Chem. Thermodyn. 2003, 35, 1335−1341. (99) Shimizu, Y.; Ohte, Y.; Yamamura, Y.; Saito, K.; Atake, T. Low-
(80) Nebig, S.; Gmehling, J. Measurements of Different Thermody- Temperature Heat Capacity of Room-Temperature Ionic Liquid, 1-
namic Properties of Systems Containing Ionic Liquids and Correlation Hexyl-3-Methylimidazolium Bis(Trifluoromethylsulfonyl)Imide. J.
of These Properties Using Modified UNIFAC (Dortmund). Fluid Phys. Chem. B 2006, 110, 13970−13975.
Phase Equilib. 2010, 294, 206−212. (100) Maginn, E. J. Atomistic Simulation of the Thermodynamic and
(81) Papaiconomou, N.; Yakelis, N.; Salminen, J.; Bergman, R.; Transport Properties of Ionic Liquids. Acc. Chem. Res. 2007, 40, 1200−
Prausnitz, J. M. Synthesis and Properties of Seven Ionic Liquids 1207.
Containing 1-Methyl-3-Octylimidazolium or 1-Butyl-4-Methylpyridi- (101) Cadena, C.; Zhao, Q.; Snurr, R. Q.; Maginn, E. J. Molecular
nium Cations. J. Chem. Eng. Data 2006, 51, 1389−1393. Modeling and Experimental Studies of the Thermodynamic and
(82) Alonso, L.; Arce, A.; Francisco, M.; Soto, A. (Liquid + Liquid) Transport Properties of Pyridinium-Based Ionic Liquids. J. Phys. Chem.
Equilibria of [C8mim][NTf2] Ionic Liquid with a Sulfur-Component B 2006, 110, 2821−2832.
and Hydrocarbons. J. Chem. Thermodyn. 2008, 40, 265−270. (102) Earle, M. J.; Esperanca, J. M. S. S.; Gilea, M. A.; Canongia
(83) Tokuda, H.; Tsuzuki, S.; Susan, M. A. B. H.; Hayamizu, K.; Lopes, J. N.; Rebelo, L. P. N.; Magee, J. W.; Seddon, K. R.; Widegren,
Watanabe, M. How Ionic Are Room-Temperature Ionic Liquids? An J. A. The Distillation and Volatility of Ionic Liquids. Nature 2006, 439,
Indicator of the Physicochemical Properties. J. Phys. Chem. B 2006, 831−834.
110, 19593−19600. (103) Santos, L. M. N. B. F.; Canongia Lopes, J. N.; Coutinho, J. A.
(84) Liu, Q.-S.; Yang, M.; Yan, P.-F.; Liu, X.-M.; Tan, Z.-C.; Welz- P.; Esperança, J. M. S. S.; Gomes, L. R.; Marrucho, I. M.; Rebelo, L. P.
Biermann, U. Density and Surface Tension of Ionic Liquids N. Ionic Liquids: First Direct Determination of Their Cohesive
[Cnpy][NTf2] (n = 2, 4, 5). J. Chem. Eng. Data 2010, 55, 4928−4930. Energy. J. Am. Chem. Soc. 2006, 129, 284−285.
(85) Sambasivarao, S. V.; Acevedo, O. Development of OPLS-AA (104) Kelkar, M. S.; Maginn, E. J. Calculating the Enthalpy of
Force Field Parameters for 68 Unique Ionic Liquids. J. Chem. Theory Vaporization for Ionic Liquid Clusters. J. Phys. Chem. B 2007, 111,
Comput. 2009, 5, 1038−1050. 9424−9427.
(86) Damm, W.; Frontera, A.; Tirado−Rives, J.; Jorgensen, W. L. (105) Armstrong, J. P.; Hurst, C.; Jones, R. G.; Licence, P.; Lovelock,
OPLS All-Atom Force Field for Carbohydrates. J. Comput. Chem. K. R. J.; Satterley, C. J.; Villar-Garcia, I. J. Vapourisation of Ionic
1997, 18, 1955−1970. Liquids. Phys. Chem. Chem. Phys. 2007, 9, 982−990.
(87) Strechan, A. A.; Paulechka, Y. U.; Blokhin, A. V.; Kabo, G. J. (106) Zaitsau, D. H.; Kabo, G. J.; Strechan, A. A.; Paulechka, Y. U.;
Tschersich, A.; Verevkin, S. P.; Heintz, A. Experimental Vapor
Low-Temperature Heat Capacity of Hydrophilic Ionic Liquids
Pressures of 1-Alkyl-3-Methylimidazolium Bis-
[BMIM][CF3COO] and [BMIM][CH3COO] and a Correlation
(Trifluoromethylsulfonyl)Imides and a Correlation Scheme for
Scheme for Estimation of Heat Capacity of Ionic Liquids. J. Chem.
Estimation of Vaporization Enthalpies of Ionic Liquids. J. Phys.
Thermodyn. 2008, 40, 632−639.
Chem. A 2006, 110, 7303−7306.
(88) Rocha, M. A. A.; Bastos, M.; Coutinho, J. A. P.; Santos, L. M. N.
(107) Zaitsau, D.; Yermalayeu, A.; Emel’yanenko, V.; Verevkin, S.;
B. F. Heat Capacities at 298.15 K of the Extended [Cnc1im][Ntf2]
Welz-Biermann, U.; Schubert, T. Structure-Property Relationships in
Ionic Liquid Series. J. Chem. Thermodyn. 2012, 53, 140−143. Ils: A Study of the Alkyl Chain Length Dependence in Vaporisation
(89) Shimizu, Y.; Ohte, Y.; Yamamura, Y.; Saito, K. Effects of Enthalpies of Pyridinium Based Ionic Liquids. Sci. China: Chem. 2012,
Thermal History on Thermal Anomaly in Solid of Ionic Liquid 55, 1525−1531.
Compound, [C4mim][Tf2N]. Chem. Lett. 2007, 36, 1484−1485. (108) Köddermann, T.; Paschek, D.; Ludwig, R. Molecular Dynamic
(90) Zhang, Z.-H.; Tan, Z.-C.; Sun, L.-X.; Jia-Zhen, Y.; Lv, X.-C.; Shi, Simulations of Ionic Liquids: A Reliable Description of Structure,
Q. Thermodynamic Investigation of Room Temperature Ionic Liquid: Thermodynamics and Dynamics. ChemPhysChem 2007, 8, 2464−
The Heat Capacity and Standard Enthalpy of Formation of EMIES. 2470.
Thermochim. Acta 2006, 447, 141−146. (109) Tokuda, H.; Hayamizu, K.; Ishii, K.; Susan, M. A. B. H.;
(91) Ge, R.; Hardacre, C.; Jacquemin, J.; Nancarrow, P.; Rooney, D. Watanabe, M. Physicochemical Properties and Structures of Room
W. Heat Capacities of Ionic Liquids as a Function of Temperature at Temperature Ionic Liquids. 2. Variation of Alkyl Chain Length in
0.1 MPa. Measurement and Prediction. J. Chem. Eng. Data 2008, 53, Imidazolium Cation. J. Phys. Chem. B 2005, 109, 6103−6110.
2148−2153. (110) Canongia Lopes, J. N.; Deschamps, J.; Pádua, A. A. H.
(92) García-Miaja, G.; Troncoso, J.; Romaní, L. Excess Enthalpy, Modeling Ionic Liquids Using a Systematic All-Atom Force Field. J.
Density, and Heat Capacity for Binary Systems of Alkylimidazolium- Phys. Chem. B 2004, 108, 2038−2047.
Based Ionic Liquids + Water. J. Chem. Thermodyn. 2009, 41, 161−166. (111) Tsuzuki, S.; Shinoda, W.; Saito, H.; Mikami, M.; Tokuda, H.;
(93) Paulechka, Y. U.; Blokhin, A. V.; Kabo, G. J.; Strechan, A. A. Watanabe, M. Molecular Dynamics Simulations of Ionic Liquids:
Thermodynamic Properties and Polymorphism of 1-Alkyl-3-Methyl- Cation and Anion Dependence of Self-Diffusion Coefficients of Ions. J.
imidazolium Bis(Triflamides). J. Chem. Thermodyn. 2007, 39, 866− Phys. Chem. B 2009, 113, 10641−10649.
877. (112) Hess, B. Determining the Shear Viscosity of Model Liquids
(94) Waliszewski, D.; Stępniak, I.; Piekarski, H.; Lewandowski, A. from Molecular Dynamics Simulations. J. Chem. Phys. 2002, 116, 209−
Heat Capacities of Ionic Liquids and Their Heats of Solution in 217.
Molecular Liquids. Thermochim. Acta 2005, 433, 149−152. (113) Hu, Z.; Margulis, C. J. On the Response of an Ionic Liquid to
(95) Crosthwaite, J. M.; Muldoon, M. J.; Dixon, J. K.; Anderson, J. L.; External Perturbations and the Calculation of Shear Viscosity. J. Phys.
Brennecke, J. F. Phase Transition and Decomposition Temperatures, Chem. B 2007, 111, 4705−4714.
Heat Capacities and Viscosities of Pyridinium Ionic Liquids. J. Chem. (114) Fendt, S.; Padmanabhan, S.; Blanch, H. W.; Prausnitz, J. M.
Thermodyn. 2005, 37, 559−568. Viscosities of Acetate or Chloride-Based Ionic Liquids and Some of
(96) Calvar, N.; Gómez, E.; Macedo, E. A.; Domínguez, Á . Thermal Their Mixtures with Water or Other Common Solvents. J. Chem. Eng.
Analysis and Heat Capacities of Pyridinium and Imidazolium Ionic Data 2010, 56, 31−34.
Liquids. Thermochim. Acta 2013, 565, 178−182. (115) Miran Beigi, A. A.; Abdouss, M.; Yousefi, M.; Pourmortazavi, S.
(97) Verevkin, S. P.; Zaitsau, D. H.; Emel’yanenko, V. N.; Ralys, R. M.; Vahid, A. Investigation on Physical and Electrochemical Properties
V.; Yermalayeu, A. V.; Schick, C. Does Alkyl Chain Length Really of Three Imidazolium Based Ionic Liquids (1-Hexyl-3-Methylimida-

5894 DOI: 10.1021/acs.jpcb.5b00689

J. Phys. Chem. B 2015, 119, 5882−5895
The Journal of Physical Chemistry B Article

zolium Tetrafluoroborate, 1-Ethyl-3-Methylimidazolium Bis-

(Trifluoromethylsulfonyl) Imide and 1-Butyl-3-Methylimidazolium
Methylsulfate). J. Mol. Liq. 2013, 177, 361−368.
(116) Andreatta, A. E.; Francisco, M.; Rodil, E.; Soto, A.; Arce, A.
Isobaric Vapour−Liquid Equilibria and Physical Properties for
Isopropyl Acetate + Isopropanol + 1-Butyl-3-Methyl-Imidazolium
Bis(Trifluoromethylsulfonyl)Imide Mixtures. Fluid Phase Equilib.
2011, 300, 162−171.
(117) Quijada-Maldonado, E.; van der Boogaart, S.; Lijbers, J. H.;
Meindersma, G. W.; de Haan, A. B. Experimental Densities, Dynamic
Viscosities and Surface Tensions of the Ionic Liquids Series 1-Ethyl-3-
Methylimidazolium Acetate and Dicyanamide and Their Binary and
Ternary Mixtures with Water and Ethanol at T = (298.15 to 343.15
K). J. Chem. Thermodyn. 2012, 51, 51−58.
(118) Arce, A.; Rodriguez, H.; Soto, A. Use of a Green and Cheap
Ionic Liquid to Purify Gasoline Octane Boosters. Green Chem. 2007, 9,
247−253.
(119) Calvar, N.; Gómez, E.; González, B.; Domínguez, Á .
Experimental Determination, Correlation, and Prediction of Physical
Properties of the Ternary Mixtures Ethanol + Water with 1-Octyl-3-
Methylimidazolium Chloride and 1-Ethyl-3-Methylimidazolium Ethyl-
sulfate. J. Chem. Eng. Data 2007, 52, 2529−2535.
(120) González, B.; Calvar, N.; González, E.; Domínguez, Á . Density
and Viscosity Experimental Data of the Ternary Mixtures 1-Propanol
or 2-Propanol + Water + 1-Ethyl-3-Methylimidazolium Ethylsulfate.
Correlation and Prediction of Physical Properties of the Ternary
Systems. J. Chem. Eng. Data 2008, 53, 881−887.
(121) Rodríguez, H.; Brennecke, J. F. Temperature and Composition
Dependence of the Density and Viscosity of Binary Mixtures of Water
+ Ionic Liquid. J. Chem. Eng. Data 2006, 51, 2145−2155.
(122) Schreiner, C.; Zugmann, S.; Hartl, R.; Gores, H. J. Fractional
Walden Rule for Ionic Liquids: Examples from Recent Measurements
and a Critique of the So-Called Ideal KCl Line for the Walden Plot. J.
Chem. Eng. Data 2009, 55, 1784−1788.
(123) Yao, H.; Zhang, S.; Wang, J.; Zhou, Q.; Dong, H.; Zhang, X.
Densities and Viscosities of the Binary Mixtures of 1-Ethyl-3-
Methylimidazolium Bis(Trifluoromethylsulfonyl)Imide with N-Meth-
yl-2-Pyrrolidone or Ethanol at T = (293.15 to 323.15) K. J. Chem. Eng.
Data 2012, 57, 875−881.
(124) Kandil, M. E.; Marsh, K. N.; Goodwin, A. R. H. Measurement
of the Viscosity, Density, and Electrical Conductivity of 1-Hexyl-3-
Methylimidazolium Bis(Trifluorosulfonyl)Imide at Temperatures
between (288 and 433) K and Pressures Below 50 MPa. J. Chem.
Eng. Data 2007, 52, 2382−2387.
(125) Ahosseini, A.; Scurto, A. Viscosity of Imidazolium-Based Ionic
Liquids at Elevated Pressures: Cation and Anion Effects. Int. J.
Thermophys. 2008, 29, 1222−1243.
(126) González, E. J.; Alonso, L.; Domínguez, Á . Physical Properties
of Binary Mixtures of the Ionic Liquid 1-Methyl-3-Octylimidazolium
Chloride with Methanol, Ethanol, and 1-Propanol at T = (298.15,
313.15, and 328.15) K and at P = 0.1 MPa. J. Chem. Eng. Data 2006,
51, 1446−1452.
(127) Alonso, L.; Arce, A.; Francisco, M.; Soto, A. Liquid−Liquid
Equilibria for [C8mim][NTf2] + Thiophene + 2,2,4-Trimethylpentane
or + Toluene. J. Chem. Eng. Data 2008, 53, 1750−1755.
(128) Andreatta, A. E.; Arce, A.; Rodil, E.; Soto, A. Physical
Properties of Binary and Ternary Mixtures of Ethyl Acetate, Ethanol,
and 1-Octyl-3-Methyl-Imidazolium Bis(Trifluoromethylsulfonyl)Imide
at 298.15 K. J. Chem. Eng. Data 2009, 54, 1022−1028.
(129) Fendt, S.; Padmanabhan, S.; Blanch, H. W.; Prausnitz, J. M.
Viscosities of Acetate or Chloride-Based Ionic Liquids and Some of
Their Mixtures with Water or Other Common Solvents. J. Chem. Eng.
Data 2011, 56, 31−34.

5895 DOI: 10.1021/acs.jpcb.5b00689

J. Phys. Chem. B 2015, 119, 5882−5895