Water solubility and dynamics of CO2 capture ionic liquids having aprotic heterocyclic anions

Fluid Phase Equilibria 368 (2014) 72–79

Contents lists available at ScienceDirect

Fluid Phase Equilibria journal homepage: www.elsevier.com/locate/fluid

Water solubility and dynamics of CO2 capture ionic liquids having aprotic heterocyclic anions Hao Wu, Edward J. Maginn ∗ Department of Chemical and Biomolecular Engineering, University of Notre Dame, Notre Dame, IN 46556, USA

a r t i c l e

i n f o

Article history: Received 5 December 2013 Received in revised form 2 February 2014 Accepted 4 February 2014 Available online 12 February 2014 Keywords: CO2 capture Water solubility Ionic liquids Molecular dynamics Simulation

a b s t r a c t A computational study was carried out to investigate the solubility and dynamics of water in five different ionic liquids capable of chemically reacting with CO2 . All the ionic liquids have a common tetrabutylphosphonium cation paired with five different aprotic heterocyclic anions. These ionic liquids have properties that make them attractive candidates for use in CO2 capture applications, but the impact of water on their properties is unknown. The simulations show that the ionic liquid having a 2-cyanopyrrolide anion is the most hydrophobic of all the liquids studied, but that upon reaction with CO2 it becomes much more hydrophilic. The other ionic liquids investigated show little change in water solubility upon reaction with CO2 . Henry’s Law constants, enthalpies and entropies of water absorption, liquid structure, hydrogen bonding and self-diffusivities were computed and found to correlate well with the strength of interaction between water and the anions and the relative hydrophobicity of the liquids. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Ionic liquids have come under intense investigation over the past several years for their use as solvents for CO2 capture [1,2]. One of the major breakthroughs for the utilization of ionic liquids in post-combustion capture applications was the tethering of a reactive primary amine to a dialkylimidazolium cation [3]. Unlike conventional ionic liquids that absorb CO2 via a physical dissolution mechanism, these “task-specific” ionic liquids (TSILs) chemically react with CO2 and are thus able to function at low CO2 partial pressures. A drawback of the first generation of TSILs is that their reaction with CO2 leads to the creation of a salt bridge network, which results in an extremely viscous liquid [4]. This phenomenon makes the TSILs difficult to use in conventional gas processing equipment. In addition, TSILs with a reactive amine on the cation tend to react with an unfavorable 2:1 TSIL:CO2 stoichiometry, while it has been found that having the reactive group on the anion leads to a more efficient 1:1 stoichiometry [5]. Other work has shown that ionic liquids having anions such as acetate, which can deprotonate an imidazolium cation, enable the formation of CO2 -carbene adducts. This significantly enhances the uptake of CO2 [6–8]. Ionic liquids having aprotic heterocyclic anions (AHAs) have been developed recently and also found to be effective for post-combustion CO2 capture [9]. By carefully tuning the strength with which CO2

∗ Corresponding author. Tel.: +1 574 6315687; fax: +1 574 6318366. E-mail addresses: [email protected], [email protected] (E.J. Maginn). http://dx.doi.org/10.1016/j.fluid.2014.02.003 0378-3812/© 2014 Elsevier B.V. All rights reserved.

reacts with the anion, the capture process can be optimized in terms of maximizing the carrying capacity of the solvent and minimizing the energy required to regenerate the ionic liquid. A key property that determines the effectiveness of an ionic liquid for CO2 capture is its affinity for water. Although ionic liquids can be used in an essentially anhydrous state (unlike conventional amine-based solvents), ionic liquids will absorb water from postcombustion flue gas. Much of this water will be evaporated during the thermal regeneration process, thereby increasing the parasitic energy load. Therefore it is generally desirable to minimize the amount of water that an ionic liquid takes up under CO2 absorption conditions. The presence of dissolved water is also known to have a profound effect on the physical properties of ionic liquids, particularly dynamical and mass transfer properties. Thus it is also important to know how these dynamical properties depend upon water concentrations and the extent of the reaction with CO2 . It is difficult to experimentally control and measure the extent of the reaction of CO2 with AHA-based ionic liquids, and it is also challenging to simultaneously measure water and CO2 uptake. For these reasons, molecular dynamics (MD) simulations have been carried out to understand water solubility trends in different AHA-based ionic liquids, as well as to probe how water solubility varies with the extent of CO2 reaction. MD simulations were also carried out to investigate how the dynamics of water and the ionic liquids depend upon the nature of the ionic liquids and their reaction with CO2 . Each ionic liquid investigated had the same tetrabutylphosphonium cation, but were paired with five distinct anions: 2cyanopyrrolide, benzo[d]imidazolide, 6-bromo-benzimidazolide,

H. Wu, E.J. Maginn / Fluid Phase Equilibria 368 (2014) 72–79 Table 1 Names and abbreviations for the cation and anions studied in this work. Name

Abbreviation

CO2 complex

Tetrabutylphosphonium 2-Cyanopyrrolide Benzo[d]imidazolide 6-Bromo-benzimidazolide 2-(Trifluoromethyl)pyrazolide Imidazolide

[P4444 ]+ [2-CNpyr]− [BnIm]− [BrBnIm]− [CF3 pyra]− [Im]−

– [2-CNpyr-CO2 ]− [BnIm-CO2 ]− [BrBnIm-CO2 ]− [CF3 pyra-CO2 ]− [Im-CO2 ]−

2-(trifluoromethyl)pyrazolide, and imidazolide. The CO2 -reacted forms of these five ionic liquids were also studied. The names and abbreviations for the cation and anions are listed in Table 1. Allatom molecular structures are presented in Fig. 1.

A standard “class I” force field of the following form was used to represent the atomic interactions. Vtot =

2

kr (r − r0 ) +

bonds

+





k ( − 0 )

2

angles

k [1 + cos(n − )]

torsions N−1 N  

+

i=1 j>i

 4εij

(NVT) simulation was run. Statistics were collected over the final 30 ns of the NVT simulation. Temperature was maintained at the desired value using a Nosé-Hoover thermostat with a damping factor of 100 fs. Pressure was maintained at 1 atm for all NPT simulations with a Parinello-Rahman barostat. The barostat damping factor was set to 700 fs. A particle–particle–particle mesh method was used for the long-range electrostatics with a precision of 10−4 . ˚ Nonbonded interactions were truncated with a cutoff of 12 A. The intramolecular Lennard–Jones and electrostatic interactions for atoms separated by three consecutive bonds were scaled by a factor of 0.5 and 0.8333, respectively. For each state point, the uncertainty of the computed properties was estimated as the standard deviation from three independent simulations. 3. Results and discussion

2. Simulation details



73



ij rij

12

 −

ij rij

6 

 +

qi qj 4ε0 rij

3.1. Density and molar volume Computed densities and molar volumes for the ten different ionic liquids are listed in Table 2. The computed densities are all on the order of 1 g/cm3 , which is similar to that of many other phosphonium-based ILs [18,19]. Based on tentative unpublished experimental data, the computed densities are likely about 3–6% too low, which is consistent with previous studies where scaled partial charges have been used to obtain better estimates of dynamical properties [20].

(1)

Most of the intramolecular parameters and Lennard–Jones parameters were taken directly from the Generalized Amber Force Field (GAFF) [10]. Ab initio gas-phase calculations were performed using the Gaussian 09 package at the B3LYP/6–311 +G(d,p) level of theory [11]. Some of the GAFF parameters were then adjusted to better match calculated ab initio gas phase structures. Atomic point partial charges were calculated from the minimized structures using the restrained electrostatic potential (RESP) approach [12]. As is often done to approximate charge transfer, the charges were scaled uniformly so that the cations and anions had a net charge of ±0.9 [13]. Lennard–Jones parameters for unlike atoms εij and  ij were obtained using Lorentz–Berthelot combining rules [14]. Water was modeled using a flexible form of the simple point charge (SPC) force field [15]. A complete listing of all the parameters is provided in the Supporting Information. All simulations were performed with the GROMACS 4.5 package [16]. Water solubility was estimated at 333 K, 373 K and 413 K and 1 bar total pressure for the unreacted and reacted ionic liquids by computing the solvation free energy of a single water molecule in a system consisting of roughly 150 ionic liquid ion pairs. The Bennett acceptance ratio (BAR) method [17] was used for the free energy calculations. Isothermal-isobaric (NPT) MD simulations were carried out using a 5 ns equilibration run followed by a 5 ns production run during which averages were taken. For each state point, three independent simulations were conducted to obtain an estimate of the uncertainty of the computed properties from the standard deviation. Additional simulations were carried out with and without water for two of the ionic liquids in the reacted and unreacted states ([P4444]− paired with [2-CNpyr]− , [2-CNpyr-CO2 ]− , [BnIm]− , and [BnIm-CO2 ]− ). Self-diffusivities, radial and spatial distribution functions and hydrogen bonding trends were calculated. The other ionic liquids in this study were also simulated in the NPT ensemble for 10 ns in order to obtain densities. For the calculations with water, 30 water molecules were added to the ionic liquids to yield a water mole fraction of x = 0.17. The systems were equilibrated in the NPT ensemble for 10 ns, after which a 35 ns canonical ensemble

3.2. Water solubility Henry’s Law constants for water in the ionic liquids, kH , were calculated from the solvation free energy of water going from an ideal gas state to being dissolved in the ionic liquid via the following expression kH (T, p) = kB T(T, p) exp

G

kB T

(2)

where  is the liquid density and kB is the Boltzmann constant. The solvation free energies of water were obtained using BAR, as implemented in Gromacs [16,17]. During the calculations, the intermolecular interaction between a single water molecule in the ideal gas state (state “0”) and the dissolved state (state “1”) was scaled by a factor which ranges from 0 to 1 U( ) = U0 (1 − ) + U1

(3)

In this work, a two-step approach was used in which the Lennard–Jones interactions and the Coulombic interactions were scaled separately. For both types of interactions, an increment of 0.05 was applied to ranging between 0 and 1. “Soft core” Lennard–Jones interactions were used to ensure that there were no numerical instabilities near = 0 [21]. Computed Henry’s Law constants for the five different ionic liquids in the unreacted and CO2 -reacted states at three temperatures are listed in Table 3. Note that a large value of kH corresponds to low water solubility and a higher degree of hydrophobicity. Given that all the cations are identical ([P4444 ]+ ), among the unreacted ILs [2-CNpyr]− is the most hydrophobic anion while [BnIm]− is the most hydrophilic anion. [BrBnIm]− , [CF3 pyra]− and [Im]− have similar water solubilities in the unreacted state in between that of [2-CNpyr]− and [BnIm]− . The reacted [2-CNpyr-CO2 ]− anion has a much smaller Henry’s Law constant than the unreacted [2-CNpyr]− anion, indicating that the reaction causes this anion to become much more hydrophilic. [CF3 pyra]− also shows an increase in water solubility upon the reaction with CO2 to form [CF3 pyra-CO2 ]− , but the change is not as large as that for [2-CNpyr-CO2 ]− . In contrast, [BnIm]− [BrBnIm]− and [Im]− exhibit very little change in water solubility upon reaction with CO2 .

74

H. Wu, E.J. Maginn / Fluid Phase Equilibria 368 (2014) 72–79

Fig. 1. All-atom molecular structures of the cation and anions studied in this work. The atom colors are as follows: hydrogen (white), carbon (gray), phosphorous (orange), nitrogen (blue), oxygen (red), bromine (dark red) and fluorine (turquoise). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Table 2 Computed liquid densities (g/cm3 ) and molar Vm = volumes (cm3 /mol) for the ionic liquids as a function of temperature. The numbers in parenthesis correspond to the uncertainty in the final digits, obtained from the standard deviation of three independent runs. Anion

T = 333 K

2-CNpyr 2-CNpyr-CO2 BnIm BnIm-CO2 BrBnIm BrBnIm-CO2 CF3 Pyra CF3 Pyra-CO2 Im Im-CO2

T = 373 K

T = 413 K



Vm



Vm



Vm

0.8955(3) 0.9581(2) 0.9209(4) 0.9811(3) 1.074(5) 1.121(6) 0.9691(3) 1.020(4) 0.8887(2) 0.9499(3)

390.83(10) 411.24(08) 408.30(17) 428.08(11) 423.52(24) 445.18(37) 406.56(19) 429.30(20) 366.84(05) 389.52(10)

0.8699(2) 0.9326(2) 0.8967(3) 0.9563(3) 1.045(4) 1.094(4) 0.9406(3) 0.9941(1) 0.8636(2) 0.9249(3)

402.33(04) 422.47(09) 419.32(10) 439.19(11) 435.47(18) 456.05(19) 418.88(14) 440.60(03) 377.50(08) 400.03(11)

0.8447(3) 0.9080(2) 0.8730(1) 0.9313(2) 1.018(4) 1.067(3) 0.9128(2) 0.9639(5) 0.8387(2) 0.8985(2)

414.36(10) 433.91(08) 430.70(05) 450.99(06) 447.09(14) 467.78(12) 431.64(03) 454.40(21) 388.71(09) 411.78(06)

3.3. Enthalpy and entropy

and the entropy was computed as

The enthalpy of solvation was determined via the following expression

S =

 H = R

∂ ln kH ∂(1/T )

 (4) p

( H − G) T

(5)

where G was obtained from BAR calculations, as described above. Table 4 shows the computed enthalpies and entropies of absorption.

H. Wu, E.J. Maginn / Fluid Phase Equilibria 368 (2014) 72–79

75

Fig. 2. Mean-squared displacements (A˚ 2 ) vs time (ns) for COMs of cations, anions and water at 413 K: (a) [P4444 ]+ [2-CNpyr]− and [P4444 ]+ [2-CNpyr-CO2 ]− (b) [P4444 ]+ [BnIm]− and [P4444 ]+ [BnIm-CO2 ]− (c) water in [P4444 ]+ [2-CNpyr]− /[P4444 ]+ [2-CNpyr-CO2 ]− and in [P4444 ]+ [BnIm]− /[P4444 ]+ [BnIm-CO2 ]− .

Table 3 Computed Henry’s Law constants (kH , bar) for water in the ionic liquids. Anion

T = 333 K

2-CNpyr 2-CNpyr-CO2 BnIm BnIm-CO2 BrBnIm BrBnIm-CO2 CF3 Pyra CF3 Pyra-CO2 Im Im-CO2

0.323 0.050 0.0103 0.0093 0.006 0.011 0.090 0.056 0.021 0.011

± ± ± ± ± ± ± ± ± ±

T = 373 K 0.019 0.007 0.0014 0.0009 0.001 0.001 0.005 0.005 0.002 0.002

1.288 0.251 0.107 0.112 0.101 0.107 0.558 0.326 0.123 0.089

± ± ± ± ± ± ± ± ± ±

T = 413 K 0.096 0.021 0.010 0.010 0.014 0.010 0.029 0.030 0.006 0.007

3.255 1.274 0.496 0.496 0.670 0.619 1.513 1.178 0.517 0.512

± ± ± ± ± ± ± ± ± ±

0.131 0.058 0.018 0.019 0.025 0.018 0.041 0.047 0.015 0.033

These results show that the enthalpy contributes the most to the solvation free energy, with values of T S being a little more than half that of H. As expected, the solvation enthalpy is negative, meaning that at a given pressure the solubility increases as temperature decreases. Given that the enthalpy of vaporization of water is on the order of 40 kJ/mol, the results indicate that in almost all cases, water interacts more strongly with the ionic liquids than it does with itself in the liquid phase. 3.4. Self-diffusivity To investigate how the dynamics of these systems depend upon water and the reaction with CO2 , additional MD simulations

76

H. Wu, E.J. Maginn / Fluid Phase Equilibria 368 (2014) 72–79

Table 4 Average enthalpy of solvation ( H) and entropy of solvation ( S) for water in ionic liquids from 333 K to 413 K. Anion

H (kJ/mol)

S (J/(mol K))

2-CNpyr 2-CNpyr-CO2 BnIm BnIm-CO2 BrBnIm BrBnIm-CO2 CF3 Pyra CF3 Pyra-CO2 Im Im-CO2

−35.1 −49.5 −55.7 −57.2 −71.4 −58.3 −40.7 −43.6 −45.6 −55.2

−52.4 −79.5 −87.1 −90.4 −127.5 −96.0 −59.5 −65.4 −62.5 −84.4

± ± ± ± ± ± ± ± ± ±

1.3 0.8 1.8 1.4 1.9 1.4 0.8 1.3 1.4 2.4

± ± ± ± ± ± ± ± ± ±

5.6 3.7 7.9 6.2 8.2 6.2 3.6 5.8 6.2 10.4

Fig. 4. Log–log MSDs (A˚ 2 ) vs time (ns) for COMs of cations, anions and water in [P4444]+ [BnIm]− and its reacted state at 413 K. The blue dotted line has a slope of 1 for comparison. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Translational dynamics were quantified in terms of meansquare displacements (MSDs) of the centers-of-mass (COMs) of the ions. The MSD of species i at time t was calculated as MSD(t) =

 N 



[ri (t) − ri (0)]2

(6)

i

where ri (t) is the position of the COM of the ion or water molecule at time t and the angle brackets indicate an average. From this, the self-diffusivity of species i, Ds,i , may be obtained from the long time slope of the MSD via the Einstein relation Fig. 3. Log–log MSDs (A˚ 2 ) vs time (ns) for COMs of cations, anions and water in [P4444]+ [2-CNpyr]− and its reacted state at 413 K. The blue dotted line has a slope of 1 for comparison. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

were carried out for [P4444 ]+ [2-CNpyr]− , [P4444 ]+ [2-CNpyr-CO2 ]− , [P4444 ]+ [BnIm]− , and [P4444 ]+ [BnIm-CO2 ] in the neat state and with a water mole fraction of x = 0.17. For the neat simulations, 150 ionic liquid ion pairs were simulated while for the “wet” simulations, 30 water molecules were added to 150 ion pairs.

Ds =

1 d lim MSD(t). 6 t→∞ dt

(7)

Fig. 2 compares the MSDs of cations, anions, and water in the unreacted and reacted systems, while Table 5 shows the computed self-diffusion coefficients. Because the dynamics of ionic liquids tends to be so slow, great care must be taken to ensure that the system is simulated for a long enough time to ensure that diffusive motion is observed. Figs. 3 and 4 show log–log plots of the MSDs. The long time slopes all approach unity, suggesting that diffusive motion has been observed over the time scales of the simulations. The self-diffusivity of water is significantly greater than that of

Table 5 Computed self-diffusion coefficients (×1011 m2 /s) with uncertainty taken as the standard deviation from three independent simulations. All results are for 413 K. Anion

xwater

Cation

Anion

H2 O

Ds

±

Ds

±

Ds

±

2-CNpyr

0 0.17

7.85 9.54

0.15 0.52

14.54 16.15

1.54 0.74

– 60.19

– 1.20

2-CNpyr-CO2

0 0.17

6.73 6.96

0.56 0.64

10.23 10.34

0.57 0.79

– 28.43

– 1.34

BnIm

0 0.17

4.53 4.75

0.30 0.39

7.51 8.99

0.28 1.25

– 21.49

– 0.04

BnIm-CO2

0 0.17

4.32 4.51

0.27 0.34

6.66 6.83

0.62 0.74

– 16.14

– 0.36

H. Wu, E.J. Maginn / Fluid Phase Equilibria 368 (2014) 72–79

77

Fig. 6. Radial distribution functions (RDFs) for the cation–anion center-of-mass (COM) at 413 K and 1 bar. The solid black curves are for the dry ionic liquids and the red dashed curves are for the ionic liquids having a water mole fraction of 0.17. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

3.5. Radial distribution function analysis

˚ at 413 K. Fig. 5. Site–site radial distribution function g(r) vs distance (A)

the ions. This is consistent with the fact that the ions are much larger than the water molecules. The anions diffuse faster than the cations for both [P4444 ]+ [2-CNpyr]− and [P4444 ]+ [BnIm]− . In many imidazolium-based ionic liquids, the cations diffuse faster than the anions. It has been argued that this is due to the fact that the imidazolium cations have preferential motion along the plane of their rings [22–24]. For the ionic liquids investigated here, the anions have rings and diffuse faster than the globular cations. This suggests that the shape of the ions plays the main role in determining relative dynamics, not charge state. The self-diffusivities of the reacted and unreacted ionic liquids are essentially the same. This indicates that these AHA-based ionic liquids will not exhibit a viscosity increase upon reaction with CO2 [9,25]. Unlike the ions, the dynamics of water in these systems is altered by the reaction with CO2 . Water diffuses much faster in [P4444 ]+ [2-CNpyr]− than in the reacted system ([P4444 ]+ [2CNpyr-CO2 ]− ). In contrast, water has only a moderately larger self-diffusivity in unreacted [P4444 ]+ [BnIm]− relative to the reacted [P4444 ]+ [BnIm-CO2 ]− . The reason for this has to do with the relative interaction strength between water and the anions. As the association strength increases between water and the slow moving anions, the dynamics of water is reduced. This is consistent with the Henry’s Law constant results, which showed that water is much more soluble in [P4444 ]+ [2-CNpyr-CO2 ]− that in [P4444 ]+ [2CNpyr]− , suggesting a stronger interaction between water and the [2-CNpyr-CO2 ]− anion. It is also consistent with the fact that the solvation enthalpy of water in [P4444 ]+ [2-CNpyr-CO2 ]− is larger than that in [P4444 ]+ [2-CNpyr]− , but is essentially the same for [P4444 ]+ [BnIm]− and [P4444 ]+ [BnIm-CO2 ]− . The difference in water association for reacted and unreacted anions can also be seen in the organization of the liquids, as shown in the next two sections.

Fig. 5 shows site-site radial distribution functions (RDFs) for the oxygen atom of water with the reactive nitrogen atom of [2CNpyr]− and [BnIm]− , as well as with an oxygen atom on the reacted [2-CNpyr-CO2 ]− and [BnIm-CO2 ]− anions. In all cases, ˚ The water associates with the anions at distances less than 3 A. peak height for [2-CNpyr-CO2 ]− is nearly double that of [2-CNpyr]− , indicating stronger association with the reacted anion. This is consistent with the trends observed for the Henry’s Law constants,

Fig. 7. Spatial distribution functions (SDFs) for water around anions at 413 K and 1 bar. The iso-density surface was drawn at 10 times the average density.

78

H. Wu, E.J. Maginn / Fluid Phase Equilibria 368 (2014) 72–79

Fig. 8. Combined distribution function representing the hydrogen bond geometry between the N atom in unreacted anion (O atom in the reacted anion) and water.

solvation enthalpies and self-diffusivities. For [BnIm]− and [BnImCO2 ]− , however, the RDFs are similar between the reacted and unreacted systems. Again, this is consistent with the solubility, enthalpy and diffusivity trends. The strong association between [2-CNpyr-CO2 ]− and water is responsible for the reduction in the water self-diffusivity in this system. Fig. 6 shows a comparison of the RDFs for each of the ionic liquids in the neat state and with a water mole fraction of 0.17. The RDFs change very little with the addition of water, indicating that water has little effect on the overall structure of the ionic liquid. This is consistent with other simulation studies where cation-anion COM RDFs for the ionic liquid [P4444 ]+ [CH3 COO]− showed little change over water concentration from x = 0.11 to x = 0.5 [26]. By integrating the RDFs up to the first solvation shell radius, coordination numbers can be computed. Results are shown in Table 6. The coordination number for an anion around the [P4444 ]+ cation is roughly four for each system. The cation has an identical four-arm structure where the central P atom forms a tetrahedron with four alkyl chains. These chains form hydrophobic regions that

exclude the anions. The anions approach the positively charged central P atom by positioning themselves at the facets of a tetrahedron. When water is added to the system, it preferentially coordinates around the anions. For the most hydrophobic system ([P4444 ]+ [2-CNpyr]− ), the coordination number is insensitive to the Table 6 First solvation shell coordination number (CN) in the ionic liquids at 413 K and 1 bar. Anion

xwater

CN of anion for each cation

CN of H2 O for each anion

2-CNpyr

0 0.17

3.94 3.92

– 0.178

2-CNpyr-CO2

0 0.17

4.13 4.10

– 0.318

BnIm

0 0.17

3.86 3.85

– 0.369

BnIm-CO2

0 0.17

4.18 4.16

– 0.376

H. Wu, E.J. Maginn / Fluid Phase Equilibria 368 (2014) 72–79

presence of water. When CO2 reacts with this ionic liquid, however, it becomes more hydrophilic and the water coordination number increases substantially. The [BnIm]-based ionic liquid is hydrophilic in both the reacted and unreacted states, and the water coordination numbers are insensitive to the reaction with CO2 . 3.6. Spatial distribution functions Additional insight into the organization of water about the anions can be obtained by computing spatial distribution functions (SDFs). Fig. 7 shows the probability of finding water molecules in the three-dimensional space around an anion. In unreacted [2-CNpyr]− , most of the water molecules are distributed around the reactive nitrogen at the 1 position of the ring; some water molecules also localize about the N atom on the CN group, and there is also a small region between the two main sites. This suggests that water hops between the two favorable sites near the nitrogen atoms. Upon reaction with CO2 , all the water tends to aggregate around the COO group on [2-CNpyr-CO2 ]− . The distribution expands broadly at the top of the COO group. It is this association between water and the anion that leads to a reduction in water self-diffusivity. In unreacted [BnIm]− , due to the symmetric structure of the ion, water molecules evenly distribute about the two reactive N atoms. In the reacted state, water moves toward the COO group, with a small residual probability observed near the unreacted N atom. 3.7. Hydrogen bond analysis The radial distribution functions and spatial distribution functions suggest a strong interaction between water molecules and the anions. To further characterize this, a hydrogen bonding analysis was carried out using the TRAVIS package [27]. Hydrogen bonds can be identified by monitoring the distance between the hydrogen atoms of water and the nitrogen or oxygen atoms of the unreacted and reacted anion, respectively. This can also be correlated with the angle formed by the nitrogen (unreacted) or oxygen (reacted) atoms of the anion and the hydrogen and oxygen atoms of water. Fig. 8 shows a two-dimensional representation of these values. The X-axis represents the distance between the hydrogen atom of a water molecule and the nitrogen atom in an unreacted anion or the oxygen atom in a reacted anion. The Y-axis shows the angle formed by the vector connecting an N or O atom in the anion with an H atom in water (Hw) and the vector connecting an O atom in water (Ow) with the H atom in water (Hw). A very intense peak is found in the region 135–180◦ /150–200 pm in all four systems. By the geometry criterion normally adopted, this corresponds to a hydrogen bond between water and the anion [28]. In Fig. 8(a), the water-anion association is diffuse in a broad range from 100 to 1500 pm. In contrast, Fig. 8(b)–(d) all show island-like regions, which shows that water molecules are localized about the anion, indicating a stronger interaction between water and the anion than that in the unreacted [P4444 ]+ [2-CNpyr]− . This is consistent with all the results presented earlier, and suggests [P4444 ]+ [2-CNpyr]− is the most hydrophobic IL studied here. 4. Conclusions Molecular dynamics simulations were carried out to investigate the thermodynamics and transport properties of five different ionic liquids capable of reacting with CO2 . Each ionic liquid had a tetrabutylphosphonium cation paired with the following anions: 2cyanopyrrolide, benzo[d]imidazolide, 6-bromo-benzimidazolide, 2-(trifluoromethyl)pyrazolide, and imidazolide. Densities as a function of temperature were predicted for all ten ionic liquids.

79

Henry’s Law constants as well as enthalpies and entropies of absorption for water were computed for the unreacted and CO2 -reacted complexes of each ionic liquid. It is known that dissolved water has a large impact on the thermodynamic and transport properties of ionic liquids and so this type of information is crucial to the design of ionic liquid solvents. The unreacted 2-cyanopyrrolide ionic liquid had the lowest water solubility, but its reacted form had quite high water solubility. Self-diffusivities for water and the ions were computed for the 2-cyanopyrrolide and benzo[d]imidazolide ionic liquids. Water has the largest self-diffusion coefficient in [P4444 ]+ [2-CNpyr]− , because this ionic liquid is the most hydrophobic ionic liquid studied and water has the weakest association with it. In all cases, water diffuses much faster than either ion. Evidence for hydrogen bond formation between water and the anions was found. Acknowledgement Support for this work was provided by the Department of Energy Innovative Materials and Processes for Advanced Carbon Capture Technology (IMPACCT), DE-FOA-0000208. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.fluid.2014.02.003. References [1] J.L. Anthony, S.N.V.K. Aki, E.J. Maginn, J.F. Brennecke, Int. J. Environ. Technol. Manage. 4 (2004) 105–115. [2] M. Ramdin, Th.W. de Loos, T.J.H. Vlugt, Ind. Chem. Eng. Res. 51 (2012) 8149–8177. [3] E.D. Bates, R.D. Mayton, I. Ntai, J.H. Davis Jr., J. Am. Chem. Soc. 124 (2002) 926–927. [4] K.E. Gutowski, E.J. Maginn, J. Am. Chem. Soc. 130 (2008) 14690–14704. [5] B.E. Gurkan, J.C. de la Fuente, E.A. Mindurup, L.E. Ficke, B.F. Goodrich, E.A. Price, W.F. Schneider, J.F. Brennecke, J. Am. Chem. Soc. 132 (2010) 2116–2117. [6] M.B. Shiflett, D.J. Kasprzak, C.P. Junk, A. Yokozeki, J. Chem. Thermodyn. 40 (2008) 25–31. [7] O. Hollóczki, D. Gerhard, K. Massone, L. Szarvas, B. Németh, T. Veszprémi, L. Nyulászi, New J. Chem. 34 (2010) 3004–3009. [8] O. Hollóczki, D.S. Firaha, J. Friedrich, M. Brehm, R. Cybik, M. Wild, A. Stark, B. Kirchner, J. Phys. Chem. B 117 (2013) 5898–5907. [9] B.E. Gurkan, B.F. Goodrich, E.M. Mindrup, L.E. Ficke, M. Massel, S. Seo, T.P. Senftle, H. Wu, M.F. Glaser, J.K. Shah, E.J. Maginn, J.F. Brennecke, W.F. Schneider, J. Phys. Chem. Lett. 1 (2010) 3494–3499. [10] J. Wang, R.M. Wolf, J.W. Caldwell, P.A. Kollman, D.A. Case, J. Comput. Chem. 25 (2004) 1157–1174. [11] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman, G. Scalmani, V. Barone, B. Mennucci, G.A. Petersson, et al., Gaussian 09, revision A, Gaussian, Inc., Wallingford, CT, 2009. [12] C.I. Bayly, P. Cieplak, W.D. Cornell, P.A. Kollman, J. Phys. Chem. 97 (1993) 10269–10280. [13] V. Chaban, Phys. Chem. Chem. Phys. 13 (2011) 16055–16062. [14] A.R. Leach, Molecular Modeling: Principles and Applications, Prentice Hall, New York, 2001. [15] Y. Wu, H.L. Tepper, G.A. Voth, J. Chem. Phys. 124 (2006) 024503. [16] H.J.C. Berendsen, D. van der Spoel, R. Drunen, Comp. Phys. Commun. 91 (1995) 43–56. [17] C.H. Bennett, J. Comput. Phys. 22 (1976) 245–268. [18] J. Zhang, S. Zhang, K. Dong, Y. Zhang, Y. Shen, X. Lv, Chem. Eur. J. 12 (2006) 4021–4026. [19] G. Zhou, X. Liu, S. Zhang, G. Yu, H. He, J. Phys. Chem. B 111 (2007) 7078–7084. [20] Y. Zhang, E.J. Maginn, J. Phys. Chem. B 116 (2012) 10036–10048. [21] J.G. Kirkwood, J. Chem. Phys. 3 (1935) 300–313. [22] A. Noda, K. Hayamizu, M. Watanabe, J. Phys. Chem. B 105 (2001) 4603–4610. [23] H. Liu, E.J. Maginn, J. Chem. Phys. 135 (2011) 124507. [24] S.M. Urahata, M.C.C. Ribeiro, J. Chem. Phys. 122 (2005) 024511. [25] H. Wu, J.K. Shah, C.M. Tenney, T.W. Rosch, E.J. Maginn, Ind. Chem. Eng. Res. 50 (2011) 8983–8993. [26] W. Shi, K. Damodaran, H.B. Nulwala, D.R. Luebke, Phys. Chem. Chem. Phys. 14 (2012) 15897–15908. [27] M. Brehm, B. Kirchner, J. Chem. Inf. Model. 51 (2011) 2007–2023. [28] R. Kumar, J.R. Schmidt, J.L. Skinner, J. Chem. Phys. 126 (2007) 204107.