60th Annual Report on Research 2015

Our initial studies of methane solvation focused on mixtures of 0 vol%, 25 vol%, 50 vol%, 75 vol%, and 100 vol% ionic liquid with water. Each mixture was simulated over a broad temperature range using replica exchange molecular dynamics at 1 atm. Methane excess chemical potentials were determined using standard particle insertion techniques.

Figure 1. Excess chemical potential (G^ex) of methane in mixtures of EAN with water (a) and BAN with water (b) as a function of temperature at atmospheric pressure. Points are simulation results, while the solid lines indicate fits to the Margules expansion.

The excess chemical potentials of methane in aqueous EAN and BAN mixtures are shown in Figure 1. Both plots indicate that the chemical potential is positive in these mixtures, indicative of low gas solubility. Moreover methane's excess chemical potential is an increasing function of temperature, indicating that dissolution is entropically unfavorable in both solvent mixtures. Interestingly the excess chemical potential appears to approach a maximum in pure water (green points), while is does not do so in the other solvents. This maximum indicates both a change in the sign of the solvation entropy in pure water that results from a significant large positive heat capacity for dissolution in water. Therefore, while methane solvation in these solvents is entropically unfavorable (a characteristic of nonpolar solvation in water, but not normal organic solvent) in all solvents, solvation in water still exhibits distinct temperature dependences not observed in the ionic liquids.

The excess chemical potential of methane in pure EAN is great than that in pure water (Figure 1a), indicting a lower solubility for methane in the ionic liquid. This result is surprising since experiments generally find the opposite trend. One potential reason for this discrepancy is the lack of polarizability in the model. We have demonstrated that the effective inclusion of polarization for alkane solvation in water through inclusion of an increased water/solute attractive interaction can account for discrepancies between experiment and simulation for simulation model parameterized to pure hydrocarbon properties. In the case of nonpolar solvation in an ionic liquid we may expect an even greater contribution from polarization due to the large electrostatic fields in the charged solvent. As the alkyl length of the ionic liquid grows, methane's solubility increases as indicated by the lowering of the chemical potential. Indeed for BAN methane's solubility is greater than in water (Figure 1b). We therefore expect that the driving force for assembly is greater in EAN than BAN. It is generally reported that assembly is weakest in BAN in agreement with the result reported in Figure 1.

To correlate the excess chemical potential of methane in these mixtures we employed a Margules-like expression for the free energy

G^ex(T, x₁) = G^ex(T, x₁=1)*x₁ + G^ex(T, x₁=0)*(1- x₁) + x₁*(1- x₁)[A(T) + B(T)*(2*x₁-1)],

where x₁ is the mole fraction of component 1 (taken to be the ionic liquid here), T is the temperature, and A and B are Margules constants. The temperature dependence of the terms G^ex(T, x₁=1), G^ex(T, x₁=0), A(T), and B(T) were assumed to follow the standard form X(T) = a + b*(T-300K) + c*Tln(T/300K), where a, b, and c are fitted coefficients. Fits of this expression are indicated by the solid lines in Figure 1. Overall this expression does an excellent job allows us to interpolate to solvent compositions that were not simulated.

In x-ray scattering study of alcohols in protic ionic liquids, Warr's group (J Phys Chem, 2014) observed large scale structuring of ethanol and longer chain alcohols in EAN and PAN mixtures. In their study they observed an increase in scattering over a wide range of length scales that was not present for either the EAN, PAN, or ethanol alone.

These results suggested the formation of self-assembled aggregates with a broad range of sizes. In figure 2 we show predicted x-ray scattering structures obtained from our simulations of ethanol mixtures with EAN and BAN.

Figure 2. Predicted x-ray scattering intensity from molecular simulation for mixtures of EAN with enthanol (a) and BAN with ethanol (b). The mole fractions (x1) reported in the figure legends correspond to the mole fraction of the ionic liquid in the mixture.

These simulations clearly show the onset of larger scale structures in both ionic liquids as indicated by the increase in scattering at low q in these figures. While EAN itself exhibits large scale structure, this peak shifts to larger scale structures in mixtures with ethanol. Given the low solubility of methane in EAN it is possible that this structure results from phase separation. When we examine molecular snapshots from these simulations, the structure appears more like a bicontinuous sponge. We have raised the temperature significantly to try to melt out these structures and permit phase separation if that is the equilibrium state. The observed simulation and x-ray structures, however, we resilient to heating suggesting we are sampling the true equilibrium state for our simulations. That said, the experimental data from Warr indicated a broad range of aggregate sizes even greater than what we observe. We are limited in our simulations to structures smaller than ~5 nm. While the ethanol to induce structuring in BAN (Figure 2b), the scattering intensity is not as pronounced as in EAN. This is consistent with observations above than methane is more soluble in BAN, thereby reducing the driving force for assembly.