Structural analysis of zwitterionic liquids vs. homologous ionic liquids

Ionic liquids (ILs) have shown great potential in industry applications due to their low vapor pressure, intrinsic electric conductivity, and tunability of physical properties.^1,2 Zwitterionic liquids (Zw-ILs) can be chemically very similar to conventional aprotic ionic liquids, except that the positive and negative charges reside on the same molecule.

In this work, we have studied the bulk liquid structure of a set of Zw-ILs and their homologous ILs using both experimental and computational methods. The Zw-ILs we study here have been developed by Kuroda et al. as enhanced solvents for cellulose dissolution while maintaining low toxicity and biocompatibility. These unique properties enable the conversion of plant cell walls to ethanol in a one-pot reaction.³ 

An earlier success of synthesis of Zw-ILs was achieved by Yoshizawa-Fujita et al., who have clearly identified the success of using ethylene oxide repeat units in the molecular structure to reduce the melting points to enable the zwitterion to be liquid at room temperature.⁴ These Zw-ILs, as well as the liquid mixtures of some other zwitterions and ionic liquids, were developed because of their propensity to yield significantly increased conductivity for Li⁺ on addition of lithium bis(trifyl)amide, which makes them of great interest as potential battery electrolytes.^4–7 Zwitterionic mixtures are also being considered for carbon dioxide capture and sequestration applications.⁸ 

The structure of neat conventional ionic liquids and ionic liquid mixtures has been studied using X-ray scattering, neutron scattering, and molecular dynamics (MD) simulation methods.^9–26 The statistics of the bulk structural properties of liquids are described by the liquid structure factor S(q), which is related to the radial pair distribution function g(r) by an inverse Fourier transform. Analysis of molecular simulation results permits a clean separation of S(q) into intramolecular correlations at short distances, which occur for larger values of the scattering vector q, and intermolecular correlations for larger separations and hence smaller values of q. Atomic liquids and non-associating molecular liquids typically present only a single peak in the intermolecular part of S(q), which correlates with the first peak in the pair distribution g(r), indicating the most probable separation between nearest neighbors. Ionic liquids are generally much more structured than neutral solvents, typically displaying two or three peaks in the intermolecular part of S(q), i.e., for values of q less than 2-3 Å⁻¹. The structure in conventional aprotic ILs results from the balance between strong charge-charge interactions vs. nonpolar or van der Waals interactions. In the case of protic ILs, the hydrogen-bonding network leads to much of the observed intermediate range order at length scales larger than the size of the ion pairs. Depending on the nature of the anionic or cationic functional groups, aprotic ILs display structures that result from adjacency of nearest neighbors, strong charge-charge ordering effects that lead to correlations between like ions, and larger length scale intermediate range order that arises from the self-aggregation of less polar or extremely hydrophobic functional groups, such as ether chains or perfluoroalkyl chains.

The structures of bulk ionic liquids with a variety of anion and cation structures have been characterized. The three peaks typically observed in the intermolecular part of ionic liquid structure factor include an adjacency peak, usually observed at q ∼ 1.4 Å⁻¹. The charge-charge correlation peak at around q = 0.8-1.0 Å⁻¹ indicates the interactions between ions and like-charged ions of the second coordination shell. This peak is a direct result of the unique charge ordering in ILs. For some types of ionic liquids or ionic liquid mixtures, a pre-peak or first sharp diffraction peak (FSDP) may appear in S(q) for q between 0.2 and 0.5 Å⁻¹, which is an indicator of the intermediate range order in ionic liquids. For example, when a cationic alkyl tail is sufficiently long^9,27–33 (or anionic tail^34,35), these nonpolar alkyl chains will form polarity-segregated nano-domains and the FSDP will arise from the outer coordination shell of opposing polar and nonpolar domains.^31,33 Recently, we have shown that ethylene oxide types of ether tails generally disrupt the domain structure.^36–38 However, when the cationic functional group is very polar, as is the case for a pentamethyldisiloxy group, the differences in chemical interactions between the ionic head groups and the functional groups can be sufficiently large as to drive nanophase segregation, leading to observation of a FSDP in S(q).³⁹ Such a FSDP can also arise from hydrogen bond networks in protic ILs where alkylammonium cations are paired with H-bond accepting anions such as nitrate, formate, or glycolate.^9,40–42 Recent reports of protic IL mixtures with strongly H-bonding neutral solvents can also lead to observations of a FSDP in S(q) for very low values of q.²² 

Because there are potential advantages of Zw-ILs relative to other electrolytes, it seems urgent to understand the structural properties of these Zw-ILs and to compare them with the structures of corresponding aprotic IL homologs. Thus, we studied the bulk structure of these ionic liquids using high-energy X-ray scattering and molecular dynamics simulations. Here we report our structural results on two novel zwitterionic liquids, 1-[2-(2-methoxyethoxy)ethyl]-3-(3-carboxypropyl)-imidazolium (OE₂imC₃C) and 1-[2-(2-methoxyethoxy)ethyl]-3-(3-carboxypentyl)-imidazolium (OE₂imC₅C).³ The aprotic IL homolog of OE₂imC₃C is 1-[2-(2-methoxyethoxy)ethyl]-3-ethyl-imidazolium acetate (OE₂eim⁺/OAc⁻). Examination of the structures shown in Fig. 1 shows that elimination of a single carbon-carbon bond in the OE₂imC₃C Zw-IL leads to the aprotic IL structure of OE₂eim⁺/OAc⁻. Because the structure of 1-ethyl-3-methylimidazolium acetate (⁠$Im2,1+$/OAc⁻) is very well characterized, we have included this IL in our study to facilitate a comparison with removal of the effect of the 2-(2-methoxyethoxy)ethyl (diether) group. Below we report our investigations of the four liquids shown in Fig. 1.

The Zw-ILs were synthesized, purified, and characterized at Kanazawa University. Both the Zw-IL, OE₂imC₃C, and the homologous IL, OE₂eim⁺/OAc⁻, were prepared as described in the work of Kuroda et al.³ OE₂imC₅C was prepared in a similar manner as OE₂imC₃C,³ with ethyl 6-bromohexanoate (purchased from TCI, purity >98.0%) replacing ethyl 4-bromobutyrate to change the product from a butanoate to hexanoate group. $Im2,1+$/OAc⁻ was purchased from IoLiTec. All samples were dried on a Schlenk vacuum-argon line at 10⁻² mbar for 48 h before use, demonstrating that the Zw-ILs have quite low vapor pressures.

The samples for X-ray scattering were filled into 3 mm diameter NMR tubes in an argon purged glovebox, temporarily sealed with the NMR cap, and promptly flame-sealed. High energy X-ray scattering experiments were performed at the Advanced Photon Source beamline 11-ID-B, as described previously.^39,43 X-ray photon energies of 58.62 keV (λ = 0.211 50 Å) were used for all experiments. The sample to detector distance was set to approximately 21 cm. The total exposure time for each sample is 5 min. The diffraction pattern is recorded using a PerkinElmer detector. The scattering pattern is integrated over q using Fit2D software,⁴⁴ and the structure factor S(q) is obtained from the corrected intensity pattern I(q) using PDFgetX2 software.⁴⁵ 

The molecular dynamics simulations were performed using GROMACS version 2016.1.^46,47 All electronic structure calculations were carried out using Gaussian09 version D.01.⁴⁸ The protocols of simulations for the bulk liquids were the same as described previously.^39,43 The initial periodic boxes containing 1000 molecules of zwitterionic liquids or 1000 ion pairs of ionic liquids were generated using the PACKMOL program.⁴⁹ A time step of 1 fs is used in the simulation. Periodic boundary conditions are applied in all directions. The cutoffs for short range Coulomb and Lennard-Jones van der Waals interactions are set to 1.5 nm. Long range electrostatic energies are calculated using the Particle Mesh Ewald (PME) method,^50,51 in which the interpolation order is 6 and FFT grid spacing is 0.08 nm. The simulation is started with fractional ionic charges using Berendsen temperature and pressure coupling,⁴⁶ which are then slowly scaled to unit charges. During a total of 8 ns of simulated annealing using Nosé-Hoover thermostat^52,53 and Parrinello-Rahman barostat,⁵⁴ the box temperatures were raised to 598 K in 500 ps, equilibrated for another 7 ns and annealed over 500 ps back to the target temperature of 298 K. The simulation box is further equilibrated at 298 K for another 10 ns. The final trajectory is run for 3 ns with data written every 1 ps.

The radial distributions are obtained from the simulation trajectories, and the structure factor S(q) is calculated from using the previous method,^31,33,55

where ρ₀ is the box number density, g_ij(r) is the partial radial distribution function between atom i and atom j, x_i is the molar fraction of atom i in the box, f_i(q) is the atomic form factor,⁵⁶ and W(r) = sin(2πr/L)/(2πr/L) is the Lorch function.^33,57,58

It has long been recognized that for binary monatomic systems, one can partition the radial pair distribution function g(r)⁵⁹ or the liquid structure factor S(q).⁶⁰ Though more complex, the same conceptual partitioning can be done for multi-component systems such as ionic liquids. For ionic liquids, the total S(q) can be partitioned into cation-cation, cation-anion, and anion-anion interactions, as previously described,^{31–33,39,43,55}

where C denote cations and A denote anions. Though for Zw-ILs there are no discrete molecular anions or cations, we can still partition the liquid structure by defining the CH₂COO⁻ group as the anion moiety and the other remaining atoms as the cation moiety of the Zw-IL. S(q) can be also partitioned into cation ring-cation tail-anion interactions by

where R denotes cationic ring atoms, T indicates cationic tail atoms, and A is for anion atoms.

Molecular simulations have been widely used to explore the charge-ordering and formation of nanoscale polar and apolar domains in ionic liquids.^27,61–63 In the first studies, much attention was given to the anisotropic arrangements of the packing of anions and cations in the periodic boxes for the molecular simulations. Introduction of strongly amphiphilic ions with cationic or anionic alkyl tails led to the appearance of significant intermediate range order²⁷ that was soon verified by X-ray scattering experiments.^28,29

To obtain a qualitative comparison between the liquid structures of Zw-ILs and ILs, it is useful to first consider the overall arrangements of ions that are revealed in snapshots of the MD simulation boxes, reminiscent of the pioneering work of Canongia Lopes and Pádua.²⁷ The equilibrated simulation boxes were found to have a density for $Im2,1+$/OAc⁻ of 1.117 g cm⁻³ at 298 K, close to the experimental value of 1.102 g cm⁻³ at this temperature.^64,65 Limited quantities of OE₂imC₃C and OE₂imC₅C samples precluded accurate density measurements for the Zw-IL samples.

In Fig. 2, we show snapshots of the periodic boxes from the equilibrated molecular simulation trajectories for our Zw-IL and homologous IL systems. Red coloring indicates the more polar cationic moiety, comprising the imidazolium ring and nearby atoms, while blue denotes the anionic moieties. Green indicates the 2-(2-methoxyethoxy)ethyl tails bonded to one of the imidazolium ring nitrogens. The spatial distributions for the OE₂imC₃C Zw-IL (top left) and its homologous IL OE₂eim⁺/OAc⁻ (top right) appear to be almost identical. The $Im2,1+$/OAc⁻ box at lower right shows the salt-like charge ordering for an IL that does not display any significant intermediate range order, while the more complex pattern observed at lower left for OE₂imC₅C illustrates that this Zw-IL can display significant intermediate range order. Compared to the other systems, the simulation box for $Im2,1+$/OAc⁻ is necessarily smaller because of the significantly smaller number of total atoms for the same number of ion pairs.

The fractional charge transfer between the anionic and cationic parts of the Zw-IL and IL systems was investigated using electronic structure calculations on a small cluster containing 8 ion pairs or molecules selected from the 1000 molecules (ion pairs) in the MD simulations. A random $CO2−$ group is selected together with the 8 nearest neighbor cationic groups and 7 nearest neighbor anionic groups. This cluster of 8 ion pairs (or molecules) is an arbitrary snapshot of the model liquid, equilibrated in GROMACS using the CL&P potentials, as described above. This cluster geometry is optimized using dispersion-corrected PM6-D3 semi-empirical Hamiltonian, and atomic dipole moment corrected Hirshfeld (ADCH) charges on the central anion are then calculated for this geometry at the B3LYP-D3/6-31G(d,p) level.^66,67 Using 7 input configurations for each Zw-IL or IL, the averaged partial charges on the anionic carboxylate group were −0.52 ± 0.07 for OE₂imC₃C and −0.45 ± 0.04 for OE₂eim⁺/OAc⁻, indicating an excess in charge density of about 15% for the Zw-IL relative to the IL. The slightly larger extent of charge transfer in OE₂eim⁺/OAc⁻ relative to OE₂imC₃C indicates a greater degree of overall H-bonding for the Zw-IL relative to the IL. Partial charge transfer between ions results both from H-bonding and polarizability of the molecular ions and is well known to affect the structure and dynamics in ionic liquids. Though we did use charge scaling in our molecular dynamics calculations, it is well known that reduction of the net ionic charge from ±1 to ±0.7–0.8 |e| will better reproduce the viscosity and diffusivities for ILs.^68–70 The greater extent of charge transfer found for our systems is most likely due to a greater number of H-bonds in these systems, as well as the fact that our anionic carboxylate group is smaller and more highly polarized anion than most common IL anions like bis(trifluoromethylsulfonyl)amide (NTf₂−) or $PF6−$⁠.

Another approach to studying the degree of intramolecular charge transfer between the anionic and cationic parts of OE₂imC₃C was to select a single molecule and obtain charges from the ADCH algorithm at the B3LYP-D3/6-31G(d,p) level using a self-consistent reaction field solvation correction with the Gaussian09 default integral equation formalism polarized continuum model⁴⁸ with a dielectric constant of 36.5. This population analysis for a single OE₂imC₃C molecule with ADCH charges showed that the partial charges on the three atoms of the carboxylate group sum to −1.00 ± 0.03e for all the stable conformations observed. This indicates that there is almost no intramolecular charge transfer between the cationic part and the anionic parts of the OE₂imC₃C molecule.

Past work has enabled us to show that liquid structure factors S(q) obtained from X-ray experiments and molecular simulations can be compared quantitatively, and excellent agreement between the two can be achieved.^33,37,39 This enables us to mine the molecular dynamics trajectories to obtain far more detailed information. Specifically, when the structure factors S(q) are factored appropriately, interference between strong positive and negative correlations provides an unambiguous marker for the underlying structure. We have taken to calling these positive and negative correlations in S(q) peaks and anti-peaks, respectively.^{11,32,33,37,39,43,55} Standard X-ray experiments observe total scattered X-ray intensities and are thus homodyne experiments for which the minimum observable signal is zero. The negative correlations, or anti-peaks,^32,33,37 are not directly observable in standard homodyne X-ray scattering intensity experiments, but because the observable is a total intensity that is a positive definite quantity, any anti-peak must be offset by an equal or larger positive-going peak, both of which arise because of positive and negative structural correlations. We have shown recently that the interferences between peaks and anti-peaks can lead to observed peaks in the total S(q) that are shifted relative to the true underlying structural features.³⁹ For this reason, we feel it is important to obtain a theoretical model for structural data to enable us to scan for the presence of anti-peaks in S(q) to avoid misinterpreting the experimental structure factors obtained from X-ray scattering data.

The experimental and simulated structure factors S(q) of the four ionic liquids are shown in Fig. 3, where one sees that good agreement is obtained. The ionic partitioning described by Eq. (2) and sub-ionic partitioning defined by Eq. (3) are shown in Figs. 4 and 5, respectively. The experimental (simulated) peak positions are summarized in Table I.

The adjacency peak in the intermolecular part of S(q) arises from nearest neighbor interactions and for these liquids appears near 1.5 Å⁻¹. From the top three graphs in Fig. 3, one observes a slight shift in the adjacency peak from 1.53 Å⁻¹ for the OE₂imC₃C Zw-IL to 1.58 Å⁻¹ for the OE₂eim⁺/OAc⁻ IL, indicating that the average distances between nearest neighbors increases slightly for the Zw-IL relative to the IL. This is as expected, since the difference of the single carbon-carbon bond between the OE₂eim⁺/OAc⁻ IL relative to the OE₂imC₃C Zw-IL can result in a slight increase in the average distance between imidazolium ring and the carboxylate tails.

The charge-charge correlation peak in S(q) is typically observed at about 0.8 < q < 0.9 Å⁻¹ and appears to be suppressed in the total liquid structure factor S(q) shown in Fig. 4. However, the ionic partitioning of S(q) given by Eq. (2) reveals that while the X-ray scattering shows only a small amplitude in the charge-charge correlation peak region, this is because the strongly interfering peaks and anti-peaks near q = 1.0 Å⁻¹ reveal the true location of the charge-charge correlation structure, while the total intensity in S(q) at this position is negligible. The IL with the smallest anion and cation, $Im2,1+$/OAc⁻, shows the sharpest structural features of the set of four liquids. The charge-charge correlation peak for the two Zw-ILs is qualitatively quite similar, though the peaks are broader for OE₂imC₃C than for OE₂imC₅C, indicating a more diffuse local structure. Comparing the S(q) for the OE₂imC₃C Zw-IL with the IL homolog for the OE₂eim⁺/OAc⁻, we can see that the charge-charge correlations are quite similar between the two species, but stronger for the IL than the Zw-IL. Thus, ionic partitioning of S(q) shows that the second-shell charge-charge correlation peak is just as important a structural feature for Zw-ILs as it is for standard ILs.

The three liquids OE₂imC₃C, OE₂imC₅C, and OE₂eim⁺/OAc⁻ show a low intensity FSDP in S(q) in the range of 0.4 < q < 0.7 Å⁻¹. $Im2,1+$/OAc⁻ does not display a FSDP because the ethyl tail is neither long enough nor nonpolar enough to lead to nanophase segregation of the alkyl chains that can lead to observation of FSDPs.^{27–29,31,33,71} All of these FSDPs have quite small amplitude relative to the sharper, more intense peaks observed for ILs with significant alkyl chain lengths.^29,31,33,71 The small amplitude observed for the FSDPs in these liquids is not unexpected because the 2-(methoxyethoxy)ethyl diether tail is expected to suppress the FSDP, as has been reported for similar systems by Triolo et al.,³⁶ Kashyap et al.,³⁷ and Hettige et al.³⁸ Figure 4 shows that only for the anion-anion component of S(q) is there even a clearly discernible contribution to the FSDP for OE₂imC₃C and OE₂imC₅C.

Analysis of the MD simulations enables us to obtain an improved interpretation of the intermediate range order for these liquids. Specifically, a sub-ionic partitioning of the structure factor S(q) is required to more fully reveal the underlying structure and this analysis is shown in Fig. 5. Structural features for the two Zw-ILs and the homologous IL are very similar for four of the six groups of structural correlations: the ring-ring, ring-tail, ring-anion, and tail-tail sub-components of S(q). Positive contributions to the FSDP in S(q) arise from ring-ring, anion-ring, and tail-tail interactions, which are offset by anti-peaks in the ring-tail correlations. Simply put, the diether tails do not coil to localize near the imidazolium rings, in contrast to other ether-substituted imidazolium systems reported by Triolo et al.³⁶ For these three liquids, subtle differences in the anion-anion and anion-tail correlations in S(q) are noted. Anion-tail interactions give rise to a moderate intensity anti-peak for the OE₂eim⁺/OAc⁻ and OE₂imC₅C Zw-ILs and a larger amplitude anti-peak for OE₂imC₃C; positive contributions to the FSDP occur for the anion-anion interactions, with the OE₂imC₃C amplitude being larger than for the other two liquids. Thus, the positive contributions to a FSDP from tail-tail interactions of the diether groups provide another contrast to what has been observed for ILs with 1-(alkoxy)-3-methylimidazolium or 1-alkoxy-1-methylpyrrolidinium cations.^36,37 The key difference for the set of liquids reported here is that the cationic ring is more centrally located in the charge distribution of the liquid, such as for OE₂imC₅C. The deconstruction of the total structure factor S(q) reveals that the overall liquid properties of the two Zw-ILs studied here are qualitatively very similar to the features of the homologous IL.

The previous discussions conclude that the FSDP in OE₂imC₃C results from nano-domain aggregation and excludes the possibility that the FSDP in OE₂eim⁺/OAc⁻ arises from the same interaction. The intermediate range order in the OE₂eim⁺/OAc⁻ IL likely results from the hydrogen bonding network, which is known to be the source of the FSDP in protic ILs such as alkylammonium nitrate.^{40–42,69,72} To explore the influence of hydrogen bonding networks in this set of Zw-ILs and ILs, we analyzed the hydrogen bond present in the MD trajectories. Networks of H-bonds were found to occur between the hydrogens on the imidazolium rings and the carboxylate groups.

The types of H-bond interactions between carboxylate oxygen atoms in acetate and imidazolium ring protons are described by Bowron et al. as single interaction, bidentate interaction, bifurcated interaction, or bridging interactions.⁶⁹ Illustrations of these H-bond types are shown in Fig. S1 of the supplementary material. To further understand the hydrogen bond types, electronic structure calculations on gas-phase ion pairs were performed at the M06-2X/6-311+G(d,p) level. The results are shown in Fig. S2 of the supplementary material. We found that single H-bond interactions between a carboxylate oxygen and an imidazolium hydrogen are the most probable structures. These geometries are further stabilized by weak interaction between another carboxylate oxygen atom and the α-hydrogen. Bidentate, bifurcated, and bridging interactions are all not as stable as single interactions, which all have higher free energies.

The molecular conformations and intramolecular hydrogen bonds for an isolated OE₂imC₃C Zw-IL molecule were also studied at the same level of electronic structure theory. We observed that an intramolecular hydrogen bond can be formed, but the conformations with intramolecular hydrogen bonds do not have a significant lower energy than those with no hydrogen bonds. That indicates that intramolecular hydrogen bonds are unlikely to be a determining factor in the overall liquid structure, which is consistent with what is shown in the spatial distribution functions from the MD simulations.

In order to locate the oxygen and hydrogen atoms with strong interactions between them, we studied the partial radial distribution functions g_ij(r) between all hydrogen and oxygen types in these ionic liquids. The graphs are shown in Fig. S3 of the supplementary material. We find strong interactions between the carboxylate O atoms and the hydrogens on imidazolium rings and some weaker interactions between carboxylate O atoms and hydrogens on alkyl carbons directly bonded to nitrogen atoms (specifically, the α-carbon hydrogen atoms). These five types of hydrogen atoms are labeled as H_A to H_E, as shown in Table II. The probability for H-bond formation is ranked H_C > H_A ≈ H_B > H_D > H_E. Only very weak hydrogen bonding interactions were observed between ether oxygens for any of the fives types of H-bond donors. These interactions have also been shown by the intermolecular and intramolecular spatial distribution of carboxylate oxygens against imidazolium rings, as shown in Fig. 6. These spatial distribution functions indicate that hydrogen bonds are formed between the three imidazolium ring protons and that most of them are intermolecular. Graphs of the data given in Table II are provided in the supplementary material as Fig. S4.

The distance-angle distributions between O–H distance and C–H–O angle for the imidazolium ring H-bond donating protons H_A, H_B, and H_C in the three liquids are also shown in Fig. S5 of the supplementary material, where O is one of the oxygens in carboxylate group. The O–H distance peaked at nearly 2.5 nm where at the same time the C–H–O angle ranged from 120° to 180°, which clearly indicates the existence of hydrogen bonds. The distribution of C–H–O angles matches the most stable conformations shown in Fig. S2 of the supplementary material.⁷³ 

In the MD trajectories, we defined the distance and angle parameters for a hydrogen bond to be r_O−H < 2.65 Å and for C–H–O angles >120°. Electronic structure calculations show that hydrogen bond lengths in these systems are about 1.9 Å, but the 2.68 Å definition for H-bond distances in the OPLS-AA force field requires us to use a larger value to parametrize H-bonds in the classical MD system. The numbers of H-bonds in the simulation box were calculated for each of the five potential H-bond donor atoms. These results are summarized in Table II, which shows how the molecule is partitioned for OE₂imC₃C, OE₂imC₅C, and OE₂eim⁺/OAc⁻. A similar analysis is also performed to determine the number of H-bonds observed for each of the carboxylate oxygens and the results are listed in Table III. From Table II we find that most hydrogen atoms favor single interactions with only rare bidentate interactions observed. Intermolecular hydrogen bonds dominate in zwitterionic liquids, with only 5%–10% of the hydrogen bonds being intramolecular, in agreement with the electronic structure calculations.

Both Tables II and III show that the numbers of hydrogen bonds for the OE₂eim⁺/OAc⁻ IL are greater than for the OE₂imC₃C Zw-IL. This is evidence that the H-bonding network in OE₂eim⁺/OAc⁻ is likely to be stronger than for the Zw-IL, indicating that the larger amplitude FSDP in OE₂eim⁺/OAc⁻ may correlate with the degree of H-bonding. We also find that the hydrogen bond numbers for OE₂imC₅C lie between those for OE₂imC₃C and OE₂eim⁺/OAc⁻, indicating that the degree of H-bonding for OE₂imC₅C is between OE₂imC₃C and OE₂eim⁺/OAc⁻. This is consistent with the preceding analysis of the liquid structure factor S(q), which indicates that increasing the distance between positively charged parts and negatively charged parts has a similar effect to changing the carbon-carbon bond in OE₂imC₃C to instead obtain OE₂eim⁺/OAc⁻.

The results comparing the structural properties of the Zw-ILs and corresponding ILs can be summarized graphically in Fig. 7. While our electronic structure calculations reveal that the charge densities of the anionic and cationic moieties are quite similar between the Zw-ILs and the ILs, the possibilities for developing H-bonding networks are more extensive for the acetate-anion ILs. By contrast, the Zw-ILs are shown to display nano-domain aggregation, as revealed by the ionic partitioning of the liquid structure factor S(q). Figure 7 illustrates these two points on the right and left sides of the image, respectively.

We have measured and compared the structure factor S(q) of zwitterionic liquid OE₂imC₃C and OE₂imC₅C and compared these with the conventional ionic liquids OE₂eim⁺/OAc⁻ and $Im2,1+$/OAc⁻. The ionic partitioning of the structure factors shows that the charge ordering is still a significant feature for zwitterionic liquids, but is weaker than in conventional ionic liquids. A first sharp diffraction peak is observed for each of the liquids (except the IL with the smallest ions, $Im2,1+$/OAc⁻), but the FSDP has subtly different origins between Zw-ILs and ILs. In OE₂imC₃C, the FSDP is mostly due to nano-domain aggregation, but in OE₂eim⁺/OAc⁻, it most likely arises from the hydrogen bonding network. The structure of OE₂imC₅C is in between that of OE₂imC₃C and OE₂eim⁺/OAc⁻, which indicates that increasing positive and negative charge separation has a similar effect to changing the carbon-carbon bond in the zwitterionic OE₂imC₃C to obtain the standard ionic liquid OE₂eim⁺/OAc⁻.

The correlations between these structural features observed in the total and deconstructed S(q) data indicate that H-bonding is an important structural motif for these liquids, which may be the reason that only the ambient temperature liquid zwitterions have an imidazolium core, as opposed to the common tetraalkyl ammonium or phosphonium cations. Note that the spatial distribution functions shown in Fig. 6 provide insight into the nature of these H-bonds. While the H_A, H_B, and H_C protons (the 2, 4, and 5 imidazolium ring positions) have interactions that span a broad range of the nearby three-dimensional space, the H-bonds from α-carbon donors on phosphonium or ammonium cations have greater angular diversity and thus likely larger configurational entropies. This is offset by the fact that the imidazolium protons are simply much more acidic, as manifested by the ¹H NMR chemical shifts. Further work on these systems is ongoing, both to introduce new structural motifs and to characterize the properties of the homologous Zw-ILs and ILs at interfaces.

See supplementary material for more details on the hydrogen bond analyses including the partial radial distribution functions.

The research at Rutgers was supported by NSF Grant No. CHE-1664809 to EWC. Research in Kanazawa was partly supported by the Advanced Low Carbon Technology Research and Development Program (ALCA), (Grant No. 2100040), and by the COI program “Construction of next-generation infrastructure using innovative composite materials—Realization of a safe and secure society that can coexist with the Earth for centuries—–,” and the Cross-ministerial Strategic Innovation Promotion (SIP) Program, and partly supported by a Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (No. 15K17867). This research used resources of the Advanced Photon Source, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357. The authors gratefully thank scientists at APS sector 11-ID-B, Dr. Olaf J. Borkiewicz, Kevin A. Beyer, and Dr. Karena W. Chapman for their help in acquiring X-ray scattering data. We thank Professor Claudio J. Margulis from the University of Iowa for helpful discussions.