MAS NMR Investigation of Molecular Order in an Ionic Liquid Crystal

The structure and molecular order in the thermotropic ionic liquid crystal (ILC), [choline][geranate(H)octanoate], an analogue of Choline And GEranate (CAGE), which has potential for use as a broad-spectrum antimicrobial and transdermal and oral delivery agent, were investigated by magic-angle spinning (MAS) nuclear magnetic resonance (NMR), polarizing optical microscopy, small-angle X-ray scattering (SAXS), and mass spectrometry. Mass spectrometry and the 1H NMR chemical shift reveal that CAGE-oct is a dynamic system, with metathesis (the exchange of interacting ions) and hydrogen exchange occurring between hydrogen-bonded/ionic complexes such as [(choline)(geranate)(H)(octanoate)], [(choline)(octanoate)2(H)], and [(choline)(geranate)2(H)]. These clusters, which are shown by mass spectrometry to be significantly more stable than expected for typical electrostatic ion clusters, involve hydrogen bonding between the carboxylic acid, carboxylate, and hydroxyl groups, with rapid hydrogen bond breaking and re-formation observed to average the 1H chemical shifts. The formation of a partial bilayer liquid crystal (LC) phase was identified by SAXS and polarizing optical microscopy at temperatures below ∼293 K. The occurrence of this transition close to room temperature could be utilized as a potential temperature-induced “switch” of the anisotropic properties for particular applications. The presence of an isotropic component of approximately 23% was observed to coexist with the LC phase, as detected by polarizing optical microscopy and quantified by both 1H–13C dipolar-chemical shift correlation (DIPSHIFT) and 1H double-quantum (DQ) MAS NMR experiments. At temperatures above the LC-to-isotropic transition, intermediate-range order (clustering of polar and nonpolar domains), a feature of many ILs, persists. Site-specific order parameters for the LC phase of CAGE-oct were obtained from the MAS NMR measurement of the partially averaged 13C–1H dipolar couplings (DCH) by cross-polarization (CP) build-up curves and DIPSHIFT experiments, and 1H–1H dipolar couplings (DHH) by double-quantum (DQ) build-up curves. The corresponding order parameters, SCH and SHH, are in the range 0–0.2 and are lower compared to those for smectic (i.e., layered) phases of conventional nonionic liquid crystals, resembling those of lamellar phases formed by lyotropic surfactant–solvent systems.


INTRODUCTION
As the name suggests, the liquid crystalline state of matter is an intermediate between liquids and solids; there is long-range order, but the molecules themselves are mobile. Liquid crystals (LCs) can form under suitable temperature (thermotropic LCs) or concentration (lyotropic LCs) conditions. A subset of LCs, ionic LCs (ILCs), carry charges and thus combine the characteristics of LCs and ionic liquids 1,2 (ILs). ILCs combine the conductivity, low vapor pressure, and tuneability (the ability to select or modify properties by the appropriate choice of the anion and cation) of ILs, with the anisotropic properties of LCs. This unique combination of properties has generated interest in ILCs, 3−5 which may be used in a broad range of fields, with applications including LC displays, 6 as electrolytes in dye-sensitized solar cells, 7−9 as ordered reaction media or templates in synthetic chemistry, 10−12 and for biological applications. 13,14 The majority of ILCs are based on typical organic nitrogen cations (e.g., imidazolium, ammonium, and pyridinium) substituted with one or more long alkyl chain and inorganic anions (e.g., halides, [BF 4 ] − , and [PF 6 ] − ), and form smectic (i.e., layered) phases, exhibiting partial positional order, in addition to orientational order. 3 They typically display considerably less orientational order than nonionic thermotropic LCs, instead resembling the lyotropic lamellar phases of surfactant−water systems. 15,16 Advancing our understanding of ILC structure and dynamics at an atomic level will enable and promote the development of novel ILCs as well as further applications for ILCs.
Choline And GEranate (CAGE) is a broad-spectrum antimicrobial and a transdermal and oral delivery agent. 17−22 The most effective formulation of CAGE is composed of choline, geranic acid, and the geranate anion in a 1:1:1 ratio ([choline][geranate 2 (H)]), 21 which has been described as both an IL and a deep eutectic solvent (DES), 20,23 a eutectic mixture of two or more components (in some cases both neutral and ionic species) that typically interact via strong hydrogen bonding. We have modified CAGE, replacing 1 equiv of geranic acid with octanoic acid, to form a novel stable thermotropic ILC, CAGE-octanoic acid (CAGE-oct; Figure  1). The LC-to-isotropic phase transition of CAGE-oct occurs close to room temperature (∼293 K), which may be exploited as a temperature-induced "switch" of the anisotropic properties for specific applications.
Considering structure and order in such materials, both the time scale and length scale are of importance. In contrast to a so-called "isotropic" liquid (where there can still be some short-range ordering), LCs exhibit long-range positional and/ or orientational order; the molecular positions and relative orientations are correlated over longer distances. Aspects of the structure (e.g., shape of the constituent molecules and nanoscale segregation of molecular components into distinct regions) determine the formation of LC phases. The segregation of incompatible parts of the molecules (e.g., nonpolar aliphatic chains from polar regions), as well as attractive forces (e.g., ionic interactions or hydrogen bonding) drive the formation of long-range positional order in ILCs.
Small-angle X-ray scattering (SAXS) and nuclear magnetic resonance (NMR) spectroscopy are complementary techniques to investigate LCs. 24 While SAXS identifies long-range order, NMR yields local atomic-level structure, e.g., about specific intermolecular hydrogen bonding interactions, and is sensitive to molecule/segment/bond orientations. Specifically, the dipolar couplings between spins report on order: the dipolar coupling depends on the time scale and amplitude of molecular motion, because motion averages the couplings. Compared to an isotropic liquid where the molecules undergo rapid tumbling and all orientations are sampled, such that the dipolar couplings are averaged to zero, anisotropic LC phases have a preferential orientation and the dipolar coupling is not completely averaged out. The ratio of the partially averaged residual dipolar coupling (RDC, D res ), to its unaveraged value (i.e., the value in the rigid limit, D rigid ) provides the order parameter, S = D res /D rigid . S ranges from 0 for isotropic systems to 1 for completely rigid systems (see section S1 for further details on the dipolar coupling and order parameter).
In solid-state NMR, magic-angle spinning (MAS) averages anisotropic interactions so that high-resolution spectra can be obtained. Radiofrequency (rf) pulses can be applied that interfere with the MAS averaging and thus recouple the interactions, and the strength of the interaction can be measured. One such approach is the much-employed solidstate NMR approach of 1 H− 13 C cross-polarization (CP), whereby long rf pulses are applied simultaneously with the same nutation frequency (under MAS, this so-called Hartmann−Hahn matching condition is adjusted by a small integer multiple of the spinning frequency) to both the 1 H and 13 C nuclear spins: 13 C− 1 H heteronuclear dipolar couplings may be obtained from 1 H− 13 C CP build-up curves. If there is a dominant dipolar interaction, the polarization transfer shows oscillatory behavior where the frequency of oscillations is directly proportional to the dipolar coupling constant. 25 CP build-up curves or variants such as Lee−Goldburg (LG) CP have been used to obtain RDCs and the related S CH order parameters in various dynamic systems, including polymers 26 and LCs. 27,28 A complement to probing 13 44 Despite the interest in ILCs, and the potential of NMR for probing structure and order, very few NMR studies have been reported on ILC materials. Recently, Di Pietro et al. 45 and Dai et al. 15,46 reported static NMR studies of ILCs containing imidazolium cations, and inorganic anions ([NO 3 ] − , Cl − , and [BF 4 ] − ). The MAS NMR techniques used in this paper have not, to our knowledge, thus far been used to investigate ILCs.
In this work, we investigate the short-range (interactions between ions) and long-range structure (the occurrence of the partial bilayer LC phase), and molecular order of CAGE-oct, by MAS NMR, SAXS, and mass spectrometry. Mass spectrometry and the 1 H NMR chemical shift provide a "snapshot" of the hydrogen-bonding structures/ionic complexes present in CAGE-oct. SAXS reveals the intermediaterange order of CAGE-oct above the LC-to-isotropic transition, and long-range order of the partial bilayer LC phase at temperatures below ∼293 K. The LC texture and LC-toisotropic phase transition was also viewed via polarizing optical microscopy. Residual 13 C− 1 H and 1 H− 1 H dipolar couplings are measured by means of 1 H− 13 C CP and 1 H doublequantum (DQ) MAS NMR build-up curves, so as to obtain order parameters, S CH and S HH , which quantitatively describe the amplitude of motion of the ions (and molecular segments/ bonds) that comprise CAGE-oct.

EXPERIMENTAL SECTION
2.1. Sample Preparation. Choline bicarbonate (80 wt % solution in water) and octanoic acid were purchased from Sigma and used as received. Geranic acid (85%) was purchased from Sigma but was recrystallized five times from cold HPLCgrade acetone (VWR) before use. A solution of choline bicarbonate (2.67 g, 12.96 mmol) in water (3 mL) was added slowly to a stirred solution of geranic acid (2.18 g, 12.96   47 2.3. Mass Spectrometry. Mass spectra were obtained by gently warming CAGE-oct until it melted and then directly injecting it into a Waters single quadrupole detector at 4.2 kV capillary voltage using electrospray ionization (both positive and negative modes were obtained).
2.4. SAXS. SAXS measurements were made by a Xenocs Xeuss 2.0 equipped with a microfocus Cu Kα source collimated with Scatterless slits. The scattering was measured by a Pilatus 300k detector with a pixel size of 0.172 mm × 0.172 mm. The distance between the detector and the sample was calibrated by using silver behenate (AgC 22 H 43 O 2 ), giving a value of 0.540(5) m. The magnitude of the scattering vector (q) is given by q = 4π sin(θ)/λ, where 2θ is the angle between the incident and scattered X-rays and λ is the wavelength of the incident X-rays. This gave a q range for the detector between 0.023 and 0.73 Å −1 in the horizontal plane. The 2D detection image was integrated as a function of q producing the 1D intensity versus q data. For measurements as a function of temperature, the samples were held between two Kapton windows in a Linkam HFSX 350 furnace. Separate measurements of the Kapton windows and furnace were used for background corrections.
2.5. MAS NMR. MAS NMR experiments were performed on a Bruker Avance III spectrometer operating at a 1 H Larmor frequency of ν 0H = 500 MHz (11.7 T), using a 3.2 mm doubleresonance MAS probe. A spinning frequency, ν R , of 5 kHz was used. The temperature was calibrated using methanol, as described in the Bruker Instruments manual, with all stated temperatures corrected by this procedure. The 1 H and 13 C 90°p ulses were of duration 2.5 and 5 μs, respectively (corresponding to nutation frequencies, ν 1,H = 100 kHz, and ν 1,C = 50 kHz).
For 1 H− 13 C CP experiments, the 1 H and 13 C nutation frequencies were approximately 52.5 and 47.5 kHz, respectively. For the 2D heteronuclear correlation (HETCOR) experiment, CP was achieved by using a 70−100% ramp 48 on the 13 C channel, whereas CP build-up curves were recorded without a ramp to allow for quantitative determination of the dipolar coupling, D CH . SPINAL-64 49 heteronuclear decoupling was applied during acquisition at a 1 H nutation frequency of 52.5 kHz, corresponding to a pulse duration of 9.0 μs. Additional parameters are provided in section S3.1.
DIPSHIFT curves were recorded using both CP and direct polarization (DP) for the initial excitation procedure. For the 1 H− 13 C CP DIPSHIFT experiment, the 1 H and 13 C nutation frequencies were 52.5 and 47.5 kHz, respectively. CP was achieved by using a 70−100% ramp 48 on the 13 C channel for a contact time of 5 ms. Eight rotor periods of recoupling (N = 8) were used. The decoupling schemes for homonuclear and heteronuclear decoupling were frequency-switched Lee−Goldburg (FSLG) 50,51 and SPINAL-64, 49 respectively, applied at a nutation frequency of 62.5 kHz. Additional parameters are provided in section S3.1.
For 1 H DQ experiments, the excitation and reconversion of 1 H DQ coherences was achieved by the POST-C7 recoupling method 52 at a 1 H nutation frequency of 35 kHz. For 1 H DQ build-up curves (see pulse sequence in section 3.7 below), DQ selection (more accurately for 4n + 2 coherence orders, although the DQ coherence order can be assumed to be dominant) was performed by a four-step phase cycle for the reconversion block (0°, 90°, 180°, 270°) while inverting the receiver phase for alternating transients (0°, 180°, 0°, 180°), so as to select Δp = ±2, where p is the coherence order. A reference experiment was recorded in the same manner, however, keeping the receiver phase constant. Additional parameters are provided in section S3.1. 1 H and 13 C chemical shifts are referenced to tetramethylsilane (TMS) at 0 ppm by using L-alanine as a secondary reference (1.1 ppm for the lower ppm 1 H resonance and 177.8 ppm for the higher ppm 13 C resonance), corresponding to adamantane at 1.85 ppm ( 1 H) 53 and 38.5 ppm ( 13 C). 54 2.6. Simulations and Fitting. CP build-up and CP T 2 -recDIPSHIFT curves were fit by a simplex minimization implemented in the opt1.0 package in SIMPSON (v4.2.1). 55,56 Further details of the simulations and error analysis and representative SIMPSON input files are provided in the Supporting Information.

RESULTS AND DISCUSSION
3.1. Small Angle X-ray Scattering. SAXS patterns for CAGE-oct at temperatures ranging from 263 to 313 K are shown in Figure 2a. The first-order scattering peak provides information on the ordering, since the layer distance by Bragg's Law is d = 2π/q max . A single peak is observed at q ≈ 0.29 Å −1 (at 273 K), corresponding to a repeat distance of 22 Å. The first-order scattering peak shifts to higher q values at higher temperatures, corresponding to a decrease in the repeat distance and indicating a negative thermal expansion upon heating ( Figure S6). At temperatures below ∼293 K, the firstorder scattering peak is narrow and intense, indicative of a liquid-crystalline phase. The repeat distance of 22 Å (at 273 K), roughly the length of two molecules of octanoic/geranic acid, is analogous to the spacing observed in bilayer membrane systems. Smectic LC phases (or lamellar phases) are the most commonly observed phase for ILCs and are characterized by the stacking of layers that exhibit some positional and orientational order (Figure 2b). The layer formation of smectic ILCs is driven by Coulombic forces and nanosegregation of the charged moieties from the hydrophobic chains: 5 forces that are expected to dominate the intermolecular interactions of CAGE-oct. In contrast, the nematic phase, which displays only orientational order, is rarely observed for ILC systems 57,58 and is unlikely to occur for The Journal of Physical Chemistry B pubs.acs.org/JPCB Article CAGE-oct owing to the lack of nematic phase favoring features such as π−π interactions (i.e., aromatic rings). 59 At temperatures above ∼293 K, the SAXS profile becomes significantly broader and closely resembles those observed in ILs based on the 1-alkyl-3-methylimidazolium cation ([C n C 1 im] + ) of various intermediate alkyl chain lengths (3 < n < 10). 60,61 While a micelle-like model for self-assembly in ILs (with alkyl chains longer than a butyl group) was initially suggested to account for the diffraction peak at low q, 61,62 more recent studies have shown that the diffraction profile may be interpreted as a bicontinuous sponge-like mesoscopic structure, 63−65 resulting from clustering of polar and nonpolar domains ( Figure 2b). This has further been evidenced by simulations for a number of ILs. 66,67 We note that even the commonly referred to "isotropic state" of many ILs displays some degree of nanoscale structure and charge ordering, although not to the same extent or length scale as for ILCs. The persistence of some short-range order above the smecticto-isotropic transition temperature has been observed in carbohydrate-based LCs, where a low enthalpy/entropy Smto-isotropic transition was suggested to be a result of hydrogen-bonded aggregates remaining in the isotropic liquid phase. 68, 69 3.2. Polarizing Optical Microscopy. Figure 3 shows a polarizing optical microscopy image of the LC phase of CAGE-oct at 278 K. There appear to be two distinct birefringent areas with larger scale and smaller scale patterning (solid red oval and dashed red oval, respectively). Both of these optical textures are conic fan textures characteristic of a smectic A (SmA) structure 70−74 and the apparently distinct patterns are likely to arise from differences in nucleation. 75 In addition, there are regions with no birefringence present between LC domains, indicating isotropic liquid regions. These results are consistent with the presence of an isotropic component coexisting with the LC phase as detected by NMR (discussed below in section 3.7 and in section S4.2). On heating, the LC phase melts to an isotropic phase between 290 and 298 K ( Figure S7); this transition is also seen in the SAXS results described above. On cooling from the isotropic phase, the first birefringence patterns appear between 291 and 293 K.
3.3. Mass Spectrometry. Careful study of the negative ion mass spectrum of CAGE-oct (see section S2 and Figures S3  and S4 for the full spectrum analysis) reveals evidence of interactions between choline and carboxylate anions. As expected, both "free" octanoate and geranate anions are observed under mass spectrometry conditions, but an ion corresponding to CAGE-oct, [(choline)(octanoate)-(geranate)] − , is also seen at a mass to charge ratio (m/z) 414 (with loss of a proton). This peak is about 70% of the intensity of the "free" octanoate peak, implying that these clusters are significantly more stable than would otherwise be expected from simple electrostatic-based ion clustering, probably due to the hydrogen bonding interactions between the hydroxyl group of the choline cation and the carboxylate anions. Further peaks at m/z 390.4 and 438.4 correspond to [(choline)(octanoate) 2 ] − and [(choline)(geranate) 2 ] − , respectively. The ratio of these three clusters is close to the statistically expected 1:2:1 even though CAGE-oct was synthesized by adding 1 molar equiv of 1-octanoic acid to in situ-generated choline geranate, implying anion metathesis is taking place. These peaks are significantly more intense (at least an order of magnitude) than those of the [(anion) 2 (H)] − species, suggesting that there is a strong association of the choline cation with the carboxylate anions. Many larger ion clusters generated by electrostatic ion clustering, for example, m/z 661.6 [(choline) 2 (octanoate) 2 (geranate)] − , were observed at <5% of the intensity of "free" octanoate, which is more typical of electrostatic ion clusters, especially those found in the mass spectra of ionic liquids. 76 3.4. 1 H Variable Temperature MAS NMR. In order to investigate the nature of the interactions between the components of neat CAGE-oct, proposed by mass spectrometry, we performed 1D 1 H MAS NMR spectroscopy ( The mechanism of this metathesis is key to establishing whether CAGE-oct (and, by extension, CAGE) can be classified as an ionic liquid or not. If the mechanism generates free octanoic acid (or geranic acid), then, by definition, the mixture is no longer "a liquid composed entirely of ions" and it cannot be considered as an ionic liquid. However, if the mechanism does not require the formation of free acid to proceed, then CAGE-oct can be considered to meet the   A single peak is observed for the COOH and OH 1 H resonances, which is indicative of fast chemical exchange on the NMR time scale between the carboxylic acid proton and the hydroxyl proton of choline. The presence of hydrogen bonding between the components is evidenced by the COOH/OH 1 H resonance appearing at 9.5 ppm (at 273 K), a value that corresponds to the average of the COOH and OH resonances, which are expected to lie at approximately 13−15 and 5−6 ppm, respectively. Note that the exact position of this peak is highly sensitive to ppm concentrations of water in the sample. The chemical shift of protons participating in hydrogen bonds is known to be temperature sensitive due to changes in the degree of excitation of the hydrogen-bondstretching vibrational mode 77−80 and has been shown to correlate with the nature and strength of the hydrogen bond. 80 (Figure 4b), is consistent with intermolecular hydrogen bonding. 83,84 The hydrogen bonding arrangement involves the carboxylic acid, carboxylate, and hydroxyl groups, possibly as hydrogen-bonded clusters such as [(choline)(geranate)(H)(octanoate)], as suggested by mass spectrometry, and is likely dynamic, where rapid formation and subsequent breaking of hydrogen bonds occurs between the carboxylic acid, carboxylate, and hydroxyl groups. The reported hydrogen bond lifetimes for ILs are typically on the order of picoseconds. 85−87 The presence of hydrogen bonding has, in some cases, been reported to drive the formation of LC phases. 88 No discontinuity in the temperature-dependent chemical shift of the COOH/OH resonance was observed at the LCisotropic phase transition temperature of CAGE-oct (as identified from the SAXS data in Figure 2a). The same temperature coefficient for the LC and isotropic phase of CAGE-oct suggests that a similar hydrogen bonding arrangement persists in the isotropic phase, consistent with reports of the behavior of other LC systems. 68,69 3.5. 2D 1 H− 13 C CP Heteronuclear Correlation MAS NMR. A 2D 1 H− 13 C CP-HETCOR MAS spectrum of CAGEoct is shown in Figure 5. The spectrum was recorded at 273 K, as were all further NMR experiments reported in this study; this temperature is below the LC-to-isotropic transition temperature indicated by SAXS and polarizing microscopy. Under 5 kHz MAS, all of the 13 C resonances are resolved. With a relatively long contact time of 2.5 ms, only a single 1 H− 13 C cross-peak was observed for all carbons with a directly bound proton corresponding to the one-bond C−H connectivities. This indicates that the dipolar couplings are highly averaged by molecular motion, implying that CAGE-oct is a dynamic system which is undergoing a rapid anion metathesis on the NMR time scale. For comparison, in rigid solids, cross-peaks corresponding to longer range C−H proximities can already be observed at contact times of only 200 μs. 89 We note the absence of any contacts to the COOH/OH proton (see Figure S8 for the full spectral range) despite the carbonyl carbons (C-1 and C-1″) showing a dipolar coupling to the protons of their adjacent carbon (H-2 and H-2′′, respectively), which are expected to be located slightly further away than the acidic proton. This suggests that the absence of CP from the COOH/OH proton is due to motion of the exchangeable proton between the ions; i.e., the COOH and OH protons do not stay within close enough proximity of any one carbon long enough for the observable buildup of the 1 H− 13 C CP signal to occur.
3.6. 1 H− 13 C MAS NMR: C−H Dipolar Couplings. In order to investigate the molecular dynamics of CAGE-oct in the LC phase, CP build-up curves were recorded to quantify the C−H dipolar couplings. The magnitude of the 13 C− 1 H dipolar coupling is dependent on both the distance and orientation of the involved nuclei. For a specific chemical moiety, e.g., a covalently bound C−H, there is a fixed internuclear distance such that the measured dipolar couplings therefore report on the extent of averaging of the dipolar coupling by motions that are fast on the time scale of the NMR experiment (>100 kHz). CP build-up curves were recorded by incrementing the CP contact time up to a maximum of 2 ms ( Figure 6). Clear oscillatory behavior was observed for all of  The Journal of Physical Chemistry B pubs.acs.org/JPCB Article the CH and CH 2 groups of the geranate and octanoate with a maximum intensity occurring between 100 and 600 μs, while smoother curves with slower initial build-up and faster relaxation were observed for the methyl groups. Dipolar couplings for all carbons to their directly bonded protons were obtained by iterative least-squares fitting to simulated build-up curves (using SIMPSON, 55,56 see sections 2.6 and S4.1) for a 2-, 3-, and 4-spin system for CH, CH 2 and CH 3 groups respectively, excluding C-1′ and C-3′ of choline for which the CP intensity was too low to accurately obtain D res . The inclusion of proton−proton dipolar couplings was found to have no significant effect on the fits (Table S1) and therefore only the CH dipolar coupling was included in the fitting (see Figure S9 and further discussion in section S4.1). The inclusion of a single exponential relaxation time as a second fitting parameter also did not significantly alter the obtained value of D res (Table S1); however, a statistically significant improvement in the fit was obtained for some curves and the relaxation term was therefore included. The resulting best-fit curves are shown by the solid lines in Figure 6 and the dipolar couplings, which ranged from 0.88 to 3.95 kHz, are given in Table 1.
In addition to the measurement of 13 C− 1 H dipolar couplings by means of CP MAS NMR build-up curves, these were also determined by the dipolar-chemical shift correlation (DIP-SHIFT) method. As the CH dipolar couplings in CAGE-oct are relatively weak, a recoupled version of the original DIPSHIFT experiment was utilized, where the phase accumulated due to the evolution under the dipolar coupling during t 1 is amplified by a REDOR-like train of π pulses. 90 43 The pulse sequence for the T 2 -recDIPSHIFT experiment is shown in Figure S10a. After excitation (CP or DP), the magnetization evolves during a variable time t 1 under the influence of the heteronuclear dipolar interaction, while the 1 H homonuclear coupling is suppressed by FSLG decoupling. DIPSHIFT curves are shown in Figure S10b. The resulting DIPSHIFT profiles ( Figure S10b) return to close to the initial maximum intensity at t 1 = Nτ r , indicating that there are no significant intermediate  Table 1. The one-dimensional spectrum shown vertically was recorded with a CP contact time of 2.5 ms.
Fitting details are given in sections 2.6, 3.6, and S4.2. b Too low to extract from the experimental data.
The Journal of Physical Chemistry B pubs.acs.org/JPCB Article time-scale motions with rates between 1 to 100 kHz in CAGEoct. 91,92 This confirms that our treatment of the reduction in the dipolar coupling described by the order parameter to provide information on the amplitude of fast motions is valid. A discussion of the differences in DP-and CP-DIPSHIFT profiles is provided in section S4.2. The values of D res obtained from fits of the CP-DIPSHIFT profiles are given in Table 1 and the resulting order parameters, S CH , are graphically compared to those obtained from CP build-up curves in Figure 7. Overall, the two methods produced values with very good agreement, within error for most of the sites. Notably, the largest discrepancies between the two methods were observed for the methyl groups (C-9, C-8, C-10, and C-8″). The CP build-up curves of the methyl groups ( Figure 6) did not show clear oscillations as were observed for the CH and CH 2 , which could account for slightly less accurate values. 3.7. 1 H DQ MAS NMR Build-up Curves: H−H Dipolar Couplings. As shown in the previous section, dipolar couplings can be obtained from 1 H− 13 C CP MAS NMR build-up curves; however, this method has some disadvantages due to the low signal-to-noise ratio of carbon detection and inaccuracies in the determination of the smaller coupling constants below 1 kHz. We therefore also looked at the measurement of residual 1 H− 1 H dipolar couplings by DQ MAS NMR. This experiment has the advantage of usually good sensitivity due to the high natural abundance and gyromagnetic ratio of protons, and for highly mobile systems such as the LC under study, good resolution is achieved with low MAS frequencies. Small 1 H− 1 H dipolar couplings, as low as 100 Hz, are also accessible from DQ build-up curves.
A 2D 1 H DQ-1 H SQ MAS NMR spectrum of CAGEoctanoic acid recorded at a MAS frequency of 5 kHz is presented in Figure 8. The excitation and reconversion of DQ coherence was achieved by the POST-C7 52 Figure S11 for full spectral range), indicating that they lack 1 H− 1 H RDCs, consistent with our observations of high mobility from 1 H− 13 C CP MAS NMR experiments. The highest intensity peaks are the auto DQ cross-peaks between like-spins (i.e., protons within CH 2 and methyl groups) for all resonances excluding the alkene protons, for which DQ peaks are observed to the aliphatic region (H-2 to H-4 and H-6 to H-8 at 5.7 + 2.1 = 7.8 and 5.1 + 1.7 = 6.8 ppm, respectively).
The partially averaged 1 H− 1 H dipolar couplings report on the amplitude of fast motions of the internuclear vector. The 1 H− 1 H dipolar couplings and corresponding order parameters, S HH , also offer complementary insight due to the different angle between the dipolar coupling and rotation axis as compared to 13 C− 1 H results. In order to extract the 1 H− 1 H dipolar coupling constants, and the related order parameter, S HH , DQ build-up curves were recorded by incrementing the τ DQ of the excitation and reconversion blocks (see pulse sequence in Figure 9a). While relative dipolar couplings are accessible from the buildup of the DQ intensity, 93 intensity normalization is necessary to obtain absolute values of the dipolar coupling. 94 This is achieved by recording a second analogous reference experiment, without the selection of DQ coherences by omitting the receiver phase alternation (for further phase cycling details see section 2.5). 34,94 The sum of the DQ (I DQ ) and reference intensities (I ref ) is termed the multiple-quantum sum intensity, I ∑MQ = I DQ + I ref . 34,94 The effect of relaxation is removed from the DQ build-up by dividing the DQ intensity by the multiple-quantum sum intensity: It is expected that, in the long-time limit, I DQ = I ref , and the normalized intensity, I nDQ , should therefore reach a plateau value of 0.5. 94 The normalized DQ build-up curves may be fit by an inverted-Gaussian function derived with the secondmoment approximation: 41 Table 1) and CP-T 2 -recDIPSHIFT experiments ( Figure  S10b, Table 1) at 273 K. The atom order corresponds to decreasing S CH as determined from CP build-up curves.
The scaling factor, k, in eq 2 is pulse-sequence specific and k C7 = 0.86/15 ≈ 0.057 for POST-C7. 97 Note that the empirical factors in eq 4 have been adjusted to take into account the scaling factor for POST-C7 recoupling, compared to those previously reported for the static case. 99 The A-l function extends the fitting range for DQ build-up curves, enabling fitting of the local maximum, extending above 0.5, which is pronounced in systems with a narrow RDC distribution. 94 For samples with a wide distribution of RDCs, the maximum vanishes, and eq 4 may be combined with a distribution of RDCs. 99 I DQ , I ref , and I nDQ for a representative peak at 1.3 ppm due to the octanoic acid alkyl chain protons H-4″-7′′, are shown in Figure 9b. The normalized DQ build-up curves for all 1 H peaks of CAGE-oct are shown in Figure 9c. The build-up curves do not reach the expected plateau level of 0.5 for any of the protons. This effect has previously been observed for elastomer systems and attributed to isotropically mobile components (short dangling chains and solvent molecules). 94 Such liquidlike components contribute to the reference intensity, but not I DQ (as they lack RDCs), meaning that the normalized intensity, I nDQ , will be lower than expected. In some cases, it has been demonstrated that these components form more slowly relaxing long-time tails of I ref , which may be fit to an exponential and subtracted to produce a well-behaving normalized build-up curve reaching the 50% intensity plateau. 94 For CAGE-oct, the long-time relaxing component is similar in both the DQ and reference intensity (Figure 9b), and it was therefore not possible to unambiguously subtract an exponential tail from the reference curve. A method to solve this problem is based on the fact that in the long-time limit, where I DQ = I ref , the more mobile fraction may be identified and fitted in a plot of I ref − I DQ vs τ DQ ( Figure S12). 100 Figure  9d shows the normalized build-up curves with the long-time exponential tail subtracted from the reference curve in this way (see Figure S12), which reach the expected plateau of 0.5. Most of the build-up curves show a pronounced maximum, indicative of a narrow RDC distribution. The build-up curves were fitted to eq 4 and the values of D res , which range from 0.5 to 3.7 kHz, are given in Table 2.
In order to probe whether the lower than expected plateau (Figure 9c) is a result of liquid-like signal contributions (i.e., lacking orientational order), we applied the experimental approach of DQ preselection. Saalwachter et al. introduced this strategy as a means to suppress such unwanted contributions arising from dangling chain ends and solvent molecules in polymer networks, as an alternative to the exponential tail correction. 94 This strategy involves applying an additional fixed DQ excitation/reconversion block, followed by a z-filter delay, τ z , preceding the incremented DQ pulse sequence of the original experiment ( Figure S13a). The first block acts as a DQ  Figure S13b,c). The resulting normalized curves reach a maximum value of between 15 and 26% higher than the curves recorded with no DQ preselection filter and are significantly closer to reaching the expected plateau of 0.5. Fits to eq 2 give D res /2π values comparable to those obtained from fits to data recorded without DQ preselection, following tail subtraction (Table S3). The difference between the DQ build-up curves with and without DQ preselection suggests that, within the liquid crystalline sample, there are some less ordered components (either unordered regions or individual ions/clusters within the ordered phase) that do not have RDCs or have very small RDCs and justifies the subtraction of the long-time tail from the data recorded with no DQ preselection. Further evidence of an isotropic component is provided by 1 H− 13 C DIPSHIFT experiments recorded with either direct polarization or 1 H− 13 C CP, given in the Supporting Information ( Figure S10b, Table S2, and additional discussion in section S4.2). We estimated a disordered component of approximately 23%, in agreement with our observations from 1 H DQ build-up experiments, and the presence of isotropic regions observed by polarizing optical microscopy. The coexistence of liquid disordered/liquid ordered phases has previously been reported in a model membrane system and quantified by 1 H− 13 C dipolar coupling MAS NMR. 101 3.8. S CH and S HH Order Parameters. Order parameters, S CH and S HH , were calculated as the ratio of the motionally averaged dipolar coupling to the rigid limit value and are reported in Tables 1 and 2, respectively. The order parameters are assigned a color on the basis of their range to show the structural correlation in Figure 10a (S CH ) and Figure 10b (S HH ). Since the dipolar coupling depends on orientation according to 1 / 2 (3 cos 2 θ − 1) (see section S1), S reports on the amplitude of fast motions (of the ions and molecular segments/bonds). S CH ranged from ∼0 to 0.224 while S HH ranged from ∼0 to 0.173 (in comparison to S = 1 for a rigid system). These order parameters are significantly lower compared to those of smectic phases of nonionic LCs, which typically display order parameters of 0.4−0.8 102−104 and are rather similar to those reported for lamellar phases formed by lyotropic mesogen/solvent systems. 105−107 Highly disordered alkyl chains corresponding to low orientational order parameters have been found for other ILCs. 15,45,46,108 The differences in order parameters between conventional nonionic LCs and ILCs such as CAGE-oct can be attributed to differences in the molecular structures and nature of the intermolecular interactions, which are the driving forces for mesophase formation. Nonionic LCs typically contain a rigid (usually aromatic) core and one or more flexible chains, with the LC phase stabilized by dispersion forces. 3 ILCs, however, consist of ions, one of which usually contains a long (n > 8) alkyl chain, and typically form smectic phases stabilized by electrostatic (Coulombic) interactions within ion-rich layers and, in some cases (likely the case for CAGE-oct), hydrogen bonding. 88 Distinct motional amplitudes were observed between the choline cation, compared to the case for the anions, and also within the ions. From the trends in these order parameters, conclusions on the structure as well as dynamics can be drawn. Choline showed lower order parameters compared to those of geranate and octanoate, indicating increased mobility. Of the choline carbons, the CH 2 adjacent to the hydroxyl group, C-2′, showed the highest order parameter, S CH = 0.039. In comparison, the adjacent CH 2 group, C-1′, and the choline methyl group, C-3′, showed much smaller dipolar couplings, indicating nearly isotropic freedom of motion (S CH ∼ 0) of the choline headgroup. The 1 H− 1 H RDC for the choline methylene group H-1′ for which the corresponding 1 H− 13 C RDC was too low to be accurately determined from the CP build-up curves, was determined to be 546 Hz, corresponding to S HH = 0.0257. We note that the calculated order parameters consider only the dominant dipolar coupling, the autocoupling between the methylene protons, for D rigid , while the measured RDC is an averaged value of all couplings. Taking into account   in the fluid lamellar (L α ) phase of magnitude 0.005 (γ; corresponding to C-3′) and 0.04 (α and β; corresponding to C-3′ and C-2′), 106,109 corresponding to purple and blue, respectively, on the scale in Figure 10a. Notably, the choline headgroup is covalently attached via the oxygen in PC, yet the adjacent α proton displays a remarkably similar order parameter to the equivalent C-2′ of the "free" choline in CAGE-oct (S CH = 0.04 and 0.039), further suggesting strong hydrogen-bonding of the choline cation to geranate/octanoate. The S CH of the aliphatic chain protons of geranate and octanoate are generally lower at the terminal end compared to those closest to the carboxylic acid group. S CH of the methylene groups of octanoic acid vary between 0.174 at C-2″ to 0.082 at C-7′′, while those of the geranate backbone show similar values and are again slightly higher at the backbone carbons closest to the carboxyl group (C-2 and C-4, compared to C-6 and C-10; Figure 10a). The methyl groups branching from the main chain of geranic acid, C-9 and C-8, have the highest S CH , which could result from the C−H vectors being more closely aligned with the long molecular axis of geranic acid, which would result in the least averaging due to axial rotation of the molecule about the long axis. The S HH for the overlapping octanoic acid alkyl chain protons H-4″-7′′ of 0.151 is close to the highest S CH obtained for these carbons (S CH = 0.136 for C-4″), as expected since the strongest dipolar couplings will dominate the buildup of DQ coherences. The S HH of 0.173 for protons H-2″, 4, 5, and 9 agrees with S CH of 0.174 of C-2″. The S HH of the terminal methyl group of octanoate (H-8″) is 0.151, compared to S CH of 0.120. As for choline, the order parameters of the alkyl chains resemble those of lyotropic phospholipid systems, which range from ∼0.24 to 0.02 along the acyl tail of PC in the L α phase (corresponding to red to purple, on the scale in Figure 10a). 109 Overall, for CAGE-oct, S HH and S CH are observed to provide the same information about the dynamics of the system (Figure 10a). The order parameters, S CH , reported above at 273 K, do not change significantly at lower temperatures (243 and 253 K; Figure S14).

CONCLUSIONS
In this work, a neat system based on choline, geranic acid, and octanoic acid has been investigated by MAS NMR, SAXS, polarizing optical microscopy, and mass spectrometry. At temperatures below ∼293 K, CAGE-oct forms a mobile bilayer liquid-crystalline phase. Mass spectrometry reveals the presence of [(choline)(geranate)(octanoate)] − , [(choline)-(octanoate) 2 ] − , and [(choline)(geranate) 2 ] − species at intensities significantly higher than expected for electrostatic ion clusters when compared to those for [(anion) 2 (H)] − species, suggesting strong association of the choline cation with the anions. 1 H MAS NMR suggests that these species are hydrogen-bonded clusters, with hydrogen bonds involving the carboxylic acid, carboxylate and hydroxyl groups. The hydrogen bonding is dynamic, with both anion metathesis and hydrogen exchange taking place, resulting in averaging of the 1 H chemical shifts. The temperature dependence of the 1 H chemical shifts suggests that a similar hydrogen bonding arrangement occurs in both the isotropic and LC phases.
The CAGE-oct ILC was further investigated by means of MAS NMR spectroscopy, which provides information about the mobility and degree of ordering of the individual ions and molecular segments, which is not accessible from other techniques such as SAXS. The dynamic order parameters, S CH and S HH , in the range 0−0.2, indicate a highly mobile LC phase, with the obtained ordered parameters being considerably lower than those reported for typical nonionic LCs, and instead resembling those of aqueous phospholipid bilayer systems. The increased mobility toward the terminal ends of geranate and octanoate alkyl chains and the choline −N-(CH 3 ) 3 headgroup are consistent with hydrogen bonding of the choline cation and anions and indicate that the CAGE-oct LC consists of mobile ionic-rich layers and regions of fluid alkyl chains. The coexistence of an isotropic component (approximately 25% of the sample) was also detected within the LC phase. This present work shows that MAS NMR is a highly promising tool for studying the site-specific structure and dynamics of mesogens, utilizing heteronuclear techniques such as the build-up in 1 H− 13 C CP or 1 H DQ MAS NMR experiments.
Additional details on the dipolar coupling and order parameter; additional spectroscopic characterization (solution phase 1 H and 13 C{ 1 H} NMR spectra, IR spectrum, and mass spectra); additional experimental details; determination of 95% confidence intervals for D CH ; temperature dependence of the layer repeat distance; polarizing optical micrographs; 1 H− 13 C HETCOR MAS NMR spectrum; additional MAS NMR discussion; further fits of experimental 1 H− 13 C CP build-up curves; effect of proton−proton dipolar couplings on simulated CP build-up curves; DIPSHIFT profiles; 1 H DQ-1 H SQ MAS NMR spectrum; plots of I ref − I DQ ; 1 H DQ build-up curves recorded with DQ preselection; temperature dependence of S CH ; example SIMPSON input files (PDF) The Journal of Physical Chemistry B pubs.acs.org/JPCB Article