Spectroscopic investigation of the structure of a pyrrolidinium-based ionic liquid at electrified interfaces

Ionic liquids (ILs) are electrolytes that solely consist of ions. Compared to conventional electrolytes, such as aqueous and organic solutions, ILs have low volatility, large electrochemical stability window (ESW), and tunable molecular structure, thus showing promise for applications in energy conversion and storage.^1,2 However, the high complexity of the molecular structure also brings challenges in understanding their electrochemical properties. In particular, at the electric double layer (EDL) region where electrochemical processes occur, the molecular configurations of ILs remain elusive.² For aqueous and organic solutions, classical theories such as Helmholtz, Gouy–Chapman, and Stern models have served as reasonable approximations of the EDL structure at typical ranges of salt concentration (0–1 M).^3,4 For ILs, such classical electrostatic theories fail to work due to the high ionic strength and strong inter-molecular interactions.^1,2,5,6 Therefore, in the past two decades, significant efforts have been devoted to developing mean-field/continuum theories to describe the EDL structure of ILs. These models have predicted unique properties of ILs, such as overscreening and crowding, and have qualitatively explained the overall trend of potential-dependent EDL capacitance of ILs.^5–9 However, these theories oversimplify the shape of the constituent ions as spheres and, thus, cannot accurately predict the molecular structure and quantitative capacitance values of EDLs of specific IL systems.

To understand the EDL structure of ILs, various in situ characterization methods have been used to probe the electrode–electrolyte interfaces, including vibrational spectroscopies such as surface-enhanced Raman and infrared spectroscopy,^10–13 as well as imaging techniques, including atomic force and scanning tunneling microscopy.^14,15 The most widely studied ILs are imidazolium-based, where the imidazolium ring in the cations contains delocalized electronic states.^10–12,16 Due to such delocalization effects, the positive charge in the imidazolium cations spreads over the whole ring.¹⁷ As a result, the cation molecules can be roughly approximated as uniformly charged spheres/ellipsoids and respond to potential changes by adjusting their distance from the electrode, as revealed in existing microscopy and spectroscopy results.^10–12,14 In recent years, however, there has been increasing awareness that the presence of an acidic C2 proton in the imidazolium ring limits the cathodic stability of these ILs, which strongly hinder their electrochemical applications.^18–20 As an alternative, pyrrolidinium-based ILs have gained increasing attention due to their larger ESW compared to the imidazolium counterparts as well as their superior thermal stability.^20–23 In contrast to the delocalized imidazolium ring, the pyrrolidinium ring contains a localized charge. Due to the strong charge inhomogeneity inside the cations, their structural response in the EDL at different electrode potentials is likely more convoluted and sensitively dependent on the exact atomic arrangement of the ionic species.

We choose to study 1-butyl-1-methylpyrrolidinium bis(trifluoromethylsulfonyl)imide (BMPy-TFSI), a pyrrolidinium-based IL that has been widely explored for electrochemical applications, such as lithium-ion batteries and supercapacitors [Fig. 1(a)].^19,24–28 We use electrochemical shell-isolated nanoparticle-enhanced Raman spectroscopy (EC-SHINERS), a special type of surface-enhanced Raman technique, to probe the molecular structure of EDLs of BMPy-TFSI on Au electrodes [Fig. 1(b)]. By depositing Au (50–60 nm)/SiO₂ (∼2 nm) core–shell nanoparticles on the Au electrode, we create nano-gaps between Au particles and Au surface, where the strong electric field can result in Raman enhancement up to ∼10⁸.²⁹ Therefore, SHINERS is highly sensitive to the liquid species present within ∼1 nm from the Au electrode surface, and the smaller the molecular group–electrode distance, the higher the Raman enhancement factor. Previously, EC-SHINERS has been used to determine the EDL structure of aqueous solutions, deep eutectic solvents, and imidazolium-based ILs.^{10,11,30–32} Here, we extend its application to pyrrolidinium-based ILs. Considering that the BMPy⁺ and TFSI⁻ ions each contain unique functional groups at different sites within the molecules, the change in the Raman peak intensity of each group is an indicator of its distance away from the electrode surface. In addition, peak position shifts, if occur, will shed light on the molecular interaction mechanisms. Therefore, by analyzing the evolution of peak intensity and position of individual molecular groups at a range of electrode potentials within the ESW, we can extract the location and orientation of each ionic species and decipher the EDL charging mechanism.

Gold(III) chloride trihydrate (99.9%, HAuCl₄ · 3H₂O, Sigma-Aldrich), trisodium citrate dihydrate (99.0%, Na₃CA, Alfa Aesar), (3-aminopropyl)trimethoxysilane (97%, APTMS; Alfa Aesar), hydrochloric acid (TraceMetal Grade, Fisher Chemical), sodium silicate solution [reagent grade, Na₂O(SiO₂)_x · xH₂O, Sigma-Aldrich], acetone (≥99.5%, Fisher Chemical), and isopropanol (99.9%, Fisher Chemical) were purchased and used without further purification. Water used in this work (18.2 MΩ cm at 25 °C) was purified by a Milli-Q system.

Au/SiO₂ nanoparticles were synthesized following a literature reported method.³³ First, 10 ml of HAuCl₄ (0.01 wt. %) aqueous solution was heated in a 20 ml glass vial in an oil bath. The solution was stirred at 300 rpm and allowed to boil, before 70 μl of Na₃CA (1 wt. %) aqueous solution was added into the boiling solution using a pipette in one shot. The solution was allowed to boil for another 30 min. The vial was taken out of the oil bath and allowed to cool down to room temperature. This Au nanoparticle suspension can be stored at room temperature in dark for weeks without quality degradation. After synthesizing Au nanosphere seeds, we proceeded to coat them with SiO₂. First, 1 mM APTMS solution and 0.54 wt. % sodium silicate solutions were freshly prepared. The 1 mM APTMS aqueous solution was obtained by adding water to APTMS. The 0.54 wt. % sodium silicate solution was prepared by adding 0.2 ml of sodium silicate solution (27%) into 3.8 ml H₂O, followed by adding 6 ml 0.01–0.02 M HCl, and then thoroughly agitating by a vortex shaker. After preparing these solutions, 3 ml of Au nanosphere solution was added into a 20 ml glass vial and mixed with 0.04 ml of APTMS (1 mM) solution. The solution was allowed to stir at room temperature at a speed of 400 rpm for 15 min. Then, 0.32 ml of sodium silicate solution (0.54 wt. %) was added into the glass vial and mixed by stirring at 400 rpm for 3 min. The glass vial was then sealed by parafilm, placed in a 90 °C oil bath, and stirred at 300 rpm. After 2 h, the reaction was quenched by placing the glass vial in an ice bath for at least 10 min. The synthesized Au/SiO₂ nanoparticles were collected using centrifugation at a speed of 2020 × g for 10 min for two times. The final Au/SiO₂ nanoparticles were re-dispersed in 200 μl water and stored at room temperature under dark conditions.

The morphology of the synthesized Au/SiO₂ nanoparticles was characterized using transmission electron microscopy. As shown in the supplementary material (Fig. S1), a uniform SiO₂ coating was observed on the Au core. We also performed cyclic voltammetry (CV) tests of the synthesized nanoparticles in aqueous solutions and observed a nearly complete blocking of Au oxidation and AuO_x reduction in the Au/SiO₂ nanoparticles (Fig. S2), which further confirms the uniform, pinhole-free SiO₂ shell coating.

BMPy-TFSI (99.9%) was purchased from Iolitec, vacuum annealed at 105 °C for ∼24 h, and stored in an Ar glovebox. Au electrodes were prepared by electron-beam deposition of Au onto a Si(100) substrate. Specifically, the Si(100) with native oxide was sonicated with acetone and isopropanol, followed by 3 min oxygen plasma cleaning. 5 nm Ti and 100 nm Au were deposited onto the cleaned Si in sequence by electron beam deposition using a Temescal e-beam evaporator. We further prepared the electrochemical cell using our previously demonstrated setup.¹⁴ We used a Pt counter-electrode and a Pt quasi-reference electrode. The working electrode, with an active area of ∼0.64 cm², consists of either the bare Au film or Au/SiO₂ nanoparticles on the Au film. The latter was prepared by drop casting the Au/SiO₂ nanoparticles on the Au film and drying under nitrogen. 100–120 μl of BMPy-TFSI was added to the cell to serve as the electrolyte.

CV was carried out inside an Ar glovebox using a CHI 760D electrochemical workstation. The working electrode was the bare Au film. The scan rate was 10 mV/s.

We used the Au/SiO₂ nanoparticle-coated Au film as the working electrode. EC-SHINERS was performed using a confocal Raman imaging system (Horiba LabRAM HR 3D-capable Raman spectroscopy) in air, right after the BMPy-TFSI was taken out of the glovebox, and each measurement lasted no more than ∼8 h. A laser with a wavelength of 633 nm and a power of ∼3.5 mW was focused on the electrode surface using a 50× objective. Except when otherwise mentioned, Raman spectra were taken at 10 s collection time, three accumulations, and 300 grooves/mm grating. The electrode was electrochemically cleaned by applying either high positive or negative potential for a few minutes, before Raman spectrum collection. We used the bulk spectrum as well as the 742 cm⁻¹ SHINERS peak of BMPy-TFSI to calibrate the obtained Raman spectra. The latter was found to be independent of either the laser focus position (either in bulk electrolyte or on electrode surface) or the electrode potential. We used the Voigt function to perform Raman peak fitting and quantitative analysis.

Raman measurements of bulk electrolytes were performed using the same confocal Raman system. In this case, the electrolyte was drop-casted on a Si substrate forming a droplet, and the laser was focused ∼100 µm above the Si surface to ensure the measured spectra correspond to the bulk electrolyte.

We first perform CV measurements. As shown in Fig. 2, we observe a large ESW of ∼−2.3 to 2.3 V (vs Pt quasi-reference electrode) for BMPy-TFSI on Au. Note that, at ∼1.5 V, the small spike in the anodic current corresponds to Au oxidation, as evident in the EC-SHINERS results showing an AuO_x peak at ∼572 cm⁻¹ (Fig. S3). Considering the range of ESW of BMPy-TFSI, Au oxidation effects at high anodic potentials, and possible Au restructuring at high cathodic potentials,^21,34,35 we choose a potential range of −0.5 to 1.1 V for EC-SHINERS measurements to ensure that the Raman peak changes are dominantly due to the reconfiguration of the EDL structure. Since the EC-SHINERS measurements are carried out in air, we also perform CV both in the Ar glovebox and in air in a small potential range and compare the results. As shown in Fig. S4, we observe very similar EDL charging behaviors. Therefore, our EC-SHINERS results should not be affected by the air atmosphere.

The overall EC-SHINERS spectra are shown in Fig. 3. As a control, we also measure and plot the Raman spectrum of bulk BMPy-TFSI. We observe clear potential-dependence of the SHINERS peaks and find that their intensity is higher than that of the bulk spectrum (despite the 20 times longer exposure time of the bulk measurements), both of which confirm the interface-sensitivity of the SHINERS results. We perform EC-SHINERS at different spots of multiple samples and observe the same peak positions and similar potential-dependence (Figs. S5 and S6). Except the low-frequency mode at ∼180 cm⁻¹, all the other observed peaks can be assigned to well-known vibrational modes of the functional groups inside the cations or anions.

We first analyze the peak intensity of the known intramolecular vibration modes at different electrode potentials and plot the results in Fig. 4 (average of three different measurements: Nos. 1, 2, and 3 as shown in Figs. 3, S5, and S6). The key components of the BMPy⁺ cation are the pyrrolidinium ring and the butyl chain, both of which have characteristic vibrational peaks in our measured spectral range. The peak at ∼903 cm⁻¹ can be assigned to the ring breathing mode of BMPy⁺,^36–39 whose intensity is found to remain unchanged as the potential varies from −0.5 to 1.1 V [Fig. 4(a)]. In contrast, the intensity of the ∼1065 cm⁻¹ peak, identified as the C–C stretching of the trans alkyl chain,^40–42 continues decreasing at more positive potentials and eventually reaches a small, constant value as the potential becomes ∼0.5 V or higher [Fig. 4(a)]. Note that the bulk spectrum shows a negligible peak at ∼1065 cm⁻¹, revealing the inherently weak Raman scattering cross section of this mode. Based on the potential-dependence of the ring breathing and alkyl stretching modes, we infer that the pyrrolidinium ring of the BMPy⁺ likely remains close to the surface of the Au electrode, while the butyl chain gradually tilts away from the surface at more positive potentials.

The intramolecular vibration peaks of TFSI⁻ are also analyzed, and their intensity vs electrode potential is plotted in Fig. 4(b). The 279 and 1243 cm⁻¹ peaks can be attributed to the rocking and symmetric stretching of the CF₃ group, respectively.^43–47 The peak at 1137 cm⁻¹ is assigned to symmetric SO₂ stretching.^43,45–47 The peak at 742 cm⁻¹, exhibiting the strongest intensity among all the observed vibrational modes, has been identified as a complex vibration due to the expansion and contraction of the entire TFSI⁻ anion.^46–49 We find that all of the intramolecular vibration modes of TFSI⁻ exhibit a similar trend of increase at more positive potentials and that the 1137 cm⁻¹ peak has a slightly more pronounced change than others. These results indicate that, at more positive potentials, TFSI⁻ molecules move closer to the Au surface due to electrostatic attraction, while the molecular orientation likely remains unchanged, resulting in a similar intensity increase of all the observed vibrational modes. The slightly higher increase of the 1137 cm⁻¹ peak indicates that the SO₂ groups are likely closest to the Au surface, thus exhibiting the strongest Raman enhancement. This is in agreement with previous reports that the oxygen atom in the SO₂ group tends to have strong interaction with the Au surface.^45,50

In contrast to the strong potential-dependence of the peak intensity of many of the intramolecular vibrational modes, the peak position of all the observed intramolecular modes is found to have negligible changes at different potentials, as evident from Figs. 3 and S5. This reveals that the intramolecular structure of both the BMPy⁺ and the TFSI⁻ likely remains the same at different electrode polarizations.

Besides the intramolecular vibrational modes with known origins, we also observe a new interfacial peak at ∼180 cm⁻¹ (Figs. 3 and S5). To the best of our knowledge, this peak has not been explained in any previous experimental or theoretical studies on BMPy-TFSI. The position and intensity of this low-wavenumber SHINERS peak are summarized in Fig. 5 (extracted based on the results in Fig. 3). We observe a strong increase in intensity at more positive potentials. In contrast to the intramolecular peaks that show negligible position changes, we find a strong blue shift of this low-frequency peak at more positive potentials, with the position changing from ∼178 cm⁻¹ at −0.5 V to ∼188 cm⁻¹ at 1.1 V. Such a large change reveals that the vibrational energy of this mode is highly dependent on the electrode potential. Considering that the intramolecular structure of cations and anions both remain unchanged (as evident from the constant vibrational frequencies of all the higher-frequency Raman modes), it is likely that this ∼180 cm⁻¹ peak is induced by intermolecular interactions. According to molecular dynamics simulations, the oxygen site of the TFSI⁻ can interact strongly with the N-methyl group of the BMPy⁺, forming ion pairs.^50–53 Therefore, we suspect that the intermolecular vibration of the ion pairs is likely responsible for the ∼180 cm⁻¹ peak. At more positive potentials, the TFSI⁻ moves closer to the electrode surface, while the N-methyl group of the BMPy⁺, directly connected to the pyrrolidinium ring, remains in close proximity to the Au surface. As a result, more ion pairs accumulate at the electrode surface, which may further lead to stronger intermolecular interaction (due to the more compact molecular packing). Therefore, higher positive potential eventually results in both an intensity increase of the ∼180 cm⁻¹ peak (due to ion pair accumulation) and its blue shift (caused by the stronger BMPy⁺–TFSI⁻ interaction).

As a control experiment, we carry out interfacial Raman measurements of BMPy-TFSI on the Au electrode coated with bare Au nanoparticles. The results, shown in Fig. S7, reveal similar low-frequency peaks and their potential-dependent position and intensity changes. Therefore, we rule out the possible contribution of the SiO₂ shell (in the Au/SiO₂ nanoparticle) to the low-frequency peak.

To further verify the intermolecular nature of the low-frequency peak, we perform additional control measurements of the bare Au electrode as well as two other TFSI-containing electrolytes, 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide (EMIM-TFSI) and water-in-salt electrolyte (21 m LiTFSI in water). The low-frequency spectra of the bare Au electrode, BMPy-TFSI/Au interface at negative potentials (same as the corresponding parts in Fig. 3), as well as three different bulk electrolytes are shown in Fig. 6. Note that the bulk BMPy-TFSI spectrum is the same as the corresponding spectrum in Fig. 3, except that the low-frequency wavenumber range is expanded and the intensity is magnified. No peak is observed on the bare Au, ruling out possible electrode contamination effects. With a long exposure time and extensive data accumulation, we are able to observe weak peaks at ∼170 and ∼164 cm⁻¹ for bulk BMPy-TFSI and EMIM-TFSI, respectively. The water-in-salt electrolyte, despite containing large concentrations of TFSI⁻, does not show any low-frequency peak even with two times more data accumulation (compared to the ionic liquid electrolytes). The 279 cm⁻¹ CF₃ rocking peak, in contrast, appears in all the TFSI-containing electrolytes at the same peak position. These results are consistent with our proposed methyl-TFSI intermolecular interaction mechanism, since methyl groups exist in both BMPy⁺ and EMIM⁺ but are absent in the water-in-salt electrolyte. The red shift of the bulk vs interfacial BMPy-TFSI reveals that the intermolecular interaction is stronger at the interface compared to the bulk. Similarly, the lower peak position of bulk EMIM-TFSI compared to bulk BMPy-TFSI indicates weaker intermolecular interaction in the former, which might be due to more charge delocalization in the imidazolium structure.

Based on the combined results of the BMPy⁺ and TFSI⁻ intramolecular peaks, as well as the low-frequency peak at ∼180 cm⁻¹, which is tentatively assigned to BMPy-TFSI intermolecular interaction, we propose a potential-dependent evolution of the EDL structure, as shown in Fig. 7. The pyrrolidinium ring of the BMPy⁺ cation is likely specifically adsorbed on the Au surface across the whole range of electrode potential we apply. At more positive potentials, the electrostatic repulsion results in the tilting of the butyl chain, which transitions from parallel to oblique orientation. Note that the previous scanning probe, neutron reflectometry, and molecular dynamics studies also predicted the specific adsorption of BMPy⁺ on Au surfaces,^21,54,55 which was explained as a result of the strong van der Waals interaction and image force effects, although the exact adsorption group was not reported. In contrast to the strong adsorption of BMPy⁺, the position of TFSI⁻ is highly sensitive to electrostatic interactions with the electrode. These anions each strongly interacts with the N-methyl group of a BMPy⁺ cation through one SO₂ group, move closer to the electrode at more positive potentials, and eventually strongly accumulate at the electrode surface with the other SO₂ group pointing down and interacting with Au.

Using in situ surface-sensitive Raman spectroscopy, we have determined the EDL structure of a pyrrolidinium-based ionic liquid, BMPy-TFSI, at different electrode potentials. By analyzing the potential dependence of the individual intra- and inter-molecular vibrational modes, we identify the molecular position and orientation on the Au electrode surface. We find specific adsorption of the pyrrolidinium ring of BMPy⁺ on Au, reorientation of the BMPy⁺ alkyl chains, position switching of the whole TFSI⁻ molecules at different electrode potentials, as well as the strong intermolecular pairing of BMPy-TFSI. These molecular insights will be important for understanding and designing the molecular structure of EDLs.

See the supplementary material for transmission electron microscopy images of the Au/SiO₂ nanoparticles, CV of Au and Au/SiO₂ nanoparticles deposited on glassy carbon electrodes, EC-SHINERS spectra of BMPy-TFSI/Au at high positive potentials, CV of BMPy-TFSI on the Au electrode in air vs Ar atmosphere, additional EC-SHINERS spectra and peak analysis of the BMPy-TFSI/Au interface from multiple measurements, and low-frequency Raman peaks of interfacial BMPy-TFSI on Au electrodes coated with bare Au nanoparticles.

Y.Z., J.K., and K.S.P. acknowledge support from the Air Force Office of Scientific Research (Award No. FA9550-22-1-0014). Y.Z., F.Z., and S.Z. acknowledge support from the Beckman Young Investigator Award provided by the Arnold and Mabel Beckman Foundation.

The authors have no conflicts to disclose.

The data that support the findings of this study are available from the corresponding author upon reasonable request.