Reduced graphene oxide/ionic liquid composites with tunable interlayer spacing for improved charge/discharge kinetics in supercapacitors

The large specific surface area and high conductivity of reduced graphene oxide (RGO) make it a promising material for supercapacitors. However, aggregation of graphene sheets into graphitic domains upon drying hampers supercapacitor performance by drastically impeding ion transport inside electrodes. Here, we present a facile approach to optimize charge storage performance in RGO-based supercapacitors by systematically tuning their micropore structure. To this end, we combine RGOs with room temperature ionic liquids during electrode processing to impede stacking of sheets into graphitic structures with small interlayer distance. In this process, RGO sheets function as the active electrode material while ionic liquid serves both as a charge carrier and a spacer to control interlayer spacing inside electrodes and form ion transport channels. We show that composite RGO/ionic liquid electrodes with larger interlayer spacing and more ordered structure exhibit improved capacitance and charging kinetics.

However, restacking and aggregation of RGO sheets during electrode assembly have impeded fabricating RGObased electrodes with large ion-accessible SSA [21,39]. As opposed to the rigid pore structure of most high surface area carbon materials, RGOs do not contain an intrinsic porous network. Consequently, the ion-accessible surface of RGO electrodes depends on how RGOs aggregate upon consolidation into electrodes. Typically, RGO-based electrode fabrication involves dispersing RGOs in a solvent, and subsequently evaporating the solvent. While RGO dispersions exhibit a high SSA, sheets collapse and aggregate upon drying due to capillary forces, and this drastically reduces their ion-accessible SSA [39,41]. A number of studies have aimed to optimize the ion-accessible SSA of RGO sheets while processing them into electrodes. [42][43][44][45][46][47][48][49][50][51]. These examples typically involve using inactive spacers to separate RGO sheets which increase electrode dead-weight [42,43]. Metal oxides and conductive polymers are used as additives to utilize their pseudocapacitance [49][50][51][52][53]. Metals and other carbon-based materials are also used as spacers [44,45,[48][49][50][51]. Using ions both as an electrolyte and a spacer between RGO sheets is another promising approach. Pope et al used a room temperature ionic liquid (RTIL) to successfully separate graphene oxide (GO) sheets and subsequently thermally reduce GO, however the electrical conductivity of the electrodes was limited due to the degree of thermal reduction of RGO controlled by the decomposition temperature of the RTIL [46]. Li et al used a solventexchange process to prevent the restacking of chemically reduced GO sheets and obtained high capacitance but the films had an amorphous structure [47].
In this work, we use RGO as the active electrode material and tune the ion accessibility of RGO sheets by introducing a RTIL as a spacer to optimize charge storage performance. Tuning the interlayer distance of RGO films with a RTIL eliminates the complications that come with carbon materials with fixed and hard pore network which require changing the chemistry of the material to tune pore structure. Our approach differs from previous studies that use ionic liquids as a spacer for graphene-based electrodes. In particular, we reduce GO prior to dispersion and consolidation into a RGO/RTIL composite electrode which improves electronic transport while also delivering composite films with more ordered structure. We show that the capacitance of electrodes comprised of the same active electrode material with varying intersheet distance distributions exhibit significant differences. We demonstrate charge storage performance can be optimized by modifying the structure of the ion transport network. We show RGO/RTIL composites with more ordered structure and larger interlayer distance have enhanced capacitance at high scan rates.

Experimental methods
2.1. Preparation of graphene oxide and reduced graphene oxide GO and RGO synthesis procedures are depicted in figure S1. GO was synthesized using the modified Hummers method [55]. 120 ml of sulfuric acid (H 2 SO 4 ) and 13 ml of phosphoric acid (H 3 PO 4 ) were mixed. 1 g of natural graphite flakes (Sigma-Aldrich) and 6 g of potassium permanganate (KMnO 4 ) were added to the mixture and mixed using a magnetic stirrer for 6 h at 50°C. The mixtures was stirred for an additional 12 h at room temperature. 2 ml of hydrogen peroxide (H 2 O 2 ) is added to the slurry to eliminate excess KMnO 4 and stirred while the mixture turned to yellow. 40 ml of hydrochloric acid (HCl) and 130 ml of deionized (DI) water were added to the mixture. The suspension was centrifuged for 15 min at 1800 rpm (Nuve). After decanting the supernatant, and the remaining material was washed with HCl and DI water for three times. The washing procedure was then repeated with DI water for three times and the filtrate was dried in a 70°C vacuum oven for 24 h, subsequently dissolved in DI water at 0.5 mg ml -1 and frozen for 24 h. The sample was then dried using a freeze dryer (Labconco Freezone 4.5). The obtained GO powder was loaded into an alumina tube and placed in a furnace (Protherm) set at 700°C for 2 min to reduce and expand the powder. The tube was then removed from the furnace and cooled down to room temperature at ambient conditions. The resulting black reduced graphene oxide (RGO1) was stored in powder form. RGO1 powder was further annealed at 1200°C using a tubular furnace (Protherm) in Ar atmosphere for 1 h. The obtained powder (RGO2) was collected and stored in a sealed container filled with Ar.

Processing of reduced graphene oxide and reduced graphene oxide/Room temperature ionic liquid composites
Three different vacuum filtration processes (one-step, twostep, three-step) were used to prepare RGO/RTIL composite films. The one-step process was comprised of mixing 60 mg RGO with 300 ml N-methyl-2-pyrrolidone (NMP) using a magnetic stirrer for 1 h, followed by 30 min tip-sonication at 20% amplitude in an ice bath and 3 h mechanical mixing using a magnetic stirrer. The process was completed by vacuum filtration through a nanoporous PVDF membrane until the solvent completely evaporated.
The two-step process involved mixing RGOs with NMP as described above, followed by vacuum filtration. Towards the end of the vacuum filtration process, when there was 10 mm free suspension on the filter cake, the vacuum was stopped and 13-130 μl RTIL (1-Ethyl-3-methylimidazolium tetrafluoro-borate (EMIMBF 4 ) or 1-Ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide (EMIMTFSI)) was added to the suspension. The suspension was left to filter through the membrane, leaving some of the RTIL between RGO sheets and forming a RGO/RTIL composite film.
The three-step process involved mixing 60 mg RGO with 100 ml NMP. 30.9 μl EMIMBF 4 or 60.8 μl EMIMTFSI was added and the solution was mixed using a magnetic stirrer for 1 h, followed by 30 min tip-sonication at 20% amplitude in an ice bath. After adding 200 ml NMP, the as-prepared solution was mixed using a magnetic stirrer for 3 h. This solution was then filtered through PVDF membrane, followed by RTIL addition to the suspension after stopping the vacuum as described above in the two-step process. The same procedure was repeated by varying the RTIL amount used in the final step.

Electrochemical characterization
Free standing 80-100 μm thick RGO or RGO/RTIL films were pressed onto Pt current collectors for electrochemical measurements. Acetonitrile (ACN) (99.8%, Sigma-Aldrich) was stored in a sealed container. EMIMBF 4 (Iolitec, 99%) and EMIMTFSI (Iolitec, 99%) were purchased at their highest available purity and used as received. 3.5 M electrolyte was prepared by mixing ACN and RTIL for 1 h using a magnetic stirrer. A 2 cm 2 platinum plate counter electrode was polished using 0.05 μm aluminum monohydroxide powder and 1 μm diamond powder, respectively, washed with DI water and sonicated in ethanol for 30 min in an ultrasonic bath. A pseudoreference electrode comprised of a platinum wire isolated in a fritted glass tube (BASi) and filled with a mixture at the same composition as the electrolyte was used. After the measurements, reference electrode was calibrated by adding 5 mM ferrocene to the electrolyte. Electrochemistry measurements were conducted using a digital potentiostat (Bio-Logic, SP-200) in a three-electrode electrochemical cell (BASi). Prior to measurements, cyclic voltammetry was performed at 5 mV s -1 for six times within the selected voltage window. Cyclic voltammetry measurements started from the open circuit potential with a 25 min wait period between scans. Experiments were performed in duplicate and the calculated capacitance values of C were within 12% of one another for the largest deviation. Impedance measurements were taken from 200 kHz to 50 mHz with 30 points in logarithmic scale at 0.0 V.

Structural characterization of RGO and RTIL/RGO films
The D-peak of the Raman spectrum of graphene-based materials is activated by defects present around aromatic sixmembered rings, i.e. vacancies, or grain boundaries/edges, while the G-peak corresponds to the oscillations of sp 2 hybridized carbon atom pairs in chains or rings [56,57]. To observe their evolution, the D and G-peaks of GO, RGO1, and RGO2 were fit using the Lorentzian (D-peak) and Breit-Wigner-Fano (G-peak) functions (figure S2) [59]. For RGO1, the ratio of the intensities of the D (1272-1406 cm −1 ) and G-bands (1500-1600 cm −1 ), ID/IG, is 1.06. This is slightly larger than the ID/IG of GO (0.95). The C/O of GO (∼2) as determined from x-ray photoelectron spectroscopy (figure S3) indicates a portion of the oxygen-containing functional groups of this material (e.g. CO, CO 2 ) is removed when it is reduced at 700°C for 2 min to form RGO1 (C/O ∼8), resulting in an increase in lattice vacancies and ID/IG. Meanwhile, ID/IG decreases from 1.06 to 0.59 when RGO1 is annealed at 1200°C for 1 h. The decrease in ID/IG upon annealing demonstrates defect density is reduced with respect to RGO1. This is consistent with the C/O of RGO2 that increases to ∼40 with the removal of oxygen-containing groups (figure S3). A distribution of sp 2 clusters with different sizes as well as bond disorder due to the presence of non-6-member rings are factors thought to contribute to the D-peak width [56,58]. The decrease in ID/IG upon annealing (RGO2) demonstrates defect density is reduced with respect to RGO1. In fact, this is consistent with the significant decrease in functional groups (∼11% to 2% oxygen) as the material is annealed for 1 h at 1200°C largely removing defects associated with oxygencontaining functional groups (figure S3).
The decrease in ID/IG also indicates during the longer heat treatment step, lattice annealing may have taken place, which is known to occur at high temperatures for graphitic materials [59]. The breadth of the D-peak (as indicated by the full width at half maximum, FWHM) decreasing from 69.7 to 17.9 cm −1 post annealing, as well as the deconvoluted XPS spectra (figure S3) are consistent with this finding. Powder XRD patterns of GO, RGO1 and RGO2 are presented in figure S4. The characteristic peak of GO is observed at a scattering angle of 2θ ∼10.8°indicating a d 0002 spacing of 8.2 Å has formed. The precise location of this peak depends on the oxidation degree of graphite flakes and the interlayer spacing is similar to what has been reported for GO prepared using the same oxidation method [60]. Upon reduction of GO to RGO1 (RGO2), the peak shifts to 2θ ∼24.3°(26.1°) corresponding to interlayer spacings of 3.7 Å(3.4 Å).
In figure 1(a), we show the thin-film XRD profiles of the dry RGO1 and RGO2 films prepared via the one-step process. RGO1 (SSA ∼ 324 m 2 g -1 ) and RGO2 (SSA ∼ 225 m 2 g -1 ) have diffraction peaks centered at 23.8°and 25.2°corresponding to d-spacings of 0.37 and 0.35 nm, respectively. RGO2ʼs peak located at a larger angle (smaller d-spacing) is attributable to the fact that RGO2 contains a smaller number of functional groups and defects due to the annealing treatment. As RGO2 is prone to more aggregation and stacking, the average spacing between RGO2 sheets is smaller compared to the spacing between RGO1 sheets, and the peak location is closer to what has been reported for graphite (∼27°) .   Figures 1(b), (c) show the XRD spectra of RGO/RTIL films prepared using the two-step process, targeting 60 wt% RTIL in RGO/RTIL composites. The peak of the scattering angle shifts to smaller values with RTIL introduction. TGA analysis of the composites prepared via the two-step process confirms the presence of RTIL in the films (figure S6). For RGO-EMIMBF 4 films, the weight loss between approximately 300°C and 450°C is mainly due to EMIMBF 4 decomposition ( figure S6). Only a small fraction of weight loss (<4 wt% for RGO1 and <1 wt% for RGO2) within this temperature range is attributable to the further reduction of RGO as indicated by the TGA profiles of RGO1 and RGO2 ( figure S6). The compositions of films are all approximated as close to the target 60 wt% RTIL with ∼57 (60) wt% EMIMBF 4 for RGO1 (RGO2) films.
The XRD spectra demonstrate that RGO2 and RGO2/ RTIL composites are more prone to stacking into graphitic domains as demonstrated by their smaller average d-spacing values compared to their RGO1 counterparts. When the same solvent exchange process is applied to RGO1 and RGO2, the shift of the scattering angle is smaller for RGO2. We attribute the smaller d-spacing of RGO2 films to their more hydrophobic nature due the presence of a smaller number of defects and functional groups. Comparison of EMIMBF 4 and EMIMTFSI shows that the increase in d-spacing is more pronounced when the more hydrophilic EMIMBF 4 is used. The surface wetting properties of RTILs comprised of BF 4 − or TFSI − ions can elucidate this behavior. Anions play an important role in determining the surface interactions between RGO and RTILs. The contact angle of a RTIL containing TFSI − anion on RGO is measured as 77°indicating strong surface dewetting, while the contact angle is only 33°when the anion is swapped with BF -4 showing that the more hydrophilic EMIMBF 4 has stronger interactions with RGO [61]. This is consistent with previous work that demonstrate EMIMBF 4 can spontaneously adsorb on GO surface [46], while EMIMTFSI phase separates [62,57] and can only adsorb on RGO surface when mixed with nonionic surfactants [58]. Nevertheless, although ionic liquid incorporation separates RGO sheets more effectively, all peaks in figure 1 are broad indicating that RTIL-filled spaces are disordered, sheet-to-sheet spacing is uneven and RGO sheets do not perfectly align with RTIL between sheets. If stacking were of the perfectly flat face-to-face type, detectable, sharp XRD peaks would be observable.
To obtain improved sheet-to-sheet alignment and more even RTIL distribution, we next use the three-step electrode consolidation process. This process involves probe ultrasonication of the RGO/organic solvent/RTIL mixture followed by filtrating the mixture through a nanoporous membrane, and when there is still free dispersion on the filter cake, adding a RTIL and vacuum filtrating until a flexible RGO/RTIL composite film is formed. Here, the motivation of the ultrasonication step is to assist the wetting of RGO sheets by RTIL molecules prior to vacuum filtration and, consequently, facilitate more uniform incorporation of ions within interlamellar spaces. We note that when the more hydrophobic EMIMTFSI is used in the probe ultrasonication step, no improvement upon the two-step process is observed. Thus, the three-step process focuses only on the more hydrophilic ionic liquid EMIMBF 4 that has stronger interactions with the oxygen-containing functional groups of RGO sheets.
The XRD spectra for the RGO1/EMIMBF 4 composite film prepared via the three-step process (figure 2(a)) shows sharper peaks that also correspond to larger d-spacing, compared to plain RGO1 film and RGO1/EMIMBF 4 film prepared via the two-step process. Particularly, the signal from RGO1/EMIMBF 4 is located at 9°corresponding to a d-spacing of 0.98 nm, compared to 16.7°(0.53 nm) for the two-step process and 23.8°(0.37 nm) for plain RGO1, respectively. This suggests that RGO1 sheets have larger interlayer spacing when prepared via the three-step process as indicated by the shift in 2θ. The sharper and narrower peaks of RGO/RTIL composites prepared via the three-step process suggest RGO1 sheets exhibit more uniform and long-range ordering when the three-step process is used for fabrication where EMIMBF 4 is added to the RGO suspension prior to ultrasonication. Meanwhile, the d-spacings for RGO2/EMIMBF 4 composites do not exhibit an obvious trend with 25.2°(0.35 nm) for plain RGO2, 18.3°(0.48 nm) for two-step, and 20.1°(0.44 nm) for three-step processes. We ascribe this to the weakened interactions between RGO and EMIMBF 4 due to a smaller number of functional groups on RGO2 sheets.

Electrochemical performance of RGO and RTIL/RGO films
where I is the current, V is the applied potential, V a and V b are the lower and upper limits of the potential range, v is the scan rate, and m is the mass of the RGO and RGO/RTIL composite film (2 ± 0.3 mg). Gravimetric capacitance (C G ) values obtained for RGO films and RGO/EMIMBF 4 films prepared via two and three-step processes are given in figures 3(b), (d), (f). At 5 mV s -1 , C G values change very little with electrode processing method for the same active electrode material, and are within 4% of each other. The relatively constant value of C G at small scan rates for films containing the same active electrode material suggests that the total ion-accessible SSA of films do not change considerably with the processing method. Hence, when electrodes are charged slow enough devoid from transport limitations, similar C G values are obtained. At higher scan rates C G depends significantly on processing method. In particular, plain RGO1 film retains only 25% of its C G at 5 mV s -1 when scan rate is increased to 500 mV s -1 . Meanwhile, capacitance retention is 37% for the  RGO1/EMIMBF 4 film prepared via the two-step process and 70% for the three-step process. We attribute the enhancement in capacitance retention to the improved structure of ion diffusion paths inside RGO/RTIL composites which increases ionic conductivity and facilitates faster charging.
To further elucidate the underlying mechanism behind the charging/discharging kinetics of composite electrodes, we report Nyquist plots obtained from electrochemical impedance spectroscopy (EIS) ( figure 4). The equivalent circuit model used to fit the impedance data is presented on figure  S7. The impedance data is consistent with double layer charging in porous electrodes exhibiting a mid-frequency region and a low frequency region with a steep but not precisely 90°slope. As shown in figure 4(a), the leftmost intersection point on the real axis of the RGO1 film is larger than that of the RGO2 film, indicating the equivalent series resistance is smaller for RGO2. As the electrolyte is identical for both cases, the shift of the intersection point is attributable to the higher degree of reduction of RGO2 that leads to a higher electronic conductivity. The impedance represented by the semicircle can be assigned to the charge transfer resistance in the electrode material and/or the contact resistance between electrode and current collector. To understand the main contribution to the semicircle diameter, we record impedance spectra at 0, 0.5, and 1 V for a sample system (RGO1/EMIMBF 4 composite prepared via the three-step process). We observe that the semicircle size does not change noticeably with applied potential indicating the semicircle is mainly governed by the interface between current collector and electrode. The semicircle diameter in the impedance spectra is smaller for plain RGO1 compared to plain RGO2, indicating the contact resistance is smaller for RGO1. When EMIMBF 4 is introduced into the electrodes via the two-step process, the semicircle diameter ( figure 4(b)) decreases from 0.69 to 0.29 Ω for RGO1/EMIMBF 4 and 1.05-0.31 Ω for RGO2/EMIMBF 4 . Meanwhile, when RGO/EMIMBF 4 composites are prepared via the three-step process ( figure 4(c)), the semicircle diameters in the high frequency region are significantly reduced to 0.15 Ω for RGO1/EMIMBF 4 and to 0.28 Ω for RGO2/EMIMBF 4 . The width of the mid-frequency region, also called the distributed resistance region, is mainly associated with resistance due to ion diffusion through the porous structure of the electrodes. The width of this region was calculated from the low-frequency extrapolated value and the location of the semicircle. The widths are larger for plain RGO films, and they decrease in RGO/EMIMBF 4 films. In particular, the smallest width is obtained for RGO1/EMIMBF 4 produced via the three-step process ( figure 4(c)). This suggests that improved accessibility of RGO sheets to EMIMBF 4 improves ion transport.

Optimization of the electrochemical performance
To determine the optimum amount of RTIL to be incorporated into RGO electrodes in the three-step process, we prepare several RGO1/EMIMBF 4 composites by tuning the amount of EMIMBF 4 added to the free suspension after stopping the vacuum. XRD profiles of the obtained films reflect an increase in d-spacing with increasing RTIL content in the film. The evolution of EMIMBF 4 content in RGO1/EMIMBF 4 composites is determined via TGA analysis (figure S5). The weight loss between 300°C and 450°C is mainly due to EMIMBF 4 decomposition as shown by the TGA profile of EMIMBF 4 in figure S5. We note that a small fraction (less than 4 wt%) of weight loss within this temperature range is also attributable to further reduction of RGO1 as indicated by the TGA profile of RGO1. Consequently, the compositions of films are determined as RGO1-79 wt% EMIMBF 4 , RGO1-57 wt% EMIMBF 4 , RGO1-48 wt% EMIMBF 4 , and RGO1-21 wt% EMIMBF 4 , respectively. While the highest d-spacing is obtained at RGO1-79 wt% EMIMBF 4 , the peak is most pronounced at RGO1-48 wt% EMIMBF 4 ( figure 5(a)).
All composites exhibit similar capacitance performance to the plain RGO1 electrode at 5 mV s -1 ( figure 5(b)). As scan rate is increased, the difference in electrode capacitance significantly changes with varying RTIL/electrode wt%/wt%. The optimum capacitance retention as a function of scan rate is obtained for the RGO1-48 wt% EMIMBF 4 electrode. In particular, when scan rate is increased from 5 to 500 mV s -1 , capacitance of this electrode decreases only by 15%. The superior scan rate performance of the RGO1-48 wt% EMIMBF 4 composite is likely due to its improved ionic and electronic conductivity. In fact, the impedance spectra in figure 5(c) shows that RGO1-48 wt% EMIMBF 4 electrode exhibits the smallest semicircle diameter, and this demonstrates the contact resistance between the electrode and current collectors is smaller at intermediate RTIL/electrode wt%/wt%. Moreover, the steepest slope of the RGO1-48 wt% EMIMBF 4 electrode in the low frequency region indicates the interconnected and evenly distributed diffusion paths inside the electrode present the smallest resistance to ion migration with improved ion transport kinetics. The decrease of the low frequency slope at lower EMIMBF 4 content is associated with ionic transport limitations due to harder accessibility of RGO sheets by ions. Meanwhile, the slope decreases at EMIMBF 4 concentration higher than 48 wt% as well. This may be attributable to an increase in electronic resistance due to bulging of electrodes with excessive EMIMBF 4 content that decreases the contact of RGO sheets to each other.

Conclusion
In this work, we used RGO as the active electrode material and optimized ion accessibility by tuning the interlayer distance using a RTIL as a spacer. We reduced GO prior to mixing with an ionic liquid, hence were able to obtain a high degree of reduction. Capacitance of composite films were similar at low scan rates for all electrode fabrication methods, but films prepared via the three-step process exhibited significantly superior scan rate performance. We showed RGO/ RTIL composites with more ordered structure exhibit significantly enhanced scan rate dependence compared to composites with more disordered structure. These results demonstrate that this strategy can offer an easy and effective way to fabricate supercapacitors based on two-dimensional carbon materials with improved performance.

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https://doi.org/10. 6084/m9.figshare.22120085.v1

Notes
The authors declare no competing financial interest.