Influence of Resorcinol to Sodium Carbonate Ratio on Carbon Xerogel Properties for Aluminium Ion Battery

Carbon xerogels were synthesized using a soft-template route with resorcinol as the carbon source and sodium carbonate as the catalyst. The influence of the resorcinol to catalyst ratio in the range of 500–20,000 on pore structure, graphitic domains, and electronic conductivity of as-prepared carbon xerogels, as well as their performance in an aluminium ion battery (AIB), was investigated. After carbonization steps of the polymers up to 800 °C, all carbon samples exhibited similar specific volumes of micropores (0.7–0.8 cm³ g−1), while samples obtained from mixtures with R/C ratios lower than 2000 led to carbon xerogels with significantly higher mesopore diameters up to 6 nm. The best results, in terms of specific surface (1000 m² g−1), average pore size (6 nm) and reversible capacity in AIB cell (28 mAh g−1 @ 0.1 A g−1), were obtained with a carbon xerogel sample synthetized at a resorcinol to catalyst ratio of R/C = 500 (CXG500). Though cyclic voltammograms of carbon xerogel samples did not exhibit any sharp peaks in the applied potential window, the presence of both oxidation and a quite wide reduction peak in CXG500–2000 cyclic voltammograms indicated pseudocapacitance behaviour induced by diffusion-controlled intercalation/de-intercalation of AlCl4− ions into/from the carbon xerogel matrix. This was confirmed by shifting of the (002) peak towards lower 2θ angle values in the XRD pattern of the CXG500 electrode after the charging step in AIB, whereas the contribution of pseudocapacitance, calculated from half-cell measurements, was limited to only 6% of overall capacitance.


Introduction
The demand for electrically rechargeable energy storage systems has increased significantly over the last few years. Apart from classical systems, such as lead-acid and nickel-metal hydride (NiMH) systems, a huge number of established and merged application devices rely on lithium ion batteries (LIB) due to their impressive cell performance in terms of cell voltage (3-3.6 V), energy density (220 Wh kg −1 /450 Wh L −1 ) and longterm cycling ability [1]. Because of the enormous demand for LIBs and the consequential rise in lithium price, as well as for critical materials such as cobalt, shortage in raw materials within the next decades appears to be inevitable [2,3]. Therefore, there is an urgent need to develop novel battery systems that rely on highly abundant, non-critical metals, such as zinc, sodium, magnesium or aluminium. One of the most promising candidates is the Al-ion battery (AIB) that typically consists of highly graphitic carbon (natural graphite [4][5][6][7][8][9][10][11][12], graphene [13][14][15][16][17], graphene oxide [18] or carbon foams [11]) as the positive electrode for Al-ion intercalation, an AlCl 3 containing ionic liquid, such as 1-ethyl-3-methylimidazoliumchloride (EMimCl), as the aprotic electrolyte, and an aluminium metal foil as the negative electrode for electrochemical Al deposition/stripping. The main benefits of AIB are related to the use of highly abundant and cheap graphite In this study, the influence of catalyst concentration during the polycondensation of resorcinol with formaldehyde on carbon xerogel morphology was systematically studied for different resorcinol to catalyst (R/C) ratios.

Synthesis of Xerogels
Fourteen different xerogel samples were prepared using resorcinol (R) to sodium carbonate as catalyst (C) ratios in the range of R/C = 50-20,000 and keeping resorcinol to formaldehyde at a molar ratio of 0.5. First, 1.061 g of sodium-carbonate (Na2CO3) (>99.99%, Sigma Aldrich, Darmstadt, Germany) was poured into ultrapure water (18.2 MΩ, ELGA Prelab ® , Celle, Germany) to obtain a standard solution of 0.1 M Na2CO3. Then, 1.0 g of resorcinol (>99%, Sigma Aldrich, Darmstadt, Germany) was dissolved in 1.95 mL of Na2CO3 catalyst solution for a given R/C molar ratio under stirring conditions. After complete dissolution, 1.355 mL of 30% formaldehyde solution (35 Vol%, Sigma Aldrich, Darmstadt, Germany) was added resulting in a molar R/F ratio of 0.5. The polymerization reaction was carried out in a water bath at 65 °C for 72 h. To remove residual water, a drying step was conducted at 125 °C and 0.1 mbar in a self-build Schlenk-line apparatus. The heating step led to a change in colour of the as-prepared polymers (see Figure 1).

Figure 1.
As-prepared resorcinol-formaldehyde resins after drying step from high catalyst concentration (R/C = 50) on the left side to low concentration (R/C = 20,000) on the right side.
Finally, carbonization of the samples was performed in a tubular oven (ROK 3/30, W. C. Heraeus, Hanau, Germany) under N2 atmosphere by applying two successive heating steps at 400 and 800 °C for 2h each and using a temperature ramp of 5 K min −1 . The yielded carbons showed differences in density where the highest and lowest values were observed at CXG50 and CXG20,000, respectively (see values in Table A1). At CXG50 especially, very hard and dense particles were formed. One possible explanation might be the higher concentration of sodium carbonate in the range of R/C = 50-1500 that favoured fast formation of large aggregates [39]. Immediately after adding formaldehyde to the R/C50 solution, the colour started to change from clear white to red/orange. After 2 h, the polymerization step was completed, resulting in a dark red colour polymer. For R/C ratios in the range of 500-1500, the whole polymerization was accomplished after 18-24 h. For R/C >1500 samples, the polymerization step took 35-45 h due to a slow reaction rate. It seems that a slow polymerization reaction favoured the formation of small aggregates which resulted after carbonization in less dense carbon particles. At very low Na2CO3 concentrations, such as those used during R/C10,000 and R/C20,000 synthesis mixtures, the solution was colourless even after 60 h reaction time.
Finally, carbonization of the samples was performed in a tubular oven (ROK 3/30, W. C. Heraeus, Hanau, Germany) under N 2 atmosphere by applying two successive heating steps at 400 and 800 • C for 2h each and using a temperature ramp of 5 K min −1 . The yielded carbons showed differences in density where the highest and lowest values were observed at CXG 50 and CXG 20,000 , respectively (see values in Table A1). At CXG 50 especially, very hard and dense particles were formed. One possible explanation might be the higher concentration of sodium carbonate in the range of R/C = 50-1500 that favoured fast formation of large aggregates [39]. Immediately after adding formaldehyde to the R/C 50 solution, the colour started to change from clear white to red/orange. After 2 h, the polymerization step was completed, resulting in a dark red colour polymer. For R/C ratios in the range of 500-1500, the whole polymerization was accomplished after 18-24 h. For R/C > 1500 samples, the polymerization step took 35-45 h due to a slow reaction rate. It seems that a slow polymerization reaction favoured the formation of small aggregates which resulted after carbonization in less dense carbon particles. At very low Na 2 CO 3 concentrations, such as those used during R/C 10,000 and R/C 20,000 synthesis mixtures, the solution was colourless even after 60 h reaction time.

Instrumental Analysis
XRD and Raman analysis of carbon powder samples was carried out in a diffractometer (D8 Advance, BRUKER, Karlsruhe, Germany) with CuKα radiation of 1.541874 Å and Raman microscope (inVia, Renishaw, Pliezhausen, Germany) at 532 nm, respectively. The intercalation of chloroaluminate into graphite or CXG matrix was studied by ex situ XRD analysis. For this purpose, a charged graphite or CXG electrode was placed in an airtight BRUKER™ specimen holder. The air tightness required to be assured because of the strong hydrolysis properties of AlCl 3 − -containing ionic liquid electrolyte. SEM images were acquired with a field emission microscope (SU5000, Hitachi, Krefeld, Germany). Conductivity measurements were performed in a pressurized device designed by ZBT (Zentrum für Brennstoffzellentechnik, Duisburg, Germany). Physisorption measurements of carbon powder materials were conducted with argon at 87 K (Autosorb iQ Station, Quantachrome / 3P Instruments, Odelzhausen, Germany). The whole mass-normalized surface area A sum and pore volume V sum were calculated by applying DFT-fitting of cylindrical and spherical pores to measured adsorption isotherms (QSDFT-fit).

Physisorption of CXGs
Plots of samples produced within 350 (a) < R/C < 2000 (b) showed a typical hysteresis for type IV isotherms indicating the presence of mesopores (Figure 2a). For R/C > 2500, named (f-n), the isotherm of carbons shifted to type 1a which indicated nearly 100% microporous textures. This was confirmed by an average pore size distribution of 6.0 nm for CXG 500 and only 1.7 nm for CXG samples with R/C > 2000. The analysis of cumulative pore volume V sum in relation to R/C showed that V sum subsequently decreased with R/C ratio from 1.0 cm 3 g −1 in CXG 500 (a) to 0.5 cm 3 g −1 in CXG 1000 (d) and then amounted to 0.2-0.3 cm 3 g −1 for R/C > 1500 (e-m), see Figure 2b. For the as-prepared carbon xerogels samples, the contribution of macropores was excluded. The micropore volume V mic and surface area A mic were calculated according to the Dubinin-Radushkevich (DR) equation [41][42][43][44]. The mesopore volume V mes and surface area A mes were consequently the textural properties is given in Appendix A Table A2.
The contribution of Vmes and Vmic in selected CXG samples is shown in Figure 2d. We found that the pore volume of CXG500 consisted of 85% Vmes and 15% Vmic. For CXG750, 80% Vmes were determined. With higher R/C values, the amount of Vmes dropped to 21% in CXG2000 while the amount of Vmic increased up to 79%.
Since carbon materials with mesoporous domains are usually preferred in electrochemical systems due to their ability to enhance better diffusion of electrolyte species to active sites and consequently overall capacity, only CXG500-2500 samples were considered further for investigation in AIB laboratory cells.  Figure 2. Characterisation of CXG by isotherms (a), pore size distribution with respect to V sum (b), contribution of microporous and mesoporous pore volume and surface area (c) and repartition of relative mesopore and micropore volume (d).
The mass-normalized surface area A sum was determined to be around 926 m 2 g −1 on average with a maximum of 1004 m 2 g −1 for CXG 500 and a minimum of 894 m 2 g −1 for CXG 1000 . The surface area of all CXG 500-10,000 samples were higher than some values reported in other investigations (300-700 m 2 g −1 ) [34,41,42,45]. Moreover, the overall pore volume V sum of 0.2-1.0 cm 3 g −1 was similar to values referred to in the literature [34,41,42].
As shown in the bar chart represented in Figure 2c, the specific volume of micropores V mic was quite similar for all CXG samples and fluctuated between 0.23-0.26 cm 3 g −1 which was in good accordance with published values [41,42]. As expected from the isotherm shapes, the mesopore volume V mes culminated in the CXG 500 sample at 0.77 cm 3 g −1 and decreased rapidly to 0.08 cm 3 g −1 in CXG 1500 . Surprisingly, the specific surface area A mes of the carbon samples did not fluctuate very much (312-141 m 2 g −1 ). A detailed overview of the textural properties is given in Appendix A Table A2.
The contribution of V mes and V mic in selected CXG samples is shown in Figure 2d. We found that the pore volume of CXG 500 consisted of 85% V mes and 15% V mic . For CXG 750 , 80% V mes were determined. With higher R/C values, the amount of V mes dropped to 21% in CXG 2000 while the amount of V mic increased up to 79%.
Since carbon materials with mesoporous domains are usually preferred in electrochemical systems due to their ability to enhance better diffusion of electrolyte species to active sites and consequently overall capacity, only CXG 500-2500 samples were considered further for investigation in AIB laboratory cells.

X-ray Diffraction
XRD diffractograms of CXG (see Figure 3a) showed three dominant broad peaks at 2θ = 22 ± 1 • , 43 ± 1 • and 80 ± 1 • . Broad peaks are typical for amorphous carbons [46,47] and are a mixture of sp 2 and sp 3 hybridized carbons [48] with a turbostratic arrangement. For comparison, a pure natural graphite (natural high conductive graphite, Prographit) is also shown with a clear visible crystal-plane pattern (green pattern). Since it is difficult to correctly interpret the crystal structure of amorphous carbons, the peaks with (hkl) nomenclature should be carefully assigned. Here we followed Maldonado-Hódar et al. [46] and assigned peaks at 2θ = 22 • and 43 • to (002) and (101) planes, respectively. The intense peak at 80 • in the CXG diffractogram could not be definitively linked to a specific crystal plane.  After synthesis and carbonization, the yielded carbons showed differences in density, with the highest for CXG50 and the lowest for CXG20,000. CXG50 especially formed very hard and dense particles. One possible explanation might be the higher concentration of sodium carbonate in the range of R/C50-1500 that favoured fast formation of big aggregates [39]. Immediately after adding formaldehyde to the R/C50 solution, the colour started to change from clear white to red/orange. After 2 h the polymerization was completed, resulting in a dark red colour. For R/C500-1500 the whole polymerization was finished after 18-24 h. For > R/C1500 the polymerization took 35-45 h due to a slow reaction rate. It appears that a slow polymerization reaction rate favoured the formation of many small aggregates which resulted after carbonization in less dense carbon particles. At very low Na2CO3 concentrations, as in the R/C10,000 and R/C20,000 synthesis mixtures, the solution was colourless even after 60 h reaction time.

Raman
The Raman spectra of the CXG samples (see Figure 4a) showed a clear D-and G-band at 1346 and 1595 cm −1 which were comparable to those for CXGs synthesized by Kakunuri et al. [47]. An intense D-band mainly arises from the polymerization reaction products of resorcinol and formaldehyde coming from numerous sp 3 orbitals of -CH3, -OH and other functional groups which result after carbonization in the formation of carbon domains with multiple defect structures [48,57]. The assignment of D-and G-bands in graphitic structures, such as natural graphite, is straightforward. The G-band around 1575 cm −1 arises from the E2g mode of all present sp²-hybridized carbons located in-plane in the carbon chains and rings and can be visualized as bond-stretching, whereas the D-band in graphitic structures can be explained by a "breathing" of the A1g mode of the carbon chains and rings [58]. Therefore, the observed G-band of the CXGs can be explained by bond-stretching of carbon rings. Natural graphite has a true second-order 2D-Band at 2750 cm −1 coming from overtones and disorder induced by in-plane modes [59][60][61][62]. 20   From the (002) peak, the layer interspacing value (d 002 ) was determined to be 3.335 Å for the natural graphite powder, a value comparable to those reported for different graphite materials (3.35-3.73 Å) in [4,8,49]. The diffractograms of CXG 500-2500 were very similar to each other with small differences of several hundred counts which indicated a similar structure. The calculated d 002 values for all CXG samples varied between 3.6-3.8 Å, values which were comparable to those extracted from plots of pyrolyzed resorcinol-formaldehyde resins (3.4-4.1 Å) [46,[50][51][52][53]. Interestingly, the interspacing values of as-prepared CXG were 8-14% higher than those of natural graphite which might be advantageous for the intercalation of large AlCl 4 − anions. The (002) peak was induced by the stacking height (L C ) of crystallite, while the (101) peak was induced by the stacking width (L A ). From lithium-ion systems, it is well-known that there is a large difference in the electrochemical activity of basal (lateral size, L A ) and edge surfaces (stacking height, L C ). Chung et al. [53], as well as Zaghib et al. [54], showed that the basal plane area contributes around 7× less to electrochemical activity than the edge sites which are important for the intercalation process. L C and L A values of different carbon samples were calculated by applying the Scherrer equation (Equation (1)) for (002) and (101) reflexes with L = stack height/width, K = Scherrer factor, λ = wavelength X-ray source, ∆(2θ) = FWHM (full-width at half maximum) of (hkl) reflex, and θ 0 the Bragg-reflex. FWHM values of CXG s were calculated by Gaussian peak deconvolution in Origin ® 2020. Graphitic carbons can have L A values >> 100 nm [54], while non-graphitic carbons (NCG) have very small particles in the low nanometre-range [48,55].  2000 and CXG 2500 , the calculated L A value was 1.45 nm. The L C value of CXG 2000 had a minimum height of 0.90 nm, whereas CXG 2500 showed a slightly higher stacking height of 0.94 nm. The maximum L A stacking was observed in CXG 500 with 1.57 nm, while the maximum L C stacking for CXG 1500 was measured as 1.06 nm. The average crystallite size in width and height was determined to be 1-2 nm, which is typical for resin-based NCGs as reported by Pfaff et al. [55]. Overall, it can be concluded that the resorcinol to carbonate ratio had almost no influence on the carbonized resins.
For comparison, the L C and L A values of crystallite in natural graphite with welldefined crystal plane patterns were calculated to be 66.3 ± 0.5 nm and 30.2 ± 0.9 nm, respectively. The overall measured intensity of the (002) plane was not only two magnitudes higher than for CXG, but also much higher than that of (101) planes, which is an indication of an L C preferred extension. Quan et al. [56] analysed different graphite samples from different mining sites and reported L C values between 10 and 18 nm and L A values between 22 and 40 nm. For the L C stacking height, however, they reported values between 10 and 18 nm, which were four to five times lower than those calculated in this investigation, while reported L A values were in good agreement with our calculations.
After synthesis and carbonization, the yielded carbons showed differences in density, with the highest for CXG 50 and the lowest for CXG 20,000 . CXG 50 especially formed very hard and dense particles. One possible explanation might be the higher concentration of sodium carbonate in the range of R/C 50-1500 that favoured fast formation of big aggregates [39]. Immediately after adding formaldehyde to the R/C 50 solution, the colour started to change from clear white to red/orange. After 2 h the polymerization was completed, resulting in a dark red colour. For R/C 500-1500 the whole polymerization was finished after 18-24 h. For >R/C 1500 the polymerization took 35-45 h due to a slow reaction rate. It appears that a slow polymerization reaction rate favoured the formation of many small aggregates which resulted after carbonization in less dense carbon particles. At very low Na 2 CO 3 concentrations, as in the R/C 10,000 and R/C 20,000 synthesis mixtures, the solution was colourless even after 60 h reaction time.

Raman
The Raman spectra of the CXG samples (see Figure 4a) showed a clear D-and G-band at 1346 and 1595 cm −1 which were comparable to those for CXGs synthesized by Kakunuri et al. [47]. An intense D-band mainly arises from the polymerization reaction products of resorcinol and formaldehyde coming from numerous sp 3 orbitals of -CH 3 , -OH and other functional groups which result after carbonization in the formation of carbon domains with multiple defect structures [48,57]. The assignment of D-and G-bands in graphitic structures, such as natural graphite, is straightforward. The G-band around 1575 cm −1 arises from the E 2g mode of all present sp 2 -hybridized carbons located in-plane in the carbon chains and rings and can be visualized as bond-stretching, whereas the D-band in graphitic structures can be explained by a "breathing" of the A 1g mode of the carbon chains and rings [58]. Therefore, the observed G-band of the CXGs can be explained by bond-stretching of carbon rings. Natural graphite has a true second-order 2D-Band at 2750 cm −1 coming from overtones and disorder induced by in-plane modes [59][60][61][62].  The intensity of a Raman signal is directly linked to the density of the material and the fourth power of the excitation frequency. Hence, since excitation frequency was kept constant over the time, sample density played a dominant role. Qualitatively, the strongest Raman intensity was obtained with the CXG2000 sample, followed by the CXG2500, CXG500 and CXG1500 samples. The intensities of the G-band assigned peak surpassed those of the D-band signals in all measurements. To obtain more information about carbon composition regarding the graphitic and non-graphitic domains, the ID/IG peak intensity and corresponding AD/AG peak surface ratio were calculated. For samples with low disorder level, the calculation of AD/AG ratio is more straightforward because it represents the probability of the whole phonon-coupling process including uncertainty [63][64][65]. For samples with high disorder level the decoupled full-width at half maximum (FWHM) values of the D-band and G-band (ΓD-Band and ΓG-Band/cm −1 ) provide more information.
Structural properties of graphitic and non-graphitic carbons, and the correlation of ΓD-Band and ΓG-Band with the lateral crystal size, have been reported by several groups [64,66,67]. Schuepfer et al. [48] demonstrated that increase in carbonization temperature up to 3000 °C strongly improved the graphitization level, especially in pitch-based samples, whereas the D-band was still visible for resin-based carbons even at this high temperature level. The authors suggested that a greater ΓG-Band value is correlated with smaller lateral crystallite sizes. We observed a median ΓD-Band ranging from 224-187 cm −1 , while for ΓG-Band, a median value of 86 cm −1 was recorded ( Figure 4b). The highest ΓD-Band of 224 cm −1 , as well as a ΓG-Band of 99 cm −1 , were measured for CXG1000, while the smallest ΓG and ΓD-Band values were calculated for CXG2500 at only 82 cm −1 and 188 cm −1 , respectively. It was found that the measured ΓD-Band, as well as the ΓG-Band, followed a similar trend. The calculated ΓG-Band values were slightly greater than those from Schuepfer et al. [48], ranging from 40-80 cm −1 for LA > 2 nm. It should be noted that the as-synthesized resins were not identical. Similar ΓG-Band values around 80 cm −1 were reported for graphene by Cancado et al. [68]. The corresponding LA values indicated sizes < 1-2 nm which fits well with the XRD results of this investigation.
The measured ΓD values > 200 cm −1 indicate LD values ≈ 2-1 nm compared with data from Ferreira et al. [64]. Since the crystallite size of our as-prepared carbon xerogels was quite similar, accurate differentiation between punctual and edge defects, as well as LD from LA values, was not possible.

Electronic Conductivity
Electronic conductivity and density values of as-prepared carbon xerogels plotted in Figure 5 followed a similar trend. The conductivity increased from 0.249 ± 0.015 S cm −1 in  The intensity of a Raman signal is directly linked to the density of the material and the fourth power of the excitation frequency. Hence, since excitation frequency was kept constant over the time, sample density played a dominant role. Qualitatively, the strongest Raman intensity was obtained with the CXG 2000 sample, followed by the CXG 2500 , CXG 500 and CXG 1500 samples. The intensities of the G-band assigned peak surpassed those of the D-band signals in all measurements. To obtain more information about carbon composition regarding the graphitic and non-graphitic domains, the I D /I G peak intensity and corresponding A D /A G peak surface ratio were calculated. For samples with low disorder level, the calculation of A D /A G ratio is more straightforward because it represents the probability of the whole phonon-coupling process including uncertainty [63][64][65]. For samples with high disorder level the decoupled full-width at half maximum (FWHM) values of the D-band and G-band (Γ D-Band and Γ G-Band /cm −1 ) provide more information.
Structural properties of graphitic and non-graphitic carbons, and the correlation of Γ D-Band and Γ G-Band with the lateral crystal size, have been reported by several groups [64,66,67]. Schuepfer et al. [48] demonstrated that increase in carbonization temperature up to 3000 • C strongly improved the graphitization level, especially in pitch-based samples, whereas the D-band was still visible for resin-based carbons even at this high temperature level. The authors suggested that a greater Γ G-Band value is correlated with smaller lateral crystallite sizes. We observed a median Γ D-Band ranging from 224-187 cm −1 , while for Γ G-Band , a median value of 86 cm −1 was recorded (Figure 4b). The highest Γ D-Band of 224 cm −1 , as well as a Γ G-Band of 99 cm −1 , were measured for CXG 1000 , while the smallest Γ G and Γ D-Band values were calculated for CXG 2500 at only 82 cm −1 and 188 cm −1 , respectively. It was found that the measured Γ D-Band , as well as the Γ G-Band , followed a similar trend. The calculated Γ G-Band values were slightly greater than those from Schuepfer et al. [48], ranging from 40-80 cm −1 for L A > 2 nm. It should be noted that the as-synthesized resins were not identical. Similar Γ G-Band values around 80 cm −1 were reported for graphene by Cancado et al. [68]. The corresponding L A values indicated sizes < 1-2 nm which fits well with the XRD results of this investigation.
The measured Γ D values > 200 cm −1 indicate L D values ≈ 2-1 nm compared with data from Ferreira et al. [64]. Since the crystallite size of our as-prepared carbon xerogels was quite similar, accurate differentiation between punctual and edge defects, as well as L D from L A values, was not possible.

Electronic Conductivity
Electronic conductivity and density values of as-prepared carbon xerogels plotted in Figure 5 followed a similar trend. The conductivity increased from 0.249 ± 0.015 S cm −1 in CXG 500 to 0.284 ± 0.023 S cm −1 in CXG 1500. The lowest conductivity was measured for CXG 750 with 0.221 ± 0.011 S cm −1 . The overall measured conductivity was around 0.260 S cm −1 which was consistent with xerogel values reported by Tondi et al. [69] (1.13-1.42 S cm −1 ), while the calculated density was relatively constant around 0.40 g cm −3 . It was observed that the conductivity fluctuated in close accordance with the Γ D and Γ G-Band values, as calculated from Raman measurements, that can be assigned to the particle size L A and point-defect distance L D . As the average Γ G was nearly constant over all CXG ratios (≈86 cm −1 ), and the measured conductivity fluctuated with little variation between 0.221 and 0.284 S cm −1 , we conclude that the through-plane conductivity of carbon xerogels was mainly influenced by sp 2 -hybridized carbons, while small changes in conductivity were a result of changes in L A and L D . For comparison, the natural graphite exhibited a conductivity of 19.3 S cm −1 which was nearly two magnitudes higher, but only 5-10% of published values for other natural graphites (152-223 S cm −1 ) [70,71]. These deviations can be explained by the difference in the analytical method used, and even more by the analysed specimen properties (e.g., type, shape, origin). Marinho et al. [72] analysed different types of graphite, graphene and multi-wall carbon nanotubes (MWCNT) at a pressure of 500 N cm −2 , which was five times higher than the pressure applied in this investigation, and they found a bulk conductivity of 21.2 S cm −1 that was comparable with our results. A detailed overview is given in Table A1.
Materials 2022, 15, x FOR PEER REVIEW 9 of 25 CXG500 to 0.284 ± 0.023 S cm −1 in CXG1500. The lowest conductivity was measured for CXG750 with 0.221 ± 0.011 S cm −1 . The overall measured conductivity was around 0.260 S cm −1 which was consistent with xerogel values reported by Tondi et al. [69] (1.13-1.42 S cm −1 ), while the calculated density was relatively constant around 0.40 g cm −3 . It was observed that the conductivity fluctuated in close accordance with the ΓD and ΓG-Band values, as calculated from Raman measurements, that can be assigned to the particle size LA and pointdefect distance LD. As the average ΓG was nearly constant over all CXG ratios (≈86 cm −1 ), and the measured conductivity fluctuated with little variation between 0.221 and 0.284 S cm −1 , we conclude that the through-plane conductivity of carbon xerogels was mainly influenced by sp²-hybridized carbons, while small changes in conductivity were a result of changes in LA and LD. For comparison, the natural graphite exhibited a conductivity of 19.3 S cm −1 which was nearly two magnitudes higher, but only 5-10% of published values for other natural graphites (152-223 S cm −1 ) [70,71]. These deviations can be explained by the difference in the analytical method used, and even more by the analysed specimen properties (e.g., type, shape, origin). Marinho et al. [72] analysed different types of graphite, graphene and multi-wall carbon nanotubes (MWCNT) at a pressure of 500 N cm −2 , which was five times higher than the pressure applied in this investigation, and they found a bulk conductivity of 21.2 S cm −1 that was comparable with our results. A detailed overview is given in Table A1.  Figure 6 shows SEM images of CXG500-2500 at 5 kV acceleration voltage and 15 k magnification. The mean size of carbon xerogel particles clearly increases with increasing R/C values from 0.079 µm in CXG500 to 0.804 µm in CXG2500. Similar surface textures have been reported elsewhere for comparable carbon xerogels [31,73,74]. In contrast, natural graphite particles consist of very wide stacked two-dimensional platelets with geometric shape >1 µm.  Figure 6 shows SEM images of CXG 500-2500 at 5 kV acceleration voltage and 15 k magnification. The mean size of carbon xerogel particles clearly increases with increasing R/C values from 0.079 µm in CXG 500 to 0.804 µm in CXG 2500 . Similar surface textures have been reported elsewhere for comparable carbon xerogels [31,73,74]. In contrast, natural graphite particles consist of very wide stacked two-dimensional platelets with geometric shape >1 µm.

Half-Cell Experiments
With the exception of CXG500 and CXG750 behaviour, the cyclic voltammograms of most CXG carbons in 1:1.5 EMiMCl:AlCl3 shown in Figure 7a do not exhibit any pronounced redox peaks in comparison to those observed for natural graphite material. The CV of mesoporous carbon CXG500 reached a maximum current density of 0.7 mA cm −2 at 2.25 V that might be assigned to the intercalation step of AlCl4 − ions into the carbon matrix shortly before the onset potential of electrolyte decomposition. The relatively small but broad reduction peak at 1.85 V (inset in Figure 7a) might be attributed to a sluggish deintercalation process. In the CV of CXG750, a slightly decreased maximum current density of 0.3 mA cm −2 at 2.25 V and a small reduction peak around 1.4 V can be seen. As shown in Figure 7b, the onset potential of the presumed AlCl4 − intercalation-related step continuously increased with increasing R/C ratio as mean pore size decreased. The CV of CXG1000 exhibited a weakly pronounced reduction peak at 1.2-1.3 V. The maximum current density produced by CXG1500-2500 samples measured at 2.25 V decreased from 0.17 mA cm −2 for CXG1500 to 0.10 mA cm −2 for CXG2500, indicating poor accessibility of active sites due to the predominantly microporous structure, as explained in Section 3.1. In contrast, the cyclic voltammograms of natural graphite (Figure 7a) show five distinct oxidation peaks during the oxidation step, where every peak corresponds to an intercalation step of chloroaluminate anions into graphene sheets. The first peak at 1.8 V can be related to a stage-6 intercalation in which in every sixth graphene sheet a chloroaluminate anion was intercalated. The assignment of the peaks is then straightforward: the main-peak at 2.3 V corresponds to stage-3 intercalation in which at every third graphene layer a chloroaluminate anion was inserted [75][76][77]. This stage also represents a fully charged graphite AIB. The peak at 2.5 V might be an overlapping of the AlCl4 − intercalation step with the onset of electrolyte oxidation. In the cathodic scan, corresponding reduction peaks related to the reversible deintercalation process are visible. Similar CVs of graphite in AIB cells have been reported elsewhere [6,78,79], though some divergences with respect to peak number and peak location may exist depending on the graphite sample used and the applied potential window.
To determine the contribution of (pseudo)capacitance and the Faradaic contribution to overall current, CVs were carried out with CXG500 and graphite materials at different scanning rates (see Figure A2a,b) and evaluated using Dunn's method where the "(pseudo)capacitance, (Q(Pseudo)cap)" designation implies here both an electrochemical double-layer capacitance EDLC (QEDLC) and a pseudocapacitive (QPseudocap) contribution. The resulting deconvoluted CVs are presented in Figure 7c,d. In the potential region of 0.5-2.0 V, the CV of CXG500 shows a broad current vs. voltage profile in comparison to graphite Stacked graphene sheets clearly visible without any porosity

Half-Cell Experiments
With the exception of CXG 500 and CXG 750 behaviour, the cyclic voltammograms of most CXG carbons in 1:1.5 EMiMCl:AlCl 3 shown in Figure 7a do not exhibit any pronounced redox peaks in comparison to those observed for natural graphite material. The CV of mesoporous carbon CXG 500 reached a maximum current density of 0.7 mA cm −2 at 2.25 V that might be assigned to the intercalation step of AlCl 4 − ions into the carbon matrix shortly before the onset potential of electrolyte decomposition. The relatively small but broad reduction peak at 1.85 V (inset in Figure 7a) might be attributed to a sluggish de-intercalation process. In the CV of CXG 750 , a slightly decreased maximum current density of 0.3 mA cm −2 at 2.25 V and a small reduction peak around 1.4 V can be seen. As shown in Figure 7b, the onset potential of the presumed AlCl 4 − intercalation-related step continuously increased with increasing R/C ratio as mean pore size decreased. The CV of CXG 1000 exhibited a weakly pronounced reduction peak at 1.2-1.3 V. The maximum current density produced by CXG 1500-2500 samples measured at 2.25 V decreased from 0.17 mA cm −2 for CXG 1500 to 0.10 mA cm −2 for CXG 2500 , indicating poor accessibility of active sites due to the predominantly microporous structure, as explained in Section 3.1. In contrast, the cyclic voltammograms of natural graphite (Figure 7a) show five distinct oxidation peaks during the oxidation step, where every peak corresponds to an intercalation step of chloroaluminate anions into graphene sheets. The first peak at 1.8 V can be related to a stage-6 intercalation in which in every sixth graphene sheet a chloroaluminate anion was intercalated. The assignment of the peaks is then straightforward: the main-peak at 2.3 V corresponds to stage-3 intercalation in which at every third graphene layer a chloroaluminate anion was inserted [75][76][77]. This stage also represents a fully charged graphite AIB. The peak at 2.5 V might be an overlapping of the AlCl 4 − intercalation step with the onset of electrolyte oxidation. In the cathodic scan, corresponding reduction peaks related to the reversible deintercalation process are visible. Similar CVs of graphite in AIB cells have been reported elsewhere [6,78,79], though some divergences with respect to peak number and peak location may exist depending on the graphite sample used and the applied potential window. tribution of the EDLC to the overall charge (0.5-2.5V) was 97.6 mC (8.4%), since the surface area of the graphite was very small compared to that of carbon xerogels. The contribution of the pseudocapacitance and the Faradaic currents could not be calculated due to the error in peak shift. The data acquisition and evaluation need to be optimised. For AIBs, typical values for the contribution of pseudocapacitance, inclusive of EDLC, are around 70-80%, while for Faradaic currents, values of 30-20% have been reported [81].

Cycling Performance of AIB Cells
The most promising carbon xerogels were further tested in the AIB cell with regards to their capacity behaviour as a function of applied current density up to 5.0 A g −1 and compared with the performance of a cell with natural graphite. As expected from half-cell experiments, AIB with CXG500 material exhibited best charge/discharge behaviour up to 28 mAh g −1 at 0.1 A g −1 with nearly 98% coulombic and 60% energy efficiency (see Figure  8). This value is in good agreement with recently published values for soft-carbon (SC) electrodes from Li et al. [82] who found a capacity of 28.2 mAh g −1 at 0.3 A g −1 current density. With increasing R/C ratio, however, the performance of AIB dropped continuously from about 15-20 mAh g −1 for cells with CXG750 and CXG1000 materials to a few mAh for the other systems. This can be ascribed to a decrease in average pore size that seems to hinder active species diffusion to intercalation sites within the carbon matrix. In comparison, the reversible capacity of the AIB cell with a natural graphite cathode reached 69 mAh g −1 . At a maximum current density of 5 A g −1 , the calculated specific capacity of AIB with CXG500 amounted to 10 mAh g −1 , one third that measured for the AIB cell with a natural graphite electrode (30 mAh g −1 ). In long-term cycling at 1.0 A g −1 , the capacity retention over 500 cycles was around 100%, with a specific measured capacity of 10-11 mAh g −1 (see Figure A1 in Appendix A). Moreover, during the first 35 cycles @ 0.1 A g −1 , the capacity of all AIBs with carbon xerogels diminished substantially, whereas the capacity of AIBs with graphite material increased. The change in capacity during the initial stage might be attributed to the irreversible intercalation of some AlCl4 − anions in the carbon structure [83] leading to a re-arrangement of the host matrix, with either capacity decrease, in the case of cells with CXG, or capacity increase (activation phase), in the case of AIBs with NG material. The activation of AIBs with natural graphite is not fully understood and is the subject of many investigations [5,77,83,84]. To determine the contribution of (pseudo)capacitance and the Faradaic contribution to overall current, CVs were carried out with CXG 500 and graphite materials at different scanning rates (see Figure A2a,b) and evaluated using Dunn's method where the "(pseudo)capacitance, (Q (Pseudo)cap )" designation implies here both an electrochemical double-layer capacitance EDLC (Q EDLC ) and a pseudocapacitive (Q Pseudocap ) contribution. The resulting deconvoluted CVs are presented in Figure 7c,d. In the potential region of 0.5-2.0 V, the CV of CXG 500 shows a broad current vs. voltage profile in comparison to graphite which can be attributed to a Q (Pseudo)cap value of 475 mC. The contributions from the Q EDLC and Q Pseudocap were determined to be 445 mC (94%) and 30 mC (6%), respectively. The latter can be assigned to insertion of chloroaluminate into the carbon host matrix. This observation is supported by the b-value of 0.86 at 2.25 V determined by Lindströms method [80] which indicates a mixed regime of fast non-diffusion limited (b-value = 1.0) and diffusion-limited (b-value = 0.5) reactions which is typical for intercalation reactions [81]. At potentials between 1.0 V and 2.0 V, the b-value was in the range of 0.97-1.01, as expected from the EDLC contribution. The plots and detailed results for determination of the b-values are shown in Figure A3a,b and Table A3. At higher anodic potentials than 2.25 V, the diffusion-limited (DL) Faradaic currents rise exponentially up to 3.4 mA cm −2 due to additional irreversible electrolyte decomposition. In the cathodic scan, a diffusion-limited reaction was observed between 2.1 V and 1.8 V and assigned to the deintercalation of chloroaluminate. The discrepancy of measured currents (black line) and experimental calculated current (green line) was less than 5%. In Figure 7d the disentangled current contributions are shown for the natural graphite. The evaluation of the data was even more complex, since the intercalation peaks drift with scan-rate. Therefore, we acquired CV scans with very low scan rates at 0.1-0.5 mV s −1 to ensure a minimal impact of the peak drift. The data shows that the measured charge was dominated by pseudocapacitive contributions. The b-values (see Figure A3c,d and Table A4) of the anodic peaks had values between 0.89 (a 1 ) and 0.67 (a 5 ) for the main intercalation peak at 2.29-2.27 V. This indicates, similarly to CXG 500 , a mixed regime of non-diffusion and diffusion limited currents, whereas for the main intercalation peak, the diffusion limitation dominates. This is supported by the deconvoluted current contributions, where, for each intercalation peak, diffusion-limited Faradaic currents are found, most intensely around the main intercalation peak. The mismatch of the measured and disentangled currents is caused by the peak drift, which was around 27 mV around the main intercalation peak. The cathodic b-values were around 0.91-0.97 which indicates nearly pure pseudocapacitive behaviour. The contribution of the EDLC to the overall charge (0.5-2.5V) was 97.6 mC (8.4%), since the surface area of the graphite was very small compared to that of carbon xerogels. The contribution of the pseudocapacitance and the Faradaic currents could not be calculated due to the error in peak shift. The data acquisition and evaluation need to be optimised. For AIBs, typical values for the contribution of pseudocapacitance, inclusive of EDLC, are around 70-80%, while for Faradaic currents, values of 30-20% have been reported [81].

Cycling Performance of AIB Cells
The most promising carbon xerogels were further tested in the AIB cell with regards to their capacity behaviour as a function of applied current density up to 5.0 A g −1 and compared with the performance of a cell with natural graphite. As expected from half-cell experiments, AIB with CXG 500 material exhibited best charge/discharge behaviour up to 28 mAh g −1 at 0.1 A g −1 with nearly 98% coulombic and 60% energy efficiency (see Figure 8). This value is in good agreement with recently published values for soft-carbon (SC) electrodes from Li et al. [82] who found a capacity of 28.2 mAh g −1 at 0.3 A g −1 current density. With increasing R/C ratio, however, the performance of AIB dropped continuously from about 15-20 mAh g −1 for cells with CXG 750 and CXG 1000 materials to a few mAh for the other systems. This can be ascribed to a decrease in average pore size that seems to hinder active species diffusion to intercalation sites within the carbon matrix. In comparison, the reversible capacity of the AIB cell with a natural graphite cathode reached 69 mAh g −1 . At a maximum current density of 5 A g −1 , the calculated specific capacity of AIB with CXG 500 amounted to 10 mAh g −1 , one third that measured for the AIB cell with a natural graphite electrode (30 mAh g −1 ). In long-term cycling at 1.0 A g −1 , the capacity retention over 500 cycles was around 100%, with a specific measured capacity of 10-11 mAh g −1 (see Figure A1 in Appendix A). Moreover, during the first 35 cycles @ 0.1 A g −1 , the capacity of all AIBs with carbon xerogels diminished substantially, whereas the capacity of AIBs with graphite material increased. The change in capacity during the initial stage might be attributed to the irreversible intercalation of some AlCl 4 − anions in the carbon structure [83] leading to a re-arrangement of the host matrix, with either capacity decrease, in the case of cells with CXG, or capacity increase (activation phase), in the case of AIBs with NG material. The activation of AIBs with natural graphite is not fully understood and is the subject of many investigations [5,77,83,84].  The 35th charge/discharge cycle carried out at 0.1 A g −1 of the different AIB cells is shown in Figure 9. Among the carbon xerogel materials, AIB with CXG500 showed a discharge capacity of 28 mAh g −1 at 0.1 A g −1 that can be attributed to its high mesoporous volume. With decrease in Vmes, the discharge capacity of AIB faded to 12-23.3 mAh g −1 in AIBs with CXG750 and CXG1000, and even lower for materials with R/C values > 1500 (7-1 mAh g −1 ). The discharge capacity of AIB with NG amounted to 69 mA g −1 . With the The 35th charge/discharge cycle carried out at 0.1 A g −1 of the different AIB cells is shown in Figure 9. Among the carbon xerogel materials, AIB with CXG 500 showed a discharge capacity of 28 mAh g −1 at 0.1 A g −1 that can be attributed to its high mesoporous volume. With decrease in V mes , the discharge capacity of AIB faded to 12-23.3 mAh g −1 in AIBs with CXG 750 and CXG 1000 , and even lower for materials with R/C values > 1500 (7-1 mAh g −1 ). The discharge capacity of AIB with NG amounted to 69 mA g −1 . With the exception of AIBs with CXG 1000 material, the cell voltage decreased/increased continuously from open-circuit down to the cut-off voltage in contrast to the polarization curves of AIBs with NG, in which distinct plateaus were visible. This correlates well with the cyclic voltammograms shown in Figure 7 and is an additional indication of different intercalation/deintercalation mechanisms of AlCl 4 − ions due to the differences in gallery/crystallite length (2 nm for CXG vs. 30 nm for NG). Since aluminium chloride is not involved in any redox process (only 3+ valence), the strong Lewis acid is assumed to coordinate with non-bonded electron pair donors [85]. exception of AIBs with CXG1000 material, the cell voltage decreased/increased continuously from open-circuit down to the cut-off voltage in contrast to the polarization curves of AIBs with NG, in which distinct plateaus were visible. This correlates well with the cyclic voltammograms shown in Figure 7 and is an additional indication of different intercalation/deintercalation mechanisms of AlCl4 − ions due to the differences in gallery/crystallite length (2 nm for CXG vs. 30 nm for NG). Since aluminium chloride is not involved in any redox process (only 3+ valence), the strong Lewis acid is assumed to coordinate with nonbonded electron pair donors [85].
The observed plateaus are related to different staging of chloroaluminate intercalation compounds into the graphite matrix [75,77]. The main plateau at 2.25 V is assigned to the stage-3 intercalation and indicates a fully charged AIB in which only one of three graphene layers is filled with chloroaluminate anions and the two intermediate layers are completely empty [75]. An ex situ XRD analysis of a fully charged natural graphite and CXG500 electrode is shown in Figure 10 and compared to that of pristine samples. The (002) peak at 26.5° in pristine graphite vanished completely, while in the charged graphite electrode, three new peaks at 16.7° (d-spacing 5.31 Å), 22.4° (d-spacing 3.98 Å) and 28.0° (d-spacing 3.18 Å) appeared which resulted from coordination of the intercalated chloroaluminate anions with delocalized π-electrons of the sp 2 orbitals from the graphite material building, forming so-called graphite intercalation compounds (GIC), as suggested in [11,75,83]. The assignment of the GIC peaks in (00l) nomenclature is straightforward: the most dominant peak is labelled as (00n+1), while the second most dominant peak is (00n+2), where (n) represents the number of a n-staged GIC. The correlation of the periodic distance (IC) of graphene layers, the intercalant gallery height (di), the gallery expansion ∆d and the intercalation stage (n) are given by the following equation [11,83,[86][87][88]: The distance 3.35 Å is given for an ideal graphite. The (l) and (dobs) are the (00l) reflex and d-spacing of the observed GIC peaks, respectively. The intercalation stage (n) can be obtained by comparing the experimental d(n+2)/d(n+1) ratios with standard values, as listed in Table 1. The observed plateaus are related to different staging of chloroaluminate intercalation compounds into the graphite matrix [75,77]. The main plateau at 2.25 V is assigned to the stage-3 intercalation and indicates a fully charged AIB in which only one of three graphene layers is filled with chloroaluminate anions and the two intermediate layers are completely empty [75].
An ex situ XRD analysis of a fully charged natural graphite and CXG 500 electrode is shown in Figure 10 and compared to that of pristine samples. The (002) peak at 26.5 • in pristine graphite vanished completely, while in the charged graphite electrode, three new peaks at 16.7 • (d-spacing 5.31 Å), 22.4 • (d-spacing 3.98 Å) and 28.0 • (d-spacing 3.18 Å) appeared which resulted from coordination of the intercalated chloroaluminate anions with delocalized π-electrons of the sp 2 orbitals from the graphite material building, forming socalled graphite intercalation compounds (GIC), as suggested in [11,75,83]. The assignment of the GIC peaks in (00l) nomenclature is straightforward: the most dominant peak is labelled as (00n+1), while the second most dominant peak is (00n+2), where (n) represents the number of a n-staged GIC. The correlation of the periodic distance (I C ) of graphene layers, the intercalant gallery height (d i ), the gallery expansion ∆d and the intercalation stage (n) are given by the following equation [11,83,[86][87][88]: The distance 3.35 Å is given for an ideal graphite. The (l) and (d obs ) are the (00l) reflex and d-spacing of the observed GIC peaks, respectively. The intercalation stage (n) can be obtained by comparing the experimental d (n+2) /d (n+1) ratios with standard values, as listed in Table 1.
(d-spacing 4.50 Å) in a charged state. This shift can be interpreted as a result of AlCl4 − intercalation into the carbon matrix since CXGs also have delocalized π-electrons from the sp2 orbitals of the aromatic resorcinol ring which are responsible for pseudo-capacitive behaviour during the intercalation step. This assumption should be confirmed by XPS analysis. Figure 10. Ex situ XRD analysis of pristine and fully charged natural graphite and CXG500 electrodes. Peaks marked with (*) indicate position of (002) peak in both pristine electrodes. The peak marked with (∆) is assumed to be associated with the (002) peak shift due to intercalated species in CXG500. Peaks indexed with (◊) are related to GIC-induced peaks in natural graphite sample.

Conclusions
In this study, the feasibility of reversible chloroaluminate anion intercalation into CXG material during AIB cycling was investigated. The most relevant results of the material texture and electrochemical charge/discharge experiments are summarized in Table  3.
Synthesis of CXG was carried out using a template-free route with resorcinol as the carbon source and formaldehyde, followed by a carbonization step at 800 °C in N2 atmosphere. The influence of the resorcinol to NaCO3 catalyst ratio in the range of 500 < R/C < 20,000 during synthesis on CXG particle morphology and electrochemical performance was studied. Physisorption analysis of carbon powder materials revealed that CXG500 had the highest mesopore volume (~77%) and pore size (6 nm) that continuously decreased with higher R/C ratios. For R/C > 2500 samples, micropore domains clearly predominated (>90%). The surface area of the prepared CXG was around 950 m 2 g −1 , with a maximum of 1000 m 2 g −1 for CXG500.
The peak positions of the D-band and G-band in Raman spectra remained unchanged at 1346 cm −1 and 1595 cm −1 , respectively, for all synthesized CXG ratios, indicating similar carbon materials. A detailed analysis of Raman spectra revealed that the synthesized CXG Peaks marked with (*) indicate position of (002) peak in both pristine electrodes. The peak marked with (∆) is assumed to be associated with the (002) peak shift due to intercalated species in CXG 500 . Peaks indexed with (♦) are related to GIC-induced peaks in natural graphite sample. Table 1. Correlation of intercalation stage (n) and theoretical d (n+2) /d (n+1) ratio according to [83,87]. 1.14 007
Inserting experimental values in equation (2), the periodic distance I C was calculated as 15.93 Å. The intercalant gallery height (d i ) for the stage-2 and stage-3 GIC were consequently determined to be 12.58 Å and 9.23 Å, respectively. The results for the gallery expansion (∆d), which were directly linked to the intercalant species, was 9.23 Å for stage-2 GIC and 5.88 Å for the stage-3 GIC. Since the radius of the intercalated AlCl 4 − was around 5.28 Å [23], the gallery expansion in stage-3 is quite compatible with the AlCl 4 − radii. The enormous expansion in stage-2 might suggest ongoing delamination of the graphene sheets. Another possibility might be the intercalation of larger Al 2 Cl 7 − anions which have twice the size of AlCl 4 − and can therefore be responsible for the extremely large expansion/delamination. An overview of our results, as well as some from the literature, is given in Table 2. It seems that there is a possible confusion between (d i ) and (∆d) in [11], because the published value for (di) differs too much between our results and already published values for the same dominant stage. The claimed (di) seems to be more logical if denoted (∆d). According to some literature sources, however, stage-2 intercalation should happen only at voltages higher than 2.6 V [75]. The observed gallery expansion (∆d) value is comparable with those from Elia et al. [83], where small discrepancies can be a result of the difference in dominant stage. Additionally, it should be noted that numerous graphite products/grades are available on the market that make accurate comparison of studies difficult. The identification of GICs in AIB graphite layers is still a challenge. By evaluating C K-edge XAS spectra of different charged graphite electrodes, Wang et al. [89] postulated that the peak at 287.3 eV results from formation of chemical Cl-C bonds after the AlCl 4 − intercalation step.
In contrast to graphite, the mechanisms of the intercalation process of AlCl 4 − species in CXGs have not yet been studied in detail. Since CXGs do not have a well-ordered crystal structure, the diffractogram did not show any sharp GIC peaks as in graphite. Nevertheless, the main peak at 2θ = 22.27 • (d-spacing 3.98 Å) of pristine CXG 500 shifted to 2θ = 19.73 • (d-spacing 4.50 Å) in a charged state. This shift can be interpreted as a result of AlCl 4 − intercalation into the carbon matrix since CXGs also have delocalized π-electrons from the sp2 orbitals of the aromatic resorcinol ring which are responsible for pseudocapacitive behaviour during the intercalation step. This assumption should be confirmed by XPS analysis.

Conclusions
In this study, the feasibility of reversible chloroaluminate anion intercalation into CXG material during AIB cycling was investigated. The most relevant results of the material texture and electrochemical charge/discharge experiments are summarized in Table 3. Synthesis of CXG was carried out using a template-free route with resorcinol as the carbon source and formaldehyde, followed by a carbonization step at 800 • C in N 2 atmosphere. The influence of the resorcinol to NaCO 3 catalyst ratio in the range of 500 < R/C < 20,000 during synthesis on CXG particle morphology and electrochemical performance was studied. Physisorption analysis of carbon powder materials revealed that CXG 500 had the highest mesopore volume (~77%) and pore size (6 nm) that continuously decreased with higher R/C ratios. For R/C > 2500 samples, micropore domains clearly predominated (>90%). The surface area of the prepared CXG was around 950 m 2 g −1 , with a maximum of 1000 m 2 g −1 for CXG 500 .
The peak positions of the D-band and G-band in Raman spectra remained unchanged at 1346 cm −1 and 1595 cm −1 , respectively, for all synthesized CXG ratios, indicating similar carbon materials. A detailed analysis of Raman spectra revealed that the synthesized CXG particles had crystallite dimensions smaller than 2 nm. Variations in Γ D-Band from 224 cm −1 (CXG 1000 ) to 188 cm −1 (CXG 2500 ) were assigned to fluctuations in L A , together with point-to-point defects L D . The Γ G-Band was almost constant at 86 cm −1 with 99 cm −1 (CXG 1000 ) and 82 cm −1 (CXG 2500 ).
Through-plane conductivity measurements of CXG 500-2500 showed conductivity lying between 0.25-0.28 S cm −1 . Variations in conductivities with respect to the CXG ratio could be ascribed to fluctuations of Γ D-Band while the Γ G-Band was understood to be beneficial for conductivity.
Crystallographic analysis of CXG showed that the stacking width L A (1.39-1.57 nm) and stacking height L C (0.94-1.06 nm) were almost independent of catalyst concentration.
The highest discharge capacity of 24-29 mAh g −1 in AIB cells was observed for CXG 500 material, compared to 69-73 mAh g −1 for AIB with a graphite electrode. With higher CXG ratios, the discharge capacity was significantly reduced compared to CXG 500 due to lower porosity.
Ex situ XRD analysis of a charged CXG electrode exhibited a small peak shift of the main peak from pristine 2θ = 22.27 of about 2.54 • towards smaller degree that might be an indication of intercalation of chloroaluminate into the carbon structure. For the charged graphite electrode, the dominant intercalation steps were determined to be stage-2 and stage-3, with gallery expansion of the adjacent graphene layers of about 5.88 Å and 9.23 Å, respectively.
In this study, we have demonstrated for the first time the feasibility of carbon xerogels prepared by a simple and cheap soft-template route as positive electrode material for aluminium ion batteries, paving the way for future investigations. Further improvements are still needed to enhance the overall capacity of CXGs, and, more particularly, the contribution of AlCl 4 − ion intercalation (pseudocapacitance) to overall capacity that might be achieved by enhancing the graphitization level of CXG through, for example, graphene doping.      The peaks named a1-a5, as well as c1-c5 in CXG500 CVs, were randomly selected to give a good overview over the whole CV. Please note that the nomenclature has nothing to do with the true intercalation or deintercalation peaks with the same nomenclature as observed with natural graphite.   Figure A2b  Since we were not able to calculate the separate contribution of (pseudo)capacitance and Faradaic currents to overall charge value, Q EDLC ratio is referred to as Q Measured .