Electrical Double Layer Capacitors Based on Steam and CO₂-Steam Co-Activated Carbon Electrodes and Ionic Liquid Electrolyte

The CV method was used to obtain the region of ideal polarizability and therefore CVs were measured at different potential scan rates ν from 1 to 500 mV s⁻¹ and in a very wide cell potential region ΔE up to 3.8 V.

Data in Fig. 2a indicates that nearly exponential increase of current density started only at ΔE ≥ 3.6 V. At higher ΔE values very weak increase of j can be explained by the mass transfer step limited faradaic process, caused by the very low concentration of O₂, H₂O and other electrochemically active residuals in EMImBF₄ liquid phase. At very high potential scan rates ν > 200 mV s⁻¹ and current densities (Fig. 2c) there is no exponential increase of j at ΔE ≥ 3.6 V, explained by slow mass transfer rate of residuals onto SiC-CDC surface. Thus, due to the high dryness and cleanness of EMImBF₄ used, the rates of electrochemically active reactants reactions are very slow.

Zoom In Zoom Out Reset image size

The gravimetric capacitance C_m;CV values were calculated from the measured CV curves, according to Eq. 1 and Eq. 2.

The Eq. 1:

where j is the current density (per geometrical surface area of electrode), ν is the potential scan rate and C is the cell capacitance. Eq. 1 is correct if the capacitance does not depend on the potential applied (C ≠ f(ΔE)) and the series resistance of the system R_s → 0. Therefore, this equation is applicable in case of very small current density values (within the region of slow v), where the ohmic potential drop is negligible and the current response is essentially equal to that of a pure capacitor.^13,14 In a two–electrode symmetrical EDLC system the specific capacitance C_m (F g⁻¹) can be obtained from the capacitance of the cell:

where m is the weight of one carbon electrode. Eq. 2 can be implemented if the positively and negatively charged electrodes have the same capacitance at the cell potential region applied.^1–3,13,29

CV curves expressed as specific capacitance (C_m;CV) vs. cell potential (ΔE) are presented in Figs. 2b–2d. In Fig. 2d for comparison the CO₂ activated material data having the best electrochemical properties, denoted as SiC-CDC 1100 A3 from our previous work, is also given.²

The EDLCs demonstrated nearly ideal capacitive behavior at cell potential scan rate ν ≤ 50 mV s⁻¹ and up to ΔE ≤ 3.6 V in EMImBF₄ (overlapping of discharging capacitance in discharging branch) (Figs. 2b–c). Even at ΔE = 3.8 V, there are only small deviations from ideal capacitive behavior. Especially at low ν some increase of faradaic capacitance has been observed, explained by the slow mass transfer process of electrochemically active reactant (H₂O) from RTIL to the electrode surface. At moderate potential scan rates 20 ˂ ν ˂ 200 mV s⁻¹, there are smallest deviations from CV shape characteristic for ideally polarizable system, explained by very low concentration of residuals in systems under study. Comparing these data to our previous work results conducted with the CO₂ activated materials,² we can see that for the steam activated materials based EDLCs we can use a bit higher potential scan rates, lower IR-drop influence on the shape of C_m vs. ΔE plots, where the capacitive behavior remains nearly ideal. It is mainly caused by the more expressed amount of bigger micropores and mesopores at steam activated materials surface, increasing the conductivity of electrolyte in the mesoporous matrix.

Analyzing data for 1 M (C₂H₅)₃CH₃NBF₄ (Et₃MeNBF₄) acetonitrile (AN) solution as the electrolyte, we discovered that although the materials differ in their activation and treatment parameters and also in their specific surface areas and pore volumes, the electrochemical behavior was nearly overlapping for all of them.³ Therefore, this confirms again that the microporous region characteristics determine the nearly linear correlation between specific surface area and capacitance, visible up to S_DFT values ∼ 1000 m² g⁻¹. However, at higher S_DFT values the logarithmic dependence has been observed and at very high S_DFT values a plateau in C_m;CV vs. S_DFT plot has been observed.¹⁷

For Et₃MeNBF₄ + AN based electrolytes^19,20 the region of ideal polarizability is weakly narrower like for EMImBF₄ + AN electrolyte based EDLCs.^{1–3,14,15,24,25}

Using EMImBF₄ as an electrolyte it can be seen that one of the materials falls out from the others – St-CO2-900 based EDLC has a smaller capacitance value and this can be explained by the smallest S_BET, V_micro and V_tot values and narrowest pore size distribution data (Table II, Fig. 1). All the other materials again differ only a little in their electrochemical behavior.

The values of C_m;CV start to increase nearly exponentially at cell potentials ΔE ≥ 3.6 V, which can be explained by the start of faradaic decomposition of an insignificant amount of residual water and reduction of O₂ in RTIL. Thus, by electrochemical reduction of H₂O and O₂ traces at a negatively charged electrode as well as by oxidation of the surface functionalities at a positively charged electrode takes place at ΔE > 3.6 V.^14,15 With increasing the cell potential scan rate v over 100 mV s⁻¹, the cyclic voltammograms and the specific capacitance vs. cell potential dependencies (Fig. 2c) become distorted in the region of potential switchover explained by the slow mass-transfer step in the porous CDC electrode matrix due to the high viscosity of EMImBF₄.^7,14,15,17

Specific capacitance C_m;CV values calculated from the linear region of the CV curves after changing the direction of the potential scan rate v = 10 mV s⁻¹ at cell potential 3.6 V are shown in Table III. As already noted St-CO2-900 has the smallest C_m;CV value (131 F g⁻¹) and St-1000 the highest (C_m;CV = 145 F g⁻¹). Previously studied CO₂ activated materials had all C_m;CV values below 130 F g⁻¹.

Capacitance vs. cell potential scan rate curves are given in Fig. 2e. Capacitance values have been calculated using the integrated total charge values, over the cell potential range ΔE = 3.2 V.²⁶ For the EDLC based on the most microporous St-CO2-900 material the capacitance values decrease fastest and in a biggest extent – at potential scan rate 20 mV s⁻¹ this system retains about 89% of its initial capacitance, but at 500 mV s⁻¹ it is only ∼ 30%. All the other systems are based on steam activated carbons with a better micro-mesoporosity ratio, which results in a slower decrease of capacitance. These systems retain ∼ 94% of its initial capacitance at potential scan rate 20 mV s⁻¹ and at 500 mV s⁻¹ ∼ 65%.

The C vs. ΔE – curves have been integrated and the charge density (σ) vs. ΔE plots (Fig. 3) have been constructed, because the σ vs. ΔE values are more informative than sometimes non-linear capacitance vs. ΔE data. Data in Fig. 3 show that within very wide potential region (up to 3.6 V) the σ vs. ΔE - plots are symmetrical. Only at ΔE ≥ 3.6 V, some non-linearity of σ vs. ΔE – plots has been established. Thus, the slow faradaic reactions started only at ΔE ≥ 3.6 V. From σ vs. ΔE – plots the integral capacitance values C = 132 F g⁻¹, in good accordance with C_{m; CV} data, have been calculated. This method eliminates the role of measurement error of current pulses in CV curves. There is no big influence of the potential scan direction on the σ_int (discharge)/ σ_int (charge) values, observed in many EDLC papers. The very high discharging/charging coulombic efficiency values n_σ = σ_disch/σ_char ≥ 97% have been calculated.

Zoom In Zoom Out Reset image size

EDLCs were tested at different fixed constant current (CC) regimes from 0.01 to 5 A g⁻¹ applying charging/discharging steps within the cell potentials from 0 to 3.0 V (Figs. 4a, 4b). This ΔE region correlates to the maximum limit of commercially available EDLCs.^16,17

Zoom In Zoom Out Reset image size

The discharge and charge capacitances, C_cc, were calculated from the data of the third cycle.²⁴ C_cc (F cm⁻²) was obtained from the slope of the discharge (or charge) curve according to Eq. 3:

where dt/d(ΔE) is the reciprocal value of the slope of discharging or charging curve. For simplicity, the discharging curves were approximated by a linear function. Therefore, the medium integral capacitance C_cc values calculated from CC/CD data are somewhat different than the C_m;CV values calculated from CVs (Table III). Slightly lower capacitance values obtained by CC/CD method can be explained by physical differences in methods applied for charging/discharging of electrodes as well as by not totally linear shape of the charging/discharging curves used for the calculation of C_cc. For more correct analysis the CC/CD curves have been integrated and the energy densities stored (E_in) and released (E_out) have been obtained. The energy efficiency values have been calculated as:

where t is time in seconds and ΔE_max and ΔE_min are maximal and minimal cell potentials applied.^17,24,25

The n_en values from 84 to 92% have been obtained, increasing from very microporous St-CO2-900 to most mesoporous material based EDLC St-1000.

The charge-discharge curves for the activated materials are nearly linear and symmetrical, demonstrating very good electrochemical reversibility and coulombic efficiency at smaller current densities j = 0.1 A g⁻¹ (Fig. 4a). Even at higher current densities j = 1 A g⁻¹ (Fig. 4b), weak deviation from linear shape and a small IR-drop can be seen. The exception is again the St-CO2-900 material based EDLC and for this system CC/CD data get somewhat distorted at higher current densities.

The coulombic efficiency has been calculated as a ratio of integrated charge released and accumulated during discharging and charging of the EDLCs.¹⁷ The calculated coulombic efficiency values for all the systems remained within the range of 97 to 98.5% (at current 0.1 A g⁻¹) and from 99.3 to 100% (at current 1 A g⁻¹), showing that the steam activated SiC-CDC materials are very promising for EDLC and various energy storage applications. It should be noted that differently from many EDLCs, the charging and discharging curves for St-CO2-900 and other electrode materials are curved upwards explained by the heavily microporous characteristics of carbons used. With the decrease of microporosity and increase of mesoporosity more linear discharging (as well as charging) curves have been measured.

Nyquist plots were measured in ac frequency range from 1 mHz to 300 kHz at fixed cell potentials from 0 to 3.8 V (Figs. 5a, 5b). From Fig. 5a it can be seen that the shape of the Nyquist plot is nearly independent of ΔE applied, if ΔE ≤ 3.2 V. At higher cell potentials the low frequency part slowly starts to drift away from the ideal −90 degree slope. The Nyquist plots (Fig. 5b) for the steam activated materials consist mainly of two parts: (1) of the so-called "porous" region with a slope of α nearly −45° in ac frequency region 0.1 < f < 1000 Hz, characteristic of the mass transfer limited process (with absorption boundary conditions) in the micro-mesoporous carbon electrode matrix of an electrode and (2) of the low frequency double-layer capacitance region with a slope of α ≈ −90° ("knee" at f ˂ 0.1 Hz), obtained by the finite length absorption effects.^9,11,25 EDLCs based on St-900-H2 and St-1000-H2 electrodes have the highest value of "knee" frequency, which indicates the easier accessibility of electrolyte ions into the pores.^8,17,27 As can be seen in Fig. 5b, all the materials show again very similar performance, except St-CO2-900 and SiC-CDC-1100 A3, which behavior looks more similar to previously studied heavily microporous CO₂ activated materials based EDLCs.^1–3 More detailed analysis of Nyquist plots shows that more microporous materials also show a third region in Nyquist plots, i.e. a small semicircle at higher frequencies f > 100 Hz,² explained by the mass transfer and charge transfer processes at macroporous surface regions.^1–4

Zoom In Zoom Out Reset image size

The specific series capacitance C_m;s values have been calculated from the impedance spectroscopy data at ac frequency f = 1 mHz and are shown in Table III and C_m;s vs. f plots are given in Fig. 5c. The C_m;s values are somewhat higher, but nevertheless in a good agreement with the values obtained using CV and CC/CD methods. The higher C_m;s values could be explained by not completely established electrochemical adsorption equilibrium if CV and CC/CD methods have been used. However, sometimes the EIS method over-estimates the capacitance values due to enhanced contribution of parasitic reactions to the electrochemical response in porous carbons.

The phase angel θ vs. ac frequency dependencies are shown in Fig. 5d. Three of the materials show similarities to the CO₂ activated material data, but still most of the steam activated materials have phase angles values between −83° to −86°, again proving the better performance and effectiveness compared to CO₂ activated materials.^1–3

Existence of long plateau in C_m;s vs. f- as well as C_m;p vs. f-plots (Figs. 5c and 6a), and overlapping of C_m;p vs. f and C_m;s vs. f plots (Fig. 6b) at f ˂ 0.05 Hz indicates that the ideal capacitance values (free from faradaic capacitance contribution) have been established.^17,28,29 All the C_m;s values calculated are quite similar for the steam activated materials (Figs. 5c, 6a, 6b). The CO₂ activated material SiC-CDC-1100 A3 data (from our previous work) shows somewhat different capacitance behavior, but it has to be noted that it was activated for 16 hours. All the other CO₂ activated materials (activated for 2 hours) showed noticeably smaller capacitance values than steam and CO₂-steam activated SiC-CDCs. The St-CO2-900 as well as SiC-CDC 1100 A3 materials based EDLCs demonstrated very slow adsorption rates due to not very well developed mesoporosity of the carbon powders.

Zoom In Zoom Out Reset image size

In an ideally polarizable system the ratio of C_m;p/C_m;s should be 1. However, it is shown in Fig. 6, that the ratio of C_m;p/C_m;s for many EDLCs is very slightly lower than 1 at very low frequencies. This indicates that the increase in capacitance from impedance spectroscopy at higher potentials ΔE > 3.6 V is not truly initiated by the increase of the electric double layer charging (increase of Gibbs adsorption value), but probably caused by very slow faradaic processes, like reduction of residual water, O₂ and other additives.^25,30

These conclusions are in very good agreement with data in Fig. 7a, where |-Z''| vs. f curves are given. At f ≥ 1000 Hz, there is weak dependence of |-Z''| on the cell potential applied and with the increase of ΔE, the |-Z''| increases, explained by the quick faradaic processes at outer surface areas of SiC-CDC electrode.

Zoom In Zoom Out Reset image size

Within moderate frequency region, nice linear section has been established (with slope −0.44 to −0.48) characteristic of finite-length mass transfer processes.^{1–4,14,15,17,33} At f ˂ 0.1 Hz, the second linear region characteristic of finite-length adsorption step limited process (with slope – 0.97 to – 0.99), demonstrating that the role of faradaic processes on the capacitance values is very low, expect at ΔE ≥ 3.6 V. However, data in Fig. 7b show that the shape of |-Z''| vs. f plots depends on the SiC-CDC materials micro-, mesoporosity and specific surface area values, i.e. on the activation conditions used. For CO₂-steam co-activated, but not H₂ reduced materials, very high |-Z''| vs. f values have been calculated at f ≥ 0.1 Hz. Thus, materials without high temperature hydrogen reduction cannot be used for completing of high energy/high power density EDLCs with cell potential ΔE > 3.2 V. The same conclusion can be reached analyzing the Nyquist plots (Fig. 6) as well as Z' vs. f –plots (not shown for shortness), demonstrating noticeably high active resistance values for St-900 and St-CO₂-900. In addition, it seems that the high |-Z''| values at f > 10 Hz can be explained by very microporous structure of the carbon powders (Table II) increasing the mass transfer resistance of RTIL ions nearly 20 times if compared with H₂ reduced materials.³³ The increase of the synthesis temperature from 900°C to 1000°C and additional reduction of powders with H₂ decreases the -Z'' values nearly 50 times. It is very important to mention that at low frequency region (f ˂ 0.1 Hz) there is no dependence of capacitance (|-Z''| i.e differential capacitance) on the synthesis and activation method used for preparation of SiC-CDC powders. There are linear areas in |-Z''| vs. f plots with the slope values varing from −0.97 to −0.99 indicating that the ideal capacitive behavior has been established for all EDLCs under study.

The characteristic time constant τ values have been calculated using imaginary and active power density vs. f plots^1–4,17 (Fig. 8). It is clearly visible that τ value increases nearly 10 times compared the best and lowest EDLC. It should be noted that for fixed EDLC, τ is independent of ΔE, if ΔE ≤ 3.6 V. Thus, this region of cell potential can be applied for real EDLC applications.

Zoom In Zoom Out Reset image size

The experimental gravimetric and volumetric Ragone plots measured (specific energy, E, vs. specific power, P, dependencies) calculated to the total material weight (including active carbon material, Al current collector and binder) or volume of two electrodes (for the EDLCs based on different SiC-CDC electrodes) have been obtained from the constant power tests within the cell potential range from 1.7 V to 3.4 V or from 1.9 V to 3.8 V, and are shown in Figs. 9a–9d.^17,31–33

Zoom In Zoom Out Reset image size

For all the materials under study very good performance has been established. At power density 1 kW kg⁻¹ the highest energy densities of 52 Wh kg⁻¹ (ΔE = 3.4 V, Fig. 9a) and 71 Wh kg⁻¹ (ΔE = 3.8 V, Fig. 9b) were calculated for the material St-1000. The excellent volumetric energy densities at P = 1 kW dm⁻³, E = 27 Wh dm⁻³ (Fig. 9c) and 39 Wh dm⁻³ (Fig. 9d) at ΔE = 3.4 V and 3.8 V, respectively, have been achieved. At higher power density 10 kW kg⁻¹, the St-CO2-900 showed the highest energy density values (28 Wh kg⁻¹ and 12 Wh dm⁻³ at ΔE = 3.4 V and 37 Wh kg⁻¹ and 17 Wh dm⁻³ at ΔE = 3.8 V). Thus, very high energy densities at high power densities can be achieved for steam and CO₂-steam activated SiC-CDC carbon electrodes and EMImBF₄ based EDLCs. Comparison with other EMImBF₄ based EDLCs, given in Fig. 9e show, that for D-glucose (GDAC), granulated white sugar derived (GWS carbon) or TiC-CDC based EDLC^3,17 higher energy density values at fixed P_min (3.6 s) or P_max (0.36 s) have been established. However, in this work very cheap raw SiC and cheap SiC-CDC synthesis methods have been used.^1–3 Fig. 9f shows a comparison between EMImBF₄ and 1 M Et₃MeNBF₄ in AN electrolytes for different carbon materials based EDLCs. It has to be noted that for the steam activated SiC-CDC (from this study) the difference between the electrolytes is the smallest. Even more, the SiC-CDC materials activated with only CO₂ (from previous study^1,2) show in ionic liquid significantly lower energy and power density values.

Life time test was carried out for St-CO2-1000 + EMImBF₄ based EDLC. CC/CD tests between cell potentials ΔE from 1.5 to 3.0 V at a constant current density of 1 A g⁻¹ have been conducted (Fig. 10). The charge/discharge curves remained practically linear after 5000 cycles (Fig. 10, inset). The IR-drop was nearly constant up to 2000^th cycle, after which it very slowly started to increase. Capacitance values decreased from 140 F g⁻¹ to 107 F g⁻¹. Energy efficiencies dropped from 96% to 94% (after 5000 cycles) and coulombic efficiencies remained during the cycling in the range of 97–99%. These results show that the steam activated SiC-CDC materials are very promising for various high energy storage applications.

Zoom In Zoom Out Reset image size