Increasing Capacitance of Zeolite-Templated Carbons in Electric Double Layer Capacitors

A low pressure (1∼10 Torr) CVD using a simple hydrocarbon (acetylene, C₂H₂) was employed for depositing carbon onto the zeolite surface.⁶⁶ During deposition ∼3 g of Zeolite Y powder (CBV100, Zeolyst International, Delfzijl, Netheraland) was placed in a 4.5 inch quartz boat, which was loaded inside of a quartz tube (1.2 inch diameter, 2 ft length) furnace (21100, Barnstead Thermolyne, USA). The zeolite was first heated to the desired temperature (700, 800 or 900°C) under an Ar gas flow. Once the temperature was stabilized, Ar flow was halted and acetylene gas was then allowed to flow at 100 sccm for desired time. After the deposition, acetylene flow was halted and Ar flow passed through the tube to avoid any possible formation of surface groups at high temperature while the furnace was cooled down for retrieval. Selected samples were additionally post-annealed under Ar. Subsequently, the carbon coated zeolite powder was submerged in a 10 wt% hydrofluoric acid (HF) solution for 24 hours to remove the zeolite template, and rinsed-off using a vacuum filtering system using a asymmetric polyethersulfone (PES) membranes (VWR, USA). Remaining impurities were removed by 50 wt% sulfuric acid and followed by repeated rinsing-off and drying under vacuum for 12 hours. Selected samples were ground by ball milling before etching the zeolite template. In these experiments, the samples were put together with 12 of 6 mm diameter stainless steel balls (Across International, USA) into a 50 ml vacuum jar (Across International, USA) assembled in an Ar-filled glove box (Innovation Technology, USA). A ball milling tool capable to control the rotating speed (QM-3SP04 Planetary Ball Mills, USA) was used to mill the samples for a desired time. Physical activation was performed on selected samples. For these experiments, the ZTC powders were heated to a desired temperature (700–900°C) in the furnace under Ar flow at a rate of 300 mL min⁻¹. Once the target temperature was achieved, the Ar flow was halted and CO₂ gas (99.9%, Air Gas, USA) was flowed at a rate of 150–300 mL min⁻¹.

ZTC powders were mixed with polytetrafluoroethylene (PTFE) as a binder in the ratio of 9:1 to prepare freestanding electrodes. The mixed and dried carbon-PTFE composite was then roll pressed to be a ∼200 μm thickness film. The produced electrode was dried in a vacuum oven at 100°C for at least 12 hours to remove any moisture left in porous carbons and cut into 1/2 inch diameter circles. The final electrode mass was around 6 mg. Coin cells were assembled inside an Ar-filled glove box (Innovation Technology, USA). Two symmetrical electrodes were separated by two overlapped 25 μm thick Gore PTFE separators (W.L. Gore and Associates, USA) and each electrode was attached to Al current collectors. Current collectors were prepared by spray coating of a thin layer of conductive graphite slurry (BW 525, Superior Graphite, USA) on a 300 μm thick Al foil to reduce interfacial resistance.⁶⁷ 1 M tetraethylammonium tetrafluoroborate (TEABF₄) solution in acetonitrile (AN) solution was employed as electrolyte. Cyclic voltammetry (CV) measurements were recorded using a Solartron 1480A MultiStat (Solartron Analytical, UK) in the voltage range –2.3 V to +2.3 V with scan rates of 1 to 500 mV s⁻¹. The carbon specific capacitance, C (F g⁻¹), was calculated according to:

where I is the current (A), dV/dt is the scan rate (V s⁻¹), and m is the mass (g) of carbon in each electrode. Charge-discharge (C-D) characteristics were recorded using an Arbin battery/supercapacitor electrochemical tester (Arbin Instruments, USA). Elecrochemical impedance spectroscopy (EIS) measurements were carried out using a Gamry reference 600 potentiostat (Gamry Instruments, USA) in the frequency range of 1 mHz – 100 kHz with a 10 mV AC amplitude. The gravimetric capacitance, C (F g⁻¹), was calculated from EIS data according to Eqn. 2, where f is the operating frequency (Hz), Im(Z) and Re(Z) are the imaginary and real parts of the total device impedance (Ohm), and m is the mass (g) of carbon in each electrode.

A field emission scanning electron microscope (FE-SEM, Zeiss Ultra60 FE-SEM, Germany) and a transmission electron microscope with scanning mode capability (Tecnai G² F30 TEM, FEI, Netherlands) were employed for observing the structural and morphological features. Crystal structures were studied using X-ray diffraction (XRD) measurements collecting peaks from 2 Θ = 2° to 70° with a step size of 0.02°. The N₂ adsorption isotherms were measured at –196°C using TriStar II 3020 surface area and porosity analyzer (Micromeritics Instrument Corporation, USA). Prior to the measurement the powder samples were degassed at 300°C for at least 4 hours using VacPrep 061 Degasser (Micromeritics Instrument Corporation, USA). The specific surface area (m² g⁻¹) was calculated using the Brunauer, Emmett, and Teller (BET) method. The pore size distribution (PSD) was determined via the density functional theory (DFT) methods using nitrogen adsorption data. Raman spectra were acquired using a Nicolet Almega XR spectrometer (Fisher scientific, usa) with an excitation wavelength of 488 nm.

In our initial studies we evaluated the influence of the temperature of carbon deposition on zeolite because it was expected to strongly affect the microstructure of a produced ZTC.⁶⁶ Figures 1a–1c show the SEM images for the ZTC (after etching the zeolite template) at carbon deposition temperatures of 700, 800 and 900°C. We see that while the 700°C (and to a slightly less extent 800°C) deposition yielded very uniform carbon particles precisely replicating the shape of the zeolite templates, the deposition at 900°C resulted in ZTC having irregularly rough surface originating from collapsing the ZTC particles during zeolite etching. Accelerated decomposition of acetylene gas at such a high temperature evidently deposited carbon mostly on the outer surface of the template. The surface reactivity was so high that the LPCVD process was limited by the diffusion kinetics instead of the surface reaction and the precursor molecules (acetylene) did not have sufficient time to diffuse into the micropores of zeolite particles prior to decomposition. Not surprisingly, such samples exhibited the smallest specific capacitance (Fig. 1d), while the carbon produced at 700°C showed the best overall performance. For this reason, carbon deposition temperature was fixed at 700°C for the following experiments.

Zoom In Zoom Out Reset image size

In order to enlarge the SSA in the produced carbon, CO₂ activation was performed on ZTC samples. The effect of activation conditions including the activation temperature, time and flow rate of CO₂ was investigated and the resulting porosities of the produced carbons are summarized in Table I. The ZTC powder without any post-synthesis treatments is named as n-ZTC. CO₂ activated samples are named as C-ZTC followed by the temperature and the duration time of activation in minutes written in parenthesis.

Activation at 900°C resulted in an over-oxidation (burning) of the samples and a decrease in their SSA and pore volume. Longer time at such a relatively high temperature yielded smaller porosity. Activation at a more moderate temperature of 800°C resulted in a noticeable improvement with an optimized activation conditions (800°C for 15 min activation) yielding SSA as large as 2390 m² g⁻¹. The high sensitivity of the achieved porosity to activation conditions (See Table I) suggested that a very delicate optimization of the post synthesis activation of ZTC is generally required to induce additional pores in ZTC and remove amorphous carbon without burning pore walls, which are mostly composed of single graphene layers.⁶⁵

Figure 2a summarizes the dependence of the specific capacitances of ZTC samples (after activation at different conditions) on the CV scan rate. Only the best performing (at each activation temperature) samples were included in this graph to make it less crowded. All samples activated at 900°C resulted in much lower specific capacitance compared to n-ZTC, which is consistent with their lower SSA and pore volume (Table I). In contrast, the samples activated at 700 and 800°C showed improved specific capacitances, particularly when measured at the slowest scan rates (Fig. 2a). The sample C-ZTC800(15m) showed the highest specific capacitances of 204 F g⁻¹ at a sweep rate of 1 mV s⁻¹. It also exhibits some of the pseudocapacitance peak near 0 V (Fig. 2b), which may increase its total capacitance, but at the expense of possible side effects, such as self-discharge.¹ Other than that, the CV curves retained a near-rectangular shape at a relatively high sweep rate of 100 mV s⁻¹, suggesting rather fast ion transport within the microporous ZTC electrodes. The possible origins of the often observed butterfly shape of the CVs (Fig. 2b) have been previously discussed in some of our former publications.¹⁴ The achieved values of the specific capacitance were found to correlate with the SSA of the produced ZTC.

Zoom In Zoom Out Reset image size

The activation commonly resulted in a small reduction in the resistance and frequency response of the EDLCs. Figure 3 compares the EIS results of n-ZTC with that of the C-ZTC800(15m). Majority of the tested devices exhibit virtually no semi-circuit loop at high frequencies (Fig. 3a), which indicate a low resistance of the interface between the current collector and electrodes.^68,69 We also see that activation resulted in lowering the equivalent series resistance, ESR (intersection of the graph with the x axis at the high frequency), from 2.7 to 1.2 Ohm cm² in the case of 15 min activation at 800°C (Fig. 3a). This ESR includes electrolyte resistance as well as the electrode resistance. Since we used identical electrolytes and electrode thicknesses we can conclude that activation reduced electrode resistance, possibly by removing some of the amorphous carbon or defects that prevented efficient compaction of the ZTC particles in the assembled electrodes. The reduced resistance allowed slightly improved frequency response in the activated samples, where a capacitance drops to 50% of its maximum value at higher frequencies than in the as-produced sample, 0.3 Hz for C-ZTC800(15m) vs. 0.2 Hz for n-ZTC (Fig. 3b).

Zoom In Zoom Out Reset image size

The pore size distribution (PSD) of the ZTC samples was derived from N₂ adsorption isotherms at –196°C. The activation process did not change the PSD of the ZTC samples significantly. For example, the sample with the highest degree of activation, C-ZTC800(15m) showed PSD similar to that of n-ZTC, but with slightly larger pore volume observed in the range of 1.5-5 nm (Fig. 4). The slightly larger volume of mesopores may contribute to the slightly better frequency response of the activated samples (Fig. 3b).

Zoom In Zoom Out Reset image size

While the size of the primary ZTC particles was only 0.2–1 μm, the size of the aggregates was larger, 3–30 μm. Large size of agglomerates may, in principle, slow down ion transport. The high resolution TEM studies revealed that the surface of the ZTC particles (even those where carbon was deposited at 700°C) was coated with the 2–5 nm (for n-ZTC produced at 700°C and larger for samples produced at higher temperatures) layer of disordered carbon. Such a layer may reduce the N₂ accessible SSA and similarly slow down the ion flux from outside of the ZTC particles to their inner pores. We hypothesized that a gentle ball milling process could break-up the agglomerates, thereby reducing the effective diffusion distance and additionally allowing electrolyte transport around the ionically resistive surface layer. In order to maintain the pore alignment (prevent collapse of the pores) and avoid complete crushing of the rather fragile ZTC particles, ball milling was conducted before etching the zeolite templates. The influence of ball mill time on the carbon porosity was investigated by N₂ adsorption analyses. The summary of which is shown in Table II.

Ball-milling enhanced samples are named as B-ZTC, followed by the milling duration time in minutes written in parenthesis. B-ZTC(15 m) indicates that 15 min of ball milling was applied to the C-coated zeolite before zeolite etching. Increasing the ball milling time up to 30 min, the BET SSA gradually increased by ∼300 m² g⁻¹ compared to the original sample, n-ZTC. Longer ball milling process did not affect the SSA (Table II).

The effect of post-ball milling process on improving the specific capacitance is shown in Figure 5a. Surprisingly, B-ZTC(30 m) sample (which possess the highest surface area, Table II) did not show significant improvement in the specific capacitance (Fig. 5a). The PSD of this sample was rather similar to that of the initial n-ZTC sample (Fig. 5b), with a small reduction in the relative volume of the smallest micropores. The reduced volume of such pores may lead to the reduced capacitance,⁹ even if the SSA is 18% higher. Interestingly, B-ZTC(120 m) sample provided the highest specific capacitances of 194 F g⁻¹ at a sweep rate of 1 mV s⁻¹. The PSD of this sample showed that the 120 min ball milling creates the highest content of small micropores (∼1 nm), which might be the reason for the great increase in the capacitance at slow scan rates as well as for a significant capacitance drop at rates higher than 100 mV s⁻¹. We hypothesize that these additional pores may originate from the activation of the C coating by the zeolite substrate induced by mechanical forces (mechanochemical reactions).

Zoom In Zoom Out Reset image size

Annealing of the ZTC samples in an inert environment at temperatures higher than that of the synthesis (carbon deposition) temperature was performed in order to test a hypothesis that such an annealing process can heal the defects in the produced carbon structure, which, in turn, might allow for better retention of the pore alignment. For this series of experiments, carbon-coated zeolite samples were annealed at temperatures of 800, 900 and 1000°C under Ar prior to zeolite etching (the finally produced ZTC samples were named A-ZTC800, A-ZTC900 and A-ZTC1000, respectively). Selected samples were additionally subjected to ball milling prior to zeolite etching (the finally produced ZTC samples were named AB-ZTC800, AB-ZTC900 and AB-ZTC1000 respectively) in order to see the sequential effect of annealing and ball milling processes.

Annealing at 900°C resulted in the most noticeable performance improvements. Figure 6a shows the effect of individual annealing (900°C) and ball milling processes as well as the combined effect of annealing followed by ball milling on improving the capacitance of ZTC. Annealing at 900°C improved the overall capacitance by a moderate ∼20 F g⁻¹, without significantly affecting the rate performance. This was in contrast to our prior studies in aqueous electrolytes, where significant improvements in both capacitance and rate performance was achieved upon implementing the post-deposition annealing.²⁵ We may conclude that larger size of ions in organic electrolytes and their stronger interactions with the walls of relatively small 1 nm pores affected the ion transport rate more significantly than concentration of defects within carbon. More importantly for practical applications, we observed a synergetic effect when annealing was conducted prior to the ball milling (AB-ZTC900). The highest value of the specific capacitance (240 F g⁻¹ at a scan rate of 1 mV s⁻¹) and improved rate performance was observed compared to the ball milled ZTC (Fig. 6a). Figure 6b shows that this best performing sample exhibited larger volume of 2–4 nm mesopores and nearly ∼1000 m² g⁻¹ higher SSA than the regular n-ZTC sample (Table III).

Zoom In Zoom Out Reset image size

Inspired by these promising results we systematically investigated the impact of the temperature of the (post-carbon deposition / pre-milling) annealing. The annealing process at all temperatures from 800 to 1000°C resulted in higher SSA and pore volume compared to both regular ball milled samples (compare Table III and II), suggesting that annealing assisted in retaining the pore structure and prevented collapse of at least some of the pores. The optimal performance, however, was achieved for the 900°C annealing. Not surprisingly (considering somewhat similar PSD), significantly enhanced SSA of the annealed samples resulted in higher values of the specific capacitance (Figure 7).

Zoom In Zoom Out Reset image size

In order to quantify the changes in the degree of carbon ordering upon annealing or ball milling we conducted Raman spectroscopy studies. However, the broad Raman bands of disordered (and mostly single-walled) carbon in ZTC samples showed only small changes among samples. We could detect some changes in the average ratio of the integrated intensity of the D (disordered carbon) and G (graphitic carbon) Raman bands in the ZTC samples upon implementation of the annealing procedure (Fig. 8a), but even these changes were relatively small. High resolution TEM studies did not show significant changes as well. The TEM micrographs showed rather disordered structure (even for annealed samples), typical for porous carbons (Fig. 9b). We hypothesize that in order to observe significant (and easily detectable) degree of ordering higher annealing temperatures could be required. Hoverer, temperatures higher than 1000°C result in a collapse of the zeolite structure and the reaction between C and zeolite and thus could not be undertaken.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We would like to emphasize that according to our XRD studies all the samples produced by LP CVD of C at 700°C (including those that undergone ball milling procedure) retained some level of pore ordering / alignment originating from that of the zeolite templates (Fig. 9a). The low angle peak at ∼6° in this figure corresponds to the carbon walls regularly spaced at ∼1.3 nm and originates from the ordering of the (111) planes of the ZTC templates increased by the post annealing process.⁶⁶ Similarly, it is important to note that the employed ball milling procedure was rather delicate and did not result in any noticeable damages in the shape of the particles (Fig. 9b). We noticed somewhat smoother surface and better zeolite shape retention in the ZTC samples that underwent annealing (including those that were annealed and ball milled prior to zeolite etching, Fig. 9b). This may qualitatively support our previous conclusion of the positive impact of annealing on pore volume retention upon zeolite etching.

In search of the best way to prepare ZTC samples, we wanted to explore if the combination of all three post-synthesis treatments may lead to the most promising performance. Therefore, sequential post-synthesis annealing, ball milling and finally activation procedures were performed. For all samples, ball mill grinding time was fixed at 30 min and activation condition was fixed at 15 min, 150 mL min⁻¹ of CO₂ flow and 800°C. Annealing temperature was varied from 800°C to 900°C. Surprisingly, three combination of post synthesis treatments resulted in lower SSA and pore volume (Table IV) compared with the original sample. Such results indicated that after ball milling (and presumably partial removal of the disordered carbon surface layer and fracturing of some of the larger agglomerates) the C-coated zeolite samples become significantly more sensitive to activation. In this case even moderate activation conditions evidently resulted in over-oxidation (burning) of carbon pore walls and noticeable particle collapse after zeolite etching. The smaller SSA resulted in smaller capacitance (results not shown for brevity).

Figure 10 demonstrates the virtually linear relationship between the specific capacitance and SSA (at least in case of high surface area samples). The low SSA of the C-ZTC samples activated at the highest temperature of 900°C suggest that they are mostly composed of the surviving thicker (e.g. multi-walled) graphitic structures, which are expected to exhibit higher free electron density and thus may show higher area-normalized capacitance.¹³ In addition, we cannot exclude the possibility of the intercalation of ions into the closed pores inaccessible by N₂ utilized in the SSA measurements.

Zoom In Zoom Out Reset image size

Carbon-based EDLC are known to exhibit a very stable performance in organic electrolytes if the purity of carbon is sufficiently high and if the remaining impurities do not induce unfavorable reactions with current collector or electrolyte.¹ In our case, several EDLCs built with ZTC samples underwent C-D tests. Even when tested at a high current density of 20 A g⁻¹ these EDLCs did not show significant signs of degradation after 22,000 cycles (Fig. 11).

Zoom In Zoom Out Reset image size