Synthesis of large surface area carbon xerogels for electrochemical double layer capacitors

Synthesis of large surface area carbon xerogels for electrochemical double

layer capacitors

Yun-Min Chang, Cheng-Yeou Wu, Pu-Wei Wu

*

Department of Materials Science and Engineering, National Chiao Tung University, 1001 Ta Hsueh Road, Hsin-Chu 300, Taiwan

h i g h l i g h t s

< Improved formula for resorcinoleformaldehyde condensation to produce porous carbon structure.

< Solvent exchange and carbon dioxide activation to fabricate carbon xerogels with surface area of 3419 m2_g
1_.

< Specific capacitance and life time comparable or better than Black Pearl 2000.

a r t i c l e i n f o

Article history: Received 4 June 2012 Received in revised form 23 August 2012

Accepted 14 September 2012 Available online 23 September 2012 Keywords:

Carbon xerogels

Electrochemical double layer capacitor Resorcinoleformaldehyde condensation Solvent exchange

Carbon dioxide activation

a b s t r a c t

A resorcinoleformaldehyde (ReF) condensation reaction catalyzed by acetic acid (C) is employed to prepare carbon xerogels for electrochemical double layer capacitors. The samples are fabricated with an R:F ratio of 1:2 and R:C ratio of 10:1, followed by solvent exchange, pyrolysis, and carbon dioxide acti-vation. The solvent exchange allows negligible structure contraction upon drying, and after pyrolysis and carbon dioxide treatment, we are able to produce porous carbons with a surface area of 3419 m2g
1. Image from scanning electron microscope reveals an interconnected foam-like structure and BET confirms the presence of excessive micro- and mesopores. Electrochemical analysis including cyclic voltammetry (CV), current reversal chronopotentiometry (CRC), and impedance spectroscopy are con-ducted using a titanium cavity electrode so relevant capacitive characteristics and kinetic parameters could be determined. Both CV and CRC results indicate specific capacitances and life time behaviors that are comparable or even better than those of Black Pearl 2000. For example, the specific capacitance is 324.8 F g
1in 0.5 M H2SO4aqueous solution at1 A g
1for potential window of 0e1 V. In contrast,

sample without carbon dioxide activation displays a similar morphology but its surface area and specific capacitance are reduced considerably to 449 m2g
1and 34.7 F g
1, respectively.

Ó 2012 Elsevier B.V. All rights reserved.

1. Introduction

Supercapacitors have attracted significant attention recently because they demonstrate impressive advantages in power output, energy density, and cycle life, as compared to conventional rechargeable batteries, for the applications in electric vehicles, power tools, and uninterrupted power systems[1e8]. From the standpoint of operation mechanism, there are two types of super-capacitors; electrochemical double layer capacitors (EDLCs) and pseudocapacitors. The EDLCs store charges via ions adsorption/ desorption at the interface between the electrode and electrolyte

[1e5]. In contrast, the pseudocapacitors entail facile faradaic reac-tions occurring on the electrode material to store charges in

different oxidation states[6e8]. Between them, the pseudocapa-citors are able to deliver larger capacitances but with cost premium and relatively finite cycle life. For the EDLCs, the capacitive responses are proportional to the effective surface area available for ions adsorption/desorption so an excessive surface area with a desirable porosity for the electrode material is necessary.

To date, considerable efforts have devoted on carbon-based materials for EDLCs since their rich varieties exhibit a wide range of pore sizes, surface areas, electric conductivities, and surface properties. In literature, carbonaceous materials including active carbons, carbon blacks, glassy carbons, and nanostructured carbons (carbon nanotubes, nanocapsules, nanofibers, gels) have been investigated with various results[9e17]. Among them, the carbon xerogels and aerogels demonstrate unique properties in extremely low density and high porosity [18,19]. The carbon xerogels and aerogels are synthesized via a solegel process in which precursors in liquid states are properly mixed, and condense to form

* Corresponding author. Tel.: þ886 3 5131227; fax: þ886 3 5724727. E-mail address:[email protected](P.-W. Wu).

Contents lists available atSciVerse ScienceDirect

Journal of Power Sources

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m/ l o ca t e / j p o w s o u r

0378-7753/$e see front matter Ó 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jpowsour.2012.09.066

continuous colloidal networks, followed by solvent removal and pyrolysis to obtain a porous carbon structure with interconnected channels. One of the solegel approaches that has been explored extensively is the polycondensation of resorcinol (R) and formal-dehyde (F)[18e20]. According to Elkhatat and Al-Muhtaseb[21], the concentrations for the precursors, catalysts, solvents, and pH play significant roles in determining the resulting morphologies and porosities of the gel structures. In addition, the solvent removal step for the wet gels is rather critical because the drag of surface tension induces contraction of the colloidal networks resulting in substantial reduction in the pore size. So far, many ReF derived porous carbons have been prepared and evaluated for possible applications in EDLCs, catalysis (as a catalyst support),filtration, gas separation, and adsorption[19,22_e25].

Previously, many RF-derived xerogels and aerogels have been synthesized and tested for capacitive behaviors[26,27]. The prep-aration of xerogels involves direct solvent evaporation that engenders notable structural contraction and possible collapse at the extreme case. As a result, after pyrolysis the carbon xerogels typically contain 25% porosity, and reveal a moderate surface area (150e900 m2_g
1_{) and small pores (1e10 nm)[28]. In contrast, the}

carbon aerogels experience a supercritical drying step that subli-mates the solvents with negligible shrinkage of carbon networks. After pyrolysis, the carbon aerogels demonstrate a substantially larger surface area and pore volume, and consequently, a larger capacitance[28,29]. Despite those merits, the carbon aerogels are of little commercial interest because the supercritical drying is energy-consuming and not adaptable for production in large quantity. On the other hand, the carbon xerogels suffer from unnecessary structure alteration and hence, their pore size and pore volume are not adequate for EDLC applications. An alternative preparation route to minimize structure contraction during solegel transformation is via repeated solvent exchanges that reduce the surface tension of the solvent sequentially in the wet gels[27]. In this way, the solvent with reduced surface tension is able to evaporate slowly rendering a dried gel known as “ambient gel” whose structure closely resembles to that of aerogel. This enables a large number of mesopores and macropores in the carbon skel-eton. This ambient gel represents a unique form of xerogels, and could be of potential interest because its desirable surface area and porosity, as well as simple drying process.

It is understood that desirable EDLCs require large surface areas with appropriate pore sizes for facile access of ions from the elec-trolyte during capacitive charging and discharging. This can be realized by fabricating suitable carbon skeletons followed by a deliberate activation step to remove atomic carbon selectively. In this work, we explore the preparation of RF-derived xerogels in conjunction with solvent exchange and carbon dioxide treatment for surface activation so an excessive surface area and suitable pore size can be obtained. The resulting materials are subjected to electrochemical measurements using a titanium cavity electrode (TCE) to determine their intrinsic capacitive behaviors and relevant kinetic parameters.

2. Experimental

The carbon xerogel was prepared by a condensation reaction between resorcinol and formaldehyde, followed by solvent exchange, pyrolysis, and carbon dioxide activation to render an excessive surface area. First, resorcinol (R) and 37 wt% formalde-hyde (F) were properly mixed in a 1:2 M ratio in deionized water (W) under ultrasonication for 10 min at 26 C. After complete dissolution of resorcinol, 99 wt% acetic acid (C) was added as a catalyst to promote the condensation reaction forming colloids in the mixture. Subsequently, the mixture was sealed in a container at

70C for 72 h with constant vibration to allow the transformation from a sol suspension to a wet gel. The wet gel maintained a cylindrical shape of

p

 (2.25)2 _{ 4.2 cm}3_{. After gelation, the}

sample was removed from the oven and washed for 72 h in which 50% of acetone was drained and replaced with fresh acetone every 24 h. Next, the process was repeated again for another 72 h with cyclohexane. These solvent exchange steps were conducted at 26C so the surface tension for the liquid phase was reduced sequentially to avoid structural contraction for the solid phase upon solvent evaporation. At this stage, the sample maintained its size without noticeable shrinkage. Afterward, the sample was dried at 26C for 24 h to evaporate remaining cyclohexane, leaving a free-standing dark orange gel in size of

p

 (2.05)2 _{ 4.1 cm}3_{. This}

corre-sponded to a volume shrinkage of 19%. In contrast, an as-prepared sample dried directly under ambient condition without solvent exchange reduced its size to

p

 (1.69)2_{ 3.4 cm}3_{, which amounted}

to 54% volume reduction. Finally, the sample was placed in a tube furnace for two-stage heat treatments. The first stage was per-formed in argon atmosphere (99.99%) at 950C for 1 h to decom-pose organic residues (pyrolysis reaction) where the dark orange sample shrunk significantly becoming a black porous carbon. The heating profile included 10_{C min}
1_{for 25e600}_{C, 6}_{C min}
1_for

600e900_{C, and 1}_{C min}
1_{for 900e950}_{C. In the second stage, at}

950C the argon was purged quickly via vacuum pump (1 min) and 99.99% carbon dioxide was introduced for 3 h to activate the surface of the porous carbon. Afterwards, the carbon dioxide was again vacuum-pumped and the tube furnace was re_{filled with argon to} allow slow cooling to room temperature for roughly 12 h. By adjusting precursor and catalyst concentration, we were able to prepare xerogels with various porosities and surface areas. The resulting samples were designated as GA, GB, and GC, respectively. The exact amounts for the precursors, catalyst, deionized water, as well as ratios for the R/C and R/W are listed inTable 1. In order to distinguish the effect of carbon dioxide activation, we also carried out identical processing steps except the carbon dioxide activation was replaced with argon treatment for another 3 h (labeled as GD as a reference sample).

The morphologies for the xerogels after pyrolysis and carbon dioxide activation were observed by an SEM (JEOL JSM-6500F). Surface area was determined by Quantachrome NOVA-1000A where a density functional theory (DFT) program was employed to estimate relevant parameters in pore size and pore distribution. The DFT is a mathematic approach to describe gas adsorption behaviors over a wide range of pressures and temperatures after taking into account of population density, energy, and equilibrium between the adsorbing molecules and substrate[30].

For electrochemical analysis to determine the double layer capacitance, a three-electrode cell was used in which a J-shaped TCE, with 2 mm in diameter and 200 mm in length, was adopted as the working electrode. The bottom of the TCE contained a cylin-drical cavity with 1 mm in diameter and 1 mm in depth for sample

Table 1

Precursors and catalyst used for the synthesis of carbon xerogels.

GA GB GC GD

Precursors R: resorcinol (g) 9.25 14.75 14.75 14.75 F: formaldehyde (g) 5.05 8.04 8.04 8.04 Catalyst C: acetic acid (mL) 965 1533 766.5 766.5 Solvent W: deionized water (g) 50 50 50 50

R:F 1:2 1:2 1:2 1:2

Molar ratio R:C 5:1 5:1 10:1 10:1

R:W 1:5.4 1:3.4 1:3.4 1:3.4

Burn-offa_{(wt%) 95.3 95.3 96.4 88.3} a_{Total weight loss of carbon xerogels after 950}_{C pyrolysis and carbon dioxide} activation.

loading. The sample was placed on a glass sheet and pressed into fine powders so they filled compactly into the bottom cavity. The exact weight was measured by a microbalance with an accuracy of 10

m

g, and in our cases, the sample weights were between 2 and 3 mg. Details on the TCE and its testing results on commercially available carbon blacks have been reported earlier [31]. Pt foil (2 5 cm2_{) and Ag/AgCl were used as the counter and reference}

electrodes, respectively. Values for the double layer capacitance were determined via cyclic voltammetry (CV) and current reversal chronopotentiometry (CRC) in 0.5 M H2SO4 aqueous solution at

26C using a Solartron SI1287 potentiostat and 1255B frequency response analyzer. The CV profiles were recorded at scan rates of 5, 10, 20, and 50 mV s
1, in potential windows of 0e0.6, 0e0.8, and 0e 1 V, respectively. The CRC measurements were carried out at 1 A g
1_{in a potential window of 0e1 V. In life time evaluation, the}

sample was subjected to repeated CRC scans at 1 A g
1 in a potential window of 0e1 V. Impedance spectra were obtained at the open circuit voltage with a sinusoidal signal of 10 mV for frequencies between 0.1 and 20 kHz. The resulting spectra were fitted by a ZView software to retrieve relevant electrochemical parameters.

3. Results and discussion

Previously, in the preparation of ReF-derived carbon aerogels, the R/C ratios were kept between 50 and 1500, and a basic catalyst such as sodium carbonate was used [21]. It is because a high R/C ratio enables the formation of large colloidal particles with limited contacts among them, and a base catalyst promotes a wider pore size distributions[20,21]. However, in their formula the concentration for the catalyst is extremely low and it becomes rather dif_{ficult to initiate condensation reaction homogeneously.} Therefore, we rationalize that a lower R/C ratio of 5 or 10 might be beneficial because the concentrated precursors engender an inti-mate structure with relatively narrower pore size distribution. In

addition, we select an acidic catalyst because Al-Muhtaseb et al. and Elkhatat et al. observed that in an acidic solution, the condensation reaction took less time to complete andfiner inter-connected mesopores were produced afterwards[20,21].

Representative SEM images for GA, GB, GC, and GD are provided in Fig. 1. Sample GA (shown inFig. 1(a) and its inset) revealed uniform spherical carbons with diameter in 7e10

m

m. These spherical carbons were contacting each other in a string-of-pearls pattern. This morphology was expected because the sample was synthesized with relatively lower precursor concentrations so the condensation reaction took place slowly forming larger primary particles. Similar behaviors were observed by Scherdei et al. where they observed the formation of spherical particles at lower precursor concentrations[32]. In sample GB exhibited inFig. 1(b), once the precursor concentrations were increased (while the R/C ratio was kept at 5:1), we observed many small irregularly-shaped carbon particles (50e500 nm in size) connecting to each other. It is realized that during solegel transformation, a higher precursor concentration often leads to a faster condensation rate and stronger cross-linking among primary particles. As a result, sample GB demonstrated a foam-like microstructure with mostly smaller primary carbon particles as compared to those of sample GA.

Fig. 1(c) displays the morphology for sample GC, where a reduced catalyst amount was used while the precursor concentrations were identical to those of sample GB. Similar to that ofFig. 1(b), there appeared a foam-like microstructure with primary carbon particles in 50_{e100 nm size. However, at a higher magnification, sample GC} revealed larger internal pores and even moderate sintering among primary carbon particles. We understood that due to a reduced catalyst loading, there were fewer nucleation sites for the solegel transformation so the resulting cross-linking became less pronounced. Morphology for sample GD is shown inFig. 1(d). Since the only difference between sample GC and GD was the treatment for carbon dioxide activation, we realized that their morphologies to be quite similar as the carbon dioxide etching effect was

expected to produce micropores on the carbon surface and those micropores were not readily discernible in SEM observations.

These carbon xerogels are known to possess a variety of pores classified as microspores (<2 nm), mesopores (2e50 nm), and macrospores (>50 nm). Their relative amounts could be estimated by nitrogen adsorption and desorption isotherms, shown inFig. 2. Interestingly, sample GA, GB, and GC demonstrated a type I behavior, indicating the predominant presence of micropores. Relevant pore properties including BET surface area (SBET),

micro-pore surface area (SMicro), external surface area (SExt), average pore

diameter, micropore volume (VMicro), and total pore volume (VPore)

are provided inTable 2. Despite their notable difference in SEM morphology, sample GA and GB exhibited micropore volumes that were rather close, implying that the R/C ratio and carbon dioxide activation were directly responsible for micropore formation. Remarkably, at identical R/C ratio of 10, sample GC revealed four times larger micropore volume over that of sample GD. Moreover, the external surface area for sample GC was increased considerably as well. It is noted that the external surface area represents the sum of macropores and mesopores. FromFig. 1, the morphologies for sample GC and sample GD were quite similar, and hence their macropore volumes were likely to be close. Therefore, the recorded large external surface area for sample GC was attributed to the formation of abundant mesopores and micropores during carbon dioxide activation. This can be confirmed by the burn-off ratio lis-ted inTable 1. With carbon dioxide activation, the samples suffered from weight loss of 95_{e96%. In contrast, without carbon dioxide} activation, sample GD retained 11.7% weight after pyrolysis. Our experimental results provide definitive evidences for carbon dioxide activation that effectively etches the carbon surface, resulting in the presence of excess micropores and mesopores. The surface area for sample GC was estimated at 3419 m2g
1, a value that is significantly larger than typical high surface area carbons like activated carbons or Black Pearl 2000 (BP2000).

In previous studies of RF-derived samples, the largest surface area reported was 3125 m2g
1from a carbon aerogel[33]. In our carbon xerogels we were able to exceed that value by combining improved formula, solvent exchange, and carbon dioxide activation simultaneously. It is noted that by employing a basic catalyst, solvent exchange, and carbon dioxide activation (5% CO2e95% N2),

Lin and Ritter obtained carbon xerogels with surface areas &1600 m2_g
1_{[34]. In contrast, our sample of 3125 m}2 _g
1_was

catalyzed by acetic acid and the carbon dioxide activation was conducted in 99.99% CO2atmosphere. Despite the acidic catalyst is

known to promote condensation reaction in RF transformation, the predominant catalyst reported in literature is basic one such as sodium carbonate. To our knowledge, the acetic acid was used once by Brandt et al. to prepare carbon aerogels free of metallic impu-rities as opposed to those using sodium carbonate[35]. They re-ported that the pore and particle size can be adjusted to a wide range contingent on the ratios of precursors to acetic acid, and the connectivity of primary particles can be exceptionally high so afine structure can be obtained. In our case, we surmised that a syner-gistic effect was occurring between our formula and processing conditions rendering a desirable porous structure which was susceptible to carbon dioxide etching. This is reflected by the intense weight loss after pyrolysis and carbon dioxide activation (88e96%) as compared to that of conventional RF-derived xerogel (50e60%). As a result, our samples revealed excessive surface areas. In general, the capacitance value for an EDLC is contingent on the electrode fabrication process which affects the resulting composition and structure, as well as measurement techniques involved. Conventional porous composite electrodes and thinfilm electrodes require certain amount of active materials to mix with binders and conductive additives. Hence, the determination of intrinsic capacitive responses for a potential EDLC material becomes a challenge. Since both binder and conductive additive are not used in TCE analysis, and only a minute amount of active material is necessary, the intrinsic capacitive behaviors can be accurately obtained as established by our earlier work[31].Fig. 3

0.0

0.2

0.4

0.6

0.8

1.0

0

200

400

600

800

1000

1200

Vo

lu

m

e ad

so

rb

ed

(cm

3

g

-1

)

Relative pressure

(PP

0 -1

)

Fig. 2. N2adsorption/desorption isotherms of sample GA (,), GB (C), GC (q), and GD (╳).

Table 2

Relevant parameters for the pore properties determined by nitrogen adsorption and desorption isotherms from GA, GB, GC, and GD, respectively.

GA GB GC GD VPore(cm3g
1) 0.99 0.86 1.36 0.23 VMicro(cm3g
1) 0.42 0.40 0.76 0.18 SBETa(m2g
1) 2293.9 2016.6 3418.6 449.1 SMicro(m2g
1) 1009 928.8 1987.4 338.4 SExtb(m2g
1) 1284.9 1087.7 1431.2 110.7 Dp(Å) 13.8 12.2 18.5 11.1 SExtratioc(%) 56 53.9 40.9 24.7 a_S

BET¼ SMicroþ SExt. b _S

Extis the surface area including mesopores and macropores. c _S

Extratio¼ SExt/SBET.

0.0

0.2

0.4

0.6

0.8

1.0

-8

-6

-4

-2

0

2

4

6

8

(d)

(c)

(b)

(a)

C

u

rren

t d

e

n

s

ity

(A

g

-1

)

Potential (V)

Fig. 3. CV profiles at 20 mV s
1_{from sample GC in potential windows of (a) 0e0.6 V,} (b) 0e0.8 V, (c) 0e1 V, and (d) empty TCE of 0e1 V, respectively.

demonstrates representative CV profiles for sample GC at potential windows of 0e0.6, 0e0.8, and 0e1 V, respectively, along with the response from an empty TCE. As shown, the TCE itself revealed negligible currents in both forward and backward scans. In contrast, the carbon xerogels exhibited quasi-rectangular responses typical of EDLCs[36]. At voltage approaching 1 V, there appeared moderate electrolysis which engendered an apparent current rise. Notably, the presence of functional groups on the carbon xerogels was rather subdued as additional redox peaks were not observed in 0.2e0.6 V. It is known that the specific capacitance could be estimated from the CV profiles by the equation listed below,

Cs ¼ _wQ

_D

_E ¼

Z icdt

w

D

E [1]

where CS, Q, w,

D

E, and ic are the specific capacitance (F g
1),

coulombs of discharge, sample weight, CV potential window, and cathodic current density, respectively. Table 3 lists the specific capacitance for the carbon xerogels at various potential windows along with commercially available carbon blacks such as Vulcan XC72R (XC72R), BP2000, and Active Carbon (AC1100). The specific capacitances for those commercial carbon blacks were determined by a TCE in identical testing conditions, and their values were re-ported in our previous work[31].

Among commercial carbon blacks, the BP2000 is known to possess a large surface area due to its fine particle size and consequently, its capacitance is considerably larger than that of XC72R (130.3 vs. 25.7 F g
1). Remarkably, the carbon xerogels of GB and GC demonstrated even larger specific capacitances, reaching values of 150.4 and 209.9 F g
1, respectively. These values, to our knowledge, are comparable to the capacitances derived from other large surface area nanostructured carbons reported in literature

[37e41]. For example, Weng et al. fabricated active carbons with a surface area of 2860 m2 g
1 and obtained a capacitance of 130 F g
1[37]. Similar results were reported by Rufford et al. where active carbons of 1788 and 2019 m2g
1were synthesized and their capacitances were recorded at 300 and 368 F g
1, respectively

[38,39]. From Table 1, sample GD was produced with identical formulation like sample GC but the carbon dioxide activation step was replaced with argon heat treatment. Without the carbon dioxide etching effect, sample GD revealed a substantially reduced micropore volume and as a result, its specific capacitance was merely 38.3 F g
1.

It is understood that the CV profiles for a double layer capacitor are affected by the scan rate imposed during CV measurements. The relation for a capacitive current from a double layer capacitor is listed below, ic ¼ vCd  1
exp 
t RsCd  [2]

where ic is defined previously, v is the scan rate, Rs represents the electrolyte resistance, t is the time, and Cdis the capacitance (F g
1).

Hence, the current rises quickly and reaches a plateau in a short

time for a true capacitor. In addition, an increasing scan rate engenders a larger current plateau.Fig. 4displays representative CV profiles for sample GC at various scan rates in a potential window of 0e1 V. The CV profiles demonstrated rectangular shapes at scan rates of 5 and 10 mV s
1, which were expected for a typical double layer capacitor[42]. Apparently, once the scan rate was increased to 20 mV s
1and above, the resulting CV curves became moderately distorted. This distortion was attributed to the compromised diffusion within the porous structure that rendered a progressively slower capacitance upon faster scan rate[31]. The specific capaci-tance is also a function of CV scans since at a sufficiently fast scan rate, some micropores become inaccessible due to diffusion limi-tation, and thus cease to contribute to the capacitive current. Hence, the largest speci_{fic capacitance is always recorded at the} lowest scan rate where ions adsorption and desorption are possible in available free surface and internal pores.Table 4provides the specific capacitance at different scan rates for the carbon xerogels and commercial carbon blacks. For the carbon xerogels except sample GA, they exhibited capacitance retention characteristics similar to those of XC72, BP2000, and AC1100. We can thus reasonably assumed that the carbon xerogels contained pores that were properly sized so access by ions was compromised moder-ately at fast scan rate.

Fig. 5provides the CRC responses at_{1 A g}
1for the carbon xerogels. It is established that the specific capacitance can be esti-mated by following equation,

Cs ¼ ic ðw  jdEjÞ dt z Q w

D

E [3]

where CS, w, ic, and Q are defined previously, and dE/dT is the slope

for discharging curve. In addition, during current reversal, there is a sudden voltage drop (iR loss) whose magnitude is proportional to the electrical resistance of carbon xerogels, TCE, and electrolyte. Values for the specific capacitance and iR loss are listed inTable 5. As listed, sample GD demonstrated the lowest voltage drop, sug-gesting its smallest electrical resistance. This might be attributed to its intimate interconnected structure and surface integrity. On the other hand, sample GA revealed the largest voltage drop which was likely caused by the finite contact areas among primary carbon particles that circumvented electron transports (see Fig. 1(a)). In

Table 3

Values of specific capacitance (F g
1_{) calculated from CV profiles in various potential} windows at 20 mV s
1scan rate.

GA GB GC GD XC72a _BP2000a _AC1100a

0e0.6 V 55.8 98 168.4 31 16.7 78 75

0e0.8 V 69.3 115.2 183.4 33 18.7 94 80.8 0e1 V 84.5 150.4 209.9 38.3 25.7 130.3 88 a_{These values were reported in reference[31].}

0.0

0.2

0.4

0.6

0.8

1.0

-12

-9

-6

-3

0

3

6

9

12

Cur

re

nt

densi

ty

(A

g

-1

)

Potential (V)

Fig. 4. CV profiles for sample GC in potential window of 0e1 V at scan rates of 5, 10, 20, and 50 mV s
1, respectively.

addition, all carbon xerogels except GD demonstrated specific capacitances larger than 200 F g
1, indicating that accessible surface area, as well as suitable microspores and mesopores structure were produced after carbon dioxide activation.

It is noted that the CRC method always leads to a higher specific capacitance as opposed to that of CV scans. It is because the CRC method is conducted at afixed potential window with a constant current density and therefore, the process allows sufficient time for ion adsorption and desorption. In contrast, the capacitance deter-mined from the CV scans is caused by a varying current so the kinetics for ion diffusion in the porous structure becomes a limiting factor. In principle, the capacitance obtained at slow scan rate during CV scans is likely to approach that of CRC method. In our measurements, indeed, the capacitance from the CRC method at 1 A g
1was rather close to the capacitance from the CV pro_{file with} a scan rate of 5 mV s
1.

To evaluate life time performance for the carbon xerogels, we carried out repeated CRC experiments between 0 and 1 V at 1 A g
1_{. The resulting capacitance variation over 3000 cycles are}

shown inFig. 6, along with that of BP2000 for comparison purpose. In life time testing, it was found that the capacitances became smaller with increasing cycles, a fact that was possibly associated with detachment of individual carbon particles upon prolonged cycling. We surmised that without binder, electrolyte inundation in the TCE cavity inevitably affected packing density and local envi-ronments of the carbon particles. As a result, the amount of carbon particles responsible for capacitive responses was decreasing with time. This undesirable effect was most severe in sample GA (with a capacity loss of 63%) since from SEM image there was limited contact between carbon particles so poor electrical conductivity and packing density were not unexpected. In contrast, sample GB revealed impressive capacitance retention with a moderate

capacity loss of 13%. In contrast, sample GC showed a capacitance loss of 27% despite its iR loss was the least fromTable 5. Never-theless, its value was still significantly better than 41% of BP2000. From Fig. 6, we concluded that except sample GA, the carbon xerogels retained reasonable EDLCs behaviors comparable or even better than that of BP2000.

Impedance analysis is a powerful technique to examine the interfacial phenomenon taking place on the carbon xerogels during capacitive responses. Our impedance spectra were ob-tained at the open circuit voltage and their values were 0.4119, 0.3327, and 0.3458 V (vs. Ag/AgCl) for GA, GB, and GC, respec-tively. Fig. 7(a) demonstrates the Nyquist plots over the entire frequency regime. Apparently, the impedance spectra were con-sisted of a semi-circle at high frequency and a Warburg diffusion at low frequency. The enlarged spectra at high frequency regime is shown inFig. 7(b) and the equilivent circuit model used tofit the impedance spectra is provided inFig. 7(c). As shown, the equi-livent circuit includes elements of Rs (ohmic resistance from

electrolyte, carbon particles, and TCE), RCT(resistance from

fara-daic charge transfer reaction), CPEp(constant phase element for

faradaic charge transfer reaction), W (Warburg impedance), and CPEEDL(constant phase element for electrochemical double layer).

For both CPEp and CPEEDL, they can be derived from following

equation[43]:

Z ¼ I

Tðj

u

Þa [4]

where Z, T, j,

u

, and

a

are impedance, capacitance of associated element, imaginary unit, angular frequency (

u

¼ 2

p

f) of the AC signal, and a dimensionless parameter for_{fitting purpose (}

a

_{¼ 0 for} a pure resistor and

a

¼ 1 for an ideal parallel plate capacitor). In

0

100

200

300

400

500

600

700

800

0.0

0.2

0.4

0.6

0.8

1.0

P

o

te

n

tia

l (V

)

Time (sec)

A

B

C

D

Fig. 5. CRC curves at1 A g
1_{in potential window of 0e1 V for sample GA, GB, GC, and} GD, respectively.

Table 5

Relevant parameters determined from the CRC measurements at1 A g
1_in potential window of 0e1 V.

GA GB GC GD Q (C) 236.3 269.2 418.6 39.60 Eþ(V) 1.00 1.00 1.00 1.00 E
(V) 0.736 0.869 0.935 0.944 iR drop (mV) 264 131 65 56 CS(F g
1) 213.2 229.4 324.8 34.7

0

500

1000

1500

2000

2500

3000

0.0

0.2

0.4

0.6

0.8

1.0

Cycles

N

o

rm

al

iz

ed capaci

ty

(

%

)

Fig. 6. Variation of specific capacitance for sample GA(,), GB(C), GC(q), GD(╳), and BP2000 (:) obtained from CRC measurements in potential window of 0e1 V at 1 A g
1_{for 3000 cycles.}

Table 4

Values of specific capacitance (F g
1_{) calculated from CV profiles at various scan} rates in potential window of 0e1 V.

mV s
1 GA GB GC GD XC72a _BP2000a _AC1100a 5 (a) 193.7 214.9 312.3 56.9 32.03 169.76 128.48 10 132.4 183.1 266.2 47.2 29.52 152.97 111.19 20 84.5 150.4 209.9 38.3 25.68 130.32 88 50 (b) 38.6 122.6 132.6 27.3 23.28 95.51 51.28 (b/a) (%) 19.9 57.1 42.5 48.8 72.68 56.26 39.91 a_{These values were reported in reference[31].}

practice, the

a

is between 0 and 1, and its exact value reflects the porous nature of the active material[44].

Table 6presentsfitting results with relevant parameters clearly identified. As listed, the Rsvalues were rather subdued confirming

the conductive nature for the carbon xerogels. The CPEpeT, WeT,

and CPEEDLeT indicate the capacitive components for CPEp,

War-burg impedance, and CPEEDL, respectively. Among these samples,

the

a

values for CPEp were consistently below 1, which was

expected for a porous electrode. In addition, the

a

values of War-burg impedance for sample GB and GC were under 1, a fact attributed to the limited diffusion in porous carbons. For sample GA, its

a

-value of Warburg impedance was undetermined because the Warburg diffusion occurring at the low frequency regime was not clearly defined. It is possible that a lower frequency AC signal is necessary to better distinguish the diffusion effect. As for the CPEEDL, interestingly, sample GC revealed a

a

-value of 1, suggesting

its response was close to an ideal capacitor.

4. Conclusions

We synthesized carbon xerogels via a condensation reaction of resorcinol and formaldehyde, followed by solvent exchange, pyrolysis, and carbon dioxide activation. The carbon xerogels experienced negligible structural contraction during solvent removal and as a result, demonstrated excessive surface area with a large number of micropores and mesopores. By selecting suitable combination of precursors and acidic catalyst, we were able to produce a porous carbon structure whose surface area reached as high as 3418 m2 g
1. Image from SEM confirmed its foam-like morphology with moderate sintering among primary carbon particles. CV scans and CRC methods indicated specific capacitances and life time behaviors that were comparable or even better than those of BP2000 in identical testing conditions. In contrast, without carbon dioxide activation, the carbon xerogels displayed a similar morphology but its surface area was reduced considerably to 449 m2g
1. Due to lack of sufficient micropores and mesopores, the resulting speci_{fic capacitance and life time behavior were} substantially inferior to samples with carbon dioxide activation.

Acknowledgements

Financial supports from National Science Council (NSC100-2221-E009-075-MY3) and Taiwan Power Company are greatly appreciated.

References

[1] B.E. Conway, Electrochemical Supercapacitors, KluwerePlenum Press, New York, 1999.

[2] P. Kossyrev, J. Power Sources 201 (2012) 347e352.

[3] M.C. Liu, L.B. Kong, P. Zhang, Y.C. Luo, L. Kang, Electrochim. Acta 60 (2012) 443e448.

[4] M. Lazzari, M. Mastragostino, A.G. Pandolfo, V. Ruiz, F. Soavi, J. Electrochem. Soc. 158 (2011) A22eA25.

[5] M. Winter, R.J. Brodd, Chem. Rev. 104 (2004) 4245e4269.

[6] C.J. Hung, J.H. Hung, P. Lin, T.Y. Tseng, J. Electrochem. Soc. 158 (2011) A942eA947.

[7] K.H. Chang, C.C. Hu, C.M. Huang, Y.L. Liu, C.I. Chang, J. Power Sources 196 (2011) 2387e2392.

[8] C.C. Hu, K.H. Chang, M.C. Lin, Y.T. Wu, Nano Lett. 6 (2006) 2690e2695. [9] Q. Li, F. Liu, L. Zhang, B.J. Nelson, S. Zhang, C. Ma, X. Tao, J. Cheng, X. Zhang,

J. Power Sources 207 (2012) 199e204.

[10] J. Ahmed, H.J. Kim, S. Kim, J. Electrochem. Soc. 159 (2012) B497eB501. [11] B. Haghighi, M.A. Tabrizi, Electrochim. Acta 56 (2011) 10101e10106. [12] I.G. Casella, D.A. Di Fonzo, Electrochim. Acta 56 (2011) 7536e7540. [13] E.L.K. Chng, M. Pumera, Electrochem. Commun. 13 (2011) 781e784. [14] Y.M. Chang, Y.C. Hsieh, P.W. Wu, Diamond Relat. Mater. 18 (2009) 501e504. [15] M.M. Hasani-Sadrabadi, I. Shabani, M. Soleimani, H. Moaddel, J. Power Sources

196 (2011) 4599e4603.

[16] L. Zou, R. Lv, F. Kang, L. Gan, W. Shen, J. Power Sources 184 (2008) 566e569. [17] J. Rooke, C.d.M. Passos, M. Chatenet, R. Sescousse, T. Budtova, S. Berthon-Fabry, R. Mosdale, F. Maillard, J. Electrochem. Soc. 158 (2011) B779eB789. [18] R.W. Pekala, US patent No. 4873218 (1989).

[19] B. Liu, S. Creager, J. Power Sources 195 (2010) 1812e1820. [20] S.A. Al-Muhtaseb, J.A. Ritter, Adv. Mater. 15 (2003) 101e114. [21] A.M. Elkhatat, S.A. Al-Muhtaseb, Adv. Mater. 23 (2011) 2887e2903. [22] C. Arbizzani, S. Beninati, E. Manferrari, F. Soavi, M. Mastragostino, J. Power

Sources 161 (2006) 826e830.

[23] A.K. Meena, G.K. Mishra, P.K. Rai, C. Rajagopal, P.N. Nagar, J. Hazard. Mater. 122 (2005) 161e170.

[24] T. Yamamoto, A. Endo, T. Ohmori, M. Nakaiwa, Carbon 42 (2004) 1671e1676.

0.00 0.05 0.10 0.15 0.20 0.00 -0.02 -0.04 -0.06 -0.08 -0.10 A B C Fitting curves

Z

'' (

c

m

2

)

Z' (

Ω

Ω

Ω

cm

2

)

Z' (

_Ω

cm

2

)

0.00 0.01 0.02 0.03 0.04 0.05 0.000 -0.005 -0.010 -0.015 -0.020 -0.025

Z'

' (

c

m

2

)

a

b

c

Fig. 7. (a) Nyquist plots for sample GA, GB, and GC at the open circuit voltage in frequency range of 0.1e20 kHz, as well as (b) the enlarged spectra at high frequency regime. (c) The equivalent circuit model used tofit the impedance spectra.

Table 6

Parameters fromfitting impedance spectra obtained at the open circuit voltage.

GA GB GC Rs(Ug
1) 0.00115 0.00181 0.00151 CPEp T (F g
1) 1.113 0.228 0.281 a 0.553 0.707 0.642 RCT(Ug
1) 0.0954 0.0339 0.0215 W R (Ug
1) e 0.0019204 0.00462 T (F g
1) e 0.69332 0.839 a e 0.741 0.417 CPEEDL T (F g
1) 77.230 117.300 130.300 a 0.428 0.552 1.000

[25] H.Y. Tian, C.E. Buckley, M. Paskevicius, D.A. Sheppard, S.B. Wang, C.J. Webb, E.M. Gray, Int. J. Hydrogen Energy 36 (2011) 10855e10860.

[26] M. Inagaki, H. Konno, O. Tanaike, J. Power Sources 195 (2010) 7880e7903. [27] M. Lazzari, F. Soavi, M. Mastragostino, J. Power Sources 178 (2008) 490e496. [28] H.T. Jespersen, K. Allermann, I. Schneider, K. Schaumburg, US patent No.

20100269731 (2010).

[29] B. Fang, L. Binder, J. Power Sources 163 (2006) 616e622.

[30] P.A. Webb, C. Orr, Analytic Methods in Fine Particle Technology, Micromeritics Instrument, Georgia, 1997.

[31] C.Y. Wu, P.W. Wu, P. Lin, J. Power Sources 195 (2010) 5122e5129. [32] R. Brandt, J. Fricke, J. Non-Cryst. Solids 350 (2004) 131e135.

[33] T.F. Baumann, M.A. Worsley, T.Y.J. Han, J.H. Satcher Jr., J. Non-Cryst. Solids 354 (2008) 3513e3515.

[34] C. Lin, J.A. Ritter, Carbon 38 (2000) 849e861.

[35] R. Brandt, R. Petricevic, H. Pröbstle, J. Fricke, J. Porous Mater. 10 (2004) 171e178.

[36] B.E. Conway, W.G. Pell, J. Power Sources 105 (2002) 169e181. [37] T.C. Weng, H. Teng, J. Electrochem. Soc. 148 (2001) A368eA373.

[38] T.E. Rufford, D. Hulicova-Jurcakova, K. Khosla, Z. Zhu, G.Q. Lu, J. Power Sources 195 (2010) 912e918.

[39] T.E. Rufford, D. Hulicova-Jurcakova, Z. Zhu, G.Q. Lu, Electrochem. Commun. 10 (2008) 1594e1597.

[40] Y. Zhu, H. Hu, W. Li, X. Zhang, Carbon 45 (2007) 160e165.

[41] J.S. Ye, X. Liu, H.F. Cui, W.D. Zhang, F.S. Sheu, T.M. Lim, Electrochem. Commun. 7 (2005) 249e255.

[42] J. Wang, M. Chen, C. Wang, J. Wang, J. Zheng, J. Power Sources 196 (2011) 550e558.

[43] X. Liu, L. Juan, L. Zhan, L. Tang, Y. Wang, W. Qiao, J. Electroanal. Chem. 642 (2010) 75e81.

[44] H.K. Song, H.Y. Hwang, K.H. Lee, L.H. Dao, Electrochim. Acta 45 (2000) 2241e2257.