14364
J. Phys. Chem. B 2008, 112, 14364–14372
Surface Morphology of Nanostructured Polymer-Based Activated Carbons† Youxin Yuan,* Israel Cabasso, and Han Liu‡ The Michael Szwarc Polymer Research Institute, Department of Chemistry, State UniVersity of New York - esf, Syracuse, New York 13210 ReceiVed: January 25, 2008; ReVised Manuscript ReceiVed: April 4, 2008
Complementary techniques, including nitrogen adsorption, small-angle X-ray scattering (SAXS), and atomic force microscopy (AFM), have been utilized to characterize the surface features of highly microporous carbon materials prepared from highly aromatic polymers. Nitrogen adsorption measurement interpreted by BET, DR, HK, and NLDFT methods reveals these nanostructured activated carbons exhibit a high surface area of up to 4000 m2/g, a micropore volume up to ∼1.75 mL/g, and an average pore size of ∼10-20 Å. A modified equation, based on Porod’s law, the Debye-Bueche equation, and fractal dimension theories, has been proposed and successfully applied to analyze the SAXS spectra and to extract the porous texture of these unique activated carbons. AFM 3D imaging combined with the Fourier transform technique has been applied to statistically quantify pore sizes on the carbon surface. Introduction Polymer-based carbon is prepared by pyrolyzing a polymer precursor in an inert atmosphere above its decomposition temperature. The formation of carbon, during the thermal decomposition, can be considered as a competition between a chain scission that generates volatile species and a condensation process that forms carbon structure. While many organic polymers can be thermally decomposed to yield polymer-based carbon, our interest is especially concentrated on aromatic polymers that generally produce high carbon yield, since the hexagonal benzene ring easily condenses into a polyhexagonal carbon layer. However, even for some aromatic polymers (e.g., polystyrene), chain scission can occur forming a volatile product before condensation takes place. Chemical or thermal cross-linking of a polymer matrix that produces a highly cross-linked network can effectively retain the radical species that are formed upon decomposition, until they condense into carbon. Such condensed carbon polyaromatic structures are highly dependent upon the parent polymer, and while the term polymer-based carbon is used, significant variation in the structure, composition, and properties can be found for each carbon.1–4 Several multiaromatic polymers have been examined by us for the comparative study of polymerbased carbon as related to its porosity and morphological properties, by different methods. From these, poly(2,6-dimethyl1,4-phenylene oxide) (PPO) has proven to be an excellent candidate. This polymer is rich in benzene rings, located in the main chain, that fuse upon thermal cross-linking and decomposition, yielding more than 50% carbon. Thus, from the same PPO-carbon matrix, several specimens from low to high surface area were prepared in order to study the properties of carbon nanostructures by three complementary methods: gas adsorption, small angle X-scattering (SAXS), and atomic force microscopy (AFM). The gas sorption method for characterization of porous carbon nanostructures employs nitrogen, and to a lesser extent CO2 and † Part of the “Janos H. Fendler Memorial Issue”. * To whom correspondence should be addressed. E-mail:
[email protected]. ‡ Current address: Giner Electrochemical Systems, LLC, 89 Ramford Avenue, Newton MA 02466.
helium, as probing molecules. On the basis of various gas/solid (absorbate/absorbent) interaction theories, such as BrunauerEmmett-Teller(BET),5 Dubinin-Radushkevich(DR),6 HorvathKawazoe (HK)7 and the thickness (t) method,8 the sorption data can be reduced to derive the porous texture. The molecular shape of the probing gas, the accessibility of pores by probing gas, the assumptions in these gas/solid interaction treatments, and other factors place the boundary constraints within which surface morphology is assessed. Small angle X-ray scattering (SAXS, corresponding to 2θ < 5 ° with Cu KR) has been widely used to study microstructures of a variety of inorganic and organic materials including carbon and coals,9,10 providing structural information unattainable by gas sorption analysis. In particular, it is effective for characterizing irregular nanostructural dimensions ranging from 1 to ∼100 nm, such as microphases, voids, and surfaces encased within materials. Although these two complementary techniques have been used in characterizing porous textures of low surface area carbons (BET surface area 0.2). Extending the activation time to 1 and 2 h (C1 and C2) facilitates the development of micropores at the expense of meso- and macropores. At a longer activation time (4 h), larger pores redeveloped, while micropore volume decreased. The surface area can be determined from the N2 isotherm by the BET method. Since the BET equation was derived based on surface coverage mechanism, the linear curve fitting is usually performed in the low relative pressure region (i.e., P/P0 < 0.3).29 All the activated PPO carbons possess a large BET surface area SBET (Table 1). The surface area increases with activation time from 530 m2/g of nonactivated carbon (C0) to ∼4000 m2/g of carbon (C2) activated for 2 h. The surface area decreases (C4 SBET ∼ 2900 m2/g), with further increase in activation time, which can be the result of fusion from pores. The surface area of C2 exceeds 2630 m2/g, which is the upper limit of the surface area of activated carbon on the basis of the
14366 J. Phys. Chem. B, Vol. 112, No. 46, 2008
Yuan et al.
[( ) ( ) ( ) (
( )
NsAsf + NfAff 1 d 9 1 d 3 P ) 0 3 σsf P κΤd3(L - 2σsf) 9 σsf
ln
9 1 d d 1 + 9 L - σsf 3 L - σsf
Figure 2. DR plots of PPO carbons. Symbols are experimental data points, and lines are fitting results.
single infinite graphene sheet structure.30 Presumably, edge termination makes up a substantial fraction of the surface area,31 although this cannot be interpreted from nitrogen adsorption measurements. Therefore, the highly microporous system, a complex pore filling mechanism, rather than the surface coverage mechanism assumed by the BET model, should be considered. Such high surface area has also been reported for carbon prepared in the nanochannels of zeolite Y (SBET ∼ 4100 m2/g)32 and for carbon aerogels (SBET ∼ 3200 m2/g).31 Yaghi et al. recently calculated that exposing the latent edges of the six-membered rings leads to significant enhancement of specific surface area to 5683 m2/g.33 The Dubinin-Radushkevich (DR) method,6 Horvath-Kawazoe (HK) method,7 and the thickness (t) technique8 are commonly used for reducing adsorption data to generate porous texture information. The DR method is based on the pore filling mechanism expressed as
( P°P )
log V ) log Vmi - k log2
)] 3
(2)
where NsAsf and NfAff represent the nitrogen/carbon and nitrogen/ nitrogen interaction energies, respectively; L is the distance between carbon atoms of opposite walls; σsf is the LennardJones effective molecular diameter for the nitrogen-carbon potential; and d is the distance of the N2 molecule to the carbon surface at the zero potential point. Figure 3 shows the impact of activation time on pore size distribution (according to eq 2). Prior to activation, the carbon contains a small amount of micropores with pore diameter e7 Å and micropore volume Ve7Å ∼ 0.17 mL/g (Table 2). When activated for 0.5 h, the pore size distribution greatly broadened, to the extent that a substantial amount of large micropores and even mesopores are formed. Longer activation time (up to 2 h) narrows pore size distribution and considerably increases Ve7Å ∼ 1.05 mL/g. Upon further activation to 4 h, Ve7Å decreases to 0.59 mL/g, while pore size distribution begins to widen again. This suggests that excessively long activation time facilitates the formation of large pores at the expense of micropores with a pore diameter of e7 Å. Similar phenomena have been observed from PET,12 Lignin,36 and PEEK4 activated carbons. The t-method is a useful tool to evaluate parameters such as external surface area (Sex) of materials. The t-plot, constructed here, is based on Halsey’s model, which is derived from graphitized carbon black.8 Instead of plotting the isotherm with adsorbed amounts versus P/Po, the pressure ratio P/Po is converted to the thickness of the adsorbate (e.g., 3.4 Å, the
(1)
where Vmi is the total micropore volume; k ) 2.303(RT/βEo)2; R is the gas constant; and β is a scaling factor to normalize the equation for different adsorbates (as in the case of nitrogen, β ) 0.33).34 Eo is characteristic adsorption energy. The average pore width (LDR) can be estimated from LDR ) 2B/Eo where B is a constant (∼12 nm kJ/mol) for nitrogen adsorption on carbon.35 Figure 2 shows the DR plot, log V vs log2(P0/P), of activated carbons. All carbons give reasonable linear DR fitting in the region of P/P0 e 0.20, except C0.5 which can only be linearly fitted at very low pressure (i.e., P/P0 e 0.03). The pore width and micropore volume of carbons activated for different time periods are listed in Table 1. The average pore width increases considerably during the first half-hour of activation from 9.0 Å for C0 to ∼23.0 Å for C0.5 but decreases to 13.6 Å with additional activation up to 2 h. When the activation time is extended to 4 h, the average pore size increases again (20.0 Å). The optimum activation time is 2 h (C2) for generating high micropore volume (Vmi ∼1.79 mL/g). The microporosity, Vmi/ Vt, of these carbons dramatically increases from 13.0% (C0.5) to 98% (C2), clearly demonstrating the impact of time on activation process. The pore size distribution in the micropore region (pore width e20 Å) was also analyzed using the HK equation assuming a slit pore model7
Figure 3. HK pore size distribution of activated carbon, assuming slit pores.
TABLE 2: Characteristic Data of Activated PPO Carbons Derived from the NLDFT Methoda carbon sample
activation time (h)
nonactivated C0.5 C1 C2 C4
0 0.5 1.0 2.0 4.0
a
SNLDFT,T (m2/g)
SNLDFT,MP (m2/g)
560 1530 2080 3200 1930
560 500 2080 3200 1500
Vmi (mL/g)
Vmi/Vt (%)
0.19 0.38 1.08 1.75 0.93
98.0 13.4 95.0 97.0 49.0
SNLDFT,MP: Micropore Surface area. SNLDFT,T: Total Surface area.
Nanostructured Polymer-Based Activated Carbons
Figure 4. t-plots of PPO activated carbon (O) C0.5, (b) C1, (∆) C2, and ([) C4.
J. Phys. Chem. B, Vol. 112, No. 46, 2008 14367 adsorption region located in the outer part of the curve.37 A sharp rise of the inner part of the curve is attributed to the high surface area and interaction energy of the micropores. The external surface area, Sex, can be estimated from the slope of the curve at t > 7 Å. This analysis is based on the assumption that the micropores are completely filled up in this pressure region, and further adsorption is attributed to adsorption of the external surface, usually originating from macropores. A curve with a large slope value, as shown in C4, indicates that the corresponding carbon has a large external surface area. The external surface area of these carbons is listed in Table 1. The very steep slope of C0.5 suggests large Sex, but since the curve continues rising without a plateau (for linear fitting), distinction between the external surface adsorption and that of the micropore cannot be clearly discerned. Such t-plots are the direct results of a relatively wide pore size distribution, as evidenced in the HK plot (Figure 3). Carbons C1 and C2 have the external surface area of ∼24 and ∼43 m2/g, respectively, suggesting that micropores are the dominant feature of these carbons and that the pore sizes fluctuate within narrow boundaries. The much higher external surface area ∼352 m2/g of C4 is attributed to the widening of the pore size distribution and roughening of the carbon surface, caused by prolonged activation. Several approaches, such as high resolution Rs-plot28 and the SPE (subtracting pore effect) method,39 have been developed for evaluating carbon surface morphology, besides the BET, DR, and HK methods. The nonlocal density function theory (NLDFT) has also been recently used to characterize microporous carbons.40 The density function theory is a molecular modeling method that calculates the specific interactions of adsorbed species with the porous solid. The interaction of inhomogeneous fluids at a solid interface is described by the grand potential function, Ω[F(r)], of the average singlet density, F(r)
Ω[F(r)] ) F[F(r)] +
∫ F(r)[Vext(r) - µb]dr
(3)
where µb is the bulk chemical potential imposed on the system and Vext(r)is the wall potential (r is the distance from the absorbate molecule to the absorbent surface). The equilibrium density F(r) of the absorbate confined in a pore of pore width L at a given chemical potential (i.e., relative pressure) and temperature is determined by minimizing Ω[F(r)] with respect to F(r), i.e., {dΩ[F(r)]}/{dF(r)} ) 0. The net quantity adsorbed, Qads, at a given pressure is then obtained by integrating F(r) from wall to wall and subtracting the amount corresponding to the nonexcluded volume of the pore at the bulk gas density F0(r)
Qads )
∫ [F(r) - F0(r)]dr
(4)
For solids having a pore size distrubtion, the integral equation of isothermal adsoption becomes Figure 5. (a) Cumulative surface area distribution, (b) cumulative pore volume distribution, and (c) pore size distribution of activated carbons determined by fitting the NLDFT Autosorb V1.53 Software (from Quantochrome).
molecular diameter of N2) from the isotherm. The t-plots of carbons reported here are shown in Figure 4. A t-plot can be divided into two regions: the micropore adsorption region located in the inner part of the curve and the external surface
Q(p) )
∫ qads(p, L)f(L)dL
(5)
where Q(p) is the total quantity of adsorbate at pressure p; f(L) is the pore surface area distribution function with respect to pore size; q(p,L) is the kernel function and describes quantity adsorped for ideally homoporous materials of pore width L (assuming slit pores). Using the nonlocal density function theory (NLDFT) model, the Q(p) function can be applied to fit the experimental adsorption isotherm data, from which the pore size distribution function f(L) and, hence, surface area and pore size
14368 J. Phys. Chem. B, Vol. 112, No. 46, 2008
Yuan et al.
TABLE 3: Porous Textures of Activated PPO Carbons Derived by SAXS carbon sample
activation time (h)
SSAXS (m2/g)
Rg (Å)
Sx/Vs (×10-4 m-1)
Ds
dAFMa (nm)
nonactivated C0.5 C1 C2 C4
0 0.5 1.0 2.0 4.0
421 1910 1870 1930 2070
6.4 6.7 6.5 6.2 6.2
5.5 150.6 10.1 38.3 152.7
2.0 2.8 2.6 2.3 2.9
25 2.5 2.3 -
a
Pore diameter determined by AFM.
decays linearly with log(q), at a rate slower than that of region I. On the basis of Porod’s law,24 the scattering intensity can be expressed as
I(q) ) (∆F)2
(
distribution can be extracted. The cumulative surface area, cumulative pore volume, and pore size distribution data derived from fitting the NLDFT Autosorb V1.53 Software (from Quantochrome) are shown in Figure 5. The porous textures of activated PPO carbons, estimated from the NLDFT method, are listed in Table 3. For all activated carbons SNLDFT is smaller than SBET (Table 1). Figure 5a shows that for highly microporous C1 and C2 the surface area originates entirely from micropores (pore diameter
