- Email: [email protected]

Note

Low-pressure argon adsorption assessment of micropore connectivities in activated carbons T. Zimny a,∗ , F. Villieras b , G. Finqueneisel a , L. Cossarutto a , J.V. Weber a a Laboratoire de Chimie et Applications, EA 3471, Université de Metz, Rue Victor Demange, 57500 Saint-Avold, France b Laboratoire Environnement et Minéralurgie, Ecole Nationale Supérieure de Géologie, UMR INPL-CNRS 7569, 15 av. du Charmois, BP 40,

54501 Vandoeuvre-lès-Nancy Cedex, France Received 19 March 2005; accepted 9 June 2005 Available online 19 August 2005

Abstract Low-pressure argon adsorption has been used to study the energetic distribution of microporous activated carbons differing by their burnoff. The collected isotherms were analyzed using the derivative isotherm summation method. Some oscillations on the experimental curves for very low partial pressures were detected. The results are analyzed and discussed according to the literature and could be attributed to local overheating caused by spontaneous mass transfer of argon through constrictions between former pores and the new opening pore or deadend pores. We used the dynamic character of the experimental method and mainly the discrepancy of the quasi-equilibrium state to deduce key parameters related to the porosity topology. 2005 Elsevier Inc. All rights reserved. Keywords: Activated carbon; Surface heterogeneity; Micropore connectivity; Low-pressure argon adsorption

Activated carbons (AC) are recognized as energetically heterogeneous materials due to the irregularities in their pore systems, which are an inherent characteristic. Techniques for the characterization of the microporous structure of AC are consequently of primary importance for the determination of potential applications. Description of their micropores is generally presented as one of the main parameters conditioning their adsorptive properties as well as surface functional groups. To reach this goal, the most common approach is using the adsorption of a range of probes differing in properties such as size and polarity. Nevertheless the heterogeneous distribution of the most energetic sites is not easy to determine. A high-resolution gas adsorption technique at very low partial pressure has been successfully developed to characterize surface heterogeneity of minerals [1] using argon or nitrogen. Some papers discuss the potentialities of the techniques to describe the gas/solid interaction of carbon– * Corresponding author. Fax: +33 387 93 91 01.

E-mail address: [email protected] (T. Zimny). 0021-9797/$ – see front matter 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.jcis.2005.06.044

mineral matrix [2,3], but very few are available for the characterization of pure carbonaceous adsorbents [4,5]. In this work, two commercial activated carbons provided by PICA Company (France) have been studied by argon adsorption at low pressure in order to study the burnoff effect on the energetic distribution of micropores. The samples were obtained by activation of coconut shells in steam at 1123 K and are respectively called NC60 and NC100 according to increasing burnoff. They both are considered mainly microporous materials and present a poor surface chemistry (see Table 1) confirmed by a low oxygen content (respectively 3.2 and 2.6% for NC60 and NC100). The isotherms of argon have been collected for both samples at very low partial pressure according to the experimental procedure previously discussed by Rouquerol et al. [6] and by Michot et al. [7]. Briefly, by slow, constant, and continuous flow, the adsorbate is introduced into the adsorption system through a microleak. The flow rate is maintained constant, at least up to the BET domain, and can be adjusted by the pressure imposed before the leak. If the introduction rate is low enough, the recorded pressures can

T. Zimny et al. / Journal of Colloid and Interface Science 293 (2006) 248–251

Table 1 Some basic characteristics of the studied samples Samples Vmicropore (cm3 /g) NC60 0.402 NC100 0.585

Vmesopores (cm3 /g) 0.053 0.082

BET surface (m2 /g) 1010 1493

Acidic groups Basic groups (meq/g) (meq/g) 0.05 0.06

0.49 0.50

by replacing 0 P kT ln E P by =

be considered as quasi-equilibrium states. Starting from the quasi-equilibrium pressure (in the range 10−4 –3 × 10−4 Pa), recorded as a function of time, the adsorption isotherm is calculated. Three high-accuracy differential pressure transducers were used for pressure measurements, allowing 0.05% accuracy of pressure read. A dynamic vacuum of 10−7 Pa is ensured on the reference side by the use of a turbomolecular vacuum pump. An accurate constant level of liquid nitrogen is maintained using a homemade electronically controlled device. The frequency of pressure recording is adjusted after each measurement to record 100–200 experimental points per unit log of relative pressure. Thus, 2000–3000 points were collected each time for relative pressures lower than 0.15. The data were treated using the derivative isotherm summation (DIS) procedure designed by Villiéras et al. [8] to examine the surface heterogeneity of the samples. Due to the large number of experimental data points acquired, the experimental derivative of the adsorbed quantity can be calculated accurately. The total derivative adsorption isotherm on a heterogeneous surface is simulated by the sum of local theoretical derivative adsorption isotherms on homogeneous surfaces, Xi θit , θt = (1) i

where θt is the total adsorption isotherm, θit is the adsorption isotherm on the different energetic domains of the surface, and Xi is its contribution to θt . θit could be written in the form θit = θi (ε)χi (ε) dε, (2) Ω

where θi (ε) is a local theoretical adsorption isotherm, and χi (ε) is the local dispersion of the adsorption energy ε in the ith domain. So and extended version of this DIS method, in which θit s are functions defined by Eq. (2), was developed. The Dubinin–Astakhov isotherm was proposed for describing adsorption in micropores [9], kT Pi0 ri ln , θit (P /P0 ) = exp − (3) Ei P where Ei is the variance of χi (ε) and ri is a parameter governing the shape of the distribution function. It is a bellshaped Gaussian-like function widened on the low-energy side for ri < 3, and widened on the high-energy side for ri > 3. Equation (3) can be generalized to take into account the effect of lateral interactions between adsorbed molecules

249

0 kT P ωθ ln − . E P E

(4)

In addition, it is also well known that the variance of the system for heterogeneous surfaces and lateral interactions are coupled parameters [10,11]. It is thus very difficult to assess this parameter properly and it is used when a best fit parameter is necessary. For ri = 2 and ω = 0, Eq. (3) becomes the Dubinin– Radushkevich isotherm. In practice, application of the Dubinin–Astakhov equation is accompanied commonly by the assumption that Pi0 means the saturated vapour pressure P0 . Though this approximation may be justified in many cases, generally Pi0 cannot be identified with P0 , and must be treated as the best fit parameter, which represents the pressure at which the largest micropores of the family are filled. In the DIS method, estimation of values of the different parameters of interest is done by considering the coordinates of the maximum and the width of the derivative. The mathematical relations between Ei , ri , and Pi0 can easily be derived from the expressions for the first and second derivatives of Eq. (3), when the second derivative is put equal to zero. The first derivative of Eq. (3) is r kT []r−1 θ ∂θt

. (5) = Erω ∂ ln P T 1 − E []r−1 θ The expression for the second derivative is relatively complicated, so its values are computed numerically. The derived isotherms (experimental and simulated) obtained for both samples are presented in Figs. 1a and 1b. They show a resolved peak close to the limited experimental window (ln P /P0 ∼ −3). As a consequence, the monolayer approach following the Dubinin–Astakhov isotherm has been adopted for the DIS modeling. This simulation has been realized using the minimum number of domains according to the best experimental curve fitting. Four domains were necessary to obtain the best description of the surface heterogeneities for both samples. The quantitative determination of the adsorbed volume is given in Fig. 1c. The positions of the peaks correspond to the maximum of the bell-shaped curve. It appears that the values of ln P /P0 for each domain are always slightly shifted toward the lowest energies according to increasing burnoff. The volume of domain I that corresponds to the most energetic sites is quite similar for both samples. In contrast, noticeable changes occur for the volumes of domains II and III, which are the microporous network concerned by the activation process. The last domain is attributed to the less energetic micropores and, in a first approach, could be considered a transition between wider micropores and the beginning of mesoporosity. In this

250

T. Zimny et al. / Journal of Colloid and Interface Science 293 (2006) 248–251

Fig. 2. Experimental derived curves of the samples.

Fig. 1. DIS modeling of first-layer argon condensation energy distributions on AC NC60 (a) and NC100 (b) and contributions of the energetic domains (c).

case, the adsorption is generally associated with cooperative effect [12], corresponding to multilayer adsorption, which was not taken into account in the present case. As a consequence, the adsorbed volumes determined for this last peak are overestimated. In the case of the adsorption study of a molecular probe, the size distribution of the pores is strongly connected to the adsorption equilibrium and kinetics. The pore connectivity is a parameter to take into consideration for the modelling of the isotherms [13–16]. Adsorption of argon in quasi-equilibrium conditions shows for the initial state of adsorption some oscillations on the experimental

curves for ln P /P0 between −14 and −10 (Fig. 2). A magnification of part of the derived curve is given in Fig. 2. It could be observed that the oscillations are of slight amplitude for NC60 and become most important for NC100 with some clear negative contributions on the derived curve. The origin of this phenomenon could be connected to a localized overheating. One possible way to check this assumption is to modify the experimental conditions of acquisition. It can be assumed that the burnoff opens closed pores connected to the others by geometric restrictions. A slow process of diffusion occurring across the constriction leads to a large mass transfer of gases for a given pressure corresponding to localized overheating. Indeed, the quasi-equilibrium technique presents a dynamic character and in the presence of a large amount of high-energy adsorption sites the thermal equilibrium of the sample can be broken due to insufficient heat transfer toward the liquid nitrogen thermostatic bath. As a consequence, two ways can be tested to limit oscillations: decreasing the number of high-energy adsorption sites in the adsorption system by using smaller sample mass or decreasing the adsorption rate by decreasing the introduction flow rate (i.e., the running pressure before the microleak). Both approaches were tested for an NC100 sample (Fig. 3), showing that such changes of the experimental conditions lead to a significant decrease of the oscillations, which is in agreement with the expected reduction of the overheating phenomenon. In addition, the shape of the mass uptake after the domain of pressure corresponding to these oscillations is similar to the initial part. This phenomenon is close to the activated diffusion described in Refs. [10,17,18]. The D–R equation has been applied on the corresponding isotherm and the obtained curves are reported in Fig. 4. Negative deviations from the DR relationship appear at very low relative pressure and could arise from the existence of activated diffusion or a molecular sieve effect. Furthermore, the inflection point at ln P /P0 = −12.5 corresponds to the beginning of the oscillations of the isotherm derived curve. The pore size concerns

T. Zimny et al. / Journal of Colloid and Interface Science 293 (2006) 248–251

Fig. 3. Experimentally derived curves of the sample NC100 under different experimental conditions.

251

To sum up, the study of the micropore network of two commercial activated carbons differing in the burnoff level has been realized using low-pressure argon adsorption. The obtained isotherms were modeled using DIS method. Four energetic domains of micropores were defined and the effect of burnoff was attributed to a moderate enlargement of the pores corresponding to 10−9 < ln P /P0 < 10−5 . Furthermore, some unusual oscillations were observed on the derived curve of NC100, which were attributed to a localized overheating due to a spontaneous mass transfer of argon through constrictions between former pores and the new opening pore or deadend pores, as suggested by the strong increases of micropore volume without significant variation of the average micropore size. This work takes advantage of the dynamic character of the experimental method and mainly the discrepancy of the quasi-equilibrium state to deduce key parameters related to the porosity topology.

Acknowledgment This research work is supported by the CNRS, in the frame of the Jumelage “Matériaux carbonés et catalytiques pour l’environnement.”

References

Fig. 4. D–R curves of the NC 100 sample under different experimental conditions.

by these domains of relative pressure are probably comparable to the molecular size and therefore molecules may never completely escape the potential field of the pore walls. Gas transport could probably occur according to an activated diffusion mechanism. It has been shown [19] that in the case where the DR equation applies, the average micropore width L0 is related to E0 by L0 [nm] = 10,800 [nm J mol−1 ]/(E0 − 11,400) [J mol−1 ]. E0 and W0 are usually obtained from the adsorption of small molecules, such as carbon dioxide in our case. This simple approach provides a rough estimate equal to 0.53 nm for NC60 and 0.57 nm for NC100. As can be seen, the values are quite similar although the total volume of micropore has increased more than 45% (Table 1). The effect of burnoff on our sample is probably to open some closed pores, leading to an increase of the entire pore structure, including some deadend pores. Recently, it has been shown [20] that these pores are associated with significant mass transport resistance, since they could be selective regarding the molecular size of argon.

[1] F. Villiéras, L. Michot, E. Bernardy, M. Chamerois, C. Legens, G. Gerard, J.M. Cases, Colloids Surf. A 146 (1999) 163. [2] F. Villiéras, R. Leboda, B. Charmas, F. Bardot, G. Gerard, W. Rudzinski, Carbon 36 (10) (1998) 1501. [3] B. Charmas, R. Leboda, G. Gerard, F. Villièras, Appl. Surf. Sci. 196 (2002) 126. [4] W. Rudzinski, K. Nieszporek, J.M. Cases, L.J. Michot, F. Villiéras, Langmuir 12 (1996) 170. [5] C. Garnier, T. Gorner, A. Razafitianamaharavo, F. Villiéras, Langmuir 21 (2005) 2838. [6] J. Rouquerol, F. Rouquerol, Y. Grillet, R.J. Ward, Characterization of Porous Solids, vol. 317, Elsevier, Amsterdam, 1988. [7] L. Michot, M. Francois, J.M. Cases, Langmuir 6 (1990) 677. [8] F. Villiéras, J.M. Cases, M. Francois, L. Michot, F. Thomas, Langmuir 8 (1992) 1789. [9] F. Villiéras, L.J. Michot, J.M. Cases, M. François, W. Rudzinski, Langmuir 13 (1997) 433. [10] J.L. Bersillon, F. Villiéras, F. Bardot, T. Gorner, J.M. Cases, J. Colloid Interface Sci. 240 (2001) 400. [11] V.S. Mamleev, F. Villiéras, J.M. Cases, Langmuir 18 (2002) 2075. [12] H. Marsh, Carbon 25 (1) (1987) 49. [13] H. Liu, L. Zhang, N.A. Seaton, Chem. Eng. Sci. 47 (1992) 4393. [14] H. Liu, N.A. Seaton, Chem. Eng. Sci. 49 (1994) 1869. [15] C. Nguyen, D.D. Do, Langmuir 16 (2000) 1868. [16] M.V. Lopez-Ramon, J. Jagiello, T. Bandosz, N.A. Seaton, Langmuir 13 (1997) 4435. [17] J.-G. Choi, D.D. Do, H.D. Do, Ind. Eng. Chem. Res. 40 (2001) 4005. [18] K.J. Masters, B. McEnaney, J. Colloid Interface Sci. 95 (1983) 340. [19] F. Stoeckli, A. Slasli, D. Hugi-Cleary, A. Guillot, Micropor. Mesopor. Mater. 51 (2002) 197. [20] S. Lagorsse, F.D. Magalhaes, A. Mendes, J. Membr. Sci. 241 (2004) 275.

Copyright © 2022 COEK.INFO. All rights reserved.