Article pubs.acs.org/Langmuir
Structure and Sorption Properties of a Zeolite-Templated Carbon with the EMT Structure Type Julien Parmentier,*,† Fabrice O. M. Gaslain,‡ Ovidiu Ersen,§ Teresa A. Centeno,∥ and Leonid A. Solovyov*,⊥ †
Institut de Science des Matériaux de Mulhouse, IS2M, UMR CNRS 7361, Université de Haute-Alsace, 15 rue Jean Starcky, BP 2488, 68057 Mulhouse Cedex, France ‡ Mines Paris, Paristech, Centre des Matériaux, UMR CNRS 7633, BP 87, 91003 Evry Cedex, France § Institut de Physique et Chimie des Matériaux de Strasbourg, UMR 7504 CNRS-ULP, 23 rue de Loess, F-67034 Strasbourg Cedex, France ∥ Instituto Nacional del Carbon (INCAR, CSIC), Apartado 73,33080 Oviedo, Spain ⊥ Institute of Chemistry and Chemical Technology, Akademgorodok 50/24, 660036 Krasnoyarsk, Russia S Supporting Information *
ABSTRACT: An ordered microporous carbon material was prepared by the nanocasting process using the EMC-2 zeolite (EMT structure type) as a hard template. X-ray diffraction (XRD) and transmission electron microscopy (TEM) revealed long-range ordering in the material that resulted from the negative replication of the host template. The carbon porous network replicating the zeolite structure was modeled by overlapped spherical voids with diameters determined from the XRD pattern that displayed up to six distinct peaks. The surface delimiting the 3D interconnected porosity of the solid has a complex morphology. The pore size distribution calculated from the XRD-derived structural model is characterized by a maximum at 1.04 nm related to the longrange-ordered microporous network. Complementary studies by immersion calorimetry revealed that most of the porosity was characterized by a size above 1.5 nm. These porous features were compared to data resulting from classical analysis (DR, DFT, BET, etc.) of the N2 (77 K) and CO2 (low and high pressure, 273 K) physisorption isotherms. The limitations of these approaches are discussed in light of the pore size distribution consistently determined by XRD and immersion calorimetry measurements.
■
incomplete filling of the mold.6 These materials have numerous potential applications because of their high surface area (>2000 m2/g) and microporosity and/or mesoporosity depending on the template, the precursor, and the processing conditions. Nevertheless, it has to be pointed out that the models for adsorption on carbon replicas of complex pore topology are not well developed and the PSD analysis is often unreliable. For instance, carbon replicas of ordered mesoporous silicas with interconnected cylindrical mesoporosity (e.g., CMK-3 type) present arrays of interconnected carbon nanorods whose PSD is not properly described by the model based on the cylindrical pore shape (e.g., BJH). Similar problems exist in the pore size analysis of ordered microporous carbons derived from microporous templates such as zeolites for which a slitlike pore-shape model is often applied irrespective of the real pore
INTRODUCTION Porous carbon materials have numerous applications in the fields of energy storage, gas storage, molecular separation, and catalysis. Usually, the activation procedures lead to materials with broad pore size distributions (PSD), which complicates the control over their performance in specific applications. The need for carbons with controlled porosity, especially in the mesoporous and microporous ranges, promoted the development of new preparation routes based on the nanocasting of porous host materials1,2 or the direct synthesis of porous carbons with the aid of pyrolyzable porogen agents.3,4 For the former process, host materials, also called exotemplates,5 are used as molds. They are infiltrated with a carbon precursor that is further converted to carbon by heat treatment. Matrix removal by chemical etching yields a negative replica of the porous structure of the host material. This nanocasting process has been widely used with various molds and carbon precursors. A large variety of carbons were obtained with porosities arising from the host material wall removal and an © 2013 American Chemical Society
Received: July 23, 2013 Revised: November 20, 2013 Published: December 20, 2013 297
dx.doi.org/10.1021/la402762v | Langmuir 2014, 30, 297−307
Langmuir
Article
between both ranges of measurements. The relative fugacity f used in this study is the fugacity defined in ref 27 divided by the saturation fugacity f ° at 273 K (2.505 MPa, from ref 28). The theoretical background for microporosity characterization was first provided by Dubiniń’s theory and its extension to immersion calorimetry. The Dubiniń−Raduskevitch (DR) equation was applied to CO2 and N2 isotherms for the microporous volume determination.29 Affinity coefficient β was 0.33 for N2 and 0.35 for CO2. The densities used for liquid N2 at 77 K and adsorbed CO2 at 273 K were 0.808 and 0.96, respectively. The latter value is still a subject of debate because it seems to depend on the nature of the porous carbon;28 it was chosen here in order to fit the N2 pore volume at high pressure. It is in the range of 0.924−1.07 g/cm3, where the values correspond to the densities of liquid CO2 and adsorbed CO2 (Dubinin’s approach), respectively.28The corresponding PSDs were calculated from the method described in refs 28 and 29 in the relative pressure range of 0− 0.4 for nitrogen adsorbate. For CO2, the refinement was obtained by minimizing the summation [(ncalculated)1/2 − (nexperimental)1/2]2 over the relative fugacity range of 0−0.8. As a result of the lack of NLDFT kernels for the EMT carbon replica, the PSDs were also calculated from N2 and CO2 isotherms (Micromeritics ASAP 2010 and Quantachrom Autosorb software, respectively) assuming a slitlike pore geometry. The total surface area was estimated from this DFT method and compared to surface areas determined by BET and DR approaches. XRD Modeling. XRD data were collected on a Panalytical X-Pert diffractometer in the Bragg−Brentano configuration using Cu Kα radiation. The XRD modeling of the density distribution in the carbon material was done by applying the continuous density function (CDF) technique.30,31 The expected pore structure of the material was approximated by overlapped spherical voids that roughly outlined the boundaries of the zeolite template framework. Figure 1 demonstrates
morphology. The PSD curves resulting from such models may be used only for comparison between samples and as indicative values.7 An alternative approach has been presented by Roussel et al.8 using an atomic simulation of nitrogen adsorption on a carbon replica derived from the FAU zeolite structure type.9 They showed that the adsorption behavior of this nanostructured carbon could be described relatively well by a model based on two stacked graphene sheets, although the carbon material displayed a 3D porosity with cages delimited by tubular walls. Nishihara et al.10 have also investigated the molecular structure of this type of carbon. They proposed for the FAU zeolite-templated carbon (ZTC) a structural model made up of buckybowl-like nanographenes assembled into a three-dimensionally regular network built mainly of sp2 carbon atoms. They also revealed different forms of defects such as oxygen functional groups probably bound to the edge sites of each buckybowl unit. In this context, other reliable and complementary tools are required to confirm the previous data and to bring about new insight into the porous and structural features of ZTC replicas. The versatility of ordered porous exotemplates in the mesoporous and microporous ranges allows a large variety of pore replica morphologies. For instance, numerous zeolite types (FAU, MOR, LTA, BEA*, and MFI) have been used as molds for the preparation of carbons exhibiting various degrees of microporosity organization.11−16 These ZTC materials have shown interesting properties for hydrogen13,17 or CO2 storage18−20 and gas separation.21 In contrast to carbon replicas obtained from ordered mesoporous silica templates,22 most of the reported ZTC materials display very poor organization and only a single broad XRD peak, which precludes the combined analysis of their porous structure by XRD and physisorption techniques. Recently, we have shown that an appropriate choice of the zeolite template (EMT structural type) has led to the first zeolite carbon replica displaying more than three XRD peaks.23 The material framework and the corresponding ordered microporosity arising from the zeolite template have been maintained during the nanocasting process. A clear correlation among the intensity of the XRD peaks, the adsorbate (e.g., N2 or H2) uptake, and the supermicropore volume arising from the zeolite wall dissolution has also been shown.15,24,25 In this study, the porous structure of a highly ordered microporous carbon replica of EMT zeolite is quantitatively characterized using a combination of XRD modeling, CO2, and N2 physisorption and immersion calorimetry. Nonlocal density functional theory (NLDFT) and other adsorption data analysis methods (BET and Dubinin−Radushkevich) are also applied in order to investigate their validity and limitations for this type of microporous material.
■
Figure 1. Approximation of EMT-zeolite framework by overlapped spheres and the respective 3D model. the respective approximation of the zeolite framework by overlapped spheres A and B corresponding to the sodalite cage and the double sixmembered rings. The density inside the spheres was set to zero, and the average density in the outer region (filled by carbon) was assumed to be constant. The symmetry of the carbon replica lattice (P63/mmc) was inherited from the zeolite template. The adjustable parameters of the model were the hexagonal unit cell constants, the diameters of spheres A and B, and the Debye−Waller factor allowing for the internal structural disorder. The parameters were refined by applying the derivative difference minimization (DDM) method,32 which allowed the XRD refinement independently of the background curve. The intensities of diffraction reflections were calculated from the numerical Fourier transform of the density function defined on a grid of the unit cell. The Rietveld33 full-profile formalism was applied in the XRD powder profile calculations. The density distribution in the material was calculated by the inverse Fourier transform based on the structure factor modules estimated from the XRD pattern with the DDM decomposition procedure34 and the initial phases derived from the refined density model function (Table 1). An XRD-based PSD was calculated by a numerical integration over the volume of the density distribution bounded by the pore surface. The fraction of pores of size D was defined as the percent of the pore volume coverable by (and accessible to) spheres of diameter D but not by spheres of diameter D+δ.35
EXPERIMENTAL SECTION
Adsorption Experiments. The synthesis of the EMT-zeolite template, a calcined EMC-2 phase with Na20[Al20Si76O192] composition, and its carbon counterpart are described in ref 23. The ZTC has been characterized by different complementary techniques such as N2 adsorption at 77 K, high-pressure CO2 adsorption at 273 K, and immersion calorimetry at 293 K. Nitrogen adsorption experiments at 77 K were performed on a Micromeritics ASAP 2010. The samples were degassed overnight at 573 K and then degassed again after the dead space determination with He gas. CO2 adsorption (273 K) at low (subatmospheric) and high pressure (3.2 MPa) was recorded on a Quantachrom Autosorb A1-LP and on a volumetric device26 in the relative fugacity ranges of 3 × 10−6−0.04 and 0.007−0.988, respectively. It must be noted that there is very good consistency 298
dx.doi.org/10.1021/la402762v | Langmuir 2014, 30, 297−307
Langmuir
Article
Table 1. Structure Factors Derived from the Observed XRD Pattern (Fo) in Comparison to the Calculated Ones (Fc)a hkl 0 0 0 0 1 0
1 0 1 1 1 1
0 2 1 2 0 3
|Fo|
Fc
84.8 100.0 35.3 7.4 15.5 13.0
84.4 −99.0 36.1 4.5 14.2 −11.5
hkl 0 1 0 0 0 0
2 1 2 0 2 1
0 2 1 4 2 4
|Fo|
Fc
0 4.6 4.7 15.7 7.6 6.4
0.3b 5.4b 8.7b 13.0b 5.0b −3.3b
W (Lc) = −
Owing to the irregular nature of the surface, the pore boundary position in the geometrical PSD calculation was assumed to reside ∼0.12 nm below the 50% level of the maximal density. This estimation of the pore surface position was done taking into account the van der Waals radius of carbon (0.17 nm) and an approximate distance of ca. 0.05 nm between the 50% density level and the average position of the surface atom centers. Immersion Calorimetry. As a thermodynamic consequence, the Dubinińs theory leads to the following expression for the enthalpy of immersion of microporous carbon in a liquid whose vapor follows the DR equation36 βEoWo(1 + αT ) π 2Vm − h iSe
(2)
In the present case, the accessibility to the porous network of the EMT-ZTC was evaluated by using, as liquid probes, CS2, C6H6, and tri-2,4-xylylphosphate (TXP) with molecular dimensions of 0.33, 0.41, and 1.50 nm, respectively. CS2 appears to be a good complement to CO2 and N2, in view of their similar molecular dimensions. Because phenol adsorbed by carbons from dilute aqueous solutions forms a monolayer on the surface of carbons, this property can be used to determine the total surface area with the help of immersion calorimetry.37,38 On the basis of a reference value of −0.105 ± 0.004 J/ m2, the enthalpy of immersion (J/g) into a 0.4 M aqueous solution of phenol leads to the total surface area Sphe of EMT-ZTC. TEM Analysis. TEM experiments were performed on a JEOL 2100F electron microscope working at 200 kV, equipped with a cs probe corrector and a GIF Tridiem energy filter. Before observation, the samples were ultrasonically dispersed in ethanol, and a drop of suspension was subsequently deposited onto a carbon membrane grid. The TEM images were recorded mainly in bright-field mode. Some selected-area electron diffraction (SAED) patterns were also taken over a 500-nm-diameter circular area. The EELS spectra were recorded in TEM mode at different locations on the platelets by averaging the EELS signal coming from an area of 200 nm lateral size. The energy resolution is about 0.7 eV. To have access to the interplanar organization between the successive carbon layers that can be easily observed in a top-view image, several cross-sectional observations were made of the thin slides obtained by cutting the sample with an ultramicrotome.
a The reliability factor RF = 7.3%; F(000) = 326. bThese values were not determined precisely because of the statistical noise, but they were included in the Fourier synthesis of the density distribution map in order to reduce artificial modulations.
−Δi H(J g −1 ) =
[Δi H(exp) − h iSe]2Vm + h iSe βEo(1 + αT ) π
■
(1)
RESULTS XRD Structural Investigations. The XRD pattern of the EMT-carbon replica demonstrated up to six distinct nonzero diffraction peaks. The consistency between the observed and calculated XRD patterns after the DDM refinement of the structural model is illustrated by Figure 2a. The difference curve exhibits minor disagreement between the observed and
where Eo, Wo, and Se correspond to parameters derived from the adsorption of small molecular probes N2 and CO2. α is the thermal expansion coefficient of the liquid, and hiSe corresponds to the wetting of the external surface Se. From eq 1, it is possible to calculate the value of micropore volume W filled by different liquids as a function of their critical molecular dimension Lc.
Figure 2. Lorentz-factor-corrected observed (−), calculated (---), and difference (···) XRD patterns for the carbon material after (a) DDM refinement and (b) DDM decomposition. The calculated reflection positions are marked by ticks. 299
dx.doi.org/10.1021/la402762v | Langmuir 2014, 30, 297−307
Langmuir
Article
sections. The continuous density areas corresponding to the infiltrated carbon material closely replicate the main pores of the zeolite template. The surface of the solid appears to be complex with concave and convex areas. The carbon framework resembles a Schwarzite structure detailed by Terrones et al.39 Its external surface at ∼0.12 nm below the 50% level of the maximal density is around 14 nm2 per unit cell. The average carbon wall thickness may roughly be estimated as ∼1 nm. A comparative analysis of the observed, calculated, and difference density distribution maps revealed some regular modulations of density along the carbon frameworks that may be viewed in Figure 3c as the darker areas on the sections. The local density maxima are observed in the thinner (0.7 nm) segments of the material framework. As a tentative interpretation, one can presume that the thinner carbon framework segments (replicating the thinner pores of the zeolite template) have a ca. 10% higher density than the thicker ones. The XRD-geometrical PSD derived from the density distribution based on the observed structure factors is shown in Figure 4. The main peak at ∼1.04 nm corresponds to the voids approximated by spheres A in the structure model (Figure 1). The broad shoulder to the left of the main peak is related to the cumulative volume of interconnecting voids B whose contribution to the PSD is small compared to the volume of the main voids A. The difference between the position of the main PSD peak and the refined diameter, 1.3 nm, of the voids A is due to the aforementioned distance, 0.12 nm, between the 50% density level and the expected pore surface position. It is worth noting that the XRD-PSD in Figure 4 is in relatively good agreement with the simulated PSD given by Builes et al. (see Figure S4 in the Supporting Information of ref 18) and calculated using the Monte Carlo scheme with nitrogen as the probe and an idealized tubular carbon structure derived from the EMT structure type. To make the XRD-PSD more adequate for the real material structure, the dispersal of the average pore surface boundary allowed for by the Debye− Waller factor (mean dispersion amplitude ca. 0.14 nm) was also taken into account. It led to a wider geometrical pore size distribution ranging from 0.7 to 1.3 with maximum around 1.0 nm. From the geometrical PSD, the expected idealized porosity of the carbon material is estimated to be 48%. As shown in the Supporting Information, a simple calculation based on the carbon content in the carbon/zeolite composite (23 wt %) and the cell volume of the EMT carbon replica (6.88 nm3) and assuming the carbon density close to that of graphite (1.8−2.2 g/cm3) leads to a higher total experimental porosity of ca. 75− 80% (denoted Xt‑exp). This suggests the presence of a complementary irregular porosity in the carbon framework. As evidenced by Fuertes6 and other authors,8 the replication of mesoporous or microporous oxide structures by the nanocasting process results, normally, in an incomplete filling of the template pores. In this case, the nonfilled template pores also contribute to the overall pore volume, and as a result, the carbon contains two kinds of pores: (i) narrowly distributed ones, related to the oxide framework dissolution, and (ii) larger, broadly distributed ones formed by the coalescence of the nonfilled pores within the regular pores of the template. A similar phenomenon could occur here because the small pore size of the EMT template (microporosity instead of mesoporosity) precludes its complete filling. Moreover, other calculations based on the grand canonical Monte Carlo
Figure 3. EMT-templated carbon material structure: (a) surface of the 50% density level, (b) a similar surface within one unit cell, and (c) the density distribution calculated from the observed structure factors. The characteristic sections are bounded by the 50% density level surface.
Figure 4. Idealized and dispersed geometrical pore size distributions for the EMT carbon replica. The dispersed curve is derived from the idealized one by taking into account the pore boundary dispersal with the mean amplitude of 0.14 nm given by the refined Debye−Waller factor value.
calculated diffraction peak intensities, which suggests that the proposed approximation of the density distribution in the material is reasonably adequate for its real framework structure. The hexagonal unit cell constants of the carbon replica lattice were determined to be a = 1.693 nm and c = 2.771 nm. The refined diameters of spherical voids A and B were 1.3 and 0.95 nm, respectively. The XRD profile fitting resulting from the DDM decomposition is shown in Figure 2b. The corresponding observed and calculated structure factors used in the density distribution calculations are listed in Table 1. The density distribution based on the observed structure factors is visualized in Figure 3 by the 50% density level surface and characteristic 300
dx.doi.org/10.1021/la402762v | Langmuir 2014, 30, 297−307
Langmuir
Article
Figure 5. Nitrogen adsorption isotherms at 77 K for the EMT carbon replica: (a) linear scale and (b) logarithmic scale.
Figure 6. Carbon dioxide adsorption isotherms at 273 K for the EMT carbon replica: (a) linear scale and (b) logarithmic scale (triangles and circles represent low- and high-pressure measurements, respectively).
simulation of carbon adsorption lead to a maximally achievable carbon loading of 43 wt % in the zeolite/carbon composite (i.e., 0.74 g of carbon per g of zeolite),16 which is notably higher than the experimental value determined here (23 wt %). This means that the host material is not completely filled with carbon and residual porosity is still present in the zeolite/ carbon composite. Therefore, the two kinds of porosity described previously could be expected for the EMT carbon replica studied. The former is related to the regular nanocasting of the zeolite framework, being ordered with the porous features revealed by XRD (PSD peak around 1 nm; 48% of idealized structured porosity). The second type of porosity, related to the inhomogeneous and incomplete filling of the template, is disordered and thus does not contribute to the observed XRD pattern. If we consider the experimental total porosity Xt‑exp calculated from the experimental data (75% < Xt‑exp
