Alkylsilane-Functionalized Microporous and Mesoporous Materials

Article pubs.acs.org/JPCC

Alkylsilane-Functionalized Microporous and Mesoporous Materials: Molecular Simulation and Experimental Analysis of Gas Adsorption S. Builes,*,† P. López-Aranguren,†,‡ J. Fraile,‡ L.F. Vega,†,§ and C. Domingo*,‡ †

MATGAS Research Center and ‡Instituto de Ciencia de Materiales de Barcelona (CSIC), Campus de la UAB s/n, Bellaterra, 08193, Spain § Carburos Metálicos, Air Products Group, C/Aragón 300, Barcelona, 08009, Spain S Supporting Information *

ABSTRACT: Solid sorbents are considered as a potentially less-energy-intensive alternative to the use of liquids for the removal and separation of liquid and gaseous fluids. The control of the surface characteristics of porous inorganic materials via the deposition of an organic layer is of great interest for tailoring the properties of the sorbent. For instance, organic functionalization of traditional solid sorbents (micro- and mesoporous silica and silicates) allows tuning their surface properties, such as hydrophilicity or hydrophobicity and surface reactivity. However, the underlying mechanism of the sorption process in highly complex organic functionalized materials is not yet fully understood. This incomplete understanding limits the possibilities of designing optimal adsorbents for different applications increasing the interest in performing complementary experimental-simulation studies. In this work, the adsorption of N2 in alkylsilane-modified disordered mesoporous silica (silica gel 40) and crystalline aluminosilicate (zeolite Y) is analyzed by a combination of experiments and simulations. The goal of the adsorption simulation study was twofold: first, to assess the ability of using grand canonical Monte Carlo to obtain quantitative predictions of the adsorption characteristics of gases on alkylsilane postfunctionalized products and, second, to provide new insights into the adsorption mechanism. A supercritical silanization experimental method was used for the postmodification of the internal surface of the studied porous substrates. This work demonstrates that even though the models of amorphous hybrid materials require simplifications related to the cell size and silane polymerization modes, it is possible to use these models to obtain an adequate insight of what happens in the macroscopic systems. These models allow us to acquire information on the mechanisms of silane functionalization and the interactions of the support and silane chains with the adsorbed gases.

1. INTRODUCTION Sorption on micro- and mesoporous solid materials has been found to be one of the most effective, potentially less-energyintensive, sorption technologies.1 Their main advantages are related to their accessibility, easy regenerability, convenient application in batch and continuous processes, and high sorption capacity. Traditionally, zeolites,2 porous siliceous minerals,3−6 and superabsorbent hydrogel polymers7 have been widely used as sorbents for pollutants in aqueous systems. However, industrial applications of these materials are limited by the low sorption capacity and selectivity for nonpolar or hydrophobic matter. Currently, a new class of solid porous sorbents based on organic−inorganic hybrids is being intensively prospected for new uses in organic matter separation, with applications ranging from sorption of oil and hydrocarbon contaminants to CO2 capture.8−11 The organic functionalization of solid materials allows tuning surface properties, such as hydrophilicity or hydrophobicity and reactivity.12−15 Therefore, porous sorbents can be tailored (or functionalized) based on the molecular structural features of target sorbate molecules. © 2012 American Chemical Society

With the advent of periodic amorphous mesoporous silica (SiO2) materials, organo-modified silica based materials have attracted more and more attention as sorbents. Furthermore, the family of organo-modified disordered mesoporous silica supports (silica gels) also deserves attention, mainly because of their low cost, large number of silanol groups on the internal surface, and high surface area and pore volume. These properties make them suitable for a large amount of bulk applications in the area of sorption of organic compounds. Herein, a disordered mesoporous silica is used as the host for the organic modifying moiety. Results on sorption capacity for this material are compared with those of an organo-modified crystalline aluminosilicate (faujasite). We used this latter material for comparison because in commercial applications of adsorption the microporous zeolite family is the most widely used adsorbent. Received: February 28, 2012 Revised: April 12, 2012 Published: April 16, 2012 10150

dx.doi.org/10.1021/jp301947v | J. Phys. Chem. C 2012, 116, 10150−10161

The Journal of Physical Chemistry C

Article

Figure 1. Used matrixes and silane reagents (raw and hydrolyzed). (A) octyltriethoxysilane, (B) octyldimethylmethoxysilane, (C) octylsilanetriol, and (D) octyldimethylsilanol.

conventional approach, the additional silica group (−SiO) from the silane molecule is taken into account in the pure silica model as an extra SiO2 layer. Consequently, this model gives better results for in situ modified materials than for the postsynthesis functionalized ones.21 For this reason, the procedure of grafting the complete surface group (R-SiO) on the surface of the pure silica model material is taken in this work.24 In the last section, information about the porosity, the pore size distribution, the surface area, and the surface chemistry of pristine and silanized materials is obtained from the adsorption data of the GCMC simulations. The equilibrium theory of simple inhomogeneous fluids is used to obtain a reliable prediction of fluid properties in crystalline microporous zeolites. The simulated results are discussed and compared with experimental data available from our laboratory. Because the main objective of this research was to obtain models for use in the design of sorbent materials, the performance of the generated structure in the simulation of adsorption compared with experimental data was the main criterion in the evaluation of the modeling method.

One of the most successful surface functionalization approaches is the impregnation and chemical grafting of long hydrocarbon chains via silanization.16 In this work, a supercritical silanization method was used for the postmodification of the internal surface of porous substrates.8,17 Monoalkyltrialkoxy (R-Si(OH)3) and trialkylmonoalkoxy (R3-SiOH) silanes were used for the study. The final objective of this study is to identify the adaptation of hybrid structures for particular applications related to sorption. To utilize these materials for maximum economical and environmental benefits, we must elucidate the relationship between physical properties (e.g., specific surface area and pore structure) and surface chemistry (e.g., surface charge, hydrophobicity). From an experimental perspective, both micro- and mesoporous structures are difficult to characterize; therefore, it is not easy to understand the sorbent structure at a molecular level. Hence, molecular simulations are used here to complement adequately the experimental efforts. The structures of crystalline solids, such as zeolites, are entirely known, and atomistic models are already available.18 However, for noncrystalline materials, such as silica gel, the link between the real and the simulated material is merely statistical. Therefore, the first step to reproduce the functionalized silica gel surface was to build a realistic model to represent the pristine porous matrix. In this study, the method for the generation of atomistic models of silica gel of MacElroy and Raghavan19,20 was used. The methodology commonly found in the literature for simulating functionalization of silica materials consists of using an algorithm to replace a number of surface silanol groups by organic moieties. A common procedure consists of adding to the surface only the organic group (R-) in the silane molecule and not the whole hydrolyzed compound (R-SiO).21−23 In that

2. EXPERIMENTAL SECTION 2.1. Materials. Commercial supports, pertaining to two groups of materials, were studied (see Figure 1): (i) a crystalline microporous Zeolite Y (ZY) of the Faujasite class, containing Si and Al atoms (Si/Al ≥1, Strem Chemicals), and (ii) an amorphous mesoporous silica gel 40 (SG40, Aldrich). The zeolite substrate was activated by calcination in a tubular oven (Carbolite 3216) at 520 °C during 48 h under N2. The matrices were functionalized with either an alkyltrialkoxysilane (octyltriethoxysilane: CH3(CH2)7Si(OCH2CH3)3) or a trialkylalkoxysilane (octyldimethylmethoxysilane: 10151



Article

Table 1. Supercritical Operating Conditions and Some Obtained Results: Textural Properties and Silane Grafting Density (ρgrafting) sample 1-SG40 2-SG40 3-SG40 4-SG40 1-SGT 2-SGT 3-SGT 4-SGT0.2 5-SGT0.2 1-ZY 2-ZY 1-ZYT 2-ZYT a

silanea

T [K]

P [bar]

t [min]

C C D

348 348 348

100 200 100

90 90 45

C C C C A A

348

200

90

SBET [m2 g−1]

VBJH [m3 g−1]

556 460 410 424 475 490 532 540 536 750 360 740 360

0.47 0.41 0.38 0.35 0.44 0.37 0.33 0.29 0.26

Vmp [m3 g−1]

Dp [nm]

ρgrafting [mmol g−1]

3.4 3.6 3.7 3.3 3.7 3.0 2.2 2.0 1.9 0.28 0.17 na na

0.41 0.58 0.59 0.4 0.6 0.4 0.6 0.60 0.59

Impregnated silane form (Figure 1) that was also used to calculate the molecular weight and thus the grafting density (ρdensity).

Second, silane chains were added considering two possible kinds of interactions with the substrate, physical (impregnation) and chemical (grafting). Finally, molecular simulations of the generated models were used to understand the N2 adsorption phenomena in pristine and silanized materials. 3.1. Structural Models of the Porous Media. 3.1.1. Zeolite Model. Zeolite Y is a microporous crystalline solid with a well-defined structure consisting of interconnected channels that forms a 3-D ordered network. Zeolite Y, with a faujasite framework, is formed by connected TO4 tetrahedral, where the T atoms are silicon or aluminum cations. This zeolite consists of almost spherical cavities of 1.3 nm in diameter accessible through tetrahedral windows of 0.74 nm. In crystalline zeolites, pore structure follows directly from the crystallographic data. Hence, only the coordinates of the zeolite Y framework atoms, obtained by X-ray diffraction, are required to simulate the structures. In this work, the faujasite structure was extracted from the zeolite database available in the Materials Studio software package.26 A ratio Si/Al = 1 was considered in the model, with no lattice defects and thus no silanol groups on the pore surface. The presence of aluminum introduces charge defects in the zeolite framework, which are compensated with Na+ nonframework counterions. The point charge values were taken from the works of Maurin et al.27 They previously reported the adsorption of different gas molecules on Faujasites at room temperature (Si: +2.4e, Al: +1.7e, O: −1.2, and Na: +0.7). The Lennard-Jones (LJ) parameters were taken from the consistent valence force field (CVFF). These parameters have been successfully used to model CO2/N2 adsorption on Faujasite.28 The model of the Zeolite Y network is shown in Figure 1. 3.1.2. Amorphous Silica Gel Model. Experimentally, mesoporous amorphous silica gels are obtained by random aggregation of primary dense silica nanospheres dispersed in a colloidal suspension (Figure 2a). The result is an amorphous porous material with a random pore size distribution and extremely high pore connectivity. The method used in this work to generate the model for the amorphous silica gel is an atomistic procedure that follows the concepts of the experimental synthetic procedure.19,20,29 Figure 2b shows the main stages involved in the simulation of the porous silica structure. Concisely, using the hard-sphere model, a cubic box with a preset void volume is filled with a chosen number of spheres and a predetermined surface area. The spheres are

(CH3(CH2)7(CH3)2SiOCH3)) both from Fluka, represented as molecules A or B, respectively, in Figure 1. CO2 (Carburos Metálicos S.A., 99.995 wt %) was used as the solvent for silanization. 2.2. Synthesis of Silane Postfunctionalized Matrices. A series of experiments was performed to prepare silanized hybrid materials with the characteristics required for comparison with the generated models. The supercritical silanization of the substrates was performed in batch mode in a high-pressure equipment described elsewhere.8 The alkylsilanes used have a relatively high solubility in scCO2 under the experimental conditions employed. Hence, experiments were performed by charging the reactor with the solid substrate, the liquid silane, and scCO2 at the desired pressure (P) and temperature (T) during a reaction time (t) (Table 1). Recovered samples after depressurization were washed with a continuous flow of scCO2 at 10 MPa and 45 °C during 30 min to remove the excess of deposited silane. 2.3. Characterization. The thermal behavior of raw and silane-treated samples as well as the quantification of the impregnated amount of silane were studied by thermogravimetric analysis (TGA). Measurements were performed in a N2 atmosphere at a heating rate of 10 °C min−1 by using a TGA TA Instruments Q5000 IR. Textural characteristics and adsorption behavior of pristine and silanized substrates were studied by low-temperature N2 adsorption at 77 K using a Micromeritics ASAP 2000 system. Prior to the measurements, mesoporous SG40 samples were dried under reduced pressure at 120 °C for 24 h, whereas ZY samples containing micropores were dried at 150 °C during 48 h. The surface area (SBET) was determined using the BET (Brunauer−Emmett−Teller) equation.25 The classical pore size model developed by Barret−Joyner−Halenda (BJH) was used for the calculation of the pore volume (VBJH) over the mesopore, using the adsorption branch of the isotherm. The displayed pore diameter (Dp) in Table 1 corresponds to the average pore diameter calculated from the equation: 4 × VBJH/SBET. The micropore volume (Vmp) was estimated by the t plot (Harkins− Jura equation).

3. COMPUTATIONAL SECTION The models for the surface-modified substrates were generated in two steps. First, the geometrical structure of the porous silica skeleton was generated together with their silanol groups. 10152



Article

Figure 2. Silica gel: (a) experimental production process, (b) schematic outline of the main stages that are involved in model development, and (c) generated model (color key: Si: yellow; bridging O: red; nonbridging O: green; H: white).

placed in the box at random positions until they fit in the simulation cell, being both the dimension of the simulation cell and the radius of the sphere found by iteration. The spheres are linked using LJ interactions, allowing them to move and minimizing the energy. Because of computational constraints, the specific model in this work was constructed using only two interconnected spheres. Although this restriction may limit the validity of some quantitative aspects of the adsorption results, the model is suitable for the thoughtful analysis of obtained data, assessed by comparison with experimental observations. The hard spheres in the box were subsequently replaced by a realistic model of pregenerated amorphous silica spheres. The

primary silica nanospheres in the model were considered as constituted by nonporous vitreous silica; the initial amorphous silica blocks were taken from the Materials Studio structures database.26 The primary silica spheres were built by carving out from the bulk vitreous silica model using a random central point and the previously calculated radius. The surface of the silica spheres obtained in this manner is composed of oxygen atoms bonded to two silicon atoms and oxygen bonded to a single silicon atom. Single-bonded nonbridging oxygens were connected to hydrogen atoms and represent the hydroxyl silanol groups, which normally exist on the surface of amorphous silica. Figure 2c shows a snapshot of the generated 10153



Article

structure, but the adsorbed alkylsilane molecules should have certain mobility in regard to the torsion and bending of the chains. The torsion and bending angles in the surface groups were handled using a coupled-decoupled configurational bias (CDCB) algorithm.32 Additionally, a pregenerated Gaussian distribution for the probabilities of generating the bending and torsion angles for the grafted molecules was used.33 The full details of the functionalization method can be found elsewhere.24 3.3. N2 Adsorption Simulations. N2 adsorption was studied in the pregenerated model materials, both pristine and hybridized with surface groups. Simulations were carried out using the Monte Carlo method, equilibrating the adsorbent with a reservoir of N2 and exchanging and moving molecules in the grand canonical ensemble (GCMC). The adsorption isotherms were calculated by simulating the average number of adsorbed N2 molecules at different sets of bulk pressure. The soft-SAFT equation of state was used to relate the pressure of the bulk fluid to the chemical potential of the adsorbate.34,35 Once the equilibrium was reached, the amount of adsorbed molecules at this bulk pressure was averaged over a number of GCMC trials. For each value of pressure, 2.0 × 107 trials were used for equilibration and 1.4 × 107 steps for data collection. The adsorption simulations took into account that in the simulation cell, periodic in three dimensions, the only rigid parts of the modeled hybrid material were the silica spheres. Conversely, the adsorbed surface chains had one end positionally fixed to the silica surface, but bending and torsion movements during the N2 adsorption were allowed. The positions of the different atoms of the silane chain were recalculated during the adsorbate equilibration using the CDCB algorithm. Finally, the adsorbate molecules were allowed to displace to new positions as well as to enter and leave the cell; this was simulated by the insertion and deletion of fluid molecules. To avoid N2 molecules being adsorbed inside the dense silica skeleton, the insertion and deletion of fluid molecules was restricted to the open pore space using a cavity bias.36 In addition to the atomic and geometrical considerations for the adsorption model, the intermolecular interactions among adsorbate molecules, silane chains, and silica surface were input parameters in the simulation. The intermolecular interactions were calculated through pair-wise-additive 12−6 LJ potentials for the repulsive and dispersive terms and Coulombic potentials for the first-order electrostatic contribution.37 The interaction parameters between LJ sites were computed according to the Lorentz−Berthelot combination rules. The interactions of the N2 molecules were modeled using the TraPPE potential.38 N2 molecules were taken to be rigid, with an N−N bond length of 0.11 nm. The parameters for the intermolecular interaction of the fluid are listed in Table 2. The parameters for the siloxane part of the organic chain are taken from the MM2 force field for silane compounds.39 The parameters for the rest of the CHn organic groups are taken from the TraPPE force field. The charges in the surface chains were adjusted to maintain electrical neutrality in the simulation cell. The parameters for the intramolecular energy of the impregnated zeolite Y with octyltriethoxysilane (Figure S1) are listed in Tables S1−S4, while the parameters for the intramolecular energy of the grafted hydrolyzed silane (Figure S2) in silica gel are listed in Tables S5−S8 (all presented in the Supporting Information).

silica structure. The textural parameters of silica gel 40 (sample 1-SG40 in Table 1) were used for computation of the model. Values of cell and sphere dimensions were found by iteration resulting in 9.07 nm length and 3.5 nm radius, respectively. The calculated silanol density in the model was 4.5 OHnm−2. The LJ parameters for the solid silica atoms were obtained from the works of MacElroy et al.19,30 in amorphous silica. These parameters have been previously used for the calculation of diffusion of different gases on models of silica gel.30 The point charges for the silica were calculated by Brodka et al.31 from semiempirical calculations for silica clusters. Moreover, these parameters have been previously used to simulate the adsorption of gases on amorphous silica.21 The parameters for the intermolecular interaction parameters of the silica gel are given in Table 2. Table 2. Lennard-Jones Interaction Parameters for the NonBonded Interaction of Silica Gel and Nitrogen in the GCMC Simulation of Adsorption site

a

σ (nm)

Si Obridging Ononbridging H

0 0.2708 0.3000 0

N COMa

0.331 0.0

ε/kB (K) Silica Gel 0 228.4 228.4 0 N2 36.0 0.0

q (e)

ref

1.283 −0.629 −0.533 0.206

30,31 30,31 30,31 30,31

−0.482 +0.964

38 38

COM: center of mass of the nitrogen molecule.

3.2. Functionalization with Silane Groups. The driving force for functionalization was selected between impregnation of nonhydrolyzed species A or B and the surface reaction with hydrolyzed forms C or D (Figure 1). The particular driving force selected for generating the models drastically depended on the concentration of surface silanols on the support. 3.2.1. Impregnation in Microporous Zeolite Y. In a perfect zeolite crystal, the internal pore surface is electrically neutral, and no silanols would be present. Hence, silane addition in this substrate would be considered as an impregnation process and emulated by performing GCMC simulations of the adsorption and desorption of nonhydrolyzed species. 3.2.2. Grafting on Mesoporous Silica Gel. Silica gel materials are characterized by having a large number of reactive silanol groups on the internal pore surface. Therefore, for these materials, it was assumed that silane functionalization was primarily driven by surface reaction. The silane chains are expected to react with the silanols tethering to the support instead of simply impregnating the surface. This hybrid material was modeled by introducing silane groups to the pregenerated model of silica gel and linking them to a certain number of silanols existing on the surface. For comparison with experimental results, the values of grafting density used to model the hybrid (0.4 and 0.6 mmol g−1 for 2-SGT and 3-SGT samples, respectively) were similar to those obtained experimentally under two different working conditions (Table 1). Each silane chain was built on the silica surface segment by segment. The oxygen atom bonded to the silica gel surface was considered as the first atom or segment in the chain (covalent grafting) after eliminating the H accompanying the silanol group, whereas the second atom was the silicon bonded to the organic chain. The silica material can be modeled as a rigid 10154



Article

Figure 3. TGA profiles of: (a) pristine substrates and (b) samples prepared by supercritical silanization with octyltriethoxysilane.

Figure 4. Results for zeolite Y: (a) octyltriethoxysilane adsorption/desorption isotherm, (b) image of the silane chains adsorbed in the crystalline framework, and N2 adsorption isotherms for the (c) pristine and (d) impregnated samples for the experimental (blue) and simulated materials (red).

4. RESULTS AND DISCUSSION The main objective of this work is to study the influence of silane functionalization on gas adsorption for different porous supports. The basic process of study is the functionalization with an alkyltrialkoxysilanes (A in Figure 1), which after hydrolysis of the alkoxy groups forms species of the type C. For comparison, the adsorption of trialkylmonoalkoxysilane (B in Figure 1), with a single hydrolyzable group that forms D, was also studied. For the systems analyzed here, the characteristics of silane interaction with the substrate, physisorption by impregnation or chemisorption by surface reaction, depended drastically on the concentration of surface silanols on the support. 4.1. Adsorptive Behavior of Zeolite Y. Silanol groups in crystalline zeolite Y are mostly present in the external surface as terminal groups. Experimentally, some internal silanols have been noticed and are taken as an indication of lattice defects. However, all together, the amount of silanols in 1-ZY sample

could be considered very small in comparison with amorphous 1-SG40, as established by TGA curves (Figure 3a). In the recorded thermograph of raw 1-ZY, the only important weight loss occurred below 200 °C, and it was attributed to desorption of physically adsorbed water. The weight loss occurring in the range 350−650 °C, assigned to silanol condensation, was insignificant. For zeolite Y due to the absence of surface silanols in the substrate, the impregnation of monomeric nonhydrolyzed species was expected to be the primary driving force for functionalization opposed to surface reaction. For the supercritically octyltriethoxysilane impregnated zeolite, TGA showed that the weight loss corresponding to silane desorption occurred mainly at temperatures below 350 °C (Figure 3b for 2-ZY sample). The loss of surface groups at this relatively low-temperature range is associated with the vaporization of physisorbed silane molecules. The TGA estimated impregnated amount was ca. 0.60 mmol g−1 (Table 1). 10155



Article

Figure 5. Experimental and simulated adsorption isotherms of N2 on pristine silica gel at 77 K for: (a) whole and (b) enlarged low pressure range.

Experimental and computed adsorption isotherms of the pristine material are both of type IV (Figure 5a), exhibiting a well-pronounced stepwise character related to a two-stage mesoporous behavior.25,40,41 In these systems, the convex part, representing the region from capillary condensation to pore filling (point D in Figure 5a), is preceded by a concave wetting transition going from an empty state to intermediate conditions characterized by layer-by-layer growth on the pore walls. The simulation yielded the absolute adsorbed amount of N2 straightforwardly correlated to the mesoporous volume inside of the cell because the model did not contemplate large mesopores or the macroporous interparticle volume. Hence, measured VBJH values for the supercritically prepared samples were used for comparison with the model. The SBET (Table 1) of model materials was calculated from the simulated adsorption isotherm applying the BET equation. The point at which pore filling occurs is controlled by the accessible porosity, the pore geometry, the pore networking, and the tortuosity. In the studied systems, the total adsorbed amount at P/P0 = 1 was higher for the synthesized material than for the simulated one (Figure 5a). Moreover, the position of total pore filling (point D in Figure 5a) occurred at lower pressure in the simulation (P/P0 ca. 0.6) than in the experiment (P/P0 ca. 0.8). Both facts indicate smaller pore size and total pore volume for the model (sample 1-SGT) with respect to the pristine material (sample 1-SG40), as reflected in the textural data (Table 1). This finding is on account of the practical limits of the simulation cell size, which did not allow the accurate replication of the complete pore size distribution (large pores). However, at P/P0 = 0.6, similar adsorbed amounts were found for the experimental and simulated isotherms. Therefore, to facilitate comparison, the adsorption at relative pressures higher than 0.6 will not be considered in further discussion, since the extra-adsorption at P/P0 > 0.6 observed for the experimental material and related to that occurring in large mesopore and macropore voids could not be represented by the model material. The microscopic details of the silica should be reflected mostly in the low-pressure range (P/P0 < 0.1) because adsorption in this region is determined by preferred adsorption sites. In mesoporous materials, the initial concave part of the adsorption isotherm is attributed to monolayer formation. Point B in Figure 5a is taken to indicate the stage at which monolayer coverage is completed and multilayer adsorption is about to begin. Figure 5b shows in more detail the low-pressure range of the isotherms. In this pressure region, the agreement between simulation and experimental data for adsorption on

The adsorptive impregnation of zeolite Y with silane A (Figure 1) was simulated by GCMC. Figure 4a shows the simulated adsorption/desorption isotherm, while Figure 4b represents an image of the chains impregnated in the crystalline zeolite. The maximum calculated loading was 0.59 mmol g−1 (sample 2-ZYT), which is a value equivalent to the one obtained experimentally (sample 2-ZY). The maximum impregnation value was limited by sterical constrains related to the length and shape of the silane molecule and the geometry of the pores and channels. The calculated adsorption/desorption isotherm (Figure 4a) indicates that the modeled 2-ZYT hybrid material was saturated at relatively low pressures, but desorbing the silane required high vacuum. Because the modeled 1-ZYT material did not contain any surface silanol group, the organic groups added were not able to form chemical bonds with the surface and were simply impregnated. However, the analysis of the desorption curve indicates that they were bonded to the surface via a strong physical interaction. Therefore, the silane impregnated inside of the zeolite would require for its removal the use of a temperature swing process. The simulated and experimental N2 adsorption isotherms for the pristine and impregnated zeolite samples (Figure 4c,d, respectively) showed an excellent agreement. The main effect of silane functionalization was the reduction of the pore volume and thus the zeolite adsorption capacity. 4.2. Adsorptive Behavior of Silica Gel. 4.2.1. Pristine Silica Gel. The model of the pristine silica gel was constructed with dimensions consistent with the structural properties of the mesoporous silica gel 40 (sample 1-SG40 in Table 1). Owing to the broad range of parameters involved when modeling the pristine amorphous silica gel, it was necessary to impose restrictions on the solid phase by limiting the number of particles in the simulation cell to two. The low number of spheres used could affect the correct reproduction of the randomness of the void space in the silica gel. Nevertheless, this was required as a compromise between the CPU requirements for the simulation and a reliable estimation of the generated pore space in the unit cell. The capability of the method to predict accurately textural and adsorption properties was evaluated by comparing N2 adsorption isotherms (at 77 K) for pristine and impregnated materials obtained by either GCMC simulation or BET measurements. The validity of the silica gel model was first verified by comparing the silanol densities in the experimental and simulated materials. From the TGA thermograph, a silanol density of 4.6 OHnm−2 was estimated for sample 1-SG40 (Figure 3a), which matches the calculated value in the model of 4.5 OHnm−2 for sample 1-SGT. 10156



Article

the pristine material was excellent. Still, the amount of adsorbed N2 was slightly larger in the experimental sample than in the simulated one, which can be corrected by applying to the simulation isotherm a scaling factor (ϕ)21,42 based on the estimated porosity values for the experiment and the model (ϕ = 0.47/0.44 in Table 1). Thereafter, the validated model of simulated silica gel can be expanded to study functionalization by considering the grafting of the silane chains inside the material. 4.2.2. Functionalized Silica Gel. In the second section of this work, the silica gel matrix was functionalized with alkylsilane chains. A supercritical CO2 silanization procedure was followed to synthesize the experimental materials. For a trialkoxysilane of the type RnSi(OR′)3 (A in Figure 1), the surface silanization reaction starts with the hydrolysis of the alkoxy OR′ groups to form silanols (RnSi(OR′)3−x(OH)x, C in Figure 1), which can then interact by hydrogen bonding with the OH groups on the surface of the porous silica material (#Si−OH···HO−Si) or with other silane molecules (Si−OH···HO−Si). Dehydration of the formed hydrogen bonds produces siloxane linkages between the silane and the surface (#Si−O−Si) and selfassembled monolayers by condensation of neighboring silane molecules (Si−O−Si). Hence, for the hydrolyzed species C, covalent grafting with the surface can be coupled to horizontal self-assembly. In the TGA plot of the supercritically silanized samples (Figure 3b), the loss of weight occurred within 350− 600 °C was related to the decomposition of strongly bonded species (highly cross-linked and/or chemisorbed silanes) by cleavage of C−C and Si−C bonds. The calculated amounts of octyltrisilanol that reacted with the substrate at two different experimental pressures were 0.42 and 0.57 mmol g−1 for the samples 2-SG40 and 3-SG40, respectively (Table 1). Experimental observations indicate that the highly reactive surface of amorphous silica gel might assist the formation of covalently bonded silanes, as surface reaction is the primary driving force for absorption versus impregnation. Moreover, the mesoporous space allows for the self-condensation reaction between adjacent silanols. Even though the steps of the experimental silanization process are well known, the complexity of the reaction makes it necessary to simplify the chemistry employed to model the functionalization. The most used generalization consists of not taking into account the formation of siloxane bridges between neighboring silane chains.21−23,43 This assumption was also employed in this work. The hybrid simulated material was generated starting from the O atom covalently bonded to the silica surface, whereas the second atom was the Si bonded to the organic chain, which was added segment by segment. Figure 6 shows an image of the resulting octylsilane chains grafted on the amorphous silica gel. The molecules were always required to react in a monodentate covalent configuration with the surface and never with neighboring silane molecules. The designed algorithm is expected to allow analyzing and understanding simultaneously the surface of silica functionalized with both tri- and monoalkoxysilanes. Actually, monodentate bonding is the expected natural experimental behavior for trialkylmonoalkoxysilane molecules with a single hydrolyzable group (B in Figure 1). After hydrolysis, octyldimethylmethoxysilane generates the form D, which reacts with the hydroxylated surface (#SiR2−O− Si). For ease of comparison, a silanized 4-SG40 sample was prepared in the laboratory with a monoalkoxysilane, containing a grafting density of 0.59 estimated by TGA (Table 1).

Figure 6. Snapshot of hydrolyzed octyltriethoxysilane functionalized silica gel (similar color key of Figure 2).

The adsorption isotherms of silica gel samples functionalized with silanes, measured for the synthesized products (C or D) and calculated with GCMC for the simulated materials (C), reflected many properties of the porous hybrid materials. The available space for N2 adsorption, measured as the VBJH, diminished with functionalization for all studied materials (Table 1). The reduction of mesopore volume was similar for the experimental (Figure 7a) and model (Figure 7b) materials. The VBJH was reduced with respect to pristine silica gel in ca. 16% by functionalizing with ca. 0.4 mmol g−1 of silane, whereas a decrease of ca. 25% was estimated for deposited silane concentrations of ca. 0.6 mmol g−1. Hence, the model was able to simulate accurately the reduction in the mesopore adsorption capacity due to the increasing void space occupied by hydrolyzed silane as the concentration increased. Conversely, the simulation slightly overpredicts the adsorption in the low-pressure region (Figure 7c,d for 0.4 and 0.6 mmol g−1 functionalization, respectively), which is the opposite behavior to that observed for the pristine materials (Figure 5b). For trialkoxysilanes, the overprediction likely reflects the absence of horizontal self-assembly in the simulated product, the importance of which increases with the functionalization degree. Whereas in the simulation the microvoids created among the randomly distributed chains can act as strong adsorption sites (Figure 8a2), the chains in the experimental material are horizontally self-assembled, leading to a more compact packing (Figure 8a1). Significant differences between the chemisorbed monoalkoxyde and the model were also observed (Figure 7d). The structural monodentate absorption in D (Figure 1) could be considered to be similar to that adopted in the model (Figure 8b and a2, respectively). However, D exposes lateral methyl groups (CH3) among chains, which shield the polar surface of the metal oxide. Therefore, the methyl groups act as a structural barrier that prevents N2 molecules interaction with the surface (Figure 8b), resulting in low adsorption values in the low pressure range.15,44 The lateral silanol moieties present in the model likely acted as stronger adsorption sites than the CH3 in the monoalkoxide. 10157



Article

Figure 7. Experimental and simulated adsorption isotherms of N2 on functionalized silica gel at 77 K for: (a) experimental materials, (b) modeled materials, and experimental and simulated materials loaded with (c) 0.4 mmol g−1 and (d) 0.6 mmol g−1 in the enlarged low-pressure range.

Figure 8. Schematic representation of the different functionalization possibilities for species: (a) A and (b) B in Figure 1. 10158



Article

Figure 9. Schematic models used in samples: (a) 2-SGT and 3-SGT and (b) 4-SGT0.2 and 5-SGT0.2. Experimental and simulated adsorption isotherms of N2 on silica gel at 77 K for samples functionalized with (c) 0.4 mmol g−1 and (d) 0.6 mmol g−1 of silane (species C in Figure 1).

Figure 10. Degree of functionalization as a function of the number of computational cycles required for the grafting simulation for: (a) octyltrialkoxysilane and (b) trialkoxysilanes with different tail lengths (ethyl 2C, butyl 4C, hexyl 6C, and octyl 8C).

original one. The alkyl chains were grafted in concentrations of either 0.4 or 0.6 mmol g−1 (samples 4-SGT0.2 and 5-SGT0.2, respectively) to the generated spheres of 3.7 nm radius. Schematic representations of the experimental and 0.2 nm model used are shown in Figure 9a,b, respectively. For these samples, the reduction in adsorption capacity with respect to sample 1-SGT was on the order of 33 and 38% for concentrations of silane chains of 0.4 and 0.6 mmol g−1, respectively (Figure 9c,d, respectively). These percentages are not comparable to the reductions observed in the experimental materials, which were on the order of 15 and 25% for samples 2-SG40 and 3-SG40, respectively. Hence, the macroscopic adsorption behavior at different functionalization degrees was not accurately described by the silica gel model with a larger radius of the spheres. Using the approach of increasing the silica radius sphere in 0.2 nm, it was considered that the functionalized hybrid was constituted by a continuous film of −Si−O-Si-O- around all primary spheres (Figure 9b). Therefore, the accuracy of the results obtained using this approach would be higher for systems with high degrees of functionalization, that is, near a monolayer. Note that at grafting densities of 0.4 mmol g−1, the deviation between

The macroscopic consequences found from this analysis were similar values of pore volume for both experimental and simulated materials but significant differences in the surface area (Table 1). The BET surface area decreased after silane adsorption for laboratory obtained samples and slightly increased for modeled compounds, indicating a large proportion of micropores in the simulated model. Geometrically the smallest pores have little contribution to the pore volume but a large contribution to the surface area. The extra micropores could be generated in the simulated material after functionalization of the mesopores. The reduction in the mean pore size was significant only for the simulated materials. In a different approach, and following the work of Schumacher et al.,21,42 the order in the −Si−O-Si-O- horizontal polymerization observed experimentally for trialkoxysilanes was simulated by adding an extra layer of SiO2 to the silica primary particles model, then building on the silica surface only the organic chain of the silane molecule. To compare both models, we calculated a new set of simulations, generating an additional amorphous silica gel model with the configuration of 1-SGT but with a radius 0.2 nm higher (the van der Waals radius of silicon and oxygen is 0.038 and 0.152 nm, respectively) than the 10159



Article

moiety, at least for concentrations lower than those corresponding to monolayer formation.

macroscopic adsorption values of the experimental and the simulated material was on the order of 55% (compare samples 2-SG40 with 4-SGT0.2), whereas this value was reduced to 35% by increasing the functionalization degree to 0.6 mmol g−1 (compare samples 3-SG40 with 5-SGT0.2). In contrast, the deviation

Alkylsilane-Functionalized Microporous and Mesoporous Materials

Recommend Documents