Acidity Characterization of Amorphous Silica–Alumina - The Journal of

Sep 18, 2012 - The general procedure for alkylamine TPD, which probes both Brønsted and Lewis acidity, is modified to increase the selectivity to str...
0 downloads 0 Views 1MB Size
Article pubs.acs.org/JPCC

Acidity Characterization of Amorphous Silica−Alumina Emiel J.M. Hensen,*,† Dilip G. Poduval,† Volkan Degirmenci, D.A J. Michel Ligthart,† Wenbin Chen,‡ Françoise Maugé,‡ Marcello S. Rigutto,§ and J.A. Rob van Veen† †

Laboratory of Inorganic Materials Chemistry, Schuit Institute of Catalysis, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands ‡ Laboratoire Catalyse et Spectrochimie, UMR CNRS-ENSICAEN, University of Caen, 6 Bd du Maréchal Juin, 14050 Caen Cedex, France § Shell Global Solutions International B.V., P.O. Box 38000, 1030 BN Amsterdam, The Netherlands S Supporting Information *

ABSTRACT: Surface characterization of amorphous silica− alumina (ASA) by COads IR, pyridineads IR, alkylamine temperature-programmed desorption (TPD), Cs+ and Cu(EDA)22+ exchange, 1H NMR, and m-xylene isomerization points to the presence of a broad range of Brønsted and Lewis acid sites. Careful interpretation of IR spectra of adsorbed CO or pyridine confirms the presence of a few very strong Brønsted acid sites (BAS), typically at concentrations lower than 10 μmol/g. The general procedure for alkylamine TPD, which probes both Brønsted and Lewis acidity, is modified to increase the selectivity to strong Brønsted acid sites. Poisoning of the m-xylene isomerization reaction by a base is presented as a novel method to quantify strong BAS. The surface also contains a weaker form of BAS, in concentrations between 50 and 150 μmol/g, which can be quantified by COads IR. Cu(EDA)22+ exchange also probes these sites. The structure of these sites remains unclear, but they might arise from the interaction of silanol groups with strong Lewis acid Al3+ sites. The surface also contains nonacidic aluminol and silanol sites (200−400 μmol/g) and two forms of Lewis acid sites: (i) a weaker form associated with segregated alumina domains containing five-coordinated Al, which make up the interface between these domains and the ASA phase and (ii) a stronger form, which are undercoordinated Al sites grafted onto the silica surface. The acid catalytic activity in bifunctional nheptane hydroconversion correlates with the concentration of strong BAS. The influence of the support electronegativity on the neopentane hydrogenolysis activity of supported Pt catalysts is considerably larger than that of the support Brønsted acidity. It is argued that strong Lewis acid sites, which are present in ASA but not in γ-alumina, are essential to transmit the Sanderson electronegativity of the oxide support to the active Pt phase.

1. INTRODUCTION

would be that the surface contains a large number of weakly acidic hydroxyl groups. A few general remarks about the surface heterogeneity of ASA are in order before discussing the origin of its acidity. Even in a typical amorphous support as silica the hydroxyl groups are present in a quite heterogeneous environment,10 so that one cannot expect to have well-defined sites in an amorphous mixed oxide such as ASA. A recently proposed surface model of ASA contains at least four different types of aluminum: (i) Al substitutions in the silica network and, at the surface, isolated (ii) tetrahedral and (iii) octahedral Al as well as (iv) more aggregated forms of Al.11 The aggregation of aluminum may be so extensive that it is better to consider the presence of a separate alumina phase. It was suggested that five-coordinated Al connects these segregated domains to the mixed silica− alumina phase.11−13 A challenge has been to resolve the

Amorphous silica−alumina (ASA) is of considerable practical importance in industrial catalysis. It is widely used as a solid acid or serves as a carrier material for finely dispersed metal sulfides or metals in a wide range of processes.1−3 ASA is often present in modern composite hydrocracking catalysts to produce middle distillates from heavy oil fractions.2 ASA is a much weaker acid catalyst than zeolite. It is well established that the strong Brønsted acidity of zeolites resides in the bridging hydroxyl group originating from the replacement of Si4+ in their crystalline framework by Al3+. Mixed silica−aluminas lack such crystallinity. A usual assumption is that ASA contains sites that are similar to the Brønsted acid sites (BAS) in zeolites but, somehow, of lower strength. Hydrocarbon cracking experiments of Haag et al.4 suggested that the low acidity of ASA might be the consequence of a very low concentration of sites of similar acidity as the bridging hydroxyl groups in zeolites. Until now, this issue has not been settled yet despite important contributions in literature.5−9 The alternative interpretation © 2012 American Chemical Society

Received: July 13, 2012 Revised: September 18, 2012 Published: September 18, 2012 21416

dx.doi.org/10.1021/jp309182f | J. Phys. Chem. C 2012, 116, 21416−21429

The Journal of Physical Chemistry C

Article

structure of the BAS in ASA in the absence of any direct spectroscopic observation of these sites. Typically, infrared spectra of ASA only contain a single sharp band due to silanol groups. Recently, selective H/D exchange combined with IR spectroscopy has been successfully employed to show that ASA contains a few strong BAS.14,15 The BAS density of zeolites, clays, and ASA of varying composition determined in this way correlates well with their acid catalytic activities. Hence, a preliminary conclusion has been that the strong acid sites in ASA are of comparable strength as those in zeolites and clays. As characterization of the surface acidity of ASA has already been extensively addressed before, it is useful to compare this new method to some of the more established ones. A widely used method to probe acidity in aluminosilicates is IR spectroscopy of adsorbed pyridine (pyridineads IR).16,17 The use of IR with adsorbed CO (COads IR) as a probe for strong acidity in zeolites is well established, and it certainly seems to have some merit in the characterization of ASA.7,18 The presence of BAS of modest acidity is apparent from the work of Cairon et al.19 On the basis of OH frequency shifts, Crépeau et al.7 have evidenced that ASA contains strong BAS, yet quantification by this method has proven difficult. Another commonly employed technique is temperature-programmed desorption (TPD) of bases,20 and especially the group of Gorte21−23 has successfully employed alkylamines to investigate zeolite acidity. Exchange of BAS with Cs+24 and Cu(ethylenediamine)22+25 might also be useful. Obviously, 1H MAS NMR has also been used in an attempt to probe strong acidity in ASA.26 By investigating how results for various methods compare with respect to the acid strength and their quantification, this work will address two questions, namely: (i) Is there a welldefined set of strong acid sites in ASA? and (ii) Can we learn anything specific about the surface heterogeneity of ASA? The results of various characterization methods will be compared to the results of FTIR spectroscopy of H/D exchanged ASAs of which the surface composition is known in detail11 as well as, where necessary, to zeolites and clays. The present set of ASAs has already been characterized in an acid catalytic test, viz., the hydroconversion of n-heptane.11 The catalytic characterization of acidity is extended here by measuring the inhibition by lutidine of the activity of some ASAs in the liquid phase isomerization of m-xylene. An attempt will be made to formulate a model for the surface acidity of ASA. Finally, we will present results of Pt-catalyzed neopentane hydrogenolysis, which point to a strong effect of the composition of the support. (The bulk of the work has been done in Eindhoven, but some aspects of COads IR and lutidine adsorption have been checked at Caen during a stay of one of us (JARV) there in November 2008 (footnote).)

Table 1. List of Aluminosilicates, Their Silica-to-Alumina Ratio (SAR), and Origin sample

SAR

origin

HY (5) USY (9.6, F)

5 9.6

USY (8.1) USY (9.3) VUSY (26.3) VUSY (33.1) XVUSY (70) XVUSY (85) MgSAP (33) MgSAP (13) ASA (5/95,x)b ASA (10/90,x)b ASA (15/85,x)b ASA (20/80,x)b ASA (5/95,3,x)b

8.1 9.3 26.3 33.1 70 85 33 13 32 15 9.6 6.8 38

ASA (10/90,3,x)b

17

faujasite zeolite (Akzo-Nobel, PA 73022) partial removal of framework Al from faujasite via (NH4)2SiF6 treatmenta steam calcined faujasite steam calcined and acid leached faujasite steam calcined and acid leached faujasite steam calcined and acid leached faujasite steam calcined and acid leached faujasite steam calcined and acid leached faujasite saponite clay (Kunimine Industries) saponite claya homogeneous deposition route homogeneous deposition route homogeneous deposition route homogeneous deposition route homogeneous deposition route, stopped at pH = 3 homogeneous deposition route, stopped at pH = 3 homogeneous deposition route, stopped at pH = 3 ASA synthesized by cogelationa

ASA (15/85,3,x)b ASA (5/95,Cogel)

12 c

32

a

Synthesized according to ref 15. bSynthesized according to ref 11, x stands for calcination temperature (773 or 1073 K). cCalcined at 923 K.

CK-300) and two silicas (Degussa, Sipernat 50 and an ultrapure silica) were also included as support materials. Several zeolites including HZSM-5 (Akzo Nobel) and ultrastabilized Y zeolites (Zeolyst International) were used as received. A stable Y zeolite was prepared through treatment of a faujasite zeolite (SAR = 5) with (NH4)2SiF6, which replaces part of the framework aluminum by silicon. Finally, two saponite clays (a commercial Mg-saponite clay material from Kunimine Industries and a Mgsaponite prepared by precipitation) were also included. Details of the acidic properties of these materials can be found elsewhere.14,15 2.2. Characterization. 2.2.1. COads IR. Infrared spectra (1200−4000 cm−1) were recorded in transmission mode in a Bruker IFS-113v FTIR spectrometer equipped with a DTGS detector. The catalyst was pressed into a self-supporting wafer (density ∼9 mg/cm2), which was placed in a controlled atmosphere transmission cell equipped with CaF2 windows. Prior to CO adsorption, the sample was heated to 823 K at a rate of 5 K/min in flowing oxygen. The sample was then kept at this temperature for 2 h and evacuated before cooling to room temperature. At room temperature, the pressure in the cell was lower than 10−6 mbar. Special measures were taken to remove residual water from the cell by heating its stainless steel body during calcination of the sample. The sample was then cooled by flowing liquid nitrogen through a capillary spiralled around the catalyst wafer. The final temperature was around 90 K. At this stage, an initial spectrum was recorded. Carbon monoxide (Praxair, 99.999%) was dosed via a sample loop connected to a six-way valve. In this manner, accurate doses of 0.04 μmol CO were administered to the cell. Dosing was carried out by a computer-controlled sequence in which a spectrum was recorded 30 s after each dose. Each spectrum was recorded by accumulating 20 scans at a resolution of 4 cm−1. Difference spectra were obtained by subtracting the initial spectrum of the

2. EXPERIMENTAL SECTION 2.1. Materials. Table 1 lists the aluminosilicates employed in this study. The preparation of these ASAs has been described elsewhere.11 Briefly, ASA was prepared by homogeneous deposition of aluminum on silica or by cogelation of a mixture of sodium silicate and aluminum chloride. Two sets of materials have been obtained from the first method with calcination at either 773 or 1073 K. The ASA obtained by cogelation was calcined at 923 K. Subsequent pretreatment for further characterization (see section 2.2) was done to dry the ASAs and was carried out at lower temperature, i.e., the standard for each of the methods employed in this study. γ-Alumina (Ketjen, 21417

dx.doi.org/10.1021/jp309182f | J. Phys. Chem. C 2012, 116, 21416−21429

The Journal of Physical Chemistry C

Article

recovered by centrifugation. The initial and final Cu 2+ concentration were determined by inductively coupled plasma measurements. 2.2.6. 1H Magic Angle Spinning (MAS) NMR. Prior to 1H MAS NMR measurement, a known amount of ASA was introduced in a special glass tube (Wilmad Glass Company) suitable for NMR measurements. After connecting the tube to a vacuum line, the sample was heated at a final pressure lower than 10−5 mbar from room temperature to 723 K at a rate of 10 K/min. The sample was kept at 723 K for 4 h. After cooling to room temperature, the glass ampule was sealed and placed in a 4 mm MAS rotor. 1H MAS NMR measurements were performed at room temperature using a Bruker AVANCE 500 MHz spectrometer. The MAS rate was 7.5 kHz. The inversion recovery pulse sequence was applied with a recycle delay of 80 s. The sample amount for NMR measurements was corrected for the initial water content of the ASA samples as determined by TGA analysis. Water was used as the reference for computing the absolute amount of protons. 2.3. Catalytic Activity Measurements. 2.3.1. Isomerization of m-Xylene. Isomerization of m-xylene was carried out in the liquid phase in a plug-flow reactor. About 1 g of aluminosilicate catalyst with a sieve fraction between 125 and 250 μm was contained between two layers of SiC in a stainless steel reactor with an average grain size of 180 μm. The sample was first dehydrated in 100 mL/min He with heating from room temperature to 823 K followed by an isothermal period of 2 h. Then, the reactor was cooled to the reaction temperature of 573 K. Liquid m-xylene was fed to the reactor by a highperformance liquid chromatography pump at a flow rate of 50 μL/min. The reaction pressure was increased and kept at 80 bar by a back-pressure regulator. Prior to use, m-xylene (Merck, purity 99%) was percolated through activated alumina (Merck) to remove peroxide impurities. To facilitate accurate gas chromatographic analysis, the effluent stream was diluted with n-heptane delivered by a second pump at a flow rate of 0.8 mL/ min. This mixture was analyzed by a HP-5890 gas chromatograph equipped with a Stabilwax column (30 m × 0.32 mm id, df = 0.5 μm) and a flame ionization detector. Poisoning of the reaction was done by introducing 2,6-dimethylpyridine (Alfa Aesar, 99%) dissolved in m-xylene via a sample loop into the feed mixture. The reaction was monitored until a stable value was attained. In a typical experiment, 2,6-dimethylpyridine was injected three times in this manner. 2.3.2. Hydrogenolysis of Neopentane. The hydrogenolysis of neopentane was carried out in a gas-phase parallel ten-flow microflowreactor system. About 100 mg of Pt-loaded catalyst was held between two quartz wool plugs in a quartz tube with an internal diameter of 4 mm. Prior to reaction, the catalysts were reduced in a mixture of 20 vol % H2 in He at a flow rate of 50 mL/min while ramping the temperature from ambient to 573 K at a rate of 10 K/min. The catalysts were then exposed to a mixture of 1 vol % neopentane in H2 at a flow rate of 50 mL/min. The reactor effluent was analyzed by online gas chromatography (Interscience CompactGC, Plot KCl/Al2O3, 10 m × 0.32 mm, df = 1 μm, TCD). The main reaction products were isobutane and methane with very small amounts of olefinic byproduct. The reaction rate was calculated on the basis of the neopentane conversion, which was lower than 8% in all cases. It was verified that the acidic supports without Pt did not convert neopentane under the reaction conditions. Pt was introduced by pore volume impregnation of the dried supports with a solution of Pt(NH3)4(NO3)2 to achieve a final

dehydrated catalyst from the spectra obtained at increasing CO coverage. 2.2.2. Pyridineads IR. A Perkin-Elmer 2000 spectrometer operating at a 4 cm−1 resolution was used to obtain infrared spectra of adsorbed pyridine. Prior to recording spectra, the sample was evacuated in vacuum (pressure lower than 10−6 mbar) at 673 K for 1 h. A background spectrum was recorded at 423 K. The sample was then exposed to 10−2 mbar pyridine at 423 K for 30 min. After removing the excess adsorbate by outgassing at the same temperature for 1 h, a first spectrum was recorded. Subsequently, the sample was heated to 673 K, kept at this temperature for 1 h, and then cooled to 423 K. At this point, a second spectrum was recorded. A comparison of the spectra before and after thermal treatment gave the amount of weakly adsorbed pyridine. 2.2.3. TPD of Isopropylamine. TPD of isopropylamine (IPAm) was carried out in a plug-flow quartz reactor. An amount of catalyst (1 g for ASA or 100 mg for zeolite, sieve fraction 125−250 μm) was loaded into a quartz reactor tube and kept between two layers of quartz wool. Prior to IPAm adsorption, the sample was calcined in a mixture of 20 vol % O2 in He at a total flow rate of 100 mL/min, while heating from room temperature to 823 K at a rate of 5 K/min (2 K/min for zeolites) followed by an isothermal period of 1 h. The temperature was then lowered to 373 K at a rate of 15 K/min. At 373 K the sample was exposed to IPAm by passing 100 mL/ min He through a saturator containing isopropylamine (Merck, purity 99%) at a partial pressure of 26 kPa. The flow of IPAm in He was led over the catalyst for 10 min. Subsequently, physisorbed IPAm was removed by purging with He at 100 mL/min for 16 h. The removal of IPAm was monitored by online mass spectrometry (quadrupole mass spectrometer, Balzers TPG-300). TPD of IPAm was started by heating the sample at a rate of 5 K/min to 773 K. The number of acidic sites was calculated from the number of propene molecules from IPAm decomposition. The total amount of propene was corrected for the fragmentation in the mass spectrometer of IPAm that was molecularly desorbed. The weight-based acid site densities were calculated taking into account physisorbed water in the samples. 2.2.4. Ion Exchange with Cs+. The exchange of strong BAS with the soft Cs+ cation was investigated. To this end, solutions of varying CsCl (Merck, purity 99.9%) concentration were prepared from an original mother solution containing 1.6 mg Cs/g. After the initial pH of the solution was recorded, 1 g of ASA previously dried at 573 K was suspended in the solution. After 5 min, the pH was measured again. Subsequently, the flasks were gently shaken for 1 h. The pH was once again recorded at the end of the experiment. The solutions were then recovered by filtration using a filter (Millipore 0.75 μm). The initial and final Cs+ concentration were determined by inductively coupled plasma measurements. Control experiments were also carried out on pure γ-alumina (Ketjen CK300) and pure silica (Sipernat-50). 2.2.5. Ion Exchange with Cu-EDA. The exchange of hydroxyl groups of strong BAS was investigated by exchange with the Cu(EDA)22+ complex. Initially, 1 M solutions of CuCl2·2H20 (Merck, purity 98%) and ethylene diamine (EDA) (Merck) were prepared. The two solutions were then mixed in a 1:2 ratio by volume (CuCl2:EDA), and from the resulting solution, a 0.03 M solution was prepared by dilution. ASA (0.25 g) previously dried at 573 K was then suspended in the solution. The flask was shaken for 30 min, and the solution was 21418

dx.doi.org/10.1021/jp309182f | J. Phys. Chem. C 2012, 116, 21416−21429

The Journal of Physical Chemistry C

Article

(LAS). Bands at 2158, 2141, and 2130 cm−1 relate, respectively, to CO coordinating to silanols, physisorbed CO, and CO coordinating to the surface via its oxygen atom. The spectra in the hydroxyl region (Figure 1b) initially show a negative feature at relatively low CO coverage due to perturbation of sites of enhanced acidity in the supercages around 3604 cm−1. With increasing CO coverage, this feature becomes a shoulder of the much more pronounced negative band at 3630 cm−1 due to perturbation of high-frequency (HF) OH groups in the supercages. The shift of the OH stretch upon interaction with CO (ΔνOH) is 360 cm−1 and 416 cm−1 for regular and enhanced sites, respectively. The higher shift for the enhanced sites is in line with their higher acidity, which has been related to the close proximinty of extraframework Al.19,27,28 The lowfrequency (LF) OH groups at 3566 cm−1 are not perturbed by CO, because CO cannot access the smaller cages. Further OH groups at 3677 and 3745 cm−1 interact with CO at higher coverage. Cairon et al. have related the presence of perturbed OH bands in the region of 3440−3475 cm−1 for Y zeolite to acidic sites of an extraframework silica−alumina phase. These bands are also observed in Figure 1. The frequency shift of about 190 cm−1 points to their intermediate acidity. Therefore, the band in the CO region at 2171 cm−1 should be linked to this weaker type of BAS. COads IR spectra of two USY zeolites with a lower Al content and MgSAP(33) are briefly discussed in the Supporting Information. Figure 2 shows IR spectra of adsorbed CO on ASA (5/95, 1073). After the first CO dose, a band at 2230 cm−1 appears, which corresponds to CO coordinating to strong LAS and a very small feature around 2177 cm−1. In the difference spectrum of the hydroxyl region (Figure 2b) a weak broad feature around 3320 cm−1 becomes visible along with a significant negative−positive feature around 3750 cm−1. The band at 2177 cm−1 grows with increasing CO coverage and shifts to 2175 cm−1. Additional bands are observed at 2158 and 2190 cm−1. The bands at 2190 and 2230 cm−1 are due to CO interacting with coordinatively unsaturated Al3+ sites, the former being of lower strength than the latter.27,29,30 The bands at 2158 and 2177 cm−1 are due to CO perturbation of silanol groups and more acidic hydroxyl groups, respectively. Examination of the perturbed hydroxyl region shows the development of broad bands at 3320, 3405, 3550, and 3670

metal loading of 0.8 wt %. After impregnation the catalysts were dried at room temperature of 1 h, dried overnight at 393 K, and calcined at 723 K for 2 h in static air. The supports form a subset of ASA samples described earlier,11,14 having equal Brønsted acidity. For comparison, SiO2 (Degussa, 190 m2/g, Al content below 0.01 wt %), SiO2:Al (Shell, 196 m2/g, Al content 0.5 wt %), γ-Al2O3 (Ketjen, 267 m2/g), and an industrial ASA (Shell, 55 wt % Al2O3, 375 m2/g) were included in these catalytic tests. The Pt metal particle size of the reduced catalysts was measured by H2 chemisorption by use of the double isotherm method on a Micromeritics ASAP2020.

3. RESULTS AND DISCUSSION 3.1. COads IR. The IR spectra upon CO adsorption on an ultrastabilized Y zeolite with a SAR of 8.1 will serve as a reference for our discussion. The first bands to appear in the carbonyl region (Figure 1a) at 2170 and 2180 cm−1 are

Figure 1. Carbonyl (left) and hydroxyl (right) stretch regions of the infrared spectra of well-dehydrated USY (8.1) at 80 K as a function of the CO coverage. The inset shows the spectrum after the first dose of CO.

respectively due to CO coordinating to the weak and strong acidic hydroxyl groups.19 At higher coverage, shoulders at 2158 and 2190 cm−1 are discerned. After saturation of the band at 2180 cm−1 additional bands at 2141 and 2130 cm−1 appear. The band just below 2200 cm−1 is due to weak Lewis acid sites

Figure 2. Carbonyl (left) and hydroxyl (right) stretch regions of the infrared spectra of well-dehydrated ASA (5/95, 1073) at 80 K as a function of the CO coverage. 21419

dx.doi.org/10.1021/jp309182f | J. Phys. Chem. C 2012, 116, 21416−21429

The Journal of Physical Chemistry C

Article

cm−1. The latter three bands were already discussed by Cairon et al.19 and Crépeau et al.7 The band at 3670 cm−1 is assigned to silanol groups perturbed by CO. The bands at 3405 and 3550 cm−1 point to the presence of OH groups with a higher acidity than silanols. Crépeau et al. concluded that ASA contains strong BAS in the form of paired (SiOH, Al) sites based on the correlation between the perturbed OH band at 3382 cm−1 and the CO stretching band at 2178 cm−1.7 The strong negative band at 3750 cm−1, which is also observed here, was explained by closure of the paired site upon CO adsorption as illustrated in Scheme 1. This explanation seems unlikely because of the very

at 2230 cm−1. This relatively strong interaction will cause a blue shift in the frequency of the silanol groups initially perturbed by the LAS. This model explains the strong negative-positive feature in the IR spectrum. This alternative explanation is also illustrated in Scheme 1. FTIR spectroscopy of partially H/D exchanged ASA has provided strong indications that ASA contains hydroxyl groups around 3635 cm−1.14 These sites are not observed in the hydroxyl region of IR spectra of dehydrated ASA because of their low concentration. They are observed, however, as the corrresponding OD bands upon selective deuteration. The IR spectra after CO adsorption on ASA contain a rather weak feature around 3320 cm−1, which can also be discerned as a weak band in the spectra after the second and third dose. We propose that this band relates to the CO stretch band at 2177 cm−1. The negative feature arising from perturbation of the suspected band at 3635 cm−1 cannot be discerned due to the presence of other perturbed bands in this region. If the unperturbed band is located at 3635 cm−1, ΔνOH is 315 cm−1, which points to Brønsted acidity of zeolitic strength. With increasing CO coverage, the band at 2177 cm−1 shifts to 2175 cm−1, which is attributed to perturbation of additional hydroxyl groups of weaker acidity. With increasing aluminum content, the trends in the IR spectra are very similar, except that the feature around 3320 cm−1 cannot be clearly discerned anymore. This seems to be due to the increasing contribution of the broad band around 3425 cm−1. The spectra for ASA (20/80, 1073) are given in Figure 3. Similar to ASA (5/95, 1073), the first CO dose results in the appearance of a band around 2230 cm−1 and a negative− positive feature around 3750 cm−1. Increasing CO coverage results in the development of a band at 2177 cm−1 and a shift of this band to lower wavenumbers. Broad perturbed hydroxyl bands are observed at 3430 and 3560 cm−1, respectively. The perturbed band at 3660 cm−1 is more difficult to discern in the spectra of the ASA (20/80, 1073). This should be due to the decrease of the silanol concentration as a result of the much higher surface aluminum loading.11 Table 2 summarizes the positions of the bands belonging to strong BAS in aluminosilicates in the CO and perturbed OH region and the shift upon CO complexation for a large set of zeolites, clays and ASA.

Scheme 1. (top) Formation of a Bridging Hydroxyl Group in the Presence of a Base as Originally Proposed by Trombetta et al.5 and (bottom) New Proposed Effect of CO on the Weak Interaction of Silanol Groups with Lewis Acid Al3+ Sites

different intensities for the negative−positive feature and the band due to perturbed OH groups. Moreover, integration of the negative-positive feature indicates that the total intensity of the spectrum does not change, where we assume that because of the small shift in frequency, the extinction coefficient does not greatly change. Our interpretation is based on a slight red shift of part of the silanol groups (weaker O−H bond) due to the interaction with Lewis acid Al3+ sites. This agrees with the increasing red shift of the silanol band with increasing aluminum content (from 3747 cm−1 for SiO2, 3745 cm−1 for ASA (5/95, 1073) down to 3737 cm−1 for ASA (20/80, 1073)). As CO interacts more strongly with LAS than with silanol groups,5 CO exposure will first lead to the appearance of a band

Figure 3. Carbonyl (left) and hydroxyl (right) stretch regions of the infrared spectra of well-dehydrated ASA (20/80, 1073) at 80 K as a function of the CO coverage. 21420

dx.doi.org/10.1021/jp309182f | J. Phys. Chem. C 2012, 116, 21416−21429

The Journal of Physical Chemistry C

Article

Table 2. Infrared Band Positions of the OH···CO Complex in the Carbonyl and Hydroxyl Regions after Adsorption of Small Doses of Carbon Monoxide at 90 Ka sample

νCO (cm−1)

νperturbed OH (cm−1)

ΔνOH (cm−1)

HY (5) USY (9.6, F) USY (8.1) VUSY (26.5) XVUSY (70) MgSAP (13) MgSAP (33) ASA (5/95,1073) ASA (10/90,1073) ASA (15/85,1073) ASA (20/80,1073) ASA (5/95,773) ASA (5/95,3,1073) ASA (5/95,Cogel)

2178 2181 2182 2181 2180 2180 2180 2177 2178 2177 2177 2175 2177 2177

3346 3285 3270 (3190b) 3276 (3190b) 3278 3260 3268 3320 n.o.c n.o. n.o. n.o. n.o. n.o.

297 345 360 (416b) 354 (416b) 356 338 337 315 n.o. n.o. n.o. n.o. n.o. n.o.

Figure 4. Deconvolution of the carbonyl stretch region of the IR spectra of CO adsorbed to (a) USY (8.1), (b) MgSAP (33) and (c) ASA (15/85,1073) after saturation of the band due to strong BAS (indicated by arrow).

a

The shifts in the hydroxyl region (ΔνOH) are relative to the positions of the (HF)OH and (HF′)OH bands for the zeolites. bEnhanced hydroxyl groups (HF′)OH in supercages. cNot observed.

Table 3. Concentration of Strong BAS (Nstrong BAS, CO IR) for Aluminosilicates based on Saturation of Infrared Carbonyl Band around 2177−2182 cm−1a

The concentration of the various hydroxyl groups was then determined by deconvolution of the CO stretch region. The molar extinction coefficient of adsorbed CO was found to be 2.6 cm/μmol in close agreement to the value of 2.7 cm/μmol reported before.19 Typically, deconvolution was started from low CO coverage, and it was determined at which coverage the band initially appearing around 2177−2182 cm−1 was saturated. This deconvolution procedure was straightforward for the faujasite zeolites and clays but not for ASA. The reason is that the band initially present at 2177 cm−1 cannot be distinguished anymore as a separate band at higher CO coverage due to the much more intense band around 2175 cm−1. In principle, this feature can be fitted by one Gaussian peak with a small shift as a function of CO coverage. An alternative is that the band at higher CO coverage is composite with a few strong BAS giving rise to the band at 2177 cm−1 and a much much larger number of sites of modest acidity to the one at 2175 cm−1. Thus, the spectra were fit by two bands with the one at 2177 cm−1 being fixed in position and width based on the initial IR spectra. Figure 4 shows such a fit for an ASA and also representative fits for a zeolite and a clay. Table 3 lists the concentrations of BAS for a number of aluminosilicates as probed by COads IR spectroscopy. As expected, the BAS concentrations for the steam calcined Y zeolites agree very well with the H/D exchange FTIR derived concentrations of BAS located in the supercages. The agreement is reasonable for the clays. The two methods give comparable values for the ASA samples. The agreement is acceptable for relatively low aluminum content, but deviations increase with the aluminum content. ASA (20/80,1073) is most acidic, which contradicts the trend in the H/D exchange FTIR results. The reason for this is that the two bands at 2175 and 2177 cm−1 overlap and, accordingly, deconvolution is not accurate. In conclusion, the COads IR spectra show that strong BAS are present, but they cannot be accurately quantified by deconvolution of the CO stretch region. The COads IR spectra provide additional insight into the ASA surface composition. At complete saturation with CO, the intensity of the carbonyl band at 2158 cm−1 (not shown) is much higher than the one around 2175 cm−1 for ASA (5/95,

sample

Nstrong BAS, CO IR (μmol/g)

Nstrong BAS, H/D (μmol/g)

USY (8.1) VUSY (26.5) XVUSY (70) MgSAP (13) MgSAP (33) ASA (5/95,1073) ASA (10/90,1073) ASA (15/85,1073) ASA (20/80,1073) ASA (5/95,773) ASA (5/95,3,1073) ASA (5/95,Cogel)

424 305 194 12 48 3.8 7.6 9.1 10.8 2.8 7.6 8.0

434 338 210 16 28 3.6 4.8 8.6 5.2 2.3 5.7 7.5

a

Values from H/D exchange FTIR (Nstrong BAS, H/D) have been included for comparison. For the zeolites, H/D exchange values correspond to number of sites located in supercages.

1073), while the reverse is true for ASA(20/80, 1073). Similar trends are found in the hydroxyl region, where the intensity of the perturbed OH band of the less acidic groups around 3450 cm−1 increases with the ASA aluminum content. The concentration of weak BAS (2175 cm−1) and silanols (2158 cm−1) was determined from spectra after complete coverage by CO. Table 4 shows that the concentration of weak BAS increases with aluminum content and is 1 order of magnitude Table 4. Concentration of Weak BAS (Nweak BAS) and Silanol Groups (Nsilanol) Determined from Infrared Spectra of Adsorbed CO at Complete Saturation sample ASA ASA ASA ASA ASA ASA ASA 21421

(5/95,1073) (10/90,1073) (15/85,1073) (20/80,1073) (5/95,773) (5/95,3,1073) (5/95,Cogel)

νCO (cm−1)

Nweak BAS (μmol/g)

Nsilanol (μmol/g)

2175 2175 2175 2175 2171 2172 2174

75 101 152 160 63 64 65

387 345 292 212 357 423 360

dx.doi.org/10.1021/jp309182f | J. Phys. Chem. C 2012, 116, 21416−21429

The Journal of Physical Chemistry C

Article

higher than the concentration of strong BAS. The silanol density in the parent silica for ASA preparation is around 2700 μmol/g, and it can thus be inferred that most of the silanol groups have been consumed during ASA synthesis. This agrees well with the surface model for ASA involving a large coverage with grafted aluminum.11 Finally, the two bands relating to Lewis acidity in ASA need to be discussed. The one at 2191 cm−1 is attributed to a σ-bond of CO with Al3+ sites with structures similar to those described by Zecchina et al. for alumina as “bulk tetrahedral Al3+ ions emergent on the surface”.27 These sites may thus be related to the segregated alumina domains on the ASA surface. The strong Lewis acidity represented by the band at 2230 cm−1 can be argued to be due to highly dispersed, likely isolated, grafted Al3+ surface sites. Figure 5 shows the intensities of the two

Figure 6. Infrared spectra of adsorbed pyridine recorded at 423 K after pyridine absorption at 423 K (0.01 mbar pyridine) followed by evacuation at 423 K for 1 h (full line) and subsequently at 673 K for 1 h (dashed line) for (a) ASA (5/95, 1073) and (b) ASA (20/80, 1073).

together with a weak band at 1544 cm−1. The shoulder at 1638 cm−1 has completely disappeared. The observation that the band related to the pyridinium ion (PyH + ) remains chemisorbed after evacuation at high temperatures points to strong acidity. Similarly, the presence of strong LAS is evident. The intensities after outgassing at 673 K were used to determine the concentration of strong BAS and LAS. The weaker form of these sites were quantified by taking the difference between concentrations determined after outgassing at 423 and 673 K. For these calculations the molar extinction coefficients as given by Emeis29 were used and the results are collected in Table 5. The ASA samples contain weak BAS and

Figure 5. Concentrations of weak (●) and strong (■) Lewis acid Al3+ sites identified by respective carbonyl stretching bands around 2191 and 2230 cm−1 as a function of the aluminum content of ASA calcined at 1073 K.

Table 5. Concentration of Brønsted and Lewis Acid Sites for ASA and a Clay from IR Spectra of Adsorbed Pyridinea

bands representing LAS as a function of the Al content. All ASA samples contain a small number of strong LAS, which does not depend strongly on the Al content. In contrast, the number of weak LAS increases strongly with the Al content. These trends are in very good agreement with the surface model of ASA.11 The COads IR spectrum of our γ-Al2O3 (Figure S3 in the Supporting Information) shows that this material does not contain the strong Lewis acid sites, while the number of weaker Lewis acid sites is significantly higher than in ASA. The results obtained in Caen on the ASA (5/95, 1073) and ASA (20/80, 1073) samples are essentially similar to the ones described above. It is indeed difficult to quantify the number of CO molecules adsorbed on strong BAS (the band at 2177 cm−1), and the perturbed OH spectra only show a clear band at 3320 cm−1 for ASA (5/95, 1073). On the other hand, there appears to be a small shift in the CO stretching frequency for the weaker BAS from 2176 cm−1 for ASA (5/95, 1073) to 2173 cm−1 for ASA (20/80, 1073), not observed in the Eindhoven data, and this may contain some additionional information on the distribution of acid sites on ASA surfaces. 3.2. Pyridineads IR. Figure 6 shows IR spectra of adsorbed pyridine on two representative ASA samples. The bands at 1623, 1492, and 1456 cm −1 correspond to pyridine coordinating to LAS. The shoulder at 1638 cm−1 and the band at 1544 cm−1 are due to the pyridinium ions, indicating the chemisorption of pyridine on strong BAS. After evacuation at 673 K for 1 h, the intensity of these bands has decreased. The bands at 1623, 1492, and 1456 cm−1 remain in the spectrum

sample

Ntotal BAS (μmol/g)

Ntotal LAS (μmol/g)

Nstrong BAS (μmol/g)

Nstrong LAS (μmol/g)

ASA (5/95,1073) ASA (5/95,3,1073) ASA (5/95,Cogel) ASA (20/80,1073) MgSAP (13)

16 21 23 17 57

59 75 77 74 182

0 2.8 2.3 2.4 17

46 62 64 62 101

a

The concentration of total Brønsted and Lewis acid sites (Ntotal BAS and Ntotal LAS) was determined after evacuation at 423 K. The concentration of strong Brønsted and Lewis acid sites (Nstrong BAS and Nstrong LAS) was determined after evacuation at 673 K.

LAS in concentrations of 15−20 μmol/g and 15 μmol/g, respectively, and strong BAS in the amount of ∼2−3 μmol/g. Although the latter value is close to the results of H/D exchange FTIR and COads IR spectroscopy, it should be noted that the pyridineads IR values show little variation for this subset of ASA samples. Moreover, pyridineads IR spectroscopy did not evidence strong BAS in ASA (5/95,1073), and the other ASA samples contain less BAS than determined by H/D exchange FTIR. A too high desorption temperature for the pyridineads IR measurements may explain these lower than expected values. Another problem might be the quite large spread in reported molar extinction coefficients.30,31 Pyridineads IR shows that the ASA samples contain a relatively high concentration of strong LAS. These sites can chemisorb pyridine up to quite high temperatures. Whereas the Lewis acidity is dominated by strong sites, the Brønsted acidity is dominated by a relatively 21422

dx.doi.org/10.1021/jp309182f | J. Phys. Chem. C 2012, 116, 21416−21429

The Journal of Physical Chemistry C

Article

Figure 7. Rate of IPAm desorption (dashed line) and decomposition (full line) of (a) ASA (5/95, 1073), (b) ASA (20/80, 1073), and (c) γ-Al2O3 as a function of temperature.

weak form. Although pyridineads IR evidence the presence of strong BAS, the data set leaves some doubt whether the values are accurate enough to distinguish differences in strong Brønsted acidity among the present set of ASA materials. 3.3. TPD of Isopropylamine. Figure 7 shows the TPD graphs for two ASA samples and γ-alumina. The Supporting Information contains graphs for three zeolites. Decomposition of adsorbed isopropylamine to propene and ammonia occurs between 550 and 640 K. It can be seen that a substantial amount of IPAm desorbs in the molecular form. IPAm desorption was observed for all aluminosilicates except for those zeolites that do not contain extraframework Al. Attempts to decrease the desorption of molecular IPAm by extending the purging time were not successful. Although IPAm desorption takes place predominantly below the temperature at which decomposition starts, the graphs show that there is a small overlap. Quantification is based on the assumption that each BAS forms a complex with one IPAm molecule.21 Table 6 collects the amounts of IPAm decomposed (Ndec) and desorbed (Ndes) for a suite of zeolites, ASA, γ-alumina, and silica. Expectedly, the amount of decomposed IPAm for HZSM-5 equals the Al density of the zeolite. To make the comparison easier, the BAS concentrations determined by H/D exchange FTIR are also included in Table 6. For faujasite zeolites, the reported H/D exchange FTIR values are those for the sites in the supercages, because it is known that IPAm only interacts with those sites. The amount of decomposed IPAm (Ndec) agrees well with the expected strong BAS concentration for the high-silica end members of the steam-calcined Y zeolites and for USY (9.6, F). For more Al-rich zeolites, however, Ndec is considerably higher than the reference value. The difference is due to the presence of an extraframework silica−alumina phase in the aluminum-rich Y zeolites. The zeolite data show that the difference increases with Ndes. The presence of IPAm adsorbed on different sites than the strong BAS may contribute in two ways to the total conversion of IPAm to propene and ammonia, namely, (i) by decomposition of IPAm strongly bound to these sites and (ii) desorption of IPAm from such sites, followed by readsorption and decomposition on strong BAS. For ASA, Ndec varies between 80 and 200 μmol/g, which is about 2 orders of magnitude higher than the strong BAS concentrations. It is immediately clear that this technique is not selective to strong BAS. The analogy with zeolites containing

Table 6. Amount of IPAm Decomposed (NIPAm, decomposed) and Desorbed (NIPAm, desorbed) during an IPAm TPD Experiment for a Set of Aluminosilicatesa sample

Nstrong BAS, H/D (μmol/g)

NIPAm, decomposed (μmol/g)

NIPAm, desorbed (μmol/g)

HZSM-5 (40) USY (9.6, F) USY (8.1) USY (9.3) VUSY (26) VUSY (33) XVUSY (70) XVUSY (85) ASA (5/95,773) ASA (10/90,773) ASA (15/85,773) ASA (20/80,773) ASA (5/95,1073) ASA (10/90,1073) ASA (15/85,1073) ASA (20/80,1073) γ-Al2O3 SiO2

n.d.b 1522 770 167 340 168 210 106 2.3 3.2 2.7 4.7 3.6 4.8 8.6 5.2 0 0

830 1700 922 244 356 167 231 112 104 112 137 128 102 87 137 197 39 0

0 0 1000 394 203 147 66 114 413 485 671 683 316 335 667 1120 366 0

a

Values from H/D exchange FTIR (Nstrong BAS, H/D) have been included for comparison. For the zeolites the value of the Nstrong BAS, H/D corresponding to concentrations in supercages is given. bNot determined.

an extraframework silica−alumina phase is evident. It agrees with the notion that ASA contains a large number of other acid sites. Figure 8 shows that the difference between the total amount of IPAm decomposed and the strong BAS concentrations by H/D exchange FTIR as a function of Ndes are well correlated. It is not clear what these sites are. Aluminol sites can be ruled out, because such sites are much more abundant in γAl2O3, i.e., typically at a concentration of 1.2 mmol/g than IPAm TPD shows for γ-Al2O3. Accordingly, silanols can also be excluded as they are even less acidic. It is then reasonable to consider BAS of modest acidity, whose concentration was estimated to be around 60−150 μmol/g for the ASA samples and approximately 100 μmol/g for the aluminum-rich Y zeolites. However, based on the pyridineads IR results it may be argued that LAS should be responsible: only a small fraction of the Brønsted acid sites is strong enough to retain pyridine at 21423

dx.doi.org/10.1021/jp309182f | J. Phys. Chem. C 2012, 116, 21416−21429

The Journal of Physical Chemistry C

Article

results for Y zeolites. Nor does the amount of adsorbed ions remain constant as a function of the Cs+ concentration. Control experiments on pure γ-alumina and pure silica did not evidence any exchange under the experimental conditions as long as the pH was below 6. At pH below 6 one expects that exchange is limited to relatively strong Brønsted acid hydroxyl groups. This was verified by monitoring the pH of the solutions at the beginning and end of each experiment (Table S1 in Supporting Information). A reason for the difficulty in determining the ionexchange capacities in this way can be the considerable variation of the pH per sample and as a function of the Cs+ concentration. Although no accurate concentrations of BAS can be determined from these experiments, it is clear that the ASA surfaces are able to exchange Cs+ in amounts in the range 10− 100 μmol/g. This implies the presence of a large number of BAS. The end pH is higher for the high aluminum content ASA samples than for other two. This effect can be explained by a buffering effect due to the presence of alumina in the former two. 3.5. Ion Exchange with Cu-EDA Complex. For clays, the cation exchange capacity can be determined by exchange with a Cu2+-EDA complex.25 The Cu2+-EDA complex does not undergo a compositional change in solutions with a pH between 6 and 832 and is able to displace most exchangeable cations and even heavy metal ions.33 The complex binds strongly to clay surfaces, possibly through covalent bond formation. An assumption is that each Cu2+-EDA complex replaces two hydroxyl groups. Table 7 lists the cation exchange

Figure 8. The amount of decomposed i-propylamine not related to strong Brønsted activity as a function of the amount of desorbed ipropylamine for (■) zeolites, (●) ASAs, and (▲) γ-Al2O3.

473 K. For example, ASA (5/95,1073) contains 75 μmol/g weak BAS from COads IR, but only 16 μmol/g of these sites are of sufficient strength to retain pyridine at 473 K. As pyridine is a stronger base than IPAm, it is expected that only a small fraction of these sites contribute to the excess decomposition. Therefore, LAS are more reasonable candidate sites for retaining IPAm up to high temperatures. Indeed, substantial amounts of pyridine remain adsorbed on ASAs up to a temperature of 673 K. For instance, ASA (5/95,1073) and ASA (20/80,1073) contain 46 and 62 μmol/g Lewis acid sites, respectively. As the strength of these sites is sufficiently high to adsorb pyridine at 673 K, it may be expected that IPAm also remains adsorbed up to temperatures high enough for its decomposition to occur. Typical values for the amount of LAS on γ-Al2O3 are around 0.1 Al3+/nm2,27 which corresponds well to a concentration of 44 μmol/g for the γ-Al2O3 employed here. In conclusion, IPAm decomposition is not able to probe selectively the strong BAS in ASA. The problem lies in the relatively strong adsorption of alkylamines to Brønsted acid sites of varying acidity and, especially, to strong Lewis acid sites. 3..4. Ion Exchange with Cs+. Figure 9 shows the relation between the fraction of Cs+ ions adsorbed by the support and the Cs+ concentration in the solution. Clearly, these plots do not represent well-behaved adsorption isotherms in contrast to

Table 7. Concentration of Strong Brønsted Acid Sites (NCu‑EDA) for Aluminosilicates Determined by Exchange of Hydroxyl Groups with the Cu-EDA Complexa sample

NCu‑EDA (μmol/g)

Nstrong BAS, H/D (μmol/g)

ASA (5/95,1073) ASA (10/90,1073) ASA (15/85,1073)

140 120 120

3.6 4.8 8.6

a

For comparison, the values from the H/D exchange FTIR are also included (Nstrong BAS, H/D).

capacities for three ASAs. These values are 2 orders of magnitude higher than the expected ones. Clearly, this method probes a wide range of hydroxyl groups on the ASA surface of varying acidic strength. 3.6. 1H MAS NMR. The 1H NMR spectra and their deconvolution into various peaks for ASA (5/95,1073) and ASA (20/80,1073) are shown in Figure S3 of the Supporting Information. The spectra were fit with peaks around 1.8, 2.3, and 3.8 ppm relative to liquid tetramethylsilane. Very small resonant peaks are noted at 0.9 and 1.2 ppm, the latter of which is due to isolated silanols in silica.26 The dominant signal at 1.8 ppm in all spectra are non-hydrogen bonded single and/or geminal silanol groups.34 The peak at 2.3 ppm is due to isolated nonacidic aluminol groups.35 Signals in the range 3.6−7 ppm have been assigned to various forms of bridging hydroxyl groups in zeolites.26 At the same time, hydroxyl groups in typical aluminum hydroxides and oxides are located in the range 2.9−5.8 ppm.35 The exact position depends on the Al coordination number of the hydroxyl group. The corresponding fit results are given in Table 8. The concentration of silanol groups decreases somewhat with increasing Al content in line with the COads IR results. The trend is similar for the isolated aluminol groups, which points to the agglomeration of Al

Figure 9. Adsorption isotherms of CsCl of (■) ASA (5/95, 1073), (●) ASA (10/90, 1073), (▲) ASA (15/85, 1073), and (▼) ASA (20/ 80, 1073). 21424

dx.doi.org/10.1021/jp309182f | J. Phys. Chem. C 2012, 116, 21416−21429

The Journal of Physical Chemistry C

Article

Table 8. Concentration of Different Protons as Derived from 1 H MAS NMR Spectra sample ASA ASA ASA ASA ASA

(5/95,1073) (10/90,1073) (15/85,1073) (20/80,1073) (5/95,Cogel)

Nδ=1.8 ppm (mmol/g)

Nδ=2.3 ppm (mmol/g)

Nδ=3.8 ppm (mmol/g)

0.57 0.55 0.40 0.41 0.57

0.22 0.17 0.11 0.12 0.19

0.21 0.21 0.20 0.20 0.20

Table 9. Concentration of Strong Brønsted Acid Sites (Nstrong BAS, isom) for Aluminosilicates Determined by Poisoning of Acid Sites with 2,6-Dimethyl Pyridine during m-Xylene Isomerisationa sample

Nstrong BAS, isom (μmol/g)

Nstrong BAS, H/D (μmol/g)

ASA (5/95, 773) ASA (10/90, 1073) ASA (5/95, Cogel)

0.4 2.6 2.7

2.0 4.8 7.5

a

Values from H/D exchange FTIR (Nstrong BAS, H/D) have been included for comparison.

species. The relatively broad feature around 3.8 ppm is due to a range of bridging hydroxyl groups attached to more than one aluminum ion and likely also to those between Si and Al atoms. The concentration of such sites is in the order of 0.2 mmol/g. Thus, it appears that the signal due to strong BAS cannot be isolated from the other contributions in the 1H NMR spectra. 3.7. Isomerization of m-Xylene. As isomerization of xylenes is catalyzed by strong BAS,36 another potential method to probe their concentration would be to measure the acid catalytic activity followed by selective poisoning of these sites. When m-xylene isomerization is carried out in the gas phase, catalyst stability is a problem. It is known that diphenylmethanes, which are intermediates in the disproportionation reaction, strongly inhibit the isomerization reaction. Further condensation of the heavy intermediates causes coke to deposit, which leads to irreversible deactivation. Takaya et al. have shown that catalyst deactivation can be avoided by carrying out the reaction in the liquid phase.37 Thus, the activity of an ASA sample was found to be stable for prolonged reaction time. The increased stability was explained by the solvation of diphenylmethanes by liquid m-xylene. Hence, in this study m-xylene isomerization was carried out in the liquid phase at 573 K. Lutidine (2,6-dimethylpyridine) was chosen to poison strong BAS, following reports38,39 that substituted pyridines react selectively with bridging hydroxyl groups in the presence of Lewis acid sites up to a temperature of 673 K. It is worthy to note that the catalyst also deactivated in our initial liquid phase experiments, but this turned out to be due to traces of peroxide stabilizers in m-xylene, which can be effectively removed by percolation over a bed of activated alumina. Figure 10 shows the catalytic activity of ASA (5/95,1073) as a function of the time on stream. Injection of calibrated amounts of lutidine results in a stepwise decrease of the mxylene conversion, which shows a linear decrease as a function of the total amount of base. Table 9 summarizes the acidity

results for three ASA samples. It is clear that the values fall in the same range as the FTIR H/D exchange ones but are generally lower. A tentative explanation may be that m-xylene isomerization probes a different fraction of strong acid sites, and this is another hint at a quite heterogeneous distribution of the acid sites in ASA. Nevertheless, selective poisoning of liquid phase m-xylene isomerization by lutidine is a promising novel method to determine the concentration of strong BAS in ASA but more quantitative work should be undertaken. At Caen, ASA (5/95, 1073) and ASA (20/80, 1073) were subjected to IR spectroscopy of adsorbed lutidine, following the procedures described by Crépeau et al.7 As is already clear,7 the amount of BAS probed by lutidine is a quite continuous function of temperature. After desorption at 423 and 573 K, the number of BAS is only slightly higher for ASA (20/80, 1073 than for ASA (5/95), while the difference in H/D exchange FTIR is about 40% (Table 3). Also, a small amount of Lewissite adsorbed lutidine is present. This will have to be taken into account if and when this selective-adsorption method is going to be further developed.

4. GENERAL DISCUSSION 4.1. Comparison of Various Techniques for Acidity Characterization. Of the various techniques presented above, COads IR spectroscopy provides quite strong indications for the presence of a very small number of BAS of zeolitic strength in ASA. First, a strongly shifted hydroxyl stretching frequency at 3320 cm−1 for ASA(5/95,1073) is observed in line with earlier findings.7 Although this band cannot be clearly distinguished anymore for ASA samples with a higher Al content, the frequency shift places the corresponding acid sites close to zeolites in terms of acidity. Second, at low CO coverage the CO stretching band for ASA is close to the value found for CO interacting with bridging hydroxyl groups in zeolites. This band

Figure 10. Liquid-phase m-xylene isomerization of dehydrated ASA (5/95, 773) (liquid phase, p = 80 bar, WHSV = 0.85 h−1): (left) conversion of m-xylene as a function of time on stream and during three consecutive pulses of lutidine and (right) as a function of the amount of lutidine introduced during the isomerization reaction. 21425

dx.doi.org/10.1021/jp309182f | J. Phys. Chem. C 2012, 116, 21416−21429

The Journal of Physical Chemistry C

Article

explained by the heterogeneity in the acid strengths of these sites. Indications that this is the case were already outlined.14 To summarize, a thorough comparison of various characterization techniques stresses the heterogeneous nature of the surface acidity of ASA. Figure 11 distinguishes three types of

can only be distinguished as a separate band at very low CO coverages. Higher CO coverage results in the appearance of a much more intense IR band at slightly lower wavenumbers, which represents a weaker form of BAS. Only under stringent fitting conditions can this band be deconvoluted into two separate contributions, providing an estimate of the concentration of strong BAS. These values are of the same order of magnitude as BAS concentrations determined by H/D exchange FTIR. Yet, this method is not very reliable, because it depends on the amount of weak BAS. COads IR spectroscopy also evidence the presence of weak and strong Lewis acid sites on the ASA surface. Analysis of the spectra as a function of CO coverage shows that the interaction of CO with strong Lewis acid sites results in a slight shift of the IR signal of a certain fraction of silanol groups. This is interpreted as coordination of CO to a SiOH···Al pair formed by a silanol group with strong isolated Lewis acid Al3+ surface sites. Here, it is proposed that these paired sites exhibit Brønsted acidity between that of silanols and the strong bridging hydroxyl groups, and these sites may thus cause the weaker form of BAS in ASA. Similarly, pyridineads IR spectroscopy probes both Brønsted and Lewis acidity. The BAS concentration after pyridine desorption at 673 K corresponds well to the H/D exchange FTIR value. A drawback of this method is that it is not clear what is the optimal desorption temperature, and there is a large spread in reported extinction coefficients of the pyridinium ion. Despite an earlier favorable report about the use of IPAm decomposition to probe strong BAS in ASA,22 it is found that this method is not selective. Most likely, Lewis acid sites adsorb IPAm so strongly during a temperature-programmed experiment that they either decompose on these sites or upon desorption are decomposed on BAS. The overestimation of the BAS concentration is also found for steam-calcined Y zeolites with an extraframework silica−alumina phase. A modification of this method, which involves a desorption step at 573 K for 4 h, was then briefly explored. In a control experiment with XVUSY (85), the concentration of strong BAS was found to be 115 μmol/g in good agreement with the original value. The strong BAS concentrations determined by this modified method for ASA (5/95,1073) and ASA (15/85,1073) were 4.7 and 5 μmol/ g, respectively. Thus, the excess IPAm on less acidic sites can be removed by flushing at a temperature below the IPAm decomposition temperature. More work is needed to turn this method into a quantitative one. Ion-exchange methods titrate a much larger number of BAS of the order of the concentration of weak BAS determined by COads IR. Cu(EDA)22+ probes a larger amount than does Cs+. The Cs+ adsorption isotherms, however, are not well behaved. 1 H MAS NMR spectra show the presence of abundant silanol and aluminol groups. A signal at 3.8 ppm is related to a more acidic form but cannot be exclusively assigned to a proton located on a bridging oxygen between silicon- and aluminumoccupied oxygen tetrahedra. Bridging hydroxyl groups between aluminum centers as also present on the surface also contribute to this broad feature. Isomerization of m-xylene can be employed to determine Brønsted acidity by using lutidine to poison the acid sites. The concentration of strong BAS determined by this method is of the same magnitude as values obtained by H/D exchange FTIR and COads and pyridineads IR, but further validation for a larger set of catalysts is required. The variation in values of strong BAS between the various techniques may at least in part be

Figure 11. Schematic representation of the three types of Brønsted acid sites in a typical ASA. Due to the amorphous nature, these sites are much less defined than the acid sites in zeolites. Methods to probe the various types of sites are discussed in Table 9.

hydroxyl groups, i.e., (i) strong Brønsted acid bridging ones in the silica network, (ii) a weaker and more abundant form, likely being silanol groups interacting with strong Lewis acid sites, and (iii) nonacidic silanol and aluminol groups. Table 10 reviews the various techniques useful to probe these sites. Besides BAS, the surface also contains Lewis acid sites. The stronger form identified by a COads IR band at 2230 cm−1, the concentration of which depends only weakly on the aluminum content, is due to grafted isolated Al3+. The weaker form gives rise to a band at 2190 cm−1. Its abundance strongly increases with Al content, which suggests that it is due to Lewis acid sites constituent of the aluminum domains. 4.2. Surface Composition ASA and Active Sites in Acid Catalysis. Figure 12 shows a simplified representation of the surface of ASA. Next to the very small amount of Al substitutions in the silica network that cause strong Brønsted acidity, the surface contains abundant isolated and slightly agglomerated Al sites. Upon dehydration, the interaction of such Lewis acid sites with silanol groups results in additional weak BAS. The extent of Al agglomeration can be such that small seggregated domains on the surface are formed that mainly consist of octahedral Al. In ASA formulations with higher Al content than explored here also free γ-alumina occurs. The five-coordinated Al as probed by NMR seems to be primarily related to the interface between the seggregated aluminum oxide domains and the ASA phase. Finally, the surface contains non- or at least very weakly acidic aluminol and silanol groups. As the present set of data evidence the presence of two forms of Brønsted acid sites in ASA, it is useful to compare the earlier correlation of the acid activity with the strong BAS concentration as probed by H/D exchanges14 to one with the concentration of weak BAS. Figure 13 shows the two relations for the acid activity in n-heptane hydroconversion of zeolites and ASA (for experimental details see reference15). It 21426

dx.doi.org/10.1021/jp309182f | J. Phys. Chem. C 2012, 116, 21416−21429

The Journal of Physical Chemistry C

Article

Table 10. Brief Overview of Various Techniques Employed to Characterize the Various Types of Hydroxyl Groups in ASA type I: strong Brønsted acid bridging hydroxyl groups (1−10 μmol/g)

II: weak Brønsted acid bridging hydroxyl groups (50−150 μmol/g)

III: Nonacidic silanol and aluminol groups (250−400 μmol/g)

technique H/D exchange IR14 COads IR7 pyridineads IR (this work) IPAm TPD (this work) COads IR (this work) pyridine FTIR (this work) 1 H NMR (this work) Cu(EDA)22+ and Cs+ exchange (this work) COads IR (this work) 1 H NMR (this work)

remarks quantitative quantification difficult because of overlap with COads bands due to type II (in some cases signature found by studying perturbed hydroxyl bands) quantification difficult because of arbitrary choice of desorption temperature quantitative after modification of method by Gorte et al.19−21 quantitative quantification difficult because of arbitrary choice of desorption temperature determines wide range of weak BAS groups, overlap with acidic aluminol groups Cs+ adsorption isotherms not well behaved quantitative (silanol and aluminol not distinguished) quantitative, silanol and nonacidic aluminol groups

catalysts are given in Table 11. The TOF of the ASA-supported catalysts is much higher than those of the alumina- and silicaTable 11. TOF of Neopentane Hydrogenolysis for a Set of Supported Pt Catalysts (T = 573 K, 1 vol % C5H12 in H2, Reduction at 573 K, WHSV = 0.01 h−1) Figure 12. Schematic representation of the surface of hydrated ASA with from left to right: bridging hydroxyl groups (strong BAS), silanol groups, grafted Al of increasing degree of agglomeration.

catalyst

TOF (h−1)

Pt/SiO2 Pt/SiO2:Al Pt/ASA (5/95,3,1073) Pt/ASA (10/90,3,1073) Pt/ASA (15/85,3,1073) Pt/ASA (20/80,1073) Pt/ASA (55/45) Pt/Al2O3

0.4 10 104 70 69 46 43 0.8

supported ones. For ASA, the activity decreases with increasing Al content. Pt/Al2O3 has a very low activity, similar to Pt/SiO2. On the other hand, Pt/SiO2−Al has a considerable higher activity than Pt/SiO2. The influence of the support acid/base properties on the activity of reduced Pt particles in aromatics hydrogenation is well known.41 Similar effects have been noted for the hydrogenolysis of neopentane42−44 and also the sulfur tolerance in this reaction.45 It is usually assumed that the increased activity is due to a more electron deficient nature of the noble metal particles on acidic supports. Changes in the Pt electronic structure have been probed by X-ray absorption spectroscopy.46 More recently, the effect of the Pt−H bond strength has been used to explain the catalytic activity differences.47 Lercher and co-workers have also found that Pt/ASA is much more active in neopentane hydrogenolysis than Pt/Al2O3.13 They explained this difference in terms of the support (Sanderson) electronegativity instead of difference in support acidity. For the present set of ASA catalysts, the Brønsted acidity of the support is constant. It follows that the activity trends in neopentane hydrogenolysis are not the result of differences in Brønsted acidity. An alternative explanation is that the Pt electronic structure is influenced by the Sanderson electronegativity of the oxide support.13,48 The more electronegative the support, the more electronegative the Pt particles and the higher the activity in C−H bond cleavage. From the substantial activity difference between the two silica-supported Pt catalysts, we infer that Al is required to transmit this electronic effect between the support and the active metal phase. Without Al, such interactions are very weak. The coverage with Al species in ASA(5/95,3,1073) is close to the monolayer coverage, and the

Figure 13. Relation between the catalytic activity in the hydroconversion of n-heptane and the concentration of acid sites (mmol·g−1) determined by H/D exchange FTIR for (■) zeolites and (▲) ASA prepared by homogeneous deposition−precipitation and (▼) ASA prepared by cogelation. The corresponding open symbols refer to the weak Brønsted acid sites as determined by COads IR. The line is a guide to the eye.

convincingly shows that the catalytic performance does not correlate with the concentration of weak BAS for ASA. 4.3. Catalytic Activity in Neopentane Hydrogenolysis. A set of Pt-containing catalysts was tested in the hydrogenolysis of neopentane. The Pt particle size was about 2 nm for Pt/ SiO2:Al and Pt/Al2O3 and 8 nm for Pt/SiO2. The reason for the larger particle size of the Pt/SiO2 catalyst is attributed to the absence of Al in the silica support. The particle size is much smaller for Pt/SiO2−Al, the support in this case having an Al content of ∼0.5 wt %. A similar difference in size of gold nanoparticles has been discussed for these two types of silica.40 The ASA-supported Pt catalysts contain metal nanoparticles around 4−5 nm. The products of neopentane hydrogenolysis are predominantly methane and isobutane (selectivity >96%). Very small amounts of isopentane (isomerization) and further cracking products were observed. The turnover frequencies (TOF) in the hydrogenolysis of neopentane for these Pt 21427

dx.doi.org/10.1021/jp309182f | J. Phys. Chem. C 2012, 116, 21416−21429

The Journal of Physical Chemistry C



Sanderson electronegativity of this support is highest among the ASA samples.48 The Sanderson electronegativity decreases with Al content concomitant with the turnover frequency. Recent work has produced rather convincing evidence that, during preparation of γ-alumina supported Pt catalysts, the Pt precursor preferably interacts with Lewis acid Al sites on the alumina surface.49 The finding that the neopentane hydrogenolysis activity of Pt/γ-Al2O3 is much lower than that of the Pt/ASA catalysts could be due to the absence of strong LAS in the former support as evidenced by COads IR. Accordingly, these data underpin the correlation between the extent of the electronic effect for supported Pt nanoparticles and the support Sanderson electronegativity and additionally show that strong Lewis acid Al sites are required to transmit this electronegativity to the active Pt phase.

AUTHOR INFORMATION

Corresponding Author

*Phone: +31-40-2475178. Fax: +31-40-2455054. E-mail: e.j.m. [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors thank Shell Global Solutions for funding this research and Prof. Johannes Lercher and Dr. Benjamin Fonfé of TU Munich for their kind help with the pyridineads IR measurements. Dr. Yejun Guan and Dr. Pieter Magusin are acknowledged for neopentane hydrogenolysis and NMR measurements.



5. CONCLUSIONS A number of techniques (COads IR, pyridineads IR, alkylamine TPD, Cs+ and Cu(EDA)22+ exchange, 1H NMR, and m-xylene isomerization) give further insight into the heterogeneous nature of the surface of amorphous silica−alumina. It contains Brønsted and Lewis acid sites of varying acidity. The number of strong Brønsted acid sites of zeolitic strength is very low (