Zooming in with QSPR on Friedel-Crafts acylation reactions over modified BEA zeolites

Molecular Catalysis 476 (2019) 110495

Contents lists available at ScienceDirect

Molecular Catalysis journal homepage: www.elsevier.com/locate/mcat

Zooming in with QSPR on Friedel-Crafts acylation reactions over modified BEA zeolites

T

Rodrigo Aleixoa, Ruben Elvas-Leitãoa,b, Filomena Martinsb, Ana P. Carvalhob, Amadeu Brigasc, Ricardo Nunesb, Auguste Fernandesd, João Rochae, Angela Martinsa,b,⁎, Nelson Nunesa,b,⁎⁎ a

Área Departamental de Engenharia Química, Instituto Superior de Engenharia de Lisboa, IPL, R. Conselheiro Emídio Navarro, 1959-007 Lisboa, Portugal Centro de Química e Bioquímica (CQB) and Centro de Química Estrutural (CQE), Faculdade de Ciências, Universidade de Lisboa, 1749-016, Lisboa, Portugal c Departamento de Química e Farmácia, Universidade do Algarve, Campus de Gambelas, 8005-139, Faro, Portugal d CQE, Instituto Superior Técnico, Av. Rovisco Pais, 1049-001, Lisboa, Portugal e Departamento de Química and CICECO, Universidade de Aveiro, 3810-193, Aveiro, Portugal b

ARTICLE INFO

ABSTRACT

Keywords: QSPR analysis Hierarchical BEA Friedel-Crafts acylation Kinetic modelling Nonlinear regression

The catalytic behaviour of hierarchical BEA zeolites with Si/Al ratio of 12.5 and 32 was studied in Friedel-Crafts acylation reactions using furan, anisole and pyrrole as substrates and acetic anhydride as acylating agent. Hierarchical BEA samples were submitted to alkaline and alkaline + acid treatments. Kinetic modelling using nonlinear regressions applied to a simplified Langmuir-Hinshelwood equation showed that the Si/Al ratio of the parent materials strongly influenced the catalytic behaviour. Catalytic results were correlated with physicochemical properties using a Quantitative Structure-Property Relationship (QSPR) methodology. This approach provided detailed information about the role of key properties on the catalytic behaviour, and pointed out which properties should be modified through direct synthesis and/or post-synthesis treatments to obtain materials with optimized catalytic performance.

Introduction Zeolites present a unique combination of properties such as high surface area, well defined porosity, high thermal and mechanical stability and customized acidity. These materials are particularly suitable for a large array of applications in catalysis and separation processes [1,2]. Although the most important technological applications of zeolites as catalysts are found in refining and petrochemistry processes, there are other emerging applications of these materials, as catalysts or catalysts’ supports, in reactions usually carried out in the presence of homogeneous catalysts which are frequently environmentally unfriendly [3–5]. Among these reactions, Friedel-Crafts processes are important synthetic routes to obtain intermediates for manufacturing fine and speciality chemicals [6]. Unfortunately, the catalysts commonly used at an industrial scale, like AlCl3 or HF, pose serious problems such as corrosion, dangerous handling and toxicity hazards, and difficult (or even impossible) separation and catalyst recovery. Numerous studies have already reported the use of zeolites aiming at replacing conventional

⁎

homogeneous catalysts in view of increasingly stricter environmental rules [7–11]. Other materials such as porous hypercrosslinked polymers (HCPs) [12], anilinotropone based nickel complexes [13] or MoO3 supported on ordered mesoporous zirconium oxophosphate [14] have also been proposed as promising catalysts. Among zeolites, several structures have been studied in Friedel-Crafts acylation reactions such as FAU, MFI or MWW [9,15,16]; however, BEA seems to be, until now, the most promising structure [17–19]. This zeolite belongs to the category of large pore size zeolites, comprising two types of linear channels with pore diameters of 7.7 × 6.6 Å and a third tortuous channel with pore diameter of 5.6 × 5.6 Å [20]. The structure is constituted by three polymorphic structures A, B and C and differs in the way the polymorphs are stacked with respect to each other. In comparison with other zeolite structures, BEA possesses unique acid properties, with a large amount of Lewis acid sites related with local defects resulting from the stacking of polymorphic units [17]. However, this zeolite structure suffers from the same drawback as all others, that is, the microporous nature of these materials limits the access of large molecules to the active sites that are mainly present inside micropores.

Corresponding author at: ADEQ, Instituto Superior de Engenharia de Lisboa, R. Conselheiro Emídio Navarro, 1959-007 Lisboa, Portugal. Corresponding author at: ADEQ, Instituto Superior de Engenharia de Lisboa, R. Conselheiro Emídio Navarro, 1959-007 Lisboa, Portugal. E-mail addresses: [email protected] (A. Martins), [email protected] (N. Nunes).

⁎⁎

https://doi.org/10.1016/j.mcat.2019.110495 Received 10 April 2019; Received in revised form 25 June 2019; Accepted 29 June 2019 Available online 25 July 2019 2468-8231/ © 2019 Elsevier B.V. All rights reserved.

Molecular Catalysis 476 (2019) 110495

R. Aleixo, et al.

The development of hierarchical zeolites, comprising mesopores in addition to the original micropores, was shown to be a good strategy to increase the accessibility to the active sites and overcome diffusion limitations [21–24]. Hierarchical zeolites can be prepared by two main methodologies: synthesis procedures and post-synthesis treatments [24]. In all cases, the common effect of these modifications is the generation of mesoporosity in addition to the pristine microporosity. The mesoporosity can be observed in two distinct situations: it can be intercrystalline, i.e. obtained by the aggregation of small and/or shape modified crystals; or intracrystalline, when it involves the enlargement of the pores inside the crystals. A combination of the two can also be found [25]. The potentialities of several hierarchical zeolites in FriedelCrafts reactions was recently reviewed by Yan et al. [26]. Regarding BEA, a few studies were also performed, using mainly nano-sized crystals [27,28]. In these cases, it was observed that the large external surface area of the small crystals exposed an increased number of pore openings and allowed a fast desorption of heavy products, reducing the catalyst deactivation and improving catalytic activity. Wang et al. [29] prepared hierarchical BEA by performing desilication treatments with NaOH followed by HCl leaching. The desilication treatment is an easy and low cost post-synthesis method to create mesoporosity and consists in an alkaline treatment with an organic or inorganic base, NaOH being the most used one [24]. Because of this treatment, the deposition of extra-framework Al species may occur leading to partial occlusion of the pore openings. In this case, an additional acid leaching is recommended to remove the deposited Al species [30,31]. Under the experimental conditions used by Wang and co-workers [29], textural and structural data obtained showed that the mesoporosity was developed mainly as a consequence of crystals’ aggregation, that is, as intercrystalline mesoporosity. In Friedel Crafts acylation of mesitylene with benzyl alcohol, an improvement of the catalytic performance was noted since the reaction occurs mainly at the external surface, due to the large size of the reactant and reaction intermediates. However, when a small molecule, like benzene, was tested no differences were observed between the treated and parent materials, because the reaction occurred mostly inside the porous structure, which remained practically unchanged upon treatments. In general, the results of these experimental approaches have been analysed by qualitative reasoning of the individual effects caused by certain characteristics of these structures and of the departing reagent used. However, this type of analysis renders difficult, if not impossible, any thorough rationalization of these processes, mostly because a single effect, encoding a given structural characteristic, can hardly model the complexity of the observed behaviours. Only the simultaneous consideration of various effects, that is a multivariate data analysis approach, will, possibly, allow the progressive unveiling of the factors that determine a certain outcome and will ultimately pinpoint the path for a judicious catalyst design. In the present work, we followed the strategy described above to modify BEA [26] but have used milder conditions attending to our previous experience in desilication treatments, especially in the case of sensitive structures (MWW) [21, 31–33], and have also applied the modified materials to Friedel-Crafts acylations [16]. Given the above, the aim of this work was to use a multiple linear regression (MLR)-based QSPR (Quantitative Structure-Property Relationships) methodology to be able to find and quantify the combined effect of changes in both zeolite and substrate properties. With these objectives in mind, the generation of mesoporosity in BEA was performed and its influence on the accessibility of simple heteroaromatic rings (furane, anisole and pyrrole) to the active sites located inside the pore structure, as well as the effect of the Si/Al ratio of the starting BEA were investigated. Kinetic parameters of the studied reactions were computed and, as stated above, their correlation with substrate and catalyst properties duly assessed.

Experimental Materials and chemicals All chemicals used in this study are commercially available and were used without further purification. The starting materials are BEA structures with Si/Al = 12.5 and 32, supplied by Zeolyst (lot. CP 814E and 1822-14, respectively) and will be designated as BEA followed by the respective Si/Al ratio (12.5 or 32). The zeolites were acquired as NH4BEA and were converted into the protonic form by calcination under dry air (6 l h−1 g−1) at 500 °C for 3 h. Post-synthesis modifications of the catalysts The parent BEA12.5 and 32 were treated with a 0.1 M NaOH (p.a.) solution using a liquid-to-solid ratio of 30 ml g−1, at 60 °C for 30 min. The solid was recovered by centrifugation, washed and dried. To account for some Na exchange during the alkaline treatment, the samples were re-converted to the protonic form by ion exchange with a 2 M NH4NO3 (p.a.) solution using a liquid-to-solid ratio of 25 ml g−1, at 80 °C during 6 h. The recovered solid was submitted to calcination at 500 °C for 3 h under dry air. The treated samples were then submitted to an acid treatment using a 0.1 M HCl (p.a. fuming 37%) solution with a liquid-to-solid ratio of 30 ml h−1, at 70 °C for 3 h. The solid was recovered and calcinated using the procedure described. The treated samples will be designated as BEA12.5 or BEA32 followed by “D” for desilicated samples plus “AT” in case the acid treatment was also applied. After each procedure step (desilication and acid treatment) the samples were weighed to account for mass loss. Physicochemical characterization The structural characterization of the parent and modified BEA samples was attained from X-ray powder diffraction (XRD) patterns that were obtained at room temperature using CuKα radiation as incident beam (Pan’Analytical PW3050/60X’Pert PRO (θ/2θ). Diffractograms were obtained in a 2θ range of 5 to 40°, with a step size of 0.017° 2θ and a time per step of 0.6 s. 29 Si magic-angle spinning (MAS) NMR spectra were recorded in Bruker Avance III 400 NMR (9.4 T) wide-bore spectrometer at 79.5 MHz on a double-resonance 7 mm Bruker probe with a 5 kHz spinning rate, 45° radio-frequency (RF) pulses, and a 60 s recycle delay. Acquisition parameters for the 29Si cross-polarization MAS experiment were the following: 1H and 29Si 90° pulses were set to 2 and 6 ms, corresponding to a RF field strength of 125 and 42 kHz, respectively. The CP step was implemented using a contact time of 2 ms with a ramp shape in the 1H channel of 50–100% and 29Si RF field strength of 66 kHz. During the acquisition, SPINAL-64 decoupling with a RF field strength of 50 kHz (SPINAL basic pulse length of 9.5 ms) was employed. Chemical shifts are quoted in ppm from tetramethylsilane (TMS). 27 Al MAS NMR spectra were recorded on a Bruker Avance III HD, 700 MHz (16.4 T) narrow-bore spectrometer at 182.4 MHz on a doubleresonance 4 mm Bruker probe with a 15 kHz spinning rate. Singlequantum (‘normal’) spectra were recorded using a short RF pulse of 0.27 μs (equivalent to a π/18 flip angle), calibrated on an aqueous solution of Al(NO3)3, with 100 kHz field strength and 0.5 s recycle delay. 27 Al 3QMAS NMR experiments were performed using the split-t1 z-filter pulse sequence [34] using RF pulse lengths of 2.9 μs and 0.9 μs for the first two hard pulses, and a 7.16 μs 90° soft pulse. RF field strengths of 125 and 10 kHz were used for the hard and soft pulses, respectively. 199 t1 points were acquired in the indirect dimension with 1200 scans each, with a recycle delay of 0.5 s. Chemical shifts are quoted in ppm from aqueous solution of Al(NO3)3. 27Al tetrahedral-to-octahedral ratios were estimated from a simple integration (limits 80 to 30 ppm and 15 to -22 ppm). 2

Molecular Catalysis 476 (2019) 110495

R. Aleixo, et al.

Transmission electron microscopy (TEM) images of parent and modified samples were acquired in a Hitachi 200 kV HD-2700b STEM equipment, performed at Aveiro University. The global Si/Al ratio was calculated from elemental analysis data, using inductively coupled plasma-optical emission spectrometry (ICPOES). The analyses were performed at “Laboratório Central de Análises da Universidade de Aveiro” using a Jobin Yvon Activa M spectrophotometer. The textural characterization of the solids was made by N2 adsorption-desorption isotherms measured at -196 °C in an automatic Micromeritics ASAP 2010 apparatus. Before the isotherms acquisition, samples (≈ 50 mg) were outgassed for 2 h at 300 °C, under vacuum better than 10−2 Pa. The micropore volume, Vmicro, and external area, Aext, were assessed applying the αS method using as reference the standard isotherm obtained in a non-porous silica [35] The micropore volume corresponds to the back extrapolation of the linear region of αS plot defined by experimental points corresponding to αS>1, and the external area obtained through the slope. The mesopore pore volume, Vmeso, was obtained from the difference Vtotal-Vmicro, where Vtotal, is the total pore volume determined according to Gurvitch rule, i.e. the amount of N2 adsorbed at p/p0 = 0.95 [35] The acidity of parent and treated samples was evaluated by pyridine adsorption followed by FTIR spectroscopy (Nicolet Nexus). Initially, the samples (10–20 mg cm−2 wafers) were outgassed at 450 °C for 3 h (vacuum of 10-6 mbar), cooled down until 150 °C and contacted pyridine vapour for 10 min. After excess pyridine removal, IR spectrum was acquired (64 scans, resolution 4 cm-1). The estimation of acid sites concentration was performed according to Emeis [36].

discarded. Thus, the reaction rate, r, can be expressed as:

r=

kKA KS [A][S] (1 + KA [A] + KS [S] + KP [P])2

(1)

where k represents the rate constant of the rate determining step, and KA, KS and KP are the adsorption equilibrium constants for each reagent and product, respectively. Since at low temperature, in a liquid-phase process, both reactant and products completely occupy the intracrystalline volume of the zeolite, then the term “1” in eq. (1) can be ignored vis-à-vis the other denominator terms, leading to a simplified version of eq. (1). Moreover, accepting that both A and S are competitively adsorbed on the zeolite acidic sites and assuming that the range of adsorption of both species, A and S, is basically comparable in the mathematical model used, at least statistically speaking (using anisole and acetic anhydride, Derouane et al. [37,38], obtained for the ratio KA/KS values between 1.4 and 1.7), eq. (1) can be further simplified by taking KA ≅ KS. On the other hand, to reduce the number of adjustable variables, it can also be considered that KA ≅ KS ≅ 1, which allows to obtain a relative KP value, Kr, from the simplified equation:

r

k [A][S] ([A] + [S] + Kr [P]) 2

(2)

where Kr represents the ratio between the adsorption equilibrium constant of the product and the normalized equilibrium constants of the reagents. Values for the adjustable variables, k and Kr, are straightforwardly obtained through nonlinear regression. As previously referred, an effective empirical quantitative strategy to deal with this type of multifactor-dependent systems relies on the simultaneous analysis of the influence of both zeolite and substrate properties which permits pinpointing the most significant contributions over k and Kr. This strategy is called QSPR and is a well-known and very successful method used to rationalize different types of molecular interactions in diverse physicochemical and biological settings [39–43]. In the context of this work, the use of QSPRs presupposes that ln k and/ or ln Kr can be expressed as linear combinations of additive and independent properties or descriptors, D, that encode the structural features of catalysts and substrates, and which can be expressed in general terms as a multiple linear regression equation (MLR) of the type:

Catalytic experiments The catalytic assays were performed in a three-necked roundbottom flask equipped with a reflux condenser and heated in a heating plate with temperature control (IKA C-MagHS7), under vigorous stirring and reflux. In a typical experiment, a mixture of either furan (purity > 99%, 0.71 g, i.e. 10.5 mmol) anisole (purity > 99%, 1.14 g, i.e., 10.5 mmol) or pyrrole (purity > 98% and previously distilled, 0.70 g, i.e 10.5 mmol) with acetic anhydride (purity > 99%, 5.4 g, i.e. 52.5 mmol) was added to a sample of parent or modified catalyst (150 mg). In all acylations, the amounts of each reactant were adjusted to obtain molar ratios between substrate and acylating agent equal to 5. The suspension was heated at 60 °C and, and at pre-defined time intervals, small aliquots were taken from the reaction mixture using a hypodermic syringe and filtered (Millipore Swinnex support with a Millipore Durapore (0.45 μm HV) membrane filter). The aliquots were quenched in an ice bath to “stop” the reaction. Each aliquot was analysed by gas chromatography in order to follow reaction progress (Perkin Elmer auto-system, equipped with a DB-5MS capillary column and FID). Product identification was achieved by comparison of retention times with those of injected standard samples.

N

ln k (or ln Kr ) = a0 +

ai Di + i=1

(3)

where ξ stands for the regression residuals, a0 is the regression independent term and ai are the regression coefficients associated with each descriptor, Di, obtained by minimizing the sum of the squared residuals [44]. Experimental data was fitted to eq. (2) using the nonlinear regression and numerical integration Solver Microsoft Excel® add-in to estimate rate constants, k, relative KP values and their corresponding coefficient parameters, and the SolverStat add-in to compute the parameters’ uncertainties. As stated before [13], this procedure allows the use of the totality of each experimental set of results, instead of using specific segments of the kinetic curves, the most frequently used methodology [37,38]. Sound estimates of the unknown parameters, even for relatively small data sets, can be obtained by this simple nonlinear regression method, as long as appropriate starting values are used. Subsequent MLR-QSPRs were performed using Microsoft Excel® by means of an in-house built combined with the Data Analysis ToolPak add-in. This Macro allows the performance of hundreds of multiple linear regressions with up to 9 variables in just a few minutes, the exclusion of undesirable outliers, and a choice of suitable regression parameters according to pre-defined, statistical criteria. In the present case, these criteria were a 95% statistical significance level for each variable, rejection of any variable combinations bearing

Kinetic and QSPR methodologies The Langmuir-Hinshelwood model was assumed for all reactions in order to account for the competition between species for the occupancy of the active sites inside the pore volume of the zeolite [37]. The reaction scheme can be represented as: A(ads) + S(ads) → P(ads) where A stands for the acylating agent, S for the substrate and P for the acylated product. As stated before, these reactions occur in the presence of a large amount of the acylating agent, in order to avoid a premature deactivation of the zeolite by the acylated product. On the other hand, the secondary product, acetic acid, is so limitedly adsorbed that it can be 3

Molecular Catalysis 476 (2019) 110495

R. Aleixo, et al.

Table 1 Degree of crystallinity (CXRD), Si/Al ratio, textural parameters and Brönsted [B]pyridine and Lewis [L]pyridine acid sites concentration. Sample

CXRDa (%)

Si/Alb

Vmicroc (cm³g−1)

Vmesoc (cm³g−1)

Aextc(m2 g−1)

[B]pyridined (mmol g−1)

[L]pyridined (mmol g−1)

BEA12.5 BEA12.5-D BEA12.5-D-AT BEA32 BEA32-D BEA32-D-AT

100 75 71 100 53 47

12.5 8.3 10.6 32.0 15.7 25.7

0.17 0.11 0.15 0.14 0.08 0.08

0.40 0.44 0.44 0.53 0.52 0.56

246 247 236 250 203 205

0.225 0.203 0.235 0.133 0.104 0.041

0.273 0.254 0.198 0.064 0.130 0.030

a b c d

Calculated from powder X-ray diffraction patterns, using BEA12.5 or 32 as references. Total Si/Al ratio estimated from elemental analysis through ICP-OES spectrometry. Textural parameters calculated from N2 adsorption isotherms: Vmicro (micropore volume); Vmeso (mesopore volume), and Aext (external area). Quantified from pyridine adsorption followed by FTIR, at 150 °C.

multicollinearity and removal of outliers if they deviate from the regression line more than 1.5 standard deviations.

BEA12.5-D, as the relative pressure approaches unit the curves of the treated and pristine samples are quite close, pointing out that the decrease of the micropore volume is counterbalanced by the expected mesoporosity development, which is not observed for sample BEA32-D as, in this case, the N2 uptake never reaches the values obtained with the starting material. The values reported in Table 1 show that in both cases the Vmicro decreased 0.06 cm3 g−1, but while the mesopore volume of sample BEA12.5-D is 0.04 cm3 g−1 higher than the value of BEA12.5, the value obtained for BEA32-D is 0.01 cm3 g−1 lower than that of BEA32. In line with the results reported for other zeolite structures [31], and also for BEA structure [29], the subsequent acid leaching had a clear unblocking effect of the microporosity in the case of sample BEA12.5-D-AT, denoted by an upward deviation of the isotherm, corresponding to the increase of Vmicro in 0.04 cm3 g−1. Conversely, the isotherm of BEA32-D-AT sample is practically coincident with the curve of BEA32-D sample, except for very high relative pressures where N2 uptakes are higher than those obtained with BEA32-D. So, in this case the acid treatment affected only the mesopore network that increased 0.04 cm3 g-1 in relation to the value of the alkaline treated sample. As a final remark it should be noted that, in agreement with results reported in literature for BEA structure [29], the starting material used in this study, the values of Vmeso are always higher than those of Vmicro. This difference increases as the structures are submitted to postsynthesis treatments, with Vmeso reaching 7 times the value of Vmicro for sample BEA32-D-AT. The mesopore size distributions presented in Fig. 1 (b) and (d) were obtained through Hybrid Density Functional Theory (DFT Plus (R) V2.01 for ASAP 2010 V5.00) assuming cylindrical pores. The results show that treated samples, as well as starting materials, have broad mesopore size distributions. In the case of BEA12.5 series, there is a very small fraction of pores widths between 2–4 nm, the large majority of pores’ apertures falling in the range 10–20 nm. In accordance with the very steep configuration of the isotherms obtained with BEA32 sample and derived materials, the mesopore size distributions present also a steep configuration indicating the presence of a progressive larger number of pores with width greater than 10 nm. The 29Si MAS-NMR spectra of parent BEA structures and treated samples (Fig. 2) are dominated by the intense resonance at ca. -111 ppm with a shoulder at ca. -114 ppm attributed to two different crystallographic Si(OSi)4 environments (Q4 sites), and a weaker peak at ca. -103 ppm. Because the latter is much enhanced upon cross-polarization (see spectra in the right side of Fig. 2) it is attributed to Si(OH)(OSi)3 environments (Q3 sites) [47], but some contribution from Si(3Si;1Al) sites cannot be discarded [29,47]. Cross-polarization enhances considerably a very week resonance at ca. 92 ppm that witnesses the presence of a small number of Si(OH)2(OSi)2 environments (Q2 sites) [47–49]. To have a deeper insight into the structural changes caused by the post-synthesis treatments, the relative amounts of Qn sites were estimated by fitting the spectra with Gaussian functions using Peak-Fit® software. The results listed in Table 2 confirm that for the parent

Results and discussion Materials characterization The X-ray powder diffraction pattern of the starting materials (see S1 in Supplementary Material) reveal the characteristic peaks of BEA structure [45] and allowed the determination of the degree of crystallinity, CXRD, based on the intensity of the two most intense peaks at 2θ around 8° and 22°, taking the respective parent sample as reference. Results (see Table 1) show that the basic treatment had a more destructive effect upon structure in the case of sample BEA-32-D, which only retained half of the CXRD of the starting material, which is in line with its higher mass loss in comparison with that observed for BEA12.5-D (28% vs 9%). The influence of the Si/Al ratio on samples’ behaviour upon desilication treatment was reported for other zeolite structures, as for example, in the study presented by Groen et al. [46] using MFI structures with a broad range of Si/Al ratios. The authors reported that samples with high Al content are less prone to desilication. In the same line, in the case of BEA structures tested, a lower crystallinity loss is verified for BEA12.5-D sample when compared with BEA32-D. The goal of the second step of the samples’ preparation protocol acid treatment – was to promote the leaching of extra-framework material formed during the basic treatment, in order to disclose the porosity. Thus, it is not surprising that in both cases the mass loss was relatively high (26% and 38% for, respectively, BEA-12.5-D-AT and BEA-32-D-AT), even though the decrease of CXRD values was not significant. This is, indeed, the most commonly reported behaviour [31], but opposite results can also be found in the literature, as in the case of the study carried out by Wang et al. [29] based on the modification of a BEA. In fact, these authors reported an increase of 7% in the crystallinity when the desilicated samples were submitted to a subsequent HCl washing. This behaviour, by comparison with the values obtained in the present study, may be the result of either the influence of a more concentrated NaOH solution, or, most probably, of the different approaches to estimate the CXRD values. Actually, while we have selected the two intense peaks covering a broad range of 2θ values, following usual procedures adopted for other zeolite structures [2], the values reported by Wang et al. [29] were assessed considering only the peak at 2θ around 8°. The N2 adsorption-desorption isotherms of the starting materials and modified structures are shown in Fig. 1 (a) and (c). The curves are typical I + IV isotherms, revealing the characteristic micropore network of zeolite structures along with an important mesopore structure, even in the case of the parent materials, what is interpreted as resulting from aggregation of very small particles. The base treatment results, in both samples, in a downward deviation of the curves denoting the decrease of the micropore volume. However, in the case of sample 4

Molecular Catalysis 476 (2019) 110495

R. Aleixo, et al.

Fig. 1. N2 adsorption-desorption isotherms (left) and mesopore size distributions (right).

samples Q4 sites prevail over Q3 and Q2 sites, not surprisingly since both samples present high Si/Al ratio values (Table 1). The evolution of the global Si/Al ratios assessed through the results of elemental analysis (Table 1) show that both series present the same, predictable, evolution. Actually, the NaOH treatment resulted in the decrease of the Si/Al ratio, especially pronounced in the case of BEA32D, and the subsequent acid leaching led to an increase of Si/Al ratio, as it promotes the removal of extra-framework debris formed during desilication. On the contrary, the analysis of the relative amounts of the different Si-type sites assessed from 29Si NMR data (Table 2) reveal that the treatments had different effects on each of the parent samples. In the case of BEA12.5, the NaOH treatment resulted in the preferential removal of Si atoms linked to Q3 sites, and some structural rearrangement seems to have occurred. As a consequence of the acid treatment, a decrease in Q4 signal along with an increase of Q3 resonance are observed. However, when sample BEA32 was tested, the desilication affected essentially Q4 sites, and no substantial change was noticed in the relative amounts of Q4 and Q3 sites after the acid leaching, which clearly illustrates the effect of Si/Al ratio of the starting materials. Generally speaking, the relative amount of Q2 sites, although much lower than that of any of the other two Si sites, shows a slight increase upon the modification treatments which is most likely linked to the increase of the mesopore volume, as demonstrated by the adsorption data previously discussed. The single-quantum 27Al MAS NMR spectra shown in Fig. 3 exhibit peaks at 54–58 ppm assigned to tetrahedral Al species, and two peaks at 0 (sharp) and -5 ppm (broad) assigned to octahedral Al. The quantification of tetrahedral-to-octahedral Al ratios (Altet/Aloct) was attempted, giving similar values, around 2, for the BEA12.5 sample set and in the

case of BEA32 a decrease from 2.4 to 1.7 is verified upon desilication and a slight decrease to 1.6 is estimated after the acid treatment. However, it must be clarified that the obtained values suffer from high uncertainty associated to the broadness of the bands and the overlapping of the signal (see Fig. 3) so they cannot give conclusive information. The spectra of samples (e) and (f) show a slightly raised baseline at ca. 30 ppm possibly due to the presence of some penta-coordinated Al. 27Al3QMAS NMR provides further insight into the Al species present (Fig. 4 and Fig. S2). Sample BEA12.5-D exhibits two tetrahedral resonances, T1 and T2, with relatively low quadrupole coupling constants, and a third peak, T3, with larger quadrupole effects. In contrast, sample BEA32 displays only T1 and T2 (see Fig.S2). The octahedral region of all samples displays two resonances with small and large quadrupole couplings. Overall, these results are in agreement with those reported by Wang et al. [29] and here we follow their spectral assignments (based on references quoted therein). Accordingly, the resonances T1 and T2 are ascribed to Al atoms residing in different Tsites. T3, in turn, is given by a defective Al framework site, witnessing a transition of the crystalline framework to an amorphous silica-alumina network. O1 is associated with distorted octahedral Al. The assignment of O2 is contentious but it may be ascribed to Al connected to 4 oxygen framework atoms and further coordinated by one water molecule and one hydronium ion. TEM images of parent and treated samples are presented on Fig. 5. The images show aggregates of small crystals, lower than 20 nm and even smaller in the case of BEA32 sample set. The lattice fringes, characteristic of BEA zeolite structure, can be visualized on parent and treated sample, especially on BEA12.5 sample set, which is in accordance with the higher preservation of microporosity as well as 5

Molecular Catalysis 476 (2019) 110495

R. Aleixo, et al.

Fig. 2. 29Si MAS-NMR spectra of (a) BEA12.5; (b) BEA12.5-D; (c) BEA12.5-D-AT; (d) BEA32; (e) BEA32-D; (f) BEA32-D-AT. CPMAS and MAS spectra of sample BEA32 are confronted in the right side of the figure. Table 2 Relative amount (%) of Qn (n = 4-2) sites ascertained from 29Si MAS-NMR, and tetrahedral-to-octahedral Al ratio (Altet/Aloct) determined from 27Al MAS NMR. Sample

Q4

Q3

Q2

BEA12.5 BEA12.5-D BEA12.5-D-AT BEA32 BEA32-D BEA32-D-AT

57 64 53 67 51 50

33 23 32 24 37 35

9 12 15 9 12 15

crystallinity observed on these samples (see Table 1). It can also be observed that, for both sample sets, the evolution of the treatments led to the development of translucid areas, that appear to be more evident in the case of BEA32 sample set, which can be attributed to the presence of mesopores; however, due to the very small size of the crystals, it is not clear from the images if those mesopores are inside the crystal (intracrystalline) or if they are originated from crystal aggregation, leading to intercrystalline mesopores or even macropores, which, in this later case could not be quantified by N2 adsorption isotherms. Fig. 6 shows the infrared spectra in the region corresponding to the adsorption of pyridine to Brönsted (B) and Lewis (L) acid sites (1300 – 1700 cm−1). All samples show the typical bands attributed to pyridine bonded to Brönsted acid sites, PyH+ at around 1545 cm−1, and PyL, at about 1455 cm−1. The sharp band at around 1490 cm−1 is attributed to pyridine bonded to both types of acid sites. Table 1 displays the relative acid sites concentration ([B]pyridine and [L]pyridine) for all samples, estimated from the integrated area of PyH+ and PyL bands using the values of the molar absorption coefficients previously determined by Emeis [36]. As can be observed, independently of the Si/Al ratio of parent BEA, there is a decrease on the concentration of Brönsted acid sites, upon desilication treatments, although this effect is much more

Fig. 3. Single-quantum 27Al MAS NMR spectra of samples (a) BEA12.5; (b) BEA12.5-D; (c) BEA12.5-D-AT; (d) BEA32; (e) BEA32-D; (f) BEA32-D-AT.

significant in the case of BEA32-D. For this sample, an increase in [L]pyridine is also noted, and can be related with the redistribution of Al, extracted during desilication treatment and further redistributed as extra-framework species [50]. Upon acid treatment, different behaviours are observed, depending on the Si/Al ratio of the pristine materials. BEA12.5-D-AT presents a [B]pyridine value that slightly exceeds the Brönsted concentration of parent BEA12.5. This result is not 6

Molecular Catalysis 476 (2019) 110495

R. Aleixo, et al.

For BEA32-D-AT, a continuous decrease on [B]pyridine is perceived, meaning that a continuous loss of Brönsted acid sites has occurred due to both basic and acid treatments. For both sample sets, a decrease in Lewis acidity is observed due to the removal of extra-framework species during the acid washing. Catalytic tests Fig. 7 (A–C) shows the product yields variation with reaction time for the two sets of catalysts. In the case of furan and anisole, the reaction gives an almost total selectivity into products 2-acethylfuran and p-methoxyacetophenone (p-MOAP), respectively, which is predictable because the α-product is preferred which is attributed to the higher stability of the α-intermediate [18]. On the other hand, two isomer products were detected and this can be explained by the higher electrophilic susceptibility of carbon-β when compared with carbon-α, in addition to a lower difference in activation hardness between α and β carbons [18]. The first observation that can be made from the analysis of the plots is that the reactivity order shown in homogeneous media, i.e., pyrrole > furan > anisole is no longer observed. This behaviour was also seen in our previous work with MCM-22 samples [16] and, in both cases, can be ascribed to the particular interactions between substrates and the active sites located inside and at the external surface of the heterogeneous catalysts. In fact, the substrates polarity seems to play a

Fig. 4. 27Al 3Q MAS NMR spectra of BEA12.5-D, and respective projections of anisotropic d2 and isotropic d1 dimensions. T1-3 and O1,2 depict the different tetrahedral and octahedral Al resonances, respectively, and the asterisks spinning side bands.

corroborated by 27Al MAS NMR data (Altet/Aloct ratio on Table 2), which could be anticipated given the uncertainty associated with the determination of these ratios, as mentioned before. So, the slight increase in [B]pyridine for BEA12.5-D-AT can possibly be attributed to the removal or displacement of some extra-framework material as a consequence of the acid treatment, thus improving the access of pyridine to the acid active sites.

Fig. 5. TEM images of parent and modified BEA samples. 7

Molecular Catalysis 476 (2019) 110495

R. Aleixo, et al.

Fig. 6. Infrared spectra of parent and treated samples after pyridine adsorption and evacuation at 150 °C.

determinant role in the reaction environment. In the case of pyrrole, the polarity (as quantified by the normalized Dimroth-Reichardt polarity parameter – ETN) is 0.627, which is significantly high when compared to the same parameter for furan (0.164) and, for that matter, also for anisole (0.198) [51], and can justify the lower product yield of pyrrole when compared to furan and hence the inversion in the reactivity order as referred above. From the analysis of the kinetic curves displayed in Fig. 7, it is evident that, although all curves present the same general trend, that is, show a significant increase in product yield for short reaction times, followed by a substantial slope attenuation after the first 10 min of reaction, different behaviours can be unveiled after a close inspection of the plots, depending on the treatment performed on the parent BEA samples. The results reported by Wang et al. [29] for the acylation of benzene with benzyl alcohol did not reveal significant alterations between parent and treated BEA samples, even though the molecular size of benzene and the main products was around 0.5 nm, like in the case of reactants and products studied in the present work (0.504 for furan, 0.584 for anisole and 0.501 for pyrrole) [51], making all these molecules able to diffuse inside the pore structure of parent and modified BEA. More distinct behaviours were solely noticed by Wang and coworkers when mesitylene was used as substrate, perhaps because its molecular size - around 0.8 nm - exceeds the micropore size of BEA, thus making the reactions to occur mainly at the external surface. In fact, the experimental conditions used by Wang and co-workers during alkaline and acid treatments of BEA led to textural and acidity modifications mostly at the external surface and, accordingly, the main differences on the catalytic behaviour were verified for molecules that could not diffuse inside the pores. However, in the present study, treatment conditions were milder and intracrystalline modifications occurred. Thus, different behaviours were observed using smaller molecules which were able to probe textural and acidity modifications inside the pore structures of the samples. For furan substrate (all tested samples), from Fig. 7A it is clear the effect of the alkaline and alkaline + acid treatment. It is also interesting to notice that distinct behaviours are observed depending on the Si/Al ratio of the parent BEA sample. For BEA12.5 sample set, the desilication treatment causes a significant decrease on product yield that is recovered after acid treatment, even surpassing the parent BEA12.5 catalyst. This behaviour can be preliminarily and qualitatively explained attending to the acidity modifications occurring on BEA12.5 samples during treatments, in line with our previous study with MCM-22 [16]. In fact, upon desilication (BEA12.5-D) some accumulation of extraframework material at the pore mouth of the zeolite porosity may occur, blocking the access of reactants to the active sites located inside the pores. Upon acid treatment the removal or displacement of the extra-framework materials improves the access to the active sites (see also the [B]pyridine trend in Table 1), causing an increase in product yield, especially pronounced for short reaction times. For BEA32

sample set, the catalytic behaviour is dissimilar since higher yields are always obtained with the parent BEA32 sample, independently of the reaction time. This can be justified by the continuous acid site loss as a consequence of alkaline and alkaline + acid treatment. However, the acidity modification cannot explain the increase in product yield verified for BEA32-D-AT, especially considering the loss of more than 50% of Brönsted acid sites when compared to BEA32-D (see Table 1). This behaviour might be explained by the reported modifications in porosity as a result of the treatments, that apparently play an important role in samples with low acid sites’ density as is the case of BEA32 sample set (see Table 1 and Fig. 1). For anisole and pyrrole (Fig. 7 B and C) product yields are substantially lower. In both cases, BEA32-D sample was not tested. In anisole, for BEA12.5 sample set, the behaviour of desilicated and desilicated + acid treated samples is the same as the one previously described for furan, but in this case the differences between the three kinetic curves are more expressive, which can be attributed to the larger dimensions of anisole and respective intermediates and reaction products. As for BEA32, there is an obvious inversion in behaviour when compared to furan, since for BEA32-D-AT product yields are higher than those for BEA32, especially for longer reaction times. Again, the larger dimension of the mesopores formed during the treatments (see Table 1) could be responsible for facilitating the desorption and the consequent diffusion of products from the pore structure of this catalyst. Interestingly for pyrrole, for the BEA12.5 sample set, the order of product yield is BEA12.5 > BEA12.5-D > BEA12.5-D-AT and for BEA32 the treated sample also presents a lower yield when compared to the parent material. This distinct behaviour, vis-à-vis the other two substrates, was already observed in our previous study with MCM-22, in which no differences were disclosed between parent and treated samples [16]. Although the situation is not the same here, it can probably be also attributed to the different zeolite active site-substrate interactions due to the higher pyrrole polarity, when compared to that of the other two substrates. Kinetic modelling The nonlinear regressions approach using the simplified LangmuirHinshelwood model previously presented in detail in an earlier work [13] was also applied in the present case and results are shown in Tables 3–5. Turnover frequencies (TOF) were calculated using the rate constant, k, and the Brönsted acid sites concentration, [B]pyridine, considering that the catalytic reactions occur mainly at the Brönsted sites, in line with what was previously assumed by Guidotti et al. [15] for this type of reactions.

TOF ( min 1 ) =

8

k (mmol min 1 g 1) [B]pyridine (mmol g 1)

(4)

Molecular Catalysis 476 (2019) 110495

R. Aleixo, et al.

display good statistical criteria (high R2 and F values and low sfit). Rate constants of pyrrole and anisole are significantly lower than those of furan and relative adsorption equilibrium constants are much higher for anisole independently of the zeolite’s treatment. For all substrates, alkaline treated materials present lower rate constants than the parent samples and, as stated in a previous work [13], this can be a consequence of extra-framework species deposited at the zeolite’s pore mouths, or of some loss of active sites. Moreover, the higher Kr observed for anisole samples, along with the brown colouring of the zeolite after the catalytic assays, indicates higher accumulation of product inside the pores, as discussed before [13]. QSPR analysis Similarly to the MLR-QSPR methodology previously used [13], ln k and ln Kr were once again correlated to a set of descriptors, of which 7 are related to zeolite properties, namely external area (Aext), micropore volume (Vmicro), mesopore volume (Vmeso), Brönsted acidity ([B]pyridine), Lewis acidity ([L]pyridine), silicon/aluminum ratio (Si/Al) and crystallinity (CXRD)), and 5 to substrate properties, specifically van der Waals volume (VvdW), dipolarity and Lewis acidity (as measured by the normalized Dimroth-Reichardt parameter, ETN), functions of the dielectric constant and the refractive index, f(ε) and g(nD), respectively, and dipole moment, μ. Given the number of data points to be fitted to the QSPR models, the maximum number of descriptors in the fitted models was limited to four, in order to ensure that we always had at least 3 data points for each descriptor. Regressions were performed using MS-Excel 365 and an in-house developed Macro that determined intercorrelations between all used descriptors as shown in Table 6. This Macro automates the fitting procedure as it fits the data to all possible combinations of the allowed descriptors, determines their intercorrelations, excludes outliers and, after one decides which combination to choose using one’s own criteria and previous knowledge of similar work, the user can, with one click, recompute the MLR chosen equation(s) from Excel’s analysis tool pack. All descriptors were previously normalized to allow subsequent comparisons between regression coefficients’ magnitudes in the various derived model equations. Table 6 shows that determination coefficients, r2, for some pairs of variables (in bold) showed values higher than 0.5, which indicates the presence of multicollinearity among the involved descriptors. This means that these descriptors should not be used simultaneously on the fitted models based on this single criterion. Once the best equation model is attained, following the methodology already detailed in a previous work [13] and founded on strict statistical criteria, a double check is performed to verify if multicollinearity produces biased estimates of the regression coefficients, through the calculation of the Variance Inflation Factors (VIFs). The commonly applied threshold of VIF < 10 was used in this work, as recommended by authors such as Hair et al. [52] and Miles [53]. For the rate constant k, the best-found equation was the following:

Fig. 7. Product yield as a function of reaction time using furane (A), anisole (B) and pyrrole (C) as substrates (the dashed curves represent calculated values obtained from the kinetic model).

where N stands for the number of points, sfit is the standard deviation of fit, R2 is the determination coefficient, R2adj is the adjusted determination coefficient (i.e., a “normalized” R2 given by an ANOVA table, which takes in due consideration both the number of variables and the number of points in a regression equation) and F is the FisherSnedecor statistics. The significance level of each descriptor’s coefficient is shown in brackets. VIFs were also computed for the four variables and they were all below 7 indicating that multicollinearity is not biasing coefficients’ estimates.

TOF values are much higher for furan than for any of the other two substrates, accompanying, in general, the variation of the rate constants, k. A comparison between BEA12.5 and BEA32 samples reveal that TOF values are usually higher for the latter, with the sharpest increase observed for BEA32-D-AT, which is consistent with the marked decrease in acidity of these samples (Table 1). The analysis of Tables 3–5 shows that all nonlinear regressions 9

Molecular Catalysis 476 (2019) 110495

R. Aleixo, et al.

Table 3 Summary of rate constants, k, turnover frequencies, TOF, and adsorption parameters of the catalytic reactions using furan as substrate. Also shown are the fit’s statistical figures of merita. Furan

BEA12.5

BEA12.5-D

BEA12.5-D-AT

BEA32

BEA32-D

BEA32-D-AT

k (mmol min−1 g−1) Kr R2 sfit F TOF (min−1)

60 ± 3 5.9 ± 0.9 0.977 0.452 348 462.4

58 ± 3 12 ± 1 0.979 0.375 339 625.3

164 ± 2 14 ± 1 0.998 0.326 3870 952.0

156 ± 3 9.9 ± 0.7 0.996 0.422 2080 2432.8

81 ± 3 12 ± 1 0.992 0.342 796 2023.9

280 ± 3 27 ± 3 0.999 0.382 8562 10753.8

a

R2, determination coefficient; sfit, standard deviation of fit; F, Fisher–Snedecor statistics.

Table 4 Summary of rate constants, k, turnover frequencies, TOF, and adsorption parameters of the catalytic reactions using anisole as substrate. Also shown are the fit’s statistical figures of merita. Anisole −1

k (mmol min Kr R2 sfit F TOF (min−1) a

−1

g

)

BEA12.5

BEA12.5-D

BEA12.5-D-AT

BEA32

BEA32-D-AT

13 ± 1 54 ± 10 0.935 0.152 86 97.5

3.8 ± 0.3 84 ± 12 0.942 0.040 98 40.9

11.9 ± 0.8 40 ± 5 0.953 0.111 145 68.9

15 ± 1 86 ± 13 0.949 0.151 151 240.2

16 ± 1 47 ± 6 0.961 0.139 177 610.8

R2, determination coefficient; sfit, standard deviation of fit; F, Fisher–Snedecor statistics.

Table 5 Summary of rate constants, k, turnover frequencies, TOF, and adsorption parameters of the catalytic reactions using pyrrole as substrate. Also shown are the fit’s statistical figures of merita. Pyrrole −1

k (mmol min Kr R2 sfit F TOF (min−1) a

−1

g

)

BEA12.5

BEA12.5-D

BEA12.5-D-AT

BEA32

BEA32-D-AT

17 ± 1 20 ± 3 0.953 0.162 144 134.6

16 ± 1 26 ± 3 0.953 0.145 144 166.7

11.4 ± 0.9 21 ± 3 0.932 0.125 97 66.3

6.0 ± 0.5 24 ± 4 0.932 0.066 83 93.8

5.7 ± 0.4 39 ± 5 0.949 0.056 112 219.2

R2, determination coefficient; sfit, standard deviation of fit; F, Fisher–Snedecor statistics.

Table 6 Descriptors’ correlation matrix for the 16 studied systems.

CXRD Si/Al Vmicro Vmeso A ext [B]pyridine [L]pyridine VvdW ETN m f(ε) g(nD)

CXRD

Si/Al

Vmicro

Vmeso

A

1.0000

0.0023 1.0000

0.6967 0.0591 1.0000

0.2297 0.6990 0.5016 1.0000

0.7837 0.0436 0.5744 0.3688 1.0000

ext

[B]pyridine

[L]pyridine

VvdW

ET N

m

f(ε)

g(nD)

0.3142 0.5521 0.6162 0.8816 0.5392 1.0000

0.1564 0.7932 0.3136 0.9520 0.3148 0.7904 1.0000

0.0022 0.0001 0.0032 0.0008 0.0049 0.0011 0.0002 1.0000

0.0035 0.0001 0.0050 0.0013 0.0077 0.0017 0.0003 0.2071 1.0000

0.0093 0.0003 0.0135 0.0035 0.0206 0.0045 0.0007 0.0239 0.6551 1.0000

0.0104 0.0004 0.0150 0.0039 0.0228 0.0050 0.0008 0.2659 0.2791 0.8579 1.0000

0.0079 0.0003 0.0114 0.0029 0.0174 0.0039 0.0006 0.0008 0.8158 0.9665 0.7082 1.0000

Fig. 8 shows the good correlation between predicted (by model Eq. 5) and observed ln k values. Open circles correspond to outliers, that is, points for which |ln kpred-ln kobs| > 1.5. These outliers were removed from the data set sequentially and the model refitted after each removal. A close inspection of the equation and a comparison with the results of a similar work based on the MCM-22 zeolite, reveal that one of the most important differences between the behaviours of both zeolites in the Friedel-Crafts acylation of the same set of substrates is the inclusion, as a relevant descriptor in the case of BEA, of the degree of crystallinity. This descriptor contributes negatively to ln k which means that the loss of crystallinity resulting from post-treatment modifications (selective

extraction of Si or, in some cases, Si and Al) improves the reaction rate. On the other hand, the microporous volume, Vmicro, has the most significant, and positive, contribution to ln k. This fact reinforces the observation that these catalytic reactions occur mainly inside the micropores (mesopores are considered important mainly to enhance diffusion). As already referred, all these substrates have dimensions that fit inside the micropores. Additionally, the Brönsted acidity, [B]pyridine also displays a negative coefficient. This probably means that further enhancing this type of acidity is detrimental for the catalyst. In fact, in the case of BEA-32-AT, a decrease of more than 50% of [B]pyridine leads for furan, for which both desilication and desilication + acid treatments were evaluated 10

Molecular Catalysis 476 (2019) 110495

R. Aleixo, et al.

The positive contribution of VvdW, reflects the effect of product bulky molecules which diffuse more slowly and become more adsorbed. As for the positive sign of the ETN coefficient, it means that the combination of the substrate’s HBA and dipolarity, as measured by this descriptor, also promotes product adsorption relative to that of the substrate. Because descriptors were normalized, a direct comparison of their magnitudes is possible, making it meaningful to establish a relative order of importance. Hence, for the adsorption process the order is: VvdW (+) > ETN (+) > Vmicro (-) >> Si/Al (+) whereas for the kinetic process it is: Vmicro (+) > f (ε) (-) > [B]pyridine (-) > CXRD (-) Results from QSPR analysis are consistent with the experimental data retrieved from the structural characterization analysis and the catalytic tests.

Fig. 8. Plot of ln kpred vs. ln kexp resulting from the application of Eq. 5.

Conclusions

Fig. 9. Plot of ln Kr

(pred)

A multivariate QSPR approach, by taking into consideration multiple changes in relevant properties of both substrate and zeolite, has again proved to be a useful tool to correlate physicochemical properties with catalytic results. The observed catalytic behaviour displays, in fact, the influence of several properties acting simultaneously. In a broader sense, the use of this methodology gives information about the key properties that could be optimized through synthesis or post-synthesis treatments to produce catalysts with improved catalytic behaviour. BEA with two Si/Al ratios (12.5 and 32) were submitted to postsynthesis modifications, desilication, followed by acid treatment, in order to modify their textural and acidic properties. The obtained materials were used as environmentally friendly heterogeneous catalysts in Friedel-Crafts acylation reactions using three different substrates, furan, anisole and pyrrole, and acetic anhydride as acylating agent. The Si/Al ratio of the parent BEA zeolites originated different physicochemical characteristics of the modified samples, which are reflected on their catalytic behavior. In the case of BEA12.5 sample set, desilication caused a decrease in product yield and rate constants that is reversed upon acid treatment. This can be attributed to the acidity modifications that occurred during the post-synthesis modification. A dissimilar behaviour is shown for BEA32 sample set where a continuous loss of acid sites is verified as the treatments were performed and, surprisingly, high yield and rate constants are obtained for the samples submitted to alkaline + acid treatment (BEA32-D-AT), that can only be attributed to textural modifications, namely larger mesopores, developed on this sample. To summarize, it can be concluded that the Si/Al ratio of the parent BEA is a key parameter that determines the properties of the treated samples, being responsible for both distinct physicochemical and catalytic behaviors.

vs. ln Kr(exp) resulting from the application of Eq. 6.

(Table 3), to a considerably increase in k (when compared to parent and desilicated samples). This suggests that even for the sample with the lowest Brönsted acidity there are enough acid sites to catalyze the reaction and a higher amount of acid sites will probably promote secondary reactions and adsorption, followed by deactivation phenomena. Finally, the substrates’ dipolarity as measured by f(ε), also lowers the rate constant, suggesting that product molecules may adsorb very strongly, resulting in the occupancy of the active sites and thus preventing further reagent adsorption. For the adsorption related equilibrium constant, Kr, the best-found equation led to following significant descriptors:

Fig. 9 displays again a good correlation between predicted (by model Eq. (6) and observed ln Kr values. As before, three outliers (open circles) were detected and sequentially removed. Also, VIF values were computed and shown to be approximately equal to one, which again, indicate that any existing collinearity is not biasing the obtained coefficients nor their standard deviations. In this case, both ETN and VvdW descriptors contribute positively to ln Kr, as in the case of MCM-22, VvdW being the most important parameter. Vmicro is also significant but affects ln Kr negatively. The Si/Al ratio has a very small positive contribution, which can be linked to the direct correlation of this factor with both acidity types and the increase of adsorption with increased acidity. The negative influence of Vmicro is related with the formation of mesopores from the destruction/enlargement of micropores. The occurrence of secondary reactions/adsorption phenomena occurs mainly inside mesopores and mesoporosity will promote the formation of bulky products that become trapped inside the mesopores and cannot diffuse.

Acknowledgments Authors acknowledge financial support from Fundação para a Ciência e a Tecnologia (FCT), Portugal, through projects PEst OE/QUI/ UI10612/2013, UID/MULTI/00612/2013, UID/Multi/00612/2019, and UID/QUI/00100/2013. Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.mcat.2019.110495. References [1] J.L. Figueiredo, Heterogeneous catalyis: an overview, in: J. faria, J.L. Figueiredo, M.-M. Pereira (Eds.), Catal. from Theory to Appl. Imprensa d, 2008, pp. 1–29. [2] F. Ramôa- Ribeiro, R.M. Guisnet, Les Zeolithes: Un Nanomonde Au Service De La

11

Molecular Catalysis 476 (2019) 110495

R. Aleixo, et al. Catalyse, EDP Science, Les Ullis, 2006. [3] R.A. Sheldon, Fine Chemicals Trough Heterogeneous Catalysis, Wiley, 2008. [4] A.R. Silva, V. Guimarães, L. Carneiro, N. Nunes, S. Borges, J. Pires, A. Martins, A.P. Carvalho, Copper(II) aza-bis(oxazoline) complex immobilized onto ITQ-2 and MCM-22 based materials as heterogeneous catalysts for the cyclopropanation of styrene, Microporous Mesoporous Mater. 179 (2013) 231–241. [5] L.M.D.R.S. Martins, A. Martins, E.C.B.A. Alegria, A.P. Carvalho, A.J.L. Pombeiro, Efficient cyclohexane oxidation with hydrogen peroxide catalysed by a C-scorpionate iron(II) complex immobilized on desilicated MOR zeolite, Appl. Catal. A Gen. 464–465 (2013) 43–50. [6] R.M.G. Sartori, Advances in Friedel-Crafts Acylation Reactions: Catalytic and Green Processes, CRC Press, Boca Raton, 2009. [7] M.G. Clerici, Zeolites for fine chemicals production, Top. Catal. 13 (2000) 373–386. [8] R. Kore, R. Srivastava, B. Satpati, Synthesis of industrially important aromatic and heterocyclic ketones using hierarchical ZSM-5 and Beta zeolites, Appl. Catal. A Gen. 493 (2015) 129–141. [9] K.V.R.I. Sreedhar, H. Kantamneni, K. Suresh Kumar Reddy, Optimal process conditions for zeolite catalyzed acylation of anisole, React. Kinet. Catal. Lett. 55 (2014) 239–242. [10] K. Smith, G.A. El-Hiti, Use of zeolites for greener and more para-selective electrophilic aromatic substitution reactions, Green Chem. 13 (2011) 1579–1608. [11] E.G. Derouane, Zeolites as solid solvents, J. Mol. Catal. A Chem. 134 (1998) 29–45. [12] J. Li, X. Wang, G. Chen, D. Li, Y. Zhou, X. Yang, J. Wang, Hypercrosslinked organic polymer based carbonaceous catalytic materials: sulfonic acid functionality and nano-confinement effect, Appl. Catal. B Environ. 176–177 (2015) 718–730. [13] W.L. Wanyuan Li, Lihua Guo, Anilinotropone nickel catalyzed tandem ethylene oligomerization and Friedel-Crafts addition to toluene, Mol. Catal. 433 (2017) 122–127. [14] S.Z. Zichao Miao, Zhenbin Li, Jinping Zhao, Weijiang Si, Jin Zhou, MoO3 supported on ordered mesoporous zirconium oxophosphate. An efficient and reusability solid acid catalyst for alkylation and esterification, Mol. Catal. 444 (2018) 10–21. [15] M. Guidotti, J.-M. Coustard, P. Magnoux, M. Guisnet, Acetylation of aromatics over acid zeolites: seeking a viable alternative to Friedel-Crafts catalysts, Pure Appl. Chem. 79 (2007) 1833–1838. [16] R. Aleixo, R. Elvas-Leitão, F. Martins, A.P. Carvalho, A. Brigas, A. Martins, N. Nunes, Kinetic study of Friedel-Crafts acylation reactions over hierarchical MCM22 zeolites, Mol. Catal. 434 (2017) 175–183. [17] J.C. Jansen, E.J. Creyghton, S.L. Njo, H. Van Koningsveld, H. Van Bekkum, On the remarkable behaviour of zeolite Beta in acid catalysis, Catal. Today 38 (1997) 205–212. [18] V.F.D. Álvaro, A.F. Brigas, E.G. Derouane, J.P. Lourenço, B.S. Santos, Mild liquidphase Friedel-Crafts acylation of heteroaromatic compounds over zeolite Beta, J. Mol. Catal. A Chem. 305 (2009) 100–103. [19] H. Wei, K. Liu, S. Xie, W. Xin, X. Li, S. Liu, L. Xu, Determination of different acid sites in Beta zeolite for anisole acylation with acetic anhydride, J. Catal. 307 (2013) 103–110. [20] Ch. Baerlocher, Lynne B. McCusker, D.H. Olson, Atlas of Zeolite Framework Types, 5th edition, Elsevier, 2001. [21] V. Paixão, A.P. Carvalho, J. Rocha, A. Fernandes, A. Martins, Modification of MOR by desilication treatments: structural, textural and acidic characterization, Microporous Mesoporous Mater. 131 (2010) 350–357. [22] J. Pérez-Ramírez, C.H. Christensen, K. Egeblad, C.H. Christensen, J.C. Groen, Hierarchical zeolites: enhanced utilisation of microporous crystals in catalysis by advances in materials design, Chem. Soc. Rev. 37 (2008) 2530–2542. [23] J.C. Groen, S. Abelló, L.A. Villaescusa, J. Pérez-Ramírez, Mesoporous beta zeolite obtained by desilication, Microporous Mesoporous Mater. 114 (2008) 93–102. [24] A. Martins, A.P. Carvalho, N. Nunes, Hierarchical zeolites: preparation, properties and catalytic applications, in: M. Aliofkhazraei (Ed.), Compr. Guid. Mesoporous Mater, Vol. 3 Prop. Dev., Nova Science Publishers, 2015, pp. 147–211. [25] S. van, D.R. Chal, C. Gérardin, M. Bulut, Overview and industrial assessmentof synthesis strategies towards zeolites with mesopores, ChemCatChem. 3 (2011) 67–81. [26] Y. Yan, X. Guo, Y. Zhang, Y. Tang, Future of nano-/hierarchical zeolites in catalysis: gaseous phase or liquid phase system, Catal. Sci. Technol. 5 (2015) 772–785. [27] E.G. Derouane, I. Schmidt, H. Lachas, C.J.H. Christensen, Improved performance of nano-size H-BEA zeolite catalysts for the Friedel-Crafts acetylation of anisole by acetic anhydride, Catal. Letters 95 (2004) 13–17. [28] X. Ji, Z. Qin, M. Dong, G. Wang, T. Dou, J. Wang, Friedel – Crafts acylation of anisole and toluene with acetic anhydride over nano-sized Beta zeolites, Catal.

Letters 117 (2007) 171–176. [29] Y. Wang, Y. Sun, C. Lancelot, C. Lamonier, J.C. Morin, B. Revel, L. Delevoye, A. Rives, Effect of post treatment on the local structure of hierarchical Beta prepared by desilication and the catalytic performance in Friedel-Crafts alkylation, Microporous Mesoporous Mater. 206 (2015) 42–51. [30] D. Verboekend, S. Mitchell, M. Milina, J.C. Groen, P. Javier, Full Compositional Flexibility in the Preparation of Mesoporous MFI Zeolites by Desilication, J. Phys. Chem. C (2011) 14193–14203. [31] V. Machado, J. Rocha, A.P. Carvalho, A. Martins, Modification of MCM-22 zeolite through sequential post-synthesis treatments. Implications on the acidic and catalytic behaviour, Appl. Catal. A Gen. 445–446 (2012) 329–338. [32] R. Monteiro, C.O. Ania, J. Rocha, A.P. Carvalho, A. Martins, Catalytic behavior of alkali-treated Pt/HMOR in n-hexane hydroisomerization, Appl. Catal. A Gen. 476 (2014) 148–157. [33] V. Paixão, R. Monteiro, M. Andrade, A. Fernandes, J. Rocha, A.P. Carvalho, A. Martins, Desilication of MOR zeolite: conventional versus microwave assisted heating, Appl. Catal. A Gen. 402 (2011) 59–68. [34] S. Brown, S.P. Heyes, S.J, Wimperis, Two-dimensional MAS multiple-quantum NMR of quadrupolar nuclei. Removal of inhomogeneous second-order broadening, J. Magn. Reson. 119 (1996) 280–284. [35] K.S.W. Sing, S.J. Gregg, Adsorption, Surface Area and Porosity, 2nd edition, Academic Press, London, 1982. [36] C.A. Emeis, Determination of integrated molar extinction coefficients for infrared absorption bands of pyridine adsorbed on solid acid catalysts, J. Catal. 141 (1993) 347–354. [37] E.G. Derouane, G. Crehan, C.J. Dillon, D. Bethell, H. He, S.B. Derouane-Abd Hamid, Zeolite catalysts as solid solvents in fine chemicals synthesis: 2. Competitive adsorption of the reactants and products in the Friedel-Crafts acetylations of anisole and toluene, J. Catal. 194 (2000) 410–423. [38] E.G. Derouane, C.J. Dillon, D. Bethell, S.B. Derouane-Abd Hamid, Zeolite catalysts as solid solvents in fine chemicals synthesis 1. Catalyst deactivation in the friedelcrafts acetylation of anisole, J. Catal. 187 (1999) 209–218. [39] F.M.C. Ventura, Application of quantitative structure–activity relationships to the modeling of antitubercular compounds. 1. The hydrazide family, J. Med. Chem. 51 (2008) 612–624. [40] F.M.M. Reis, R.E. Leitão, Enthalpies of solution of 1-butyl-3-methylimidazolium tetrafluoroborate in 15 solvents at 298.15 K, J. Chem. Eng. Data 55 (2010) 616–620. [41] M.R.F. Martins, S. Santos, C. Ventura, R. Elvas-Leitão, L. Santos, S. Vitorino, J.V. Miranda, H.E. Correia, J. Aires-de-Sousa, V. Kovalishyn, D.A.R.S. Latino, M.V. Ramos, Design, synthesis and biological evaluation of novel isoniazid derivatives with potent antitubercular activity, Eur. J. Med. Chem. 81 (2014) 119–138. [42] F.M. M.H. Abraham, Human skin permeation and partition: general linear freeenergy relationship analyses, J. Pharm. Sci.-Us. 93 (2004). [43] L.M.F. Martins, R.E. Leitão, Solvation effects in the heterolyses of 3-X-3-methylpentanes (X = Cl, Br, I), J. Phys. Org. Chem. 17 (2004) 1061–1066. [44] F. Martins, L. Moreira, L. Moreira, R. Leitao, Solvent effects on solution enthalpies of adamantyl derivatives, J. Therm. Anal. Calorim. 100 (2010) 483–491. [45] M.M. Treacy, J.B. Higgins, Collection of Simulated XRD Powder Patterns for Zeolites, Elsevier, London, 2001. [46] J.C. Groen, J.C. Jansen, J.A. Moulijn, J. Pe, Optimal Aluminum-Assisted Mesoporosity Development in MFI Zeolites by Desilication, (2004), pp. 13062–13065. [47] R. Hajjar, Y. Millot, P.P. Man, M. Che, S. Dzwigaj, Two kinds of framework Al sites studied in BEA zeolite by X-ray diffraction, Fourier Transform Infrared spectroscopy, NMR techniques, and V probe, J. Phys. Chem. C. 112 (2008) 20167–20175. [48] K. Leng, Y. Wang, C. Hou, C. Lancelot, C. Lamonier, A. Rives, Y. Sun, Enhancement of catalytic performance in the benzylation of benzene with benzyl alcohol over hierarchical mordenite, J. Catal. 306 (2013) 100–108. [49] G.L.H.M. McMillan, J.S. Brinen, J.D. Carruthers, A 29Si NMR investigation of the structure of amorphous silica—alumina supports, Colloids Surf. 33 (1989) 133–148. [50] P. Matias, J.M. Lopes, S. Laforge, P. Magnoux, M. Guisnet, F. Ramôa Ribeiro, nHeptane transformation over a HMCM-22 zeolite: catalytic role of the pore systems, Appl. Catal. A Gen. 351 (2008) 174–183. [51] Y. Marcus, The Properties of Solvents, Wiley, 1998. [52] J.F. Hair Jr, W.C. Black, B.J. Babin, R.E. Anderson, Multivariate Data Analysis, 7th edition, Prentice Hall, 2010. [53] J. Miles, Tolerance and Variance Inflation Factor, Wiley StatsRef, 2014, https://doi. org/10.1002/9781118445112.stat06593 Statistics Reference Online.

12