Palladium-doped hierarchical ZSM-5 for catalytic selective oxidation of allylic and benzylic alcohols

Hierarchical zeolites have the potential to provide a breakthrough in transport limitation, which hinders pristine microporous zeolites and thus may broaden their range of applications. We have explored the use of Pd-doped hierarchical ZSM-5 zeolites for aerobic selective oxidation (selox) of cinnamyl alcohol and benzyl alcohol to their corresponding aldehydes. Hierarchical ZSM-5 with differing acidity (H-form and Na-form) were employed and compared with two microporous ZSM-5 equivalents. Characterization of the four catalysts by X-ray diffraction, nitrogen porosimetry, NH3 temperature-programmed desorption, CO chemisorption, high-resolution scanning transmission electron microscopy, X-ray photoelectron spectroscopy and X-ray absorption spectroscopy allowed investigation of their porosity, acidity, as well as Pd active sites. The incorporation of complementary mesoporosity, within the hierarchical zeolites, enhances both active site dispersion and PdO active site generation. Likewise, alcohol conversion was also improved with the presence of secondary mesoporosity, while strong Brønsted acidity, present solely within the H-form systems, negatively impacted overall selectivity through undesirable self-etherification. Therefore, tuning support porosity and acidity alongside active site dispersion is paramount for optimal aldehyde production.

mesopore generation. Milina et al. [25] reported that the conversion of alkylation between benzyl alcohol and toluene was up to 90% over a hierarchical ZSM-5, while only 7% over commercial microporous ZSM-5. Moreover, hierarchical zeolites also provide additional pores for metal particle deposition, extending the application of zeolites beyond petrochemicals and small-molecule activation [22]. Martens et al. [26] used Pt-doped hierarchical ZSM-22 for hydro-isomerization of nonadecane and pristane, the latter requiring a lower temperature to obtain 20% conversion compared to Pt-doped microporous ZSM-22 (180°C versus 230°C).
Here, we report an investigation of Pd-doped ZSM-5 for aerobic selox of alcohols. The Pd-doped microporous-mesoporous hierarchical ZSM-5 in H-form (acidic) and Na-form (relatively neutral), prepared via desilication, are compared with corresponding Pd-doped microporous ZSM-5 to elucidate the role of ZSM-5 pore structure as well as intrinsic acidity in Pd impregnation and selox activity.

Preparation of ZSM-5 supports
NH 4 -form ZSM-5 zeolite (Si/Al = 25) was calcined at 550°C for 5 h (ramp rate 5°C min -1 ) to convert it to the H-form. The zeolite was denoted as HMic-Z.
The H-form ZSM-5 (Si/Al = 25) was ion-exchanged with an aqueous 0.2 M Na 2 CO 3 solution at 60°C for 0.5 h (1 g zeolite per 30 ml solution). The solid was isolated by filtration and washed to pH 8 with deionized water. The solid was dried at 100°C for 6 h and calcined at 550°C for 4 h (ramp rate 5°C min -1 ). The zeolite was denoted as NaMic-Z.
The desilicated ZSM-5 was prepared by the methods reported by Perez-Ramirez and co-workers [27]. NH 4 -form ZSM-5 zeolite (Si/Al = 40) was calcined at 550°C for 5 h (ramp rate 5°C min −1 ) to convert it to the H-form. Six grams of H-ZSM-5 was vigorously stirred in aqueous NaOH solution (0.2 M, 30 ml g −1 of zeolite) at 65°C for 0.5 h. The solid was recovered by centrifugation and was washed with deionized water to pH 8. Ion exchange was carried out in an aqueous NH 4 NO 3 solution (0.1 M, 30 ml g −1 of zeolite) at 80°C for 3 h. The product was isolated by centrifugation and washed to a neutral pH. The NH 4 ion exchange process was repeated twice. The material was dried at 100°C for 6 h and calcinated at 550°C for 4 h (ramp rate 5°C min -1 ). The zeolite was denoted as HDe-Z. The same procedure was employed with the omission of the NH 4 ion exchange step to produce a Na-form denoted as NaDe-Z.

Pd impregnation
The microporous and hierarchical ZSM-5 supports were functionalized with Pd by wet impregnation. One gram of the support was stirred in an aqueous Pd(NH 3 ) 4 (NO 3 ) 2 solution, with salt concentration adjusted to give a 1 wt% Pd loading. The slurries were stirred vigorously at 25°C for 18 h before being heated to 50°C to dry. The solids were calcined in air at 500°C for 2 h (ramp rate 1°C min −1 ), prior to reduction at 400°C for 2 h (ramp rate 10°C min −1 ) under flowing H 2 (10 ml min −1 ). Pd contents were confirmed by inductively coupled plasma-optical emission spectroscopy, with Pd loading of 1.03 wt% Pd/NaDe-Z, 0.96 wt% Pd/NaMic-Z, 1.02 wt% Pd/HDe-Z and 1.04 wt% Pd/HMic-Z.
Nitrogen adsorption/desorption isotherms of materials were measured by using a Micromeritics ASAP 2000 analyser. The materials were degassed under vacuum at 350°C for 12 h prior to nitrogen adsorption at -196°C. BET surface area was calculated over 0.01-0.15 relative pressure, and the microporous area was calculated over 0.4-0.6 relative pressure.
Ammonia temperature-programmed desorption (NH 3 -TPD) analysis was conducted with a Micromeritics AutoChem II 2920 chemisorption analyser. Catalysts (approx. 200 mg) were degassed and dried at 550°C for 1 h and then cooled to 50°C under He. A gas mixture of NH 3 in He (10% : 90%, 30 cm 3 min −1 ) was then introduced to saturate the catalyst followed by He purge (60 cm 3 min −1 ) at 50°C for 2 h to remove the physically adsorbed NH 3 . Finally, NH 3 -TPD was performed by heating the catalyst from 50 to 600°C at 10°C min −1 under He (30 cm 3 min −1 ), and the desorbed NH 3 was monitored by gas chromatography (GC) using a thermal conductivity detector.
CO pulse chemisorption was performed with a Quantachrome chemBET 3000 system to measure Pd metal surface area. Samples were degassed at 150°C for 1 h under He (20 ml min −1 ) before reduction at 100°C for 1 h under H 2 . The mild reduction condition can avoid additional particle sintering. The CO chemisorption analysis was carried out at room temperature with a CO/Pd stoichiometry of 1 : 1 [10].
X-ray photoelectron spectroscopy (XPS) was conducted with a Kratos Axis SUPRA using monochromated Al kα (1486.69 eV) X-rays at 15 mA emission and 12 kV HT (180 W) and a spot size/ analysis area of 700 × 300 µm. Spectral fitting was conducted using CasaXPS v. 2.3.19PR1.0. The binding energy of each sample was calibrated based on the Si 2p peak at 103.4 eV.
Fluorescence Pd K-edge (24.35 keV) XAS was conducted at Diamond Light Source on beamline B18 using a Si [3 1 1] monochromator, Pt-coated mirrors and a Vortex multichannel fluorescence detector. Spectra were processed using Demeter v. 0.9.26. Athena was used for normalization, background subtraction and linear combination fitting of XANES, and Artemis for extended X-ray absorption fine structure (EXAFS) fitting. Reference spectra of a Pd foil and PdO were also collected.
High-resolution transmission electron microscopy (TEM) was conducted with an FEI Tecnai F20 FEG TEM operating at 200 kV equipped with an Oxford Instruments X-Max SSD EDX detector (10 nm spot size). High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images were recorded on an FEI Titan3 Themis G2 operated at an accelerating voltage of 300 kV, equipped with a field-emission gun (X-FEG) operating at an extraction voltage of 4.5 kV, a monochromator and an FEI Super-X 4-detector EDX system.

Aerobic selective oxidation of alcohols
Catalyst screening was conducted in a two-neck 50 ml round-bottom glass flask at 90°C fitted with a water condenser. Cinnamyl alcohol (or benzyl alcohol) (1.1 g, 8.4 mmol) was dissolved in toluene (10 ml) with mesitylene (0.1 ml) employed as an internal standard. The reaction solution was heated to 90°C with agitation at 800 r.p.m. The catalyst (50 mg) was added to the reactor. Periodic reaction sampling was conducted by withdrawing an aliquot (0.25 ml). The aliquots were filtered into a vial and diluted with toluene (1.75 ml). The diluted extracted solutions were analysed by GC.

Catalyst support properties
The properties of the zeolite supports were evaluated to confirm the retention of the zeolite structure in the hierarchal zeolites and compare physico-chemical properties of HMic-Z, NaMic-Z, HDe-Z and NaDe-Z. As presented in figure 1a, XRD patterns for all four samples show well-resolved peaks in the range of 5-75°, which are indexed to the characteristic pattern of the MFI zeolite structure (JCPDS data, card no. 01-084-0385), and confirm the retention of the zeolite cage structure within the hierarchical desilicated systems with long-range crystal ordering preserved [27].
As shown in figure 1b, nitrogen porosimetry reveals that HMic-Z and NaMic-Z both exhibit Type I isotherms, reflecting monolayer adsorption. According to IUPAC, this corresponds to microporous royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 8: 211086 solids with relatively small external surfaces, such as zeolites [28]. By contrast, HDe-Z and NaDe-Z present Type IV isotherms with a characteristic hysteresis loop due to the capillary condensation occurring in mesopores. BJH pore size distributions (figure 1c) show mesopores in HDe-Z and NaDe-Z concentrated at around 15 and 25 nm, respectively, which are considerably larger and more abundant than pores within HMic-Z and NaMic-Z. Quantitative physical properties from nitrogen porosimetry are reported in table 1. Comparison of BET surface areas of hierarchical mesoporousmicroporous desilicated ZSM-5 against microporous ZSM-5, either H form or Na form, shows both exhibit similar total surface area. However, there is a significant decrease in micropore surface areas for the hierarchical support materials, which indicates that a degree of the micropores is converted into mesopores during desilication. This is also apparent from the increased pore volume. We proposed that these mesopores will prove beneficial during the subsequent catalysis through enhanced accessibility of reactants to the active sites and shortened micropore diffusion path length [29]. Comparing the two hierarchical supports, HDe-Z possesses a higher mesopore area than NaDe-Z (335 versus 179 m 2 g −1 ). This may arise from the dissolution of Al debris (resulting from desilication) during the NH 4 NO 3 step [30] and may also explain the mesopore size distribution differences.  royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 8: 211086 The acidity of the support materials was probed by NH 3 -TPD, as illustrated in figure 1d. Both HMic-Z and HDe-Z show a peak at approximately 350°C, indicating strong acidity from the Brønsted acid sites of Si-O(H + )-Al groups of zeolite framework [27]. HMic-Z shows a slightly greater intensity as the desilication process inevitably disrupts part of the zeolite framework. By contrast, the acid strengths of the two Na-form ZSM-5 are much weaker. The weak acidity in the four catalysts resulted from extra-framework Al, which yields predominately Lewis acid sites [30]. Abad et al. deposited Au onto nanocrystalline CeO 2 to generate Lewis acidity and evaluated the effect for aerobic selox of 3-octanol [31]. Here, Lewis acidity enhanced the TOF from 130 to 420 h −1 . However, to our knowledge, there is no evidence that the Lewis acidity in extra-framework Al can likewise improve Pd-catalysed selox, and given the comparable levels of Lewis acidity within our systems is a parameter we are unable to assess here.

Pd-doped zeolite characterization
The four zeolite supports were impregnated with 1 wt% palladium to produce a series of ZSM-5-based selox catalysts. Powder XRD patterns (electronic supplementary material, figure S1) show retention of the parent zeolite framework, and thus the impregnations and thermal processing showed no negative impact. An additional peak at 40°, corresponding to the Pd(111) reflection, is apparent in all samples except 1 wt% Pd/NaDe-Z. The absence indicates average Pd particle size of less than 2 nm [32]. Scherrer volume-average particle sizes of the Pd nanoparticles for the other three samples are reported in table 2. For comparison, Pd particle sizes estimated by STEM and CO chemisorption are also provided. These confirm a general decrease in Pd size within hierarchical mesoporous-microporous zeolites and Na-form zeolites. However, we do point out that inherent limitations hinder all techniques. These are lower size detection limits by XRD, potential for inaccessibility to probe molecules within porous architectures during CO chemisorption due to pore blockage, and minute sample size by STEM. Since only CO chemisorption reflects truly accessible Pd, this can be considered the more appropriate method for comparison of the four catalysts. N 2 porosimetry of the Pd-doped ZSM-5 catalysts (electronic supplementary material, figure S2) shows isotherm and hysteresis (where present) types, along with pore size distributions, are generally unchanged after Pd impregnation. However, as shown in table 2, Pd incorporation results in a decrease in micropore surface area. Some blockage of micropores by palladium nanoparticles may be the cause of this [10]. To further investigate the porosity of the four Pd-doped zeolites, the highresolution TEM images are shown in figure 2. The structures of ZSM-5 particles in Pd/NaMic-Z and Pd/HMic-Z are comparable to each other but not to the equivalent forms subjected to desilication. For these (Pd/NaDe-Z and Pd/HDe-Z), the presence of complementary intracrystalline mesopores is clear, yielding hierarchical mesoporous-microporous ZSM-5. In figure 2b, for Pd/HDe-Z, clear lattice fringes of the zeolite are apparent, thus zeolite crystallinity is preserved after selective extraction of Si from ZSM-5 framework and Pd impregnation, which concurs with wide-angle XRD.
The acidity of the Pd/ZSM-5 materials, assessed by NH 3 -TPD, is reported in the electronic supplementary material, figure S3, and shows Pd impregnation had only a minor influence on acidity. While acid strength across all four zeolites is unaffected, the concentration of weak acid sites (less than 250°C) does reduce slightly in all four catalysts. Weak acid sites are attributed to extraframework Al [33,34], so these might provide an anchoring site for a small proportion of Pd nanoparticles. However, this decrease is negligible, with a significant degree of weak acidity still present.
Given the significant difference in atomic mass between Pd and the support elements, the z contrasting nature of HAADF-STEM is ideal for observing Pd sites within the ZSM-5 frameworks. Representative images are shown in figure 3 and electronic supplementary material, figure S4, while EDX images with corresponding STEM images are presented in the electronic supplementary material, figure S5. Pd nanoparticles on the H form of the zeolites, both microporous only and hierarchical, displayed a high proportion of Pd nanoparticle diameters spanning 1-3 nm; there is also a small degree of highly sintered species (approx. 100 nm). Pd/NaMic-Z exhibits a slightly broad average size, ranging from 2 to 10 nm, while the Pd on the hierarchal equivalent showed smaller average particle sizes, centred at approximately 2 nm and spanning 1-7 nm. Large (greater than 20 nm) Pd agglomerations in the Na form are not witnessed. The optimal dispersion of Pd/NaDe-Z was further confirmed by CO chemisorption, as presented in table 2, and is consistent with the XRD results. Pd dispersion of the remaining catalysts follows the following trend: Pd/HDe-Z > Pd/NaMic-Z > Pd/HMic-Z. This suggests the open nature of the hierarchical zeolite enables either greater dispersion, in contrast with XRD (which may be skewed due to size limitations of the technique and the presence of a small proportion royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 8: 211086  4), with the Pd 3d doublet in all samples deconvoluted to two oxidation states, metallic (Pd 0 at 335.4 eV) and an oxidized state (Pd 2+ at 337.3 eV) [35]. Pd 2+ : Pd 0 , also presented in figure 4, shows Pd/HMic-Z to contain the highest Pd 2+ content, followed closely by Pd/HDe-Z and reflects the high degree of surface PdO formation on the small Pd nanoparticles, which increases with increasing surface-to-bulk ratio [10]. The Na forms show lower Pd 2+ content, which is inconsistent with the higher metal dispersion previously discussed. However, XPS will preferentially probe Pd on the external surface of the zeolite.
In contrast with XPS, XAS provides bulk characterization allowing evaluation of Pd on the external surface and within the pore framework. Normalized XAS spectra of the four catalysts are shown in figure 5, with XANES shown in the insert. The XANES linear combination fitting results, using reference PdO and Pd foil spectra, allowed the evaluation of Pd 2+ and Pd 0 content (reported in  table 3). Here, we observed a correlation of PdO content with dispersion, with the hierarchical zeolite catalysts containing significantly higher PdO content, with the Na form displaying the highest proportion, while the two microporous-only zeolite systems possess comparable PdO density.
Further insight was gained from EXAFS, as shown in figure 6 and electronic supplementary material, table S1. The dominant scatterers in Pd/NaDe-Z are oxygen at 1.996 Å in the first shell and Pd in the second shell at 2.997 Å, consistent with the structure of PdO [36], although with reduced average coordination number (CN) of 2.8 and 1, respectively. A small contribution at 2.731 Å, attributed to metallic Pd, was also present, albeit with a CN much reduced from that of Pd foil (3.67 versus 12). These results were similar to those of our early work for highly dispersed Pd over Al-grafted SBA-15 [37], in which we considered the Pd nanoparticles to consist of a metallic core encapsulated by a PdO shell, surface oxidation arising due to exposure to air during storage. The low CNs here reflect minute average particle sizes in agreement with CO chemisorption. Pd/HDe-Z showed similar results, although with an increase in CN to 6.3 for the metallic phase, suggesting a larger metal core. Pd in the two microporous ZSM-5 samples was dominated by metallic Pd scatterers, with CN of the order   [36,38,39]. Guillemot et al. [38] have proposed the relationship between first shell CN and nanoparticle size, from which we assess the diameter of Pd/HMic-Z and Pd/ NaMic-Z to be 2.2 and 2.9 nm, respectively.

Alcohol selective oxidation
Cinnamyl alcohol selox was employed to screen and benchmark the four Pd/ZSM-5 catalytic materials. Figure 7 shows the introduction of complementary mesoporous structure within the hierarchical mesoporous-microporous zeolites, to impart enhanced conversion relative to the equivalent microporous catalyst; 1 wt% Pd/NaDe-Z showed the highest conversion of approximately 70% at 6 h and 60% selectivity to the desired aldehyde (electronic supplementary material, figure S6). The main by-product was hydrocinnamic acid (19%). This arises from over-oxidation of aldehyde and concurrent C=C hydrogenation. While hydrogenation under oxidizing conditions appears contradictory, under insufficient oxygen supply to the catalyst surface this has been previously observed over model Pd(111) [40] and 'real' Pd/SiO 2 systems [10].
While 1 wt% Pd/NaMic-Z resulted in similarly high aldehyde selectivity, conversion was servery limited. We attribute to the reduced accessibility of the Pd active sites within the micropore framework and reduced Pd dispersion according to CO chemisorption. The catalysts based on H-form zeolites displayed a similar enhancement with mesopore incorporation. However, cinnamaldehyde yields were much reduced, to around 20%, with the major product consisting of a high boiling point compound(s). GC-MS (electronic supplementary material, figure S7) suggested dicinnamyl ether formation, a previously reported by-product during cinnamyl alcohol Friedel-Crafts alkylation of toluene over zeolite Y [41]. To further investigate the etherification side reaction and the possibility of royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 8: 211086 strong acid sites playing a role, 1 wt% Pd/HDe-Z was treated with Na 2 CO 3 in mild condition to remove strong Brønsted acidity. As shown in the electronic supplementary material, figure S8, the capacity for cinnamaldehyde production was significantly enhanced and closely matches that of 1 wt % Pd/NaDe-Z. This results from switching off of etherification and thus indicating strong acidity to    Benzyl alcohol selox performance, as shown in the electronic supplementary material, figure S10, displays a similar enhancement with the incorporation of mesoporosity. Selectivity to the desired aldehyde is still favourable over systems without strong acidity, with benzaldehyde selectivity reaching approximately 90% for both 1 wt% Pd/NaDe-Z and 1 wt% Pd/NaMic-Z, compared to 50% for the two H-forms. The increased unsaturated aldehyde selectivity reflects a reduction in undesirable ether by-product formation. Etherification was again confirmed by GC-MS, identifying dibenzyl ether (electronic supplementary material, figure S11). This confirmation of the nature of the unknown high boiling point compounds, present in both reactions, indicates a competing reaction over the co-present strong Brønsted acids sites in HDe-Z and HMic-Z (electronic supplementary material, figure S3). Similarly, mesopores are also shown to escalate this undesired side reaction rate, which can again be attributed to superior internal substrate diffusion. Thus for optimal selox performance, strong zeolite acidity is undesirable, while complementary mesopores and highly dispersed Pd sites are desirable, the latter leading to enhanced nanoparticle surface PdO concentration, as shown in table 4.  1 wt% Pd/HDe-Z 1 wt% Pd/HMic-Z 1 wt% Pd/NaDe-Z 1 wt% Pd/NaMic-Z 1 wt% Pd/HDe-Z 1 wt% Pd/HMic-Z 1 wt% Pd/NaDe-Z 1 wt% Pd/NaMic-Z