Controlling Pd site location within hierarchical ZSM-5Hierarchical H+ form ZSM-5 (H-MZSM5) arises from the post-synthetic treatment of the parent zeolite (H-ZSM-5) with NaOH, evidenced by a characteristic Type IV isotherm, Barrett–Joyner–Halenda (BJH) pore size analysis, and electron microscopy, as shown in Fig. 1 and Supplementary Fig. 1. The formation of mesopores, with an average diameter of 8.6 nm (Fig. 1b), gives rise to a fivefold increase in mesopore volume, while a relative crystallinity of 80.6% and comparable micropore volume confirm preservation of micropore network (Supplementary Table 1 and Supplementary Fig. 1b). The presence of mesopores is subsequently exploited for the selective deposition of metal NPs solely within these domains through size exclusion, as illustrated in Fig. 1a. Pd NPs, possessing an average diameter of 4.6 ± 1.3 nm (Supplementary Fig. 2), are deposited exclusively outside of the micropores of the hierarchical architecture (PdNP/H-MZSM5). This spatial control is governed by the difference in pore diameter, 8.6 nm (meso) vs 0.54–0.56 nm (micro), thus, preformed Pd NPs can reside only at the external surface and/or within the mesopores of H-MZSM5. Furthermore, spatial separation of Pd and acid sites is realised with Brønsted acid sites (i.e. Al-O(H)-Si) principally located within the retained micropores. Conversely, conventional impregnation and thermal processing of a Pd salt (Pdimp/H-MZSM5) cannot achieve this active site spatial separation (Fig. 1).Fig. 1: Catalyst synthesis and physicochemical properties.a Graphic illustration of catalyst synthesis. b Pore size distribution based on the BJH method with Faas correction using the adsorption branch of N2 physisorption isotherms. c Pd:Si molar ratio by X-ray photoelectron spectroscopy (XPS), hard X-ray photoelectron spectroscopy (HAXPES) and inductively coupled plasma optical emission spectroscopy (ICP). d TEM image of PdNP/H-MZSM5. e HAADF-STEM image of PdNP/H-MZSM5 (inset: Pd particle size distribution). f HAADF-STEM image of Pdimp/H-MZSM5 (inset: Pd particle size distribution). The error on the average particle sizes reported is the standard deviation of the dataset.High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) of PdNP/H-MZSM5 and Pdimp/H-MZSM5 shows a distinct difference in the distribution and size of Pd within the porous support framework. The conventional preparation route, namely Pdimp/H-MZSM5, produces highly dispersed small Pd NPs (2.2 ± 0.6 nm) in conjunction with sub-1-nm Pd clusters, as shown in Fig. 1f and Supplementary Fig. 1. These species are randomly distributed throughout the three-dimensional hierarchical ZSM-5 skeleton, including within micropores, as confirmed by HAADF-STEM energy dispersive x-ray spectroscopy (EDX), consistent with previous investigations in which metal nanoparticles larger than the micropore diameter have been confirmed inside the zeolite framework44. Thus, such a synthesis approach endorses Pd integration within the micro and mesopores, i.e. uncontrollably across the entire hierarchical pore network and hence non-selectively45. The combination of Pd NPs and clusters results in a Pd dispersion of 26.5% (Supplementary Table 2). In contrast, PdNP/H-MZSM5 displays only Pd NPs, with an average diameter of 4.4 ± 0.6 nm (Fig. 1e), which are consistent with the parent unsupported NPs and naturally possess a lower dispersion (11.9%, Supplementary Table 2). Therefore, given their sizes and the size of the two distinct pore domains, and in conjunction with the imaging, Pd NPs reside selectively and exclusively in the mesopores and outer surface of the support matrix, as evidenced by Fig. 1d, e, for PdNP/H-MZSM5, whilst Pd is present in the micro and mesopores in Pdimp/H-MZSM5 (Fig. 1f). It is worth pointing out that incorporating Pd, via either route, does not lead to micropore blockage. Comparable micropore properties to the parent zeolite (Supplementary Table 1) provides evidence of the highly accessible hierarchical framework structure.Further evidence of the difference in Pd distribution within the two catalysts is afforded through the combination of inductively coupled plasma (ICP) (Supplementary Table 2), X-ray photoelectron spectroscopy (XPS) and hard X-ray photoelectron spectroscopy (HAXPES) (Fig. 1c and Supplementary Fig. 3). Bulk elemental analysis (ICP) confirms comparable Pd:Si ratios, 0.0045 for 0.81 wt% PdNP/H-MZSM5 and 0.0057 for 1.02 wt% Pdimp/H-MZSM5, respectively. Due to the inherent sampling depths of XPS and HAXPES, ~9.6 nm for Al Kα vs ~36 nm for Ga Kα, for which 63% of the signal originates from the top 3.2 and 12 nm, respectively, they provide pseudo surface vs pseudo bulk analysis46, i.e. they probe depth-dependence of Pd species across the two catalysts. A significantly higher Pd concentration at the surface (external and within directly accessible mesopores) is identified by XPS for PdNP/H-MZSM-5, whilst HAXPES reveals an intermediate Pd:Si ratio of the three techniques. This trend completely flips for Pdimp/H-MZSM5, with comparable ratios for HAXPES and ICP and a significant drop for XPS, indicative of significant levels of Pd migration into the bulk of the zeolite skeleton during impregnation, a feat not observed of PdNP/H-MZSM5. Both catalytic materials consist of Pd metal, with only a trace amount of PdO detected in Pdimp/H-MZSM5, again reflecting the smaller NP size and, thus, increase in surface-to-bulk ratio and, therefore, greater surface PdO concentration47.Three-dimensional tomography reconstructions of PdNP/H-MZSM5 and Pdimp/H-MZSM5, produced from tilt-series HAADF-STEM imaging, were established to study the distribution of Pd NPs through the two catalysts (Fig. 2 and Supplementary Videos 1–4) and further validate the distinct difference in Pd location between the two catalysts. Semi and non-transparent rendering of the zeolite framework, Fig. 2a–d, shows a degree of Pd located on the external surface within PdNP/H-MZSM5, whilst the same is not apparent in Pdimp/H-MZSM5. Consistent with our expected deposition locations within the two systems, Pd within Pdimp/H-MZSM5 is more uniformly distributed through the meso and micropores, while Pd in PdNP/H-MZSM5 is compartmentalised at external sites and within directly accessible mesopores. Cross-section slices through the reconstructions (Fig. 2e, f and Supplementary Fig. 4), taken from the centre of the zeolite particle, reinforce our deduction that Pd in PdNP/H-MZSM5 is located within mesopores close to the external surface of the zeolite crystal, whereas Pdimp/H-MZSM5 comprises Pd NPs within the interior region of the framework, with species present within the micropores and mesopores. For Pd residing within the micropores, the NPs can span multiple micropores or can induce the partial collapse of the localised framework. Thus, Pd NPs larger than a single micropore exists but are encapsulated within the micropore framework, i.e. accessible only via a micropore44. Identification of Pd in the same location in the cross-section slices above and below those in Fig. 2, as shown in Supplementary Fig. 4, verifies our identification of Pd within the tomogram and the corresponding slices.Fig. 2: STEM tomogram reconstructions.a, c PdNP/H-MZSM5 and b, d Pdimp/H-MZSM5, with the Zeolite set as semi-transparent in (a, b) and non-transparent in (c, d). Cross-section slices from the centre of the catalyst are presented in e PdNP/H-MZSM5 and f Pdimp/H-MZSM5.Mutually exclusive sites within hierarchical ZSM-5Having developed a protocol to selectively deposit Pd NPs within mesopores of hierarchical ZSM-5, the attention can turn to the intrinsic Brønsted acid sites within ZSM-5. The aim to truly segregate the two different sites is complicated given the capacity for mesopores, in addition to micropores, to accommodate surface acid sites, both Brønsted and Lewis, the latter due to extra-framework aluminium species48. Therefore, removing such surface acidity within the mesopores is paramount to completely segregating two active species. The deployment of a bulky chelating agent, diethylenetriaminepentaacetic acid (DPTA), with a molecule diameter of ~0.9 nm (Supplementary Fig. 5) permits selective extraction of Al species (via complexation) from the mesopores/outer surface areas only49,50, as depicted in Fig. 3a. Post-DPTA treatment (H-MZSM5-DA) crystallinity and porosity of the zeolite framework are unaffected, as evidenced by XRD and N2 porosimetry (Supplementary Fig. 6 and Supplementary Table 1). Furthermore, this dealumination treatment restores Si:Al close to that of the parent ZSM-5, 34 (H-MZSM5-DA) and 40 (H-ZSM-5), an increase from 24 for H-MZSM5 (resulting from the initial desilication).Fig. 3: Investigation of acid site location in H-MZSM5 and H-MZSM5-DA.a Schematic of H-MZSM5 dealumination. b Schematic of using 2,6-di-tert-butylpyridine (DTBPy), 3,5-dimethylbenzylamine (DBAM) and ethylenediamine (EDA) as probes to titrate acid sites. c ATR-IR spectra for the DTBPy-titrated H-MZSM5 and H-MZSM5-DA and their pure zeolite equivalents. d ATR-IR spectra for DBAM- and EDA-titrated H-MZSM5 and H-MZSM5-DA. e NMR relaxation T1/T2 ratios of n-octane and water in H-MZSM5 and H-MZSM5-DA and corresponding (T1/T2)n-octane/(T1/T2)water. The relative error on T1/T2 measurements is ~±5%.Acid-base titration using 2,6-ditertbutylpyridine (DTBPy), 3,5-dimethylbenzylamine (DBAM in acetone), and ethylenediamine (EDA) enables either partial or total titrations of the acid sites within the hierarchical zeolite catalysts due to steric limitations for DTBPy and DBAM relative to EDA. The large molecular dimension of DTBPy, kinetic diameter of 1.05 nm51, limits its interaction to solely acid sites within the mesopores and on the external surface. Attenuated total reflection infrared (ATR-IR), shown in Fig. 3c, confirms DTBPy only adsorbs on H-MZSM5. The band at ~1530 cm−1 is attributed to the DTBPyH+ ion52 from the interaction of DTBPy with Brønsted acid sites53, thus validating the absence of acid sites at the external surface and within mesopores for the DPTA-treated zeolite (H-MZSM5-DA). In comparison, in addition to titrating external and mesopore-bound acid sites, DBAM, based on the pendant amine chain length, probes acid sites at a depth of up to ~0.3 nm into a micropore (Fig. 3b). CH stretching bands are observed in the region of 2800–3000 cm−1 by ATR-IR, shown in Fig. 3d, with the intensity of these bands for H-MZSM5-DA greatly reduced relative to the parent material. This reduced acid site density is consistent with the successful removal of the mesopore and external surface acidity, with DBAM adsorbed at acid sites within the micropores close to the mesopores and external surface, i.e. within the first zeolite cage from the mesopores or external surface. In contrast, EDA is unselectively adsorbed, titrating sites in the micropores, mesopores, and outer surfaces. CH stretch intensities reveal only a slight drop post-DPTA treatment (Fig. 3d), which again is consistent with the selective removal of sites from the mesopores and external surface only. Based on the combined results from the acid-base titrations, it can be concluded that DPTA treatment of H-MZSM-5 selectively extracts acidity from the mesopores and external surfaces only, with the characteristic micropore acidity of the zeolite, even in the micropores adjoining the mesopores and external zeolite surface, retained.NH3-TPD (Supplementary Fig. 6c) further confirms the decrease in total acidity. Weak and strong acid peak intensity, evaluated by desorption below and above 300 °C, respectively, reveals H-MZSM5-DA possesses lower weak (0.219 vs 0.320 mmol g−1) and strong acidity (0.340 vs 0.396 mmol g−1), with the loss of weak site symptomatic of the removal of Al from outside the micropore framework. A side effect of removing acidity is the potential change in the hydrophobic-hydrophile nature of the material54, with NMR relaxation used to probe internal hydrophobicity/hydrophilicity55,56. As shown in Fig. 3e, Supplementary Fig. 7, and Supplementary Table 3, T1 and T2 NMR relaxation times of water and n-octane in H-MZSM5 and H-MZSM5-DA probed variations in hydrophobicity. After post-treatment with DPTA, the zeolite framework becomes more hydrophobic, confirmed by a 16% reduction of the T1/T2 of water, i.e. 1.5 vs 1.8. Conversely, the interaction with the organic non-polar hydrocarbon probe, n-octane, is enhanced, with a T1/T2 ratio of 15.4 for H-MZSM5-DA vs 10.6 for H-MZSM5. The subsequent ratio of the T1/T2 values of n-octane to water is indicative of the relative surface affinity towards hydrocarbons relative to water57,58. As shown in Fig. 3e, DPTA removal of Al increases hydrophobicity (and correspondingly the affinity for hydrocarbons), which has been shown to be beneficial for catalytic reactions in which water is the by-product through its displacement from the catalyst59,60,61. The interaction of a model substrate and intermediate, 1-nonanoic acid and 1-nonanol, respectively, with the zeolites (H-MZSM5 and H-MZSM5-DA) can be considered to arise from a combination of the non-polar hydrocarbon chain and polar functional group, i.e. an intermediate of the two initial probes (water and n-octane) rather than one dominating over the other (shown in Supplementary Fig. 8 and Supplementary Table 3). Dealumination results in an increase in interaction for both model substrate and intermediate, reflecting a significant affinity between the hydrocarbon chain and the increased hydrophobic mesopore surface rather than a decreased interaction due to the repulsion of the polar functional group.Having successfully extracted acidity from the mesopores and external surface to produce H-MZSM5-DA, the selective deposition of Pd, via the route deployed for PdNP/H-MSM5, yields a spatially orthogonal system with true active site compartmentalisation, i.e. acid sites only in micropores and Pd solely in mesopores and on external surfaces (PdNP/H-MSM5-DA). Supplementary Fig. 9 and Supplementary Table 1 reveal its analogous nature to its parent zeolite (where applicable) and PdNP/H-MZSM-5, with equivalent porosity (surface areas, pore volumes and pore diameter), PdNP size (4.5 ± 0.9 nm from HAADF-STEM), and Pd:Si atomic ratios from surface, semi-bulk, and bulk analysis (0.0125 for XPS, 0.0118 for HAXPES, and 0.0048 for bulk analysis, to be compared with Fig. 1c) substantiating their comparable physicochemical nature. However, close inspection of the Pd 3d XPS reveals a subtle shift to lower binding energy. Given the identical Pd particle sizes for PdNP/H-MSM5 and PdNP/H-MSM5-DA, this shift in energy must arise from the interaction with the acid site and is not a result of size effects. This shift results from interactions between the mesopore-located acid sites (both Brønsted and Lewis) and the Pd NPs within PdNP/H-MZSM-5. The acidity imparts an δ+ charge on the Pd NPs, and hence a higher binding energy62,63. Thus, removing these acid sites eradicates this influence on the electronic nature of the deposited Pd NPs, i.e. further evidence of the absence of mesopore acid sites in PdNP/H-MSM5-DA. XPS of Pdimp/H-MZSM5 reveals a larger impact on Pd binding energy, attributed to the smaller PdNP size, here being both a particle size effect and a greater Pd-acid site interaction. PdNPsmall/H-MZSM5 (Supplementary Fig. 9), prepared using pre-reduced Pd NPs with a diameter of only 2.7 ± 0.5 nm, shows an identical degree of mesopore acid site influence.Catalytic cascade HDO of Lauric acid to alkanesWe propose that the spatial segregation within the reported catalytic systems should be beneficial to a cascade hydrodeoxygenation that proceeds stepwise as follows: reduction (carboxylic acid to alcohol) in the mesopores over Pd, dehydration (alcohol to alkene) in the micropores over acid sites, and finally hydrogenation (alkene to alkane) over Pd in the mesopores, with the alkane product diffusing to the bulk reaction media, i.e. cascade intermediates are contained within the zeolite framework and subsequently consumed until the final product is generated. The stepwise cascade pathway is illustrated in Fig. 4, which is also observed for Ni/ZSM-5 and Mo/ZSM-2264,65. Observing cascade intermediate species is dependent on the catalyst configuration and species reactivity, e.g. carbonyl and alkene intermediates are often not observed due to rapid conversion36,37, i.e. they are consumed before they can accumulate in the bulk solution.Fig. 4: HDO cascade reaction pathway of lauric acid to dodecane.a HDO of lauric acid monitored from dodecane yield over PdNP/H-MZSM5-DA, PdNP/H-MZSM5, Pdimp/H-MZSM5, PdNPsmall/H-MZSM5 and physical mixture of H-MZSM5 + PdNP/Na-MZSM5. Reaction conditions: 100 mg 1 wt% Pd-doped catalyst, 300 mg lauric acid, 0.1 cm3 of nonane as the internal standard, 40 cm3 of hexane as the solvent, at 200 °C, 30 bar and 1000 rpm for 6 h. b Dehydration of lauryl alcohol of H-MZSM5 and H-MZSM5-DA to evaluate the rate of intermediate conversion, with the inset showing the location of dehydration reactions within the zeolite. Error bars in (a, b) correspond to the standard deviation of the average taken over at least two independent measurements.HDO of lauric acid (dodecanoic acid) evaluates the effectiveness of PdNP/H-MZSM-5 and PdNP/H-MZSM5-DA relative to Pdimp/H-MZSM5 and a physical mixture of monofunctional catalysts, H-MZSM5 (acid only) and PdNP/Na-MZSM5 (Pd only). The latter was prepared by applying the same synthesis strategy as PdNP/H-MZSM5 to the Na+ form equivalent zeolite, which possesses similar Pd properties (Supplementary Table 2 and Supplementary Fig. 10), albeit with a reduced Pd binding energy due to electron donation from the alkali metal66. Optimised reaction conditions are established on dodecane total yield (combination of n-dodecane, n-C12, and iso-dodecane, iso-C12), shown in Supplementary Table 4 and Supplementary Fig. 11. The process also yields undecane (C11), through undesirable DCOx, this being the main by-product detected (Supplementary Fig. 12), while trace amounts of lauryl laurate from the esterification reaction between the fatty alcohol and acid are also detected. The concentration of the latter does not increase significantly over the duration of the reaction, suggesting that its formation is a minor pathway. The impact of pressure shows at ≤20 bar, only selectivity is influenced whilst conversion is comparable, whereas increasing pressure to 30 bar drives up conversion while maintaining higher C12 selectivity. The impact of stirring rate shows no effect, as presented in Supplementary Fig. 11, ruling out bulk mass diffusion effects.The synergy between the two sites, viz. metal and acid, is paramount to enabling HDO. Control experiments over one-site-only systems, i.e. H-MZSM-5 and PdNP/Na-MZSM5 (Table 1), reveal acidity alone is unable to convert the starting substrate. In comparison, the sole deployment of Pd via PdNP/Na-MZSM5 produces only the first intermediate product (1-dodecanol) at low levels, consistent with the initial reduction step of the cascade67. A physical mixture of these two materials does produce a C12 alkane yield (Fig. 4a), thus confirming the criticality of a union of metal and acid sites for the catalytic cascade HDO. While also demonstrating the power of coupling these two sites in driving the cascade reaction forward, i.e. consumption of the alcohol by the acid sites drives the equilibrium increasing acid reduction, hence higher lauric acid conversion relative to PdNP/Na-MZSM5. However, a physical mixture does not permit governance over the order in which the different catalytic active species are encountered and is thus far from optimal9,24,25,26. When this reaction parameter is controlled by deploying PdNP/H-MZSM5, it yields a significant enhancement, with dodecane yield increasing by ~20% and a sevenfold increase in production rate, a direct consequence of dictating the reaction sequence of the cascade. Decreasing PdNP size whilst retaining spatial compartmentalisation, i.e. PdNPsmall/H-MZSM5, does not reveal an anticipated kinetic enhancement from the increased Pd dispersion, i.e. faster lauric acid conversion from increased surface Pd sites. In fact, it results in reduced C12 selectivity. Thus, decreasing Pd particle size harms overall HDO performance. This correlation between HDO performance and PdNP size is further evident from the further reduced HDO performance of Pdimp/H-MZSM5, with a greater impact, at least partially, due to accessibility, i.e. Pd in micropore being less accessible. The drop in the capacity towards HDO, including lauric acid conversion, is in conjunction with an increase in C11 selectivity. The latter arising from the proximity, i.e. adjacent, of Pd and acid sites, with their interaction favouring decarboxylation over HDO68, and the decrease in Pd size69. Thus, decreasing Pd particles is detrimental to overall HDO performance, exacerbated by contiguous Pd and acid sites, i.e. when co-located within the micropores or mesopores of the zeolite framework. However, comparing HDO performance for materials with comparable average particle sizes, impregnation vs small Pd NPs, further highlights the benefit of spatial segregation on the HDO cascade process, with reduced C11 formation resulting from suppressing interaction between Pd and acid sites.Table 1 Lauric acid HDO performance over conventional prepared and spatial compartmentalised catalysts.aWhile it is expected that compartmentalisation of Pd into the mesopores of PdNP/H-MZSM5 is likely beneficial for substrate diffusion, given the dimension of lauric acid (Supplementary Fig. 13), we acknowledge that a degree (albeit minor based on the NH3-TPD) of acid sites co-exist within the mesopores. This raises the question regarding which sites are responsible for the alcohol dehydration step, i.e. acid sites within mesopores or micropores. The dehydration of lauryl alcohol over H-MZSM5 and H-MZSM5-DA (acid only zeolites), as shown in Fig. 4b, reveals almost identical reaction rates (1.62 and 1.74 mmol h−1, respectively), confirming that Brønsted acid sites within the micropores of the zeolite framework drive dehydration. Thus, Pd and acid sites separated at the nanoscale operate individually for their respective steps within the cascade process but collaborate for overall optimal HDO performance. Furthermore, as shown in Fig. 4b and Supplementary Fig. 14, the rate of dehydration is significantly greater than that of lauric acid consumption (~0.3 mmol h−1) for PdNP/H-MZSM5. The initial reduction step of the acid to the alcohol is, therefore, slower than the subsequent alcohol dehydration step, which is further supported given that cascade intermediate species are not observed, i.e. intermediates are consumed rapidly within the pore architecture of the bifunctional catalysts as substrates for the subsequent step in the cascade before having time to diffuse into the bulk reaction media. Likewise, alkene hydrogenation is another facile step in the cascade and is thus consumed before building up in the reaction media36,37.HDO activity further escalates with true active site compartmentalisation, with PdNP/H-MZSM5-DA showing the highest rate of dodecane production (30.6 moldodecane molPd−1 h−1), a 2.5-fold increase on PdNP/H-MZSM5, and C12 selectivity. Furthermore, the rate of lauric acid consumption increases by a comparable factor (~2). As previously observed in the non-dealumianted system, the incorporation of small Pd NPs (PdNPsmall/H-MZSM5-DA) with a greater number of surface sites is not conducive to improved activity relative to its larger NP equivalent (PdNP/H-MZSM5-DA). Likewise, it induces increasing C11 formation, although now solely due to reduced NP size69. That said, achieving true compartmentalisation once again is highly beneficial, as evident by the amplified activity and conversion relative to the impregnated and non-dealuminated equivalents (Pdimp/H-MZSM5-DA and PdNPsmall/H-MZSM5, respectively). While PdNPsmall/H-MZSM5 does give rise to a significant increase in performance relative to Pdimp/H-MZSM5, it does not reach the level of PdNP/H-MZSM5-DA or PdNPsmall/H-MZSM5-DA, confirming the benefit of spatial orthogonality of active species on catalytic HDO. The origin of the enhancement from spatial separation of the active species resulting from the removal of the mesopore acidity is attributed to the coeffect of; (i) the elimination of the undesirable interaction between acid sites and Pd NPs, and (ii) the increased mesopore hydrophobicity, which aids the expulsion of water (the by-product of acid reduction), driving the localised equilibrium forward increasing alcohol production. Furthermore, H2 dissociation on the PGM sites has been shown to be hindered by the presence of water70. Concurrently, increasing hydrophobicity within PdNP/H-MZSM5-DA does not negatively impact interaction with the fatty acid (discussed above) and, thus, would not negatively impact cascade process efficiency. The inherent benefit of active site spatial compartmentalisation within PdNP/H-MZSM5-DA is further evident from its capacity for HDO of other fatty acids, including palmitic (C16) and stearic (C18) (Supplementary Table 5).Catalyst recycling studies show that PdNP/H-MZSM5-DA exhibits good recyclability (Supplementary Figs. 15, 16) without requiring reactivation between runs, whilst PdNP/H-MZSM5 shows reasonable recyclability (TON drop from 688 to 574 over three cycles). In contrast, PdNP/H-ZSM-5, comprising Pd NPs deposited on the external surface of micropore-only ZSM-5, exhibits poor catalyst recyclability, deactivating by ~90% over only two runs. HAADF-STEM imaging confirms Pd agglomeration is the root cause of this, which arises from the lack of mesopore confinement of the Pd NPs. Thus, the catalyst configuration possessed by PdNP/H-MZSM5 and PdNP/H-MZSM5-DA bestows yet another highly beneficial asset on our catalytic materials. Moreover, the formation of coke, another common deactivation pathway in zeolite catalysis, can also be ruled out. Thermogravimetric analysis of PdNP/H-MZSM5-DA (Supplementary Fig. 17) shows mass losses only below 300 °C, which we attributed to arise from the desorption of adsorbed substrate (dodecanoic acid b.p. 299 °C), intermediates (dodecanal b.p. 242 °C, dodecanol b.p. 259 °C, and dodecene b.p. 213 °C), and product (dodecane b.p. 216 °C), whereas combustion of coke species requires temperatures ≥500 °C. So, while species are adsorbed, which accounts for some of the mass balance of the process (the remaining missing mass may reflect low solubility and precipitation of the substrate during sampling), it is reasonable to assume they remain as species that partake in the cascade process. This varies from studies into unsaturated triglycerides where coking and oligomerisation are apparent due to reactivity from unsaturation in the initial substrate71.To further explore this approach and evaluate the impact of intrinsic zeolite acidity, the optimal synthetic route was applied to ZSM-5 with Si:Al ratios of 25 and 15. Increasing the inherent acidity of the catalytic system shows little effect on overall conversion (Supplementary Table 6), note the higher substrate concentration and lower metal loading, with only a decrease observed at the lowest Si:Al ratio. In contrast, selectivity towards HDO is compromised by the increasing acidity, which negatively impacts C12 alkane productivity. Furthermore, switching to other zeolite frameworks, namely USY and Beta, further demonstrates the generic nature of the approach for catalyst synthesis. While resulting in superior acid conversion, it is at the expense of C12 alkane selectivity.In conclusion, the successfully developed a hierarchical ZSM-5 catalyst involving spatial compartmentalisation of Pd (mesopores) and acid (micropores) sites, and its influence over the one-pot HDO of lauric acid to dodecane has been demonstrated. Compared with conventional catalyst production, via impregnating with Pd salts, Pd within the spatially separated catalysts resides within mesopores and on the outer surfaces, as evidenced by electron tomography. This facilitates greater accessibility for the initial acid reduction step of the cascade, whereas the dehydration of lauric acid, the second step in the cascade, is shown to occur over acid sites within the micropores of hierarchical ZSM-5, potentially within micropores bordering mesopores. Furthermore, the segregation of Pd and acid sites impedes undesirable side reactions, i.e. decarboxylation and decarbonylation, while the elimination of mesopore acidity, tunes the hydrophobicity of the mesopores and interactions between Pd and acid sites, which further accelerates overall HDO performance. Moreover, this approach prohibits Pd blocking/binding to the acid sites, another undesirable aspect of producing such bifunctional materials. However, we do point out that here, within our Pdimp/H-MZSM5, the acid site of the parent zeolite significantly outweighs the amount of Pd impregnated, and thus, the resulting surface Pd atoms could, in theory, only block a maximum of 3.5% of the acid sites present, i.e. the impact of such an effect can be considered inconsequential.The benefit of employing a spatially orthogonal catalyst, through active site compartmentalisation, for lauric acid HDO is further evident when it is benchmarked against the current state-of-the-art HDO catalysts (Supplementary Table 7), in which Pd is deployed either as an active species, promotor or stabiliser. Compared with mono-metallic Pd systems29, the higher selectivity of the catalyst reported here drives greater HDO performance, as while higher reaction temperatures deployed in the literature enhance conversion, this is at the expense of favouring DCOx over HDO. Furthermore, the benefit of spatial orthogonality within PdNP/H-MZSM5-DA is further evident from its superior performance to Pd supported randomly on acidic alumina silicates (Pd/Al-SBA-15, Pd@Al-SiO2)29,33, and zeotype materials (Pd/SAPO-31)29, which again is due to superior selectivity. However, relative to acidity co-deposited on silicas (Pd-Nb2O5/SiO2 and Pd/HPA-SiO2)34,72 and bimetallic systems (Pd/CuZnAl + ZSM-5 and Pd-Re/C + zeolite A)40,43 it is the greater mol normalised activity of PdNP/H-MZSM5-DA that gives rise to greater HDO productivity. Comparisons are also made with Pt-based systems, with either comparable (Pt/Nb2O5) or superior (Pt/ZSM-5, Pt/TiO2, Pt-Re/TiO2) alkane productivities observed. Therefore, it is a combination of activity and selectivity that results in the optimal HDO performance of PdNP/H-MZSM5-DA, which equals or outperforms the current literature demonstrating the advantage of nanoscale active species segregation in catalytic cascade HDO processes. Thus, this study should inspire future HDO catalyst development while also highlighting the potential of spatial separation of active species within multifunctional materials for one-pot cascade reactions, i.e. catalysts that can facilitate the combination of multiple steps of a chemical production process within a single reactor. Furthermore, it demonstrates a generic synthesis approach for the production of bifunctional materials, in which different functionalities are compartmentalised within two distinct regions.
