Shape-persistent porous organic cage supported palladium nanoparticles as heterogeneous catalytic materials †

Porous Organic Cages (POCs) are an emerging class of self-assembling, porous materials with novel properties. They offer a key advantage over other porous materials in permitting facile solution processing and re-assembly. The combination of POCs with metal nanoparticles (NPs) unlocks applications in the area of catalysis. In this context, POCs can function as both the template of ultra-small NPs and a porous, but reprocessable, heterogeneous catalyst support. Here, we demonstrate the synthesis of ultra-small Pd NPs with an imine linked POC known as ‘CC3’, and show that hydrogen gas can be used to form metallic NPs at ∼200 °C without the reduction of the organic cage (and the accompanying, unwanted loss of crystallinity). The resulting materials are characterized using a range of techniques (including powder diffraction, scanning transmission electron microscopy and synchrotron X-ray absorption spectroscopy) and shown to be recrystallizable following dissolution in organic solvent. Their catalytic efficacy is demonstrated using the widely studied carbon monoxide oxidation reaction. This demonstration paves the way for using ultra-small NPs synthesized with POCs as solution-processable, self-assembling porous catalytic materials.


Introduction
Porous materials such as zeolites, metal organic frameworks (MOFs), 1 covalent organic frameworks (COFs) 2 and porous polymers 3 attract considerable attention for applications such as gas storage, separations, sensing and catalysis. 4 These materials can be rationally designed to exhibit desirable properties such as porosity and selectivity. 5 The synthetic diversity of this class of materials permits the incorporation of numerous and highly diverse chemical functionalities in the pore structure. 6 When employed as catalysts or catalyst supports, porous materials' high surface areas, tuneable porosity and chemical functionality can enhance performance, as well as improving separation and recovery relative to homogeneous catalysts. 7 Many industrial catalysts comprise nanoscale metal nanoparticles (NPs) on inorganic porous materials such as high surface area alumina or fumed-silica. 8 Metal NPs confined in well-defined pore structures, such as metal organic frameworks (MOFs), have recently begun to be investigated as heterogeneous catalysts. 9 The pore cavities have the potential to confine the growth of metal NPs, affording excellent particle size control and, in turn, improving the surface area and possibly selectivity and stability of the catalyst. Additionally, synergistic interactions between the metal catalyst and porous support have been suggested to enhance catalytic performance. 10 However, taking advantage of porous materials in this manner, and ensuring good control of the size, composition, spatial distribution, confinement and accessibility of ultrasmall metal NPs in the pore structures remains challenging. 10 Self-assembly of discrete, shape-persistent cage moleculesknown as porous organic cages (POCs)is a rapidly expanding field of study. 11 These materials differ from other porous materials in that they comprise individual molecular building blocks, the cages, which assemble into a macroscopic structure. This is in clear contrast to zeolites or MOFs that are typi-cally synthesized as an extended structure. This decoupling of assembly from synthesis is a transformative difference, because it enables solution processing of POC-derived materials: they can be readily dissolved in solvents to their molecular form, then re-assembled as solids, membranes or thin films. 12 Such materials are already of widespread interest for applications such as gas storage, gas separation and molecular recognition. 5 The use of POCs to control the synthesis and assembly of ultra-small metal NPs is therefore highly appealing as a route to make ultra-small NPs, but also to enable facile solution processing compared to other porous material templates. 10 'Mix and match' assembly strategies using more than one cage molecule type can permit the fabrication of both porous organic alloys, 13 and functional composites. 14 The possibility of including multiple components in these self-assembling materials unlocks further possibilities for controlled and tuneable fabrication of bimetallic or multicomponent systems in the context of nanoparticle and catalyst syntheses.
To date, there are only a few attempts to fabricate ultrasmall metal NPs using POCs as stabilizing and templating agents. [15][16][17][18][19][20] These have allowed metal NPs to be homogenized in solution or used as a heterogeneous catalyst for a range of liquid phase reactions (Suzuki coupling, 18 nitrophenol hydrogenation, 16 cyanation of aryl halides, 17 amine borane decomposition, 16 photoreduction of nitroarene to azobenzenes 20 ) and have demonstrated the potential of organic cage molecules to template metal NPs with uniform size. However, the two types of organic cage systems employed in these examples are either not crystalline/prone to self-assembly or contain imine linkages that are reduced to amines by NaBH 4 (a reducing agent being necessary to reduce the metal salt precursor). This conversion of the imines to amines is known to lead to the POCs losing their shape persistence, crystallinity and porosity. 21 It therefore remains an important challenge to prepare metal NPs templated by shape persistent POCs, without loss of their crystallinity. It is their crystallinity that is crucial for applications requiring facile processability and reassembly of these highly ordered materials.
Here, we describe a simple and efficient synthetic approach that enables production of metal NPs with controlled size (around 1.6 nm) using the shape-persistent POC CC3 ( Fig. 1(a)). This is achieved with the material and NPs retaining the ability to self-assemble into porous crystalline form, demonstrating their potential for easy post-synthesis processing. Fig. 1 (a) Molecular structure of porous organic cage -CC3. (b) Crystal structure depictions of thermodynamic CC3α polymorph to show the pore network connectivity around one molecule (the accessible pore structure is seen from the connolly surfaces in red generated using a probe radius of 1.82 Å), the local cage packing motif, and the wider crystal structure. (c) Schematic representation of the overall synthesis of Pd NPs@CC3 material (DCM = dichloromethane, rectangles are the CC3 cage material, the 0.5 wt% Pd(NO 3 ) 2 @CC3 is presumed to contain cages that incorporate Pd 2+ species).

Results and discussion
As can be seen from Fig. 1(a), CC3 is an imine linked cage with a tetrahedral shape. The cage has four windows allowing access for guest species to the internal void. When crystallized in the absence of guest species, the thermodynamically most stable polymorph, denoted CC3α, is formed. CC3α shows a window-to-window cage-packing motif, which results in an interconnected three dimensional (3D) pore network as shown in Fig. 1(b). The pores comprise both the internal void of the cage but also extrinsic voids that result from inefficient packing of the cages, leaving spaces between the CC3 molecules. The limiting pore size that will govern access for guest species to this pore network can, however, be approximated by the cage windows. In this work, Pd(NO 3 ) 2 was consequently selected as the metal precursor because, in contrast to larger metal precursors, the size of one Pd(NO 3 ) 2 complex means it is likely to fit into the CC3 molecule ( Fig. S1 †), via these cage windows. The initial encapsulation of Pd(NO 3 ) 2 inside the cage cavities in solution was anticipated to be important in enabling the metal to become distributed within the porous solid upon crystallization. By way of contrast, similar experiments with the larger Pd(acac) 2 metal precursor, which is not expected to fit within the cage cavities, produced a material which exhibited very poor Pd particle size control (demonstrating the critical importance of the size of the Pd(NO 3 ) 2 relative to the size of the cage molecule cavity, Fig. S2 † vs. Fig. 3).
The synthesis of the Pd NPs@CC3 is summarized schematically in Fig. 1(c). First, CC3 was prepared using previously reported methodology (the [4 + 6] cycloimination reaction of 1,3,5-triformylbenzene with (1R,2R)-1,2-diaminocyclohexane). 11 The successful synthesis of CC3 was confirmed by 1 H and 13 C NMR (Fig. S3 †). Next, the Pd(NO 3 ) 2 metal precursor was incorporated into CC3 in solution, and subsequently crystallized together to form Pd(NO 3 ) 2 @CC3, before reducing in H 2 gas to form ∼1.6 nm metal NPs (Pd NPs@CC3). Gas-phase H 2 reduction of the metal salt impregnated crystalline cages enables formation of metal NPs without hydrogenation of the imine linkages (still seen by NMR, Fig. S4 †), in contrast to liquid phase reduction with NaBH 4 reported in previous studies. [15][16][17][18][19] The Pd NPs@CC3 materials were shown to be catalytically competent in the oxidation of carbon monoxide, a widely used reaction for the benchmarking of conventional heterogeneous catalysts. 22 The Pd NPs@CC3 material could also be reproducibly recrystallized as a single (cubic) polymorph in which the (XRD silent) NPs appear to be held within the bulk of CC3 particles. We note that the absence of reflections for Pd metal in the XRD pattern is a confirmation that significant amounts of larger size particles are not present; for the small (<2 nm) particles observed by TEM, peak broadening effects mean that scattering from Pd would not be observable from the data, particularly at the low loadings present. 23 The inclusion of the Pd NPs within the crystalline POC matrix is an interesting observation, recrystallization normally being used as a purification procedurepossible reasons for this behaviour will be discussed later in the section. The recrystallization in the final step is conducted in DCM, conditions under which the Pd-free CC3 is known to be fully dissolved such that crystallisation into other polymorphs is also possible. 24 Overall this procedure yields size controlled, ultra-small NPs in a high quality POC matrix.
Pd(NO 3 ) 2 @CC3 crystals (light red) were obtained after slow evaporation of the solvent (the slow crystallization being known to produce pure CC3 crystals with few defects). 25,26 The Pd loadings, which ranged from 0.5 to 2.0 wt%, were verified by Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES, Table S2 †).
Having crystallized the Pd precursor together with CC3 (Pd(NO 3 ) 2 @CC3), we then examined the most suitable conditions to conduct the reduction under H 2 to yield ultra-small metal NPs (Pd NPs@CC3), without chemical damage to the cage structure itself. Thermogravimetric analysis (TGA) was employed to follow the changes that occurred on heating Pd(NO 3 ) 2 @CC3 samples in both an inert atmosphere and 10% hydrogen ( Fig. 2(a)). In N 2 only small changes, possibly due to residual solvent loss are seen below 300°C, at which temperature decomposition occurs. In 10% H 2 /N 2 , however, mass loss that exceeds that for the sample in N 2 was seen to have an onset as low as 150°C, with eventual complete decomposition at ∼400°C. The H 2 -reduction temperature being substantially below the temperature of decomposition in N 2 mirrors the behaviour already known for bulk Pd(NO 3 ) 2 . 27 To investigate the process, two samples were prepared by reduction for 1 h at either 150 or 200°C. The color of the samples (Fig. S5 †) shows a change has occurred with the original light orange-red color changing to yellow and then dark grey on reduction at 150 and 200°C, respectively. X-ray Absorption Spectroscopy (XAS) was used to monitor the reduction process more carefully, confirming little change at 150°C but a change to the metallic Pd 0 oxidation state after reduction at 200°C. The expanded near edge XAS region ( Fig. 2(b)) clearly shows a shift in edge energy between the 150°C reduced sample (which is similar to the unreduced sample) and the 200°C reduced sample, which is very similar to a metallic Pd foil reference sample (full spectra are shown in Fig. S6 †). Analysis of the Extended-edge X-ray Absorption Fine Structure (EXAFS) spectra obtained from Pd(NO 3 ) 2 @CC3 or the 150°C reduced sample indicated no significant Pd-Pd interactions in the first co-ordination sphere ( Fig. S7-9, Tables S4 and 5 †), while the sample reduced at 200°C (Pd NPs@CC3) showed an average first shell Pd-Pd co-ordination number of around 5 palladium atoms per Pd scatterer. This is indicative of ultra-small NPs around 1 nm in size (even allowing for different possible particle morphologies 28 ).
Scanning electron microscopy (SEM) of the Pd NPs@CC3 (reduced at 200°C) shows the octahedral shape of the Pd containing organic crystals is preserved with respect to their Pdfree counterparts, which also exhibit an octahedral microscopic mophology (Fig. S10 †). Higher resolution imaging using Scanning Transmission Electron Microscopy (STEM) is shown in Fig. 3(a-d). Small near monodisperse Pd NPs can be seen with a mean diameter and corresponding size distribution of 1.6 ± 0.3 nm. No NPs were seen for reduction at lower temperature. The size determined from STEM imaging is broadly in agreement with the analysis of the EXAFS spectra, any small difference could either result from the tendency of microscopy to overestimate the size due to the fact larger particles are easier to see, or else to the fact a small amount of residual Pd 2+ remains as isolated sites and lowers the co-ordination number derived from XAS. If the latter, only a very small fraction of isolated Pd 2+ can remain based on XANES oxidation state analysis. STEM images of a higher Pd loading sample (Fig. S11 †) indicate the size of metal NPs does not change significantly from that present in the 0.5 wt% sample.
The STEM images show the NPs are embedded within the POC crystal matrix. If NPs were present only at the surface of the POC crystals, when projected into 2D by the TEM most particles would appear decorating only the crystal edges (with apparent particle densities lower on the top and bottom faces), as seen for NPs on the surface of polystyrene spheres. 29 This is not the case for Pd NPs@CC3 (Fig. 3), which exhibit an even spatial distribution in the STEM images, providing evidence of incorporation in the material rather than merely on its surface.
Gas sorption analysis was performed on CC3 and 0.5 wt% Pd NPs@CC3 as shown in Fig. S12. † The CC3 material was found to have a Brunauer-Emmett-Teller (BET) surface area of 400 m 2 g −1 , consistent with the reported value of 409 m 2 g −1 for highly crystalline CC3. 26 0.5 wt% Pd NPs@CC3 was found to have a similar Type-I N 2 sorption isotherm with a BET surface area of 350 m 2 g −1 . Pd NPs@CC3 also had a total adsorption of 4.7 mmol g −1 N 2 at 77 K and a relative pressure (P/P 0 ) of 0.998, compared to 5.5 mmol g −1 for pure CC3. The key point is the 0.5 wt% Pd NPs@CC3 material's porosity doesn't substantially decrease compared to that of the CC3 parent material, meaning that the porosity necessary for reagent access during catalysis is preserved. Any slight decrease in accessible surface area could be indicative of pores blocked by the Pd NPs, but since the decrease is only small this should not be over-interpreted.
A number of attempts were made to identify interactions between either the Pd salt and the cage functional groups in Pd(NO 3 ) 2 @CC3 or Pd NPs and the cage in Pd NPs@CC3, but were not successful. Neither 1 H NMR (Fig. S4 †), nor infrared spectroscopy (Fig. S13 †) showed any discernable differences in the cage structure before or after reduction as a result of incor- porating Pd, possibly because of the low ratio of Pd : CC3 (1 Pd atom : 20 cage molecules). The similarity of the organic structures confirm the cages have remained intact and are not reduced (the imine linkages therefore appear likely too far away or too hindered for hydrogenation to be catalyzed by the nascent Pd NPs during reduction). There are also no major differences discernible by powder X-ray diffraction (PXRD) (Fig. 4 and S14 †) between CC3 and the samples in which Pd is present (before or after reduction). Notably, there is no evidence for Pd(NO 3 ) 2 forming an independent precipitated phase. The PXRD pattern of the pure CC3 cage material before incorporation of Pd shows that the crystals possess cubic F4 1 32 symmetry with a ≈ 25 Å, in good agreement with the reported single crystal structure of CC3α ( Fig. 4 and Fig. S14 †).
The only change seen on going from CC3 to Pd(NO 3 ) 2 @CC3 is that the PXRD pattern of Pd(NO 3 ) 2 @CC3 contains some weak additional reflectionsmost notably at 7.6°2θ, although changes at 11°and 19.5°2θ can be identified in a Pawley fit to the data. 30 The presence of a feature as low as 7.6°2θ points to a unit cell with at least one dimension greater than 11 Å, and therefore the possible presence of a second polymorph of the organic cage material. These features are seen to persist in the structure of the Pd NPs@CC3 material after gas phase reduction. The crystal packing motif adopted by CC3 is (like many compounds) known to be sensitive to surface functionalization or guest species (including solvents) present during their assembly. 24 In addition to the CC3α form, CC3 has also been shown to form a second polymorph, CC3β (space group P3), and several gas or solvent inclusion crystals, such as (CC3).(Et 2 O) 3 .CH 2 Cl 2 . 24 This low concentration component likely results from similar inclusion crystals of either the Pd(NO 3 ) 2 or the directed inclusion of other compounds (e.g. solvent) present due to difference in ionic strength. The additional features are not present if the CC3 is subject to the same DCM/acetone recrystallization in the absence of Pd(NO 3 ) 2 , Fig. S15. † This second phase was only formed in low quantities; attempts to load substantially more palladium resulted in less crystallinity and attempts at single crystal analysis showed the crystals were unstable (evidenced by a rapid color change) in intense X-rays, hence the precise identification of this second polymorph has remained elusive. As will be seen below this does not alter the use or potential applications since the material (cages and NPs) can be recrystallized easily from DCM into the usual CC3α motif, Fig. 4.
The observation that only a single phase (CC3α) is formed when Pd NPs@CC3 is dissolved in DCM and recrystallized (Fig. 4 & S14 †) is consistent with the fact that on reduction to form NPs, the Pd species do not interact with a sufficient number of cage molecules to influence the packing motif adopted (for hypothetically 100 Pd atom NPs at 0.5 wt% Pd loading there is one NP for every ∼1900 cage units). The ability to recrystallize the NPs dispersed within the CC3 material is crucial for applications. Since STEM images of Pd NPs@CC3 were obtained after dissolving and effectively recrystallizing the sample by evaporation on the TEM grid, this confirms the NPs remain dispersed throughout the material upon solution processing. It is interesting that recrystallization is typically employed as a purification procedure but here the NPs clearly become embedded in the CC3 material. This can be rationalised by the very high surface free energies of the Pd NPs being stabilised by (at the very least) van der Waals interactions with the CC3 (in the absence of any other stabilising agent), and thus remaining bound to CC3 molecules. 31 Speculatively, the CC3 molecules already attached to the Pd NPs during crystallisation can then act as a 'Trojan horse' ligand to mimic the expected supramolecular interactions during crystal epitaxy, as found for the inclusion of NPs with appropriate ligands in liquid crystals. 32 A number of reports have argued or postulated the NPs form within the cages, 15,18-20 their relative size range (about 1.3-1.9 nm) makes this option unrealistic given the small size of the cavity (diameter ∼0.8 nm); this is also true for the pore cavity between two cage molecules with window-towindow packing (again ∼0.8 nm). This is illustrated in Fig. S16. † Instead, given the similarity of the particle size with the size of the cage (max diameter = 2 nm), it more likely becomes a guest/defect occupying the space of one cage molecule within the porous crystalline structure (given that by TEM they are not strongly clustered or only at the surface of the material, Fig. 3). The role of the cages is therefore in the controlled assembly of the ultra-small NPs and their subsequent external stabilization, rather than as conventional templates.
CO oxidation was used as a benchmark reaction, to demonstrate the use of Pd NPs@CC3 materials as catalysts. Fig. 5 shows light-off data (conversion as a function of increasing temperature) for the CO/O 2 reaction conducted in a fixed packed-bed reactor over the temperature range 100-200°C. In order to measure the rate of reaction under gradientless conditions the reactor was operated under conditions of differential CO conversion (<20%). The CO oxidation reaction cata- lyzed by the 0.5 wt% Pd NPs@CC3 catalyst proceeds at ∼20 mmol CO 2 mol Pd −1 s −1 at 200°C (∼6% CO conversion).
Additionally, Fig. 5 shows points for a typical reference Pd/Al 2 O 3 catalyst and two literature catalysts. 33,34 It can be seen that the Pd NPs@CC3 exhibit broadly comparable catalytic activity to other Pd catalysts (normalised to Pd amount). § However, the other catalysts shown are not as amenable to solution phase processing techniques as Pd NPs@CC3, but once assembled are fixed in form.
The stability of the catalyst during reaction was also demonstrated by repeat cycling of the material in temperature (5 times, Fig. S17 †), which shows very little change and is consistent with both TGA in air (Fig. S18 †) and variable temperature XRD (Fig. S19 †), both of which show stability above these temperatures. In addition to using low conversions that avoid a significant change in concentration along the reactor bed, a series of experiments at varying flow rates were conducted confirming the absence of mass transfer effects along the reactor, which if present could lead to erroneous catalytic performance data (Fig. S20 †).

Conclusions
We have developed a synthetic approach that enables controlled assembly of metal NPs and their incorporation with shape persistent POC materials. The Pd NP containing material remains soluble and recrystallizable after NP incorporation. The cage molecules act as a stabilizer and effective catalyst support for the ultra-small Pd NPs produced (∼1.6 nm as determined by STEM). A range of techniques (including PXRD, TGA, XAS) was used to give detailed insight into the synthesis of ultra-small Pd NPs@CC3 via gas phase reduction in hydrogen. The resulting Pd-embedded porous organic material has been shown to be an effective CO oxidation catalyst. In contrast to other recent reports of organic cage templated NPs, liquid phase reduction of the organic cage was avoided by use of the gas-phase reductant H 2 for the first time, preventing loss of cage bonding that inhibits crystal formation/assembly. This proof-of-concept demonstration paves the way for taking advantage of the solution processability and facile directed assembly of porous organic cage materials to assemble and process ultra-small nanoparticle catalysts.
Synthesis of CC3. The synthetic procedure has been described in the previous study. 11 CC3 was synthesized as follows: benzene-1,3,5-tricarboxaldehyde (100 mg, 0.62 mmol) was dissolved in DCM (8 mL), and added to a solution of (1R,2R)-(−)-1,2-diaminocyclohexane (106 mg, 0.93 mmol) in DCM (8 mL) in an oven-dried round bottomed flask. The flask was then sealed and left to stand for 7 days at room temperature, with no stirring. After 7 days the solvent was removed in vacuo, the crude product was filtered and washed with methanol to give CC3 as a white powder (122 mg, 0.11 mmol, 71%). Synthesis of Pd(NO 3 ) 2 @CC3. CC3 (10 mg, 8.95 × 10 −3 mmol) was dissolved in DCM (2 mL) in a glass sample vial. Next, a stock solution of Pd(NO 3 ) 2 in acetone (3 mg mL −1 ) was prepared. The corresponding Pd(NO 3 ) 2 solution was added into CC3 solution to form Pd(NO 3 ) 2 @CC3 crystals. Four Pd(NO 3 ) 2 @CC3 samples were prepared with a Pd(NO 3 ) 2 loading at 1 wt% (Pd(NO 3 ) 2 ), 2 wt%, 4 wt% and 8 wt%, respectively (see amounts given in ESI, Table S1 †). The mixed solution was stirred for 1 h. The products were obtained by slowly evaporating the solvents. The lid was placed on the vial, but not Light-off data: red: CO oxidation rate as a function of temperature catalyzed by Pd NPs@CC3 (<10% conversion; 3% CO, 15% O 2 , balance He, total flow 60 mL min −1 (NPT); ambient pressure; 71.5 mg of 0.5 wt% Pd catalyst). Black: reference/literature data (all measured at <20% conversion) for typical Pd catalysts: Ω, Pd triptycene based microporous polymer; 33 Σ, Pd/SiO 2 (literature); 34 Φ, Pd/Al 2 O 3 commercial reference sample (Alfa Aesar) (as a sample of similar particle size could not be sourced, a sample with around 3 m 2 g −1 metal surface area was employed-about double the geometric surface area assuming spherical particles in the CC3 derived sample-and so the data point for this sample only was normalized for known relative surface area to aid in comparison of this sample with our system) tested in same apparatus. § It should be noted that Pd with reducible oxides, e.g. Ceria, 35 and Au based systems 36 have much lower light off temperatures, but this is due to a different mechanism at play. Pure Pd catalysts are known to be strongly inhibited by absorption of CO, which blocks oxygen access, below around 150°C. fully closed to allow solvent to escape. The solvent was then allowed to evaporate off under ambient conditions. Once all the solvent had disappeared, the sample was further de-solvated by placing the sample in an oven at 70°C under vacuum overnight. Accurate Pd loading was verified by ICP-OES was 0.5 wt%, 1 wt%, 2 wt% and 4 wt% (ESI , Table S2 †). It should be noted that ICP-OES is a preferred method for low loading samples, as the residual mass by TGA is a very small fractional amount of the material being measured is less reliable than ICP over a ∼3 h experiment or contain residual carbonaceous ash.
Synthesis of Pd NPs@CC3. Pd(NO 3 ) 2 @CC3 crystals were reduced using a mixture containing 10% H 2 in N 2 , with a flow rate of 30 mL min −1 in the furnace of a PerkinElmer Pyris 1 TGA. Samples were heated to 200°C with a heating rate of 5°C min −1 and then held at that temperature for 1 h to form Pd NPs@CC3.

Characterization
Nuclear magnetic resonance (NMR) spectroscopy. Solution phase 1 H NMR spectra were recorded using a Brucker Avance 400 spectrometer at 400 MHz, at a temperature of 296 K, and using CDCl 3 as the solvent. 13 C NMR spectra were recorded at 400 MHz. Chemical shifts are reported in ppm relative to TMS, and coupling constants are in Hz.
Fourier transform infrared spectroscopy (FTIR). IR spectra were collected on a PerkinElmer Frontier IR, with a UATR attachment. Samples were prepared by placing dry powder onto the ATR attachment, and screwing the vice down onto the sample until the force gauge indicator reading was approximately 120 N. Samples were analysed for 4 scans with a resolution of 4 cm −1 .
Thermogravimetric analysis (TGA). TGA experiments were performed using a PerkinElmer Pyris 1 TGA. For reduction experiments, a mixture containing 10% H 2 in N 2 or pure N 2 was used, with a flow rate of 30 mL min −1 . Samples were heated to 400°C with a heating rate of 5°C min −1 under either N 2 or 10% H 2 in N 2 with a flow rate of 30 mL min −1 . For the stability study, samples were heated to 800°C with a heating rate of 5°C min −1 under an air flow.
Powder X-ray diffraction (PXRD). Room temperature PXRD patterns were acquired using a Bruker AXS D8 Advance equipped with a Lynx-Eye PSD Detector and Cu K α1 /K α2 radiation. The diffractometer was used in Bragg-Brentano mode with a Ni filter, variable slits to give a 6 mm beam on the sample, and sample changer. Samples were prepared by applying a small amount of Vaseline to a Si zero-background sample holder, and then lightly applying the as-prepared powder. Each pattern was recorded over a 2θ range of 5-45°at room temperature for 60 min using a step size of 0.02°. VT-PXRD were collected in capillary mode on a Bruker D8 diffractometer equipped with a Lynx-Eye detector and an Oxford Cryosystems Cryostream Plus device using Mo Kα radiation (λ = 0.7093 Å). Ground sample was loaded into a 0.7 mm borosilicate glass capillary, placed on a goniometer head and rotated during the data collection. The sample was cooled and warmed between 500 K and 110 K at a rate of 20 K h −1 .
A series of 15 min PXRD patterns were recorded over the 2θ range 5-30°using a step size of 0.02°.
Transmission electron microscopy (TEM). TEM images were acquired using a JEOL 2100F FEG TEM. Samples were prepared by dissolving ∼5 mg 0.5 wt% Pd NP@CC3 in 1 mL DCM. A small amount of MeOH (∼0.1 μL) was added in order to enhance the solubility of Pd NP@CC3 catalyst. Several drops of the solution were then placed onto a holey carbon grid. Imaging was conducted at a working voltage of 200 kV. STEM samples were prepared in a similar fashion.
Scanning transmission electron microscopy (STEM). Scanning transmission electron microscopy was performed using a JEOL 2100F FEG TEM operated at 200 kV in scanning TEM mode, STEM images were obtained using a JEOL dark field detector and Gatan bright field detector (latter at the entrance aperture for the Gatan Tridiem spectrometer). STEM samples were prepared by dissolving ∼5 mg 0.5 wt% Pd NP@CC3 in 1 mL DCM. A small amount of MeOH (∼0.1 μL) was added in order to enhance the solubility of Pd NP@CC3 catalyst. Several drops of the solution were then placed onto a holey carbon grid. STEM for 1 wt% Pd NP@CC3 was prepared by the same procedure.
Scanning electron microscopy (SEM). SEM images were acquired using a focused ion-beam microscopy (FEI) Helios Nanolab 600. Samples were prepared by depositing dry powder on 15 mm aluminium stubs using an adhesive high purity carbon tab. The samples were then coated with a 12 nm layer of gold using a gold coater. Imaging was conducted at a working distance of 5.0 mm, and a working voltage of 3 kV.
Mass spectrometry. ASAP mass spectrometry was carried out on a LCT Premier XE mass spectrometer. Samples were prepared by dissolving approximately 1 mg of solid in 1 mL of DCM.
Inductively coupled plasma optical emission spectroscopy (ICP-OES). ICP-OES was performed on a Jobin Yvon Horiba -Ultima 2 Radial, with a vertically orientated torch and a sequential monochromator. A Pd ICP-OES standard was used, based on Pd emission lines at 324.270 nm and 340.458 nm. Samples were prepared by heating 5 mg of solid in the minimum volume of dilute nitric acid at 80°C for 30 min. The solution was then transferred to a 5 mL volumetric flask and was made up to 5 mL using deionized water.
Gas sorption analysis. Surface areas were measured by nitrogen adsorption and desorption at 77 K using a Micromeritics ASAP 2020 volumetric adsorption analyzer. Samples were degassed at offline at 100°C for 15 h under dynamic vacuum before analysis, followed by degassing on the analysis port, also at 100°C.
X-ray absorption spectroscopy (XAS). XAS experiments were performed at beamline B18, Diamond Light Source, Oxfordshire, UK. A double-crystal Si(311) monochromator was used to scan X-ray energy from −200 to 1000 eV relative to Pd K edge (24 350 eV). Each sample (sufficient that the absorption coefficient jump at the edge was satisfactory for transmission measurements) was prepared by pressing finely ground powders into self-supporting disks using an IR die press and then supporting on Kapton tape. These were employed for transmission XAS measurements with pure Pd metal foil measured concurrently between a second and third ionisation chamber, enabling X-ray energy calibration and data alignment. Further details of the analysis of data processing methods are given in the ESI. †

Conflicts of interest
There are no conflicts to declare.