Hydrogenation on Palladium Nanoparticles Supported by Graphene Nanoplatelets

Pd nanoparticles (1 wt %; mean size ∼4 nm) were supported on ∼2 μm sized, but few nanometers thick, graphene nanoplatelets (GNPs) and compared to 1 wt % Pd on activated carbon or γ-alumina. Catalyst morphology, specific surface area, and Pd particle size were characterized by SEM, BET, and TEM, respectively. H2-TPD indicated that GNPs intercalated hydrogen, which may provide additional H2 supply to the Pd nanoparticles during C2H4 hydrogenation. Whereas the two types of Pd/GNPs (NaOH vs calcinated) catalysts were less active than Pd/C and Pd/Al2O3 below 40 °C, at 55 °C they were about 3–4 times more active. As for example Pd/GNPs (NaOH) and Pd/Al2O3 exhibited not too different mean Pd particle size (3.7 vs 2.5 nm, respectively), the higher activity is attributed to the additional hydrogen supply likely by the metal/support interface, as suggested by the varying C2H4 and H2 orders on the different supports. Operando XANES measurements during C2H4 hydrogenation revealed the presence of Pd hydride. The Pd hydride was more stable for Pd/GNPs (NaOH) than for Pd/C, once more pointing to a better hydrogen supply by graphene nanoplatelets.


INTRODUCTION
Supported Pd catalysts are particularly important for industrial hydrogenation reactions, including fine chemicals synthesis. 1−3 Palladium is considered the most selective among the platinum metals, as hydrogenation is much faster than dehydrogenation to undesired carbonaceous species. High dispersion (mean Pd particle sizes below 3 nm) is a prime asset of these catalysts. Clearly, the nature of the support has a strong influence on performance, with carbon-based materials, alumina, and silica being most frequently applied. Resulting from their practical importance, technological as well as model catalysts of Pd/ Al 2 O 3 , Pd/Fe 3 O 4 , or Pd/C (activated carbon, nanotubes, planar graphene layers, highly oriented pyrolytic graphite (HOPG), etc.) have been repeatedly studied for the (selective) hydrogenation of for example ethylene, 4−9 1,3-butadiene, 10−13 1-butene, 14 acetylene, 5,15 1-propyne, 16,17 unsaturated aldehydes, 18 and others. 4,19−21 For reviews and more detailed accounts we refer to refs 21−33. When carbon was used as catalyst support, many different types and morphologies have been examined, 1 including activated carbon, carbon black, graphite, and various forms of graphene (GN), such as (single and multiwalled) nanotubes, (exfoliated) sheets, nanoplatelets (GNP), etc. Graphenesupported Pt nanoparticles have received particular attention because of their use in polymer electrolyte membrane (PEM) fuel cells (hydrogen or methanol oxidation). They were reported to exhibit higher electrocatalytic activity, stability, and poisoning tolerance than Pt/activated carbon, which also enabled to reduce the Pt loading. 34−36 For graphene-supported Pd there are fewer studies, but similarly, electrocatalytic oxidation and hydrogenation activity were higher than for traditional carbon supports. 37−39 Carbon and Pd/GN(P) received further attention for hydrogen storage, again relevant for fuel cell technology. 40−42 In this contribution, we have exploited graphene nanoplatelets (GNPs) as support for Pd nanoparticles and contrasted them to Pd/activated carbon and Pd/Al 2 O 3 .
Ethylene hydrogenation, following Langmuir−Hinshelwood kinetics via stepwise hydrogenation (a mechanism proposed by Horiuti and Polanyi in 1934), is a prototype test reaction providing valuable information about catalytic performance, reaction mechanisms, and deactivation processes. 5 For reaction temperatures above 40°C, higher activities (turnover frequencies TOFs) were observed for Pd nanoparticles supported by GNPs, which may be related to hydrogen storage/intercalation by the graphene nanoplatelets.
Operando studies under reaction conditions were performed by X-ray absorption near-edge structure (XANES) spectroscopy, revealing Pd hydride. The higher activity was attributed to the additional hydrogen supply, as corroborated by the varying C 2 H 4 and H 2 orders on the different supports. Clearly, more detailed (preferentially operando) studies are required, but the current results already indicate the high potential of Pd supported on graphene nanoplatelets as a hydrogenation catalyst.

Materials and Catalyst Preparation.
Graphene nanoplatelets (GNPs) and palladium acetate as metal precursor 45 (both from Sigma-Aldrich) were used for two types of synthesis: (i) wet impregnation with formaldehyde as reducing agent 46 in basic medium (NaOH; pH 12) 47 and (ii) wet impregnation followed by calcination in air.
For wet impregnation (i), specific amounts of GNPs and precursor (Pd 2+ ac) were suspended in 50 mL of water (with a small amount of ethanol for better dispersion) and stirred at room temperature for 30 min, after which formaldehyde was added as reducing agent. Sodium hydroxide (pH 12) was added, as it was reported to improve particle nucleation. Finally, the precatalyst was obtained by filtration and drying at 100°C for 3 h (Pd/GNPs (NaOH)).
For comparison and benchmarking, commercial 1% Pd/ activated carbon and 1% Pd/γ-alumina (both from Sigma-Aldrich) were examined. 17 The activation treatment of all catalysts is described below.
2.2. Catalyst Characterization. Characterization was carried by various techniques: SEM/EDX and TEM/HAADF-STEM: Scanning electron microscopy (SEM) was used to determine the morphology of the graphene nanoplatelets. Using energy dispersive X-ray fluorescence (EDX), we confirmed the absence of impurities. To determine the metal nanoparticle size distribution, transmission and high angle annular dark field scanning transmission electron microscopy (TEM/HAADF-STEM) were performed (FEI-TECNAI F20; "ImageJ" software) for pretreated catalysts. N 2 physisorption (BET): For determining the specific surface area (SSA) of the samples, N 2 physisorption was performed at −196°C (ASAP 2020 Micromeretics). 37 Using ca. 100 mg sample, we performed degassing at 350°C for 4 h (temperature ramp of 10°C/min) to remove adsorbed water. Data analysis was performed by using the Brunauer−Emmett− Teller (BET) method.
CO chemisorption: As alternative measurement of the average Pd nanoparticle diameter (apart from TEM), quantitative chemisorption of CO was performed for pretreated catalysts. The particle diameter was calculated by assuming hemispherical particles, with a stoichiometric factor (per Pd surface atom) of 1 for CO. A 0.5 g sample was placed into a quartz tube for measurements at 35°C (10−860 mbar) and analyzed as described by Canton et al. 48 H 2 -TPD: Temperature-programmed desorption (TPD) of H 2 was performed by using a 0.1 g sample of support or (pretreated) catalysts. Hydrogen (100 mbar) was dosed at room temperature for 30 min, followed by evacuation for 30 min. For TPD, a heating rate of 10°C/min was applied.
ζ-potential: To investigate the effect of the pH value, the surface potential of the sample was determined (ζpotential 49,50 ). Palladium ions (Pd 2+ ) that were used as precursor likely interact stronger with a negatively charged support surface, created for example by adding sodium hydroxide to increase the pH to 12. 51 X-ray absorption near-edge structure (XANES): Operando X-ray absorption spectroscopy (focusing on the near-edge structure) was performed in transmission mode at the ALBA-Synchrotron beamline (CLAESS, Barcelona). 52,53 XANES measurements of pellets (sample diluted 1:10 with BN) were compared with models of hydrogen absorption reported in refs 52 and 53.
2.3. Catalytic Reaction. For kinetic measurements of ethylene hydrogenation, a fully automated plug-flow reactor was used ("Microeffi" from PID Comp.), coupled to a micro-GC (Inficon) with two columns (PlotQ and Molsieve). Because of the high activity of supported Pd, about 5 mg of catalyst was diluted with 995 mg of SiO 2 . For kinetic tests, 10 mg of the "diluted" catalyst was used. For catalyst activation (pretreatment) before the reaction, the carbon-supported catalysts (Pd/GNPs (NaOH) and Pd/GNPs (calc), Pd/C) were reduced in 8% H 2 in He (total flow: 100 mL/min) at 200°C (2 h), whereas Pd/γ-alumina was activated in 20% O 2 in He (total flow: 100 mL/min) at 500°C (2 h), followed by reduction in 8% H 2 in He (total flow: 100 mL/min) at 200°C (2 h).
For the ethylene hydrogenation, a mixture of 2 vol % C 2 H 4 and 20 vol % H 2 in He at a total flow of 100 mL/min was used. GC measurements were taken from 25 to 55°C, in steps of 5°C . The reaction rate was calculated via the following equation: reaction rate [mol/(s g)] = (ethylene conversion × ethylene flow [mol/s])/(amount of catalyst (without SiO 2 ) [g]). Additionally, rates were normalized per gram of Pd or per Pd surface atom (turnover frequency, TOF). Reaction orders were determined at 55°C for C 2 H 4 (from 1 to 8 vol % C 2 H 4 in 0.5 vol % steps at constant 20 vol % H 2 ) and H 2 (from 3 to 18 vol % in 3 vol % steps at constant 2 vol % C 2 H 4 ).

RESULTS AND DISCUSSION
3.1. Graphene Nanoplatelets (GNPs). SEM and TEM images of the graphene nanoplatelets (GNPs) are shown in Figures 1a and 1b, respectively. The GNPs were typically less than 2 μm in diameter and a few nanaometers in thickness but often seem thicker due to stacking/aggregation. EDX detected only C, Na, and O but no (metal) impurities. Because of nonoxidizing manufacturing, the GNPs have a pristine graphitic surface with sp 2 carbon. N 2 physisorption ( Figure  1c) indicated a SSA of 304 and 184 m 2 /g without and with NaOH treatment, respectively (activated carbon is included for comparison). The effect of NaOH treatment will be discussed below. H 2 -TPD from GNPs is contrasted to activated carbon in Figure 1d. The graphene nanoplatelets "stored" significant amounts of hydrogen, likely by intercalation between the nanosheets (macroporosity), which then desorbed above 50°C . Activated carbon (and Al 2 O 3 ) did not show this effect. The Journal of Physical Chemistry C pubs.acs.org/JPCC Article 3.2. Pd Nanoparticles Supported by Graphene Nanoplatelets (Pd/GNPs). As described in section 2.2, the GNPs were used as support for Pd nanoparticles, employing different synthesis protocols. Figure 2 shows HAADF-STEM and SEM images, as well as the corresponding Pd particle size distributions, for 1% Pd/GNPs (NaOH), 1% Pd/GNPs (calc.), and commercial 1% Pd/C and 1% Pd/Al 2 O 3 . All structural parameters, including results from BET and chemisorption, are summarized in Table 1. H 2 -TPD from (pretreated) Pd/GNPs in Figure 1d showed desorption over a broad temperature range, apparently with contributions from both GNPs and Pd nanoparticles (cf. Pd/C and Pd/Al 2 O 3 ).
As mentioned, route i was impregnation with reduction by formaldehyde in basic (NaOH) medium. BET indicated that NaOH reduced the SSA (from 304 to 184 m 2 /g) but increased the ζ-potential of GNPs (−3.3 mV at pH = 7 to −17.4 mV at pH = 12). Consequently, Pd 2+ ions of the precursor interacted stronger with the negatively charged GNPs surface, creating higher Pd dispersion (∼38%) with a mean Pd particle size of 3.7 nm (with TEM and chemisorption agreeing well; Table 1).
Furthermore, NaOH may also create surface defects acting as nucleation centers. 54 Route ii included impregnation, calcination in air at 300°C, and H 2 reduction at 200°C. This maintained a higher SSA (278 m 2 /g), but the Pd nanoparticles were larger (mean size 5.7 nm; ∼26% dispersion), again with TEM in agreement with chemisorption (Table 1). Apparently, calcination at 300°C already induces Pd particle sintering on GNPs.
Commercial 1% Pd/C had a mean Pd particle size of 2.3 nm (dispersion ∼56%), similar to commercial Pd/Al 2 O 3 (mean Pd particle size of 2.5 nm, dispersion ∼53%). For the smaller particle sizes, TEM and chemisorption somewhat deviated, which has been repeatedly reported. 55,56 TEM may "overlook" the smallest sizes, and some of the assumptions of chemisorption analysis (specific adsorption on the metal, fixed CO:Pd adsorption ratio and geometry) may no longer be valid.
3.3. Kinetics of C 2 H 4 Hydrogenation on Pd/GNPs. The Pd nanoparticles (1 wt %) on different supports were tested for ethylene hydrogenation. Minor deactivation (approximately 1−5% of total conversion, depending on reaction temperature)    Figure 3 and Table 2. Reaction rates are normalized to the catalyst weight (ignoring the dilution by inert SiO 2 ) per gram of Pd and per Pd surface atom. Conversions ranged from 0.5 to 51%. Activated carbon exhibited no activity at all, whereas pure graphene nanoplatelets had low activity at 55°C (alkene hydrogenation on GNPs at 110°C has been reported in Primo et al. 57 ). As mentioned, EDX analysis of GNPs indicated no metal impurities, but it cannot be excluded that metal traces were responsible for the small activity.
Below 40°C, the Pd nanoparticles supported on GNPs (for both preparation routes) were less active than Pd/C or Pd/ Al 2 O 3 . Above 40°C, the activities reversed, with Pd/GNPs being ∼3−4 times more active than Pd/C or Pd/Al 2 O 3 . This behavior led to apparently much higher activation energies on Pd/GNPs (Table 2). Comparing the E a values for the GNPsupported Pd against that of Pd/Al 2 O 3 and Pd/C indicates differences in the surface chemistry when GNP are used to support Pd crystallites. This outcome suggests different metal/ The Journal of Physical Chemistry C pubs.acs.org/JPCC Article support interactions are active with the Pd/GNP catalyst. The origin of the higher activity of Pd/GNPs may be related to the larger Pd particle size (∼4 nm vs ∼2 nm; with smaller particles being more easily deactivated) and/or the nature of the support (GNPs vs C or Al 2 O 3 ). Also, sodium seems to have a beneficial effect (cf. Pd/GNPs(NaOH) vs Pd/GNPs(calc)), which may be due to higher dispersion. Indeed, with respect to turnover frequencies (Figure 3c), the two GNPs-supported catalysts reversed, and 1% Pd/GNPs (calc) was the most active.
To further investigate differences between the catalysts, the C 2 H 4 and H 2 reaction orders were determined at 55°C (Table  2). For Pd/C and Pd/Al 2 O 3 , the C 2 H 4 orders are (slightly) negative (in agreement with model studies; −0.3 in ref 58), which indicates that under these conditions ethylene already blocks the Pd surface. In contrast, for Pd/GNPs the C 2 H 4 order is clearly positive, which indicates that the Pd surface was not yet saturated with C 2 H 4 . Concerning H 2 order, it is ∼1 for Pd/C and Pd/Al 2 O 3 but ∼0.75 for Pd/GNPs. This also confirms that the Pd surface of Pd/GNPs is more readily available for the reactants. For Pd single crystal and Pd/Al 2 O 3 , C 2 H 4 and H 2 reaction orders of approximately 0 and 1 were reported in the literature, 59,60 respectively. 61 Several effects may account for these observations. Most importantly, the GR nanoplatelets intercalated hydrogen and released it around 40°C (note that this is the temperature of the activity reversal), which creates additional hydrogen supply to a Pd particle, in addition to the gas phase route (a similar effect was reported by Rupprechter and Somorjai 62 ). Although the GNPs supported Pd particles are a bit larger, one cannot exclude electronic metal−support interactions (especially at the phase boundary). Indeed, graphene has been shown to be an excellent support for CuZn nanoparticles for methanol synthesis. 63 Last but not least, the smaller Pd nanoparticles on activated carbon or alumina may be more prone to partial deactivation by carbonaceous species (CH x ), as reported by Lennon and co-workers. 17 3.4. Operando Studies of C 2 H 4 Hydrogenation on Pd/ GNPs. Figure 4a shows in situ XANES spectra of (the smaller) Pd nanoparticles on GNPs (NaOH), before and after H 2 pretreatment (without intermittent exposure to the atmosphere). Pd foil and PdO were measured in a vacuum as reference. Before H 2 pretreatment, the Pd particles were partially oxidized but metallic after reduction (with cooldown in He to avoid Pd hydride formation). 6,62 C 2 H 4 hydrogenation was then performed under the mentioned reaction conditions at 30, 50, and 100°C, with the gas composition monitored by mass spectroscopy (MS), shown in Figure 4b. Operando XANES spectra of Pd/GNPs (NaOH) and Pd/C are displayed in Figures 4c and 4d, respectively (Pd foil is again added just as reference; i.e., it is not measured under reaction conditions). Apparently, both Pd/GNPs and Pd/C showed the presence of Pd hydride in the entire temperature range (reaction flow conditions equivalent to 200 mbar of H 2 ). Thermal desorption spectroscopy (TDS) of model catalysts had suggested a strongly enhanced hydrogenation probability in the presence of Pd hydrides, for both Pd nanoparticles and Pd single crystals (the conversion of C 2 H 4 being nearly 100% in the presence of palladium hydrides). 4,6,43 For Pd/C, the hydride phase seems more pronounced at lower temperature (30°C) than at higher temperature (100°C), pointing to hydride decomposition, which may be facilitated for smaller particles.
A direct comparison of Pd/GNPs and Pd/C at 50 and 100°C ( Figure 4e,f) reveals almost identical spectra at 50°C, but at 100°C the hydride is still present for Pd/GNPs but seems (more) decomposed for Pd/C. This indicates a Pd particle size and/or support effect (with larger particles and GNPs supplying more hydrogen to the Pd particles).
Parker, Albers, and co-workers have used the technique of inelastic neutron scattering (INS) to investigate the matter of hydride formation in supported Pd catalysts. Application of carbon-supported Pd catalysts for the hydrogenation of nitroarenes showed that carbon exhibiting enhanced sp 2 character templated the binding of the Pd particles at the edges of carbon particles. Catalytic activity was attributed to (i) hydrogen storage capability and (ii) the availability of that reservoir of hydrogen. 64 Follow-on work established the influence of the support material, which critically can affect the rate of release of stored hydrogen from the Pd particles. 65 Thus, in addition to verifying a role for Pd hydride in heterogeneously catalyzed hydrogenation reactions, the INS studies also reveal how important the nature and form of the metal/support interface are in affecting hydrogen supply for subsequent reaction at the Pd surface.
Given that Figure 4 shows the extent of Pd hydride in the Pd/GNPs and Pd/C samples to be broadly comparable, this indicates that it is not simply the presence or absence of a hydride phase that is affecting the favorable hydrogenation performance of Pd/GNPs compared to Pd/C as evidenced in Figure 3. However, with reference to the aforementioned INS studies, this apparent discrepancy hints at a discrete role for hydrogen at the metal/support interface and the relative availability of that hydrogen for reaction. Chesters and coworkers used 1 H NMR spectroscopy to examine hydrogen chemisorption over a silica-supported Pt catalyst and report the presence of a resonance that is assigned to hydrogen present at the interface between the Pt particles and the support material. 66 Further variable temperature NMR studies concentrating on the EuroPt-1 reference catalyst endorse the generality of the models proposed and provide insight into the dynamics of hydrogen chemisorbed over supported metal particles. 67 Linking these strands together, the origin of the enhanced hydrogenation performance observed for the Pd/GNP ( Figure  3) is thought to be that the graphene nanoplatelets foster binding to the Pd crystallites that can additionally accommodate hydrogen atom transfer at the metal/support interface. Then, warming the catalyst above 40°C increases the mobility of this interfacial hydrogen such that it becomes available for reaction. Thus, although possessing a comparable Pd particle size to that of Pd/C (section 3.2), it is the nature of the graphene nanoplatelets of the Pd/GNPs that facilitates hydrogen availability at the metal/support interface. The ethene hydrogenation reaction profiles presented in Figure 3 indicate that the activated carbon used for the Pd/C catalyst cannot similarly buffer surface the hydrogen supply in this way.

CONCLUSIONS
Pd nanoparticles, with mean sizes around 4 and 2 nm, were supported on either graphene nanoplatelets (GNPs), activated carbon, or γ-alumina. Apart from apparent differences in specific surface area (SSA), H 2 -TPD indicated that GNPs intercalated hydrogen which may provide an additional supply of hydrogen to the Pd nanoparticles (apart from the direct adsorption from the gas phase). For ethylene hydrogenation, The Journal of Physical Chemistry C pubs.acs.org/JPCC Article Pd/GNPs (NaOH and calc) were less active than Pd/C and Pd/Al 2 O 3 below 40°C, but at 55°C they were about 3−4 times more active. As Pd/GNPs (NaOH) and Pd/Al 2 O 3 exhibited not too different mean Pd particle size (3.7 vs 2.5 nm, respectively), this effect seems related to the additional hydrogen supply (likely at the metal/support interface), as corroborated by the measured C 2 H 4 and H 2 orders of the reaction. Operando XANES measurements during ethylene hydrogenation revealed the presence of Pd hydride. However, the Pd hydride was more stable for Pd/GNPs (NaOH) than for Pd/C, once more pointing to a better hydrogen supply by graphene nanoplatelets. Pd/GNPs will be employed for further reaction studies in the near future. Moreover, given the emphasis of temperature-dependent interfacial effects revealed here, that work will include theoretical studies (density functional theory) alongside experimental testing.