Hydrogen spillover and its relation to catalysis: observations on structurally dened single-atom sites

Hydrogen spillover, involving the transfer of H atoms from metal sites onto the catalyst support, is ubiquitous in chemical processes such as catalytic hydrogenation and hydrogen storage and is therefore of tremendous fundamental and technological interest. However, atomic level information concerning the kinetics of this process, the structural evolution of the catalysts during hydrogen spillover, as well as the nature of participation of the spilled over H in catalysis, remain vastly lacking. Here, we provide insights to those questions with the development of a solubilised polyoxometalate-supported single-atom catalyst which allows for the use of characterisation techniques generally inaccessible to study heterogeneous catalysis. Hydrogenation kinetics together with poisoning studies further reveal that hydrogen spillover can be either detrimental or beneficial for catalysis – the direction and magnitude of which depends predominantly on the nature of the reducible bond in the substrate. Similar trends were observed on one of the most prototypical hydrogen spillover catalysts, Pt/WO 3 , supporting the generalisability of the observations.


Introduction
Previously studied catalysts exhibiting hydrogen spillover only offer limited insights due to their lack of well-defined structures, insufficient tools to precisely characterise spillover stoichiometries and kinetics, as well as the lack of information on the role of spilled over hydrogen during catalysis. 1 For example, conventional nanoparticle catalysts on metal oxides such as the prototypical Pt/WO3 comprise ill-defined metallic hydrogen splitting sites as well as complex metal support interfaces. Furthermore, ambiguities exist about whether spilled over H should be rather regarded as hydrogen atoms or protons. 2 Although surface science studies offer deeper insights, correlating hydrogen spillover with catalytic activity remains difficult. [3][4][5] In most cases, the support covered in spilled over hydrogen is viewed as a H reservoir with the metallic site as the main driver for catalysis, though in certain cases the direct participation of the H-covered support in hydrogenation reactions has been hypothesised. 1,6 Besides constraints by the catalyst materials, techniques to study the hydrogen spillover phenomenon are additionally restricted to solid state NMR, 7 temperature-programmed reactivity studies, 8,9 neutron scattering, 10 resonant photoemission 11 and X-ray absorption spectroscopy 12 , among others, all with known limitations in the investigation of heterogeneous catalysts. DFT calculations can offer additional details but the relevance of those results is compromised by the lack of definedness of the spillover catalysts.
Due to these limitations, several major questions remain unanswered about hydrogen spillover: (i) the kinetics of the process, i.e. how the rate of H spillover is affected by an intermediate spillover state, (ii) the composition and properties of the catalyst after H spillover, and (iii) how the spilled over H contributes to heterogeneous hydrogenation catalysis. Herein, we provide insights to those questions by characterising the hydrogen spillover behaviour of several solubilised lacunary polyoxometalate (POM)-based single-atom catalysts (SACs). Previously, we reported similar catalysts to identify active sites during hydrogenation, as well as the CO and benzyl alcohol oxidation reactions. [13][14][15][16] Herein, we find a unique interplay between positively charged Pd and a reducible Mo-based support for hydrogen spillover. Due to the solubility of the SAC in polar solvents, techniques like electrospray ionisation mass spectrometry (ESI-MS), liquid phase 1 H NMR and UV-Vis spectroscopy become viable tools for the characterisation of the process. Based on reaction kinetics, catalyst poisoning, and ESI-MS, we identified subtleties in the contribution of spilled over H to the hydrogenation of different substrates.
These molecularly defined catalyst model structures with experimentally proven reaction intermediates provide an ideal system for DFT calculations to understand both thermodynamic and kinetic behaviour of hydrogen spillover. Combined with analogous experiments with Pt/WO3 catalysts, it appears that our findings are not only true for SACs but may be generally applicable.

Hydrogen spillover on PMo11O39Pd1
SACs based on lacunary POMs offer advantages such as easy synthesis, stable structures in 1:1 metal support ratios, as well as their high solubility in polar solvents. 16 PMo11O39 was synthesised by the lithium carbonate-mediated removal of one MoO subunit from phosphomolybdic acid ( Supplementary   Fig. 1) while PW11O39 was synthesised following the procedure described in our previous study. 16 Mixing equimolar ratios of the lacunary POM and metal salts with weakly coordinating ligands yielded in situ the active site, PMo11O39Pd1, which can be easily and unambiguously characterised by  Supplementary Fig. 2). We did not observe any species with higher H coverage suggesting that this is the highest hydrogen spillover state of PMo11O39Pd1 under the reaction conditions. The lack of colour changes and unaltered ESI-MS spectra of PMo11O39, PW11O39Pd1, PMo11O39Rh1, and PMo11O39Pt1 after H2 exposure revealed that none of those exhibit a similar behaviour suggesting that the combination of Mo-based POMs and Pd has unique properties (Extended Data Fig. 1). UV-Vis spectroscopy of the Pd-Mo POM system showed the formation of two distinct states with absorption maxima at 700 and 316 nm, respectively. This indicates significant changes in the electronic state of the POM-based SAC as should be expected from the oxidation state change of Mo. Reduction of Mo(VI) to Mo(V) often leads to absorption at 600-800 nm 17,18 while more significant reduction of the POM changes the colour to a deep blue in this case with the absorption maximum in the UV region.
The occurrence of two distinct maxima observed by UV-Vis spectroscopy implies the formation of two sets of species during hydrogen spillover (Fig. 1c) . 1b and Supplementary Fig. 2

Kinetics of the hydrogen spillover:
The distinction of two related spillover species by UV-Vis provides us with a direct and non-destructive tool to follow hydrogen spillover kinetics. The formation of the first set of stable spillover intermediates with its UV-Vis absorbance at 700 nm has a positive order towards H2 concentration (1.52) confirming that H is involved in the rate-determining step and that the spillover entails multiple elementary steps ( Fig. 2a). We further observed a negative order towards the concentration of PMo11O39Pd1 (-0.99) (Fig.   2b). With this, the apparent activation enthalpy and entropy were determined to be 55.2 (±1.7) kJ mol -1 and -133.5 (±7.8) J mol -1 K -1 , respectively (Fig. 2c). Hydrogen spillover overall is an entropically disfavoured process and thus it is reasonable to assume that the apparent activation entropy is also negative.  Fig. 3). Noticeably, the hydrogen splitting mechanism also changes from a heterolytic to a homolytic with higher H coverage.
At higher coverage, the activation barriers for H2 splitting become almost negligible and thus we can assume the transfer of H from the Pd site to the support or the diffusion of H on the support to be ratedetermining. In contrast, the diffusivity of H on spillover catalysts has been shown to be inversely proportional to the extent of hydrogen coverage. 21 Those two competing effects might lead to the observation of only a few distinct hydrogen spillover intermediates. Overall, an important finding here is that the initial spillover of hydrogen benefits the subsequent activation and spillover of H, i.e., it is a self-catalytic process, which is consistent with the occurrence of an induction period that is sometimes observed for hydrogen spillover. 8

Fig. 2 | Kinetics of the hydrogen spillover on PMo11O39Pd1. Reaction orders towards a, H2 and b,
PMo11O39Pd1, during the hydrogen spillover process. c, Eyring plot for the first hydrogen spillover step with the absorbance at 700 nm.

Relation of hydrogen spillover to catalysis:
To identify the relation of hydrogen spillover to catalysis, we tested the nitrobenzene (NB) hydrogenation during H2 exposure of the catalyst. Remarkably, NB conversion increased in an accelerated manner with time (Fig. 3a). The reaction rate acceleration appeared to follow an approximately two-stage linear behaviour with constants of 0.040 and 0.263 min -2 , respectively ( The correlated reaction rates are thus increasing linearly for the unpoisoned case while they stayed constant for the case where the poison was added. Noticeably, the reaction rates were dependent on the time point of benzyl mercaptan addition indicating that the extent of spillover affects reaction kinetics ( Fig. 3d and Extended Data Fig. 4a). We observed similar trends for the hydrogenation of acetophenone (ACP) with increasing or constant reaction rates when no or two equivalents of benzyl mercaptan were added, respectively ( Fig. 3e and Extended Data Fig. 4b). For the hydrogenation of vinyl acetate (VA) comprising an unsaturated C=C bond the conversion increased linearly with time and the related reaction rates were constant. Upon addition of benzyl mercaptan, reaction rates immediately dropped to zero and thus the conversion did not increase anymore within 30 min (Fig. 3f and Extended Data Fig.   4c). This strongly suggests that spilled over H did not participate in the hydrogenation of C=C bonds.
Overall, there is a tendency where polar bonds can be hydrogenated by spilled over H while less polar bonds rely on H directly from the metal active sites.  We measured the initial hydrogenation turnover frequencies for NB and VA with PMo11O39Pd1 after different times of H2 exposure to obtain a clearer correlation between the extent of hydrogen spillover and the hydrogenation kinetics. A 33-fold increase in NB hydrogenation activity after 90 min initial hydrogen exposure was observed (Fig. 4b). In sharp contrast, the reaction rate dropped by a factor of 3.3 after 90 min H2 exposure of the catalyst for VA hydrogenation (Fig. 4a). This indicates that not only does hydrogen spillover contribute to the hydrogenation of various reducible groups differently, but also can hydrogen spillover be detrimental to catalysis in some cases.
Since experimental data provide compelling evidence that hydrogenation may occur both on the POM support through spilled over hydrogen and on the metal site through Pd-mediated hydrogen transfer, and high H coverages, respectively ( Fig. 4c and Extended Data Fig. 6). In contrast, the hydrogenation of polar bonds such as C=O in ACP only proceeds with low activation barriers when a catalyst contains a high H coverage. Low hydrogen coverages increase the reaction barriers and render the hydrogenation far less exothermic ( Fig. 4d and Extended Data Fig. 7). In addition, the local H density, depending on the location of H atoms, significantly impacted the hydrogenation reaction barriers (Supplementary Fig.   10). The hydrogenation of NB should be assumed to depend on the extent of hydrogen spillover as well, since the proclivity for forming oxygen vacancies is H coverage-dependent (Supplementary Table 2).
Kinetic barriers for the deoxygenation of ACP are high on Mo POM-based Pd SACs, in accordance with one of our previous studies ( Supplementary Fig. 11). 22

Implications for conventional H spillover catalysts:
To verify whether the findings related to hydrogen spillover are specific to POM-supported SACs or more generally applicable, we performed analogous investigations on the archetypical hydrogen spillover catalyst Pt/WO3, which undergoes a colour change from yellow to blue from this process (Extended Data Fig. 8). Plotting the initial reaction rates for the VA and ACP hydrogenation reaction against time after different durations of H2 treatment reveal a picture comparable to the case of PMo11O39Pd1. Hydrogenation rates drop by a factor of 7.8 for VA when Pt/WO3 was exposed to H2 for 90 min while the opposite is true for NB with a rate increase by a factor of 3.6 (Figs. 5a, b). Thus, spilled over H positively contributes to NB reduction while hinders VA hydrogenation. Further verification of this hypothesis was obtained from isotope labelling experiments. Provided that the spilled over H exchanges protons with protic solvents and considering that the rate of hydrogen spillover is considerably larger than its reversal, differences in the D incorporation should be expected when D2O and H2 are used during the hydrogenation reaction. Indeed, significant deuteration of the ACP hydrogenation product 1-phenylethanol was observed while the D incorporation was negligible when VA was used as substrate (Figs. 5c, d). This strongly supports the assumption that H/D from the support is involved in the hydrogenation of the polar C=O bond while only H from the Pt nanoparticles are transferred to VA.

Fig. 5 | Correlation between hydrogen spillover and the hydrogenation activity on Pt/WO3.
Substrate conversion for the hydrogenation of a, VA and b, NB after hydrogen spillover for 5 or 90 min. Deuterium isotope-labelling during the hydrogenation of c, VA and d, ACP in D2O as solvent after treatment with H2 as reductant for 120 min. The MS spectra of standard samples of the reaction products are shown for comparison.
Studies on bimetallic single-atom alloys have revealed certain factors affecting spillover catalysis but the nature of H on metallic surfaces is very different compared to more conventionally investigated metal oxides. 5,23,24 Understanding the impact of hydrogen spillover on the electronic structure of the catalyst as well as its hydrogenation activity might have implications for reactions that do not involve H atom transfer directly but which may either form H species as side-product or benefit from the cofeeding of H2 such as shown recently for epoxidation reactions. 25 For reactions like the CO2 or N2 hydrogenation, the contribution of hydrogen spillover to catalysis has been vividly discussed recently. [26][27][28][29] Based on our results, it is very likely that hydrogen spillover plays a crucial role in forming oxygen vacancies and transferring H to polar molecules like CO2. Care should be taken for correlating hydrogen spillover and enhanced catalytic activity since in some cases limiting the spillover of H on the support appears to be beneficial. We developed a related heterogeneous single-atom catalyst and observed comparable hydrogen spillover behaviour as well as hydrogenation kinetics for solid-gas biphasic processes and we intend to communicate those data soon.

Conclusions
In summary, the unique soluble SAC-POM catalyst system allows for the investigation of hydrogen spillover to a previously unattainable level of detail. Our observations based on kinetics, spectroscopy, spectrometry, and DFT calculations reveal that hydrogen spillover is an autocatalytic process, that the

Catalyst synthesis
The lacunary POMs were synthesised according to previously published procedures. 16 For K7PW11O39, in short, 18.15 g sodium tungstate dihydrate (SCR), 5 mL phosphoric acid (VWR, reagent grade) and 8.8 mL glacial acetic acid (Merck, reagent grade) dissolved in 30 mL deionised water and heated to reflux for 120 min. After the mixture was cooled down, 6 g potassium chloride (Sigma Aldrich, ACS reagent) were added, and the precipitation was continued for 60 min at room temperature. The obtained solid was separated by filtration and washed three times with small amounts of hot deionised water.
Product purity was improved by recrystallisation in 5 mL deionised water with 1 g potassium chloride.
This reprecipitation procedure was repeated three times and the final solid was dried at 95 °C overnight.
SACs were synthesised in situ by mixing appropriate amounts of K7PW11O39 and metal nitrate salt.