Intentional Formation of Persistent Surface Redox Mediators by Adsorption of Polyconjugated Carbonyl Complexes to Pd Nanoparticles

Abstract

Adsorbing polyconjugated carbonyl and aromatic species to Pd nanoparticles forms persistent intermediates that mediate reactions between hydrogen and oxygen-derived species. These surface redox mediators form in situ and increase selectivities toward H₂O₂ formation (∼65–85%) compared to unmodified Pd nanoparticles (∼45%). Infrared spectroscopy, temperature-programmed oxidation measurements, and ab initio calculations show that these species adsorb irreversibly to Pd surfaces and persist over extended periods of catalysis. Combined rates and kinetic isotope effect measurements and simulations suggest that carbonyl groups of bound organics react heterolytically with hydrogen to form partially hydrogenated oxygenated complexes. Subsequently, these organic species transfer proton–electron pairs to O₂-derived surface species via pathways that favor H₂O₂ over H₂O formation on Pd nanoparticles. Computational and experimental measurements show redox pathways mediated by partially hydrogenated carbonyl species form H₂O₂ with lower barriers than competing processes while also obstructing O–O bond dissociation during H₂O formation. For example, adsorption and hydrogenation of hexaketocyclohexane on Pd forms species that react with oxygen with high H₂O₂ selectivities (85 ± 8%) for 130 h on stream in flowing water without additional promoters or cosolvents. These paths resemble the anthraquinone auto-oxidation process (AAOP) used for industrial H₂O₂ production. These surface-bound species form partially hydrogenated intermediates that mediate H₂O₂ formation with high rates and selectivities, comparable to AAOP but on a single catalytic nanoparticle in pure water without organic solvents or multiunit reaction-separation chains. The molecular insights developed herein provide strategies to avoid organic solvents in selective processes and circumvent their associated process costs and environmental impacts.

1. Introduction

The coordination of organic molecules with metal atoms influences catalytic reactions by altering the electronic and geometric structure of active sites, changing interactions with solvent molecules, and reacting directly with bound species. − These phenomena are critical to the design of organometallic centers in homogeneous and enzymatic catalysts. , In heterogeneous systems, organic species (e.g., capping agents) can control the size, − shape, − and composition of colloidal nanoparticles , but also block active sites − and alter rates and selectivities. − For example, phosphine ligands inhibit the over-reduction of product molecules and improve selectivity for the partial hydrogenation of alkynes, carbonyls, and aromatics over late transition metals (e.g., Rh, Ni, Pd). − However, it is unclear how the interface of organic species and metal surfaces impacts the active sites and mechanisms involved in converting small molecules at solid–liquid interfaces. ,,−

Organic ligands can take on many forms, such as surfactants, ionic liquids, simple small molecules, or coadsorbates. ,, These species can transfer charge with catalytic surfaces and reactant molecules and alter adsorption, activation, and desorption of surface intermediates, ,− which impacts rates and selectivities for catalytic reactions. ,− For example, organic species can bind to metal surfaces, disrupt contiguous ensembles of metal atoms, and block sites that mediate undesired reaction pathways. − Such species may introduce bulky organic functions that affect the solvation sphere of catalytic surfaces , or sterically hinder the binding of reactive intermediates. ,,, Furthermore, ligands can induce electric fields and introduce acid–base functionalities − that stabilize or react with intermediates. These features offer several tactics to modify the reactivity of solid catalysts, but multiple interactions often coincide and obscure the origins of these effects.

Noble metals (e.g., Pd, Pt) are effective catalysts for hydrogenating oxygen and carbonyl-containing substrates, particularly in the presence of water molecules that aid the oxidation of adsorbed hydrogen atoms to protons and electrons during the reduction of such substrates. , Notably, reactions among H₂ and O₂ over metal nanoparticles respond sensitively to the presence of organic species within protic solvents. For example, the adsorption of hexadecyl-2-hydroxyethyl-dimethylammonium dihydrogen phosphate (HHDMA) on Pd improves the selectivity to form H₂O₂ (∼80%) by increasing barriers of O–O bond dissociation relative to clean Pd surfaces. Here, the authors hypothesized that HHDMA binds strongly to Pd and induces steric hindrance, stabilizing dioxygen intermediates (e.g., O₂*, OOH*, H₂O₂*). Later work examined a wider range of organic ligands and gave evidence that −OH functional groups present upon adsorbed ligands improved selectivities, consistent with observations for PVA-capped nanoparticles of Pd. , These observations supported the interpretation that associated hydrogen-bonding networks inhibit O–O bond dissociation paths. However, subsequent computation indicated that the −OH functions of H-bonding thiolates (e.g., ethanethiol, β-mercaptoethanol, and thioglycolic acid) bound to metal surfaces preferentially stabilize monatomic oxygen (e.g., O*) over dioxygen (O₂*) intermediates, suggesting that hydrogen-bonding networks should favor O–O bond cleavage and subsequent H₂O formation over H₂O₂ formation. Instead, these authors hypothesized that −OH functional groups of adsorbates may present alternative proton-coupled electron transfer (PCET) mechanisms for O–O bond dissociation. Taken together, the reasons for differences in rates and selectivities for reactions among H₂ and O₂ caused by the addition of organic surface species remain controversial and may involve site blocking, hydrogen-bonding of surface species, solvation at the surface, and the introduction of additional PCET pathways at solid–liquid interfaces.

Recent findings from our group contributed to a further understanding of these complex systems. We demonstrated that organic solvent molecules (e.g., methanol) spontaneously react with metal surfaces to form surface complexes that mediate reactions among H₂ and O₂ on Pd nanoparticles. A combination of computation, isotopic, and kinetic measurements show that methanol binds to Pd to form hydroxymethyl (e.g., CH₂OH*) redox mediators that provide low barrier pathways that reduce O₂-derived intermediates by PCET steps. Moreover, these reactions form formaldehyde (CH₂O*) that reacts spontaneously with H* to regenerate CH₂OH* and propagates a cycle of cocatalyzed heterolytic reactions of H₂ and O₂ with lower barriers than on clean surfaces of Pd. These findings align with other reports showing that organic species with carbonyl or hydroxyl functions facilitate selective hydrogen transfer reactions over metal nanoparticles by heterolytic processes. ,,− Thus, these carbonyl and hydroxyl intermediates act as surface-bound mediators that cocatalyze hydrogenation reactions analogous to homogeneous outer sphere mediators such as TEMPO in electrocatalytic systems and NADH in enzymatic reactions. −

The present study builds upon our prior work while drawing inspiration from the anthraquinone auto-oxidation process (AAOP), the dominant method for industrial production of H₂O₂. , In these systems, the carbonyl functions of quinone molecules react with H₂ over Pd surfaces, generating partially hydrogenated quinone species that react with O₂ and form H₂O₂ in separate homogeneous steps within organic media. Here, we demonstrate the intentional creation of persistent surface redox mediators that increase both rates and selectivities for H₂O₂ formation during steady-state catalysis in flowing water. Following ex situ adsorption of mediators from the solution, the modified Pd catalysts were tested within a fixed-bed reactor under a continuous flow of the reactants (H₂ and O₂) and pure liquid water. Measured rates, kinetic isotope effects, and catalyst stabilities, along with density functional theory (DFT) derived reaction energetics, suggest that PCET steps mediated by quinone-derived surface redox mediators on Pd nanoparticles allow H₂ and O₂ to react together through pathways that favor the production of H₂O₂. These surface-bound complexes increase H₂O₂ selectivities (up to 85%) relative to unmodified Pd (45%). Comparisons between experimentally measured and calculated activation barriers show that electrophilic mediators not only facilitate low-barrier pathways to form H₂O₂ but also inhibit O–O bond dissociation steps that yield H₂O. The organic surface redox mediators persist on Pd surfaces, leading to high rates and selectivities for long reaction periods (>130 h on stream; turnover numbers above 10⁶), consistent with the postreaction characterization of the used catalysts and highly exothermic adsorption energies of the mediators. Consequently, these findings demonstrate the intentional creation of surface redox mediators that cocatalyze reactions of H₂ and O₂ in water, inhibit undesired reaction pathways, and eliminate the need for organic solvents for H₂O₂ production and related chemistries.

2. Experimental Methods

The concise details of the experimental methods are discussed below. More thorough descriptions are provided in the Supporting Experimental Methods section, reporting additional details of experiments, relevant chemicals, purities, and vendors.

2.1. Catalyst Preparation

The Pd-SiO₂ catalysts were synthesized by strong electrostatic adsorption of cationic Pd precursors (Pd­(NO₃)₂ or (NH₄)₂PdCl₄) onto a porous SiO₂ support. Briefly, the SiO₂ was soaked in a solution of DI H₂O and NH₄OH and then mixed with a dissolved solution of the Pd precursor. The resulting mixture was stirred intermittently for 1h and left overnight. The solids were washed with DI water and vacuum-filtered for 24 h. The dried material was heated within a quartz tube furnace to 673 K for 4 h within a mixture of 2:1 He/Air. Then, the sample was purged with He and heated at 573 K for 4 h within a mixture of 4:1 He/H₂. The sample reached room temperature and was passivated in a mixture of 199:1 He/Air for 1 h before removal from the furnace.

The resulting Pd-SiO₂ catalyst was treated with organic species (e.g., benzoquinone derivatives), in which 230 mM of the organic were typically dissolved in dioxane, a solvent that negligibly affects rates (vide infra; Figure S17). These concentrations saturate the surface of Pd nanoparticles with the organic (vide infra; Figure ). These solutions were sparged with a mixture of 4:1 He/H₂ for at least 10 min before adding the Pd-SiO₂, which equilibrated for 1 h under continuously flowing H₂. The samples were then sealed and soaked for an additional 3 h. The resulting solids were then vacuum-filtered overnight. Some excess organic material was mixed within the catalyst, which was removed by flowing water (35 cm³ min^–1) during catalysis.

5.

Average (a) rates and (b) selectivities of H₂O₂ formation over SiO₂-supported Pd nanoparticles functionalized with increasing concentrations of 1,4-benzoquinone (blue), 1,4-naphthoquinone (red), 1,4-anthraquinone (black) before catalysis. Measurements used DI H₂O as the solvent within a fixed-bed reactor (200 kPa H₂, 60 kPa O₂, 278 K). Figures S11–S13 show the corresponding time-on-stream measurements.

2.2. Characterization of Catalytic Materials

2.2.1. Pd Nanoparticle Size and Composition

The numerical average of Pd nanoparticle diameters (<d _TEM,N >) were calculated from the mean diameter of particle size distributions obtained by transmission electron microscopy (TEM; Hitachi, H-9500, and ThermoFisher Themis Z) of at least 100 nanoparticles. Each sample was prepared by grinding the catalyst into a fine powder, dispersed in ethanol, and dripped onto a Cu holey-carbon TEM grid. The surface area normalized average diameter (<d _TEM,S >) for each catalyst was calculated using eq .

$$<d_{\mathrm{TEM},\mathrm{S}}>=\frac{\sum _i{n}_i{d}_i^3}{\sum _i{n}_i{d}_i^2}$$ 1

where n _i is the number of nanoparticles with the diameter d _i . Figure shows a representative image of a 4 nm Pd nanoparticle with a particle-size distribution histogram.

1.

Representative TEM image of 0.04 wt % Pd nanoparticles supported on SiO₂, in which an inset histogram shows the particle size distribution. Figure S1 shows corresponding images and histograms of particle diameters for the other Pd-based catalysts treated with organic species.

2.2.2. Examination of Pd Surface by Infrared Spectroscopy

Ex situ infrared spectra were obtained from Pd samples treated with and without organics, using adsorbed CO as a probe molecule. First, the catalyst was pelletized into a self-supporting disk and loaded into a transmission cell with CaF₂ windows. ,, Background spectra of samples were collected under flowing He, regulated by mass flow controllers. Samples were reduced by a mixture of 4:1 He/H₂ while heated at 373 K for 1 h. The sample was then cooled in pure He to 303 K, and new background spectra were collected. Then, the gas composition was changed to 50:1 He/CO, and steady-state spectra were collected. Next, the sample was oxidized at 573 K for 1 h in 4:1 He/O₂. The sample was again cooled in pure He to 303 K, and new background spectra were collected. Then, the gas composition was again changed to 50:1 He/CO, and steady-state spectra were collected. Data were normalized by Si–O–Si overtones as an internal standard from the SiO₂ support.

In situ infrared spectra were obtained using attenuated total reflectance. Here, 10 wt % Pd-SiO₂ was deposited on a ZnSe crystal and loaded into the instrument. The sample was pretreated at 573 K for 1 h in 4:1 He/O₂. The sample was then purged with He and treated with 4:1 He/H₂ at 373 K for 1 h before cooling to room temperature. Separately, two reservoirs of DI H₂O were sparged with H₂ and O₂, which were pumped over the catalyst with liquid flow rates of 3:1, respectively (10 mL min^–1, 75 kPa H₂, 25 kPa O₂, 298 K). Spectra were collected until the sample reached a steady state. Afterward, the pump connected to the H₂-sparged DI H₂O was switched to a 5 mM hexaketocyclohexane solution sparged with H₂ until a new steady state was reached.

2.2.3. Temperature-Programmed Oxidation and Desorption

Temperature-programmed oxidation (TPO) and desorption (TPD) profiles quantified the moles of organic species adsorbed to the catalyst surface. Samples were loaded onto a quartz frit within a fused quartz tube with an embedded thermocouple. A split tube furnace heated the material under 5% O₂/N₂ or Ar gas for TPO or TPD, respectively, regulated by a mass flow controller. A mass spectrometer analyzed the effluent gas to quantify CO₂ evolution (44 m/z). The sample was purged overnight with the gas at room temperature before measurements. Samples were heated to 973 K at 5 K min^–1 while monitoring the CO₂ evolution, which was calibrated using known quantities of NaHCO₃ (Figure S2) and integrated for the total moles evolved.

2.2.4. Determination of Elemental Composition by EDXRF

The metal loadings were determined by energy-dispersive X-ray fluorescence spectroscopy of Pd-SiO₂ samples (Table S1). Samples were prepared by loading catalytic materials into sample cups sealed with mylar film. The sample chamber was purged with a He environment and given at least 2 min to purge before conducting the measurement. The molar compositions of Si and Pd were quantified and converted into elemental weight loadings of Pd by assuming all elemental contributions come from Si and Pd signals.

2.3. Rate Measurements Using a Fixed-Bed Reactor

Rate and selectivities of H₂O₂ formation were measured in a continuous-flow fixed-bed reactor with a cooling jacket (Figure S3). ,, The reactor was loaded with catalyst, SiO₂, glass wool, and glass rods, which were secured and sealed by fritted gaskets. The temperature was controlled by flowing aqueous ethylene glycol through a cooling jacket from a recirculating bath and monitored by a thermocouple. H₂ and O₂ compositions in the reactor were controlled by flowing 25% H₂/N₂, 99.9% D₂, or 5% O₂/N₂, regulated by mass-flow controllers. An HPLC pump delivered the solvent, which was mixed with gas before contacting the catalyst. The pressure was maintained by a back-pressure regulator controlled by an electronic pressure regulator.

The reactor effluent entered a gas–liquid separator, and the gas was analyzed by a gas chromatograph with a capillary column and thermal conductivity detector using an Ar reference gas. Reactant conversions were compared to calibrated compositions of gas. The liquid fraction was periodically drained by an electronic valve and pushed into a 10-port valve, which injected the liquid and a colorimetric titrant (CuSO₄, neocuproine) into test tubes. The mixtures were analyzed by UV–vis to quantify H₂O₂ concentration using appropriate calibrations. Reported rates were normalized by the total metal content of the materials, and H₂O₂ selectivities were calculated by dividing the rate of H₂O₂ formation by the rate of H₂ consumption (r _H2O2 /–r _H2 ).

Control measurements were conducted using blank SiO₂, which does not form H₂O₂ (Figure S4). Measurements were also performed with SiO₂ prepared with 1,4-benzoquinone and hexaketocyclohexane equivalent to Pd-SiO₂ samples (Section ). The redox-active mediators react with the indicator but completely wash away after ∼1 h on stream. All experiments were conducted at a liquid flow rate of 35 cm³ min^–1 to avoid external mass transfer limitations. All samples satisfy the Madon-Boudart Criterion, enabling meaningful comparisons of catalytic rates and stability without mass transfer constraints. − Barrier measurements were corrected by accounting for the deactivation rates.

2.4. Rate Measurements Using Semibatch Reactors

H₂O₂ formation was measured in round-bottom three-neck flasks in a semibatch configuration (Figure S5). The flasks were connected to a condenser chilled to 273 K by a recirculating bath. The reactant gases of H₂ and 5% O₂/N₂ were fed through gas dispersion tubes submerged in the solvent and regulated by mass-flow controllers. Measurements were conducted at ambient temperatures and pressures, in which the catalyst slurry was agitated at 1500 rpm by magnetic stir bars to avoid mass transfer limitations. The solvent was sparged for 10 min before adding the catalyst. Aliquots were extracted periodically to quantify H₂O₂ concentrations using the CuSO₄ indicator (vide supra). Concentration profiles were fit to determine rate constants.

2.5. Computational Methods

Concise details of the computational methods used are discussed below. More thorough descriptions are provided in the Supporting Computational Methods section.

Homogeneous aqueous phase DFT calculations were carried out using the Gaussian 16 software package to determine the electronic properties of various organic mediators. The M06–2X exchange-correlation functional, was used along with the 6–311++G­(d,p) basis set, and an implicit SMD solvation model for water.

Periodic plane-wave DFT calculations were carried out using the Vienna Ab initio Simulation Package (VASP) to model the mediated reactions of H₂ and O₂ on Pd. , A slab comprised of a 4 × 4 unit cell with four layers of the Pd(111) surface was used as a model of the exposed Pd facets of the supported Pd nanoparticles. The bottom two layers of the surface were held fixed to the bulk Pd positions while the top two layers were allowed to relax. Previous experimental and computational results suggest steady state coverages of ∼5/16 monolayer (ML) of oxygen (O*) and ∼5/16 ML of subsurface hydrogen (H_s*), which are similarly considered. , As shown in our prior work, the dissociative adsorption of H₂ has a small barrier and is highly reversible; hence, one surface hydrogen with other surface adsorbates (O*, mediators, O₂*) has been chosen as a resting state for calculations.

Ab initio molecular dynamics (AIMD) simulations were carried out to incorporate explicit solvent molecules in the model and provide more faithful models for the solution phase structures and energies. NVT ensemble dynamics at 300 K were performed with the Nosé-Hover thermostat. A box size of 10 × 10 × 10 Å was first filled with solvent and the individual mediator(s) as an initial guess for the bulk solvation phase of the mediator. AIMD simulations were then carried out for five picoseconds to obtain the bulk solvation phase of the mediators. This equilibrated bulk solvated phase was then used as an initial guess for the solvent and mediator interacting with the Pd(111) surface with the surface O* adsorbates. The condensed phase was simulated by filling the 15 Å vacuum region between the slabs with water molecules with a density of 1 g cm^–3. This initial interfacial structure was equilibrated for another five picoseconds to allow the explicit solvent molecules to relax around the surface species.

Periodic DFT calculations were carried out using a plane-wave energy cutoff of 400 eV. The Perdew–Burke–Ernzerhof (PBE) functional form of the generalized gradient approximation (GGA) was employed to determine the corrections to the energy due to exchange and correlation effects. PAW pseudopotentials were used to describe the interactions between the valence and core electrons. Noncovalent long-range interactions in the system were modeled with the D3 dispersion corrections by Grimme. The wave functions were self-consistently optimized to within 10^–6 eV using a 2 × 2 x 1 γ-centered k-point mesh. The atomic positions were iteratively optimized until the maximum force was less than 0.05 eV/Å. The nudged elastic band (NEB) method was initially used to locate transition states. Transition states were further refined using the dimer method.

3. Results and Discussion

3.1. Effect of Organic Species on Rates of H₂O₂ Formation

Figure shows rates and selectivities of H₂O₂ formation as functions of time-on-stream for Pd-SiO₂ and two materials formed by treating Pd-SiO₂ either with 1,4-benzoquinone (BQ) or hexaketocyclohexane (HKH) (200 kPa H₂, 60 kPa O₂, 278 K). These materials exhibit an induction period during which rates and selectivities approach nearly constant values over a period of 20 h, suggesting structural or compositional changes in the form of the catalyst. Such changes reflect the aggregation of Pd nanoparticles (Figure S1) or the evolution of organic adsorbates bound to Pd surfaces, similar to our prior findings. Untreated Pd nanoparticles present greater steady-state turnover rates of H₂ consumption (−r _H2 ) than samples treated with BQ (r _H2 /r _H2 = 0.51) or HKH (r _H2 /_H2 = 0.75), which suggests BQ- and HKH-derived species inhibit steps for H₂ and O₂ activation. Figure a shows H₂O₂ formation rates, however, are greater on HKH-treated (r _H2O2 /r _H2O2 = 1.40) and lower on BQ-treated samples (r _H2O2 /r _H2O2 = 0.75) in comparison to untreated Pd nanoparticles. Moreover, Figure b shows selectivities of H₂O₂ formation over 20 h on stream (200 kPa H₂, 60 kPa O₂, 278 K), which reach mean values of 44% for untreated, 64% for BQ-treated, and 83% for HKH-treated Pd nanoparticles in the final 4 h of the measurement. Consequently, adsorbed BQ and HKH-derived species block sites and inhibit O–O bond dissociation paths, consistent with prior observations with organic adsorbates. ,,, Still, the diminished rates of H₂ activation and increased rates of H₂O₂ formation on HKH-treated samples suggest that these organics block surface active sites, but the remaining sites present between the adsorbed organics are much more reactive for forming H₂O₂ relative to those on untreated Pd. These observations agree qualitatively with the measured effects of quinone compounds on H₂O₂ formation rates and yields upon Pd nanoparticles supported on zeolite β in batch reactors. ,

2.

(a) Rates and (b) selectivities of H₂O₂ formation as a function of time over SiO₂-supported Pd nanoparticles without treatment (black) and following treatment with 1,4-benzoquinone (blue) or hexaketocyclohexane (red). Measurements used DI H₂O as the solvent within a fixed-bed reactor (200 kPa H₂, 60 kPa O₂, 278 K).

Figure shows the rates and selectivities of H₂O₂ formation over 130 h on stream. The mean H₂O₂ selectivity increases to ∼90% in the final 8 h of the measurement on HKH-treated samples. However, the total rate of H₂O₂ formation decreased by 32% over this interval, consistent with a loss of Pd content (0.025 to 0.012%, Table S1). In comparison, rates, selectivities, and Pd contents of untreated and BQ-treated Pd catalysts appear stable during reactions in water over similar periods (20 h, Table S1). These findings suggest that HKH leads to the formation of soluble molecular complexes that leach Pd from the catalyst, similarly observed during the dissolution of noble metals in the presence of chelating organic compounds (e.g., acetylacetone) and strong oxidants (e.g., H₂O₂, O₂). ,

3.

(a) Rates and (b) selectivities of H₂O₂ formation as functions of time over SiO₂-supported Pd nanoparticles without treatment using a 70% (v/v) mixture of methanol and water as the solvent (black) and treated with hexaketocyclohexane using DI H₂O as the solvent (red). Measurements were within a fixed-bed reactor (200 kPa H₂, 60 kPa O₂, 278 K).

Prior studies identified a mixture of 70% by volume methanol in water as an optimal solvent for H₂O₂ synthesis in batch reactors over short periods (∼0.5 h). − Although alcohol and aqueous alcohol solvent yield higher initial rates of H₂ consumption in comparison to water, alcohol-derived residues accumulate on Pd nanoparticles and inhibit reactions of H₂ and O₂. ,,, For example, rates of H₂O₂ formation on Pd-SiO₂ decrease by 78% over 130 h in 70% vol. methanol in water, while giving low mean H₂O₂ selectivities (28%) in the final 8 h of the experiment. In comparison, HKH-modified Pd-SiO₂ provides greater rates, selectivities, and stability than Pd-SiO₂ catalysts operating within methanol solvent or aqueous methanol solutions. Thus, modifying Pd nanoparticles with organic species provides a strategy to decouple the advantages of organic solutions (higher initial rates and selectivities of H₂O₂ formation) from their drawbacks (rapid deactivation, costly and environmentally impactful organic solvent) over much longer reaction intervals.

In contrast to reactions conducted in excess hydrogen, O₂-rich conditions (60 kPa H₂, 100 kPa O₂, 278 K; Figures S6 and S7) result in lower selectivities to H₂O₂ on untreated Pd (18%), BQ-treated (24%), and HKH-treated (27%) Pd nanoparticles in the final 4 h of the 20 h time-on-stream measurement. These observations correlate with in situ X-ray absorption spectra showing excess O₂ stabilizes the metallic Pd while excess H₂ favors PdH _x formation, − which in our prior studies was found to stabilize O–O bonds and improve H₂O₂ selectivity. The presence of BQ- and HKH-derived species still leads to greater H₂O₂ selectivities; however, the absolute selectivities and the fractional improvements are more significant in H₂-rich conditions (200 kPa H₂, 60 kPa O₂, 278 K; Figure b), which suggests the higher coverages of reactive oxygen found upon metallic Pd surfaces decompose or destabilize the BQ- and HKH-derived organic species. Cyclic changes between H₂- (200 kPa H₂, 60 kPa O₂) and O₂-rich (60 kPa H₂, 100 kPa O₂) conditions at 278 K leads to initially lower H₂O₂ selectivities on organic-treated samples (e.g., 83 to 53% for HKH-Pd-SiO₂) upon returning to H₂-rich conditions for 2 h (Figure S8). However, the H₂O₂ selectivities return to the previous steady-state values after a 20 h period (Figure S9), which indicates the regeneration of the selective form of the catalyst. In contrast, untreated Pd samples recover their steady-state selectivities within less than 1 h after returning to H₂-rich conditions. These comparisons suggest that there may be a pool of quinone- and hydroquinone-derived intermediates on the Pd surface that undergoes reversible changes upon exposure to high coverages of oxygen and hydrogen-derived surface species.

Still, the species that show the greatest H₂O₂ formation rates and selectivities only form at high H₂ pressure and require many hours to recover the composition and catalytic structures responsible for the selective reduction of O₂ to H₂O₂ on Pd.

In contrast to expectations, H₂O₂ formation rates and selectivities differ only weakly between catalysts that use distinct 1,4-benzoquinone analogs, with the exceptions of chlorine and additional carbonyl groups (vide infra). Furthermore, polyaromatic species (e.g., 1,4-naphthoquinone, 1,4-anthraquinone; Figures S11–S13) do not introduce significant differences relative to 1,4-benzoquinone. These comparisons suggest that inductive effects that impact the extent of electron withdrawal from carbonyl groups do not significantly impact catalysis or that these functional groups speciate in situ (e.g., hydrogenolysis of C–X bonds results in loss of heteroatoms). By comparison, catalysts formed by adsorption of benzene analogs (C₆X₄H₂; X = H, CH₃, F, Cl) to Pd nanoparticles also improve H₂O₂ selectivities (Figures S14 and S15), despite the absence of carbonyl groups needed to provide redox activity from the initial polyconjugated structure. Similarly, the adsorption of C₁–C₃ aldehydes and ketones (e.g., formalin, acetone) increases H₂O₂ selectivities relative to unmodified Pd (Figure S16), albeit to a lesser extent than observed on the cyclic organic compounds. In contrast, Pd nanoparticles treated with aliphatic or ether species (e.g., n-hexane, tetrahydrofuran, or dioxane) show negligible differences in reactivity compared to untreated samples (Figure S17).

Taken together, these observations give evidence that carbonyl groups, aromatics, and polyconjugated rings facilitate strong adsorption of organic species to Pd surfaces where redox active complexes form in situ by pathways that cleave C–X bonds and form C–O bonds to produce surface complexes with a chemical function analogous to quinones. This interpretation agrees with prior reports that Pd nanoparticles catalyze the oxidation of benzene to phenol (423–473 K) − and the reduction of quinones to hydroquinones (313–343 K). ,, These findings suggest that Pd introduces oxygen to aromatic rings and subsequently forms hydroxyl groups. Indeed, these conclusions agree with our theoretical calculations showing that aromatics bind strongly to Pd surfaces and transform quickly to hydroxylated species (vide infra). Consequently, these results are consistent with quinones, aromatics, and carbonyl compounds reacting with Pd nanoparticles to form persistent organic surface species that mediate reactions of H₂ and O₂ to produce H₂O₂.

While most analogs of benzene and 1,4-BQ provide similar selectivities for H₂O₂ formation, adsorption of tetrachloro-1,4-benzoquinone, 1,2,4,5-tetrachlorobenzene, or HKH leads to statistically significant increases in H₂O₂ selectivities (Figures and S10). These differences implicate the in situ formation of adsorbate structures distinct from other polyconjugated and aromatic compounds. For example, adsorption of chloride and bromide to Pd nanoparticles increases H₂O₂ selectivities, ,− potentially due to the disruption of contiguous ensembles of Pd atoms needed to dissociate O–O bonds. Here, the modest increases in H₂O₂ selectivities are consistent with the in situ speciation of chlorine (i.e., breakage of C–Cl bonds) from the chlorinated BQ analogs and their subsequent adsorption to Pd nanoparticle surfaces (Figure ). Note, however, that fluorine atoms speciated from fluorinated analogs have much smaller ionic radii and weaker effects on selectivity, a trend similar to those reported in studies comparing halides during H₂O₂ formation on Pd. Regardless, this speciation hypothesis does not account for greater rates and selectivities for H₂O₂ formation with HKH (i.e., in the presence of additional carbonyl moieties). Rather, these observations suggest that the quinones react with hydrogen to form partially hydrogenated quinones (HQ) with −OH functional groups. These species likely stabilize or react with O₂-derived intermediates (e.g., O₂*, OOH*) and promote the formation of H₂O₂, as was described for the in situ formation of hydroxymethyl species on Pd derived from methanol.

4.

Average rates (black) and selectivities (red) of H₂O₂ formation over SiO₂-supported Pd nanoparticles without treatment and following treatment with analogs of 1,4-benzoquinone (C₆X₄O₂) with different functional groups (X = H, CH₃, F, Cl, O). Measurements used DI H₂O as the solvent within a fixed-bed reactor (200 kPa H₂, 60 kPa O₂, 278 K), in which rates were averaged over 1200 min on stream. Figure S10 shows the corresponding time-on-stream measurements.

3.2. Quinone Species Adsorb Strongly to Pd Surfaces

Figure shows rates and selectivities of H₂O₂ formation averaged over 1200 min on stream, which increase with increasing concentrations of quinone species (1,4-benzoquinone, 1,4-naphthoquinone, 1,4-anthraquinone). These rates and selectivities approach constant values with sufficiently large concentrations of quinones, suggesting these species form complete monolayer coverages on the Pd surfaces. Such observations agree with temperature-programmed oxidation (TPO) conducted on materials following reactions for 1200 min on stream (200 kPa H₂, 60 kPa O₂, 278 K), showing that BQ-treated and HKH-treated Pd nanoparticles (Figure S18a) evolve ∼10-times the quantity of CO₂ (303–423 K) as observed for untreated Pd-SiO₂ materials, indicating the oxidation of persistent organic species from Pd nanoparticles. Control experiments (Figure S19) show that the TPO of untreated SiO₂ samples evolves CO₂ only at temperatures above 423 K (Figure S19a), consistent with the oxidation of adventitious organics on the silica surface. Moreover, temperature-programmed desorption (TPD) shows that similar quantities of CO₂ evolve (303–423 K) from organic-treated and untreated samples (Figure S18b) when heated in flowing helium, which reflects desorption of atmospheric CO₂ or oxidation of organic species by strongly bound oxygen on Pd surfaces but not SiO₂ (Figure S19b). Still, TPD profiles show the evolution of significantly fewer moles of CO₂ than in TPO profiles, indicating that most of the CO₂ evolution on organic-treated samples comes from the oxidation of BQ- or HKH-derived species. Thus, organic species saturate the surface of Pd nanoparticles and persist over extended periods of catalysis.

Figure shows ex situ infrared spectra of CO adsorbed on Pd-SiO₂ samples after catalysis. Before the adsorption of CO, each material was treated in flowing H₂ (21 kPa H₂, 373 K, 1 h) to reduce oxidized Pd and remove H₂O. Multiple vibrational features appear between 2015–2175 cm^–1 that correspond to CO adsorbed at atop sites on Pd and features below 2000 cm^–1 that reflect CO bound to bridge sites between Pd atoms. ,, Spectra of adsorbed CO on each material following catalysis (200 kPa H₂, 60 kPa O₂, 278 K, 20 h) appear similar, but the relative intensity of atop and bridging CO features differ compared to untreated Pd samples. Specifically, the line shape and peak centers of atop CO show minor changes between fresh Pd (2088 cm^–1) and spent samples of Pd (2090 cm^–1), BQ-Pd (2092 cm^–1), and HKH-Pd (2093 cm^–1). Similarly, the bridging CO feature (∼1910 cm^–1) appears broad and unimodal on all catalysts. Thus, organic residues may form upon the Pd surfaces either from the intentional addition of organic species or by adventitious carbon compounds that bind to untreated surfaces.

6.

Ex situ infrared spectra of CO-saturated surfaces of SiO₂-supported Pd nanoparticles (0.04 wt % Pd, 0.02 kPa CO, 303 K) freshly prepared (gray) and after catalysis within a fixed-bed reactor (200 kPa H₂, 60 kPa O₂, 278 K) for 20 h. Spent samples include those loaded in the reactor without treatment (black) and with treatment of 1,4-benzoquinone (blue) or hexaketocyclohexane (red). Spectra were collected after initial reduction (20 kPa H₂, 373 K) for 1 h (dotted lines) and subsequent oxidation (20 kPa O₂, 573 K) and reduction (20 kPa H₂, 573 K) for 1 h each (solid lines).

Next, samples were oxidized (21 kPa O₂) and reduced (21 kPa H₂) at 573 K to remove organic species remaining on the sample. Figure S20 shows the difference in background spectra before and after this subsequent thermal treatment, showing the removal of adventitious carbon and H₂O remaining on the material. However, the thermal pretreatments used to clean the Pd metal before FTIR measurements may irreversibly change the organic species on Pd. Upon CO exposure, all catalysts show changes in line shape and an increase in total peak area compared to the initial reductive treatment, which is consistent with the emergence of a greater quantity of sites that bind CO after high-temperature oxidation. Notably, new features emerge that agree with CO adsorbed atop oxidized atoms of Pd (∼2133 cm^–1) and CO bridging undercoordinated ensembles of Pd (∼1990 cm^–1). − Since these sites only appear after high-temperature oxidation, they likely bind adsorbates strongly and may be blocked by organic species during catalysis, whether added intentionally or not. Nevertheless, the line shape and ratio of the integrated peak area of bridging and atop CO are nearly identical for each spent sample (Figure and Table S2), consistent with the similar particle size distribution of these catalysts (Figure S1). Still, this ratio increases between fresh and spent materials, corroborating the agglomeration of Pd nanoparticles from a diameter of ∼4 to ∼9 nm after catalysis (Figure S1). As such, the Pd nanoparticles of the final samples are morphologically similar irrespective of their modification with HKH or BQ, suggesting they do not influence the core structure of nanoparticles. Thus, the electronic state of Pd is likely similar between each sample, indicating quinones affect catalysis by altering reactions on the nanoparticle surface.

Furthermore, we sought to better understand the in situ transformation of these species upon adding mediators. Figure S21 shows that the addition of HKH leads to growth of broad features consistent with alkoxy (υ­(C–O) at 1060 cm^–1), aromatic (υ­(CC) at 1540, 1640, and 1690 cm^–1), carbonyl (υ­(CO) at 1770 cm^–1), and hydroxyl functions (υ­(O–H) at 3200 cm^–1), distinct from molecular HKH. Taken together, these spectra suggest that HKH adsorbs to the surface of Pd and reacts with hydrogen to generate hydroxyl functions. However, HKH intermediates may also decompose a variety of carbonylic and other organic species on the Pd surface, which react with H₂ and O₂, similarly observed for methanol speciation in our prior work.

3.3. Dependence of Rates on Isotopic Compositions

These results show that organic species bind to Pd nanoparticles and modify oxygen reduction paths over extended periods of catalysis. Still, these data do not distinguish whether these changes result from site blocking or the introduction of new reaction pathways. Thus, we performed isotopic measurements to understand if the adsorption of organic species influences the mechanism of proton transfer reactions, as reported in our prior work. Table shows the relative rates of hydrogen activation and oxygen reduction paths over HKH-treated and untreated Pd nanoparticles using H₂ and D₂ reactants and H₂O and D₂O solvents. Rates of oxygen reduction decrease when using D₂ versus H₂ as the reductant (200 kPa H₂ or 200 kPa D₂, 60 kPa O₂, 278 K), which implicates kinetically relevant reactions of H₂ (Figure S22). Here, the kinetic isotope effects of H₂O₂ formation on HKH-treated (k _H2 /k _D2 = 1.17 ± 0.03) and untreated (k _H2 /k _D2 = 1.18 ± 0.01) surfaces of Pd appear similar. Yet, D₂ lowers rates of H₂O formation by a greater extent for HKH-treated (k _H2 /k _D2 = 2.5 ± 0.5) than untreated (k _H2 /k _D2 = 1.5 ± 0.2) samples. Additionally, rates of H₂O₂ formation (k _H2O/k _D2O = 1.0 ± 0.1) and decomposition (k _H2O/k _D2O = 1.0 ± 0.2) on untreated Pd do not respond to the isotope substitution of the solvent water. Together, these observations on untreated Pd nanoparticles indicate that heterolytic hydrogen activation and oxidation occur in kinetically relevant processes; however, proton transfer steps from water do not affect rates. , In contrast, rates on HKH-treated Pd samples respond to changes from perhydrogenated to perdeuterated water solvent and give clear kinetic isotope effects for H₂O₂ formation (k _H2O/k _D2O= 1.3 ± 0.4) and decomposition (k _H2O/k _D2O = 1.8 ± 0.7). These differences imply that the presence of HKH-derived residues introduces new reaction pathways that involve kinetically relevant proton transfer steps that do not occur in water.

1. Effect of Isotopic Substitution on Rate Constants for the Formation of H₂O₂ and H₂O on Silica-Supported Pd Nanoparticles before and after Treatment with Hexaketocyclohexane Using H₂ or D₂ Reactants within a Fixed-Bed Reactor (200 kPa H₂ or D₂, 60 kPa O₂, 278 K; Figure S22)­ .

	H2/D2		H2O/D2O
kH/kD	Pd	HKH-Pd	Pd	HKH-Pd
H2 + O2 → H2O2	1.18 ± 0.01	1.17 ± 0.03	1.0 ± 0.1	1.3 ± 0.4
H2 +1/2O2 → H2O	1.5 ± 0.2	2.5 ± 0.5
H2O2 + H2 → 2H2O			1.0 ± 0.2	1.8 ± 0.7

^aComplementary H₂O₂ formation and decomposition measurements on the same material using H₂O or D₂O solvents (80 mL) within a semibatch reactor (4.8 kPa H₂, 4.8 kPa O₂, 298 K). Figure S23 shows the transient concentration profiles of H₂O₂ formation used to fit these rate constants.

These observations align with our theoretical calculations (vide infra) and observed differences in catalytic performance and kinetic isotope effects observed between reactions among H₂ and O₂ on Pd nanoparticles in pure water and in aqueous organic solutions. Those earlier results demonstrated product formation rates depend sensitively on isotope substitution (i.e., H₂O/D₂O, and H₂/D₂), which implicated hydroxymethyl intermediates (CH₂OH*) in reactions that mediated PCET steps that reduced O₂*, OOH*, and O* surface intermediates. Here, the carbonyl groups of HKH (and similar functions of other organic species examined) seem likely to convert to −OH/D functions by reaction with H*/D*, and these functions can rapidly exchange protons and deuterons with the solvent (H₂O, D₂O). Consequently, the reduction of O₂-derived surface species to hydrogen peroxide or water occurs at lower rates in D₂O due to the kinetically relevant transfer of deuterons from HKH-derived species. Without organic residues, rates in D₂O and H₂O are indistinguishable. Indeed, recent findings show measurable kinetic isotope effects for oxygen reduction with related quinone compounds over M–N–C catalysts within solutions of H₂O or D₂O, which corroborate this interpretation. These observations support the hypothesis that HKH and other polyconjugated carbonyl and aromatic compounds adsorb to Pd nanoparticles and form surface redox mediators (e.g., (CO)₆H*) that provide new reaction pathways for the catalytic reduction of O₂ with H₂ that favor the retention of O–O bonds and formation of H₂O₂.

3.4. Effect of Organic Adsorbates on Reaction Barriers

Quinone and carbonyl species block sites and introduce new proton transfer paths that favor H₂O₂ formation, consistent with these species stabilizing transition states that convert OOH* to H₂O₂. Rates of H₂O₂ and H₂O formation were measured as a function of temperature to calculate apparent activation enthalpy barriers for H₂O₂ (ΔH _H2O2 ) and H₂O (ΔH _H2O ) formation, which were then compared to DFT-calculated values using model Pd surfaces, representative of experiments.

Quantitative descriptions of reaction rates help elucidate the mechanistic interpretation of these barrier measurements (Figure S24), as shown by

$$-r_{{\mathrm{H}}_2}=r_{{\mathrm{H}}_2\mathrm{O}}+r_{{\mathrm{H}}_2{\mathrm{O}}_2}$$ 2

where the rates of H₂O₂ (r _H2O2 ) and H₂O (r _H2O) formation equal that of H₂ consumption (−r _H2 ). Under the conditions used here, prior work suggests that these rates can be described by

$$r_{{\mathrm{H}}_2{\mathrm{O}}_2}=k_{{\mathrm{H}}_2{\mathrm{O}}_2}[{\mathrm{O}}_2]$$ 3

$$r_{{\mathrm{H}}_2\mathrm{O}}=k_{{\mathrm{H}}_2\mathrm{O}}[{\mathrm{O}}_2]$$ 4

where r _H2O2 and r _H2O depend on the apparent rate constants of H₂O₂ (k _H2O2 ) and H₂O (k _H2O) formation, increase in proportion to the pressure of O₂ (i.e., [O₂]), and remain independent of the pressure of H₂. Complete derivations of eqs – are presented in detail elsewhere. , Briefly, the values of k _H2O2 and k _{H 2 O} reflect the relative preference of hydroperoxyl (OOH*) to reduce to H₂O₂ or dissociate to H₂O, respectively. Here, all paths that form H₂O occur irreversibly, as evidenced by studies showing that mixtures of H₂, ¹⁶O₂, and ¹⁸O₂ form H¹⁶O¹⁶OH and H¹⁸O¹⁸OH but not H¹⁶O¹⁸OH. Thus, the formation of H₂O₂ requires the preservation of O–O bonds.

The tenets of transition state theory describe the relationship between the ratio of H₂O₂ and H₂O formation rates and the free energy differences between the apparent activation free energies as

$$\frac{r_{{\mathrm{H}}_2{\mathrm{O}}_2}}{r_{{\mathrm{H}}_2\mathrm{O}}}=\frac{k_{{\mathrm{H}}_2{\mathrm{O}}_2}}{k_{{\mathrm{H}}_2\mathrm{O}}}=\frac{{\mathrm{e}}^{-\mathrm{\Delta }{G}_{{\mathrm{H}}_2{\mathrm{O}}_2}^{‡}/\mathit{RT}}}{{\mathrm{e}}^{-\mathrm{\Delta }{G}_{{\mathrm{H}}_2\mathrm{O}}^{‡}/\mathit{RT}}}={\mathrm{e}}^{\mathrm{\Delta }{S}_{{\mathrm{H}}_2{\mathrm{O}}_2}^{‡}-\mathrm{\Delta }{S}_{{\mathrm{H}}_2\mathrm{O}}^{‡}/R}•{\mathrm{e}}^{\mathrm{\Delta }{\mathrm{H}}_{{\mathrm{H}}_2\mathrm{O}}^{‡}-\mathrm{\Delta }{\mathrm{H}}_{{\mathrm{H}}_2{\mathrm{O}}_2}^{‡}/\mathit{RT}}$$ 5

where the difference in transition state Gibbs free energies of H₂O₂ (G _H2O2 ) and H₂O (ΔG _H2O ) formation reflects the difference in the apparent values of ΔH _H2O2 and ΔH _H2O (i.e., ΔΔH ^‡ = ΔH _H2O – ΔH _H2O2 ) and their entropic contributions (ΔS _H2O2 ,ΔS _H2O ).

Similar forms of expressions describe the rate of H₂ consumption at high coverages of H*, which appears as

$$-r_{{\mathrm{H}}_2}=k_{{\mathrm{H}}_2}[{\mathrm{O}}_2]\propto {\mathrm{e}}^{-\mathrm{\Delta }{G}_{{\mathrm{H}}_2}^{‡}/\mathit{RT}}[{\mathrm{O}}_2]\propto {\mathrm{e}}^{-\mathrm{\Delta }{\mathrm{H}}_{{\mathrm{H}}_2}^{‡}/\mathit{RT}}[{\mathrm{O}}_2]$$ 6

where the H₂O₂ and H₂O formation rates sum to the apparent rate constant of H₂ reactivity (k _H2 ), reflecting the apparent Gibbs free energy (ΔG _H2 ) and enthalpy (ΔH _H2 ) of activation for H₂ consumption.

Table shows values for apparent activation enthalpies for H₂O₂ and H₂O formation and H₂ consumption on untreated Pd nanoparticles and those treated with BQ or HKH. Measurements were conducted at reaction conditions where rates do not depend on the pressure of H₂ (200 kPa H₂, 60 kPa O₂, 278–308 K), which align with the forms of eqs –. Values of ΔH _H2 are greater in the presence of either BQ- or HKH-derived surface species, which suggests these complexes inhibit the activation of H₂. Comparisons among values of ΔH _H2O2 and ΔH _H2O demonstrate that barriers for H₂O₂ formation remain nearly constant within the uncertainty of the measurements. However, barriers for H₂O formation via steps that cleave O–O bonds increase significantly. These differences agree with slight differences in H₂O₂ formation rates and notable increases in H₂O₂ selectivities caused by the adsorption of HKH and other organic intermediates (Figures –). These observations may reflect the strong adsorption of organic species that interrupt contiguous ensembles of Pd atoms and likely destabilize transition states and product states for reactions that cleave O–O bonds. However, differences in measured kinetic isotope effects suggest that organic species also modify the mechanisms for reactions of H₂ and O₂ by introducing steps that transfer protons and electrons through these complexes.

2. Apparent Activation Enthalpies of H₂O₂ (ΔH _{H 2 O 2} ) and H₂O (ΔH _{H 2 O} ) Formation and H₂ (ΔH _{H 2} ) Activation over Untreated Pd Nanoparticles and Pd Nanoparticles Treated with Either 1,4-Benzoquinone (BQ) or Hexaketocyclohexane (HKH) .

material	ΔH H2O2 (kJ mol–1)	ΔH H2O (kJ mol–1)	ΔH H2 (kJ mol–1)
untreated Pd-SiO2	8 ± 3	19 ± 3	15 ± 2
BQ-treated Pd-SiO2	9 ± 1	36 ± 5	20 ± 2
HKH-treated Pd-SiO2	12 ± 5	46 ± 10	27 ± 9

^aMeasurements used DI H₂O as the solvent within a fixed-bed reactor (200 kPa H₂, 60 kPa O₂, 278–308 K). Figure S24 shows the corresponding rate measurements as a function of temperature.

Plausible reaction mechanisms and the effects of quinone adsorption on surface reactions were evaluated with DFT calculations carried out on a model Pd(111) surface (the dominant facet of Pd nanoparticles). The calculations examined surfaces that possessed adsorbed surface O* and subsurface H_s* (similar to adlayer structures determined in our prior work (Figure S25) together with the reactive quinone species (e.g., C₆O₆*)). The adsorption of these mediators on the surface was initially examined using implicit water solvation, after which the surfaces were solvated with explicit water molecules, and AIMD simulations were conducted to determine the lowest-energy structures for reactive surface species and hydrogen bonding networks at the solid–liquid interface. The models constructed in this manner were used to determine reaction energies (ΔE _rxn) and intrinsic barriers (ΔE ^⧧) for relevant elementary steps (Figures and ).

7.

Reaction coordinate diagrams showing DFT-calculated energies for forming partially hydrogenated quinone-based redox mediator by adsorption and reaction with surface hydrogen by water-mediated steps on Pd(111)-based surfaces. The simulated Pd(111) surfaces contain adsorbed O* (5/16 ML) and subsurface H_s* (5/16 ML) to represent the state of Pd nanoparticles at reaction conditions determined previously by operando EXAFS. The green circles show hydrogen atoms heterolytically oxidized to form protons (and electrons), and the yellow circles show the proton (and electron) transfer.

8.

Reaction coordinate diagrams showing DFT-calculated energies for the catalytic formation of H₂O₂ (black solid lines) and H₂O (red dashed lines) mediated by (a) partially hydrogenated 1,4-benzoquinone, (b) partially hydrogenated hexaketocyclohexane, and (c) solution-phase water molecules on Pd(111)-based surfaces. The simulated Pd(111) surfaces contain adsorbed O* (5/16 ML) and subsurface H_s* (5/16 ML) to represent the state of Pd nanoparticles at reaction conditions determined previously by operando EXAFS. The green circles show hydrogen atoms heterolytically oxidized to form protons (and electrons), and the yellow circles show the proton (and electron) transfer.

Quinones and other polyconjugated carbonyl and aromatic species bind to the surface by Pd–C (hydroxy) or Pd–O (alkoxy) bonds. Here, the hydroxy species were more reactive and formed more readily, a conclusion similar to that drawn for reactions of methanol-derived adsorbates. Figures S26–S28 show the binding modes and adsorption energies of aliphatic and polyconjugated compounds (e.g., quinones, aromatics) to the Pd surface (i.e., Q + * → Q*), showing highly exothermic adsorption. HKH presents the most exothermic adsorption energy (ΔE _ads = −156 kJ mol^–1) of all quinone compounds, while BQ binds slightly weaker (ΔE _ads = −143 kJ mol^–1). Aromatics (e.g., benzene) adsorb with comparable but slightly weaker energies (ΔE _ads = −125 kJ mol^–1). Moreover, these aromatics readily form hydroxylated species (e.g., phenols, quinones) by direct OH-transfer paths (ΔE ^⧧ = 74 kJ mol^–1; Figure S29b), in which the net reaction is highly exothermic (ΔE _rxn = −202 kJ mol^–1; Figure S29a). Thus, DFT calculations support the strong binding of polyconjugated carbonyl and aromatic species to Pd surfaces, which agree with their persistence during extended periods of reaction (Sections and 3.2).

Hydrogenation of one or more of the carbonyl bonds of the surface-bound quinones (Q*) can form the partially hydrogenated (HQ*) or multihydrogenated (e.g., H₂Q*, H_nQ*) quinone species. Hydrogenation steps can occur by direct reactions of the quinone and surface-bound hydrogen (i.e., H* + Q* → HQ* + *) or by PCET steps mediated by water molecules, i.e., H* + H₂O + Q* → HQ* + H₂O + *; Figures S30a,b, S31a,b, and S7. Mechanisms involving direct surface hydrogenation of the quinone occur more readily for HKH (ΔE ^⧧ = 28 kJ mol^–1) than for BQ (ΔE ^⧧ = 39 kJ mol^–1). In comparison, barriers for hydrogenation by water-mediated PCET steps show similar trends (i.e., HKH < BQ) but present significantly lower values for HKH (ΔE ^⧧ = 0 kJ mol^–1) and BQ (ΔE ^⧧ = 27 kJ mol^–1) than for direct hydrogenation mechanisms (Figure ). These findings suggest that rates of quinone hydrogenation (reduction) by paths mediated by water molecules exceed those of direct hydrogenation by H atoms, an implication consistent with measured kinetic isotope effects (Table ). Notably, H₂O₂ and H₂O formation rates in perhydrogenated water exceed those in perdeuterated water, suggesting that PCET processes with kinetically relevant proton transfer prevail over other paths in the presence of HKH. This conclusion agrees with computational studies (Figures S33–S36), which predict weak kinetic isotope effects between H₂O/D₂O (k _H/k _D = 1.02) and H₂/D₂ (k _H/k _D = 1.18) without a mediator. Similarly, computation suggests that HKH can significantly increase the maximum possible kinetic isotope effect (k _H/k _D = 5.74). However, the equilibration of protons and deuterons with the mediators likely attenuates this effect, but it still agrees with the emergence of kinetic isotope effects when adding mediators.

The Pd surface binds H* and O₂* species that can react to form OOH* via direct surface hydrogenation (i.e., H* + O₂* → OOH* + *), PCET mediated via a partially hydrogenated surface-bound quinone (i.e., QH* + O₂* → Q* + OOH*), or by PCET mediated by water molecules (i.e., H* + H₂O + O₂* → OOH* + H₂O + *). Again, the direct hydrogenation of O₂* from surface H* presents the highest barrier for OOH* formation (ΔE ^⧧ = 51 kJ mol^–1; Figure S32a). The reduction of O₂ by the partially hydrogenated quinone-mediated steps shows barriers more than 20 kJ mol^–1 lower (Figures a,b, S30c, and S31c). Here, the barrier for reacting O₂* with partially hydrogenated BQ (ΔE ^⧧ = 0 kJ mol^–1) remains lower than that for reactions with partially hydrogenated HKH (ΔE ^⧧ = 27 kJ mol^–1). The reduction of O₂* by water-mediated PCET steps shows higher barriers (ΔE ^⧧ = 40 kJ mol^–1; Figures c and S32b). Thus, PCET mechanisms that involve quinone-derived surface redox mediators show the most accessible pathways to reduce O₂* on the surface of Pd. Moreover, it is likely that partially hydrogenated quinones are rapidly consumed by these fast O₂ reduction steps (vide infra); hence, multihydrogenated species were not considered further.

Subsequent steps reduce OOH* species to H₂O₂ or H₂O by similar PCET mechanisms (Figure b). However, the quinone-mediated oxygen reduction routes are less exothermic than the water-mediated routes since most of the energy loss occurs during the initial hydrogenation of the mediator (Figure ). Regarding selectivity, the water-mediated PCET steps show similar barriers to forming both H₂O₂ (ΔE _H2O2 = 26 kJ mol^–1) and H₂O (ΔE _H2O = 28 kJ mol^–1) (Figures c and S32c,d) without quinone-derived intermediates. By comparison, the partially hydrogenated HKH-derived mediator offers PCET pathways with a barrier for H₂O₂ formation (ΔE _H2O2 = 26 kJ mol^–1) equal to that of water; however, the barrier for HKH to mediate H₂O formation (ΔE _H2O = 48 kJ mol^–1) significantly exceeds that for water-mediated paths and all other partially hydrogenated quinone species (Figures a,b, S30d,e, and S31d,e). Furthermore, the differences between the barriers to form H₂O versus H₂O₂ with the partially hydrogenated HKH mediator (ΔΔE ^⧧ = ΔE _H2O – ΔE _H2O2 = 22 kJ mol^–1) yield the greatest kinetic preference for selective H₂O₂ formation. These energetics represent a considerable increase over the change in barriers to form H₂O₂ and H₂O via water-mediated PCET steps (ΔΔE ^⧧ = 2 kJ mol^–1).

These calculated values for the differences in the barriers of H₂O₂ and H₂O formation (ΔΔE ^⧧) agree with the trends observed for experimentally measured barriers. Table compares the differences between apparent activation enthalpies to form H₂O₂ and H₂O for HKH- (ΔΔH ^⧧ = ΔH _H2O – ΔH _H2O2 = 34 kJ mol^–1), BQ- (ΔΔH ^⧧ = 27 kJ mol^–1), and water-mediated paths (ΔΔH ^⧧ = 11 kJ mol^–1). However, mathematical treatments of the apparent barriers of reactions reflect an interdependence of H₂ adsorption, quinone reduction, and oxygen reduction steps, which complicate direct quantitative comparisons between the absolute values of experimental (ΔH _H2O2 ) and computed activation barriers (e.g., ΔE _H2O2 ). Regardless, experimental barriers, kinetic isotope measurements, and the interpretation of these values informed by ab initio calculations provide strong evidence that quinones bind to catalyst surfaces and form mediators, which present favorable barriers that enable the greatest rates and selectivities of H₂O₂ formation compared to untreated Pd nanoparticles (Figures and ).

3.5. Influence of Mediator Chemical Structure

First-principles homogeneous-phase DFT calculations were subsequently used to provide insight into how the chemical structure of the mediators relates to their catalytic performance (Figures and S37–S38 and Table S3). In general, electron-donating substituents on BQ decrease the reduction potential of BQ derivatives, thus lowering the thermodynamic favorability of adding an electron to the mediator. Comparisons of one- and two-electron reduction potentials show that HKH possesses the most positive reduction potential of all mediators investigated and indicate that HKH most readily accepts electrons. HKH also shows the smallest energy gap between the lowest unoccupied molecular orbital (LUMO) and the highest occupied molecular orbital (HOMO), which further suggests that HKH is the most electrophilic among these mediators.

9.

Top and side views of the lowest unoccupied molecular orbitals and the resulting distinct surface chemisorption modes for (a) BQ and (b) HKH, in which palladium, oxygen, carbon, and hydrogen are represented by the teal, red, gray, and white spheres, respectively. The orbital colors, green and maroon, show the opposite phases of the wave function in each region of the mediators. The plots below show the partial density of states for physisorbed and chemisorbed BQ and HKH for (5/16) ML O* and (5/16) ML H_s* coverage in an implicit solvent. The total DOS are black, the PDOS for d states of Pd are blue, the p states of O are red, the p states of C are green, and the s states of H are maroon. The shading under each region is proportional to the total number of electrons that can occupy each type of orbital.

The trends among these characteristics align with comparisons of the partial densities of states (PDOS) for physisorbed and chemisorbed mediators bound to Pd in the presence of coadsorbed O* and subsurface H_s* (Figures and S39–S43). Specifically, chemisorption of HKH results in a significant intensification, shift, and delocalization of the molecular contributions of the p-states from C- and O-atoms that overlap with d-band states of Pd near the Fermi level. These changes exceed those that occur when BQ chemisorbs to an equivalent surface (Figures S39–S43). These comparisons show that HKH gives the greatest extent of orbital intermixing between Pd and the organic adsorbate of the mediators examined, in line with its most electrophilic nature established previously, which explains the highly exothermic adsorption of HKH to Pd surfaces (Figure S26).

The electronic properties of the mediators correlate to intrinsic barriers for PCET reactions with coadsorbed surface species. The most electrophilic mediators (i.e., HKH) present the lowest barriers for the heterolytic oxidation of surface H* species and concurrently give the highest barriers for PCET to reduce O₂* and OOH* species compared to less electrophilic species (Figure ). These trends reflect the greater extent of electron withdrawal from the Pd surface, which decreases the length of C–Pd and O–Pd bonds between Pd and mediators, consistent with stronger adsorption energies. These differences in charge transfer between electrophilic mediators and surface Pd atoms lower the extent of electron back-donation from Pd to the 2π* orbitals of O₂*, which shorten O–O bond lengths, increase O–O bond dissociation barriers, and increase H₂O₂ selectivities.

Ab initio calculations also explain the need to operate at H₂-rich conditions to maintain the surface redox mediators on Pd. Increases in the surface coverage of O* from 2/16 ML to 8/16 ML weaken the adsorption energies of all mediator species (Figure S26) by differences that span from 43 to 116 kJ mol^–1. Higher coverages of O* block more surface sites and reduce the number of bonds between Pd and the mediator, diminishing the covalent interaction between the mediator and the surface. The PDOS calculations support this interpretation and show that lower coverages of O* atoms delocalize (i.e., broaden) the molecular contributions of p-states from C- and O-atoms of HKH and BQ that overlap with the d-band states of Pd near the Fermi level (Figures S39 and S42–S43). Thus, low coverages of O* result in greater hybridization among the p orbitals of the mediator with the d orbitals of Pd surface atoms. By comparison, subsurface H_s* has a much weaker effect on the adsorption energies of mediators (Figure S26) and PDOS features of these species (Figures S39–S41). Although O* destabilizes the redox mediators, O* does not affect the density of surface electrons or reduce the extent of electron back-donation to π* 2p orbitals of O₂* as significantly as the mediators. Such findings seem consistent with the lower H₂O₂ selectivities under O₂-rich conditions, even in the presence of surface-bound mediators (Figure S8). Thus, H₂-rich conditions maintain selective PdH _x phases and also preserve the strong binding and electronic structure of bound redox mediators that favor the reduction of O₂ without cleaving O–O bonds.

These electronic structure calculations agree with prior findings, kinetic isotope measurements (Table ), and the requirement of high hydrogen pressures to promote H₂O₂ formation (Figures , S6, and S8). Together, these findings appear most consistent with a mechanism in which mediators adsorb strongly to Pd surfaces, react readily with surface H* species, form O–H moieties, and transfer protons and electrons to O₂* and OOH* species during the formation of H₂O₂. These calculations demonstrate several key points. First, PCET paths universally show lower barriers for transferring hydrogen atoms to the bound oxygenates (i.e., quinones, O₂-derived species) than direct hydrogenation routes. Second, more electrophilic mediators (e.g., HKH) bind most strongly to Pd, react most readily with H* species, and best preserve O–O bonds for the selective formation of H₂O₂. Third, high coverages of O* species block sites and disrupt the selective electronic state of quinone mediators, necessitating H₂-rich conditions to maintain the promoting effects of organic adsorbates.

4. Conclusions

Rate measurements, kinetic isotope effects, and calculated reaction pathways show that quinone and carbonyl species modify Pd surfaces and undergo cocatalytic redox reactions with H₂ and O₂, leading to greater selectivities of H₂O₂ formation (∼65–85%) than on unmodified surfaces (∼45%). Specifically, FTIR spectra, TPO profiles, and DFT calculations show the adsorption of organic molecules onto Pd nanoparticles that persist over extended periods of catalysis (>130 h). These species stabilize O–O bonds under H₂-rich conditions (200 kPa H₂, 60 kPa O₂) but not O₂-rich conditions (60 kPa H₂, 100 kPa O₂), consistent with oxygen disrupting the binding of organic mediators and surface electron density needed for selective H₂O₂ formation under H₂-rich conditions. Moreover, these species lead to the emergence of kinetic isotope effects distinct from those on untreated Pd catalysts, consistent with organic moieties facilitating kinetically relevant proton transfer reactions. Such findings agree with DFT calculations showing that the carbonyl functions of quinones heterolytically oxidize hydrogen atoms that subsequently transfer to oxygen species by PCET mechanisms. Furthermore, transition state calculations and temperature dependence measurements show that these paths present favorable barriers of forming H₂O₂ while obstructing O–O dissociation reactions that generate H₂O. Thus, this approach enables stable and selective H₂O₂ formation using H₂O as the solvent, allowing sustainable H₂O₂ synthesis without organic solvents.

These findings provide a strategy to introduce persistent cocatalytic moieties onto metal nanoparticles that activate H₂ and O₂, which guide the design of catalysts for H₂O₂ synthesis and other redox reactions. This work, combined with prior studies, shows how quinone, carbonyl, and alcohol molecules adsorb to Pd nanoparticles and act as redox mediators. Generally, these species must (1) adsorb to catalytic surfaces more strongly than they dissolve into solution, (2) remain kinetically stable under both reductive and oxidative conditions, (3) withdraw electron density from metal surfaces to stabilize O–O bonds, and (4) introduce low-barrier PCET paths at solid–liquid interfaces. Such species may enable similar redox reactions on transition metal surfaces or alloys. Moreover, these species may facilitate other heterolytic hydrogenation and hydrogenolysis steps, which may reduce other polar heteroatoms, such as C–N, C–O, and N–O bonds in thermal and electrochemical systems. Thus, this work presents opportunities to tailor the active sites of metallic nanoparticles by introducing surface mediators that cocatalyze a broad range of chemical transformations.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.4c15874.

• Additional characterization of materials by transmission electron microscopy, energy dispersive X-ray fluorescence, infrared spectroscopy, and temperature-programmed desorption and oxidation measurements. Details of experimental rate studies, including process flow diagrams, additional controls, time-on-stream measurements, kinetic isotope studies, and Arrhenius plots. Details of theoretical calculations, including modeled surface structures, molecular electronic structures, redox potentials, partial density of states, kinetic isotope effects, and calculations of energies of adsorption, reaction, and transition states (PDF)

• Computational structure files for this work (ZIP)

∥.

Department of Chemical and Biomolecular Engineering, Rice University, 6100 Main St., Houston, Texas 77030, United States

⊥.

Amogy, 19 Morris Ave, Brooklyn, New York 11250, United States.

The manuscript was written with contributions of all authors.

The authors declare no competing financial interest.