Hydrogen peroxide (H2O2) is a green oxidant with diverse applications including disinfection, wastewater treatment, and chemical synthesis. 1,2 The current "anthraquinone process" for chemical production of H2O2 is energy intensive, highly centralized and requires hazardous transportation of concentrated H2O2 solutions. This hazard is unnecessary when considering that many applications use dilute H2O2 solutions, or low concentrations of transient H2O2. 3 The electrosynthesis of H2O2 via the two-electron oxygen reduction reaction (2e -ORR) could be a promising alternative, especially for applications requiring on-demand or continuous production of low concentrations of H2O2, if catalysts with high activity, selectivity and stability for 2e -ORR (versus the four electron oxygen reduction reaction to water, 4e -ORR) can be identified. Even though there are many recent reports on carbon-based and metal compound catalysts that are active in alkaline (or acidic) media, [4][5][6][7][8][9][10][11][12] reports of catalysts selective for 2e -ORR in neutral media, in which H2O2 could be stored and directly utilized without post-treatment, have been more limited, [13][14][15][16][17] especially in buffered neutral media where the activity is not a result of a shift towards alkaline pH during operation. 7 Metal compounds are particularly interesting for this catalytic challenge on account of their natural active site separation, which can serve as a kinetic barrier to the 4e -ORR, and their tunable structures that can enable systematic characterization of the structure-property relationships. 7 A crucial consideration for the practical electrosynthesis of H2O2 is the stability of these emerging metal-containing electrocatalysts. Electrocatalysts for 2e -ORR must be (i) stable against leaching of metals and other non-metal elements (ORR inert sites), (ii) prevent activity degradation and (iii) decomposition of the electrosynthesized H2O2. 4 Preventing leaching is especially important for wastewater treatment to ensure the catalyst does not lead to further contamination.
The Environmental Protection Agency (EPA) limits Se in drinking water to 0.05 ppm as well as limits (or requires concentration monitoring of) metals such as Co and Cu. Metal leaching can also lead to degradation of catalytic activity during continuous operation and accelerate the reduction of accumulated H2O2. 18,19 Compounds containing noble metals (such as Au, Rh, and Pd) promise better stability based on their bulk Pourbaix diagrams and DFT calculations predicted some to have high activity towards 2e -ORR. 20 The higher cost of noble metals compared to earth-abundant metals could be necessary for applications where greater stability is crucial, particularly on-site applications where the produced H2O2 will be used directly without further purification.
Specifically, recent experimental studies on amorphous Pd, 21,22 amorphous Pd-Se nanoparticles, 23 and Pd4Se nanoparticles for 2e -ORR, 24 as well as the theoretical prediction that some Pd-Se compounds are stable and active for 2e -ORR suggest that the separation of Pd atoms by Se atoms likely leads to both high selectivity for 2e -ORR 7 and good stability due to selenium's improved resistance to oxidation (relative to sulfides). 25 Among the 10+ possible Pd-Se compounds and polymorphs, 26,27 the most stable orthorhombic polymorph of PdSe2 has a unique two-dimensional (2D) pentagonal structure where Se2 2-dumbbells form five-membered rings with Pd in puckered layers (Figure 1a). 2D PdSe2 has been increasingly exploited for optoelectronic device applications due to its interesting electronic properties, 28,29 layer-number tunable bandgap, [30][31][32] and optical nonlinearity. 33 Several characteristics of this unique layered PdSe2 structure type also make it promising for 2e -ORR: as the most Se rich compound in the Pd-Se phase diagram, PdSe2 has a large Pd-Pd site separation (>4 Å in and out of plane) that helps to promote the selectivity for 2e -ORR. 2 The dichalcogenide motif is also reminiscent of the dichalcogenide dumbells in the metal pyrite structures (e.g., CoS2, CoSe2, and NiSe2) that have been demonstrated as highly active 2e -ORR catalysts. 4,25,34 The uncommon d 8 square-planar coordinated Pd embedded in a layered material is similar to the 2D metal organic frameworks (and SACs) that are active for 2e -ORR in neutral media. 15,35,36 More broadly, several 2D electrocatalysts were recently identified as promising 2e -ORR catalysts [37][38][39] and these compounds could be exfoliated from the bulk crystals for electrode fabrication. with the (001) basal plane highlighted in pink and shown on the top right (as a monolayer for clarity) and the (100) plane highlighted in light blue and shown on the bottom right. b) Free energy diagrams of the reaction intermediates on PdSe2 (001) and (100) surfaces at the calculated equilibrium potential with free energies of intermediates on NiSe2 shown for comparison. 4 The dashed blue line represents OOH* in the presence of specifically adsorbed water (see Computational Methods for the validation of including a specifically adsorbed water molecule).
Here, we computationally and experimentally study 2D PdSe2 and reveal its high catalytic activity and selectivity for the 2e -ORR and good stability in neutral media. Theoretical calculations predict that both the basal plane and edge facets to be thermodynamically viable surfaces for 2e -ORR, but with different kinetic behaviors. Electrochemical studies of hydrothermally synthesized PdSe2 nanoplates demonstrate that 2D PdSe2 is uniquely active and selective for 2e -ORR in neutral media over other crystalline Pd-Se phases. Effective bulk electrosynthesis of H2O2 in neutral media has been achieved using PdSe2 catalyst supported on a gas diffusion electrode and the stability is confirmed by long-term electrolysis operation, elemental analysis of the electrolyte and various ex-situ and in-situ structural characterizations including synchrotron X-ray absorption spectroscopy (XAS).
Prediction of 2e -ORR Catalytic Behaviors on PdSe 2 . Given the layered structure of PdSe2, we expect that the dominant electrochemically active facets of interest will be the (100)/(200) edges and the (001)/(002) basal planes (Figure 1a). To understand the electrochemical stability of PdSe2, we first calculated the surface Pourbaix diagrams of the PdSe2 (001) and (100) surfaces (Figure S1 and Table S1-S2) by using the computational hydrogen electrode (CHE) method 40 under the assumption that the surfaces are in equilibrium with bulk water. 25 (Equation 1)
Particularly interesting is the fact that the two surfaces show very different trends in the stability of adsorbates. The (001) surface is predicted to be free of water-related reduction products over a wide range of electrode potential. On the other hand, the potential window for a "clean" (100)
surface is predicted to be narrower, although it is predicted to persist under the relevant applied potential. Note that the Pd-Se bond breaking is not required to generate the basal (001) surface with only van der Waals interaction. On the other hand, generation of the (100) surface requires cleavage of Pd-Se bonds, which is expected to make this surface more reactive. Our calculations also suggest that specific adsorption of one water molecule per unit cell is favored on the (100) surface (-0.38 eV).
Next, we calculated the reaction free energy diagrams for 2e -ORR on both surfaces at the calculated equilibrium potential using the CHE method (Figure 1b, Figure S2). Given the substantial differences in the stability of the (001) and bare, water-free, (100) surfaces, it is therefore not surprising that stability of OOH* varies significantly between the two surfaces; at the calculated equilibrium potential of 2e -ORR, the first proton-coupled electron transfer (PCET) step (Equation 2) is uphill by 0.48 eV and downhill by 0.49 eV on (001) and (100) surfaces, respectively, whose magnitude determines the thermodynamic liming potential. Since the (100)
surface features an equilibrium coverage of water molecules, we also found the adsorption energy of OOH* on the (100) surface is destabilized by ~0.3 eV in the presence of this adsorbed water (Fig. 1b dash line), leading to a substantial decrease in the predicted overpotential.
Because the standard reduction potential of 4e -ORR (1.23 V) is higher than that of 2e - ORR (0.69 V), the selectivity to H2O2 is determined by the competition between the kinetics of 2e - and 4e -ORR. To achieve a more comprehensive picture, we therefore considered the following PCET reactions,
along with the following thermal (non-electrochemical) steps,
𝑂𝑂𝑂𝑂𝐻𝐻 * → 𝑂𝑂 * + 𝑂𝑂𝐻𝐻 * (7)
We used "constant-charge" calculations to estimate the barriers to the various PCET processes, reactions (2)- (8). Note that the calculated potentials of zero charge of the 3-layered (001) and 5-layered (100) surfaces are around 0.3 V vs SHE, where all the steps involved in 2e -ORR become downhill. The use of an implicit solvation model and a large supercell makes the shift in work function small throughout the reactions (see Computational Details for the validation of the constant-charge calculations). For the PCET processes, we expect the calculated energy barriers to be lower bounds to the true barriers, since the bulk protons will be further stabilized via additional hydration as compared to our interfacial model. Nonetheless, the relative PCET barriers can be safely compared to estimate the relative selectivity of various PCET processes.
On the (001) facet, reduction of O2 to OOH* via PCET [reaction (2)] is expected to proceed with minimal to no barrier under the applied potential, and thus strongly outcompete thermal dissociation of O2* [reaction (6)]. The resulting OOH* can again be reduced via PCET to H2O2
[reaction (3)], undergo reductive elimination via PCET [reaction (4)] or thermal dissociation to O* and OH* [reaction (7)]. Our results suggest that the PCET barrier for the reduction to form H2O2 is at least 0.3 eV lower than that for reductive elimination, which in turn is smaller than the barriers for thermal dissociation. The same considerations also hold for the (100) surface, suggesting a high selectivity on that surface as well. As such, even in the absence of a detailed microkinetic model, high selectivity to H2O2 should be expected, however with perhaps higher activity on the (100) relative to (001) given the reduced thermodynamic limiting potential.
Beyond the initial formation of H2O2, we also need to consider the possible reduction of the accumulated H2O2 both at the electrode surface and in bulk solution, because it ultimately limits the highest achievable concentrations of H2O2. 7 Here, it is important to note that the adsorption free energy of H2O2(aq) highly depends on the surface, varying from 0.28 eV to -0.12 eV on the (001) and (100) surfaces, respectively. Once readsorbed, the H2O2 can be consumed via both reductive elimination [reaction (5)] or thermal dissociation [reaction (8)] (see these barriers in Table S3 and S4). In contrast to the "inert" (001) surface, both thermal dissociation and reductive elimination of H2O2 appear quite feasible on the (100) surface, and thus could limit the accumulation of H2O2(aq).
Synthesis and Characterization of PdSe 2 Nanoplates. We synthesized PdSe2 nanoplates using a facile hydrothermal reaction of palladium chloride and selenourea in near-stoichiometric ratios (see Materials and Methods for details). The powder X-ray diffraction pattern (PXRD) of the product matched the standard pattern of pentagonal 2D PdSe2 (Figure 2a) without major impurities from other known Pd-Se phases. Scanning electron microscopy (SEM) images show that the PdSe2 product has a rectangular nanoplate-like morphology with typical lateral dimension of 200-2000 nm, with enriched surface area in the (001) basal plane compared to the (100) edges (Figure 2b).
Transmission electron microscopy (TEM) images (Figure 1c, Figure S3) also show the morphology of the nanoplates. Furthermore, the indexed selected area electron diffraction (SAED, Figure 1d) along the [001] zone axis confirms the orientation of the plates lying flat on the TEM grid is perpendicular to the [001] crystallographic direction. This shows that the major exposed facet of these PdSe2 plates lying parallel to the substrate is the (001) basal plane, which also means that the other edges of the rectangularly shaped plates correspond to the (100) edge facets. The Raman spectrum matched with that previously reported 30 for bulk PdSe2 (Figure 2e), confirming that the these nanoplates do not approach few-layer thicknesses. Therefore, these PdSe2 nanoplates should have metallic behavior, rather than the large bandgaps expected for near-monolayer PdSe2, 30 which should be beneficial for electrocatalysis. Energy dispersive spectroscopy (EDS) via SEM showed a slightly less than ideal Se/Pd ratio of 1.6 from mapping results. More careful point-by-point EDS analysis of individual nanoplates (Figure S4) showed an average Se/Pd ratio of 1.7 ± 0.2. From a bulk elemental analysis by inductively coupled plasma optical emission spectroscopy (ICP-OES) measurements of digested samples we found the Se/Pd ratio to be 1.98.
This suggests that the deviation of the EDS analysis is likely due to typical errors intrinsically associated with EDS without standardization and its relatively small sampling size. X-ray photoelectron spectroscopy (XPS) on the as-synthesized PdSe2 powder suggested the surfaces of the nanoplates were more Se rich (Figure S5). Since PdSe2 is the most Se rich phase in the Pd-Se phase diagram, this could be due to the potential residual elemental Se on the surface from the synthesis process. diffraction pattern of PdSe2 compared to the standard pattern of orthorhombic PdSe2 (PDF No. 01-072-1197). b) SEM micrograph of PdSe2 nanoplates. c) TEM micrograph of a PdSe2 plate viewed To elucidate the electronic structure of the probable ORR active sites of PdSe2, we conducted X-ray absorption spectroscopy X-ray absorption near edge spectroscopy (XANES) at the Pd L3-edge (Figure 2f, Figure S6). The position of the white line and the white line intensity for PdSe2 are most similar to that of PdO (Pd 2+ ), falling between the intensities/edge positions of Pd foil (Pd 0 ) and K2PdCl6 (Pd 4+ ), suggesting that the oxidation state of Pd in PdSe2 is close to +2.
Se K-edge XANES (Figure S7d) show similarity in peak shapes to Se2 2-in CoSe2, 25 further confirming that PdSe2 and pyrite phase CoSe2 have similar Se-Se dumbbells. XANES at the Pd L2-edge similarly matched a Pd 2+ oxidation state (Figure S6) for PdSe2. For comparison, we also synthesized nanomaterials of tetragonal Pd4Se and cubic Pd17Se15 without major impurities using similar hydrothermal methods (Figure S8). Pd17Se15 also has a similar oxidation state, while Pd4Se exhibits an oxidation state between Pd(0) and Pd(I) (Figure S6). SEM images of the Pd17Se15 and Pd4Se products (Figure S9) reveal a high surface area morphology suitable for electrocatalysis, although without well-defined faceting.
2e -ORR Activity and Selectivity of PdSe 2 . To evaluate the 2e -ORR activity, selectivity, and stability of PdSe2, we first conducted electrochemical tests using a rotating ring disk electrode (RRDE) in a neutral buffer solution (0.05 M NaPi = 0.025 M Na2HPO4/0.025 M NaH2PO4, pH = 6.8-6.9). A high selectivity for 2e -ORR above 80% can be maintained across a large overpotential window (Figure 3a blue curve) for suitable catalyst loading. If we would like to optimize the 2e - ORR catalytic performance, a high ring current of >1 mA/cm 2 disk was observed at high overpotential (~0.2 V vs. RHE), and the selectivity remained >60% at the peak ring current (Figure 3a cyan curve) for higher effective mass loading. In comparison, RRDE measurements of Pd4Se and Pd17Se15 further illustrate the unique behavior of PdSe2 in the family of crystalline Pd-Se phases: even though all three Pd-Se phases exhibit significant catalytic activity in neutral solutions, only PdSe2 displays high selectivity for 2e -ORR (Figure 3b). In contrast to other metal compound 2e -ORR electrocatalysts, the activity of PdSe2 in buffered neutral conditions and that in acidic conditions are quite similar (Figure S10), and the high selectivity is maintained in both pH conditions at potentials up to 0.05 V vs. RHE. properties of PdSe2 nanoplates with various mass loadings in O2-saturated conditions with 0.05 M NaPi buffer (pH ~ 6.9) and a rotation rate of 1600 rpm, and b) in comparison with other Pd-Se Even though selectivity remained high across many samples, the ring current was found to be highly dependent on the double layer capacitance (Cdl) which is proportional to surface area (Figure S11-S13) and ultimately determined by catalyst mass loading. Although this effect is observed for other metal compound catalysts, 18,34 it is more drastic in 2D PdSe2 possibly due to the additional dependence of edge vs. basal plane exposure. 41 Since RRDE samples depend on the dispersion of the catalyst films as well as the edge/basal plane ratio in the catalyst dispersion, the mass loading, resultant capacitance, and thus the observed activity/selectivity can be highly variable for similar mass loadings.
To evaluate the practical performance of PdSe2 for electrosynthesis of H2O2, we prepared a gas diffusion electrode (GDE) loaded with PdSe2 catalyst (Figure 4a) and applied a constant current in a two-compartment flow cell to rapidly accumulate H2O2 (see Methods and Figure S14 for details). We initially measured CV on the PdSe2/GDE electrode and observed a cathodic current onset of ~0.6 V vs. RHE, and a large increasing current up to almost -90 mA (or -16 mA cm -2 ) at 0 V vs. RHE (Figure 4b). of 0.25 M Na2HPO4/0.25 M NaH2PO4 (initial pH ~ 6.5). a) Schematic of the 3-layer GDE/flowcell assembly with each layer annotated and a representative SEM image of the PdSe2 catalyst dispersed on GDE electrode. b) Cyclic voltammogram of the PdSe2 loaded GDE before the electrosynthesis process, c) potential vs. time curve (continuous line, left-axis) and Faradaic efficiency (dots, right axis) of the PdSe2/GDE during constant current electrosynthesis, and d) the corresponding H2O2 accumulation during the H2O2 electrosynthesis processes. Based on the CV, we performed an extended electrosynthesis test at a fixed current of -30 mA (Figure 4c cyan curves). After an initial onset curve, the potential remained fairly constant at ~0.3 V vs. RHE over a 3-hour electrolysis, and yielded an accumulated H2O2 concentration of 870 ppm (Figure 4d, cyan curve). Rapid accumulation of H2O2 was also achieved over the PdSe2/GDE The drop in Faradaic efficiency over the duration of bulk electrolysis is similar to that observed for cubic NiSe2. 4 To understand this, we comprehensively considered both the production and decomposition of the H2O2 and also computationally compare the expected catalytic properties of 2D pentagonal PdSe2 against those of pyrite phase NiSe2 with a cubic symmetry and identical (100) and (001) surfaces (Figure 1b). NiSe2 is predicted to be more active than the PdSe2 based on the stability of OOH* (Figure 1b). The PdSe2 (100), PdSe2 (001), and NiSe2 surfaces each have large barriers for the thermal bond dissociation of OOH* [reaction (7), Table S3], which is beneficial for 2e -ORR. Since both PdSe2 (100)/(001) and NiSe2 show barrierless PCET steps for 2e -ORR with larger barriers for reaction (4), both should favor 2e -ORR over 4e -ORR. However, the barrier for the further reduction of H2O2 [reaction (5)] over NiSe2 lies between those of PdSe2 (100) and PdSe2 (001) surfaces (Table S4), which can be explained by the trend of the relative stability of OH* to H2O2* (or the Brønsted-Evans-Polanyi (BEP) principle). Therefore, the PdSe2 (001) surface is expected to demonstrate better selectivity and accumulation of H2O2, and the kinetic behavior of NiSe2 and PdSe2 (100) edges may dictate the overall Faradaic efficiency of electrosynthesis over these surfaces.
Notably, the PdSe2/GDE electrosynthesis performance is highly dependent on the loading and morphology of the deposited catalyst film -a PdSe2 film with higher loading on the GDE severely limited the Faradaic efficiency and accumulation of H2O2 during electrolysis (Figure S16). This was most likely due to inhibited O2 transport to the triple phase boundary. Despite this, all GDE flow cell electrosyntheses resulted in significantly more H2O2 accumulation in a much shorter time than a traditional H-cell electrosynthesis (Figure S17), which highlights the importance of cell design in evaluating the performance of 2e -ORR catalysts. Full results of various electrosynthesis runs on the same electrode are summarized in Table S5 and also compared to the reported accumulation of H2O2 over various other 2e -ORR catalysts in neutral media in Table S6. The high H2O2 production rate (7.9 mol gcatalyst -1 h -1 ) at 0 V vs. RHE is notably among the highest to reach practically useful concentrations of H2O2 in a conventional flow cell device.
To confirm the predicted stability of PdSe2, we conducted several post-electrolysis characterization experiments. First, in order to evaluate the viability for direct applications of the produced H2O2 where the concentrations of leached metal and Se must be low (i.e. wastewater treatment, semiconductor cleaning), we used ICP-OES to measure the amount of Pd and Se leached into the electrolytes during GDE electrosynthesis conducted with applied constant currents of -30 mA, and -60 mA, and -85 mA (Figure 5a). The leaching of Pd into the solution from the PdSe2 catalyst was negligible at all measured operating currents, but the leaching rate increased with increasing current density from 0.05 ± 0.05 µg gcat -1
h -1 at -30 mA to 0.5 ± 0.5 µg gcat -1 h -1 at -85 mA. The Se leaching rate at an operating current of -30 mA (12.4 ± 0.5 µg gcat -1 h -1 ) was higher than that of Pd. Additionally, we compare several other metrics of stability across metal chalcogenide catalysts under different operating currents (Table S9). One such stability metric is an adaption of the "stability number" (S-number) previously defined for OER catalysts, 42 which we define as the molar ratios of the produced H2O2 to leached metal or Se. We find that PdSe2 has a substantially higher metal S-number than NiSe2 (~10 5 vs.
~10 3 ), and PdSe2 and NiSe2 have Se S-numbers on the same order of magnitude (~600 to ~900 depending on conditions). chemical and structural characterizations. a) Leaching rates of Pd and Se in electrolytes measured by ICP-OES after electrosynthesis of H2O2 for 3 hours at -30 mA, 15 minutes at -60 mA, and 15 minutes at -85 mA. b) Raman spectra, c) XRD (after washing, peak intensities beyond 2θ = 30° multiplied by 10 for visibility), d) XPS, e) XANES (fluorescence mode signal), and f) EXAFS
(transmission mode signal) of the PdSe2/GDE electrode after several electrosynthesis tests (as described in Figure 4 and Table S5) in comparison to the pristine PdSe2 sample.
We then investigated the chemical stability of the PdSe2 catalyst. Very little change was observed in the post-electrolysis Raman spectrum (Figure 5b, Figure S18), showing that the catalyst surface structure is largely maintained. Given the previously observed dependence of the Raman spectrum on PdSe2 layer number, 43 this also suggests that PdSe2 does not significantly delaminate. The XRD pattern of the PdSe2/GDE after the electrolysis and rinsing with nanopure water closely matched a combination of the XRD patterns of PdSe2 and the bare GDE (Figure 5c).
X-ray photoelectron spectroscopy (XPS) indicated that the valence states of Pd (Figure 5d) remained unchanged, suggesting that the Pd 2+ active sites maintained their surface electronic environment. While there were minor changes to the Se XPS (Figure S19), these changes did not suggest formation of SeO2, which was observed for other metal chalcogenides. 25 Furthermore, the peak shifts of the Pd 3d and Se 3d XPS due to partial oxidization of PdSe2 were not observed.
Similarly, the Pd-O bond 44 was not observed in the extended Raman spectrum (Figure S18) for both pristine PdSe2 powder and the post-electrosynthesis PdSe2/GDE electrode. The XPS measurements revealed the Se:Pd ratio decreased post-electrolysis, but was still in excess of a 2:1 ratio (Table S10), thus precluding the formation of Pd17Se15 triggered by Se leaching. 43 These results confirm that the active site for 2e -ORR is indeed the square planar Pd 2+ motif, rather than the higher valent Pd or Pd-O bonds formed due to oxidation, as observed for other 2e -ORR catalysts. 21 To further investigate the changes to the electronic structure of Se, we conducted ex-situ X-ray absorption spectroscopy (XANES) at the Se K-edge (see Methods). The similar ratio between the main white line peak and the post-white line peak (at ~12660 eV and ~12770 eV) confirms that Se remains anionic in character and precludes the formation of bulk Se oxides.
Namely, the peak near 12770 eV was still clearly present, unlike in Se(0) powder, and cationic Se compounds (such as SeO2) which show a clearly blue shifted white line peak. Additionally, the Se K-edge extended X-ray absorption fine structure (EXAFS) spectroscopy of post-electrosynthesis PdSe2/GDE showed little change in the first two major peaks (Figure 5f). Fitting the EXAFS data (see details in Methods and Table S11 and S12) suggested a minimal decrease in Se-Se coordination (from 1 to 0.8 ± 0.1) but larger decreases in direct Se-Pd coordination (from 2 to 1.0 ± 0.1) and Se-Pd interlayer coordination (from 1 to 0.3 ± 0.1). Similar results were achieved with a simplified fitting model (Table S11). Given the Se-rich samples and the lack of significant Pd leaching, these changes in Se-Pd paths could be indicative of some subtle local restructuring of the Pd sites that does not disrupt the overall layered motif of the PdSe2 catalyst.
To further confirm the operational stability of PdSe2 catalyst on a longer timescale, we fabricated a GDE electrode with a high loading of Nafion to prevent electrolyte flooding through the gas diffusion electrode. The absolute current observed (Figure S20a) was similar to that of other PdSe2/GDE electrodes measured. By tuning the catalyst microenvironment, we could continuously operate the PdSe2/GDE electrode for 48 hours at -75 mA to accumulate more than 600 ppm of H2O2 (Figure S20b). In fact, the device operation was only stopped due to the overload at the OER counter electrode. Even after this much longer operation time, the PdSe2 catalyst retained high S-numbers of 2.07 × 10 4 for Pd and 416 for Se. Such long term operational catalyst stability suggests that the H2O2 concentration buildup is likely the primary mechanism of Se leaching from PdSe2. This high stability would enable repeated use of the PdSe2 catalyst in separate solutions, or in a single pass flow cell configuration. After several tests, this PdSe2/GDE electrode was digested for ICP-OES measurement, which showed a decrease of the Se:Pd ratio from 1.98:1 (as measured in the pristine electrode) to 1.30:1 (Table S10). Taken together with the more surface sensitive XPS/EDS results (Table S10) and the bulk sensitive EXAFS results (Figure 5f), this further suggests differences between the catalytic active surface and the bulk of the PdSe2 catalyst triggered by faster Se loss.
After confirming ex-situ that PdSe2 maintained its bulk structure and valence states after electrolysis, we also sought to probe the dynamic electronic and local structure changes under the operating conditions of 2e -ORR using in-situ XAS (see details in Methods and Figures S7 and S21). Interestingly, the Pd and Se K-edge XANES peak shapes and edge positions were unchanged from open circuit potential (EOC) up to a potential of 0 V vs. RHE (Figure S7c and S7d). This suggests that there is no substantial adsorption of oxygenates on the Se sites. Furthermore, Se Kedge EXAFS exhibited very minimal changes in peak intensities or positions under similar conditions (Figure S7e and S7f), indicating no major changes to the local structure of PdSe2 under operating conditions. These in-situ experiments further confirm that the PdSe2 catalyst is stable with minimal local structure changes under ORR operating conditions and high concentrations of H2O2.
In conclusion, our joint computational and experimental study shows crystalline 2D pentagonal PdSe2 to be an active and stable 2e -ORR electrocatalyst for electrosynthesis of H2O2 in neutral solutions. Computational free energy diagrams reveal both the basal plane and edge of PdSe2 nanoplates to be active and selective towards 2e -ORR but with different surface speciation in solution and H2O2 degradation kinetics. Hydrothermally synthesized nanoplates of PdSe2 can achieve >1 mA/cm 2 disk ring currents in RRDE measurements while maintaining high selectivity at high overpotentials, which is significantly better than other crystalline Pd-Se phases examined (Pd4Se and Pd17Se15). Using gas diffusion electrodes loaded with the PdSe2 catalyst, >800 ppm of H2O2 can be electrochemically synthesized in buffered neutral solution within 3 hours using a constant current procedure, or even more rapidly (15 min) at larger current densities/overpotentials. After bulk electrolysis, analysis of the electrode via XRD, Raman, and XPS confirmed that PdSe2 maintains its structure, and ICP-OES analysis of the operation solutions showed negligible Pd loss and slow Se loss. The PdSe2/GDE electrode can be continuously operated under a constant current in a flow cell for at least 48 hours. Ex-situ and in-situ XANES and EXAFS measurements of the PdSe2/GDE confirm subtle changes around the Se sites and stability over a wide potential range during ORR electrolysis conditions. We anticipate that further tuning of the edge/basal plane ratio by using facet-selective growth or exfoliation could further tune the catalytic properties of PdSe2. This work establishes a stable electrocatalyst with a unique pentagonal structural motif to achieve rapid practical electrosynthesis of H2O2, opening up further studies of a new and unique class of 2D electrocatalysts.
The Vienna Ab initio Simulation Package (VASP) 45,46,47 interfaced with the Atomic Simulation Environment (ASE) 48 was used for energies and geometry optimization for all adsorbates. The core electrons are described by projector augmented wave (PAW) pseudopotentials 49 and the Perdue-Burke-Ernzerhof (PBE) 50 functional was used to treat exchange and correlation. The Tkatchenko-Scheffler(TS) method 51 was utilized to treat dispersion because PBE-TS method is known to give reasonable lattice constants of bulk PdSe2 52 . Note that PBE-D3 method was used for the NiSe2 system. 4 Each electronic self-consistent field (SCF) calculation was converged below 10 -5 eV and the surface adsorbates were allowed to relax until forces became below 0.01 eV/Å.
Solvation effects are described by using the implicit solvation model, VASPsol 53,54 . The effective surface tension is set to zero due to the numerical stability 55,56 , which is validated by the fact that the difference in adsorption energy between the default value (0.525 meV/Å 2 ) and 0 meV/Å 2 is less than 0.01 eV. The Brillouin zone was sampled using 10×10×10, 10×10×1, and 10×8×1 Γcentered Monkhorst-Pack mesh 57 for bulk, (001), and (100) facets calculations, respectively. When larger supercells were used, the corresponding Brillouin zones were sampled. The (001) surface of PdSe2 was modeled as a 4-layered 1×1 unit cell slab with two bottom layers fixed, which corresponds to 8 Pd atoms and 16 Se atoms. The (100) surface was modeled as a 5-layered 1×1 unit cell slab with two bottom layers fixed, which corresponds to 10 Pd atoms and 20 Se atoms.
The choice of the number of layers was validated by the convergence of adsorption energies of O*, OH* and OOH* with respect to the number of layers. For the electrochemical activation barrier calculations, 3-layered PdSe2 (001) slabs and 4-layered (100) slabs were used for efficiency. Since simple implicit solvation is unable to capture the steric and electronic interactions with such adsorbed water, 58 we included also one specifically adsorbed water in our subsequent calculations and confirmed that the presence of the adsorbed water does not change the conclusion although the stability of intermediates could be changed a little bit.
The free energy of H2O(l) was calculated using the experimental free energy difference between H2O(l) and H2O(g). The free energy of O2 was determined by setting it to give the experimental standard reduction potential (1.229 V) of the reaction, 0.5 O2(g) + H2 → H2O(l). The free energies of adsorbates were calculated using G = EDFT + Uthermal + ZPE -TS, where EDFT is the energy obtained from DFT calculations, Uthermal, ZPE and S are the contribution from thermal internal energy, the zero-point energy, and entropy, respectively, which are calculated under the harmonic approximation. The free energies of gas-phase/aqueous H2O2 are calculated by using Gaussian 16 59 with the SMD continuum solvation model 60 . The calculated standard reduction potential of 2e -ORR is 0.83 V, which is slightly higher than the experimental value (0.69 V) and close to other computed values. 25,34 The proton-coupled electron transfer (PCET) reaction of a general adsorbate can be written as:
The electrochemical potentials of a proton and an electron pair were calculated by using the computational hydrogen electrode (CHE) method 40 (Equation 10):
where 𝜇𝜇 𝐻𝐻 + � and 𝜇𝜇 𝑒𝑒 - � are the electrochemical potentials of a proton and an electron, and 𝜇𝜇 𝐻𝐻 2 is the chemical potential of hydrogen gas at the standard state, and URHE refers to the applied potential vs RHE. Transition states for thermal bond dissociation were searched using the nudged elastic band (NEB) 61 method and the dimer method 62 . The Eigen ion (H9O4 + ) was used as a model for a solvated proton for an electrochemical process because the use of the Eigen ion was validated for HER on Pt. 63 Furthermore, because the purpose of this calculation is the comparison of activation energy, we believe it is safe to rely on the static minimum energy path and ignore the solvent fluctuations and entropic effects, which are normally avoided by considering a backward barrier and adding the reaction free energy obtained from the CHE method to avoid the description of a solvated proton at the interface. 64 By using the nudged elastic band (NEB) method 61 and the dimer method 62 , we calculated the "constant-charge" activation energy, not constant-potential activation energy. We believed it is sufficient to consider just the "constant-charge" activation energy to compare competing PCET steps because the competing steps are likely to experience similar electrochemical environments. 65 This is particularly important for the (001) surface because the (001) surface has a non-zero band gap, i.e., the assumption of the simple capacitor model cannot be valid, and furthermore, its potential of zero charge (PZC) or work function, which is important to convert a constant charge barrier into a constant potential barrier, highly depends on the interlayer distance, which is also highly sensitive to the presence of the dispersion correction method. Fortunately, the comparison of the PdSe2 (100), PdSe2 (001), and NiSe2 surfaces with the constant charge barriers can be reasonable by the fact that the calculated work function values of the range between 4.5 to 4.8 eV, which implies that the adsorbates experience similar electrochemical conditions during PCET processes. In addition, we believe the trend remains the same even if the proton donor changes from a solvated proton to water.
Chemicals and materials. Palladium chloride (≥99.9%) was obtained from Sigma-Aldrich.
Selenourea (98+%) was obtained from Acros Organics. Carbon fiber paper (CFP, Toray: 5% wet proofing) was obtained from Fuel Cell Earth and plasma treated and heated at 700 °C in air for 5 minutes prior to use. Sigracet 28 BC gas diffusion electrodes (GDE) were used for all GDE electrosynthesis measurements.
Materials synthesis. The PdSe2 nanoplates were synthesized by a hydrothermal method. Briefly, 1.0 mmol of PdCl2 (0.18 g) and 2.0 mmol selenourea (0.25 g) were dissolved in 9 mL of nanopure water and added to a Teflon lined 20 mL stainless steel autoclave. The autoclave was sealed and heated at 220 °C for 12 hours and allowed to cool in air. The resulting powder was washed with nanopure water and ethanol before being dried under vacuum at 60 °C. Nanoparticles of Pd4Se and Pd17Se15 were synthesized by similar methods with appropriate stoichiometric amounts of PdCl2 and selenourea.
Materials characterization. Powder X-ray diffraction (PXRD) patterns were collected on a Bruker D8 ADVANCE powder X-ray diffractometer using Cu Kα radiation with a 0.6 mm slit.
Scanning electron microscopy (SEM) images were collected on a Zeiss SUPRA 55VP field emission scanning electron microscope at a typical accelerating voltage of 1-3 kV for imaging.
Electron dispersive spectroscopy (EDS) spectra were collected on the same microscope using an accelerating voltage of 22 kV and a Thermo Scientific UltraDry EDS Detector, and the Pd and Se L-lines were used for quantification. X-ray photoelectron spectroscopy (XPS) was performed on a Thermo Scientific K-Alpha XPS system with an Al Kα X-ray source. Raman spectra were collected on a ThermoFisher Scientific DXR3xi Raman imaging microscope using a 532 nm laser with a 10 mW laser power. Transmission electron microscopy (TEM) imaging was performed on a FEI Tecnai TF-30 TEM instrument (300 kV). The PdSe2 nanoplates were dispersed in nanopure water with a concentration of 2.5 mg/mL and dropcast onto lacey carbon supported Cu TEM grids (TEM-LC200CU25, Sigma-Aldrich).
XAS experiments and data analysis. X-ray absorption spectroscopy (XAS) was collected at The cathode was mounted onto a Teflon flow field and contacted using gold mesh or copper foil and the area of the cathode exposed to electrolyte was 5.7 cm 2 . The anolyte flow plate chamber, separated by a Nafion-117 membrane, contained a plasma-treated carbon fiber paper electrode coated with ~25 mg cm -2 of platinized carbon (20 wt% Pt) for shorter term testing (Figure 4) and a GDE coated with ~2 mg cm -2 IrO2 (>99.99%, <20 nm particle size, Fuel Cell Store) was used for longer term testing (Figure S20b). Neutral buffered electrolyte (0.25 M Na2HPO4/0.25 M NaHPO4 = 0.5 M NaPi, pH ~6.5) was continuously cycled through both the cathode and anode chamber at a flow rate of 100 mL/h using a peristaltic pump. The O2 flow rate was fixed at 100 sccm using a mass flow controller. 10 mL of buffered electrolyte was used in the cathode chamber.
Aliquots of electrolyte solution were continuously taken during the electrolysis process by sampling through a septum inserted in the flow path, and then added to solutions of ~0.4 mM Ce(SO4)2 in 0.5 M H2SO4. The resulting change in absorbance of the solutions at 318 nm were then measured shortly thereafter by UV-Vis on a JASCO V-570 UV/Vis/NIR spectrophotometer.
The observed absorbance values were converted to the corresponding concentration of H2O2 via the following equations: In-situ X-ray absorption spectroscopy measurements. In-situ XAS measurements were collected using a custom-built fluorescence detection H-cell (Figure S21) where the working electrode was mounted on the front using an acrylic or aluminum face plate, with the catalyst side facing inward to the electrolyte and the bare side facing back toward the incident X-rays and the fluorescence detector. To contact the working electrode under the faceplate, a small piece of copper foil was used. A leakless Ag/AgCl reference electrode (Innovative Instruments Inc.) was secured in the front chamber with the working electrode and a graphite rod was used as the counter electrode in the back chamber, which was separated from the front chamber by a Nafion-117 membrane to minimize H2O2 crossover and decomposition. 2.5 mL of 0.5 M NaPi electrolyte was added to the front chamber for each measurement, and was kept under a continuous stream of O2 gas throughout the duration of the measurement via a port in the top of the cell. The anode chamber was filled with the same electrolyte. XAS spectra were continuously collected during the electrolysis processes under various electrochemical potentials, with each displayed spectrum representing the average of at least three raw spectra collected under the same conditions. Email: jin@chem.wisc.edu