Electrocatalytic Glycerol Oxidation with Concurrent Hydrogen Evolution Utilizing an Efficient MoOx /Pt Catalyst.

Glycerol electrolysis affords a green and energetically favorable route for the production of value-added chemicals at the anode and H2 production in parallel at the cathode. Here, a facile method for trapping Pt nanoparticles at oxygen vacancies of molybdenum oxide (MoOx ) nanosheets, yielding a high-performance MoOx /Pt composite electrocatalyst for both the glycerol oxidation reaction (GOR) and the hydrogen evolution reaction (HER) in alkaline electrolytes, is reported. Combined electrochemical experiments and theoretical calculations reveal the important role of MoOx nanosheets for the adsorption of glycerol molecules in GOR and the dissociation of water molecules in HER, as well as the strong electronic interaction with Pt. The MoOx /Pt composite thus significantly enhances the specific mass activity of Pt and the kinetics for both reactions. With MoOx /Pt electrodes serving as both cathode and anode, two-electrode glycerol electrolysis is achieved at a cell voltage of 0.70 V to reach a current density of 10 mA cm-2 , which is 0.90 V less than that required for water electrolysis.


Introduction
Glycerol is a by-product in biodiesel manufacturing, where 1 ton of biodiesel production yields roughly 110 kg of crude glycerol (or 100 kg of pure glycerol). 1,2The rapid development of the biodiesel industry during the last decades has resulted in a significant oversupply of glycerol on the world market. 3As a consequence, the glycerol price has dropped drastically.Fortunately, glycerol is a highly functionalized molecule with three hydroxyl groups, which makes it a promising candidate for conversion into more valuable fine chemicals and products. 4Most of the oxygenated species obtained from the glycerol oxidation reaction (GOR) are commercially relevant; 5,6 especially C3 products, such as dihydroxyacetone (DHA), hydroxypyruvic acid (HYDP), glyceraldehyde (GD), glyceric acid (GA), tartronic acid (TA) that are all 100-1000 times more expensive than glycerol. 7However, they have a limited market impact currently as they are often produced through expensive processes. 1Compared with the complex stoichiometric oxidation and enzymatic or microbial fermentation processes, electrochemical valorization of glycerol has attracted more interest because it displays many advantages. 8First, the reaction can be carried out in aqueous media at low temperature and ambient pressure, making the process environmentally friendly.
Second, the selectivity to a given product can be precisely tuned by the formulation of catalysts or regulation of the electrochemical process (e.g.electrode potential, reaction time, or electrolyte pH). 8][10] From this viewpoint, 'glycerol electrolysis' also termed 'electrochemical reforming of glycerol', can thus at a lower cost convert low-value waste (glycerol) into high-value products at the anode with concomitant cathodic H 2 production. 11 is a widely accepted efficient catalyst for both alcohol oxidation and the hydrogen evolution reaction (HER) and is thus a perfect choice for glycerol electrolysis. 9,12It has been widely studied for electrocatalytic GOR at different pH.The oxidation pathway on Pt electrodes is rather independent of the pH. 13However, the current densities of GOR on Pt electrodes in acidic or neutral media are often 10 times lower than at alkaline conditions, mainly due to the sluggish deprotonation step in acidic or neutral media which significantly suppresses the reaction efficiency. 13,14For this reason, electrochemical GOR is often conducted in alkaline media for production of useful oxygenated products.However, Pt electrodes often exhibit lower activity during alkaline HER than in acidic media due to the sluggish water dissociation step. 15us, finding a good balance between the GOR and HER is a big challenge when designing Pt catalysts for glycerol electrolysis.
Here, we report a facile method for trapping and confining Pt nanoparticles (Pt NPs) at oxygen vacancies of molybdenum oxide (MoO x ) nanosheets.The resultant MoO x /Pt composite exhibits good electrocatalytic performance for both the GOR and HER in alkaline electrolyte.The introduction of MoO x nanosheets not only plays a significant role for the adsorption of glycerol in GOR and the dissociation of water in HER but also provides enhanced electronic interaction with Pt NPs, thus substantially promoting the intrinsic activity and thereby the specific mass activity of Pt.As a consequence, the MoO x /Pt anode displays 83 mV lower peak potential and two times higher mass activity than that of pure Pt NPs in catalytic GOR in 1.0 M KOH electrolyte with 0.1 M glycerol, accompanied by up to 50% selectivity to glycerate (the salt form of GA).In addition, the MoO x /Pt cathode also exhibits significantly improved activity and kinetics as compared with the Pt NPs electrode in catalytic HER, along with a Faradaic efficiency of H 2 production close to 100% in 1.0 M KOH electrolyte.Finally, the MoO x /Pt catalyst is integrated into a twoelectrode alkaline electrolyzer for glycerol electrolysis.A current density of 10 mA cm -2 can be obtained at the extremely low cell voltage of 0.70 V, which is 0.90 V less than that required for water electrolysis.
The working principles revealed by the model Pt catalyst for promotion of both GOR and HER can also be used for designing other electrocatalysts to produce high-value products at both cathode and anode.

Results
Preparation and morphology of the typical MoO x /Pt composite.The preparation procedures of the MoO x /Pt composite are schematically depicted in Fig. 1a.Commercial MoO 3 powder was used as precursor and first characterized by powder X-ray diffraction (PXRD) that displays sharp and well-defined peaks fitting well with orthorhombic α-MoO 3 (JCPDS 01-080-3491), see Fig. 1b.The crystal structure of α-MoO 3 has a distinctive layered structure parallel to the (011) plane with each layer comprising two sublayers of edge-sharing [MoO 6 ] distorted octahedra that extend along the [010] direction and connect to each other via corner-sharing along the [001] direction.Thus, it has an organization with strong covalent bonds within the layers and weak van der Waals forces between them. 16The commercial α-MoO 3 precursor exhibits a lateral size range from several micrometers to tens of micrometers with thicknesses from 2 to 5 μm as determined by scanning electron microscopy (SEM) imaging, see Fig. 1c.In order to functionalize the MoO 3 bulk crystals, a novel and facile method is introduced by using sodium borohydride (NaBH 4 ) aqueous solution to partially reduce MoO 3 crystals and create oxygen vacancies at the surface.When introducing oxygen vacancies in the MoO 3 lattice, some Mo 6+ sites are reduced to a lower valence state (Mo 5+ ) which results in the injection of electrons into the conduction band; these additional electrons from such Mo 5+ sites will delocalize within the layers and thus enhance the reactivity of these sites by increasing the density of states. 17In addition, the generated H 2 from the reaction between the NaBH 4 and oxides can also help to overcome the van der Waals forces between layers and thus exfoliate the layered MoO 3 bulk crystal into thinner nanosheets (see Fig. 1d), which are denoted MoO x .The MoO x nanosheets exhibited reduced lateral sizes along with a significantly reduced thickness to about 5 nm, see atomic force microscopy (AFM) profiles in Supplementary Figs.1a-b.High-resolution transmission electron microscopy (HRTEM) imaging of an individual MoO x nanosheet displays a lattice fringe with a d-spacing of 0.35 nm (Supplementary Figs.1c-d), which can be ascribed to the (400) lattice plane of α-MoO 3 .The selected area electron diffraction (SAED) pattern of the MoO x nanosheet confirms the HRTEM result, see Supplementary Fig. 1e, in which the inner halo diffraction ring corresponds to the (400) plane of MoO 3 .The SAED pattern also indicates that the MoO x nanosheet has lost the initial MoO 3 crystalline structure, which is further confirmed by PXRD pattern, see Fig. 1b.Only broad peaks appear in the PXRD pattern, which does not show any sharp peaks related to the initial MoO 3 precursor, suggesting that MoO x has a ''nano-amorphous'' structure.This is possible due to that the presence of oxygen vacancies induces increased local disorder.The prepared MoO x powder is also very easy to disperse in water and yield a clear suspension even without assistance of sonication or magnetic stirring (see Supplementary Fig. 2).In contrast, the MoO 3 precursor has a lower dispersibility in water.This significant difference may be primarily due to the significantly reduced size and thickness of the MoO x nanosheets.Considering that the vacancies at the MoO x nanosheets may also benefit from trapping heterogeneous components, the diluted chloroplatinic acid solution was added drop by drop via a peristaltic pump to make a uniform dispersion of [PtCl 6 ] 2-ions on the MoO x nanosheets; the [PtCl 6 ] 2-ions were finally reduced by NaBH 4 aqueous solution to form the MoO x /Pt composite.The content of Pt in the composite was determined by inductively coupled plasma mass spectrometry (ICP-MS) to be about 10.3 wt%.The TEM image of the MoO x /Pt composite demonstrated uniform dispersion of Pt NPs on the MoO x nanosheets, see Figs. 1e-1f.It should be noted here that the obtained Pt NPs on the MoO x nanosheets showed less aggregation even without the help of a protection agent (e.g.poly(vinyl pyrrolidone) PVP or poly(vinyl alcohol) PVA) upon reduction.This would be attributed to the initial Pt seeds being strongly trapped by the oxygen vacancies of the MoO x nanosheets, which limits the growth process and results in uniform dispersion and size distribution.The average diameter of the Pt NPs is about 3.2 nm with a narrow size distribution, see Fig. 1g.The HRTEM image of Pt NPs on the MoO x nanosheets displays a lattice fringe with a d-spacing of 0.23 nm, corresponding to the Pt (111) facet, see Fig. 1h.This result is consistent with the PXRD pattern (Fig. 1b), which shows distinctive peaks at around 2θ = 39.9 o and 46.precursor, the MoO x nanosheet, and the MoO x /Pt composite were characterized by Raman and X-ray photoelectron spectra (XPS).The Raman spectrum of the bulk α-MoO 3 precursor displays well-defined sharp peaks in the range of 0 to 1000 cm -1 , see Fig. 2a, in accordance with its highly ordered and perfect crystalline structure.The detailed assignment of each peak is listed in Supplementary Table 1.Here, only some important peaks are emphasized.The sharp Raman bands at 996, 819, and 666 cm -1 in the MoO 3 spectrum are ascribed to the stretching vibration modes of the terminal oxygen ν(Mo=O), doubly-connected bridge-oxygen ν(Mo 2 -O), and triply-connected bridge-oxygen ν(Mo 3 -O), respectively. 18,19In contrast, the corresponding bands for the MoO x sample become broad and less sharp.The ν(Mo=O) band in MoO x is red-shifted to 935 cm -1 , indicating a distortion of the terminal MoO 6 octahedra.The ν(Mo 2 -O) and ν(Mo 3 -O) bands are instead blue-shifted to 837 and 685 cm -1 respectively, suggesting that the local structure around Mo atoms has changed because of the presence of oxygen vacancies. 19,20The ν(Mo=O) band has the largest Raman shift (61 cm -1 ), indicating that the reductant (NaBH 4 ) preferably interacts with the terminal oxygen in MoO 6 octahedral and leaves oxygen vacancies in these sites. 16For the MoO x /Pt composite, the ν(Mo=O) and ν(Mo 2 -O) bands are slightly red-and blue-shifted respectively, compared to those for the bare MoO x sample, indicating that part of oxygen vacancies are occupied by Pt and/or there are electronic interactions between the MoO x and the Pt constituents.
In the XPS spectra, the MoO 3 precursor exhibits well-defined peaks at 235.7 and 232.6 eV in the Mo 3d XPS spectra, see Fig. 2b, which are assigned to the 3d 3/2 and 3d 5/2 of Mo cations in high oxidation states (Mo 6+ ). 21The split is 3.1 eV, which is consistent with previous reports. 20,22For the MoO x sample, the related Mo peaks broaden and shift to lower binding energies; the broadening indicates that the MoO x sample has mixed oxidation states of Mo 5+ and Mo 6+ , as also supported by the peak fit shown in Fig. 2b.The fitted peaks located at 234.8 and 231.7 eV are ascribed to Mo 5+ 3d 3/2 and 3d 5/2 , respectively. 235][26] By measuring the peak area, the ratio is calculated to be 4:1 for Mo 5+ /Mo 6+ , thus the molecular formula of MoO x should be defined as MoO 2.6 , in which the average valence state of Mo is +5.2 and with about 13.3% of introduced oxygen vacancies.The O 1s XPS spectra give further evidence for the existence of oxygen vacancies in the MoO x sample, see Fig. 2c.For the MoO 3 sample, the O 1s peak is located at 531.6 eV, which can be assigned as lattice oxygen (O 2-). 27For the MoO x sample, the O 1s peak is significantly shifted to lower binding energy and can be deconvoluted into three peaks.The main peak ascribed to O 2-is shifted to 530.5 eV, indicating a significant change in the coordination of Mo with O. 21 The other two fitted peaks at 531.3 and 532.8 eV can be assigned to surface-adsorbed oxygen species (e.g.OH -, O -) and adsorbed water molecules respectively. 21,28The Mo 3d and O 1s XPS shifts were also computed for models of the MoO 3 and MoO x samples and compared to experiment, see Supplementary Table 2.The calculated Mo 3d and O 1s shifts for the MoO x as compared with initial MoO 3 are -0.95 and -0.87 eV, respectively, which are consistent with the experimentally observed -0.90 and -1.10 eV (Figs.2b-c).The experimental XPS shifts of the MoO x /Pt composite are compared with the bare MoO x and Pt NPs constituents, see Figs. 2d-e.The MoO x /Pt composite displays a further negative binding energy shift in the Mo 3d XPS spectra compared with the bare MoO x (Fig. 2d), indicating an increase in the electron density around Mo atoms in the MoO x /Pt composite.However, compared to bare Pt NPs, it shows a positive shift of binding energy in the Pt 0 4f XPS spectrum, indicating loss of electron density around Pt in the MoO x /Pt composite.These combined XPS results suggest that there is electron transfer from Pt to MoO x in the composite, which would induce strong electronic interaction at the interface and thereby making the interfacial region more active for adsorption of reactants and/or intermediates. 29 further clarify the interaction between MoO x and Pt in the composite, density functional theory (DFT) calculations were performed.The optimized structures of Pt, MoO 3 , and MoO x in DFT models are displayed in Supplementary Fig. 4. The free energy of adsorption of a single Pt atom on MoO x was calculated to determine the preferable combination of Pt with MoO x .Two possible cases were investigated: adsorption on two oxygen atoms (MoO 3 -Pt) or at a vacant site (MoO x /Pt), see atomic structures in Supplementary Fig. 5.The calculated free energies for these two cases are plotted in Fig. 2f, which shows that for the Pt atom it is more energetically favorable to adsorb at a vacant site (ΔG ad = -0.14eV) than on two oxygen atoms (ΔG ad = 0.21 eV), verifying the hypothesis that Pt NPs are likely to nucleate at the oxygen vacancies of MoO x , thus resulting in the observed narrow size distribution of Pt NPs on the MoO x nanosheets.Based on the fact that Pt NPs are more likely to start growing at the introduced oxygen vacancies, a new atomic model was constructed in order to model a realistic situation, where we can simulate the top and edge atoms of a Pt NP supported on MoO x .The optimized DFT model for the MoO x /Pt composite is displayed in Supplementary Fig. 6, for which the charge difference isosurfaces at the MoO x and Pt interface were calculated, see Fig. 2g.Substantial charge is accumulated at the MoO x and Pt interface, especially between Mo and Pt atoms, indicating bond formation and confirming the XPS results that there is electron transfer between these two constituents.Such electronic interactions at the interface would improve the interfacial activity and thereby promote the catalytic performance of the composite.Pt for hydrogen desorption/absorption are observed in the potential region of 0.1 to 0.5 V (vs.RHE). 15The oxidation of the Pt surface starts from around 0.70 V, and the reduction of Pt oxides occurs from 0.90 V in the backward scan.The performance for electrochemical GOR over the different electrodes was evaluated in terms of current density normalized by mass loading of Pt, see Fig.After the peak potential is reached, the Pt surfaces on the MoO x /Pt composite electrode are gradually oxidized (see Fig. 3a), thereby the current for GOR is sharply reduced (see Fig. 3b).Subsequently, a broad oxidation peak appears at around 1.06 V, see Fig. 3b, which is attributed to further oxidation of the products adsorbed on the electrode surface.In the negative scan, the Pt oxides formed in the positive scan are gradually reduced, leading to the formation of free Pt sites that are again available for reacting with glycerol molecules. 13The reactivation of the Pt surface starts at around 0.80 V and the peak current occurs at 0.63 V for the MoO x /Pt electrode.

Glycerol oxidation over the
The CV profiles of GOR at different scan rates over the MoO x /Pt and Pt NPs electrodes are displayed in Fig. 3c and Supplementary Fig. 8.By increasing the scan rates from 10 to 100 mV s -1 , the forward peak currents increase gradually.A linear relationship between the forward peak current (J f ) and the square root of the scan rate (ν 1/2 ) is clearly seen in Fig. 3d, suggesting that the oxidation of glycerol on the MoO x /Pt and the Pt NPs electrodes is governed by a diffusion-controlled process. 30The much steeper slope for the MoO x /Pt electrode (87.4) over the pure Pt NPs electrode (34.0) suggests enhanced electro-oxidation kinetics for the MoO x /Pt electrode in catalytic GOR.
The GOR involves complicated oxidation pathways, which would produce C3 products via direct oxidation or C2/C1 products through C-C cleavage. 31The quantitative analysis of the products from GOR over the different electrodes was performed by high-performance liquid chromatography (HPLC).A typical HPLC spectrum recorded from the product solution of the MoO x /Pt electrode is shown in Supplementary Fig. 9.The relationship between the concentration of products and reaction time over the MoO x /Pt electrode is plotted in Fig. 3e.All the product concentrations increase with increasing reaction time from 1 to 4 h at a constant potential of 0.85 V. Notably, glycerate, glycolate, and formate are the main C3, C2, and C1 products respectively.Here, the oxygenated products from GOR in alkaline media are in the form of salts.
The influence of electrode potential on the product distributions was also investigated, see Fig. 3f NPs electrodes have the highest selectivity to glycerate at 0.85 V, indicating that the observed high peak current at around 0.80 V from CV profiles should mostly come from the oxidation of glyceraldehyde to glycerate. 14When further increasing the potential to 1.25 V, the selectivity to glycerate on the MoO x /Pt electrode is decreased, while the selectivity to all C3 products is increased.This can reasonably explain the broad shoulder in the CV profiles at around 1.10 V (see Fig. 3b), which is ascribed to the further oxidation of adsorbed products.
Based on the products observed by HPLC analysis and the CV profiles in the standard product solutions (see Supplementary Fig. 10), the proposed reaction pathway of GOR on the MoO x /Pt electrode is illustrated in Fig. 3g.First, a glycerol molecule is adsorbed on the catalyst surface, it is deprotonated and oxidized to GD or DHA with coordination of catalyst and base. 32,33These two intermediates undergo a reversible interconversion in the base, 34,35 they can further go through alkali-catalyzed dehydration to form pyruvaldehyde or 2-hydropropenal, 14 which can subsequently interconvert into lactate via Cannizzaro rearrangement. 36,37The lactate can further transform into acetate or formate via C-C cleavage. 36From the HPLC results showing that the main product is glycerate, we note that this pathway should be a minor contribution.Most of the products originate from the pathway involving oxidation of GD, in which it is first catalytically oxidized to form glycerate, and then glycerate can be further oxidized to tartronate or undergo C-C cleavage to evolve glycolate and formate. 14,36The detected oxalate from the HPLC analysis would originate from the further oxidation of glycolate. 35The glycolate can also go through C-C breaking resulting in formate.The total carbon balance from all C3, C2, and C1 products is calculated from ten experiments and measured to be in the range 95% to 98%, the missing 2% to 5% carbon is most likely due to adsorption of reactants or products on the electrode surface.
To further investigate the origin of the superior performance of the MoO x /Pt electrode over the pure Pt NPs electrode in catalytic GOR, DFT calculations on the binding energy with reactants based on the different catalyst surfaces were performed, see Fig. 4a.The initial MoO 3 surface presents unfavorable interaction with all the reactants (OH, H 2 O, glycerol), especially for OH species, and they are not likely to form on the MoO 3 surface.However, the MoO x surface displays substantially stronger binding with all the reactants compared with the MoO 3 , indicating a significantly improved surface affinity and reactivity of the MoO x in the presence of oxygen vacancies.Particularly, the interaction energy of glycerol with MoO x is surprisingly more negative (ΔE BE = -0.89eV) than that of the pure Pt surface (ΔE BE = -0.59eV), suggesting a strong ability of MoO x to adsorb glycerol molecules.This result further confirms the hypothesis that MoO x plays an important role in helping to attract glycerol molecules close to Pt NPs during GOR.As expected, with the assistance of the MoO x substrate, the MoO x /Pt composite also interacts more strongly The free energy diagrams for both the MoO x /Pt composite and the pure Pt surfaces on the DHA and GD pathways were calculated for comparison, see Fig. 4b.For the first deprotonation step, the Pt surface prefers to go to the GD pathway, while the MoO x /Pt composite has a comparable tendency to both GD and DHA.On the second deprotonation step, both the GD and DHA pathways become sluggish on the pure Pt surface; however, they are more energetically favorable on the MoO x /Pt composite surface.Such observations verify the higher activity of the MoO x /Pt composite than the pure Pt for catalytic GOR.For the following steps, we do not make further calculations, because the final products would be more influenced by experimental factors, i.e., applied potential, reaction time, electrolyte pH, etc.The configurations of the optimized initial states for four cases are shown in Fig. 4c.For glycerol adsorbed on Pt, the shortest oxygen-to-surface distance is 3.1 and 3.5 Å for the GD and DHA pathways, respectively.
However, the optimized structure at the MoO x /Pt composite surface has glycerol at the edge of the interface, which we attribute to that the electron transfer at the interface makes the interfacial edge region more active for adsorption of reactants. 29The bond distance between glycerol and the MoO x /Pt surface is 2.0 and 2.5 Å for the GD and DHA pathways, respectively, which is much smaller than the corresponding distance on the pure Pt surface.These observations are consistent with the results shown in Figs.4a-b, and also verify the superiority of the MoO x /Pt composite for adsorption of glycerol.In addition, the local density of states (LDOS) based on the MoO x /Pt composite surface was calculated, see Fig. 4d.The Fermi level (0 eV) is mainly populated by Pt states, indicating that the Pt constituent plays the main role in catalytic reactions, which is consistent with the electrochemical results shown in Figs.3a-b.The MoO x states mainly dominate above 0.5 eV, suggesting it would be a good Lewis acid and may contribute to interacting with reactants (e.g.OH -, glycerol), in agreement with the calculated strong binding interaction between MoO x and all the reactants (Fig. 4a).
Based on the experimental results and DFT calculations, the GOR mechanism on the MoO x /Pt composite catalyst is concluded and schematically illustrated in Fig. 4e.First, the MoO x nanosheets help to attract glycerol molecules because of their abundant oxygen vacancies and strong ability for adsorption of reactants.Then the glycerol molecules react with the nearby Pt clusters.Due to the electronic interaction at the interface, the interfacial region becomes more reactive.Glycerol molecules adsorbed on the MoO x /Pt surface will first undergo two subsequent deprotonation steps and convert to glyceraldehyde at lower oxidation potential.Successively, the generated glyceraldehyde will further deprotonate with the addition of a water molecule and then mainly oxidize to glycerate.

Hydrogen evolution and combined two-electrode glycerol electrolysis.
Pt is an efficient catalyst for HER because of its high ability for adsorption of hydrogen intermediates (H ad ), 12 however, it is generally inefficient in the first step of water dissociation in alkaline media, which results in a lower catalytic activity and thereby a lower reaction efficiency for alkaline HER. 38,39Considering that MoO x nanosheets have abundant active edge sites and oxygen vacancies, and have been verified theoretically to have an energetically favorable adsorption binding energy with water and hydroxyl, it is reasonable to assume that they would also enhance the performance of Pt for catalytic alkaline HER.The HER performance of the Co(OH) 2 ) for Pt in alkaline HER, 38,42,43 that can help to dissociate water because of its abundant active edge sites and oxygen vacancies, as well as the theoretically verified strong adsorption ability for water.Second, the electronic interaction between MoO x and Pt constituents would activate the interfacial Pt clusters to become more active for adsorption of H ad , 44,45 thus significantly increasing the specific mass activity of Pt  production of H 2 .In addition to lowering the required cell voltage and energy cost, another advantage from glycerol electrolysis over water electrolysis is the transformation of low-value waste (glycerol) from biodiesel to high-value products (e.g.glycerate) which are desirable for the industrial market.What should be noted here is that the high selectivity (50%) of the MoO x /Pt electrode to glycerate (or GA) makes it very promising and interesting in industrial applications, because the glycerate (or GA) not only has a high price ratio of about 1000 compared with glycerol, but also has significant and wide applications in medicine. 3It is an important metabolite in the glycolysis cycle and an intermediate in the synthesis of amino acids, and can also be used for the treatment of skin disorders. 1,46Based on the above-mentioned results, the schematic diagram of the two-electrode alkaline electrolyzer for glycerol electrolysis on the MoO x /Pt electrode is illustrated in Fig. 5f.In the anode compartment, the electrode surface donates electrons and gradually oxidizes glycerol to glycerate, mainly following the reaction formula: CH 2 OH-CHOH-CH 2 OH + 5OH -→ CH 2 OH-CHOH-COO -+ 4H 2 O + 4e -.In the cathode compartment, the electrode surface instead accepts electrons coming from the external circuit to reduce H 2 O into H 2 gas by following the reaction formula: The OH -ions will cross over the membrane from the cathode side, where water is reduced, towards the anode side, where glycerol is oxidized.Finally, the overall reaction formula is

Discussion
In Then copious amounts of water were added to dilute the above system and it was left stirring for 1 h.
Successively, the product was washed several times by water and acetone via a centrifugation process.The collected MoO x was subsequently dried in vacuum overnight at 60 o C. It was first cut into electrode shape with 1 × 1 cm 2 effective surface area and then washed with 3 M HCl by sonication for 15 min, followed by washing with water and drying naturally.Finally, 50 μL of catalyst ink was dropped on the CFP electrode for common electrochemical tests, and dried in a fume hood, making the final catalyst loading to be 0.25 mg cm -2 .200 μL of catalyst ink was dropped on the CFP electrode for the tests in HPLC, faradaic efficiency, and two-electrode glycerol electrolysis, making the mass loading of the catalyst to be 1 mg cm -2 .

Synthesis of
Electrochemical measurements.GOR tests were performed in a three-electrode system by using an SP-  HER tests were performed in a similar three-electrode setup with GOR, except for that the electrolyte was changed to 1.0 M KOH aqueous solution.To minimize the capacitive current, the scan rate for the linear sweep voltammetry (LSV) curve is 1 mV s -1 .The overpotential (η) of HER was calculated by using the equation: η = 0 -E RHE .The Tafel plots were obtained by transforming the LSV curve into log J vs. E curve, in which the current density (J) was corrected for back reaction current followed by Conway's work. 40The stability tests for HER were conducted at a constant potential without iR compensation.The Faradaic efficiency (FE) of HER was calculated by measuring the ratio between practically generated H 2 with theoretically generated H 2 .The practically generated H 2 is measured by the gas displacement in a sealed divided cell and finally converted to the amount by the ideal gas law.The theoretically generated H 2 was calculated by integrating the passed charge during the reaction.Glycerol electrolysis was conducted in a divided two-electrode system with the MoO x /Pt electrode serving as both cathode and anode.The cathode chamber was filled with 1.0 M KOH electrolyte, the anode chamber was filled with 1.0 M KOH plus 0.1 M glycerol.The two chambers are separated by an anion exchange membrane.
Computational methods.Electronic structure calculations were performed using density functional theory.The exchange-correlation functional was approximated using the Bayesian Error Estimation Functional (BEEF-vdW) 47 which has been previously shown to describe methanol oxidation on the molybdenum trioxide (MoO 3 ) surface. 48We have performed calculations based on the plane wave method and all calculations were performed using the Vienna Ab initio Simulation Package (VASP). 49The ionic cores were described by projector-augmented wave (PAW) potentials, 50,51 whereas the valence electrons were treated by a plane-wave basis set with 450 eV cut-off.For all structures, the Brillouin-zone sampling was performed using a Γ-centered Monkhorst-Pack 3×3×1 k-point mesh.The Methfessel-Paxton 52 approach with a Gaussian width of 0.2 eV was employed and all the energies were extrapolated to T = 0 K.
A vacuum space of 25 Å was included in all calculations.Dipole corrections were used in the out-of-plane (z) direction in order to avoid unphysical electrostatic interactions between periodic images.Spin-polarized calculations were carried out, and the most stable spin state for all systems was reported in this work.For all atomic structures, structural relaxations were first conducted with the criteria for energy and atom force convergence set to 10 −5 eV and 0.01 eV Å −1 , respectively.
All the surface atomic structures were simulated by slab models constructed in the Atomic Simulator Environment (ASE). 53Pure platinum surfaces were modeled using a (4 × 4 × 4)-fcc(111) slab at the optimized lattice constant of 3.99 Å (the experimental bulk lattice constant is equal to 3.92 Å).For MoO 3 an orthorhombic unit cell with the Pbmn space group was used, and the slab was built in the [100] direction.
The calculated cell parameters were found to be: {a = 14.BEEF-vdW thus overestimates the lattice constant by less than 2.4 %, which we believe will not significantly affect the surface energetics.In order to simulate the MoO x nanosheet, one oxygen atom was removed on the surface, resulting in 11 % vacancy on the surface.For both MoO 3 and MoO x , the structures were built in two layers interacting through weak van der Waals forces.For the slab calculations, the bottom atoms were kept fixed.For the pure platinum slab, the bottom two atomic layers were kept fixed.
The MoO x /Pt composite models were designed as the interface between MoO where  +  ,    , and    are the obtained energies for the slab system containing the adsorbate, the energy of the slab and the energy of the adsorbate in vacuum, respectively.Here "ads" represent chemisorbed OH, H 2 O, and glycerol species.Δn is the difference in terms of hydrogen atoms (particularly used for OH species) and   2  is the energy of the hydrogen molecule.
Chemical potential values for glycerol and Pt are 0.78 and 1.20 eV, respectively.
We note that for Pt chemical deposition calculations, only one Pt atom was adsorbed onto the MoO 3 and MoO x nanosheets, because we are only trying to analyze the stability of the platinum in the presence/absence of the vacancy at the surface.As has been discussed in the main text, the vacancy results in the chemisorption of Pt atom being favorable at the surface.In order to avoid using the Pt ions as a reference in free energy calculations, the DFT calculations were evaluated using the Pt fcc bulk structure.
∆  + + − in Eq 3 represents the chemical potential of protons and electrons and was described based on the computational hydrogen electrode method. 55For every proton-electron transfer step, it is given as where  RHE is the electrode potential, e is the elementary charge, and  H 2 ° is the chemical potential of H 2,(g) at 1 bar and 25°C.
O 1s and Mo 3d XPS peak shifts were calculated based on the MoO 3 and MoO x .The absolute energy position of the spectra was determined with ∆ (Kohn-Sham) calculations, and the core-level binding energies were calculated in the final state approximation as implemented in VASP. 56The onset of the computed spectra depends on the exchange-correlation functional while relative peak-positions have been shown to be independent of the selected functional. 57

Fig. 1
Fig. 1 Preparation and morphology characterization.a Schematic illustration of the crystal structure of MoO 3 and preparation procedures of the MoO x /Pt composite.b PXRD patterns of MoO 3 , MoO x , and MoO x /Pt powders.c SEM image of the commercial MoO 3 powder.d TEM image of the MoO x nanosheet.e-f TEM images of the MoO x /Pt composite.g Histograms of the size distribution of Pt NPs on the MoO x nanosheets recorded from panel f.h HRTEM image showing the lattice fringes of Pt in the MoO x /Pt composite.i High-angle annular dark-field (HAADF) STEM image and corresponding EDS mapping of the MoO x /Pt composite.

Fig. 2
Fig. 2 Structural characterizations.a Raman spectra recorded from MoO 3 , MoO x , and MoO x /Pt powders.b Mo 3d and c O 1s XPS spectra of MoO 3 and MoO x .d Mo 3d XPS spectra of MoO x and MoO x /Pt.e Pt 4f XPS spectra of MoO x /Pt and Pt NPs.f Calculated adsorption free energy of Pt on MoO 3 and MoO x .g Calculated charge difference isosurfaces at the MoO x and Pt interface; the isovalue is 0.0017 electrons/(bohr) 3 ; the cyan and yellow isosurfaces represent electron depletion and accumulation, respectively.For the DFT models in b, f, and g, purple, red, and gray balls depict Mo, O, and Pt atoms, respectively.
MoO x /Pt electrode.The electrocatalytic performance of GOR over the MoO x /Pt composite was studied in 1.0 M KOH electrolyte both in absence and in presence of 0.1 M glycerol and tested in a standard three-electrode setup.Comparison was made with a pure Pt NPs catalyst.All the catalyst powders were first dispersed in a solvent to make catalyst inks and then dropped on carbon fiber paper (CFP, 1 × 1 cm 2 effective surface area) serving as working electrodes.The cyclic voltammogram (CV) profiles in 1.0 M KOH over different electrodes are presented in Fig. 3a.The characteristic peaks of

Fig. 3
Fig. 3 Electrocatalytic glycerol oxidation.a CV profiles of Pt NPs, MoO x , and MoO x /Pt electrodes in 1.0 M KOH electrolyte.b iR-corrected CV profiles of Pt NPs and MoO x /Pt electrodes for glycerol oxidation in the electrolyte of 1.0 M KOH with 0.1 M glycerol.c iR-corrected CV profiles of the MoO x /Pt electrode for glycerol oxidation at different scan rates.d Relationship between forward peak current and the square root of scan rate at the Pt NPs and MoO x /Pt electrodes.e Relationship between the concentration of products and reaction time at the MoO x /Pt electrode.f Comparison of products from glycerol oxidation on the MoO x /Pt and Pt NPs electrodes at different operation potentials based on HPLC analysis.g) Proposed GOR pathway on the MoO x /Pt composite electrode.
3b.In the forward scan, the oxidation of glycerol over the MoO x /Pt electrode begins at approximately 0.50 V.The peak current for the MoO x /Pt electrode occurs at 0.78 V; this potential is 83 mV lower than that for the pure Pt NPs electrode, indicating preferential oxidation of glycerol on the MoO x /Pt electrode.The maximum specific mass activity of the MoO x /Pt electrode is 1056 mA mg -1 Pt , about 2 times that for the pure Pt NPs electrode (541 mA mg -1 Pt ).A bare MoO x electrode was also tested for comparison, its contribution to the oxidation current is negligible compared with the Pt-based catalysts, see Supplementary Fig. 7.However, the role of MoO x in the MoO x /Pt composite and in catalytic GOR cannot be ignored.First, during preparation, it works as an active substrate to trap Pt seeds and confine the size of Pt NPs upon reduction.Second, it helps to concentrate and attract glycerol molecules close to Pt active sites because of the abundant edges and oxygen vacancies, which would contribute to reducing the onset potential of the MoO x /Pt composite for catalytic GOR.Third, the electronic interaction between MoO x and Pt constituents in the composite would also improve the interfacial activity of the composite.The positive role of MoO x from different aspects finally promotes efficient utilization of Pt atoms and thereby the high specific mass activity of Pt in the composite.

Fig. 4
Fig. 4 DFT calculations.a) Calculated binding energy of OH, H 2 O, and glycerol with different catalyst surfaces.b) Calculated free energy diagram for the GD and DHA pathways on the Pt(111) and MoO x (100)/Pt(111) surfaces.c) Calculated configurations of the bonding between glycerol and different catalyst surfaces.d) Local density of states (LDOS) of the MoO x /Pt composite.e) Proposed GOR mechanism on the MoO x /Pt composite catalyst.
MoO x /Pt electrode was tested in a standard three-electrode setup with 1.0 M KOH as electrolyte.The comparison was made with pure Pt NPs and commercial PtC (10 wt% of Pt on activated carbon) electrodes and the current density was normalized by mass loading of Pt, see Fig.5a.As expected, the linear sweep voltammetry (LSV) profile of the MoO x /Pt electrode exhibits superior activity for catalytic HER with a small overpotential of 63 mV to reach a high current density of 1000 mA mg-1   Pt .The overpotential is significantly smaller than that required for pure Pt NPs (η = ~350 mV) or commercial PtC (η = 97 mV) to reach the same mass activity.In addition, Tafel plots are analyzed to compare the kinetics, see Fig.5b.The three electrodes all exhibit linear region at lower overpotential and with curvature beyond 100 mV.The plot shape is similar to that of the anodically activated Pt electrode reported by Conway,40 indicating the HER kinetics on these electrodes all follow the Volmer to Heyrovsky process.41The MoO x /Pt electrode displays a small Tafel slope of 54 mV dec -1 , which is lower than that for the pure Pt NPs (65 mV dec -1 ) or the commercial PtC (64 mV dec -1 ) electrode, suggesting improved HER kinetics on the MoO x /Pt composite electrode.A control MoO x electrode was also prepared and tested for comparison.It exhibited negligible current density compared with the Pt-based catalysts, see Supplementary Fig. 11, indicating its poor activity for catalytic HER.However, the significantly improved activity and kinetics of the MoO x /Pt composite electrode compared to the pure Pt NPs and commercial PtC electrodes suggest the special role of MoO x in catalytic alkaline HER.First, the MoO x may have a function similar to metal hydroxide (e.g.Ni(OH) 2 ,

Fig. 5
Fig. 5 Hydrogen evolution and glycerol electrolysis.a iR-corrected LSV curves, b Tafel plots of the MoO x /Pt, Pt NPs, and PtC electrodes for catalytic HER in 1.0 M KOH electrolyte.c Measured and calculated amount of H 2 and Faradaic efficiency of the MoO x /Pt electrode for catalytic HER.d Schematic illustration of the reaction mechanism of alkaline HER on the MoO x /Pt composite catalyst.e Comparison of two-electrode alkaline electrolyzer for glycerol electrolysis and water electrolysis.f Schematic diagram of two-electrode glycerol electrolysis based on the MoO x /Pt electrode serving as both cathode and anode.

5e.
It requires a cell voltage of 1.60 V to reach a current density of 10 mA cm -2 for water electrolysis on the MoO x /Pt electrode.However, it only takes a cell voltage of 0.70 V for glycerol electrolysis to reach the same current density, and thus about 0.90 V voltage (mainly in anode potential) is saved by using glycerol electrolysis to replace water electrolysis.The dramatic reduction in energy consumption indicates the potential application of substituting water oxidation with glycerol oxidation in clean industrial mass summary, an efficient catalyst based on a MoO x /Pt composite was demonstrated by trapping and confining Pt NPs at the oxygen vacancies of MoO x nanosheets, resulting in high activity and kinetics for both catalytic GOR and HER.The combined experimental results and DFT calculations revealed the special role of MoO x in these two reactions: (i) It helps to adsorb the glycerol molecules close to the nearby Pt NPs and thus substantially reduces the overpotential in catalytic GOR.(ii) It also promotes the sluggish prior step of water dissociation and thereby improves the reaction activity and kinetics in catalytic HER.(iii) The electronic interaction between MoO x and Pt constituents also significantly improves the interfacial reactivity and the intrinsic specific mass activity of Pt for both reactions.Due to the superior activities of the MoO x /Pt composite for both GOR and HER, this enables the two-electrode glycerol electrolysis to take place at 0.90 V lower cell voltage than needed in water electrolysis to reach the current density of 10 mA cm -2 .This work underlines the advantages of glycerol electrolysis compared to classic water electrolysis for the generation of value-added products at the anode with concurrent H 2 production at the cathode.The working principles revealed from the model Pt catalysts in our work can also apply to design other electrocatalysts for the same purpose.MethodsSynthesis of MoO x nanosheets.In a typical process, NaBH 4 aqueous solution (0.1 g in 10 mL H 2 O) was drop by drop added to 1 g MoO 3 commercial powder with vigorously magnetic stirring in an open bottle.
MoO x /Pt composite.In a typical process, 0.5 g MoO x powder was dispersed in 100 mL ultrapure water under magnetic stirring.In a seperate bottle, 4 mL concentrated H 2 PtCl 6 •xH 2 O aqueous solution (10 mg/mL) was diluted to 100 mL and then fed into the MoO x suspension by a peristaltic pump with a feed rate of 10 mL/h.The mixture was kept stirring overnight to ensure sufficient anchoring of [PtCl 6 ] 2-ions on the oxygen vacancies of MoO x .Finally, [PtCl 6 ] 2-ions were reduced by addition of 0.1 M NaBH 4 aqueous solution (the total amount is 5 times the amount of Pt).The resultant composite was then washed several times by water and acetone via a centrifugation process.Afterward, the collected composite was dried in vacuum at 60 o C for 6 h.Synthesis of Pt NPs: For comparison, the pure Pt NPs were synthesized by a similar chemical reduction process in a solution of 10 mM H 2 PtCl 6 •xH 2 O (10 mL) + 1 wt% polyvinylpyrrolidone (PVP, Mw. 10 000) + 0.1 M NaBH 4 (10 mL) for 30 min, in which PVP was introduced as the protection agent to avoid aggregation of Pt NPs upon reduction.Then the sample was washed by water, ethanol, and acetone several times via centrifugation to remove the PVP surfactant and residual salts, and finally the powder was dried in vacuum at 60 o C for 6 h.Synthesis of catalyst modified electrode.First, for preparing catalyst ink, 5 mg catalyst powders (e.g.MoO x , Pt, MoO x /Pt, or PtC) were dispersed in 720 μL ultrapure water, 240 μL isopropanol, and 40 μL Nafion (5 wt%) solution.The mixture was then sonicated for 10 min and subsequently stirred for several hours to form a homogenous ink.A carbon fiber paper (CFP) was used as the working electrode substrate.
50 potentiostat (Biologic, France).The catalyst ink-modified carbon fiber paper electrodes were used as working electrodes.A platinum mesh and a saturated calomel electrode (SCE) were used as the counter and reference electrode, respectively.GOR tests were carried out in a single compartment electrolytic cell with Ar saturated 1.0 M KOH aqueous solution (pH 13.7) as electrolyte with 0.1 M glycerol.All electrodes were first applied for 10 cyclic voltammetry (CV) cycles in pure 1.0 M KOH solution to reach a stable state before transferring into solutions with glycerol.The iR drop was directly compensated by the potentiostat by 85% level.When collecting product solutions for HPLC tests, the experiments were carried out in a divided cell without iR compensation by using the chronoamperometry (CA) technique.The divided cell was separated by using an anion exchange membrane (Fumasep FAA-3-50) in the middle which was fixed by a stainless steel clip.The cathode chamber consists of the counter electrode with 1.0 M KOH electrolyte; the anode chamber consists of working and reference electrodes with an electrolyte of 1.0 M KOH with 0.
20 Å, b = 3.72 Å, c = 4.03 Å, α = β = γ = 90°}, which compares well with the experimental values {a = 13.86Å, b = 3.69 Å, c = 3.96 Å , α = β = γ = 90°}. 54 x and Pt.Initially, the Pt(111) and MoO 3 (100) were piled forming the MoO 3 (100)/Pt(111) interface region.Afterward, two oxygen atoms were replaced by two platinum atoms (referring to 11% of vacancy).The structure was optimized allowing only the platinum atoms to move.The calculated binding energies (Δ  ) shown in Fig.4awere evaluated based on the following equation: . At 0.85 V, the selectivity to glycerate, glycolate, and formate is 50%, 8%, and 18% respectively on the MoO x /Pt electrode.The main products from the Pt NPs electrode at the same working potential are similar to what is observed from the MoO x /Pt electrode.Its selectivity to glycerate, glycolate, and formate is 40%, 7%, and 9% respectively.Its selectivity to another C3 product (lactate) is relatively high and goes up to 34%.The MoO x /Pt electrode has higher selectivity to glycerate than the Pt NPs electrode.Both the MoO x /Pt and Pt