Selective electrochemical production of hydrogen peroxide at zigzag edges of exfoliated molybdenum telluride nanoflakes

Abstract The two-electron reduction of molecular oxygen represents an effective strategy to enable the green, mild and on-demand synthesis of hydrogen peroxide. Its practical viability, however, hinges on the development of advanced electrocatalysts, preferably composed of non-precious elements, to selectively expedite this reaction, particularly in acidic medium. Our study here introduces 2H-MoTe2 for the first time as the efficient non-precious-metal-based electrocatalyst for the electrochemical production of hydrogen peroxide in acids. We show that exfoliated 2H-MoTe2 nanoflakes have high activity (onset overpotential ∼140 mV and large mass activity of 27 A g−1 at 0.4 V versus reversible hydrogen electrode), great selectivity (H2O2 percentage up to 93%) and decent stability in 0.5 M H2SO4. Theoretical simulations evidence that the high activity and selectivity of 2H-MoTe2 arise from the proper binding energies of HOO* and O* at its zigzag edges that jointly favor the two-electron reduction instead of the four-electron reduction of molecular oxygen.


INTRODUCTION
Hydrogen peroxide (H 2 O 2 ) is a potential energy carrier and an important commodity chemical with high industrial value [1,2]. Its low-concentration (3-9 wt%) solution is also widely used for a vast range of environmental, medical and household applications. At present, over 99% of H 2 O 2 is produced via the energy-intensive anthraquinone oxidation process [3,4]. For economic reasons, this process only operates in centralized reactors on a large scale, and produces highly concentrated H 2 O 2 that often has to be distributed to, and diluted at, the site of use, bringing additional complexity and challenges [5]. In addition to the anthraquinone oxidation process, H 2 O 2 can also be directly converted from H 2 and O 2 in the presence of Pd-based catalysts [4,6]. The practical viability of this high-pressure conversion, however, is seriously undermined by its potential hazard of explosion. There is a pressing call to explore other effective methods for the green, mild and on-demand production of H 2 O 2 .
The electrochemical synthesis of H 2 O 2 from oxygen reduction reaction (ORR) represents an ideal solution. ORR can take place via a four-electron pathway or a two-electron one. The former yields water as the reduction product, and is the key reaction at the cathode of fuel cells and aqueous metal-air batteries [7][8][9]. The latter enables the direct production of H 2 O 2 at ambient conditions, but was substantially less investigated until very recently [5,10]. Due to the competitive nature of these two pathways, suitable electrocatalysts are required to selectively promote the two-electron ORR (2e-ORR) process. Previous theoretical study established the binding energy of HOO * ( G HOO * ) on catalyst surface as the activity descriptor, with the highest activity C The Author(s) 2020. Published by Oxford University Press on behalf of China Science Publishing & Media Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. achieved at the optimal G HOO * ∼ 4.2 eV [11]. This selection rule was used to guide the computational screening of new alloy electrocatalysts and validated through experiments. At present, the state-of-the-art 2e-ORR electrocatalysts are Pt-Hg and Pd-Hg alloys [11,12], followed by Au alloys [13][14][15]. Despite their relatively high mass activity and selectivity in acids, these precious metal alloys are unlikely to be used on a large scale due to their prohibitive costs and sometimes high toxicity. On the other hand, carbon-based materials (e.g. oxidized carbon nanotubes and reduced graphene oxide) exhibit appreciable 2e-ORR activity and selectivity in alkaline solution but generally very poor performance in neutral or acidic solution [16][17][18]. Their potentials are also limited since H 2 O 2 is subjected to rapid decomposition in alkaline medium. We are therefore in great need of developing non-precious-metal-based electrocatalysts with outstanding selectivity and activity for 2e-ORR in acids.
Over recent years, two-dimensional (2D) transition metal dichalcogenides (TMDs) have attracted intense research interest owing to their unique anisotropic structures and intriguing physical and chemical properties [19][20][21]. Many of them (as best exemplified by MoS 2 ) are well known to be active for electrocatalytic hydrogen evolution reaction (HER) [22][23][24]. Computational and experimental studies suggest that their active sites are often located at the edges [25], and that the catalytic activities can be tuned via proper alloying, doping, and defect and strain engineering [26][27][28][29]. However, the potential of TMDs for other electrocatalytic reactions beyond HER remains to be explored. In this study, we demonstrate that 2H-phase molybdenum telluride (MoTe 2 ) nanoflakes, synthesized from bulk powder via ultrasonication-assisted liquid phase exfoliation, acts as an efficient 2e-ORR electrocatalyst in acids. They are measured to catalyze the electrochemical production of H 2 O 2 with high activity, selectivity and stability. Our theoretical calculations reveal that the high activity and selectivity can be attributed to the favorable binding of HOO * and weak binding of O * at the zigzag edges.

Exfoliation and characterizations of MoTe 2 nanoflakes
Even though synthesis of high-quality MoTe 2 nanosheets or nanoflakes by chemical vapor deposition (CVD) has been reported in literature [30,31], such a high-temperature bottom-up approach is seriously limited by its complexity and low production yield. Here, top-down liquid phase exfoliation (LPE) was employed for its notable simplicity, scalability and reproducibility [32,33]. Commercial crystalline MoTe 2 powders were ultrasonicated in N-methylpyrrolidone (NMP) (see Experimental Method in the Supplementary data for details). During the ultrasonication, the weak interlayer van der Waals interactions were interrupted; the 2D bulk crystal was then exfoliated to few-layered nanoflakes stabilized by NMP (Fig. 1a). These nanoflakes were collected by centrifugation and could be readily re-dispersed in common solvents. Figure 1b showed exfoliated MoTe 2 nanoflakes re-dispersed in ethanol. The dispersion was stable for weeks without obvious sediment. Powder X-ray diffraction (XRD) pattern of the product confirmed that it was composed of hexagonal 2H MoTe 2 , which is the most stable form of MoTe 2 at low temperatures. Under scanning electron microscopy (SEM), the product was revealed to consist of nanoflakes ( Fig. 1d and e). Additional SEM image and low-magnification transmission electron microscopy (TEM) images were supplemented in Fig. S1. Based on TEM statistics, the lateral size of MoTe 2 nanoflakes was analyzed to be in the range of 50-350 nm with the mean length ∼143 nm (Fig. S1). Raman spectrum of the product displayed two pronounced peaks at 170 and 233 cm −1 (Fig. 1f), which are characteristic to the A 1g and E 2g modes of 2H MoTe 2 , respectively [34,35].
To elucidate the atomic structure of exfoliated MoTe 2 nanoflakes, aberration-corrected scanning transmission electron microscopy (STEM) was utilized. Figure 2a showed a typical annular dark field (ADF) image of the product, which could be identified as 2H MoTe 2 along the c-axis based on the honeycomb pattern of the atomic sites and the corresponding fast Fourier transform (FFT) pattern. The assignment was further corroborated by the excellent agreement between the experimental image and the simulated ADF image of 2H MoTe 2 (0001) plane ( Fig. 2b and c). Cross-sectional STEM-ADF image of MoTe 2 nanoflakes was also acquired as shown in Fig. 2d. Its corresponding FFT pattern matched with the diffraction pattern of 2H MoTe 2 along the [0110] zone axis. The simulated ADF image along this zone axis also agreed well with the experimental image ( Fig. 2e and f). Furthermore, since the intensity of STEM-ADF image increased approximately linearly with the number of layers in thin 2D crystals, we could determine the nanoflake thickness for up to seven layers (Fig. S2) and directly examine the stacking mode between layers by quantifying the image intensity. As summarized in in the bilayer honeycomb lattice showed very similar image intensity, a characteristic feature for 2H stacking with overlapped Mo+Te 2 sites. This intrinsic feature of 2H-stacking MoTe 2 , with 180 • rotation between two adjacent layers, could help us identify the edge orientations in multilayer flakes. The exposed edges, though not atomically sharp, were mostly along the zigzag directions with abundant unsaturated Mo and Te bonds (Fig. 2g), which might be responsible for the observed high electrocatalytic activity as described in the following part [36].

Electrochemical production of H 2 O 2
To assess the electrocatalytic performance of MoTe 2 nanoflakes for 2e-ORR to H 2 O 2 , they were physically mixed with graphene nanosheets exfoliated from graphite powders as the conductive additive (Fig. S4, see Experimental Method in the Supplementary data for details). Graphene nanosheets were used here instead of conventional carbon black because of their superior electrical conductivity and similar 2D geometry that could form better contacts with MoTe 2 nanoflakes [37]. The electrocatalyst mixture was then loaded onto the glassy carbon disk of a rotating ring disk electrode (RRDE) with an active material loading of 10 μg cm −2 . RRDE voltammograms were separately carried out in N 2 -saturated and O 2 -saturated 0.5 M H 2 SO 4 at the electrode rotating speed of 1600 rpm (Fig. S5). Corresponding ORR polarization curves were then derived from their differences, and summarized in the lower panel of Fig. 3a. Exfoliated graphene nanosheets alone were electrochemically inert and exhibited no apparent cathodic current density till at <0 V (versus reversible hydrogen electrode or RHE, the same hereafter). Bulk MoTe 2 mixed with graphene nanosheets also had a negligible activity. By stark contrast, exfoliated MoTe 2 nanoflakes demonstrated a dramatically improved performance with an onset potential as positive as ∼0.56 V (corresponding to an overpotential of ∼140 mV), close to the state-of-the-art PtHg 4 alloy reported in literature (∼0.6 V) [11]. Their cathodic current density continuously increased beyond the onset, and reached ∼1.9 mA cm −2 at 0 V. The broad wave centered at ∼0.3 V was likely associated with the reduction of the surface oxide. Based on the concurrently measured ring current, we further derived the H 2 O 2 percentage in the product from MoTe 2 nanoflakes (upper panel of Fig. 3a). It was shown to stay >80% at almost the entire potential region with a recorded peak value of ∼93%. Such remarkable selectivity was also comparable to Pt-Hg and Pd-Hg alloys [11,12]. In addition, the effect of catalyst loading on the geometric current density and H 2 O 2 percentage was investigated. Increasing the loading was found to slightly enlarge the cathodic current density but adversely compromise the reaction selectivity (Fig. S6). We speculated that the higher catalyst loading helped retain produced H 2 O 2 , and caused it to be further reduced to H 2 O via another two-electron pathway, therefore giving rise to lower H 2 O 2 selectivity.
In order to allow direct activity comparison with other reported 2e-ORR electrocatalysts in acids, we normalized the H 2 O 2 partial current of MoTe 2 nanoflakes over the mass of active material as presented in Fig. 3b. The mass activity was calculated to be in the range of ∼10-10 2 A g −1 between 0.3-0.45 V, which, although still not as magnificent as the state-of-the-art Pt-Hg and Pd-Hg alloys [11,12], was superior to Au alloys [14,15] and carbon-based materials [38,39]. For example, the mass activity of MoTe 2 nanosheets at 0.4 V was 27 A g −1 -which is ∼7-10 times greater than those of Au-Pd alloys [14,15] and N-doped carbon [38,39]. Interestingly, MoTe 2 seemed to be quite unique among TMD materials for the electrochemical production of H 2 O 2 . Its sulfide and selenide analogues-MoS 2 and MoSe 2 nanoflakes likewise exfoliated from corresponding bulk powders and hybridized with graphene nanosheets-exhibited significantly worse activities (Fig. 3c).
Stability is another key parameter in the electrocatalyst assessment. Despite the concern over the susceptibility of MoTe 2 to oxidation, we found that exfoliated MoTe 2 nanoflakes had very decent electrochemical stability under the 2e-ORR working condition, at least for several days. To demonstrate this, our electrocatalyst was subjected to an accelerated durability test by rapidly cycling between 0 and 0.3 V at 100 mV s −1 in O 2 -saturated electrolyte for a predetermined number of cycles, and then measuring its polarization curves. Figure 3d compared the polarization curves at the initial state, after 2500 cycles and after 5000 cycles. The measured disk polarization curves remained largely similar except for the reduction wave at ∼0.3 V presumably due to the partial surface oxidation during the durability test. There was only a slight decrease in the potential-dependent H 2 O 2 percentage likely as a result of the catalyst oxidation. Moreover, even after MoTe 2 nanoflakes were aged overnight (>12 h) in the O 2 -saturated electrolyte at the open-circuit potential, only slight current decay occurred and no apparent selectivity decay were observed as compared to the fresh electrode (Fig. S7).

DFT calculations
In order to understand the origin of the high activity and selectivity of MoTe 2 nanoflakes, spin-unrestricted density functional theory (DFT) calculations were carried out to simulate the 2e-ORR process on both the basal plane and edge site of 2H MoTe 2 as well as several other materials. The basal plane slab was modeled using a 4 × 4 × 1 supercell (Fig. S8). Previous experimental studies on MoS 2 demonstrated that its most stable edges were zigzag-type with the Mo atoms covered by 50% S coverage [40,41]. As a result, the Mo-edge slab in our study was constructed with 50% Te coverage and a periodicity of 3 Mo atoms, which was determined to be the most stable configuration (Fig. S9). 2e-ORR to H 2 O 2 generally involves two elemental steps: molecular O 2 is first transformed to HOO * via a proton-coupled electron transfer, followed by the protonation and reduction of HOO * to yield H 2 O 2 . Previous study established that the binding energy of HOO * ( G HOO * ) was an effective activity descriptor, and its optimal value was found to be ∼4.2 eV [42]. Following this guiding principle, G HOO * on different surface sites was computed and compared. Figure 4a showed the optimized structures of HOO * adsorbed on the basal plane or edge of 2H MoTe 2 . G HOO * calculations suggested that the H 2 O 2 formation was overwhelmingly challenging on the basal plane ( G HOO * = 5.34 eV). Introduction of Te vacancies to the basal plane was also not helpful. For example, HOO * could not even be produced over the single Te vacancy on the basal plane as it would be spontaneously dissociated to O * adsorbed on the Te vacancy site and HO * adsorbed on the nearby Te site (Fig. S10). By contrast, the reaction could readily proceed at the zigzag edge with G HOO * = 4.32 eV. Such an observation was interestingly reminiscent of the structure-dependent HER activities of 2H-MoS 2 and many other TMD materials. Furthermore, the theoretical overpotential (η t ) of 2e-ORR at the edge of MoTe 2 was estimated by | G HOO * /e − 4.22 V| and calculated to be 100 mV, which agreed reasonably well with the experimental value (∼140 mV), and was sufficiently close to that of PtHg 4 (110) sur-face ( G HOO * = 4.28 eV and η t = 60 mV) [11]. It unambiguously corroborated the high 2e-ORR activity of 2H MoTe 2 . Worth noting is that the armchair edge of MoTe 2 (Fig. S11) was not catalytically active for 2e-ORR because it bound HOO * too weakly ( G HOO * = 4.60 eV). For the purpose of comparison, the HOO * adsorption on Pt(111) was calculated to be considerably stronger than the ideal ( G HOO * = 3.98 eV), and the HOO * adsorption on MoS 2 and MoSe 2 edges significantly weaker than the ideal ( G HOO * = 4.60 eV and 4.59 eV on MoS 2 and MoSe 2 , respectively, Fig. 4b).
The observed inefficiency of MoS 2 and MoSe 2 for 2e-ORR was therefore rationalized in spite of their structural similarity to MoTe 2 .
Of note, a favorable G HOO * value does not necessarily guarantee high H 2 O 2 selectivity since 2e-ORR and 4e-ORR share the same initial step. The further transformation of HOO * to O * (4e-ORR pathway) would have to be effectively suppressed in order to achieve high H 2 O 2 selectivity. As a result, weak O * binding energy ( G O * ) on the catalyst surface becomes a prerequisite. Based on this rationale, we could understand the predominant 4e-ORR selectivity on Pt owing to its very strong O * affinity ( G O * = 1.58 eV on Pt(111)), as well as the predominant 2e-ORR selectivity on PtHg 4 owing to its very weak O * affinity ( G O * = 2.75 eV on PtHg 4 (110)) ( Fig. 4c) [11]. Our calculations showed that the edge of 2H MoTe 2 had G O * = 2.33 eV, which, albeit still not as positive as PtHg 4 , was sufficiently large to render MoTe 2 H 2 O 2 -selective. It is also worth mentioning that MoS 2 relatively favored 4e-ORR, and MoSe 2 relatively favored 2e-ORR even though both of them had negligible activities. At last, based on the previous volcano for 2e-ORR built upon the sole activity descriptor G HOO * [11], we further compiled our computation results on a 2D heat map as shown in Fig. 4d. We implicitly assumed that the hottest spot was located at G HOO * = 4.22 eV and G O * = 2.75 eV. The G O * of PtHg 4 (2.75 eV) was chosen as a reference given the fact that, in contrast to HOO * , there is not explicit criteria for denoting the optimal O * affinity. According to this definition, the closer to the hottest spot, the better 2e-ORR performance, and vice versa. Among different materials under investigation, PtHg 4 was in the region of the greatest activity and followed by MoTe 2 , while MoSe 2 , MoS 2 and Pt were quite far away from the hot area.

CONCLUSION
In summary, we for the first time introduced 2H-MoTe 2 nanoflakes as a high-performance catalyst for electrochemical H 2 O 2 production in acids. MoTe 2 nanoflakes were exfoliated from the bulk powder via LPE in NMP. They were determined to have small thickness of a few nm, lateral sizes of 100∼500 nm and with preferentially exposed zigzag edges. When physically mixed with graphene nanosheets as the conductive additive, the MoTe 2 nanoflakes demonstrated excellent activity and selectivity for 2e-ORR in 0.5 M H 2 SO 4 , with an onset potential at ∼0.56 V (η ∼140 mV), a large mass activity of 27 A g −1 at 0.4 V and high H 2 O 2 selectivity up to ∼93%. Such a performance was far superior to those of Au alloys and N-doped carbon reported in literature, and approaching that of the state-of-the-art PtHg 4 alloy. MoTe 2 nanoflakes also exhibited impressive chemical and electrochemical stability in an accelerated durability test and overnight aging experiment. Finally, detailed DFT calculations showed that the high activity and selectivity of 2H MoTe 2 originated from the favorable binding of HOO * and weak binding of O * at the zigzag edges, and thereby directly correlated the electrocatalytic performance with the unique anisotropic structure of MoTe 2 . Our study here unveiled the unexpected potential of MoTe 2 nanoflakes as a non-preciousmetal-based electrocatalyst for H 2 O 2 production in acids, and might open a new pathway toward the catalyst design for this challenging electrochemical reaction.

SUPPLEMENTARY DATA
Supplementary data are available at NSR online.

AUTHOR CONTRIBUTIONS
Y.G.L. conceived the project and designed the experiments. X.Z. and X.W. prepared the material and conducted electrochemical measurements. X.H. and J.N.C. contributed to the material preparation method. T.W. assisted in electrochemical measurements. Y.D., M.X. and W.Z. performed the STEM analysis. Y.W. and Y.F.L. conducted the theoretical calculations. X.Z., Y.W., Y.D. and Y.G.L. co-wrote the paper. All authors discussed the results and commented on the manuscript.