The rapid electrochemical activation of MoTe2 for the hydrogen evolution reaction

The electrochemical generation of hydrogen is a key enabling technology for the production of sustainable fuels. Transition metal chalcogenides show considerable promise as catalysts for this reaction, but to date there are very few reports of tellurides in this context, and none of these transition metal telluride catalysts are especially active. Here, we show that the catalytic performance of metallic 1T′-MoTe2 is improved dramatically when the electrode is held at cathodic bias. As a result, the overpotential required to maintain a current density of 10 mA cm−2 decreases from 320 mV to just 178 mV. We show that this rapid and reversible activation process has its origins in adsorption of H onto Te sites on the surface of 1T′-MoTe2. This activation process highlights the importance of subtle changes in the electronic structure of an electrode material and how these can influence the subsequent electrocatalytic activity that is displayed.

Reaction between Mo and Te at temperatures of 400 °C allows for the formation of a black material with a diffraction pattern ( Supplementary Fig. 1) similar to the one collected on a sample prepared by regrinding single crystals of the high temperature monoclinic 1Tꞌ-MoTe 2 phase and literature reports. 13,14,15 The pattern also matches well with the nanocrystalline MoTe 2 products synthesised recently via a solution based route. 16 It is apparent that the low synthetic temperature results in a more disordered material (referred to as nanocrystalline 1Tꞌ-MoTe 2 throughout the text), which is easily distinguishable from its high temperature  Fig. 3b).
Electron diffraction patterns collected from the crystallites confirm the validity of the proposed model, e.g. 1Tꞌ-MoTe 2 with the monoclinic distortion. Thus, the formation of 1T'-MoTe 2 polymorph is confirmed ( Supplementary Fig. 3c-e) and is in good agreement with the data reported for nanocrystalline 1Tꞌ-MoTe 2 by Schaak's group. 16 Further twins with twin boundary (100) and respective splitting of planes along a*-direction were found which are prototypical for the para-variant twin of 1T'-MoTe 2 . 21 In comparison, the crystalline 1Tꞌ-MoTe 2 exhibited substantially larger crystallites evident as platelets with the sizes in the range of several micrometres (Supplementary Fig. 4a). Electron diffraction patterns were found in good agreement with expected monoclinic structure as well ( Supplementary Fig. 4b-d).  Fig. 12).

Exchange current density, Turnover Frequencies and Surface Coverage:
The exchange current densities (i 0 ) for 1Tꞌ-MoTe 2 were estimated by extrapolation the Tafel slopes on 0 mV. In the case of the non-activated, initial material the observed value was 1.03×10 4 mA cm 2 . The activated material showed a substantially higher value of 5.03×10 1 mA cm 2 . It should be mentioned that the current densities were calculated per geometrical area of the electrode which was 0.071 cm 2 .
Catalytic activity can be quantified in terms of the turnover frequency (TOF A further assumption was made that only 1/3 of the surface would be available for electrocatalytic reaction due to overlaps between platelets within the material. As such, A is an estimate rather than an exact value.
The TOF is measured in molecules per second and despite being expressed as a fractional number it implies that only a certain number of sites on the surface of the MoTe 2 is available for catalytic reaction. Therefore, it is better to express the TOF through the surface coverage which would represent the integer number of sites available. For example, the surface coverage of 12.3 % for product after 100 cycles would correspond to roughly one of every eight catalytic sites available for catalytic reaction at a given time.

Computational studies
The bonding energy of a surface to a hydrogen atom can be directly correlated to the activation energy of the reduction step. 12 From an experimental standpoint this is expressed by the electrochemical overpotential. Therefore, it is useful to compute the binding energies to infer and predict the optimum catalyst and additionally obtain an atomistic insight into the catalytic process. A 2x2 supercell of a monolayer slab was adapted as working models of 1Tꞌ-MoTe 2 (Fig. 4a). The optimised unit cell parameters (a = 649.2 pm, b = 337.5 pm) for the monolayer surface are comparable with experimentally reported (a = 633 pm, b = 346.9 pm) and calculated values (a = 652 pm, b = 352 pm). 13 The slight expansion of the a-parameter is likely due to solvation effects. The values for the binding energies (Supplementary Table   4) should be examined with caution since they result from some approxiamtions on the surface binding site. For instance, charge effects caused by the electrode potential are absent. 24  The ighest Occupied Crystal Orbital ( OCO) at k={0,π/(4b)} is displayed in Supplementary   Fig. 21 and it is a zig-zag linear combination of in-phase 4d-orbitals of σ symmetry. The Lowest Unoccupied counterpart (LUCO) is a linear combination of 4d-orbitals of π symmetry perpendicular to the surface plane.
At k={π/a,0} another set of 4d orbitals are present, the OCO being similar in nature to the one at k={0,π/(4b)} except they are slightly tilted above and below the surface plane in a zigzag manner which is the result of a slight admixing with the π symmetry set of AOs. The LUCO is altogether different in composition and is a linear combination involving metal to ligand π-backdonation. This virtual crystal orbital belongs to a band which travels across the

Additional Experimental Details of Computational Studies
More accurate single point energy evaluations of the obtained stationary points were calculated using a denser 5×9 k point grid on the optimised unit cells (Scheme 1).
Scheme 1: Sketch of the real space unit cell and the corresponding main symmetry points in the Brillouin zone for an approximately rectangular lattice where a>b. The reciprocal lattice was sampled with 5 k points along a -1 and 9 k points along b 1 . a b (π/a, π/b) (0,π/b) (π/a,0) (0,0) Crystal orbital overlap populations were calculated using a grid of 2000 points in the range of 0.7 to +0.4 a.u. below and above the Fermi levels respectively.

El ctr ch mical activati n f crystallin '-MoTe 2
Upon cycling crystalline 1T'-MoTe 2 under reducing potentials, a similar activation is observed, albeit to a lesser extent most likely due to the bulk plate-like microcrystals which may hinder access of protons to the active sites ( Supplementary Fig. 11). Therefore, it is evident that the morphology of 1T'-MoTe 2 plays a key role in the activation process.
Furthermore, EIS indicates higher conductivity for the nanocrystalline 1T'-MoTe 2 compared with the crystalline phase ( Supplementary Fig. 14). This is in line with optical spectroscopy absorbance values, which are proportional to the optical conductivity and thus indicate a conductivity increase when going from the crystalline to the nanocrystalline phase ( Supplementary Fig. 8). This means that electron transport is different between the two phases, likely due to the shorter interlayer pathways associated with the disordered nature of the nanocrystalline 1T'-MoTe 2 material.