The Synergistic Effects of Alloying on the Performance and Stability of Co3Mo and Co7Mo6 for the Electrocatalytic Hydrogen Evolution Reaction

Metal alloys have become a ubiquitous choice as catalysts for electrochemical hydrogen evolution in alkaline media. However, scarce and expensive Pt remains the key electrocatalyst in acidic electrolytes, making the search for earth-abundant and cheaper alternatives important. Herein, we present a facile and efficient synthetic route towards polycrystalline Co3Mo and Co7Mo6 alloys. The single-phased nature of the alloys is confirmed by X-ray diffraction and electron microscopy. When electrochemically tested, they achieve competitively low overpotentials of 115 mV (Co3Mo) and 160 mV (Co7Mo6) at 10 mA cm−2 in 0.5 M H2SO4, and 120 mV (Co3Mo) and 160 mV (Co7Mo6) at 10 mA cm−2 in 1 M KOH. Both alloys outperform Co and Mo metals, which showed significantly higher overpotentials and lower current densities when tested under identical conditions, confirming the synergistic effect of the alloying. However, the low overpotential in Co3Mo comes at the price of stability. It rapidly becomes inactive when tested under applied potential bias. On the other hand, Co7Mo6 retains the current density over time without evidence of current decay. The findings demonstrate that even in free-standing form and without nanostructuring, polycrystalline bimetallic electrocatalysts could challenge the dominance of Pt in acidic media if ways for improving their stability were found.


Introduction
The share of wind and solar in global electricity output has reached 9% in Q1 2020 [1,2]. It is largely anticipated that the output from renewables will continue to grow over the next decade. This presents a substantial challenge to distribution networks due to the intermittent nature of renewable energy. Batteries have been considered an efficient way of moderating daily changes in electricity production, but they discharge quickly and hence are currently unable to provide a long-term solution for managing the electricity output between the seasons [3]. Storing the excess renewable electricity directly as fuels, for example, from photovoltaic cells during summer and then using these fuels later during cold and short winter days could be a practical way to address this challenge.
Hydrogen (with a world demand of over 70 Mt per year in 2018 [4][5][6]) is an excellent fuel and key feedstock in the production of chemicals and fertilisers. However, less than 0.1% of global hydrogen currently comes from water electrolysis. This can change rapidly if electrolysis is used as a storage platform for mitigating seasonal variations in electricity output rather than as a stand-alone method of H 2 production [7]. In an electrochemical cell (which is an integral part of any industrial electrolyser), water electrolysis yields best its results in alkaline and acidic electrolytes. Hydrogen evolves at a cathode, and thus the choice of cathode materials (which simultaneously act as electrocatalysts) is crucial for driving the hydrogen evolution reaction (HER) from water at appreciable voltage and current 12 density rates [8]. Industrial alkaline electrolysers (AEL) with an energy output as high as 6 MW [7] have already benefited from the early adoption of low-cost Ni-based alloys as cathode materials [5][6][7][8][9][10]. Pt-group metals, on the other hand, remain the only viable choice in acidic conditions [11,12]. Platinum contributes to the costs of proton-exchange membrane electrolysers (PEMEL), which are almost twice as expensive per KWh as AEL [7]. With the PEMEL offering short start-up times and load-flexibility (which are significant parameters for the integration with renewable energy networks), reducing the cost of PEMEL is important. Finding earth-abundant alternatives to Pt in acidic conditions is thus the key to future integration of PEMEL into renewable networks and more research is needed in this context.
Metal alloys show electrocatalytic performance superior to their individual metal counterparts [13][14][15]. This synergetic effect has been successfully exploited in alkaline electrolytes [7,9,15], but only limited attempts have been made to investigate the electrocatalytic performance of polycrystalline products with well-defined structures and compositions in acidic media [16][17][18][19]. A growing interest in nanostructured materials appeals for a better understanding of the electrochemical properties of pure-phased polycrystalline alloys, and in this respect the Co-Mo system remains relatively unexplored. Coupled with the recent reports of impressive performance of Co 3 Mo towards HER in 1M KOH [20,21], the investigation of electrochemical properties of alloys in the Co-Mo system is key to understanding how cheaper and more efficient electrocatalysts could be designed and made.
In this work, we investigated two binary compounds within the Co-Mo system and tested them as electrocatalysts for HER. Phase-pure, polycrystalline Co 3 Mo, and Co 7 Mo 6 were prepared as powders and electrochemically tested for HER in acidic and alkaline media. The tests were carried out on free-standing products directly on a glassy carbon electrode to minimize the effect of the substrate and possible impurities on their catalytic performance and electrochemical stability. There was a pronounced synergistic effect for both Co 3 Mo and Co 7 Mo 6 that showed substantially higher current densities compared with Co and Mo metals. However, polycrystalline free-standing Co 3 Mo has a poor electrochemical stability. This makes it a suboptimal choice as an electrocatalyst compared with Co 7 Mo 6 , which retained its performance after prolonged hours of testing.

Synthesis
CoMoO 4 precursor was made from Na 2 MoO 4 ·2H 2 O (Sigma-Aldrich, Gillingham, UK 99.99%) and CoCl 2 ·6H 2 O (Alfa Aesar, Lancashire, UK 99.9%). First, Na 2 MoO 4 ·2H 2 O (1 g; 4.13 mmol; 1.00 eq.) was dissolved in 75 mL of deionized water in a 200 mL beaker and stirred at 400 rpm to produce a clear, 0.55 M solution. In a separate 200 mL beaker, CoCl 2 ·6H 2 O (0.983 g, 4.13 mmol, 1.00 eq.) was dissolved in 75 mL of water to produce a 0.55 M solution as well. The CoCl 2 ·6H 2 O solution was slowly added to the solution of Na 2 MoO 4 ·2H 2 O while stirring. The resulting solution was stirred in ambient conditions for 4 h. The solid product was then isolated on a cellulose acetate membrane filter (0.2 µm pore diameter, Sartorius, Epsom, UK) from the aqueous solution via vacuum filtration and washed with 1 L of water. The resulting powder (proved as a phase-pure CoMoO 4 by XRD) was dried overnight on the filter in ambient conditions and used without any additional temperature treatment. Co 7 Mo 6 was prepared in a single-step reaction between Co(OH) 2 (0.0080 g, 0.0861 mmol, Alfa Aesar, 99.9%) and CoMoO 4 (0.11294 mg, 0.5160 mmol). Co(OH) 2 and CoMoO 4 powders were mixed together with mortar and pestle. The powder mixture was sandwiched between two silica wool pieces inside a silica tube (8 mm diameter and 20 mm long). The silica tube was then placed inside a tube furnace. The reaction was carried out at 950 • C for 10 hours in 5% H 2 in Ar gas flow.
Co 3 Mo was made in two steps. First, Co(OH) 2 (0.08492 g, 0.9136 mmol Alfa Aesar, 99.9%) and CoMoO 4 (0.100 g, 0.4569 mmol) were mixed together with mortar and pestle. The powder mixture was sandwiched between two silica wool pieces inside a silica tube (8 mm diameter and 20 mm long). The silica tube was then placed inside a tube furnace. The reaction was carried out at 700 • C for 10 h in 5% H 2 in Ar gas flow. The resulting powder was reground with mortar and pestle, placed back into the silica tube, and reannealed at 850 • C for 10 h in 5% H 2 in Ar gas flow.
Mo metal was made by reduction of MoO 3 (0.15 g, Sigm-Aldrich, 99.9%). MoO 3 powder was sandwiched between two silica wool pieces inside a silica tube (8 mm diameter and 20 mm long). The silica tube was then placed inside a tube furnace. The reaction was carried out at 950 • C for 10 h in 5% H 2 in Ar gas stream.
Co metal was prepared by heating Co(OH) 2 (0.15 g, Alfa Aesar, 99.9%). Co(OH) 2 powder was sandwiched between two silica wool pieces inside a silica tube (8 mm diameter and 20 mm long). The silica tube was then placed inside a tube furnace. The reaction was carried out at 350 • C for 5 h in 5% H 2 Ar stream.

Materials Characterisation
Powder X-ray diffraction (PXRD) was carried out on a Panalytical Xpert-Pro diffractometer with the Cu Kα (λ = 1.54184 Å) source operating in the Bragg-Brentano geometry. A sample was carefully spread over a zero-background holder and flattened with a glass slide. The simulation of the XRD patterns from ICSD data was carried out using Mercury 4 software [22]. The stability studies were carried out by PXRD directly on glassy carbon electrode using the fabricated bracket [23] and depicted in Figure S1.
Scanning Electron Microscopy (SEM) was performed using a Phillips XL30 ESEM instrument equipped with an Oxford Instruments X-act spectrometer for Energy Dispersive X-ray Spectroscopy (EDXS) measurements. The INCA software was used for the data analysis. Cu foil was used as the calibration standard for the EDXS. The sample preparation involved dispersing ca. 1 mg of alloy powder in 1 mL of hexane with sonication for 5 minutes. A 100 µL drop of resulting solution was casted on Cu foil, producing a well-spread and homogenous coating.

Electrochemical Characterisation
Electrocatalysts were prepared in the form of inks, which were loaded onto polished glassy carbon electrodes, for catalytic testing. To prepare the catalytic ink, 8 mg of a synthesised catalyst was added to a mixture of 1.00 mL of DMF (99.9%, Sigma-Aldrich) and 50 µL of Nafion (5 wt. % in mixture of lower aliphatic alcohols and water, Sigma-Aldrich) followed by sonication for half an hour. Then, 30 µL of the prepared inks were applied onto the surface of polished glassy carbon electrodes with a surface area of 0.071 cm 2 (corresponding to sample loading of ca. 3.3 mg cm −2 geometric ) and left to dry overnight. A Biologic SP-150 potentiostat with a three-electrode setup was used to investigate the electrochemical performance. The prepared electrode was used as a working electrode, and carbon felt and 3 M Ag/AgCl were used as counter and reference electrodes, respectively. The catalytic activity towards the HER was tested in 0.5 M H 2 SO 4 (prepared by dilution of 95% H 2 SO 4 , Fischer, Wallingford, UK, with deionised water) and 1 M KOH (prepared from 85% KOH, Honeywell, UK) for all materials. The electrode potentials were converted to reversible hydrogen electrode (RHE) by E(RHE) = E(Ag/AgCl) + 0.209V + 0.059 × pH value, and the ohmic resistances were compensated. Polarization curves (obtained using linear sweep voltammetry (LSV) and cyclic voltammetry (CV)) were set up in a single-compartment electrochemical cell, and performed with a scan rate of 5 mV s −1 and 100 mV s −1 . Tafel plots were obtained from the polarisation data.

Gas Chromatography Measurements
Gas chromatography (GC) was used to confirm the production of hydrogen using an Agilent GC 7890A instrument with a thermal conductivity detector. The column used was a 30 meter-long 0.320 mm widebore HP-molesieve column (Agilent). The GC oven temperature was set to 27 • C, and the carrier gas was Ar. The front inlet was set to 100 • C. The hydrogen production was experimentally measured using a single cell two-electrode set up ( Figure S2), which involved the catalyst-deposited on a glassy carbon working electrode and carbon felt counter electrode attached to a silver wire immersed in 0.5 M H 2 SO 4 . A constant volume of electrolyte was used in order to keep the cell headspace accurate. The cell was then sealed with a septum and flushed with Ar for 30 minutes. A representative 25 µL sample of the headspace was taken with a syringe and injected into the GC. No hydrogen peak was detected on the chromatogram at this stage. The flow of Ar was stopped, and galvanostatic electrolysis was then performed with an applied current of −0.24 mA. At regular intervals, a representative 25 µL sample of the headspace was taken with a syringe and probed by the GC. The experimental volume% of H 2 in the cell headspace was calculated from the comparison of the hydrogen peak areas detected by GC and the peak area detected by GC on certified standards of hydrogen at various % concentrations in Ar (CK Gas Products Limited, UK) before the electrolysis. The detection on certified standards was carried out as follows. The cell sealed with a septum was flushed with a standard for 30 min, and then a representative 25 µL sample of the headspace was taken with a syringe and injected into the GC. The expected volume percentages of hydrogen in the headspace was calculated by converting the charge passed to an expected number of moles of gas, taking the volume of 1 mole of an ideal gas at room temperature and pressure to be 24.5 L.

Results and Discussion
The phase diagram of Co-Mo is well established [24][25][26][27]. Initial attempts were made to make Co 3 Mo from CoMoO 4 as described in [20]. However, only formation of Co 7 Mo 6 with Mo as an impurity (in line with the nominal Co:Mo = 1:1 composition in CoMoO 4 ) was observed within tested temperature ranges of 700-950 • C. Therefore, the reaction of mixtures of CoMoO 4 and Co(OH) 2 was used instead of a single CoMoO 4 precursor. This proved to be a successful strategy for synthesis, and produced phase-pure products. Therefore, all samples reported in this work were made by this route as also described in the experimental section.
The PXRD pattern of Co 3 Mo ( Figure 1a) prepared at 850 • C shows a very good match with the simulated patterns for hexagonal Co 3 Mo [28]. The attempts to prepare Co 3 Mo at higher temperatures (to match the reaction temperature with the one used for synthesis of Co 7 Mo 6 for a better comparison between two alloys) led to phase separation with Co and Co 7 Mo 6 as impurities. Therefore, the temperature of 850 • C was deemed optimal. The attempt to carry out the reaction at higher temperatures was motivated by the assumption that products prepared at similar temperatures are more likely to have similar morphologies and surface areas. Hence, a comparatively more reliable assessment of their electrochemical performance would be possible.
Hydrogen 2020, 1, FOR PEER REVIEW 4 hydrogen peak was detected on the chromatogram at this stage. The flow of Ar was stopped, and galvanostatic electrolysis was then performed with an applied current of −0.24 mA. At regular intervals, a representative 25 µL sample of the headspace was taken with a syringe and probed by the GC. The experimental volume% of H2 in the cell headspace was calculated from the comparison of the hydrogen peak areas detected by GC and the peak area detected by GC on certified standards of hydrogen at various % concentrations in Ar (CK Gas Products Limited, UK) before the electrolysis. The detection on certified standards was carried out as follows. The cell sealed with a septum was flushed with a standard for 30 min, and then a representative 25 µL sample of the headspace was taken with a syringe and injected into the GC. The expected volume percentages of hydrogen in the headspace was calculated by converting the charge passed to an expected number of moles of gas, taking the volume of 1 mole of an ideal gas at room temperature and pressure to be 24.5 L.

Results and Discussion
The phase diagram of Co-Mo is well established [24][25][26][27]. Initial attempts were made to make Co3Mo from CoMoO4 as described in [20]. However, only formation of Co7Mo6 with Mo as an impurity (in line with the nominal Co:Mo = 1:1 composition in CoMoO4) was observed within tested temperature ranges of 700-950 °C. Therefore, the reaction of mixtures of CoMoO4 and Co(OH)2 was used instead of a single CoMoO4 precursor. This proved to be a successful strategy for synthesis, and produced phase-pure products. Therefore, all samples reported in this work were made by this route as also described in the experimental section.
The PXRD pattern of Co3Mo ( Figure 1a) prepared at 850 °C shows a very good match with the simulated patterns for hexagonal Co3Mo [28]. The attempts to prepare Co3Mo at higher temperatures (to match the reaction temperature with the one used for synthesis of Co7Mo6 for a better comparison between two alloys) led to phase separation with Co and Co7Mo6 as impurities. Therefore, the temperature of 850 °C was deemed optimal. The attempt to carry out the reaction at higher temperatures was motivated by the assumption that products prepared at similar temperatures are more likely to have similar morphologies and surface areas. Hence, a comparatively more reliable assessment of their electrochemical performance would be possible. The formation of phase-pure rhombohedral Co7Mo6 was only successful at relatively high temperature of 950 °C, as evident from the powder X-ray diffraction, which matched very well with the peak positions of the simulated pattern for Co7Mo6 from the inorganic crystal structure database (ICSD) [29]. The lower intensity of some peaks, especially the peak at ca. 42.5°, when compared with the simulated pattern was probably due to peak broadening caused by disorder (stacking faults) within the rhombohedral structure of Co7Mo6 [30][31][32]. Lower reaction temperatures used for synthesis of Co7Mo6 always resulted in elemental Mo impurities. The formation of phase-pure rhombohedral Co 7 Mo 6 was only successful at relatively high temperature of 950 • C, as evident from the powder X-ray diffraction, which matched very well with the peak positions of the simulated pattern for Co 7 Mo 6 from the inorganic crystal structure database (ICSD) [29]. The lower intensity of some peaks, especially the peak at ca. 42.5 • , when compared with the simulated pattern was probably due to peak broadening caused by disorder (stacking faults) within the rhombohedral structure of Co 7 Mo 6 [30][31][32]. Lower reaction temperatures used for synthesis of Co 7 Mo 6 always resulted in elemental Mo impurities.
The reader may have noticed from the experimental part that both Co and Mo used as controls were prepared by reduction in H 2 /Ar stream from Co(OH) 2 and MoO 3 (rather than using elemental Co and Mo purchased from the suppliers of chemicals). The rationale behind this approach was based on an assumption that it would provide a more adequate comparison with Co 7 Mo 6 and Co 3 Mo, which were made under reductive conditions from CoMoO 4 and Co(OH) 2 . Both Co and Mo are single phase products according to PXRD (Figures S3 and S4).
The morphology of Co 3 Mo from the SEM (Figure 2a and Figure S5 for a higher magnification image) could be described as consisting of spherical-shaped particles with a diameter of ca. 300-500 nm. The microstructure of Co 3 Mo is substantially different from the morphology of CoMoO 4 precursor, which consisted of well-defined prismatic microcrystals ( Figure S6). On the other hand, there is a certain degree of similarity between the morphology of Co 3 Mo synthesized in this work and Co 3 Mo prepared on Cu-substrate by dealloying Cu from an arc-melted product in sulfuric acid [21] and by reduction of freeze-dried metal oxides in 8% H 2 /Ar gas mixture [32]. The SEM images in [21,32] display the aggregates of the spherical particles similar to those observed in the current work.
Hydrogen 2020, 1, FOR PEER REVIEW 5 The reader may have noticed from the experimental part that both Co and Mo used as controls were prepared by reduction in H2/Ar stream from Co(OH)2 and MoO3 (rather than using elemental Co and Mo purchased from the suppliers of chemicals). The rationale behind this approach was based on an assumption that it would provide a more adequate comparison with Co7Mo6 and Co3Mo, which were made under reductive conditions from CoMoO4 and Co(OH)2. Both Co and Mo are single phase products according to PXRD (Figures S3-S4).
The morphology of Co3Mo from the SEM (Figure 2a and Figure S5 for a higher magnification image) could be described as consisting of spherical-shaped particles with a diameter of ca. 300-500 nm. The microstructure of Co3Mo is substantially different from the morphology of CoMoO4 precursor, which consisted of well-defined prismatic microcrystals ( Figure S6). On the other hand, there is a certain degree of similarity between the morphology of Co3Mo synthesized in this work and Co3Mo prepared on Cu-substrate by dealloying Cu from an arc-melted product in sulfuric acid [21] and by reduction of freeze-dried metal oxides in 8% H2/Ar gas mixture [32]. The SEM images in [21,32] display the aggregates of the spherical particles similar to those observed in the current work.
Co7Mo6 showed similar microstructure to Co3Mo, although with less-defined crystallite shapes and less apparent surface roughness (Figure 2b). This similarity between Co3Mo and Co7Mo6 surfaces is even more pronounced at lower magnification ( Figure S7), with Co3Mo appearing more porous. The difference is unsurprising, given that the higher synthetic temperature was used for Co7Mo6, which could have led to sintering and coalescence of particles [32]. Energy dispersive X-ray spectroscopy (EDXS) confirmed that Co3Mo and Co7Mo6 are homogenous ( Table 1). The average compositions are in line with the expected theoretical compositions within the error of the measurements. The higher Mo content within the samples could be explained by the overlap between Cu (0.93 eV, Lα) and Co (0.76 eV, Lα), peaks which lead to underestimation of Co peak areas ( Figure S8). The Cu-peak was due to the X-ray beam penetrating down to Cu foil, which was used as a support for drop-casted sample in EDXS characterization. After establishing the purity of the samples, linear sweep voltammetry (LSV) was used to investigate electrocatalytic activity of Co3Mo and Co7Mo6 (Figure 3a). At the current density of 10 mA cm -2 (which has been suggested as a useful benchmark in conjunction with photovoltaic applications Co 7 Mo 6 showed similar microstructure to Co 3 Mo, although with less-defined crystallite shapes and less apparent surface roughness (Figure 2b). This similarity between Co 3 Mo and Co 7 Mo 6 surfaces is even more pronounced at lower magnification ( Figure S7), with Co 3 Mo appearing more porous. The difference is unsurprising, given that the higher synthetic temperature was used for Co 7 Mo 6, which could have led to sintering and coalescence of particles [32].
Energy dispersive X-ray spectroscopy (EDXS) confirmed that Co 3 Mo and Co 7 Mo 6 are homogenous ( Table 1). The average compositions are in line with the expected theoretical compositions within the error of the measurements. The higher Mo content within the samples could be explained by the overlap between Cu (0.93 eV, L α ) and Co (0.76 eV, L α ), peaks which lead to underestimation of Co peak areas ( Figure S8). The Cu-peak was due to the X-ray beam penetrating down to Cu foil, which was used as a support for drop-casted sample in EDXS characterization.  (Figure 3a). At the current density of 10 mA cm −2 (which has been suggested as a useful benchmark in conjunction with photovoltaic applications [8]), the overpotential values of 115 ± 8 mV and 160 ± 5 mV are achieved for Co 3 Mo and Co 7 Mo 6, respectively (Figure 3a). At the same current density, Mo and Co show significantly higher overpotentials of 373 ± 7 mV and 411 ± 5 mV, respectively. Furthermore, upon application of a reductive bias in 0.5 H 2 SO 4 both alloys achieved much higher current densities ( Figure S9) than elemental Co and Mo powders. The observed overpotentials confidently place both Co 3 Mo and Co 7 Mo 6 among the best catalysts for HER from water in acidic electrolytes [8,14,15].
Hydrogen 2020, 1, FOR PEER REVIEW 6 [8]), the overpotential values of 115 ± 8 mV and 160 ± 5 mV are achieved for Co3Mo and Co7Mo6, respectively (Figure 3a). At the same current density, Mo and Co show significantly higher overpotentials of 373 ± 7 mV and 411 ± 5 mV, respectively. Furthermore, upon application of a reductive bias in 0.5 H2SO4 both alloys achieved much higher current densities ( Figure S9) than elemental Co and Mo powders. The observed overpotentials confidently place both Co3Mo and Co7Mo6 among the best catalysts for HER from water in acidic electrolytes [8,14,15]. Tafel plots were studied to investigate the possible kinetics of the reaction (Figure 3b). However, the resulting Tafel slope values of 71 ± 7 mV dec -1 (Co3Mo) and 84 ± 5 mV dec -1 (Co7Mo6) are outside the values for defined reaction mechanisms, i.e., 40 mV dec -1 (Volmer-Heyrovsky) and 120 mV dec -1 (Volmer) [8,33]. The deviation from ideal Tafel slopes is a common situation with powdered materials and Co3Mo in particular [8,20,33], as they do not always follow perfect charge-transfer behaviour (Tafel behaviour) due to mass transport effects [34]. Proper accounting for mass transport effects requires comprehensive numerical simulation procedures that are beyond the scope of this manuscript. An initial assessment of morphology by SEM indicated a denser Co7Mo6 sample. This could lead to a relative underperformance of Co7Mo6 in comparison with more porous Co3Mo due to a lower surface area, which leads to a lower specific activity [34]. We attempted measurements of surface area using N2 gas adsorption. However, the sorption characteristics of both samples were below the detection limit of the instrument due to the low surface area of the samples. An alternative option would be to investigate the surface area using Kr gas (instead of N2), which was shown as an efficient approach in the case of Ni-Mo alloys [35]. However, our group lacks access to such equipment. Therefore, we used values of double-layer capacitance (CDL) for the comparison of the relative surface areas between the two alloys. CDL has been deemed to be proportional to roughness factor and therefore is routinely used as an approximation for the electrochemical surface area of a sample in the absence of direct physical data [36][37][38]. The CDL values were evaluated at a constant potential from the difference of current densities with the scan rate (Figure 4a).
The ratio between CDL(Co3Mo):CDL(Co7Mo6) = 2.7 indicates that the surface area of Co3Mo could be three times higher than that of Co7Mo6, although in the absence of direct measurements of surface area (such as gas sorption) this value should be treated with great caution [39]. For example, Csernica et al. noticed that although the ratio between Kr BET surface area of their Mo7Ni7 and Ni0.92Mo0.08 varied almost by a factor of 4, the CDL ratio between these sample was only 1.6 [35]. Still, in the absence of Kr BET data, the comparison between CDL numbers appears as a good approximation of surface areas (as it seems to give numbers of the same magnitude as BET data). The comparison of CDL also corroborates with SEM results, which pointed out that Co3Mo is more porous than Co7Mo6. Therefore, Tafel plots were studied to investigate the possible kinetics of the reaction (Figure 3b). However, the resulting Tafel slope values of 71 ± 7 mV dec −1 (Co 3 Mo) and 84 ± 5 mV dec −1 (Co 7 Mo 6 ) are outside the values for defined reaction mechanisms, i.e., 40 mV dec −1 (Volmer-Heyrovsky) and 120 mV dec −1 (Volmer) [8,33]. The deviation from ideal Tafel slopes is a common situation with powdered materials and Co 3 Mo in particular [8,20,33], as they do not always follow perfect charge-transfer behaviour (Tafel behaviour) due to mass transport effects [34]. Proper accounting for mass transport effects requires comprehensive numerical simulation procedures that are beyond the scope of this manuscript.
An initial assessment of morphology by SEM indicated a denser Co 7 Mo 6 sample. This could lead to a relative underperformance of Co 7 Mo 6 in comparison with more porous Co 3 Mo due to a lower surface area, which leads to a lower specific activity [34]. We attempted measurements of surface area using N 2 gas adsorption. However, the sorption characteristics of both samples were below the detection limit of the instrument due to the low surface area of the samples. An alternative option would be to investigate the surface area using Kr gas (instead of N 2 ), which was shown as an efficient approach in the case of Ni-Mo alloys [35]. However, our group lacks access to such equipment. Therefore, we used values of double-layer capacitance (C DL ) for the comparison of the relative surface areas between the two alloys. C DL has been deemed to be proportional to roughness factor and therefore is routinely used as an approximation for the electrochemical surface area of a sample in the absence of direct physical data [36][37][38]. The C DL values were evaluated at a constant potential from the difference of current densities with the scan rate (Figure 4a). the relative surface area of Co3Mo is 2.7 times higher than Co7Mo6. Thus, the specific activity of Co7Mo6 (mA cm -2 catalyst) is 2.7 times higher. In other words, based on this approximation Co7Mo6 can deliver 2.7 times higher current densities, as demonstrated in Figures S10 and S11, which also leads to an improved overpotential of 123 mV at 10 mA cm -2 . This is an interesting result, especially given that recent DFT calculations on Co3Mo have shown that Mo-atoms are the catalytic sites in Co3Mo [20]. One can assume that increasing the amount of Mo would lead to improved catalytic performance. However, due to differences in crystal structures of Co3Mo and Co7Mo6, it is hard to draw definitive conclusions at this stage and further computational studies are required. The stability of the catalysts was tested by applying a constant potential over time (Figure 4b).
There is a clear decline in current densities for Co3Mo, while Co7Mo6 retained the original values over 16 hours of testing without any evidence of decay. Co7Mo6 also shows a good stability upon cycling for 1000 times to the current densities of over 100 mA cm -2 at a scan rate of 100 mV s -1 ( Figure S12). The stability of Co3Mo was additionally investigated by powder XRD with the product measured before and after electrolysis directly on a glassy carbon electrode ( Figure S13). The disappearance of a peak at ca. 43.8° and significant reduction of peak at ca. 46.3° suggest the decomposition of the product. We can hypothesize that the prolonged electrolysis leads to formation of metal hydrides and dissolution of the product based on a previous report for Ni-based catalysts [40].
To the best of our knowledge, this is the first report of the electrocatalytic performance of Co7Mo6, and therefore, it is important to prove that the LSV measurements correspond to the actual reduction of protons to hydrogen. A representative trace of gas produced in an airtight cell and probed by gas chromatography is shown in Figure S14, which confirms the production of hydrogen using Co7Mo6.
There are currently no electrochemical studies of Co3Mo in acidic conditions, and to the best of our knowledge no studies of Co7Mo6 in acidic or alkaline media. Therefore, to put the results into perspective we studied Co3Mo and Co7Mo6 by LSV in 1 M KOH ( Figure S15). The alloys show overpotentials of 120 ± 5 mV (Co3Mo) and 160 ± 5 mV (Co7Mo6) at 10 mA cm -2 . In comparison with recently reported Co3Mo on nickel foam, the observed overpotential seems high [20]. Chen et al. reported the overpotentials of 78 mV (CDL = 15.8 mF cm -2 ) and 75 mV (CDL = 17.5 mF cm -2 ). However, the CDL = 3.2 mF cm -2 observed for Co3Mo in this work (based on double-layer capacitance in 1M KOH) is lower, and this could explain the higher overpotential ( Figure S16). The comparison also points out the limitation of using double-layer capacitance (CDL) as a proxy for the catalytic surface area. Without direct measurement of surface area, it is hard to tell whether it is the substrate or other factors (such as substantially higher surface area due to nanostructuring) that led to such a difference in the overpotential values. Still CDL is a useful parameter, as it at least provides some background information that enables a comparison between samples prepared in different labs. For example, The ratio between C DL (Co 3 Mo):C DL (Co 7 Mo 6 ) = 2.7 indicates that the surface area of Co 3 Mo could be three times higher than that of Co 7 Mo 6 , although in the absence of direct measurements of surface area (such as gas sorption) this value should be treated with great caution [39]. For example, Csernica et al. noticed that although the ratio between Kr BET surface area of their Mo 7 Ni 7 and Ni 0.92 Mo 0.08 varied almost by a factor of 4, the C DL ratio between these sample was only 1.6 [35]. Still, in the absence of Kr BET data, the comparison between C DL numbers appears as a good approximation of surface areas (as it seems to give numbers of the same magnitude as BET data). The comparison of C DL also corroborates with SEM results, which pointed out that Co 3 Mo is more porous than Co 7 Mo 6 . Therefore, given the fact that the mass loadings and electrode surface area were identical, we can assume that the relative surface area of Co 3 Mo is 2.7 times higher than Co 7 Mo 6 . Thus, the specific activity of Co 7 Mo 6 (mA cm −2 catalyst ) is 2.7 times higher. In other words, based on this approximation Co 7 Mo 6 can deliver 2.7 times higher current densities, as demonstrated in Figures S10 and S11, which also leads to an improved overpotential of 123 mV at 10 mA cm −2 . This is an interesting result, especially given that recent DFT calculations on Co 3 Mo have shown that Mo-atoms are the catalytic sites in Co 3 Mo [20]. One can assume that increasing the amount of Mo would lead to improved catalytic performance. However, due to differences in crystal structures of Co 3 Mo and Co 7 Mo 6, it is hard to draw definitive conclusions at this stage and further computational studies are required.
The stability of the catalysts was tested by applying a constant potential over time (Figure 4b). There is a clear decline in current densities for Co 3 Mo, while Co 7 Mo 6 retained the original values over 16 h of testing without any evidence of decay. Co 7 Mo 6 also shows a good stability upon cycling for 1000 times to the current densities of over 100 mA cm −2 at a scan rate of 100 mV s −1 ( Figure S12). The stability of Co 3 Mo was additionally investigated by powder XRD with the product measured before and after electrolysis directly on a glassy carbon electrode ( Figure S13). The disappearance of a peak at ca. 43.8 • and significant reduction of peak at ca. 46.3 • suggest the decomposition of the product. We can hypothesize that the prolonged electrolysis leads to formation of metal hydrides and dissolution of the product based on a previous report for Ni-based catalysts [40].
To the best of our knowledge, this is the first report of the electrocatalytic performance of Co 7 Mo 6, and therefore, it is important to prove that the LSV measurements correspond to the actual reduction of protons to hydrogen. A representative trace of gas produced in an airtight cell and probed by gas chromatography is shown in Figure S14, which confirms the production of hydrogen using Co 7 Mo 6 .
There are currently no electrochemical studies of Co 3 Mo in acidic conditions, and to the best of our knowledge no studies of Co 7 Mo 6 in acidic or alkaline media. Therefore, to put the results into perspective we studied Co 3 Mo and Co 7 Mo 6 by LSV in 1 M KOH ( Figure S15). The alloys show overpotentials of 120 ± 5 mV (Co 3 Mo) and 160 ± 5 mV (Co 7 Mo 6 ) at 10 mA cm −2 . In comparison with recently reported Co 3 Mo on nickel foam, the observed overpotential seems high [20]. Chen et al. reported the overpotentials of 78 mV (C DL = 15.8 mF cm −2 ) and 75 mV (C DL = 17.5 mF cm −2 ). However, the C DL = 3.2 mF cm −2 observed for Co 3 Mo in this work (based on double-layer capacitance in 1M KOH) is lower, and this could explain the higher overpotential ( Figure S16). The comparison also points out the limitation of using double-layer capacitance (C DL ) as a proxy for the catalytic surface area. Without direct measurement of surface area, it is hard to tell whether it is the substrate or other factors (such as substantially higher surface area due to nanostructuring) that led to such a difference in the overpotential values. Still C DL is a useful parameter, as it at least provides some background information that enables a comparison between samples prepared in different labs. For example, Co 3 Mo prepared by arc-melting showed an overpotential of 340 mV at 10 mA cm −2 in 1M KOH [41]. However, since this work is missing C DL values, it is difficult to explain the high value of overpotential. Conversely, nanoporous Co 3 Mo on Cu substrate achieved 65 mV cm −2 at 100 mA cm −2 [21] but the material had a C DL = 324 mF cm −2 . The Tafel slope of 40 mA cm −2 suggests an overpotential of just 25 mV cm −2 at 10 mA cm −2 in 1 M KOH. However, given a large C DL it appears that the overpotential may be underestimated. Therefore, it is often difficult to decouple the role of the substrate and morphology from the actual performance of the material, and more research on free-standing catalysts is needed. Remarkably, Co 7 Mo 6 achieved substantially higher current densities in 1 M KOH (well in excess of 120 mA cm −2 ) than in 0.5 H 2 SO 4 ( Figure S17). It also outperformed, in this respect, Co 3 Mo as well as Co and Mo powders.
We also studied Tafel plots of Co 3 Mo and Co 7 Mo 6 ( Figure S18), which showed the Tafel slopes of 117 mV dec −1 (Co 3 Mo) and 106 mV dec −1 (Co 7 Mo 6 ). The values suggest that the electron transfer reaction leading to H-adsorption on the surface of the catalyst is the rate determining step [8,33]. The Tafel slope for Co 3 Mo is broadly consistent with arc-melted Co 3 Mo (105 mV dec −1 ) [41] and Co 3 Mo on Ni-foam [20], although it should be noted that in [20] the Tafel slopes varied quite significantly depending on the synthesis reaction times and C DL values. Conversely, nanostructured Co 3 Mo on Cu substrate showed substantially different Tafel slope of 40 mV dec −1 [21]. This reinforces the fact that the morphology and support material could play a vital role (as well as adding an extra layer of complexity) when it comes to comparison between products made in different labs.
Finally, we also evaluated the stability of free-standing Co 3 Mo in 1M KOH to see how it compared with the products prepared on metal supports ( Figure S19). Contrary to the previous reports [20,21] (which reported good stability of Co 3 Mo on Ni-foam and Cu substrate), we observed a clear fading of the current with time. This suggests that substrate plays a significant role. Therefore, this work (of free-standing systems decoupled from the support) provides an important study that should provide a better general understanding of the actual catalyst's performance.

Conclusions
In conclusion, phase-pure Co 3 Mo and Co 7 Mo 6 can be prepared through the reaction of stoichiometric amounts of Co(OH) 2 and CoMoO 4 . The evaluation of electrochemical performance points out that Co 3 Mo in the free-standing form could routinely show good catalytic performance similar to previously studied classes of materials in acidic conditions. To the best of our knowledge, this is the first report on the activity of Co 7 Mo 6 towards HER either in acidic or alkaline conditions. When adjusted for the C DL ratios, Co 7 Mo 6 demonstrated improved current densities and an overpotential of 123 mV at 10 mA cm −2 tentatively, suggesting that it can be as electrocatalytically active as Co 3 Mo while showing a better stability in acidic media. However, without accounting for surface areas by other methods, these results should be treated with caution. This highlights a set of challenges for researchers working on nanostructured catalysts. The issue is quite evident from the comparison between free-standing Co 3 Mo investigated in this work and Co 3 Mo fabricated on Ni-foam or Cu substrates when tested in 1 M KOH. Therefore, the estimation of surface area by direct physical methods (i.e., through gas adsorption which is independent of sample history, oxidation state and solvent effects) is strongly recommended for an adequate comparison between products made in different labs. Computational studies are also promising for clarification whether the difference in comparative performance of Co 3 Mo and Co 7 Mo 6 fundamentally persists, and these will be explored in our future work. In particular, the studies of the mechanisms of charge-transfer and proton reduction behavior are considered as the most appropriate.
Supplementary Materials: The following are available online at http://www.mdpi.com/2673-4141/1/1/2/s1, Figure S1: PRXD bracket for testing of catalyst directly on the electrode. Figure S2: Electrochemical cell used for hydrogen production measurements. Figure S3: PXRD patterns of Mo. Figure S4: Powder XRD patterns of Co. Figure S5: Higher magnification SEM image of Co 3 Mo. Figure S6: SEM image of CoMoO 4 precursor. Figure S7: Low magnification SEM image of Co 3 Mo and Co 7 Mo 6 . Figure S8: A representative EDXS spectrum of Co 3 Mo sample. Figure S9: Comparison of current densities achieved by Co 3 Mo, Co 7 Mo, Co and Mo. Figure S10: Comparison of current densities achieved by Co 3 Mo, Co 7 Mo 6 and Co 7 Mo 6 when adjusted for results of double-layer capacitance (C DL ). Figure S11: Magnified version of Figure S10. Figure S12: Comparison of current densities achieved by Co 7 Mo 6 in 0.5M H 2 SO 4 before and after 1000 scans at the scan rate of 100 mv s −1 . Figure S13: Comparison of XRD pattern achieved by Co 3 Mo in 0.5M H 2 SO 4 before and after 1000 scans at the scan rate of 100 mv s −1 . Figure S14: A representative trace of the proportion of H 2 (in vol. %) in the single-cell headspace during the electrolysis of Co 7 Mo 6 . Figure S15: Comparison of current densities achieved by Co 3 Mo, Co 7 Mo 6 as well as Mo and Co powders in 1M KOH. Figure S16: Evaluation of double-layer capacitance of Co 3 Mo and Co 7 Mo 6 achieved in 1 M KOH. Figure S17: Extended range of the Graph in Figure S15. Figure S18: Tafel plots and corresponding Tafel slopes of Co 3 Mo and Co 7 Mo 6 in 1M KOH. Figure