Best Practice for Evaluating Electrocatalysts for the Hydrogen Economy

Screening new electrocatalysts is key to the development of new materials for next-generation energy devices such as fuel cells and electrolysers. The counter electrodes used in such tests are often made from materials such as Pt and Au, which can dissolve during testing and deposit onto test electrocatalysts, resulting in inaccurate results. The most common strategy for preventing this effect is to separate the counter electrode from the test material using an ion-transporting Nafion membrane. Here, we use X-ray photoelectron spectroscopy, energy-dispersive X-ray analysis, mass spectrometry, and voltammetry to demonstrate the limitations of this approach during constant-current, extended stability testing of electrocatalysts for H 2 evolution. We show that Nafion membranes cannot prevent contamination of carbon electrocatalysts by Pt and Au counter electrodes, leading to an apparent increase in electrocatalytic activity of the carbon. We then demonstrate that carbon counter electrodes in undivided cells can contaminate and deactivate Pt and Au electrocatalysts for H 2 evolution. We show that use of a setup comprising a glass frit separating a carbon counter electrode from the test electrocatalyst can prevent these effects. Finally, we discuss these phenomena using H 2 evolution at MoS 2 and at a K 6 [P 2 W 18 O 62 ](H 2 O) 14 /carbon nanotube composite as test reactions.


INTRODUCTION
The development of earth-abundant electrocatalysts is a key driver in the development of nextgeneration hydrogen fuel cells and water electrolysers -devices that are expected to play key roles in the emerging hydrogen economy. The development of such systems requires that methods for accurately measuring the activity of electrocatalysts are available, and the method-of-choice is rotating-disk electrode voltammetry. [1][2][3][4][5][6][7][8][9] Glassy carbon (GC) is usually used as the substrate for electrocatalyst testing, due to its wide electrochemical window and low electrocatalytic activity towards reactions such as the oxygenevolution reaction (OER), oxygen-reduction reaction (ORR), hydrogen-evolution reaction (HER), and hydrogen-oxidation reaction (HOR). 6,10,11 Counter electrodes in electrochemical cells are usually fabricated from materials such as Pt and Au, due to their perceived electrochemical inertness. 12 Indeed, Pt has been recommended as a counterelectrode material in guides to electrochemistry and electrocatalysis. 2,9,13 However, if the potential of the counter electrodes becomes sufficiently high during analysis, dissolution of the metal can occur. 14 This effect is exacerbated when the potential of the test electrocatalyst is cycled -a common method for preactivating electrocatalysts 2 and examining their long-term stability. 3,8,15 For example, if the potential of a Pt electrode is cycled between those at which the ORR and OER occur, repeated reduction of electrogenerated Pt-O species exposes readily-solubilized, low-coordinate Pt. 12,14,[16][17][18][19][20] Dissolved Pt ions can reach and become deposited on the electrocatalyst surface, increasing the apparent activity of the test material. 3,7,8,12,[21][22][23][24][25][26] Increasing the surface area of the counter electrode decreases the current density it experiences, and while this can slow the dissolution process, it cannot stop it. 27 Dissolved Clions (either from the supporting electrolyte or leaking from a reference electrode) can also increase the rate of dissolution of Pt and Au electrodes through the formation of metal-chloride complexes. 12,25 One of the most common strategies for mitigating the effects of dissolution of Pt counter electrodes is to separate the test electrocatalyst from the counter electrode using a Nafion membrane. 12,22,[27][28][29] However, a search of the literature does not reveal strong evidence for the effectiveness of this strategy. There is, on the other hand, evidence that Nafion is permeable to dissolved Pt ions. For example, Pt ions dissolving from fuel-cell cathodes can enter and deposit in Nafion proton-exchange membranes during operation. [30][31][32][33] Transport of dissolved Pt ions through Nafion membranes has even been used to fabricate membrane-electrode assemblies. 34 Perhaps the most obvious solution to the problem of counter-electrode dissolution is to use non-metal counter electrodes, such as graphite rods. [3][4][5]7,8,11,24,35,36 Some work has probed the effectiveness of this strategy but, as with earlier studies on Pt dissolution, it was probed using extended potential-cycling tests. 8,24 Real devices would not experience the kind of potential excursions used in accelerated-stability tests and are expected to operate at or near constant currents or potentials.
Consequently, testing the long-term stability of electrocatalysts at constant currents is also important. 5 In this contribution, we first show that Pt counter electrodes contaminate and activate GC electrocatalysts during constant-current HER in undivided cells containing acidic electrolytes. After 24 h, the electrocatalytic activity of the contaminated GC increases such that it is similar to that of pure Pt. We then show that use of a Nafion membrane to separate Pt and Au counter electrodes from the GC is ineffective; microscopic and spectroscopic analysis shows that GC surfaces become covered with Pt and Au after 24 h of HER electrolysis. Electrochemical analysis shows that the use of carbon counter electrodes in undivided cells can lead to the opposite problem; deposition of particulate carbon onto Pt and Au electrocatalysts reduces their electrocatalytic activity for the HER. We show that these problems can be avoided by using a carbon counter electrode separated from the test electrocatalyst by a glass frit. We discuss this Electrochemical testing was carried out using a Model CHI760C potentiostat (CH Instruments) coupled with a modulated speed rotator from Pine Research. X-ray photoelectron spectroscopy (XPS) was carried out using a Kratos AXIS DLD instrument equipped with an Al Kα X-ray source (1486.6 eV).
CasaXPS software was used with Kratos sensitivity factors to determine atomic concentrations. A Shirley background correction was applied to all spectra prior to analysis. All spectra were charge corrected to  just 0.125 V more negative than when using a pure Pt electrocatalyst for the HER, 5 demonstrating the drastic activation of the GC that occurred over the test period. SEM images of the GC surface after the stability test in which a Pt electrode was used ( Figure 1C) revealed bright 700-1600 nm spots on the GC surfaces. The spots were identified as Pt islands using EDX analysis.

Contamination of Working Electrodes in Undivided
Au dendrites were observed on the GC surface ( Figure 1D)  to decrease the onset potential for the HER to about -0.5 V and -0.4 V, respectively. The gray line of Figure 2C shows that use of the glass frit resulted in a larger decrease of the HER onset potential to about -0.2 V, due to the transport of more dissolved Pt across the frit. The HER overpotential at −10 mA cm −2 was −0.42 V and −0.67 V after using the glass frit and Nafion membrane, respectively, to separate the GC 8 from the Pt counter electrode. Use of the Nafion membrane to separate the Au counter electrode from the GC resulted in an HER overpotential at -10 mA cm −2 of -0.54 V after 24 h of electrolysis. These data prove that Nafion membranes, while more effective than glass frits, could not prevent contamination and activation of the GC surfaces by dissolved Pt and Au from counter electrodes during the HER.   Figures 4A,B), and EDX analysis confirmed they were Pt ( Figure 4C). The presence of the metals on the Nafion can be attributed to reduction of metal ions crossing the membranes by H2 generated at the GC surfaces over 24 h. A similar phenomenon has been observed during "post-mortem" analysis of Nafion membranes from proton-exchange membrane fuel cells. 41,42 Oxidation of Pt at fuel-cell cathodes releases mobile Pt 2+ , which is reduced in the Nafion by H2 supplied to the anode. That significantly less Pt than Au was detected on our membranes is presumably due to a lower concentration and/or slower reduction of dissolved Pt 2+ ions in our systems. Indeed, XPS analysis of 3 random spots on the membrane surface failed to identify any Pt, due to its localization on the membrane surface (as evident from Figure 4A

Use of Carbon Counter Electrodes to Mitigate against Metal Contamination.
Given the problems associated with the use of metal counter electrodes, avoiding the use of metal seems a logical solution. It is generally thought that the use of graphite counter electrodes has no (or very little) influence on the activity of test electrocatalysts 8,22,24 and, consequently, this has been proposed as best practice for testing new materials. [3][4][5]7,8,11,24,35,36 However, as far as we can tell, no study has specifically reported on the use of graphite counter electrodes for extended constant-current HER tests, though some report the  We first tested the effect of using of a graphite counter electrode in an undivided cell containing a polished Pt electrocatalyst for the HER. The black dashed line in Figure 5A shows the voltammogram recorded at the beginning of the test period. The HER onset potential was about 0.0 V, as expected for a Pt electrocatalyst. The solid black line shows the voltammogram recorded after 24 h. Remarkably, the magnitude of the HER overpotential at -10 mA cm -2 increased by more than 0.1 V over the test period, demonstrating that the electrocatalytic activity of the Pt surface decreased over the test period. Microscopic analysis of the surface of Pt after the stability test revealed that the surface had become contaminated with carbonaceous spots (Figure 6). That we observe a large apparent decrease in the activity of Pt over 24 h of electrolysis using a carbon counter electrode, and such an effect has not been reported previously, may be a result of the HER occurring at a constant current density during the aging period, rather than by potential cycling.  We then tested the effect of using a Nafion separator to isolate the carbon counter electrode from the Pt, and the resulting voltammogram is also presented in Figure 5A (gray solid line, which overlaps with the dashed line). No change in activity of the Pt electrode was observed during the test, indicating that the membrane had effectively prevented the crossover of contaminants from the graphite counter electrode. We also tested the effectiveness of using a glass frit to separate the carbon counter electrode from Pt and Au electrocatalysts, a common approach to electrocatalyst testing 5 (though to the best of our knowledge no published data has demonstrated the usefulness of this strategy). Figure 5B shows that after 24 h of HER, the voltammograms recorded using each electrocatalyst almost overlaid each other, demonstrating that the glass frit could effectively prevent deactivation of the electrocatalysts when graphite counter electrodes were used. We also investigated whether it was possible to use a GC rod instead of graphite as counter electrode in an undivided cell. In this case, a glass frit was also required to prevent contamination and deactivation of an Au electrocatalyst for the HER, a factor that could be attributed to flaking of the GC electrodes due to oxidation of graphitic domains. 43

Electrocatalysis of H2 Evolution at MoS2 and K6[P2W18O62](H2O)14@SWNT. The implications
of our observations were examined using MoS2, which has been used as an earth-abundant HER electrocatalyst. 44,45 Testing and optimising such high-surface-area, earth-abundant electrocatalysts is of significant interest to those interested in developing cost-effective materials for the H2 generation. HER voltammograms were recorded using MoS2 deposited onto a GC electrode before and after 24 h of constant-current HER using a Pt and graphite counter electrode (Figure 7). The graphite rod counter electrode was used with and without a glass frit separating it from the MoS2. Before the extended stability test, the HER onset potential was about -0.25 V, which is typical of that expected for this electrocatalyst, 46   This material has been described recently and represents a class of materials that offer unique opportunities for electrochemical applications. 47 A drop-cast film of this material on a GC electrode was cycled in Ar-purged 1.0 mol dm −3 HCl at 100 mV s −1 . The black voltammogram in Figure 8A shows 3 sets of redox waves in the potential region between −0.1 and −0.7 V, attributable to the following redox couples (from most positive to most negative): The bell shape of the voltammetric peaks in Figure 8 can be attributed to confinement of the composite on the electrode surface (that is, the peaks show non-diffusional responses). After fewer than 200 electrochemical cycles in an undivided cell containing a Pt counter electrode ( Figure 8A), the peak currents decreased due to slow loss of the material from the electrode. However, by the 300 th cycle, a large increase in cathodic current appeared at potentials negative of about -0.6 V. This effect has been observed previously during solution-phase electrochemistry of POMs and was originally attributed to the formation of a POM-derived HER electrocatalyst upon potential cycling. 48,49 It has since been shown that this increase in electrocatalytic HER activity was due to migration of dissolved Pt from Pt counter electrodes onto the electrode surface, which catalyzed the HER. 26 Zhang et al. have used this phenomenon to produce highly active silicotungstate POM electrocatalysts. 40 We repeated the cycling experiments using our K6[P2W18O62](H2O)14@SWNT composite electrode and a GC counter electrode instead of a Pt counter electrode ( Figure 8B), and found that the HER current at potentials negative of −0.6 V no longer appeared. The only change in the voltammograms was the decrease in the peak areas, due to loss of the active POMs from the electrode. Furthermore, we found that after 330 cycles in the cell containing the Pt counter electrode, there was significant H2 bubble formation on the electrode surface due to the HER, preventing meaningful voltammograms from being recorded. In contrast, when a GC counter electrode was used no bubble formed at the electrode surface, further demonstrating that the use of a carbon electrode in this system was necessary to avoid the problems associated with the use of Pt counter electrodes. to isolate test electrocatalysts was demonstrated using a MoS2 electrocatalyst for H2 evolution. We also showed that potential-cycling of polyoxometalate-based composite electrodes is sensitive to contamination by Pt from counter electrodes, but this can be prevented using a carbon counter electrode. We recommend that divided cells containing carbon-based counter electrodes are used when analysing the performance of electrocatalysts for H2 evolution, to avoid the reporting of erroneously high or low performance metrics for electrocatalysts for H2 evolution. This approach will save time and resources as we face the important challenge of finding new materials for the emerging hydrogen economy.

AUTHOR CONTRIBUTIONS
The manuscript was written by all authors. All authors have given approval to the final version of the manuscript.