Determining the Viability of Hydroxide-Mediated Bifunctional HER/HOR Mechanisms through Single-Crystal Voltammetry and Microkinetic Modeling

The slow kinetics of the hydrogen oxidation and hydrogen evolution reactions (HER and HOR) in alkaline compared to acidic media remain a fundamental conundrum in modern electrocatalysis. Recent efforts have proposed that OH, as well as H, must bind optimally for improved kinetics, but the exact role of adsorbed OH is not yet known. In this work, we combine steady-state single-crystal voltammetry and microkinetic modeling to determine the roles of adsorbed hydroxide and the so-called bifunctional mechanism in alkaline HER and HOR kinetics. We consider both a direct Volmer mechanism, in which H and OH compete for sites on Pt (110), and an OH-mediated mechanism, in which Pt (111) adsorbs H while transition metal clusters adsorb OH. Our experimental and computational results show that on a thermodynamic coverage basis, increasing OH adsorption strength cannot promote faster HER/HOR kinetics. Only changes to the kinetic rate constants can explain experimental observations. We speculate that adequate electrocatalyst design in alkaline media additionally requires manipulation of interfacial water structure to lower energetic barriers for HER and HOR. © The Author(s) 2018. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0271815jes]

The hydrogen evolution and oxidation reactions (HER/HOR) are known to be several orders of magnitude slower in basic than acidic electrolytes, but the reasons behind such pH dependence are not yet understood. Identifying the root of this phenomenon is key to the rational design of catalysts for many applications, including CO 2 reduction, oxygen evolution, and nitrogen electrochemistry. Beyond its practical importance, the pH dependence of HER/HOR kinetics is one of the long-standing fundamental conundrums in electrochemistry.
There have been several proposed explanations for the effect of pH on hydrogen electrocatalysis. Traditionally, the hydrogen binding energy (HBE) alone has been used to describe HER/HOR kinetics. 1,2 One hypothesis is that OH − stabilizes the Pt-H bond, 3,4 leading to stronger H binding and slower catalysis. This hypothesis is based on a measured shift with pH in peak potential for the hydrogen underpotential deposition (H-UPD) on Pt (110) and Pt (100) sites. However, a similar shift is not seen on Pt (111) sites, even though a drop in HER/HOR activity is still observed. [5][6][7] Furthermore, weak-binding catalysts (such as gold-group metals) do not experience improved activity with higher pH, 8,9 as the HBE mechanism predicts.
Another explanation for the pH dependence of HER/HOR kinetics considers OH adsorption more explicitly. Multiple groups have shown that the H-UPD peaks on Pt (110) and (100) are instead representative of hydrogen/hydroxide exchange (H/OH-X). [10][11][12] The presence of alkali cations weakens the OH adsorption strength, and so the shift with pH in the H-OH/X peak is therefore instead caused by weaker competitive OH adsorption, allowing H to remain adsorbed up to higher potentials. The corresponding direct Volmer mechanism, in which H and OH compete for sites, is shown in Figure 1. Related to this phenomenon, the Markovic group reported an improvement in HOR when using Pt 0.1 Ru 0.9 over bare Pt (111), and in HER when doping Pt (111) with Ni(OH) 2 clusters. [13][14][15][16] The improved reaction kinetics were attributed to the introduction of oxophilic sites, which were hypothesized to reduce the energetic cost of dissociating water. In the resulting bifunctional mechanism, both OH ad and the required H ad intermediate are formed on their respective sites, as shown in Figure 2. Li et al. 17 provided spectroscopic evidence for the presence of OH ad on Ru sites. However, the presence of OH on the surface does not automatically infer participation in the reaction.
A third possibility, independent of OH adsorption, points not to thermodynamic adsorption energies, but to kinetic parameters caused by water orientation near the electrode interface. Recent experiments   show that, unlike in base, H/OH-X and HER in acid occur close to the potential of zero charge (PZC). 10,18 This observation implies that in alkaline media, water interacts strongly with the interfacial electric field and is therefore more difficult to reorganize and accommodate charge migration. The Koper group showed that Ni(OH) 2 clusters on Pt (111) shift the PZC in base by about −25 mV, 7 closer to that in acid. It was suggested that at the PZC, a more dynamic water structure would allow rapid solvent reorganization and facilitate charge transfer processes. 7,10 While the beneficial effects of transition metal (TM) clusters are experimentally confirmed, the contrasting explanations for their fundamental effect, as well as the asymmetric effects on HER and HOR, shows that the role played by OH ad is still unclear. Our previous work 19 has shown that on a single-site catalyst such as Pt (110), an OH-mediated mechanism (analogous to Rxns. 5-8, but in a Fig. 1 setting) is energetically unrealistic and qualitatively inconsistent with experimental trends. However, neither the possibility of a two-site mechanism nor the effects on overall HER/HOR were considered.
In this work, we specifically investigate the effect of OH adsorption strength on HER and HOR kinetics. We develop a microkinetic model of the alkaline hydrogen reaction and relate the OH coverage to reaction kinetics via its adsorption strength, G OH . We focus our study on two different reaction schemes: (1) a single-site model proceeding through the direct Volmer step ( Fig. 1) and (2) a bifunctional model proceeding through an OH-mediated water recombination/dissociation step (Fig. 2). The models are compared to the experimental systems of bare Pt (110) and a Pt (111) surface doped with oxophilic TM clusters, respectively. The single-site model is analogous to a bare Pt (110) facet because the overlapping H and OH adsorption energies result in competition for adsorption sites. In contrast, the benchmark for the bifunctional model is a Pt (111) surface because the H and OH adsorption energies are so different that competition for Pt surface sites does not occur on Pt (111). A direct Volmer step on a bifunctional catalyst is not considered as it provides no opportunity for adsorbed H and OH species to interact. Further, our previous work has shown that the bifunctional mechanism where adsorbed OH is an active participant is not energetically viable on a single-site catalyst. 19 Therefore, the microkinetic analysis was limited to the two reaction schemes described above. The predicted effect of stronger OH binding is compared with experiment to determine the role of adsorbed OH on reaction kinetics.

Experimental
Cell and electrolyte preparation.-Electrolyte solutions were prepared using KOH pellets (99.99% [metal basis], Sigma) or high purity HClO 4 (70 wt%, Suprapur). All electrochemical experiments were carried out in a FEP cell. Prior to its use, the FEP cell was soaked in 50% HNO 3 /50% H 2 SO 4 for 2 hours followed by repeated boiling and rinsing in Millipore water. A Pt wire was used as the counter electrode and a Ag/AgCl reference electrode (BASi) was used for the experiments. All potentials reported in this work are calibrated against the reversible hydrogen electrode (RHE) scale. (111) and (110) disks were annealed at 1100 • C in 3% H 2 /Ar flow for 10 minutes followed by cooling for 5 minutes in the same gas. Each disk was then protected with a drop of Millipore water before being mounted in the RDE holder. A drop of 0.05 M solution of perchloric salt of Ni, Mn or Co was drop-casted onto the disk for 5 minutes after which it was rinsed with Millipore water. The disk then transferred to the cell under potential control at 0.08 V.

Electrode preparation and TM deposition.-Pt
Electrochemical measurements.-The electrochemical measurements were made with an Autolab PGSTAT302N potentiostat in a rotating disk electrode (RDE) setup. The electrolyte in the cell was purged with Ar (Research grade, Airgas) and the potential was cycled between 0.08-0.9 V to obtain the cyclic voltammetry (CV) profiles. HER and HOR polarization curves were obtained through potential cycling between −0.3 -0.9 V in H 2 -purged electrolyte at a rotation rate of 1600 rpm. The CV profile and HER-HOR polarization curves were obtained in 0.1 M KOH. The coverage of the TM clusters on the Pt (111) surface was obtained by the difference in H-UPD between the bare Pt (111) and the disk decorated with the clusters (approx. 30-40% coverage). After the measurements, the Pt (111) disk was cycled between 0.1-1.1 V in 0.1 M HClO 4 to dissolve the TM clusters and clean the disk.

Microkinetic Model Development
Previous alkaline hydrogen reaction models have neglected the competition between H and OH, [20][21][22][23] or used only thermodynamic descriptors without considering kinetics. 12,24 Here, following our previous methodology, 19 we relate the equilibrium potential of elementary steps to the free energy of adsorption, while considering possible reaction mechanisms on single-site and bifunctional catalysts. For each of the reaction schemes, the hydrogen reaction may progress via either a Tafel-Volmer or Heyrovsky-Volmer mechanism. Assuming quasi-equilibrium for non-rate-determining steps (RDS) then yields eight possible mechanisms. The resulting rate laws are summarized in Tables I and II. For illustration, one mechanism is derived below.
Considering a bifunctional catalyst where water recombination is relatively fast and Heyrovsky is the RDS, the corresponding Butler-Volmer 25 rate expression is: Because there are two possible binding sites with exclusive adsorption, there is no site competition between H and OH and the fraction of empty sites is represented as 1-θ H instead of 1-θ H -θ OH . The activities of hydroxide, water, and hydrogen are assumed to be unity in this work. Following the previous model development, 19 the rate expression can be ultimately written as: The equations above represent the rate expression intrinsic to this reaction mechanism (Heyrovsky RDS). Here, A f is a rate constant with units of s −1 that includes an exponential activation energy dependence as well as a frequency factor with contributions from interfacial water structure and the time constant for solvent reorganization. [26][27][28][29] Adsorption energies are taken from literature while A f is used to fit simulations to experimental current densities measured at ± 0.1 V. The transfer coefficient β relates to the symmetry of the charge-transfer reaction barrier, and is assumed to be equal to 0.5 for all elementary steps. The theoretical maximum surface charge, Q tot , is 240 μC/cm 2 for Pt (111), 6 and is assumed to be independent of catalyst loading, although increasing the density of TM clusters will decrease Q tot . Assuming all other reactions involved are fast and in equilibrium (i.e. r OHads = r R = 0, and U OHads = V), the H and OH coverages can be expressed in terms of potential and binding energies using a similar Butler-Volmer approach. 25 Adsorption of OH is always assumed to be fast, due to the high concentration of OH − ions in solution. Thus: The overall change in free energy of OH adsorption is affected by the enthalpy and entropy of adsorption, changes in zero-point energy, and the configurational entropy. 2,24 The enthalpy of adsorption, zero-point energy, and entropy of adsorption can be calculated via DFT simulations, while the configurational entropy is given by the logarithmic term in Eq. 4. Again, competition is eliminated by representing the Volmer fast, Heyrovsky RDS Volmer RDS, Tafel fast Volmer fast, Tafel RDS fraction of empty sites as 1-θ OH . Solving for OH coverage thus gives: [5] Water recombination/dissociation is a chemical step, so by assuming it is in equilibrium (r R = 0), H coverage is obtained through the equilibrium constant: Substituting in θ OH and the equilibrium constant, and simplifying: Eq. 8 shows that, after simplification, θ H is independent of G OH , but is still a function of G H and potential. The obtained θ H is used to calculate the rate of reaction for this mechanism.
The derivation outlined above is applicable to all reaction mechanisms in this study. All relevant equations are summarized in Tables I and II. For a single-site reaction (Pt (110)), the maximum surface charge is 147 μC/cm 2 . 6 Moreover, site competition between H and OH must be considered, and so the site balance used is θ H + θ OH + θ * = 1. This interdependence can have significant effects on the predicted current density, to be later discussed.
Water Recombination fast, Heyrovsky RDS Water Recombination RDS, Tafel fast Water Recombination fast, Tafel RDS Simulated polarization curves calibration.-Experimental HER and HOR data for bare Pt (110) and Pt (111) in KOH was used to calibrate the simulated single-site and bifunctional polarization curves, respectively. The kinetic current from HOR data was extracted using Koutecky-Levich to adjust for hydrogen mass transport limitations: The experimental current densities measured at ±0.1 V were then used as reference points to fit all simulated HER and HOR polarization curves. The rate constants used for these calibrations are displayed in Table III. As reported by other works, 6,12,26 the equilibrium adsorption potentials of adsorbing species depend significantly on the catalyst surface structure. Here, H and OH adsorption strengths of G H = −0.37 eV and G OH = 0.18 eV (for Pt (110)) or G H = −0.2 eV and G OH = 0.80 eV (for Pt (111)) were used as the benchmark for all simulated polarization curves 6,12 and are assumed to be independent of coverage. Equilibrium adsorption potentials can depend significantly on coverage due to interactions between adsorbates. On a Pt (111) facet, strong H ad -H ad repulsions lead to significant peak broadening. 12 If adsorbates interact, the enthalpy of adsorption could be expressed as H = a + bθ, where b > 0 corresponds to  repulsive interactions. However, coverage dependence is much less pronounced on stepped surfaces such as Pt (110), 24 so it is appropriate neglect enthalpic interactions between adsorbates. More importantly, including the coverage dependence of adsorption strength would not alter the qualitative trends of HER/HOR activity with OH adsorption strength, which is the focus of this study.  Figure 3 shows HER/HOR polarization curves on Pt (110) in either LiOH or KOH. While the difference between the curves is small, it is highly reproducible on Pt(110), and is also observed on polycrystalline Pt. Because adsorbed Li has weaker effects on OH adsorption strength than adsorbed K, G OH is stronger in LiOH than KOH. 12 The difference in the curves thus reveals that stronger OH binding improves HER and HOR kinetics. The improved HER/HOR kinetics in the presence of LiOH differs from our previous findings in which the Volmer step was found to proceed more rapidly in KOH than LiOH. 19 However, such trends can be rationalized if the Volmer step of HER/HOR is not the RDS.

Effect of G OH on single-site reaction.-Experimental
To determine if the experimental effects of LiOH/KOH can be described by OH binding strength alone, the various models of Figure  1 were applied to the experimental data. This approach yields four possible reaction mechanisms for the single-site reaction scheme, where HER and HOR are considered separately as they may follow different mechanisms. Figure 4 compares the experimental and simulated HER and HOR polarization curves for all mechanisms. The simulated polarization curves that best fit experimental data correspond to the mechanism where Volmer is the RDS and Tafel is fast. However, experimental evidence on H/OH-X in LiOH and KOH indicates that Volmer cannot be the RDS, as explained above. In addition, Shinagawa et al. demonstrated that a Tafel slope of 120 mV dec −1 , often considered a Volmer RDS reaction, can also be observed for other mechanisms such as Heyrovsky RDS, 32 suggesting that Tafel slopes can be inaccurate descriptors of surface electrochemistry. Therefore, Tafel plot analyses might not be adequate indicators of the reaction pathway. Importantly, the curves of Figure 4 are generated assuming that all steps other than the RDS are in complete equilibrium, and that only Tafel or Heyrovsky contributes to current. Better fits between the simulated and experimental HER/HOR curves for other mechanisms may be possible by relaxing these assumptions. Regardless, the aim of this work is not to quantitatively match experimental polarization curves nor to identify a precise mechanism, but rather to qualitatively shed light onto the effect of OH adsorption strength on kinetics. We Simulated HER and HOR on single-site catalyst.-In the first case, the direct Volmer step is the RDS, and Heyrovsky and OH adsorption are fast. Tafel plots of the simulated baseline HER/HOR, as well as the effect of G OH , are shown in Figure 5. The H and OH coverages at different potentials with varying G OH are also shown. Experimentally, H is known to desorb at higher potentials in Ar-purged electrolyte. In this case, however, the assumption that the Heyrovsky step is fast implies that H binds more strongly at higher potentials instead of desorbing from the surface (Rxns. 4, 8). Therefore, both H and OH coverages increase at higher potentials. Due to the strength of H adsorption ( G H = −0.37 eV) and site competition, the catalyst surface is predominantly occupied by H for potentials above −0.37 V. It is only for very strong OH adsorption (e.g. G OH < −0.2 eV) that OH can begin to adsorb onto the surface to a noticeable degree. However, the Volmer RDS requires only H ad for oxidation or empty sites for reduction. Therefore, increasing OH adsorption strength results in surface poisoning and slows the kinetics of both HER and HOR.
Similar trends are observed for the case in which Volmer is the RDS and Tafel is fast (Fig. 6). Because Tafel is a chemical, rather than an electrochemical step, for it to be in equilibrium implies that the driving force for H adsorption is independent of potential. However, H coverage is still affected by competitive OH adsorption. If there were no competition for sites, the strength of H adsorption would result in the surface being almost entirely covered by H for all potentials (i.e. θ H ≈ 1). With site competition, H begins to desorb at higher potentials as it is displaced by OH. For stronger OH binding strengths, this H/OH exchange occurs at lower potentials. Because the Volmer step requires only H ad for oxidation or empty sites for reduction, increasing OH adsorption strength once again results in surface poisoning and downgrades the kinetics of both HER and HOR for this mechanism as well.
If the Volmer step is instead in equilibrium and Heyrovsky is the RDS, the results are quite different from the previous mechanisms, as shown in Figure 7. The most immediate feature is the shift in OCP as a function of G OH , as also seen in Figure 4. Because both Volmer and OH adsorption are in equilibrium, the contributions from site competition and configurational entropies lead to an added sensitivity to potential and adsorption strengths, manifested in the denominator of Eqs. 17, 18, 23, and 24. The OCP therefore shifts toward lower potentials as OH binding becomes stronger. In this mechanism, H desorbs from the surface at high potentials since the Volmer step is fast. Due to the convoluted H and OH adsorption behavior arising from competition, stronger OH binding also shifts the potential at which H spontaneously adsorbs or desorbs from the surface. As OH binding strength increases, H is desorbed from the surface at lower potentials. The Heyrovsky reaction requires H ad as well as empty sites to progress. Hence, stronger OH binding results in slower HER kinetics and faster HOR kinetics. An identical trend in which stronger OH adsorption shifts the OCP to improve HOR only is observed for the mechanism where Volmer is fast and Tafel is the RDS. The overpotential needed to reach a given current is a common metric of comparing electrocatalysts. 33 To summarize the trends observed for the single-site reactions, the overpotential needed to reach ± 10 mAcm −2 as a function of OH binding strength for each mechanism is shown in Figure 8. An overpotential of smaller magnitude corresponds to higher catalyst activity and faster kinetics. When Volmer is the RDS, as G OH decreases and the catalyst binds OH more strongly the required overpotentials are seen to increase, in contrast to what is observed experimentally (Fig. 3). Faster HOR kinetics can only be achieved via stronger OH binding strength if the Volmer step is fast, a case which agrees with our previous experimental results. However, our simulations, which use only thermodynamic descriptors and surface coverages to calculate current, predict that this increase in HOR kinetics should be accompanied by a corresponding decrease in HER kinetics. From the trends shown in Fig. 8, however, stronger OH adsorption cannot improve HER kinetics, regardless of the reaction mechanism. Thus, based solely on the relationships between thermodynamic adsorption strength and surface coverage, OH binding strength proves itself to be an insufficient descriptor for predicting HER and HOR kinetics on a single-site catalyst surface.  34 As shown in Figure 9, HER is seen to improve considerably with the introduction of TM clusters with varying oxophilicity, usually attributed to the presence of OH ad on the surface. Conversely, HOR is seen to decrease slightly, but this effect results from the decrease in available Pt (111) surface sites needed for H ad formation. Below a coverage of 40%, HOR current was identical to bare Pt (111). Thus, introducing oxophilic sites has little or no effect on HOR.

Effect of G OH on bifunctional reaction.-Experimental
Due to the improved activity in the presence of oxophilic sites, some studies have concluded that OH ad must be an active participant in the hydrogen reactions through the bifunctional mechanism (Fig. 2). If so, then there should exist an OH adsorption strength that optimizes HER/HOR kinetics. To determine if the experimental effects from Figure 9 can be described by OH binding strength alone, the Figure 2 models were applied to the experimental data. Here, H and OH adsorption strengths of G H = −0.20 eV and G OH = 0.80 eV (for Pt (111)) 6,12 were used as the benchmark for the simulated polarization curves, and the rate constant A f was again adjusted to match experiment. Assuming that H and OH bind exclusively onto their respective sites eliminates effects of site competition. Because H and OH have widely different adsorption strengths on Pt (111), it is appropriate to assume that only H binds to the bare Pt (111) surface. Meanwhile, the non-noble TMs are more likely to be hydroxylated than hydrogenated at HER/HOR potentials, so it is assumed that only OH binds to the oxophilic sites. However, the model makes no assumptions on the oxidation state of the oxophilic clusters. The TM clusters only provide an adsorption site other than bare Pt (111) onto which OH can adsorb non-competitively. Figure 10 compares experimental and simulated HER and HOR polarization curves for all four of these reaction  mechanisms. None of the simulated polarization curves experience a shift in OCP because there is no site competition, but for mechanisms in which the chemical step of Tafel or water recombination is limiting, the current is insensitive to overpotential, in contrast to experiment.
Simulated HER and HOR on bifunctional catalyst.-For the case in which water recombination is the RDS and Heyrovsky is fast, Tafel plots of the simulated HER and HOR, along with the effect of G OH , are shown in Figure 11. The trends in H and OH coverages with G OH are also shown. Because there is no site competition between H and OH, the potential at which each species can adsorb onto their corresponding surfaces is simply dictated by their individual adsorption strengths (Fig. 11c). The Heyrovsky step is fast for this mechanism, so once again H adsorption becomes favorable at higher potentials. Therefore, due to the strength of H adsorption ( G H = −0.37 eV), the H coverage is essentially unity for potentials above −0.2 V. Importantly, water recombination is a chemical, rather than an electrochemical step. Thus, for this potential window, the HOR reaction kinetics is mainly dictated by the OH coverage. Once both the H and OH coverages reach unity at high potentials, current becomes independent of potential, as seen in Figure 11b. For this mechanism, stronger OH binding (decreasing G OH ) depletes HOR reaction kinetics because it shifts the equilibrium constant to favor the reactants of H ad and OH ad . Increasing the OH binding strength reaches the limiting current faster, but decreases the limiting current value significantly due to the shift in the equilibrium constant.
For moderate OH adsorption strengths ( G OH > 0 eV), OH coverage is essentially zero for negative potentials, and the HER current density is dictated by the desorption of H. To observe the effect of OH adsorption strength, G OH must be close to G H , below which it becomes increasingly difficult to remove OH from the surface, thereby deteriorating HER kinetics. Hence, for this mechanism, stronger OH binding only worsens the hydrogen reaction kinetics, which disagrees with what is observed experimentally. An identical effect is observed for the mechanism where water recombination is the RDS and Tafel is in equilibrium (Fig. 12). Because Tafel is a chemical step, it being in equilibrium implies that H coverage is independent of potential. Further, as there is no site competition in a bifunctional mechanism, H coverage will be essentially unity for all potentials due to the equilibrium constant, itself dictated by the strength of H adsorption (Fig. 12c). Therefore, only OH coverage has an impact on the simulated current. The observed effect of G OH is identical to the one discussed above (Fig. 11), where stronger OH binding depletes both HOR and HER kinetics.
If the water recombination step is in equilibrium and either Heyrovsky or Tafel is the RDS, varying the OH adsorption strength has no effect on the simulated current density, as shown in Figure 13. The lack of dependence on OH adsorption occurs because the reaction kinetics for Heyrovsky or Tafel depend only on the H coverage. As there is no site competition on a bifunctional catalyst, altering the OH coverage by varying its adsorption strength has no effect on the simulated current density. This agrees with experimental observations for HOR (Fig. 9).
To summarize the trends observed for the bifunctional reactions, the overpotential needed to reach ±10 mAcm −2 as a function of OH binding strength for each mechanism is shown in Figure 14. For a bifunctional reaction, decreasing G OH either leads to an increase in overpotential or has no effect. Our simulations therefore predict that stronger OH binding cannot improve HER/HOR kinetics, regardless of the reaction mechanism. The Figure 2 mechanisms all assume that the hydrogen reaction requires OH ad to proceed, an assumption which ultimately fails to match what is observed experimentally (Fig. 9). These results strongly suggest that, for a bifunctional catalyst containing oxophilic sites, OH ad cannot be an active participant in the alkaline hydrogen reaction.
The purpose of this work is to qualitatively investigate the effect of OH adsorption strength on hydrogen reaction kinetics, which  was modeled on two different surfaces (Figs. 1, 2). All mechanisms studied herein therefore include OH ad as an active adsorbate, and the effect of stronger OH binding is integrated using only thermodynamic adsorption energies. Table IV summarizes the comparison of experimental and simulated kinetic trends with varying OH adsorption strength. In the single-site model, simulations predict that stronger OH binding can only lead to faster HOR kinetics if the Volmer step is assumed to be fast, and that such change should be accompanied by a decrease in HER kinetics. Stronger OH binding, however, never improves both HER and HOR simultaneously, as opposed to what is observed experimentally (Fig. 3). Thus, based only on thermodynamic adsorption strength and surface coverages,  Figure 12. Effect of G OH on HER (a) and HOR (b) polarization curves for the mechanism where water recombination is the RDS and Tafel is fast. The corresponding H (solid) and OH (dashed) surface coverages are also shown (c). The H coverage curves collapse for all G OH due to no site competition.
OH binding strength proves itself to be an insufficient descriptor for predicting HER and HOR kinetics on a single-site catalyst surface. For the bifunctional model, simulations match experimental HOR observations for the trivial case where water recombination is assumed to be fast and OH adsorption strength has no effect. However, none of the mechanisms predict that stronger OH adsorption leads to faster HER kinetics, which is experimentally the far more significant observation.
While it is tempting to deduce that these results disprove the bifunctional mechanism, they rather re-define its meaning. Clearly, the presence of oxophilic sites on the surface is beneficial 6,7,12-17,34 (Fig. 9), and so the bifunctional mechanism here forth implies that OH ad must improve reaction kinetics indirectly. In this regard, the results here are consistent with the Koper group's evidence that Ni(OH) 2 clusters on Pt (111) shift the PZC in base closer to that in acid. 5 Unlike in acid, the PZC of bare Pt (111) in alkaline media is far from the reversible hydrogen potential, which is hypothesized to lead to a more rigid water orientation and a larger barrier for solvent reorganization required for charge transfer through the double layer. It may be that the high activity of Pt (111) with oxophilic sites stems not from facile water dissociation, but from a more dynamic interfacial water structure.
In the model presented here, the dynamics of solvent reorganization are reflected in the rate constant A f . However, solvent reorganization is not the only phenomenon affecting A f . For example, Chen et al recently showed that the solvation geometry of hydronium results in more frequent and concerted rearrangement of hydrogen bond networks than that of hydroxide. 35 Regardless of the reason for lower or higher values of A f , the results of this study indicate clearly that this kinetic parameter, rather than the thermodynamic quantities of adsorption strength, is critical to explaining the trends between catalysts and electrolytes.

Conclusions
We have explicitly modeled the effect of OH adsorption strength on both single-site and bifunctional alkaline HER/HOR catalysts in order to interpret experimental voltammograms on single-crystal Pt (110) and TM-doped Pt (111). After considering all possible combinations of Tafel, Volmer, Heyrovsky, and water recombination as rate-limiting steps, we conclude that the thermodynamic effects of stronger OH binding do not lead to improved HER nor HOR kinetics in alkaline media. For the single-site case, experimental voltammetry J3220 Journal of The Electrochemical Society, 165 (15) J3209-J3221 (2018) Figure 13. Effect of G OH on HER (a) and HOR (b) polarization curves for the mechanisms where water recombination is fast and either Heyrovsky is the RDS (blue) or Tafel is the RDS (red). The corresponding H (solid) and OH (dashed) surface coverages are also shown (c). When either Heyrovsky or Tafel is the RDS, G OH has no effect on current, so all polarization curves collapse for all G OH . The H coverage curves also collapse for all G OH due to no site competition. Figure 14. Trends in overpotential for the simulated HER (negative overpotentials) and HOR (positive overpotentials) bifunctional mechanisms as a function of G OH for the studied mechanisms: water recombination RDS, Heyrovsky fast (red squares), water recombination fast, Heyrovsky RDS (blue triangles), water recombination RDS, Tafel fast (green diamonds), and water recombination fast, Tafel RDS (black circles).
of Pt (110) in electrolytes with different cations shows faster kinetics for both HER and HOR when OH binding is stronger. However, microkinetic models predict that increasing OH binding strength cannot achieve faster HER kinetics, and can only lead to improved HOR kinetics by shifting the OCP. In general, stronger OH binding leads to surface poisoning rather than increased reaction kinetics due to site competition. For the bifunctional case, experimental voltammograms of Pt (111) with TM surface clusters varying in OH affinity shows that stronger OH binding is associated with faster HER kinetics only. However, the simulations predict that HER activity is always either deteriorated or unchanged by increasing OH adsorption strength. Stronger OH binding shifts equilibrium away from products during HOR and poisons the surface during HER. Comparing theory and experiment therefore concludes that OH ad cannot be an active participant in the hydrogen reactions in alkaline media, regardless of the catalyst structure or kinetic pathway. This work therefore emphasizes that thermodynamic adsorption energies are insufficient for describing hydrogen reaction kinetics. Instead, adsorbed OH must affect the intrinsic rate constants indirectly, either through stabilizing a transition state, by modifying the interfacial electronic structure, or perhaps by affecting the orientation and dynamics of water molecules near the catalyst surface. These phenomena, which must also be responsible for the kinetic effects of pH on HER and HOR in acid and alkaline media, must be considered when designing electrocatalysts to be used in alkaline environments.

Acknowledgments
This work was supported by NSF-1602886.