Water electrolysis provides a means to store renewable energy through the production of hydrogen1. Typically, the electrolyte — and therefore the pH — is the same at both the cathode (where hydrogen is produced) and the anode (where oxygen is produced). However, having a different pH at each electrode has some advantages, including utilization of active and stable noble-metal-free catalysts2. To maintain such a pH gradient, a bipolar membrane (BPM) containing a proton exchange part and a hydroxyl-ion exchange part between the electrodes is needed (Fig. 1a). Inside the BPM, water dissociation (WD) occurs as an electric-field-driven heterogenous chemical reaction (\({{\rm{H}}}_{2}{\rm{O}}\rightleftharpoons {{\rm{H}}}^{+}+{{\rm{OH}}}^{-}\,\)). Protons generated are transported towards the cathode while hydroxyl ions generated are transported towards the anode3.

Fig. 1: Compensation effect as a limiting factor for water dissociation and for electrocatalytic reactions in an electrolyser with a bipolar membrane.
figure 1

a, H2 and O2 production in an acido-alkaline electrolyser with a bipolar membrane (BPM) and water dissociation (WD) catalyst under constant overpotential (η). b, Illustration of the compensation effect for two different cases. Line 1 is the case where it is possible to significantly reduce activation energy (Eact) and how that impacts activity (j1). Line 2 is the case where it is possible to significantly enlarge the prefactor (log A) and how that impacts activity (j2). The red cross qualitatively indicates desirable catalyst performance (that is, low Eact and high log A), suggesting that the most promising catalysts are those where a relatively large gain in log A and high activity (j2) can be obtained on the basis of a relatively small increase in Eact.

Full size image

As is the case for the hydrogen evolution reaction (HER) and the oxygen evolution reaction, the water dissociation reaction also requires a catalyst. The key task of a catalyst is to accelerate reactions by reducing the activation energy (Eact), which can be experimentally obtained using the well-known Arrhenius equation. This relation dictates that the slope of the temperature dependence of the reaction rate is proportional to Eact, while the intercept is equal to the logarithm of the pre-exponential frequency factor, log A. This latter term indicates the frequency of collisions between the reactants and the catalyst surface that will result in formation of the product.

Following the Arrhenius equation, the reaction rate can be enhanced by reducing Eact or enlarging log A (Fig. 1b). In recent decades, research focus was predominantly on Eact because it can be directly related to the adsorption energy of intermediates, while the nature of log A has remained challenging to comprehend. The main obstacle that has been hindering acceleration of WD by reducing Eact is the existence of the so-called compensation effect — an empirical relationship indicating that Eact is directly proportional to log A. In other words, if we reduce Eact we simultaneously reduce log A. This phenomenon is known in thermal catalysis4 and it was recently shown to be valid in electrocatalysis2. Nevertheless, systematic datasets detailing Eact and log A are rare in electrocatalysis because they require carefully executed high-temperature electrochemistry experiments. Therefore, our understanding of the compensation effect has been limited.

Now, writing in Nature Energy, Sebastian Oener and colleagues at the Fritz-Haber Institute of the Max Planck Society report a systematic study of the compensation effect in WD catalysis occurring in BPMs5. They find that Eact and log A are directly related to the capacitance of the double layer (the interfacial region between the electrode and the bulk electrolyte) and attribute the impact of the capacitance on log A to a change in the activation entropy. Understanding of the intrinsic link between double layer capacitance on one side and Eact or log A on the other side can help us to understand how to minimize the compensation effect, and in principle allow us to develop better catalytic materials.

In the past, it was proposed that the surface charge could be responsible for the existence of a balance between Eact and log A (refs. 2,6), but no experimental data supported this hypothesis. Oener and colleagues now confirm these predictions with a substantial amount of data for catalysed WD in BPMs and also for HER. The researchers investigated many important parameters that determine the kinetics of WD, including: the applied bias, the loading of the WD catalyst, the potential of zero charge of WD catalysts as well as the frequency of the electric field used to probe the amount of the capacitive charge stored in the double layer.

The team’s data suggests that the WD rate can be enhanced by increasing the catalyst loading, which reduces Eact. At the same time, the WD rate was also improved by strengthening the interfacial electric field, achieved by increasing the applied bias or by increasing the potential of zero charge of the catalyst. In both cases, the increased WD rate stemmed from an increase in log A. The influence of the frequency of the electric field used to estimate the capacitance of the double layer is intriguing: the impact of the double layer capacitance on Eact is noticeable at much higher frequencies than the impact of the capacitance on log A. This suggests that the processes contributing to the formation of the activated complex (that is, the pre-organized structure formed at the catalyst/electrolyte interface triggered by intermediate adsorption, prior to product formation) are intrinsically much faster than the processes that are contributing to collisions of the reactants with the catalyst surface.

The main claim in the researchers’ work — that the bias-dependent relationship between Eact and log A is linked to interfacial capacitance — is well grounded. The accumulated capacitive charge influences the entropy in the double layer, which originates predominately from the motion of the solvent (that is, water)7. The characteristic water motions contribute to effective collisions between reactant and catalyst active sites and/or BPM functional groups, and in that way contribute to log A. The team also propose that the double layer capacitance impacts Eact only if the applied bias is high enough to trigger adsorption of intermediates, which seems to be intuitive.

Another important claim is that OH– solvation is kinetically more demanding than H+ solvation and that the solvation kinetics are independent of the catalyst properties. For the case of WD in a pristine BPM, when the bias is altered, the slope of log A versus Eact is identical to the slope for HER in alkaline media at a metallic electrocatalyst. The essential difference between these two reactions is the absence of the generated proton in the case of HER in alkaline media. Based on that, the researchers propose that the WD kinetics are predominantly determined by the hydroxyl ion solvation.

Oener and colleagues indeed report a number of valuable phenomenological relations, yet some still require clarification. In the future, it would be useful to comprehend the physical meaning of the slopes of log A versus Eact plots for various conditions (for example, different catalysts, electrolyte pH, applied bias and so on). Nevertheless, to use the equality of the slopes of log A versus Eact in the case of two different reactions at two different surfaces (that is, WD at pristine BPM and HER in alkaline media at metallic electrocatalyst) to be an indicator of ion solvation kinetics may be an oversimplification: both Eact and log A depend on a multitude of material and interfacial properties, and ion solvation is just one of them. It is worth bearing in mind that the slopes of log A versus Eact plots for HER in acidic and alkaline media on various metals are also identical2 despite different ions participating in the (de)solvation process. This suggests that the slope of log A versus Eact perhaps cannot be attributed to the (de)solvation energy of one ionic species alone.

The important results reported by Oener and colleagues offer ample scope for extension in the future. For example, it would be valuable to investigate why the impact of capacitive charge on Eact becomes apparent at relatively high-frequency electric fields, while the impact of capacitive charge on log A is apparent at lower frequencies. This could shed light on what processes are controlling the link between Eact and log A. Furthermore, knowing that the activation entropy2,8,9 is just one part of the pre-exponential factor, the components of log A should be thoroughly investigated, because it seems that enhancement of log A is the most promising way to accelerate conversion reactions. In the meantime, the work from Oener and colleagues provides progressive insight into interfacial structural dynamics, which is essential for understanding of energy conversion processes and for creating a sound basis for advanced catalyst design.