Reversible and irreversible transformations of Ni-based electrocatalysts during the oxygen evolution reaction

Nickel-based catalysts for the alkaline oxygen evolution reaction (OER) demonstrate excellent catalytic performance and stability. However, a lack of fundamental understanding of the dynamic electronic and structural changes that occur under OER conditions inhibits the rational design of new materials. Recent advances in operando spectroscopy and computational modeling techniques have helped to elucidate the electrochemically-driven transformations of Ni-based materials. For reversible transformations, this encompasses an increased understanding of the redox transformations of Ni/Fe centers, the adsorption and desorption of reaction intermediates, oxygen vacancy dynamics, phase transformations, and the mechanism of dissolution and redeposition of surface atoms. Likewise, there have been great advances in scientific understanding of irreversible transformations including phase transformations related to ageing, as well as operando surface reconstruction which involves the growth of new OER active phases.


Introduction
As societal understanding of the adverse implications of global warming is increasing, so is the dependence on renewable energy sources such as solar and wind [1,2].Water electrolysis provides a means to store this intermittently produced energy in the form of hydrogen, thus facilitating the global transition away from fossil fuels [3,4].While the hydrogen evolution reaction (HER) is a relatively fast and efficient process, its anodic counterpart, the oxygen evolution reaction (OER), suffers from slow kinetics due to the complex, four-electron transfer pathway [5,6].Commercial OER catalysts for polymer electrolyte water electrolyzers (PEWEs) are based on non-abundant, thus expensive, Ir and Ru; consequently, research efforts target the development of efficient and stable non-noble metal catalysts for alkaline water electrolyzers (AWEs) [7,8].Transition metal oxides containing Ni, Co and Fe are promising candidates for alkaline OER catalysis due to their relative abundance, tunable 3d electron configuration, and the versatility of available crystal structures [2,9,10].In particular, Ni-based materials possess high activity and stability and, unlike their Co-based counterparts, have a comparatively cleaner supply chain with reduced geopolitical risk [11].
The rational design of new Ni-based catalysts requires the development of structureeactivity relationships in order to correlate the electronic properties, local and long-range structure, and morphology of materials with their catalytic performance.However, it is well established that Ni and Co-based catalysts undergo such significant transformations of surface and sub-surface atoms under OER conditions that structureeactivity relationships are difficult to qualify.Indeed, the assynthesized structure is considered merely a "pre-catalyst," which undergoes dynamic reconstruction under oxidative conditions to form an amorphous, active surface known as the oxyhydroxide layer [12,13].Simultaneously, many other reconstruction processes occur under OER conditions.Reversible processes include other potential-dependent phase transformations, the electrochemically-driven dissolution and re-deposition of surface atoms, and the adsorption and desorption of OER intermediates during the catalytic process (accompanied by the associated redox transformations of catalytic centers, and vacancy generation and refilling).Differently, irreversible transformations can include phase transformations and morphological or structural changes [14].This review will examine the following operando transformations of Ni-based OER catalysts: phase changes involving the formation of a surface layer with a new crystalline structure; oxidation state changes of interfacial cations including Ni and Fe; the extent of lattice oxygen participation in the OER; and the uptake of Fe from an impure electrolyte into the crystal lattice.

Phase transformations of Ni-based materials
Ni oxides undergo phase transformations as a function of applied potential, with the reversible formation of an OER active surface layer under oxidative conditions being a key prerequisite for the high activity of these materials [15].The electrochemical stabilities of Ni metal and its oxide, hydroxide and oxyhydroxide derivatives have been calculated by Huang et al. using standard Gibbs free energies of formation (D f G) obtained both experimentally and using DFT, across a range of pH values [16].The resulting Pourbaix diagrams (Figure 1a  NiOOH from their initial NiMoO 4 nano-flower catalyst, without an intermediate hydroxide step [21].Conversely, with comparable reaction conditions and spectra acquisition time, Saguı `et al. used operando Raman to identify an initial, irreversible transformation of their Ta-doped NiO films to a-Ni(OH) 2 upon immersion in the alkaline electrolyte, followed by a reversible transformation to g-NiOOH with applied anodic potential [22].Likewise for Fe/Co-based materials (hydr)oxide catalysts undergo a similar potentialdependent surface transformation to form the OER active oxyhydroxide phase [23,24].For example, D. Grumelli et al. used operando X-ray diffraction to observe the surface reconstruction of Fe 3 O 4 in OER conditions [24].For the highly active spinel Co 3 O 4 , this process is also dependent on the Co-ion geometry: only the tetrahedral Co 2þ is capable of releasing electrons under applied potential to form the surface CoOOH layer [25].[28].Post-mortem lowenergy electron diffraction (LEED) revealed that this surface reconstruction is associated with the disappearance of the perovskite diffraction pattern.While the authors hypothesized an irreversible loss of long-range order in the material, it is important to note that LEED probes approximately a < 1 nm depth.Conversely, Liu et al. used scanning transmission electron microscopy (STEM), probing the entire sample, to identify the localized amorphization of only two LNO surface layers after OER catalysis [29].While the structural amorphization of LNO is a fully irreversible process, the formation of the eOOH layer is theoretically reversible with applied potential, according to the thermodynamic stability windows outlined in Pourbaix diagrams calculated by Huang and Zhou [16,30].However, Fabbri et al. used operando X-ray absorption spectroscopy (XAS) to correlate the oxyhydroxide layer formation on the perovskite Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 O 3- d (BSCF) with an irreversible oxidation of Co atoms [12].This implies that surface eOOH adsorption is in fact not fully reversible, though an analogous claim has not yet been proven for Ni perovskites.

Electronic transformations of cations
Under OER conditions, interfacial cations including Ni and Fe (where present) will undergo reversible changes in oxidation state.For nickel oxide, hydroxide and perovskite catalysts, recent evidence has pointed to the reversible oxidation of Ni centers Ni (2þ) / Ni (3þ) / Ni (4þ) during OER catalysis [12,28,31e33].The formation of Ni (4þ) OO under OER conditions was first identified by Diaz-Morales et al. using in situ SERS combined with an 18 O-labeled electrolyte [34].Due to its high oxidation state, the 3d orbitals in Ni (IV) are lowered in energy to allow optimal overlap with the O 2p orbitals.This, in turn, energetically favors OH -adsorption onto the material surface [27,35].Similar transformations occur in Co-based oxides, as evidenced by Hu et al., who used in situ XAS to identify the formation of surface Co 4þ sites on CoOOH as a function of applied potential [36].Their findings were echoed by operando XAS studies conducted by Zhang et al., who identified the transformation of CoOOH to Co 3þ / 4þ OOH 1-x under oxidative conditions via a potential dependent deprotonation reaction [37].Ni-based materials are often doped with Fe to enhance the OER activity, though the reversible electronic transformations of Fe under OER conditions are still widely debated [38] in an anion exchange membrane electrolyzer, and reported a current of 2 A cm À2 at 2.046 V and 50 C, a performance on par with proton exchange membrane alternatives [46].Initial rotating disk electrode (RDE) studies revealed a positive shift in the Ni 2þ /Ni 3þ redox peak potential attributed to the Ni to Fe charge transfer, and the trigonal distortion of the octahedral symmetry that arises from Ni-O bond contraction as a consequence of Fe doping [47,48].A similar distortion was also observed for Fe-doped LaNiO 3 films by Bak et al., who demonstrated the oxygen-octahedron distortion results in a significant increase of the DOS of both the O 2p and Ni/ Fe 3d orbitals near the Fermi level, facilitating the charge transfer from transition metals to adsorbates via oxygen (Ni 3þ -O(OH*) / Ni 4þ -OO*) [49].Wang et al. used DFT þ U calculations to model the OER mechanism on the (001) facet of g-Fe 0.25 Ni 0.75 OOH (Figure 2a), and created a free energy diagram from the reaction energies of each elementary step, as shown in Figure 2b.The most favorable surface configuration was achieved with a local arrangement of one Fe 3þ and two Ni 4þ atoms, resulting in a low overpotential of 0.57 V (Figure 2c).However, they provided no experimental evidence that such a high local concentration of strongly oxidized Ni 4þ could be achieved, and it appears that they did not consider the possibility of Fe 4þ .Besides, it is vital to consider that irreversible, OER-driven processes such as cation dissolution or surface amorphization will likely evolve the local surface structure when evaluating DFT facet calculations.

Oxygen vacancy dynamics
There are two main classes of OER mechanism: the conventional adsorbate evolution mechanism (AEM), in which all oxygen-containing intermediates originate from the electrolyte, and the lattice-oxygen mediated mechanism (LOM), in which the lattice oxygen participates in the reaction [50].The latter is associated with an increased reactivity for Ni-based perovskites, oxides and hydroxides; thus, developing surfacesensitive spectroscopic techniques that can directly detect oxygen vacancies in situ is paramount for optimizing future catalyst design [51e54].Ex situ neutron diffraction has provided indirect evidence for the oxygen vacancy content of OER catalysts, though it has crucial limitations: highly crystalline samples are required, surface changes in vacancy concentration are difficult to detect with bulk sensitivity, and a large amount of material is necessary, which precludes the possibility of operando studies [55,56].Soft X-ray absorption spectroscopy (sXAS) at the oxygen K-edge likewise indirectly monitors oxygen vacancy dynamics; in combination with total electron yield (TEY) acquisition, a high surface sensitivity (up to 5 nm) is possible [57,58].Thus far, the design of an operando cell with negligible electrolyte interference in the oxygen Kedge absorption spectra limits definitive sXAS data interpretation.Recently, Mom et al. used sXAS to study IrO x thin films of 100 nm thickness while eliminating electrolyte oxygen contribution, through a backcontacted electrolyte/IrO x interface [59].Nevertheless, the use of TEY detection at the 'front' negates the concept of surface sensitivity, and illustrates the multifaceted problem of optimal cell design.
As a result, conclusions about the extent of oxygen vacancy participation in Ni-based materials must be examined with caution based on the experimental methods employed.For instance, Zhang et al. used sXAS of the oxygen K-edge to determine that oxygen vacancy participation in FeCoCrNi thin films can be promoted by the reversible formation of Ni 4þ [57].The generation of Ni 4þ under OER conditions results in downshifted Ni 3d orbitals (as predicted by partial density of states (PDOS) calculations); this induces the formation of oxygen ligands with localized holes in their p-orbitals (i.e.O (2Àd)-) [57].As outlined by Nong et al., the enhanced electrophilic character of these oxygen ligands increases their reactivity towards nucleophilic acidebase-type O-O bond formation (i.e.nucleophilic attack of electron-deficient O (2Àd)-ligands), facilitating the LOM and improving OER activity [60].However, obtaining these ex situ sXAS measurements involved freeze-quenching the post-mortem samples in liquid N 2 , before transferring them in air to the vacuum chamber for measurement; the impact of this preparation process on the surface oxygen is unknown.Moreover, systematic studies performed by Cheng et al. concluded that varying the oxygen vacancy content in Ni and Co-based perovskites was accompanied by changes in other physiochemical properties including conductivity, degree of structural (dis)order and cation oxidation state; the predominant mechanism is determined by the combination of these and other factors [55,61,62].

The effect of electrolyte Fe impurities
Ni-based electrocatalysts uptake trace Fe impurities from unpurified KOH electrolyte, resulting in significantly enhanced OER activity and cycling stability [63,64].Therefore, it is vital to understand the mechanism of this Fe incorporation and its role in improving OER kinetics in order to decouple and optimize the intrinsic activity of Ni catalysts.Kuai et al. used operando XAS to determine that electrolyte Fe incorporation into 2D Ni(OH) 2 nanosheets is an electrochemically driven process, occurring at the OER reactive potential during the anodic CV sweep [65].Furthermore, by using X-ray fluorescence microscopy (XFM) to generate elemental distribution maps, they determined that Fe incorporation occurs predominantly at edge sites, which feature a higher concentration of oxygen vacancies and show higher OER reactivity [65].The group identified a nonlinear increase in OER current with Fe atomic ratio, and attributed this to the irreversible formation of a separate, insulating FeOOH phase at high overpotentials, although other disruptive electronic or structural effects may also influence OER activity.Furthermore, operando soft XAS suggested that Fe surface uptake enhances the reducibility of Ni, increasing the concentration of oxygen vacancies and improving the OER activity [65].
The surface Fe sites exist in a dynamic equilibrium, undergoing reversible dissolution and re-incorporation in aqueous KOH [63].Chung et al. investigated the effect of electrolyte Fe concentration on the OER activity of MO x H y (M = Ni, Fe, Co), and determined that Fe adsorption saturates at electrolyte concentrations as low as 0.1 ppm (Figure 3a, b) and the OER activity increases linearly with Fe surface coverage (Figure 3c).Importantly, maintaining the high OER activity was conditional on achieving "dynamically stable" surface Fe with continued re-deposition after dissolution (Figure 3e) [63].Farhat et al. reported that Ni(Fe)O x H y subsequently cycled in purified (Fe-free) KOH will experience a loss of OER activity as the active, surface Fe atoms move into inactive bulk sites, though they provide only electrochemical evidence to support their claims [66].The authors hypothesize that Ni atoms at the surface are now able to dissolve and redeposit in the structure preferentially with respect to Fe in so-called "Fe-free" KOH, as the purification process involves dissolving nickel nitrate in the electrolyte [66,67].However, Chung et al. used ICP-MS combined with stationary probe rotating disk electrode studies (SPRDE) to prove that Ni sites have a comparatively high stability with respect to Fe in (Ni/Fe)O x H y , and Ni dissolution is negligible [63].

Conclusion
In summary, Ni-based OER catalysts experience significant operando transformations, often leading to a highly active final state.Reversible changes include phase transformations, oxygen vacancy dynamics, intermediate adsorption/desorption, and surface atom dissolution/redeposition.Irreversible changes comprise structural and morphological transformations.This begets the obvious question for further research: is the initial 'precatalyst' structure a prerequisite of the high activity final state, or should the future development of Ni-based materials focus on direct synthesis of this final state?Accordingly, gaining a greater fundamental understanding of the OER mechanism is a broad but pressing concern.In particular, this should involve developing operando methods and innovative cell designs to enable direct observation of oxygen vacancy dynamics (with bulk and surface sensitivity, a high time resolution, and minimal interference from atmospheric, electrolytic, or binder oxygen).Alongside the expansion of fundamental mechanistic studies, further exploration of new classes of materials with less well-known structureactivity correlations would advance the field in new directions.In particular, Ni-based metal-organic framework (MOF) catalysts, noble metal-free high entropy alloys, and amorphous Ni-based materials provide underexplored yet promising avenues for research [68e71].In addition, while it is worthwhile to decouple the intrinsic activity of Ni-based catalysts from changes induced by operando uptake of Fe from unpurified electrolyte, shifting the focus of future research to optimize Ni-based catalysts for operation in Fe-doped electrolyte may be a valuable, application-oriented approach.Finally, approaching DFT calculations with careful consideration of the reversible and irreversible changes of Ni-based catalysts during operation will enable this valuable computational method to direct and enhance experimental work, rather than acting as an addendum.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
, b), experimentally verified by electrochemical impedance spectroscopy (EIS) and surfaceenhanced Raman spectroscopy (SERS), illustrate the formation of new phases at the surface of NiO with applied anodic potential Ni (2þ) O / Ni (2þ) (OH) 2 / Ni (3þ) OOH / Ni (4þ) O 2 , and the instability of these phases at low pH.The limitations of such DFT calculations are exposed by the discrepancy in experimentally determined and DFT-obtained phase stability windows and Ni(OH) 2 oxidation potentials; though the latter would likely arise from an irreversible phase transformation of the disordered, hydrous a-Ni(OH) 2 to the crystalline, anhydrous b-Ni(OH) 2 polymorph with electrochemical cycling.In addition, the study proposed probability profiles (Figure 1c) that indicate multiple phases are present at the NiO-NiOOH boundary including Ni 3 O 4 , Ni 2 O 3 , NiO 2 , NiO, and Ni(OH) 2 , and this may contribute to discrepancies in reported experimental oxidation potentials of NiO or Ni(OH) 2 [16].These electrochemically-driven phase transformations are associated with significant structural changes.The Ni (2þ) (OH) 2 "precatalyst," with the brucite structure (P3m1), exists as two polymorps: b-Ni(OH) 2 , which consists of Ni 2þ and OH -ions in a hexagonal close packed arrangement, and a-Ni(OH) 2 , which comprises planes of b-Ni(OH) 2 with intercalated H 2 O and electrolyte ions.Under OER conditions, a-Ni(OH) 2 and b-Ni(OH) 2 experience reversible phase transformations to form their corresponding oxyhydroxides, g-NiOOH (with a Ni oxidation state of 3.3e3.67þ),and b-NiOOH (with a Ni oxidation state of 2.7e3.0þ),respectively [17e20].While the high g-NiOOH oxidation state range can be attributed to the presence of Ni 4þ , that of b-NiOOH is less easily understood [20].Additionally, in concentrated alkaline solutions, the g-NiOOH phase can form from the irreversible overcharging of b-NiOOH.Attempts to track the formation of these eOH and eOOH species with Raman spectroscopy have produced differing results, suggesting that the exact phase transformations experienced by a material depend on the precursor structure and specific heteroatom doping.For instance, Du ¨rr et al. used operando Raman to identify the direct and irreversible formation of g-

Figure 1 (
Figure 1 Similar to the phase transformations of Ni(OH) 2 polymorphs to the corresponding NiOOH structures, other Ni-based structures such as perovskite or spinel oxides experience an electrochemically-driven surface reconstruction associated with the dynamic formation of an oxyhydroxide surface layer [26e28].Baeumer et al. used operando UV-vis spectroelectrochemistry to investigate the behavior of Ni-terminated LaNiO 3 (LNO) films, and discovered the formation of a 4 A ˚-thick NiOOH surface layer at the Ni 2þ /Ni 3þ redox peak potential during the anodic CV sweep . For instance, XAS studies conducted by Friebel et al. and Go ¨rlin et al. determined Fe 3þ to be the maximum oxidation state of Fe under OER conditions, whereas Hunter et al. used Mo ¨ssbauer and UV-vis spectroscopy to detect Fe 4þ states stabilized as *Fe 4þ =O, a finding supported by DFT calculations performed by Martirez et al. [39e43].Despite the discrepancies in reported electronic structure, it is evident that synergistic electronic interactions between Ni and Fe result in reduced overpotentials for the OER on Fe-doped materials.The bridging oxygen (meO) in Ni-O-Fe bonds can facilitate partial electron transfer as p-donation between the Ni/Fe d-orbitals, and the Ni-Fe synergy might enable a bimetallic mechanism to proceed through bridging O 2 intermediates Ni-*O-O*-Fe [44].Aside from electronic effects, Abbott et al. used operando XAS to observe that the Fe-doping of nickel oxides increases the structural stability of the b-Ni(OH) 2 phase [45].The transformation of b-Ni(OH) 2 to a layered b-NiOOH structure (and disordered a/gphases) facilitates ionic diffusion to previously inaccessible metal centers, increasing the electrochemically active surface area.However, since Fe-doping enhances the intrinsic activity of the available active sites in b-Ni(OH) 2 , this transformation is no longer a prerequisite of high activity.Wang et al. studied the effect of Fe doping of a-Ni(OH) 2