Evidence of Mars‐Van‐Krevelen Mechanism in the Electrochemical Oxygen Evolution on Ni‐Based Catalysts

Abstract Water oxidation is a crucial reaction for renewable energy conversion and storage. Among the alkaline oxygen evolution reaction (OER) catalysts, NiFe based oxyhydroxides show the highest catalytic activity. However, the details of their OER mechanism are still unclear, due to the elusive nature of the OER intermediates. Here, using a novel differential electrochemical mass spectrometry (DEMS) cell interface, we performed isotope‐labelling experiments in 18O‐labelled aqueous alkaline electrolyte on Ni(OH)2 and NiFe layered double hydroxide nanocatalysts. Our experiments confirm the occurrence of Mars‐van‐Krevelen lattice oxygen evolution reaction mechanism in both catalysts to various degrees, which involves the coupling of oxygen atoms from the catalyst and the electrolyte. The quantitative charge analysis suggests that the participating lattice oxygen atoms belong exclusively to the catalyst surface, confirming DFT computational hypotheses. Also, DEMS data suggest a fundamental correlation between the magnitude of the lattice oxygen mechanism and the faradaic efficiency of oxygen controlled by pseudocapacitive oxidative metal redox charges.


Introduction
Future large-scale reductive electrochemical generation of fuels and chemicals at electrolyzer cathodes will require reactions and catalysts at the counter anode that facilitate the release of protons (H + )a nd electrons (e À )w ith maximum efficiency.T he electrocatalytic oxygen evolution reaction (OER) or electrochemical water oxidation reaction leading to the formation of O 2 ,e lectrons,a nd protons is such ak ey counter reaction. TheOER catalysts activity is assumed to be kinetically controlled by surface binding energies between catalytic active surface sites and reactive oxygenated intermediate species. [1] Amolecular understanding of energy-and cost efficient catalysts for the OER is vital for the design of advanced anodes for water electrolyzers [2,3] or CO 2 -water coelectrolyzers for the generation of hydrogen or carbonaceous products,respectively. [4,5] Ni-based materials are among the most active and energyefficient OER catalysts for alkaline electrolyzer anodes.T he catalytic active state of Ni-based catalysts is their g-phase that forms from the inactive a-o rb-phase.T he activity of g-NiOOH was shown to increase drastically after the addition of Fe under anodic conditions [6] with reported optimal molar Fe :Ni ratios ranging between 0.1 and 0.5. [7] Thec atalytic active g-NiFeO ER catalyst presents al ayered crystalline structure with intercalated cations between brucite-type metal oxide layers,i nw hich the metal atoms occupy the center of edge-connected octahedra. In contrast, the nonactive a-phase is characterized by intercalated anions,and it is referred to as a-N iFel ayered double hydroxide (LDH). By structural analogy with g-NiOOH, the OER active deprotonated phase that forms by oxidation of a-NiFeL DH is referred to as g-NiFeL DH. Over the years,d istinct hypotheses about the nature of the catalytically active site on the brucite layers were put forward. [1b,8] Earlier hypotheses include single Fe active sites [8c] or single Ni sites. [9] Later, there emerged heightened interest in the mechanistic role of the oxygen ligands of the edge-connected M-O octahedra of the 2D brucite layers. [10] Recently,anew model of the bulk structure of the catalytically active g-NiFeL DH phase and of its active surface sites was brought to the fore. [1b] What distinguished this latest model from earlier ones was the primary catalytic role of specific m 2 -oxygen bridges between neighboring Ni and Fe sites,w hile explicitly accounting for i) ar eversible cation and water intercalation in the interlayer space and ii)n oncovalent interactions between the interlayer species and the Ni/FeOx brucite layers. [1b] Thereactivity of this structural and mechanistic model relied on ad irect involvement of surface lattice oxygen ligands in the elementary catalytic reaction mechanisms,a lso referred to as as urface lattice oxygen evolution reaction (LOER) mechanism. ForN iFeL DH, ap articular surface LOER mechanism has been proposed where,f ollowing aM ars-van-Krevelen-typem echanism, the surface lattice oxygen atom combines with oxygen from the electrolyte forming molecular oxygen and leaving av acancy, which is then filled by ahydroxide ion from the electrolyte. [1b] TheL OER character of state-of-art NiFeL DH anodes of alkaline water electrolyzers has remained acontentious issue and calls for experimental verification or dismissal using atomic-level analytics.
To gain atomic-level experimental insights into operating NiFe-based OER catalysts,m ost earlier studies employed acombination of in situ/operando X-ray absorption spectroscopy (XAS) [8c, 11] Mçssbauer spectroscopy, [12] and voltammetric techniques.However,this set of techniques is not suitable to discriminate whether or not aL OER is present. [13] To achieve this,Differential Electrochemical Mass Spectrometry (DEMS), where product sampling occurs directly from the electrified liquid-solid interface is most suitable.E arlier DEMS studies on ab imetallic NiFem ass-selected nanoparticle model catalyst [14] as well as asolvothermally prepared NiFeL DH OER catalyst revealed important details on the stability,activity,and the faradaic efficiency. [11a,15] In almost all previous DEMS studies,t he electrochemical flow cell architecture involved amono-or adual thin-electrolyte layer type. In these DEMS cells,volatile products generated inside avery thin electrolyte film between electrode and ah ydrophobic membrane transition into the differentially pumped vacuum system for detection. Thin layer cells suffer from severe mass transport limitations and dontsustain large catalytic current densities.M oreover,t hey require large electrolyte volumes, which is problematic for studies with expensive isotope labelled electrolytes available only at micro-or milliliter scale.This is why innovative bulk electrolyte DEMS cells with small-volume designs are needed.
In this contribution, we use an ovel DEMS cell interface to experimentally test ar ecently reported computational hypothesis regarding the participation of surface lattice oxygen ligands,w hich is aL OER, in the OER catalysis on a-/g-NiFeL DH catalysts.I sotope labelling results indeed suggest the presence of aL OER for both liquid precursorderived g-NiFeL DH catalysts as well as aF e-free b-/g-NiO x H y reference catalyst. Crystalline Ir reference oxides showed no LOER. We further unravel and discuss ap reviously overlooked correlation between the faradaic efficiency and the contribution of the LOER for various catalyst systems.This relation calls for chemical or synthetic measures to minimize the LOER character of OER catalysts in order to maintain high faradaic efficiency.Finally,wedemonstrate the ability of the present time-resolved DEMS technique to accurately deconvolute faradaic charge stored in molecular oxygen from pseudocapacitive charge stored in the catalyst surface.T his previously inaccessible charge balance analysis quantifies the anodic Ni 3+ and Ni 4+ redox charges as afunction of applied electrode potential under catalytic operating conditions.

Results and Discussion
Differential Electrochemical Mass Spectrometry in aH anging Droplet Cell An ew differential electrochemical mass spectrometry (DEMS) setup (LIQUIDLOOP GmbH) ( Figure S1) was used in this work employing two distinct electrochemical liquid/vacuum cell interfaces (Figure 1a nd Figure S2). The dual thin-layer electrolyte cell ("thin layer cell") design is based on earlier similar approaches. [16] It is ar obust and reliable interface design that requires relatively large electrolyte volumes.Itconsists of two horizontal parallel disk-shaped compartments with connecting liquid channels (Figure 1a The" hanging droplet DEMS flow cell" (Figure 1b)i s an ew design (see Figure S2 and S4). It was developed for isotope labelling experiments because it allows experiments with microliter scale electrolyte volumes (typically 20-50 mL). This capability is useful whenever expensive solvents or electrolytes are,f or instance,i sotope-labelled compounds. Thee lectrochemical measurements are performed inside ah anging electrolyte droplet, the volume of which is maintained under constant in/out flow conditions.T he outlet flow ranged typically at 1 mLs À1 .T he inlet tube was placed at 2mmfrom the electrode.Reaction products were withdrawn through acapillary placed at 500 mmdistance from electrode. Thec apillary is ac oncentric tube inside the inlet flow tube. Thereaction products are collected together with the electrolyte solution and introduced into adisk-shaped compartment where aP TFE membrane acts as the interface between the liquid and vacuum. AAg/AgCl electrode served as reference electrode.

Mass Spectrometric and Faradaic Voltammetry in Non-Labelled Conditions
TheOER catalysts addressed here comprised a a-/g-NiFe LDH and a b-Ni(OH) 2 / g-NiO(OH) x powder thin film catalysts and were compared to acrystalline Ir oxide electrocatalyst.
At the outset of this study,t he DEMS thin film cell was used to characterize the surface voltammetry and O 2 faradaic efficiency ("FE O2 ") of the catalysts in non-isotope-enriched (non-labelled) environments with natural 16 O/ 18 Oa bundances.A fter ac yclic voltammetric (CV) activation protocol, faradic voltammograms and simultaneous mass spectrometric cyclic voltammograms (MSCV) were recorded between + 0.5 V RHE and the point when 6mAcm À2 was reached at as can rate of 5mVs À1 (Figure 2). MSCVs tracked the three mass currents of m/z = 32 of 16  Thef aradaic CVs obtained in the DEMS cell were nearly identical in shape to those reported previously in conventional three electrode cell and electrode setups,which validates the cell and electrode design. [17] In the anodic scan direction, the MSCVs closely traced the CVs,w hile on the cathodic scans mass currents revealed characteristic tailings.
Such tailings may originate from slow diffusional O 2 transport out of/across the porous catalyst film. [11a, 18] However,w e observed some tailing in non-porous thin Ni oxide layers,a s well (see Supporting Note 1), implying that the origin of the tailings might be at least in part related to the charge (hole) storage mechanism and slow discharge in Ni-based OER catalysts.Noother volatile products than oxygen (e.g. CO 2 at m/z 44) were detected.
Forthe Ni(OH) 2 catalyst, the m/z = 32 MSCV featured an unexpected quite cathodic ("low") onset potential of 16 O 2 formation near + 1.41 V RHE (green line in Figure 2b)t racing closely the well-documented Ni(II +)(OH) 2 ! Ni-(III +)OOH redox wave (red line). This 16 O 2 generation, however, appears transient in nature,w hich points to an incomplete reduction of Ni(III +)OOH to Ni(II +)(OH) 2 during an earlier cathodic scan resulting in oxidative charge trapped by the formation of poorly conductive Ni(OH) 2 domains. [19] Once the electrode potential was swept anodically again, the Ni(OH) 2 ! NiOOH oxidation re-occurred restoring ac onductive catalyst layer. As ar esult of this,t he trapped hole charges could now discharge by reacting with the electrolyte molecules,r esulting in the transient evolution of molecular oxygen.
In bimetallic NiFeL DH catalysts,t he partial overlap of the anodic Ni(OH) 2 ! NiOOH wave with the voltammetric OER onset makes an accurate estimate of the onset potential of sustained O 2 evolution during potential sweep measurements difficult. MSCVs,h owever, are able to provide them accurately.F igure 2a shows that the 16 O 2 evolution onset of the NiFeLDH catalyst occurred at + 1.47 V RHE, hence slightly more cathodic compared to Ni(OH) 2 In Figure 2, the mean O 2 faradic efficiency,FE O2 ,ofNiFe LDH (obtained by integrating the MSCV from and back to + 0.5 V RHE ,w hile integrating the anodic faradaic current only) was 90 %, while that of Ni(OH) 2 was 62 %(see details in Supporting Note 2).
Thesignals of the other two oxygen isotopes ( 16 O 18 Oand 18 O 2 )were at least 2orders of magnitude smaller than that of 16 O 2 .While the 16 O 18 Osignal displayed aweak rise at the most anodic potentials,t he signal of 18 O 2 was too low to discern adetailed current/potential response. 18

OIsotope-Enriched Mass Spectrometric and Faradaic Voltammetry
To obtain deeper insight into the character of the OER reaction mechanism on the two Ni-based OER catalysts,w e performed DEMS experiments,d uring which the nonlabelled 16 Table S1). Thee lectrode potential was swept three times from + 0.8 V RHE to    (Figure 3c and 3d inset). Thec oupling of lattice oxygen atoms with solvent oxygen atoms appears kinetically preferred, which may have to do with the ready initial availability of 16 Oligands on the catalyst surface,w hile the 18 O 18 Op roduct requires the adsorption of two solvent molecules.T he formation of m/z = 32 16 O 16 O remained at noise level at all times,s howing that the direct coupling of surface or bulk lattice oxygen atoms is unlikely.
We conclude the existence of akinetically favored lattice oxygen evolution reaction (LOER) process on the two Nibased OER electrocatalysts.N ote we use the term LOER regardless whether the oxygenated ligand (OH, O) from the catalyst lattice belonged to the surface or to the bulk.
To get further insight in the gradual exchange of oxygen atoms between the catalyst surface and electrolyte,w ek ept tracing the oxygen isotope ratios in non-labelled electrolyte after an electrolyte exchange.N on-labelled electrolyte was continuously flown over the catalyst surface to ensure acomplete exchange of the electrolyte. Figure 4s hows the evolution of the experimental (blue bars) vs.t he natural (hashed bars) atomic abundance of 18 O during four potential cycles for both Ni-based catalysts.T he data revealed anearly 3-fold and more than 4-fold higher 18 O abundance on the first cycle for Ni(OH) 2 and NiFeL DH, respectively,w hich now reflects the opposite oxygen isotope exchange between H 2 16 Oe lectrolyte and catalyst, following

Angewandte Chemie
Forschungsartikel the measurements in H 2 18 O-based electrolyte.T he absolute 18 Oi sotope excess remained lower than that of 16 Ob efore, which indicates that only af raction of the catalyst surface is actually contributing to the oxygen exchange processes.T he faster depletion in 18 OofNiFeLDH vs. 18 OofNi(OH) 2 is fully consistent with its higher OER catalytic activity (Figure 2).
From the isotope labelling experiments,w ed erived the percentage of catalyst oxygen atoms that participated in the LOER mechanism. Over one potential cycle,this ratio ranged from 2.9 %f or NiFeL DH to 3.6 %f or Ni(OH) 2 (see Supporting Note 3). These numbers suggest am inute contribution of the lattice oxygens of the catalysts,c onceivably due to alimited accessibility of metal/oxygen moieties at the surface of the catalyst. Forc omparison, the ratio of electrochemically reactive Nickel atoms was evaluated from the precatalytic anodic voltammetric charge under the assumption of a1e lectron transfer per Nickel center (see Supporting Note 3). Fort he NiFeL DH catalyst, the estimate of the electrochemically redox active Ni amounted to 1.7 %o ft he total Nickel atoms evidencing limited accessibility.I nt he case of Ni(OH) 2 ,t he ratio exceeded unity under the 1electron/Ni atom assumption, evidencing redox transition from Ni 2+ to Ni 3+ and Ni 4+ .This analysis suggests aquite distinct electrochemical Ni accessibility between the two electrocatalysts.Arole of the distinctly difference morphologies in terms of nanoplatelet size combined with ar ole of Fe appears plausible.T his analysis is consistent with the existence of aLOER process occurring only at the surface of the nanoplatelets (and at al imited number of internal sites accessible through cracks, defects and edges of their polycrystalline domain structure [1b] ), in agreement with the absence of bulk lattice involvement in the OER mechanism demonstrated by Roy et al. [14] Ab ulk LOER process,i nvolving sites deep inside the LDH interlayer regions in the center of the crystalline domains, would result in amuch larger lattice oxygen contribution than that observed here.
To compare the contribution of the LOER mechanism of the Ni based catalysts with another benchmark catalyst, we performed similar isotope-labelling DEMS OER experiments on an Iridium oxide catalyst in HCl electrolyte. [18] Figure 5s hows the data analysis of the Ir oxide DEMS experiments.I nc omparison to the theoretical m/z = 34 mass current that is expected based on the 18 Oisotope enrichment of the electrolyte (dotted black line in Figure 5a,center), the experimental m/z = 34 mass current (grey line in Figure 5a, center) was essentially am atch. This evidenced an egligible contribution of lattice oxygen atoms and thus,unlike the Nibased catalysts,s uggested no significant contribution of aLOER mechanism for the Ir catalyst.
While studies of LOER mechanisms on Ir catalysts are sparse,there are anumber of LOER studies for hydrous and crystalline Ru oxides. [20] Ap resence of LOER was reported on porous RuO x [20c] and nanocrystalline RuO x , [20b] but not on crystalline rutile RuO 2 nanoparticles of % 50 nm, and neither on well-defined (100), (110), (101), and (111)-oriented rutile RuO 2 surfaces. [21] ForRucatalysts,structural arguments were put forward, that is,LOER mechanisms are likely to occur on amorphous or hydrous phases with their large number of lattice defects,t heir undercoordinated sites and their high degree of redox-active surface hydroxylation giving rise to large pseudo capacitance. [21] Supporting this hypothesis further, amore pronounced presence of LOER was observed for Ru 0.9 Ni 0.1 O 2Àd than for RuO 2 : [20b] Ni leaching is known to result in alattice-defective,redox-active hydroxylated surface with an elevated ratio of undercoordinated sites.I nv iew of the Ru results,the crystalline nature of our IrOx catalyst may explain the absence of aLOER mechanism.
Similar structure-mechanism relations can be invoked to account for the presence of aL OER mechanism on NiFe LDH and Ni(OH) 2 .E ven though the prepared NiFeL DH and Ni(OH) 2 catalysts displayed bulk crystallinity (Fig-Figure 4. The evolution in the atomic fraction of 18 Oofthe total DEMS charge of evolved oxygen measured in non-enriched H 2 16 O-based electrolyte for a) NiFeLDH and b) Ni(OH) 2 .The catalysts were previously cycled into the OER range in 18 Oisotope-enriched electrolyte. Arinsing step with non-enriched H 2 16 Ow as conducted before the experiments shown here. Shown are data obtained from MSCVso ver the first four cycles (see also Figure S8 and S9). The bars represent the experimental 18 Ofraction (full blue bars) and the expected 18 Ofraction based on the natural isotope abundance( hashed bars).

Angewandte Chemie
Forschungsartikel ure S10), their catalytic active g-NiFeL DH and g-NiOOH structures feature water intercalation and hydroxylated surfaces, [1b] the nanoplate morphology of which favored undercoordinated edge sites.S imilar to our results,S hao-Horn and co-workers demonstrated LOER mechanisms in highly covalent perovskites that show pH-dependent activity (La 0.5 Sr 0.5 CoO 3Àd ,P r 0.5 Ba 0.5 CoO 3Àd and SrCoO 3Àd ), while less covalent and pH-independent LaCoO 3 lacked LOER. [20a] Those perovskites contain alkaline earth metals (Ba and Sr), which easily dissolve in the electrolyte and result in the formation of surface oxyhydroxides of amorphous and, possibly,h ydrous nature,w ith high number of undercoordinated sites,asithas been shown for BaSrCoFeperovskite. [22] Finally,L OER was also observed for spinel Co 3 O 4 , [23] the surface of which reconstructed under OER in amorphous CoOOH. [24] Along these lines,D oyle et al. [25] suggested that hydrous transition metal oxides show pH-dependent activity. Indeed, NiFeL DH follows a( super-Nernstian) pH-dependence activity, [17,26] and so does g-NiOOH [27] and both are characterized by aL OER.
Following the most recent structural hypothesis as to the active surface site and ligand on g-NiFeL DH and g-NiOOH, [1b] the successful observation of aLOER mechanism requires the presence of catalytic active surface lattice m 2 -OH ligands that perform aO -O coupling with incoming electrolyte molecules.A ni nitial, very rapid exchange of lattice OH with OH À or H 2 Ofrom the electrolyte,however,atthe outset of voltammetric scans and outside the OER potential range may prevent the experimental observation of the LOER. To ensure the analytical detection of an existing LOER, the following conditions need to apply:i )t he catalyst has to continuously expose pristine surface facets with not-yet exchanged oxygen ligands due to am orphological decomposition, ii)the time of the voltammetric pretreatment should be kept at am inimum, and iii)t he time resolution of the DEMS analysis has to be sufficiently high. Otherwise,failure to detect aLOER remains inconclusive.
Based on these arguments,wetend to attribute the lack of aL OER mechanism on the surface of electrochemically activated NiFea lloy nanoparticles [14] to their stable bulk structure combined with rapid ligand exchange prior to DEMS detection (see Supporting Note 4), even though differences in pretreatment protocols may have arole,aswell. Indeed, differences in the XAS-determined local structure between NiFeLDH and an electrochemically Fe-activated Ni (hydr)oxide have been recently reported by Hu and coworkers, [28] suggesting that local structural differences might exist which will have implications in the OER mechanism.
Apart from quantitative estimates of the contributions of LOER mechanisms,derived from the ratios of Q MS 34 /Q MS 36 in Figure 3e,o ur DEMS analysis revealed another previously overlooked kinetic-mechanistic correlation, as shown in Figure 5b.T he contribution of the LOER mechanisms of the catalysts scaled very closely with their faradaic efficiencyo f O 2 ,F E O2 .T he catalyst with larger LOER contribution suffered from lower efficiency, that is,m ore holes injected in the catalyst were stored as oxidative pseudocapacitive charge in redox-active metal centers,r ather than being used to generate oxygen. Excess pseudocapacitive anodic charge, however, is known to promote undesired catalyst corrosion pathways. [29] In conclusion, from ac harge efficiency point of view,s ignificant LOER contributions appear undesirable as they appear to be linked to low FE O2 .

DEMS Based Deconvolution of Pseudocapacitive Charge and the Effective Chemical State of Ni under OER
To learn more from the DEMS data about the chemical state of the Ni catalyst during OER, we conducted am ore detailed charge analysis of the CV, i F ,and the faradaic MSCV, i DEMS F; O2 ,o fb-Ni(OH) 2 /g-NiO(OH). From the faradaic current density i F the total anodic oxidative charge Q tot F injected into the catalyst can be estimated. Q tot F can be deconvoluted into three different components:1 )t he faradic charge Q DEMS F;OER , (grey area) associated with O 2 evolution from solvent molecules,2 )t he faradic charge Q DEMS F;LOER (cyana rea) associated with evolution of mixed isotope O 2 due to LOER, 3) the Ni oxidation charge, Q F;Ni ,consumed for redox state changes of the Ni centers. Figure 6s hows the deconvolution of the faradaic current i F and the faradic mass spectrometric current i DEMS F; O2 for the first potential cycle (cf.F igure 3e)a nd their respective charges Q tot F and Q DEMS F;O2 . Q DEMS

F;O2
splits into the charge Q DEMS F;LOER associated with the lattice oxygen mechanism (cyan area), and into the charge Q DEMS F;OER associated with oxygen formed from the electrolyte (grey area). Thef ollowing relations for the mean faradic efficiencyhold: .T he percentage of Q DEMS F;LOER in respect to the total Q DEMS F;O2 might represent alower limit due to the exchange of surface lattice hydroxides with electrolyte). The faradaic O 2 efficiency, FE O2 ,is dependent on the potential window considered in the integration of i F : FE O2 = 82 %f or purple potentialr ange, that is, without the Ni redox wave 1; FE O2 = 62 %for pink potentialr ange including the Ni redox waves. Only anodic faradaic currents were included in the analysis to exclusively account for anodic processes (molecular O 2 ).

Angewandte
Where x LOER is the percentage ratio of the lattice oxygen with respect to all oxygen. From the number ratio l = Q 34 /Q 36 (cf. Figure 3e)weestimate x LOER to be 14 %(Supporting Note 3). Mean FE O2 values were calculated to 62 %(pink potential window in Figure 6). In other words,38% of Q tot F is oxidative charge Q F; Ni that was injected into Ni atoms and served to increase the Ni redox state.Ifthe Ni +2/+3 redox charge of peak "1" was excluded from Q tot F by narrowing the integrated potential window,aFE O2 of 82 %e nsued (purple potential window in Figure 6). However,t his still left 18 %o fQ tot F unaccounted for, which was evidently used for the further oxidation of Ni 3+ under peak "2". We split the total metal charge Q F;Ni into the Ni 2+/3+ transition (charge under peak 1) and the subsequent Ni 3+/4+ transition (convoluted with the OER charge under peak 2ofi F )according: From data in Figure 6a nd the relations in Supporting Note 2w eobtain Our charge balance analysis implies the formation of Ni 4+ ; more importantly,itsuggests that more than half and almost 2 = 3 of all Ni centers of the Ni(OH) 2 catalyst have reached the Ni 4+ state inside the OER range.T his is excellent agreement with independent measurements of the mean Ni oxidation state of + 3.6 for g-NiOOH by previous XAS studies [6c, 25,30] and fully consistent with the classical Bode redox model of Ni oxyhydroxides. [30] Fort he NiFeL DH catalyst, the faradaic contribution of the evolved O 2 was distinctly different ( Figure S11). The DEMS-based evaluation of Q F;Ni2þ=3þ was no longer possible, since the Ni 2+/3+ redox process had merged with the OER voltammetric profile.I ndeed, the oxidation states of Ni and Fe during catalytic OER are still being debated. Earlier XAS measurements on unsupported NiFeL DH suggested al arge portion of Ni to remain in a + II state during OER, while Ni 4+ remained below 4%. [31] By contrast, higher pH, supported catalysts or very thin catalyst films showed increased ratios of Ni 4+ .I no ur present study of unsupported NiFeL DH, the catalyst exhibited al arge mean FE O2 = 90 %, which implied little metal redox charge,which is consistent with the low Ni 4+ ratios reported by Gçrlin et al. for unsupported NiFe-based catalyst in 0.1 MK OH. [11a, 26, 31] In summary,the DEMS-based faradaic oxygen efficiency and charge analysis is able to deconvolute faradaic molecular oxygen charge from pseudocapacitive redox metal charge.I t can be used to extract independent estimates of the chemical state of the catalyst under catalytic reaction conditions.T he metal centers of the NiFeLDH catalyst appear to be in aless oxidized state compared to the Ni centers of the Fe-free Ni(OH) 2 ;h owever, NiFeL DH outperforms the Fe-free catalyst in catalytic reactivity (Figure 2), which speaks to the high intrinsic activity of the NiFeL DH active sites.

Conclusion
Thepresent work has revealed new mechanistic aspects of the oxygen evolution process on the surface of Ni-based OER electrocatalysts in alkaline environments.
To achieve these insights,wehave first presented aversatile new differential electrochemical mass spectrometry (DEMS) liquid/vacuum cell interface,referred to as "hanging droplet cell". Thecell design addresses the need of minimum electrolyte flows where expensive isotope-labelled reagents or solvents are involved. Theu sefulness of the new DEMS cell was demonstrated in the study of the OER mechanism of aN i(OH) 2 and aN iFeL DH catalyst in 18 O-labelled electrolyte.
Characteristic 16 O 18 Oi sotope DEMS data suggested that the mechanism of catalytic OÀObond formation involves,to as mall portion, lattice oxygen atoms at the catalyst surface. This observation was more evident in Ni(OH) 2 than in NiFe LDH, even though valid for both catalysts.D uring this socalled lattice OER (LOER) mechanism, oxygen atoms from the catalyst lattice are continuously consumed. In the present case of aL OER Mars-Van-Krevelen mechanism, the lattice oxygen atoms are continuously substituted by oxygen atoms from the electrolyte.I np arallel to the LOER mechanism, oxygen evolves from H 2 18 Or esulting in 18 O 18 O. Ar elation between LOER, faradic efficiency,t he amorphous/hydrous catalyst structure,a nd its pH-dependent activity is hypothesized and discussed. Thecase of aMars-Van-Krevelen LOER mechanism has important implications for future designs or models of OER electrocatalysts that now have to consider the role and the binding of lattice atoms ligands,a sw ell. This study highlights the importance of understanding the surface atomic structure of oxides to tune their catalytic activity.