Cation insertion to break the activity/stability relationship for highly active oxygen evolution reaction catalyst

The production of hydrogen at a large scale by the environmentally-friendly electrolysis process is currently hampered by the slow kinetics of the oxygen evolution reaction (OER). We report a solid electrocatalyst α-Li2IrO3 which upon oxidation/delithiation chemically reacts with water to form a hydrated birnessite phase, the OER activity of which is five times greater than its non-reacted counterpart. This reaction enlists a bulk redox process during which hydrated potassium ions from the alkaline electrolyte are inserted into the structure while water is oxidized and oxygen evolved. This singular charge balance process for which the electrocatalyst is solid but the reaction is homogeneous in nature allows stabilizing the surface of the catalyst while ensuring stable OER performances, thus breaking the activity/stability tradeoff normally encountered for OER catalysts.


Supplementary Figures
. XRD patterns and Rietveld refinements for α-Li2IrO3, α-Li1IrO3, β-Li2IrO3, and β-Li1IrO3. The α-Li2IrO3 and its oxidized counterpart α-Li1IrO3 possess an expandable 2D layered structure, while both β-Li2IrO3 and β-Li1IrO3 have a relatively rigid 3D structure. This structural difference between these two polymorphs enables us to study the effect of the crystal structure on the ion-sieving effect for cation absorption/desorption during the dynamic OER process. After 50 cycles at a potential range between 1.1 -1.7 V vs. RHE, the β-Li2IrO3 catalyst transforms to β-Li1IrO3 but no further evolution to a new phase is recorded, unlike for -Li2IrO3. (a) deconvolution of the 1 H spectrum with three proton environments, with residual protons accounting for 9%, structural OH groups accounting for 11% and structural water for 80%. (b) Upon heating in the spectrometer, the intensity for structural water peak at 1.9 ppm slowly decreases indicating that structural water is loosely bond to the lattice. (c) Under vacuum, such as encountered in the microscope for TEM, the peak for structural water at 1.9 ppm keeps decreasing, indicating that the hydrated birnessite phase transforms into a dehydrated intermediate under vacuum. Supplementary Fig. 17. XRD patterns demonstrating the instability of the birnessite O3-type phase upon washing. XRD diffractograms showing the structural evolution of the hydrated birnessite phase after repeated washing by DI water (a) and vacuum drying (b). (1) as-prepared pristine birnessite sample washed by a mixed H2O/acetone (1:1) solution, (2) washed with DI water for 3 to 5 times till the pH reaches ~9, (3) washed with DI water till the pH reaches neutral, (4) asprepared pristine birnessite sample washed by mixed H2O/acetone (1:1), and dried at R.T., (5) sample dried in a vacuum oven at a pressure of 50 mbar at 40 °C for 4 hours. The α-Li1IrO3 powder is prepared by oxidizing α-Li2IrO3 in H2O-free Li-ion battery up to ~4.0 V. The powder is then washed with DMC solvent and ~50 mg of the obtained α-Li1IrO3 powder was kept in a sealed three-neck Swagelok cell and filled with Ar gas (a). 2 mL of Ar-saturated aqueous alkaline solutions with different KOH or NaOH concentrations was then injected into the Swagelok cell through a silicon rubber stopper. The system was further kept for 24 hours and the gas evolution inside the system was monitored by an online mass spectrometer (c) Background test was performed in order to ensure that no oxygen was detected through leakage from the cell (b). showing the modification of the interlayer distance, related to the intercalation of K + and Na + upon OER. The structure of LiCoO2 is found to evolved upon cycling in KOH and NaOH with two different new phases forming, as indicated by the modification of the (003) peak at low angle. This result indicates that both K + and Na + intercalates into LiCoO2 upon OER, leading to a cation dependent OER. These results demonstrate the universality of the EC mechanism described for -Li2IrO3 that can be applied to other transition metal oxides.    As shown in Supplementary Fig. 2, the α-Li1IrO3 and β-Li1IrO3 can be formed at a voltage of ~ 4.05 and 4.20 V vs. Li + /Li, respectively, during the galvanostatic charging process in Li-ion battery after x = 1 lithium removal from the structure. These voltages are above the voltage of E = 3.503 V vs. Li + /Li (as indicated in the above figure) which corresponds to the OER reversible potential at E = 1.23 V vs. RHE in alkaline solutions at pH = 13.

Supplementary Note 2.
As observed in Supplementary Fig. 3 When using α-Li2IrO3 as OER electrocatalyst in alkaline media (KOH or NaOH solutions), an oxidation event corresponding to the delithiation occurs in the potential range of ~1.0 -1.5 V vs. RHE with approximately x =1.0 equivalent of lithium being removed from the structure. The formation of α-Li1IrO3 and -Li1IrO3 after this oxidation event is then confirmed by ex situ XRD (Supplementary Fig. 7-8).
Interestingly, the electrochemically formed α-Li1IrO3 demonstrates different capacitive behavior in KOH than in NaOH. The capacitive envelope is found almost constant in the potential range 13-1.4 V, indicative that the electrochemically active surface area is not drastically modified by changing the supporting electrolyte and therefore cannot explain the enlarged OER activity measured with KOH (around 5 times greater at fixed current density) compared to NaOH. Only the pseudocapacitive behavior observed by the peaks below 1.25 V and above 1.4 V and presumably associated with cation adsorption/intercalation for α-Li1IrO3 in KOH increases by a factor of 1.5-2 for  comprised between 0.25 and 0.3 V but cannot account for the modification of the OER by a factor of ≈ 5.

Supplementary Note 3.
As observed in Supplementary Fig. 9, Starting from -Li2IrO3, a first oxidation event occurs to form -Li1IrO3 (blue pattern). Then, when the electrode potential is held at 1.55 V vs. RHE for 10 hours, a new phase starts to form (red pattern) with a characteristic peak appearing at 2θ = 12.6 deg., suggesting an expansion of the layered space from d = 4.7 Å to ≈ 6.9 Å during the OER.

Supplementary Note 4.
As observed in Supplementary Fig. 10, XRD measurements reveal that the hydrated birnessite phase α-Li1KxIrO30.7H2O formed by electrochemical cycling of α-Li2IrO3 catalyst in KOH solution as seen in Supplementary Fig. 8 can also be prepared in a chemical way. By soaking the α-Li1IrO3 powder in different alkaline solutions of 1.0 M KOH, NaOH, and LiOH, only the KOH solution was found to trigger the chemical reaction leading to the structural modification and the formation of the hydrated birnessite phase.

Supplementary Note 5
As observed in Supplementary Fig. 15, the mass loss starting below 100°C for the birnessite α-LiK0.3IrO30.7H2O is associated with the loss of structural water. Knowing the potassium amount to be 0.3 per iridium atom and assuming x to be equal to 1 for lithium, the loss of 4.2 w% corresponds to approximatively 0.7 H2O per formual unit. The loss at around 450°C for the birnessite and 550°C for α-Li1IrO3 corresponds to the decomposition of the phase and the formation of IrO2.
Supplementary Note 6 1 H solid-state NMR (ssNMR) analysis of the hydrated birnessite phase in Supplementary Fig. 16 shows that 89% of the protons detected is coming from water, and only 11% of proton is coming from OH -. Based on the TGA analysis ( Figure S18) and Rietveld analysis of the synchrotron XRD (Supplementary Table 2), the total amount of structural water per formula unit is 0.7. Considering that 11% of this structural water comes from OH -, as seen by ssNMR, it would mean that 0.077 OHare inserted into the catalyst. This is not enough to counterbalance the 0.3 K + that are found to be intercalated. Hence, the charge neutrality is not kept and the phase is reduced upon chemical OER.

Supplementary Note 7
As observed in Supplementary Fig. 23, the birnessite phase was first prepared by soaking -Li1IrO3 in KOH solution and then, the powder was recovered before to be assembled in the operando XRD cell with a 0.25 M K2HPO4 electrolyte. The potential was linearly scanned from OCV (~0.5 V) to 1.1 V vs. NHE at a scan rate of 0.1 mV s −1 . (b) selected XRD diffractograms of the as-prepared birnessite phase before (black curve) and after (red curve) charging in K2HPO4 solution. The characteristic peak at 2 = 12.6° is found to be completely suppressed after the oxidation peak starting at around 0.7 V vs NHE, which is therefore associated with the deintercalation of K + from the birnessite phase and the regeneration of the K + -free catalyst.

Supplementary Note 8
As observed in Supplementary Fig. 25, RRDE measurements 15,16 were conducted to better understand the cation effect on the anodic current, and more specifically the current resulting from the oxygen evolution on the surface of the catalyst in different alkaline solutions. The RRDE electrode consists of a glassy carbon working electrode loaded with 50 g cm -2 geo of −Li2IrO3 catalyst and a Pt ring electrode. The collection efficiency is calculated as = × 100%. By calculating the collection efficiency = ( / ) at a fixed potential (E = 1.58 V vs. RHE) at which the oxygen evolution occurs, the effective current contributing from oxygen evolution can thus be estimated. 16 While the theoretical collection efficiency would be 25 % based on the geometry of the RRDE electrodes, the use of drop-casted powder is known to lead to lower collection efficiency. Values ranging from 2 to 19% were previously reported in the literature for various OER catalysts. [15][16][17][18][19] As revealed by the RRDE analyses, the collection efficiency obtained in KOH and NaOH are 11.2% and 10.7%, respectively. As a result, we can conclude that the enhanced anodic current obtained in KOH is indeed related to a higher O2 gas generation, suggesting an intrinsic high OER activity. RRDE measurement was further carried out to better understand the effect of [KOH] on the anodic current and O2 evolution efficiency. The collection efficiency obtained in 0.1 M and 1.0 M KOH are evaluated to be 11.2% and 9.6%, respectively, at E = 1.58 V vs. RHE. An intrinsic high OER activity can be obtained with high concentration of KOH when using the birnessite as catalyst.

Supplementary Note 9
As observed in Supplementary Fig. 27, the electrochemical behavior measured in NaOH and KOH are intrinsically different. Indeed, we further tested the effect of K + concentration in the electrolyte by adding K2SO4 salt and could confirm that the OER activity is dependent on the concentration of K + , and that this concentration effect affects both the high overpotential mass transport limited region as well as the Tafel region at ower overpotential. Therefore, K + is involved in the rate determining step and is decoupled from the electron exchange. It is therefore in equilibrium between the structure and the solution and it is continuously shuttling (intercalate and deintercalate) on the solid catalyst/electrolyte interface between the bulk and the solution. No such effect was observed for Na + , confirming that Na + is not involved into the OER reaction, unlike K + .

Supplementary Note 10
For Supplementary Table 2, the relative occupancies were determined with the following procedure: i) there are 0.7 water molecules per Ir, ii) the potassium content (which is the heaviest atom contributing therefore the most to the diffracted intensity) is freely refined, iii) the Li ratio is fixed by considering that the 9d positon should be at maximum occupied at 1/3, and that the resulting chemical formulae should make sense regarding the Ir oxidation state comprised between 4+ and 5+ as determined by XAS.
The chemical composition as determined by Rietveld analysis for the O3 phase shows greater potassium content than what was determined by chemical analysis (EDX). This is explained by the formation of the O1 domains during the washing step. Indeed, while the chemical analysis probe the whole sample particles and therefore gives an average of the O3 and O1 domains created during the washing step, the Rietveld analysis only take into account the O3 domains and therefore a greater potassium amount is found.
Nevertheless, as our operando XRD study reveals (Fig. 3F in the manuscript), no such O1 domains are found during the electrochemical formation of the birnessite phase. Therefore, care must be exercised when considering the exact chemical composition for the birnessite phase which is constantly evolving depending on the conditions and the characterizations performed.

Possible Li and O surface vacancies in Li1IrO3
As shown in Fig 6 of the manuscript several different surface terminations are observed experimentally. We therefore consider the three different surfaces shown in Supplementary Fig.  29. Given the strongly oxidising conditions we furthermore consider the number of Li atoms in the subsurface layer of the [001] surface and near the step edge of the stepped surfaces. The Li vacancy formation energies for the surfaces are given in Supplementary Table 4. The values are calculated relative to Li(s) and corrected for the free energy of the reaction Li(s) → Li + (aq, 1M) (-3.04 eV). For comparison the Li vacancy formation energy in a 2x1x3 unit cell of bulk -Li1IrO3 is 0.20eV. The concentration of Li + in solution is not known and therefore the exact amount of Li at the surface can not be determined, however based on the loading of the catalyst and the volume of electrolyte a concentration of Li + of 4µM can be estimated. The free energy of dissolving Li(s) will then be reduced by kTln(4e-6) = 0.32 eV at room temperature. This suggests that 1-2 additional Li atoms could be removed from all three surfaces. The [001] surface is different from the steps since the vacancies are in the subsurface layer, and for our unit cell the removal of 4 Li atoms could result in a collapse of the subsurface Li layer. We therefore investigated this situation, however the vacancy formation energy was found to be 0.49 eV/vacancy making such a structure unfavourable.
As we start from an O-rich surface we consider the possibility of removing some of these to form surface oxygen vacancies. The Ovac formation energies at U=0V with H2O as reference are given in Supplementary Table 5. Given that the OER will run at potentials above 1.23 eV only position 1 at the A-step and position 1 at the Z-step are plausible surface structures.

Possible K and O surface vacancies in Li0.75K0.25(H2O)0.50IrO3
For the intercalated structure we consider the formation energy of oxygen and potassium vacancies at the surface and step edges, however limiting the number of structures based on our findings for -Li1IrO3. Thus, we do not consider oxygen vacancies in position 2, since they are less favourable than vacancies at position 1, or sub-surface vacancies of the [001] surface, since they were found to have limited effect on the OER overpotential.
The structure of the subsurface layer means that there are in principle many different O sites, however the position relative to the subsurface layer is assumed to have limited effect on the binding energy compared with the surrounding IrOx environment, and we therefore consider one site only. Furthermore, we only consider K vacancies at the step edges since there are no Li atoms right at the edge. The calculated vacancy formation energies are given in Supplementary Table 6, assuming a concentration of 1M K + for the K vacancies, and a potential of U=0V for the O vacancies. The Kvac formation energies are much higher than the Livac formation energies of -Li1IrO3, and since the experiment is performed in 0.1M KOH the concentration of K + is higher than the concentration of Li + , altogether making the formation of K vacancies unlikely. The formation of oxygen surface vacancies is also less favourable than on -Li1IrO3, meaning that it is less likely for OER to happen close to another Ovac.

Active sites and OER activities
Since we start from an O-terminated surface, the formation of an active site requires the initial removal of an oxygen atom (i.e. formation of a 'surface oxygen vacancy'). While the O atoms at the surface of the [001]-terminated -LiIrO3 are equivalent, there are several inequivalent O atoms at the step edges. In addition, the configuration of oxygen atoms around the Ir atoms at the step edges could differ from the octahedral configuration found in the bulk IrO3 layers. Some of the inequivalent step sites considered here are illustrated in Supplementary Fig. 31. Structures with Li vacancies are not shown as these are in the subsurface or at the edge where they are difficult to visualize.
The OER limiting potentials for the calculated sites are plotted as a function of ∆G2 in Supplementary Fig. 32a. The figure illustrates that the [001] termination has poor activity, and this is largely independent of the number of Li atoms removed from the subsurface layer. The stepped surfaces show much better activities. Supplementary Figure 31b shows the same results as Supplementary Fig. 32a but with the estimated corrections for ZPE, entropy and solvation added. The corrections result in some change in the values of the overpotentials, but the three structures that results in the lowest overpotentials for -Li1IrO3 remain the same. The best limiting potential of 1.60 V is found for the Z-step with an additional lithium vacancy. The formation energy for this vacancy is 0.13 eV in 1M Li + (c.f. Supplementary Table 4), making it a realistic surface structure at the experimental conditions. The second and third most active sites are the step sites with oxygen vacancies, which are also plausible according to the formation energies in Supplementary Table 5.
For the intercalated Li0.75K0.25(H2O)0.50IrO3 we calculate the OER limiting potential for some of the most plausible surface structures, i.e. the Z-step and the A-step with and without oxygen vacancies and the [001] surface. We also consider the less plausible Z-step with a K atom removed from the step edge for comparison with the best performing structure of -Li1IrO3, the Z-step with an Li vacancy. The results are plotted as red markers in Supplementary Fig. 32, showing that all the sites at the steps have a high activity, while the [001] surface has a poor activity, comparable to the [001] surface of -Li1IrO3. There are however no clear trend in the change of the adsorption energies and overpotential when comparing similar sites on the -Li1IrO3 and the disordered surfaces.

Table of calculated adsorption energies
∆G1-∆G4 for all the surface sites considered are given in Supplementary Table 7. The results include the corrections for ZPE, entropy and solvation, and the limiting step in the reaction is highlighted. It is found that most of the sites are limited by the 2 nd or 3 rd step which define the activity volcano. However, if the corrections for ZPE, entropy and solvation are not included the energy of the *OOH intermediate is decreased and several of the sites are then limited by the 4th reaction step. We also note that including the corrections decrease the energy difference Gads(OOH)-Gads(OH), thus making it possible to achieve a better overpotential within the scaling relations.

Supplementary Methods
Quantification of K + intercalation. The atomic ratio of K and Ir elements in the birnessite phase was evaluated by EDX analysis performed on a SEM FEI Quanta FEG 250 microscope. The amount of K intercalated into the layered structure is calculated based on the following equation: Supplementary equation 1: in which, is the amount of K (mol), is the amount of Ir (mol), / is the atomic ratio between K and Ir determined by the SEM-EDX, -1 3 is the mass of the as-prepared α-Li1IrO3 powder, and -1 3 is the molar mass of α-Li1IrO3 (247.16 g/mol). It is worth to stress out that the K amount is very sensitive to the experimental conditions for preparing the birnessite phase (α-Li1KxIrO30.7H2O), since the K cations can be easily leaching out from the structure during the washing step by using DI water. Hence, in order to limit the K leaching, a mixed solution of acetone and water (v/v, 1:1) was used for the washing step. Dertermination of O2 gas evolution. O2 gas evolution when soaking the as-prepared α-Li1IrO3 powder in KOH solution was analyzed by the use of online mass spectrometry. The evolved O2 gas amount is estimated using ideal gas law: Supplementary Equation 2: 2 = 2 • = 2 ′ ′ • with 2 being the amount of O2 gas evolved (mol), 2 the partial pressure of the O2 gas in the Swagelok-type cell. 2 ′ is the partial pressure of the m/z = 32 signal and ′ is the gas pressure measured after the capillary inlet by the instrument. is the pressure of the Swagelok-type cell monitored by an external pressure sensor. Doing so, we assume that 2 = 2 ′ ′ .

Computational details
Further details on the setup of the surfaces. The three surfaces considered for -Li1IrO3 are shown in Supplementary Fig. 29; a 1x2 model of the [001] surface which corresponds to a termination that follows the layers ( Supplementary Fig. 29a), a [102] termination which results in an armchair-type edge of the hexagonal IrO3 layer (A-step, Supplementary Fig. 29b), and a [111] termination which results in a zigzag-type edge of the IrO3 layer (Z-step, Supplementary Fig. 29c). All surfaces are modelled by 3 layer slabs of the bulk structure. The bottom of the slab is terminated by half a layer of Li which, along with the bottom layer of IrO3, is fixed in the bulk position to resemble the bulk of the material. The step sizes are chosen such that the unit cell dimensions in the xy-plane are ca. 10x10Å.
In order to compare the performance of -LiIrO3 with that of the disordered material obtained in KOH solution we make a model of the intercalated structure. This is not a straightforward procedure given the many degrees of freedom, e.g in the ordering of K + , Li + and H2O in the spacer layers between the IrO3 layers and in the orientation of the water molecules. Several different optimisation attempts were therefore performed in order to arrive at a structure where no significant rearrangements in the spacer layers occurred when different adsorbates were added to the surface.
The surface structure is based on the experimentally determined crystal structure, but expanded in a (23x3x1/3) supercell, such that each Ir layer contains 4 Ir atoms, 2 Li atoms and 12 O atoms. The reduction in the c lattice parameter to 1/3 is necessary in order to create a stepped surface, since the layers may otherwise be translated relative to each other during initial optimisation, creating a mismatch at the periodic boundaries. In the spacer layers 2 K atoms, 2 Li atoms and 2 H2O molecules are initially positioned in an ordered fashion, then half of the Li and K atoms are removed to create a structure with stoichiometry Li0.75K0.25(H2O)0.50IrO3 corresponding reasonably well with the experimentally determined stoichiometry of LixK0.3(H2O)0.70IrO3 (x  1) and having the same charge balance as Li1IrO3. This structure forms the basis for the surfaces, which are made to correspond with the surfaces of -Li1IrO3, i.e. a [001] surface, a zigzag step and an armchair step. The optimised structures of the surfaces, shown in Supplementary Fig. 30a-c, are found to have layer spacings of ca. 7.2 Å in reasonable agreement with the experimentally determined layer spacing of 6.94 Å. The structure of the spacer layer in the optimised surfaces is highlighted in Supplementary Fig. 30d. The oxygen atoms of the water molecules are found close to the positively charged K and Li atoms, and the hydrogens point towards the oxygen atoms of the IrO3 layers. We note that this structure might not be the global minimum, but an extended search covering all degrees of freedom is beyond the scope of this paper. To limit the openness of the stepped structure the steps are terminated by K atoms.