Crystal-facet-dependent surface transformation dictates the oxygen evolution reaction activity in lanthanum nickelate

Electrocatalysts are the cornerstone in the transition to sustainable energy technologies and chemical processes. Surface transformations under operation conditions dictate the activity and stability. However, the dependence of the surface structure and transformation on the exposed crystallographic facet remains elusive, impeding rational catalyst design. We investigate the (001), (110) and (111) facets of a LaNiO3−δ electrocatalyst for water oxidation using electrochemical measurements, X-ray spectroscopy, and density functional theory calculations with a Hubbard U term. We reveal that the (111) overpotential is ≈ 30−60 mV lower than for the other facets. While a surface transformation into oxyhydroxide-like NiOO(H) may occur for all three orientations, it is more pronounced for (111). A structural mismatch of the transformed layer with the underlying perovskite for (001) and (110) influences the ratio of Ni2+ and Ni3+ to Ni4+ sites during the reaction and thereby the binding energy of reaction intermediates, resulting in the distinct catalytic activities of the transformed facets.

1 Supplementary Note 1: Additional structural and compositional analysis of the as-prepared state.
Supplementary 2 Supplementary Note 2: X-ray absorption spectroscopy In this section, we discuss the peak positions and related interatomic distances in the EXAFS analysis.The peak positions in Fig. 4c of the main text are lower by about 0.2 to 0.5 Å than the interatomic distances due to a phase shift δ originating from the scattering atoms (note that the EXAFS oscillations are modeled as sin(2kR + δ) with wavenumber k) [2].Peak (i) is assigned to Ni-O with interatomic distance 1.92 Å, while peak (iii) is assigned to Ni-La with interatomic distances of about 3.1 and 3.3 Å and peak (iv) is assigned to Ni-Ni in corner-sharing octahedra with interatomic distance of 3.95 Å, indicative of the perovskite structure [3,4].These values are in acceptable agreement with expected values for LaNiO 3−δ (where strain and point defects might increase the lattice parameter compared to the bulk value of 3.84 Å) [5] and with our DFT+U simulation.Supplementary Figure 5: Fourier transform of the EXAFS of the Ni-K edge collected on (111)-oriented LaNiO 3−δ samples.(a) As prepared.There is spectral intensity near peak (ii).But it is almost fully explained by sidelobes of the Ni-La interactions (peak (iii)).Other contributions could be Ni-O interactions that were not modeled.Fit in Supplementary Table 1.(b, c) 16 h CA at 2.5 V vs. RHE.In (b), we removed the Ni-Ni interaction at 3.2 Å, which results in a worse fit than the one in Fig. 4c 5.This fit is also worse than the one in Fig. 4c of the main text.These considerations indicate that the spectral intensity near 2.5 Å reduced distance are mainly due to sidelobes for the as prepared sample but due to a (hydr)oxide phase for the sample after 16 h CA at 2.5 V. Data are shown as circles and EXAFS fits as lines.Fourier transforms were performed between 35 and 550 eV (above E 0 = 8333 eV) with a cosine window covering the first and last 10% of the data range.
The EXAFS of LaNiO 3−δ after 16 h at 2.5 V is markedly different.The Ni-O peak (i) is found at larger distance, corresponding to an interatomic distance of 2.01 Å.This increase in Ni-O bond length is indicative of a reduction of Ni.We note that the actual interatomic Ni-O distance in oxyhydroxides (Ni (3+δ)+ ) is shorter (1.88 Å) [6] than for the perovskite LaNiO 3−δ (1.92 ± 0.01 Å in Supplementary Table 1 to 1.93 Å [3]) so that the larger distance after 16 h at 2.5 V vs RHE supports a reduced hydroxide or oxide, e.g., Ni(OH) 2 or NiO.This is expected in our ex-situ measurement, where the samples were returned to open circuit voltage (OCV) and exposed to air before X-ray absorption analysis.According to the Ni-O-H Pourbaix diagram, this would lead to Ni(OH) 2 or NiO surface phases after returning to OCV.The interatomic Ni-O distance of 2.01 ± 0.01 Å is slightly shorter than expected for NiO (2.05 -2.08 Å [6]), suggesting slight oxidation and possibly protonation as in Ni(OH) 2−x .
Note that each EXAFS peak has sidelobes, and intensity near the distances marked as peak (ii) also results from the sidelopes of peaks (i) and (iii) (Supplementary Fig. 5).[7] For the sample after 16 h CA, there are no major peaks at similar distance, so that a sidelobe can be excluded.The metal-metal distance at peak (ii) is similar to the edge-sharing reference LiNiO 2 (Fig. 4b of the main text).Thus, we assigned the new peak/distance to edge-sharing Ni octahedra at 2.98 Å in LaNiO 3−δ after 16 h (Supplementary Fig. 5b,c) and note that the corner-sharing Ni octahedra, i.e., peak (iv), are resolved simultaneously.Supplementary Figure 6: Fourier transform of the EXAFS of the Ni-K edge collected on (001)-oriented LaNiO 3−δ samples.Data are shown as circles and EXAFS fits as lines.Fourier transforms were performed between 35 and 550 eV (above E 0 = 8333 eV) with a cosine window covering the first and last 10 % of the data range.Fits in Supplementary Tables 6, 7).The region around 2.5 Å is not well fit, which may suggest that (hydr)oxide is present on the as prepared sample and after 1 h CA at 1.75 V vs RHE.
3 Supplementary Note 3: Additional structural and compositional analysis after OER To investigate structural and compositional changes as a function of operation time, we performed XRD, XPS and STEM analysis after up to 100 h operation time.
Supplementary Figure 7: Cation ratio for LaNiO 3−δ (a) (111) and (b) (001) films before and after chronoamperometry (CA) for mean escape depths of d = 1.1 nm and d = 2.1 nm.Only minor changes in cation composition of the surface layer is observed in the first 16 h of operation.

XPS:
We compare the surface and subsurface composition, obtained from angle-dependent XPS measurements (resulting in different mean information depths d = 1.1 nm and d = 2.1 nm, Supplementary Fig. 6).The overall Ni/La ratio at d = 1.1 nm is higher than for d = 2.1 nm for all samples, indicating predominant Ni termination in the as-prepared state for both (100) and (111) orientations.For operation up to 16 h, the Ni/La ratio remains constant, as shown exemplarily for the (111) orientation 7a.This indicates that on average, the transformed surface is as Ni-rich as the predominantly Ni-terminated asprepared state.Only after additional stability testing up to a total operation time of 100 h, we find that for both (100) and (111) samples, the Ni/La ratio changes: While the surface remains slightly Ni-rich, the overall cation ratio approaches 1:1.But we note that the XPS signal intensity decreased by a factor Supplementary Figure 8: HAADF STEM micrographs of LaNiO 3−δ films after 100 h CA for (001) orientation (a, c) and ( 111) orientation (b,d).Pit-like corrosion is observed after this long-time stability tests (highlighted with red and yellow ellipses in panels a and b).In addition, we still observe the extended defects in the bulk of the film, similar to the samples before CA (examples highlighted with red arrows). of roughly 10 from the pristine state to the 100 h CA state, indicative of an increased surface roughening (in line with the XRR and TEM results discussed below), making Ni/La ratios extracted from the area-integrated technique after 100 h CA less reliable.
STEM: We performed high-angle annular dark-field imaging (HAADF) STEM analysis of (001) and (111) LaNiO 3−δ samples before and after 100 h CA, to track the most severe changes during stability tests (Supplementary Fig. 7).We note that cross-sectional STEM relies on preparation of lamella using a focussed ion beam (FIB).While we used a protective carbon layer during lamella preparation, the top surface ( 1-2 unit cells) may be damaged during FIB.Nevertheless, the STEM analysis provides valuable additional information: Even after 100 h of operation, the bulk of the film remained intact (in line with the XRD analysis below).Near the surface, we observe pit-like corrosion: The entire lamellas for both orientations were decorated with nm-sized amorphous regions with a depth of up to 3 nm.These observations are an interesting starting point for future investigations of stability and degradation mechanisms in perovskite-type electrocatalysts, but the observed pits are not directly related to our observations regarding activity.Supplementary Figure 9: XRR and XRD analysis before and after 100 h CA for (001) orientation (a, c) and (111) orientation (b,d).
XRD/XRR: We performed X-ray reflectivity (XRR) and X-ray diffraction (XRD) before and after 100 h CA (Supplementary Fig. 8).Overall, the data confirm that the bulk of the films did not change to a large extent, in line with the STEM analysis.The XRR data is consistent with an increased surface roughening and a similar thickness for all films (with a 1 uc thicker (111) film after 100 h).The nm-sized amorphous regions observed in TEM make the extraction of quantitative figures from XRR difficult.The (bulk-sensitive) XRD data indicates a similarly high crystallinity (similarly pronounced Laue fringes) and similar lattice constant before and after the stability test, with small thickness variations (variations in the period of the Laue fringes).
Summary: We observe local near-surface amorphization after extended stability tests, but otherwise the LaNiO 3−δ samples remain highly-ordered model systems in the bulk.We conclude that the long-time stability observed for the bulk of the films (XRD, STEM) and the comparably stable surface cation ratio (for at least 16 h of operation) form the basis for our detailed analysis of the transformed surface using XAS and DFT+U .for both transformed facets (Supplementary Fig. 15).For * O, we observe a higher density of O 2p states close to the Fermi edge for U = 2.0 eV.As the Ni states shift to lower energies with increasing U , the hybridization with the O 2p states is reduced, resulting in a weaker bonding, and thus higher ∆G B * O .This effect is stronger for the (001)-orientation.In the case of * OH and * OOH the oxygen 2p states are more uniformly distributed and the hybridization with the Ni 3d states is not or only marginally altered with change of U .In other words, the covalency of the Ni-O bonds changes with the U value, where a more ionic bond leads to weak bonding of the * O intermediate.

Supplementary Tables
Supplementary Table 1: EXAFS fit results of (111) LaNiO 3−δ as prepared.Phase functions generated from LaNiO 3 were used.An asterisk indicates that the parameter was fixed during the fit between 35 -550 eV in k-space.The filtered square R-factor [8] (0 -4.5 Å) of this fit was 5.3.Fit errors were calculated as described in the experimental section.Peak position refers to Fig. 4.
1.92 ± 0.01 3.13 ± 0.02 3.29 ± 0.01 3.95 ± 0.01 4.15 ± 0.04 N 6.1 ± 0.5 1.8 ± 0.9 4.4 ± 1.0 9.9 ± 1. Phase functions generated from LaNiO 3 were used.An asterisk indicates that the parameter was fixed during the fit between 35 -550 eV in k-space.The filtered square R-factor [8] (0 -4.5 Å) of this fit was 6.0.Fit errors were calculated as described in the experimental section.Peak position refers to Fig. 4. Phase functions generated from LaNiO 3 were used.An asterisk indicates that the parameter was fixed during the fit between 35 -550 eV in k-space.The filtered square R-factor [8] (0 -4.5 Å) of this fit was 13.3.Fit errors were calculated as described in the experimental section.Peak position refers to Fig. 4.

Figure 1 :
(a) RHEED intensity evolution during PLD growth of LaNiO 3−δ on (110)-and (111)-oriented SrTiO 3 substrates.(b, c) RHEED patterns after growth, confirming two-dimensional growth.Supplementary Figure 2: HAADF STEM micrographs of LaNiO 3−δ films on (a) (001)-and (b)(111)-oriented SrTiO 3 substrates.We observe extended defects in the bulk of the film (examples highlighted with red arrows).These were observed before,[1] connected to the chosen synthesis temperature, which was in turn necessary to obtain the desired Ni-termination in the as-prepared state.Supplementary Figure 3: (a)-(c) La 4d XP spectra and (d)-(f ), Ni 3p spectra for different orientations.Data for all orientations were normalized to the La 4d peaks.Open circles show the data points, lines are a guide to the eye obtained by 5-point adjacent average smoothing.Angle-dependent measurements resulting in different mean escape depths d qualitatively indicate predominant Ni-termination for the (001) and (111) orientations.
of the main text.No new parameter optimization was performed after removal of the Ni-Ni interaction from the fit.In (c), we performed parameter optimization (i.e., a new fit), still excluding the Ni-Ni interaction, fit results are shown in Supplementary Table

4
Supplementary Note 4: Additional DFT+U results.Supplementary Figure 10: Relative stability of NiOOH x on (a) LaO-terminated LaNiO 3 (001) and LaNiO-and O 2 -terminated LaNiO 3 (110) and (a) LaO 3 -and Ni-terminated LaNiO 3 (111) as a function of applied voltage, at pH = 13.x denotes the number of hydrogen atoms intercalated between the transformed NiOO layer and the underlying perovskite.The Hubbard parameter on the Ni 3d states is U = 5.5 eV in all cases.Supplementary Figure 11: Surface Pourbaix diagrams showing the most stable coverages with functional groups as a function of applied voltage and pH for (a) untransformed, Ni-terminated LaNiO 3 (111).(b) a NiOO-layer on LaO 3terminated LaNiO 3 (111).Note that the voltage in experiment is measured vs RHE, therefore the reaction conditions correspond to φ = (1.68 − 0.0591 • pH) = 0.91 V vs SHE and pH = 13 in the diagram.Above and below, side views of the coverages are shown, with frame colors matching the respective areas in the diagram.Note that the calculations for the untransformed and transformed surface are performed with U = 2.9 eV and U = 5.5 eV, respectively (see Supplementary Discussion).Using U = 2.9 eV for the latter leads to the same surface termination under OER conditions despite some changes in regions of stability of the terminations.Supplementary Figure 12: Surface Pourbaix diagrams showing the most stable coverages with functional groups as a function of applied voltage and pH for (a) untransformed, LaNiO-terminated LaNiO 3 (110) (the O 2 -termination is treated as a possible coverage).(b) a NiOO-layer on LaNiO-terminated LaNiO 3 (110).Note that the voltage in experiment is measured vs RHE, therefore the reaction conditions correspond to φ = (1.72 − 0.0591 • pH) = 0.95 V vs SHE and pH = 13 in the diagram.Above and below, side views of the coverages are shown, with frame colors matching the respective areas in the diagram.Note that the calculations for the untransformed and transformed surface are performed with Hubbard U = 2.0 eV and U = 5.5 eV, respectively (see Supplementary Discussion).Supplementary Figure 13: Cumulative reaction free energies of NiO 2 -terminated (001), LaNiO-terminated (110) and Ni-terminated (111) untransformed perovskite surfaces for Hubbard parameter U = 2.0 eV.At the (001)-surface all Ni sites are covered by * OH, while a quarter of the O sites are covered by H, as shown in [1].For (110), the reaction intermediates are adsorbed between one Ni and two La atoms.For (111), the reaction intermediates are adsorbed between one Ni and one La atom.Two surface terminations with functional groups, 0.75 ML * OH + 0.25 ML * OOH and 1 ML * OOH, were taken into account for (111) as both appear close to the experimental overpotential (see Supplementary Fig. 12a).Ideal ∆G B i are indicated by dashed black lines.Supplementary Figure 14: (a) Dependence of OER overpotential on the Hubbard U parameter.For NiOO/LaNiO 3 (001), the most favorable reaction sites for each U value are shown (see Supplementary Table 9 ). Gray, dotted horizontal lines indicate the experimental overpotential of 0.51 V for (001) and 0.45 V for (111) (at 1 mA/cm 2 ).(b) Binding energy of OER intermediates as a function of Hubbard parameter.∆G B O (solid lines) shows a notably stronger dependence on U than ∆G B OOH (dotted) and ∆G B OH (dashed).Supplementary Figure 15: Density of states of oxygen 2p and the 3d orbitals of the three Ni neighbors for the different OER intermediates at (a) one oxyhydroxide-like layer on LaNiO 3 (001) and (b) NiOOH/NiOOH 0.75 on LaNiO 3 (111).At lower U value, there is a stronger hybridization of O and Ni 3d states.For U = 5.5 eV, the occupied Ni 3d bands are shifted to lower energies, which reduces the hybridization with the O 2p states in the region just below the Fermi edge, especially for the * O intermediate.For the * OH and * OOH intermediates, the effect is much weaker.

:
Binding energies of OER intermediates and overpotential for different lattice oxygen reaction sites for NiOO on LaNiO-terminated LaNiO 3 (110) at Hubbard U = 5.5 eV.The transformed surface is covered by 1/3 ML H under reaction conditions, the sites with hydrogen are marked in the table.Bond lengths to the three Ni neighbors of the reaction site are given, as well as Ni magnetic moments.Bond lengths are marked in bold, if they are substantially increased due to distortion, magnetic moments of ∼ 1 µ B are marked in bold.The accompanying figure shows the positions of the reaction sites.To emphasize elongated Ni-O bonds, other bonds are shown semi-transparent.Site d Ni−O ( Å), M Ni (µ B ) ∆G B * OH (eV) ∆G B * O (eV) ∆G B * OOH (eV) η (V) NiOO/LaNiO The choice of the Hubbard U -parameter applied to the Ni 3d orbitals turns out to have a considerable effect on the calculated OER overpotentials, as shown for four sample systems in Supplementary Fig.14a.To gain more insight, we plotted the binding energies of the OER intermediates ∆G B i as a function of U in Supplementary Fig.14b.On both transformed facets there is only a slight increase of ∆G B * OH and ∆G B * OOH with increasing U , but a markedly stronger one for ∆G B * O .This implies that the height of the * OH → * O step increases, while that of * O → * OOH decreases with U .While for the transformed (111) surface the PDS remains * O → * OOH for all U values, for the (001) transformed surface the PDF is * OH → * O at U = 5.5 eV, but there is a crossover to O is strongly influenced by U , we analyzed the projected density of states of oxygen and its Ni neighbors for the three intermediates ∆G B * O → * OOH and a lower overpotential for an increasing number of reaction sites below U = 2.9 eV.To gain further insight into the U dependence of η and why mainly ∆G B * * OH , ∆G B * O and ∆G B * OOH

Table 6 :
[8]FS fit results of (001) LaNiO 3−δ as prepared.Phase functions generated from LaNiO 3 were used.An asterisk indicates that the parameter was fixed during the fit between 35 -550 eV in k-space.The filtered square R-factor[8](0 -4.5 Å) of this fit was 7.3.Fit errors were calculated as described in the experimental section.Peak position refers to Fig.4.

Table 8 :
[1]ding energies of the OER intermediates and overpotential on untransformed LaNiO 3 surfaces.All Ni sites at the (001) surface are covered by * OH, while a quarter of the O sites are covered by * H, as shown in[1].For (110), the functional groups covering the surface are adsorbed between one Ni and two La ions.For (111), functional groups are adsorbed between one Ni and one La ions.

Table 9 :
Binding energies of OER intermediates and overpotential for different lattice oxygen reaction sites at NiOO/LaNiO 3 (001) at Hubbard U = 5.5 eV.Bond lengths to the three Ni neighbors of the reaction site are given, as well as the Ni magnetic moments.Bond lengths are marked in bold, if they are substantially increased due to distortion, magnetic moments of ∼ 1 µ B are marked in bold.The accompanying figure shows the positions of the reaction sites.To emphasize elongated Ni-O bonds, other bonds are shown semi-transparent.