Nature of S-States in the Oxygen-Evolving Complex Resolved by High-Energy Resolution Fluorescence Detected X-ray Absorption Spectroscopy

Abstract

Photosystem II, the water splitting enzyme of photosynthesis, utilizes the energy of sunlight to drive the four-electron oxidation of water to dioxygen at the oxygen-evolving complex (OEC). The OEC harbors a Mn₄CaO₅ cluster that cycles through five oxidation states S_i (i = 0–4). The S₃ state is the last metastable state before the O₂ evolution. Its electronic structure and nature of the S₂ → S₃ transition are key topics of persisting controversy. Most spectroscopic studies suggest that the S₃ state consists of four Mn(IV) ions, compared to the Mn(III)Mn(IV)₃ of the S₂ state. However, recent crystallographic data have received conflicting interpretations, suggesting either metal- or ligand-based oxidation, the latter leading to an oxyl radical or a peroxo moiety in the S₃ state. Herein, we utilize high-energy resolution fluorescence detected (HERFD) X-ray absorption spectroscopy to obtain a highly resolved description of the Mn K pre-edge region for all S-states, paying special attention to use chemically unperturbed S₃ state samples. In combination with quantum chemical calculations, we achieve assignment of specific spectroscopic features to geometric and electronic structures for all S-states. These data are used to confidently discriminate between the various suggestions concerning the electronic structure and the nature of oxidation events in all observable catalytic intermediates of the OEC. Our results do not support the presence of either peroxo or oxyl in the active configuration of the S₃ state. This establishes Mn-centered storage of oxidative equivalents in all observable catalytic transitions and constrains the onset of the O–O bond formation until after the final light-driven oxidation event.

Introduction

Photosystem II (PSII) of plants, algae, and cyanobacteria converts sunlight into chemical energy in the form of “high energy” electrons derived from water. These electrons are used in CO₂ reduction to glucose, which is the primary form of food for all living organisms. PSII is thus a key enzyme in the flow of energy into the biosphere, and its unique ability to oxidize water makes its study of paramount importance both for basic research and as the only model system for developing artificial water-splitting catalysts. Water splitting takes place at the Mn₄CaO₅ cluster of the oxygen-evolving complex (OEC) (Figure 1a), embedded in the protein scaffold of PSII.¹⁻⁵

Figure 1.

(a) Crystallographic structure of the OEC in the S₁ state.³⁵ (b) Catalytic cycle of water oxidation by the OEC with commonly accepted oxidation states of the Mn ions in the S₀–S₂ states and the currently discussed alternatives for the S₃ state.

The cluster cycles through a progression of four light-induced sequential oxidations, S₀ → S₁ → S₂ → S₃ → [S₄], mediated by redox active tyrosine residue Tyr161 (Y_Z). Dioxygen is formed and released during the last S₃ → [S₄] → S₀ transition, where S₄ is a transient state (Figure 1b). The catalytic cycle includes two water binding events, probably during or after the S₂ → S₃ transition and prior to reconstitution of the S₀ state, but the details of water binding as well as the position and protonation states of the substrates for dioxygen formation remain uncertain. Commonly accepted formal oxidation states of the four Mn ions in the S₀, S₁, and S₂ states are Mn(III)₃Mn(IV), Mn(III)₂Mn(IV)₂, and Mn(III)Mn(IV)₃, respectively.⁶⁻¹¹ Therefore, the S₀–S₂ oxidation events are unanimously considered to be Mn-based. However, the localization of the S₂ → S₃ oxidation remains a fundamental mechanistic question since differing suggestions exist about the nature of the transition as well as the electronic structure of the S₃-state itself.^10,12−18

All interpretations of electron paramagnetic resonance (EPR) observations for all different signals associated with the S₃ state are consistent with Mn-centered oxidation in the S₂ → S₃ transition, leading to a Mn(IV)₄ assignment for the S₃ state.¹⁹⁻²⁵ However, in X-ray absorption near edge spectroscopy (XANES) experiments, the shift of the Mn K-edge to higher energy in the S₂ → S₃ transition was found to be smaller than that in the S₀ → S₁ and S₁ → S₂ transitions. This observation along with K_β X-ray emission spectra during the S₂ → S₃ transition^9,15,26 has received two alternative interpretations: either ligand-based oxidation during this transition,¹⁵ or a coordination sphere change from a five-coordinate Mn(III) ion to an octahedral Mn(IV) ion^10,17 as a result of water binding (Figure 2, oxo-hydroxo). The first scenario would mean that the Mn oxidation states in the S₃ state remain Mn(III)Mn(IV)₃ and that a ligand radical species is formed instead (Figure 2, oxyl-oxo). This scenario appears consistent with two-flash (2F) structural models derived by recent serial femtosecond X-ray free electron laser crystallography (SFX-XFEL),²⁷⁻³⁰ Crystallographic data have received conflicting interpretations, ranging from an all-Mn(IV) oxo-hydroxo,²⁹⁻³¹ to a ligand oxidized oxyl-oxo or peroxo species (Figure 2).^{27,28,32−34}

Figure 2.

Different models discussed for the S₃ state, showing the range of possible electronic structures.

The above distinct formulations for the S₃ state have profound implications for the water oxidation mechanism. The possibility of oxygen-radical formation after the second flash (2F) led to the hypothesis of an “early-onset” O–O bond formation already from the S₃ state.^{18,32,36−40} By contrast, if the S₃ state is Mn(IV)₄, the O–O bond formation must take place after the last flash of the cycle (3F), either in a transient S₄ state or after formation of the S₃Y_Z^• intermediate. Computationally, all three types of models have been proposed as part of possible water oxidation mechanisms.^32,39,41,99 However, the relative energies between computational models of the different S₃ variants and the correspondence to the spectroscopic observations remain inconclusive.

X-ray spectroscopy provides information on core–valence transitions and can probe Mn oxidation states and the ligand environment. In previous XAS studies,^15,17 conclusions were derived only from the XANES edge region, based on comparison with experimental results from Mn model complexes with known structures and oxidation states. The pre-edge region corresponds to the electric dipole forbidden 1s → 3d transitions, which gain intensity from metal d–p orbital mixing when the metal ion symmetry deviates from centrosymmetry, making this region sensitive to the ligand field. Correlation of the pre-edge transition energies and intensities with the structure of investigated compounds can be performed using quantum chemistry (QM) calculations.⁴² However, the resolution of partial fluorescence XAS is limited by the 1s core-hole (created upon Mn 1s excitation in XANES) lifetime broadening, and higher resolution data are required for rigorous and quantitative comparison with calculated spectra.

Herein, Mn K_α high-energy resolution fluorescence detected (HERFD) spectra of the S₀–S₃ states are presented. The data provide for the first time a highly resolved picture of the pre-edge region on all S-states. Importantly, contrary to previous studies,^15,17,43 we use glycerol-free samples to avoid the glycerol-induced nonphysiological nonphysiological heterogeneity in the S₃ state.^21,44,45 By employing time-dependent density functional theory (TDDFT) to compute XAS spectra of a wide variety of OEC models, we achieve a detailed mapping between structural and spectroscopic features. Among other important conclusions regarding the chemical properties of distinct intermediates, correlation of the Mn XAS pre-edge region between experiment and theory shows that the HERFD spectra are consistent only with all-Mn(IV) oxo-hydroxo models of the S₃ state, consistent with magnetic resonance studies. This result rules out ligand-based oxidation during the S₂ → S₃ transition and locates the onset of the formation of the O–O bond past the final light-driven oxidation in the catalytic cycle.

Methodology

Photosystem II Purification and Preparation of EPR-XAS Samples

Photosystem II was purified from the thermophilic cyanobacterium Thermosynechococcus vestitus (previously named Thermosynechococcus elongatus) according to Kuhl et al.⁴⁶ Based on the final purification step, highly active PSII dimers from several preparations were pooled to reach a homogeneous sample base. The biochemical quality of the PSII dimers is routinely verified by Native PAGE electrophoresis, and oxygen evolution measurements are taken. The typical oxygen evolution activity is 5000 (±500) μmol O₂/h · mg Chl at +30 °C. Further verification of active PSII is obtained via S-state signal progression in the EPR data. For the present batch of samples, the standard purification protocols were followed and verified by Native PAGE and EPR, but explicit O₂ activity assays were not made for the present preparation. Given the identical EPR, however, which evidences S state advancement, a similar activity can be assumed. The final buffer contains 500 mM mannitol, 40 mM MES (pH = 6.5), 10 mM CaCl₂, 10 mM MgCl₂, and 0.03% v/v n-dodecyl β-D-maltoside. Phenyl-para-benzoquinone (PPBQ) dissolved in dimethyl sulfoxide (DMSO) was added as electron acceptor at a final concentration of 0.5 mM. Photosystem II (1.5 mM Mn, 15 μL) was loaded in XAS cells. The cells were designed to fit in the EPR resonator and the cryostat utilized for HERFD XAS measurements. Both surfaces of the cell were open and sealed with Mylar tape windows. The sample dimensions in the cell were 2.5 mm width and 8 mm length and 0.8 mm depth. Samples were synchronized in the S₁ state by a “pre-flash” and dark incubation at room temperature for 1 h. All samples were illuminated by a Nd:YAG laser (wavelength 532 nm, 10 ns pulse of 300 mJ) with 0, 1, 2, or 3 flashes and frozen in liquid N₂. The laser beam was split in two and the sample was illuminated from both sides.

EPR Quantification of S-States

The S-state advancement is imperfect.⁴⁷⁻⁵⁵ By using short intense laser pulses, we can avoid “double hits”, but it is impossible to avoid “misses”. The miss factor was calculated using the intensity of the multiline EPR signal of S₂ state in all samples (Figure S1), as previously reported by Messinger et al.¹⁵ Measurements were carried out by using a Bruker E500 spectrometer equipped with a Bruker ER 4116DM resonator and an Oxford Instruments ESR 935 cryostat. From the peak to peak intensity of the multiline (lines used marked in Figure S1a), the population of each S-state after 0, 1, 2, and 3 flashes was calculated (Tables S1, S2) and used in order to deconvolute the pure S₁, S₂, S₃, and S₀ XAS spectra. We note that after the fourth flash the intensity of the S₂ multiline signal is 80% compared to the first cycle (Figure S1b). The very good conversion factor confirms the high activity of the samples. Given that inactive OEC centers would have released Mn(II) in solution, all samples were checked for the presence of Mn(II) EPR signals and batches with high Mn(II) content were discarded. Crucially, the homogeneity of our S₃ state samples has been confirmed by W-band EPR spectroscopy (Figure S2).⁴⁴

HERFD Measurements

Mn Kα HERFD measurements were performed at beamline 6–2 of the Stanford Synchrotron Radiation Lightsource (SSRL).⁵⁶ The SPEAR storage ring was operating at 3 GeV with a current of 500 mA. BL6–2 utilizes a 56-pole 0.9 T wiggler insertion device. Two Rh-coated Si crystals (one flat for vertical collimating and one cylindrical for focusing, upstream and downstream of the monochromator) were utilized to focus the beam. The incident beam was monochromatized using a pair of cryogenically cooled Si(311) crystals, giving an energy resolution of ∼0.2 eV. The flux was estimated to be ∼2.9 × 10¹¹ photons/s in a 420 × 200 μm beam spot. The energy of the incident beam was calibrated by setting the maximum of the pre-edge peak of KMnO₄ to 6543.21 eV. Kα-HERFD and Kα XES data were collected by utilizing a Johann-type XES spectrometer equipped with five Ge(111) crystals and an energy-dispersive silicon drift detector. The spectrometer resolution at the elastic peak was 0.8 eV. The sample temperature during measurements was poised at 10 K by using liquid He-cooled flow cryostat. The beam shape was Gaussian with vertical full width at half height of 200 μm and horizontal of 420 μm. By using a vertical step of 300 μm and horizontal step of 630 μm, it was possible to measure ∼50 separate spots per sample. The Kα HERFD scans were collected in an energy range of 6530–6570 eV, while longer 6530–6800 eV scans were measured to facilitate proper normalization. The energy step was varied across the measurement and was 0.15 eV for the pre-edge region, which is the primary focus of the present study, and 0.2 eV across the edge. The Kα XES spectra were collected to assess the Kα maxima of each S state, and hence, short scans were used in the energy range of 5896–5904 eV.

Damage Studies

Detailed damage studies were performed before the actual measurement in all S states (Figure S3). They involve measurement of XAS scans in the range 6535–6570 eV in order to establish the maximum dwell time per sample spot. Due to the low sample concentrations, we exposed to radiation a number of spots and collected consecutive scans in each spot. We compared the average of all the first scans, to the average of all the second scans and so on, to establish the time point at which a change in the pre-edge and edge region is observed. Attenuation of the incident beam by using foils was also required. A range of attenuation factors was tested, and we concluded that 6% of the total flux was not damaging the sample during a scan time of ∼60s, while reasonable signal/noise was maintained. Additionally, time scans were performed at a constant incident energy of 6551 eV, the energy point that was considered as the most sensitive probe of damage during the initial damage scans. The dwell time for S₁ and S₂ states was determined as ∼100 s/spot while for S₃ and S₀ was ∼75 s/spot and was used during the actual measurement. Higher oxidation states of the Mn₄Ca cluster are expected to damage faster. This fact was verified by the damage studies, and a shorter scan (75 s vs 100 s/spot) was used for the S₃-state. We note that while a shorter dwell time for the formally more reduced S₀ may seem counterintuitive, this is due to the fact that the S₀ samples as prepared contain a significant amount of S₃ as quantified by EPR. A dose of 1.7 × 10⁷ photons/μm² was used for the S₁ state, 1.4 × 10⁷ photons/μm² for the S₂ state and 1.2 × 10⁷ photons/μm² for the S₃ and S₀ states.

HERFD Data Processing

Every scan was normalized to the incident flux I₀ and subsequently scans of the same flash number were averaged in PyMCA in order to create the average of 0, 1, 2, and 3 flash spectra (ca. 150 scans were averaged for the S₁ and S₂, while ca. 220–230 scans for the S₃ and S₀ states). All other processing was performed in Matlab. The short scans (6530–6570 eV) were normalized to the edge jump by using the long scans (6530–6800 eV). Spectra of pure S_i states were acquired by subtraction of the appropriate percentage of pure S_i–1 (and S_i–2 if needed) as calculated by EPR quantification of S-states after each flash (Tables S1 and S2). The rising edge of the pure spectra was fitted by interpolating a spline polynomial in the region 6530–6550 eV. By subtraction of the XAS spectrum minus the fitted line, the isolated pre-edge spectra of the S-states were acquired. Fitting of the pre-edge peaks was performed using Voigt curves (Lorentzian has only a small contribution) and least-squares fitting. The fact that the peaks represent Gaussian shape is expected since the 1s core hole lifetime broadening (Lorentzian) is effectively eliminated in a HERFD experiment, and the energy resolution is thus dominated by the instrumental broadening (Gaussian).

Evaluation of the Error

The error of the 0, 1, 2, and 3 flash spectra was estimated by the standard error of the mean of averaged spectra (Figure S4). This is a measure of how far our sample mean, i.e., scans of our experiment, is likely to be from the true population mean that would be infinite scans that would give the real spectrum. The standard error is calculated as where is the standard deviation, x̅ is the mean of scans x₁–x_n, and n is the total number of scans. After subtraction of the appropriate percentage of the S_i–1 state determined by EPR from the raw spectra in order to acquire the pure S_i state spectra and renormalization, the error was recalculated as , where is the standard error of the pure spectrum of the S_i state, X_Si is the fraction of the S_i state in the sample, while X_Si–1 is the fraction of the “missed” centers in the S_i–1 state that was subtracted, σ_Si is the standard error of the spectrum of S_i, and σ_Si–1 is the standard error of the S_i–1 spectrum.

Computational Details

Cluster models of the OEC were constructed from the X-ray crystallographic coordinates of the 5B66 structure monomer B.³⁵ The models include the inorganic core Mn₄CaO₅, terminal water molecules W1–W4, and first coordination sphere amino acids His332, Glu189, Asp342, Ala344, CP43-Glu354, Asp170, and Glu333. Second coordination sphere residues include His337, CP43-Arg357, Asp61, Ser169, Leu343, Tyr161, His190, Asn298, Gln165, Val185, and Phe186, and 13 crystallographic water molecules: 6 from the O1 water channel, 3 hydrogen bonding to W3 of the Ca²⁺ ion, 2 from the Cl^– water channel, and 2 from the O4 water channel. Overall, the S₁ state model (with W2 = H₂O) includes 329 atoms and is shown in Figure S5.

All calculations were performed with Orca 5.⁵⁷ Geometry optimizations were carried out using the B3LYP^58,59 functional with the resolution of the identity (RI) approximation including D4 dispersion corrections.⁶⁰ Optimizations were performed in the respective broken symmetry states for each model, i.e. α–β–α–β for S₁A, S₁HA, and S₀O5, β–α-α–α for S₁B, S₁HB, S₃P, and S₀HO4, α–β–β–α for S₁O5H, S₂A, S₂HA, and S₀O4, α–α–α–β for S₂B, S₂HB, S₂O4H, S₃AW, and S₃BW, α–β–α–α for S₀HO4, and α–α–α-α–β (O5–O6 radical antiferromagnetically coupled to Mn ions) for S₃O. Relativistic effects were considered throughout using the zeroth-order regular approximation (ZORA).⁶¹⁻⁶³ The scalar-relativistically recontracted⁶⁴ ZORA-def2-TZVP(-f) basis sets⁶⁵ were used for all atoms except C and H for which the ZORA-def2-SVP basis sets were used. In all calculations, the CPCM solvation⁶⁶ with dielectric constant ε = 6.0 was used to simulate the effect of the protein surrounding.

XAS spectra were calculated with the TD-DFT method employing the Tamm–Dancoff approximation.^67,68 The TPSSh functional⁶⁹ was used and the computed excitation energies were shifted to higher energies by 36.3 eV, to account for systematic methodological deviations, based on previous benchmarking studies on Mn monomers and dimers.^70,71 The accuracy of the method has been quantified in previous benchmark studies on mononuclear Mn complexes using the computed R² of the linear relationship between the calculated and experimentally determined transition energies (R² = 0.94) and intensities (R² = 0.88).⁷¹ The RI and chain of spheres (RIJCOSX) approximations⁷² were employed to speed up the calculations. The XAS spectra of each Mn ion were calculated with 150 roots, and the spectra of each model were plotted by adding the spectra of the four Mn ions and using a peak Gaussian broadening of 1.1 eV. In order to compare with the normalized intensities of the experimental spectra, all computed intensities were multiplied by 0.02, based on agreement between the maximum intensity of the 0F state with the computed intensities of the S₁ states.

Results and Discussion

HERFD Experiment

The Mn Kα HERFD X-ray absorption spectra of the dark-adapted S₁ state, the one-flash (S₂), two-flash (S₃), and three-flash (S₀) illuminated states of PSII samples from T. vestitus were measured at beamline 6–2 of the Stanford Synchrotron Radiation Lightsource (SSRL) (Methods section, SI) and are shown in Figure 3a. All spectra are normalized to the edge jump, as described in the Methods. Functionality of the samples is indicated by the period four oscillation in the S₂ state multiline g ≈ 2 EPR signal (Figure S1).

Figure 3.

(a) HERFD spectra of the dark-adapted samples (0F) and after 1–3 flashes (1–3F). Spectra are normalized to the edge jump. 3-point adjacent averaging was used. (b) Pure spectra of all S-states after the appropriate subtractions based on EPR quantification.

Progression of PSII samples to the next S-state upon illumination is not perfect; thus, quantification of the population of PS II centers of the sample in each S-state is essential. EPR spectroscopy was employed following a previously suggested protocol.¹⁵ Quantification of the flash-induced S-state progression efficiency was performed based on the intensity of the multiline g ∼ 2 EPR signal of the S₂ state after each flash.¹⁵ The S₂ population after each visible light flash was quantified, and a model considering misses for every flash illumination was fitted to the EPR data. The miss factor was calculated to be 6%. In other studies, miss factors of 6–18% have been reported.^{10,15,17,19,43} In a separate experiment, a singly flashed sample was illuminated continuously at 200 K in order to obtain the maximum of S₂ and no increase in the multiline signal intensity was observed. Double hits did not improve the fitting. A considerable population of PS II centers (15%) remains in the S₂ state and does not progress to S₃ and S₀ upon illumination. This is caused by acceptor side deficiency: the acceptor side in some PSII centers cannot accept more electrons in order for the OEC to proceed further than S₂. Single turnover values between ∼5 and 10% of PSII centers have been reported elsewhere.^10,15 A percentage (∼5%) of centers in S₂ state in the dark adapted sample has also been considered, but the inclusion of such population did not improve the fitting (see also SI, section 1 and Table S1). The population of each S-state after each laser pulse is presented in Table 1. After “0, 1, 2, 3 flashes”, there is 100% S₁, 94% S₂, 74% S₃, and 70% S₀, respectively.

Table 1. S-State Content (%) after a Given Number of Flashes.

flash	S1	S2	S3	S0
0	100
1	6	94
2	1	25	74
3	0	16	14	70

The optimal fitting with the experimental data was achieved by a miss factor of 6% and assuming that a 15% population of the OEC centers remains blocked in the S₂ state. The pure S-state spectra derived after deconvolution according to the S-state quantification are shown in Figure 3b. The spectra consist of the rising edge at 6545–6560 eV, that represent 1s → 4p transitions, and the pre-edge at 6538–6545 eV, attributed to dipole forbidden 1s → 3d transitions. Each of the regions is examined separately in the next sections.

Analysis of the Edge Region

For systems with similar coordination environment, the energy of the edge is dependent on the oxidation state of the metal ion: as the oxidation state increases the edge shifts toward higher energy. Visual inspection of the edge regions in Figure 3b shows that the edge shifts to higher energies upon the S₀ → S₁ and S₁ → S₂ transitions. For the S₁ → S₂ transition, the edge shift can be attributed only to the more positive charge of the cluster since the structures of S₁ and S₂ are similar, as proposed by EXAFS^17,43 and most recently by the XFEL crystal structures.²⁹ The S₀ → S₁ transition involves deprotonation in addition to oxidation; therefore, the observed edge shift is attributed to Mn-centered oxidation as well as small structural rearrangements. Thus, in agreement with previous works,^15,17,43 positive edge energy shifts are observed during the S₀ → S₁ and S₁ → S₂ transitions, which reflect Mn-based oxidations.

The edge shift during the S₂ → S₃ transition is not equally obvious from visual inspection because the shapes of the edge of S₂ and S₃ states are significantly different, as shown also in the first derivatives of the spectra (Figure S6). Those differences possibly stem from significant structural changes involved in the S₂ → S₃ transition, which include deprotonation and possibly the insertion of a water molecule in the cluster. Thus, in S₂ → S₃, the edge shift is dependent on the oxidation state changes as well as the ligand field change; since the magnitudes of these effects are comparable, this complicates the interpretation of the edge as it has been shown in Mn model systems.⁷³ A change in the coordination environment of a Mn ion from 5-coordinate Mn(III) to 6-coordinate Mn(IV) was proposed based on comparison of Kβ emission difference spectra of the S₃ minus the S₂ state with the difference spectra of synthetic models with Mn(III)L₅, Mn(IV)L₅, and Mn(IV)L₆.⁹

Three different methods were used for assessing the energy of the edge in XAS experiments: (a) the position at the half intensity of the normalized spectra,⁷⁴ (b) the integral method,⁷⁵ and (c) the energy at the inflection point.¹⁵ In Table 2, the energies of the edge and the edge shifts determined using all of the above methods are given. The half intensity energy point, after proper normalization using the long scans up to 6800 eV, shifts ∼0.7 eV with each S-state transition from S₀ to S₃. Similar results are obtained with the integral method, which gives a shift of ∼0.6 eV. By contrast, the inflection point of the edge, determined by the second derivative (Figure S7), is almost the same for the S₂ and S₃ states. Notably, the present results for the inflection point values are similar to those reported by Messinger et al.,¹⁵ while those reported with the integral method are consistent with Haumann et al.¹⁷ As shown in Figures S8 and S9, the present XANES spectra are very similar in the rising edge region to the previously reported XANES measured on spinach^15,17 and cyanobacteria.⁴³ The inflection point method is more sensitive to the shape of the edge since it is defined by a single point. Thus, the shift of the inflection point energy is not consistent with the other two methods and the different methods used by the different groups for assessing the edge shift may cause the discrepancy in the literature regarding the presence^17,74,76 or near absence^15,77 of edge shift during S₂ → S₃ and thus if Mn-centered or ligand-based oxidation occurs during this transition.^{12,22,32,38,41,78−81}

Table 2. Edge Energies and Edge Shifts (S_i–S_i-1, in Parentheses) Obtained with Three Different Methods^a.

	half of normalized Intensity	integral method	inflection point
S0	6550.8	6551.6	6550.5
S1	6551.4 (0.6)	6552.1 (0.5)	6553.1 (2.6)
S2	6552.2 (0.8)	6552.7 (0.6)	6553.7 (0.6)
S3	6552.9 (0.7)	6553.2 (0.5)	6553.9 (0.2)

^aAll energies in eV.

Overall, all methods of determining the edge energy confirm Mn-based oxidation in the S₀ → S₁ and S₁ → S₂ transitions. On the other hand, diverging results are obtained in the case of the S₂ → S₃ transition by different methods, which has given rise to seemingly contradictory interpretations among different groups, despite their data being almost identical. Given the additional complication of structural differences between the S₂ and S₃ states, the locus of oxidation in this transition cannot be conclusively determined from the analysis of the edge region. Therefore, in the next sections we focus on the analysis of the pre-edge region, which more directly correlates with the electronic structure, and where the high resolution of the HERFD spectra combined with QM calculations provide an unambiguous interpretation.

Analysis of the Pre-edge Region

Figure 4 shows the expanded pre-edge region for all S-states. Before analyzing the current data in more detail, it is important to compare these spectra to previous XAS data. As discussed above, the XANES region of the HERFD spectra are essentially identical to previous reports.^15,17,43 However, as seen in Figure S8, the ratio of intensities in the pre-edge region is modulated relative to previous reports. First, we note that the present data represent the highest resolution XANES data available for all S states. While 1s2p RIXS planes of the S-states were previously reported by Glatzel et al.,⁸² these data do not include the full XANES and thus do not allow for proper normalization. Further, the previous data were collected with a Si(111) monochromator relative to the Si(311) monochromator used in the current study. Hence, we estimate the previous report would have a resolution of ∼1.3 eV at constant emission energy relative to the ∼0.8 eV resolution in the current study. These differences in resolution, however, are unlikely to account for the differences in the pre-edge ratios. More likely, the observed differences are attributed to the fact that no glycerol was used in the present study. As glycerol is known to change the equilibrium of the different configurations of the S states, we hypothesize that this may be the origin of the modulations in the pre-edges. This is consistent with previous studies by Yano et al.,⁴⁵ which showed that near-infrared conversion of the S₂ low spin multiline signal to a high-spin EPR signal is correlated with a decrease in the intensity of the first pre-edge feature. These observations thus further motivate the importance of having the present data set for all S-states in the absence of glycerol. In order to quantitatively correlate the present data with theory, we proceeded to a detailed fitting analysis.

Figure 4.

(a) Fitting of the pre-edge region with Gaussian peaks and the rising edge background. (b) Pre-edge region of each S-state after the subtraction of the rising edge background. (c) S-state pre-edge XAS spectral differences. The dashed lines show the zero value for each difference spectrum.

The pre-edge region of each state was fitted by Gaussian curves, and the optimal fittings are shown in Figure 4a. The pre-edge regions of the S₀ and S₁ states can be best described by three features: a lower intensity peak at 6540 eV and two higher intensity peaks at 6541 and 6543 eV. The S₂ and S₃ pre-edge were fitted by only two peaks at 6541 and 6543 eV. Alternative fittings are given in Figure S10 and Table S3.

Changes in the pre-edge region with S-state progression can be quantified based on the areas of the individual peaks and of the overall pre-edge region, given in Table 3. The intensity of the first peak decreases from the S₀ to the S₁ peak, and the peak is not detectable at all at S₂ and S₃. The second peak at 6541 eV becomes more intense in S₁ than in S₀ and has a similar intensity in S₁–S₃. A subtle peak is also observed at ∼6546 eV in the S₀ spectrum partially covered by the rising edge. The third peak at 6543 eV increases from S₀ to S₁ and from S₂ to S₃, but not in S₁ to S₂. In order to compare solely the contributions to the XANES spectra due to the 1s → 3d transitions, the contribution of the rising edge to the pre-edge was subtracted. The spectra of S₀–S₃ with the subtracted baseline are shown in Figure 4b. The pre-edge region changes with S-state progression are shown in the difference spectra in Figure 4c. Figure S11 indicates the standard errors of the difference spectra.

Table 3. Energy in eV and Area (in parentheses) for Each Fitted Peak in Each S-State Pre-edge Spectrum and Intensity Weighted Average Energy (IWAE) and Area of the Total Pre-edge Region.

	peak 1	peak 2	peak 3	total
S0	6539.9(0.031)	6540.9(0.082)	6542.5(0.079)	6541.4(0.208)
S1	6539.7(0.008)	6540.9(0.118)	6542.7(0.115)	6541.6(0.235)
S2		6540.8(0.108)	6542.8(0.108)	6541.8(0.217)
S3		6541.0(0.113)	6542.9(0.121)	6542.0(0.238)

The changes in the overall shape of the pre-edge region during the S₀–S₃ state transitions can be expressed through the intensity weighted average energies (IWAEs). The IWEAs shift 0.2 eV higher after each transition, which imply that higher energy transitions gain intensity with each Mn oxidation. Notably, those changes cannot be explained solely in terms of Mn oxidation during each transition, as in the case of the edge region. Apart from indirect correlations between pre-edge intensities and the presence of specific groups,⁸³ a detailed correlation between structure and spectroscopy requires QM calculations.^{70,71,83−86} Therefore, in the next sections, we attempt to correlate the pre-edge features with structural features using QM-derived models of all states.

Correlation with Structural Models

To correlate the XAS spectra with the structure of the OEC in each state, we optimized models of several variants of the S₀–S₃ states, and we calculated their XAS spectra. The core parts of all optimized structures are shown in Figure 5, along with explanations of the labeling used in this work. Additionally, the inorganic cores of all models are depicted schematically in Figure S12 where the Jahn–Teller axis orientations for all Mn(III) ions are depicted and important distances are indicated. For the S₁ state, we considered five models with variations in the protonation states of terminal water-derived ligands on Mn4 and the O5 bridge, as well as in the orientation of pseudo-Jahn–Teller elongation axes of the Mn(III) ions.⁸⁷ In all optimized S₁ state models, the Mn oxidation states are III–IV–IV-III. For the S₂ state, we considered five models that take into account all different ideas discussed in the literature regarding protonation states of terminal water-derived ligands and the O4 bridge,⁸⁸ as well as valence isomerism (i.e., Mn oxidation states III–IV–IV−IV and IV–IV–IV−III).⁸⁹ For the S₃ state, we considered all possibilities discussed in the introduction regarding Mn and ligand oxidation states, i.e., Mn IV–IV–IV−IV with oxo and hydroxyl O5 and O6 ligands, III–IV–IV−IV with oxo-oxyl radical, or III–IV–IV−III with peroxo O–O bond formation, in total 5 models. Finally, for the S₀ state we considered 4 models differing in the protonation states of W2 and of the O4 and O5 bridges.^90,91 The TD-DFT calculated XAS spectra of all 19 optimized structures are shown in Figure S13, compared to the experimental spectra for each S-state.

Figure 5.

Optimized core structures of the computational models of S₀–S₃ states considered in this work. For clarity, only the inorganic core and a few selected coordinating ligands (of the actual ∼330 atom models) are depicted in this figure. Mn(III) ions are shown in light pink, and Mn(IV) ions are in dark purple. Label superscripts A and B are for “open” and “closed” cubane geometries, respectively; H superscript denotes that W2 is in the aquo form (H₂O), otherwise, it is in the hydroxo form, and O4 and O5 denote that bridging oxo ligands O4 and O5, respectively, are protonated in the specific models. The models whose labels are in colored boxes turned out to be in best agreement with experiment, as explained in the following sections.

Analysis of Difference Spectra

The difference spectra between successive S-states are substantially more informative compared to the individual spectra, as it removes systematic errors in the intensity evaluation and explicitly isolates the features that change from one state to the next. The results show that in some cases where the similarities in computed spectra are significant, the difference spectra enable the confident selection of the best fitting structural models.

Figure 6 compares the S₂–S₁ difference spectra between experiment and theory for the selected computational models. Importantly, the comparison of difference spectra using distinct computational models is so sensitive that it is possible to distinguish between S₁ state isomers in which the orientation of the Jahn–Teller axis of the Mn4(III) ion is either perpendicular or collinear to the Mn1(III) Jahn–Teller axis (S₁^A,H and S₁^B,H, respectively),⁸⁷ but also between different protonation states of the terminal W2 ligand in both the S₁ and S₂-state models. For example, the S₂–S₁ difference spectra for the S₁ models with perpendicular Jahn–Teller axes (models S₁^A and S₁^A,H) show a large negative intensity difference at 6542–6543 eV, inconsistent with experiments, whereas this is well reproduced with model S₁^B,H. Interestingly, the observed negative intensity difference in the region 6539–6540 eV is only reproduced by S₁ models that have W2 in the aquo form (in agreement with FTIR studies⁹²), but not by those with W2 in the hydroxo form. This is associated with the shoulder at 6539.5 eV that is observed for the S₁ models where W2 is H₂O (Figure S13a).

Figure 6.

Experimental and calculated S₂–S₁ spectral differences for different S₁ and S₂ isomers.

The models that best reproduce the experimental S₂–S₁ difference spectra are S₁^B,H and S₂^A,H (Figure 6a, blue line). The comparison shows the excellent agreement between the experiment and calculations. It is worth pointing out that the calculated XAS spectrum of S₂^B,H is slightly lower in intensity than S₂^A,H (Figure S13b), similar to the S₂ high-spin and low-spin forms reported by Chatterjee et al.⁴⁵ By contrast, the computed spectrum of the S₂ state model with protonated O4, (S₂^O4H), which has been suggested as a possible interpretation of the high-spin S₂ form,^88,93 is more intense than S₂^A,H (Figure S13b). Therefore, our experimental pre-edge results combined with TD-DFT calculations are most consistent with valence isomerism⁸⁹ giving rise to the low and high spin S₂ state signals, respectively. As the samples were not treated with glycerol,⁹⁴ about 20% of the centers are in the high-spin S₂ state (quantification based on EPR, see SI). Since the fraction of high-spin S₂ is small, the effect on the S₂–S₁ difference spectra in the pre-edge region is negligible (Figure S14) and thus we use only the dominant S₂^A,H spectrum for the difference spectra plots.

Comparison of the calculated S₁–S₀ difference spectra using different computational models, in Figure S15, shows that model S₀^O4 exhibits the best fitting with experiment among the different S₀ variants. Notably, light-induced Fourier transform infrared (FTIR) S₁–S₀ difference spectra reported by Yamamoto et al.⁹⁵ also indicate O4 protonation in the S₀ state; however, that study also supports a fully protonated W2, which would be represented by our S₀^O4,H model. This would indeed be more consistent with one electron oxidation and one deprotonation (of O4) during the S₀ → S₁ transition, and it needs to be further investigated. Previous benchmarking studies on Mn oxo-bridged compounds have also demonstrated the sensitivity of Mn XAS to oxo-ligand protonation states.⁷⁰ Therefore, analysis of the S₀–S₂ state transitions shows that the pre-edge region combined with quantum chemistry simulations not simply confirms Mn-based oxidation but directly constrains the possible structural features of each state.

Even more pronounced than the S₁ and S₂ states are the differences observed among computational models of the S₃ state in the case of S₃–S₂ difference spectra shown in Figure 7. The comparison reveals that the Mn(IV)₄ oxo-hydroxo models (S₃^A,W and S₃^B,W) are clearly in better agreement with experiment than the Mn(III)Mn(IV)₃ oxyl-oxo (S₃^oxyl–oxo) and Mn(III)₂Mn(IV)₂ peroxo (S₃^peroxo) models. The S₃^oxyl–oxo and S₃^peroxo are completely inconsistent because S₃^oxyl–oxo shows much higher intensity than S₂ between 6541 and 6544 eV, whereas S₃^peroxo shows much lower intensity. Among the Mn(IV)₄ oxo-hydroxo structural isomers, S₃^A,W and S₃^B,W, the open cubane S₃^A,W shows the closest agreement with experiment. The above results strongly disfavor ligand-based oxidation during the S₂ to S₃ transition.

Figure 7.

Experimental S₃–S₂ data and calculated S₃–S₂ difference spectra for models S₃^A, W, S₃^B, WS₃^oxyl–oxo, S₃^peroxo, and S₂^A, H.

Analysis of S-State Progression

In Figure 8, we show the calculated spectra of the most experimentally consistent models of each state, as identified from the analysis of the difference spectra (structures S₁^B,H, S₂^A,H, S₃^A,W, and S₀^O4, as labeled in Figure 5). The most experimentally consistent model of the S₃ state (S₃^A,W) is an all-Mn(IV) model, and hence, the XAS pre-edge region is consistent with Mn-based oxidation in each S-state transition. Importantly, the same model is the one that fully explains observations by electron–electron double resonance (ELDOR) detected nuclear magnetic resonance experiments (EDNMR).^20,21

Figure 8.

Calculated Mn K-edge XAS spectra for the most experimentally consistent models of states S₀ to S₃.

The calculated spectra in Figure 5 reproduce all of the major features of the experimental spectra. Specifically, (a) the shoulder at 6539.5 eV is smaller in the S₁ state than in the S₀ state and is absent in the S₂ and S₃ states, as in experiment; (b) the peak above 6541 eV has the smallest intensity in the S₀ state and the largest in the S₃ state; (c) the intensity in the region between 6542 and 6544 eV increases with each S-state transition; and (d) the intersection points of all spectra at 6540.5 and 6544 eV, and of the S₁ and S₂ states spectra at 6542.7 eV are also reproduced. Notably, small deviations from experiments in the relative intensities among the S-states are within the accuracy limits of our methodology and can be more easily distinguished in the corresponding difference spectra in Figures 6, 7 and Figure S15.

It is important to note that direct visual comparison of the calculated spectra of a specific state with experimental spectra is not expected to lead to useful conclusions; instead, it is meaningful to compare the total pre-edge area as well as the IWAEs to experiment, as stressed in previous studies.^70,71,96 Each spectrum results from the distribution of energies and intensities of a vast number of transitions from different Mn ions, and the height and width of each apparent peak of the spectrum are very sensitive to small variations in the relative energies of the transitions it contains. Moreover, it is amply established that errors in the (relative) intensities of the transitions are larger than errors in the energies.⁷¹ It is precisely for these reasons that the pre-edge area is considered a far better parameter than individual peak heights to assess agreement with experiment.⁹⁶ Indeed, an excellent correlation is observed between summed calculated intensities and experimental pre-edge areas, as presented in Figure 9a. The linear relationship (R² = 0.98) shows a strong correspondence between experiment and theory. It also implies that intensity errors are similar among different S-states, which further justifies the use of difference spectra to probe S-state progression. By contrast, the transition intensities of the ligand-based oxidized alternatives for the S₃ state, S₃^oxyl–oxo and S₃^peroxo (Figure 2), deviate strongly. The computed spectra also reproduce the shift of the IWAEs to higher energy with the S-state progression (Figure 9b).

Figure 9.

(a) Correlation of summed calculated intensities and experimental pre-edge areas and (b) correlation of calculated and experimental intensity-weighted average energies for different OEC models in different S-states. The correlations show that only Mn-centered oxidation in the S₂ → S₃ transition is consistent with the HERFD data.

Electronic Structure Origin of the Observed Spectral Features

The contributions of each Mn ion to the calculated XAS spectra of the most experimentally consistent models of each S-state are shown in Figure 10, where the stick spectra of each Mn ion are plotted in different colors for each structure. Analysis of the TD-DFT natural transition orbitals (NTOs) reveals the nature of the underlying transitions that form the pre-edge region. The NTOs of S₁^B,H, S₂^A,H, S₃^A,W, and S₀^O4 as well as of the redox isomers of the S₃ state, S₃^A,W, S₃^oxyl–oxo, and S₃^peroxo, are shown in Figures S16–S21. We observe that the nature of the transitions in each energy region remains essentially the same with S-state progression; thus, in Figure 10 we focus on the origins of the differences between successive S-states, and we show the NTOs that contribute the most to those differences.

Figure 10.

Assignment of the calculated pre-edge XAS spectrum based on the NTOs associated with the transitions for models S₀^O4, S₁^B,H, S₂^A,H, and S₃^A,W. In the stick spectra, transitions from 1s core orbitals of Mn1 are shown in green sticks, of Mn2 in red, of Mn3 in blue, and of Mn4 in orange. NTOs for transitions from 1s core orbitals of Mn1 are shown in green boxes, of Mn2 in red, of Mn3 in blue, and of Mn4 in the orange box.

Specifically, transitions with energy lower than ∼6542 eV are mostly attributed to excitations from the 1s core orbitals to 3d with dominant local Mn character, while transitions with energy higher than 6542 eV also have metal-to-metal charge transfer character. Transitions from the 1s to nonbonding t_2g orbitals have very low intensity due to negligible metal 4p mixing, while transitions to the antibonding d_z² and d_x²–y² orbitals contribute significantly to the intensity due to increased metal 4p mixing, which is covalently mediated by σ interacting oxygen 2p orbitals. The most intense spectral differences are observed in regions A–D, denoted with pink, purple, and blue hues in the stick spectra plots in Figure 10.

In S₀^O4, the peak around 6540 eV (region A) is mostly attributed to local Mn4(III) and Mn3(III) 1s → 3d transitions, whereas in S₁^B,H the latter is blue-shifted due to Mn3(III) oxidation to Mn3(IV), and it appears in region B. In addition, Mn3(IV) transitions in regions B, C, and D gain intensity in S₁^B,H, which explains the positive S₁–S₀ difference spectra in those regions (Figure S15).

As in the S₀ → S₁ transition, Mn4(III) oxidation to Mn4(IV) during the S₁ → S₂ transition blue-shifts the local Mn4 1s → 3d transition from region A in S₁^B,H to region B in S₂^A,H, and local Mn4(IV) transitions gain intensity in regions B and C. However, the local Mn3(IV) transition at 6541.5 eV has decreased intensity in S₂^A,H relative to S₁^B,H, probably due to decreased Mn 3d–4p orbital mixing as a result of coordination geometry changes, leading to negative S₂–S₁ difference spectra in region B. Furthermore, a metal-to-metal charge transfer Mn3(IV) 1s → Mn4(IV) 3d transition arises in region D. This qualitative analysis reveals that the effect of metal-based oxidation on the XAS pre-edge region of the OEC is essentially due to increased intensity of transitions that involve the oxidized Mn(IV) ion, consistently with previous reports,⁹⁶ as well as due to coordination geometry changes on the rest of the Mn ions of the cluster.

During the S₂ → S₃ transition, Mn1(III) oxidation to Mn1(IV) blue shifts local Mn1 transitions, leading to negative S₃–S₂ difference spectra at 6540–6541 eV. Mn1(IV) transitions are more intense in S₃^A,W than S₂^A,H in regions B, C and D, whereas Mn4(IV) transitions are less intense. The effect of those changes is reflected in the calculated S₃–S₂ difference spectra, which is in excellent agreement with experiment (Figure 7). Interestingly, the predicted pre-edge regions of the alternative S₃-state models S₃^oxyl–oxo and S₃^peroxo are strikingly different.

The stick spectra of S₃ state isomers S₃^A,W, S₃^oxyl–oxo, and S₃^peroxo are compared in Figure 11. Comparison of the stick spectra reveals that the large intensity difference between S₃^oxyl–oxo and S₃^A,W peaks is attributed to regions B, C, and D. Notably, region A has a local Mn4(III) character. The NTOs that correspond to peaks B–D are shown next to each plot in Figure 11. NTOs that correspond to excitations from Mn1 1s core orbitals are shown in green boxes and from Mn4 in the orange box. In addition, the calculated spectra of individual Mn ions are compared in Figure S22. The higher intensity of the transitions of the S₃^oxyl–oxo model compared to the S₃^A,W in the 6541–6544 eV region is attributed mostly to excitations from Mn1 and Mn4 1s core orbitals (Figure S22). Comparison of the NTOs of the B, C and D peaks of S₃^oxyl–oxo and S₃^A,W shows that (a) the local 1s to 3d transitions of Mn1 (first row of S₃^oxyl–oxo NTOs in Figure 11) are more dipole-allowed due to the increased p character of the acceptor orbitals and are thus more intense than the corresponding S₃^A,W peaks, and (b) S₃^oxyl–oxo additionally has three intense charge transfer excitations from Mn1 and Mn4 to O5 and O6 p orbitals (second row of S₃^oxyl–oxo NTOs). Thus, examination of the NTOs of S₃^A,W and S₃^oxyl–oxo shows that the higher calculated intensity for the S₃^oxyl–oxo model is attributed to charge transfer channels enabled by the oxyl-oxo radical.

Figure 11.

Assignment of the calculated pre-edge XAS spectrum based on the NTOs associated with the transitions for models S₃^A,W, S₃^oxyl–oxo, and S₃^peroxo. In the stick spectra, transitions from 1s core orbitals of Mn1 are shown in green sticks, of Mn2 in red, of Mn3 in blue, and of Mn4 in orange. NTOs for transitions from 1s core orbitals of Mn1 are shown in green boxes, of Mn3 in blue, and of Mn4 in orange box.

The lower intensity of the transitions in S₃^peroxo compared to S₃^A,W in the 6541–6544 eV region is traced to the same B, C, and D peaks of S₃^A,W as in the comparison between S₃^A,W and S₃^oxyl–oxo. The corresponding peaks in S₃^peroxo have lower intensity, in the case of B and C, or do not exist at all, in the case of D. In the S₃^peroxo acceptor, orbitals of local excitations of Mn1(III) and Mn3(IV) have less p character than the corresponding excitation in S₃^A,W, indicating that the peroxo unit hinders charge transfer channels, contrary to the oxyl-oxo and oxo-hydroxo groups.

In summary, a comparison of the experimental and calculated pre-edge XAS spectra supports Mn(III) to Mn(IV) oxidation in each S-state transition. Ligand-based oxidation in the S₂ → S₃ transition would result in either significantly more intense (oxyl formation) or significantly less intense (peroxo formation) transitions, therefore this possibility can be safely excluded based on the present HERFD data.

Conclusions

We have presented HERFD spectra of all S-states of PSII from T. vestitus. Spectra were collected as “0, 1, 2, and 3 flashed” samples and deconvoluted to pure S₁, S₂, S₃ and S_0, respectively, with the aid of EPR spectroscopy. The energy of the edge shifts to a higher energy during the S₀ to S₁ and S₁ to S₂ transitions, which clearly reflects Mn-based oxidation. However, the edge shift cannot be used reliably to determine the locus of oxidation during the S₂ to S₃ transition, due to inherent complexity of this oxidation reaction as well as due to quantification method limitations. Therefore, we focus on the pre-edge region, leveraging the high-resolution achieved by the HERFD technique and quantum chemical calculations on large DFT-optimized cluster models of the OEC.

The relative intensities of the pre-edge peaks change clearly with the S-state progression. Using quantum mechanical results, we compared different structural variants of each state on the basis of the difference spectra, the sum of intensities, and the IWAEs. Our results demonstrate that this QM-supported analysis of the pre-edge is highly sensitive and can discriminate even between structures with minor differences such as terminal water ligands in different protonation states. The best fitting model for the S₁ state is one with collinear pseudo-Jahn–Teller axes on Mn1(III) and Mn4(III) ions and with both the W1 and W2 terminal ligands of the Mn4 ion in the aquo form. The S₂ state is most consistent with an “open-cubane” conformation with an Mn1(III) ion and both Mn4-bound W1 and W2 in the aquo form. In the S₀ state, protonation of the O4 gives a better fitting with the XAS spectra than protonation of the O5. Crucially, the computed Mn pre-edge spectra are starkly different for distinct variants of the S₃ state and support exclusively the Mn(IV)₄ oxo-hydroxo formulation as the species observed experimentally, excluding the Mn(IV)₃Mn(III) oxyl-oxo and Mn(IV)₂Mn(III)₂ peroxo models. Electronic structure analysis shows that if an oxyl-oxo group were present, the pre-edge intensity would be significantly higher because it would enable charge-transfer channels from the 1s core orbitals of Mn1 and Mn4 to the corresponding valence 3d orbitals, which would have higher p mixing. By contrast, the presence of a peroxo group in the S₃ state would have exactly the opposite effect on the transition intensities, due to blocking of Mn1 and Mn3 charge transfer transitions attributed to lower p mixing. Importantly, this is fully consistent with magnetic resonance data, where the manganese hyperfine tensors are all nearly isotropic, which is only possible if they are all Mn(IV).²⁰

By combining high-resolution XAS data with QM calculations, we have clearly shown that recent controversies regarding the nature of the S₃ state can be rigorously resolved and that presence of either peroxo or oxo/oxyl level intermediate can be ruled out. An all-Mn(IV) S₃ state should be considered as the starting intermediate in the interpretation of experimental data on the S₃ to S₀ transition.^97,98 Overall, the results of this work clearly show Mn-based oxidation during the S₂ to S₃ transition and constrain the possible mechanisms of the formation of the O–O bond to those that initiate after the final light-driven oxidation. Beyond the importance for biological water oxidation, the cumulative metal-centered storage of oxidizing equivalents supported by the present study directs our attention to synthetic water oxidation catalysts that leverage multimetallic cooperativity.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.3c06046.

• Additional experimental details; representative EPR spectra and experimental multiline signal intensity after 1–5 flashes (Figure S1); calculated multiline intensity in comparison with the experimental using different combinations of fitting parameters (Table S1); calculated S state content after a given number of flashes (Table S2); inhomogeneity of the S₂ and S₃ states as indicated by X-band and W-band EPR spectroscopy (Figure S2); damage test in 0, 1, 2, 3 flashed samples (Figure S3); standard error of the 0, 1, 2, 3 flashed spectra and of the pure S₁, S₂, S₃, S₀ spectra (Figure S4); QM model of the S₁ state (Figure S5); first derivative of the XANES spectra (Figure S6); second derivative of the XANES spectra (Figure S7); HERFD spectra of this work in comparison with literature data (Figure S8 and S9); fitting of the pre-edge experimental peaks with Voigt curves (Figure S10); parameters of the fitting of the pre-edge region (Table S3); standard error of the S_i–S_i-1 differences (Figure S11); schematic depiction of the inorganic cores of all optimized S₀–S₃ models (Figure S12); calculated XAS spectra of all models (Figure S13); comparison of calculated S₂–S₁ difference spectra for models S₁^B,H and S₂^A,H to those for models S₁^B,H and a mixture of 80% S₂^A,H and 20% S₂^B,H (Figure S14); experimental and calculated S₁–S₀ spectral differences for different of S₀ and S₁ isomers (Figure S15); assignment of the calculated pre-edge XAS spectra based on NTOs for models S₁^B and S₁^B,H (Figure S16), model S₂^A (Figure S17), model S₃^A,W (Figure S18), model S₃^oxyl–oxo (Figure S19), model S₃^peroxo (Figure S20), and model S₀^O4 (Figure S21); calculated XAS spectra for individual Mn ions for S₃^A,W, S₃^B,W, S₃^oxyl–oxo, and S₃^peroxo (Figure S22); example of ORCA input file (PDF)

• Cartesian coordinates of all structural models discussed in this work (TXT)

Open access funded by Max Planck Society.

The authors declare no competing financial interest.