Site-specific electronic structure of covalently linked bimetallic dyads from nitrogen K-edge x-ray absorption spectroscopy

Molecular donor/acceptor assemblies for solar energy capture and fuel production are often built of combined chromophore, bridge, and catalytic subunits designed to promote the unidirectional charge separation and multiple charge accumulation necessary for artificial photosynthesis.^1–8 The efficacy of these molecular assemblies as multi-electron multi-proton photocatalysts depends not only on the functionality of the individual parts but also on the intricate donor–acceptor interactions between the chromophore–bridge–catalytic subunits, as well as on how these interactions are modulated by the presence of electrons and protons on the complex.^9,10 Transition metals are readily incorporated into the design of such assemblies due to their tunable electronic structures and geometries, resulting in broad interest in assemblies of transition metal complex chromophores and catalysts, covalently linked using a conductive bridge, to facilitate efficient light-induced directional electron transfer.^2,4–7

Understanding the mechanistic roles of the molecular donor, acceptor, and bridging subunits in solar energy conversion necessitates understanding the electronic structure in both the ground state and reactive photo-excited states with high spatial sensitivity (localization to molecular moieties) as well as specificity to the donor and acceptor valence orbitals that participate in charge transfer. The mechanisms for charge transfer that underpin energy conversion processes for large, multinuclear molecular assemblies are often studied with optical spectroscopy methods.^11–21 While optical spectroscopy can be applied with high time resolution to capture the ultrafast dynamics of charge migration, optical spectra are not explicitly site selective as the probed transitions are between delocalized valence orbitals. Furthermore, the interpretation of such measurements can be complicated by the overlapping optical features of the donor and acceptor moieties¹³ or by the absence of visible wavelength spectral features for some commonly employed molecular catalysts.^11,22,23 Here, we instead pursue N K-edge x-ray absorption near-edge structure (XANES) spectroscopy as a probe of electronic structure for covalently bridged chromophore/catalyst assemblies, as conceptually illustrated in Fig. 1. We demonstrate site-specific spectral features to separately probe the unoccupied ligand-centered orbitals of the chromophore, catalyst, and linking bridge moieties that directly participate in the light-driven electron transfer and charge accumulation in these systems.

XANES is an element-specific technique that probes transitions between atomic core levels and the unoccupied valence orbitals of a molecule. Metal L-edge and K-edge XANES spectroscopies (along with other metal edge x-ray spectroscopies) have been successfully employed to probe the excited state charge redistribution between two distinct metal sites in heterobimetallic donor–acceptor assemblies.^{13,19,24–26} Here, we instead focus on the use of ligand atom K-edge XANES spectroscopy, which provides an additional tool to monitor charge localization beyond the metal coordination spheres and can resolve the intra-ligand charge transfer processes that underpin directional charge transfer across the assembly. Nitrogen K-edge spectroscopy is especially relevant to the field of molecular solar energy conversion due to the ubiquity of N-containing ligands in synthetically designed charge-transfer systems, both in metal coordinating ligands and in conductive linker molecules.^2,9 The N K-edge XANES spectra of aromatic heterocyclic ligands are characterized by a sharp and high intensity pre-edge peak attributed to the transitions of N 1s core electrons to the lowest unoccupied π* orbital of the ligand (illustrated schematically in Fig. 1). This makes the N K-edge pre-edge feature a direct probe of the electron-accepting orbitals involved in metal–ligand or ligand–ligand charge transfer. N K-edge XANES has been successfully applied to investigate the electronic structure in the ground and/or excited states of electron-accepting ligand-based orbitals in metalloporphyrins and metal–polypyridyl complexes.^27–33 The energies and intensities of the N pre-edge absorption features of transition metal complexes have been assessed to reveal the metal-dependence of local charge on ligand N atoms,^27,28 changes in metal–ligand covalency,^30,34 and time-dependent changes in ligand electronic structures following optical excitation of charge transfer excited states.^29,31,35

Herein, we extend N K-edge XANES to multi-valent systems with multiple distinct nitrogen sites contained in both metal-bound and bridging polypyridyl ligands. Specifically, we investigate donor/acceptor assemblies that contain the conductive bridging ligand tetrapyrido[3,2-a:2′,3′-c:3″,2″-h:2‴,3‴-j]phenazine (tpphz), which has been used extensively to link transition metal chromophore and catalytic complexes.^9,15,17,36 The tpphz bridge is characterized by a low-lying acceptor orbital localized on the central phenazine motif.^15,17,37 This serves to enable electron-hopping-type charge-transfer mechanisms in tpphz-bridged dyads, as well as to facilitate the accumulation of multiple electrons within the assembly. The presence of the extended π-system in tpphz has also been found to improve ligand stability against dissociation.⁵⁸ Given the centrally located N atoms of the tpphz bridge, we aim to identify if N K-edge XANES can provide a local probe of the electron-accepting orbitals on the bridge phenazine motif, distinct from the acceptor orbitals on the metal-bound ligands. We present a N K-edge XANES survey of the bimetallic tpphz-bridged assemblies, as shown in Fig. 2 (labeled according to their metal centers as M-M′), as well as the monometallic molecular subunits that comprise the dyads (with and without bridge-mimicking phenazine-containing ligands) and the corresponding unbound ligands. We observe distinct pre-edge absorption peaks that can be mapped to different N sites in the assemblies and identify the origins of their energetic separation by comparison with time-dependent density functional theory (TD-DFT) spectral simulations. We find that N K-edge XANES not only is sensitive to metal-bound vs unbound N atoms in the dyads but also is a sensitive probe of the degree of ligand-to-metal σ-donation for metal-bound polypyridyl ligands and structural distortion of the tpphz bridge.

The molecular structures of all studied compounds and their name abbreviations are presented in Table S1. The following compounds were synthesized and characterized according to previously published procedures: mes-phen (2,9-dimesityl-1,10-phenanthroline),³⁸ tpphz (tetrapyrido[3,2-a:2′,3′-c:3″,2″-h:2‴,3‴-j]phenazine),^13,36,39 [Cu(mes-phen)(phen)]PF₆ (where phen = 1,10-phenantrholine),³⁸ [Cu(mes-phen)(taptp)]PF₆ (where taptp = 4,5,9,18-tetraazaphenanthreno-[9,10-b] triphenylene),¹³ [Ru(bpy)₂(dppz)](PF₆)₂ (where bpy = 2,2′-bipyridine and dppz = dipyrido[3,2-a:2′,3′-c]phenazine),⁴⁰ [Ru–Ru]⁴⁺ ([(bpy)₂Ru(tpphz)Ru(bpy)₂](PF₆)₄),¹³ [Ru–Cu]³⁺ ([(bpy)₂Ru(tpphz)Cu(mes-phen)](PF₆)₃),¹³ [Cu–Cu]²⁺ ([(mes-phen)Cu(tpphz)Cu(mes-phen)](PF₆)₂),¹³ [Cu–Os]³⁺ ([(bpy)₂Os(tpphz)Cu(mes-phen)](PF₆)₃),³⁹ [Os–Os]⁴⁺ ([(bpy)₂Os(tpphz)Os(bpy)₂](PF₆)₄),³⁹ [Ru–Pt]²⁺ ([(tbbpy)₂Ru(tpphz)PtCl₂](PF₆)₂),⁴¹ [Ru–Pd]²⁺ ([(tbbpy)₂Ru(tpphz)PdCl₂](PF₆)₂),⁴² [Ru–Rh]³⁺ ([(tbbpy)₂Ru(tpphz)Rh(Cp*)Cl](PF₆)₃),²² [PtCl₂(dppz)],⁴³ and [RhCl(Cp*)(phen)]Cl⁴⁴ (where Cp* = pentamethylcyclopentadienyl). The following compounds were prepared and characterized as described in the supplementary material: [Cu^II(mes-phen)(phen)](BF₄)₂, [PtCl₂(phen)], [PdCl₂(phen)], [PdCl₂(dppz)], and [RhCl(Cp*)(dppz)]Cl. The following compounds were commercially available and used as purchased: [Ru(bpy)₃](PF₆)₂ (CAS 60804-74-2) and [Ru(bpy)₃]Cl₂ (6H₂O)(CAS 50525-27-4).

N K-edge XANES measurements were performed at two beamlines: beamline 7.3.1 at the Advanced Light Source (ALS) and beamline 8-2 at the Stanford Synchrotron Radiation Lightsource (SSRL). The solid samples were ground to a fine powder, applied to a carbon tape, and mounted on an Al sample stick. All data were recorded at room temperature in a vacuum chamber by scanning under incident x-ray energy and detecting the total electron yield (TEY) by measuring the drain current through the conductive sample mount. The spectra are normalized to the incident x-ray flux measured before taking the sample and represent an average of 3–9 measurements per sample. Both end stations employ spherical grating monochromators with an energy resolution of ∼0.2 eV.

The scan-to-scan drift of the monochromator was corrected by applying an energy shift determined from simultaneous measurement of an in-line reference sample (SSRL data) or by measurement of a reference sample between each set of three scans (ALS data). Due to monochromator drift during data collection, we assign a ±0.05 eV confidence interval to the peak energies. Periods of larger drift were excluded from the data to avoid energy drift effects within a single scan (scan-to-scan shift of >0.1 eV excluded). Absolute energy calibration was achieved by setting the lowest energy peak of [Ru(bpy)₃]²⁺ to 398.5 eV, which was determined against the internal reference NiF₂ using the second harmonic of the Ni L₃-edge peak at 426.35 eV (Fig. S1). The pre-edge peak energies were determined by fitting pseudo-Voigt distributions to the pre-edge peak region of 395–402 eV. The uncertainty from the peak-fitting is significantly smaller than the uncertainty of 0.05 eV due to monochromator instability. Additional details of the fitting procedure are provided in the supplementary material.

All calculations were performed using the GAUSSIAN 16⁴⁵ package. We used the ωB97X-D functional⁴⁶ with the LanL2DZ effective core potential (ECP) and the corresponding basis set⁴⁷ on metal atoms and 6-31G(d) on non-metal atoms for geometry optimization and N K-edge simulations. The choice of the functional and basis set has been proven suitable for bimetallic systems in previous studies.^26,39,48 The optimized structures were confirmed as true minima by frequency calculations. Natural transition orbital (NTO)⁴⁹ and natural population⁵⁰ analyses were carried out to reveal the origin of excitations and peak energy differences, respectively. Energy-specific TD-DFT⁵¹ was used to compute N K-edge spectra. All calculated spectra are shifted by 9.81 eV to match the experimental [Ru(bpy)₃]²⁺ pre-edge peak energy. The N K-edge peak spacing is slightly overestimated in the simulations compared to the experiment, which is attributed in part to the intrinsic self-interaction errors of TD-DFT.⁵² 

For multiple occurrences of N within a single tpphz-bridged donor/acceptor assembly, we observe that the unique sets of metal-bound N atoms and tpphz phenazine N atoms have distinct N K-edge XANES pre-edge features, as shown in Fig. 3. This enables one to map the unoccupied ligand-based electronic structure with a very high spatial specificity, separately probing the electron accepting valence orbitals of the donor, bridge, and acceptor motifs involved in the excited state electron transfer.

Figures 3(a) and 3(b) show the spectra of a tpphz-bridged [Ru–Cu]³⁺ assembly, as recently synthesized by the Mulfort group and investigated for its excited state electron transfer processes.¹³ [Ru–Cu]³⁺ is selected as an illustrative example due to the clear energetic spacing between the three pre-edge peaks of the three sets of unique N atoms: Cu-bound N, Ru-bound N, and bridge phenazine N atoms. The experimental [Fig. 3(a)] and calculated [Fig. 3(b)] N K-edge XANES spectra of [Ru–Cu]³⁺ (black trace) and its molecular building blocks (colored traces) are shown separately. The assignment of each pre-edge peak to a specific set of N atoms can be made directly by a comparison of the measured spectra of [Ru–Cu]³⁺ with those of the individual molecular components. The highest energy pre-edge peak of [Ru–Cu]³⁺ [Fig. 3(a), black] at 398.51 eV is due to the six near-degenerate Ru-bound N atoms, as evidenced by the comparison with the corresponding [Ru(bpy)₃]²⁺ monomer (orange) and its single pre-edge peak at 398.53 eV. The middle [Ru–Cu]³⁺ peak at 397.86 eV is due to the four Cu-bound N atoms, as can be seen from the alignment with the corresponding single pre-edge peak in the [Cu(phen)(mes-phen)]⁺ monomer (green) at 398.05 eV. The lowest energy peak in the [Ru–Cu]³⁺ spectrum at 397.26 eV is due to the two degenerate N atoms on the phenazine portion of the tpphz bridge. This is clear from the appearance of the low energy peak in [Cu(phen)(taptp)]⁺ (blue) at 397.23 eV upon the addition of the phenazine-containing taptp ligand (relative to [Cu(phen)(mes-phen)]⁺).

These mappings of the pre-edge peaks to distinct sets of N atoms are also confirmed by the TD-DFT calculations of N K-edge XANES spectra for the same [Ru–Cu]³⁺ dyad and molecular sub-units, which reproduce the measured spectra as shown in Fig. 3(b). The transitions underlying the simulated spectra [sticks in Fig. 3(b)] are investigated in more detail to identify the acceptor orbitals contributing to each absorption peak. This is illustrated by the natural transition orbitals (NTOs), which are generated from the acceptor orbitals involved in each of the electronic transitions and are shown below the [Ru–Cu]³⁺ spectrum in Fig. 3(b). From the NTO analysis, we determine that the phenazine N peak at 397.23 eV originates from transitions of the phenazine N 1s electrons to the lowest lying unoccupied orbital of the molecular dyad, a π* orbital localized across the three rings of the phenazine subunit of tpphz. The Cu-bound N peak at 398.05 eV is due to transitions of the Cu-bound N 1s electrons to an unoccupied π* orbital delocalized across the phenanthroline portions of the mes-phen and tpphz ligands. We find that the calculated transitions corresponding to the mes-phen ligand cannot be resolved from those of the Cu-bound phenanthroline portion of tpphz. The Ru-bound N peak at 398.53 eV is due to the transitions of the Ru-bound N 1s electrons to the unoccupied π* orbitals of the bipyridine ligands and closest phenanthroline portion of tpphz. Again, we are unable to resolve separate absorption features distinguishing transitions of the Ru-bound bipyridine ligands vs the Ru-bound phenanthroline portion of tpphz. The findings of this NTO analysis are generally applicable across the full range of complexes. The measured spectra of the additional tpphz-bridged bimetallic dyads are shown in Fig. 3(c), with their calculated spectra and peak energies/assignments presented in Fig. S3 and Table S2.

The TD-DFT calculations were further used to investigate the source of the energetic splitting of the three peaks in the [Ru–Cu]³⁺ spectrum by calculating the transition energy and the energy of the underlying N 1s and π* acceptor orbitals. We find that the stabilization of the tpphz phenazine N peak relative to the metal-bound N peaks is primarily driven by a difference in the energies of their respective π* acceptor orbitals. Consistent with previous work,^15,17 the lowest unoccupied orbital of the dyads is delocalized across the phenazine portion of the tpphz bridge, resulting in its lower energy pre-edge peak. This is generalizable across all dyads measured in this work [as shown in Fig. 3(c)]. In contrast, we find that the transitions of the different metal-bonded N atoms appear at distinct energies primarily due to varying degrees of stabilization of the metal-bound N 1s orbitals. This trend of N 1s stabilization is generalizable across the range of 3d-to-5d metals measured herein but is metal-dependent, as discussed in Sec. III B.

It is apparent from the dyad spectra shown in Fig. 3(c) that the pre-edge peak energies corresponding to the metal-bound N atoms of the bimetallic dyads are highly metal-dependent and increase for 3d < 4d < 5d metal-bound N sites of the dyads. Here, we explore the underlying reason for this observed metal dependence. The trend is investigated more explicitly, as shown in Fig. 4(a), which plots the fitted N XANES pre-edge peak energies for N atoms bonded directly to the metal center but only considers the monometallic sub-units or homo-bimetallic dyads to avoid any challenges associated with fitting closely spaced peaks in the hetero-bimetallic dyads. Both experimental (solid black circles) and simulated (open red circles) peak energies are shown. Additionally, the peak energy of the unbound mes-phen ligand is included to highlight the influence of metal binding more generally. The associated experimental and simulated spectra for representative homo-bimetallic complexes of 3d ([Cu–Cu]²⁺), 4d ([Ru–Ru]⁴⁺), and 5d ([Os–Os]⁴⁺) metal systems are shown in Fig. 4(b). In general, simulated spectra show excellent agreement with the experimental trends. All additional measured and simulated spectra for the other molecules plotted in Fig. 4(a) are shown in Fig. S2 (experimental) and Fig. S3 (calculated). The same general trend noted for the heterobimetallic dyads is shown in Fig. 4(a): metal coordination shifts the N pre-edge peak toward a higher energy and the extent of that shift increases as one goes from N atoms bound to 3d < 4d < 5d metals. Given the agreement between the calculated and experimental results, the TD-DFT calculated transitions are investigated in detail to understand the origin of the observed trend.

Generally, the TD-DFT calculations show that both core and unoccupied orbitals are significantly stabilized upon metal binding (and increasingly from 3d-to-5d metal binding) through a combination of long-range electrostatic and local chemical bonding effects. The electrostatic effect was isolated by simulating the spectra of the Ru-bound N atoms of [Ru(bpy)₂(tpphz)]²⁺ in the presence of a point charge to mimic the introduction of a second positively charged metal ion, as shown in Fig. S6. The presence of the point charge does not significantly change the calculated transition energy (compared to the magnitude of change shown in Fig. 4) but results in significant stabilization of both the N 1s and ligand π* orbitals. In contrast, we experimentally observe that metal binding to the ligand N atoms shifts the pre-edge peak toward a higher energy, indicating that the N 1s stabilization is greater than the ligand π* stabilization [Fig. 4(c)]. This difference from the purely electrostatic behavior is attributed to the local chemical bonding interactions between the ligand N and bonded metal.

Overall, we observe that this larger degree of stabilization of the N 1s orbital relative to the ligand π* is responsible for the increasing transition energies of the unbound < 3d < 4d < 5d metal-bonded ligands. This trend is consistent with the dominant σ-donating character of polypyridyl ligands,⁵³ such that the ligand-to-metal σ donation increases the effective charge of the ligand N atoms and stabilizes the N 1s core levels (due to decreased shielding in nuclear charge). As stronger metal–ligand interactions are expected for the more diffuse 5d and 4d metal orbitals,⁵⁴ increasing the σ donation with respect to the 3d metals is reflected by the increasing degree of N 1s stabilization (and the resulting increase in the pre-edge peak energies). The calculated natural charges of the metal-bound N atoms partially support the correlation between the effective charge of the N atoms and the measured and calculated transition energies, as shown in Fig. S5. A large positive increase in N effective charge is predicted for the 4d and 5d bound N atoms relative to the Cu-bound N atoms; however, no systematic trend is predicted between the 4d vs 5d bound N atoms. Weaker contributions of metal-to-ligand π backdonation, which also increases for the 5d metals, could also affect the N pre-edge peak energies. However, we suggest that the resulting spectral shift would be minor as the destabilization of the π* accepter orbital and reduction of the N atom effective charge (leading to destabilization of the N 1s core level) would have opposing effects on the N 1s-to-π* transition energy. We note that neither trend is predicted by the TD-DFT results, as shown in Fig. 4(c). The effect of π backdonation could be investigated by a rigorous comparison of the N k-edge for a single metal with a series of increasingly π-accepting ligands.

Although the present study focuses on a limited subset of metal–polypyridyl complexes, many of the trends reported here are consistent with the previous reports of N K-edge XANES for similar classes of molecules. For example, metalation of porphyrin, bipyridine, and phenanthroline ligands is known to shift the pre-edge peak toward a higher energy, consistent with the comparison of un-bound and metal-bound ligands reported here.^28,32,34 While only complexes of a single 3d metal (Cu) were measured here, the previous work demonstrated a large spread of ∼0.5 eV in the N pre-edge peak energies of metalloporphyrins across the 3d periodic row.²⁷ The N pre-edge was found to increase (Mn < Fe < Co) and then decrease (Co > Ni > Cu > Zn), moving across the row and the transition energy correlated with N 1s stabilization and increasing the effective charge of the porphryin N atoms, assigned on the basis of comparison with DFT calculations, and in agreement with the origins of the trends presented herein. We briefly explore the variation in the 3d metal bound N peak energies by varying the oxidation state of Cu in [Cu(mes-phen)(phen)]ⁿ⁺ from Cu(I) to Cu(II) shown in Fig. 4(a). As expected, Cu oxidation results in a shift toward a higher energy of the N pre-edge peak (by 0.13 eV) due to the increased ligand-to-metal σ donation (in this case, due to stabilization of the 3d orbitals) and resulting stabilization of the N 1s energy. We did not observe large variations or obvious trends within the few 4d or 5d metal complexes investigated here (d⁶ to d⁸). Instead, the largest variations observed in our survey are found as one moves down the period from 3d to 4d to 5d metals. The same trend was reported in the comparison of the N spectroscopy of Fe and Ru polypyridyl and porphyrin complexes,²⁷ and also appears to be generally observed in a large survey of metal–tris(bipyridine) complexes by Lukens et al.,^28,30 although not reported specifically. That the origin of the metal-dependent transition energies is attributed to N 1s stabilization here is further supported by a previous N K-edge x-ray photoelectron spectroscopy study, observing an increasing binding energy of the N 1s electron for Os > Ru > Fe [M(bpy)₃]²⁺ complexes.⁵⁵ 

In summary, we find that the N K-edge XANES spectra have sufficient specificity to distinguish unique N sites in tpphz-bridged bimetallic dyads. The presence of the low energy acceptor orbital on the tpphz phenazine motif results in the significant stabilization of the phenazine N pre-edge peak, making it a well-resolved probe of bridge electronic structures. In addition, the sensitivity of the metal-bound N pre-edge peak to the presence and strength of metal–ligand bonding interactions enables sufficient separation to specifically probe the 3d/4d/5d-bound ligands that participate in metal–ligand and ligand–bridge electron transfer. This specificity to unique N sites will now enable the evaluation of electronic couplings of the molecular constituents across a range of tpphz-bridged donor/acceptor assemblies. In the future, this spectroscopic distinction could additionally be used in the time-domain to probe the intra-ligand excited state electron transfer processes that drive charge separation and the effects of multiple charge accumulation on the assemblies during catalysis.

With the above-mentioned assignments of the N K-edge XANES features, we now investigate how the combination of molecular subunits affects their intrinsic electronic structures. We check for electronic coupling between the unoccupied states of the phenazine and phenanthroline motifs by looking for changes in their respective pre-edge peaks when combined.

First, the effect of adding the conjugated phenazine moiety on the metal-bound phenanthroline ligands was evaluated. The metal-bound N pre-edge peak position was compared for monometallic complexes with either a phenanthroline (phen) or a dipyridophenazine (dppz) ligand, where dppz extends the phenanthroline ligand with a phenazine motif that emulates the central structure of the tpphz bridge. For a series of M-phen and M-dppz bound complexes with M = Cu, Ru, Rh, Pd, Pt (the full chemical structures are presented in Table S1), we find that addition of the phenazine bridge motif has a negligible effect (<0.05 eV) on the energy of the metal-bound N pre-edge peak (peak positions are presented in Table S2). Second, for a series of heterobimetallic Ru–M′ dyads ([Ru–Ru]⁴⁺, [Ru–Cu]³⁺, [Ru–Pt]²⁺, [Ru–Pd]²⁺, [Ru–Rh]³⁺), we investigated the influence of the second metal (M′) linkage on the Ru-bound N pre-edge peak energies. Again, no significant effects (<0.07 eV) were observed upon variation of the M′ acceptor motif or metal (peak positions are presented in Table S2). The TD-DFT calculated N peak energies agree with the experimental results. From the lack of spectral changes observed, we infer that there is no significant ground state electronic coupling (above our ∼50 meV energy uncertainty) between the metal-bound ligands of the donor and acceptor sites or with the phenazine unit of the tpphz bridge. This is consistent with the previous conclusions drawn from optical and electrochemical characterization, finding that neither the metal-to-ligand charge transfer band in UV–Visible absorption nor the redox potentials of the Ru–polypyridyl chromophores are influenced by the addition of the acceptor moiety in [Ru–Cu]³⁺,¹³ [Ru–Rh]³⁺,²² [Ru–Pt]²⁺,⁵⁶ [Ru–Pd]²⁺,⁴² or [Ru–Os]⁴⁺.³⁶ 

Surprisingly, we do observe a significant variation in the lower energy pre-edge peaks of the tpphz phenazine N atoms as a function of metal ligation and metal identity in the tpphz-bridged M-M′ dyads, as shown in Fig. 5. We find that any metal complexation to tpphz shifts the phenazine N pre-edge peak toward a lower energy compared to the unbound ligand, as observed in both measured and simulated spectra. Analysis of the orbital energies underlying the predicted transitions shows that the presence of metal ions has the same electrostatic effect described above for the metal-bound N peaks, stabilizing both the N 1s and lowest unoccupied valence orbitals. A slightly larger stabilization of the valence orbital relative to the 1s results in a small decrease in the transition energy. This decrease was also observed in the point-charge model (Fig. S6), which found that the unoccupied valence orbitals were slightly more stabilized than the N 1s in the presence of a positive point charge.

Further variation in the phenazine N pre-edge peak position is observed as a function of the donor/acceptor metal complex identities. However, looking across the series of bimetallic complexes, there exists no obvious trend as a function of metal identity (3d vs 4d vs 5d). Furthermore, no similar metal-dependent trend is observed in the phenazine N peaks of the monometallic M-dppz bound complexes (variation <0.08 eV, Table S2). Thus, we consider that this is not an electronic effect based on the specific metal attached but perhaps instead is a structural effect based on the coordination geometries of the two metals and the resulting intermolecular interactions in the solid state. Indeed, we find that the observed trend can be understood when the complexes are grouped by molecular geometries, as shown in Fig. 5(a). We hypothesize two possible reasons for this structural effect as follows: (1) strain induced on the tpphz ligand or (2) π-stacking interactions between tpphz ligands.

First, in the crystalline forms of the dyads measured herein, distortions of the tpphz planarity depend on the geometry of the bound metal complexes. This was demonstrated by crystallography, finding waving or twisting distortions across tpphz for [Ru–Ru]⁴⁺ and [Os–Os]⁴⁺, respectively [as shown in Fig. 5(b)], and a primarily bowing distortion across tpphz for [Ru–Cu]³⁺ and [Os–Cu]³⁺.³⁹ Although the crystal structures of the octahedral-square planar [Ru–Rh]³⁺, [Ru–Pd]²⁺, and [Ru–Pt]²⁺ dyads have not been reported, the structure of an analogous [Ru–Au]³⁺ dyad [as shown in Fig. 5(b)] shows that the square planar motifs place little steric constraint on the bridge, resulting in minimal distortions from planarity.²³ The TD-DFT simulated spectra do not reproduce the experimental trend. However, this is not surprising as DFT was previously shown to underestimate the geometric distortion across the tpphz bridge as compared to the crystallographic geometries for these systems, as the distortions are largely induced by inter-molecular interactions in the crystal.³⁹ Ergo, it is possible that increasing tpphz structural distortion away from planarity decreases the energy of the bridge phenazine N pre-edge peak.

Second, intermolecular π interactions between tpphz ligands were demonstrated by x-ray crystallography for [Ru(bpy)₂(tpphz)]²⁺, which formed π-stacked dimers.⁵⁷ While these dimers may be hindered for complexes containing bulky octahedral and tetrahedral ligand environments, they may occur more readily in the octahedral-square planar dyads. We suggest that it is also possible that π-stacking interactions between dyads increase the energy of the phenazine N pre-edge peak.

A N K-edge XANES survey of tpphz-bridged heterobimetallic assemblies that couple chromophore and catalyst transition metal complexes for light driven catalysis, as well as their individual molecular constituents, is presented herein. We demonstrate a high specificity to the unique N sites in the N pre-edge XANES features, which are energetically well-separated for the tpphz bridge phenazine N atoms, and the donor and acceptor metal inner coordination sphere ligands. By comparison with the TD-DFT calculated spectra, we determine the origins of the differentiable spectral features observed. In a complement to the previous work probing N XANES in 3d^5–10 coordination complexes,²⁷ we find that metal coordination generates large shifts toward higher energy for the metal-bound N atoms, with increasing shift for 3d < 4d < 5d metal bonding. This is attributed to increasing ligand-to-metal σ donation that increases the effective charge of the bound N atoms and stabilizes the N 1s core electrons. In contrast, the tpphz phenazine N pre-edge peak is found at a lower energy due to its low energy electron-accepting orbital localized on the phenazine motif. While the spectra do not indicate any sensitivity to the electronic coupling between molecular components, they are sensitive to structural distortions of the tpphz bridge away from planarity.

The mapping of distinct N XANES peaks to bridges, donor ligands, or acceptor ligand N atoms enables one to separately probe the unoccupied electronic structure of the acceptor orbitals that participate in the metal–ligand and intra-ligand electron transfer processes that underpin the excited state charge separation in these photocatalytic assemblies. This report sets the stage for time-resolved N XANES experiments, which could potentially follow the redistribution of charge across chromophores, bridges, and catalyst ligands in real time. While several former studies demonstrated the use of time-resolved N K-edge XANES to track charge localization on the ligand following optical excitation of monometallic species,^29,31,35 our results here show that these efforts can be extended to heterobimetallic dyads, maintaining a high spatial specificity even in systems with multiple unique sets of N atoms. Future work will make use of the well-separated pre-edge peaks of the donor, bridge, and acceptor motifs to monitor the dynamics and localization of charge across the tpphz ligand during the first steps of photocatalysis.

Additional methodological details of materials synthesis and characterization and experimental energy calibration and peak fitting, as well as a complete tabulation of the measured and calculated spectra and fitting parameters, and additional computational results can be found in the online supplementary material.

This work was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, Chemical Sciences, Geosciences, and Biosciences Division, through the SLAC National Accelerator Laboratory under Contract No. DE-AC02-76SF00515. Z.-L.X., K.L.M., X.L., and X.L. were supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, Chemical Sciences, Geosciences, and Biosciences Division through Argonne National Laboratory under Contract No. DE-AC02-06CH11357. Work from A.K.M., S.S.F., K.S., B.D.I., and S.R. was funded by the Deutsche Forschungsgemeinschaft, DFG (German Research Foundation) - TRR 234 CataLight - 364549901, Project Nos. A1 and B2. The use of the Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Contract No. DE-AC02-76SF00515. This research used resources of the Advanced Light Source, which is a DOE Office of Science User Facility under Contract No. DE-AC02-05CH11231.

The authors have no conflicts to disclose.

Elizabeth S. Ryland: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Methodology (equal); Validation (lead); Writing – original draft (lead); Writing – review & editing (lead). Xiaolin Liu: Data curation (equal); Formal analysis (equal); Methodology (equal); Validation (equal); Writing – original draft (equal); Writing – review & editing (equal). Gaurav Kumar: Conceptualization (equal); Data curation (equal); Methodology (equal); Writing – review & editing (supporting). Sumana L. Raj: Data curation (supporting); Methodology (supporting); Writing – review & editing (supporting). Zhu-Lin Xie: Methodology (equal); Writing – review & editing (supporting). Alexander K. Mengele: Methodology (equal); Writing – review & editing (supporting). Sven S. Fauth: Methodology (supporting). Kevin Siewerth: Methodology (supporting). Benjamin Dietzek-Ivanšić: Funding acquisition (supporting); Methodology (supporting); Supervision (supporting); Writing – review & editing (supporting). Sven Rau: Funding acquisition (supporting); Methodology (supporting); Supervision (supporting); Writing – review & editing (supporting). Karen L. Mulfort: Conceptualization (supporting); Funding acquisition (equal); Methodology (supporting); Supervision (supporting); Writing – review & editing (supporting). Xiaosong Li: Conceptualization (supporting); Funding acquisition (equal); Methodology (supporting); Supervision (supporting); Writing – review & editing (supporting). Amy A. Cordones: Conceptualization (lead); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Methodology (equal); Supervision (lead); Validation (lead); Writing – original draft (lead); Writing – review & editing (lead).

The data that support the findings of this study are available from the corresponding author upon reasonable request.