Site-Selective Orbital Interactions in an Ultrathin Iron-Carbene Photosensitizer Film

We present the first experimental study of the frontier orbitals in an ultrathin film of the novel hexa-carbene photosensitizer [Fe(btz)3]3+, where btz is 3,3′-dimethyl-1,1′-bis(p-tolyl)-4,4′-bis(1,2,3-triazol-5-ylidene). Resonant photoelectron spectroscopy (RPES) was used to probe the electronic structure of films where the molecular and oxidative integrities had been confirmed with optical and X-ray spectroscopies. In combination with density functional theory calculations, RPES measurements provided direct and site-selective information about localization and interactions of occupied and unoccupied molecular orbitals. Fe 2p, N 1s, and C 1s measurements selectively probed the metal, carbene, and side-group contributions revealing strong metal–ligand orbital mixing of the frontier orbitals. This helps explain the remarkable photophysical properties of iron-carbenes in terms of unconventional electronic structure properties and favorable metal–ligand bonding interactions—important for the continued development of these type of complexes toward light-harvesting and light-emitting applications.

Optical measurements used a Perkin Elmer Lambda 1050 Spectrophotometer. Absorbance measurements of solutions were performed in standard 10 x 10 mm cuvettes in transmission geometry.
A cuvette filled with pure solvent was used as reference. Measurements on the films were performed in reflection geometry by placing the samples on the entrance port of a 150 mm integrating sphere detection addon. Reference measurements were performed on a bare substrate.
Lab source XPS used a Physical Electronics (PHI) X-Tool equipped with a Monochromatic Al Kα (1487 eV) source. The acquisition area was 200 µm with the sample 45° from normal emission. A pass energy of 224 eV was used.
Synchrotron experiments were carried out at the HE-SGM beamline/endstation 1 at the Helmholtz-Zentrum Berlin BESSYII facility. This bending magnet beamline produced a ~1 mm x ~200 μm X-ray spot with energy range 200-800 eV. The analysis chamber, (pressure 1e-9 mbar), is equipped with a Scienta R3000 hemispherical analyser. RPES was measured with a pass energy of 200 eV in fixed mode. Carbon and nitrogen NEXAFS were measured using a partial electron yield MCP detector with a 150 V retardation potential. Iron L-edge NEXAFS was measured with a 470V retardation potential.
Binding energy scales were calibrated using the Au 4f peak at 83.91 eV -measured on a clean Au(111) crystal and calibrated to the fermi edge (defining 0 eV binding energy). Photon energy scales were calibrated using the kinetic energy difference between Au 4f photoemission peaks generated from 1st and 2nd order light where possible.

Optical Spectroscopy
Immobilization on glass For additional check on the stability thins film were also dropcasted onto glass substrate. The films were deposited simultaneously to the films on gold, i.e., from the same solution and under nitrogen flow. As substrate standard borosilicate cover glasses were used. In case of the Fe 3+ oxidation state (Fig S2a) a behavior similar to the samples deposited on gold is observed. The absorbance maxima are slightly red shifted compared to the positions in solution. At the same time the ratio of the LMCT to LC transition changes. In contrast to the sample on gold, however, here the resulting ratio is almost 1:1 rather than being 4:1. These findings strengthen our statement of the preserved oxidation state and indicates that the surface has an influence on the electronic structure. This is further affirmed by spectra of the Fe 2+ oxidation state deposited on gold (Fig S2b). In contrast to the sample on gold, the spectral shape now resembles that of the solution very well. This indicates that the oxidation state is retained. Figure S2: Absorbance of Fe III (btz) and Fe II (btz) thin films on glass. a) Absorbance of Fe III (btz) dissolved in acetonitrile (colored area) compared to absorbance of a thin film deposited on glass (line and circles). For comparison the absorbance is normalized to the peak at lowest energy.
Inset: photograph of the sample with 0.5mm scale bar. b) Absorbance of Fe II (btz) dissolved in acetonitrile (colored area) compared to absorbance of a thin film deposited on glass (line and circles). For comparison the absorbance of the thin film is scaled by a factor of 20. Inset: photograph of the sample with 0.5mm scale bar.

Fe 2+ oxidation state on gold
To underline the change in oxidation state of the Fe II (btz) complex deposited on gold, we compare the absorbance spectrum to calculated spectra that would result from a mixture of Fe 3+ and Fe 2+ oxidation state molecules. The resulting comparison is given in Fig S3. Most prominent indication for a change of oxidation state is the absorbance peak at 550 nm that corresponds to the Fe 3+ state. Additionally the LC centered absorbance around 400 nm exhibits a blue shift or grows a shoulder towards 350 nm, again resembling the spectral shape of the Fe 3+ state rather than the original Fe 2+ state. Figure S3: Absorbance of Fe II (btz) on gold (purple line and circles). a) Absorbance of solution (colored area) and thin film (circles and lines) of [Fe(btz) 3 ] 2+ . Each curve is normalized to the peak at the longest wavelength. Inset: Photo of the as prepared thin film with 1 mm scale bar. b) Absorbance of mixed sample ranging from 100 % Fe II (btz) and 0 % Fe III (btz) (blue dotted) to 0 % Fe II (btz) and 100 % Fe III (btz) (red dotted) contribution. The spectra calculated based on the absorbance measured for the pure oxidation state solutions.

X-ray spectroscopy
Oxidation state discussion Fe 2p XPS is presented in Figure S4 for the Fe(II) variant of the Fe(btz) 3 complex alongside the spectra of the Fe(III) variant (presented in the main manuscript) for comparison. A good summary of Fe 2p oxidation state data can be found at reference 2 which includes a review of binding energy positions of many Fe compounds and discusses the charge transfer satellite features that are a fingerprint of Fe(III) species. Satellite features should be apparent for Fe(II) but appear as a shoulder to the main peak (unlike Fe(III) where they are distinctly separated). This summary is also supported by other literature 3,4 .
The spectra for [Fe(btz)] 3+ has a Fe2p 3/2 peak at ~711 eV binding energy with satellite structure extending until ~720 eV. Although we are unable to rule out that there is an Fe(II) component in this spectra, we believe that based on the literature discussed above, binding energy position and the extent of the satellite structure, the [Fe(btz)] 3+ sample is dominantly Fe(III). This is supported by Fe L-edge NEXAFS ( Figure S7) where the shape of the measured NEXAFS matches well with the shape of the calculated NEXAFS for [Fe(btz)] 3+ . The [Fe(btz)] 2+ sample is less clear. It does not show the satellite shoulder associated with Fe(II) and the binding energy position of the Fe 2p peaks is slightly lower than the above literature reports for Fe(II) -the position is nearer the expected value for Fe 0 . Given the optical spectroscopy, it is possible that the oxidation state of this complex is not maintained. This sample however is not the focus of this manuscript -it is included as an interesting addition.

Supporting XPS
Note. We would like to explicitly state that the measurements were of samples prepared ex-situ so some amount contamination is inevitable. We would also note that the samples were drop cast from an acetonitrile solution -although the samples were dried in vacuum some contamination from which may remain. Neither of which can be meaningfully factored into curve fitting so the analysis below should be taken as an indication the spectral shape being reasonable rather than definitive binding energy positions of the components.   Fe L-Edge NEXAFS and beam-damage study Figure S8. Red: Fe L-edge NEXAFS measurement of Fresh molecules with no prior X-ray exposure prior to starting the measurement. Purple: Fe L-edge NEXAFS on an area of sample exposed to X-ray radiation for several hours. Here we see extensive evidence of radiation damage. Green: CTM4XAS NEXAFS simulation (details below) For the Fe L-edge NEXAFS measurements, the nearby F k-edge NEXAFS from the PF6 counterion was used as a reference as the beamline does not produce 2nd order light at these high energies. This allows comparison of the two spectra but the precise absolute photon energy position of the spectra should be ignored. Fe L-edge NEXAFS was measured with the partial yield detector with a retardation potential of 470 V.
Regarding the NEXAFS simulation: For the charge transfer multiplet model, a threeconfiguration calculation with an extra LMCT and an additional MLCT configuration is performed with Cowan code 5 . The ligand field splitting 10Dq=3.3 eV is used, for the configuration separations, EG2=0.5 eV, EF2=0 eV, EG3=-0.5 eV, and EF3=-1.0 eV. (EG2/EF2 is the average energy separation between d 4 L n+1 (MLCT) and d 5  Electronic structure calculations used the B3LYP* modification of the B3LYP hybrid functional 6 with a standard SDD ECP basis set 7 on a structure that had been optimized geometry with the same functional and with a standard 6-311G(d) basis set 8 . Both computational steps carried out with a polarizable continuum model (PCM) of acetonitrile solvent 9 . The solvent model ensures compatibility with previous calculations 10,11 , although it can be noted that this represents a somewhat different environment compared to the dried films used experimentally. We still consider inclusion of a PCM a more realistic description of the film environment compared to a pure vacuum calculations of isolated metal complexes carrying a net ion charge. All calculations were performed using Gaussian09. 12 Details of the valence electronic description obtained from the quantum chemical calculation used in the underlying analysis for the results presented in the main text are given in Figures S9-S12 and Table S1.  S10. Decomposition of valence alpha MOs in the energy range -10 -+2 eV according to atomic type. The displayed contributions for each orbital (from left to right) is: grey: iron, yellow: N1, light blue: N2, green: N3, dark blue: carbene, dark brown: 2nd carbon on carbene ring, dark grey: C on methyl side-group on carbene ring, light brown: toluene carbons, medium blue: hydrogens. Figure S11. Decomposition of valence beta MOs in the energy range -10 -+2 eV according to atomic type. The displayed contributions for each orbital (from left to right) is: grey: iron, yellow: N1, light blue: N2, green: N3, dark blue: carbene, dark brown: 2nd carbon on carbene ring, dark grey: C on methyl side-group on carbene ring, light brown: toluene carbons, medium blue: hydrogens. Table S01. Valence MO summed PDOS contributions from different atoms and functional groups. The two leftmost columns give the MO number and MO energy (in eV). The listed atoms and functional group PDOS include Fe, N_1's, N_2's, N_3's, the Fe-bonded carbons on the carbene rings (C_carb), the second carbons on the carbene rings, the methyl side-group carbons on the carbene rings, all the carbons on the toluene side-groups, and all the hydrogens.

Explicit XAS simulations
The Carbon and Nitrogen K edge X-ray absorption spectra (XAS) have been calculated explicitly using DFT/restricted open configuration interaction with singles (DFT/ROCIS) method with ORCA 13,14 . The def2-VTZP basis set is used for all the elements. Both the standard B3LYP (20% Hartree-Fock exchange (HFX)) and modified B3LYP (15%HFX) have been used to simulate the K-edge XAS 15,16 . The role of HFX in the simulation of XAS has effects on the absolute excitation energy, but with little effects on the peak shape, see Figure S13. For C kedge XAS, the shift energy of 11.77 eV and 13.45 eV are used for standard and modified B3LYP respectively. For N k-edge XAS, the shift energy of 14.20 eV and 16.25 eV are individually used for standard and modified B3LYP respectively. The underestimation of excitation energy comes from the limited ability of DFT to describe potentials near the nucleus, which means the energy of 1s core orbitals of C and N are too high relative to the valence orbitals 17 . However, the DFT/ROCIS approach used here can provide relative excitation energies and intensities of the peaks with an accuracy that allows for a direct comparison with experiment. The relative excitation energies of the peaks are determined by the molecular valence orbitals and, therefore, not affected significantly by errors in the description of the core orbital. The calculated spectra have been broadened with Gaussian function with full width at half maximum (FWHM) = 0.2, 0.4, 0.6, and 0.8 eV. The comparison is shown in Figure S14. The peak decomposition analysis and assignments are detailed in Figure S15. The spectral features of C 1s K edge XAS have been investigated by including different number of roots in the calculation. The intensity of high-energy spectral feature converge slower compare to the lower-energy spectral features, which means more roots are required in order to get full description of spectral features at high energy position.    Figures S17 to S22 show selected molecular orbitals from the explicit XAS calculations at the B3LYP*/ def2-VTZP level of theory used to validate the correspondence of MO characters with the ground state molecular orbital scheme and XAS/RPES assignments discussed in the main manuscript. In order to connect the calculated XAS with the valence orbital character, the orbital assignments were carried out by selecting the strongest transitions which contribute most of the intensities in the spectra. The orbital coefficients larger than 10% in the configuration are presented below: Figure S17. Selected valence molecular orbitals from ground state calculations of 2 [Fe(btz) 3 ] 3+ with def2-VTZP basis set and modified B3LYP (15%HFX) functional. This diagram shows the similar information as the Figure 2 in the main manuscript. Figure S18. Selected molecular orbitals for the C Toluene peak: 1s orbitals of C on the Toluene excited to valence orbitals from LUMO+9 to LUMO+20. Most of the final orbitals are * character locating on the toluene. Figure S19. Selected molecular orbitals for the C a peak (the C on the C2N3 ring not bonded to the Fe): The C a character peak comes from 1s of C on the C2N3 ring excited to the * btz and * toluene character orbitals. Figure S20. Selected molecular orbitals for the C b peak (the methyl C on the C2N3 ring): The C b character peak comes from 1s of the methyl C on the C 2 N 3 ring excited to the * btz character orbitals. Figure S21. Selected molecular orbitals for the C Fe peak. The C Fe character peak comes from 1s of the C atom bonded to Fe center excited to the * toluene character orbitals.

Figure S22
Molecular orbitals for N K-edge XAS. The XAS intensity comes from the three N 1s atoms excited to these * btz character orbitals.