Competing dynamics of intramolecular deactivation and bimolecular charge transfer processes in luminescent Fe(iii) N-heterocyclic carbene complexes

Steady state and ultrafast spectroscopy on [FeIII(phtmeimb)2]PF6 (phtmeimb = phenyl(tris(3-methylimidazol-2-ylidene))borate) was performed over a broad range of temperatures. The intramolecular deactivation dynamics of the luminescent doublet ligand-to-metal charge-transfer (2LMCT) state was established based on Arrhenius analysis, indicating the direct deactivation of the 2LMCT state to the doublet ground state as a key limitation to the lifetime. In selected solvent environments photoinduced disproportionation generating short-lived Fe(iv) and Fe(ii) complex pairs that subsequently undergo bimolecular recombination was observed. The forward charge separation process is found to be temperature-independent with a rate of ∼1 ps−1. Subsequent charge recombination takes place in the inverted Marcus region with an effective barrier of 60 meV (483 cm−1). Overall, the photoinduced intermolecular charge separation efficiently outcompetes the intramolecular deactivation over a broad range of temperatures, highlighting the potential of [FeIII(phtmeimb)2]PF6 to perform photocatalytic bimolecular reactions.


Sample Concentrations & Solvents
For sample preparation [Fe (III) (phtmeimb)2]PF6 powder and the respective solvent was mixed in standard glass vials under ambient conditions following procedures established in prior studies. 1 In case of butyronitrile, propionitrile and the methanol and ethanol mixture, the amount of solvent and solute were estimated to give a targeted concentration. For the mixtures with 1-propanol and 2propanol, the added amount of solute was higher than the solubility limit. To ensure maximum concentration, the solutions were sonicated for ~20min, filtered using a 0.45µm syringe filter (PFTE) and then topped up with additional 10% of pure solvent to avoid spontaneous precipitation.
To estimate the concentration of [Fe (III) (phtmeimb)2] + in the various samples prepared for the measurements, we performed standard UV/Vis measurements on each sample directly after preparation of the sample. The absorbance at 502 nm is then used to calculate the concentration based on the extinction coefficient at that wavelength (2950 M -1 cm -1 ). 1 An overview of the concentrations in all solvents used for different experiments is given in Table S1. In Table S2 we have collected relevant physical properties for all solvent systems used in the manuscript.

Steady State Absorption Spectroscopy
Steady-state absorption measurements in the UV-VIS region were performed in a Perkin Elmer Lambda 1050 Spectrophotometer. An Oxford Instruments Optistat DN bath cryostat was put into the absorption spectrometer. The temperature of the sample could be controlled from 80 to 300 K, from utilizing N2(l) in this cryostat. Absorbance of all prepared samples was measured in a standard quartzglass cuvette of path length 1 mm (Hellma -High Performance Quartz Glass). For reference the same cuvette with pure solvent was used for each measurement.  For fitting of the absorption spectra, the spectra were first converted from wavelength scale to energy scale. Assuming that a vibronic progression is responsible for the observed multi-peak structure, the spectra are fitted by series of Gaussian peaks that share the same width and spacing to one another.
With the general formula: Where ΔE represents the spacing between the single peaks and ω the width. The results of these fits are given in Figure S3 for Butyronitrile as well as the MeOH:EtOH solvent mixture. In both cases the fits yield an overall agreement with the observed spectra. However, the distinct peaks of the experimental data are not well reproduced by the fits. Consequently, some of the restrains on the Gaussian peaks were lifted subsequently to improve the quality of the fit. Best results are obtained when using three Gaussian peaks that share the same width and spacing and two additional peaks with slightly different parameters. Resulting fits are shown exemplarily for the low temperature absorption spectra in Butyronitrile and MeOH:EtOH solvent mixture in Figure S4. For fitting of the whole temperature series, the parameters obtained at the lowest temperature were used as starting parameters for the next higher temperature and so forth. The fitting parameters obtained for the whole temperature series are summarized in Table S3 and Table S4. MeOH:EtOH (volume ratio 1:4). Also included is the fit to the absorption spectrum using 5 Gaussian functions of shared width and spacing, modelling a vibrational progression. It is apparent that this model does not fit the measured data. Figure S4. Steady state absorption spectrum at the lowest temperature measured of Fe(phtmeimb)2]PF6 in a) butyronitrile and b) MeOH:EtOH (volume ratio 1:4). Also included is the fit to the absorption spectrum using 5 Gaussian functions with less restrictions to width and spacing compared to fit results presented in Figure S3. It is apparent that this model provides a better fit to the experimental data. The parameters obtained by the fits are listed in Table S3 and S4.  Table S4. Fitting parameters of the best obtained 5 Gaussian function fitting to the steady state absorption spectrum of Fe(phtmeimb)2]PF6 in butyronitrile. The fit result is visualized in Figure S4 and Figure 1 (main manuscript).

Steady State Emission Spectroscopy
Prior to emission measurements the sample quality was checked by steady-state absorption and no significant differences from the known absorption spectra were found. Emission measurements were performed on a Horiba Fluorolog spectrofluorometer in the front face geometry, using a quartz cuvette with 1 cm or 1 mm optical path length (Hellma -QS-glass). To suppress stray light from sample excitation, long-pass filters were inserted for some measurements at the entrance slit of the detection monochromator. An Oxford Instruments Optistat DN bath cryostat was put into the emission spectrometer. The temperature of the sample could be controlled from 80 to 300 K, from utilizing N2(l) in this cryostat. The recorded temperature dependent emission spectra over the whole accessible temperature range are plotted in Figure S5. For fitting of the emission spectra, the spectra were first converted from wavelength scale to energy scale including the Jacobian conversion factor. 6 Subsequently the spectra are fitted by a series of Gaussian peaks. The best fitting results were obtained using 4 individual peaks with equal spacing and shared width. Resulting fits are shown exemplarily for the low temperature emission spectra in Butyronitrile and MeOH:EtOH solvent mixture in Figure S6. For fitting of the whole temperature series, the parameters obtained at the lowest temperature were used as starting parameters for the next higher temperature and so forth. The obtained fitting parameters for the whole temperature series are summarized in Table S5 and Table S6.  Table S5 and S6.  Table S6. Fitting parameters of the best obtained 4 Gaussian function fitting to the steady state emission spectrum of Fe(phtmeimb)2]PF6 in butyronitrile. The fit result is visualized in Figure 1 (main manuscript) and Figure S6. To give an overview of the temperature dependent changes obtained by the fits, the individual peak positions relative to the lowest temperature position are plotted in Figure S7. Additionally, Figure S8 compares the temperature dependence of the fitted peak width in both solvent systems.  In addition to fitting the spectra by a series of Gaussians, we also perform a Franck-Condon type analysis. 7 For this, the spectra are fitted by a series of Gaussian peaks that represent a vibronic progression with shared spacing and width. In contrast to the previous fit, where the individual amplitudes were used as fitting parameters, the amplitudes of the vibronic progression are scaled by the Huang-Rhys factor (S) that describes the coupling between electron and vibrational states and is related to the distortion between ground and excited state. 8,9 Best fits were obtained when using a model with two vibrational modes that is represented by: Here ℏ 1 and ℏ 2 give the energy of the vibronic modes that are involved. The factor Δ 1 2 gives the full width at half maximum of the observed Gaussian. The sum indices 1 and 2 define how many vibronic replicas are considered. In our case, a maximum of = 10 was used to fit the spectra. The parameters 1 and 2 are the Huang-Rhys parameters describing the coupling to the respective vibrational modes. The energy 00 corresponds to the lowest energy transition between ground and excited state of the molecule.
Two exemplary fits of the lowest temperature emission spectra in butyronitrile and MeOH:EtOH solvent mixture are given in Figure S9. The overall shape of the observed emission spectra is reproduced well by the Franck-Condon analysis. However, compared to the fits by a simple series of Gaussian peaks, the Franck-Condon analysis fails to adequately reproduce simultaneously both the high-energy and low-energy part of the emission spectrum.  Table S7 and S8, respectively.

Transient Absorption Spectroscopy
Transient absorption (TA) spectroscopy was performed using an in-house built setup. Basis of this setup is a Spitfire Pro XP (Spectra Physics) laser amplifier system that produces ~80 fs pulses at a central wavelength of 796 nm at 1 kHz repetition rate. The amplifier output is divided into two parts that each pump collinear optical parametric amplifiers (TOPAS-C, Light Conversion). One of the TOPAS generates the pump beam (wavelength roughly set to the absorption maximum of each sample), while the other one generates a NIR beam (1350 nm) that is focused onto a 5 mm CaF2 crystal to generate a supercontinuum probe beam. The delay between pump and probe beams is introduced by a computer-controlled delay stage (Aerotech) placed in the probe beam's path. After supercontinuum generation the probe pulses are split into two parts: the former being focused to ~100 µm spot size and overlapping with the pump pulse in the sample volume, and the latter serving as a reference. The sample is placed inside an Oxford Instruments Optistat DN bath cryostat and put into the spectrometer. The temperature of the sample could be controlled from 80 to 300 K, from utilizing N2(l) in this cryostat. After passing the sample the probe beam is collimated again and relayed onto the entrance slit of a prism spectrograph. The reference beam is directly relayed on said spectrograph. Both beams are then dispersed onto a double photodiode array, each holding 512 elements (Pascher Instruments). The excitation power of the pulses was set to 1 mW at ~500 nm.
Mutual polarization between pump and probe beams was set to the magic angle (54.7°) by placing a Berek compensator in the pump beam. Time-resolution of the setup after dispersion correction is estimated to be ≤150 fs.
The solution of Fe(phtmeimb)2]PF6 in different solvents (from Sigma-Aldrich Sure/Seal Bottle) was filled in 1 mm optical path length cuvettes (Hellma -Optical Special Glass). The measured samples were translated after each scan to avoid photodegradation. To check for stability of each sample steady-state absorption spectra were measured before and after TA experiments, and they were found to be the same. Before analysis the measured data were corrected for group velocity dispersion (GVD -"chirp"). Note that data is not normalized, but is represented as measured. Note that data is not normalized, but is represented as measured.    Note that data is not normalized, but is represented as measured.    The resulting kinetics for three different solvents at all investigated temperatures are shown in Figure S18. The normalized kinetics were fitted by a sum of exponential decays of the general formula: Below the freezing point of the solvent the second exponential decay had to be introduced to accurately reproduce the measured data. The resulting time constants and amplitudes at all measured temperatures and for the three solvents are summarized in Figure S19.

Arrhenius Analysis and Decomposition
The time resolved photoluminescence data obtained by TCSPC has been analyzed following a standard Arrhenius-type model: 10 The model includes two activated terms with scaling factor A1 and A2, and one term that accounts for

Radiative and Non-Radiative Rates & Emission Quantum Yield
Based on the integrated extinction coefficient we estimate the radiative recombination rate (kr) following the Strickler-Berg formalism: With max being the wavenumber of the absorption maximum, the refractive index of the solvent (n) and  the extinction coefficient plotted versus wavenumber. 13 In combination with the measured luminescence lifetime (TCSPC) we can then estimate the non-radiative recombination rate (knr) according to: Using the combination of radiative and non-radiative rate we estimate the emission quantum yield () by: The extinction coefficients are integrated from the absorption spectra in Figure S2 over the spectral region corresponding to the LMCT transition (400 nm -600 nm). The resulting values for two different solvent systems are given in Figure S22. The increase in extinction coefficient appears to be independent from the solvent system. Below 100 K the extinction coefficient is no longer increasing but leveling off. The calculated temperature dependent radiative and non-radiative recombination rates are presented in Figure S23 whereas the lifetimes are taken from Figure S19. For both solvents the radiative rates are found to be very similar. Additionally, the radiative rate increases only slightly from 1.5·10 7 s -1 at RT to 2.5·10 7 s -1 at 80 K. In the high-and low-temperature limit, the non-radiative rates are also similar for the two solvent systems. However, the non-radiative rate decreases by a factor of 10 from  Finally, the calculated temperature dependent emission quantum yield is presented in Figure S24. At RT the quantum yield is found to be ~3 % which is in good agreement with the ~2 % reported previously. 1 It is noteworthy that the quantum yield values presented here are based on new sets of experimental data calculated based on a different formalism, compared to the previously reported value. At 80 K the quantum yield increased to ~25 %. Whereby the major rise appears below the freezing point of the solvent and is mainly related to the increase in luminescence lifetime.
For comparison, the relative emission quantum yield is calculated based on the temperature dependent luminescence intensity measurements (c.f. Figure 4) as well as the temperature dependent absorption. For this, the luminescence intensity is measured at various temperatures with fixed excitation wavelength (500 nm) and excitation density. Then the luminescence intensity is integrated over the whole recorded spectrum. Additionally, the temperature depended absorption spectra are integrated over a ±5 nm window around 500 nm. At 275 K, the integrated luminescence intensity is assumed to correspond to 3 % quantum yield. For the other temperatures the integrated emission intensities are first scaled according to the change in integrated extinction coefficient, in order to normalize them for the number of absorbed photons. Afterwards, the integrated emission intensity is converted to relative quantum yield based on the integrated emission intensity at 275 K.
The resulting relative emission quantum yield is in good agreement with the values calculated based on the Strickler-Berg formalism.
In case of the MeOH:EtOH solvent mixture the situation is different. Due to the additional quenching mechanism introduced by the photochemical charge separation (c.f. Figure 4), the luminescence intensity is no longer directly proportional to the emission quantum yield. Hence, the relative quantum yield cannot be extracted from the luminescence intensity. Figure S24. with the def2-TZVP/C auxiliary basis set, although vacuum was applied during the minimization. An experimental IR spectrum was recorded from the neat compound on a Bruker, Alpha-P FTIR spectrometer (see Figure S25). A quantum chemical frequency calculation was also performed on the optimized ground state geometry and compared with the measured IR spectrum ( Figure S25 it can be noted that modes with negative frequencies were discarded from further consideration after identification as spurious low-energy stretching modes. The energy and nature of the frequencies are listed in Table S12. The vibrations in the 1000-1300 cm -1 range are mainly characterized by C-H bond bending and stretching in both Imidazol and Phenyl group frames. Interesting stretching involving boron atoms were also identified in some vibrations specially at 1218.52 cm -1 and 1219.20 cm -1 (171 and 172 vibrational modes). The bond distances of the TD-DFT relaxed 2 LMCT (Table S9) show a clear distorted geometry regarding both ligands L1 and L2 respect to the L1/2 (L1 and L2 are equivalent) geometries in the GS. L1 ligand is farther distorted than L2 such that B-N2 bond is 0.03 Å more elongated (the atom numbering is displayed in Figure S26) and    Table S12)   first absorption band lies in the visible range (502 nm).