Iron(III) Carbene Complexes with Tunable Excited State Energies for Photoredox and Upconversion

Abstract

Substituting precious elements in luminophores and photocatalysts by abundant first-row transition metals remains a significant challenge, and iron continues to be particularly attractive owing to its high natural abundance and low cost. Most iron complexes known to date face severe limitations due to undesirably efficient deactivation of luminescent and photoredox-active excited states. Two new iron(III) complexes with structurally simple chelate ligands enable straightforward tuning of ground and excited state properties, contrasting recent examples, in which chemical modification had a minor impact. Crude samples feature two luminescence bands strongly reminiscent of a recent iron(III) complex, in which this observation was attributed to dual luminescence, but in our case, there is clear-cut evidence that the higher-energy luminescence stems from an impurity and only the red photoluminescence from a doublet ligand-to-metal charge transfer (²LMCT) excited state is genuine. Photoinduced oxidative and reductive electron transfer reactions with methyl viologen and 10-methylphenothiazine occur with nearly diffusion-limited kinetics. Photocatalytic reactions not previously reported for this compound class, in particular the C–H arylation of diazonium salts and the aerobic hydroxylation of boronic acids, were achieved with low-energy red light excitation. Doublet–triplet energy transfer (DTET) from the luminescent ²LMCT state to an anthracene annihilator permits the proof of principle for triplet–triplet annihilation upconversion based on a molecular iron photosensitizer. These findings are relevant for the development of iron complexes featuring photophysical and photochemical properties competitive with noble-metal-based compounds.

Introduction

Coordination complexes and organometallic compounds based on precious metals such as ruthenium, iridium, platinum, or gold remain among the most widely used compounds for applications in lighting,¹⁻⁸ sensing,^9,10 photocatalysis,¹¹⁻¹⁷ solar energy conversion,¹⁸⁻²² dye-sensitized solar cells,²³ artificial photosynthesis,²⁴⁻²⁶ and phototherapy.²⁷⁻²⁹ Further exploration of precious-metal-based compounds remains crucially important, due to their stability and sometimes remarkable performance,³⁰⁻³⁶ but aspects such as greater abundance and cost-effectiveness drive a substantial portion of current research in the direction of first-row transition metals. This new focus also represents an opportunity for fundamentally new discoveries, as the photophysics and photochemistry of first-row transition metal complexes can be very different from those of the more widely investigated precious metals named above.^37,38

Traditionally, considerable focus has been placed on copper(I), primarily because of its completely filled 3d¹⁰ subshell.³⁹⁻⁴¹ This particular electron configuration, responsible for the semiprecious chemical character of copper(I), precludes the presence of energetically low-lying metal-centered (MC) states that can negatively affect the photophysical and photochemical behavior of charge transfer (CT) excited states.^42,43 Consequently, long-lived and brightly luminescent CT excited states in copper(I) complexes are comparatively easily accessible,⁴⁴⁻⁴⁶ because they are less susceptible to unwanted nonradiative relaxation involving metal-centered (MC) states. This stands in strong contrast to their d⁸ (Ni^II)^13,47−55 or d⁶ (Cr⁰, Mn^I, Fe^II, Co^III)⁵⁶⁻⁶⁹ counterparts, where low-lying MC states can be present.

Among first-row transition metal complexes with partially filled d-orbitals, iron(II) compounds are particularly popular,^61,70−78 and furthermore Cr(III) complexes have received substantial attention in recent years.⁷⁹⁻⁸⁸ The lowest metal-to-ligand charge transfer (MLCT) excited states of iron(II) complexes have been investigated intensively,^63,66,89,90 due to the prominent role played by these states in the luminescence behavior of isoelectronic ruthenium(II) complexes and for the photoredox properties of iridium(III) compounds. However, in iron(II) complexes, the lowest MLCT states commonly deactivate rapidly via lower-lying MC states, due to comparatively weak ligand fields created by common polypyridine ligands.^66,92,93 Strongly electron-donating ligands such as N-heterocyclic carbenes (NHC) can raise the energy of the MC states and can therefore increase the MLCT excited state lifetimes.^94,95 This approach led to the Fe^II complex [Fe(btz)₃]²⁺ (Figure 1a), for which an MLCT lifetime of 0.5 ns in solution at room temperature was reached,⁷⁸ and this represented a record value at the time of publication, exceeding the MLCT lifetime of prototypical [Fe(bpy)₃]²⁺ (bpy = 2,2′-bipyridine) by 4 orders of magnitude.⁷¹

Figure 1.

Molecular structures of previously investigated iron(III) complexes (a–e) along with the new complexes investigated herein (f–h).^{57,78,96−100} (g) X–ray crystal structure of [Fe(ImPP)₂][HCOO] with 50% thermal ellipsoids. Hydrogen atoms and counterion are omitted for clarity.

The isostructural Fe(III) compound, [Fe(btz)₃]³⁺, turned out to be luminescent from a ligand-to-metal charge transfer (LMCT) excited state with a lifetime of 0.1 ns in acetonitrile at 20 °C.⁹⁶ This unanticipated discovery sparked broader interest in the investigation of photoactive 3d⁵ Fe(III) complexes; all of a sudden, LMCT states in the 3d⁵ configuration became an interesting alternative to MLCT states in 3d⁶ complexes. The use of a tridentate scorpionate ligand in the [Fe(phtmeimb)₂]⁺ complex (Figure 1b) subsequently allowed the elongation of the LMCT lifetime to 2 ns and a luminescence quantum yield of 2.1%.⁹⁷ Decoration of this complex with different chemical substituents led to negligible changes in the LMCT excited state behavior (Figure 1b, R = MeO, COOH, Br).⁹⁸ The favorable photophysical properties of the [Fe(phtmeimb)₂]⁺ and [Fe(btz)₃]³⁺ complexes subsequently enabled their use for symmetry-breaking charge separation and for photoredox catalysis.^101−105 Recently, the first heteroleptic luminescent Fe^III complex ([Fe(pzTp)(CN)₃]⁻, Figure 1d) was reported, featuring a ²LMCT excited state lifetime of 80 ps in solution at room temperature.⁹⁹ The recent claim of a Janus-type dual MLCT and LMCT luminescence from the iron(III) complex [Fe(ImP)₂]⁺ (Figure 1e) represented another unanticipated development,^100,106 because genuine dual luminescence is a rare phenomenon that is usually in conflict with Kasha’s rule, one of the most fundamental principles of photophysics.^107,108 Other studies had found that Fe^III complexes can lead to low cage escape quantum yields in photoinduced electron transfer reactions,^109,110 and that aspect does not seem to be fully understood yet. Herein, we report two new Fe^III complexes, [Fe(ImPP)₂]⁺ and [Fe(ImPAr₂)₂]⁺ (Figure 1f–h), in which we managed to tune the energy of the photoactive LMCT excited state, contrasting the above-mentioned earlier studies that failed to accomplish this with Fe^III carbene complexes.⁹⁸ Our two new iron(III) complexes are structurally similar to the [Fe(ImP)₂]⁺ complex, for which dual luminescence was recently reported, including supposed blue MLCT emission in addition to red LMCT luminescence.^100,106 We observe similar apparent dual emission from one of our complexes, but can trace the blue emission back to an impurity, while the red luminescence is genuinely due to the iron(III) complex.¹¹¹ While the diarylamino substituent was introduced in the hope to tune excited state energies,¹¹² the additional phenyl substituent present in [Fe(ImPP)₂]⁺ with respect to [Fe(ImP)₂]⁺ was anticipated to lower excited state distortions, owing to greater π-delocalization.^{65,67,113,114} Despite the short lifetime of roughly 267 ps, the luminescent LMCT excited state of our [Fe(ImPP)₂]⁺ complex undergoes efficient photoinduced electron transfer with various donors and acceptors, which enabled its application in both oxidative and reductive photoredox catalysis, whereby high product yields were achieved under mild reaction conditions using red light as energy input. Moreover, one of the new complexes permits sensitized triplet–triplet annihilation upconversion (sTTA-UC) from the red into the blue spectral region; this represents the first case of molecular iron-based upconversion to the best of our knowledge.

Results and Discussion

Synthesis and Characterization

[Fe(ImPP)₂]⁺ and [Fe(ImPAr₂)₂]⁺ were synthesized as shown in Figure 2a, by reacting 1.0 equiv of the ligand precursors 1 (HImPP = 3,5-bis(3-ethyl-1H-3λ⁴-imidazol-1-yl)-1,1′-biphenyl) or 2 (HImPAr₂ = 3,5-bis(3-ethyl-1H-3λ⁴-imidazol-1-yl)-N,N-bis(4-methoxyphenyl)aniline) with 2.5 equiv of Zr(NMe₂)₄ and 0.5 equiv of FeBr₂. This procedure leads to the air- and water-stable Fe^III complexes as bromide salts after oxidation during the workup under air. Further purification via reversed-phase preparative high-performance liquid chromatography (HPLC) was performed to remove the excess of ligand precursor and other luminescent impurities. Both complexes were characterized by ¹H and ¹³C nuclear magnetic resonance (NMR) spectroscopy in CD₃CN (Figure 2d,e) and (high-resolution) mass spectrometry (Figures S34 and S35), elemental analysis, Mössbauer spectroscopy (Figure 2b,c), and cyclic voltammetry. [Fe(ImPP)₂][HCOO] was additionally characterized by single-crystal X-ray diffraction (Figure 1g). The ⁵⁷Fe Mössbauer spectrum of [Fe(ImPP)₂]⁺ (Figure 2b) measured at 80 K shows one quadrupole doublet with an isomer shift of −0.06 mm s^–1 and a quadrupole splitting of 1.41 mm s^–1. The spectrum of [Fe(ImPAr₂)₂]⁺ has its doublet at a similar isomer shift of −0.07 mm s^–1 while its quadrupole splitting is increased to 1.79 mm s^–1. Both spectra are compatible with carbene-ligated Fe^III in d⁵ low spin configuration, pointing to an S = 1/2 spin state with two pairs of d-electrons and one unpaired d-electron on the metal.^96,100

Figure 2.

(a) Synthesis route using a zirconium reagent to activate the ligand and subsequent transmetalation with iron(II) bromide to obtain both iron(III) complexes. (b) ⁵⁷Fe Mössbauer spectrum of [Fe(ImPP)₂]⁺ at 80 K. (c) ⁵⁷Fe Mössbauer spectrum of [Fe(ImPAr₂)₂]⁺ at 80 K. (d, e) ¹H NMR spectra of [Fe(ImPAr₂)₂][PF₆] (top) and [Fe(ImPP)₂][PF₆] (bottom) at 293 K in CD₃CN with the assignment of the individual resonances.

The ¹H NMR spectra of [Fe(ImPP)₂]⁺ and [Fe(ImPAr₂)₂]⁺ in CD₃CN further corroborate the structural integrity of these compounds (Figure 2d,e). Both complexes showed proton resonances in the range of −2.5 to 28 ppm because of the paramagnetic nature of these compounds. Comparing the chemical shifts of the protons of the cyclometalated phenyl core between both complexes (labeled “H_d” in Figure 2d,e) highlights the electron-donating character of the diarylamino substituent, manifesting in an upfield shift of the H_d proton resonance from 27.1 ppm in [Fe(ImPP)₂]⁺ to 23.0 ppm in [Fe(ImPAr₂)₂]⁺. A similar, though weaker trend is observed for the imidazole alkyl protons labeled H_g and H_h.

Dark green single crystals of [Fe(ImPP)₂][HCOO] suitable for X-ray diffraction were obtained from layering n-hexane on a saturated tetrahydrofuran (THF) solution at room temperature, resulting in the structure shown in Figure 1g. The average C_NHC–Fe–C_NHC bite angle of 155.6(3)° closely resembles the distorted octahedral geometry observed in other iron(III) and iron(II) complexes with similar ligands.^{94,100,115,116} The average dihedral angle between the two tridentate chelates is 89.8°. The Fe–C bond lengths of 1.940(7) Å for Fe–C_Ph and 1.980(9) Å for Fe–C_NHC are similar to the previously reported [Fe(ImP)₂]⁺ complex (Figure 1e).¹⁰⁰

UV–Vis Absorption and Photoluminescence Properties

[Fe(ImPP)₂]⁺ and [Fe(ImPAr₂)₂]⁺ in THF (Figure 3a) both feature a similar absorption band at 375 nm, which is obviously unaffected by the electron-donating substituents of the [Fe(ImPAr₂)₂]⁺ complex.

Figure 3.

(a) UV–vis absorption of [Fe(ImPP)₂]⁺ (solid red trace) and [Fe(ImPAr₂)₂]⁺ (solid green trace) in dry, deaerated THF at 293 K. The star (*) marks an anomaly caused by the detector change of the spectrometer. (b) Normalized luminescence spectra of a crude sample of [Fe(ImPP)₂]⁺ following excitation at 635 nm (solid red trace) and at 375 nm (solid blue trace) recorded in dry, deaerated THF at 293 K. (c) Jablonski diagram with key excitation processes and luminescence transitions. The black arrows indicate excitation at 375 and 635 nm, and the red arrow indicates LMCT emission. The gray crossed arrow indicates the absence of genuine MLCT or any other higher-excited state emission in this compound. (d) Normalized luminescence spectra of a sample of [Fe(ImPP)₂]⁺ at different purification steps. The emissions were normalized to 1 at 725 nm, corresponding to the maximum of the genuine ²LMCT luminescence (see the inset showing the same data sets on a magnified y-scale). RM, reaction mixture; Ext., after extraction; HPLC, after purification with preparative reversed-phase HPLC. The data in (d) shows that the 450 nm luminescence band is due to an emissive impurity.

This absorption band is likely due to a superposition of π–π* transitions on the aromatic backbones of both ligands (²LC) and ²MLCT transitions. The low-energy bands (630 nm for [Fe(ImPP)₂]⁺, 800 nm for [Fe(ImPAr₂)₂]⁺) are expected to have substantial LMCT character, analogously to recently studied iron(III) carbene complexes.^{96,97,100,111} In the simplest interpretation framework, the redshift of the lowest-energy absorption band between [Fe(ImPP)₂]⁺ and [Fe(ImPAr₂)₂]⁺ reflects the greater ease of ligand oxidation in the more electron-rich amine-decorated complex. However, in [Fe(ImPAr₂)₂]⁺ electronic transitions directly involving the diarylamino group could furthermore contribute, as reported for precious-metal-based complexes containing ligands with similar substituents.^117,118 The higher oscillator strength associated with the lowest-energy absorption band in [Fe(ImPAr₂)₂]⁺ relative to [Fe(ImPP)₂]⁺ seems compatible with this hypothesis, though other iron(III) carbene complexes lacking electron-donating substituents have similarly high molar extinction coefficients associated with electronic transitions to the lowest-energy ²LMCT excited state (Table 1).

Table 1. Summary of the Photophysical Properties of Fe^III Complexes.

entry	compound	λmax, abs, LMCT (ε) [nm (M–1 cm–1)]	λmax, em [nm]	τ0 [ps]	ELMCT [eV]	Φ [%]
1	[Fe(btz)3]3+a	558 (1200)	600	100	2.07	0.03
2	[Fe(phtmeimb)2]+b	502 (3000)	655	2000	1.89	2.1
3	[Fe(bimca)2]+c	825 (12 000)	-	1.3	-	-
4	[Fe(pzTp)(CN)3]−d	420 (2500)	600	80	2.10	0.02
5	[Fe(Imp)2]+e	600 (540)	735	240	1.69	<1i
6	[Fe(ImPP)2]+f	635 (660)	725	267	1.82g	0.07
7	[Fe(ImPAr2)2]+f	790 (3610)	-	6.5	1.41h	-

^aIn deaerated CH₃CN at 20 °C.⁹⁶

^bIn deaerated CH₃CN at 20 °C.⁹⁷

^cIn deaerated CH₃CN at 22 °C.⁵⁷

^dIn deaerated CHCl₃ at 20 °C.⁹⁹

^eIn deaerated CH₃CN at 20 °C.¹⁰⁰

^fIn dry, deaerated THF at 20 °C. λ_max, abs, wavelength of the LMCT absorption band maximum; λ_max, em, wavelength of the LMCT emission band maximum; τ₀, LMCT excited state lifetime; E_LMCT, energy of LMCT excited state; Φ, quantum yield of LMCT emission in solution at room temperature.

^gE_LMCT was estimated from the wavelength, at which 10% of the emission band maximum intensity was reached.

^hE_LMCT was estimated from the wavelength, at which 10% of the extinction at the ²LMCT absorption band maximum was reached, because no emission was detectable for this compound.

ⁱEstimated upper limit of the integrated LMCT and presumed MLCT double emission.¹⁰⁰

Photoexcitation of [Fe(ImPP)₂]⁺ into the ²LMCT absorption band (λ_exc = 635 nm) results in photoluminescence peaking at 725 nm in deaerated THF at room temperature (solid red trace in Figure 3b). The photoluminescence quantum yield (Φ) under these conditions is 0.07% and the luminescence lifetime is 256 ± 10 ps, whereby the latter is unaffected by oxygen from air (entry 6 of Table 1). The luminescence excitation spectrum matches the UV–vis absorption spectrum of [Fe(ImPP)₂]⁺ (Figure S39), supporting the view that the observed LMCT photoluminescence originates from the complex. The energy difference between the ²LMCT absorption and the ²LMCT emission band maximum is only 0.13 eV, considerably less than what is often observed in d⁶ luminophores between MLCT absorption and emission bands (0.5–0.8 eV), because in those cases, intersystem crossing from the initially excited ¹MLCT to emissive ³MLCT excited states causes additional energy losses.¹¹⁹

After excitation into the ²MLCT/²LC absorption band at 375 nm, an additional luminescence band at 450 nm was initially observed (solid blue trace in Figure 3b). Previously, an analogous double emission was reported for a structurally similar Fe(III) complex (Figure 1e),¹⁰⁰ where the higher-energy luminescence band maximizing near 450 nm was assigned to a transition originating from a higher-lying MLCT excited state, leading to the claim of dual LMCT and MLCT luminescence from the same compound.^100,106 We noticed that in our [Fe(ImPP)₂]⁺ sample, the relative intensity between the blue and red emission intensities strongly varies from sample to sample. With each additional purification step, less of the blue luminescence remained detectable (dotted, dashed, and solid blue traces in Figure 3d). In the reaction mixture resulting from the iron complexation step, the blue emission is dominant and the red ²LMCT luminescence is hardly seen in comparison (dotted blue trace in Figure 3d, labeled “RM”). After extraction, the intensity of the blue emission is decreased by roughly a factor of 3 (dashed blue trace in Figure 3d, labeled “Ext.”), using the red ²LMCT emission band maximum at 725 nm as a reference point (inset of Figure 3d). After reversed-phase HPLC, the red ²LMCT emission becomes dominant, and the blue emission is no longer discernible on the scale used in the main plot of Figure 3d (solid blue trace), yet remains detectable on the magnified scale of the inset of Figure 3d.

The solid blue trace in Figure 3d was obtained after repeated HPLC chromatography, and while each repetition of the chromatography step led to diminished blue emission intensity relative to the red luminescence, the blue emission could never be fully eliminated. There are no significant changes in the shape or position of the blue emission band detectable in the sequence of purification steps, suggesting that the blue emission originates from the same species in all cases, and is not a property of [Fe(ImPP)₂]⁺ (with trivalent iron), but instead originates from an impurity. Based on the ²LMCT excited state lifetime of 267 ps and the ²LMCT photoluminescence quantum yield of 0.07% for [Fe(ImPP)₂]⁺, the rate constant for radiative ²LMCT decay to the electronic ground state is k_r = 2.6 × 10⁶ s^–1. This value agrees well with the value obtained by Strickler–Berg analysis of the oscillator strength of the corresponding ²LMCT absorption band, yielding k°_r = 3.6 × 10⁶ s^–1 (Table 2, and Supporting Information page S36).^120,121 Applying the same analysis to the (integrated) ²MLCT/²LC absorption band at 375 nm leads to a radiative rate constant of 1.2 × 10¹⁴ s^–1, which is several orders of magnitudes higher than typical radiative fluorescence rate constants, which seems unphysical.¹⁰⁷

Table 2. Emission Spectral Fitting Parameters and Excited State Decay Rate Constants^a.

entry	compound	Ea [cm–1]	kr [s–1]	k°r [s–1]	knr [s–1]	knr,MC [s–1]	E0 [cm–1]	SM	h·ωM [cm–1]	Δν1/2 [cm–1]
1	[Fe(ImPP)2]+	390	2.6 × 106	3.6 × 106	3.7 × 109	3.2 × 109	13 675	0.95	1237	1300
2	[Ru(bpy)3]2+	3600	6.5 × 104	-	1.1 × 109	-	16 320	1.10	1350	1750

^aE_a, estimated activation barrier between the luminescent ²LMCT and an energetically nearby MC state based on temperature-dependent luminescence experiments (Figure 4a). k_r, experimentally determined radiative rate constant based on luminescence quantum yield and lifetime experiments at room temperature. k°_r, radiative rate constant obtained from Strickler–Berg analysis of room temperature data. k_nr, experimentally determined nonradiative ²LMCT excited state rate constant based on room-temperature luminescence lifetime and quantum yield data. k_nr,MC, rate constant for nonradiative ²LMCT deactivation via the MC channel. E₀ difference between the zero-point energies of the ground and the ²LMCT excited state. S_M, Huang–Rhys parameter obtained from spectral fitting of the room-temperature luminescence spectrum (Figure 4b). ℏ·ω_M, average energy of the vibrational mode coupling to the luminescence transition, as obtained from the spectral fitting in Figure 4b. Δν_1/2, bandwidth associated with the individual vibronic transitions observable in Figure 4b. The data in entry 2 is from refs (126) and (131).

The Jablonski diagram in Figure 3c summarizes all relevant excitation processes as well as genuine and nongenuine luminescence transitions. Following excitation at 375 nm, [Fe(ImPP)₂]⁺ undergoes internal conversion to its lowest-lying ²LMCT excited state, which emits red light. The mixed ²MLCT/²LC states of [Fe(ImPP)₂]⁺ in the blue spectral region are not emissive. No luminescence was detectable from the amine-decorated [Fe(ImPAr₂)₂]⁺ complex upon excitation in the visible region. This is in line with the energy gap law, according to which a decrease in the energy gap between the excited state and the ground state results in increased nonradiative relaxation.¹²² Assuming that the radiative ²LMCT decay rate constants of [Fe(ImPAr₂)₂]⁺ and [Fe(ImPP)₂]⁺ are identical, a luminescence quantum yield below 0.002% would be maximally expectable for [Fe(ImPAr₂)₂]⁺ on the basis of its 6.5 ps ²LMCT lifetime (see the UV–Vis Spectro-Electrochemistry and Transient UV–Vis Absorption Spectroscopy section).

Some key photophysical properties of the Fe^III complexes reported herein along with previously reported related compounds are summarized in Table 1. The ²LMCT absorption band maximum of [Fe(ImPP)₂]⁺ is slightly red-shifted compared to the previously investigated complex [Fe(ImP)₂]⁺ (entries 5 and 6 of Table 1), due to the more π-conjugated ligand structure resulting from the additional phenyl ring. Somewhat counterintuitively, in the [Fe(btz)₃]³⁺ complex (Figure 1a), in which the ²LMCT energy is higher (entry 1 of Table 1), a shorter luminescence lifetime is observed. This can be rationalized by a 2.5-fold faster nonradiative ²LMCT decay (k_nr = 9.9 × 10⁹ s^–1) in comparison to [Fe(ImPP)₂]⁺ despite a 0.25 eV increase in energy gap. The opposite trend is observed for the scorpionate complex [Fe(phtmeimb)₂]⁺ (Figure 1b), in which a slower nonradiative decay (k_nr = 4.9 × 10⁷ s^–1) combined with a 0.07 eV higher ²LMCT energy leads to a 2.0 ns lifetime (entry 2 of Table 1). The first luminescent heteroleptic complex ([Fe(pzTp)(CN)₃]⁻, Figure 1d) has LMCT absorption and emission band maxima at essentially the same wavelengths as [Fe(btz)₃]³⁺ and features a similar excited state lifetime of 80 ps (entry 4 of Table 1), in accordance with the energy gap law. Evidently, radiative and nonradiative ²LMCT excited state decay rate constants both vary considerably among the small family of iron(III) carbene complexes investigated until now.

Temperature-dependent emission lifetime measurements (Figure 4a) reveal an increase of the excited state lifetime of [Fe(ImPP)₂]⁺ by a factor of 3.5 (from 0.27 to 0.87 ns) upon decreasing the temperature from 298 to 77 K, similar to what was reported for other luminescent iron(III) complexes.^96,97,123 The experimentally determined lifetimes were fitted to an Arrhenius model (eq 1) describing a two-state thermally equilibrated system.^124−126

1

Figure 4.

(a) Decay rate constants (k_obs = τ^–1) extracted from ²LMCT luminescence decays of [Fe(ImPP)₂]⁺ in 2-methyl-THF at different temperatures, measured after excitation at 635 nm, detected at 730 nm (Figure S44). Solid red trace: Fit of the experimental data with eq 1 and the parameters given in the text. (b) Normalized luminescence spectrum of [Fe(ImPP)₂]⁺ at 20 °C following excitation at 635 nm (solid red trace) and simulated emission spectrum obtained with eq 2 and the parameters given in the text (dotted black traces).

The best fit was obtained with k_LMCT = 1.13 × 10⁹ s^–1, k_MC = 2.48 × 10¹⁰ s^–1, and an activation barrier (E_a) of 4.7 kJ mol^–1 (390 cm^–1). When assuming identical degeneracies for the ²LMCT and MC states, the Boltzmann population of the MC state at 293 K is approximately 12.7% of the total excited state population under these conditions.

The rate for nonradiative depopulation through the MC channel then becomes k_nr,MC = 0.127 × k_MC = 3.2 × 10⁹ s^–1, which is close to the value of k_nr = 3.7 × 10⁹ s^–1 determined on the basis of room-temperature luminescence lifetime and quantum yield measurements (Table 2). This simplistic two-state model (eq 1) therefore suggests that the main nonradiative ²LMCT relaxation pathway involves the MC channel. For comparison, in [Ru(bpy)₃]²⁺, the activation barrier to populate a higher-lying MC excited state is 43 kJ mol^–1 (3600 cm^–1) from the emissive ³MLCT state, almost 1 order of magnitude higher than the ²LMCT-MC activation barrier in [Fe(ImPP)₂]⁺.¹²⁶

The estimation of the activation barrier for nonradiative ²LMCT excited state via MC state population is a simplistic approach,¹²⁷ but is commonly used to analyze indirect nonradiative depopulation of luminescent charge transfer excited states via MC states. To gain insight into direct nonradiative ²LMCT depopulation to the electronic ground state (not involving any MC excited states), analysis of the luminescence band shape is useful to characterize the ²LMCT excited state distortion.¹²⁸ The respective analysis was performed using the room-temperature emission spectrum of [Fe(ImPP)₂]⁺ (solid red trace in Figure 4b). The experimental emission spectrum can be simulated by considering all individual vibronic transitions between the electronically excited state and the ground state.¹²⁹ The Franck–Condon factors, serving as indicators of the extent of overlap between the vibrational wave functions in the initial and final states of the relevant electronic transition, govern the intensity of each vibronic transition. The emission intensity I(ν) as a function of the energy ν can be described by eq 2, wherein E₀ represents the difference between the zero-point energies of the ground and excited states, and ℏ·ω_M represents the mean energy of the vibrational mode (with quantum number ν_M) that couples to the luminescence transition.^113,130S_M denotes the Huang–Rhys parameter, describing the extent of molecular distortion occurring between the two respective electronic states. The term Δν_1/2 is the homogeneously broadened bandwidth of the individual vibronic transitions.

2

The experimental luminescence spectrum was simulated by letting the index ν_M run over the number of relevant vibrational levels of ℏ·ω_M, serving as final states in the electronic ground state. To reach a satisfactory fit that matches the experimentally observed emission spectrum, we summed over ν_M values from 0 to 3 and adapted E₀, S_M, and ℏ·ω_M manually. The result of the simulation is included in Figure 4b as a dashed black trace, and the fitting parameters are summarized in Table 2 along with those reported previously for the [Ru(bpy)₃]²⁺ complex.¹³¹ For [Fe(ImPP)₂]⁺, the Huang–Rhys parameter S_M is lower by ∼15% compared to [Ru(bpy)₃]²⁺, which indicates a smaller degree of excited state distortion relative to the ground state. The ℏ·ω_M value of 1237 cm^–1 obtained by our simulations suggests that mostly aromatic C=C stretching vibrations (∼1400 cm^–1) along with aromatic C=N vibrations (∼1200 cm^–1) are contributing to the weighted average of all modes responsible for the distortion. The relatively weak ²LMCT excited state distortion suggests that nonradiative depopulation predominantly occurs via the nearby MC state, rather than via direct relaxation to the ground state, in line with the outcome of the temperature-dependent luminescence lifetime studies. Recent studies of [Fe(phtmeimb)₂]⁺ indicate that in this complex nonradiative relaxation from the ²LMCT excited state directly to the ground state is a key limitation to the lifetime.¹²³

Cyclic Voltammetry

Cyclic voltammetry measurements of [Fe(ImPP)₂]⁺ and [Fe(ImPAr₂)₂]⁺ in CH₃CN (Figure 5a) revealed two one-electron waves that can be attributed to the Fe^III/II and Fe^IV/III couples, in analogy to the previously reported [Fe(ImP)₂]⁺ parent complex from Figure 1a.¹⁰⁰ In the new [Fe(ImPP)₂]⁺ complex (solid red trace in Figure 5a), the one-electron reduction to Fe^II occurs at a half-wave potential of E_1/2 = −0.75 V vs SCE, whereas one-electron oxidation to Fe^IV takes place at 0.4 V vs SCE. An additional irreversible oxidation wave was observed at more positive potentials, likely attributable to the ligand. In [Fe(ImPAr₂)₂]⁺ (solid green trace), the reduction wave to Fe^II is at essentially the same potential as in [Fe(ImPP)₂]⁺. By contrast, the first electron oxidation wave is shifted to 0.2 V vs SCE in the presence of the electron-donating diarylamino group of [Fe(ImPAr₂)₂]⁺. Triarylamine oxidations typically occur at potentials near 0.8 V vs SCE, even when para-substituted with methoxy groups;^132,133 hence, it seems plausible to assign the wave observable for [Fe(ImPAr₂)₂]⁺ at 0.2 V vs SCE to the Fe^IV/Fe^III couple.

Figure 5.

(a) Cyclic voltammograms of 1 mM [Fe(ImPP)₂]⁺ (solid red trace) and 1 mM [Fe(ImPAr₂)₂]⁺ (solid green trace) in dry deaerated CH₃CN at 20 °C with 0.1 M [NBu₄][PF₆] as a supporting electrolyte. The scan rate was 0.1 V/s in both cases. Latimer diagrams of [Fe(ImPP)₂]⁺ (b) and [Fe(ImPAr₂)₂]⁺ (c) based on the voltammograms shown in (a) and (b), including the energies of the photoactive ²LMCT excited states from Table 1 (E_LMCT). Redox potentials for both complexes are reported in V versus SCE.

A summary of the ground state redox potentials and excited state energies is given in the Latimer diagrams in Figure 5b,c. The ²LMCT excited state energies of both complexes were estimated from their emission and absorption spectra, from the wavelengths at which 10% of the emission and absorption maxima was reached (see the footnote of Table 1 for details).

UV–Vis Spectro-Electrochemistry and Transient UV–Vis Absorption Spectroscopy

UV–vis spectro-electrochemisty data showing the spectral changes associated with Fe^III reduction to Fe^II and Fe^III oxidation to Fe^IV were obtained via bulk electrolysis for both [Fe(ImPP)₂]⁺ and [Fe(ImPAr₂)₂]⁺. The difference absorption spectrum resulting from the oxidation of Fe^III to Fe^IV and leading from [Fe(ImPP)₂]⁺ to [Fe(ImPP)₂]²⁺ (blue trace in Figure 6a) contains two broad bands peaking at 525 nm and >800 nm.

Figure 6.

(a, b) UV–vis absorption changes following metal-based oxidation (blue, applied potential: 0.5 V vs SCE) and reduction (brown, applied potential: −0.8 V vs SCE) of 1 mM [Fe(ImPP)₂]⁺ (a) and 1 mM [Fe(ImPAr₂)₂]⁺ (b) in deaerated CH₃CN at 20 °C. The electrolyte was 0.1 M [NBu₄][PF₆] in all spectro-electrochemical experiments. (c) Transient UV–vis absorption spectra of 1.5 mM [Fe(ImPP)₂]⁺ (solid red trace) and 0.25 mM [Fe(ImPAr₂)₂]⁺ (solid green trace) in deaerated THF at 20 °C. The data was obtained after excitation at 635 nm (solid red trace) and 700 nm (solid green trace) with a time delay of 1 ps (for both [Fe(ImPP)₂]⁺ and [Fe(ImPAr₂)₂]⁺). The inset contains the decays of the ESA signal at λ_obs = 433 nm (for [Fe(ImPP)₂]⁺) and λ_obs = 480 nm (for [Fe(ImPAr₂)₂]⁺), respectively.

In the difference spectrum resulting from the reduction of Fe^III to Fe^II and leading from [Fe(ImPP)₂]⁺ to [Fe(ImPP)₂], only a single broad absorption band with a maximum at 450 nm is detectable in the visible spectral range (brown trace in Figure 6a). In the corresponding Fe^III/Fe^IV difference spectrum for the amine-decorated [Fe(ImPAr₂)₂]⁺ complex (blue trace in Figure 6b), the two bands also seen in Figure 6a appear to be blue-shifted to 400 and 600 nm. Furthermore, a bleach at 775 nm becomes evident, because the [Fe(ImPAr₂)₂]⁺ complex has its lowest ²LMCT absorption band red-shifted in comparison to [Fe(ImPP)₂]⁺ (Figure 3a). Upon reduction of [Fe(ImPAr₂)₂]⁺ to [Fe(ImPAr₂)₂] (brown trace in Figure 6b), the respective bleach in the red spectral range is also observed (though less pronounced) along with an absorption band at 425 nm. The difference spectra in Figure 6 are useful to confirm the LMCT character of the lowest electronically excited states in [Fe(ImPP)₂]⁺ and [Fe(ImPAr₂)₂]⁺, as discussed in the following.

The transient UV–vis absorption spectra of [Fe(ImPP)₂]⁺ (solid red trace) and [Fe(ImPAr₂)₂]⁺ (solid green trace) in THF recorded after excitation at 635 and 800 nm into the respective ²LMCT absorption bands are in Figure 6c. A time delay of 1 ps was applied for [Fe(ImPP)₂]⁺ and [Fe(ImPAr₂)₂]⁺ in order to detect the fully developed transient absorption spectra of the photoactive ²LMCT state. In the spectrum of [Fe(ImPP)₂]⁺, the expected ground state bleach (GSB) in the spectral region around 600 nm is overcompensated by a broad excited state absorption (ESA). The ESA band with a maximum near 433 nm (red trace in Figure 6c) is considerably narrower than the difference absorption spectrum associated with the Fe^III to Fe^II reduction (brown trace in Figure 6a), yet this seems in line with the LMCT assignment of the electronically excited state probed by the transient absorption experiments. An additional ESA band peaking at 700 nm and tailing into the near-infrared region (red trace in Figure 6c) can tentatively be attributed to the ligand radical, though we have been unable to probe ligand oxidation by UV–vis spectro-electrochemical experiments because Fe^III to Fe^IV oxidation occurs at a lower potential than ligand oxidation (Figure 5a,b).

For [Fe(ImPAr₂)₂]⁺, the main ESA band is red-shifted from 433 to 480 nm with respect to [Fe(ImPP)₂]⁺ (green trace in Figure 6c). The observable weak GSB above 700 nm matches well with the expected disappearance of the ²LMCT absorption of the electronic ground state (green trace in Figure 3a and blue trace in Figure 6b). The ESA band at 480 nm in the transient absorption spectrum (green trace in Figure 6c) matches less well with the difference spectrum obtained for Fe^III to Fe^II reduction in the spectro-electrochemical measurements (brown trace in Figure 6b). Evidently, the correspondence between transient UV–vis absorption and spectro-electrochemical data for the photoactive LMCT excited states of our iron(III) complexes is considerably less good than what is typically found for the MLCT excited states of d⁶ metal complexes.¹³⁴ Clear-cut comparisons between spectro-electrochemical and transient UV–vis absorption spectra in iron(III) carbene complexes seem relatively scarce yet, and studies by other experimental (X-ray) techniques probing the nature of the photoactive excited state would seem worthwhile.^73,135

The transient absorption kinetics of both complexes at the ESA band maxima (Figure 6c, inset) show a single exponential decay with lifetimes of 277 ± 10 ps for [Fe(ImPP)₂]⁺ (solid red trace) and 6.5 ps for [Fe(ImPAr₂)₂]⁺ (solid green trace), respectively. The 277 ps lifetime is in good agreement with the ²LMCT emission lifetime of [Fe(ImPP)₂]⁺ (256 ps) obtained by time-correlated single photon counting (TCSPC, Figure S42), leading to the average lifetime of 267 ps reported in Table 1. The significantly shorter ²LMCT excited state lifetime of [Fe(ImPAr₂)₂]⁺ (6.5 ps) can be understood on the basis of the reduced energy gap to the electronic ground state in this compound relative to [Fe(ImPP)₂]⁺, which commonly leads to accelerated nonradiative ²LMCT relaxation, at least in the better explored ³MLCT excited states of d⁶ metal complexes.¹²² The reduced energy gap in [Fe(ImPAr₂)₂]⁺ compared to [Fe(ImPP)₂]⁺ can be rationalized by the strongly electron-donating diarylamino moiety, which decreases the energy between the relevant filled π-orbitals of the ligand and the t_2g-orbital vacancy involved in the lowest ²LMCT excitation. Somewhat more quantitative insight is possible by the energy gap law relationship according to eq 3, where ΔE is the energy of the lowest excited state, and ℏω_M is the mean energy of the vibrational mode that couples to the luminescence transition.⁸⁸ γ is a molecule-specific property describing the degree to which the relevant vibrational mode contributes to the Stokes shift.^113,136

3

Plotting ln(k_nr) against the excited state energies (ΔE) of [Fe(ImPP)₂]⁺ and [Fe(ImPAr₂)₂]⁺ (Figure S46) results in a slope of −0.001 cm^–1 (−8.1 eV^–1) when connecting the respective two data points. According to eq 3, this slope corresponds to −γ/ℏωM and the obtained numerical value is only −0.6 eV^–1 steeper than what is found typically for MLCT-emissive Ru^II polypyridine complexes (−7.5 eV^–1).¹²² This slight difference is likely related to the lower value of ℏω_M obtained for the Fe^III complexes investigated here (1237 cm^–1, Table 2) in comparison with the ℏω_M value typically used for the Ru^II polypyridines (1350 cm^–1).¹³¹ When taking this aspect into account by multiplying the two slightly different slopes by the relevant ℏω_M values, comparable γ values result for the ²LMCT excited states in iron(III) complexes and the ³MLCT excited states in ruthenium(II) complexes (1.24 vs 1.26). This crude analysis suggests that the nonradiative relaxation from charge transfer excited states with opposite charge transfer direction shows an essentially identical dependence on the energy gap to the electronic ground state.

Photostability

In view of possible applications in photocatalysis and upconversion (see below), the investigation of the photostability following a recently described method seemed relevant,¹²⁴ particularly for the [Fe(ImPP)₂]⁺ complex, owing to its longer ²LMCT excited state lifetime and its photoluminescence. The luminescence intensity of this compound at 725 nm was monitored as a function of irradiation time (Figure 7) with continuous-wave (cw) lasers at 405 nm (0.5 W) and 635 nm (1.0 W). The initial absorption of each sample was adjusted to 0.2 at the excitation wavelength, and the temperature was kept at 20 °C during the whole experiment. Using red excitation at 635 nm, no significant decrease in luminescence intensity for [Fe(ImPP)₂]⁺ was observable in all solvents over 5 h (red traces), despite the high excitation power of 1.0 W used, indicating very high photostability of [Fe(ImPP)₂]⁺ upon direct ²LMCT excitation, similar to recently reported iron(III) ²LMCT emitters based on other ligands.^97,137

Figure 7.

Photostability of [Fe(ImPP)₂]⁺. Luminescence of [Fe(ImPP)₂]⁺ at 725 nm upon irradiation with either a blue cw laser (405 nm, 0.5 W, blue traces) or a red cw laser (635 nm, 1.0 W, red traces) in different deaerated solvents at 20 °C.

Using blue excitation at 405 nm with only half the power (0.5 instead of 1.0 W), complete decomposition of the [Fe(ImPP)₂]⁺ complex is observed over time in CH₂Cl₂ and THF, whereas in acetonitrile and dimethyl sulfoxide (DMSO), the complex appears to remain intact (blue traces in Figure 7). Higher-energy excitation usually populates excited states with more dissociative character than lower-energy radiation, which could account for the differences observed in photostability when using blue and red light. The solvent dependence is more difficult to rationalize, as there is no obvious correlation between the coordinating character of the solvent and the photodecomposition rate, as well as between the solvent redox potentials and the photodecomposition rate. Nonetheless, it seems possible that photogenerated organic radicals originating from the solvent or from solvent impurities could negatively affect the long-term integrity of the iron(III) complex.

Photoinduced Electron Transfer

Based on the Latimer diagrams in Figure 5b, the ²LMCT-excited [Fe(ImPP)₂]⁺ complex should be a good oxidant (1.07 V vs SCE) and an even stronger excited state reductant (−1.42 V vs SCE). To test whether these excited state potential estimates are indeed reasonable, and to assess the application potential of [Fe(ImPP)₂]⁺ for both oxidative and reductive photoredox catalysis, it seemed meaningful to conduct nanosecond flash photolysis experiments with two benchmark redox reagents. As an electron donor, we chose methyl viologen (MV²⁺) featuring a one-electron reduction potential of −0.42 V vs SCE,¹³⁸ and as an electron acceptor, we opted for 10-methylphenothiazine (Me-PTZ), for which an oxidation potential of 0.71 V vs SCE has been reported.¹³⁹ Based on these redox potentials, we anticipate reaction free energies (ΔG°) of −1.0 eV for photoinduced electron transfer from [Fe(ImPP)₂]⁺ to MV²⁺, and −0.36 eV for electron transfer from Me-PTZ to ²LMCT-excited [Fe(ImPP)₂]⁺. Experiments were performed in acetonitrile (MV²⁺) and dichloromethane (Me-PTZ), respectively, owing to the different solubility of these two redox reagents. The anticipated electron transfer products MV^•+140 and Me-PTZ^•+141 were readily observed by transient absorption spectroscopy after excitation of [Fe(ImPP)₂]⁺ with a 532 nm pulsed laser (Figure 8a,b, red traces), as confirmed by complementary UV–Vis spectro-electrochemical measurements (blue traces in Figure 8a,b). Control experiments performed in the absence of the iron(III) complex underpinned the formation of the organic radicals by photoinduced electron transfer (PET) involving [Fe(ImPP)₂]⁺ (Figure 8a,b, black traces).

Figure 8.

(a) UV–vis transient absorption difference spectra obtained from a mixture of 0.5 mM [Fe(ImPP)₂]⁺ and 500 mM [MV][PF₆]₂ (red trace) and from a solution containing only 500 mM [MV][PF₆]₂ but no [Fe(ImPP)₂]⁺ (black trace) in deaerated CH₃CN at 20 °C. Both spectra were obtained after excitation with a 532 nm pulsed laser (100 mJ/pulse) using a delay time of 100 ns. UV–vis difference absorption spectrum obtained upon electrochemical reduction of MV²⁺ to MV^•+ at a potential of −0.7 V vs SCE in acetonitrile (blue trace). Inset: Stern–Volmer plot for 0.3 mM [Fe(ImPP)₂]⁺ with increasing [MV][PF₆]₂ concentrations in deaerated CH₃CN at 20 °C, based on 635 nm pulsed excitation and detection of the time-resolved ²LMCT luminescence at 730 nm using the TCSPC technique. (b) UV–vis transient absorption difference spectra obtained from a mixture of 0.5 mM [Fe(ImPP)₂]⁺ and 500 mM 10-methylphenothiazine (Me-PTZ) (red trace) and from a solution containing only 500 mM Me-PTZ but no [Fe(ImPP)₂]⁺ (black trace) in deaerated CH₂Cl₂ at 20 °C. Both spectra were obtained after excitation with a 532 nm pulsed laser (100 mJ/pulse) using a delay time of 100 ns. UV–vis difference absorption spectrum obtained upon electrochemical oxidation of Me-PTZ to Me-PTZ^•+ at a potential of 1.2 V vs SCE in acetonitrile (blue trace). Inset: Stern–Volmer plot for 0.1 mM [Fe(ImPP)₂]⁺ with increasing Me-PTZ concentrations in deaerated CH₂Cl₂ at 20 °C, based on 635 nm pulsed excitation and detection of the time-resolved ²LMCT luminescence at 730 nm using the TCSPC technique.

The rate constants for photoinduced electron transfer were determined by monitoring the ²LMCT luminescence lifetime quenching as a function of MV²⁺ and Me-PTZ concentration. The resulting Stern–Volmer plots (insets of Figure 8a,b) yield in both cases a value for dynamic quenching close to the diffusion-controlled limit, namely, k_q = 2.8 × 10⁹ M^–1 s^–1 for MV²⁺ and k_q = 9.5 × 10⁹ M^–1 s^–1 for Me-PTZ. The diffusion limits under the relevant conditions are 1.9 × 10¹⁰ M^–1 s^–1 in acetonitrile and 1.5 × 10¹⁰ M^–1 s^–1 in dichloromethane, respectively.¹⁴² Further (steady-state) luminescence intensity quenching experiments show an additional static quenching component for both electron transfers (Figure S47), suggesting that there is preassociation between the iron(III) complex and the reaction partners.⁹⁹

Reductive and Oxidative Photocatalysis

To explore the application potential of the Fe^III complex [Fe(ImPP)₂]⁺ in reductive photocatalysis, the light-driven C–H arylation of a diazonium salt was chosen as a benchmark reaction (Figure 9a). The oxidative ²LMCT excited state quenching of [Fe(ImPP)₂]⁺ with 4-methoxyphenyl diazonium is exergonic by 1.02 eV based on the data in Figure 5b and based on the known reduction potential of this substrate.¹⁴³ The reaction is expected to proceed via single-electron transfer (SET) from the excited Fe^III complex to the substrate, which forms a carbon-centered radical after N₂ loss.¹⁴⁴ Subsequently, the resulting organic radical is expected to react with furan, to form the C–C coupled product after deprotonation (for a detailed reaction mechanism, see Figure S51). In addition to its high driving force, the irreversible nature of the reaction seemed helpful in view of the short ²LMCT excited state lifetime of [Fe(ImPP)₂]⁺. The same reaction has been investigated previously using Eosin-Y as a photocatalyst under 530 nm light irradiation,¹⁴⁵ but in this case, the photoactive excited state was roughly 8 times longer lived (2.1 ns).¹⁴⁶ A mixture of [Fe(ImPP)₂]⁺ (0.5 mol %) with 10 equiv of furan was irradiated with red light (LED λ = 623 nm, 3.6 W output power, ca. 0.4 W reaching the sample) in DMSO-d₆ under argon.

Figure 9.

(a) Photocatalytic reductive C–H arylation of 4-methoxyphenyl diazonium salt with furan in deaerated DMSO-d₆. The reaction was performed in a water-bath-cooled cooled NMR tube irradiated with a red LED (623 nm, output power of 3.6 W). The yield and the conversion were determined via ¹H NMR spectroscopy using 1,3,5-trimethoxybenzene as the internal standard. (b) Photocatalytic aerobic oxidative hydroxylation of different phenylboronic acids in an oxygen-saturated deuterated acetonitrile/water mixture. The individual reactions were performed in a water-bath-cooled NMR tube irradiated with a red LED (623 nm, output power of 3.6 W, power reaching the sample was ca. 0.4 W). The yields and conversions were determined based on ¹H NMR spectroscopy using phenyl trimethylsilane as the internal standard.

The reaction progress was monitored by ¹H NMR spectroscopy, and C–H arylation product yields were referenced to an internal standard. After 2 h of reaction time, full conversion of the substrate was observed with an overall product yield of 59%. Further screening of other substrates with more positive reduction potentials led to higher C–C coupling product yields but also increased background reactions, whereas using more challenging substrates with less positive reduction potentials decreased the yield and the conversion significantly (data not shown).

The key finding is that reductive photocatalysis with [Fe(ImPP)₂]⁺ requires highly activated substrates with very positive reduction potentials undergoing irreversible photoinduced electron transfer, owing to the short (267 ps) lifetime of the ²LMCT excited state. High substrate concentrations are furthermore needed to make bimolecular electron transfer competitive with inherent ²LMCT decay.

Given the photo-oxidizing ability of the [Fe(ImPP)₂]⁺ complex seen with Me-PTZ (Figure 8b), it seemed attractive to furthermore explore its application in oxidative photoredox chemistry. For that purpose, the photocatalytic aerobic hydroxylation of phenylboronic acids to phenols was investigated (Figure 9b, top), knowing that the ²LMCT excited state of [Fe(ImPP)₂]⁺ is not directly quenched by ³O₂ (see above). This reaction has been explored previously with [Ru(bpy)₃]²⁺,¹⁴⁷ a Cr^III polypyridine complex,¹⁴⁸ Cu@N₃C₄,¹⁴⁹ and also metal-free versions using methylene blue,¹⁵⁰ N-substituted acridones,¹⁵¹ Eosin-Y,¹⁵² and Rose Bengal were used.¹⁵³ Recently, an efficient hydroxylation procedure was accomplished using chlorophyll extracted from spinach in ethanol.¹⁵⁴

First experiments with [Fe(ImPP)₂]⁺ (2 mol %) were performed in an acetonitrile/water mixture (4:1) with triethylamine (TEA) as a sacrificial electron donor under air with red light irradiation (LED λ = 623 nm, 3.6 W output power; the power at the sample position was ca. 0.4 W). This procedure led to a maximum yield of 30% of the 4-methoxyphenol product starting from 4-methoxyphenylboronic acid after 16 h of irradiation. Changing to an oxygen atmosphere increased the yield to >95% under otherwise identical conditions. A significant influence of the oxygen concentration was already previously observed in the same reaction catalyzed by [Cr(dqp)₂](PF₆)₃ (dqp = 2,6-bis(8′-quinolinyl)pyridine).¹⁴⁸ The substrate scope of the hydroxylation reaction was examined for various other phenylboronic acids bearing either electron-donating or -withdrawing groups (Figure 9b, bottom). There was no noticeable difference in the required reaction time for the various substrates, and full conversion and high yields were achieved after 16 h of irradiation in all cases.

The previously used photocatalysts for this reaction type typically had far longer excited state lifetimes than [Fe(ImPP)₂]⁺. For instance, the photoactive excited state of the recently employed [Cr(dqp)₂]³⁺ complex has a lifetime of roughly 30 μs in aerated solution,¹⁴⁸ more than 5 orders of magnitude longer than the ²LMCT lifetime of [Fe(ImPP)₂]⁺. Against this background, the photoreactivity observed in Figure 9b seems quite remarkable. Photoexcited perylene diimide radical anion (PDI^•–) has a similarly short excited state lifetime as [Fe(ImPP)₂]⁺,^155,156 and was invoked as a catalytically relevant species in photoredox catalytic reactions.¹⁵⁷ Mechanistic studies questioned whether such short-lived species can indeed be competent mediators of light-driven reactions,^158−160 and similar concerns seem well taken for a rather broad range of organic radical anions and their energetically lowest-lying electronically excited states.¹⁶¹ Often, the key question is whether the ps-lived species is indeed the catalytically relevant species, or whether a degradation or follow-up product of the radical anion is responsible for the main reactivity.^162−165 Based on the combination of UV–vis transient absorption and photocatalysis studies presented herein, the case for [Fe(ImPP)₂]⁺ seems comparatively clear-cut, underscoring genuine photoreactivity from its short-lived ²LMCT state. The example of [Fe(ImPP)₂]⁺ illustrates however the rather stringent set of conditions to be fulfilled in order to promote successful bimolecular reactions with very short-lived excited states, namely, high substrate concentrations and large driving forces.¹⁶⁶ Inner-sphere reactions (involving substrate coordination to Fe^III) do not suffer from this limitation.^167−171

Red to Green Upconversion

In principle, the photoactive ²LMCT excited state of [Fe(ImPP)₂]⁺ might be able to sensitize triplet–triplet annihilation upconversion (TTA-UC), but there are two key challenges not faced in most previously investigated upconversion systems.^172−178 First, most photosensitizers used for this purpose so far have very long-lived (≥100 ns) triplet excited states, which undergo efficient triplet–triplet energy transfer (TTET) to a fluorescent annihilator.^179−183 The ²LMCT lifetime of [Fe(ImPP)₂]⁺ is only ∼267 ps; hence, the inherent (unimolecular) excited state decay in this case will be far more competitive with bimolecular energy transfer than in the vast majority of previously explored cases. Second, the luminescent LMCT state of [Fe(ImPP)₂]⁺ has doublet spin multiplicity, and doublet-triplet energy transfer (DTET) is comparatively little explored and seems yet relatively poorly understood, despite important recent progress indicating that DTET can enable upconversion with high quantum yields.^80,184 Additional challenges are to be expected owing to the limited solubility of typical polyaromatic hydrocarbon annihilators in organic solvents, because the short ²LMCT lifetime of [Fe(ImPP)₂]⁺ asks for a large annihilator concentration, to make bimolecular DTET kinetically competitive. Against this background, the present study aimed mostly at providing the proof of concept for upconversion with a molecular iron-based photosensitizer.

We identified perylene with a triplet energy (E_T) of 1.53 eV¹⁴² as a potentially suitable annihilator because this energy acceptor can be expected to result in a DTET reaction that is exergonic by roughly 0.3 eV from the ²LMCT excited state of the Fe^III sensitizer (E_LMCT = 1.82 eV). Above this driving-force threshold, the better explored triplet–triplet energy transfer (TTET) reactions typically occur with diffusion-limited kinetics.¹⁸⁵ Saturated solutions of perylene in THF were explored with 0.1 mM of [Fe(ImPP)₂]⁺, but strong excimer emission was observed due to the high concentration of the annihilator. Excimer emission is usually undesirable for upconversion because this greatly limits the maximally achievable (pseudo) anti-Stokes shift, i. e. the difference between the excitation energy and the energy of the upconversion luminescence band maximum.^175,186 Therefore, 9,10-bis(phenylethynyl)anthracene (BPEA, Figure 10d) was chosen as an alternative annihilator, because it is less prone to excimer formation and provides even greater driving force for DTET (E_T = 1.26 eV).¹⁸⁷ Furthermore, the molar extinction of the iron(III) photosensitizer is low in the spectral region of the BPEA singlet emission, which limits unwanted reabsorption of the upconversion luminescence, one of the most challenging aspects in general for sTTA-UC.¹⁷⁵ A solution containing 0.1 mM of [Fe(ImPP)₂]⁺ and 18 mM BPEA in deaerated 1,2-dichloroethane (DCE) was excited at 635 nm with a cw laser. This led to delayed BPEA fluorescence with a maximum at 508 nm (blue traces in Figure 10a,e), as well as prompt ²LMCT emission of [Fe(ImPP)₂]⁺ maximizing at 725 nm (red traces in Figure 10a). The UC-emission maximum is blue-shifted by 0.52 eV relative to the excitation wavelength of the laser, which corresponds to a somewhat modest, yet appreciable pseudo anti-Stokes shift,^178,188 considering the substantial driving force (0.56 eV) associated with DTET in this sensitizer-annihilator couple. Expectedly, in the absence of [Fe(ImPP)₂]⁺ no BPEA fluorescence occurs under these conditions (Figure S65). The blue upconversion emission is in good agreement with the prompt fluorescence spectrum of BPEA obtained after direct excitation at 450 nm (Figure S64), confirming that this emission originates from the annihilator. Time-resolved luminescence spectroscopy confirms the delayed nature of the BPEA upconversion fluorescence, yielding a lifetime of 50 μs (Figure S63) upon excitation at 532 nm, whereas the prompt ¹BPEA fluorescence excited at 450 nm has a lifetime of 3.9 ns under identical conditions (Figure S66).¹⁸⁹ The lifetime of the T₁ state of BPEA is roughly 100 μs (Figure S62), twice as long as the lifetime of the delayed upconversion fluorescence, as expected in the so-called weak annihilation limit.^188,190,191

Figure 10.

Sensitized triplet–triplet annihilation upconversion (sTTA-UC) using [Fe(ImPP)₂]⁺ as the photosensitizer and 9,10-bis(phenylethynyl)anthracene (BPEA) as the annihilator. (a) Steady-state emission spectra recorded from a solution containing 0.1 mM [Fe(ImPP)₂]⁺ exclusively (red traces) and a solution containing 18 mM BPEA in addition to 1 mM [Fe(ImPP)₂]⁺ (blue traces) in deaerated 1,2-dichloroethane (DCE) at 20 °C, both following cw laser excitation at 635 nm using different excitation densities under otherwise identical conditions. In the absence of BPEA (red traces), only ²LMCT luminescence emitted by the iron(III) complex is detected, whereas in the presence of 18 mM BPEA, the upconverted delayed becomes detectable (blue traces). (b) Upconversion luminescence quantum yield as a function of excitation power density at 635 nm, determined relative to the ²LMCT luminescence quantum yield of [Fe(ImPP)₂]⁺ (Supporting Information (SI) S34–S35). (c) Normalized upconversion luminescence intensity as a function of excitation power density, to estimate the threshold excitation power density of the upconversion system. (d) Energy-level diagram of the [Fe(ImPP)₂]⁺ sensitizer and the BPEA annihilator showing a summary of the key photophysical processes in the upconversion system. (e) Photograph of the upconverting sample with [Fe(ImPP)₂]⁺ and BPEA (right) and the sample containing only [Fe(ImPP)₂]⁺ but no BPEA (left) under 635 nm laser excitation (laser power = 1.0 W).

Excitation power density-dependent measurements were performed in parallel with the sensitizer-annihilator couple (blue traces in Figure 10a) and for a solution containing only [Fe(ImPP)₂]⁺ but no BPEA (red traces in Figure 10a). The integrated prompt and upconverted emission intensities were compared to determine the excitation power density dependence of the upconversion luminescence quantum yield (Φ_UC, Figure 10b). Given a ²LMCT luminescence quantum yield (Φ) of 0.07% in deaerated DCE at 20 °C (Table S3), this analysis yields a maximal upconversion luminescence quantum yield of 0.019% (relative to a theoretical limit of 50%,¹⁹²Figure 10b), reachable at excitation power densities above 80 W/cm². This is a very high value for sTTA-UC systems,¹⁷⁷ but in molecular lanthanide complexes operating through different upconversion mechanisms, excitation power densities on this order of magnitude do not appear to be uncommon.^193−196 By plotting the normalized upconversion emission intensity as a function of the excitation power density (Figure 10c), the expected quadratic power dependency is obtained (blue line, slope 2.0) in the low power regime.¹⁹⁷ At higher excitation power densities, the strong annihilation limit is reached (red line, slope 1.0), with a threshold (Φ_th) of ∼46 W/cm² for the crossover between the two different regimes. These upconversion performance factors (Φ_UC, Φ_th) are expectedly somewhat modest, given the short ²LMCT excited state lifetimes of [Fe(ImPP)₂]⁺ compared to what is achievable with traditionally employed triplet sensitizers.¹⁷⁸ The maximum solubility of BPEA in DCE at 20 °C is 18 mM, and at this concentration, one can expect a maximum efficiency of ∼4% for DTET from the short-lived ²LMCT excited state (267 ps, Table 1) to BEPA assuming diffusion-controlled reaction (8.5 × 10⁹ M^–1 s^–1 for DCE at 25 °C).¹⁴² Therefore, it seems plausible to conclude that DTET is a key upconversion performance-limiting factor, though the DTET rate constant could not be determined with straightforward ²LMCT luminescence quenching experiments; at the very high concentrations of annihilator necessary for such experiments, an additional long-lived emission attributable to trace impurities in the BPEA sample interfered.

Conclusions

Following the unanticipated discovery of a luminescent iron(III) carbene complex in 2017 (Figure 1a),⁹⁶ a handful of structurally related iron(III) compounds featuring a photoactive ²LMCT excited state were reported (Figure 1b,c).^{57,96,97,99,100,111} Recent work on scorpionate carbene ligands (Figure 1b) inspired by earlier studies on closely related Fe^III198 and Mn^IV199 complexes demonstrated that the luminescence color emitted by the ²LMCT excited state is somewhat difficult to tune in this class of compounds.⁹⁸ Our study demonstrates that the ²LMCT energy in iron(III) carbene complexes is indeed tunable, manifesting in a shift of the absorption band maximum by 3100 cm^–1 (Figure 3a) between [Fe(ImPP)₂]⁺ (Figure 1f) and [Fe(ImPAr₂)₂]⁺ (Figure 1h). This redshift is accompanied by an approximately 40-fold reduction of the ²LMCT lifetime from 267 to 6.5 ps (Table 1). A simple analysis reveals that the rate for nonradiative relaxation from the LMCT state in these d⁵ iron(III) complexes depends very similarly on the energy gap to the ground state as the MLCT excited states of d⁶ noble metal complexes. In other words, nonradiative relaxation from excited states with opposite charge transfer directions shows a very similar energy gap dependence, a fundamental aspect that so far received limited attention. Spectral emission band shape analysis for the [Fe(ImPP)₂]⁺ complex (Figure 4b) furthermore leads to the new insight that the ²LMCT excited state of this complex is less distorted with respect to the electronic ground state than the luminescent ³MLCT excited state in prototypical d⁶ metal compounds.^113,131 The activation barrier for thermally activated depopulation of the luminescent ²LMCT excited state via a metal-centered excited state (MC) in [Fe(ImPP)₂]⁺ is similarly low as in previously investigated Fe^III carbenes (390 cm^–1 or 4.8 kJ/mol; Table 2);^96,97 hence, this deactivation channel remains a major unwanted excited state relaxation pathway also in our compounds. Evidently, the activation barrier for ²LMCT depopulation in Fe^III carbenes via a higher-lying MC state is roughly 3–6 times smaller than the activation barrier for ³MLCT deactivation via higher-lying MC states in Cr⁰ arylisocyanides,^65,67 and roughly 10 times smaller than for Ru^II polypyridines and Mo⁰ isocyanides.^126,201,202

Following excitation at 375 nm, samples of [Fe(ImPP)₂]⁺ feature apparent dual emission very similar as reported recently for the structurally closely related [Fe(ImP)₂]⁺ complex (compare Figure 1e,f),¹⁰⁰ but the blue emission in our case is clearly attributable to an impurity (Figure 3d). Moreover, a simple (Strickler–Berg) analysis of the observed oscillator strength of the respective ²MLCT/²LC absorption band of [Fe(ImPP)₂]⁺ points to an unphysical rate constant for radiative relaxation from the respective state in the blue spectral range (1.2 × 10¹⁴ s^–1). These two combined findings are in line with a very recent study that appeared while our work was still in progress,¹¹¹ which reached the conclusion that the dual emission reported for [Fe(ImP)₂]⁺ is not genuine.^100,106

Combined time-resolved luminescence and transient absorption studies (Figure 8) demonstrate that dynamic ²LMCT excited state quenching by photoinduced electron transfer is readily possible despite the subnanosecond lifetime of the relevant excited state. Many organic radical ions have similarly or even shorter-lived electronically excited states,^155,156,203 and substantial controversy emerged over the question to what extent these very short-lived species can indeed undergo productive photoreactions, and to what extent (photo)degradation products, side products, or preaggregated adducts perhaps play a more dominant role.^{157,158,161,164,204−206} For the Fe^III carbene complexes, the situation seems comparatively clear-cut with unambiguous photoreactivity from the luminescent ²LMCT excited state, where the main preconditions for [Fe(ImPP)₂]⁺ are high concentrations of the redox partners (≥200 mM) and driving forces of at least 0.3 eV, to allow for bimolecular reactions near the diffusion-controlled limit. Given the remarkable photostability (Figure 7) and favorable redox properties of [Fe(ImPP)₂]⁺ (Figure 5), both oxidative redox catalysis and reductive redox catalysis were achieved with high yields. The reductive activation of aryl diazonium salts has been previously accomplished with LMCT-excited iron(III) complexes, but usually UV, blue, or green excitation was used,^103,105,207 whereas the C–H arylation reported herein (Figure 9a) was driven by red light. Longer-wavelength irradiation is generally considered as milder, causing less photodamage and can benefit from enhanced light penetration depths into reaction vessels.^{163,208−210}

The aerobic oxidative hydroxylation of phenylboronic acids (Figure 9b) was previously accomplished with a Cr^III-based photocatalyst featuring an excited state lifetime that was more than 5 orders of magnitude longer than the ²LMCT lifetime of [Fe(ImPP)₂]⁺.¹⁴⁸ The finding that these reactions run similarly well with [Fe(ImPP)₂]⁺ is testimony of the high photoreactivity of short-lived ²LMCT excited states in Fe^III carbenes.

Last but not least, the [Fe(ImPP)₂]⁺ complex allowed for the first proof of concept for triplet–triplet annihilation upconversion sensitized by a molecular iron complex (Figure 10). The upconversion performance reachable so far is limited mainly by the short ²LMCT lifetime, restricting the efficiency of doublet to triplet energy transfer to the annihilator to merely 4% under optimal conditions, despite more than 0.5 eV driving force. Triplet reservoir effects as known from Ru^II-pyrene and Ir^III-naphthalene bichromophores could potentially elongate the ²LMCT lifetime of Fe^III carbene complexes substantially.^{188,211−216} Previous studies have demonstrated that this bichromophore approach can have a strong positive impact on the upconversion performance,^{188,217−220} which might open new avenues for iron-based upconversion. Collectively our findings deepen the current understanding of the photophysical behavior and the photochemical application potential of Fe^III carbene complexes.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.4c00605.

• General procedures and equipment details, synthesis and characterization of ligand precursors and organometallic complexes, X-ray data, NMR and HR-ESI mass spectra, Mössbauer spectra, and photophysical and photochemical data (PDF)

Author Present Address

^§ School of Chemical Sciences, Indian Institute of Technology (IIT) Mandi, Kamand, Himachal Pradesh 175005, India

Swiss National Science Foundation (grant numbers 200020_207329 and 187043); Deutsche Forschungsgemeinschaft (DFG) in the framework of the Priority Program SPP 2102 (project number 404391096)

The authors declare no competing financial interest.