Cobalt(III) Carbene Complex with an Electronic Excited-State Structure Similar to Cyclometalated Iridium(III) Compounds.

Many organometallic iridium(III) complexes have photoactive excited states with mixed metal-to-ligand and intraligand charge transfer (MLCT/ILCT) character, which form the basis for numerous applications in photophysics and photochemistry. Cobalt(III) complexes with analogous MLCT excited-state properties seem to be unknown yet, despite the fact that iridium(III) and cobalt(III) can adopt identical low-spin d6 valence electron configurations due to their close chemical relationship. Using a rigid tridentate chelate ligand (LCNC), in which a central amido π-donor is flanked by two σ-donating N-heterocyclic carbene subunits, we obtained a robust homoleptic complex [Co(LCNC)2](PF6), featuring a photoactive excited state with substantial MLCT character. Compared to the vast majority of isoelectronic iron(II) complexes, the MLCT state of [Co(LCNC)2](PF6) is long-lived because it does not deactivate as efficiently into lower-lying metal-centered excited states; furthermore, it engages directly in photoinduced electron transfer reactions. The comparison with [Fe(LCNC)2](PF6), as well as structural, electrochemical, and UV-vis transient absorption studies, provides insight into new ligand design principles for first-row transition-metal complexes with photophysical and photochemical properties reminiscent of those known from the platinum group metals.

Introduction

Coordination complexes and organometallic compounds of ruthenium(II) and iridium(III) are among the most widely used substance classes for applications in lighting,[1,2] sensing,[3] photocatalysis,[4−9] upconversion,[10,11] solar energy conversion, and phototherapy.[12−18] Research on precious-metal-based compounds continues to be important because they are comparatively robust and often promise excellent performance.[19−35] Nevertheless, there is now an increasing interest in photoactive complexes made from first-row transition metals.[36,37] Higher abundance and lower cost are often quoted as key motivators for research in this direction,[38] but aside from such practical aspects, there is still much room for fundamental and groundbreaking discovery in the photophysics and the photochemistry of first-row transition-metal complexes.[39−44]

Complexes with semiprecious copper(I) have been investigated particularly thoroughly.[45−56] Given their filled 3d10 subshell, long-lived excited states are far more easily established in copper(I) complexes than in complexes with a partially filled 3d subshell because there are no low-lying metal-centered (MC) states that facilitate undesired nonradiative excited-state relaxation. Among first-row transition-metal compounds with partially filled d-orbitals, six-coordinate chromium(III) complexes are popular[57] because the lowest spin-flip MC state of octahedral d3 complexes is only weakly distorted relative to the electronic ground state, which is helpful to obtain unusually long-lived luminescence.[58−62] By contrast, MLCT excited states of nickel(II) (3d8)[63−66] and iron(II) (3d6)[44,67−71] complexes typically deactivate very rapidly, making it very difficult to use these compounds for similar applications as second- or third-row d8 and d6 metal complexes, in which long-lived MLCT states govern the photophysics and the photochemistry.[67−73]

Photoactive iron(II) complexes have been investigated for several decades, with much focus on excited-state deactivation pathways,[74−76] the establishment of longer-lived MLCT states,[77−83] and the use of such compounds in photoredox catalysis.[84−87] By contrast, the photophysics and photochemistry of isoelectronic cobalt(III) complexes have received surprisingly little attention, and the comparatively few examples investigated in detail until now either have photoactive MC or ligand-to-metal charge transfer (LMCT) excited states.[88,89] Here, we disclose an organometallic cobalt(III) complex featuring a photoactive excited state with substantial MLCT character, similar to what is well known from many cyclometalated iridium(III) compounds. This MLCT excited state does not relax as efficiently into lower-lying MC states as is typically observed in isoelectronic iron(II) complexes,[36,55−57,74−85,90−95] and it can undergo photoinduced electron transfer in an analogous fashion as many ruthenium(II) and iridium(III) complexes.

The higher oxidation state of cobalt(III) with respect to isoelectronic iron(II) causes a stronger ligand field, and this is helpful in terms of suppressing unwanted nonradiative relaxation from distorted MC excited states.[17,97] Owing to the particularly strong ligand field of [Co(CN)6]3–, its lowest MC state (3T1) is sufficiently high in energy for luminescence to become a competitive deactivation pathway.[98,99] Building on these early findings, the use of an anionic tripodal borate ligand with three strongly σ-donating carbene subunits recently led to a cobalt(III) complex with enhanced luminescence properties from that 3T1 state (Figure a).[88] In separate studies by other investigators who used tridentate imine σ-donor ligands, two cobalt(III) complexes featuring emission from a state with substantial LMCT character were obtained (Figure b),[89] contributing to the recent surge of interest in LMCT luminophores and photocatalysts.[100−109]

Figure 1

Molecular structures of previously investigated pertinent complexes along with the compounds investigated herein. (a) Cobalt(III) complex emitting from an MC state;[88] (b) cobalt(III) complex emitting from an excited state with substantial LMCT character;[89] (c) iron(II) complex with amido π-donor ligand units;[78,96] (d) new [Co(LCNC)2]+ and [Fe(LCNC)2]+ compounds investigated herein.

In our work aiming at cobalt(III) complexes with a photoactive MLCT state, it seemed desirable to explore ligands that increase the electron density at the metal center (to lower the CoIV/III potential). Furthermore, it seemed necessary to have ligands with low-lying π* orbitals (as landing spots for MLCT-excited electrons) while maintaining a strong ligand field (to shift MC states to higher energies). In recently explored iron(II) complexes, amido π-donor ligands markedly lowered the metal oxidation potential (Figure c),[78,96] and therefore, we focused on a tridentate ligand (Figure d), in which a central anionic carbazolate bears two N-heterocyclic carbene units (LCNC). This chelate ligand follows the abovementioned design principles; furthermore, it can potentially provide rigid coordination of cobalt(III) with bond angles close to ideal octahedral symmetry, which is known to counteract unwanted nonradiative relaxation.[37,57,110,111] Until now, this ligand (LCNC) had been used for square-planar d8 complexes based on rhodium(I),[112] iridium(I),[113] nickel(II),[114] palladium(II),[114,115] and platinum(II),[114] mostly in contexts not related to photophysics, except for platinum(II) alkynyl compounds with luminescent MLCT states.[116,117] Although the main focus of our work was on cobalt(III), it seemed useful to explore at the same time an isostructural iron(III) complex as a reference compound. Indeed, the comparative investigation of the 3d6 and 3d5 compounds [Co(LCNC)2](PF6) and [Fe(LCNC)2](PF6) provides insight into new ligand design principles for first-row transition-metal complexes with photophysical and photochemical properties reminiscent of those known from precious metal compounds.

Results and Discussion

Synthesis and Characterization of the Complexes

The ligand precursor (H3LCNC)(PF6)2 was synthesized via an anion exchange reaction from (H3LCNC)(I)2. The latter compound was prepared by adapting a previously published procedure (see Supporting Information Page S3).[114] The reaction of 1 equiv of (H3LCNC)(PF6)2 and 0.5 equiv of CoCl2 or FeBr2 in the presence of 3.3 equiv of potassium tert-butoxide in tetrahydrofuran at 55 °C for 18 h, followed by subsequent oxidation under air, resulted in the formation of the [Co(LCNC)2](PF6) and [Fe(LCNC)2](PF6) compounds (Scheme ) as orange and green solids, respectively. Both substances are insensitive to air and moisture.

Scheme 1

Synthesis of the Cobalt(III) and Iron(III) Compounds Investigated Herein

The diamagnetic cobalt(III) compound was characterized by NMR spectroscopy, mass spectrometry, combustion analysis, and X-ray crystallography (Supporting Information Page S4). Upon formation of [Co(LCNC)2](PF6), the characteristic N–H proton and the (N)C–H(N) protons of the imidazolium groups of (H3LCNC)(PF6)2 at 11.45 and 9.72 ppm, respectively, are no longer observable in the 1H NMR spectrum. Instead, a new resonance appears at 172.0 ppm in the 13C NMR spectrum of [Co(LCNC)2](PF6), attributable to the ligating carbon atom of the N-heterocyclic carbene (NHC) units, in line with previously reported cobalt(III)-NHC complexes.[88] Orange-colored single crystals of [Co(LCNC)2](PF6)·2C2H4Cl2 suitable for an X-ray diffraction study were obtained by slow vapor diffusion of pentane into a saturated 1,2-dichloroethane solution of [Co(LCNC)2](PF6) at ambient temperature. Structure analysis confirmed the formation of the expected diamidotetracarbene cobalt(III) complex cation [Co(LCNC)2]+ (Figure a), which has a close-to-perfect octahedral geometry around the cobalt(III) center. The Namido–Co–Namido bond angle is 179.5(2)°, and the CNHC–Co–CNHC bond angle is 176.67(14)°. The Co–Namido bond length (1.910(3) Å) is shorter than the Co–CNHC bond lengths (1.997(4) and 2.000(4) Å), and this is compatible with the more covalent character of the Co–Namido bond in the present CoIII complex.[96,118] The two imidazolidene rings in LCNC are not coplanar but rather tilted from the central carbazole plane, and hence, the LCNC ligand adopts a helical twist compatible with P or M chirality according to Cahn–Ingold–Prelog notation.[119] In the crystal structure, the CoIII complex cation exists as a pair of PP-[Co(LCNC)2]+/MM-[Co(LCNC)2]+ enantiomers, similar to recently reported CrIII complexes.[59,62,118,120]

Figure 2

Molecular structures of the complex cations in [Co(LCNC)2](PF6)·2C2H4Cl2 (a) and in [Fe(LCNC)2](PF6)·2C2H4Cl2 (b); 50% probability ellipsoids used in both cases. Hydrogen atoms, the counter anion, and the solvent molecules have been omitted for clarity. Selected bond lengths (Å) and angles (°) for [Co(LCNC)2]+: Co1–C1, 1.997(4); Co1–C28, 2.000(4); Co1–N3, 1.910(3); C1–Co1–C28, 176.67(14)°; N3–Co1–N3′, 179.5(2); and for [Fe(LCNC)2]+: Fe1–C2, 2.016(5); Fe1–C25, 2.013(4); Fe1–N3, 1.910(4); C2–Fe–C25, 175.9(2)°; N3–Fe–N3′, 179.6(2)°. Different enantiomers are shown for the two complexes.

The paramagnetic [Fe(LCNC)2](PF6) compound was characterized by combustion analysis, high-resolution electrospray ionization (HR-ESI) mass spectrometry, EPR spectroscopy, and X-ray crystallography (Supporting Information Page S5). The X-band EPR spectrum of [Fe(LCNC)2](PF6) in acetonitrile at 77 K shows a broad isotropic signal with a g-value of 2.07 (Figure S1), compatible with the low-spin d5 valence electron configuration that was previously observable for related iron(III) compounds.[41,78,107,121] Green-colored single crystals of [Fe(LCNC)2](PF6)·C2H4Cl2 suitable for an X-ray diffraction study were grown by slow vapor diffusion of pentane into a saturated 1,2-dichloroethane solution of [Fe(LCNC)2](PF6) at ambient temperature. Structure analysis revealed the near-perfect octahedral structure of the diamidotetracarbene iron(III) complex cation [Fe(LCNC)2]+ (Figure b). Analogously to [Co(LCNC)2]+, in [Fe(LCNC)2]+, the metal–Namido bond length (1.910(4) Å) is shorter than the metal–CNHC bond lengths (2.013(4) and 2.016(5) Å). Key bond lengths and angles are in the expectable range of values according to previously published iron(III) complex structures containing Fe–Namido or Fe–CNHC bonds.[78,107] Similar to [Co(LCNC)2]+, [Fe(LCNC)2]+ also exists as PP- and MM-enantiomers in the crystal structure.

Electrochemistry

Two reversible oxidation waves appear in the cyclic voltammogram of [Co(LCNC)2](PF6) recorded in acetonitrile with 0.1 M tetra-n-butylammonium hexafluorophosphate (Bu4N)(PF6) at 22 °C (Figure a). The first oxidation wave at 0.42 V versus Fc+/0 is attributed to the CoIV/III couple based on a comparison with a recently reported (less electron-rich) cobalt(III)–hexacarbene complex (Figure a), in which the CoIV/III redox couple appeared at 0.96 V versus Fc+/0 (entry 2 of Table ).[88] This shift of the CoIV/III potential to a substantially less positive value is attributed to the combined σ- and π-donor properties of the LCNC ligand, including the anionic nature of the carbazolate subunit,[112] and this is further supported below by additional experimental results and computational studies. The second oxidation wave at 0.72 V versus Fc+/0 is ascribed to ligand oxidation in the [Co(LCNC)2]+ complex, whereas the irreversible wave at −2.21 V versus Fc+/0 in Figure a is attributed to ligand reduction. Unlike in cobalt(III) complexes with less electron-rich ligands,[89,122,123] a CoIII/II redox wave is not detectable for [Co(LCNC)2](PF6) in the electrochemical window of acetonitrile.[124] Further experimental evidence that the first oxidation of [Co(LCNC)2]+ is a metal-centered CoIV/III process is given in the Supporting Information (Figure S18b).

Figure 3

Cyclic voltammograms of (a) 1 mM [Co(LCNC)2](PF6) and (b) 1 mM [Fe(LCNC)2](PF6) in deaerated acetonitrile at 22 °C. 0.1 M (Bu4N) (PF6) was used as the electrolyte, and the potential scan rate was 0.1 V/s in both cases.

Table 1

Electrochemical Potentials (E1/2 in V vs Fc+/0) of [Co(LCNC)2](PF6), [Fe(LCNC)2](PF6), and a Few Pertinent NHC Complexes of Cobalt(III) and Iron(III)

entry	compound	E1/2(MIV/III)a	E1/2(L•+/0)	E1/2(L0/•–)	E1/2(FeIII/II)
1	[Co(LCNC)2]+b	0.42	0.72	–2.21c	N/A
2	[Co(PhB(MeIm)3)2]+d	0.96	1.55		N/A
3	[Fe(LCNC)2]+b	0.05	0.56		–1.38
4	[Fe(ImP)2]+e	0.08	1.23		–1.16
5	[Fe(phtmeimb)2]+f	0.25	1.67c		–1.16
6	[Fe(btz)3]3+g		1.16c		–0.58

M = Co or Fe.

This work.

Irreversible wave.

PhB(MeIm)3 = tris(3-methylimidazolin-2-ylidene) (phenyl)borate (Figure a).[88]

Imp = deprotonated form of 1,1’-(1,3-phenylene)bis(3-methyl-1-imidazol-2-ylidene).[121]

Phtmeimb = phenyl[tris(3-methylimidazol-1-ylidene)]borate.[107]

Btz = 3,3′-dimethyl-1,1′-bis(p-tolyl)-4,4′-bis(1,2,3-triazol-5-ylidene).[41]

In the cyclic voltammogram of [Fe(LCNC)2](PF6) recorded under identical conditions (Figure b), the first oxidation wave appears at 0.05 V versus Fc+/0. This is in line with the potential reported for the FeIV/III redox couple in a related iron(III) complex (entry 4 of Table )[121] and consequently is attributed to a metal-centered oxidation event. The second oxidation wave appearing at 0.56 V versus Fc+/0 is attributed to ligand oxidation (entry 3 of Table ), which appears to be cathodically shifted by 0.16 V relative to the cobalt(III) complex. A metal-based oxidation process leading to iron(V) does not seem plausible in this case here.[125] On the reductive side, the FeIII/II redox couple is readily detectable at −1.38 V versus Fc+/0, in line with the potentials reported for two iron(III) complexes with related coordination environments (entries 4 & 5 in Table ).[107,121] The fact that the first reduction event in [Fe(LCNC)2](PF6) is metal-centered and leads to iron(II) then evidently precludes the detection of the same ligand-centered reduction event as for the isostructural cobalt(III) complex, at least in acetonitrile.

Turning our attention back to the [Co(LCNC)2](PF6) compound, we note that the first oxidation event is metal-based, whereas the first reduction process is ligand-centered. Based on the relevant CoIV/III and L0/•– potentials of 0.42 and −2.21 V versus Fc+/0 (entry 1 of Table ), respectively, one can expect an MLCT excited-state energy of approximately 2.6 eV. Furthermore, given a ligand-centered oxidation at 0.72 V versus Fc+/0, an intraligand charge transfer (ILCT) excited state can be anticipated at roughly 2.9 eV. The combined experimental and computational studies presented in the following will demonstrate that these two expectations regarding the lowest MLCT and ILCT states are largely fulfilled.

UV–Vis Absorption Spectroscopy

The (H3LCNC)(PF6)2 ligand precursor is optically transparent over large parts of the visible spectrum but features several π–π* absorption bands at wavelengths shorter than 450 nm (Figure a). The [Co(LCNC)2]+ complex (Figure b) exhibits a pronounced absorption band with a maximum at 430 nm and a molar extinction coefficient (ε430) of 14,000 M–1 cm–1, which is substantially greater than twice the ε430 value of the (H3LCNC) (PF6)2 ligand precursor (2 × 2200 M–1 cm–1). It follows that the longest-wavelength UV–vis absorption band of [Co(LCNC)2](PF6) is most likely due to a different type of electronic transition than in the ligand precursor. The onset of the respective band is near 450 nm, which corresponds to 2.75 eV and thus falls into the range, in which MLCT and ILCT transitions are expectable on the basis of the electrochemical investigations. Time-dependent density functional theory (TD-DFT) calculations support this interpretation and suggest that the absorption band at 430 nm is caused by a transition with substantial MLCT character (transition 3 in Table S3), whereas the shoulder at 408 nm (transition 9 in Table S3) is predominantly of ILCT type (vertical bars in Figure b, see Supporting Information Page S24 for details). The energy difference between these two absorption bands is 0.16 eV, which seems reasonably close to the value of 0.3 eV estimated on the basis of the redox potentials in Table .[126] The calculated difference electron density plot for the lowest-energy transition (Figure d) illustrates its substantial MLCT character, showing depletion of electron density (colored in purple) at the metal center and the amido N atoms, along with an increase of electron density (colored in light blue) at the NHC moieties (in particular the ligating C atoms) and the carbazole backbone. For fragment contributions and a more detailed discussion of the HOMO–LUMO transition, see Tables S4/S5 in the Supporting Information.

Figure 4

UV–vis absorption spectra of the (a) ligand precursors (H3LCNC) (PF6)2, (b) [Co(LCNC)2](PF6), and (c) [Fe(LCNC)2](PF6) in acetonitrile at 22 °C. Molar extinction coefficients at 430 nm (ε430 in units of M–1 cm–1) are indicated. The blue bars (Figure b) mark the energies of TD-DFT-calculated vertical transitions for the CoIII complex, and the heights of these bars reflect relative oscillator strengths. (d) Electron density difference plot for the lowest-energetic transition of [Co(LCNC)2](PF6) at 430 nm with a high oscillator strength of 0.224, based on TD-DFT calculations, with purple color marking a depletion and light-blue color indicating a gain of electron density.

[Fe(LCNC)2](PF6) shows multiple absorption bands in the blue–green and in the red spectral range (Figure c), similar to recently reported iron(III) and iron(II) compounds,[78,96,121] for which the concept of “HOMO inversion” was put forward.[127] Applied to the [Fe(LCNC)2]+ complex, this concept would predict that the filled amido N(2p) orbitals mix with the occupied metal 3d-orbitals, resulting in a HOMO of combined metal–ligand character at somewhat more elevated energy than the pure iron-based 3d orbitals in the absence of mixing. Such metal–ligand orbital mixing is indeed what we have found (Figure d) for the lowest-energy transition of [Co(LCNC)2](PF6). In [Fe(LCNC)2](PF6), there are additional absorption bands between 600 and 900 nm, which are likely due to LMCT transitions, expectable near 1.94 eV (639 nm) based on the electrochemical data in Table (E1/2(L•+/0) – E1/2(FeIII/II)).

Transient Absorption and Excited-State Dynamics

The combination of picosecond UV–vis transient absorption spectroscopy with spectro-electrochemical investigations further corroborates the substantial MLCT character of the lowest excited state of [Co(LCNC)2](PF6), as seen in the following. Upon excitation at 430 nm with a picosecond laser, the UV–vis transient absorption difference spectrum shown in Figure a is obtained, featuring the disappearance of the lowest-energy ground-state absorption band at 430 nm, along with an excited-state absorption (ESA) band peaking at 380 nm and a weaker ESA band near 460 nm. The overall appearance of this spectrum is similar to the transient absorption difference spectrum known from 3MLCT-excited ruthenium(II) polypyridine complexes, which includes essentially the same two ESA bands (in similar intensity ratio) in combination with the MLCT bleach.[128−130] Spectro-electrochemistry at a potential of 0.42 V versus Fc+/0 (inducing oxidation of cobalt(III) to cobalt(IV) according to Table ) yields the UV–vis difference spectrum presented in Figure b, for which the UV–vis absorption spectrum prior to applying any potential served as a baseline. The disappearance of the lowest-energy absorption band of [Co(LCNC)2]+ at 430 nm is evident, whereas the [Co(LCNC)2]2+ complex has prominent absorption bands near 380 nm and in the red spectral range (even beyond the accessible spectral region of the picosecond transient absorption experiment of Figure a). Reductive spectro-electrochemistry was performed in an analogous manner by applying a potential of −2.3 V versus Fc+/0, which leads to ligand reduction according to Table . The 430 nm band of [Co(LCNC)2]+ also disappears under these conditions (Figure c), whereas [Co(LCNC) (LCNC)•–] features prominent absorptions near 380 and 460 nm. The UV–vis transient absorption difference spectrum in Figure a is in reasonably good agreement with a 1:1 superposition of the spectro-electrochemical UV–vis difference spectra in Figure b/c, which monitor the effects of metal oxidation and ligand reduction. Such good agreement strongly supports the view that the lowest-energetic excited state of [Co(LCNC)2]+ has substantial MLCT character.[131]

Figure 5

(a) Picosecond transient absorption spectrum of 100 μM [Co(LCNC)2](PF6) in deaerated acetonitrile at 22 °C obtained after excitation at 430 nm, time-integrated over 2 ns. The inset in the upper right corner shows the decay of the ESA signal at 380 nm and the recovery of the ground-state absorption bleach at 430 nm (both decay traces normalized to a ΔOD value of 1 at time = 0). (b) UV–vis absorption changes following metal-based oxidation of [Co(LCNC)2](PF6) at 0.42 V versus Fc+/0 [= E1/2(CoIV/III)] in acetonitrile at 22 °C. (c) UV–vis absorption changes upon ligand-centered reduction of [Co(LCNC)2](PF6) at −2.3 V versus Fc+/0 (

Following pulsed excitation at 430 nm, the ESA signal at 380 nm disappears with the same time constant (1.3 ns in deaerated acetonitrile at 22 °C) as the ground-state bleach at 430 nm recovers (inset of Figure a). This observation suggests that the lowest excited state decays directly to the electronic ground state, or at least it does not lead to a noticeable population of any MC excited state. This is an important difference to the excited-state dynamics in the vast majority of previously investigated isoelectronic iron(II) polypyridine and NHC complexes, in which the lowest MLCT state relaxes nonradiatively via lower-lying (and longer-lived) MC states.[36,55−57,74−85,90−95] Consequently, whereas photochemical reactions typically occur from MC states in iron(II) polypyridines,[84−86,132] for [Co(LCNC)2]+, one can expect that the MLCT state is indeed the main photoreactive excited state. It seems plausible that the stronger ligand field experienced by the more highly charged cobalt(III) in our complex (relative to isoelectronic iron(II) in various coordination environments) plays a key role in leading to this favorable behavior.

Given a lifetime of 1.3 ns for the photoactive excited state of [Co(LCNC)2](PF6), radiative relaxation could in principle become a competitive decay pathway; yet our repeated attempts to detect unambiguous photoluminescence in solution at room temperature were unsuccessful. However, this MLCT state is unusually long-lived in comparison to iron(II) complexes, for which MLCT lifetimes in the picosecond time range are typical, with less than a handful of cases on (or approaching) the nanosecond timescale.[77,78,87] Owing to its comparatively long lifetime and its substantial MLCT character, the lowest excited state of [Co(LCNC)2](PF6) is able to undergo photoinduced electron transfer, as demonstrated further below.

Before turning to photochemistry, however, we recall that the electrochemical, UV–vis absorption, and TD-DFT studies presented above pointed to an ILCT state at roughly 0.16–0.30 eV higher energy than the MLCT state. The results obtained by femtosecond UV–vis transient absorption spectroscopy (as presented in the following) can indeed be interpreted in terms of internal conversion (IC) from an initially coexcited ILCT to the lower-lying MLCT state. Following excitation at 430 nm, global fit analysis to the experimental UV–vis transient difference data (Figure S21) result in the two decay-associated amplitude spectra (DAS) shown in Figure . DAS1 decays with a lifetime (τ1) of 96 ± 25 ps, whereas DAS2 has a lifetime (τ2) of 1.24 ± 0.10 ns. The appearance of DAS2 and its lifetime are in good agreement with the picosecond data in Figure a, and consequently, DAS2 is attributable to the MLCT state discussed above. DAS1 has a weaker bleach at 430 nm and a more pronounced ESA band at 445 nm than DAS2 (black arrows), extending to somewhat shorter wavelengths. In a UV–vis spectro-electrochemical experiment, in which one of the ligands of [Co(LCNC)2](PF6) is oxidized at a potential of 0.72 V versus Fc+/0 [E1/2(L•+/0) in Table ], an absorption band appears quite prominently in exactly this spectral region around 445 nm (Figure S19). Furthermore, ligand reduction at −2.3 V versus Fc+/0 [E1/2(L0/•–) in Table ] causes absorption bands around 460 nm which extend well beyond 600 nm (Figure S20). On this basis, it is tempting to attribute DAS1 to the ILCT state and further conclude that τ1 (96 ± 25 ps) reflects the kinetics for IC from the ILCT to the MLCT state. However, a timescale of roughly 100 ps seems slow in comparison to IC and other excited-state relaxation processes studied for example in nickel(II) complexes.[64,65,133,134]

Figure 6

Decay-associated amplitude spectra (DAS) obtained from femtosecond transient absorption experiments with 260 μM (1 mm cuvette was used) [Co(LCNC)2](PF6) in acetonitrile at 22 °C following excitation at 430 nm.

Nevertheless, the combined electrochemical, UV–vis (transient) absorption, and computational studies point to the orbital picture in Figure a and the single-configuration coordinate diagram in Figure b. The π* LUMO of [Co(LCNC)2](PF6) extends over the NHC moieties and the carbazole backbone of the ligands, whereas the HOMO has substantial metal character, labeled in Figure a simply with “t2g” (thereby neglecting the contribution of amido N and other atoms seen for the HOMO–LUMO transition in Figure d). The HOMO – 1 is largely based on the two amido N atoms, and consequently, an ILCT transition is energetically close to the MLCT-like transition. In the potential well picture of Figure b, the time constant for the supposed IC from the ILCT to the 1/3MLCT state is included along with the decay time of the lowest MLCT state. The population of 3MC (or even 5MC)[135,136] states remains undetectable for [Co(LCNC)2](PF6). If such MC states are involved in the excited-state relaxation pathway, for example, by thermal population from the 1/3MLCT manifold, then the lifetimes of the respective MC states would have to be considerably shorter than what is often observed in isoelectronic iron(II) polypyridines.

Figure 7

(a) Simplified frontier orbital picture highlighting two of the key electronic transitions occurring in [Co(LCNC)2](PF6). (b) Single configurational coordinate diagram with energetically low-lying charge-transfer and MC excited states, along with the time constants of two relaxation processes. g. s. = ground state.

Femtosecond UV–vis transient absorption studies of the [Fe(LCNC)2](PF6) reference compound recorded after excitation at 430 nm and at 700 nm (Figures S25–S28) provide evidence for the population of a 2LMCT state, in analogy to what has been recently reported for two related iron(III) compounds.[41,92,107] For [Fe(LCNC)2](PF6), the lifetime of that 2LMCT state in acetonitrile at 22 °C is only 1.3 ps, which is between 2 and 3 orders of magnitude shorter than in the two prior key studies.[41,107]

Photoinduced Electron Transfer

A key characteristic of MLCT excited states is their predisposition to undergo electron transfer reactions,[24,137] as nowadays widely exploited in photoredox catalysis[5−7,13,138] and in solar energy conversion.[8,9,139−144] To further corroborate the finding of a photoactive excited state with substantial MLCT character in [Co(LCNC)2](PF6), it therefore seemed useful to investigate to what extent this state can indeed trigger photoinduced electron transfer reactions. Toward this end, 30 μM [Co(LCNC)2](PF6) in acetonitrile was excited selectively at 430 nm in the presence of 150 mM methyl viologen hexafluorophosphate (MV(PF6)2). The resulting nanosecond UV–vis transient absorption difference spectrum in Figure a (green trace) shows the diagnostic spectral features of the methyl viologen radical monocation (MV•+) with maxima at 395 and 605 nm, along with a bleach around 430 nm, which coincides with the disappearing absorption band upon oxidation of cobalt(III) to cobalt(IV) in Figure b. Thus, the data in Figure a provides unambiguous evidence for photoinduced electron transfer from [CoIII(LCNC)2]+ to MV2+, resulting in [CoIV(LCNC)2]2+ and MV•+.

Figure 8

(a) UV–vis transient absorption difference spectra providing direct evidence for PET from [Co(LCNC)2]+ to MV2+. In the presence of 150 mM MV(PF6)2, 30 μM [Co(LCNC)2](PF6) and 110 μM ligand precursor (H3LCNC)(PF6)2 were excited at 430 nm (green and red traces, respectively) with laser pulses of ca. 10 ns duration, and the signals were time-integrated over 1.5 μs with a delay time of 100 ns. The solvent was deaerated acetonitrile at 22 °C. Blue trace: UV–vis difference absorption spectrum obtained upon electrochemical reduction of MV2+ to MV•+ at a potential of −0.7 V versus SCE in acetonitrile. Inset: kinetic decays at 395 and 430 nm, respectively, obtained from the transient absorption spectrum (green trace) for PET from [Co(LCNC)2]+ to MV2+, showing the disappearance of [CoIV(LCNC)2]2+ and MV•+. (b) Quantum yield (ϕ) determination for the formation of MV•+ by a relative actinometry experiment as described in the main text. The orange trace reflects the 3MLCT excited-state decay of [Ru(bpy)3]2+ upon excitation at 440 nm (monitored at 455 nm) and serves to quantify the number of absorbed photons. The green trace reflects the disappearance of MV•+, monitored at 395 nm after excitation of [CoIV(LCNC)2]+ at 440 nm, and serves to quantify the number of productive electron transfer events. Inset: UV–vis absorption spectra of [Co(LCNC)2](PF6)/150 mM MV(PF6)2 in deaerated acetonitrile (green trace) and [Ru(bpy)3]2+ in water (orange trace), adjusted to an identical optical density at 440 nm to ensure the absorption of an equal amount of photons by both solutions.

The formation of MV•+ was quantified by a relative actinometry experiment using [Ru(bpy)3]2+ as a reference. Specifically, an aqueous solution of [Ru(bpy)3]Cl2 and an acetonitrile solution of [Co(LCNC)2](PF6) with 150 mM of MV(PF6)2 were prepared such that both solutions have identical absorbance values at the excitation wavelength (440 nm, in this case, inset of Figure b). The two isoabsorptive solutions were then excited at 440 nm under strictly identical conditions. The [Ru(bpy)3]Cl2 solution served to quantify the number of photons absorbed at 440 nm, whereas the [Co(LCNC)2](PF6)/MV(PF6)2 solution was used to determine the amount of MV•+ electron transfer products formed. The formation of the 3MLCT excited state of [Ru(bpy)3]2+ occurs in a quantitative fashion and leads to a change in extinction coefficient Δε at 455 nm of −10,100 M–1 cm–1,[145] the well-known MLCT (ground state) bleach. By monitoring the change in optical density of the [Ru(bpy)3]Cl2 solution at 455 nm, one can therefore determine the number of photons absorbed at 440 nm, based on the (valid)[130] assumption that the 3MLCT state of [Ru(bpy)3]2+ is the dominant photoproduct following 1MLCT excitation. Since the [Ru(bpy)3]Cl2 and the [Co(LCNC)2](PF6)/MV(PF6)2 solutions both have the same optical density at 440 nm, both solutions will absorb the same number of photons under identical excitation conditions. The molar extinction coefficient ε at 395 nm for MV•+ is 41,800 M–1 cm–1;[146] hence, when monitoring the change in optical density at that wavelength, the amount of productive electron transfer events becomes quantifiable. The main plot of Figure b shows the decays of the relative concentrations of 3MLCT-excited [Ru(bpy)3]2+ and MV•+. Expectedly, the two photoproducts decay with very different kinetics (orange and green horizontal axes), but the key point is the relative signal intensity at time = 0, which is proportional to the relative concentrations of 3MLCT-excited [Ru(bpy)3]2+ and MV•+ immediately after excitation. The main finding is that the quantum yield for the formation of MV•+ is about 2% (dashed black line in Figure b).

In general, the quantum yields for photoinduced electron transfer (PET) can depend on many different factors, among which the competition between geminate recombination and cage-escape plays a crucially important role.[147] Ruthenium(II) polypyridine complexes typically feature cage-escape yields between 5 and 60%.[148−152] Excited-state electron transfers between organic electron donors (dimethylaniline, dimethyltoluidine, and tritolylamine) and an iron(III) photosensitizer, [Fe(phtmeimb)2]+, were recently studied in detail, and it was reported that the yields for the charge-separated states are in the range of 1–7% in polar solvents such as acetonitrile and dimethylformamide.[100,107] The 2% quantum yield for the formation of the charge-separated state in our case seems much in line with these earlier findings.

Whereas the initial PET step occurs within the duration of the 10 ns laser pulses, bimolecular reverse electron transfer from MV•+ to [CoIV(LCNC)2]2+ then occurs on a timescale of roughly 500 μs (inset of Figure a).

Photostability

Although low-spin 3d6 complexes along with 3d3 compounds are among the most substitution-inert first-row transition-metal complexes in the electronic ground state, the situation can be vastly different in some of their electronically excited states. For instance, carbonyl complexes of manganese(I) or chromium(0) (3d6) undergo rapid dissociation of CO ligands upon photoexcitation,[153−155] and similarly, CrIII (3d3) complexes are prone to ligand dissociation in some of their distorted MC excited states.[57,156] Even well-known MLCT emitters such as [Ru(bpy)3]2+ and related 4d6/5d6 luminophores undergo rather facile photodegradation,[129,157] typically as a result of the thermal population of MC states with metal–ligand dissociative character.[17,158−160] Against this background, it seemed interesting to test the inherent photostability of [Co(LCNC)2](PF6) under continuous photoirradiation.

For this purpose, deaerated acetonitrile solutions of [Co(LCNC)2](PF6), [Ru(bpy)3](PF6)2, and fac-[Ir(ppy)3] (ppy = 2-phenylpyridine) with equal optical densities at 447 nm were prepared, and these isoabsorptive solutions were then irradiated by a continuous-wave laser with an output power of 1.1 W at that wavelength. For the emissive [Ru(bpy)3](PF6)2 and fac-[Ir(ppy)3] compounds, the photoluminescence intensity was monitored directly as a function of irradiation time (red and blue lines in Figure ), whereas for the nonemissive [Co(LCNC)2](PF6) compound, UV–vis absorption spectra were recorded in time intervals of 10 min (green circles in Figure ). Under these conditions of very intense irradiation in a coordinating solvent, [Ru(bpy)3]2+ decomposes nearly completely within 10 min, whereas fac-[Ir(ppy)3] persists for 200 min before it reaches a similar level of decomposition. In a recent study, photodegradation quantum yields of 2.76·10–2% and 1.74·10–3% were found for these two compounds under very similar conditions.[129] [Co(LCNC)2]+ remains essentially intact over the same irradiation period and no significant changes in the UV–vis absorption spectra are detectable over time (Figure S29h), very much in contrast to [Ru(bpy)3]2+ and fac-[Ir(ppy)3] (Figure S29b/e).

Figure 9

Comparative photostability studies of [Co(LCNC)2](PF6), [Ru(bpy)3](PF6)2, and fac-[Ir(ppy)3] in deaerated acetonitrile at 22 °C. Solutions with identical optical densities at 447 nm were irradiated with a continuous-wave laser providing an output power of 1.1 W at that wavelength. Red and blue traces: normalized photoluminescence intensities of [Ru(bpy)3](PF6)2 at 620 nm and fac-[Ir(ppy)3] at 525 nm, respectively. Green circles: normalized absorbance of [Co(LCNC)2](PF6) at 430 nm; UV–vis spectra recorded in time intervals of 10 min. UV–vis absorption spectra of all compounds before and after irradiation of all three compounds are shown in the Supporting Information (Figure S29).

The astonishing photorobustness of [Co(LCNC)2](PF6) could have its origin in the fact that the rigid tridentate LCNC ligand provides a nearly geometry-optimized coordination environment (Figure a) and simultaneously ensures a particularly tight binding of the metal trication through two carbenes and one anionic amido donor atom. In this particular binding situation, the extent of metal–ligand bond covalence should be substantially higher than in [Ru(bpy)3]2+, as well as in many other d6 metal complexes, and it seems possible that this contributes to the photostability of [Co(LCNC)2](PF6). Another key contributor is certainly the comparatively short excited-state lifetime of [Co(LCNC)2]+ (1.2 ns) in comparison to [Ru(bpy)3]2+ (830 ns)[161] and fac-[Ir(ppy)3] (1.9 μs),[162] which limits the probability for deleterious side reactions from the photoactive excited state.[129] This principle plays a key role for OLED triplet emitters, for which one typically targets high radiative excited-state decay rates to maximize the emission output and to minimize photodecomposition.[48,50,56,163,164]

Recently, a tris(diisocyanide)molybdenum(0) complex already showed good long-term photostability under high-power red (laser) irradiation at 635 nm,[165] and thus, the photorobustness of [Co(LCNC)2](PF6) represents further evidence that d6 complexes of earth-abundant transition metals can in fact be remarkably photostable.

Conclusions

Photoactive MLCT excited states are currently intensively investigated for the entire 3d6 series from chromium(0)[166−168] to manganese(I)[168,169] and iron(II),[36,55−57,74−87,90−95,168,169] whereas cobalt(III) complexes seem underexplored in this regard.[36,37] The comparatively high oxidation state of cobalt(III) usually favors low-lying LMCT excited states[89,104] and further can induce sufficiently strong ligand fields for the lowest MC state to become photoactive.[88,98,99] Regarding the photophysics of cobalt(III) compounds, the two main conceptual novelties of [Co(LCNC)2](PF6) are (1) the fact that its energetically lowest-lying excited state has substantial MLCT character and (2) the observation that this excited state decays to the electronic ground state without a noticeable population of any MC state. These findings are in strong contrast to the vast majority of iron(II) complexes, in which lower-lying (and longer-lived) MC states typically deactivate the MLCT state very rapidly and efficiently.[36,55−57,74−87,90−95,168,169] Controlling such excited-state decay paths is a key challenge in the first-row transition-metal complexes with d6 or d8 configurations, for which obtaining long-lived MLCT states is inherently more difficult than for copper(I) compounds with their filled 3d[10] subshell.[44] If an MC state is involved in the excited-state relaxation pathway of [Co(LCNC)2]+ (e.g., by thermal population from the MLCT/ILCT manifold, as illustrated in Figure b), then the respective MC state would have to be very short-lived. In this case, the MC population could escape detection because the MC state decays faster than it is formed. This scenario could potentially explain the lack of photoluminescence in [Co(LCNC)2](PF6). In addition to its substantial MLCT character, the photoactive excited state of [Co(LCNC)2](PF6) has an intraligand contribution, making its overall composition qualitatively similar to that of the lowest excited state in many cyclometalated iridium(III) compounds.[4,12,13,20,170] Based on DFT calculations, the MLCT character of the photoactive excited state of [Co(LCNC)2]+ is limited (on the order of 10%). However, the cyclic voltammetry, spectro-electrochemical, and transient UV–vis absorption data point to much stronger metal involvement (Supporting Information Page S30). In particular, the comparison of picosecond transient absorption data with spectro-electrochemical data monitoring metal oxidation and ligand reduction (Figure ) are much in line with substantial MLCT character. Furthermore, femtosecond transient absorption data (Figure ) in comparison with spectro-electrochemical experiments (Figure S19) reveal noticeable differences between (predominant) MLCT and ILCT spectral signatures.

Except for its surprising photostability, the photophysical and photochemical behavior of [Co(LCNC)2](PF6) is largely the anticipated result of careful ligand design. In this context, our study provides some insights (or at least hints) that seem relevant far beyond cobalt(III) complexes and which could affect a much broader range of transition-metal compounds. Recent investigations of photoactive first-row transition-metal complexes set much emphasis on the importance of ligand field strength (usually captured by the ligand field parameter 10 Dq), for which the σ-donor and π-acceptor properties of the ligands were optimized.[36,57,58,77,81,110,111,171,172] However, in addition to the ligand field strength, the metal–ligand bond covalence determines the energy of MC states (typically quantified by the so-called Racah B parameter), yet the importance of covalence could have been underestimated so far in the design of new photoactive transition-metal compounds.[96,118,173] The cobalt(III)–amido N(2p) bond in [Co(LCNC)2](PF6) has substantial covalent character owing to the π-donor properties of the anionic carbazolate. This situation, in combination with the strong σ-donor properties of the four NHC units, is of key importance to obtain sufficient electron density at the metal center for a state with substantial MLCT character to become the lowest excited state. Furthermore, a high extent of metal–ligand bond covalence can contribute to a favorably high ratio between 10 Dq and B, leading to the scenario in which the MLCT state is not efficiently depopulated by lower-lying MC states as a result of their shift to very high energies. Our recent investigation of a chromium(III) complex with a mixed carbazolate-polypyridine ligand provided a remarkably low Racah B parameter,[118] which supports our speculation concerning the favorable 10 Dq/B ratio in [Co(LCNC)2](PF6).

Computational and experimental work by other investigators already emphasized the importance of π-donation,[78,96,123,127,174] which is however often intuitively associated with a weakening of the ligand field strength.[92] We speculate that for this reason, π-donor ligands did so far not receive the amount of attention they would perhaps deserve for the design of new photoactive transition-metal compounds. Furthermore, it seems that metal–ligand bond covalence might be a somewhat underappreciated design factor in comparison to the more frequently considered strategies of establishing strong ligand fields, structural rigidity, and coordination geometry optimization.[37]