Titanium-based catalytic transformations¹ are highly intriguing for a variety of reasons. Firstly, Ti belongs to the ten most abundant elements of the earth's crust, thus, rendering these processes more sustainable and more cost-effective as compared to those relying on precious metals. In addition, Ti itself is non-toxic and hence, catalyst and by-products can be expected to be environmentally benign.² Finally, Ti-catalysis has matured into a versatile tool for synthesizing fine chemicals by making use of two complementary reactivity patterns; namely, hydro-functionalization, requiring the metal only at its highest oxidation state, +IV, and redox transformations, which toggle the metal between Ti^IV and its low-valent forms, Ti^II or Ti^III.^1a-1d

Ti-catalysis in SETs via the Ti^IV/Ti^III redox couple is particularly compelling because it offers facile chemoselective access to radical intermediates via inner-sphere electron transfer (ET) to a metal-bound substrate.³ A superb Ti^III catalyst is the coordinatively unsaturated species Ti^III(Cp)₂Cl, originally introduced by Nugent and RajanBabu.⁴ Owing to its remarkable functional group tolerance,² it has emerged as an ideal green catalyst for engaging a wide range of substrates such as epoxides, enones, nitriles, alkyl halides, and others.^{1a-1d, 2, 3, 5} It is prepared in situ from the higher valent dichlorido precursor, Ti^IV(Cp)₂(Cl)₂, using a metal reductant⁶ and coexists in solution with its homodimer.⁷ Enantiomerically pure titanocene precursors even opened up the avenue to enantioselective transition-metal catalyzed radical reactions, which can then be utilized to reductively de-symmetrize meso-epoxides⁸ and to facilitate regiodivergent ring-openings of racemic or enantiomerically pure epoxides.⁹ Finally, the outstanding utility of Ti^III(Cp)₂Cl as both, a stoichiometric reagent and a catalyst, is showcased by numerous applications in natural product synthesis.¹⁰

The sustainability of Ti-based SET-catalysis can be improved further by avoiding the stoichiometric metal reductant. To this end, Ti-catalyzed reactions were recently coupled to PR-catalysis.¹¹ Apart from replacing the metal with photons, this strategy allowed also for overcoming the limited oxidizing power of the Ti^IV radical intermediate, which may impede a desired reductive elimination.¹² However, as is common in the field, bipyridine complexes of Ru and Ir were used as PR-co-catalysts and thus, sustainability was gained only at the expense of utilizing a precious metal in the photocycle.

Here, we elucidate the photochemical principles of PR-catalysis based exclusively on the Ti^IV/Ti^III redox couple, i.e. without any PR-co-catalyst, as we proposed earlier.¹³ Specifically, we reveal here the elementary dynamics following optical excitation of a titanocene PR-catalyst and in particular, the critical events of entry into the catalytic cycle that transforms a targeted substrate. Apart from time-resolved fluorescence spectroscopy, we deploy UV/MIR-spectroscopy on femtosecond-to-microsecond time scales and take full advantage of this technique's unique sensitivity to changes of the molecular and electronic structures of dynamically evolving systems featuring transition metals¹⁴ and/or systems implicated in organocatalytic cycles.¹⁵

UV-MIR-spectroscopy requires the PR-catalyst to feature distinctive vibrations that are IR-active in a spectral region accessible with state-of-the-art ultrafast laser sources. Unfortunately, this is not the case for Ti^IV(Cp)₂(Cl)₂, whose Cl−Ti−Cl stretching and Cp-ring modes are located well below 1000 cm⁻¹. We therefore investigated instead the IR-amenable model, Ti^IV(Cp)₂(NCS)₂ (1), in which the chlorido ligands are substituted by isothiocyanato ligands. Their stretching motions give rise to two characteristic IR-bands in the Fourier-transform infrared (FTIR) spectrum (cf. Figure 1a).¹⁶ The resonance at 2022 cm⁻¹ results from the out-of-phase combination (ν_op) of the two N=C=S-oscillators whereas the band at 2059 cm⁻¹ is due to their in-phase combination (ν_ip).

The electronic absorption spectrum of 1 in acetonitrile (MeCN) solution at 295 K is displayed in Figure 1b (blue curve). It consists of three bands peaking at 435 nm, 349 nm, and 288 nm, and upon expanding the spectrum vertically (dashed blue), a distinct shoulder at longer wavelengths is noted, which can be fitted very well with a Gaussian profile (pale blue) centered at 549 nm. Compound 1 has a d⁰-configured Ti-center and thus, transitions in the UV-to-visible (UV/Vis) region have nearly pure ligand-to-metal charge-transfer (LMCT) character.

According to time-dependent (TD) density functional theory (DFT, cf. Supporting Information, SI), the shoulder is due to the S₀→S₁ transition whereas the 435 nm-band originates from the S₀ → S₃ transition. Importantly, TD-DFT also confirms that the electronic structure of 1 is fully analogous to that of Ti^IV(Cp)₂(Cl)₂ (see SI). Finally, we also tested the catalytic activities of (−)-menthyl-substituted¹⁷ Ti^IV(Cp)₂(L)₂ with L=Cl and NCS (cf. Figure 2) in the atom-economical arylation of epoxide-derived radicals and found for irradiation with 453 nm-light that switching from chloride to pseudohalide even enhances the conversion efficiency slightly. Thus, complex 1 is a very sensible IR-model system.

Our model was reported to be non-emissive in room temperature solutions but luminescence attributed to the triplet ground state, T₁, could be collected from cryomatrices at 77 K.¹⁸ To understand this behavior, we initially examined the luminescence of 1 in MeCN at 287 K with time-correlated single-photon counting (TCSPC, see SI) and excitation at 450 nm. We were able to record emission traces at selected detection wavelengths, , between 540 nm and 750 nm, from which the stationary emission spectrum was then reconstructed as shown in Figure 1b (open circles, see SI for data processing). Again, the data can be fitted very well with a Gaussian profile (shaded pale red) peaking at 653 nm, in accordance with the cryomatrix data.¹⁸ The Stokes-shift amounts to ≈2900 cm⁻¹ and is thus much smaller than previously thought.¹⁸ A TCSPC-trace recorded at 650 nm is shown in Figure 3a (red dots only, label 0) together with a least squares fit to a biexponential decay. Time constants of (157±20) ps and (10.1±0.1) ns with relative amplitudes of 0.956±0.005 and 0.044±0.005 were found for the fast and slow components, respectively. We attribute the biexponential kinetics to thermally activated delayed fluorescence (TADF) as sketched in the Jablonski diagram in Figure 3b.

The 450 nm-excitation accesses S₃, which relaxes within the time resolution to the S₁-state. Emitted photons are then collected from S₁ and the prompt sub-nanosecond decay of the TCSPC trace results primarily from its depopulation due to intersystem crossing (ISC) to T₁ and internal conversion (IC) to the ground state, S₀. The slow decay on the 10 ns-time scale is caused mostly by the thermally-induced (uphill) intersystem re-crossing (ISrC) from T₁ to S₁. As shown in the SI, the time constants of the fast and slow decays together with their relative amplitudes can be used to determine the rate constants for ISC and ISrC, k_f and k_b, respectively, as well the IC-rate, k_IC. We find 1/k_f=(223±40) ps, 1/k_b=(2.64±0.26) ns, and 1/k_IC=(593±20) ps, which allows us to extract a value for the triplet-singlet gap according to ΔE_S1→T1=E(T₁)−E(S₁)=−RT⋅ln(k_f/3k_b)=−(3.3±0.2) kJ/mol.

Most importantly, the T₁-state appears as a suitable entry into the redox-catalytic cycle via reductive quenching with an electron donor (Figure 2). To confirm this SET, the TCSPC-experiment was repeated in the presence of various equivalents of triphenylamine and results are shown in Figure 3a. Clearly, increasing the quencher concentration, [Q], accelerates the luminescence tail while at the same time, leaving the fast decay unaffected. This observation can be taken as indirect evidence (cf. SI) that T₁ reacts indeed with NPh₃ presumably according to

()

where the rate constant, k_q, quantifies the bimolecular kinetics. A full kinetic analysis of all processes sketched in Figure 3b including the quenching reaction is given in the SI. A Stern–Volmer plot of the slow luminescence decay rate (cf. Figure 3b) yields a value for k_q of (9.4±0.2) ⋅ 10⁹ M⁻¹ s⁻¹, which is only about a factor of two smaller than an estimate of the diffusion-limited on-contact ET-rate (cf. SI).

However, the quenching reaction (1) still needs to be verified by directly monitoring the T₁- population and the putative radical anion product, Ti^III(Cp)₂(NCS)₂⋅⁻ (2), must be identified. To this end, we carried out next UV/MIR-spectroscopy with 450 nm-pump pulses. Representative results are shown in Figure 4a and the full spectro-temporal evolution from 200 fs all the way up the 300 μs is presented in the SI as an animation.

At early delays (Figure 4a), two negative bands are detected that spectrally coincide with the FTIR-resonances (gray). These features (aka ground-state bleach, GSB) result from the pump-induced depletion of S₀. Two positive bands (termed induced absorption, IA) initially peaking at 1997 cm⁻¹ and 2042 cm⁻¹ are also observed, which are fingerprints of an electronically excited state. As the time delay is increased to about 300 ps, the 1997 cm⁻¹-band gradually decays while a new IA-band at 2007 cm⁻¹ appears. This evolution partially suppresses the GSB at 2022 cm⁻¹ but leaves the GSB at 2059 cm⁻¹ unaffected. Furthermore, two isosbestic points (cf. asterisks) at 2004 cm⁻¹ and 2034 cm⁻¹ are formed, which indicates that a simple kinetic two-state interconversion occurs within the first 300 ps after excitation.

From ca. 0.5 ns onwards (cf. Figure 4b and SI), the UV/MIR-spectrum vanishes completely while retaining its overall shape. During this period, three new isosbestic wavenumbers emerge; namely, 2015 cm⁻¹, 2036 cm⁻¹, and 2050 cm⁻¹. These findings are indicative of yet another two-state interconversion, however, this time involving the return to S₀ as evidenced by a simultaneous recovery of the two GSB-bands.

To test for compliance of the UV/MIR-data with the coupled rate equations derived from the Jablonski diagram (Figure 3b) a global target analysis was carried out.¹⁹ As shown in the SI, the global fit involving only the S₀, S₁, and T₁-states is virtually indistinguishable from the data thereby lending ultimate credence to our kinetic model. The target fit yields the MIR-spectra of 1 in the S₁ and T₁-states (cf. Figure 4c). Qualitatively, these are very similar to the stationary FTIR-spectrum except for a distinct frequency-downshift, which can be rationalized by a diminished electron density on the ligands upon LMCT and a concomitant bond softening in the NCS-moieties. Importantly, the successive downshift in the order S₀-T₁-S₁ is reproduced by TD-DFT ( cf. Figure 4d).

The global target analysis yields also the rate constants, k_f=1/[(138±10) ps], k_b=1/[(1.67±0.05) ns], and k_IC=1/[(284±20) ps] and hence, the time-dependent occupations of the three states (cf. Figure 4e). Just like the TCSPC-traces, the S₁-population decays bi-exponentially thus, fully corroborating our assignment of the luminescence to TADF. The forward and backward rates yield ΔE_S1→T1=−(3.6±0.2) kJ/mol, fully consistent with the gap obtained from TCSPC. However, the rates are larger than those obtained from time-resolved fluorescence, which can be attributed to the higher temperature (295 K vs 287 K) used to collect the UV/MIR-data.

The target analysis renders the pump-probe study directly sensitive to the occupation of T₁; quite in contrast to TCSPC. In MeCN at 295 K, the T₁-population builds up with a time constant of ≈90 ps and subsequently decays with a lifetime of 5.3 ns (note that the T₁-lifetime is a complex function of k_f, k_b, and k_IC; cf. SI). This is the time window, which can be exploited for reductive quenching with an amine. Having confirmed full consistency of the UV/MIR-data with the TCSPC-results and the Jablonski diagram of Figure 3b, we move on to identify the Ti^III-product of the SET quenching reaction. To this end, the UV/MIR-experiment was repeated in the presence of 10 equivalents of NPh₃ and representative results are depicted in Figure 4f, g (cf. SI for an animation).

During the first 300 ps, the MIR-response is very similar to that of pure 1, which demonstrates that the quencher does not perturb the initial S₁-T₁ equilibration significantly. However, on longer time scales, the amine profoundly affects the UV/MIR-response. Whereas in the absence of the quencher, all GSB and IA bands decay concertedly to zero within 10 ns as a result of the non-radiative return to S₀, this is no longer the case when NPh₃ is present. Although the IA-bands at 2005 cm⁻¹ and 2045 cm⁻¹ still disappear, a long-lasting GSB is now noticed. This proves that the T₁-state decays without replenishing S₀ via S₁. Instead, a new distinctly split IA-band peaking at 2080 cm⁻¹ and 2091 cm⁻¹ gradually appears within a few nanoseconds. Clearly, this feature must be attributed to the product of the quenching reaction, presumably the radical anion, Ti^III(Cp)₂(NCS)₂⋅⁻ (2). A target analysis was carried out again, but now taking into account the bimolecular quenching reaction with the rate, k_q⋅[Q]. The global fit is once again superb (cf. SI) and provides us with the MIR-spectrum of the product (cf. Figure 4c, green) as well as the populations of all species under quenching conditions (cf. Figure 4h). On much longer time scales (≈100 μs, data shown in the SI), the split product absorption finally vanishes while simultaneously, the two GSB-bands fully recover, thus, demonstrating that in the absence of an organic substrate (e.g. an epoxide), the putative radical anion decays by back-ET to replenish the parent's ground state.

To test our tentative assignment we computed the IR-spectrum of 2 as shown in Figure 4d (green). DFT predicts that the radical anion's absorption is indeed frequency-upshifted relative to that of the neutral and that it is characteristically split into two bands of equal intensity, again originating from the in-phase and out-of-phase combinations of the two local NCS-modes. The superb agreement between experiment and theory constitutes the final and unambiguous evidence that the product of the quenching reaction with NPh₃ is indeed Ti^III(Cp)₂(NCS)₂⋅⁻.

Unfortunately, TD-DFT falls short in predicting the triplet-singlet gap. Depending upon the exchange-correlation functional, values for ΔE_S1→T1 between −14 kJ/mol and −22 kJ/mol were obtained; i.e. roughly a factor of four larger than the experimental value, and sadly, even using coupled-cluster theory including spin-orbit coupling does not lead to a significant improvement (see SI).

At any rate, our results clearly emphasize the importance of a long T₁-lifetime for an efficient entry into the redox catalytic cycle utilizing titanocene di(pseudo)halides. The lifetime in turn depends critically on the magnitude of ΔE_S1→T1 in relation to the temperature (see SI). Indeed, for a given ΔE_S1→T1, a decrease of temperature should lower the ISrC rate, k_b, by virtue of detailed balance. According to the TCSPC-data, the T₁-lifetime in MeCN is 3.2 ns at room temperature whereas it shortens to 2.8 ns by raising T by only 30 K. Yet, this small perturbation impedes the catalytic efficiency measurably as evidenced by a smaller overall chemical conversion for a given reaction time (see Figure 2). Working as close as possible to the solvent's freezing point might appear favorable; however, the consecutive steps within the redox cycle (e.g. substrate binding and H-atom transfer) may then become too slow. Electronic structure theory is required here to develop general ligand design principles for (i) optimizing the T₁-S₁ gap and (ii) enhancing the extinction coefficient for Vis-absorption without (iii) sacrificing the near-diffusion-limited character of the reductive ET quenching reaction ultimately leading to the Ti^III intermediate. Unfortunately, sufficiently accurate predictions of the spin-state energetics of transition-metal complexes remains still a formidable challenge for DFT and TD-DFT.