Decoding Key Transient Inter-Catalyst Interactions in a Reductive Metallaphotoredox-Catalyzed Allylation Reaction.

Metallaphotoredox chemistry has recently witnessed a surge in interest within the field of synthetic organic chemistry through the use of abundant first-row transition metals combined with suitable photocatalysts. The intricate details arising from the combination of two (or more) catalytic components during the reaction and especially the inter-catalyst interactions remain poorly understood. As a representative example of a catalytic process featuring such intricacies, we here present a meticulous study of the mechanism of a cobalt-organophotoredox catalyzed allylation of aldehydes. Importantly, the commonly proposed elementary steps in reductive metallaphotoredox chemistry are more complex than previously assumed. After initial reductive quenching, a transient charge-transfer complex forms that interacts with both the transition-metal catalyst and the catalytic base. Surprisingly, the former interaction leads to deactivation due to induced charge recombination, while the latter promotes deprotonation of the electron donor, which is the crucial step to initiate productive catalysis but is often neglected. Due to the low efficiency of this latter process, the overall catalytic reaction is photon-limited and the cobalt catalyst remains in a dual resting state, awaiting photoinduced reduction. These new insights are of general importance to the synthetic community, as metallaphotoredox chemistry has become a powerful tool used in the formation of elusive compounds through carbon-carbon bond formations. Understanding the underlying aspects that determine the efficiency of such reactions provides a conceptually stronger reactivity paradigm to empower future approaches to synthetic challenges that rely on dual metallaphotoredox catalysis.

Introduction

Transition-metal catalysis has been paramount in the development of stereoselective carbon–carbon bond-formation reactions that are essential for the synthesis of natural products and novel pharmaceutical scaffolds. Historically, precious second- and third-row transition metals have been key catalyst components in such processes, since the reactivity and synthetic use of the more earth-abundant 3d transition metals have been much harder to control and implement. Within the era of “dual” catalytic metallaphotoredox catalysis,[1−3] the utilization of 3d transition metal-based catalysts has witnessed a remarkable revival and delivered new reactivity paradigms that are enabled through a delicate control over their available redox states, allowing productively for single-electron transfer (SET) pathways and radical-centered coupling reactions complementing the photoredox processes. In addition, the use of photoredox catalysis is beneficial as a way to avoid the preparation of organometallic precursors by generating the organometallic species in situ as an intermediate. This additionally amplifies the functionality in the precursors and opposes the utilization of stoichiometric reducing zerovalent metal additives (e.g., Zn or Mn), which are generally less tolerant toward substrate functionality.[4,5]

The addition of a second catalytic system (or multiple systems) complicates the mechanism of the reaction,[6−10] and therefore a simplified model of the photoredox catalysis cycle is usually presented (Figure a), in which electrons are directly available for SET to the transition-metal catalyst following reductive quenching. However, the nature of the interactions between the photoredox catalyst and the transition-metal catalyst is likely more complex.[6,10] Furthermore, photoredox catalytic reactions are often limited by the photon flux, and thus, the usual rate-limiting steps for the transition metal-driven cycle are unlikely to apply, which has important consequences for the overall kinetics of the reaction, and therefore, for process scale-up. In order for photocatalytic processes to become generic tools in synthetic organic chemistry, it is vital to gain a deeper understanding of how the catalytic systems are intertwined and uncover the mutual influence of the separate catalytic components in relation to the overall performance of the process. A detailed knowledge of such multicatalytic approaches offers a way to create new opportunities for various transformations that operate under similar mechanistic regimes, thereby advancing this chemistry beyond empirical process optimization.

Figure 1

(a) Commonly proposed single-electron transfer (SET) step for reductive quenching, where an electron is transferred immediately from the quencher (Q) to the photocatalyst (PC), and the products (Q•+ and PC•–) are considered solubilized. (b) Important transient inter-catalyst interactions determined in this work that determine the efficiency of the representative reaction. (c) Representative metallaphotoredox catalytic cross-electrophile coupling process studied in this work.

Interest in such processes was initially focused on the combination of nickel or copper catalysis with photoredox catalysis, but recently cobalt has emerged as a potent, alternative metal in dual catalytic procedures.[11−14] While mechanistic studies for cobalt catalysis have been thoroughly probed under reductive conditions using chemical reductants[15−19] or via electrochemistry,[20] it remains uncertain whether dual metal/photoredox approaches proceed through the occurrence of similar intermediates. Several features of Co catalysis under reductive conditions remain open for debate, including the intermediacy of either CoI or Co0-species[5,20,21] and in dual Co/photoredox catalysis, particularly, how the effectiveness of different catalyst components and additives can be rationally optimized.

Recently, our group and several others reported stereoselective cobalt-catalyzed allylation of aldehydes as a promising new way to synthesize highly functional compounds through C–C bond formation.[21−25] Such processes are mechanistically reminiscent to other important C–C bond-formation reactions in synthetic chemistry, for which metallaphotoredox catalysis recently demonstrated to be a powerful methodology.[12,13,26−31] Here, we describe a detailed mechanistic analysis of a stereoselective C–C bond-formation reaction that can be considered as a representative example for various types of reductive dual metallaphotoredox catalytic C–C bond-formation reactions, and we give special attention to the inter-catalyst interactions (Figure b). Specifically, we elucidated the full mechanism of the allylation of aldehydes involving photoredox, cobalt, and base catalysis (Figure c). Herein, we show that the initially formed charge-separated state formed after excited-state SET is a key intermediate that requires a catalytic base to produce turnover. The same intermediate reacts counter-productively with the cobalt catalyst through an energy-wasting charge recombination path (Figure b). Furthermore, we demonstrate that the cobalt cycle is fast compared to the number of photons absorbed per unit time and therefore the reaction is limited by the light intensity and, as a consequence, dictated by the inter-catalyst interactions.

Results and Discussion

Kinetics

We started our investigations by studying the kinetics of the reaction presented in Figure c to determine the order of reagents and the rate-limiting step. Under optimized turnover conditions as described previously,[24] using K3PO4 as a base, the disappearance of 1 was monitored by the evolution of its characteristic IR-band at 1815 cm–1 (Figure a). We first confirmed by NMR that the data obtained by IR corresponds to a 1:1 transformation of 1 to 3 without any observable intermediates, i.e., that conversion of 1 corresponds to the formation of 3 (Figure a gray data). A striking observation is that the reaction rate increases until 1 has been exhausted, after which it necessarily comes to a sudden halt. This indicates, first, that the reaction is likely zero order in the substrates, and second, that a species that speeds up the reaction is formed over time. Initially, we considered that a preactivation of the catalyst was responsible for the acceleration of the reaction. However, as discussed below, the starting cobalt complex, [Co(DPEPhos)Cl2], is part of the catalytic cycle and therefore this cannot be the cause. By changing the base from K3PO4 to 2,4,6-collidine, we observed a higher initial rate and the reaction acceleration was less evident (Figure b). When the byproduct of the reaction, HEPy, is added as an additive, an intermediate rate is observed, whereas no significant differences are observed using product 3. Therefore, the reaction follows an autocatalytic pathway, and its rate depends on the basic species and their concentration. Indeed, when plotting the quantum yield of the reaction as a function of the amount of HEPy produced (assuming a 1:1 production of HEPy from the consumption of 1), a linear increase in efficiency is observed until the reaction is exhausted (i.e., full consumption of 1), see Figure c. The rate seems to follow the same linear behavior if the reaction is started with additional HEPy present, which further confirms this explanation.

Figure 2

(a) Kinetic profile of the reaction when K3PO4 is used as a base. Conditions: 1 (0.1 mmol), 2 (0.15 mmol), 1,4-dihydro-2,6-dimethyl-3,5-pyridinedicarboxylic acid diethyl ester (HEH) (0.15 mmol), [CoII] (10 μmol), PC (2 μmol), K3PO4 (0.15 mmol), THF (1 mL), and blue light-emitting diode (LED) irradiation (λem = 445 nm, 0.7 A, 1.2 μeinstein s–1). Red dots: data obtained through in situ monitoring the C=O IR stretch frequency of 1 at 1815 cm–1. Black line: fit to the kinetic model. Gray circles and gray triangles: remaining 1 and produced 3 as observed by 1H NMR at different times with 1,3,5-mesitylene as internal standard. (b) Kinetic profile in the presence of different bases. The blue and purple triangles are in the presence of HEPy or 3, respectively, in addition to K3PO4. The green squares are obtained by substituting K3PO4 with 0.15 mmol 2,4,6-collidine. (c) Quantum yield derived (using a Savitzky–Golay filter) from the kinetic trace in (b) as a function of HEPy produced during the reaction to display the autocatalytic behavior (colors match those in (b)). (d) Initial quantum yield of the reaction for the two bases as a function of the concentration of cobalt catalyst [CoII], fit to the equation y = 1/(a + bx), a simple model for competitive kinetics from the intermediate 3[PC•–:HEH•+] (see the Supporting Information (SI)). (e) Kinetic profiles of the reaction with 2,4,6-collidine as base and different concentrations of [CoII]. Solid lines: fits to the kinetic model.

We then set out to determine the influence of the light intensity and the order of substrates 1 and 2, HEH, and [Co(DPEPhos)Cl2] ([CoII]). Decreasing or increasing the light intensity speeds up or slows down the reaction, indicating that the reaction is photon-limited (see the SI). The order in the substrates was further scrutinized by varying their concentrations. Neither diluting the mixture (effectively decreasing the concentration of all reagents) nor increasing the concentration of 1, 2 or increasing the scale leads to any notable difference in the kinetic trace (see the SI). Therefore, 0th order in all reagents is confirmed, caused by the photon-limited regime. Importantly, this means that the amount of material that can be converted per unit time is limited by the light intensity. The initial quantum yield of the reaction using K3PO4 as the base is 1.3%, which can be increased to 7% using collidine, making the reaction more efficient and allowing more facile scale-up of the reaction mixture.

When the concentration of [CoII] was varied, we did observe a change in the kinetics. Counterintuitively, increasing the concentration of [CoII] leads to a slower reaction and vice versa (Figure d,e). The effect is observed both when using K3PO4 or collidine as the base. We initially considered that quenching of the excited state by [CoII] might be responsible but quickly discarded this on the basis of Stern–Volmer analysis (see the SI), showing that [CoII] does not compete with HEH for quenching of the excited state under turnover conditions. Additionally, the cobalt complex does not absorb at the irradiation wavelength, so it does not compete with light absorption. Interestingly, the lower reaction rate that results from doubling the concentration of [CoII] can be counteracted by doubling the concentration of the base (see Figure e), suggesting that a competition between [CoII] and the base exists in an intermediary step of the reaction mechanism.

Initial Photoreactions

To study the competitive effect between the base and [CoII], we investigated how the reductive equivalents required for the cross-electrophile reaction are produced. Due to the use of thermally activated delayed fluorescence (TADF) photocatalyst 4CzIPN (PC), the excited-state system is complex, comprising an initial prompt fluorescent state that converts an initial singlet state to an equilibrium mixture of triplet and singlet (see the SI).[32−34] We studied the photocatalytic system using microsecond transient absorption spectroscopy (TAS). Our attention was initially focused on two specific wavelengths: 420 and 810 nm. At the former, a ground-state bleach is seen (see the SI), whereas, at the latter, a transient signal is observed (see the SI) that we assigned to the absorbance of the 3CT-state of the excited photocatalyst, as reported previously.[32] The decay-constants (τ0 = 3.89 ± 0.05 μs) of these signals match the decay of the TADF observed by time-correlated single-photon counting (see the SI). Interestingly, upon addition of HEH, the signal at 420 nm shows the formation of a transient species that subsequently vanishes (τ = 8.2 μs, see Figure a; red trace indicated by the arrow and SI). At 810 nm, we only observed quenching of the 3CT-state (τ = 0.99 ± 0.02 μs, see the SI). The transient species is assigned to a triplet charge-separated cage complex formed from the initial photoproducts: PC•– and HEH•+. Due to the charged nature of these species, they remain in close proximity as a cage complex, 3[PC•–:HEH•+], and eventually recombine to reform PC and HEH without any observable permanent charge transfer (i.e., cage escape). The same experiment performed in the presence of collidine instead shows an irreversible transfer of electrons from HEH to PC, without the recovery of the ground state (Figure b; red trace). It is clear that the base therefore prevents charge recombination by deprotonation of the 3[PC•–:HEH•+] intermediate. It is often assumed that the HEH•+-species is highly acidic and will lose a proton rapidly; here, we show that this is not generally the case, and hence the reason why photoredox catalytic reactions using HEH, or similar reductive quenchers, often require (sub)stoichiometric amounts of the base to provide productive process turnover.[12,14,35−38]

Figure 3

(a) Transient absorption spectroscopy kinetic traces at 420 nm. Conditions: PC (50 μM), HEH (200 μM) without (red) or with (green) [CoII] (170 μM) in THF (3 mL). (b) Like in panel (a) but in the additional presence of 2,4,6-collidine (50 mM). (c) Simplified molecular orbital (MO) diagram showing that charge recombination from the 3[PC•–:HEH•+] is spin forbidden and that in the presence of [CoII], charge recombination can occur to yield an excited-state 2[CoII] complex that subsequently relaxes to the ground-state 4[CoII]. Td = ligand field splitting in a tetrahedral environment. SOMO = Singly occupied molecular orbital. (d) Difference transient absorbance spectrum of 3[PC•–:HEH•+] measured at 5 μs delay. (e) Transients at 525 nm (at the absorption maximum of 3[PC•–:HEH•+]) in the absence or presence of a variety of metal complexes (170 μM).

To probe the influence of the acidity of oxidized electron donor on the overall reaction, various other electron donors with more acidic protons were tried (see the SI). Only the 4-phenyl-substituted analogue PhHEH[39] yielded 60% of product 3 after 17 h of irradiation, indicating that this electron donor performs worse than HEH. Indeed, even though the proton is more acidic (calculated pKa = −7.9 for PhHEH•+vs −6.3 for HEH•+), a base is required because the solvent is not sufficiently basic (calculated pKa = −15.5 for THF-H+, see the SI). We assign the slower reaction with PhHEH to mismatches in the redox potentials (see the SI).

The competition between the base and [CoII] observed during the kinetic studies led us to hypothesize that the charge-separated state 3[PC•–:HEH•+] could also react with [CoII]. Therefore, the traces at 420 nm of a mixture of PC and HEH were compared in the presence and absence of low concentrations of [CoII]. In the presence of the cobalt complex, the charge-separated state 3[PC•–:HEH•+] is no longer observed (Figure a, green data), indicating a fast reaction with this state. No significant change in the decay of the TA-signal at 810 nm is observed (see the SI), therefore, excluding the possibility of a (significant) reaction between the excited-state 3PC with the cobalt complex prior to the formation of 3[PC•–:HEH•+]. When collidine is also present in the sample, less permanent electron transfer is observed than in the absence of [CoII], showing that the reaction of 3[PC•–:HEH•+] with the cobalt complex is detrimental to the overall productivity of the reaction, in agreement with the observed reaction kinetics (Figure d,e).

We propose that the Co complex reacts with the charge-separated cage state to give quantitative charge recombination. The reason that this cage complex leads to fast charge recombination is that recombination can occur through a spin-allowed process to produce an excited-state cobalt(II) complex, removing the need for a direct spin-flip (Figure c). Density functional theory (DFT) calculations on [CoII] indeed show that the excited (doublet) structure is only 1.1 eV (26 kcal mol–1) in energy above the ground-state (quartet) structure (cf. the energy of the photons of 2.8 eV), which is further confirmed by the presence of absorption peaks in the NIR of the [CoII]-spectrum (Figure S19). Such low-lying metal excited states may limit the quantum efficiency in metallaphotoredox chemistry involving first-row transition metals by facilitating charge recombination, reminiscent of the ultrafast deactivation of metal-to-ligand charge-transfer (MLCT) excited states in first-row transition metal complexes compared to their second- and third-row counterparts.[40−42] Indeed, we assessed the generality of the process by probing the transient absorption signal at 525 nm (the wavelength maximum of 3[PC•–:HEH•+], see Figure d) in the presence of a variety of cobalt complexes and a nickel complex widely employed in metallaphotoredox catalysis (Figure e; with more example complexes in the SI). All but one of the 6 complexes reacted similarly with 3[PC•–:HEH•+] with varying rates, indicating the general nature of this energy-wasting process.

Reduction of Cobalt

Having these results in hand, the reduction of [CoII] was studied in detail. In recent years, several authors have suggested that the reaction follows a CoII–CoI–CoIII–CoII cycle,[21−23] but compelling proof using in situ spectroscopy remains elusive. A cyclic voltammogram (CV) of the complex shows a reduction peak at Ep = −1.3 V vs Fc+|Fc (Figure a; left), which can be readily reduced by PC•– (E1/2 = −1.63 V vs Fc+|Fc). The reduction is electrochemically irreversible, but the initial [CoII]-complex can be regenerated by applying more positive potentials (reoxidation peak at Ep = −0.75 V). Upon reduction to [CoI], the complex likely rapidly loses a Cl– ligand (at least up to 0.5 V s–1, no direct reoxidation peak is observed) that is then replaced either by the solvent,[43,44] or the complex dimerizes to [Co(DPEPhos)(μ-Cl)]2.[18,44] Spectroelectrochemistry confirmed that the process is chemically reversible, as the initial characteristic spectrum of [Co(DPEPhos)Cl2] can be recovered after a full CV (see the SI). After the reduction wave, the ultraviolet–visible (UV–vis) spectrum shows an altered structure in the 550–800 nm range in addition to an intense band appearing at 390 nm with clear isosbestic points at 550, 660, and 715 nm. The changed spectrum was therefore assigned to a [CoI]-complex.[44] Interestingly, if the CV is measured in the presence of 1, the reduction wave becomes completely irreversible at slow scan rates, indicating that [CoI] is responsible for the oxidative addition of 1 to form a [CoIII-allyl] species (Figure a; right), confirming previous hypotheses.[22,23] At faster scan rates, the reoxidation peak reappears, allowing the determination of the rate constant of oxidative addition, koa = 1.0 M–1 s–1 (see the SI). The characteristic UV–vis spectrum of the species after oxidative addition was also obtained by spectroelectrochemistry, which looks very similar to that of [CoI]. Importantly, a rate constant of this order of magnitude means that oxidative addition is not expected to limit the reaction at the initial stage when compared to a lower overall rate limited by the photon flux.

Figure 4

(a) Cyclic voltammograms of [CoII] (5 mM) in the absence (left) and presence (right) of 1 (50 mM). (b) Evolution of the UV–vis spectrum under turnover conditions: 1 (0.1 mmol), 2 (0.15 mmol), HEH (0.15 mmol), [CoII] (10 μmol), PC (2 μmol), collidine (0.15 mmol), THF (3 mL), blue LED irradiation (λem = 445 nm), and residual after a multivariate curve resolution fit using three spectral species. (c) Spectral species obtained from multivariate curve fit of the UV–vis data. As a reference, the spectrum obtained by spectroelectrochemistry in the presence of 1 is plotted as a dashed gray line. (d) Concentration profile of the spectral species in (c) (colors matching). The simulated cobalt species ([CoII] and a mixture of [CoI] and [CoIII-allyl]) concentrations from the kinetic model are indicated as a dashed gray line (see the SI for details). The arrow indicates the moment the spectra become noisy due to the precipitation of black metallic cobalt, coinciding with the moment the experiment and the model start to differ significantly due to exhaustion of 1.

Using in situ UV–vis spectroscopy, we then monitored the spectral changes occurring under irradiation. When a mixture of [CoII], PC, HEH, and collidine was irradiated, a fast conversion to a species that resembles the spectrum obtained for [CoI] by spectroelectrochemistry (see the SI) was observed. After this species is formed, further reduction occurs that leads to the precipitation of a black magnetic solid that we assigned to metallic cobalt. In the absence of collidine, no significant changes in the UV–vis spectrum are observed, confirming the results obtained by TAS, showing the requirement and key role of the base in the initial processes leading to the reduction of [CoII] to [CoI]. Next, the spectrum was monitored under turnover conditions (in the presence of 1 and 2 and collidine, see Figure b). Multivariate curve resolution of the spectra shows that a photostationary state is rapidly formed, consisting of mainly two species: [CoII] and an intermediate species, that matches the [CoIII-allyl] spectrum obtained from spectroelectrochemistry in the presence of 1 (cf. the blue to the dashed gray spectrum in Figure c), although small contributions of [CoI] are likely (see the SI), as the spectra of these two species might not be distinguishable. The photostationary state evolves with the intermediate species becoming more pronounced until substrate 1 is exhausted (Figure d), after which precipitation of metallic cobalt occurs (indicated by the arrow in Figure d). A third species (green curve), which we assign to [Co0], becomes significant only when the reaction is complete. These observations again indicate that the catalytic cycle is fast and therefore limited by the production of reductive equivalents through photoinduced electron transfer and thus by the light intensity. Indeed, after a further reduction of [CoIII-allyl] to [CoII-allyl], the formation of the product and the regeneration of [CoII] are fast (vide infra), and as such, the cobalt catalyst resting state is a mixture of mainly two species, viz. [CoIII-allyl] and the starting complex [CoII], which both await further photoinduced reduction. As substrate 1 is nearing exhaustion, the rate of oxidative addition decreases, leading to an increase in the concentration of [CoI]. Satisfyingly, the fitted kinetic model (vide infra and SI), shown by the dashed gray curves in Figure d captures the evolution of cobalt species during the reaction remarkably well when the intermediate species is assumed to be the sum of [CoIII-allyl] and [CoI], whose spectra are hard to distinguish (see the SI).

Diastereoselectivity

Density functional theory (B3LYP-D3 in THF solvent, see the SI for full details) was employed to elucidate the final steps of the mechanism starting from [CoII-allyl] B and explain the formation of the 1,3-diol 3 in a high 95:5 syn/anti ratio. Three different key steps beyond the [CoII-allyl] intermediate B were considered: isomerization of the allylic moiety (i.e., B to B and vice versa, going through tertiary-allyl species A, see the SI), C–C bond formation, and a protodemetalation stage. For the C–C bond formation, we optimized the pathway through well-known Zimmerman–Traxler-type transition states (TSs). In principle, four different stereoisomeric pathways can exist upon coordination of substrate 2. The allyl species could have either an E or Z configuration, and substrate 2 could be positioned pseudo-equatorial of pseudo-axial. However, due to the steric nature of the DPEPhos ligand, we excluded the possibility of a pseudo-axial positioning of substrate 2 due to unfavorable steric 1,3-diaxial interactions.[45] Therefore, only the pathways with pseudo-equatorial aldehyde coordination and using either E or Z isomers of the [CoII-allyl] B were considered. In addition, we considered a pathway where the pendent alcohol function was left deprotonated, but such structures lead to unfeasibly high energies for intermediates B and C and were therefore excluded (see the SI). The overall barriers for C–C bond formation are low (15.9 and 17.8 kcal mol–1 for forming the syn and anti-product, respectively), in agreement with a fast reaction that was observed experimentally. The protodemetalation step for the anti-isomer (see the SI) presents a slightly higher barrier than for the preceding C–C bond formation step, but this seems to be a minor computational artifact, as only minimal changes in the diastereoisomer ratio (dr) were observed experimentally when comparing different bases (see the SI for discussion).[24]

On the basis of the energies of the path (Figure and SI), we therefore conclude that the stereochemistry is determined by differences in the steric requirements of the E and Z-allylic isomeric forms of the Zimmerman–Traxler TS. Isomerization of the allyl is comparable in energy to C–C bond formation, allowing interconversion of the E to the Z allyl species to give a higher amount of the syn product through a Curtin–Hammett scenario, in line with other reports.[46] The diastereoselectivity predicted from microkinetic modeling (96:4 dr, see the SI) aligns well with the experimental value (95:5 dr), supporting the view that the Zimmerman–Traxler TS determines the stereoselective outcome of the transformation. It should be noted that the energies of the three steps are close enough that small alterations to the system might lead to a different stereodetermining step, and therefore a large difference in dr. Indeed, the ligand, substrate, and the base have all been shown to influence the dr in varying degrees.[24] Overall, this means that although the diastereoselectivity is controlled by the cobalt cycle, the overall rate of the reaction is not.

Figure 5

Full elucidated mechanism of the photoredox cycle (top) and cobalt cycle (bottom) with DFT energies (kcal mol–1) in gray for selected steps. Techniques used to probe various steps in the mechanism are indicated. Only the pathway from the Z isomer toward the major product (syn) is shown; a corresponding higher energy pathway exists for the E-allyl isomer.

Kinetic Modeling

Combining all observations and measured rate constants, an overall mechanism and kinetic model are proposed (Figure with detailed information in the SI). In addition to the known photophysical/photochemical reactions, an additional charge-separated state (1[PC•–:HEH•+]) is introduced that is formed by reductive quenching of the singlet state, which is a prevalent pathway under turnover conditions (88% of the excitations are quenched in the initial singlet state by HEH, see the SI). This species can undergo fast intersystem crossing (ISC) to form 3[PC•–:HEH•+] or decay through charge recombination. Because the interconversion of 1[PC•–:HEH•+] and 3[PC•–:HEH•+] was not observed by TAS, these reactions are expected to be much faster than the subsequent reactions from the 3[PC•–:HEH•+]-complex, and therefore the ratio of 3[PC•–:HEH•+] and 1[PC•–:HEH•+] are in pre-equilibrium. We therefore used the yield of intersystem crossing from the singlet state (i.e., the part of 1[PC•–:HEH•+] reacting through ISC vs charge recombination) as a fitting constant. Other unknown parameters were the rates of deprotonation by the three bases (i.e., K3PO4, HEPy, or collidine), as well as the rate of deactivation of 3[PC•–:HEH•+] with [Co(DPEPhos)Cl2]. All other unknown rate constants did not alter the overall process as long as they were sufficiently high, which can be reasonably assumed (see the SI). A global fit was performed on the in situ IR kinetic data, where all rate constants were shared as parameters. The excellent fit to the experimental data (see Figures a,e, 4d and SI) shows that the model can be validated as it well captures the important parts of the reaction path, predicts the intrinsic bottlenecks, and correctly simulates the [Co]-intermediates in the UV–vis data that was not part of the fitting dataset. Importantly, the interactions in the initial charge-separated state with either base or transition-metal catalyst are key for the efficiency of the reaction. Indeed, many of the absorbed photons are “wasted” due to recombination processes from the 3[PC•–:HEH•+]-intermediate. Such interactions are likely to be prevalent in many other metallaphotoredox catalysis processes, as similar reactivity was observed using various metal complexes with TAS and consequently would lead to unproductive pathways, especially when low-lying excited states are available on the transition-metal catalyst.

Conclusions

In summary, key inter-catalyst reactions determine the efficiency of this representative metallaphotoredox-catalyzed C–C bond-formation reaction. The base deprotonates the photochemically generated transient charge-separated cage complex, preventing rapid charge recombination that occurs in the absence of the base and thus enabling further electron transfer. The byproduct of the reaction, HEPy, is itself a base, and therefore the reaction follows an autocatalytic kinetic profile. We further show that the cobalt complex reacts competitively with the transient charge-separated state, leading to catalyst-induced charge recombination and therefore an unintuitive negative order in [CoII]. Such counterproductive interactions are expected to be generally present in reductive metallaphotoredox catalytic reactions employing 3d metals as charge recombination in cage complexes might be facilitated by low-lying excited states of the transition-metal complex. These intricacies occurring within the SET step can easily inhibit reactions and therefore should be considered when metallaphotoredox-promoted reactions are investigated. Elucidation of the underlying mechanistic complexity of multi-catalyst systems will help to further advance the field of metallaphotoredox chemistry as a powerful tool to uncover new transformations in synthetic chemistry.