Formation and degradation of strongly reducing cyanoarene-based radical anions towards efficient radical anion-mediated photoredox catalysis

Cyanoarene-based photocatalysts (PCs) have attracted significant interest owing to their superior catalytic performance for radical anion mediated photoredox catalysis. However, the factors affecting the formation and degradation of cyanoarene-based PC radical anion (PC•‒) are still insufficiently understood. Herein, we therefore investigate the formation and degradation of cyanoarene-based PC•‒ under widely-used photoredox-mediated reaction conditions. By screening various cyanoarene-based PCs, we elucidate strategies to efficiently generate PC•‒ with adequate excited-state reduction potentials (Ered*) via supra-efficient generation of long-lived triplet excited states (T1). To thoroughly investigate the behavior of PC•‒ in actual photoredox-mediated reactions, a reductive dehalogenation is carried out as a model reaction and identified the dominant photodegradation pathways of the PC•‒. Dehalogenation and photodegradation of PC•‒ are coexistent depending on the rate of electron transfer (ET) to the substrate and the photodegradation strongly depends on the electronic and steric properties of the PCs. Based on the understanding of both the formation and photodegradation of PC•‒, we demonstrate that the efficient generation of highly reducing PC•‒ allows for the highly efficient photoredox catalyzed dehalogenation of aryl/alkyl halides at a PC loading as low as 0.001 mol% with a high oxygen tolerance. The present work provides new insights into the reactions of cyanoarene-based PC•‒ in photoredox-mediated reactions.

) via supra-efficient generation of long-lived triplet excited states (T 1 ). To thoroughly investigate the behavior of PC •in actual photoredox-mediated reactions, a reductive dehalogenation is carried out as a model reaction and identified the dominant photodegradation pathways of the PC •-. Dehalogenation and photodegradation of PC •are coexistent depending on the rate of electron transfer (ET) to the substrate and the photodegradation strongly depends on the electronic and steric properties of the PCs. Based on the understanding of both the formation and photodegradation of PC •-, we demonstrate that the efficient generation of highly reducing PC •allows for the highly efficient photoredox catalyzed dehalogenation of aryl/alkyl halides at a PC loading as low as 0.001 mol% with a high oxygen tolerance. The present work provides new insights into the reactions of cyanoarene-based PC •in photoredox-mediated reactions.
To further enhance the efficiency and expand the reaction scope of visible light-driven photoredox catalysis, it is essential to maximize This can mostly be facilitated in a reductive quenching cycle wherein a one-electron-reduced PC (i.e., PC •-) commonly acts as an active PC species because PC •usually exhibits a far longer lifetime than the optically excited PC species (i.e., 1,3 PC * ) acting as an active PC intermediate in an oxidative quenching cycle 21 . Moreover, PC •is regarded as a core intermediate for the recently proposed multiphoton excitation catalysis mechanism based on consecutive photoinduced electron transfer (ConPET) [22][23][24] and electrophotocatalysis [25][26][27] , therefore, merits special attention.
The concentration of PC •in the photostationary state and the ground state reduction potential of the PC (E red 0 (PC)) both play critical roles in photoredox-mediated catalytic reactions employing PC •as an active species 21 . A high concentration of PC •implies a high collision frequency with the substrate under illumination by visible light, which facilitates ET events. In addition, a more negative E red 0 indicates an increase in the driving force for ET, thus accelerating ET processes. To achieve a highly negative E red 0 of a PC, a high energy lowest unoccupied molecular orbital (LUMO) is required 15 . Furthermore, to ensure visible light absorption by such a PC, the energy of the highest occupied molecular orbital (HOMO) should scale with that of the LUMO; however, the accompanying decrease in E red * is detrimental to the photoinduced electron transfer (PET) between a PC and a reductant. In other words, it is very difficult to target PCs combining the following properties: (i) good visible light absorption, (ii) adequate initial PET with a sacrificial reductant, and iii) a highly negative reduction potential. Thus, the generation of PC •is normally targeted in PCs with a less negative reduction potential (e.g., perylene diimide 22,28 , Acr-Mes + BF 4 -8,29 , Rh6G 30 , and Ru(bpy) 3 Cl 2 7,[31][32][33] ) or under special reaction conditions 23,34,35 (Supplementary Table 1 in the Supplementary Information (SI)).
As exemplified by the recent reports by the groups of Zhang 36 , Zeitler 37 , Kwon 15 , and others 11,[38][39][40][41][42][43][44] , cyanoarenes have emerged as attractive organic PCs. Such PCs exhibit excellent catalytic performances for a variety of visible light-driven organic reactions [38][39][40] and polymerizations [45][46][47][48][49] . Among them, 4DP-IPN and its analogs have attracted considerable interest owing to their superior catalytic performance for radical anion-mediated photoredox catalysis. For example, Wickens et al. reported the successful photocatalyzed reductive cleavage of strong C(sp 2 )-N and C(sp 2 )-O bonds by the electrochemically generated 4DP-IPN •-41 . More recently, the groups of Wickens 42 and Wu 43 used 4DP-IPN analogs as PCs to perform the phosphonylation, borylation, and hydroarylation of highly inactivated aryl chlorides. Through careful characterization of the radical anion of 4DP-IPN, they proposed that its high reducing power (E red 0 = −1.66 V vs SCE, all redox potentials in the current work are against saturated calomel electrode (SCE) unless otherwise noted) is the crucial factor.
Meanwhile, recent studies on cyanoarene-based PC •have revealed that a photodegradation of PCs is involved in photoredox catalysis [50][51][52][53] , which might induce unwanted catalytic activities of the photodegraded adducts. However, despite these research efforts, it is still unclear which factors affect the formation and degradation of the radical anion of cyanoarene-based PCs. This lack of understanding can lead to inefficiencies in radical anion-mediated photoredox reactions such as inappropriate choice of PC, excessive PC loading, and inadequate selection of the excitation source. However, no studies have focused on the in-depth investigation of the cyanoarene-based PC •-.
Herein, we investigate the formation and degradation of the cyanoarene-based PC •under widely-used photoredox-mediated reaction conditions (Fig. 1). Through the investigation of various cyanoarene-based PCs with different redox potentials and abilities to generate triplet excited states (T 1 ), we found that organic PCs exhibiting both the ultra-efficient generation of long-lived T 1 and adequately positive excited state reduction potentials (E red * (PC)) enable these PCs to efficiently form strongly reducing PC •under mild visible light illumination. During the screening of these cyanoarenes, we also found that the different photodegradation behaviors of PCs depend on their electronic and steric properties. We also identified a strong correlation between the photodegradation reaction of cyanoarenes and the abilities of PCs to be one-electron reduced. To further investigate the photodegradation behavior of PC •in actual photoredox catalysis, we carried out the reductive dehalogenation of aryl halides as a model reaction. From in situ monitoring of the reaction, we revealed that the dehalogenation and photodegradation of PCs are codependent on the rate of the ET process. Furthermore, we demonstrated the highly efficient dehalogenation of aryl/alkyl halides at a very small loading of 4DP-IPN with a high oxygen tolerance; it thus outperformed other conventional PCs that were used as controls.

PET event between 4DP-IPN and sacrificial reducing agents
We first investigated the initial PET from amine-based sacrificial donors to 4DP-IPN; tertiary amines were used because they are the most commonly used sacrificial reducing agents. Given the ground state oxidation potentials of the prepared amines (E ox 0 (DIPEA) = 0.68 V, E ox 0 (TBA) = 0.88 V, and E ox 0 (TEA) = 0.96 V, where DIPEA, TBA, and TEA represent diisopropylethylamine, triethylamine, and tributylamine, respectively) 54,55 , and the excited state reduction potential of the PC (E red * (4DP-IPN) = 0.63 V), the PET is expected to be very slow owing to its unfavorable thermodynamics. To monitor the PET, we conducted photoluminescence (PL) decay quenching experiments using timecorrelated single-photon counting (TCSPC). The UV-Vis absorption and PL emission spectra of 4DP-IPN are shown in Supplementary Fig. 3. Under degassed conditions without a sacrificial electron donor, the decay lifetimes of the prompt and delayed components of 4DP-IPN Here, ISC, T 1 , PET, and Sub denote intersystem crossing, triplet excited state, photoinduced electron transfer, and substituent, respectively.
were measured to be 3.3 ns and 104 μs, respectively ( Supplementary  Fig. 5). Adding tertiary amines significantly shortened the decay lifetimes of the delayed components ( Supplementary Fig. 6), whereas those of the prompt components changed negligibly ( Supplementary  Fig. 7), implying that T 1 is mainly responsible for the PET events. In fact, the measured rate for PET is larger than the reverse intersystem crossing (RISC) rate and smaller than the ISC rate, further supporting this argument (Fig. 2b). Among the sacrificial donors, DIPEA exhibited the strongest quenching effect ( Fig. 2e and Supplementary Fig. 6). The PET rate constants (k PET ) increased with an increasing thermodynamic driving force (−ΔG PET ), consistent with Marcus normal region behavior ( Supplementary Fig. 6) 56 . As a control, we performed a PL quenching experiment with DIPEA on Ru(bpy) 3 Cl 2 , which is widely used as a PC for reductive quenching photocatalytic cycles [31][32][33] . In the presence of the same amount of DIPEA, k PET for Ru(bpy) 3 Cl 2 (9.7 × 10 6 M −1 s −1 ) was higher than that for 4DP-IPN (1.1 × 10 6 M −1 s −1 ), consistent with the trend in the excited state reduction potentials of the PCs (E red * (Ru(bpy) 3 Cl 2 ) = 0.77 V) 57 . Interestingly, for the same amount of DIPEA, the PL decay quenching in 4DP-IPN changed to a greater extent than that in Ru(bpy) 3 Cl 2 . This indicates that the rate of PET (v PET ) from DIPEA to the PC is faster in 4DP-IPN than that in Ru(bpy) 3 Cl 2 , although the thermodynamic driving force for PET is less favorable in 4DP-IPN. We attribute this finding to the highly efficient long-lived T 1 generation by 4DP-IPN. In fact, the excited state population of PC is a crucial factor for generating the PC •because the molar rate of PET, v PET (in M s −1 ), is described by where [PC * ] is the concentration of the PC in the excited state and [Q] is the concentration of the quencher (i.e., DIPEA). We then modeled the time-dependent excited state concentrations of selected PCs after turning on irradiation; this model was based on the rate law to estimate the PET ability of 4DP-IPN and to compare it to those of other wellknown PCs ( Fig. 2f; see also Supplementary Fig. 13 in the SI for details of the simulation and the rate equations employed). The rate constants used in the kinetics simulations were obtained from either our experiments or the literature. In the photostationary state, the concentration of 4DP-IPN molecules in T 1 was approximately 10 2 times higher than those of both Ru(bpy) 3 Cl 2 and Ir(ppy) 3 and was approximately 10 5 times higher than that of 10-phenylphenothiazine (PTH), which has commonly been used as a highly reducing PC (E ox * (PTH) = −2.10 V) 58 . This indeed suggests that 4DP-IPN is superior for PET with sacrificial agents, despite its unfavorable thermodynamics.

Generation of PC •of cyanoarene-based PCs
The formation of the PC •of 4DP-IPN can be directly monitored by UV-Vis absorption spectroscopy. As the PC •is known to be sensitive to oxygen and moisture 8,59 , all samples were prepared inside a glove box. A degassed solution of 4DP-IPN was first irradiated with a 515 nm light emitting diode (LED) for 1 min in the presence of an excess amount of DIPEA; in this context, it should be noted that exogenous tertiary amines have been reported to be able to suppress back electron transfer (BET) events 60 . The UV-Vis absorption spectra of the 4DP-IPN solution were recorded before and immediately after irradiation (Fig. 2c). Spectral changes were observed, in agreement with a color change that could be observed by the naked eye (Fig. 2c, inset). A decrease in the absorption peak at 470 nm implied the depletion of 4DP-IPN, while a new broad absorption band appeared at 500-750 nm, indicating the generation of 4DP-IPN •-; perfect isosbestic points (at 428 and 493 nm) also appeared for the reaction. Once the resulting PC •was exposed to air, close to 100% of the 4DP-IPN was rapidly regenerated, demonstrating the nearly perfect reversibility of this transformation ( Supplementary Fig. 8a). The stability of the PC •was also measured by recording changes in the UV-Vis absorption spectrum of freshly generated PC •every 2 min under dark conditions (orange line in Fig. 2d). A gradual decay was observed in the dark over more than 20 min, which is approximately two times longer than that of Ir(dtbby) •-(ppy) 2 PF 6 reported by König et al. under similar conditions (green line in Fig. 2d) 23 . This superior stability of 4DP-IPN •is likely due to retarded BET in the triplet contact radical ion pair involving the 4DP-IPN •-, which is related to the lack of heavy atoms and, thus, a decreased spin-orbit coupling.
The formation of the PC •was further confirmed by (timedependent) density functional theory, (TD-)DFT, calculations (using the B3LYP functional, 6-311++G* basis set, and the polarizable continuum model, PCM, with acetonitrile as the solvent); such calculations provide insights into the underlying electronic situation. In the neutral form of 4DP-IPN (black line in Fig. 2c), the low-energy absorption band consisted of a number of transitions with partial charge transfer (CT) character and a relatively low oscillator strength f ( Supplementary  Fig. 14); for instance, S 1 was well described by a transition from the HOMO to the LUMO (HOMO→LUMO), as depicted in Fig. 2a. Furthermore, the TD-DFT calculations correctly reproduced the appearance of a red-shifted absorption band for the PC •-(orange line in Fig. 2c), which was shown to consist of a multitude of CT and locally excited (LE) transitions with a small f; here, the D 2 state corresponded to the excitation from the singly occupied molecular orbital (SOMO) to the LUMO+1, which was of LE character but had a small differential overlap, thus generating a small f ( Fig. 2a and Supplementary Fig. 14). Note, however, that the theoretical evidence clearly indicates the appearance of a near-infrared (NIR) band for the PC •-, which corresponded to the D 1 state (SOMO→LUMO; Fig. 2b and Supplementary Fig. 14), independent of the chosen DFT functional. At the B3LYP level of theory, it appeared at 1454 nm. We thus recorded the absorption spectrum in the NIR range to 1500 nm but did not find evidence for such absorption in this range ( Supplementary Fig. 10). However, an absorption band might possibly have appeared at a wavelength >1500 nm and was therefore not detected. Nonetheless, note that the energy difference between the bands at 1450 and 1600 nm, for example, is only~0.1 eV, which is within the error of the DFT calculations.
Next, we investigated the formation of PC •for a set of cyanoarene-based PCs with different redox potentials and abilities to generate T 1 ; nine additional PCs were synthesized with different donor moieties ( Fig. 3 and Supplementary Fig. 11). In most cases, the lowenergy absorption bands appeared to red-shift relative to those of the PC; even 4-p,p-DCDP-IPN shows broad absorption band in the NIR region (Fig. 3j). These results indicate the broad applicability of our strategy. Furthermore, to verify the relevance of T 1 for PC •formation, two 4DP-IPN analogs (4-p-MCDP-IPN and 4-o,p-DCDP-IPN) with better E red * (i.e., a lower HOMO) but a negligibly small concentration of T 1, were prepared (Fig. 3, Supplementary Fig. 5, and Supplementary Table 4); this finding might be due to fast RISC mediated by vibronic coupling 61 , and further in-depth investigations are currently underway. Interestingly, PC •was not noticeably generated for such PCs in our experimental conditions, clearly confirming that the long-lived T 1 generation of a PC is crucial for the formation of PC •-. Nevertheless, in most strongly twisted donor-acceptor structures, 3 PC * is efficiently generated; furthermore, by changing the donor and/or acceptor moieties, the redox potentials are delicately controlled over a broad range, enabling the use of radical ions with tailored redox potentials for a variety of highly efficient conventional photoredox catalysis, multiphoton excitation catalysis, and photoelectrocatalysis 13,15 .

Photodegradation behavior of 4DP-IPN
The PC •formation was determined to be sensitive to the wavelength and intensity of irradiation. Spectral changes similar to those in the solutions of 4DP-IPN and DIPEA under 515 nm irradiation were observed when irradiated with a 455 nm LED for 1 min; however, the spectrum did not fully recover even after exposure to air, implying that the PC is likely to be degraded by 455 nm irradiation ( Supplementary  Fig. 8). In fact, continuous 455 nm LED irradiation over 5 min resulted in the photodegradation of 4DP-IPN ( Fig. 4a and Supplementary  Fig. 8c). These results suggest that the PC •can be efficiently generated under 455 nm LED irradiation, although followed by molecular degradation or unwanted chemical reactions [50][51][52][53] .
We carefully monitored the degradation of 4DP-IPN in the presence of the sacrificial reductant (i.e., DIPEA) alone under both 455 and 515 nm LED irradiation, and the degradation under 455 nm illumination was much faster than that under 515 nm illumination. This can possibly be ascribed to differences in the absorption efficiencies (ε = 9.0 × 10 3 M −1 cm −1 at 455 nm and ε = 4.5 × 10 2 M −1 cm −1 at 515 nm). As illustrated in Fig. 4a, 4DP-IPN decomposed into two green luminescent compounds. After scaling up the reaction, we successfully isolated the photodegraded adducts. Intensive structural characterization via 1D/2D NMR analyses and mass spectroscopy clearly confirmed that one of the two CN groups of 4DP-IPN was substituted with a methyl group or hydrogen atom to yield 4DP-Me-BN (major product) and 4DP-H-BN (minor product), respectively (Supplementary Figs. 20,21).
Control experiments were performed in combination with DFT calculations to investigate the mechanistic pathway of the 4DP-IPN degradation under the given conditions ( Supplementary  Figs. 18 and 19). Interestingly, no degradation could be identified by thin-layer chromatography (TLC) in the absence of DIPEA ( Supplementary Fig. 18a), suggesting that DIPEA plays an important and specific role in the photodegradation process of 4DP-IPN. We thus assumed that long-lived 4DP-IPN •first formed in the presence of DIPEA, followed by the methyl and hydrogen substitution reaction of the PC •-. Here, the methyl and hydrogen seem to have been provided through the β-scission of the oneelectron-oxidized adduct of DIPEA (DIPEA •+ ). In fact, the C−C (and C−H) bonds located in the β-position of DIPEA •+ are well known to be substantially weaker than those of neutral DIPEA; hence, β-scission normally occurs to generate the radical species 42,62-66 , which was also well reproduced by our DFT calculations (Fig. 4b). Moreover, the possibility that the methyl and hydrogen originated from the solvent (CH 3 CN) could be ruled out because in deuterated acetonitrile (CD 3 CN), the CH 3 substitution reaction was still observed instead of the CD 3 substitution (Fig. 4a).
We further investigated the photodegradation of other cyanoarene-based PCs. As illustrated in Fig. 5, very interesting patterns emerged, which relied on the electronic properties (i.e., propensity for PC •formation) and structural features (i.e., steric environments nearby the CN group) of PCs. As in 4DP-IPN, methyl substitution at one of the two CN groups was observed for 3DP-Cz-IPN and 3DP-DCDP-IPN, both of which have similar electronic and structural properties compared with 4DP-IPN (Fig. 5a). However, in 4Cz-IPN and 4tCz-IPN, ethyl substitution occurred with a small amount of methyl substitution, suggesting that the steric environment near the CN group is a crucial factor in the substitution reaction (Fig. 5b). The ethyl group likely originates from the C−N bond cleavage of DIPEA •+ assisted by a 1,2methyl shift. Indeed, the use of diisopropylmethylamine instead of DIPEA generated a CH 3 -substituted adduct as a major product, which clearly supports our hypothesis.   degradation mixtures were formed in 4-p-MCDP-IPN. These results imply that in such PCs, no well-defined degradation pathway through the PC •intermediate exists, and thus, the photodegradation behavior is determined by the intrinsic photostability of the PCs. Finally, we examined the photodegradation behavior of 3DP-F-IPN and 4-p,p-DCDP-IPN, which effectively generated PC •and contain additional labile groups such as C−F bonds or other types of C−CN bonds. Complex reaction mixtures formed for both PCs, which was probably due to the degradation of these labile groups (Fig. 5d). Further indepth investigations are currently underway to fully understand the photodegradation behaviors of cyanoarene-based PCs.

Dehalogenation of activated aryl/alkyl halides
Based on the experimental results discussed above, we assumed that the supra-efficient PC •generation of 4DP-IPN and its highly negative E red 0 of −1.66 V would enable the very efficient dehalogenation of aryl/ alkyl halides. To that end, we first examined the feasibility of reducing 4-bromobenzonitrile (E red 0 = −1.83 V), which was chosen as a substrate owing to its moderate reactivity, thereby allowing us to compare the catalytic performance of 4DP-IPN with those of well-known PCs. The reaction conditions were optimized by irradiating a mixture of 4bromobenzonitrile, 4DP-IPN (0.005 mol%), and tertiary amines in degassed acetonitrile with a 455 nm LED light at room temperature. Although no conversion was achieved in the presence of five equivalents of TEA, the reduction product, benzonitrile, was obtained in 100% yield after 8 h using ten equivalents of DIPEA (Supplementary  Table 5). This result is consistent with the PL quenching experiments, clearly indicating that the efficient generation of PC •is a key factor in proceeding with the dehalogenation reaction. Notably, 455 nm LED irradiation gave better results (i.e., faster reaction kinetics) than 515 nm LED irradiation, which seems to contradict the UV-Vis absorption results concerning the photodegradation of the PC •-. This inconsistency was most likely due to the fact that the ET between the PC •and the substrate occurred much faster than the photodegradation of the PC •-. To confirm our hypothesis, we investigated the photodegradation behavior of 4DP-IPN in actual reactions to dehalogenate various aryl halides (Fig. 6a) in which ten equivalents of DIPEA were used as a sacrificial agent. Aliquots of each reaction mixture were taken at a given time to monitor the progress of the reactions as well as the PC degradation. As only a small amount of PC (here, 1 mol%) was used, the degradation of the PC was monitored by TLC, whereas the reaction was tracked by 1 H NMR. As shown in Fig. 6a: Reaction A, an 100% yield was obtained for 4-bromobenzonitrile in 1.5 h, while only a trace amount of the photodegradation product, 4DP-Me-BN, appeared; this implies that 4DP-IPN is an active PC for the dehalogenation of 4-bromobenzonitrile. More interestingly, the degradation of the PC was significantly retarded during the dehalogenation reaction compared to that in the presence of DIPEA alone (Fig. 6a: TLC). This was presumably due to the fact that the PC degradation competes with the dehalogenation reaction, as described in Fig. 6c. In other words, in the presence of 4-bromobenzonitrile, the ET from 4DP-IPN •to 4-bromobenzonitrile seemed to be significantly faster than (i) the formation of 4DP-IPN •and (ii) the substitution reaction to form 4DP-Me-BN, which results in a substantial delay in the PC degradation. In fact, faster PC degradation occurred in the dehalogenation reactions of more challenging substrates (i.e., 4chlorobenzonitrile, 4-iodoanisole, and 4-bromoanisole in Fig. 6a: Reaction B, C, and D, respectively), further supporting our hypothesis.
In a further step, we examined the catalytic activity of the PC degradation adduct, 4DP-Me-BN (Fig. 6b), using 1 mol% of 4DP-Me-BN to proceed with the dehalogenation reactions of 4-bromobenzonitrile and 4-bromoanisole; note that the reactions proceeded under the   (Fig. 6b, right). More interestingly, when the continuous irradiation was prolonged to 12 h after the completion of the dehalogenation reaction of 4-bromobenzonitrile, only a small amount of the PC degradation adduct appeared, suggesting that the PC degradation was still retarded in the absence of 4-bromobenzonitrile ( Fig. 6a: Reaction A, TLC). This might have been due to the fact that the reaction product, benzonitrile (E red°= −2.40 V) 67 , can act as an electron acceptor instead of 4-bromobenzonitrile, competing with the PC degradation process.
Finally, the molecular origin of the lower catalytic activity of 4DP-Me-BN was studied. According to DFT calculations, the substitution of CN by CH 3 did not significantly change the overall three-dimensional structure with respect to that of 4DP-IPN, thus still providing a strongly twisted donor-acceptor conformation (Fig. 7a). Nevertheless, the electronic structure substantially changed in 4DP-Me-BN. The substitution of the strongly electron-withdrawing CN group with the weakly electron-donating CH 3 in the central acceptor moiety increased both the HOMO and LUMO energies. However, the increase of 0.47 eV in the LUMO energy was considerably larger than that of the HOMO energy (0.11 eV) because the acceptor (where the LUMO is located; see Figs. 2a and 7a) was directly affected by CH 3 substitution, while the HOMO (mainly located on the donor) was only indirectly affected by Photophysical and electrochemical measurements of PCs species are well in accordance with their DFT calculations. Strongly delayed fluorescence with a long decay lifetime of 153 μs was seen in 4DP-Me-BN (Fig. 7b), indicating an intact thermally activated delayed fluorescence cycle not too different from that in 4DP-IPN. The UV-Vis absorption and PL emission spectra of 4DP-Me-BN showed a distinct blue shift with respect to those of 4DP-IPN (Fig. 7c, d), confirming the steeper increase in the LUMO energy upon CH 3 substitution than that of the HOMO. Finally, the lower excited state reduction potential (E red * = 0.54 V) and ground state reduction potential (E red 0 = −2.05 V) of 4DP-Me-BN compared to those of 4DP-IPN ( Supplementary Fig. 3) were also well supported by the DFT results, which revealed that 4DP-Me-BN exhibited higher HOMO and LUMO energies.
The fact that the catalytic performance of 4DP-Me-BN was worse than that of 4DP-IPN might be surprising, considering that 4DP-Me-BN has a (i) highly negative reduction potential, (ii) decent long-lived T 1 generation, and (iii) greatly improved stability. However, we concluded that the catalytic activity might have been impaired by i) the very low light absorption of 4DP-Me-BN (ε = 1.1 × 10 3 M −1 cm −1 at 455 nm), as observed in the UV-Vis absorption spectrum (Fig. 7d), and (ii) a decrease in the PET rate between 4DP-Me-BN and DIPEA owing to the lowered excited state reduction potential of 4DP-Me-BN. In fact, the k PET of 4DP-Me-BN in the presence of DIPEA measured by a PL quenching experiment was 8.6 × 10 4 M −1 s −1 , approximately one order of magnitude lower than that of 4DP-IPN, which was consistent with Marcus normal region behavior described above (Fig. 7e), thus supporting our hypothesis.
Using the optimized conditions, we screened different PCs. As shown in Table 1, 4DP-IPN outperformed the other PCs used for comparison. Ru(bpy 3 )Cl 2 and Ru(phen) 3 (PF 6 ) 2 , which are commonly used as PCs for reductive cycles, gave very low yields of 13 and 0%, respectively, even at high catalyst loadings of 5 mol%; this might have been caused by the less negative reduction potential of the catalysts (E red 0 (Ru(bpy) 3 57 . Rh6G and PDI employed as PCs for ConPET showed yields of 2 and 4%, respectively, at 5 mol% PC loadings. Although the one-electron-reduced forms of those PCs are known to have highly negative excited state oxidation potentials (E ox * (PDI •-) = −1.87 V and E ox * (Rh6G •-) = −2.40 V) 28,30 , the excited state populations of Rh6G •and PDI •were low because of their short excited state lifetimes; this might have led to inefficient ET with 4-bromobenzonitrile and hence very low yields. On the other hand, Ir(ppy) 3 , known as a PC for oxidative cycles 9 , showed decent catalytic activity, but to give comparable yields, it required 100 times higher PC loading (0.5 mol%) compared to that of 4DP-IPN. The higher required catalyst loading might have been due to the relatively shorter excited state lifetime of Ir(ppy) 3 compared with that of 4DP-IPN •-, despite the highly negative excited state oxidation potential of Ir(ppy) 3 (E ox * (Ir(ppy) 3 ) = −1.73 V). Finally, the reaction scope was explored in the presence of 4DP-IPN (0.05-0.005 mol%) for a range of differently substituted aryl/alkyl halides with ground state reduction potentials of less than −2.2 V, which gave corresponding reduction products in nearly quantitative yields (Fig. 9, Activated aryl/alkyl halides).

Oxygen tolerance
At a higher 4DP-IPN concentration of 0.05 mol%, the reaction of 4-bromobenzonitrile gave a quantitative yield without degassing (Fig. 8b), which enabled a facile gram-scaled reaction under ambient conditions without any preceding degassing (Fig. 9). To understand the origin of this high oxygen tolerance, the kinetics of the reaction with and without degassing were monitored. An inhibition period clearly appeared under non-degassed conditions but not under fully degassed conditions (Fig. 8c). These results imply that dissolved oxygen in the mixture solutions was fully consumed before starting the reductive dehalogenation reaction. According to the previous reports [68][69][70] , oxygen could be consumed through the generation of reactive oxygen species (i.e., singlet oxygen and superoxide anion) and their subsequent reaction with DIPEA and/or its cationic intermediates. The kinetics of the photoredox catalytic reactions of 4-bromobenzonitrile with and without degassing process were then simulated using the mechanistic model described in Fig. 8a (see Supplementary Fig. 15 and Supplementary Table 7 in the SI for details of the kinetics simulation); all parameters including the rate constants used in the kinetics simulations were obtained from the experiments or from the literature. As shown in Fig. 8c (green line with open squares), the experimental trends were quite well reproduced by the kinetics simulations, supporting our hypothesis. Although these trends were well simulated, a substantial discrepancy appeared between the simulated and experimental kinetics, which was likely due to the fact that ET was overestimated, and BET was not considered in the model. In fact, the nonlinearity observed in the experimental kinetics was adequately reproduced in the simulations by considering the adequate BET for a slower ET ( Fig. 8c; green line with open circles). However, significant discrepancies were still observed for dehalogenation reactions in the presence of oxygen (Fig. 8c, right). This is probably due to the depletion of the tertiary amines by reactive oxygen species, which is not considered in model 69,70 . indeed, in air, the oxygen tolerance was determined depending on the amount of DIPEA (Supplementary Table 8).
The experimental results thus far suggest that the dehalogenation of activated aryl/alkyl halides at very low PC loadings and the high oxygen tolerance arise from the highly efficient generation of both the long-lived T 1 of 4DP-IPN and the PC •with a highly negative E red 0 (4DP-IPN) of −1.66 V. In particular, high PC •concentrations apparently led to the highly efficient dehalogenation of the halides by overcoming the potential barrier of~0.5 V without involving the ConPET process. In fact, the addition of 4-bromobenzonitrile into the solution of 4DP-IPN •induced the perfect regeneration of 4DP-IPN, clearly indicating that direct ET occurs between 4DP-IPN •and 4-bromobenzonitrile, and thus, that ConPET was not involved in this case (Supplementary Fig. 12).

Dehalogenation of inactivated aryl/alkyl halides
Finally, the dehalogenation of inactivated aryl/alkyl halides was explored. The E red 0 values of the substrates were ranged from −2.2 to −3.0 V, which were expected to be difficult to overcome by conventional photoredox catalysis. However, despite the very high potential energy barrier, moderate to high yields were mostly attained at higher PC loadings (5-0.5 mol%). We thus presumed that ConPET was involved in the dehalogenation of inactivated halides. In fact, the excited state oxidation potential of 4DP-IPN •was estimated to be −3.38 V according to Eq. (2), which corresponds to the D 2 state; relying on Koopmans theorem, this corresponds to an energy of −1.30 eV. According to the (TD-)DFT calculations, the D 1 state could also contribute to ConPET. However, considering the energies and corresponding reducing powers (i.e., E ox * (4DP-IPN •at D 1 ) = −2.51 V), the D 2 state was likely to be responsible for ConPET. In fact, the excited state lifetime of D 2 was expected to be sufficiently long, as the energy gap between D 2 and D 1 (0.87 eV from E theo ( Supplementary Fig. 14) was similar to that between D 1 and D 0 (0.85 eV, Supplementary Fig. 14); this should effectively slow down the internal conversion from D 2 and thus enhance the D 2 lifetime, not considering here any barriers to internal conversion via conical intersections, which could additionally stabilize the D 2 state.
Thus, the excited state oxidation potential of 4DP-IPN •was high enough to reduce the inactivated aryl/alkyl halides. Despite similar reduction potentials, aryl chloride gave a lower yield than aryl bromide, even with higher PC loadings (i.e., chlorobenzene vs 4-bromoanisole, see Fig. 9). This might be attributed to the slow bond dissociation of the aryl chloride anion owing to the strong C-Cl bond, facilitating BET from the aryl chloride anion to PC or DIPEA derivatives.
The participation of ConPET in the case of 4-bromoanisole was corroborated by the following semi-quantitative quenching experiments ( Supplementary Fig. 12). In fact, the addition of an excess amount of 4-bromobenzonitrile (E red 0 = −1.83 V) to the preformed solution of 4DP-IPN •distinctly changed the color from dark green to yellow in a few seconds, which is indicative of the regeneration of 4DP-IPN. In contrast, the solution of 4DP-IPN •showed no evident color change in the presence of excess 4-bromoanisole (E red 0 = −2.70 V). This suggests that fast ET between 4DP-IPN •and 4-bromoanisole might be ruled out and that a multiphoton excitation process (i.e., ConPET) might be involved. However, given the low to moderate yields of inactivated aryl chlorides, it seems that the ConPET process was not very efficient here (Fig. 9). This was presumably due to the short excited state lifetime of 4DP-IPN •-, which was supported by the fact that steady-state PL emission was not observed for the solution of 4DP-IPN •- (Supplementary Fig. 10). Of course, other possible pathways, such as a halogen atom transfer (XAT) by α-aminoalkyl radicals recently proposed by Leonori et al. 71 , cannot be excluded. Further investigation of the excited state dynamics of PC •and related ConPET processes is underway.

Discussion
We investigated the formation and degradation of one-electron reduced cyanoarene-based PCs under widely-used photoredoxmediated reaction conditions and found that the cyanoarenes exhibiting both ultra-efficient generation of long-lived T 1 and adequately positive E red * enabled these PCs to efficiently form PC •under mildly visible light illumination. We further investigated (i) the photodegradation of various cyanoarene-based PCs with different electronic and steric properties and (ii) the photodegradation of 4DP-IPN in actual dehalogenation reactions with different substrates, providing significant insights into the photoredox catalysis carried out by cyanoarene-based PCs. Based on the in-depth analysis of photodegradation behaviors, we also demonstrated highly efficient reductive dehalogenation, indicating that it is possible to significantly reduce PC loading when photodegradation processes are not involved. We believe that our work will provide a better understanding of the formation, degradation, and photochemical/electrochemical features of PC •-. Moreover, our results pave the way to in-depth knowledge of the reductive cycles and the stability of PCs in purely organic PCdriven photoredox catalysis, which can be applied to address many challenging issues in a variety of photoredox-mediated organic reactions and polymerizations.

Methods
General procedure for generation of PC   c Experimental (filled squares) and simulated (empty squares) reaction kinetics of photoredox-mediated reductive dehalogenation of 4-bromobenzonitrile with 4DP-IPN (0.03 mol%) under Ar and air atmospheres in a closed glass vial; simulated (empty circles) reaction kinetics indicate that the BET process (k BET = 1 × 10 10 M −1 s −1 ) and a slower ET process (k PET = 4.9 × 10 6 M −1 s −1 ) are involved. Kinetics simulations were performed based on the proposed mechanism described in Supplementary  Fig. 15. The rate constants for all processes were evaluated from experiments or calculations (see Supplementary Fig. 15 and Supplementary Table 7 for the full details of the kinetics simulation). screw cap and sealed with parafilm. Subsequently, the reaction was carried out for 2 h under the illumination of four 3 W 455 nm LEDs at RT without degassing process 53 . Afterward, the reaction mixture was evaporated under low pressure and the concentrated crude products were further purified by column chromatography on silica gel.

General procedure for photoredox reductive dehalogenation
Inside the glove box, a 20 mL glass vial equipped with a stirring bar was charged with aryl halides (0.1 mmol), DIPEA (174 μL, 1 mmol), PC (5-0.001 mol% to relative aryl halides), 1,3,5-trimethoxybenzene (33.6 mg, 0.2 mmol) as an internal standard for GC-FID and 1 H NMR, and anhydrous CH 3 CN (1 mL, 0.1 M of aryl halides) as solvent. After, the vial was capped with a rubber septum or screw cap and sealed with parafilm. The reaction batches were purged with Ar (99.9999%) (or with air or with O 2 (99.995%)) for 30 min outside the glove box. Subsequently, the reaction was carried out for hours under the illumination of two 3 W 455 nm LEDs at RT. Without a work-up process, the aliquots of the crude reaction mixture were analyzed by GC-FID or 1 H NMR to obtain yields of dehalogenated products. In preparation for the reaction, pre-prepared stock solutions of the PCs were used for the higher reproducibility of results.