Multi‐Resonant Thermally Activated Delayed Fluorescent (MR‐TADF) Compounds as Photocatalysts

Abstract Donor‐acceptor (D−A) thermally activated delayed fluorescent (TADF) compounds, such as 4CzIPN, have become a widely used sub‐class of organic photocatalysts for a plethora of photocatalytic reactions. Multi‐resonant TADF (MR‐TADF) compounds, a subclass of TADF emitters that are rigid nanographene derivatives, such as DiKTa and Mes3DiKTa, have to date not been explored as photocatalysts. In this study both DiKTa and Mes3DiKTa were found to give comparable or better product yield than 4CzIPN in a range of photocatalytic processes that rely upon reductive quenching, oxidative quenching, energy transfer and dual photocatalytic processes. In a model oxidative quench process, DiKTa and Mes3DiKTa gave increased reaction rates in comparison to 4CzIPN, with DiKTa being of particular interest due to the lower material cost (£0.94/mmol) compared to that of 4CzIPN (£3.26/mmol). These results suggest that DiKTa and Mes3DiKTa would be excellent additions to any chemist's collection of photocatalysts.


Introduction
Visible-light photocatalysis has become a widely used technique in synthetic organic chemistry. Photocatalysis functions by harnessing the electronic excited state of a photocatalyst (PC*), generated via the absorption of a photon, to interact with an organic substrate through either electron or energy transfer. Photocatalytic photoinduced single electron transfer (PET), commonly known as photoredox catalysis, proceeds either via a reductive or an oxidative quenching mechanism, dependent on whether the PC* gains or loses an electron, respectively, during the initial PET. [1] Alternatively, photoinduced energy transfer (PEnT) implicates the transfer of energy from the PC* to the substrate through either a Dexter or Förster energy transfer mechanism, regenerating the ground state photocatalyst (PC). [2] Organometallic complexes based on Ru(II) and Ir(III) are the most widely used PCs (Figure 1a). They possess an attractive suite of properties including suitably long-lived stable excited states, absorption that extends into the visible region where most organic substrates are transparent, plus (especially for Ir(III) complexes), the capacity to modulate both the ground and excited state redox properties through ligand variation. [3] However, the scarcity, toxicity and cost of the noble metals employed has spurred intense efforts to find alternative PCs. There are now many established examples of Earth-abundant metal complexes [4] and metal-free organic photocatalysts, [5] and numerous examples where these perform comparably to the noble metal PCs. While organic photocatalysts, such as xanthene dyes, phenothiazines, and acridinium-based compounds are commonplace (Figure 1b), [5][6][7][8] their ground and excited state redox potentials are difficult to tune. Donoracceptor (DÀ A) thermally activated delayed fluorescent (TADF) PCs, most widely exemplified by the compound 4CzIPN, have rapidly been adopted in the field as their properties are readily tuneable through substituent variation (Figure 1c). [9,10] 4CzIPN, initially developed as an emitter for organic light-emitting diodes a decade ago, luminesces via a TADF mechanism. [11] As a result, 4CzIPN possesses microsecond-long emission lifetimes that, coupled with similar redox properties to that of the widely used [Ir(dF(CF 3 )ppy) 2 (dtbbpy)]PF 6 , endows it with similar photochemical reactivity. 4CzIPN was first used as a photocatalyst by Luo and Zhang to achieve a dual catalysed (C)sp 3 À (C)sp 2 crosscoupling reaction. [12] Subsequent work by Speckmeier et al. [13] demonstrated the versatility of this class of DÀ A TADF photocatalyst and the tunability of their redox potentials by varying the nature and number of electron-donating and electronaccepting groups. A growing number of structurally related DÀ A PCs have since been reported. [16] TADF operates when the energy gap between the lowest energy singlet and triplet excited states, ΔE ST , is sufficiently small such that there is an endothermic upconversion of triplet excitons into singlets by reverse intersystem crossing (RISC). This is possible when the exchange integral between the frontier orbitals involved in the emissive excited state is sufficiently small, which occurs in DÀ A compounds where the donor and acceptor groups are poorly conjugated such as when they adopt a highly twisted conformation, as is the case for 4CzIPN and its derivatives. An alternative molecular design strategy to reduce the exchange integral is based on the exploitation of opposing resonance effects of p-and n-dopants in nanographenes that is embodied in multi-resonance TADF (MR-TADF) emitters. [17] Herein we present the use of two MR-TADF compounds as photocatalysts for the first time, using DiKTa and Mes 3 DiKTa, previously reported by us as emitters in OLEDs, [15] as typical examples ( Figure 1d). Owing to their rigid structure, MR-TADF compounds typically show much narrower emission profiles and smaller Stokes shifts while also exhibiting larger molar absorptivities for the low-energy short-range charge transfer (SRCT) absorption band (see Figure S2). The emissive excited state also shows SRCT character, which is identifiable due to the modest positive solvatochromism, in contrast to the large positive solvatochromism observed for DÀ A TADF compounds (see Figures S3-S4). [15] Enhanced molar absorptivity and reduced positive solvatochromism are expected to have positive implications for photocatalysis reactivity. The higher molar absorptivity of the band that is being targeted for photoexcitation could translate to faster reaction rates and lower required photocatalyst loadings. The attenuated positive solvatochromism of MR-TADF  [14] Ground state redox potentials of DiKTa and Mes 3 DiKTa in MeCN. [15] Excited state redox properties of DiKTa and Mes 3 DiKTa calculated from the experimentally determined E ox /E red and E 0,0 values in MeCN, using E ox (PC * + /PC*) = E ox À E 0,0 and E red (PC*/PC *À ) = E red + E 0,0 . [15] compounds implies that less energy is lost due to stabilization of the excited state by solvent, potentially leading to greater reactivity of the PC, particularly in commonly used polar aprotic solvents such as MeCN and DMF. DiKTa and its mesitylated analogue Mes 3 DiKTa were chosen for investigation as photocatalysts because of their similar redox potentials to those of 4CzIPN ( Figure 1e). [13,15] An additional benefit is that the raw material cost per mmol is significantly lower for DiKTa (£0.94/ mmol) than for 4CzIPN (£3.26/mmol, see Supporting Information). These PCs were assessed across a diverse range of transformations including reductive quenching reactions, oxidative quenching reactions, energy transfer reactions, nickel dual catalysis and hydrogen atom transfer (HAT) dual catalysis. The result of this assessment shows that DiKTa and Mes 3 DiKTa are attractive alternatives to the widely used 4CzIPN.

Reductive quench
Our investigations began with a decarboxylative photo-Giese reaction. This process has previously been reported by Ji et al. [8] for their comparison of the effectiveness of different acridinium photocatalysts and also by Speckmeier et al. [13] for their comparison of the suitability of alternative DÀ A photocatalysts. In the latter study Speckmeier et al. found that when 4CzIPN was used, a superior isolated yield of 80 % is achieved compared to the previously reported best acridinium photocatalyst, which produced an isolated yield of 73 %. Using N-Cbz protected proline 1 a as the carboxylic acid substrate and diethyl maleate 2 as the electron deficient alkene, both DiKTa and Mes 3 DiKTa gave comparable NMR yields to that of 4CzIPN (Table 1, [19] were also investigated in order to differentiate the photooxidation ability of the PCs. With isobutyric acid both DiKTa and Mes 3 DiKTa showed improved NMR yields of 78 % and 79 %, respectively (Table 1, entries 4 and 5), relative to the 64 % achieved using 4CzIPN (Table 1, entry 6). Changing to the primary radical formed when using propanoic acid resulted in lower yields for all three PCs (Table 1, entries 7-9). This is likely due to the decreased nucleophilicity of primary radicals relative to secondary radicals, leading to alternative and undesired reaction pathways becoming competitive. Notwithstanding the lower yields, both DiKTa and Mes 3 DiKTa still outperformed 4CzIPN (Table 1, entries 7-9). 4CzIPN in the presence of similar carboxylic acids has been shown by König et al. [14] to undergo photosubstitution with the radical formed after decarboxylation. This prompted us to investigate the stability of DiKTa and Mes 3 DiKTa under similar conditions (see Supporting Information for details). Unfortunately, 4CzIPN, DiKTa, and Mes 3 DiKTa all show a blue-shift in their absorption spectra after irradiation in the presence of 1 a, which suggests DiKTa and Mes 3 DiKTa are also unstable under these conditions; however, the products of this decomposition were not identified.
PC loading was next investigated as a discriminating parameter. The reaction using N-Cbz-proline 1 a as starting material was therefore repeated at 1 mol%, 0.5 mol%, 0.25 mol% and 0.1 mol% (Table 2). Yields remained largely the  same for all three PCs down to 0.5 mol% (

Oxidative quench
Subsequent studies assessed these PCs in an oxidative quenching process based upon the atom transfer radical addition (ATRA) reaction developed by Pirtsch et al. [20] Using perfluorobutyl iodide 4 a (E p red = À 1.42 V vs. SCE in MeCN) [21] and tertbutyl-N-allyl carbamate 5 as the substrates, 4CzIPN produced the desired ATRA product 6 a in 83 % yield ( Table 3, entry 1). Interestingly, when using DiKTa and Mes 3 DiKTa, near quantitative yields of 97 % and 93 %, respectively, could be achieved (Table 3, (Table 3, entry 4-6). However, when using 4 c, 4CzIPN gave marginally improved product yields, affording 6 c in 76 % compared to 69 % and 71 % for DiKTa and Mes 3 DiKTa, respectively, although the difference between them is within the error observed for this reaction (Table 3, entries 7-9). The reduction potential stated for 4 b by Pirtsch et al. of E p red = À 0.49 V vs. SCE was originally reported by Tanner et al. [22] and is commonly used in the literature; however, this value is erroneous and occurs only as a result of electrochemical degradation ( Figure S6). Indeed, a previous report by Bahamonde et al. [23] found that the peak reduction potential is significantly more negative E p red = À 1.39 vs. SCE in MeCN, and our own measurements have found it to be E p red = À 1.21 V vs. SCE in DMF from the peak of the differential pulse voltammogram (DPV). Similarly, literature values for the reduction potential of 4 c vary significantly from E p red = À 1.0 V vs. SCE [24] in DMF to E p red = À 1.74 vs. SCE [23] in MeCN; our own measurements indicate that E p red = À 1.41 V vs. SCE in DMF from the peak of the DPV. Such a negative reduction potential would be predicted to be beyond the ability of any of these photocatalysts to reduce via oxidative quenching; however, quenching experiments with both 4CzIPN and DiKTa (see Figures S8-S11) show quenching to occur in the presence of 4 c in DMF, showing that while endergonic, oxidative quenching does occur.

Photoinduced energy transfer (PEnT)
Having shown that DiKTa and Mes 3 DiKTa are capable photoredox catalysts, attention turned to their application in PEnT. One of the simplest examples of PEnT processes is the (E)/(Z) isomerization of alkenes. The isomerization of (E)-stilbene 7 a was used effectively by Lu et al. to compare various DÀ A TADF fluorophores and then correlate the results with their triplet energies (E T ) relative to that of (E)-stilbene (E T = 2.2 eV) and (Z)stilbene (E T = 2.5 eV). [25,26] The crucial determinant for the efficiency of Dexter PEnT reactions is the degree of spectral overlap between the emission of the PC and the spin-forbidden absorption of the substrate. [2] A cross-comparison of the triplet energies of the reactant and PC are typically used as a crude handle to assess whether the reaction is likely to proceed. Thus, to maximize the yield for the isomerization of the substrate by limiting the reverse reaction, E T (Substrate) � E T (PC) < E T (Product). Using similar reaction conditions to those of Lu et al. [Ru-(bpy) 3 ](PF 6 ) 2 (E T = 2.13 eV) [27] achieved a (Z)/(E) ratio of 94 : 6 and 4CzIPN (E T = 2.53 eV) [25] performed comparably with a (Z)/(E) ratio of 92 : 8 (Table 4, entry 1 and 2). When using DiKTa (E T = 2.61 eV) [15] as the photocatalyst a (Z)/(E) ratio of only 59 : 41 was observed (Table 4, entry 3), consistent with the high triplet energy of this PC, although a (Z)/(E) ratio of 61 : 39 was achieved with Mes 3 DiKTa (E T = 2.49 eV), [15] despite having a similar triplet energy to that of 4CzIPN (  used as PEnT photocatalysts. Diisopropyl fumarate 7 b (E T = 2.7 eV) [25] was used by Lu et al. as a more challenging substrate due to its higher E T . [Ir(dF(CF 3 )ppy) 2 (dtbbpy)]PF 6 (E T = 2.67 eV) [27] was used as the reference PC for this reaction and achieved a (Z)/(E) ratio of 95 : 5 (

Dual Photoredox Catalysis
Metallaphotoredox catalysis is a fast growing area of research as it often offers a mild alternative to existing transition metal catalytic reactions and give access to different redox couples of the co-catalyst, resulting in new reactivity. [28] Nickel in particular has been paired with photocatalysts for a wide range of different coupling reactions with Luo and Zhang [12] reporting the first use of 4CzIPN as a photocatalyst in a dual-mode catalysed (C)sp 3 À (C)sp 2 cross-coupling. Employing a modified version of this reaction to assess the performance of DiKTa and Mes 3 DiKTa, using 4CzIPN, the coupling reaction between 1 a and aryl bromide 9 gave the desired product 10 in 78 % yield (  [29] used the alkyl radicals generated after the reductive quenching between PC* and triethylamine to abstract iodine atoms from alkyl iodides 11 to generate alkyl radicals that typically would require a far more potent reductant (E red (RÀ I) < À 2 V). These alkyl radicals can then be trapped by a thiol HAT catalyst to generate the dehalogenation products 12 (Table 6). Under the literature conditions using 4CzIPN, an 85 % yield of 12 was obtained ( Table 6, entry 1). Both DiKTa and Mes 3 DiKTa were able to achieve comparable average yields of 88 % and 86 %, respectively ( Table 6, entries 2 and 3).

Reaction profile and rates analysis
Reactions in the previous sections have used the yield of the reaction after a given time to compare the efficiency of the PCs. While useful, this provides an incomplete picture of these reaction processes as it does not allow comparison of relative rates of product formation. This prompted an investigation into the reaction kinetics of the PCs for a model transformation, with the ATRA reaction between 4 a and 5 shown in Table 3 chosen.    Inspired by Yi et al. [30] in situ NMR was used to monitor the generation of product employing each of the PCs (Figure 2). Notably, using this experimental set-up, the reaction reached completion for all three PCs in less than 3 h, significantly faster than using a photoreactor, presumably due to more efficient irradiation within the in situ NMR set-up. Furthermore, when catalysed by Mes 3 DiKTa or DiKTa the rate of product formation is significantly enhanced than with 4CzIPN. To quantify these differences, initial rates of the reaction with each photocatalyst were calculated and compared, with the use of DiKTa giving a slightly larger initial rate (5 × 10 À 4 M s À 1 ) than Mes 3 DiKTa (3.7 × 10 À 4 M s À 1 ), but with both an order of magnitude larger than that of 4CzIPN (0.6 × 10 À 4 M s À 1 ) ( Table 7). While our original hypothesis was that increased molar absorptivity at the excitation wavelength for DiKTa and Mes 3 DiKTa compared to 4CzIPN would lead to increased reaction rates at 455 nm the molar absorptivity for DiKTa and 4CzIPN are similar (see Figure S2). It is not evident what is the cause for the divergence in reaction rates between the two photocatalysts.

Conclusion
In summary this manuscript demonstrates the potential of MR-TADF compounds as a new class of photocatalyst, using DiKTa and Mes 3 DiKTa as examples compared with 4CzIPN as a prototypical donor-acceptor TADF benchmark. Compared to other photocatalysts DiKTa stands out for its wide redox window, low molecular weight, and low cost. Multiple different classes of photocatalytic reaction were tested, including, oxidative and reductive quenching reactions, energy transfer reactions and dual catalytic reactions. Both DiKTa and Mes 3 DiKTa were shown to be comparable or superior in all examples of photoredox catalysis, particularly at low catalyst loadings, and complementary for energy transfer reactions compared to 4CzIPN due to their higher triplet energies. In situ NMR studies were used to probe reaction kinetics of the ATRA reaction between a perfluorinated alkyl halide and an unactivated alkene and showed a significant enhancement of the rate of reaction when using either DiKTa or Mes 3 DiKTa over 4CzIPN. To the best of our knowledge this work documents the first instance of MR-TADF compounds being used in photocatalysis and we expect further work to expand to other MR-TADF compounds of different structural classes.

Experimental Section
Decarboxylative photo-Giese reaction: Carboxylic acid (0.15 mmol, 1.0 equiv.), K 2 HPO 4 (31.4 mg, 0.17 mmol, 1.1 equiv.), photocatalyst (3.0 μmol, 0.02 equiv.) and diethyl maleate (27 μl, 0.18 mmol, 1.2 equiv.) were added to a vial in the photoreactor then evacuated and backfilled with nitrogen three times. In a separate Schlenk flask acetonitrile was sparged for 10 minutes and then added to the reaction vial (3 mL). The reaction was then stirred under 440 nm irradiation at rt for 16 h. Water (5 mL) was added then the mixture was extracted with CH 2 Cl 2 (3 × 5 mL). The organic phases were combined and dried (Na 2 SO 4 ) then concentrated in vacuo. Purification by silica column chromatography EtOAc:Pet. Ether afforded the desired products.
Oxidative quench ATRA reaction for 4 a/b: tert-Butyl allyl carbamate (39.3 mg, 0.25 mmol, 1.0 equiv.), alkyl halide (0.50 mmol, 2.0 equiv.) and the photocatalyst (2.5 μmol, 0.01 equiv.) were added to a vial in the photoreactor then evacuated and backfilled with nitrogen three times. In a separate Schlenk flask a DMF/H 2 O (1 : 2) solution was sparged for 10 minutes and then added to the reaction vial (0.6 mL). The reaction was then stirred under 440 nm irradiation at rt for 16 h. Water (5 mL) and EtOAc (5 mL) were added, and the mixture was extracted with EtOAc (3 × 5 mL). The organic phases were combined and dried (Na 2 SO 4 ) then concentrated in vacuo. Purification by silica column chromatography EtOAc:Pet. Ether afforded the desired products.
Oxidative quench ATRA reaction for 4 c: tert-Butyl allyl carbamate (47.2 mg, 0.30 mmol, 1.0 equiv.), alkyl halide (0.60 mmol, 2.0 equiv.) and the photocatalyst (3.0 μmol, 0.01 equiv.) were added to a vial  Table 3, replacing DCE with CD 2 Cl 2 as solvent using 455 nm LED and 1,4-bis(trimethylsilyl)benzene as internal standard. (a) Concentration of 5 over time. (b) Concentration of 6 a over time.  ) and the photocatalyst (3.75 μmol, 0.01 equiv.) were all added to a vial in the photoreactor and evacuated then backfilled with nitrogen three times. In a separate Schlenk flask DMF was sparged for 10 minutes with nitrogen and then added to the reaction vial (3.5 mL). The reaction was then stirred under 440 nm irradiation at rt for 16 h. Water (5 mL) and EtOAc (5 mL) were added, and the mixture was extracted with EtOAc (3 × 5 mL). The organic phases were combined and dried (Na 2 SO 4 ) then concentrated in vacuo. Purification by silica column chromatography EtOAc : Pet. Ether afforded the desired product.
Dual catalysed deiodination: 4-Iodo-N-boc-piperidine (46.7 mg, 0.15 mmol, 1.0 equiv.) and the photocatalyst (7.50 μmol, 0.05 equiv.) were added to a vial in the photoreactor and evacuated then backfilled with nitrogen three times. In a separate Schlenk flask a solution of triethylamine (62.7 μl, 0.45 mmol, 3.0 equiv.), acetonitrile (1.25 mL) and water (0.25 mL) was degassed via three freeze-pump-thaw cycles then transferred to the vial in the photoreactor. In a separate vial methyl thioglycolate (2.7 μl, 30 μmol. 0.2 equiv.) was sparged with nitrogen for 5 minutes before being added to the vial in the photoreactor. The reaction was then stirred under 440 nm irradiation at rt for 4 h. Then 0.5 mL of a stock solution of 1,3,5-trimethoxybenzene in acetonitrile (0.1 m) was added and the solution stirred for 5 minutes. A sample was taken and concentrated in vacuo then dissolved in CDCl 3 to obtain an NMR yield. A reference sample was prepared for comparison from piperidine using a known literature procedure. [31]