Optimization study. With these considerations in mind, diphenyl ether (1a) and 4-methylbenzoic acid (2a) were studied as model substrates. Under our developed conditions for the intramolecular C–O bond cleavage,55 using PDI64 or Acr+-Mes ClO4– (PC 1)44-46 as PC, with 10 mol% K2HPO4 as base, under 450–455 nm blue LEDs irradiation, only less than 10% of phenyl 4-methylbenzoate (3a) and phenol (4a) were obtained (Table 1, entries 1, 2). Thereafter, a series of Lewis acids such as Cu(OAc)2, Cu(acac)2, Cu(OTf)2, Ni(acac)2, Fe(acac)2, Zn(acac)2, and Cu(TMHD)2 were studied (Table S1, entries 1–6, Table 1, entry 3). Cu(TMHD)2 slightly promoted the transformation. Adjustment of the wavelength of the blue LEDs to the maximum absorption of PC 1 (425–430 nm) induced a slightly increased reactivity (Table 1, entry 4). Other solvents such as MeOH, DCE, EtOAc, and acetone did not give any better results (Table S1, entries 7–10). As Acr+-Mes ClO4– is susceptible to degradation in the presence of oxygen-centered radicals,48 and de-N-methylation is also possible,47 it is deduced that the generated carboxylic acid radical may induce the degradation of Acr+-Mes ClO4–.
Subsequently, PC 2 and PC 3 were tried,48 in which the N-phenyl were used instead of the N-methyl, and also with the tert-butyl on the 3- and 6-positions of the acridinium in the latter case. Both factors induced distinct higher efficiency (Table 1, entries 5 and 6, 3a in 29% and 55%, with 4a in 22% and 49%).
Furthermore, the influence of the substitutes on the 9-aryl ring was investigated. Since the complex procedure for synthesis of PC 3,48 PC 4–PC 9, with a similar skeleton to that of PC 3 but with different substituents on the 9-aryl ring, by a two-step synthetic procedure,52 were investigated (Table 1, entries 7–12). Notably, the aryl rings with electron-withdrawing groups instead of mesitylene, typically used in other acridinium catalysts,44-54 resulted in noticeably high efficiencies. To our delight, PC 9 with 2’-Cl and 4’-F on the 9-aryl ring resulted in 80% of 3a with 71% of 4a (entry 12). Although the smaller group of 2’-Cl compared with the methyl groups in mesitylene was used, X-ray crystallography of PC 9 unambiguously confirmed the angle of torsion between the 9-aryl ring and the acridinium ring (Fig. 1B), which is closely related with a longer fluorescence lifetime.44,47 In addition, variation of the tert-butyl groups from the 2,7-positions to the 3,6-positions resulted in decreased efficiency (entry 13). Further variation of the substituents on the 10-aryl ring revealed that the unsubstituted phenyl gave a slightly higher yield (entry 15). Decreasing the amount of Cu(TMHD)2 resulted in obviously lower efficiency (entry 16). The amount of base did not influence the reaction efficiency, even without base (entries 17, 18). Other Lewis acids, such as Cu(OAc)2, Cu(acac)2, and Fe(acac)2 instead of Cu(TMHD)2, were investigated once again (entries 19–22), as slight differences during the initial investigation. The distinctly positive influence of Cu(TMHD)2 was further confirmed. Without base and Cu(TMHD)2, no reactivity was observed (entry 23). Control experiments indicate that a PC and visible light irradiation are essential (entries 25, 26). The fluorescence lifetime and the redox potentials of PC 1–PC 12 were determined (Table S2). The data do not provide clear insight regarding the higher efficiency achieved using PC 9.
Table 1 Optimization of the Reaction Conditions.

Evaluation of substrate scope. With the optimized reaction conditions, the substrate scope was investigated (Fig. 3). First, the influence of various substituents on the benzoic acid was investigated. 4-Methoxy, 4-tert-butyl gave decreased reaction efficiencies (3ab, 3ac in 61–65% with 4a in 56–60%). The benzoic acid afforded 3ad in 55% with 4a in 50%. 4-Fluoro, 4-chloro, and 4-bromo induced excellent yields (3ae–3ag in 87–91% with 4a in 82–84%). 4-Nitro, 4-aldehyde also resulted in high efficiencies (3ah, 3ai in 71–77% with 4a in 65–70%). The substituents on the 3-position were also studied. As with the substituents on the 4-position, fluoro and chloro resulted in high yields (3ak, 3al in 81–86% with 4a in 74–76%). Methyl gave a decreased yield (3aj in 61%). When the methyl was installed on the 2-position, no product was observed. 2-Fluoro resulted in lower efficiency (3ap in 63%). The phenomenon should be influenced by the steric factor.
Next, the substituents on the aryl ring of the diphenyl ethers were investigated (Fig. 4). With electron-withdrawing groups such as methyl ester, trifluoromethyl, nitro, cyan, or acetyl on the 4-position of the aryl ring, 3a (65–74%) and the corresponding phenols with these electron-withdrawing groups (4b-4f, 62–72%) were selectively obtained in high yields. The results agree with the designed pathway of the electrophilic attack of the generated carboxylic acid radical, including the selective attack on the electron-rich aryl ring of the diphenyl ethers. 4-Bromo afforded two esters 3a (22%) and 3aq (64%), and two phenols 4g (21%) and 4a (54%). The reason for the result may be the synergic effect of induction and conjugation of the bromo. For the 4-phenyl group, 3ar and 4a were selectively obtained in 84% yields. The reason for the selectivity may be the 4-phenyl group stabilizing the generated radical intermediate after the electrophilic attack. 4-Methyl and 4-methoxy resulted in very low efficiency (<10%). Generally, it maybe be explained by their comparatively lower oxidation potentials, inhibiting generation of the aryl carboxylic radical (Table S3). Methyl, methoxyl, dimethyl substituents on other positions afforded 3as–3au in 68–80% yields, and comparable yields of 4a. Symmetric dimethyl, dimethoxyl, dibromo, and dichloro, and asymmetric dichloro on the 3- or 4-positions resulted in good to high efficiencies (3av, 3aw, 52–90%, 4h–4k, 50–86%). When the methyl and methoxy on 2- or 3-position, with 4¢-ester, 4¢-cynao or 4¢-trifluoro, were investigated, the esters with methyl or methoxyl, as well as phenols with these electron-withdrawing groups, were obtained selectively in high yields (3as–3ax, 72–82%, 4b, 4e, 4c, 70–82%).
Synthetic application. To demonstrate the potential application, a gram-scale reaction of 1a with 2e in a flow reactor and a following one-pot hydrolysis was conducted. 4a was obtained in 80% yield, with 2e in 88% recovery rate (Fig. 5A). Meanwhile, the model of 4-O-5 lignin linkage (1t)30 afforded 4a (71%) and 4l (75%) in high efficiency, with 2e in 82% recovery rate (Fig. 5B).
Mechanism studies. To gain insight into the reaction mechanism, a series of experiments were conducted. First, UV−vis absorption spectra of each component and the reaction mixture confirmed that PC 9 acts as a PC (Fig. 6A). Second, luminescent quenching experiments were conducted (Fig. 6B). The anion of 2a (4-MePhCO2–nBu4N+), 2a, and 1a quenched the excited state PC 9*. The anion of 2a displayed an obviously larger quenching rate. Third, the pH value of the reaction mixture was determined as 3.61 or 4.30, with or without 10 mol% Cu(TMHD)2. Based on these pH values, Cu(TMHD)2 should promote the ionization of 2a by the salt effect.65 In addition, under base free conditions, the high reactivity of 80% ester 3a with Cu(TMHD)2 (entry 18), and 13–56% ester 3a with Cu(OAc)2, Cu(acac)2, Ni(acac)2, Fe(acac)2 (entries 19−22) in comparison with no production of 3a without any these metal salts (entry 23), these results suggest that the function of Cu(TMHD)2 also as a Lewis acid to promote the transformation.
Furthermore, with addition of TEMPO as oxidant, compounds 5, instead of 3 and 4, were obtained via the possible intermediates C48 (Fig. 7). This result suggests the generation of intermediate B’ to afford 3a and 4a under the optimized conditions.
The thermodynamic feasibility of the photo-induced SET was analyzed based on the oxidation–reduction potentials. The oxidation potential of E4-CH3PhCO2•/4-CH3PhCO2–, E1a+•/1a, and the reduction potential of EPC 9/PC 9–• in CH3CN were determined as +1.45 V vs. SCE, +1.86 V vs. SCE66, and –0.47 V vs. SCE (Figures S9, S10, and S34), respectively. The excited-state energy E0,0 of PC 9 was determined as 2.63 eV (Fig. S22). Therefore, the reduction potential of EPC 9*/PC 9–• was calculated as + 2.16 V vs. SCE (EPC*/PC–•=EPC/PC–•+ E0,0) (Fig. S34). These reduction potentials indicate the prior formation of PC–• and the carboxylic acid radical55-60 by a SET between PC* and the carboxylic acid anion. Furthermore, a quantum yield value of φ=0.20 was determined. Thus, at this stage, whether the reaction proceeds via a photoredox catalytic pathway or a radical chain pathway could not be reached.67
Based on these results, the reaction mechanism is proposed as shown in Fig. 8. First, Cu(TMHD)2 promotes the ionization of 2a to afford 2a– and a proton. Meanwhile, irradiation of PC with blue LEDs leads to the excited state PC*. A SET occurs between PC* and 2a– to generate the carboxylic acid radical A¢ and PC–•. An electrophilic attack of A¢ occurs on the electron-rich aryl ring of diphenyl ethers to form intermediate B¢. A SET between B¢ and PC–• in the presence of a proton, and the promotion of Cu(TMHD)2 as a Lewis acid afford 3a, 4a, with the regeneration of PC.
In summary, we have developed visible-light photoredox-catalyzed C–O bond cleavage of diaryl ethers by an acidolysis and a following one-pot hydrolysis at rt. Two phenols are obtained from a diaryl ether in high efficiency. The aryl carboxylic acid used for the acidolysis can be recovered. The key to success of the acidolysis is merging visible-light photoredox catalysis with a new acridinium photocatalyst and Lewis acid catalysis with Cu(TMHD)2. The transformation is applied to a gram-scale reaction and the model of 4-O-5 lignin linkages. In comparison with the developed selective hydrogenolysis in recent years, using H2O instead of a large amount of reductant affords two more valuable phenols at rt. This approach would inspire further visible-light photoredox-catalyzed C–O bond cleavage of 4-O-5 lignin linkages in native biomass for utilization of lignin as renewable aryl sources.