Heterogeneous Photocatalysis as a Potent Tool for Organic Synthesis: Cross-Dehydrogenative C–C Coupling of N-Heterocycles with Ethers Employing TiO2/N-Hydroxyphthalimide System under Visible Light

Despite the obvious advantages of heterogeneous photocatalysts (availability, stability, recyclability, the ease of separation from products and safety) their application in organic synthesis faces serious challenges: generally low efficiency and selectivity compared to homogeneous photocatalytic systems. The development of strategies for improving the catalytic properties of semiconductor materials is the key to their introduction into organic synthesis. In the present work, a hybrid photocatalytic system involving both heterogeneous catalyst (TiO2) and homogeneous organocatalyst (N-hydroxyphthalimide, NHPI) was proposed for the cross-dehydrogenative C–C coupling of electron-deficient N-heterocycles with ethers employing t-BuOOH as the terminal oxidant. It should be noted that each of the catalysts is completely ineffective when used separately under visible light in this transformation. The occurrence of visible light absorption upon the interaction of NHPI with the TiO2 surface and the generation of reactive phthalimide-N-oxyl (PINO) radicals upon irradiation with visible light are considered to be the main factors determining the high catalytic efficiency. The proposed method is suitable for the coupling of π-deficient pyridine, quinoline, pyrazine, and quinoxaline heteroarenes with various non-activated ethers.


Introduction
Heterogeneous photocatalysis in organic synthesis is a young and fast-growing area [1][2][3][4][5]. The semiconductor materials used in photocatalysis are inexpensive and widely available; their advantages include the ease of separation from organic products, stability and recyclability [1,5]. However, the development of this area is still hindered by several formidable obstacles, such as low catalytic efficiency due to the low degree of charge separation in photoexcited states and the fast recombination of electron-hole pairs [6,7], low visible light absorption and low selectivity due to the strong oxidation power of photogenerated valence-band (VB) holes in popular semiconductors (TiO 2 , ZnO, Bi 2 O 3 , WO 3 , etc.) [1,8]. This situation is reflected in the comparatively low number of synthetic methods in fine organic synthesis based on heterogeneous photocatalytic systems compared to the mainstream applications of heterogeneous photocatalysis: oxidative destruction of pollutants [9][10][11], hydrogen generation [12,13], CO 2 reduction [14][15][16] and water splitting [17].
UV irradiation, which is used frequently for the excitation of heterogeneous photocatalysts, is inconvenient due to safety issues, the comparatively high cost of UV light sources, incompatibility with common laboratory glassware (UV-transparent quartz is necessary) and possible side reactions due to the high energy of the light. The modification of heterogeneous photocatalysts, such as TiO2, in order to shift their photoactivity spectrum from UV to visible light [10,[34][35][36][37] is the key task for expanding the scope of their applications in organic synthesis, increasing selectivity and making the of use cheap and available light sources for catalyst activation possible. At present, the following modification approaches have been proposed: the immobilization of dyes (organic compounds or metal complexes) on the photocatalyst surface [34,[38][39][40][41], doping with metal ions or nonmetal elements [42,43], semiconductor coupling [7,[44][45][46][47][48][49] and modification with organic molecules bearing hydroxyl or carboxyl groups [34,[50][51][52][53][54][55][56], which demonstrate the occurrence of visible light absorption when adsorbed on the surface of a semiconductor.
NHPI/TiO2 is one of the efficient catalytic systems activated by visible light based on industrially available substances (Scheme 1). The interaction of NHPI with the TiO2 surface leads to the occurrence of visible light absorption, resulting in the photogeneration of phthalimide-N-oxyl radicals (PINO) [20,22]. In our previous work [20], we demonstrated that the NHPI/TiO2 system could be successfully applied to the aerobic oxidation of alkylarenes under visible light irradiation (Scheme 1A). The conceptual novelty of this system arises from the conjunction of heterogeneous photocatalysis with homogeneous radical chain organocatalysis. A distinguishing feature of this system is the migration of PINO into the volume of solution, where the PINO/NHPI catalyzed radical chain process, once initiated on the TiO2 surface, produces the target product without the need for additional light absorption [20]. Thus, the energy efficiency of photocatalysis is fundamentally improved by combining heterogeneous photocatalysis with homogeneous organocatalysis. In the presence of additional organocatalyst (2,2,6,6-Tetramethylpiperidin-1-yl)oxyl (TEMPO) the effective oxidative homocoupling of benzylamines [22] [20], oxidative homocoupling of benzylamines (B) [22], and Minisi-type corss-dehydrogenative C-C coupling reported in the present work (C). Scheme 1. Applications of NHPI/TiO 2 photocatalytic system in organic synthesis: CH-oxygenation (A) [20], oxidative homocoupling of benzylamines (B) [22], and Minisi-type corss-dehydrogenative C-C coupling reported in the present work (C).
In the present study, we demonstrate the successful application of the NHPI/TiO 2 system to a more challenging cross-dehydrogenative C-C coupling process (Scheme 1C). In this case, previously reported CH-oxygenation processes [20] should be suppressed, which is a difficult task. In addition, the process of C-O coupling between NHPI-derived PINO radicals and CH-reagents [57][58][59] must be avoided. The oxidative coupling of ethers with π-deficient N-heteroaromatic compounds (a Minisci-type reaction) was chosen as a model reaction due to the practical importance for the functionalization of N-containing heterocycles with C-C bond formation. Minisci-type reactions [60][61][62][63][64][65][66][67][68] are based on the addition of nucleophilic C-centered radicals to electron-deficient arenes and represent one of the most important methods for the functionalization of such arenes, along with the nucleophilic aromatic substitution of hydrogen [69][70][71], and functionalization via transitionmetal-catalyzed C(sp 2 )-H bond activation [72][73][74][75][76]. The products of the Minisci reaction are of great value for medicinal chemistry [61,64]. Thus, the development of new, milder, more efficient methods tolerant to a large number of functional groups based on Minisci chemistry remains a hot research topic.

Optimization of Photocatalytic System Composition
Based on our previous work [20], TiO 2 with high specific surface area (anatase nanopowder, Hombikat UV100) and industrially available N-hydroxyphthalimide were chosen as the components of the photochemical system. Blue LEDs (455 nm) with an input power of 10 W were used as light sources. In the first step, we optimized the conditions of the photochemical cross-dehydrogenative Minisci reaction between 4-methylquinoline 1a and tetrahydrofuran 2a (Table 1). Tert-butyl hydroperoxide (TBHP) was used as an inexpensive, easily available and metal-free oxidant.
The starting conditions (10 mg of TiO 2 , 20 mol.% of NHPI, 4 mmol of TBHP, 5 h, run 1) yielded 45% of the product 3aa. The absence of either TiO 2 or NHPI resulted in the zero conversion of 1a (runs 2, 3), proving that both components of the catalytic system are essential. Without t-BuOOH, the reaction proceeded with low efficiency: only trace amounts of the product were formed (run 4). As a rule, the addition of a strong Brønsted acid, such as HCl [85] or TFA [77,79,82,84,86], increases the efficiency of the Minisci reaction. Acids protonate π-deficient N-containing heterocycles, making them more susceptible to attack by nucleophilic C-centered radicals [67]. However, in our case, the addition of trifluoroacetic acid (TFA, run 5) had no significant effect on the yield and conversion. The addition of 0.5 mL of water resulted in a drop in 3aa yield (run 6). Water breaks down the stable suspension of TiO 2 in THF, causing the catalyst particles to aggregate in the water droplets. Both an increase and a decrease in the amount of THF lead to a decrease in the yield of 3aa (runs 7,8). The dilution of the reaction mixture with such co-solvents as hexafluoroisopropanol (HFIP, run 9) and acetonitrile (MeCN, run 10) slowed down the reaction, and dilution with dichloroethane (DCE, run 11) led to the complete suppression of the target process. It is known that hydrogen peroxide can be used as the oxidant for the photocatalytic Minisci reaction [85]. However, the change of the oxidant from TBHP to aqueous H 2 O 2 led to a dramatic drop in the yield (run 12). The lower efficiency of H 2 O 2 compared to TBHP can be explained by the fact that H 2 O 2 can not only initiate free-radical reactions but can also be an inhibitor via the formation of HOO• radicals [92][93][94]. The use of other organic peroxides, such as meta-chloroperoxybenzoic acid (m-CPBA, run 13), cumene hydroperoxide (run 14) and dicumyl peroxide (run 15) led to low yields or did not provide the product at all. Dibenzoylperoxide (BzOOBz, run 16) showed a yield comparable to TBHP, but the formation of a large amount of benzoic acid, which is poorly soluble in the system, complicates the isolation of the products and limits the scalability of the procedure. Therefore, TBHP was chosen as the optimal oxidant. The standard version of the Minisci reaction often uses inorganic persulfates as oxidants. In our system, the use of persulfates was less efficient than TBHP, and led to a significant drop in yield with increasing reaction time, presumably due to the overoxidation of the product (runs [17][18][19][20]. An inert atmosphere did not increase the selectivity of the process (run 21), so we decided to carry out the reaction under air. Table 1. Influence of photocatalytic system composition, irradiation power, and nature of oxidant on the conversion of 4-methylquinoline 1a and yield of 3aa in photocatalytic Minisci reaction.

Optimization of Photocatalytic System Composition
Based on our previous work [20], TiO2 with high specific surface area (anatase nanopowder, Hombikat UV100) and industrially available N-hydroxyphthalimide were chosen as the components of the photochemical system. Blue LEDs (455 nm) with an input power of 10 W were used as light sources. In the first step, we optimized the conditions of the photochemical cross-dehydrogenative Minisci reaction between 4-methylquinoline 1a and tetrahydrofuran 2a (Table 1). Tert-butyl hydroperoxide (TBHP) was used as an inexpensive, easily available and metal-free oxidant. Table 1. Influence of photocatalytic system composition, irradiation power, and nature of oxidant on the conversion of 4-methylquinoline 1a and yield of 3aa in photocatalytic Minisci reaction.

Run
Changes Argon atmosphere 44 39 a The conversion of 1a and the yield of 3aa were determined by 1 H NMR using C 2 H 2 Cl 4 as an internal standard. b instead of TBHP. c 1 mL of water was used as co-solvent to dissolve the persulfate.
In the next step, we optimized the NHPI/TiO 2 /TBHP ratio and irradiation time to achieve the maximum yield of the coupling product 3aa (Table 2).
Increasing the amount of TiO 2 increases the yield of 3aa (runs 1-4). However, when switching from the TiO 2 loading of 20 mg to 40 mg, the efficiency increased only slightly. Therefore, the TiO 2 loading of 20 mg was chosen as the optimal amount. Similarly, large loadings of NHPI resulted in an increase in the 3aa yield (runs 5-8), but the step from 20 to 40 mol.% of NHPI increased the yield of 3aa slightly, and a slight drop in selectivity was observed. The optimum excess of THBP was 4 mmol per 1 mmol of 1a (runs 9-11). The reaction proceeded with almost complete conversion in 8 h (run 15). It should be noted that visible-light-active heterogeneous photocatalyst g-C 3 N 4 was ineffective for the model coupling reaction under the same conditions (run 16). The conditions of experiment 15 were chosen as optimal for further studies of the substrate scope for the developed method.  a The conversion of 1a and the yield of 3aa were determined by 1 H NMR using C2H2Cl4 as an internal standard. b Bulk g-C3N4 (20 mg) was used instead of TiO2 as heterogeneous photocatalyst.
Increasing the amount of TiO2 increases the yield of 3aa (runs 1-4). However, when switching from the TiO2 loading of 20 mg to 40 mg, the efficiency increased only slightly. Therefore, the TiO2 loading of 20 mg was chosen as the optimal amount. Similarly, large loadings of NHPI resulted in an increase in the 3aa yield (runs 5-8), but the step from 20 to 40 mol.% of NHPI increased the yield of 3aa slightly, and a slight drop in selectivity was observed. The optimum excess of THBP was 4 mmol per 1 mmol of 1a (runs 9-11). The reaction proceeded with almost complete conversion in 8 h (run 15). It should be noted that visible-light-active heterogeneous photocatalyst g-C3N4 was ineffective for the model coupling reaction under the same conditions (run 16). The conditions of experiment 15 were chosen as optimal for further studies of the substrate scope for the developed method.

Application of the Designed Photocatalytic NHPI/TiO2 System to the Minisci Reaction
With the optimal conditions in hand ( Table 2, run 15), we have synthesized a wide range of coupling products between N-heterocycles and ethers. The scope of ethers was explored first (Scheme 2). For substrates demonstrating lower conversions compared to 1a, the reaction time increased in some cases up to 48 h (the reaction times and conversions are given in Scheme 2). a The conversion of 1a and the yield of 3aa were determined by 1 H NMR using C 2 H 2 Cl 4 as an internal standard. b Bulk g-C 3 N 4 (20 mg) was used instead of TiO 2 as heterogeneous photocatalyst.

Application of the Designed Photocatalytic NHPI/TiO 2 System to the Minisci Reaction
With the optimal conditions in hand ( Table 2, run 15), we have synthesized a wide range of coupling products between N-heterocycles and ethers. The scope of ethers was explored first (Scheme 2). For substrates demonstrating lower conversions compared to 1a, the reaction time increased in some cases up to 48 h (the reaction times and conversions are given in Scheme 2).
Among the tested ethers, we obtained the best result with THF: after 8 h of reaction, the almost complete conversion of 4-methylquinoline 1a and a high yield of product 3aa (89%) were observed. As a rule, the reaction proceeds more slowly and with lower selectivity for other ethers. In the reaction of 4-methylquinoline with 2-methyltetrahydrofuran 2b, a mixture of products 3ab (as a diastereomeric mixture, major) and 3ab' (minor) was observed. The observed regioselectivity can be explained by the fact that although the hydrogen atom abstraction is most favored from the weakest tertiary CH-bond (position 2 of 2-methyltetrahydrofuran) [95], the resulting C-centered radical is more stable and sterically hindered than the secondary radical and reacts less efficiently with 4-methylquinoline. For 1,3-dioxolane 2c, two isomeric products 3ac and 3ac' were formed, and the major product 3ac corresponds to the breaking of the weakest C2-H bond in 1,3-dioxolane. With dioxane and tetrahydropyran, the reaction proceeded more slowly, but with a longer reaction time, its selectivity decreased simultaneously with an increase in conversion. With glyme, the dehydrogenative coupling product was not observed even after 24 h of reaction.
In the case of diethyl ether as a substrate, the reaction under the standard conditions was not effective due to the immiscibility of Et 2 O and H 2 O contained in TBHP (70% aq.), which led to the aggregation of TiO 2 particles in water droplets and the low conversion of 1a. The solution to the problem was the use of anhydrous TBHP, prepared before the reaction (See experimental details for Scheme 2). The same problem limited the reaction time for the coupling of 1a with Et 2 O since the water generated during TBHP reduction accumulated in the reaction mixture and made the TiO 2 suspension unstable. Among the tested ethers, we obtained the best result with THF: after 8 h of reaction, the almost complete conversion of 4-methylquinoline 1a and a high yield of product 3aa (89%) were observed. As a rule, the reaction proceeds more slowly and with lower selectivity for other ethers. In the reaction of 4-methylquinoline with 2-methyltetrahydrofuran 2b, a mixture of products 3ab (as a diastereomeric mixture, major) and 3ab' (minor) was observed. The observed regioselectivity can be explained by the fact that although the hydrogen atom abstraction is most favored from the weakest tertiary CH-bond (position 2 of 2-methyltetrahydrofuran) [95], the resulting C-centered radical is more stable and sterically hindered than the secondary radical and reacts less efficiently with 4methylquinoline. For 1,3-dioxolane 2c, two isomeric products 3ac and 3ac' were formed, and the major product 3ac corresponds to the breaking of the weakest C2-H bond in 1,3dioxolane. With dioxane and tetrahydropyran, the reaction proceeded more slowly, but with a longer reaction time, its selectivity decreased simultaneously with an increase in conversion. With glyme, the dehydrogenative coupling product was not observed even after 24 h of reaction.
In the case of diethyl ether as a substrate, the reaction under the standard conditions was not effective due to the immiscibility of Et2O and H2O contained in TBHP (70% aq.), which led to the aggregation of TiO2 particles in water droplets and the low conversion of 1a. The solution to the problem was the use of anhydrous TBHP, prepared before the reaction (See experimental details for Scheme 2). The same problem limited the reaction time for the coupling of 1a with Et2O since the water generated during TBHP reduction accumulated in the reaction mixture and made the TiO2 suspension unstable.
In the next step, the scope of the electron-deficient N-heterocycles was tested (Scheme 3).

Scheme 2. Scope of ethers for the photocatalytic Minisci reaction with 4-methylquinoline 1a.
In the next step, the scope of the electron-deficient N-heterocycles was tested (Scheme 3). N-heterocycles with electron-donor groups reacted slower compared to substrates with electron-withdrawing groups, but at the same time, higher selectivity was observed (products 3ba, 3ea in comparison with 3ca). The reaction is sensitive to steric hindrance: 2-chloro-5-bromoquinoline 2d did not yield the target product of 3da, presumably due to the presence of a bulky Br substituent near the 4th position of the quinoline. Our photochemical system is also applicable to quinoxalines and pyrazines. It is worth noting that the products of 3ga and 3ha have not been previously reported (See Supplementary Materials for additional information). In general, the reaction is inefficient for pyridines with no substituents or with electron-donor substituents (pyridine, picolines, lutidine), but good yields have been obtained for pyridines with electron-acceptor substituents, such as pyridine-3-carboxylic acid methyl ester (product 3ia). 4-Methylquinoline-N-oxide reacted with the preservation of the N-oxide function (product 3ja). Good yields have also been obtained in the reaction with isoquinoline (product 3ka). In the reaction with imidazo [1,2-a]pyridine 2l, it was only possible to isolate the product of deep oxidation with the destruction of the ring-3la'. It should also be noted that the addition of acid (TFA) afforded increased yields in some cases (products 3ba, 3ca, 3ea, 3ga, 3ha,3ja and 3ka).
It turned out that carrying out the reaction to complete the conversion of π-deficient arenes in the NHPI/TiO 2 photochemical system leads to a sharp drop in selectivity for target product 3. We assumed that product 3 could undergo further oxidation under the reaction conditions. To find out what role the individual components of the system play in oxidation, we performed control experiments in which the pure reaction product 3aa was placed under standard reaction conditions or irradiated in an inert atmosphere in the absence of NHPI or TBHP (Scheme 4). N-heterocycles with electron-donor groups reacted slower compared to substrates with electron-withdrawing groups, but at the same time, higher selectivity was observed (products 3ba, 3ea in comparison with 3ca). The reaction is sensitive to steric hindrance: 2-chloro-5-bromoquinoline 2d did not yield the target product of 3da, presumably due to the presence of a bulky Br substituent near the 4th position of the quinoline. Our photochemical system is also applicable to quinoxalines and pyrazines. It is worth noting that the products of 3ga and 3ha have not been previously reported (See Supplementary Materials for additional information). In general, the reaction is inefficient for pyridines with no substituents or with electron-donor substituents (pyridine, picolines, lutidine), but good yields have been obtained for pyridines with electron-acceptor substituents, such as pyridine-3-carboxylic acid methyl ester (product 3ia). 4-Methylquinoline-N-oxide reacted with the preservation of the N-oxide function (product 3ja). Good yields have also been obtained in the reaction with isoquinoline (product 3ka). In the reaction with imidazo [1,2a]pyridine 2l, it was only possible to isolate the product of deep oxidation with the destruction of the ring-3la'. It should also be noted that the addition of acid (TFA) afforded increased yields in some cases (products 3ba, 3ca, 3ea, 3ga, 3ha,3ja and 3ka).
It turned out that carrying out the reaction to complete the conversion of π-deficient arenes in the NHPI/TiO2 photochemical system leads to a sharp drop in selectivity for target product 3. We assumed that product 3 could undergo further oxidation under the reaction conditions. To find out what role the individual components of the system play  Under the standard conditions, an 86% conversion of 3aa was observed in 8 h (Scheme 4, A). In the absence of TBHP under an air atmosphere, the product is also oxidized (88% conversion, Scheme 4, B), which suggests that a significant role in the decomposition of the product is played by air as an oxidant. The primary oxidation product was hydroperoxide 3aa', which was detected in a mixture of oxidation products by 13 C NMR and was confirmed by HRMS (See Supplementary Materials). The 13 C signal with chemical shift typical for geminal alkoxyhydroperoxide fragment was observed [96]. However, carrying out the reaction under an argon atmosphere (Scheme 4, C) does not completely suppress the oxidation of product 3aa since TBHP or residual amounts of oxygen can Under the standard conditions, an 86% conversion of 3aa was observed in 8 h (Scheme 4, A). In the absence of TBHP under an air atmosphere, the product is also oxidized (88% conversion, Scheme 4, B), which suggests that a significant role in the decomposition of the product is played by air as an oxidant. The primary oxidation product was hydroperoxide 3aa', which was detected in a mixture of oxidation products by 13 C NMR and was confirmed by HRMS (See Supplementary Materials). The 13 C signal with chemical shift typical for geminal alkoxyhydroperoxide fragment was observed [96]. However, carrying out the reaction under an argon atmosphere (Scheme 4, C) does not completely suppress the oxidation of product 3aa since TBHP or residual amounts of oxygen can serve as oxidants. The lowest conversion of the product was observed when the reaction was carried out in an argon atmosphere without the addition of NHPI (Scheme 4, D), implying that NHPI-derived PINO radicals play an important role in 3aa oxidation.
Based on the collected data, we proposed the following mechanism (Scheme 5). Upon irradiation with visible light, PINO radicals are generated from NHPI on the TiO 2 surface. Simultaneously, the tert-butyl hydroperoxide decomposes on the TiO 2 surface with the formation of tert-butoxyl radicals. Tert-butoxyl radicals can regenerate PINO by abstracting a hydrogen atom from the NHPI in solution [59]. Tert-butoxyl radicals can also generate tertbutylperoxy radicals from t-BuOOH [97,98]. Either tert-butoxy, tert-butylperoxy [99][100][101], or PINO radicals [59,95] can abstract a hydrogen atom from the α-CH bond in ether to form C-centered radical A. However, considering the fact that no cross-dehydrogenative coupling was observed without the addition of NHPI, the main role in H-atom abstraction is assumed to be played by the PINO radicals. Then, radical A undergoes addition to a heteroarene with the formation of the intermediate radical B, which is further subjected to HAT with the retrieval of aromaticity.
Experimental details for Table 2 4-methylquinoline 1a (1 mmol, 143. If needed, another 4 mmol of the reaction t-BuOOH was added, and the reaction mixture was irradiated for another 8 h. At the end of the required time, the reaction mixture was poured into 20 mL of water and extracted with 3×15 mL of CH 2 Cl 2 . The combined organic extracts were washed with 2 × 20 mL of NaHCO 3 saturated solution. The extracts were dried over MgSO 4, and the solvent was evaporated in a vacuum membrane pump. The residue was purified using column chromatography to afford products 3aa-3ka. For the reaction of 1a with Et 2 O, anhydrous t-BuOOH was prepared. t-BuOOH 70% aq. (12 mmol, 1545 mg) was extracted with CH 2 Cl 2 (10 mL). The organic layer was dried over MgSO 4 , and the solvent was rotary evaporated. The obtained anhydrous t-BuOOH was used instead of t-BuOOH 70% aq. For the longer reaction times, the new portion of anhydrous t-BuOOH (4 mmol, 360 mg) was added each 8 h.

Conclusions
In this work, a new visible-light active heterogeneous photocatalyst system based on industrially available and non-toxic TiO 2 and NHPI was proposed for the cross-dehydrogenative C-C coupling of electron-deficient N-heterocycles with ethers. In this photocatalytic system, phthalimide-N-oxyl radicals photogenerated on the surface of titanium oxide become active mediators of the reaction, which leads to 1) an increase in efficiency due to the homogeneous organocatalytic process in solution and 2) allows the selective cleavage of the weak CH bonds. We have proposed a new mild method for the generation of C-centered radicals from non-activated esters for the Minisci reaction. Despite the fact that acidic additives are frequently used in Minisci-type reactions, the addition of acid was not necessary in our procedure in the case of several substrates. Optimal conditions were chosen for the Minisci reaction between π-deficient pyridine, quinoline, pyrazine, and quinoxaline heteroarenes with non-activated ethers.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/molecules28030934/s1, copies of NMR spectra of the synthesized products, the comparison of the developed method with the literature procedure, the determination of the side products of the studied reaction.