Titania-Promoted Carboxylic Acid Alkylations of Alkenes and Cascade Addition–Cyclizations

Photochemical reactions employing TiO2 and carboxylic acids under dry anaerobic conditions led to several types of C–C bond-forming processes with electron-deficient alkenes. The efficiency of alkylation varied appreciably with substituents in the carboxylic acids. The reactions of aryloxyacetic acids with maleimides resulted in a cascade process in which a pyrrolochromene derivative accompanied the alkylated succinimide. The selectivity for one or other of these products could be tuned to some extent by employing the photoredox catalyst under different conditions. Aryloxyacetic acids adapted for intramolecular ring closures by inclusion of 2-alkenyl, 2-aryl, or 2-oximinyl functionality reacted rather poorly. Profiles of reactant consumption and product formation for these systems were obtained by an in situ NMR monitoring technique. An array of different catalyst forms were tested for efficiency and ease of use. The proposed mechanism, involving hole capture at the TiO2 surface by the carboxylates followed by CO2 loss, was supported by EPR spectroscopic evidence of the intermediates. Deuterium labeling indicated that the titania likely donates protons from surface hydroxyl groups as well as supplying electrons and holes, thus acting as both a catalyst and a reaction partner.


■ INTRODUCTION
Carboxylic acids are pervasive in nature and are produced industrially on a large scale. Therefore, methods of employing them in organic preparations have received a lot of attention. Hard UV (λ ≤ 250 nm) irradiations generate carbon-centered radicals but also induce much degradation. 1,2 Conventionally, decarboxylative alkylations have relied on preparations of unappealing precursors such as peroxides, 3 Barton esters,4 or Hunsdieker salts. 5 Recently attention shifted toward discovering catalytic systems capable of utilizing them in target-oriented syntheses. Decarboxylative copper-catalyzed reactions of carboxylic acids with carbonyls 6 and with aldimines, 7 and of amino acids with alkynes 8 have been reported. Photoredox methods have also been used for this purpose 9 but can suffer from difficulties in removing the catalyst. Interest in heterogeneous photoredox methods for carboxylates dates back to the 1970s and the photo-Kolbéreaction. 10 Semiconductor photoredox catalysis (SCPC) offers tangible advantages for organic preparations. The chosen semiconductor can be nontoxic and can easily be removed by filtration, and when visible or soft UVA light is used, the procedure can be convenient and benign. Most applications of TiO 2 SCPC have utilized aerobic conditions, resulting in selective oxidations of organic substrates 11 or more complete substrate mineralizations. 12 Under anhydrous, anaerobic conditions and with certain substrates, TiO 2 SCPC can lead to the generation of specific radicals adapted for molecular assembly applications. To date, however, only a modest number of such processes have been identified and developed. 13 Notably, SCPC additions of enol ethers to various acceptors have been investigated, 14 and successful additions of tertiary amines to electron-deficient alkenes, some bearing chiral auxiliaries, have been described. 15 We recently discovered that under dry, anaerobic conditions, TiO 2 mediation with carboxylic acid precursors could result in carbon−carbon bond-forming processes. Alkylations and annulations were achieved when photolyses of certain acid precursors were carried out in the presence of suitable electrondeficient alkenes. 16 This paper reports advances in crucial aspects of this process in the following areas: (1) structural features in the carboxylic acids that are necessary and sufficient; (2) the range of functionality in the alkene acceptors that can be tolerated; (3) how the reactions respond to modifications to the catalyst form and the catalyst support; and (4) the mechanism followed by the reaction.

■ RESULTS AND DISCUSSION
In our preliminary study, 16 we discovered an experimental protocol with reasonably wide applicability utilizing Degussa (Evonik) P25 as a 1−5 mg mL −1 dispersion in dried acetonitrile. This dispersion, with added carboxylic acid and substrate, in an oven-dried Pyrex tube was purged with argon and then irradiated by two face-to-face sunlamp arrays (UVA) at ambient temperature. After photolysis, the catalyst was removed by filtration and the products were isolated by conventional organic techniques.
1. Alkylations of Maleimide with Aliphatic Carboxylic Acids. During initial investigations, N-phenylmaleimide (2) was found to be an ideal radical acceptor in this system, as it is symmetrical and photostable, its photodimerization is negligible (under our experimental conditions), and its electron-withdrawing character favors addition by weakly nucleophilic radicals. This alkene was therefore chosen to test the applicabliity of our process with simple aliphatic carboxylic acids 1. Standard unoptimized conditions were adopted so a clear comparison of behavior from acid to acid could be obtained (Table 1). In all instances, 0.1 mmol of 1 and 0.2 mmol of 2 were photolyzed in a suspension of the P25 catalyst (12 mg, 0.15 mmol) in CH 3 CN (12 mL) for 18 h. The reactions did indeed lead to the formation of adducts 3 incorporating the R moiety of the carboxylic acid and an additional H atom, together with significant amounts of succinimide 4 (Scheme 1).
For acids that decarboxylated to give methyl (entry 1) or primary radicals (entries 2 and 3), the yields were low; however, increased yields were obtained with secondary and tertiary radicals (entries 4 and 5). Negligible adduct was detected for the destabilized, σ-type CF 3 radical (entry 6).
The efficiency of the process evidently increased as the stabilization energy of the released radical increased (see the Supporting Information for a plot of radical stabilization energy vs yield). The observation of a significant amount of 4 in each case suggested that alkene 2 plays the role of an electron sink, consuming electrons in two sequential reduction−protonations.
2. Additions of Stabilized Radicals to N-Phenylmaleimide. We next examined a set of addition reactions with N-phenylmaleimide involving more diverse carboxylic acids (Scheme 2). Primary alkyl radicals stabilized by α-alkoxy groups proved to be a good deal more reactive than the simple alkyl radicals. Methoxy-and tert-butoxyacetic acids gave the corresponding adducts (3g and 3h) in reasonable yields of 54% and 57%, respectively, while 2-tetrahydrofuroic acid furnished a 1:1 mixture of the two diastereomers of 3i in a very pleasing yield of 75%. However, methylthioacetic acid returned only a disappointing 34% yield of the desired product 3j. Moderate and poor yields were obtained with benzyl (57%) and 2thienylmethyl acids (22%), but these adducts (3k and 3l) were accompanied by the dimers bibenzyl (18%) and 1,2-bis-(thiophen-2-yl)ethane (27%), respectively.
Decarboxylative additions also worked with Boc-protected αamino acids, and the bottom row in Scheme 2 displays the adducts and the useful yields obtained. While Boc-proline was transformed into adduct 3n in a fine yield of 75%, Boc-alanine and the unnatural α-amino acid Boc-piperidinecarboxylic acid gave much more modest yields of 38% and 29%, respectively. As expected, these adducts were 1:1 mixtures of stereoisomers. No retention of chirality was observed when optically pure amino acids were employed because the released radicals have planar, or close to planar, configurations at their reactive centers. Surprisingly, vinylacetic, ethynylacetic, and cyanoacetic acids failed to react, even though the released radicals would be strongly stabilized.

Comparison of Catalyst Efficiencies.
To compare the efficiencies of different forms of the titania catalyst, the reaction of phenoxyacetic acid with acrylamide, which was clean and efficient (82% adduct isolated) 16 was chosen. For each photolysis, suboptimum conditions of phenoxyacetic acid (0.1 mmol) and acrylamide (0.2 mmol) with irradiation for 17 h in anhydrous CH 3 CN (12 mL) were employed. Where applicable, TiO 2 (12 mg) was employed in a 1 mg mL −1 dispersion (see the Supporting Information for full details). With P25 itself ( Table 2, entry 1), a 47% yield of the 4-phenoxybutanamide adduct was recorded. With Millenium PC500 titania, having about 6 times the surface area, 17 the yield dropped slightly (entry 2). Reactions were also carried out in Schlenk tubes coated with a fine internal layer of TiO 2 by a sol−gel process (see the Supporting Information for details). These proved to be very efficient and led to the best yield (entry 3). TiO 2 Photospheres consisting of hollow Pyrex beads coated externally with rutile titania 18 delivered lower conversion and adduct yields (entry 4). This was attributed to difficulties in dispersing the Photospheres satisfactorily throughout the reaction flask. Because of their buoyancy, they tended to accumulate at the top, and this was compounded by the fact that the high rate of magnetic stirring used in attempts to overcome this tended to result in their becoming damaged. An isolated yield of 68% was achieved upon scale-up of the reaction in the coated tube (entry 3). Incorporation of 4-tertbutyl and 4-trifluoromethyl substituents in the aryl ring of the acid component furnished the corresponding adducts in similarly pleasing yields of 68% and 66%, respectively. Although the sol−gel-coated tubes gave the best yields and conversions, the coatings tended to detach, so they could only be used about three times. We therefore concluded that for ease of handling with conventional organic techniques, scale-up, and efficient product formation, the orthodox P25 catalyst was the best compromise.
4. Addition−Cyclizations with Aryloxyacetic Acids. The reaction of aryloxyacetic acids (5) with N-substituted maleimides (2) led to N-substituted-3,4-dihydrochromeno- [3,4]pyrrole-1,3-dione derivatives 6, each accompanied by a significant amount of the expected adduct 7 (Scheme 3). Tricyclics 6 presumably resulted from cyclization of the initial adduct radical onto the aryl ring followed by rearomatization. Good to excellent overall yields were obtained under all conditions (Table 3). A moderate amount of succinimide (11− 43% yield) was also formed in each case. Table 3 shows that the process worked well with both electron-releasing and electron-withdrawing substituents in the aromatic ring and for methyl-and phenyl-substituted maleimides. The reactions with just P25 generally gave a slight excess of the chromenopyrrole 6 (entries 1, 4, 6, 8, and 10), except in the case of the CF 3 substituent (entry 12), for which an excess of adduct 7 was observed. The selectivity for chromenopyrroles 6 was improved by using a more dense dispersion of P25 (entries 2, 7, and 11). On the other hand, photolyses in the sol−gel TiO 2 -coated tubes (entries 3, 5, 9, and 13) led to a reversal in selectivity with predominant formation of adduct 7. The possibility to tune the selectivity of reactions in this manner is an advantage of the technique. The 1 H NMR spectra of all 6 had 3 J H−H coupling constants between the two protons at the junction of the pyran and dihydropyrrole rings within the range of 9.2−9.3 Hz. This indicated a cis arrangement, 19 in agreement with the conclusion that the reaction selectively formed the cis isomer in every case. Very few syntheses of dihydrochromeno [3,4]pyrrole-1,3-diones 6 have been reported, 20 but compounds with the corresponding reduced ring system, chromenopyrrolidines, are well-known biologically active species. 21 5. Intramolecular Ring Closures with Acceptor-Functionalized Aryloxyacetic Acids. Three types of acid framework with different acceptor groups were prepared to test the applicability of P25-mediated ring closures. 2-(2-Vinylphenoxy)acetic acids having alkene acceptors with ester (8a), nitrile (8b), and ketone (8c) substituents (Scheme 4) were obtained by treatment of 2-formylphenoxyacetic acid with the appropriate phosphorane (see the Experimental Section). 2-(Biphenyl-2-yloxy)acetic acids 11a−c having aromatic acceptors were prepared by coupling of the appropriate 1,1′-biphenyl-2-ol with methyl bromoacetate and subsequent hydrolysis of the ester with LiOH in MeOH/H 2 O. Acids 13a and 13b having oxime ether acceptor groups were prepared by condensation of 2-formylphenoxyacetic acid with the appropriate alkoxylamine hydrochloride. Reactions with these acids were carried out in two ways. The first used P25 under standard conditions as described above. The second used 1 mL aliquots of acid solutions (10 mM) in CD 3 CN (5 mL) in sol−gel-coated NMR tubes. The tubes were purged with argon and then placed in an overhead stirrer (turned on its side), which was then spun at 250 rpm. The tube was held at a slight angle with respect to horizontal in the "chuck" of the overhead stirrer so that, when rotated, it described a small (ca. 2 cm) circle at its non-chuckheld end so as to agitate the reaction solution. Irradiation was performed using a photoreactor with twelve 8 W black light blue (BLB) UV lamps (λ max = 365 nm) (see the Supporting Information for a graphic). Periodically the tube was removed for NMR analysis. This was repeated until all of the starting material had been consumed or the product concentration plateaued. For acids 8a−c with alkene acceptors, control photolyses in the absence of TiO 2 showed that E/Z isomerization was significant (52−79%). In conventional P25-mediated reactions, 8a−c all gave rather complex product mixtures containing unreacted alkene as a mixture of E and Z isomers (8/9) as well as the 5-exo ring-closed dihydrobenzofurans 10. None of the 4-substituted chroman derivatives from 6-endo cyclizations were detected. Disappointing yields of 10a and 10b were isolated. The acids with aromatic acceptors 11a− c all reacted very poorly under both sets of conditions. NMR and GC−MS analyses showed complex product mixtures containing possible trace amounts of the cyclized 6Hbenzo[c]chromenes 12a−c. The 2,3-dihydrobenzofuran core of 10 is found in many natural products and pharmaceutically relevant molecules and has generated much interest. 22 The reaction profile from irradiation of acid 8a in a sol−gelcoated tube obtained by NMR monitoring (Figure 1) shows the sharp decrease in 8a (the E isomer), which was all consumed in <150 min. The Z isomer 9a initially increased, passed through a maximum at about 25 min, and then steadily decreased as it too was converted to cyclic product 10a. The latter steadily increased, reaching a plateau by about 200 min. Analogous reaction profiles were obtained for the sol−gel tube reactions of 8a−c, 11b, and 13a. No cyclized product was identified in the sol−gel tube reactions of 11a and 11b. A significant yield (28%) of the cyclized product N-(2,3dihydrobenzofuranyl)-O-methylhydroxylamine (14a) was found in the sol−gel tube reaction of oxime ether-functionalized acid 13a (Scheme 4). However, attempts to isolate and fully characterize this alkoxyamine were not successful.
6. Mechanism of the Reaction. Hole capture at the TiO 2 surface by a carboxylate creates the corresponding acyloxyl radical. Decarboxylation of this species furnishes the corresponding alkyl radical. Radicals of this type were directly observed by CW X-band EPR spectroscopy of frozen suspensions of P25 and t-BuCO 2 H in CH 3 CN. 16 Transient t-Bu • radicals and PhOCH 2 • radicals were observed during UV irradiation of the acids t-BuCO 2 H and PhOCH 2 CO 2 H, respectively, with PC-500 23 in fluid PhH at 300 K. 16 In similar EPR experiments with vinylacetic acid and 2,2,2-triphenylacetic acid, we were also able to detect and characterize allyl and triphenylmethyl radicals, respectively (see the Supporting Information). The isotropic character of the solution EPR spectra of all the RXCH 2 • radicals established that in the main they were freely tumbling and not attached to the TiO 2 surface (Scheme 5).
Literature precedents imply that weakly nucleophilic RXCH 2 • radicals should add rapidly to the electron-deficient double bonds of maleimides (and similar acceptors). 24 Most likely the resulting adduct radicals 15 will be converted to  • radicals containing aromatic rings, addition produces electrophilic imidoalkyl radicals 15, for which a competition exists between reduction to adducts 7 or homolytic closure (6endo-trig) onto the aryl ring. As radicals 15 are weakly electrophilic in character, this annulation should be favored by increasing the electron density in the ring and, conversely, disfavored by decreasing it. 26 In agreement with this, the proportion of 6 increased when an electron-releasing tert-butyl group was introduced to the ring, whereas it decreased when an electron-withdrawing trifluoromethyl group was introduced. The resonance-stabilized cyclohexadienyl-type radicals 17 will rearomatize to yield the functionalized chromenes 6. This aromatization could result from hole capture from TiO 2 by the radicals 17 and subsequent proton loss as shown. Alternatively, electron transfer to maleimide might take place, as suggested in related work by Hoffmann. 15b,c Protonation of the resulting maleimide radical anions, followed by further electron capture and protonation steps, would explain the significant yields of succinimides obtained in our reactions. However, as greater amounts of succinimides than cyclized products 6 were generally formed, it is thought that direct reduction of the maleimide acceptor by the conduction band of the P25 also initiates this process.
From the results described in Schemes 1 and 2 and Table 1, it can be deduced that the stability of the initial radicals is an important factor in this system. Carboxylic acids dissociate on the TiO 2 surface to give the corresponding surface-bound carboxylate and a proton. 27 Photolysis with UVA and resultant photoexcitation generates an electron−hole pair, which can migrate through the bulk of the semiconductor to the surface. The TiO 2 surface is extensively hydroxylated, 28 and it is thought that these species act as surface trapping sites for the valence-band holes. 29 Electron transfer from the π system of the carboxylate to a hole trap site furnishes the surface-bound RXCH 2 −CO 2 • radical. We propose that a competition exists between β-scission of this surface-bound species to yield the desired free RXCH 2 • radical and back transfer of the electron, either from the trap site or from trapped conduction-band electrons, to the carboxylate. For acid precursors that generate stabilized RXCH 2 • radicals, such as those depicted in Scheme 2, the decarboxylation step is more favorable, thus explaining the improved yields and conversions recorded. Conversely, for the simple aliphatic acids of Table 1, the radical stabilization energy from the RX groups is minimal, so back transfer of an electron to the TiO 2 will take precedence over the loss of CO 2 . This is illustrated in Scheme 6.
It should be noted that while we believe radical stability to be a key aspect in this system, it is clear that there are other factors at play as well and that these take precedence in deciding the reaction outcome in a few instances. For example, vinylacetic acid would be expected to generate the allyl radical very readily in the presence of photoexcited TiO 2 . While we did observe the allyl radical in our EPR experiments, no products at all were observed in the preparative photolysis involving vinylacetic acid and 2. In this case, other as yet unknown factors at the TiO 2 surface must come into play.
A crucial feature of the process is that in the formation of adducts 3 and 7 a hydrogen atom is gained from some source within the dispersion. This is not the proton lost f rom the aromatic ring because reduced adducts were formed equally well from acids lacking aromatic rings. Several deuterium labeling experiments were carried out to try to identify the source. The reaction of phenoxyacetic acid with acrylamide was again chosen as a test-bed process (Scheme 7) because it was clean and practically quantitative.
When this reaction was carried out in d 3 -CH 3 CN solvent, the isolated adduct was screened for D incorporation by 1 H and 2 H NMR spectroscopy and GC−MS. However, no deuterium could be detected in the 4-phenoxybutanamide (19H). Reaction with d 1 -phenoxyacetic acid also yielded adduct with no detectable D incorporation. Thus, unless the acidic D atom of 5a rapidly and completely exchanges, the source of the additional H atom is neither the acid nor the solvent. The most likely source therefore is the P25 semiconductor. Obviously it is counterintuitive that a metal oxide would supply protons. However, it is well-established that H 2 O molecules and HO groups are attached to the surface of the TiO 2 particles. 28a Drying P25 in a vacuum at 150°C removes a significant amount of the surface H 2 O while leaving the chemically attached HO groups. When P25 dried in this way was used, the yield of 18 was scarcely reduced, suggesting that the OH groups were likely the proton donors. Attempts were made to deuterate the H 2 O and OH groups on the P25 surface by refluxing in degassed D 2 O in a glovebox. However, it was found that back exchange occurred immediately upon exposure to air and rapidly with any moisture traces in solvents or on surfaces. Only partly deuterated P25 (ca. 80% as judged by IR) could be obtained. Definitive experiments were difficult to achieve because of this. When the partly deuterated P25D was photolyzed with the undeuterated acid, the phenoxybutanamide product contained no deuterium. It seemed possible, however, that P25 surface OD groups could have rapidly exchanged with protons from the COOH groups of 5a. Experiments with partly deuterated P25D and 5a-COOD were also carried out, but deuterated 19D was again not found. In view of the partially deuterated nature of the P25D and the ease with which it reverts to the protiated form, making handling difficult, our tentative conclusion is still that the P25 is the source of the H atoms.

■ CONCLUSIONS
We have found that photochemical reactions employing P25 and carboxylic acids under dry anaerobic conditions generate C-centered radicals that can be deployed in several types of C− C bond-forming processes. Simple aliphatic carboxylic acids alkylated N-phenylmaleimide, albeit rather inefficiently. Acids that yielded alkoxyalkyl, alkylthioalkyl, and benzyl radicals alkylated electron-deficient alkenes in moderate to good yields. Boc-protected amino acids furnished aminoalkyl radicals that alkylated N-phenylmaleimide in useful yields. The reactions of aryloxyacetic acids with maleimides resulted in a cascade process in which a pyrrolochromene derivative accompanied the alkylated succinimide. Selectivity for one or other of these products could be tuned to some extent by employing the TiO 2 catalyst under different conditions. Aryloxyacetic acids adapted for intramolecular ring closures by the inclusion of 2-alkenyl, 2aryl, or 2-oximinyl functionality reacted rather poorly, especially the 2-aryl acceptor types.
The mechanism (Scheme 5) is supported by EPR spectroscopic evidence of the intermediates. An intriguing aspect of the process is the identity of the proton donor. Experiments with deuterium-labeled solvent and reactants implied that this was the hydroxyl groups on the surface of the P25. It appears that the titania supplies electrons and holes upon photostimulation but also donates protons from surface hydroxyl groups, thus acting as both a catalyst and a reaction partner.
These photoredox reactions are cleanly, safely, and cheaply carried out in the laboratory, and the heterogeneous catalyst is simply filtered off during workup. Thus, with the exception mentioned above, they constitute a useful synthetic protocol.

■ EXPERIMENTAL SECTION
General Procedure for TiO 2 Photoredox Experiments. To a suspension of semiconductor (1−5 mg mL −1 ) in MeCN (freshly distilled over CaH 2 ) in an oven-dried Pyrex tube were added known amounts of the desired carboxylic acid and alkene. The resulting suspension was degassed by bubbling with argon for 20 min. The mixture was then irradiated with eight 29 cm 15 W Philips Cleo tubes (λ = 350 nm) for the desired reaction time at ambient temperature. Following irradiation, the semiconductor powder was removed by filtration through a Celite pad. The solvent was removed under reduced pressure, and the reaction mixture was purified by chromatography. Yields were determined from the amounts of isolated products and/or from the 1 H NMR spectra by reference to CH 2 Br 2 as an internal standard. Additional general procedures are described in the Supporting Information.
NMR Tube Irradiations. Irradiations were carried out using titania sol−gel-coated NMR tubes. A photoreactor consisting of two lots of six 8 W BLB UV lamps (emission λ max = 365 nm), each arranged in a semicylinder with an aluminum reflector, was used to irradiate the contents of the NMR tube. The two photoreactor semicylinders were brought together to surround the NMR tube (see the Supporting Information for a graphic). For a typical reaction, a stock solution comprising acid (10 mM) in d 3 -CH 3 CN was prepared. This solution was purged with argon for 5 min, and then 1 mL was pipetted into an argon-flushed sol−gel-coated NMR tube. The cap was quickly replaced on the tube and sealed with Parafilm. The loaded NMR tube was placed in an overhead stirrer (turned on its side), which was spun at 250 rpm. The tube was held at a slight angle with respect to horizontal in the chuck of the overhead stirrer so that, when rotated, it described a small (ca. 2 cm) circle at its non-chuck-held end so as to agitate the reaction solution. Once the NMR tube was in place, the irradiation was started, and the tube was periodically removed and analyzed by NMR.
Deuterium Labeling Experiments. These experiments were carried out in a similar fashion to the photoredox experiments outlined above. Following photolysis, the reaction mixtures were scrutinized for deuterium incorporation by 1 H and 2 H NMR spectroscopy and GC− MS. Further details of these experiments are provided in the Supporting Information.