Visible Light Mediated [4+2] Annulation of N-Cyclobutylanilines with Alkynes Catalyzed by Self-Doped Ti³⁺@TiO₂

Abstract

We herein report a visible light mediated heterogeneous [4+2] annulation of N-cyclobutylanilines with alkynes catalyzed by self doped Ti³⁺@TiO₂. The self doped Ti³⁺@TiO₂ is stable under photooxidation conditions, easy to recycle, and can be used multiple times without appreciable loss of activity. Extensive mechanistic studies suggest that the annulation reaction is mediated by singlet oxygen that is generated via photosensitization of oxygen in the air by the self doped Ti³⁺@TiO₂. In contrast, the homogeneous variant catalyzed by a far more expensive iridium complex proceeds under an inert atmosphere, which indicates a different mechanism. The substrate scopes of the two processes are comparable.

Graphical abstract

Introduction

Since Fujishima and Honda’s landmark work on discovering TiO₂’s ability as a photochemical water-splitting catalyst in 1971,^[1] TiO₂ has been developed into a major class of photocatalysts for a variety of applications ranging from water splitting to detoxification of water.^[2] The popularity of TiO₂ as photocatalysts is driven by some of its exceptional features, such as earth abundance, stability, and non-toxicity. Unfortunately, since pure TiO₂ (anatase phase) has a band gap of 3.2 eV, its photoactivation can be achieved only by ultraviolet (UV) light. This shortcoming has certainly limited pure TiO₂’s applications, particularly in catalysis, as functional group compatibility and chemoselectivity under UV light is generally worse than lower-energy visible light.^[3] A number of methods have been developed to narrow the band gap to the level that can be overcome by visible light.^[4] Among them, doping proves very effective for narrowing the band gap by producing extra energy states within the band gap.^[5] Compared to doping by addition of a foreign metal or nonmetal element to TiO₂, self-doping by partial reduction of TiO₂ to generate Ti³⁺ ion or the oxygen vacancy, provides an alternative and potentially better approach for conversion of TiO₂ to a visible light photocatalyst.^[6] Feng’s group has developed a simple, economical, and one-step method for self-doped Ti³⁺@TiO₂.^[7] High-quality, uniform self-doped Ti³⁺@TiO₂ with tunable concentration of Ti³⁺ ion or the oxygen vacancy can be easily obtained. Since the Ti³⁺ ion is buried inside TiO₂, it is stable upon exposure to air and water under light irradiation. The self-doped Ti³⁺@TiO₂ has also been shown highly thermo- and photo-stable. These features are all critically important for catalysis applications. Not surprisingly, the developed self-doped Ti³⁺@TiO₂ materials were used successfully to catalyze water splitting^[8] and CO₂ reduction^[9] under visible light irradiation.

Encouraged by the initial success of self-doped Ti³⁺@TiO₂ as visible light photocatalysts, we started to investigate its use to catalyze organic reactions. TiO₂, which is arguably the most studied semiconductor photocatalyst, is also appealing to synthetic chemists for the same reasons (earth-abundant, stable, and non-toxic) that others find TiO₂ attractive. Equally importantly, TiO₂ is used as a heterogeneous catalyst, which makes its recycle trivial. Moreover, since TiO₂-catalyzed reactions occur at the semiconductor/liquid interface, ample evidence from the literature suggests different reactivity and/or pathways between homogeneous and heterogeneous reactions can occur for the same substrates.^[10] Despite all the advantages possessed by TiO₂, its use as photocatalyst in organic synthesis has remained limited.³ Early work on TiO₂’s synthetic applications mostly focused on UV light irradiation.^{[3a], [11]} Unfortunately, this work was often plagued by overoxidation and poor chemoselectivity, and thus later refocused on wastewater treatment.^[12] Now armed with better understanding about how TiO₂ works in photocatalysis^[13] and availability of a large number of visible-light-active TiO₂^[14], we strongly felt that it might be time to revisit the topic of using TiO₂ to catalyze organic reactions particularly from the angle of visible light photocatalysis.

The [4+2] annulation reaction of cyclobutylanilines with alkynes was chosen as the first model reaction to examine self-doped Ti³⁺@TiO₂’s potential as visible light photocatalyst. We developed the homogeneous variant of this reaction in which an iridium complex was used as photocatalyst in 2015 (Scheme 1).^[15] To the best of our knowledge, this work was the first reported example of this type of reaction. With a relatively unactivated cyclobutyl ring, we were able to successfully apply one-electron photooxidation of cyclobutylanilines to the amine radical cations to open the ring, producing reactive distonic radical cations that underwent the annulation with alkynes. This one-electron oxidation strategy proved to be fruitful and quite general, as a variety of aniline-substituted cyclobutanes successfully underwent the desired opening and thereby broadened the use of cyclobutanes as synthetic building blocks.

Scheme 1.

[4+2] Annulation of N-Cyclobutylanilines with Alkynes.

We questioned whether the iridium complex could be replaced by self-doped Ti³⁺@TiO₂ in the [4+2] annulation reaction. In 1986, the Fox group reported that upon UV irradiation, primary amines were converted to imines on bare TiO₂ powder suspended in oxygenated CH₃CN (Scheme 2).^{[16], [17]} The reaction was proposed to proceed through two one-electron oxidation steps via amine radical cations. The proposed pathway was corroborated by cyclic voltammetry, which revealed two irreversible oxidation waves presumably corresponding to the two oxidations. The necessity of oxygen in the reaction suggested that it played a critical role. In this role, oxygen was believed to act as an electron acceptor and thereby minimize unproductive back electron transfer in which amine radical cations were reduced to the starting amines. O₂ was presumably reduced to superoxide, which is one of reactive oxygen species (ROS). Since O₂ is commonly used as a terminal oxidant in photooxidation by semiconductors and ROS plays a significant role in biology, medicine, and atmospheric chemistry, photogeneration of ROS on the semiconductor surface has been studied quite extensively.^{[12a], [18]} In addition to superoxide, ¹O₂ is another ROS that is often suggested to be produced by semiconductor photocatalysis.^[19] However, detection of ¹O₂ in suspension poses various issues.^[20] Two main pathways have been proposed for the formation of ¹O₂. One is energy transfer between photoexcited semiconductors and oxygen,^[21] the other is hole oxidation of superoxide.^[22] Although both pathways have been supported by experimental evidence, it remains debatable about which pathway is dominant. Additionally, ¹O₂ has been shown to directly oxidize amines, leading to various types of products.^[23] This poses an interesting and significant question in our proposed studies, whether N-cyclobutylanilines are oxidized by the hole or ¹O₂. Furthermore, how self-doped Ti³⁺@TiO₂ excited by visible light affects the pathways is another significant question. Herein we report our findings on the [4+2] annulation of cyclobutylanilines with alkynes catalyzed by self-doped Ti³⁺@TiO₂ including the viability of Ti³⁺@TiO₂ as visible light photocatalyst and the involvement of ¹O₂ in the reaction.

Scheme 2.

Oxidation of Primary Amines to Imines on UV-irradiated TiO₂.

Results and Discussion

We chose cyclobutylaniline 1a and phenylacetylene 2a as representative substrates as well as self-doped Ti³⁺@TiO₂ with 16.7 µmol/g of Ti³⁺(referred to as “T3”) as the photocatalyst initially to optimize the catalyst system. Using the conditions optimized for the homogeneous annulation except the photocatalyst and air rather than a nitrogen atmosphere, a negligible amount (3%) of product 3a was detected (Table 1, entry 1). This was not surprising, since oxidation of MeOH on the TiO₂ surface was known to be facile^[24] and thereby might compromise the desired annulation reaction. Examination of other solvents revealed that DMSO and ^tBuOH were the two best, although the latter was slightly better (entries 2–5). The reason why ^tBuOH was superior to MeOH was probably because that it lacks hydrogen at the carbon alpha to the oxygen atom, which makes the resulting alkoxy radical inert towards TiO₂.^[25] Air or oxygen in the air was found to be essential as the reaction barely proceeded under degassing conditions (entry 6). On the other hand, under an oxygen atmosphere, the [4+2] annulation reaction was completely suppressed, and a new product, endoperoxide was produced instead (entry 7).^[26] It is worth noting that the endoperoxide was formed only in an oxygen atmosphere and was not detected under air. Control studies showed that both light and the catalyst were required for the annulation reaction to occur. Very little reaction was observed in the dark (entry 8). In comparison, without the catalyst, although the majority of cyclobutylaniline 1a was consumed (72% conversion), product 3a was detected in 9% yield (entry 9).

Table 1.

Reaction Optimization

Entry[a]	Condition	Solvent	t [h]	Conversion 1a [%][b]	Yield 3a [%][b]
1	Ti3+@TiO2	MeOH	14	40	3
2	Ti3+@TiO2	CH3NO2	14	28	2
3	Ti3+@TiO2	DMF	14	18	14
4	Ti3+@TiO2	DMSO	14	97	78
5	Ti3+@TiO2	tBuOH	14	100	83
6[c]	Ti3+@TiO2	tBuOH	14	16	8
7[d]	Ti3+@TiO2	tBuOH	14	100	0[e]
8	Ti3+@TiO2, dark	tBuOH	14	5	0
9	No catalyst	tBuOH	14	72	9

^[a]Reaction conditions: 1a (0.2 mmol, 0.1 M in solvent), 2a (0.6 mmol), 4a (2 mg), irradiation with two 18 W LED light bulb at room temperature.

^[b]Measured by GC using dodecane as an internal standard.

^[c]Degassed and refilled with N₂.

^[d]Degassed and refilled with O₂.

^[e]Endoperoxide was isolated in 57% yield.

We anticipated that Ti³⁺’s concentration in self-doped Ti³⁺@TiO₂ would potentially affect TiO₂’s photoactivity as it was shown to enhance visible light absorption.⁷ A series of Ti³⁺@TiO₂ that differs in the concentration of Ti³⁺ were then prepared based on our previously reported methods and subsequently examined using the model reaction (Table 2). We found out that although the Ti³⁺’s concentration greatly affected the formation of 3a, it was not the sole factor determining the catalyst’s reactivity. The yield of 3a increased initially with the increased concentration of Ti³⁺ (entries 1–3). Then, it decreased significantly when self-doped TiO₂ with the highest Ti³⁺ concentration was used (entry 4). Coincidentally, in our previously study of water splitting catalyzed by self-doped Ti³⁺@TiO₂ with various concentrations of Ti³⁺, the optimal TiO₂ didn’t have the highest concentration of Ti³⁺, either.⁸ The optimal catalyst for water splitting turned out to be also the best for the annulation reaction (entry 3). It has been suggested that better crystallinity and less defect favor water splitting, while surface area is generally more important than crystallinity for photodecomposition of organic pollutants.⁸ Our examination of the self doped TiO₂ catalysts suggested that in addition to Ti³⁺’s concentration, other parameters of the forming TiO₂ crystals such as phase, size, crystallinity, BET surface areas, and defect may also affect the Ti³⁺@TiO₂’s photocatalytic activity. However, more mechanistic studies will be required to establish this thesis. For comparison, we examined commercially available TiO₂ (P25) in the annulation reaction (entry 5). It was inferior to all self-doped Ti³⁺@TiO₂ examined, which strengthened our argument for studying self-doped Ti³⁺@TiO₂ to catalyze organic reactions.

Table 2.

Evaluation of self-doped Ti³⁺@TiO₂ with various concentrations of Ti³⁺.

Entry[a]	Catalyst [4]	Concentration of Ti3+ (µmol/g)	t [h]	Conversion 1a [%][b]	Yield 3a [%][b]
1	Ti3+@TiO2 (T1)	0.3	14	70	21
2	Ti3+@TiO2 (T2)	4.5	14	83	39
3	Ti3+@TiO2 (4a) (T3)	16.7	14	100	82
4	Ti3+@TiO2 (T4)	25.5	14	82	45
5	TiO2 (P25)	0	14	72	20

^[a]Reaction conditions: 1a (0.2 mmol, 0.1 M in ^tBuOH), 2a (0.6 mmol), 4 (2 mg) irriadiation with two 18 W LED lightbulb at room temperature.

^[b]Measured by GC using dodecane as an internal standard.

Substrate scope

Using the optimized conditions (entry 5, Table 1), we examined the scope of cyclobutylanilines and alkynes (Chart 1). The annulation catalyzed by T3 displayed the scope similar to the homogeneous variant catalyzed by Ir(ppy)₂(dtbbpy)PF₆, although somewhat lower yields (34–85%) than the latter (42–93%) were generally observed. Also similar to the homogeneous variant, the substituents on the N-aryl group were well tolerated. Electron donating groups (e.g., p-^tBu 3a, p-OTBS 3d, and o-Ph 3e), electron withdrawing substituents (e.g., p-CF₃ 3c), and bulky groups (e.g., o-ⁱPr 3f) were all amenable to the annulation, as were heterocycles such as pyridine (3g and 3h). For the other annulation partner, alkynes, the confinement was that they had to bear with at least one substituent capable of stabilizing radicals (e.g., Ph or CO₂Me). The reactivity pattern displayed by alkynes was consistent with intermolecular addition of alkyl radicals to alkynes, one of the key steps in what we proposed for the Ir-catalyzed [4+2] annulation as well as the one catalyzed by T3 (see Scheme 3). However, diphenylacetylene, which was not reactive in the Ir-catalyzed annulation, surprisingly participated in the annulation catalyzed by self-doped Ti³⁺@TiO₂, providing product 3o in modest yield. This result highlighted the difference of reactivity between homogeneous photocatalysis and semiconductor photocatalysis, as we discussed in the introduction.

Chart 1.

Substrate Scope of the [4+2] Annulation Catalyzed by Self-doped Ti³⁺@TiO₂^a,b

[a] Reaction condition: substrate 1 (0.2 mmol, 0.1 M in ^tBuOH), 2 (0.6 mmol), T3 (2 mg, 10 mol%), irradiation with two 18 W white LED light bulbs. [b] Isolated yield after silica gel chromatograph.

Scheme 3.

Proposed Catalytic Cycle for the [4+2] Annulation Catalyzed by Self-doped Ti³⁺@TiO₂.

Catalyst reuse

Facile separation and repeated use of catalysts are two of the significant strengths that heterogeneous catalysis holds over homogeneous catalysis. To exploit these two strengths in our self doped Ti³⁺@TiO₂ system, we performed the annulation of cyclobutylaniline 1a and phenylacetylene 2a catalyzed by the same batch of T3 for 5 cycles. A representative procedure of recycling T3 for each run was as follows: after 14 h of irradiation, the reaction tube was centrifuged for 5 min; the clear liquid was removed for determination of GC yield of product 3a; the solid residue was washed with ^tBuOH (0.5 mL) three times and then used for the next cycle. The GC yield of 3a for each cycle was held steady (Figure 1), suggesting that there was no appreciable loss of activity for T3 over 5 cycles. The recycling experiments were also conducted for 6 h when the average conversion for 5 cycles was 62%, and similar GC yields of 3a were obtained (see Supporting Information). To further support this thesis, we took TEM images of T3 before and after the five cycles. The size, shape, and morphology of T3 changed very little after the five uses (Figures 2 and 3), further confirming the stability of T3 as photocatalyst.

Figure 1.

Catalyst Recycle.

Figure 2.

TEM Images of Self-doped Ti³⁺@TiO₂ before Reactions.

Figure 3.

TEM Images of Self-doped Ti³⁺@TiO₂ after 5 Runs.

Mechanistic studies

Our working model for the [4+2] annulation catalyzed by self-doped Ti³⁺@TiO₂ was similar to the one we previously proposed for the Ir-catalyzed annulation (Scheme 3).¹⁵ The noticeable changes between the two catalytic cycles were ascribed to the role of O₂ played in the reaction. Singlet oxygen, generated by visible light irradiation in the presence of a photosensitizer such as Rose Bengal or porphyrin derivatives, was shown to oxidize amines to the corresponding amine radical cations, which were subsequently converted to synthetically useful intermediates such as iminium ions or alpha-amino radicals.²³ We anticipated that upon visible light irradiation, self-doped Ti³⁺@TiO₂ sensitizes triplet oxygen to singlet oxygen, which affects the initial oxidation of cyclobutylaniline 4 to amine radical cation 5 in the place of the photoexcited Ir complex. Following the oxidation step, the two cycles shared the same elementary steps, such as ring opening of amine radical cation 5 to distonic radical cation 6, intermolecular addition of distonic radical cation 6 to alkyne, and intramolecular addition of vinyl radical 7 to the iminium ion to form the cyclohexenyl ring. To close the catalytic cycle, we proposed a chain process in which the product radical cation 8 is reduced to the product 9 with concomitant oxidation of cyclobutylaniline 4 to the amine radical cation 5. However, other pathways such as reduction of the product radical cation 8 by superoxide or the electron in the conduction band of TiO₂ might also operate to close the catalytic cycle. We also measured the diffuse reflectance spectra of T3 only, T3 complexed with phenyl acetylene 2a, and T3 complexed with cyclobutylaniline 1a (see SI). Compared to the white LED’s emission spectrum, the excitation of T3 or T3 complexed with phenyl acetylene 2a or cyclobutylaniline 1a by the white LED is very reasonable.

A series of radical intermediates were proposed in the catalytic cycle. In order to lend indirect support of their existence, we decided to chemically probe them. TEMPO, a radical scavenger shut down the reaction (Scheme 4a). So did 1,4-cyclohexadiene, a good hydrogen donor (Scheme 4a). Both data supported the involvement of radical intermediates. Since ¹O₂ was proposed to be the actual oxidant oxidizing cyclobutylanilines and triggering the serial radical events, we envisioned that self-doped Ti³⁺@TiO₂ could be replaced by another ¹O₂ photosensitizer, such as Rose Bengal. Indeed, Rose Bengal catalyzed the annulation rather efficiently under visible light irradiation, and afforded product 3a in 64% yield (Scheme 4b). On the other hand, DABCO, a ¹O₂ quencher,^[27] almost completely inhibited the annulation reaction (Scheme 4a). These two results were consistent with our proposed catalytic cycle involving ¹O₂.

Scheme 4.

Chemical Probes for the Involvement of ¹O₂ and Radical Intermediates.

Although we obtained rather strong evidence to indirectly support the involvement of ¹O₂, the important roles played by ¹O₂ in a variety of fields^{12a, 18, 19} and complexity of its detection particularly on the semiconductor surface²⁰ prompted us to further investigate the formation of ¹O₂. We focused on spectroscopic methods that permit us to detect ¹O₂. Two methods, EPR and SOSG^[28], were chosen because of their robustness, sensitivity, and especially capability for differentiation of ROS. 2,2,6,6-Tetramethylpiperidine (TEMP), a spin trap specifically for detection of ¹O₂, was used in our EPR studies. Two sets of data over a period of 10 min were collected. In the first set, a suspension of T3 and TEMP in ^tBuOH was irradiated by a 8 W LED, and EPR spectra were then recorded after 0 min, 1 min, 5 min, and 10 min. The signals corresponding to 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO), the product of ¹O₂ and TEMP, increased as the time progressed (Figure 4A). In the second set, cyclobutylaniline 1a was added in addition to T3 and TEMP in ^tBuOH. EPR spectra were recorded after 0 min, 1 min, 5 min, and 10 min irradiation by the 8 W LED. The signals of TEMPO initially grew and then depleted (Figure 4B). This observation could be rationalized by the following: the concentration of ¹O₂ grew upon more irradiation and then began to deplete when it started to react with 1a and/or TEMPO reacted with the radical intermediates in the proposed catalytic cycle (Scheme 3). We also performed a control study in which TEMP in ^tBuOH without T3 was irradiated. A lower level of TEMPO was detected, and more importantly, the intensity of the signals changed little over the time (Figure 4C). Overall, the EPR data provided another strong evidence to support the involvement of ¹O₂.

Figure 4.

EPR Study. A). Trapping of ¹O₂ generated from Ti³⁺@TiO₂ by TEMP without cyclobutylaniline 1a. B). Trapping of ¹O₂ generated from Ti³⁺@TiO₂ by TEMP with cyclobutylaniline 1a. C). Control: Trapping of ¹O₂ Generated by Visible Light Irradiation without Ti³⁺@TiO₂.

The commercially available Singlet Oxygen Sensor Green (SOSG),²⁸ which has been popularized by biochemists for detection of ¹O₂ in plants and other biological systems,^[29] is a highly specific molecular probe for ¹O₂ and does not show any noticeable response towards hydroxyl radical or superoxide. Recently, it was applied to detect ¹O₂ generated on photoirradiated surface of surfaced-modified TiO₂ nanoparticles and metal nanoparticles.^[21b],[30] We selected SOSG to qualitatively measure the efficiency of ¹O₂ generation by self-doped Ti³⁺@TiO₂ with various concentrations of Ti³⁺. Four self-doped TiO₂ samples (T1, T2, T3, and T4) were mixed with SOSG separately in 1:1 ^tBuOH : D₂O. The resulting suspensions were irradiated by a 18 W white LED. As expected, strong fluorescence (λ_em = 535 nm) was observed for the four TiO₂ samples (Figure 5). But unfortunately, they all produced fluorescence in similar intensity that could not be differentiated by the fluorometer. We also performed two control studies: one was T3 with SOSG in the dark, while the other was SOSG without the self-doped TiO₂. The ¹O₂ generation was negligible in both studies. Again, although the SOSG studies didn’t meet our goal to differentiate the TiO₂ samples’ efficiency in generating ¹O₂, they nevertheless unequivocally supported its formation.

Figure 5.

Fluorescence Spectra of SOSG in the Presence of Self-Doped Ti³⁺@TiO₂ with light, without light, or in the Absence of Self-Doped Ti³⁺@TiO₂.

Comparison with Rose Bengal

In our mechanistic studies (vide supra), Rose Bengal was shown to catalyze the [4+2] annulation reaction in the place of self-doped Ti³⁺@TiO₂ (T3), affording product 3a in slightly lower yield. Since Rose Bengal is one of the widely used photosensitizers for ¹O₂ generation, to demonstrate that T3 is a more efficient and recyclable alternative to Rose Bengal, we decided to benchmark T3 against the latter in the [4+2] annulation reaction. We performed two annulation reactions, and in both cases, T3 gave the products in higher yields (Scheme 5).

Scheme 5.

Comparison of Self-doped Ti³⁺@TiO₂ against Rose Bengal.

Conclusions

We have successfully developed a heterogeneous [4+2] annulation of cyclobutylanilines with alkynes catalyzed by self-doped Ti³⁺@TiO₂. The substrate scope is comparable to that of the homogeneous variant catalyzed by a far more expensive iridium complex. Self-doped Ti³⁺@TiO₂ has shown excellent stability under the conditions of the annulation reaction. No appreciable loss of activity has been observed after it was reused to catalyze the annulation for 5 consecutive times. Recycling the TiO₂ is trivial. The catalyst is collected by centrifuge, washed by fresh solvent three times, and then used directly in the next reaction. Mechanistically, we have conducted a series of experiments to probe the [4+2] annulation. All the data support that upon visible light irradiation, self-doped Ti³⁺@TiO₂ sensitizes triplet O₂ to singlet O₂, which then oxidizes cyclobutylanilines to the amine radical cations and initializes the serial radical events en route to the annulation product. Interestingly, the most active self-doped TiO₂ doesn’t have the highest concentration of Ti³⁺. Self-doped TiO₂ has been compared favorably against Rose Bengal as ¹O₂ photosensitizer in terms of efficiency and recyclability. More studies on improving self-doped TiO₂’s catalytic activity and developing new reactions are ongoing.

Experimental Section

General Procedure

An oven-dried test tube equipped with a stir bar was charged with Ti³⁺@TiO₂ (2 mg) and N-cyclobutylaniline derivative (0.2 mmol), alkyne derivative (0.6 mmol), and ^tBuOH (2 mL). The test tube was capped with a Teflon screw cap and the reaction mixture was sonicated for 3 min followed by irradiation with two LED (18 watts) positioned 6 cm from the test tube. After the reaction was complete, monitored by TLC, the mixture was diluted with diethyl ether and filtered through a short pad of silica gel. The solution was concentrated and the residue was purified by silica gel flash chromatography to afford the corresponding annulation products.