Photocatalytically Active Mechanically Interlocked Carbon Nanotubes

ABSTRACT

This work presents the heterogenization of phenothiazine‐based photocatalysts onto carbon nanotubes (CNTs) through mechanical interlocking. Two different photocatalysts featuring a phenothiazine core were synthesized. Both derivatives were successfully clipped onto single‐walled carbon nanotubes (SWNTs) of different diameters by means of ring‐closing metathesis. Comprehensive characterization of the resulting mechanically interlocked carbon nanotubes (MINTs) bearing the photocatalysts, using spectroscopic, microscopic, and computational techniques, confirms the effectiveness of the procedure, achieving a functionalization degree of 10%–14% with photoactive materials. These photocatalyst‐MINT hybrids were subsequently evaluated in the photocatalytic borylation of aryl halides. Both MINT materials exhibited similar or improved performance compared to their homogeneous counterparts, while also demonstrating enhanced substrate tolerance, recyclability, and robustness. Furthermore, the mechanism of the photocatalytic reaction mediated by the MINT materials was investigated through both experimental and computational studies. The results highlight the importance of electronic communication between the nanotube and the molecular photocatalyst for achieving the observed catalytic behavior.

Mechanically interlocked CNTs (MINTs) are studied as heterogeneous photocatalysts. Two different motifs are presented, which own enhanced activity versus homogeneous systems at very low loadings, high TON, excellent recyclability, varied scope and no leaching. Efficient CNT–photocatalyst interaction, demonstrated both theoretically and experimentally, enables efficient charge transfer for the visible‐light‐driven borylation of aryl halides.

1. Introduction

Single Walled carbon NanoTubes (SWNTs) are sp² carbon allotropes constituted of an individual rolled up graphene sheet that leads to a nanocylinder structure [1, 2]. Discovered by Iijima in 1993 [3], SWNTs present incredible properties in terms of electron and thermal conductivity, mechanical strength and robustness, and high surface area too [4, 5, 6, 7]. In the chemical realm, this set of properties is exploited, among other applications, for the development of heterogeneous catalysts [8]. Indeed, the intrinsic SWNTs electronic conductivity is able to assist in redox reactions [9], and pristine SWNTs are able to trigger chemical reactions where electron transfer steps are required, such as oxidations [10], oxidative dehydrogenation reactions [11], cycloadditions or electrocatalytic processes (Scheme 1a) [12, 13]. Nevertheless, the individual tube aggregation and the absence of active sites in those pristine structures strongly hinder the widespread catalytic character of SWNTs [14]. These problems shall be amended by functionalization procedures [15], such as, among others, the formation of Mechanical Interlocked NanoTubes (MINTs) [16]. MINTs are based on organic macrocycles or precursors that are threaded around the axis of an individual SWNT in a rotaxane configuration, usually by clipping the terminal functional groups of a motif able to interact by supramolecular π interactions with the outer surface of the nanotube (U‐shaped molecules) rather than tossing the macrocycle directly across the tube end. Typically, the clipping reactions employed for MINTs synthesis include ring closing metathesis (RCM) [17], thiol oxidation [18], coordination chemistry or hydrogen bonding [19, 20]. As a result, a decorated nanotube is obtained, where the ideally infinite structure of SWNT prevents unthreading events, which provides stability to the whole system because the mechanical interlocking can be considered a combination of supramolecular and covalent functionalizations [21]. As a result, functionalized and more detached nanotubes arise. Their new properties will be a function of the size and chemical entities present in the macrocycle and the electronic communication among partners. Thus, MINTs have found applications as polymer reinforcers [22, 23, 24, 25], as artificial olfactory systems [26], as spintronic devices or as catalysts in (electro)chemical processes, such as the reduction of nitrocompounds with allosteric control as a function of the electronic character of the macrocycle, the oxygen reduction reaction and the Knoevenagel condensation using N‐functionalized and therefore basic MINTs (Scheme 1b) [27, 28, 29, 30, 31]. Nevertheless, even considering that MINTs have been reported with photoactive moieties as pyrenes [17], perylene‐diimides or porphyrins [32, 33], to the best of our knowledge, a MINT has never been used as photocatalyst in organic synthesis. Therefore, the design and application of a photocatalytic MINT would be highly desirable for the development of novel hybrid catalysts, compared with those synthesized so far using more traditional strategies based on supramolecular or covalent functionalization.

SCHEME 1.

(a) Different examples of SWNTs employed in catalysis; (b) MINTs employed in catalysis compared to (c) This work: photocatalytic application of MINTs in organic synthesis.

This work proposes the synthesis of photocatalytic MINTs by taking advantage of the aromatic phenyl–phenothiazine (PTH) core (Scheme 1c), which is a well‐known and efficient organic photocatalyst. At the same time, photocatalytic MINTs shall provide an answer for the typical self‐aggregation of this organic photocatalyst in solution that causes autoquenching phenomena [34, 35]. This photocatalyst exhibits a strong reduction potential (E° (PTH•+/PTH*) = −2.1 V vs standard calomel electrode (SCE); E° (PTH•+/PTH) = 0.68 V vs SCE) mainly employed for the reduction of carbon‐halogen bonds [36]. Thus, modifications to prepare PTH‐based U‐shaped molecules are undertaken for further clipping around SWNTs. To demonstrate the versatility of the clipping strategy, two different molecules have been synthetized to be interlocked with nanotubes of different diameters. On the one hand, the PTH core was expanded with phenoxy rings that coupled the aliphatic arms able to clip with SWNTs of 0.8 nm in diameter. On the other hand, a larger molecule containing two PTH cores linked by a spacer was synthesized and mechanically interlocked with SWNTs of 1.4 nm in diameter. Computational optimization of the molecules facilitated the selection of the nanotubes and confirmed the obtained geometries that matched the characterization of the interlocked hybrids unambiguously. Finally, the materials were tested as photocatalysts for visible‐light‐driven dehalogenation and subsequent borylation of aromatic compounds. In addition, the mechanism of the photocatalytic transformation was supported by computational studies.

2. Results and Discussion

2.1. Synthesis and Characterization

The synthesis of PTH‐based MINTs requires the derivatization of the aromatic phenyl–phenothiazine to introduce functional groups with adequate reactivity to sustain the macrocycle clipping‐threading event over the nanotube. Thus, our study began with the synthesis of the molecules required to perform the mechanical interlocking functionalization of SWNT (see Scheme 2a and Supplementary Information for further details). In order to synthetize a PTH U‐shaped molecule, a Buchwald‐Hartwig amination between phenothiazine and tolyl bromide was accomplished to obtain 1 in 83% yield. Thus, the most reactive position toward NBS‐mediated bromination is blocked [37], affording the dibrominated derivative 2 in 63% yield. The newly reactive positions allowed a double Suzuki coupling, which resulted in the expansion of the PTH core with two additional phenoxy rings. Substrate 3, obtained at 60% yield, was used as a precursor of the U‐shaped molecules. On the one hand, a single Williamson alkylation with 9‐bromo‐non‐1‐ene afforded substrate 4a in 40% yield, where the spacer dibromo‐p‐xylene was then introduced in a second Williamson alkylation to yield the U‐shaped molecule 5 in 18% yield. On the other hand, a double Williamson alkylation drove the synthesis of the U‐shaped molecule 4b at a 40% yield.

SCHEME 2.

(a) Synthesis of U‐shaped molecules; (b) Conformation study performed by DFT calculations.

We carried out quantum chemistry calculations to optimize the geometry of molecules 1, 4b and 5 and to simulate their UV‐Vis spectra (see Scheme 2b and Supplementary Information for details). Molecule 1 adopts a butterfly conformation, where the apical tolyl moiety is disposed perpendicular to the phenothiazine core. This conformation was also detected for molecules 4b and 5, but more importantly, the optimized conformation informs in both cases that an extended configuration without bending caused by eventual intramolecular π stacking is obtained. The extended size of molecule 4b is therefore estimated as 3.3 nm. Molecule 5, with the presence of the aromatic spacer and two PTH cores, has an extension of 5.6 nm.

The size of the molecules becomes critical for the functionalization of the nanotubes. Considering that aromatic π‐π interactions in graphite require 0.34 nm of space [38], the size of molecule 4b fits SWNTs of 0.8 nm in diameter (perimeter = 2.5 nm), and the dimensions of molecule 5 are adequate to interlock SWNTs of 1.4 nm in diameter (perimeter = 4.4 nm). Thus, we mixed the substrate with the corresponding acid‐purified SWNT (Scheme 3) in the presence of the second generation Grubbs catalyst to promote the RCM clipping reaction. Hence, we obtained the resultant MINT‐1 and MINT‐2 samples after profuse washing by filtration to eliminate unreacted and non‐interlocked organic moieties (see experimental part for details). The structures of both interlocked samples proposed in Scheme 3 are fully supported by the computed optimized geometries and experimental data (see below and Supplementary Information). Supramolecular complexes between the U‐shaped molecules and the nanotubes, prepared without subjecting them to the RCM clipping reaction, SUPRA‐1 and SUPRA‐2 (their synthetic details and characterization data can be found in the Supplementary Information) were also prepared for comparative purposes (Scheme 3) [28], and also macrocycle M2 (see Supplementary Information) arising from the homogeneous RCM of molecule 5.

SCHEME 3.

Interlocking routes with molecules 4b and 5 with the corresponding SWNT to yield samples MINT‐1 and MINT‐2, including the synthesis of supramolecular controls SUPRA‐1 and SUPRA‐2.

Pristine nanotubes and MINT materials were characterized by elemental analysis to estimate the degree of functionalization (Table 1). On the one hand, the purified SWNTs used for the interlocking possessed high carbon content but negligible to low amounts of nitrogen and sulfur, which are the heteroatoms used to track the PTH core. Conversely, both interlocked materials showed an increase in the corresponding levels of N and S, indicating that the functionalization has been successful. Since MINT‐2 contains two PTH cores, higher amounts of N and S are detected compared to MINT‐1 (0.4 and 0.5 for MINT‐1, and 1.0 and 0.7 for MINT‐2, respectively). Using these values, the functionalization degree of sample MINT‐1 was estimated as 10%, while sample MINT‐2 was functionalized to a 14% extent. The functionalization degree was cross‐checked by TGA analysis (see Figure S2), where additional weight losses at 400°C compared to the corresponding pristine nanotubes of 9% for sample MINT‐1 and 13% for sample MINT‐2 were observed in the corresponding thermal profiles. The temperature region near 400°C is usually assigned as that in which the organic moieties that functionalize carbon nanomaterials decompose and clearly supports the elemental analysis data [39].

TABLE 1.

Elemental analysis of samples under study, including the functionalization degree.

Sample	C a	H a	N a	S a	Functionalization degree b
o‐SWNT‐0.8	77.5	1.8	0.5	0.1	—
MINT‐1	76.0	3.3	0.9	0.6	10
o‐SWNT‐1.4	83.1	1.1	0.1	0.1	—
MINT‐2	76.6	3.5	1.1	0.8	14

^a Values expressed in %wt.

^b %.

The mechanical interlocking was spectroscopically analyzed by Raman spectroscopy (Figure 1). The Raman spectra of pristine materials showed the typical pattern of SWNT [40]. The most prominent peak was found at ∼1600 cm⁻¹ of Raman shift, and it is assigned as the in‐plane C sp² vibration signal (G band). The defects band (D band), associated with the characteristic sp³ vibration, emerged at 1340 cm⁻¹. Then, the harmonic and second order vibrations are found at higher Raman shift, while the radial breathing mode (RBM), an expansion‐contraction vibration of the whole nanotube skeleton, and directly associated with the tube diameter, appears at ∼200 cm⁻¹. Interestingly, the Raman functionalization marker I_D/I_G ratio of both pristine materials was estimated as 0.09 for o‐SWNT‐0.8 and 0.07 for o‐SWNT‐1.4 [41]. After submitting the nanotubes to the mechanical interlocking with the corresponding U‐shaped molecules, very similar values to the untreated materials were obtained, with mean values of 0.10 for sample MINT‐1 and 0.08 for sample MINT‐2. The conservation of the I_D/I_G ratio after reaction indicates supramolecular functionalization [42], while the unaltered position of the RBM indicates external modification. It is worth mentioning that for both interlocked samples, a blue shift of the G band of ca. 8–10 cm⁻¹ of Raman shift is detected, which can result from the charge transfer injection from the PTH core to the nanotube [43], which may modify the electronics of the hybrid materials compared to the pristine nanotubes by electron doping. Similar effects have been reported before for covalent, encapsulation and interlocked configurations [28, 44, 45]. We can also confirm that overall homogeneous procedures have occurred, since the Raman map of the MINT materials sharply resembles the original nanotube one.

FIGURE 1.

Raman analysis of samples under study. (a) and (b) Optical and Raman map of sample o‐SWNT‐0.8 and MINT‐1, including the mean Raman spectrum of each sample. (c and d) Optical and Raman map of sample o‐SWNT‐1.4 and MINT‐2, including the mean Raman spectrum of each sample.

The synthesis of the MINT materials was monitored employing TEM methods. Figure 2 shows the representative image of each MINT material, accompanied by the in situ collected EDX spectrum. In addition, in the Supplementary Information we show the images of the corresponding pristine SWNT (Figure S1), where smooth, clean and defect‐free nanotubes without objects, disruptions or protrusions are observed, in agreement with previous reports [44, 45, 46, 47]. Figure 2a describes a region of sample MINT‐1, where an individual nanotube bridges two big aggregations of nanotube bundles. While observing any phenomenon inside the bundle is very difficult, the individual nanotube presented a 3.8 nm wide circular object clearly ringing the main axe of the tube. The in situ EDX spectrum informed on the presence of N and S elements in that particular region, which means that the object may correspond to the macrocycle resulting from the RCM reaction of molecule 4b and o‐SWNT‐0.8. A similar trend was observed for MINT‐2. Indeed, in Figure 2b an individual SWNT is found bridging two arrays of aggregated nanotubes with two circular objects threading it. These objects exhibit greater complexity than those discussed previously for sample MINT‐1, as they are larger, with a mean size of 5.0 nm. The chemical analysis performed by in situ EDX also revealed the presence of N and S, which suggests that the RCM of molecule 5 is performed using the o‐SWNT‐1.4 as the template has succeeded. In very close agreement with previous reports [21, 48], this cyclization reaction seems to be favored in the presence of nanotubes to draw closer the reactive groups even considering that the most stable conformation is an extended configuration (see above). In addition, the intimate contact between macrocycle and nanotube might play a role in the modification of the electronics of the hybrid. The absorption of the MINT samples was analyzed by means of UV‐Vis spectroscopy, employing suspensions of the materials in D₂O with 1% of sodium dodecylsulfate as surfactant. With these conditions, the spectrum of the MINT‐1 and MINT‐2 could be fairly recorded (see Figure S3). Indeed, both samples absorbed light from 800 to 250 nm, and besides the characteristic transitions of the nanotubes [49], the features of the PTH motifs could be distinguished in MINT‐1 as red‐shifted tiny shoulders compared to homogeneous molecules 4b, respectively, due to the functionalization (see Figure S24) [50]. Sample MINT‐2 presented a similar profile to molecule 5 (see Figure S27) but individual features were hardly distinguishable in the broad profile of the heterogeneous sample.

FIGURE 2.

TEM images of (a) MINT‐1 and (b) MINT‐2, including digital zoom regions in the marked squares and, in the lower panels, the corresponding in situ EDX spectrum.

Supramolecular complexes SUPRA‐1 and SUPRA‐2 were also characterized. Indeed, similar N and S contents were observed in the elemental analysis of the SUPRA samples compared to the mechanically interlocked materials (Table S1). The Raman analysis also showed the same trend in the bathochromic shift of the G band (Figure S4 and Table S2). In this case, both samples, SUPRA‐1 and SUPRA‐2, also experienced a blue shift of ca. 8–10 cm⁻¹ by comparing with the pristine nanotube, thus indicating that the electronic communication among partners might be established by the close contact provided through the supramolecular configuration [28]. Finally, TEM observations afforded nanotubes surrounded by organic material, but with a different aspect compared to the MINT samples [28], and clear threading events could not be distinguished (Figure S5).

2.2. Photocatalytic Activity

The photocatalytic activity of our MINT materials was tested in the light‐driven borylation of aromatic compounds 6, and compared to homogeneous photocatalysts 1, 4b, 5 and macrocycle M2. Non‐threaded SUPRA‐1 and SUPRA‐2 materials were also studied as control experiments. As a reaction (see Table 2), we studied the homogeneous protocols for the photocatalytic borylation of p‐iodobenzonitrile 6a [51]. Indeed, the reaction was carried out with an excess of diboron bispinacolate (B₂pin₂) and 1,8‐diazabicyclo[5.4.0]undec‐7‐ene (DBU) as the base, under an inert atmosphere in dimethyl sulfoxide (DMSO) at room temperature, and irradiated with a 410 nm blue‐purple LED, corresponding to the region nearest the visible spectrum where the photocatalyst absorbs (see Table S2). Within the selected conditions, catalysts MINT‐1 and MINT‐2 were able to photocatalytic trigger the formation of the borylated product in 73% and 72% yield, respectively (Table 2, entries 1 and 2). The reaction required the presence of a photocatalyst, as neither pristine SWNTs nor the uncatalyzed reaction showed any reactivity toward the formation of the borylated product (Table 2 entries 12 and 13). The performance of the MINT materials was superior to the homogeneous catalysts 1, 4b, 5 and M2 in a 5 mol% catalyst loading scale, with a similar performance in the borylation of substrate 6a. Indeed, both catalysts 1 and 5 gave a 64% yield of 8a in 18 h under light irradiation, while sample 4b reached 67% yield (Table 2 entries 3, 4, 6 and 8). We also carried out the reaction with 4b and 5 using the same catalytic loading as that determined for MINT‐1 and MINT‐2 (0.01 mol%) [52], yielding 8a in 26% and 29%, respectively. M2, even with a much higher catalytic loading than MINT‐2 with similar architecture, presented a poorer behavior with 42% yield at the 5 mol% catalyst loading experiment (Table 2 entry 8). Therefore, the direct comparison indicates that the MINT materials owned a much better performance as photocatalyst of this reaction, possibly due to electronic interactions between the PTH motifs and the nanotube as depicted by the characterization techniques, computational details (see below), and previous reports [28, 44, 45, 46]. Electronic communication was also observed using control materials SUPRA‐1 and SUPRA‐2. These materials exhibited intermediate activity between the MINT materials and homogeneous photocatalyst 1 at comparable PTH loading (50% and 39% yield; Table 2, entries 9 and 10), indicating that the supramolecular configuration slightly enhances the catalytic activity, although catalyst recovery and reuse are not possible due to its supramolecular nature (see below). Nevertheless, the mechanical bond seems to be necessary to effectively enhance the activity compared to the homogeneous catalysts. Hence, the U‐shaped molecules can only stick to the nanotube by labile π interactions, which can be broken in the dynamic reaction media, thus justifying the intermediate activity observed.

TABLE 2.

Model reaction analysis with samples under study. ^a

Entry	Catalyst	Yield % b
1	MINT‐1	73
2	MINT‐2	72
3	1 (5 mol%)	64
4	4b (5 mol%)	67
5	4b (0.01 mol%)	26
6	5 (5 mol%)	64
7	5 (0.01 mol%)	29
8	M2 (5 mol%)	42
9	SUPRA‐1	50
10	SUPRA‐2	39
11	o‐SWNT‐0.8	0 c
12	No catalyst	0 c

^a Reaction conditions: 6a (0.1 mmol), 1 mg of heterogeneous catalyst or different amounts of homogeneous catalysts (0.01 and 5 mol%), 0.2 mmol of DBU and 0.6 mmol of B₂pin₂ in 1 mL of DMSO at 20°C, under inert atmosphere illuminated during 18 h with a 410 nm wavelength LED.

^{b 1}H‐NMR yield determined with MeNO₂ as internal standard.

^c No reaction.

The kinetics of this photocatalytic borylation were monitored by ¹H‐NMR spectroscopy, following the consumption of the starting material 6a (Figure 3a). On the one hand, sigmoidal kinetics were observed in both cases monitoring the activity of catalysts MINT‐1 and MINT‐2. Indeed, iodo derivative 6a began to be consumed after 3 h of illumination, and the borylation product 8a was detected. At this point, the consumption of the starting material increased sharply, reaching ∼20% in 5 h, 30% in 6 h and nearly 80% in 7.5 h. It is worth mentioning, in agreement with the single 18 h run (see Table 2), MINT‐1 exhibited faster reaction rate than MINT‐2, as it achieved higher catalytic values in this region of the profile. After 10 h under light, the consumption rate of 6a decreased, where full consumption was achieved after 24 h reaction time. As a whole, the interlocked catalysts are showing turnover number (TON) values of ca. 730 and 720 for MINT‐1 and MINT‐2, respectively. In addition, the sigmoidal shape of the profile totally matched previous photocatalytic kinetics presented by other types of PTH hybrids [44, 45, 46, 53]. Furthermore, the activity was found comparable or superior to other state‐of‐the‐art catalysts for similar transformations (see benchmark Table S4). On the other hand, homogeneous catalysts 1, 4b and 5, presented a different behavior, closer to pseudo‐first order kinetics and clearly different to the MINT materials (Figure 3a). A 5 mol% loading of catalyst 1 achieved 70% conversion in just 1 h and completed the reaction in 7.5 h. The rates of catalysts 4b and 5, however, were different: 4b converted 37% of 6a in the first hour, whereas 5 converted 85%, indicating faster kinetics. Therefore, MINT‐1 and MINT‐2 exhibited slightly slower performance than the homogeneous catalysts; however, the photoactive motif is heterogenized at lower loadings, and the catalysts display induction periods characteristic of this type of heterogeneous system.

FIGURE 3.

(a) Temporal evolution of the borylation of 6a photocatalyzed by the samples under study. (b) Recycling study on the photoborylation of 6a.

It is also important to evaluate the stability of the MINT materials during the catalytic activity. To discard eventual leaching, “hot filtration” experiments were performed at different levels of conversion (see Figure S8). Therefore, after 6 h of reaction in the presence of MINT‐1 and MINT‐2, the mixture was filtered, and the filtrate was returned to the same reaction conditions under light irradiation [46, 53]. In both cases, no increase in the amount of 8a was observed, indicating that no active species were present in solution during the catalytic run related to leaching events, being consistent with the results in Table 2 too. This observation underlines the heterogeneous character of the MINT photocatalysts and the high stability and robustness of the mechanical bond. This functionalization technique provides stable materials during operation because it prevents the existence of active species in solutions by design. In addition, these two materials were found to be robust systems, capable of being recycled over four consecutive runs without any decrease in the catalytic activity. (Figure 3b), in contrast with the homogeneous catalyst 1 and the SUPRA‐1 and SUPRA‐2 materials that we could not recover after reaction [28]. Therefore, the mechanical interlocking results in a reliable strategy for the development of stable and robust photocatalytic materials. In addition, the MINT materials were analyzed after the catalytic runs by TEM and Raman spectroscopy (see Supplementary Information), finding very similar samples compared to the fresh materials, indicating that the catalytic procedure did not damage the interlocked MINT photocatalysts. Again, the stability of MINT‐1 and MINT‐2 as robust systems is highlighted.

The versatility of the MINT materials in this photocatalytic borylation was evaluated by changing the properties of the substrates, using MINT‐1 as catalyst due to the slightly better photocatalytic performance (Table 3). The analysis of the scope of the reaction began modifying the electronic character of substrate 6 and interestingly, the reaction tolerated aryl iodides with neutral, electron rich and electron poor substituents at para and ortho positions. Indeed, phenyl iodide 6b was borylated in a moderate 50% yield, while o‐tolyl boronpinacolate 8c and p‐tolyl boronpinacolate 8d were obtained in good yield, which means that the electronic character and the relative position of the substituent in the aromatic ring did not rule the outcome of the transformation. Then, p‐aminoiodobencene 6e and p‐methoxyiodobencene 6f were transformed into the corresponding borylated products with similar performance (67% and 45% yield, respectively). Another different electron poor substituent, such as the methyl p‐iodobenzoate 6 g, gave the best result with an 87% yield. In addition, the reaction was also tolerant to heterocyclic substrates, obtaining 4‐boronpinacolate‐1H‐indole 8 g in 77% yield. With aromatic bromides, the obtained yield values were lower because this particular radical is more difficult to be formed. Nevertheless, useful yields with electrowithdrawing groups of 57% yield of 8a and 65% yield of 8 g were achieved. With neutral or electron donating groups, more challenging substrates, catalyst MINT‐1 provided poorer results as example of method limitation (6% yield for 8f and 26% yield for 8 g). Overall, MINT‐1 was a very versatile photocatalyst able to borylate iodides and bromides with very different characteristics.

TABLE 3.

Scope of the reaction.

2.3. Mechanism Investigation

To gain insight into the reaction mechanism, we investigated the photocatalytic borylation promoted by the MINT materials. In homogeneous catalysis, it has been reported that the reaction starts with the generation of the aryl radical by photocatalytic reduction of the C‐I bond (see Figure 4a) [54]. The aryl radical is then responsible for the breaking of the B‐B bond, which can be assisted by OH⁻ or I⁻ directly coordinated to one of the B atoms, to generate the radical anion of the product. The final product is formed either through oxidation by the radical cation of the base—generated during the single‐electron transfer regeneration step of the catalyst—or directly by the radical cation of the catalyst. We set different borylations of 6a varying the conditions in the presence of the MINT materials to confirm the described previous mechanism (see Figure 4b and Supplementary Information). Firstly, the photocatalytic generation of the aryl radical was studied (step II). Hence, dark reaction conditions afforded negligible conversion, indicating that the system requires light to proceed. In addition, the reaction without base did not afford the borylated product, suggesting that the base takes an important role in the catalytic cycle. Then, additives were added to the reaction to modify the reaction conditions (see Figure 4c). In particular, we studied different H‐atom transfer (HAT) agents, such as Ph₃SnH, formic acid and Ph‐SH, and also radical trappers such as TEMPO [55, 56, 57]. For the Ph₃SnH, the reaction evolved to the reduction product instead of the borylated compound. Conversely, the addition of Ph‐SH as well as formic acid, both weaker HAT donors, diminished the amount of borylated product 8a to 44% and 38% yield, respectively, in comparison with the standard run. These observations indicate that the aryl radical was formed via reduction of the C– I bond. In the case of the radical scavenger TEMPO, more aromatic signals were detected (see Supplementary Information), and a mass compatible with the radical‐TEMPO adduct was observed by mass spectrometry (see Supplementary Information). As a whole, we can conclude that the MINT materials are able to form the aryl radical in the standard reaction conditions and therefore replicate the homogeneous borylation.

FIGURE 4.

(a) Proposed mechanism cycle. (b) Mechanistic tests against the proposed cycle. The data is compared to the activity of the homogeneous catalyst 1. (c) Mechanistic proofs.

Simulations have been carried out to provide further information about the photocatalytic activity of the synthesized MINT material and the reaction mechanism (see computational details in the Supplementary Information). The structure of MINT‐1 was optimized by performing DFT simulations imposing periodic boundary conditions along the nanotube using the VASP code (see Scheme 2b) [58, 59, 60, 61, 62]. From the calculated electronic structure, we propose a photoexcitation mechanism (see Figure 5), in which an electron is first excited from the highest occupied electronic state to the lowest unoccupied electronic state of the photocatalyst, separated by about 3 eV‐gap, which lies in the range of the wavelength experimentally employed. Then, the excited electron decays into the π band of the nanotube (see Figure 5b). These simulations were performed using a wave packet propagation approach [63], which allows us to calculate the resonant wave function and thus localize the decay mechanism towards the π orbital of the nanotube. We also computed the lifetime of an excited electron in the lowest unoccupied state located on the photocatalyst. An ultrafast decay of about 30 fs was obtained, showing that the photoexcitation is followed by a very efficient deexcitation mechanism towards the π band of the nanotube. The transfer of the photoexcited electron to the nanotube is essential to increasing the efficiency of the catalytic process. Basically, the electron transfer cross‐section to the reactant is much larger, since the entire surface of the nanotube interacts with it, thus triggering the subsequent steps of the reaction.

FIGURE 5.

(a) Computed photoexcitation mechanism. The electronic density associated to the states involved in the excitation are indicated with arrows. Density of States (DOS) referred to the Fermi level (in eV) showing the electronic structure of MINT‐1. The two bars indicate the energy of the highest occupied electronic state and the lowest unoccupied electronic state in the photocatalyst (b) Deexcitation of the excited electron towards the nanotube π electronic band.

Calculations performed with DFT, including solvent effects [64], also showed that, in accordance with previous reports [54], the aryl radical is responsible for breaking the B‐B bond of the B₂pin₂ 7 reagent (step III). To facilitate the breaking, an OH⁻ moiety must bond to the B, because the reaction performed without base, as discussed above, resulted in negligible conversion towards the borylated product. The simulations shown in Figure 6a demonstrate that the bonding of OH⁻ to one of the boron atoms is a very energetically favorable process. Furthermore, the new structure formed appears more open, with less steric hindrance in the B‐B bond, so that the aryl radical can attack more easily, as shown in Figure 6b, with a very small, almost negligible, barrier and leading to a very stable intermediate, an anion that is oxidized in the final step. In addition, a reaction performed in dry DMSO also resulted in very poor activity (7% yield of 8a, Figure 4b), indicating that the presence of OH⁻ is necessary to drive the reactivity to the target product, thus confirming the computational prediction and the key role of the basic media. Furthermore, the eventual role of the iodine anion (I⁻) was ruled out because we run a reaction in the presence of S₂O₈ ²⁻ which yielded the borylation product in identical fashion than the standard reaction conditions [55]. Also, simulations showed that the binding of I⁻ to B₂pin₂ is an energetically very unfavorable process (see Figure S13), thus ruling out the activation of the B‐B bond by the iodide. The generation of the final product was performed through the corresponding oxidation, which is mediated by the radical cation of the catalyst, (step IV). To confirm it, we studied a reaction using Na₂CO₃ as base, as the result was comparable to the standard conditions (66% yield of 8a). Thus, DBU might only be acting as base, and the catalyst should be able to perform a second redox reaction to close the cycle and get ready for a further catalytic turnover.

FIGURE 6.

Computed energy in two key steps of the proposed mechanism (a) insertion of OH⁻ in the B₂pin₂ and (b) following attack of the aryl radical.

3. Conclusions

We report, for the first time, mechanically interlocked nanotubes (MINTs) as photocatalysts for the light‐driven borylation of aryl halides with B₂pin₂. Using ring‐closing metathesis, U‐shaped molecules bearing the phenyl‐phenothiazine photocatalyst were threaded onto single‐walled carbon nanotubes (SWNTs) of tailored diameters (0.8 and 1.4 nm), yielding samples MINT‐1 and MINT‐2. Different techniques confirmed 10%–14% functionalization while preserving nanotube integrity and doped character. The photoactive macrocycles efficiently harness light along the nanotube surface, enabling superior photocatalytic activity compared to analogous homogeneous systems. The reactions displayed sigmoidal kinetics, broad substrate scope, high recyclability, and yields up to 87%. Mechanistic studies revealed that upon illumination, the MINT catalyst generates aryl radicals, while the base activates the B–B bond, which is cleaved by the radical to form the product and regenerate the catalytic cycle. Enhanced molecular–nanotube communication was key to the observed catalytic efficiency. This work demonstrates mechanical interlocking as a versatile strategy for designing active photocatalysts and provides a platform to study the interplay between molecular and nanotube components.

4. Experimental

4.1. General

All chemicals, including solvents (analytical grade or higher purity was employed) and SWNTs, have been acquired from commercial sources and directly employed without further purification unless otherwise stated. Gas sensitive reactions were performed under N₂ atmosphere and anhydrous conditions using a sieve with a porous size of 3 Å. The synthesis of the organic products under study, products 1–5, M2 and 8, is described in full detail in the Supplementary Information. Purification of these products, unless otherwise stated, was accomplished under column chromatography, using Gelduran Silicagel 60 (40‐63 µm). The photocatalytic experiments, characterization techniques and computational details were also described in detail in the Supplementary Information. Additional and supporting characterization, catalytic and computational data are gathered in the Supplementary Information too.

4.2. Synthesis of MINT Samples

The procedure employed for the synthesis of MINT materials was adapted from the literature [16]. Briefly, 10 mg of carbon nanotubes (CNTs) previously treated (o‐SWNT‐0.8 for MINT‐1 and o‐SWNT‐1.4 for MINT‐2), 5 mg compound 4b for MINT‐1 and 6 mg compound 5 for MINT‐2, and 10 mol% of the second generation Grubbs catalyst were suspended in 40 mL of tetrachloroethane (TCE) in a flask. The reaction was allowed to proceed at room temperature for 3 days. After that, the reaction mixture was diluted in dichloromethane (DCM), sonicated for 10 min and filtrated using a polytetrafluoroethylene membrane filter of 0.45 µm. This procedure was repeated 3 times and, finally, the materials were dried under vacuum, obtaining the corresponding MINT‐1 and MINT‐2 samples.

4.3. Photocatalysis

The typical photocatalytic experiment was performed as follows: A vial was charged with a magnet, 1 mg of the heterogeneous catalyst, 0.1 mmol of 6, 0.2 mmol of 1,8‐diazabicyclo[5.4.0]undec‐7‐ene (DBU) and 0.6 mmol of diboronbispinacolate (B₂pin₂), dissolved in 1 mL of dimethylsulfoxide (DMSO) [51]. 3 freeze‐pump‐thaw‐N₂ cycles were done, and the reaction was carried out over 18 h under 410 nm blue LED illumination with stirring. After this time, the material was filtered and the liquid phase was extracted with DCM and H₂O. The organic phase was dried in the presence of MgSO₄ and concentrated under vacuum. The yield of the reaction was measured by ¹H‐NMR using as an internal pattern MeNO₂ (0.1 mmol, 5.4 µL). The material recovery was done by filtration, and it was washed twice with 20 mL of DCM, twice with 20 mL of THF and finally twice with acetone. Drying under vacuum afforded the recovered material. “Hot filtration” experiments were performed as follows: A general reaction was set, and after 6 h of reaction (∼50% conversion), the reaction was filtered off, the filtrate was returned to the light, and the crude was monitored with NMR spectroscopy [46, 53]. Finally, mechanistic studies were performed under general conditions by adding various amounts of the selected agent.

Author Contributions

Sergio Díaz‐tendero: investigation, formal analysis, writing – review and editing. Andrei G. Borisov: formal analysis, writing – review and editing. José Alemán: conceptualization, writing – review and editing, project administration, funding acquisition, resources. Fernando Aguilar‐galindo: investigation, validation, formal analysis. Paula Blanco‐Caamano: investigation, methodology, formal analysis, data curation, validation, visualization. Matías Blanco: conceptualization, investigation, validation, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting Information

Supporting Information is available from the Wiley Online Library or from the author.

Supporting information

Supporting File: smll74640‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.