Methods for the chemical synthesis of carbon nanotubes: an approach based on hemispherical polyarene templates

Synthesis of a (5,5) carbon nanotube end-cap

Figure 1 outlines our original strategy for synthesizing a C₅₀H₁₀ geodesic polyarene (3) from corannulene (1) and for using it as a small hydrocarbon template from which to grow (5,5) CNTs [10]. After considerable experimentation, the 5-fold indenoannulation of corannulene (1) was achieved in three steps, using microwave-induced, Pd-catalyzed, intramolecular arylation reactions to close all the new 5-membered rings (Fig. 2) [11], [12]. Each cyclization reaction introduces additional strain in the molecule and further accentuates the curvature of the hydrocarbon bowl [13]. The X-ray crystal structure of pentaindenocorannulene (2) reveals that the curvature of 2 (POAV angle of the pyramidalized hub carbon atoms=12.1°) exceeds not only that of corannulene (1, hub POAV angle=8.3°) but also even that of the smallest stable fullerene, C₆₀ (POAV angle=11.6°) [11], [12].

Fig. 1:

Strategy for synthesizing (5,5) CNTs from corannulene (1), using an end-cap (3) as the template for determining the diameter and rim structure of the CNT. (a) 5-fold indenoannulation (b) stitching the arms together (c) repetitive benzannulation.

Fig. 2:

(a) ICl, CH₂Cl₂, 25°C, 48 h, 67% yield. (b) Pd₂(dba)₃, 1,3-bis(2,6-diisopropylphenyl)imidazolium chloride, Cs₂CO₃, dioxane, 80°C, 48 h, 48% yield; (c) Pd(PCy₃)₂Cl₂, DBU, DMAc, 180°C (microwave), 45 min, 35% yield.

In principle, it should be possible to stitch together the arms of pentaindenocorannulene (2) by flash vacuum pyrolysis (FVP); however, our years of experience with this technique had taught us the value of incorporating aryl radical precursors (Br or Cl substituents) at the sites of the intended C–C bond formation [14]. Brute force thermal cyclodehydrogenations of unfunctionallized polycyclic aromatic hydrocarbons have been reported, but they rarely give good yields of cyclized products [15], [16]. Consequently, rather than focus on trying to convert 2 to 3, we modified the plan and incorporated five additional chlorine atoms in our synthetic intermediates. In this way, we succeeded in synthesizing the (5,5) CNT end-cap 3 in three steps from corannulene, as shown in Fig. 3 [17], [18]. Attempts to convert intermediate 4 to 3 by Pd-catalyzed, intramolecular arylation reactions (cfFig. 2) gave hydrocarbon products in which up to 6 of the 10 desired bonds had formed, as judged by mass spectral analysis [19], but only FVP closed all of the new rings to give 3.

Fig. 3:

Three-step synthesis of (5,5) CNT end-cap (3) from corannulene (1). (a) B₂pin₂ (5.2 eq), [Ir(OMe)COD] ₂, 4,4′-dimethyl-2,2′-bipyridyl, t-BuOK, THF, 85°C (sealed tube), 4 d, 70% yield. (b) Pd(dppf)Cl₂, Cs₂CO₃, THF, 95°C (sealed tube), 2 days, 70% yield. (c) FVP, 1100°C, 0.25 torr, 3% yield.

The low yield of the FVP step makes it difficult to prepare more than a few milligrams of 3 at a time, but we were able to obtain enough of this C₅₀H₁₀ geodesic polyarene to grow crystals for X-ray analysis [17]. Fortunately, the strategy of growing CNTs from small hydrocarbon templates does not demand large quantities of the templates. Elongating end-cap 3 to a length of even 1 μ would provide gram quantities of pure (5,5) CNTs, and 1 μ does not seem unreasonable, considering that individual CNTs have been grown by other methods to lengths of more than 20 cm! [20] Our efforts to grow long CNTs from template 3 are described in a later section below.

Attempted syntheses of a C_3v (6,6) carbon nanotube end-cap

Figure 4 outlines a strategy we explored to synthesize a C_3v C₆₀H₁₂ geodesic polyarene (6) for use as a small hydrocarbon template from which to grow (6,6) CNTs [21].

Fig. 4:

Attempted synthesis of (6,6) CNT end-cap (6). (a) p-TsOH·H₂O, propionic acid, o-DCB, 180 °C, 79%.

This plan relies on 1,3-peri-shifts of hydrogen atoms (Fig. 5) in the aryl radicals generated by homolysis of the C–Cl bonds closest to the hub in order to initiate closures to the bowl [22]. Spontaneous thermal cyclodehydrogenations analogous to those that enabled our C₆₀ synthesis [23], [24], [25] and radical cyclizations initiated by homolysis of the other three C–Cl bonds would be required to close the remaining new rings. Unfortunately, FVP of the C₆₀H₂₄Cl₆ synthetic precursor (5) at a variety of temperatures and pressures gave only small amounts of the desired C_3v C₆₀H₁₂ geodesic polyarene (6), or none at all, depending on the FVP conditions. From the most encouraging experiments, the ¹H NMR, UV-vis, and mass spectra recorded on the unpurified pyrolysates all showed peaks in good agreement with those calculated for 6, but the low solubilities of the components in the mixture rendered isolation and purification of the CNT end-cap hopeless [22].

Fig. 5:

Thermal 1,3-peri-shift of a hydrogen atom.

An alternative, solution-phase route to (6,6) CNTs from a slightly larger C_3v C₆₆H₂₄ geodesic polyarene (8) is outlined in Fig. 6 [26], [27]. In this approach, all six 5-membered rings of the final target compound (8) are already present in the penultimate intermediate (7). It was anticipated that the challenge of stitching the arms together in this end-cap precursor might be less demanding than in 5 and not require FVP, because 7 already contains much of the pyramidalization strain and molecular curvature of the desired hemispherical polyarene (8). Unfortunately, all attempts to convert 7 to 8 by oxidative cyclodehydrogenation reactions (FeCl₃ or AlCl₃/Cu(OTf)₂ or MoCl₅), even at elevated temperatures, gave back only unchanged starting material and, in some cases, small amounts of partially chlorinated derivatives of 7 [26], [27]. Anionic cyclodehydrogenation attempts (alkali metals or potassium naphthalenide or tetrapotassium corannulenide), even with microwave irradiation, likewise gave mostly unchanged starting material (7), after the reactions were quenched with O₂ [26], [27]. We speculate that the oxidative cyclodehydrogenation reactions may have failed because corannulenes and other curved polyarenes are more difficult to oxidize than planar polyarenes. On the other hand, curved polyarenes are more easily reduced than their planar counterparts. The failure of our anionic cyclodehydrogenation attempts may have resulted from over-reduction of 7 to a stable polyanion. If so, then milder reduction under more controlled conditions (e.g. electrochemical) might have a better chance. Notwithstanding the reasons for our failures to stitch up synthetic intermediates 5 and 7, we never succeeded in preparing a (6,6) CNT end-cap in isolable amounts. Fortunately, we were able to prepare enough of the (5,5) CNT end-cap (3) to move on to the final hurdle.

Fig. 6:

(a) BBr₃, Cl₂CHCHCl₂, 147°C, 6%. (b) FeCl₃ or AlCl₃/Cu(OTf)₂ or MoCl₅. (c) alkali metals or potassium naphthalenide or tetrapotassium corannulenide.

“Masked Acetylene” Diels-Alder cycloaddition/aromatization strategy

The rims of armchair CNTs and their end-caps (e.g. 3) are characterized by continuous strings of polyarene bay regions. One strategy for growing an armchair CNT from an end-cap such as 3 requires a reagent that will add two carbon atoms across the mouths of all the bay regions and convert them into new, unsubstituted benzene rings. In principle, Diels-Alder cycloaddition of acetylene, followed by aromatization of the new 6-membered ring (symmetry-allowed thermal 1,4-loss of H₂, Fig. 7) would produce an elongated CNT with a continuous string of bay regions on the new rim created by the new benzene rings [6]. As long as all the steps are purely thermal and require no intervention by the chemist, the process should repeat itself indefinitely, until the supply of reagent is exhausted. In this scenario, the growing hydrogen-terminated CNT is a living polymer, and the iterative Diels-Alder cycloaddition/aromatization process is a polymerization that requires only a small hydrocarbon template, such as end-cap 3, to get started.

Fig. 7:

Diels-Alder cycloaddition/aromatization strategy for elongating armchair CNT templates.

Despite the attractiveness of this strategy on a formal level, two serious problems make it look rather unrealistic on a practical level. One problem is that the “diene” partners of these Diels-Alder cycloadditions are buried in benzene rings, the aromaticity of which will be energetically costly to disrupt by the conversion of sp² carbon atoms to sp³ carbon atoms, even though just temporarily. The second problem is that acetylene (HC≡CH) is a notoriously poor dienophile for Diels-Alder cycloaddition reactions [28]. Nevertheless, we decided to pursue this strategy.

Diels-Alder cycloadditions on biphenyl and phenanthrene are unknown, even with the most potent dienophiles, because the accompanying loss of aromaticity is simply too much to overcome. On the other hand, Diels-Alder cycloadditions at the bay regions of larger polyarenes, though rare, are feasible [29], because the loss of aromaticity in just two rings out of many becomes energetically less costly as the size of the polyarene increases (Fig. 8) [6].

Fig. 8:

Calculated activation energies (B3LYP/6-31G*) for Diels-Alder cycloadditions of acetylene to the bay regions of polycyclic aromatic hydrocarbons.

To validate the predictions of these calculations, we prepared a soluble, air-stable derivative of bisanthene (9) and heated it with a standard dienophile, diethyl acetylenedicarboxylate. We were rewarded by production of the anticipated ovalene derivative (10) formed by two-fold Diels-Alder cycloaddition/aromatization (Fig. 9) [6]. We did not verify that the hydrogen atoms originating in the bay regions of 9 were actually lost as molecular H₂, but we searched for reduced dienophile (diethyl maleate and diethyl fumarate) and confirmed that none was formed. In a competition experiment between 9 and perylene (the next smaller polyarene in Fig. 8), only the bisanthene (9) reacted with the diethyl acetylenedicarboxylate; the perylene was recovered unchanged [6].

Fig. 9:

Diels-Alder cycloaddition/aromatization of a soluble bisanthene derivative (9). (a) Sealed tube, toluene, 120°C, 1 day, −2 H₂, >95% conversion.

Encouraged that Diels-Alder cycloadditions should be possible on the rims of hydrogen-terminated CNTs, we turned our attention to the second problem, the poor dienophilicity of acetylene (HC≡CH). One solution to this problem would be to use a “masked acetylene” as the dienophile, rather than acetylene itself, i.e. a dienophile that is more potent than HC≡CH but would still leave new HC=CH bridges across the original bay regions after completion of all the post-cycloaddition thermal reactions. Paquette introduced phenyl vinyl sulfoxide as a masked acetylene for Diels-Alder cycloadditions several decades ago [30], and it is commercially available, so we studied its Diels-Alder reactivity with the bay regions of bisanthene 9. Unfortunately, phenyl vinyl sulfoxide proved to be a much weaker dienophile than diethyl acetylenedicarboxylate, although it did give low yields of the two-fold Diels-Alder cycloaddition/aromatization product from 9 under forcing conditions (180-fold molar excess of phenyl vinyl sulfoxide, 155°C, 6 days) [6], [31].

Clearly, a more potent masked acetylene is needed for Diels-Alder cycloadditions to polyarene bay regions. Guided by computational experiments, [31] our attention was drawn to nitroethylene as an attractive candidate. Nitroethylene polymerizes readily and is not commercially available, but methods for generating it in situ and trapping it in Diels-Alder cycloadditions with aliphatic dienes have been known for many years [32]. To our delight, bisanthene 9 reacts readily with nitroethylene to give excellent yields of the two-fold Diels-Alder cycloaddition/aromatization product (11), in which the original bay regions have been spanned by HC=CH bridges and transformed thereby into new, unsubstituted benzene rings [33]. Figure 10 shows the steps that presumably account for this new annulation reaction. The final thermal loss of HONO resembles a Cope elimination but is not a common reaction for ordinary nitroalkanes. In this case, however, several special structural features facilitate the thermal elimination. Specifically, the C–N and C–H bonds that break during this elimination are both benzylic and, therefore, much weaker than those in ordinary nitroalkanes. Furthermore, the new C=C bond being formed completes the π-system of a new benzene ring, which is also energetically favorable.

Fig. 10:

Nitroethylene is a potent dienophile that serves as a “masked acetylene” for the conversion of polycyclic aromatic hydrocarbon bay regions into new, unsubstituted benzene rings by a Diels-Alder cycloaddition/aromatization process.

It should be noted that the Diels-Alder cycloaddition/aromatization method ought to work not only for the growth of armchair CNTs but also for the growth of chiral CNTs (Fig. 11) [7], [33]. As long as the conversion of a bay region into a new unsubstituted benzene ring creates a new bay region, the growth process should continue indefinitely. Additionally, the small hydrocarbon templates suitable for elongation by the Diels-Alder cycloaddition/aromatization strategy need not be limited to hemispherical polyarenes; aromatic belts that are sufficiently wide to enable bay region Diels-Alder cycloadditions should serve well, too.

Fig. 11:

Proposed elongation of chiral a CNT by the Diels-Alder cycloaddition/aromatization process.

During the course of the computational experiments that led us to nitroethylene, we were intrigued to discover that the reactivity of unsubstituted acetylene (HC≡CH) in bay region Diels-Alder cycloadditions is predicted to be only marginally less than that of phenyl vinyl sulfoxide [31]. Would it be possible to use acetylene itself, perhaps under high pressure and/or elevated temperatures, rather than nitroethylene or some other masked acetylene, to grow CNTs from small hydrocarbon templates? We felt compelled to look into this question experimentally and were amazed to find that heating a solution of bisanthene 9 in dimethylformamide under 1.8 atm of acetylene gas in a pressure-regulated reactor gives the same primary Diels-Alder cycloaddition/aromatization product as the one formed in the nitroethylene reaction (Fig. 12) [34]! Subsequent Diels-Alder cycloaddition to the second bay region is actually more difficult and was not observed under the conditions employed. Nevertheless, with sufficiently reactive hemispherical polyarenes or aromatic belts, acetylene might prove to be the most attractive 2carbon annulation reagent of all for the growth of CNTs from small hydrocarbon templates.

Fig. 12:

In favorable cases, even unsubstituted acetylene (HC≡CH) can serve as the dienophile to convert polycyclic aromatic hydrocarbon bay regions into new, unsubstituted benzene rings by the Diels-Alder cycloaddition/aromatization process: 1.8 atm HC≡CH, 140°C, DMF, 21% conversion in 48 h.

In our efforts to elongate the C₅₀H₁₀ hemispherical polyarene (3) into (5,5) CNTs, we chose to use the most reactive dienophile of those studied, nitroethylene. When template 3 was exposed to nitroethylene under the same conditions that work well for the double annulation of bisanthene 9 (Fig. 10), however, the starting material (3) was recovered unchanged. All of the nitroethylene simply polymerized. We tried increasing the amount of nitroethylene, raising the temperature, and running the reaction longer, but polyarene 3 never showed any signs of engaging in even one Diels-Alder cycloaddition (Fig. 13). For some reason, the bay regions of 3 are far less reactive toward nitroethylene than the bay regions of the bisanthene practice compound (9).

Fig. 13:

Unsuccessful elongation of (5,5) CNT template 3 by the nitroethylene Diels-Alder cycloaddition/aromatization method.

Theoretical calculations revealed the reason for the low reactivity of 3. At the B3LYP/6-31G* level of theory, the activation energy for Diels-Alder cycloaddition of acetylene in the bay region of bisanthene 9 is 24.2 kcal/mol, whereas that for polyarene 3 is 55.1 kcal/mol [31]. We suspected that the curvature of the rim of 3 might be responsible for much of this difference, and that hypothesis was borne out by further calculations. Replacing the (5,5) CNT template (3) with a less severely curved (6,6) CNT template lowered the activation energy to 52.9 kcal/mol, and for a (10,10) CNT template the activation energy dropped down to 27.1 kcal/mol [31], [35]. Reaching the transition state for Diels-Alder cycloaddition on the rim of a hemispherical polyarene apparently requires the introduction of considerably more strain than is required for the corresponding reaction at the bay region of a planar polyarene. It is worth noting that the activation energy for Diels-Alder cycloaddition on the rim of a (10,10) CNT template lies very close to that for Diels-Alder cycloaddition to bisanthene 9, which occurs readily. Thus, although the growth strategy using nitroethylene fails with our (5,5) CNT template (3), the method still holds promise for elongating large-diameter (≥2 nm) hydrocarbon templates, both hemispherical polyarenes and wide aromatic belts. A synthesis of a (10,10) CNT end-cap has been initiated but remains unfinished [10], [36].

Benzyne Diels-Alder cycloaddition/aromatization/cyclodehydrogenation strategy

As an alternative to the repetitive 2-carbon annulation strategy described in the previous section, we also considered a 6-carbon annulation strategy based on bay region Diels-Alder cycloadditions of benzyne (Fig. 14) [37].

Fig. 14:

Proposed use of benzyne for elongating armchair CNT templates by a Diels-Alder cycloaddition/aromatization/cyclodehydrogenation strategy.

For this “polymerization” to work, thermal cyclodehydrogenations in the newly formed fjord regions on the rim must join together the new 6-carbon units after they have become attached in order to restore an armchair rim on the growing CNT [37]. Fortunately, such uncatalyzed thermal cyclodehydrogenations of [5] helicenes are well known to occur at high temperatures [38].

Reports of benzyne cycloadditions in the bay regions of polycyclic aromatic hydrocarbons are scarce [39], but we found that benzyne adds readily to our bisanthene (9) to give the dibenzovalene derivative (12, Fig. 15) [31], [40].

Fig. 15:

Conversion of polycyclic aromatic hydrocarbon bay regions into new benzene rings by the Diels-Alder cycloaddition/aromatization method using benzyne, generated from o-trimethylsilylphenyl triflate with Bu₄NF, o-DCB, 150°C, 2 h, 82% yield.

These Diels-Alder cycloadditions and all of those in the previous section have been performed in solution. There may be some concern, however, about the ability of the synthetic CNTs to continue growing once they reach a size that makes them insoluble. To these concerns we have two responses. First, even an insoluble CNT should be able to continue engaging in Diels-Alder cycloadditions at the solid-liquid interface, as long as the bay regions on its rim are still exposed to the solution that contains the dienophile. Furthermore, there may be no need for solvent at all. Perhaps the Diels-Alder cycloaddition/aromatization strategy could be carried out at high temperatures under solvent-free conditions, similar to the ways that CNTs are prepared by previous methods. To test this latter concept, we studied the feasibility of benzyne Diels-Alder cycloaddition/aromatization in the bay region of perylene under solvent-free conditions. FVP of phthalic anhydride has long been known as a clean thermal source of benzyne, [41], [42], [43] so we sublimed a mixture of phthalic anhydride and perylene through a FVP tube. As anticipated, this co-pyrolysis gives the same annulated hydrocarbon (13) as the one formed when benzyne (from a different precursor) adds to perylene in solution (Fig. 16) [31], [39].

Fig. 16:

Conversion of a polycyclic aromatic hydrocarbon bay region into a new benzene ring by the Diels-Alder cycloaddition/aromatization method using benzyne under solvent-free conditions in the gas phase.

The high reactivity of benzyne as a dienophile gave us hope that we might be able to elongate our (5,5) CNT template (3) by this strategy. Unfortunately, all attempts to add benzyne to 3 failed and gave back the starting material unchanged (Fig. 17). Thus, even with powerful dienophiles, the bay regions on the rim of 3 resist Diels-Alder cycloadditions. Larger diameter templates will no doubt be required in order to make this strategy work, too.

Fig. 17:

Unsuccessful Diels-Alder cycloaddition of (5,5) CNT template 3 with benzyne, generated from o-trimethylsilylphenyl triflate and Bu₄NF.

Surface-catalyzed growth strategy

It was always our ambition to develop methods for the chemical synthesis of CNTs from hemispherical polyarenes that rely entirely on metal-free organic reactions, such as those described above. As a backup plan, however, we knew that CNTs could be grown on nanoparticles of catalytically active metals [3] and hypothesized that attachment of hemispherical polyarenes to such metal surfaces should lead to the preprogramed growth of single-index CNTs. The structures of CNTs grown in this manner would be dictated by the template chosen. Attaching a template such as 3 to a metal surface by replacement of the hydrogens with carbon-metal bonds, however, is not a trivial task.

Inspired by the work of Nuckolls et al. in 2007 [44] and Otero et al. in 2008 [45] on surface-catalyzed cyclodehydrogenation of polyarenes, we reasoned that the best way to attach a CNT end-cap to a catalytically active metal surface might be to create the hemispherical polyarene in situ by surface-catalyzed cyclodehydrogenation of an adsorbed planar or quasiplanar precursor. Still focusing on syntheses of (5,5) CNTs, we identified pentabenzocorannulene (14) as the ideal precursor (Fig. 18) [46].

Fig. 18:

Proposed synthesis of (5,5) CNTs from pentabenzocorannulene (14) on a catalytically active metal surface, such as Pt(111).

Theoretical calculations predict that the central core of 14 should adopt a bowl-shaped conformation that is more shallow than that of corannulene (1) and that the barrier for bowl-to-bowl inversion of 14 should be much lower that that of 1 [47], [48]. Thus, no matter which face of 14 (convex vs. concave) approaches the metal surface, the polyarene should flatten out on the surface to maximize its contact. Surface-catalyzed cyclodehydrogenation at elevated temperatures should then yield an attached hemispherical template, ready for elongation into a CNT.

We completed two different syntheses of pentabenzocorannulene (14) several years ago, but neither has yet been published. Some encouraging results have also been obtained from experiments to deposit 14 onto a Pt(111) surface under UHV conditions, transform it thermally into a metal-bound (5,5) CNT end-cap by surface-catalyzed cyclodehydrogenation, and grow those into long CNTs by successive insertions of 2-carbon units. This work is ongoing but is still too preliminary to report in detail here. We are, however, quite optimistic.

Our optimism about the viability of this approach received a major boost in August of 2014, when Fasel et al. reported its successful application to a template-directed synthesis of (6,6) CNTs from a 90-carbon polyarene on a Pt(111) surface under UHV conditions [49]. To the best of our knowledge, that work represents the first and only controlled chemical synthesis to date of single-index CNTs from a small hydrocarbon template.