Fullerenes are a unique family of carbon-based cage molecules, which attract interest because of their remarkable properties and potential applications. Most effort so far has been focused on the study of C60 and C70, whereas other members of the huge fullerene family remain poorly explored. One of the main challenges in this field is the developing of the synthetic methods, which are suitable for the production of these unique materials in isomer-pure form in macroscopic amounts. Here, we review studies toward the rational synthesis of fullerenes from molecular precursors that have been published to date. The scope and limitation of the zipping strategy are discussed. The relevance and prospects for construction of the fullerene cages and related carbon-based nanostructures via cyclodehydrofluorination (C─F bond activation) are highlighted.
- bottom-up synthesis
- zipping approach
- C─F bond activation
Fullerenes represent a unique class of sp2-carbon-based compounds with a spherical π-system, displaying intriguing physical and chemical properties [1, 2]. The conventional methods of fullerene production are based on various graphite evaporation techniques which result in the heterogeneous fullerene containing soot [1, 2]. Unfortunately, none of the existing techniques provide access to fullerenes in isomer-pure form, which is crucial for their possible applications. The random nature of the fullerene formation results in a highly heterogeneous mixture containing mainly the most stable C60-
2. The concept of the rational synthesis of fullerenes
Rational synthesis of fullerenes is of practical interest as a promising method for the synthesis of isomer-pure fullerenes, including fullerenes which cannot be obtained by conventional techniques. One of the first attempts of the “rational synthesis“ of fullerenes was undertaken in 1993 . It has been proposed that C60-
The general strategy of the rational synthesis of fullerenes is based on the synthesis of specially “designed” PAH-precursor molecules containing the carbon framework required for the formation of the target fullerene cage. The respective precursor can be “zipped” to the desired fullerene by the tandem intramolecular Aryl─Aryl coupling. The characteristic feature of this approach is a zipper mechanism of cyclization in the final synthetic step. Since the regiospecificity of each condensation step can be strictly predefined by the precursor design, the cyclization will lead to the preprogrammed isomer of fullerene exclusively. This final cyclization step represents the key transformation in the rational synthesis of fullerenes by the zipping approach.
2.1. Wet-chemical synthesis
Since a big number of new C─C bonds have to be established during precursor zipping, the efficiency of each condensation has to be very high in order to achieve reasonable conversion. Unfortunately, methods developed for the synthesis of the planar PAHs via intramolecular Aryl–Aryl coupling are usually uneffective in the case of highly strained fullerenes. The strain energy of the fullerenes stems mainly due to pyramidalization of carbon atoms, as a result of deviation from energetically ideal values observed in graphene lattice. The perturbation of the π-system causes enormous enhancement of the energy, which has to be overcome during cyclization. Thus for example, various polyphenylenes can readily undergo an intramolecular cyclization under UV irradiation in the presence of an oxidant . The photocyclization is an effective and frequently used technique for the synthesis of large planar polycyclic systems. Alternatively, oxidative intramolecular cyclodehydrogenation can be carried out under Scholl condition . Recently, this approach has become of high practical interest for the synthesis of large PAHs such as nanographenes and nanoribbons [17, 18]. Although both cyclodehydrogenation methods show excellent efficiency in the case of planar PAHs, these techniques appear to be not effective for the synthesis of bowl-shaped PAHs and fullerenes. Several alternative intermolecular aryl–aryl coupling techniques have been developed to overcome these disadvantages. Here the palladium(0)-catalyzed direct arylation utilizing bromo and chloro derivatives has been found to be effective for the synthesis of small fullerene fragments . The largest fullerene fragments obtained so far contain 50 carbon atoms and formally represent more than 83% of the C60 fullerene surface (Figure 3) .
Although the largest synthesized C50 bowl is more curved than C60 , the rational synthesis of fullerene following this route still remains a challenge. This is mainly connected with difficulties to perform the Pd-catalyzed cyclization in a domino fashion. Secondly, the reaction conditions appear to be aggressive with respect to fullerene species. The recent discovery of effective intramolecular aryl–aryl coupling via C-F bond activation under transition-metal free conditions might be a solution to this problem. Recently, Siegel et al. , Ichikawa et al. , and Amsharov et al. [23, 24] have demonstrated the efficiency of the C─F bond activation strategy for the synthesis of various bowl-shaped PAHs via formal HF elimination step (cyclodehydrofluorination). In the last case, the effective cyclization was achieved utilizing easily accessible aluminum oxide, which makes the approach highly attractive for preparative-scale synthesis. Moreover, it has been found that intramolecular Aryl-Aryl coupling of fluoroarenes via γ-aluminum oxide-mediated C─F bond activation can be realized under mild conditions with unprecedentedly high chemoselectivity [25, 26]. Using this technique an extended bowl-shaped PAHs representing more than 75% of the C60 fullerene connectivity was obtained with close to quantitative yield (Figure 4) , demonstrating high potential for generating extended nonplanar carbon-based nanostructures including fullerenes. Although the rather big fullerene fragments can be successfully produced by wet-chemical methods, the solution-synthesis of fullerenes remains a challenge.
2.2. Surface-assisted synthesis
The direct on-surface synthesis of nanostructures is a highly promising strategy. The on-surface studies are usually conducted on single-crystal surfaces under ultra-high vacuum conditions that offer extraordinary degree of control in the experiments . Moreover, the usage of atomically clean and flat crystal surfaces allows the observation of the reaction process at the single molecule level by means of scanning tunneling microscopy (STM) . Recent progress in this field demonstrates a high potential for the synthesis of complex architectures via Aryl-Aryl coupling of organic precursors on single-crystal metal surfaces [29, 30]. In the case of planar PAHs, the intramolecular cyclodehydrogenation can be performed effectively on Cu(111), Ag(111), and Au(111) surfaces. Intramolecular cyclization on Au(111) surfaces typically occurs at around 400°C . On the more reactive copper surfaces, cyclization takes place earlier at 200°C . However, attempts to perform a similar cyclization to form non-planar PAHs were unsuccessful because of the intermolecular merging of the precursor molecules. Therefore, the synthesis of fullerenes requires harder conditions, including using more reactive Pt-group metal surfaces and applying higher temperatures of 450–600°C . The first surface-assisted synthesis of the bowl-shaped PAH was reported by Rim et al. on the example of the hemispherical PAH C48, which was generated on the Ru(0001) surface via exhaustive cyclodehydrogenation of hexabenzocoronene precursor (Figure 5) . The six-fold cyclodehydrogenation on the periphery of precursor was achieved by annealing of the precursor molecule at 600°C. Note that Ru surface serves not only as a catalyst for the cyclodehydrogenation it also prevents intermolecular coupling due to efficient bonding to the surface. Finally, the Ru surface acts as a support for precursors and products enabling STM investigation. The STM analysis has revealed a high selectivity and exceptionally high conversion ratio of the deposited precursors into non-planar PAH structure (Figure 5). However, due to high reactivity of the Ru surface and high temperature applied, the dehydrogenation process does not stop on the formation of the C48H12 bowl. The bowl-shaped PAH undergoes further dehydrogenation resulting in all-carbon C48 bowl covalently bonded to the Ru surface by 12 C─Ru bonds (Figure 5) .
Soon after, the efficiency of the surface-assisted approach has been demonstrated by Otero et al. on the synthesis of C60 fullerene on Pt surface . In their study, C60H30 fullerene precursor molecules (Figure 6, precursor A) were deposited onto a Pt(111) surface and annealed at around 480°C. The annealing step led to the formation of C60 cages with an efficiency close to 100%. The STM analysis shows that the molecular coverage does not change during temperature-induced reaction, and all of the precursor molecules converted into fullerenes . Later it was shown that the process is highly selective in nature and is not accomplished by skeletal rearrangements. This important question has been addressed by the investigation of surface-catalyzed cyclodehydrogenation of a specifically designed precursor molecule (Figure 6, precursor B) “preprogrammed” to the formation of the C60-bowl-shaped PAH with unique triangular shape . The on-surface cyclodehydrogenation of two isomeric C60H30 precursors (Figure 7) gave convincing evidence that the reaction occurs in a highly selective manner . In contrast to the condensation of the precursor A whose cyclodehydrogenation leads to the spherical C60, the zipping of the precursor B resulted in a triangular-shaped C60-bowls whose formation was distinguished by the STM analysis.
Those works demonstrated that the surface-assisted cyclodehydrogenation on Pt-group metal surfaces is highly selective and can be used for fabrication of various specifically designed carbon-based nanostructures including fullerenes.
Finally, the approach was successfully applied for the first rational synthesis of isomer-pure higher fullerene C84(20) . Upon annealing of specially designed C84H42 precursor at 550°C on Pt(111) surface, the shape of all of precursor molecules changed and all species became spherical with a lateral size expected for C84 fullerene (Figure 8). The formation of the C84 fullerene cages was additionally confirmed by MS analysis . Moreover, the approach was found to be suitable for the synthesis of highly interesting hetero fullerenes. The fabrication of an azafullerene by surface-assisted approach has been demonstrated by Otero et al. on the example of C57N3 synthesis. In this study, a structurally related precursor with three incorporated nitrogen atoms was quantitatively converted to the hetero fullerene C57N3 upon annealing on Pt(111) surface at 477°C .
Although the technique allows to perform effective zipping to the target carbon nanostructures [37, 38], the resulting nano-carbons are covalently bonded to the metal surface and cannot be desorbed without decomposition. Those difficulties could be probably overcome by using less active metal oxide surfaces. Since the metal oxides show low activity in cyclodehydrogenation, the cyclization on such surfaces would require much higher temperatures leading to loss of selectivity. However, introduction of “activating” groups into the precursor structures could significantly reduce the activation barriers. In particular, it has been found that fluorine functionality appears to be an excellent ring-closure promoter, which has demonstrated unexpectedly high efficiency. It has been found that coupling can be realized in fluoroarenes on thermally activated aluminum oxide at very mild conditions [23, 24, 25, 26]. Moreover, the alumina-supported cyclodehydrofluorination (cyclization via formal HF elimination) has demonstrated unprecedentedly high chemoselectivity and regiospecificity. The intramolecular Aryl–Aryl coupling on aluminum oxide was achieved with close to quantitative yield even in the case of large bowl-shaped PAHs. Since fluorine can promote the ring closure only if hydrogen is placed neighboring in space in the precursor structure, it was possible to perform the cyclization in domino fashion. This makes the cyclodehydrofluorination more attractive in comparison to cyclodehydrogenation where the selectivity must be predefined by the rigid structure of precursor. Further, it was found that metal oxides of II, III and IV groups are active in the cyclodehydrofluorination of fluoroarenes. In particular, zirconium, titanium and indium oxides were found to be effective for cyclization via C─F bond activation . These results point the way for fabrication of higher fullerenes and other related carbon-based nanostructures in a fully controllable manner directly on metal oxide surfaces.
2.3. Gas phase synthesis
The rational construction of fullerenes in the gas phase seems to be very logic, since in all conventional techniques the cage fabrication occurs in the gas phase. First of all, fullerenes, including higher, non-IPR and heterofullerenes, are stable in the gas phase. Second, the undesired intermolecular interactions can be effectively suppressed in the gas phase, providing excellent conditions for intramolecular processes. Not surprising that gas-phase synthesis appears to be the most promising approach.
The flash vacuum pyrolysis (FVP) represents one of the most powerful techniques for the synthesis of highly strained architectures via intramolecular cyclization. It consists in the heating of precursor molecules in the gas phase to a very high temperature for a very short time. The high temperature applied in FVP causes hemolytic cleavage of C─H bonds leading to the formation of radicals possessing enough energy to cause intramolecular cyclization yielding highly strained products. Pioneering by Scott et al., variety of small bowl-shaped PAHs (fullerene fragments) have been successfully synthesized in macroscopic amounts via intramolecular cyclization of quasi-planar PAH precursors under FVP conditions [40, 41, 42]. It was found that introduction of chlorine or bromine functionalities into the initial precursor is essential to facilitate the cyclization via free radical mechanism (Figure 9). Finally, this methodology has been successfully applied for the first rational synthesis of C60 utilizing partially chlorinated C60 fullerene precursor (Figure 9) . Despite the low yield of C60 (0.1–1%) the power of the strategy has been clearly demonstrated.
The possibility of the synthesis of higher fullerenes by FVP has been shown on the examples of C78  and C84  by pyrolysis of specially programmed precursors. In both cases the size-selective formation of fullerenes was observed as indicated by the MS analysis (Figure 10). However, the rates of conversion to the target molecules have remained disappointingly low because of the low efficiency of intramolecular condensation of non-activated precursors. The usually employed for activation chlorine reach their limits in the case of large molecules because of decomposition of precursors during sublimation. The molar masses will generally be high, since a large number of new C─C bonds that need to be formed necessitate the introduction of a big number of promoter groups. Moreover, the radical nature of the condensation could drastically affect the selectivity of the process.
Although the FVP approach has been proven to be very prolific for the synthesis of small non-planar PAHs, the high-yield synthesis of large bowls and isomerically pure fullerenes has remained challenging. Recently, an efficient intramolecular fluorine-promoted ring closure in benzo[c]phenanthrenes under FVP conditions via HF elimination has been reported . It has been shown that HF elimination is a synchronous process leading directly to the C─C bond formation without any intermediates, thus producing no side products . The small size and low molecular weight of fluorine, as well as high thermostability of the C─F bond make fluorine a perfect activating group for the rational synthesis of fullerene by zipping approach. The feasibility of the approach has been demonstrated on the successful synthesis of C60 fullerene from the respective fluorinated precursor which was effectively folded into the fullerene under laser ionization. The respective 15-fold cyclization was realized in a domino fashion via consecutive HF elimination without observation of side reactions or undesired fragmentations . As it is shown in Figure 11, the precursor molecule undergoes highly selective HF loss (with simultaneous C─C bond formation) leading to the open-cage which spontaneously collapses to the target C60 via H2 loss. The regioselectivity of the domino-like cyclization and the formation of bowl-shaped intermediates and finally spherical C60-
Due to exceptionally high efficiency, it was assumed that the same approach can be extended to the synthesis of higher fullerenes as well as to the synthesis of highly attractive non-IPR cages. Recently, it has been demonstrated that non-IPR fragments can be indeed fabricated via cyclodehydrofluorination in the gas phase. It was shown that despite high strain the respective precursor undergoes rather smooth four-fold cyclodehydrofluorination, leading to the non-IPR fullerene fragment C26H10 (Figure 12) . These results demonstrate the general possibility of rational fabrication of the non-IPR architectures via cyclodehydrofluorination of the specially preprogrammed fluoroarene-based precursors. Two possible precursors for generation of non-IPR fullerenes C38 and C42 are shown as example. Taking into account the present state-of-the-art of MS-based synthesis/deposition technique, which allows multilayer deposition of such compounds, the future scale-up appears to be feasible. Among non-IPR cages, the laser-induced cyclodehydrofluorination approach seems to be also applicable for the rational synthesis of higher and non-classical carbon cages.
The zipping strategy had been proved to be prolific for the rational synthesis of the fullerene fragments and the fullerene cages. The progress in this field shows the general possibility to fabricate elusive fullerenes in a fully controllable manner. Production of these unique materials in isomer-pure form in macroscopic amounts does no longer appear elusive, and there is promising prospect that such compounds will become available in the near future. Since there are virtually no limiting factors, there are no principle obstacles standing against extending the zipping approach to highly interesting related carbon-based architectures.
The author thanks the Deutsche Forschungsgemeinschaft (DFG-SFB 953 “Synthetic Carbon Allotropes”).