Exploring Silyl Protecting Groups for the Synthesis of Carbon Nanohoops

The synthesis of topological molecular nanocarbons, such as hoop-like [ n ]cycloparaphenylenes, requires the use of spatially prearranged, pro-aromatic units to overcome a build-up of large molecular strain in their curved structures. The used cyclohexadienyl units, however, contain tertiary alcohols that need protection to prevent side re- actions until the aromatization step that affords the final curved hydro-carbon. Although alkyl and triethylsilyl groups have been successfully applied as protecting groups for this purpose, each suffers from specific drawbacks. Here, we explore the potential of sterically more crowded silyl groups, namely, tert -butyldimethylsilyl and triisopropylsilyl, as al-ternatives to the established protection strategies. We show that tert - butyldimethylsilyl can be easily installed and removed under mild conditions, displaying markedly higher resistance towards acids or bases than the triethylsilyl group used to date. Unlike in the case of alkyl groups, tert -butyldimethylsilyl also preserves a high stereoselectivity during the nucleophilic additions of ArLi. Furthermore, we demonstrate that both tert -butyldimethylsilyl and triethylsilyl groups can be installed on the same substrate, and that the latter be selectively deprotected. Thus, the high stereoselectivity, improved stability, and easy deprotec- tion make tert -butyldimethylsilyl an excellent protecting group for the synthesis of carbon nanohoops.

Protecting groups are an indispensable tool to control the selectivity of chemical transformations in modern natural product synthesis or the synthesis of organic materials. An ideal protecting group needs to match several criteria, such as high-yielding introduction and removal steps, suffi-cient stability and highly specific deprotection conditions to afford orthogonality towards other protecting groups. Particularly useful types of scaffolds bearing tertiary alcohols, which require protection, are A and B (Figure 1). For example, compounds 1 and 2 (Figure 2), that contain scaffolds A and B, are important building blocks for the synthesis of topological molecular nanocarbons that involve carbon nanohoops, such as [n]cycloparaphenylenes ([n]CPPs). 1 Carbon nanohoops exhibit unique size-dependent optoelectronic properties 2 and host-guest chemistry, 3 giving rise to applications in bioimaging, optoelectronic materials and supramolecular carbon-rich nanomaterials. 4 Although envisioned decades ago, the first synthesis of these strained molecules was accomplished only in 2008 by Jasti and Bertozzi. 5 Their strategy relies on masking a p-phenylene unit as a 1,4-dimethoxycyclohexa-2,5-diene-1,4-diyl moiety. Here, the two tertiary alcohols prearrange the geometry in scaffold B with cis configuration for an effective macrocyclization. The reductive aromatization of the cyclohexa-2,5-diene unit then provides the necessary driving force to build up the strain in the final step. The exergonicity of the aromatization, however, requires that the tertiary alcohols be protected throughout the multistep synthesis to avoid side reactions. The electronic nature and the size of the methyl group used originally, however, result in a poor diastereoselectivity of the addition of organolithium reagents, typically ArLi, to ketone scaffold A. 5,6 Moreover, their removal during the aromatization often necessitates the use of harsh reagents, such as lithium/sodium naphthalide. Although the diastereoselectivity of the addition was found to be mostly dictated by electrostatics, 7,8 Yamago and co-workers 9,10 proposed the use of bulkier triethylsilyl (TES) as the protecting group for the tertiary alcohols in A or B. TES improves the stereoselectivity of ArLi addition and the exclusive formation of the cis-diastereomer of B is typically observed. TES is also relatively straightforward to deprotect with tetra-n-butylammonium fluoride (TBAF) before the aromatization of the ensuing 1,4-dihydroxycyclohexa-2,5-diene-1,4-diyl, performed under mild conditions with H 2 SnCl 4 . 8 In our own experience, however, the sensitivity of the TES group towards acids compromises the stability of compounds that contain the structural motif B. Such compounds are prone to decomposition, either during the purification step after their synthesis or when stored, even at low temperatures, although such negative results are rarely reported in the literature. 1k,4d,11 Because of the individual drawbacks of the methyl and TES groups, we decided to search for an alternative protecting group that would (a) be easy to introduce, (b) would undergo stereoselective addition of ArLi to ketones A, (c) would be significantly more stable than TES, and (d) would be easy to remove to allow for a mild aromatization. Such a protecting group could introduce additional orthogonality to the synthesis of topological molecular nanocarbons.
The steric bulk around the silicon determines the stability of a silyl protecting group and permits selective protection/deprotection in the presence of another silyl ether. [12][13][14] The stability of the TES ethers used in A and B is relatively low in comparison to other silyl ethers in the presence of both acids and bases. For example, the half-lives of TESprotected p-cresol in the presence of 1% hydrochloric acid or 5% sodium hydroxide are <1 minute, while the half-lives of tert-butyldimethylsilyl (TBDMS)-protected p-cresol are 273 minutes and 3.5 minutes, respectively. 13 In general, the relative stability of different silyl groups to acids increase in the following order: trimethylsilyl (TMS) (1) < TES (64) < TBDMS (20,000) < triisopropylsilyl (TIPS) (700,000) < tertbutyldiphenylsilyl (TBDPS) (5,000,000)

Special Topic Synthesis
order is: TMS (1) < TES (10-100) < TBDMS ~ TBDPS (20,000) < TIPS (100,000). 14 Therefore, protecting groups with additional steric bulk, such as TBDMS or TIPS, represent great candidates as considerably more stable protecting groups. In addition, the extra steric bulk of the silyl ether in A is expected to display excellent diastereoselectivity for an ArLi nucleophilic addition to form the motif B. Protection of a tertiary alcohol, however, becomes more challenging with increasing steric bulk of the silyl protecting group. 15 We selected model alcohols 1a and 2a (Scheme 1) as proxies for the motifs A and B, respectively. In fact, these two alcohols are frequently used in the synthesis of carbon nanohoops, including CPPs, 1,[5][6][7][8]16 and their protection with TES to give 1b and 2b can be achieved in 92% 1f and ≥82% 8 yields, respectively, in a clean transformation.
We first tested the standard silylation of 1a with tertbutyldimethylsilyl chloride (TBDMSCl) and imidazole 17 in CH 2 Cl 2 at room temperature (Table 1, entry 1; see also Table  S1, entry 1 in the Supporting Information) and obtained full conversion of 1a. However, we observed the formation of a complicated mixture of unknown products. We assumed that the steric bulk slows down the nucleophilic substitution of the chloride in TBDMSCl, allowing other reactions to compete. We thus exchanged CH 2 Cl 2 for significantly more polar DMF, which we expected to stabilize the alcoholate that would be generated from 1a and imidazole in a very small amount. Such attempts resulted only in a slow conversion of 1a into an unknown product, even when larger amounts of the reagents and an elevated temperature were used (Table 1, entry 2). Diol 2a was completely inert under the same reaction conditions (Table 1, entry 5). Similarly, alcohol 1a turned unreactive when bulkier 2,6-lutidine with a similar basicity was used (Table 1, entry 3). Likely, the observed side reactions are catalyzed by a general base. To accelerate the nucleophilic substitution at silicon, we decided to replace the chloride in TBDMSCl for triflate, which is an excellent nucleofuge. The combination of 2,6-lutidine with tert-butyldimethylsilyl trifluoromethanesulfonate (TBDMSOTf) in CH 2 Cl 2 , however, afforded a full conversion of 1a into 3c and 4c (Table S1, entry 4). Product 3c is most probably formed through a rearrangement involving a 1,2aryl shift (Scheme 2), while 4c ensues from the excess TBDMSOTf reacting with 3c. The observed transformation nicely illustrates the power of aromatization to drive rearrangements of cyclohexadienones. 18 To our satisfaction, we observed the formation of 1c when the reaction was repeated under the same conditions but in polar DMF, albeit the reaction appeared significantly slower (Table S1, entry 5).

Special Topic Synthesis
Increasing the amounts of the TBDMSOTf and 2,6-lutidine and increasing the temperature provided a faster reaction, which was, however, accompanied by the formation of 4c (Table S1, entries 6-10). The optimal temperature was found to be 70 °C because it resulted in a full conversion of 1a in 18 hours (incomplete at 60 °C) and a lower yield of 4c compared to the reaction at 80 °C. The reaction at 70 °C afforded 1c and 4c in 58% and 8% isolated yields, respectively, after an easy separation by column chromatography. Finally, reducing the amounts of the reagents and the solvent (Table S1, entries 11-17) provided the optimal conditions that allowed us to isolate 1c in 71% yield on one-gram scale ( Table 1, entry 4). We also tested different bases in combination with TBDMSOTf (Table S1, entries 18-23), however, the reaction was slower or no significant improvement of the 1c:4c ratio or the isolated yield of 1c could be achieved.
We next attempted the protection of diol 2a with TBDMSOTf (Scheme 1, Reaction B). The optimal conditions found for the protection of 1a led to a full conversion of 2a (Table 1, entry 6). In addition to the desired product 2c, however, formation of a significant portion of 5c was observed. Similar to the formation of 3c, this rearrangement involves a 1,2-aryl shift aromatizing the cyclohexa-2,5-diene-1,4-diyl unit (Scheme 2). Increasing the amount of TBDMSOTf and 2,6-lutidine improved the relative yield of 2c (see Table S2, entry 3 in the Supporting Information). In analogy to our experiments with 1a, we hypothesized that the rate-limiting step in the rearrangement has a higher barrier. We noticed almost exclusive formation of 5c at 100°C , while the reaction at 40 °C showed the opposite effect and provided mostly 2c (Table S2, entries 4 and 5). At room temperature, however, a small amount of the monoprotected compound 6c was observed (Table S2, entry 6). Finally, the reaction performed at 50 °C with reduced amounts of TBDMSOTf and 2,6-lutidine provided the best result (Table 1, entry 7). Compounds 2c and 5c are relatively nonpolar with similar retention factors. Nevertheless, we achieved their full separation by column chromatography and isolated 2c in 70% yield. Although the isolated yields upon protection with TBDMS are somewhat lower than those with TES, the optimized protocols provide satisfactory results that improve with the reaction scale and allow for using TBDMS protection in a multistep synthesis of hoop-like molecular nanocarbons.
We were then interested whether the steric bulk on silicon could further be increased. We thus attempted protecting 1a with TIPSOTf by employing the base and solvent used in the case of TBDMSOTf (Table 1, entry 4). We reached an incomplete conversion after 18 hours and isolated the desired TIPS-protected 1d in only 3% yield along with the rearranged product 4d in 34% yield ( Table 1, entry 8). A similar unsatisfactory result was obtained when 2a was reacted with TIPSOTf (Table 1, entry 9). In this case, no formation of doubly protected 2d was observed. Instead, the reaction afforded the monoprotected product 6d in 39% isolated yield, together with 6% of the rearranged product 5d. Although increasing the amounts of TIPSOTf and the base and elevating the temperature allowed us to improve the conversion (see Table S3, entries 3 and 4 in the Supporting Information), heating favored the formation of 5d, the exclusive product at 100 °C. Clearly, the bulkiness of the TIPS group leads to its relatively difficult introduction to 1a, and precludes installing two TIPS protecting groups on 2a due to their proximity in 2d.
The nearly exclusive formation of 5c at 100 °C (see Table  S2, entry 4 in the Supporting Information) was rather intriguing because we expected the formation of 2c and 5c in similar amounts if the protection and the rearrangement were parallel processes. Instead, the result suggests that it is compound 2c that is transformed into 5c in a subsequent reaction. We thus conducted a series of experiments to determine the conditions that promote the rearrangement and that may have a detrimental effect on the stability of 2c (Table 2).

Special Topic Synthesis
When a DMF solution of 2c was stirred at 100 °C, with or without addition of 2,6-lutidine, no conversion of 2c was observed ( Table 2, entries 1 and 2). An excess of TBDMSOTf in the absence of base converted 2c fully into 5c, which we isolated in 83% yield (Table 2, entry 3). Using an excess of TBDMSCl, a weaker Lewis acid, did not show any sign of rearrangement (Table 2, entry 4). This indicates that TBDM-SOTf may act in the rearrangement as a Lewis acid. 11 We suspected that TBDMSOTf may contain traces of triflic acid (TfOH) that could catalyze the rearrangement. Indeed, a catalytic amount (10 mol%) of TfOH led to a clean rearrangement of 2c at 100 °C to furnish 5c, which we isolated in 84% yield (Table 2, entry 5). It is worth noting that the second TBDMS group in 5c was not cleaved during the reaction. The TBDMS group is known to be more resistant to acidic conditions than the TES group, 14 which motivated us to compare their stabilities in 2b and 2c in the presence of acids such as TFA, HCl, and TfOH. In a typical experiment, a DMF solution of 2b or 2c was stirred with a catalytic amount of the selected acid at 35 °C for 18 hours, and the reaction progress was monitored by 1 H NMR spectroscopy (Table 2, entries 6-9). Compound 2c was inert to the presence of TFA (pK a = 0.52 in water 19 ), while partial deprotection of one of the TES groups in 2b was observed. The much stronger acid HCl (pK a = -5.9 20 ) affected neither 2b nor 2c when only 1 mol% was used. Increasing the amount of HCl to 10 mol%, however, promoted deprotection of the TES groups, while the TBDMS was inert. No rearrangement was observed in any of these attempts. Both 2b and 2c reacted in the presence of 10 mol% of the very strong acid TfOH (pK a = -14.7 20 ). Only a partial conversion of 2c was reached after 18 hours at 35 °C compared to the full rearrangement into 5c at 100 °C (Table 2, entries 5 and 9), and no TBDMS deprotection could be detected. On the other hand, compound 2b was fully transformed into a mixture of unknown products. These results indicate that 2c is considerably more resistant to acids than 2b. In our experience, some TES-protected moieties similar to 2b (see 2 in Figure 2) are surprisingly labile. 11 Their stability is influenced by the solvent and traces of impurities. For example, we even observed a rearrangement analogous to 2c → 5c in the presence of Mg 2+ ions in one specific case. 21 We thus expect that

Special Topic Synthesis
many issues associated with the stability of the building blocks (1 and 2 in Figure 2) used in the synthesis of hooplike molecular nanocarbons that rely on the strategy employing triethylsilyl protecting groups (TES) can now be avoided by using TBDMS. Clearly, the 1,2-aryl shift of TBDMS-protected 2c in DMF requires a very strong Lewis or Brønsted acid and an elevated temperature to proceed on a reasonable timescale (see Scheme 2 for the proposed mechanism). However, we cannot exclude that the formation of 5c first involves the rapid formation of 6c, although our results with HCl indicate that this process is likely not particularly facile.
We next aimed to demonstrate that the TBDMS protecting group in 1c preserves the high stereoselectivity of ArLi addition to the carbonyl group and that both TBDMS and TES groups could be installed on a single 1,4-dihydroxycyclohexa-2,5-diene-1,4-diyl moiety, as in compound 2e (Scheme 3), with the latter being selectively removed in a subsequent step. The addition was accomplished via the reaction of monolithiated 1,4-dibromobenzene with 1c at -78 °C, followed by protection of the resulting alcohol with a TES group to provide 2e (1.7 g, 78% yield). We noticed the formation of a single cis diastereomer, which confirms that TBDMS displays the same high stereoselectivity as observed for 1b bearing a TES group. We fully deprotected both silyl groups with TBAF at room temperature, isolating 2a in 83% yield. Similarly, both TBDMS groups in 2c could be cleaved in 73% yield. When the temperature was decreased to -45°C , we successfully achieved the selective deprotection of TES in 2e to afford 6c in 76% yield. Similarly, the selective deprotection of TES could also be accomplished in 82% yield when 2e was stirred with an excess of K 2 CO 3 in refluxing methanol. The selective deprotection of the TES group can thus be easily achieved under mild conditions with two complementary methods. As such, combination of both silyl groups represents an attractive strategy to construct versatile building blocks that may prove useful beyond the synthesis of macrocycles found in topological molecular nanocarbons.
Finally, we evaluated the effect of the steric bulk of the protecting groups on the 'bite' angle defined by the two phenylenes attached to the central cyclohexadienyl unit in 2. Building blocks such as 2 are key intermediates in the synthesis of a variety of carbon nanohoops and the size of the bite angle may affect the efficiency of the macrocyclization step. The steric hindrance between two bulky protecting groups, such as TBDMS, could decrease this angle and prevent macrocycle formation if a wider bite angle is required. Therefore, we compared the impact of the size of the protecting group (Me, TMS, TBDMS) in 2 on the bite angle using DFT calculations. We found that the value of the bite angle in 2 (65° ± 5°) is not particularly sensitive to the type of the protecting group (see Table S4 in the Supporting Information). Analysis of the few reported 16b,22 crystal structures of compounds analogous to 2 (see Figures S1-S3 and Table S4) to validate the accuracy of the selected DFT functionals confirmed that our calculations reproduce the bite angles well. In addition, analysis of the crystal structures further revealed that the phenylenes can adopt surprisingly acute bite angles (47°). As a result, it can be expected that 2c with bulky TBDMS protecting groups can be used in place of 2b (or Me-protected 2a) in the synthesis of carbon nanohoops. However, in cases where the TES introduction or removal steps prevent successful synthesis of a carbon nanohoop, the use of TBDMS will likely lead to the same outcome.
In conclusion, we have demonstrated that tert-butyldimethylsilyl is a versatile protecting group in the preparation of important building blocks used in the synthesis of topological molecular nanocarbons, such as hoop-like CPPs. We developed the methodology to install and cleave tertbutyldimethylsilyl under mild conditions, minimizing un-

Special Topic Synthesis
desired rearrangements driven by aromatization of cyclohexa-2,5-diene-1,4-diyl units. The tested tert-butyldimethylsilyl ethers were substantially more stable towards acids and bases than the corresponding triethylsilyl groups frequently used to date. The steric bulk in the tert-butyldimethylsilyl group preserves the high stereoselectively of ArLi additions to afford spatially prearranged building blocks used in CPP synthesis. We also explored the conditions that trigger the undesired rearrangements involving 1,2-aryl shifts. In addition, we successfully prepared a compound with both tert-butyldimethyl and triethylsilyl ethers, and identified conditions that permit removing the triethylsilyl chemoselectively. We anticipate that the strategies developed in this work will not only find applications in the synthesis of carbon nanohoops, such as CPPs and their derivatives, but also of other macrocycles, e.g., unprecedented macrocyclic drugs that incorporate a biphenyl or a terphenyl moiety.
Unless otherwise stated, all glassware used to perform moisture-sensitive reactions was oven-dried at 120 °C overnight, assembled hot and allowed to cool to room temperature under a stream of argon, or flame-dried under high vacuum and filled with argon. All reactions that require heating were conducted in an oil bath and the indicated temperature corresponds to the temperature of the oil bath. To obtain a temperature of -78 °C or -45 °C, a bath of acetone or acetonitrile, respectively, was cooled with dry ice. All commercially available reagents and solvents were used directly without purification unless stated otherwise. Flash column chromatography was performed using silica gel 60 Å (230-400 mesh particle size) from Supelco ® . Thinlayer chromatography (TLC) was performed on silica gel plates F 254 60 (aluminum-supported) from Supelco ® using UV (254 nm) visualization. 1 H and 13 C NMR spectra were recorded on a Bruker Avance II 400 or a Bruker Avance III HD 400 spectrometer ( 1 H: 400 MHz, 13 C: 101 MHz), or a Bruker Avance III HD 300 spectrometer ( 1 H: 300 MHz, 13 C: 75 MHz). Chemical shifts () are reported in parts per million (ppm) referenced to the residual solvent peak ( 1 H: 7.26 ppm for CDCl 3 , 5.32 ppm for CD 2 Cl 2 ; 13 C: 77.16 ppm for CDCl 3 , 53.84 ppm for CD 2 Cl 2 ). Multiplicities are given as s (singlet), d (doublet), t (triplet), q (quartet), m (multiplet), and br (broad). High-resolution mass spectra (HRMS) were recorded on a ThermoScientific LTQ Orbitrap XL mass instrument using nanoelectrospray (NSI-MS) or electron ionization (EI-MS). Elemental analysis was performed in triplicate on a Thermo Scientific Flash 2000 Series instrument using the CHN method with sulfanilamide or cyclohexanone as the reference. Melting points were determined on a Büchi B-545 apparatus and are uncorrected.
Yellow oil; R f = 0.33 (SiO 2 , EtOAc/n-pentane = 1:19 Elemental analysis was not performed due to an insufficient amount of the material.