Stereochemical significance of O to N atom interchanges within cationic helicenes: experimental and computational evidence of near racemization to remarkable enantiospecificity

Oxygen atoms of cationic dioxa and azaoxa [6]helicenes can be exchanged by amino groups and these reactions lead to either near racemization (es 3%) or to a remarkable enantiospecificity (es up to 97%). Based on detailed investigations, the chirality transfer of the first transformation can be improved to 61%.


Introduction
Pioneered by Baeyer and Piccard more than a century ago, 1-3 the formation of pyridinium ions by treatment of pyrylium salts with primary amines is recognized as a method of choice for the construction of cationic aza aromatics (Fig. 1, top). [4][5][6] The three step reaction mechanism, i.e., (i) the nucleophilic attack of the amine, (ii) the O-ring opening yielding an intermediate ketone and (iii) the N-ring closure on the carbonyl, 7-10 classes this reaction as an ANRORC process. 11 The driving force of the transformation is the better stabilization of the positive charge by the more electron-donating amino group. When the pyrylium unit is embedded in a polyaromatic scaffold, the reaction suffers from a breaking of the overall aromaticity upon O-ring opening. The transformation of xanthylium salts to acridinium ions is thus more challenging but it remains possible as exemplied by the groups of Laursen, Müllen or recently Nicewicz, usually in moderate yields. [12][13][14][15][16] The amenability of this transformation to non-planar chiral polyaromatic systems, and helicenes in particular, was questionable. In fact, with strained substrates like helicenes, the occurrence of more stable (sterically less-demanding) ring-opened biaryl intermediates might disfavor the nal ring closure. Conversely, extension of this strategy to cationic oxa helicenes would allow the rapid construction of libraries of aza helicenes, [17][18][19] compounds that are attractive in regard of their applications in chemistry and at the interface of biology and physics. [20][21][22][23][24][25] This methodology would also be an interesting addition to the eld of late-stage functionalization that remains challenging in polyaromatics. [26][27][28][29][30][31][32][33] Finally, enantiospecic substitutions of O by N atoms within the helicene outer rim would be a novelty in the helicene eld. Herein the pyrylium to pyridinium route is extended to cationic [6]helicenes, 34,35 and the direct transformation of dioxa 1 into azaoxa 2 and ultimately into diaza 3 is demonstrated (Fig. 1, bottom). The reaction conditions are sufficiently mild to introduce tailored and/or chemically sensitive functional groups (24 examples). From 2 to 3, the reaction is enantiospecic (es up to 97%, retention of conguration), which is in stark contrast to the conversion of 1 into 2 for which an almost racemization occurs (es 3%). In-depth experimental and computational investigations of the reaction proles reveal, in this latter case, the occurrence of propeller intermediates that can racemize through a series of two ring ip processes. 36 Aer adaptation of the experimental conditions to limit such deleterious pathways, reaction 1 to 2 then occurs with an effective transfer of the stereochemical information (es 57-61%).

Results and discussion
In the initial report, 34 hetero helicenes 1, 2 and 3 were prepared using a common acyclic triarylcarbenium precursor and then three different ring closing routes, each at very high temperatures (>170 C, Schemes S1 and S2 †). The synthesis of dioxa 1 was particularly efficient as it was obtained in 95% yield in 5 min in molted Pyr$HCl. For the construction of aza helicenes 2 and 3, it was only possible to introduce amines with linear alkyl chains in low to moderate yields; amines substituted with (sensitive) functional groups were incompatible. In this context, the possibility to use dioxa 1 as a unique precursor to prepare derivatives 2 and 3 in one step via a "pyrylium to pyridinium" strategy was deemed not only interesting but also advantageous considering the ready availability of 1 in large scale and high yield by the previous method. This was put to the test. Experimentally the addition of an excess of n-propyl amine (25 equiv.) to a solution of dioxa 1 in NMP resulted in an immediate color change from red to blue. Upon heating at 70 C for 7 h, the formation of diaza 3a was evidenced (ESI-MS monitoring, 13% isolated yield). Aer this result, reaction conditions were optimized (Tables S1-S3 †). A reduction of the amount of amine (3 equiv.) and the use of benzoic acid as additive (1.5 equiv.) afforded the preferential formation of azaoxa 2a in 3 h at 60 C in 38% yield. 37 Upon increasing the reagents' stoichiometry (25 equiv. of amine and 12.5 equiv. of benzoic acid) along with a prolonged reaction time (7 h) at 70 C, dioxa 1 was converted into diaza 3a in 47% yield. The same reaction conditions also afforded diaza 3a from azaoxa 2a in 36% yield ( Fig. 1, R ¼ n-propyl).
With the mild conditions developed above, the scope of the reaction was extended. The preparation of azaoxa derivatives 2b to 2i was investigated (Fig. 2). A focus was given to the introduction of residues which were not tolerated before. For instance, n-octyl 2b was obtained in 26% yield. More sterically demanding i-propyl amine yielded 2c (20% yield). Redox sensitive benzyl and allyl amines were readily introduced (compounds 2d and 2e, 20% and 23% respectively). 15,38,39 Alcohol, carboxylic acid and ester functional groups were also compatible with the reaction conditions (compounds 2f-2h, 29%, 29% and 25% respectively). Finally, PEG functionalized azaoxa helicene 2i was obtained in 25% yield.
Importantly, reactions 1 / 2a, 1 / 3a and 2a / 3a were then performed with enantioenriched substrates to determine, as mentioned earlier, if the substitution(s) of the O by N atoms could occur enantiospecically. This was in fact debatable as the S N (ANRORC) mechanism offers a priori many possibilities for a loss of enantiomeric purity aer ring opening steps in particular. Practically, rac-1, rac-2a and rac-3a were resolved by chiral stationary phase (CSP) HPLC into the corresponding M Fig. 1 Top: conversion of pyrylium to pyridinium ions via sequences of ring openings and ring closures. Bottom: enantiospecific transformations of cationic dioxa and mono oxa [6]helicenes to the corresponding aza derivatives.  and P enantiomers (ee 97-99%). Compounds 1, 2a and 3a are congurationally stable at elevated temperatures, up to 110, 150 and >180 C specically. 34,40 The absolute conguration of the enantiomers of 1 and 2a was unknown prior to this study. It was established by vibrational circular dichroism (VCD) thanks to the conformational rigidity of the compounds that helped the determination (see ESI, Fig. S18-S23 †). Interestingly, in this series which also includes 3a, 40 the dextro and levorotatory enantiomers are of P and M congurations irrespectively of their dioxa, azaoxa or diaza nature.
Then, attempts to prepare 2a starting from enantiopure (+) and (À)-1 led unfortunately to the formation of (+)-P-and (À)-M-2a in only marginal enantiomeric excesses (ee 3%). Similarly, 3a was obtained from enantiopure (+)-and (À)-1 in almost racemic form (4% and 0% ee for the (+)-and (À)-enantiomers respectively). In striking contrast, conversion of azaoxa (+)-P-and (À)-M-2a (ee 99 and 98%) provided diaza (+)-P-and (À)-M-3a in 96 and 93% ee, indicating this time a very high level of enantiospecicity for the O to N exchange (es 97-95%). The three transformations detailed above, and 2a / 3a in particular, occur hence with retention of conguration. 41 With these results in hand, experimental and theoretical investigations were performed to determine the mechanism. The aim was to understand the reactivity pathways including their stereochemistry, measure the importance of additives like benzoic acid or offer prospects on how to limit the racemization. Various intermediates, products of trapping experiments and reaction processes were identied. They are summarized in Fig. 5 and specic details are given the following paragraphs.

Mechanistic evidence
First, dioxa helicene 1 was treated at ambient temperature with an excess of n PrNH 2 which led to one intermediate, named A, almost instantly and quantitatively. In the absence of benzoic acid and in the given reaction time (3 h), the reaction did not proceed further. To our surprise, any attempt to isolate this intermediate aer aqueous workup led to the complete recovery of starting dioxa 1. The reaction 1 $ A is thus fully reversible in the presence of water. 42 Detailed in situ NMR spectroscopy and MS analyses revealed that A possesses a biaryl structure with the N atom connected to the naphthyl ring as a result of a regioselective O-ring opening (see ESI, Fig. S1 †). The structure of A was indirectly conrmed by the isolation and X-ray structural analysis of leuco compound 4 (see ESI and CCDC 1908256 †); this derivative results from the addition of an excess of triuoroacetyl chloride to A (Fig. 5). Also, if a different amine is added to A, then full substitution of the rst amine by the second is evidenced in 1 H NMR spectroscopy. This exchange forming intermediate A 0 shows (again) the reversibility of amine additions as observed experimentally by the replacement of i Pr by n Pr residues (A $ A 0 , Fig. S2 and S3 †).
Of particular interest for the stereochemistry of the process, (S)-methyl-tert-butylamine was added to enantiopure (À)-Mdioxa [6]helicene 1. Upon addition, a single set of NMR signals was observed for intermediate A (Fig. 6, spectrum (a), see also Fig. S4 in the ESI †). For instance, signals of protons H 7 and H 9 appeared as simple doublets at 6.23 and 5.98 ppm respectively. However, aer 3 h at 60 C (Fig. 6, spectrum (b)), all the mentioned doublets, and some other signals, were split in two. The crude, which was composed initially of a single species (spectrum (a)) has now become a mixture of two stereoisomers in a 88 : 12 ratio. 43 Importantly, spectrum (b) is similar and opposite to spectrum (c) that is obtained upon the reaction of (S)-methyl-tert-butylamine with (+)-P-1. In this latter case, the  presence of two diastereoisomers also occurs aer some time at 60 C. The diastereomeric ratio of 88 : 12 (spectrum (b)) is now inverted to a 21 : 79 value (spectrum (c), see also Fig. S5 in the ESI †). Clearly, in the span of the reaction, an epimerization/ racemization process occurs at the stage of intermediate A but only to a minimal extent; some further steps must happen to account for the almost complete racemization observed at the end of the process. In Fig. 6 are represented the geometry of the stereoisomers of A obtained upon addition of the chiral amine onto (À)-M-1; the conguration of the epimers was determined in the course of the study (vide infra).
Finally, it was established that A is a productive intermediate as, in the presence of benzoic acid (1.5 equiv.) at 60 C (3 h), A transforms itself into the corresponding helicene 2a. The presence of the carboxylic acid is essential for the way forward and its role in the promotion of the aza ring closure will be clearly demonstrated by the theoretical analysis (vide infra). Conversely the use of sodium benzoate alone provided 2a in only 4% yield (see Table  S3 †). Other acids like n PrNH 3 $Cl (3 equiv.) did not induce the formation of the azaoxa [6]helicene.
In conclusion for this section, experimental evidence was found to shed some light on the mechanism of the transformation of 1 to 2a, and on the existence of intermediates of type A or A 0 . Structural and energetical aspects of the process were however hard to apprehend. Care was thus taken to perform an extensive computational survey of the reaction pathway which is depicted in Fig. 7 starting from (M)-1 (energy prole). 44

Theoretical study
Computationally, the experimental excess of amine (3 equiv.) was modelled using a bis-amine cluster. The nucleophilic attack, which occurs exclusively on the naphthyl subunits, can happen on the Re and Si faces of 1, and both approaches were simulated. The reaction that occurs on the Re face of (M)-dioxa 1 proceeds through a late transition state TS1 (DE ‡ ¼ +15.0 kcal mol À1 ) which models the dearomatization of the naphthyl moiety and leads to the unstable ammonium adduct B (DE ¼ +14.6 kcal mol À1 ). The following carbon oxygen bond cleavage occurs via a 6-membered transition state TS2 at 21.8 kcal mol À1 and it generates stabilized intermediate (aR)-A that possesses an ortho quino methide character (DE ¼ À6.0 kcal mol À1 ). 45 Importantly, a strong association with the bis-amine cluster occurs and the positive charge is now localized on the H-bonded ammonium adduct. The attack on the Si face follows a similar behavior although in this case, intermediate A is highly distorted and clearly unfavorable energetically (DE(A Re ) ¼ À6.0 kcal mol À1 vs. DE(A Si ) ¼ +5.7 kcal mol À1 , Fig. S24 †). These observations are in adequacy with the experimental results and the existence of a highly stereospecic formation of A, at the beginning of the reaction in particular. It is rooted in the facial Re selectivity of the O-ring opening of dioxa (M)-1 and, hence, in the stereocontroled formation of biaryl (aR)-A.
For the subsequent conversion of intermediate A to azaoxa 2, an explicit carboxylic acid molecule is required computationally to reproduce well the experimental observations. Calculations show that the replacement of the ammonium ion by the carboxylic acid destabilizes the supramolecular adduct, from À6.0 to 1.1 kcal mol À1 . Once the acid coordinated, the desired N-ring closure happens via TS3 (DE ‡ ¼ +22.9 kcal mol À1 ) and forms C. During the C-N bond formation, the carboxylic acid acts as a proton shuttle helping the approach of the nucleophile by promoting simultaneously the deprotonation of the amino group and the protonation of the oxygen atom. Then, dehydration of C occurs in a barrier-less process and provides azaoxa (M)-2 that is more stable than the starting dioxa (M)-1 by 14.4 kcal mol À1 . 46 Overall, the transformation from the ammonium adduct (aR)-A$ + H 3 NR to C constitutes the rate determining step (DE ‡ ¼ +28.9 kcal mol À1 ). An analogous computational study was performed starting from (P)-1 and leading to the formation of (P)-2; throughout the study, an While this theoretical approach (Fig. 7) accounted well for the reactivity prole of the reaction, the above calculations did not rationalize the isolation of 2 in almost racemic form starting from either (M)-1 or (P)-1. In fact, as detailed above, the O-ring opening of dioxa (M) or (P)-1 is enantiospecic and leads to intermediates (aR) and (aS)-A which in turn are converted to azaoxa (M) and (P)-2 respectively. Also, as shown in the experiments with (S)-methyl-tert-butylamine (Fig. 6), intermediates of type A are quite congurationally stable in basic medium at 60 C. The origin of the racemization was thus debatable and the possibility of an equilibrium between intermediates (aR)-A and (aS)-A induced by an acidication was considered (Fig. 8). As a matter of fact, upon protonation, (aR)-A is transformed into (aR)-A$H + in which the positive charge is localized on the xanthenium core. This species is highly electrophilic and, as a consequence, it reacts further with the excess of amine cleaving the second ether linkage. Again, this attack can formally occur from either Re and Si faces of (aR)-A$H + leading to the formation of two three-bladed propellers, named L-D and D-E; the L and D prexes corresponding to the le and right handed propeller congurations respectively as the additions behave stereoselectively. In D, the two amino groups point inward, i.e. toward the central phenyl ring while in E, one amino group is directed toward the center and the other outward. The global energy proles, depictured in Fig. 8, clearly indicate the preferential kinetic formation of L-D over D-E (see the ESI for more details, Fig. S26 and S27 †). 47 However, the enantiomerization of L-D into D-D, through TS5, is highly disfavored (DE ¼ +35.3 kcal mol À1 ) and follows an unusual one ring ip route in which the transition state deviates from ideality; the distortion is brought by the steric hindrance of the naphthyl units. 36,48 Alternatively, stepwise isomerization of L-D into D-D was computed through a rst facile isomerization of L-D into D-E that follows a two ring ip mechanism, via TS6 (DE ‡ ¼ +6.7 kcal mol À1 ), in which one of the larger naphthyl group remains in the main plane. Subsequent enantiomerization from D-E to L-E, also modeled through a two ring ip mechanism with TS7 located at +16.9 kcal mol À1 , then accounts for the racemization observed under acid conditions. In fact, (aS)-A$H + is readily obtained via the antipodal route from L-E and D-D leading to a globally facile (aR)-A $ (aS)-A equilibrium under acidic conditions. 49 Noteworthy, the isomerizations of L-D into D-D and of D-E into L-E were also computed in the absence of the extra H-bonded amine; quite higher energies for the associated TS were obtained in both cases (see ESI, Fig. S28-S30 †). 50 It is also worth to mention that propeller intermediates D and E were detected by in situ mass spectrometry and their presence explain as well the formation of dibenzoacridinium 5 which can be isolated as a side product (10-30% yield, Fig. 9). 51 In fact, a relatively simple pathway was computed for its formation from D-E via intermediate F and an acid-promoted cyclization/deamination (Fig. 9); the overall kinetic barrier is compatible with the experimental observation of 5 (DE ‡ ¼ +17.5 kcal mol À1 ).

Toward higher level of enantiospecificity
According to the mechanistic scheme detailed above (Fig. 8), the racemization observed during the formation of azaoxa 2 occurs via three-bladed propeller intermediates of type D and E that necessitates both an acid activation and an excess of primary amine to cleave all ether linkages. To overcome this deleterious effect, it was decided to employ a strict stoichiometric amount of primary amine (for the O to N ring exchange) along with an excess of non-nucleophilic Hunig's base to buffer the medium. Furthermore, to avoid the build-up of species A (responsible for the racemization), a slow addition of this combination of nucleophilic and non-nucleophilic amines was considered. With this set of conditions (Fig. 10), the enantiospecicity of the transformation from 1 to 2a was in part recovered (es 61% and 57% with (M)-1 and (P) À1 respectively, Fig. S12 and S13 †). The reactions occurred with retention of the conguration as it could be expected from the computational study. Unfortunately, the yield of (M) or (P)-2a was drastically lowered to 6% only.
In addition, as mentioned earlier, conversion of (+)-P-and (À)-M-2a provides diaza (+)-P-and (À)-M-3a with very high enantiospecicity (es 97-95%) for the O to N exchange. Computationally, the reaction proceeds in a similar fashion to that from 1 to 2. Ring opening of M-2 with an excess of primary amine leads to the formation of intermediate (aR)-G located at À4.4 kcal mol À1 (Fig. 11a and S34 † for details). 52 This ring opening is regioselective (Fig. S8 †) and it was evidenced experimentally by the trapping of G with 4-chlorobenzoyl chloride to afford 6 (63%, Fig. 11b and CCDC 1908257 †). This step is clearly stereoselective as evidenced by 1 H-NMR monitoring upon addition of (S)-methyl-tert-butylamine acting as stereochemical probe. As detailed in Fig. S9, † the NMR spectrum of the addition adduct displays signals for a single stereoisomer which is stable at 70 C for 7 h. Not surprisingly, the subsequent aza ring closure from (aR)-G to M-3 (located at À16.6 kcal mol À1 ) is also facilitated by the presence of benzoic acid that promotes the proton transfer and the nal dehydration. Yet, the 2 / 3 transformation differs from 1 / 2 in two aspects. First, the O-ring opening is rate determining this time (TS10 +27.9 kcal mol À1 ). Second, and of utmost importance for the enantiospecicity, protonated intermediate G$H + does not react with an excess of primary amine (Fig. 11c). Both Re and Si face approaches and subsequent reactivities would proceed with very high activation energies (+48.6 and +54.3 kcal mol À1 , respectively) and would be thermodynamically disfavored with corresponding propeller adducts located at +43.1 and +40.5 kcal mol À1 (see ESI, Fig. S36 and S37 †). This explains the large stereochemical difference between reactions 1 / 2 and 2 / 3something that was hard to predict at the onset of the project.

Conclusions
Herein the transformation of pyrylium rings into pyridinium cycles has been extended to the eld of cationic helicenes. The replacement of oxygen atom(s) at the outer rim by stabilizing amino groups proceeds via S N Ar mechanisms. The reported reaction conditions are sufficiently mild to be benecial for the introduction of functional groups that were not tolerated previously (24 examples). In-depth mechanistic studies, experimental and theoretical, demonstrate that the process is enan-tiospecic (es up to 97% for 2 / 3) and occurs with retention of conguration. Near racemization can however happen via three-bladed propeller intermediates, if starting from dioxa 1 in particular. Based on the mechanistic understanding, corrective measures were applied to recover a large amount of the enan-tiospecicity for the 1 / 2 process (es 57-61%). Care will be taken to utilize this late-stage functionalization strategy for future developments and applications of these original chromophores and uorophores. 53-61

Conflicts of interest
There are no conicts to declare.