Pnictogen-bonding catalysis: brevetoxin-type polyether cyclizations† †Electronic supplementary information (ESI) available: Detailed procedures and results for all reported experiments. CCDC 1999316–1999326. For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/d0sc02551h

This study marks chemical space available for pnictogen-bonding catalysis, and demonstrates that reactivity accessible in this space is unique.

Pnictogen and tetrel bonds refer to non-covalent interactions 1-7 between electron-rich acceptors and s holes on group V (15) and group IV (14) atoms, respectively . 3,4 s Holes are regions with positive electrostatic potential appearing at the side opposite to s bonds to electron-withdrawing substituents R. Compared to better established halogen 5 and chalcogen bonds, 6 pnictogen-and, although less important in this study, also tetrel-bond donors are of higher valency and thus offer more s holes. Moreover, pnictogen-bond donors can be interconversion-free 8 stereogenic centers 9 and at the origin of chiral axes. 10 s-Hole interactions are primarily electrostatic. They strengthen with the depth of the s hole, which relates to polarizability, thus increases downward and toward the le in the periodic table. 1, 2 Here, we suggest to dene pnictogen-bonding catalysis as the non-covalent, supramolecular counterpart of classical covalent Lewis acid catalysis ( Fig. 1d and e). This is analogous to hydrogen-bonding and Brønsted acid catalysis, with interactions that become too strong transfer their proton and form new covalent bonds ( Fig. 1f and g). Similarly overachieving cation-p and anion-p interactions can continue into electrophilic and nucleophilic aromatic substitution, respectively. 11 Group 15 Fig. 1 (a) Region of interest in the periodic table. (b and c) General structure of tetrel and pnictogen-bond donors (D; D III : trivalent; D V : pentavalent) interacting with their acceptors (A); blue circles, s holes; red orbitals, lone pairs. (d) Pnictogen-bonding catalysis defined as a non-covalent counterpart of (e) Lewis acid catalysis (La), analogous to (f) hydrogen-bonding and (g) Brønsted acid catalysis (Ba, conjugate base: Bb À ). S, substrate; P, product; etc.*: ligand (L) exchange, proton release from S upon addition to La, etc.
Lewis acids, however, have been studied exhaustively as reagents and catalysts. 2,[12][13][14] Except for a few recent examples, 2,12,13 possible contributions from pnictogen bonds to these activities were either ignored or alluded to from different points of view. 14 The question thus arises whether or not pnictogenbonding catalysis is just a weak form of Lewis acid catalysis and thus essentially trivial. The differences between hydrogenbonding and Brønsted acid catalysis are understood. The differences in structure and charge distribution between noncovalent pnictogen bonding and covalent ligand addition/ exchange ( Fig. 1d and e) further support that pnictogenbonding catalysis should exist and matter. In the following, we show that this is indeed the case.
Most catalyst candidates 1-13 were readily accessible in a few steps from commercially available substrates (Fig. 2a, Schemes S1-S3, † X-ray structures: Fig. 2b, S75-S86 †). 2 Only Bi 7 was too unstable in our hands. 15 Stibine 1 was obtained by nucleophilic substitution of SbCl 3 with aryl anions derived from bromobenzene 14 (Fig. 2a). Sb(III) 1 was oxidized with chloranil (Ch) 15 12 to give stiborane 2. Molecular electrostatic potential surfaces (MEP, BP86-D3/def2-TZVP level) conrmed 12 that this oxidation converts the three deep s holes on Sb(III) 1 into one deep s hole on Sb(V) 2 ( Fig. 2b and c). Consistent with increasing polarizability, 1,2 Sn(IV) 3 excelled with four deep s holes, whereas the s holes of the smaller Ge(IV) 4 were not accessible. In 1-4, the ortho uorines of the original per-uorinated 5 2 were replaced by hydrogens because the crystal structure of 5 indicated the existence of Sb-F pnictogen bonds that weaken and obstruct all s holes (Fig. 2b). The acidic ortho hydrogens in 1-4 should further assist s-hole interactions with proximal C-H/A bonds (see below).
The structural complexity of epoxide-opening ether cyclizations 16-18 was considered as ideal to identify possible differences between pnictogen-bonding and Lewis acid catalysis. Initial studies focused on monomers 16-19 (Fig. 3a). According to the Baldwin (B) rules, their 5-exo-tet cyclization into oxolanes 20-23 is preferred over 6-endo-tet anti-Baldwin (A) oxanes 24-27. 16-18 Aer one day under standard conditions, Sb(FP 345 ) 3 1 converted 81% of cis epoxide 17 17,18 into (B)-21 (Table 1, entry 1). Reactions were much slower with Bi 6, Sn 3 and Ge 4 (entries 2-5). However, tetrel-bonding Sn 3 remained operational as catalyst, as conrmed with high conversion at 20 mol% (entry 4). FP 2-6 5 and 8 were unstable, supporting that the ortho hydrogens in FP 345 minimize not only s-hole obstruction but also catalyst decomposition.  With the permethylated monomer 19, 17,18 stibine 1 produced signicant amounts of anti-Baldwin product (B/A z 6:4, entry 6, 7). Access to anti-Baldwin selectivity depended on substrate (18 > 19 > 17, 16) and catalyst structures (2 > 13 > 12 > 1-9 > 11 > 10, Tables 1, S2-S6 †). Some small but signicant irregularities in dependence of endo/exo selectivity on catalyst structure nicely illustrated the inuence of the specic environment in the respective binding pockets with pnictogen-bonding catalysis. Proximity effects in binding pockets is a hallmark of supramolecular catalysis, much appreciated in hydrogen-bonding catalysis to access stereoselectivity, and absent in covalent "general" Brønsted acid catalysis, which is independent of the acid used. Very fast conversion on the "cyclopean" s hole of Sb(V) stiborane 2 allowed for meaningful studies at lower temperature as well as lower catalyst loading, which caused the expected increase in anti-Baldwin selectivity (entry 8,9). The ortho-uorinated Gabbaï original 11 12 failed to break the Baldwin rules, as did Lewis and Brønsted acid controls (entries 10-12).
Access to anti-Baldwin cascade cyclizations was of general interest also because, in nature, Baldwin oligomers such as the monensin-like ionophores are complemented by the rich family of brevetoxin-like ladder oligomers. [16][17][18] Minimalist cascade cyclizations were explored with a cis-trans mixture of diepoxide 28 to maximize the number of constitutional and stereoisomers contributing to catalyst ngerprints (Fig. 3b and c). 1 H NMR spectroscopy and X-ray analyses of at least partially puried products and comparison with literature data 17 allowed us to assign NMR signals to isomers 29-33 (Fig. 3b-e, S10-S13 †). The endo/exo selectivity was estimated from the ratio of characteristic peaks in the spectra of the product mixtures. Isolated, easy to integrate peaks of cis,cis-(BB)-30 and trans,trans-(AA)-33 were selected because they originate from the same substrate isomer, i.e., trans,syn-28 (Fig. S13 †). The results are described as BB/AA ratios (Table S11 †).
The cyclizations of trimers 34 and tetramers 35 were characterized mainly by comparing their 1 H NMR ngerprint to those of dimers (Fig. 4, S18 and S20 †). The products obtained from 34 with AcOH showed a cluster of signals between 3.75 $ 4.10 ppm, characteristic of Baldwin products (Fig. 4b). With pnictogenbonding catalysts 1 and 2, the appearance of up-eld shied peaks revealed anti-Baldwin selective cyclizations. NMR ngerprints of cascade cyclized tetramer 35 showed the same trends at increased complexity, containing up to 16 constitutional isomers, from (B 4 )-36 to (A 4 )-37 (Fig. S17 †). In NMR ngerprints beyond dimers, differences between pnictogen-bonding and Lewis acid catalysis remained visible but became increasingly difficult to quantify. Gas chromatography (GC) proved more revealing, con-rming the lessons learned on the dimer level: the reactivity of supramolecular pnictogen-bonding catalysts Sb(III) 1 and Sb(V) 2 differs from covalent Lewis acid catalysts like SbCl 3 , and the former excel with an almost complete suppression of all-Baldwin products (Fig. 4c, S19 and S21 †).    (c) GC fingerprints of 34 converted with AcOH, SbCl 3 , 1 and 2 (c, Baldwin products: t R 6.5-6.9 min, anti-Baldwin: t R > 6.9 min, substrates: t R < 4.0 min).
Computational studies were complicated by the high number of possible stereochemical and conformational isomers (Fig. S22-S28 †). However, signicant isolate observations could nevertheless be secured. Firstly, the binding of epoxide 19 to Sn 3 revealed a formal tetrel bond, 1,4 shorter than the sum of vdW radii (3.69Å) and longer than covalent bonds (2.03Å, Fig. 5a). The smaller Ge 4 preserved the bidentate CH/ O interactions but lost the tetrel bond (3.66Å, vdW 3.62Å). These ndings were consistent with accessible s holes on the MEP surface of Sn 3 but not Ge 4 (Fig. 2c). Although weak and presumably precedented in the Lewis acid literature, 1,4,14 the cyclization of 17 with Sn 3 could thus be considered as one of the rst examples of explicit tetrel-bonding catalysis (Table 1, entry 3, 4). Also worth noting were more than one tetrel bond with oligoepoxides ( Fig. S27 †), and four intermolecular tetrel bonds in the crystal structure of Sn 3 but not of Ge 4 (Fig. S77 †).
Most important were pnictogen bonds between epoxides of 35 to all three s holes of 1 (Fig. 5b). This was not trivial because each pnictogen bond formed weakens the remaining s holes. 7 This nding thus supported contributions from multivalency, including entropy-driven substrate destabilization, 17-19 to catalysis. The last epoxide is engaged in lonepair-p interactions, ready to occupy the s hole liberated by the rst ring formed. Finally, a single pnictogen bond to 34 conrmed the loss of multivalency of 2, which was compensated by CH/O and lonepair-p interactions. Affinity gradients in the resulting "triad" would be compatible with 2 crawling along the antiparallel epoxides in snake-like foldamer tracks (Fig. 5c).
In summary, with the hypersensitive epoxide-opening polyether cascade cyclizations, we show that pnictogen-bonding catalysts are more than just weak Lewis acids. Naturally slower and not autocatalytic like on p-acidic aromatic surfaces ( Fig. S6-S8 †), 17,18 the distinctive characteristic of pnictogenbonding catalysis is the breaking of the Baldwin rules. Important differences in regio-and stereoselectivity exist also between multivalent Sb(III) and hypervalent Sb(V) pnictogen-bonding catalysts. These initial results on pnictogen-bonding catalysis thus support the general expectation that the integration of unorthodox interactions 20 will provide access to new reactivity. Attractive perspectives include antimony as stereogenic center 9,10 combined with multivalency, and the integration into more advanced functional systems. 11,21 The discussion about the difference between pnictogenbonding and Lewis acid catalysis launched in this report will continue and spread into other, less affected s-hole interactions.
The need for such a distinction will have to be conrmed, and the tantalizing question where and how to draw the line will persist, particularly considering the underlying continuum and the dependence on the involved partner, either pnictogen-bond acceptor or Lewis base (i.e., every weak enough pnictogen-bond acceptor will turn also a strong Lewis acid like SbCl 3 into a pnictogen-bond donor 22 ). Differences in bond length, changes in geometry, charge distribution or deprotonation (Fig. 1) are all convincing but indirect measures to draw this line; direct functional differences as identied in this study will ultimately be needed. What remains for certain is that the IUPAC denition restricts Lewis acids to reactions and covalency, 23 while extrapolation from halogen bonding 24 denes pnictogen-bond donors as the supramolecular counterpart, i.e., electrophilic regions that interact non-covalently, rather than react covalently like Lewis acids. The comparison with non-covalent hydrogen-bonding catalysis and covalent Brønsted acids, presumably applicable to all s-hole catalysis, 25 could thus help in this situation because the same ambiguities exist but they are understood and appreciated. However, despite all compelling analogies, only the future will tell if s-hole catalysis in general and pnictogen-bonding catalysis in particular will also become as important as the complementary hydrogenbonding catalysis.

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