An unusually short intermolecular N—H⋯N hydrogen bond in crystals of the hemi-hydrochloride salt of 1-exo-acetamidopyrrolizidine

In the crystal, two AcAP molecules related by a crystallographic twofold axis link to H+ and Cl− ions lying on the rotation axis, thereby forming N—H⋯N and N—H⋯Cl⋯H—N hydrogen bonds. The first of these has an unusually short N⋯N separation of 2.616 (2) Å: refinement of different models against the present data set could not distinguish between a symmetrical hydrogen bond (H atom lying on the twofold axis and equidistant from the N atoms) or static or dynamic disorder models (i.e. N—H⋯N + N⋯H—N).


Chemical context
In the course of our ongoing studies on the biosynthesis of loline alkaloids (Schardl et al., 2007;Pan et al., 2018), we had occasion to prepare 1-exo-acetamidopyrrolizidine (C 9 H 16 N 2 O; AcAP) by reduction of the corresponding oxime ( Fig. 1) (Pan et al., 2014). As part of our effort to prove unambiguously that the major diastereomer obtained in this reaction was indeed the exo diastereomer, we attempted to obtain crystals of AcAP that were suitable for X-ray analysis. We obtained sufficiently high-quality crystals by recrystallization from CHCl 3 , but to our surprise, the analysis showed that the crystals were not AcAP, but 2AcAPÁHCl, with the HCl presumably originating from amine-promoted decomposition of CHCl 3 . This paper describes the structure of crystalline 2AcAPÁHCl, which features an unusually short + N-HÁ Á ÁN interaction.

Supramolecular features
The primary structural motif within crystals of 2AcAPÁHCl consists of a pair of homochiral AcAP molecules hydrogen bonded to H + and Cl À ions about a crystallographic twofold axis of space group C2/c. Based on electron density alone, the H + cation and Cl À anion both appear to be located on the twofold axis. As a result of the twofold symmetry axis, the hydrogen bond N4-H4NÁ Á ÁN4 i [symmetry code: (i) Àx + 1, y, Àz + 1 2 ] appears to be symmetric, and since d NÁ Á ÁN = 2.616 (2) Å (Table 1) An ellipsoid plot (50% probability) showing the molecular structure of 2AcAPÁHCl. Unlabeled atoms are related to their labeled counterparts by a crystallographic twofold rotation (Àx + 1, y, Àz + 1 2 ). The unusually short N-HÁ Á ÁN hydrogen bond between the pyrrolizidine N atoms is highlighted by open dashed lines. Weaker N-HÁ Á ÁCl hydrogen bonds between the acetamide NH group and the Cl À anion are shown as dotted lines.

Figure 1
Synthesis of AcAP.

Figure 4
A difference-Fourier map showing elongated electron density for the hydrogen atom involved in the short N-HÁ Á ÁN hydrogen bond of 2AcAPÁHCl.
the twofold axis and is equidistant from N4 and N4 i , the refined NÁ Á ÁH interatomic distance appears unusually long at 1.3080 (12) Å . The whole hydrogen-bonded ensemble makes an R 3 3 (12) supramolecular motif. For the short N4-H4NÁ Á ÁN4 i hydrogen bond, a difference-Fourier map (Fig. 4) clearly shows an elongated region of electron density centered on the twofold axis corresponding to the position of the refined H4N hydrogen atom. The true location of H4N, however, remains ambiguous based on X-ray data alone. Possibilities include strictly symmetric (i.e., on the twofold axis exactly halfway between N4 and N4 i ), statically disordered (i.e., 50% on each of N4 and N4 i ) or dynamically disordered (i.e., exchanging rapidly between N4 and N4 i ). Alternative strategies for H4N inclusion in the model refined equally well (see Refinement, below), so we settled on the simplest approach, following the recommendations of Fá bry (2018). Nonetheless, the ambiguity prompted us to analyze the potential energy surface of the H4N position via computational methods for single versus double energy well character (see Computational analysis, below). In addition to the strong N-HÁ Á ÁN hydrogen bond, weaker N2-H2-Cl1-H2 i -N2 i interactions link the twofold-related acetamide groups to the Cl À anion [NÁ Á ÁCl = 3.2263 (11) Å ]. The twist of the twofoldrelated hydrogen-bonded pyrrolizidine moieties relative to each other, as defined by the torsion angle C8-N4Á Á ÁN4 i -C8 i is À69.82 (16) . The almost planar acetamide group forms a dihedral angle with its twofold-related counterpart of 20.18 (3) . The only other intramolecular interactions are van der Waals contacts. Estimates of the relative fractions of intermolecular contacts between individual atom types derived from a Hirshfeld-surface analysis using Crystal-Explorer (Turner et al., 2017;Tan et al., 2019) are complicated by the ring disorder and by the N4-H4NÁ Á ÁN4 i hydrogen bond. Nevertheless, all contacts appear to involve hydrogen atoms, with the overwhelming majority being HÁ Á ÁH (65.3%). The Cl À anion and the acetamide O atom each reside in pockets surrounded by hydrogen atoms, giving HÁ Á ÁCl/ClÁ Á ÁH (16.2%) and HÁ Á ÁO/OÁ Á ÁH (12.4%), with the remainder being NÁ Á ÁH/HÁ Á ÁN and CÁ Á ÁH/HÁ Á ÁC contacts.

Database survey
A search of the Cambridge Structure Database (version 5.40, Nov. 2018;Groom et al., 2016) on the bicyclic pyrrolizidine core of AcAP yielded 584 hits. Of these, 41 are protonated at the ring N atom, but only three of those bear a substituent (other than H) at the 1-position of the pyrrolizidine ring system (assuming standard IUPAC ring numbering). CSD entry BRPYLZ (Wilson & Sawicki, 1979) is a bromide salt of a bromine derivative, and CPYRZD (Soderberg, 1971) is a zwitterion, with a carboxylate group at the 1-position and a bromo substituent at the 2-position. The relative stereochemistry of BRPYLZ and CPYRZD, however, are different from AcAP. CSD entry EDOTUP (Bhardwaj et al., 2017) was a precursor to AcAP, and is the most closely related currently deposited structure.
The most striking feature of 2AcAPÁHCl is the unusually short N-HÁ Á ÁN hydrogen bond. A CSD search on the fragment 'C-N(X)-H-N(X)-C 0 , where 'X 0 denotes 'any group', gave 45 hits. Rejection of cases where apparent close NÁ Á ÁN distances were due to disorder (four entries), and those in which the N-bound H atom was missing from the model (two entries), left 39 structures, of which three were duplicates. In the remaining 36 structures the N-HÁ Á ÁN hydrogen bonds are intramolecular in 22 and intermolecular in 14. The closest NÁ Á ÁN separations occur in the intramolecular N-HÁ Á ÁN hydrogen-bonded structures, the shortest being 2.419 Å in EBOKOV (Wilkes et al., 2000). However, in these intramolecular cases, the NÁ Á ÁN separation is largely dictated by the intramolecular geometry, which effectively forces the donor and acceptor N atoms into close proximity. Of the 14 CSD entries from the search that have intermolecular N-HÁ Á ÁN hydrogen bonding, the closest NÁ Á ÁN separation occurs in BECHOG (Glidewell & Holden, 1982), in which a bis(4methylpyridine)hydrogen(I) cation sits on an inversion centre, giving an apparently symmetric N-HÁ Á ÁN hydrogen bond with NÁ Á ÁN distance of 2.610 (15) Å , and ROHTIR (Bock et al., 1997), for which the asymmetric unit contains two separate halves of a methylammonium-methylamine cation, [CH 3 NH 2 -H-NH 2 CH 3 ] + , each sitting on centres of inversion, giving NÁ Á ÁN distances of 2.620 and 2.641 Å . The NÁ Á ÁN separation in 2AcAPÁHCl is similar, at 2.616 (2) Å , although the difference is not significant, and well within the quoted precision estimate of BECHOG and the accuracy limits imposed by the spherical-atom scattering-factor approximation (see e.g., Dawson, 1964). Computed charge-density line profile from nitrogen to nitrogen in the N-HÁ Á ÁN hydrogen bond. charge density. In the case where symmetry constraints were absent, a small displacement (0.06 Å ) was applied to the hydrogen atom in the N-HÁ Á ÁN hydrogen bond to break symmetry in the initial geometry. The relaxations led to two structures, one with constrained twofold symmetry (A), and one unconstrained (B). The volume difference between these theoretical models (calculations assumed absolute zero) was negligible (A vol = 1918.85 Å 3 versus B vol = 1919.63 Å 3 ). In the symmetric model, the N-HÁ Á ÁN hydrogen atom (corresponding to H4 in the crystallographic model) is equidistant between the two nitrogen atoms (N-H = 1.290 Å ), whereas in model B the N-H distances differ (N-H = 1.194 and 1.406 Å ). This is in agreement with the computed chargedensity line profile of N-HÁ Á ÁN in structure B, as shown in Fig. 5. Structure B is calculated to be slightly more stable, but the energy difference (4.7 meV per unit cell) is vanishingly small (Fig. 6). The low energy barrier suggests dynamic disorder of the N-HÁ Á ÁN hydrogen atom.

Computational analysis
Computational details: For this periodic system, density functional theory (DFT) calculations were carried out using the Vienna ab initio simulation Package (VASP) with Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional (Kresse & Furthmü ller, 1996a,b;Kresse et al., 1994;Perdew et al., 1996). The electron-ion interactions were described with the projector augmented-wave (PAW) method (Blö chl, 1994; Kresse & Joubert, 1999). The valence electronic wavefunctions were expanded on a plane-wave basis with a kinetic energy cutoff at 520 eV, and Gaussian smearing with a width of 0.05 eV was employed. The convergence criterion of the total energy was set to 10 À5 eV in the self-consistent field loop. The Brillouin zone was sampled with a 1Â2Â2 À-centered grid. The experimental structures were relaxed until the Hellman-Feynman forces for each site were less than 0.005 eV Å À1 , and Grimme's DFT-D3 dispersion correction was applied with Becke-Johnson damping (Grimme et al., 2010(Grimme et al., , 2011.

Synthesis and crystallization
AcAP was synthesized and purified according to the published procedure (Pan et al., 2014). Crystals of 2AcAPÁHCl were obtained in the form of colorless plates by dissolving 20 mg of AcAP in 1 ml of CHCl 3 in a 10 ml round-bottom flask and allowing the solution to stand in a refrigerator for about a month.

Refinement
Crystal data, data collection, and structure refinement details are given in Table 2. Non-disordered carbon-bound H atoms were found in difference-Fourier maps, but subsequently included in the refinement using riding models, with constrained distances set to 0.98 Å (RCH 3 ), 0.99 Å (R 2 CH 2 ) and 1.00 Å (R 3 CH). Following the advice of Fá bry (2018), the hydrogen atom involved in the short N-HÁ Á ÁN hydrogen bond (H4N) was placed into difference-Fourier electron density and refined, albeit constrained to the twofold axis. An alternative model in which this H atom was allowed to ride at 50% occupancy on both N4 and N4 i [symmetry code: (i) Àx + 1, y, Àz + 1 2 ] refined equally well: the X-ray data alone being insufficient to establish a preference. The amide N-H hydrogen atom (H2N) was refined freely. U iso (H) parameters for nitrogen-bound hydrogen atoms were refined, while for Computed potential-energy surface as a function of N-HÁ Á ÁN hydrogenatom displacement from the midpoint of the two nitrogen atoms. carbon-bound H atoms, U iso (H) were set to values of either 1.2U eq (R 3 CH, R 2 H 2 ) or 1.5U eq (RCH 3 ) of the attached atom. The refined displacement parameters for the Cl À anion (e.g., Figs. 2 and 4) appear a little small compared to the rest of the structure. In addition, the largest residual difference map peaks (0.67 and 0.65 e Å À3 ) were close (0.37 and 0.47 Å respectively) to Cl1. Refinement of the anion as mixed Cl and Br gave an occupancy ratio of 0.934 (2):0.066 (2), a lower Rvalue (3.02%), and a flatter difference map (Á = 0.29/-0.19 e Å À3 ). However, the reaction included no known source of Br À , so the mixed anion model was not retained.
To ensure satisfactory refinement for disordered groups in the structure, a combination of constraints and restraints were employed. The constraints (SHELXL commands EXYZ and EADP) were used to fix overlapping fragments. Restraints were used to ensure the integrity of ill-defined or disordered groups (SHELXL commands SAME, SIMU, and RIGU). An alternative model using space group Cc (50:50 inversion twinned) was considered but rejected as it required hefty restraints and did not resolve the H4N atom ambiguity.

Special details
Experimental. The crystal was mounted using polyisobutene oil on the tip of a fine glass fibre, which was fastened in a copper mounting pin with electrical solder. It was placed directly into the cold gas stream of a liquid-nitrogen based cryostat (Hope, 1994;Parkin & Hope, 1998). Diffraction data were collected with the crystal at 90K, which is standard practice in this laboratory for the majority of flash-cooled crystals. Geometry. All esds (except the esd in the dihedral angle between two l.s. planes) are estimated using the full covariance matrix. The cell esds are taken into account individually in the estimation of esds in distances, angles and torsion angles; correlations between esds in cell parameters are only used when they are defined by crystal symmetry. An approximate (isotropic) treatment of cell esds is used for estimating esds involving l.s. planes. Refinement. Refinement progress was checked using PLATON (Spek, 2009) and by an R-tensor (Parkin, 2000 (5)  C2 0.0266 (7) 0.0247 (7) 0.0185 (6) −0.0007 (5) 0.0071 (5) 0.0009 (5)  C3 0.0240 (7) 0.0244 (7) 0.0214 (7) −0.0021 (5) 0.0041 (5) 0.0056 (5)  0.0246 (7) 0.0178 (7) 0.0239 (7) 0.0004 (5)