HBr or not HBr? That is the question: crystal structure of 6-hydroxy-1,4-diazepane-1,4-diium dibromide redetermined

A probably wrong interpretation of a previously determined structure of the title compound is discussed and an alternative more convincing description is presented.


Introduction
6-Hydroxy-1,4-diazepane (dazol) was first synthesized by Saari et al. (1971) and has been used as a tridentate facially coordinating metal-complexing agent (Liu et al., 1997a). The free ligand has been isolated as a dihydrogen bromide, C 5 H 12 N 2 OÁ2HBr, and its crystal structure has been reported [Liu et al., 1996;Cambridge Structural Database (CSD;Groom et al., 2016) refcode TOKTIW]. The authors postulated crystallization of the neutral diazepane as a free base together with two HBr molecules. From a chemical point of view, the formation of discrete HBr molecules beside a basic entity is surprising, even taking into account that the situation in the solid state does not necessarily reflect the well-known acid-base properties in aqueous solution. However, a search for molecular H-Br in the CSD (Version 5.20, 2018) gave a total of 69 hits and hence some support for the molecular model. To shed light on this discrepancy, we have: (i) investigated the chemistry of dazol in aqueous solution using potentiometric titration experiments and pD-dependent 1 H ISSN 2053ISSN -2296 NMR spectroscopy, and (ii) prepared a crystalline sample of the title compound (see Scheme), repeated the structure determination reported by Liu et al. (1996) and performed additional thermogravimetric measurements to elucidate the solid-state properties.

Synthesis and crystallization
The synthesis of the title compound was performed following the protocol given by Saari et al. (1971) with minor modifications. N,N 0 -Dazolbis(toluenesulfonamide) was prepared as described by the reaction of N,N 0 -ethylenebis(toluenesulfonamide) with 2,3-dibromopropan-1-ol. However, in the next step, the detour via the acetate proved not necessary. The two toluenesulfonamide groups could be removed directly without any loss of yield by heating the bis(toluenesulfonamide) suspended in 48% aqueous HBr to 125 C for 3 h. The clear bright-yellow solution was allowed to cool to room temperature and was then evaporated to dryness under reduced pressure. The resulting solid was washed with diethyl ether and ethanol to yield the title compound as a pale-gray solid (91%). Crystals were grown by slow diffusion of EtOH into an aqueous solution of the product which has been acidified with additional HBr. Liu et al. (1996) reported an unambiguous location of the C-, N-, O-and Br-atom positions of one dazol moiety and two crystallographically independent Br atoms. It is clear that a reliable assignment of H-atom positions is more difficult, and might even be a highly questionable task, if the high electron density of the two heavy Br atoms is considered. Unfortunately, the data set of Liu et al. (1996) is of rather poor quality. According to the CSD, the data set was recorded at room temperature. Moreover, the information provided by these authors is not really conclusive. In their Table 1 (and in the CIF available from the CSD), the H-atom positions are all listed without standard deviations and the authors stated that 'H atoms were located by geometric method except the hydroxyl one, which was oriented from difference Fourier map.' This statement seemingly indicates that the H(-Br), H(-C) and H(-N) positions have not been taken from a difference Fourier map. It is confusing that some of the C-H (1.13 Å ) and N-H (1.15 Å ) distances and H-C-H (93.2 ) angles do not fall in expected ranges. Obviously, the authors did not apply the usual riding model with fixed angles and distances. The two H-Br lengths of 1.04 and 1.08 Å are also rather short.

Refinement
To improve the quality of the data set, we performed a data collection at À131 C. The space group and lattice parameters, as well as the positional parameters of the non-H atoms, were all in agreement with Liu's report (Liu et al., 1996). Crystal data, data collection and structure refinement details are summarized in Table 1. Inspection of a difference Fourier map unambiguously yielded all of the H(-C) and H (-O) hydrogens with meaningful bond lengths and angles. The positional parameters of these H atoms were now included in a subsequent refinement using a riding model, with an appropriate restraint for H(-O) and appropriate constraints for the H(-C) atoms. At this stage, the refinement provided agreement factors of R 1 = 3.94% and wR 2 = 7.47%. The two most intense peaks, with electron densities of 0.72 and 0.70 e Å À3 , in the new difference Fourier map were located in proximity to atoms N2 and N1, respectively. They were interpreted as H(-N) positions. Further refinement confirmed this assignment and gave a slight drop of the R 1 (3.85%) and wR 2 (6.64%) values. A subsequent difference Fourier map exhibited more than 30 unassigned peaks with electron densities in the range 0.62-0.42 e Å À3 . The four peaks with highest intensities (Q1-Q4) were located in proximity to atoms N1, Br2, N2 and Br1 (in this order). At this stage, two different models were considered for the final refinement. Model A comprised Q1 and Q3, which were both interpreted as H(-N) positions; Q2 and Q4 were disregarded. Model B comprised Q2 and Q4 as H(-Br) positions, with Q1 and Q3 now being neglected. Free refinement of model A resulted in a stable and meaningful result, yielding two NH 2 + groups, whereas the free refinement of model B collapsed: the H(-Br) atoms moved to positions with very short Br-H distances (<0.3 Å ). The agreement factors of model A (R 1 = 3.72% and wR 2 = 6.31%) were marginally better than those of model B (R 1 = 3.84% and wR 2 = 6.63%). These results clearly show that the crystal structure analysis alone does not allow discrimination with certainty between the two models. However, the stable refinement of model A and the slightly better agreement factors may be regarded as a first sign for the ionic structure. The observed electron density in proximity to the Br atoms (Q2 and Q4) could be understood as well-known series-termination errors in the Fourier synthesis (Glusker et al., 1994).
In the final refinement (model A), a riding model was used for the C-bonded H atoms. As suggested by Mü ller et al. (2006), the positional parameters of the O-and N-bonded H atoms were refined using isotropic displacement parameters, which were set at 1.5U eq (O) or 1.2U eq (N) of the pivot atom. In addition, restraints of 0.84 and 0.88 Å were used for the O-H and N-H distances, respectively.

Results and discussion
3.1. Chemical context Liu et al. (1996) postulated the presence of neutral diazepane as a free base, together with two molecular HBr units in the solid-state structure. At first glance, such an interpretation is amazing, since HBr is known to react as a very strong acid, and the diazepane moiety -as an alicyclic diamine -is expected to react as a base. We investigated the protonation behaviour of dazol in aqueous solution (25 C) and found -as expected -that an uptake of two protons occurred readily upon addition of acid. A series of potentiometric titration experiments ( Fig. 1) revealed two pK a values of 6.01 and 9.05 (0.1 M KCl) or 6.37 and 9.28 (1 M KNO 3 ) for H 2 dazol 2+ . These values are in agreement with those reported for related amino alcohols (Martincigh & Marsicano, 1995). In addition, we also performed a 1 H NMR titration experiment in D 2 O (  Titration curve (25 C, 0.1 M KCl) of H 2 dazol 2+ . Squares refer to the measured values. The red line was calculated using the deprotonation constants Àlog K a,1 = 6.01 and Àlog K a,2 = 9.05.

Figure 2
pH* dependence of the various 1 H resonances for the nonlabile protons (pH* refers to the uncorrected pH-meter reading in D 2 O for an electrode calibrated in H 2 O). The observed resonances ( obs ) are shown as closed circles: H1 (black), H2 (blue) and H3 (red). The lines were calculated [minimization of AE( obs À calc ) 2 ]. Inset: The pattern of H1 at (i) pH* = 5, (ii) pH* = 10 and (iii) in 5 mol l À1 NaOD. observed characteristic pD-dependent resonances for the H(-C) protons upon addition of NaOD. This pD dependency could again be interpreted as a twofold deprotonation reaction of the dication. The evaluated pK a values in this medium are 6.35 and 9.62. All these characteristics clearly indicate that addition of two equivalents of HBr to an aqueous solution of dazol results in a complete transformation into the H 2 dazol 2+ dication. Crystal growth of the title compound has indeed been performed in such an acidic aqueous medium. However, one must of course be aware that -in general -the solid state does not necessarily depict the equilibrium composition in solution.
A solid sample of the title compound was therefore investigated by IR spectroscopy, looking at around 2600 cm À1 for any H-Br stretch vibration. However, these measurements were not conclusive, since a possible H-Br peak was covered by the intense and broad absorption between 2200 and 3500 cm À1 , caused by the various associated N-H and O-H stretching vibrations. Thermogravimetric measurements combined with an IR analysis of the gaseous products was more instructive (Fig. 3). A 20 mg sample was heated by a rate of 10 C min À1 from room temperature up to 800 C and exposed to a steady stream of N 2 (20 ml min À1 ). Complete degradation occurred almost quantitatively (>90%) in one single step in the range 300-400 C. Evolution of HBr could readily be recognized in the IR spectrum (2400-2700 cm À1 ) by its characteristic pattern for the two isotopomers with resolved transitions for the various rotamers (NIST, 2019). In addition, an organic component (O-H, N-H and C-H, and possibly C-C and C-O stretching vibrations, but no CO 2 ) was formed. These findings are in agreement with a predominant sublimation of the product. It is well known that NH 4 Br and its organic derivatives, such as methylammonium bromide sublimate in the range 300-400 C (Ivanov et al., 2019). The remaining small nonvolatile residue (5-10%) probably indicates some minor decomposition during the sublimation process (formation of elemental carbon, as indicated by a black coating inside the crucible). If a 1:1 stream of N 2 and O 2 (each 20 ml min À1 ) is applied to the sample during heating, the degradation occurred in more than one step. Again, less than 1% weight loss was noted below 200 C. Up to 250 C, the sample weight decreased slightly by about 3%. A first significant step of decomposition was then observed in the range of about 250-330 C, with a corresponding weight drop of 21%. This value is clearly smaller than the 29% required for a dissociation of one HBr molecule. The final part of the decomposition reaction occurred in two steps at 350-400 (À50%) and 400-550 C (À22%). The IR spectra of the Thermogravimetric analyses of H 2 dazolBr 2 (%weight versus T) together with IR characteristics of the evolving gases at temperatures as indicated. Heating of the sample (a) in a steady stream of N 2 only and (b) in a 1:1 mixture of N 2 and O 2 . The inset in (b) shows an enlargement for the range 2400-2700 cm À1 , indicating the formation of HBr in small traces. evolving gases showed the formation of H 2 O during the entire decomposition reaction. At elevated temperatures (>300 C), increasing formation of CO 2 and of some organic components (observation of C-H stretching vibrations) was also noted. Above 450 C, CO 2 remained as the most significant decomposition product. Inspection of the spectra did again exhibit that small traces of HBr have been formed. Maximum HBr production was found around 350-400 C. These observations do not indicate a simple and quantitative loss of HBr at low temperature. It rather appears that small amounts of HBr are formed in situ during the entire decomposition process, particularly at elevated temperatures. As a conclusion, evolution of HBr appears generally to be combined with a complete breakdown of the entire structure. Eliel et al. (1965) proposed high conformational flexibility for cycloheptane, with a twist-chair (TC) conformation being of lowest energy. We previously studied 6-amino-1,4-diazepane (daza) as a metal-complexing agent, and the adoption of such a TC conformation for the seven-membered 1,4-diazepane ring of H 3 daza 3+ could indeed be confirmed by crystal structure analysis (Romba et al., 2006;Neis et al., 2010). Its 1 H NMR spectrum exhibited a total of five resonances for the H(-C) protons, indicating a rapid interconversion of different conformations, yielding an averaged structure of higher symmetry (Longuet-Higgins, 2002). The molecular structure of daza and dazol is closely related and similar 1 H NMR characteristics have also been observed for dazol. The coupling pattern of the H-(C-X) proton (H 3 daza 3+ : X = NH 3 + ; H 2 dazol 2+ : X = OH) revealed, however, some characteristic differences for the two compounds. For H 3 daza 3+ , this signal appeared as a triplet of triplets with one large coupling constant of 10.5 Hz and a second much smaller constant of 3.1 Hz. The large coupling of 10.5 Hz is indicative of a staggered orientation of the H-C-C(NH 3 + )-H fragment, with a torsion angle close to 180 . Obviously, the primary ammonium group of H 3 daza 3+ is placed in an equatorial position. However, for H x dazol x+ (x = 0, 1, 2), the corresponding coupling constants are significantly smaller, with a value of 5.4/1.4 Hz at pD 5.5 and 5.8/4.4 Hz at pD 10 ( Fig. 2). It thus appears that in solution the hydroxy group of H x dazol x+ (x = 0, 1, 2) is positioned axially. Interestingly, at a very high base concentration (5 mol l À1 NaOD), the signal of this proton is again shifted by about 0.2 ppm to lower frequency and appeared as a quintet with one unique coupling constant of 4.0 Hz. Obviously, the hydroxy group of dazol becomes deprotonated in such a highly alkaline medium.

Structural commentary
In agreement with our NMR study, the H 2 dazol 2+ cation also adopted a TC conformation in the crystal structure, with the pivot atom N1 located in the isoclinal position. The puckering parameters (Boessenkool & Boeyens, 1980) of the seven-membered diazepane ring are Q = 0.821 (7) Å , Q2 = 0.506 (7), Q3 = 0.647 (7), ' 2 = 86.5 (8) and ' 3 = 92.0 (6) . Also, in accordance with the solution structure, the hydroxy group adopted an axial position. The different orientation of the NH 3 + group of H 3 daza 3+ and the OH group of H x dazol x+ (x = 0, 1, 2) is remarkable. Since this difference is observed in the solid state, as well as in solution, the peculiar structure may be explained by the well-known attractive gauche effect (Entrena et al., 1997), which proposes the preferential adoption of a gauche rather than a trans conformation for such an X-C-C-OR fragment (X = O, N).
Similar to the work performed by Liu et al. (1996), the crystal structure analysis presented here exhibited low precision, i.e. large standard uncertainties for bond angles and distances. An inspection of the displacement parameters of the C, N and O atoms indicated some significant deviations for the displacement ellipsoids from a spherical shape (Fig. 4). Considering the high conformational flexibility of the sevenmembered diazepane frame, we attribute these large deviations to minor disorder rather than to thermal motion (in contrast to Liu's work, we performed data collection at À131 C). It was, however, not possible to resolve this slight disorder in terms of a superposition of distinct individual conformers.

Supramolecular features
The cationic H 2 dazol 2+ gravicentres (dgcs) are arranged into layers oriented parallel to the crystallographic ac plane. In The molecular structure of the H 2 dazol 2+ Á2Br À unit, with the atomnumbering scheme and displacement ellipsoids drawn at the 50% probability level.

Figure 5
The distorted anticuboctahedron formed by the gravicentres of 12 H 2 dazol 2+ cations (gray spheres) together with the central H 2 dazol 2+ cation. The two crystallographically independent Br À anions are placed in 50% of the trigonal holes of a hexagonal layer (Br1 = green spheres) and in all of the octahedral holes (Br2 = blue spheres). these layers, each dgc is surrounded by six neighbouring dgcs, forming a distorted hexagon. The layers are stacked along b in a staggered fashion (ABABAB . . . ) and can thus be regarded as a distorted hexagonal packing. If the two adjacent layers are taken into account, each dgc receives 12 dgc neighbours, which form an anticuboctahedron (Fig. 5). However, some characteristic deviations from a regular shape are noted for this polyhedron. One reason for the distortion originates from the significant deviation of the H 2 dazol 2+ cations from a spherical shape; these cations should be regarded as disks rather than spheres. Consequently, the (dgc) 12 anticuboctahedron is compressed along the pseudohexagonal axis, as is expressed by the unequal edge lengths (5.59-7.87 Å ) and short interlayer distances. Further distortion originates from the general position of the dgc and the reduced crystallographic symmetry. As a consequence, the dgc layers (and thus the equators of the anticuboctahedra) are puckered. The symmetry class of this layer group is p12 1 1 (International Tables for Crystallography, 2002;Shubnikov & Koptsik, 1974). Interestingly, the Br1 ions (green spheres in Figs. 5 and 6) are located neither in the tetrahedral nor in the octahedral holes of this packing. They are placed almost straight within the pseudohexagonal dgc planes. Each Br1 anion is thus surrounded by three dazol dications. Only 50% of these triangular holes are occupied. Such a packing becomes understandable if the huge difference in size between Br À and H 2 dazol 2+ is considered (Fig. 6). The entire geometry and coordination number of Br1 becomes evident if the staggered arrangement of the dgc layers is taken into account. Beside the three dgc neighbours of the triangular hole, the Br1 anions receive two additional dgc neighbours from adjacent dgc layers and the coordination polyhedron of Br1 can thus be described as a trigonal bipyramid. The Br2 ions (blue spheres) are located in the octahedral holes of the pseudohexagonal dgc packing. The coordination number of Br2 is thus six and the coordination geometry is a distorted octahedron. The H 2 dazol 2+ cations, in turn, are surrounded by 11 Br anions with BrÁ Á Ádgc distances ranging from 4.0 to 6.5 Å (Fig. 7a). The resulting Br 11 structure can be described as a distorted Edshammar polyhedron ( Fig. 7b; Edshammar, 1969). Notably, the regular Edshammar polyhedron adopts D 3h symmetry (Fig. 7c) and is a space filler (Lidin et al., 1992). Each H 2 dazol 2+ entity forms N-HÁ Á ÁBr and O-HÁ Á ÁBr hydrogen bonds (Table 2). Br1 is bonded to H1N-N1, H3N-N2(Àx + 1, y + 1 2 , Àz + 1 2 ) and H4N-N2(Àx + 1 2 , Ày + 1, z À 1 2 ). Further Br1Á Á ÁH interactions (Br1Á Á ÁH2N-N1, Br1Á Á ÁH1-C1, Br1-H2A-C2 and Br1Á Á ÁH3B-C3), with BrÁ Á ÁH-N or BrÁ Á ÁH-C angles of 131-140 and BrÁ Á ÁH distances of 2.95-3.05 Å , must be regarded as very weak hydrogen bonding if they are to be regarded as hydrogen bonding at all. The sum of the van der Waals radii of Br and H is 2.95 Å (Bondi, 1964). Br2 forms only two unambiguous hydrogen bonds to H1O-O1 and H2N-N1(x + 1 2 , Ày + 3 2 , Àz + 1). Again, the contacts Br2Á Á ÁH3A-C3, Br2Á Á ÁH4A-C4, Br2Á Á ÁH3N-N2, Br2Á Á Á H2N-N1, Br2Á Á ÁH2B-C2, Br2Á Á ÁH5A-C5, Br2Á Á ÁH3A-C3, with BrÁ Á ÁH distances of 2.88-3.05 Å and BrÁ Á ÁH-N or BrÁ Á ÁH-C angles of 117-162 may be considered as further weak interactions which stabilize the structure. A graph-set analysis (Bernstein et al., 1995) shows the hydrogen-bonding network in the [110] direction (Fig. 8). The Br1 and Br2 anions are 3 -acceptors and the N-H and O-H hydrogens are donors in three distinct ring systems, i.e. R 4 7 (20), R 2 4 (8) and R 3 5 (16). Notably, there is no direct H 2 dazol 2+ Á Á ÁH 2 dazol 2+ hydrogen bonding and, consequently, the hydroxy group of H 2 dazol 2+ does not act as an acceptor. All these structural features correspond well to the packing of large charged molecular entities, as observed for instance in Zintl phases (Lidin et al., 1992), and thus provide support for the ionic model.

Database survey
By studying some related literature, we became aware that the report of Liu et al. (1996)  View of the (010) plane (space-filling model). The H 2 dazol 2+ cations form a pseudohexagonal layer with the Br1 anions (green spheres) located in every second trigonal hole of this layer. The Br2 anions are shown as light-blue spheres and are placed in the octahedral holes above and below this layer. C, H, N and O atoms are shown as black, light-gray, dark-blue and red spheres, respectively. Table 2 Hydrogen-bond geometry (Å , ).  (3) 3.355 (6) 153 (6) atom by a single bond), revealed a total of 69 entries (see supporting information for the full list). Some of the listed entries do not provide three-dimensional coordinates and are thus not relevant for this discussion. Furthermore, it appears that the technical term 'hydrobromide' may have led to confusion, since it has been used by some of the authors for a salt of a protonated organic molecule and bromide as counterion, but appeared in our search as molecular H-Br.  (Liu et al., 1997b), SOCZUH (Banothu et al., 2014), UNESAI (Zhang & Shen, 2011) and YOTSIJ (Monte et al., 1995)], where the simultaneous presence of HBr with a basic moiety (in KEKQAS, HBr and OH À !) is proposed, with H-Br bond lengths ranging from 0.79 to 1.84 Å . It is clearly beyond the scope of this contribution to decide whether these structural assignments are correct. However, it appears that: (i) the maintainers of the database should carefully clarify whether the expression 'hydrobromide' refers to an ionic or rather a molecular model, and (ii) reporting such a crystal structure, authors should take the required care to clarify the bonding mode when postulating incorporation of undissociated HBr together with a basic moiety. Similar considerations might also be applicable for the so called 'hydrochlorides' (292 entries for a H-Cl molecule) and 'hydroiodides' (20 entries for a H-I molecule).

Conclusion
The solution and solid-state properties of the title compound do not provide evidence for a simple incorporation of HBr as intact molecules into the solid-state structure. An ionic model,

Figure 8
The comprising one H 2 dazol 2+ cation and two Br À anions is clearly a better explanation. According to the structure postulated by Liu et al. (1996), the two HBr molecules would not be bonded to other moieties by strong interactions (such as Coulombic forces or hydrogen bonding). The (Br1-)HÁ Á ÁH(-Br2) separation of 2.24 Å roughly corresponds to the sum of the van der Waals radii of two H atoms (2.20 Å ; Bondi, 1964) and, as a consequence, the two HBr molecules would not act as H-atom donors in hydrogen bonds. One could thus expect that liberation of HBr should occur readily even at moderate temperatures.
For further clarification of this question, we grew single crystals of the title compound and repeated the X-ray analysis. We have shown that it is possible to refine the crystal structure in terms of an ordinary ammonium salt (see x2.2, Refinement). There is thus no need to postulate the rather exotic incorporation of molecular HBr into the crystal structure. We think that, in such a case, it is more advisable to chose the model which also directly explains the observed chemical properties (acid-base behaviour and breakdown of the structure upon HBr elimination above 300 C). program(s) used to solve structure: SHELXT2014 (Sheldrick, 2015a); program(s) used to refine structure: SHELXL2014 (Sheldrick, 2015b); molecular graphics: DIAMOND (Brandenburg, 2007); software used to prepare material for publication: PLATON (Spek, 2009).  (2) Special details 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.