One molecule, three crystal structures: conformational trimorphism of N-[(1S)-1-phenylethyl]benzamide

The conformational trimorphism of a chiral amide is described, in space groups P21 and P212121, with different orientations of the supramolecular one-dimensional structures with respect to the twofold screw axis.


Chemical context
The study of polymorphism is paramount in the field of organic materials, especially in the design of new active pharmaceutical ingredients, either for tailoring their bioavailability, or for legal reasons related to patent rights and intellectual property. Walter McCrone (1965) famously stated more than 50 years ago that 'the number of [polymorphic] forms known for a given compound is proportional to the time and money spent in research on that compound'. Today, it seems that this statement still holds true, and that a large proportion of the discovered polymorphs are obtained in a non-planned way. In the current situation, the rules allowing (or avoiding) a molecular system to crystallize with several forms are not fully understood, although assessing the risk of polymorphism is workable to some extent. For example, the CSD-Materials module available in Mercury can perform predictions on a polymorphic target compound, through an estimation of its hydrogen-bonding landscape (Feeder et al., 2015;Macrae et al., 2020).
A recent survey of the CSD (Groom et al., 2016) showed that polymorphism prevalence among single-component organic anhydrates constitutes about 1.22% of compounds for which at least one crystal structure is known (Kersten et al., 2018). A similar figure was obtained using the Merck index as a source of data: for 10330 compounds present in the 12th edition (1996), 1.4% were polymorphic (Stahly, 2007).
However, compounds appearing only once in the CSD might exist in other polymorphic forms that have still not been crystallized. It also seems hard to believe that all molecules should be necessarily polymorphous, as sometimes claimed. For example, huge amounts of ibuprofen [2-(4-isobutylphenyl)propanoic acid] have been produced since its introduction as a painkiller in 1969. Notwithstanding the numerous studies carried out on this small molecule, only one crystalline form is known. But there is no doubt that from a statistical point of view, trimorphic systems are much less common than dimorphic systems, tetramorphic systems are in turn much less common than trimorphic systems, etc. A rule of thumb is that a tenfold drop is observed for the prevalence of n-morphism in comparison to (n-1)-morphism (n ! 3). It is not surprising that, for example, well-characterized hexamorphism is exceptional (Yu et al., 2000). Another empirical observation is that more polymorphs are reported for small molecules (less than 30 C atoms per formula) compared to large ones, because of the correlation between molecular complexity and the difficulty of synthesizing large molecules. These observations are in line with McCrone's statement, and today there is a consensus that polymorphism is a pervasive phenomenon, which occurs on a random basis and remains poorly predictable (Cruz-Cabeza et al., 2015).
Within this context, we report a case of trimorphism, for a low-molecular-weight chiral molecule, for which the crystal structure was never established, even though many researchers have used it as a reagent since its first reported synthesis (Bezruchko et al., 1967).

Molecular and crystal structures
We used N-[(1S)-1-phenylethyl]benzamide as a component for co-crystallization with other small molecules having a high hydrogen-bond propensity. While probing a variety of solvents for the crystallization of the free amide, we recovered three non-solvated polymorphs, in a reproducible manner. Form I (P2 1 ) was obtained from acetonitrile, and its measured melting point and angle of optical rotation match data reported by other groups (e.g. Karnik & Kamath, 2008). Forms II (P2 1 ) and III (P2 1 2 1 2 1 ) were obtained as concomitant crystals, by using ethanol-water, toluene-ethanol or THF-methanol mixtures. Simulated X-ray powder patterns are clearly different for each form, confirming that true polymorphs were crystallized.
Forms I and II share the same crystal symmetry (Table 5), but have very different densities, 1.157 and 1.208 g cm À3 , respectively. It can therefore be predicted that molecules are packed in the solid state in a more efficient manner for II, compared to I. However, both forms display the same supramolecular structure, based on the classical C 1 1 (4) chain motif, which is the most common for amide derivatives (Figs. 1 and 2). The N-HÁ Á ÁO hydrogen bond is stronger for I, while an opposite situation should be expected if one considers crystal densities (Tables 1 and 2). The factor triggering polymorphism is, in this case, related to the molecular structure. The conformation of the molecule is modified by rotation of the phenyl ring C3-C8 bonded to the chiral centre, while the Part of the crystal structure of form I, with the asymmetric unit displayed with displacement ellipsoids for non-H atoms at the 30% probability level. C-bound H atoms are omitted for clarity, and hydrogen bonds forming the infinite C 1 1 (4) chains are drawn as dashed lines. The inset is the crystal structure viewed down the chain axis, parallel to the crystallographic a axis. Grey and green molecules are related by the 2 1 symmetry elements parallel to [010] in space group P2 1 .

Figure 2
Part of the crystal structure of form II, using the same style as for Fig. 1. The labelling scheme is as in I. In the inset, the projection axis is [010]. position of the other peripheral phenyl group, C10-C15, remains almost unchanged with respect to the amide group. Dihedral angles involved in the molecular conformation are given in Table 4: angle N1-C9-C10-C15 is modified by ca 3 between the two forms, while the other angle, N1-C2-C3-C4, is modified by ca 14 . As a consequence, the dihedral angle between the phenyl rings is 23.1 (2) and 56.2 (1) in I and II, respectively.
The conformational modification leads to different arrangements for the infinite C(4) chains in the crystals. In I, the 1D motif is running in the [100] direction, and is thus normal to the twofold screw axis (Fig. 1, inset). The 2 1 symmetry element relates neighbouring chains in the crystal, resulting in a relative orientation of the chains that is unfavourable for the packing of the phenyl rings: inter-chain dihedral angles between phenyl groups are close to 90 : 1!1 0 = 88.4 (3) , 1!2 0 = 84.9 (2) and 2!2 0 = 70.8 (2) , where 1 and 2 stand for rings C3-C8 and C10-C15, while a primed ring is related to a non-primed ring through the symmetry element 2 1 . These angles were calculated using PLATON (Spek, 2020), and only non-parallel rings are considered. In contrast, the crystal structure of form II is built on C(4) chains parallel to the screw axis, in the [010] direction. As in the previous case, two neighbouring chains are related through the 2 1 axis. However, given that chains and symmetry elements share the same direction, some inter-chain interactions feature phenyl rings in a less perpendicular arrangement: 1!1 0 = 85.2 (2) , 1!2 0 = 80.3 (2) , 2!1 0 = 56.2 (2) and 2!2 0 = 64.7 (1) . Chains are then more densely packed, to afford a material with higher density (Fig. 2, inset). These different packing structures, in the same space group, are also reflected in different Kitaigorodskii packing index: 0.638 for I and 0.670 for II (Spek, 2020).
The third polymorph, III, includes two independent molecules in the asymmetric unit of an orthorhombic cell, each one forming a supramolecular structure identical to those of forms I and II [infinite C 1 1 (4) chains parallel to the a axis for molecules A and B, see Table 3 and Fig. 3]. The molecular conformation is similar for A and B molecules, and can be described as intermediary between conformations stabilized in crystals I and II: the phenyl ring bonded to the chiral C atom is configured as in crystal II, while the other phenyl group is oriented as in crystal I (Table 4). The intramolecular dihedral angle between phenyl rings is therefore also midway: 47.0 (1) for molecules A and 47.4 (1) for molecules B.
With such a configuration, it is not surprising to obtain a crystal structure for III in space group P2 1 2 1 2 1 simultaneously reminiscent of those observed for I and II (Fig. 4). The twofold screw axis parallel to [100] gives an arrangement similar to that described in form II, with two neighbouring C(4) chains including molecules from the same family, A/A or B/B, closely packed around this symmetry element. On the other hand, the packing in directions perpendicular to the chain axis is based on screw axes along [010] and [001], and is thus similar to that observed in form I with regard to neighbouring crystallographically independent molecules, A/B or B/A. The orthorhombic form III with Z 0 = 2 can be seen as a mixture combining features of Z 0 = 1 monoclinic forms I and II. This is consistent with metrics directly related to packing efficiency, which fall between those of phases I and II: the calculated density for III is 1.199 g cm À3 , the Kitaigorodskii packing index is 0.666, and large intermolecular dihedral angles p!q 0 between phenyl rings in neighbouring chains are in the range 70.1 (2) to 89.7 (2) . Part of the crystal structure of form III. Left: unit-cell content is represented, as in Figs. 1 and 2. The labelling scheme is identical, with A and B suffixes for the two independent molecules. Right: two neighbouring C 1 1 (4) chains based on independent molecules A and B are represented. Table 1 Hydrogen-bond geometry (Å , ) for form I.

Table 4
Intramolecular dihedral angles describing the conformations of the title compound in forms I, II and III.

Database survey
From the previous description, it is clear that the conformational trimorphism for the title compound is a consequence of the rotation of the peripheral phenyl rings, which changes their environment, affecting the packing of the C(4) chains. This molecular flexibility is confirmed by the crystal-structure determination of the unique co-crystal reported to date including the title molecule (Tinsley et al., 2017): the conformation is far from that observed in the free amide we report, and one phenyl is even disordered by rotation. Polymorphism can then occur, although the hydrogenbonded pattern remains unaltered. Such a behaviour has been invoked to rationalize the crystallization of the highly metastable orthorhombic form of benzamide, for which the space group is still controversial (Pba2: Blagden et al., 2005;Fdd2: Johansson & van de Streek, 2016). In the same way, the twisting between the nitrophenyl and the thiophene rings in the pharmaceutical intermediate 5-methyl-2-[(2-nitrophenyl)amino]-3-thiophenecarbonitrile is related to the rich polymorphism of this compound: six forms have been structurally characterized in this case, with a variety of colours and shapes (Yu et al., 2000;Price et al., 2005).
Regarding the supramolecular structure observed in the title compound, the imposed supramolecular motif limits the scope for polymorphism. Indeed, the frequency of infinite chains in the crystal structures of amides is as high as 28.2% in the CSD (version 5.41, updated May 2020; both organic and metal-organic amides were considered), and is probably higher for non-sterically hindered amides, such as the title compound. Moreover, the title amide having only one donor and one acceptor sites, any variation of the supramolecular structure is very unlikely. However, it should be noted that this 1D structure is easily propagated through a screw axis in the crystal state. A survey of the organic amides crystallizing in Sohncke (i.e. non-enantiogenic) space groups reveals that for 449 hits, 83% are reported in space groups P2 1 and P2 1 2 1 2 1 , while the combined frequency of these groups over the whole CSD database is only 12%. It thus seems that any space group including rototranslations can fit a polymorphic form of a small amide, either enantiopure or achiral, regardless of the rigidity of the supramolecular structure. We could anticipate that crystallization of other forms of the title compound could be achieved, for example, in space groups P2 1 2 1 2, or P3 1 , among others.

Refinement
Crystal data, data collection and structure refinement details are summarized in Table 5. C-bound hydrogen atoms were placed in calculated positions and refined using a riding model with U iso (H) = 1.2U eq (C), C-H = 0.93 Å , and C-H = 0.96 Å , U iso (H) = 1.5U eq (C), for aromatic and methyl hydrogen atoms, respectively. Amide hydrogen atoms (H1 in I and II; H1A and H1B in III) were found in difference maps and their coordinates were freely refined with U iso (H) = 1.2U eq (N). Part of the crystal structure of form III, viewed down the chain axis, parallel to the crystallographic a axis. Red and green molecules belong to the A and B families, respectively. All symmetry elements of space group P2 1 2 1 2 1 are positioned. H atoms treated by a mixture of independent and constrained refinement Á max , Á min (e Å À3 ) 0.11, À0.12 0.12, À0.14 0.11, À0.10 program(s) used to refine structure: SHELXL2018/3 (Sheldrick, 2015b); molecular graphics: Mercury (Macrae et al., 2020); software used to prepare material for publication: publCIF (Westrip, 2010).    (7) 0.0555 (9) 0.0003 (7 (2)