Received 10 July 2013
Creation of a ternary complex between a crown ether, 4-aminobenzoic acid and 3,5-dinitrobenzoic acid
The creation of ternary multi-component crystals through the introduction of 18-crown-6 to direct the hydrogen-bonding motifs of the other molecular components was investigated for 3,5-dinitrobenzoic acid (3,5-dnba) with 4-aminobenzoic acid (4-aba). The creation of a binary complex between 18-crown-6 and 4-aba (C12H24O6·2C7H7NO2)2 and a ternary salt between 3,5-dnba, 18-crown-6 and 4-aba (C12H24O6·C7H8NO2+·C7H3N2O6-·C7H4N2O6) were confirmed by single-crystal structure determination. In both structures, the amino molecules bind to the crown ether through N-HO hydrogen bonds, leaving available only a single O atom site on the crown with restricted geometry to potentially accept a hydrogen bond from 3,5-dnba. While 3,5-dnba and 4-aba form a binary co-crystal containing neutral molecules, the shape-selective nature of 18-crown-6 preferentially binds protonated amino molecules, thereby leading to the formation of the ternary salt, despite the predicted low concentration of the protonated species in the crystallizing solution. Thus, through the choice of crown ether it may be possible to control both location and nature of the available bonding sites for the designed creation of ternary crystals.
A key aim of crystal engineering is the designed formation of novel functional materials through the control and application of intermolecular interactions. One route for this involves the creation of multi-component crystalline materials such as salts or co-crystals, and research in this area has recently expanded due to interest in the application of such materials as new formulations for active pharmaceutical ingredients (APIs; Almarsson & Zaworotko, 2004; Caira, 2007; Babu et al., 2008; Berry et al., 2008; Childs et al., 2008; Ainouz et al., 2009; Rager & Hilfiker, 2010; Braga et al., 2012). While the creation of binary systems has been frequently investigated (Etter & Reutzel, 1991; Lynch et al., 1994; Aakeröy et al., 2003; Bis et al., 2007; Corvis et al., 2010) and an understanding of the structural and experimental factors that contribute to the successful formation of such materials is beginning to be constructed (Nehm et al., 2006; Chiarella et al., 2007; Shattock et al., 2008; ter Horst et al., 2009; Issa et al., 2009; Karamertzanis et al., 2009; Boyd et al., 2010; Leyssens et al., 2012; Croker et al., 2013), the development of ternary and higher complexes is significantly less developed.
The major design process in the creation of multi-component materials is the consideration of supramolecular synthons between the components (Desiraju, 1995). These are frequently occurring intermolecular interactions that are expected to bind the components together in the resulting crystal structure. In the case of a binary complex, a single supramolecular synthon is required between the two components, while for a ternary system two such synthons are required. This can be achieved either by a central molecule exhibiting two binding sites for the other components or through the formation of a framework by two components with the third component included within the structure. The first route has been utilized more frequently as it exists as a straightforward extension of the binary complex creation process. However, it does require the two binding sites of the bridging molecule to be able to distinguish successfully between the two other components and reduce the potential for the formation of a 2:1 composition binary co-crystal. This has been achieved by the use of hydrogen-bond acceptor/donor sites of differing strengths, and second and third components of differing strengths (Aakeröy et al., 2001, 2005, 2006, 2007; Aakeröy & Salmon, 2005; Bhogala et al., 2005; Tan et al., 2006; Dabros et al., 2007; Bhogala & Nangia, 2008), or by the use of differing but complementary types of interaction such as charge transfer and hydrogen bonding (Thomas et al., 2010; Seaton et al., 2013). Examples of the second approach are rarer but the creation of co-crystal frameworks with interchangeable solvent molecules in the lattice has been demonstrated (Friscic et al., 2006). Related to this is the inclusion of NH4+/18-crown-6 or HONH3+/18-crown-6 into the framework of 1,3-cis-5-cis-cylcohexanetricarboxylic acid, forming the inverse concept (single-component framework and binary inclusion; Bhogala & Nangia, 2006). Indeed, building on this methodology, another potential route for ternary co-crystal formation is the inclusion of a co-crystal-forming dimer into a macrocyclic container. This would require a bifunctional co-crystal former that could selectively bind to both the macrocycle and to the other molecular component.
4-Aminobenzoic acid (4-aba) forms binary co-crystals with both 3,5-dinitrobenzoic acid (3,5-dnba; Etter & Frankenbach, 1989; Chadwick et al., 2009) and 18-crown-6 (Fig. 1; Elbasyouny et al., 1983). The binary acid/acid co-crystal is formed through a heteromolecular R22(8) acid dimer. Only the unit-cell dimensions of the complex with 18-crown-6 were reported. However, the related methyl ester complex is known to bind through its amino group to the O atoms of the crown ether through N-HO hydrogen bonds (Elbasyouny et al., 1983). Thus, the hydrogen-bonding interactions between the three components are potentially complementary, allowing for the formation of the desired ternary co-crystal (Fig. 2). Additionally, the crown ether may play a secondary role in directing the interactions between the remaining components. The initial aim of this project was to synthesize the binary co-crystal between 4-aba and 18-crown-6 (1), in order to confirm the expected hydrogen-bonding interactions, followed by the creation of a ternary complex between 3,5-dnba, 4-aba and 18-crown-6 (2).
| || Figure 1 |
Molecular structures for compounds used in this study.
| || Figure 2 |
Envisioned intermolecular interactions between species in a ternary complex.
All starting materials were purchased from Sigma-Aldrich and used as received.
18-Crown-6 (0.27 g, 0.001 mol) in methanol (2 ml) and 4-aminobenzoic acid (0.3 g, 0.002 mol) in methanol (5 ml) were mixed and heated until all the solid was dissolved. The solution was left to cool and clear block-shaped crystals were formed. Comparison of powder X-ray diffraction data (PXRD) data with the simulated pattern from the determined crystal structure confirmed the identity of the bulk sample.
IR: max (cm-1) 3445 and 3363 (s, NH2), 3258 (s br, OH), 2886 (m, CH2), 1656 (s, C=O), 1595 (s, NH2), 1107 (s, CH2).
18-Crown-6 (0.27 g, 0.001 mol) in methanol (2 ml) and 4-aminobenzoic acid (0.14 g, 0.001 mol) in methanol (7 ml) were mixed and heated until all the solid was dissolved. To this solution, 3,5-dinitrobenzoic acid (0.42 g, 0.002 mol) was added. The resulting solution was left to cool and colourless block-shaped crystals were formed. Comparison of PXRD data with the simulated pattern from the determined crystal structure confirmed the identity of the bulk sample.
IR: max (cm-1) 3450 (s br, OH), 3370 (m, NH3+), 1536 (s, C=O), 1347 (s, NO2).
The crystal structures of (1) and (2) were determined using intensity data collected on a Bruker APEX II X8 diffractometer using Mo K radiation at 173 K (Table 1).1 H atoms bound to C atoms were treated as riding, while those bound to N or O atoms were located in the difference Fourier map and refined freely, except for H1D in (1), which displayed significant shortening of the O-H bond on refinement and so was placed in an idealized position and treated as riding. For atoms H1F and H1G in (1), which are involved in a carboxylic acid dimer interaction, the O-H bonds appear elongated on account of the hydrogen-bonding interaction.
X-ray powder diffraction data were collected on a Bruker D8 FOCUS instrument in Bragg-Brentano - geometry with Cu K radiation using a zero-background Si holder and a scintillation counter.
Complex (1) has a 2:1 acid/crown-ether ratio and packs in space group P21/n with three symmetry-independent trimers [denoted (1a), (1b), (1c); Fig. 3 and Table 2]. The molecular components of the asymmetric unit are labelled A-I for identification (A, H and I are the crown ethers, B, C, D, E, F and G are the 4-aba molecules). Comparison of the geometries of each crown ether molecule within the asymmetric unit (Table 2) shows that all conformations differ from the expected range of torsion angles for the commonly observed D3d geometry (OCCO = 180°, COCC = 70°; Watson et al., 1984). The crown ether binds to the two 4-aba molecules through N-HO hydrogen bonding. In each case, five of the six ether O atoms are involved in hydrogen bonds (Table 2). In crown ether A, the O atom that is not involved in hydrogen bonding (O4A) and the neighbouring methylene group (C4A) are disordered over two conformations [refined site occupancy 0.673 (12):0.327 (12)]. Residual electron density suggests that there may be further disorder in crown ether I, but attempts to refine a disordered model for this molecule were unstable, so this was not considered further. The flexibility in conformation for the crown ethers is further demonstrated by the geometry of binding for the 4-aba molecules, where the 4-aba molecule in trimer (1a) is pulled closer to the centre of the crown ether ring and features shorter hydrogen bonding than in trimers (1b) and (1c) (Table 2). The hydrogen bonding occurs from both 4-aba molecules to O atoms on either side of the ring, with a bifurcated bond forming for one 4-aba to use the five acceptor sites.
| || Figure 3 |
Molecular conformations of the three trimers in complex (1), with displacement ellipsoids shown at 50% probability for non-H atoms. The atomic labelling is also shown. Trimer (1a) is formed from 4-aba (C), crown ether (A) and 4-aba (G). Trimer (1b) is formed from 4-aba (D), crown ether (H) and 4-aba (E). Trimer (1c) is formed by 4-aba (F), crown ether (I) and 4-aba (B). C atoms are coloured green, O red, N blue and H grey.
The three symmetry-independent trimers are bound together through R22(8) carboxylic acid dimers to generate a one-dimensional ribbon motif (Fig. 4) with the molecules in the order E-H-D-C-A-G-F-I-B-B-I-F-G-A-C-D-H-E. As with the binding between the acid and the crown ether, a range of geometries is observed for this interaction (Table 3). Two of these carboxylic acid dimers (4-aba B and E) are generated around crystallographic inversion centres, while the remaining pairs are created between symmetry-unrelated molecules (4-aba D/C and F/G). In general, the observed distances are shorter than those observed in the -form of 4-aminobenozic acid [OO distances: 2.650 (4) and 2.616 (4) Å, respectively, Z' = 2; Athimoolam & Natarajan, 2007], suggesting that complexation of the crown ether causes a stronger hydrogen bond to be formed. This is supported by the apparent disorder in the H-atom location within the dimers, especially between complexes F and G where the proton appears to be centrally located between the O atoms. While the refined positions of these H atoms may be uncertain, the similarity in C-O distances within the F/G carboxylic acid groups supports this interpretation (Table 4). The complete crystal structure is assembled through the packing of these one-dimensional ribbons along the a and c axes through weaker C-HO and C-H interactions, forming a herringbone motif.
| || Figure 4 |
View of the one-dimensional ribbon motif constructed through acid/acid dimers in the crystal structure of (1). The packing of two chains is displayed. The colour scheme is the same as Fig. 3.
Complex (2) forms with a 1:1:2 4-aminobenzoic acid/crown ether/3,5-dinitrobenzoic acid composition in space group , with a single complex in the asymmetric unit. The crown ether displays a closer match to the expected D3d geometry (COCC = 178, 177, 173, 173, 172, 177°; OCCO = 59, -62, 65, -65, 61, -63°) compared to that in complex (1). The geometry of the benzoic acid molecules shows no unusual features. Unlike the binary co-crystal formed between 3,5-dnba and 4-aba (Etter & Frankenbach, 1989; Chadwick et al., 2009), complex (2) forms a salt between the acids through the deprotonation of 3,5-dnba and the protonation of the amino group of 4-aba. The remaining 3,5-dnba molecule remains protonated and so 3,5-dnba is present in two protonation states in the crystal structure. The pKa for these acids is 2.05, which is on the boundary for salt formation according to empirical rules (Childs et al., 2007). However, the difference in composition does limit the ability to compare the binary and ternary systems, and recent work has also demonstrated the importance of local crystal structure on proton transfer in hydrogen-bonded systems; these structural changes may have a greater influence than changes in pKa (Wilson, 2007; Thomas et al., 2010; Jones et al., 2013; Seaton et al., 2013).
The protonated amino group on the 4-aba molecule binds through N+-HO hydrogen bonds strengthened by the symmetrical match between the crown ether and the tetrahedral ammonium group (Table 5). The charged 3,5-dnba molecule bridges the 4-aba/18-crown-6 complex and the neutral 3,5-dnba molecule, through a pair of O-HO- hydrogen bonds (Fig. 5). These clusters are bound into a two-dimensional sheet in the (100) plane through a combination of weaker C-HO interactions (Carom-HOnitro chains along the c-axis, Cether-HOnitro along the b-axis, Fig. 6). The sheets feature alternating layers of crown ether and 3,5-dinitrobenzoic acid molecules, and they are stacked along the a-axis through C-H and C-HO interactions to form the complete crystal structure.
| || Figure 5 |
Formation of a four-component motif in complex (2). The colour scheme is the same as Fig. 3.
| || Figure 6 |
Formation of a two-dimensional sheet in the (100) plane of complex (2), viewed down the a-axis. A combination of strong and weak hydrogen bonds combine to create the sheet structure. The hydrogens of the crown ether have been removed for clarity.
The solution speciation in this system was determined using literature values for the pKa of the carboxylic acid groups in 3,5-dinitrobenzoic acid and 4-aminobenozic acid in methanol (Rived et al., 1998), while the pKa of the amino group of 4-aba in methanol was estimated from the literature aqueous value by the methods detailed by Rived et al. The speciation plots (Fig. 7) indicate that under the conditions of the experiment, very little of either species would be present in an ionized form and so the selective binding by the crown ether promotes the formation of the salt phase observed. However, the presence of any water in the system would readily shift the distribution of these curves, with the pKa of 3,5-dnba in water being 2.82 compared with 7.38 in methanol (Rived et al., 1998). As the vials were open to the air during crystallization, variable levels of water may be present, influencing the solution speciation. Understanding the influence of moisture on crystallization of these systems is a topic for future study.
| || Figure 7 |
Calculated speciation curves for (a) 3,5-dnba and (b) 4-aba in methanol. Lines between points have been added as a guide to the eye.
The frequency of occurrence of the acidcarboxylate motif displayed in complex (2) was investigated through a database search. The Cambridge Structural Database (Version 5.33 with four updates; Allen, 2002) was searched for systems with a close contact between the O atoms of a carboxylic acid and a carboxylate group, with the negative charge on either O atom. The search was limited to only organic systems, with no disorder and no errors, and with powder structure determinations excluded. A total of 1357 structures were identified, of which only 134 featured this interaction between different acids (9%). As these systems must feature both a carboxylic acid and a carboxylate, positively charged species must also be present in all cases for charge balance. This is achieved either by the formation of a salt where different charges are present on different species or a co-crystal between a neutral acid and zwitterionic species, frequently an amino acid.
Of the 134 identified systems, the results display an even split between salts (53%) and co-crystals of zwitterions (45%), with the remaining 2% comprising a mixture of the two types. To identify any structural preferences for salt or co-crystal formation, those acids that were involved in five or more systems were studied further: oxalic acid, tartaric acid, 4-hydroxybenzoic acid, mandelic acid, maleic acid, trichloroacetic acid, trifluoroacetic acid, fumaric acid, 2-phenyloxypropionic acid and (18-crown-6)-2, 3, 11, 12-tetracarboxylic acid. As the strength of the acid increases (lower pKa) the number of salts increases (Fig. 8). For the strong acids, with pKa < 2 (oxalic, maleic, tricholoracetic and trifluoroacetic acid), only salts were obtained. Thus it appears that pKa may be a relevant factor in designing such motifs. However, these results refer only to a selected subset, and further consideration of the systems without the desired motif would need to be studied in order to draw final conclusions.
| || Figure 8 |
Fraction of zwitterionic co-crystals identified in the CSD search plotted against the pKa of the acid used in the co-crystallization.
The improved match between the crown ether and the protonated amino group would be expected to drive the selection of the salt form. To confirm this, a computational study into the binding preferences of the 18-crown-6 ring was carried out through optimization of the binding energy between a protonated and an unprotonated 4-aminobenzoic acid molecule with 18-crown-6, using the differential evolution global optimization algorithm (DE; Storn & Price, 1997; Price et al., 2006). The calculations were performed using locally developed code. DE is an evolutionary algorithm (EA) and operates on similar principles to genetic algorithms or evolutionary strategies. It works through the evolution of a population of trial solutions to the problem through recombination, mutation and selection. The difference from traditional EAs is that in DE the recombination and mutation processes are combined into a single procedure and a deterministic selection is used. Each member of the population is a real-valued vector (Pi) representing a solution to the problem. Initially each member of the population is randomly generated. The size of the population is a user-defined variable (Np). For each member of the population, a child solution is generated by application of
where , , are randomly selected members of the population, K and F are user-defined parameters controlling the level of recombination and mutation, respectively. Boundary conditions on each parameter may be introduced at this stage to ensure that the values remain meaningful. The better solution out of the child and parent is then retained in the population. This process is repeated until the population converges onto one solution or a user-defined number of generations (Gmax) have passed. DE is a general purpose optimization method and has been successfully applied to many problems in science and engineering including crystal-structure determination from powder diffraction data (Seaton & Tremayne, 2002; Tremayne et al., 2002; Chong & Tremayne, 2006; Chong et al., 2006), the optimization of atomic clusters and hydrocarbon isomers (Ali et al., 2006), modelling E-coli metabolism (Ceric & Kurtanjek, 2006) and the analysis of X-ray absorption fine structure (Dimakis & Bunker, 2006). The DE methodology was selected for this problem since a large number of potential minima would be expected and an efficient optimization algorithm is required to interrogate such a complicated hypersurface. Given the large number of energy evaluations required, the use of ab initio quantum mechanical calculations would be prohibitive and so the use of an empirical forcefield was required. While this may limit the accuracy of the absolute energies calculated, the relative trends and structural features would be expected to be suitable to draw general conclusions.
For this problem, trial structures were encoded as the position and orientation of the 4-aba molecule relative to the centre of the 18-crown-6 molecule (x, y, z, , , ). Boundaries of -15 x, y, z 15 Å and 0 , , 360° on the six parameters were maintained during the runs. The intermolecular binding energy between the molecules was minimized. This was determined using the non-bonding interactions of the force field described by No et al. (1995) with the electrostatic interaction calculated from single-point charges on each atom. The point charges were calculated by fitting to the electrostatic potential calculated at the PBE/def2-TZVP/def2-TZVP/J level (Schäfer et al., 1992; Weigend & Ahlrichs, 2005) using the RIJCOSX option in the ORCA program (Neese, 2012). The original model for each molecule was extracted from the crystal structure of the dimer, optimized using the QuickOpt option and then the single-point energy was calculated at the above level. The DE control parameters were K, F, Gmax, Np = 0.1, 0.5, 5000, 60. The DE calculation was performed ten times for each pair and the same solution was located each time.
The optimal solution in each case has the amino group within the crown ether as expected (Fig. 9), with the protonated system displaying significantly stronger binding than the unprotonated amino group (E4abaH+ = -170.00 kJ mol-1, E4aba = -68.74 kJ mol-1). This indicates a strong preference for the protonated system over the deprotonated one. Therefore, even a low level of salt within the solution should be preferentially extracted. Optimization of three component clusters between 4-aba/18c6 and either 4abaH+, 4aba or 3,5-dnba offers insight into the competition between hydrogen-bonding sites. For this calculation, the translational boundaries were increased to -25 x, y, z 25 Å and Np to 120. In the case of 4aba/4-aba/18c6, the second acid forms the expected acid/acid dimer with a total energy of -152.30 kJ mol-1. For 4aba/4abaH+/18c6, the crown ether selects the protonated acid and an acid/acid dimer is formed between the two 4-aba molecules with a total energy of -264.85 kJ mol-1 (Fig. 10). The difference in energy between the dimer pairs and the trimer clusters gives the binding energy for the acid/acid dimer, which are relatively similar in this case (E4aba/4aba = -83.60 kJ mol-1, E4abaH+/4aba = -94.85 kJ mol-1).
| || Figure 9 |
Optimized geometries for interactions between 18-crown-6 and (a) protonated and (b) neutral 4-aba.
| || Figure 10 |
Comparison of optimized trimers between 18-crown-6, 4-aba and (a) protonated and (b) neutral 4-aba. In the protonated case the third H atom of the ammonium group points into the page.
Optimization of the 3,5-dnba/4-aba/18-crown-6 trimer proved to be more problematic. Initial runs failed to converge, often suffering from convergence of the majority of the population to a single solution but a small proportion of the population stuck in a different solution with only a slightly higher energy. As the DE generates trial structures based on the difference of randomly selected members of the population, this situation causes stagnation of the population, as the majority of the differences generated for the higher-energy solution are close to zero and so no movement occurs. To overcome this stagnation, optimization with the recently developed drift-bias free differential evolution (DBFDE) algorithm was attempted (Price, 2008). In this case, the generation of trial structures proceeds by equation (2) instead of equation (1)
F and P are user-defined parameters in the range 0-1 and D is the number of parameters to be determined. This formulation has been shown to remove bias in the generation and selection of trial positions and ensure that each operation has a unique function during the search (mutation explores the surface, recombination homogenizes the population and selection improves the population; Price, 2008). This methodology has been shown to be effective on complex multi-modal surfaces with a high dimensionality, where the traditional DE can stagnate (Price, 2008).
DBFDE was run using control parameters P, F, Gmax, Np of 0.5, 0.98, 5000, 120, respectively, and displayed a more robust search for the 3,5-dnba/4-aba/18-crown-6 case than the traditional DE. Ten independent runs all successfully converged, with the 4-aba molecule binding to the crown ether through N-HO hydrogen bonds. The interaction of 3,5-dnba with this dimer was more varied, however, including O-HO hydrogen bonds and stacking interactions (Fig. 11), with interaction energies in the range -148.16 to -161.67 kJ mol-1, corresponding to acid/acid interactions of -79.41 to -92.93 kJ mol-1, comparable to those obtained in the 4-aba/4-aba case. This result suggests that a wider number of binding modes are possible for 18-crown-6 and 3,5-dnba, unlike the previous systems where a single low-energy solution was located in all runs. Given the successful convergence of the each run of the DBFDE method compared with the traditional DE, this is likely due to the genuine existence of numerous minima on the hypersurface of similar energy rather than a limitation of the search method. This change in hypersurface morphology can be a contributing factor for the selectivity, as the binding of 4-abaH+ or 4-aba to an existing 4-aba/crown ether complex follows a straightforward route to a single low-energy solution, while the 4-aba /3,5-dnba system has a number of competing routes.
| || Figure 11 |
Structures of the three lowest-energy minima on the 3,5-dnba/4-aba/18-crown-6 hypersurface with associated total interaction energy.
The creation of a ternary multi-component crystal by the binding of a binary co-crystal through additional hydrogen bonding to the crown ether has been demonstrated for 18-crown-6 with the binary pairs of 4-aminobenzoic acid/3,5-dinitrobenzoic acid. The selectivity of the crown ether for the protonated amino group is highlighted by the selection of the salt form of 4-aminobenzoic acid/3,5-dinitrobenzoic acid in favour of the neutral co-crystal that is obtained from solutions lacking the crown ether. This system also displays a carboxylic acid/carboxylate hydrogen bond between differing benzoic acids. This is a relatively rare supramolecular synthon, however, variations in crystal type (e.g. salt or co-crystal) appear to be selective depending on the strength of the acid used. Computational studies support the bonding preference for protonated 4-aba, with the crown ether showing a lower energy of binding and preferential selection when both options are present. Binding of 3,5-dnba to the created 4-aba/crown ether complex is complicated by a number of potential binding modes of similar energy. However, the selective binding of the crown ether to the protonated 4-aminobenzoic acid over the deprotonated form does suggest a potential route for the creation of new materials through the design of acceptors with preferential binding a desired form of the other components.
The computational portion of this work was undertaken as part of the Solid-State Pharmaceutical Cluster supported by the Science Foundation Ireland (Grant 07/SRC/B1158).
Aakeröy, C. B., Beatty, A. M. & Helfrich, B. A. (2001). Angew. Chem. Int. Ed. 40, 3240-3242.
Aakeröy, C. B., Beatty, A. M., Helfrich, B. A. & Nieuwenhuyzen, M. (2003). Cryst. Growth Des. 3, 159-165.
Aakeröy, C. B., Desper, J. & Scott, B. M. T. (2006). Chem. Commun. pp. 1445-1447.
Aakeröy, C. B., Desper, J. & Smith, M. M. (2007). Chem. Commun. pp. 3936-3938.
Aakeröy, C. B., Desper, J. & Urbina, J. F. (2005). Chem. Commun. pp. 2820-2822.
Aakeröy, C. B. & Salmon, D. J. (2005). CrystEngComm, 7, 439-448.
Ainouz, A., Authelin, J. R., Billot, P. & Lieberman, H. (2009). Int. J. Pharm. 374, 82-89.
Ali, M. M., Smith, R. & Hobday, S. (2006). Comput. Phys. Commun. 175, 451-464.
Allen, F. H. (2002). Acta Cryst. B58, 380-388.
Almarsson, Ö. & Zaworotko, M. J. (2004). Chem. Commun. pp. 1889-1896.
Athimoolam, S. & Natarajan, S. (2007). Acta Cryst. C63, o514-o517.
Babu, N. J., Reddy, L. S., Aitipamula, S. & Nangia, A. (2008). Chem. Asian J. 3, 1122-1133.
Berry, D. J., Seaton, C. C., Clegg, W., Harrington, R. W., Coles, S. J., Horton, P. N., Hursthouse, M. B., Storey, R., Jones, W., Frisc ic, T. & Blagden, N. (2008). Cryst. Growth Des. 8, 1697-1712.
Bhogala, B. R., Basavoju, S. & Nangia, A. (2005). CrystEngComm, 7, 551-562.
Bhogala, B. R. & Nangia, A. (2006). Cryst. Growth Des. 6, 32-35.
Bhogala, B. R. & Nangia, A. (2008). New J. Chem. 32, 800-807.
Bis, J. A., Vishweshwar, P., Weyna, D. R. & Zaworotko, M. J. (2007). Mol. Pharm. 4, 401-416.
Boyd, S., Back, K. R., Chadwick, K., Davey, R. J. & Seaton, C. C. (2010). J. Pharm. Sci. 99, 3779-3786.
Braga, D., Grepioni, F., Maini, L., Capucci, D., Nanna, S., Wouters, J., Aerts, L. & Quéré, L. (2012). Chem. Commun. 48, 8219-8221.
Bruker (2008). APEX2, SADABS and SAINT. Bruker AXS, Madison, Wisconsin, USA.
Caira, M. R. (2007). Mol. Pharm. 4, 310-316.
Ceric, S. & Kurtanjek, Z. (2006). Chem. Biochem. Eng. Q. 20, 243-253.
Chadwick, K., Sadiq, G., Davey, R. J., Seaton, C. C., Pritchard, R. G. & Parkin, A. (2009). Cryst. Growth Des. 9, 1278-1279.
Chiarella, R. A., Davey, R. J. & Peterson, M. L. (2007). Cryst. Growth Des. 7, 1223-1226.
Childs, S. L., Rodríguez-Hornedo, N., Reddy, L. S., Jayasankar, A., Maheshwari, C., McCausland, L., Shipplett, R. & Stahly, B. C. (2008). CrystEngComm, 10, 856-864.
Childs, S. L., Stahly, G. P. & Park, A. (2007). Mol. Pharm. 4, 323-338.
Chong, S. Y., Seaton, C. C., Kariuki, B. M. & Tremayne, M. (2006). Acta Cryst. B62, 864-874.
Chong, S. Y. & Tremayne, M. (2006). Chem. Commun. pp. 4078-4080.
Corvis, Y., Négrier, P., Lazerges, M., Massip, S., Léger, J. M. & Espeau, P. (2010). J. Phys. Chem. B, 114, 5420-5426.
Croker, D. M., Davey, R. J., Rasmuson, Å. C. & Seaton, C. C. (2013). CrystEngComm, 15, 2044-2047.
Dabros, M., Emery, P. R. & Thalladi, V. R. (2007). Angew. Chem. Int. Ed. 46, 4132-4135.
Desiraju, G. R. (1995). Angew. Chem. Int. Ed. 34, 2311-2327.
Dimakis, N. & Bunker, G. (2006). Biophys. J. 91, L87-L89.
Elbasyouny, A., Brugge, H. J., von Deuten, K., Dickel, M., Knöchel, A., Koch, K. U., Kopf, J., Melzer, D. & Rudolph, G. (1983). J. Am. Chem. Soc. 105, 6568-6577.
Etter, M. C. & Frankenbach, G. M. (1989). Chem. Mater. 1, 10-12.
Etter, M. C. & Reutzel, S. M. (1991). J. Am. Chem. Soc. 113, 2586-2598.
Friscic, T., Trask, A. V., Jones, W. & Motherwell, W. D. S. (2006). Angew. Chem. Int. Ed. 45, 7546-7550.
ter Horst, J. H., Deij, M. A. & Cains, P. W. (2009). Cryst. Growth Des. 9, 1531-1537.
Issa, N., Karamertzanis, P. G., Welch, G. W. A. & Price, S. L. (2009). Cryst. Growth Des. 9, 442-453.
Jones, A. O. F., Blagden, N., McIntyre, G. J., Parkin, A., Seaton, C. C., Thomas, L. H. & Wilson, C. C. (2013). Cryst. Growth Des. 13, 497-509.
Karamertzanis, P. G., Kazantsev, A. V., Issa, N., Welch, G. W. A., Adjiman, C. S., Pantelides, C. C. & Price, S. L. (2009). J. Chem. Theory Comput. 5, 1432-1448.
Leyssens, T., Springuel, G., Montis, R., Candoni, N. & Veesler, S. (2012). Cryst. Growth Des. 12, 1520-1530.
Lynch, D. E., Smith, G., Freney, D., Byriel, K. A. & Kennard, C. H. (1994). Aust. J. Chem. 47, 1097-1115.
Macrae, C. F., Edgington, P. R., McCabe, P., Pidcock, E., Shields, G. P., Taylor, R., Towler, M. & van de Streek, J. (2006). J. Appl. Cryst. 39, 453-457.
Neese, F. (2012). WIREs Comput. Mol. Sci. 2, 73-78.
Nehm, S. J., Rodríguez-Spong, B. & Rodríguez-Hornedo, N. (2006). Cryst. Growth Des. 6, 592-600.
No, K. T., Kwon, O. Y., Kim, S. Y., Cho, K. H., Yoon, C. N., Kang, Y. K., Gibson, K. D., Jhon, M. S. & Scheraga, H. A. (1995). J. Phys. Chem. 99, 13019-13027.
Palmer, D. (2005). CrystalMaker. CrystalMaker Software Ltd, Yarnton, Oxfordshire, England.
Price, K. V. (2008). Advances in Differential Evolution, edited by U. K. Chakraborty, pp. 33-38. Berlin: Springer.
Price, K. V., Storn, R. & Lampinen, J. (2006). Differential Evolution: A Practical Approach to Global Optimization. Berlin: Springer.
Rager, T. & Hilfiker, R. (2010). Cryst. Growth Des. 10, 3237-3241.
Rived, F., Rosés, M. & Bosch, E. (1998). Anal. Chim. Acta, 374, 309-324.
Schäfer, A., Horn, H. & Ahlrichs, R. (1992). J. Chem. Phys. 97, 2571.
Seaton, C. C., Blagden, N., Munshi, T. & Scowen, I. J. (2013). Chem. Eur. J. 19, 10663-10671.
Seaton, C. C. & Tremayne, M. (2002). Chem. Commun. pp. 880-881.
Shattock, T. R., Arora, K. K., Vishweshwar, P. & Zaworotko, M. J. (2008). Cryst. Growth Des. 8, 4533-4545.
Sheldrick, G. M. (2008). Acta Cryst. A64, 112-122.
Storn, R. & Price, K. V. (1997). J. Glob. Optim. 11, 341-359.
Tan, T.-F., Han, J., Pang, M., Song, H., Ma, Y. & Meng, J. (2006). Cryst. Growth Des. 6, 1186-1193.
Thomas, L. H., Blagden, N., Gutmann, M. J., Kallay, A. A., Parkin, A., Seaton, C. C. & Wilson, C. C. (2010). Cryst. Growth Des. 10, 2770-2774.
Tremayne, M., Seaton, C. C. & Glidewell, C. (2002). Acta Cryst. B58, 823-834.
Watson, W. H., Galloy, J., Grossie, D. A., Vogtle, F. & Müller, W. M. (1984). J. Org. Chem. 49, 347-353.
Weigend, F. & Ahlrichs, R. (2005). Phys. Chem. Chem. Phys. 7, 3297-3305.
Wilson, C. C. (2007). Crystallogr. Rev. 13, 143-198.