Crystal structure prediction solvates

In this section, a number of examples of crystal structure prediction of small organic molecules, co-crystals, solvates and salts will be presented. 

From a computational point of view, the prediction of the crystal structures of solvates poses the same issues as the prediction of co-crystal structures. Crystal structure prediction studies on solvates have only recently been attempted. 

Novel techniques for the creation of co-crystals and solvates such as neat and liquid assisted grinding have challenged the ability of crystal structure prediction to predict stoichiometry from first principles. Recent work has addressed the problem of predicting solvate stoichiometry of acetic acid (the solvent) with various molecules including carbamazepine (CBZ) and its 10,11-dihydro derivative (DHCBZ), urea and theobromine(see Figure 4.8). 

The crystal structures of two anhydrous polymorphs and the two solvates were determined experimentally by single crystal X-ray diffraction. Single crystals of sufficient size and quality could not be obtained to determine the crystal structures of the other two anhydrous forms in the same manner. However, low quality X-ray powder diffraction patterns were available and a crystal structure prediction study was carried out to generate possible crystal structures for these two anhydrous forms. 

Nevertheless, as in many previous observations, the clathrate formation by dipolar host compounds could not have been predicted in advance. In fact, there are no channels in the crystal structures of hydrated moxnidazole hydrochloride (closely related species to furaltadone hydrochloride) and of hydrated furaltadone base (Fig. 13)37). Rather, the latter two structures are best described as solvates with the H20 molecules contained in local voids between adjacent moieties of the host. 

Ab initio methods for polymorph, hydrate and solvate prediction are highly prized by the industry and good progress has been made in this field in recent years. This work is still a number of years from routine commercial application however, and polymorph screening experiments together with crystal structure determination, remain critical tasks for today s Pharmaceutical companies. 

Three-membered ring carbenoids have been structurally characterized for species where X = OR or NR2. The crystal structure of the diethyl ether-solvated a-(dimethylamino)-benzyllithium dimer 2 shows crystallographically equivalent three-membered Li-C-N rings, with (Lij-C) = 2.475(6) A, and . As predicted by theory, the anionic carbon sits closer to the lithium in the other three-membered ring (,7(LiiA-C) = 2.230(7) A) than to Lij. This seems to present a means for the anionic carbon to delocalize its charge. 

The identification, structural and thermal characterization of new polymorphs is an important topic in solid-state chemistry and requires a battery of techniques that includes X-ray diffraction and spectroscopic methods, in addition to thermal analysis methods and dissolution techniques to determine solubility trends. Such studies are described by Caira in Chapter 16, as well as more recent theoretical techniques aimed at the prediction of the crystal structures of new polymorphs. Crystal polymorphism is particularly important in pharmaceutical products, so there is an emphasis on this area. Systems displaying solvatomorphism (the ability of a substance to exist in two or more crystalline phases arising from differences in their solvation states) molecular inclusion and isostructurality (the inverse of polymorphism) are also given due attention in this chapter. 

Computational methods could, therefore, potentially impact on polymorph screening, salt or co-crystal selection, as well as the avoidance of solvates. The outline of such a possibility should not be viewed as an over-optimistic assessment of current capabilities, but rather a goal towards which developments should aim. Given that progress is being made on flexible molecules and multi-component crystal structures, the methods that are necessary for such in silico screens are attainable, although it is difficult to predict the rate of progress and, therefore, when such calculations will be practical for the typical pharmaceutical molecule. Moreover, results will always need to be interpreted with care and a realistic view of the approximations and limitations of the methods. This is where continued assessment of methods on well-characterized systems is needed, to inform our level of confidence in the calculations. 

---