Prior Likelihoods and Space-Group Preferences of Solvates

For a range of organic solvents, the likelihood of the solvent forming solvates has been estimated using the recrystallization solvent (RS) data in the Cambridge Structural Database (CSD). Although RS data are viewed with caution by some crystallographers, most of the likelihood estimates are shown to have good precision. Strong trends are apparent in the results. For example, high likelihoods are found for aromatic solvents with electron-withdrawing substituents and low likelihoods for acyclic aliphatic hydrocarbons. Results for different CSD subsets, such as organic and metalloorganic, are highly correlated. The likelihood that a solvent will form solvates is almost always higher when the solvent is part of a mixture than when it is pure. The likelihood of two solvents forming a heterosolvate (i.e., both solvents in the structure) can be well estimated by the product of the likelihoods of the solvents forming normal solvates (i.e., only one solvent in the structure). The space-group preferences of solvates vary significantly with the nature of the cocrystallized solvent. Those of nonsolvates vary significantly with the solvent(s) from which they were crystallized. Solvents with inversion centers favor solvate crystallization in centrosymmetric space groups, and solvents with 2-fold rotational symmetry promote crystallization in space groups with 2-fold proper rotational axes. The inclusion of cyclohexane and carbon tetrachloride in a lattice can facilitate crystallization in trigonal and tetragonal space groups, respectively. Our results can: (a) guide solvent selection when solvates are undesired; (b) assist in predicting solvate formation, e.g., using Bayesian algorithms; (c) assist in the choice of space groups for solvate crystal structure prediction; and (d) suggest ways in which solvent incorporation can be used to influence space groups.