No abstract

Stabilizing the unstable: In textbooks barbituric acid is always drawn in its keto tautomeric form, which is indeed preferred in solution and in most polymorphic phases. However, phase IV, obtained by grinding, consists of molecules in the enol form, as shown by neutron powder diffraction. This phase is found to be the most stable one at room temperature; the “unstable” enol tautomer is stabilized by a higher number of hydrogen bonds.

Quantum chemical calculations at the BP86/def2-SVP levels of theory have been carried out for the reaction pathways of the [Co(L)] (+)-catalyzed Diels-Alder reaction of isoprene with phenylacetylene, with L = dppe, iminA, iminB. The calculations suggest that the reactions take place in a stepwise fashion, starting with the formation of the complex [Co(L)(isoprene)(phenylacetylene)] (+) as precursor for the consecutive C-C bond formation. The actual Diels-Alder ring-closing reaction proceeds as an intramolecular addition of the ligands isoprene and phenylacetylene, yielding a metallacyclic intermediate after generation of the first carbon-carbon bond, which determines the regioselectivity of the reaction. There are four different conformations of the starting complexes [Co(L)(isoprene)(phenylacetylene)] (+) which initiate four different pathways yielding the 1,3-cyclohexadiene product. The energetically most stable conformations do not lead to the reaction pathways that have the lowest activation energies. All conformations and the associated pathways must be considered in order to obtain the kinetically most favorable reaction course. The calculated values for the regioselectivities of the [Co(L)] (+)-catalyzed Diels-Alder reaction agree exceptionally well with the experimental values. The calculations concur with the experimental finding that the para product is kinetically favored for L = dppe while the formation of the meta product is kinetically favored when L = iminA or iminB. The different regioselectivies for L = dppe and L = iminA or iminB come from (a) the steric interactions of the bidentate ligands with the isoprene and phenylacetylene moieties in [Co(L)(isoprene)(phenylacetylene)] (+), which determine the distance between the carbon atoms forming the C-C bond, and (b) the relative energies of the different starting complexes. The first C-C bond formed in the rate-determing step of the [Co(dppe)] (+)-catalyzed reaction yielding the para product is the C4-C1' bond, and for the meta product it is the C1-C1' bond. The opposite order is found for the [Co(iminA)] (+)- and [Co(iminB)] (+)-catalyzed reactions, where the C1-C2' bond formation is the initial step toward the para product, while the C4-C2' bond is first formed in the reaction yielding the meta product. The calculations suggest that a less polar solvent should reduce the preference for formation of the meta product in the [Co(iminA)] (+)- and [Co(iminB)] (+)-catalyzed reactions but would enhance the formation of the para product in the [Co(dppe)] (+)-catalyzed reaction. Experimental tests using toluene as solvent instead of dichloromethane confirm the theoretical predictions.

A method to refine organic crystal structures from powder diffraction data with incorrect lattice parameters has been developed. The method is based on the similarity measure developed by de Gelder et al. [J. Comput. Chem. (2001), 22, 273-289], using the cross- and auto-correlation functions of a simulated and an experimental powder pattern. The lattice parameters, molecular position, molecular orientation and selected intramolecular degrees of freedom are optimized until the similarity measure reaches a maximum; subsequently, a Rietveld refinement is carried out. The program FIDEL (FIt with DEviating Lattice parameters) implements this method. The procedure is also suitable for unindexed powder data, powder diagrams of very low quality and powder diagrams of non-phase-pure samples. Various applications are shown, including structure determinations from powder data using crystal structure predictions by standard force-field methods. Other useful applications include the automatic structure determination from powder data starting from the crystal structures of isostructural compounds (e.g. a solvate, hydrate or chemical derivative), or from crystal data measured at a different temperature or pressure.

A crystallographic quantum-mechanical/molecular-mechanical model (c-QM/MM model) with full space-group symmetry has been developed for molecular crystals. The lattice energy was calculated by quantum-mechanical methods for short-range interactions and force-field methods for long-range interactions. The quantum-mechanical calculations covered the interactions within the molecule and the interactions of a reference molecule with each of the surrounding 12-15 molecules. The interactions with all other molecules were treated by force-field methods. In each optimization step the energies in the QM and MM shells were calculated separately as single-point energies; after adding both energy contributions, the crystal structure (including the lattice parameters) was optimized accordingly. The space-group symmetry was maintained throughout. Crystal structures with more than one molecule per asymmetric unit, e.g. structures with Z' = 2, hydrates and solvates, have been optimized as well. Test calculations with different quantum-mechanical methods on nine small organic molecules revealed that the density functional theory methods with dispersion correction using the B97-D functional with 6-31G* basis set in combination with the DREIDING force field reproduced the experimental crystal structures with good accuracy. Subsequently the c-QM/MM method was applied to nine compounds from the CCDC blind tests resulting in good energy rankings and excellent geometric accuracies.