A test of crystal structure prediction of small organic molecules

Abstract

A collaborative workshop was held in May 1999 at the Cambridge Crystallographic Data Centre to test how well currently available methods of crystal structure prediction perform when given only the atomic connectivity for an organic compound. A blind test was conducted on a selection of four compounds and a wide range of methodologies representing, the principal computer programs currently available were used. There were 11 participants who were allowed to propose at most three structures for each compound. No program gave consistently reliable results. However, seven proposed structures were close to an experimental one and were classified as "correct". One compound occurred in two polymorphs, but only one form was predicted correctly among the calculated structures. The basic problem with lattice energy based methods of crystal structure prediction is that many structures are found within a few kJ mol(-1) of the global minimum. The fine detail of the force-field methodology and parametrization influences the energy ranking within each method. Nevertheless, present methods may be useful in providing a set of structures as possible polymorphs for a given molecular structure.

Crystal structure predictions for disordered halobenzenes

“…Only 1,2,4,5-tetrachlorobenzene and 2,3,6,7-tetrabromonaphthalene would have been found in a genuine crystal structure prediction. In two blind tests organized recently 11,12 a ranking up to three was accepted; this would have doubled the present success rate to 4 structures out of 24. This corresponds roughly to the general results of the blind tests, showing that the present state of the art leaves much to be desired.…”

Force Fields and Point Charges for Crystal Structure Modeling

“…Promising results have been obtained for neutral, fairly rigid molecules, but there are still limitations in the application to flexible molecules, polar molecules -especially when hydrogen bonds are involved -hydrates and salts 4,6 . In a series of benchmark tests 4,7,8 , invited experts using a wide variety of methodologies were challenged to propose candidate structures for each of a set of compounds, for which they only knew the molecular formula. The overall conclusion from the first three tests is that no technique gives consistently reliable results, with major obstacles being how to deal with structures with more than one molecule in the asymmetric unit, and the treatment of the conformational energy which is very important for flexible molecules.…”

“…Crystal structure predictions for a particular compound usually produce several hundred structures 13 within relevant lattice energy and density boundaries. If these structures are ranked in order of lattice energy or density, frequently the experimentally known crystal structure(s) will not be found at the top, possibly not even being found at all 4,13 . Poor energy models and search algorithms are partly to blame for this, but it is also possible that the nucleation and/or growth of certain thermodynamically favoured structures is hampered for kinetic reasons.…”

“…One of the most intriguing and challenging goals of molecular modelling is to predict possible crystal packing arrangements from nothing more than the molecular formula 4 . In the processing of crystalline organic fine chemicals and pharmaceuticals, crystal polymorphism is of tremendous importance; a large proportion of all organic compounds exhibit polymorphism, and different solid forms of a substance display different solid state properties, e.g.…”

DL_MULTI—A molecular dynamics program to use distributed multipole electrostatic models to simulate the dynamics of organic crystals

“…There are still many issues to be solved before polymorphism can be predicted [12]. Methods that have had some success in the international blind tests of crystal structure prediction [13,14] have been based on searching for the global minimum in the static lattice energy. However, most such searches find far more distinct crystal structures within the energy range of possible polymorphism than known polymorphs [15].…”