Multipoles versus charges in the 1999 crystal structure prediction test

Abstract

Developments towards a new generally applicable force field for the organic solid state are reported. The new force field combines atomic multipoles with an existing generic force field, iz. Dreiding, thus integrating an accurate description of electrostatic interactions with full molecular flexibility. This model was validated by means of crystal structure prediction for the set of molecules which was considered in the 1999 blind test on crystal structure prediction. Results are compared to the 1999 submission based on standard Dreiding with atom-centered point charges derived by fitting to the electrostatic potential. The improvements seen when using atomic multipoles are promising. For two out of five structures the charge-based predictions were already accurate enough to find the experimental structure among the three “best” structures, and these findings are reproduced by the multipole model. For two other compounds the multipole model results in significant improvements in the energy ordering of the possible crystal packings. Only one structure is still problematic. According to the rules of the test, the multipoles/Dreiding model has resulted in a correct prediction for four out of the five crystal structures involved, in comparison to two out of five for the charges/Dreiding approach.