Crystal Structure Prediction and the Cambridge Structural Database

Abstract

A computer program has been written to attempt prediction of likely polymorphs of small organic molecules, using a genetic algorithm. The cost function used is based not on energy but the intermolecular atom pair distances as seen in the Cambridge database, for selected similar molecules to the target. Some successful results are described, showing possible applications to molecules not easily treated with empirical energy potentials.

Key Words:

• crystal structure prediction
• database
• frequency curves