Crystal Polymorph Prediction
M. Leslie
CSE Department, CLRC Daresbury Laboratory
J.B.O. Mitchell and S.L. Price
Department of Chemistry, University College London
D. Buttar and R.J.Roberts
AstraZeneca/Avicia
The program DMAREL [1] has found widespread usage in modelling the
crystal structures of polar organic molecules, where the crystal
packing is sensitive to the electrostatic forces [2]. Recent
enhancements include the following:
- The code has been extended to include anisotropy in the short range
repulsive interactions of the crystal forces. This has proved
necessary for molecules containing halogen atoms in particular [3]
- The program has been extended to calculate phonons. This may
be used to evaluate the statistical thermodynamics of the target
crystals to improve the quality of the predictions.
The anisotropic short range potentials have been used to investigate
cyanuric chloride (C3N3Cl3) and two
other more complex chlorinated azaaromatics [3]. An overlap model was
used to derive non-empirical parameters for short-range repulsive
atom-atom potentials. The model assumes a power law relationship
between the total repulsive energy Erep and the overlap
Sr
Erep = K[Sr]g
where the exponent g is just less than 1.
The overlap is subdivided into atom-atom contributions which are
then fitted to an analytical model. The short range repulsions are
then combined with an accurate distributed multipole electrostatic
model and an atom-atom dispersion model to give totally non-empirical
intermolecular pair potentials.
The crystal structure of cyanuric chloride has C-Cl...N interactions
which geometrically resemble hydrogen bonds. The Cl...N intermolecular
contacts are considerably shorter than the sum of the van der Waal
radii. One of these bonds is linear and two others have an angle of
173o. A potential model based on an empirical fit gives all
three bonds linear and a space group Rc rather than the experimentally
observed C2/c. The models using non-empirical potentials, whether
isotropic or anisotropic, all give the correct space group with the
C-Cl...N bond angle about 173o. There is a marked
improvement in the reproduction of the cell constants going from the
isotropic to the anisotropic model, the former gives a r.m.s error of
2.3%, the latter 1.3%.
To test the transferability of the potential parameters, two
other compounds were investigated. Some small changes were made to the
potential. First, the multipole charge distribution was calculated
using SCF rather than MP2. Secondly, the small anisotropy in the short
range potential on the N and C atoms in cyanuric chloride was
removed. This was to allow for the different environments for C and N
atoms in the larger molecules. When cyanuric chloride was recalculated
using this potential model a small improvement was
found. (r.m.s. 0.99%). The crystal structure of one of the two new
molecules was reproduced well, the other was not as successful.
References
[1] D.J. Willock, S.L. Price, M. Leslie, C.R.A. Catlow, J. Comput. Chem. 16 (1995) 628.
[2] D.S. Coombes, S.L. Price, D.J. Willock, M. Leslie, J. Phys. Chem. 100 (1996) 7352
[3] J. B. O. Mitchell, S. L. Price, M. Leslie, D. Buttar and
R. J. Roberts, The Journal of Physical Chemistry A, 9961-9971 (2001).
