1. Obtain crystal structure in .res (shellx) format.
Can be obtained from Cambridge Structural Database by choosing to save in this format within Mercury. If necessary, edit to have entire molecules in asymmetric unit.

2. Obtain molecular coordinates in local axis system, using NEIGHCRYS.
Requires bondlengths file to define all intramolecular bonds, and dmacrys.axis file to define direction of molecular fixed axes in terms of the atom labels. Choose an axis system defined by the atoms in the molecule that will be around the centre of the molecule and uses any approximate molecular symmetry. NEIGHCRYS can adjust the lengths of any bonds to hydrogen in X-rays structures to standard neutron values.
Running NEIGHCRYS with these files produces a *.dmain input file for DMACRYS, without the intermolecular potential. Extract the molecular structure in the correct axis system from fort.21, by editing the section on the molecule into *.geom, using the co-ordinates in Angstrom.

3. Run ab initio calculation on the experimental molecular structure
Edit the structure from *.geom into a GAUSSIAN input file *.com. Run a e.g. MP2 631G(d,p) calculation at that geometry, ensuring that you save the charge density (*.Fchk) file.

4. Analyse the charge density to obtain the distributed multipole model. Use GDMA2 to analyse the Test.Fchk and obtain the set of atomic multipoles as specified by dmaMP2.dat to give gdma.dma. Since GAUSSIAN strips off the atom identifier information, it is necessary to restore this using GDMANEIGHCRYS and the *.geom to create a *.punch file containing the atom labels, coordinates and atomic multipoles relative to the local axis system.

5. Set up and run the lattice energy minimisation. This requires another run of NEIGHCRYS to set up a *.dmain file, as in (2) but also supplying the *.punch file to define the electrostatic model, and a MODEL.pots file containing the parameters for the Buckingham repulsion-dispersion potential. Submit DMACRYS to minimise within the value of MAXI iterations that you specify. (Note that to use the Williams 2001 potential, it is necessary to adapt the procedure, including the calculation of the multipoles, to allow for the hydrogen sites not corresponding to the proton positions.)

6. Check the results for a true minimum. Ensure that the minimization is VALID and that it has converged to a true minima, rather than a transition state. In the latter case, the minimisation should be re-run removing the symmetry element that gave rise to the negative eigenvalue. The minimised structure in .res format (fort.16) can be compared with the input structure in Mercury. Rerun DMACRYS with the STAR PROP command for accurate second derivative properties.