The crystal structures of 4,10-dibromo-anthanthrone (Pigment Red 168; 4,10-dibromo-dibenzo[def,mno]chrysene-6,12-dione), 4,10-dichloro- and 4,10-diiodo-anthanthrone have been determined by single-crystal X-ray analyses. The dibromo and diiodo derivatives crystallize in P2(1)/c, Z = 2, the dichloro derivative in P1, Z = 1. The molecular structures are almost identical and the unit-cell parameters show some similarities for all three compounds, but the crystal structures are neither isotypic to another nor to the unsubstituted anthanthrone, which crystallizes in P2(1)/c, Z = 8. In order to explain why the four anthanthrone derivatives have four different crystal structures, lattice-energy minimizations were performed using anisotropic atom-atom model potentials as well as using the semi-classical density sums (SCDS-Pixel) approach. The calculations showed the crystal structures of the dichloro and the diiodo derivatives to be the most stable ones for the corresponding compound; whereas for dibromo-anthanthrone the calculations suggest that the dichloro and diiodo structure types should be more stable than the experimentally observed structure. An experimental search for new polymorphs of dibromo-anthanthrone was carried out, but the experiments were hampered by the remarkable insolubility of the compound. A metastable nanocrystalline second polymorph of the dibromo derivative does exist, but it is not isostructural to the dichloro or diiodo compound. In order to determine the crystal structure of this phase, crystal structure predictions were performed in various space groups, using anisotropic atom-atom potentials. For all low-energy structures, X-ray powder patterns were calculated and compared with the experimental diagram, which consisted of a few broad lines only. It turned out that the crystallinity of this phase was not sufficient to determine which of the calculated structures corresponds to the actual structure of this nanocrystalline polymorph.