Stacking energies for average B-DNA structures from the combined Density Functional Theory and Symmetry-Adapted Perturbation Theory Approach - Supporting Information - Annamaria Fiethen, 1 Georg Jansen, 1,* Andreas Heßelmann, 2 and Martin Sch¨ utz 3 (1) Fachbereich Chemie, Universit¨at Duisburg-Essen, Campus Essen, Universit¨atsstr. 5, D-45117 Essen, GERMANY (2) Lehrstuhl f¨ ur Theoretische Chemie, Universit¨at Erlangen-N¨ urnberg, Egerlandstr. 3, D- 91058 Erlangen, GERMANY (3) Institut f¨ ur Physikalische und Theoretische Chemie, Universit¨at Regensburg, Univer- sit¨atsstr. 31, D-93040 Regensburg, GERMANY (*) e-mail: georg.jansen@uni-due.de Further computational details Geometries Atomic coordinates generated by the 3DNA program (cf. Ref. 10 of the main text) normally do not contain those of the hydrogen atoms which, in general, are experimen- tally inaccessible. To obtain them first all atoms belonging to the backbone of DNA were deleted in the atomic coordinate files of the various nucleotides. Then the non-hydrogen atomic positions of the AT and CG dimers were generated with 3DNA using the appropriate complementary base pair parameters (values given in main text). Subsequently the positions of manually added hydrogen atoms were refined in geometry optimizations using the TZVPP basis set (cf. Ref. 12) in frozen-core MP2 calculations. This level of theory is expected to yield reliable CH and NH bond lengths and fairly accurate hydrogen bridge geometries, thus also providing a starting point for future investigations of the hydrogen bonds in average B-DNA. The positions of all non-hydrogen atoms were kept fixed during the geometry optimizations. Finally, the optimized hydrogen atom positions were added to the atomic coordinate files of the various nucleotides used in 3DNA and the tetramer structures were generated with the appropriate base pair step parameters (values given in main text). The corresponding structures are shown in Figure S1, full lists of atomic coordinates can be found in Table S6. Exchange-correlation potential (Cited publications listed as Ref. 18 in the main text.) The LPBE0AC xc potential used in the DFT-SAPT calculations is a localized version of the PBE0AC xc potential. The latter was combined from the PBE xc potential of Perdew, Burke and Ernzerhof in its PBE0 modification by Adamo and Barone and the asymptotically correct LB94 xc potential of van Leeuwen and Baerends through the gradient-regulated seamless connection technique of Gr¨ uning et al., taking a fraction of 3 4 v xc (LB94) in the long-range part. The remaining 1 4 of the desired asymptotic 1/r behaviour is already taken into account through the exact exchange part of the PBE0 potential. The shift parameter required for the seamless connection of the asymptotic correction was determined from the ionization 1