Can DFT and ab initio methods adequately describe binding energies in strongly interacting C₆X₆⋯C₂X_n π–π complexes?

We calculate the CCSD(T)/CBS complexation energies of C₆X₆⋯C₂X_n complexes (X = F, Cl; n = 2, 4) by means of the W1-F12 and CCSD(T)/MP2(CBS) procedures. These complexes involve π–π stacking, charge-transfer, and van der Waals interactions and their complexation energies range between 11.1 (C₆F₆⋯C₂F₂) and 34.0 (C₆Cl₆⋯C₂Cl₄) kJ mol^{− 1}. We use our best CCSD(T)/CBS data to assess the performance of DFT, double-hybrid DFT (DHDFT), and standard/composite ab initio methods. The G4 and G4(MP2) composite methods show relatively poor performance with root-mean-square deviations (RMSDs) of 8.7 and 6.3 kJ mol^{− 1}, respectively. With the main exception of the Minnesota functionals, DFT functionals without a dispersion correction do not predict binding in these complexes. Most of the conventional DFT procedures attain RMSDs above the ‘chemical accuracy’ threshold. The best performing functionals with RMSDs ≤2.0 kJ mol^{− 1} are: B3LYP-D3, PW6B95-D3, LC-ωPBE-D3, PWPB95-D3, B2GP-PLYP-D3, and B2-PLYP-D3.