Attempts to model neutral solitons in polyacetylene by ab initio and density functional methods.: The nature of the spin distribution in polyenyl radicals

Different quantum chemical methods have been examined, in order to test their abilities to model long polyenyl radicals and, in the limit, a neutral soliton in polyacetylene. For polyenyl radicals up to C9H11, it was possible to use a range of methods up to CCSD(T). A number of interesting features of the geometries and spin distributions were found in these radicals. Coupled cluster methods reproduced rather well the experimental ratios of the spins at each of the carbons in allyl, pentadienyl, and heptatrienyl radicals. In contrast, both restricted and unrestricted Hartree-Fock methods gave very poor results. CASSCF gave good geometries and pi spin distributions, but only if all the pi MOs were included in the active space. Both pure (BLYP) and hybrid (B3LYP) DFT methods were found to over-estimate the widths of solitons in long polyenyl chains.