Improving upon CCSD(T): LambdaCCSD(T). I. Potential energy surfaces.

Despite the successes of spin-restricted coupled-cluster singles, doubles, and perturbative triples [CCSD(T)], it fails for systems away from equilibria, which might raise questions about transition states, e.g. In an attempt to remedy this failure, we implemented LambdaCCSD(T) in a general purpose form for open and closed shells and with arbitrary single determinant reference functions, and applied it to a wide variety of bond-breaking examples. LambdaCCSD(T) is shown to substantially improve the behavior of CCSD(T) at long bond lengths without significantly affecting the quality of the equilibrium results. Weighted average nonparallelity errors for HF, C(2), N(2), and H(2)O are reduced from 22 mhartree for CCSD(T) to 10 mhartree for LambdaCCSD(T). Using LambdaCCSD(T) with a Brueckner reference provides the best single reference coupled-cluster description of N(2)'s dissociation curve to date. Application of CCSD(T) and LambdaCCSD(T) to the concerted transition state of 1,3,5-trinitrohexahydro-1,3,5-triazine (RDX) indicates that this transition state is well described by either methods, and indicates that the activation barrier is too high for it to be a major pathway of decomposition.