Prediction of correlation energies using variational subspace valence bond.

In the variational subspace valence bond (VSVB) [G. D. Fletcher, J.

Valence : A Massively Parallel Implementation of the Variational Subspace Valence Bond Method.

This work describes the software package, Valence, for the calculation of molecular energies using the variational subspace valence bond (VSVB) method. VSVB is an ab initio electronic structure method based on nonorthogonal orbitals. Important features of practical value include high parallel scalability, wave functions that can be constructed automatically by combining orbitals from previous calculations, and ground and excited states that can be modeled with a single configuration or determinant.

The variational subspace valence bond method.

The variational subspace valence bond (VSVB) method based on overlapping orbitals is introduced. VSVB provides variational support against collapse for the optimization of overlapping linear combinations of atomic orbitals (OLCAOs) using modified orbital expansions, without recourse to orthogonalization. OLCAO have the advantage of being naturally localized, chemically intuitive (to individually model bonds and lone pairs, for example), and transferrable between different molecular systems.

Large-Scale MP2 Calculations on the Blue Gene Architecture Using the Fragment Molecular Orbital Method.

Benchmark timings are presented for the fragment molecular orbital method on a Blue Gene/P computer. Algorithmic modifications that lead to enhanced performance on the Blue Gene/P architecture include strategies for the storage of fragment density matrices by process subgroups in the global address space. The computation of the atomic forces for a system with more than 3000 atoms and 44 000 basis functions, using second order perturbation theory and an augmented and polarized double-ζ basis set, takes ∼7 min on 131 072 cores.

A systematic multireference perturbation-theory study of the low-lying states of SiC3.

The three known lowest-energy isomers of SiC(3), two cyclic singlets (2s and 3s) and a linear triplet (1t), have been reinvestigated using multireference second-order perturbation theory (MRPT2). The dependence of the relative energies of the isomers upon the quality of the basis sets and the sizes of the reference active spaces is explored. When using a complete-active-space self-consistent-field reference wave function with 12 electrons in 11 orbitals [CASSCF (12, 11)] together with basis sets that increase in size up to the correlation-consistent polarized core-valence quadruple zeta basis set (cc-pCVQZ), the MRPT2 method consistently predicts the linear triplet to be the most stable isomer.

Scalable implementation of analytic gradients for second-order Z-averaged perturbation theory using the distributed data interface.

The analytic gradient expression for second-order Z-averaged perturbation theory is revised and its parallel implementation is described in detail. The distributed data interface is used to access molecular-orbital integral arrays stored in distributed memory. The algorithm is designed to maximize the use of local data and reduce communication costs.