Critical assessment of the performance of the semiempirical divide and conquer method for single point calculations and geometry optimizations of large chemical systems

van der Vaart, Arjan; Suárez, Dimas; Merz, Kenneth M.

doi:doi:10.1063/1.1323257

Volume/Page
Keyword
DOI
Citation
Advanced

Volume: Page/Article:

Search Content Type

article doi:

Citation:

Quantitative harmonization of the three molecular orbital, valence bond, and broken symmetry approaches to the exchange coupling constant: Corrections and discussion

Three current methods, used to evaluate exchange coupling constants in molecular magnetism, i.e., the molecular orbital (MO) model [Hay et al., J. Am. Chem. Soc. 94, 4884 (1975)], the valence bond (VB) model [Kahn and Briat, J. Chem. Soc. Trans. II 72, 268 (1976)], and the broken symmetry (BS) model [Noodleman, J. Chem. Phys. 74, 5737 (1981)], have been revisited. In effect, the three published antiferromagnetic contributions seem mutually inconsistent, as far as their magnitudes are concerned. ...

Rovibrational Hamiltonians for general polyatomic molecules in spherical polar parametrization. III. Global vs local axis system and angular coordinates

We study different parametrizations of the angular space of polyatomic molecules for an orthogonal description of the molecular geometry. Kinetic energy operators for pentatomic molecules, given by compact and computationally useful forms in a global and a local formulation of the axis system, are compared and discussed. A new decomposition of for sequentially bonded pentatomic molecules in conjunction with a basis of Wigner and associated Legendre functions provides kinetic energy matrix elem...

We present a detailed analysis of the performance of the semiempirical divide and conquer method as compared with standard semiempirical MO calculations. The influence of different subsetting schemes involving dual buffer regions on the magnitude of the errors in energies and computational cost of the calculations are discussed. In addition, the results of geometry optimizations on several protein systems (453 to 4088 atoms) driven by a quasi-Newton algorithm are also presented. These results indicate that the divide and conquer approach gives reliable energies and gradients and suggest that protein geometry optimization using semiempirical methods can be routinely feasible using current computational resources. © 2000 American Institute of Physics.