Generalized gradient approximations (GGA’s) seek to improve upon the accuracy of the local-spin-density (LSD) approximation in electronic-structure calculations. Perdew and Wang have developed a GGA based on real-space cutoff of the spurious long-range components of the second-order gradient expansion for the exchange-correlation hole. We have found that this density functional performs well in numerical tests for a variety of systems: (1) Total energies of 30 atoms are highly accurate. (2) Ionization energies and electron affinities are improved in a statistical sense, although significant interconfigurational and interterm errors remain. (3) Accurate atomization energies are found for seven hydrocarbon molecules, with a rms error per bond of 0.1 eV, compared with 0.7 eV for the LSD approximation and 2.4 eV for the Hartree-Fock approximation. (4) For atoms and molecules, there is a cancellation of error between density functionals for exchange and correlation, which is most striking whenever the Hartree-Fock result is furthest from experiment. (5) The surprising LSD underestimation of the lattice constants of Li and Na by 3–4 % is corrected, and the magnetic ground state of solid Fe is restored. (6) The work function, surface energy (neglecting the long-range contribution), and curvature energy of a metallic surface are all slightly reduced in comparison with LSD. Taking account of the positive long-range contribution, we find surface and curvature energies in good agreement with experimental or exact values. Finally, a way is found to visualize and understand the nonlocality of exchange and correlation, its origins, and its physical effects.
A variety of new and old ideas have been combined to develop a variational algorithm for generating an integration mesh for use in electronic-structure calculations. While our interests are in deriving an efficient yet accurate mesh for local-orbital calculations, the method which we introduce is independent of basis set, very efficient, and allows systematic refinements. The method incorporates a new algorithm for integrating over arbitrary interstitial regions and an improved atomic-spheres integration technique (errors decrease exponentially with the number of radial points). The mesh points and volume elements are optimized to ensure that a large class of integrals are evaluated accurately. We employ two new ideas in optimizing the mesh. First, we introduce a set of one-dimensional Gaussian-quadrature schemes depending continuously on a variational parameter. The introduction of this variational parameter leads to substantially more efficient integration meshes compared to traditional Gaussian-quadrature schemes. Second, we show how force algorithms may be used to refine the resulting meshes. The accuracy and efficiency of our scheme is assessed by presenting charge-density integrations on a variety of large isolated clusters. The number of mesh points required using this scheme is as much as an order of magnitude less than what is required with conventional special-points algorithms. For example, for the tetrahedral ground state of a seventeen-atom crystal fragment, a total of 6081 points is needed to integrate the total charge and total energy to six-decimal-place accuracy.
We make use of the existing formalism for calculating atomic forces within the local-density approximation (LDA) to determine forces in an all-electron, local-orbital framework. The forces are calculated as the proper total-energy derivatives, including the necessary basis-set corrections. In our computational technique the LDA potential is evaluated exactly on a variationally determined integration mesh which allows all integrals relevant to the electronic-structure problem to be computed to any desired accuracy. We apply our technique to the molybdenum dimer, the ethylene molecule, and the NaCl monomer to demonstrate the high accuracy of the calculated forces. In addition, we investigate directly the effects of integration and self-consistency errors on the accuracy of the forces, and use the results of these tests to suggest strategies for the efficient use of force information in structural studies.
We have used a recently developed computational technique based on density-functional theory to study the Raman-active modes of amorphous and Vibrational modes and the associated Raman activities for three cluster building blocks of the glasses are calculated directly from first principles. The positions of the calculated symmetric-stretch modes in the cluster models are in excellent agreement with sharp features in the observed spectra. Moreover, simulated spectra based on the cluster results are in good agreement with experiment, accounting for all the observed features in the bond-stretch region of the spectra. The cluster results suggest a new interpretation for the 250 mode appearing in the spectra of Ge-rich samples in the family.
We describe a novel method for the structural optimization of molecular systems. Similar to genetic algorithms (GA), our approach involves an evolving population in which new members are formed by cutting and pasting operations on existing members. Unlike previous GA's, however, the population in each generation has a single parent only. This scheme has been used to optimize Si clusters with 13–23 atoms. We have found a number of new isomers that are lower in energy than any previously reported and have properties in much better agreement with experimental data.