Geometry Optimization

In ADF the (first) derivatives of the energy with respect to nuclear displacements are calculated analytically for a particular geometry at the end of the SCF procedure. They are used to identify stationary points in the energy surface, notably for the automatic optimization of the molecular structure. This is carried out by a Newton-type iterative procedure, based on a quadratic local approximation of the energy surface and an initial estimate of the Hessian. This estimate is based on a force-field approximation incorporated in the program or read-in from a restart file, for instance, obtained from an explicit ADF calculation of harmonic frequencies. At each step in the optimization the estimate of the true Hessian is improved by an updating procedure using the difference of current and previous gradients in relation to the difference in geometries. Various Hessian update schemes are available in ADF. The default (for normal geometry optimization) is the BFGS (Broyden Fletcher Goldfarb Shanno) method, which contains a bias to keep the Hessian positive-definite, which should be the case near an energy minimum.

• initial Hessian estimate, constraints (combined), restraints possible
• Cartesian, internal and delocalized coordinates

Links

ADF User Documentation: geometry optimization (in delocalized, Cartesian, or internal coordinates)
ADF-GUI: structure and reactivity, Tutorial: geometry optimization, tutorial structure building
Examples: geometry optimization