Contents: Table of Contents   Finding Minima  Chapter 4  Introduction

Calculating one energy gradient

This is not an operation which the user will generally want to do, however it is included for completeness. Consider the following dataset.

TITLE
Double zeta basis set for water....one gradient
SYMMETRY
CNV 2
END
BASIS DZ
ATOMS
OXYGEN     8.0   0.0   0.0         0.0
HYDROGEN   1.0   0.0   1.431534    1.1094114
END
GRADIENT
START
FINISH

This simple dataset calculates the gradient with respect to nuclear geometry changes for a closed-shell SCF example. Since the gradient requires the wavefunction, the SCF energy is automatically evaluated first. The final output looks like this :

 GRADIENT OF THE ENERGY
 ATOM         dE/dx          dE/dy          dE/dz
    1         0.00000000     0.00000000    -0.02466938
    2         0.00000000     0.00657215     0.01233469
    3         0.00000000    -0.00657215     0.01233469

ie the cartesian components of the gradient are given (in atomic units) for each atom.

Analytic gradients are available for the following types of calculation (which are also the types of calculation for which automated geometry optimisations are possible) :-

All types of SCF ie RHF, ROHF, UHF and GRHF,

All types of MP2, ie RHF, ROHF-MP2 and UHF-MP2,

RHF and UHF MP3,

All types of Density Functional Theory,

The BD and BD(T) coupled cluster methods.

In all, cases the appropriate keyword is GRADIENT. If one does a gradient (this also applies to geometry optimisations) certain accuracy thresholds in the program e.g. the SCF convergence, are automatically increased slightly from those appropriate for energy calculations.