Geometry optimization using non-variational wave functions

DALTON does not have any support for the calculation of molecular gradients and Hessians for the non-variational wave functions CI and NEVPT2. However, in order to exploit the facilities of the first-order geometry optimization routines in DALTON, a numerical gradient based on energies will be calculated if a geometry optimization is invoked for a non-variational wave function. As a simple example, to optimize the MP2 geometry of a molecule using numerical gradients^7.1, the only input needed is

**DALTON INPUT
.OPTIMIZE
**WAVE FUNCTIONS
.HF
.MP2
**END OF DALTON INPUT

The size of the displacements used during the evaluation of the numerical gradient can be controlled through the keyword .DISPLA in the *OPTIMI input module. Default value is $1.0\cdot 10^{-3}$ a.u. By default, the threshold for convergence of the geometry will be changed because of estimated inaccuracies in the numerical gradients. However, if the threshold for convergence is altered manually, this user supplied threshold for convergence will be used also in geometry optimizations using numerical gradients. Note that due to the possibility of larger numerical errors in the gradient, too tight convergence criteria for an optimized geometry may make it difficult for the program to obtain a converged geometry.