manual   quickstart   instguide   update   basis

Next: 46.2.17 Optimizing energy variables Up: 46.2 Directives for OPTG Previous: 46.2.15 Line searching (LINESEARCH)   Contents   Index   PDF

46.2.16 Reaction path following options (OPTION)

OPTION,key=param;

where key can be

IDIR
If starting at a transition state (or near a transition state) determine where to take the first step. If IDIR=0 is chosen, the first step will be towards the transition state. This is the default. If IDIR=1 is given in the input the first optimization step will be along the ''transition vector'' i.e. the hessian eigenvector to the smallest eigenvalue which points down towards the minimum. If using a larger IDIR parameter, the first step will be larger; if using a negative value, the first step will be in the opposite direction.
STPTOL
If using an updated hessian matrix, this parameter determines what update to take. If the step size between two subsequent points on which the steepest decent lines are puzzled together is smaller than stptol (i.e. if we are close to a minimum) the BFGS update is used, otherwise it is Powell update. The default value of stptol is $1.d-6$. Note that the stepsize is also affected by the STEPMAX option.
SLMAX
This option is only valid with the old version of the reaction path following algorithm (i.e. METHOD,SRSTEEP). In this algorithm slmax determines the frequency of the recalculation of the numerical hessian. If the total step size of the last steps exceeds slmax the hessian will be recalculated, otherwise it will be updated. By default slmax is two times the maximum step size of the optimization step steplength (see STEP section 46.2.12). If you are using METHOD,QSD, the SLMAX option is obsolete and the NUMHES command (see above) should be used instead.

These options can also be given on the OPTG command line.



Next: 46.2.17 Optimizing energy variables Up: 46.2 Directives for OPTG Previous: 46.2.15 Line searching (LINESEARCH)   Contents   Index   PDF

manual   quickstart   instguide   update   basis

molpro@molpro.net 2017-12-15