constrained optimization on the tutorial example fails

Posted by Darko Babic on January 24, 2003 at 19:49:15:

I'm having problems with constrained optimization, using
Molcas 5.4/Linux/AMD compiled with PGI 4.0. I know from
the documentation that it may produce difficulties, but
this one seems to be relatively simple: constraint is
put on only one bond. What makes me suspicious is the
following part of the output (at the first structure
iteration step):

Rational Function Optimization: Lambda/2= -3.5283117832661719E-002

Norm of the Step --> -nan<--
Diverging in INTOCA!!!
1 nrc012 -nan
rMax= -nan
rOld= -nan

I have tried the example from the Molcas tutorial
(water molecule constrained to be linear), and it
also fails even after increasing number of itera-
tions in SLAPAF to 1000. It does not produce -nan-s,
as the above case, but still makes me wondering:

Rational Function Optimization: Lambda/2= -4.9124896155974511E-006

Norm of the Step --> 4.4138446453849722E-003<--
IntoCa: Too many iteration, no convergency!

This happens after 5 successful structure iterations,
both in Molcas 5.2 and 5.4.

I appreciate any hint/comment/suggestion.

Follow Ups:

Post a Followup

Name:
E-Mail:

Subject:

Comments:
: I'm having problems with constrained optimization, using : Molcas 5.4/Linux/AMD compiled with PGI 4.0. I know from : the documentation that it may produce difficulties, but : this one seems to be relatively simple: constraint is : put on only one bond. What makes me suspicious is the : following part of the output (at the first structure : iteration step): : Rational Function Optimization: Lambda/2= -3.5283117832661719E-002 : : Norm of the Step --> -nan<-- : Diverging in INTOCA!!! : 1 nrc012 -nan : rMax= -nan : rOld= -nan : I have tried the example from the Molcas tutorial : (water molecule constrained to be linear), and it : also fails even after increasing number of itera- : tions in SLAPAF to 1000. It does not produce -nan-s, : as the above case, but still makes me wondering: : Rational Function Optimization: Lambda/2= -4.9124896155974511E-006 : : Norm of the Step --> 4.4138446453849722E-003<-- : IntoCa: Too many iteration, no convergency! : : This happens after 5 successful structure iterations, : both in Molcas 5.2 and 5.4. : I appreciate any hint/comment/suggestion.