[molpro-user] RS2 gradient error: WRONG E0 FROM CIGAM0I

Seth Olsen seth.olsen at uq.edu.au
Sun Sep 24 02:09:35 CEST 2017

Hi Molpro Users,

I have run into the error same previously brought up by Artur Nenov (see email copied below), which arises in the RS2 routine when gradients are calculated.  I have been unable to find further mention of it in the mailing list, and wanted to know if there had been follow-up or if a work-around had been found.  Can the threshold for matching E0 be tuned without recompilation?

Many Thanks,

******Original Message from Artur Nenov dated Oct 18 2011******

Dear Molpro developers,

I have encountered an error while trying an RS2 optimization in Molpro. It
occurs while Molpro tries to calculate the gradient in a SS-PT2 or MS-MR-PT2
The error output reads

WRONG E0 FROM CIGAM0I: -34.84434593   -34.84434594 (exemplary numbers)

and comes after the RESULTS FOR STATE list. The values given after the error
message give the Zeroth-order valence energy, which is calculated once in the
begining of the RS2 routine (without level shift) and then recalculated when a
gradient calculation is requested (this time adding the level shift). The
error occurs also when no error shift given.
The error can be straced back to the files cigam2ppi.f (SS-PT2) and
cigam2ppi_ms.f (MS-MR-PT2), where an accuracy of 1.d-9 for the delta of both
values is needed, otherwise the program aborts. Obviously, in the given case
the delta is higher. I all situation where I had this error so far both values
differ by one or two digits in the very last printed decimal place. Further so
far this error has not occured with smaller molecules like ethylene or
cyclohexadiene, only with bigger molecules.

My question is whether these small discrepances between the Zeroth-order
valence energies before and after the PT2 correction are due to some numerical
inconsistence or there is more to it.
Can I ignore this error (comment the abortion execution signal in the two files
above) or reduce the accuracy? Is this error somehow connected to the accuracy
of the calculated gradient?

Best regards,
Artur Nenov

Seth Olsen, PhD.
Honorary Fellow
School of Mathematics & Physics
The University of Queensland
QLD 4072  Australia
Ph: +61 7 3365 2816
A PGP public key for this address has been uploaded to the key servers.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20170924/114ed0d2/attachment.html>

More information about the Molpro-user mailing list