Posted by Björn Roos on October 30, 2007 at 16:52:10:
In Reply to: CASPT2 energy 6.2 & 6.4 posted by Piero Altoe' on October 29, 2007 at 18:03:06:
: I've done the same calculation using MOLCAS 6.2 & 6.4 the CASPT2 energy differs. while the 6.2 version gives the correct result the 6.4 is wrong (the difference is 0.53 eV).
: I've compiled the 6.4 using ifort 9.1 on a opteron machine (both case safe and fast).
: Do you have any explanation for this strange result?
: Thanks in advance for any help
Dear Dr. Altoe,
There is a fundamental change in CASPT2 from 6.4 and 7.0 compared to earlier version. The zeroth order Hamiltonian has been changed such that open and closed shell systems are treated in a more balanced way. As a result excitation energies are usually shifted upwards somewhat and dissociation energies are larger. It happens that agreement with experiment is not improved but this is usually due to a "fortunate" cancellation of errors in the old code. If one wants to get the old results back one has to use the keyword IPEA with the value 0.0 in CASPT2. Read more about this in the manual and also in: CPL,396, 142(2004).
Best Regards
Björn Roos