lowest state of determinantal CI cannot be reproduced

[1234zou]

• [ ] Summary of the issue and expected results. I performed a MRMP2 calculation based on the CASSCF(8,8) reference. The converged CASSCF(8,8) orbitals were provided as the $VEC group in the .inp file. I wrote NSTATE=10 in the $DET group because for my studied molecule, to find the lowest singlet state, the determinantal CI program needs 5~10 roots. The strange thing is that the first CASCI found the lowest singlet successfully within 10 roots, but the CASCI before the MRMP2 could not found any singlet. These two CASCI actually utilizes the same one determinantal CI program. Is it possible that the CASCI code forgets to read in the canonical MOs, but keep NOs in memory?

• [ ] Description of the run environment:

  • operating system CentOS 7.6
  • compiler Intel 2019 Update 5
  • math library Intel MKL
  • communication library sockets
  • GAMESS version 2017-Apr(R1)

• [ ] An input file (if applicable) to reproduce the issue.
  Please see the attached files.

• A copy of your install.info file. Please see the attached files. GMS_CASCI_bug.zip

[saromleang]

This is the observed behavior of GAMESS (version August 2011) with your input.

[1234zou]

This is the observed behavior of GAMESS (version August 2011) with your input.

Hi, if my understanding is correct, the above reply is a reproduction of the behavior on other GAMESS version, not a solution to the problem, right? Please tell me if I miss the point. Thanks!

[saromleang]

Correct.

Often times the question is asked if any issue is due to recent changes in GAMESS since transitioning to GitHub.

This issue persisted before that transition and stems all the back to 2011. Which would suggested that it is a behavior of the original code.

We will try our best to track down the original author of the code to have a look at this issue.