Molcas Forum

huangtf

Member

Registered: 2017-04-23

Posts: 18

Deleted orbitals

Dear Ignacio,

Thanks a lot for your kind and very useful reply last time! I have four questions to consult you.

1. As shown below, which came from my .log file, do you think if I should use "DELEted" keyword in &rasscf to remove some secondary orbitals?

++    Orbital specifications:
      -----------------------

      Symmetry species                           1
                                                 a
      Frozen orbitals                            0
      Inactive orbitals                         47
      Active orbitals                           12
      RAS1 orbitals                              0
      RAS2 orbitals                             12
      RAS3 orbitals                              0
      Secondary orbitals                       121
      Deleted orbitals                           0
      Number of basis functions                180
--

2. "DELEted" in &rasscf specifies the number of deleted orbitals in each symmetry. These orbitals will not be allowed to mix into the occupied orbitals. (page 301 of the manual) In general, under what situations should I use this keyword?

3. Will using "DELEted" in &rasscf to delete orbitals reduce the calculation accuracy and save calculation time compared to without?

4. Is 4096 MB of memory enough for my serial caspt2 job?

Thank you very much!

Sincerely yours,
huangtf

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,087

Re: Deleted orbitals

This is not not directly related to "Write input file for caspt2 optimization". Please write appropriate titles for your questions and create new topics when/if necessary. I've moved this to a separate thread.

There is generally no reason to use "Deleted" orbitals in RASSCF. Only when you have a very large basis set, or problems with convergence due to near-linear dependencies, or if you want to play tricks (like prevent some particular orbital being mixed with the active space). Deleted orbitals should have a similar effect to reducing your basis set: faster and less accurate calculation. If necessary and smartly executed, you can save time with minimal accuracy impact. In the limit, if you start from a converged calculation you can delete ALL virtual orbitals without affecting the accuracy at all, because those orbitals are actually playing no role. But as soon as you modify your conditions, if you don't allow those orbitals to be mixed with the occupied orbitals you are degrading you calculation.

I imagine 4GB would be more than enough, but you should try. If the CASPT2 program wants more memory it will tell you.

huangtf

Member

Registered: 2017-04-23

Posts: 18

Re: Deleted orbitals

Dear Ignacio,

Thank you very much for your detailed reply! I got it, and 4GB is more than enough (see below).

Estimated memory requirements:
  POLY3 :                 54392696
  RHS:                     1478720
  SIGMA :                  2552872
  PRPCTL:                  4912604
 Available workspace:    482978920

Best regards,
huangtf