Molcas Forum

Anugraha

Member

Registered: 2025-12-17

Posts: 22

caspt2 warning Too large irlxroot,clarification of ciroot and staverag

Dear Molcas developers,
I am performing a single-state caspt2 calculation on state 2 of symmetry 1

&SEWARD

&RASSCF
 Fileorb = Project.ScfOrb
 symmetry =1
 Iterations = 200 100
 Spin = 1
 nActEl = 10 0 0
 Inactive = 25 0
 Ras2 = 0 8
 staverage = 2

&CASPT2
 Multistate
 1 2
 IPEAshift = 0.00

In the CASPT2 module, I receive the following warning:

WARNING: Too large iRlxRoot 
Reset to NSTATE=                    1

And the wavefunction specification in caspt2 module is :

Wave function specifications:
      -----------------------------

      Number of closed shell electrons          34
      Number of electrons in active shells      10
      Max number of holes in RAS1 space          0
      Max number of electrons in RAS3 space      0
      Number of inactive orbitals               17
      Number of active orbitals                  8
      Number of secondary orbitals             123
      Spin quantum number                      0.0
      State symmetry                             1
      Number of CSFs                          1176
      Number of CASSCF root(s) available         2
      CASPT2 state passed to geometry opt.       1
      A file JOBMIX will be created
      This is a CASSCF or RASSCF reference function
--

Despite this warning the CASPT2 energy for the state 2 appears to be calculated.

My doubts are :
1. Will this warning affect the accuracy of my caspt2 energy for state 2 ?

2. In casscf module i have tried ciroot and staverage separately, and i observed small differences in energy depending upon which one i have used. Should i use Staverage or ciroot ?

I would be thankful if anybody can comment on these doubts.

Thank you
Anugraha

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,204

Re: caspt2 warning Too large irlxroot,clarification of ciroot and staverag

1. RlxRoot does not affect single-point calculations, it's only relevant for gradient (geometry optimization or dynamics). By default it's set to the highest available root in RASSCF, and the value is passed to CASPT2. Since you are computing fewer roots in CASPT2 than RASSCF, hence the warning.

2. "StAverage = 2" should be equivalent to "CIRoot = 2 2 1". If you see differences, please show input/output files.