Molcas Forum

daishudaishu

Member

Registered: 2019-02-24

Posts: 22

Weird RASSI energy

Dear MOLCAS users,

I am trying to calculate the XAS of CH3I with OpenMolcas. However, I just notice that RASSI will dramatically change the valence excitation energy.

Valence excitations:
::    MS-CASPT2 Root  1     Total energy: -7150.59619163
::    MS-CASPT2 Root  2     Total energy:  -7150.40051633
::    MS-CASPT2 Root  3     Total energy:  -7150.40020228
::    MS-CASPT2 Root  4     Total energy:  -7150.32736642

Excitations with core hole in 4d orbital of I.
::    MS-CASPT2 Root  1     Total energy:  -7148.67266158
::    MS-CASPT2 Root  2     Total energy:  -7148.67262147
::    MS-CASPT2 Root  3     Total energy:  -7148.66578712
::    MS-CASPT2 Root  4     Total energy:  -7148.66543060
::    MS-CASPT2 Root  5     Total energy:  -7148.66103915

RASSI result:
::    RASSI State    1     Total energy:  -7159.96523826
::    RASSI State    2     Total energy:  -7159.59468956
::    RASSI State    3     Total energy:  -7150.77664451
::    RASSI State    4     Total energy:  -7150.37392533
::    RASSI State    5     Total energy:  -7148.67266106
::    RASSI State    6     Total energy:  -7148.64583640
::    RASSI State    7     Total energy:  -7148.46690822
::    RASSI State    8     Total energy:  -7139.49897556
::    RASSI State    9     Total energy:  -7139.12525989
::    SO-RASSI State    1     Total energy:  -7159.96523826
::    SO-RASSI State    2     Total energy:  -7159.59468956
::    SO-RASSI State    3     Total energy:  -7150.77664451
::    SO-RASSI State    4     Total energy:  -7150.37392533
::    SO-RASSI State    5     Total energy:  -7148.67266106
::    SO-RASSI State    6     Total energy:  -7148.64583640
::    SO-RASSI State    7     Total energy:  -7148.46690822
::    SO-RASSI State    8     Total energy:  -7139.49897556
::    SO-RASSI State    9     Total energy:  -7139.12525989

We can see that red-flagged state energy change a lot (about 9 Hartree)! I believe that the result is already wrong. If we only have the valence excitations, the RASSI energy only changes about 0.01 Hartree compared with MS-CASPT2 Root.

The following is my input file.

&GATEWAY  &END
  Title= molecule
  coord  = CH3I-001.xyz
  basis  =  ANO-RCC-VDZ ;  group=  E
  AMFI
SDIP
ANGM
 0. 0. 0.
End of Input
&SEWARD
&SCF
&RASSCF  &END
 LumOrb
  Spin=  1;  Nactel=  18  1 0;
ALTER
3
1 20   25
1 21   26
1 22   27
SUPSYM
1
5 20 21 22 23 24
  Inactive=  22 ;
  Ras1 = 5
  Ras2 =  7
  Ras3 = 0
  CiRoot=  4 4     1
End of Input
>> COPY  $Project.JobIph   $Project.S0.JobIph
>> COPY  $Project.rasscf.molden $Project.S0.molden

&CASPT2  &END
IMAGinary
0.15
MAXITER
100
  *Frozen=  52
  MultiState
  all
End of Input
>> COPY $Project.JobMix $Project.S0.JobMix
&RASSCF  &END
 LumOrb
  Spin=  1;  Nactel=  18  1 0;
SUPSYM
1
5  23 24 25 26 27
HEXS
1
1
  Inactive=  22 ;
  Ras1 = 5
  Ras2 = 7
  Ras3 = 0
  CiRoot=  5 5    1
End of Input
>> COPY  $Project.JobIph   $Project.S1x.JobIph
>> COPY  $Project.rasscf.molden $Project.S1x.molden
&CASPT2  &END
IMAGinary
0.2
MAXITER
200
  *Frozen=  52
  MultiState
  all
End of Input
>> COPY $Project.JobMix $Project.S1x.JobMix

>>COPY  $Project.S0.JobMix  JOB001
>>COPY  $Project.S1x.JobMix  JOB002

&RASSI  &END
TRDI
 Nr  of  JobIph=  2  4 5;  1 2 3 4;  1  2  3  4  5
  Spinorbit
  EJob
End of Input

The following is the xyz file.

     5
 geometry
 C                     0.00001425    -0.00000038     1.81970885
 H                     1.03457785    -0.00000026     2.15172706
 I                    -0.00000045    -0.00000005    -0.32779794
 H                    -0.51727735     0.89595583     2.15171140
 H                    -0.51727738    -0.89595646     2.15171254

I just find that if I remove EJob, the RASSI energy will be very close to CASPT2 energy. However, according to the manual, EJOB is necessary for the Spin-orbit calculation. In a different test where the spin is 3 in S1x calculation, the RASSI energy is close to CASPT2 energy.

I was wondering if I find a bug in RASSI. When Spin=1 in S1x calculation, all states with the core hole and without core hole are all singlet states. Maybe RASSI somehow couples them and gets wrong energy. Removing EJob will require the RASSI to recalculate the energy, which will fix the energy issue. When Spin=3 in S1x calculation, core hole states are triplet, so they cannot couple with the valence excited singlet state. So the RASSI energy looks good.

Please help me out!

Thanks a lot!

Last edited by daishudaishu (2019-02-26 08:50:09)