Molcas Forum

moabe

Member

Registered: 2018-06-03

Posts: 23

Molecular orbitals exchange in RASSCF calculation

Hi Everyone,

I'm trying to do CASSCF calculations for a ring-opening reaction. To do so, it is necessary to include sigma and sigma* orbitals of the breaking bond. My active space is 10 electrons in 10 orbitals. I've found those sigma orbitals away from my active space and then tried to exchange those with ones in the active space using ALTEr keyword. After RASSCF calculation, surprisingly no sigma orbital observed. Maybe I'm doing this, incorrectly. Does anyone have experience in this case to help me? In the following, I put my input file. Here, I'm using localized orbitals as initial guess and trying to exchange orbital 43 (HOMO-4) by orbital 31 (sigma). 

&Gateway
coord=mystructure.xyz
Basis=cc-pvdz
Group=C1
>>> Do while
&SEWARD
MEDIUM CHOLESKY
&SCF
KSDFT=B3LYP

&LOCALISATION
  File = $Project.ScfOrb
  Occupied

&LOCALISATION
  File = $Project.LocOrb
  Virtual
>copy $Project.LocOrb INPORB

&RasScf 
Nactel
 10   0   0
Spin
  1
Inactive
  28 
Ras2
  10 
Frozen
  14
CiRoot
  10 10 1
Alter
1
1 31 43
>>> EndDo

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Molecular orbitals exchange in RASSCF calculation

The exchanges specified by the ALTER keyword are done before the calculation, on the starting orbitals. Are all your localized orbitals in the right position (except for #31)? Even if you select your active space "right", there is no guarantee it will stay as you want it in the RASSCF calculation, as there may be orbital rotations that lower the energy. Try starting from a distorted geometry where your sigma bond is half broken. If you get a stable active space you can then use those orbitals as initial guess for geometries closer to your desired structure.

moabe

Member

Registered: 2018-06-03

Posts: 23

Re: Molecular orbitals exchange in RASSCF calculation

Dear Ignacio,

Thank you so much for your suggestion. I distorted my initial structure and repeated the RASSCF calculation. A stable active space is obtained with sigma and sigma* orbitals. Now, I want to come back to the initial structure and use RASSCF orbitals as initial guess. But, I wonder how to modify the input file in which the program takes my .RasOrb file. Could you please modify my posted input file for this case? I really appreciate your help.

Best,

Mostafa

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Molecular orbitals exchange in RASSCF calculation

Just add:

FileOrb = your_RasOrb_filename

to your &RASSCF block (and remove the ALTER part):

&Gateway
coord=mystructure.xyz
Basis=cc-pvdz
Group=C1
&SEWARD
MEDIUM CHOLESKY

&RasScf
FileOrb
  your_RasOrb_filename
Nactel
 10   0   0
Spin
  1
CiRoot
  10 10 1

The orbital specification (frozen, inactive, RAS2) should be correctly read from the RasOrb file.

moabe

Member

Registered: 2018-06-03

Posts: 23

Re: Molecular orbitals exchange in RASSCF calculation

Hi Ignacio,

Can I use more reliable MS-CASPT2 orbitals (Pt2Orb files), instead of RASSCF orbitals (RasOrb file) for the CASSCF optimization? If YES, how could it be done? I mean, for the RASSCF orbitals we have .RasOrb file contains orbitals for all roots and use "FileOrb = your_RasOrb_filename" in &RASSCF block. But, for the MS-CASPT2 there are individual files for each root. How does the input file look like for this case?

Many Thanks

Best,

Mostafa

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Molecular orbitals exchange in RASSCF calculation

You can use any set of orbitals as input orbitals. Just choose one of Pt2Orb files instead of the RasOrb file. But I don't think they are more reliable. In the limit where there is no geometry change, the orbitals from RasOrb will be already "perfect" and optimized.

moabe

Member

Registered: 2018-06-03

Posts: 23

Re: Molecular orbitals exchange in RASSCF calculation

Thank you Ignacio for your reply. Actually, the main reason that makes me to think about the MS-CASPT2 orbitals is my serious problem when I use RASSCF orbitals. From the single point CASSCF calculation, I've figured out which root is my desired state (low-lying singlet excited state; root 3) and chosen as Rlxroot. But, during the optimization, it starts to switch to other root and the program optimizes the wrong root. Even with TRACk keyword in &SLAPAF block, the problem doesn't solve. Do you have any idea or suggestion for this issue?  Many Thanks.

&Gateway
RICD 
coord=MyStructure.xyz
Basis=cc-pVDZ
Group=C1
Constraint
  a  =  Bond  C5  C10
  Value
  a  =  1.57759 angstrom
End  of  Constraints

>> EXPORT MOLCAS_REDUCE_PRT=NO
>> EXPORT MOLCAS_MAXITER=500

>>>  Do  while

&SEWARD
DoAnalytical
&RasScf
FileOrb = /directory/DHA.RasOrb
Spin=1; Nactel= 10 0 0; Inactive= 42; Ras2= 10
CiRoot
  4 4 1
Rlxroot=  3
&Alaska
NUMErical
&SLAPAF
TRACk
CARTesian

>>>  EndDo

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Molecular orbitals exchange in RASSCF calculation

You could try optimizing the intersection between the problematic states. It could tell you how the PES looks like around it and how to get away from it, if possible.