Molcas Forum

gmahesh

Member

Registered: 2024-07-02

Posts: 2

Restarting CASSCF from converged CI results in different energies

Hi all,

I optimized a benzene S0/S1 minimum-energy crossing point (CAS(6,6), 6 π electrons in 6 π orbitals) and it converged smoothly to a proper CI. When I redo a single point energy calculation at the optimized geometry and the same job's RasOrb file, I get noticeably different, non-degenerate energies. Input files of both jobs are provided below. Would appreciate it if someone could point out what I'm missing. Please let me know if any additional information is required.

Job 1(Initial CI optimization)

&Gateway
XBAS=6-311G**
ricd
ZMAT
 c1
 c2   1  1.46623
 c3    2  1.39487    1  115.579
 c4    3  1.39487    2  112.962      1   6.980
 c5    4  1.46623    3  115.579      2   -6.98
 c6    5  1.45406    4  120.561      3  -41.83
 h7    5  1.07699    6  119.884      1 -113.84
 h8    6  1.07546    5  119.776      4 -167.85
 h9    4  1.07375    5  120.650      6 141.906
 h10    3  1.07378    2  123.473      1 -176.43
 h11    1  1.07699    2  119.334      3 -143.58
 h12    2  1.07375    3  123.658      4 -169.16
End of z-matrix

Constraints
    a = Ediff 1 2
  Values
    a = 0.000
  End of constraints

>> Copy /home/mggr/Projects_2/Mahesh_azaborines/Regular_Benzene/cas_active_s0s1_ci_3_state_xms_sp_opt.RasOrb  INPORB

&SEWARD
&RASSCF
  LumOrb
  spin = 1
  nactel = 6 0 0
  inactive = 18
  ras2 = 6
  ciroot = 2 2 1

>> Copy $Project.rasscf.molden sa2_cas_s0s1_check.molden
>> Copy $Project.RasOrb INPORB
>>>  EXPORT  MOLCAS_MAXITER=200
>>> Do while

&SEWARD
&RASSCF
  LumOrb
  spin = 1
  nactel = 6 0 0
  inactive = 18
  ras2 = 6
  ciroot = 2 2 1
  Rlxroot= 2
&SLAPAF

>> Copy $Project.RasOrb INPORB

>>> EndDo

>> Copy $Project.RasOrb INPORB
>> Copy $Project.rasscf.molden sa2_cas_s0s1_minima.molden

S0 and S1 energies at the optmized CI from Job 1

:: RASSCF root number 1 Total energy:   -230.62051565
:: RASSCF root number 2 Total energy:   -230.62051559

Job 2 (Restarting with the optimized CI zmat)

&Gateway
XBAS=6-311G**
ricd
ZMAT
 C1
 C2       1     1.465861
 C3       2     1.393405   1   115.596944
 C4       3     1.393405   2   112.958133   1    -7.094152
 C5       4     1.465862   3   115.596942   2     7.094232
 C6       5     1.454255   4   120.463698   3    41.832441
 H7       5     1.077204   6   119.880424   1   113.867476
 H8       6     1.076054   5   120.097936   4   167.349344
 H9       4     1.074015   5   120.622128   6  -141.679248
 H10      3     1.073820   2   123.470198   1   176.457857
 H11      1     1.077203   2   119.423749   3   143.710780
 H12      2     1.074015   3   123.680476   4   169.273822
End of z-matrix

>> Copy /home/mggr/Projects_2/Mahesh_azaborines/Regular_Benzene/Clean/sa2_cas_s0s1_minima.RasOrb  INPORB

&SEWARD
&RASSCF
  LumOrb
  spin = 1
  nactel = 6 0 0
  inactive = 18
  ras2 = 6
  ciroot = 2 2 1
>> Copy $Project.rasscf.molden sa2_cas_s0s1_check.molden
>> Copy $Project.RasOrb INPORB

S0 and S1 energies from single point energy calculation, at optimized CI, from Job2

:: RASSCF root number 1 Total energy:   -230.61167437
:: RASSCF root number 2 Total energy:   -230.60109951

Thanks
Mahesh

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,204

Re: Restarting CASSCF from converged CI results in different energies

a) The Z-matrix has no information about the global orientation of the molecule, there's no guarantee that using a ZMAT as input will end up in the same orientation as the ZMAT you get from some arbitrary Cartesian coordinates. Then the INPORB might be completely wrong.

b) It seems in this case the main problem is that you're probably taking the Z-matrix reported after convergence, but there's a note:

  NOTE: on convergence the final predicted structure will be printed here.
  This is not identical to the structure printed in the head of the output.

If you take the geometry printed in the last RASSCF calculation (or the Z-matrix printed in the second-to-last SLAPAF), it should work better.