Molcas Forum

Support and discussions for Molcas and OpenMolcas users and developers

You are not logged in.

Announcement

Welcome to the Molcas forum.

Please note: The forum's URL has changed. The new URL is: https://molcasforum.univie.ac.at. Please update your bookmarks!

You can choose an avatar and change the default style by going to "Profile" → "Personality" or "Display".

Pages: 1

#1 2024-11-05 00:47:17

taras
Member
Registered: 2024-11-05
Posts: 3

Different treatment of AMFI for RASSI

Hi!

In short, I am trying to calculate Spin-Orbit effects with different AMFI integrals on RASSI level. Here's a more detailed description of my problem:

What I have:
A Yb(III)-complex, for which I converged CASSCF(13,7) wavefunction. I am using ANO-RCC-VTZP basis set for Yb. After that I ran a MS-CASPT2 calculation for all 7 roots.
So, in terms of files, its $Project.RasOrb $Project.JobIph $Project.JobMix files.

What I want to calculate:
I want to obtain the Oscillator strengths (one-body matrix elements) between all my 7 roots with spin orbit coupling corrections. Specifically, I want to try different orders of DKH integrals (I.e. RELA keyword in SEWARD), and X2C/BSS hamiltonians.

What I tried:

The most naive approach is to invoke different AMFI in GATEWAY:

&gateway
coord=thf3-fac.xyz
group=nosym
basis=ano-rcc-vdz&gateway
coord=thf3-fac.xyz
group=nosym
basis=ano-rcc-vdz
basis=yb.ano-rcc-vtzp
basis=h.ano-rcc-mb
amfi

&seward
noguessorb
cholesky

>>COPY $CurrDir/$Project.JobMix $WorkDir/$Project.JobIph

&RASSI
  Spin  Orbit
  QIALL
  DIPR = 1.0D-10
  HEFF
  EJOB

&gateway
coord=thf3-fac.xyz
group=nosym
basis=ano-rcc-vdz
basis=yb.ano-rcc-vtzp
basis=h.ano-rcc-mb
amfi
rx2c

&seward
noguessorb
cholesky

&RASSI
  Spin  Orbit
  QIALL
  DIPR = 1.0D-10
  HEFF
  EJOB

&gateway
coord=thf3-fac.xyz
group=nosym
basis=ano-rcc-vdz
basis=yb.ano-rcc-vtzp
basis=h.ano-rcc-mb
amfi
rbss

&seward
noguessorb
cholesky

&RASSI
  Spin  Orbit
  QIALL
  DIPR = 1.0D-10
  HEFF
  EJOB

basis=yb.ano-rcc-vtzp
basis=h.ano-rcc-mb
amfi

&seward
noguessorb
cholesky

>>COPY $CurrDir/$Project.JobMix $WorkDir/$Project.JobIph

&RASSI
  Spin  Orbit
  QIALL
  DIPR = 1.0D-10
  HEFF
  EJOB

&gateway
coord=thf3-fac.xyz
group=nosym
basis=ano-rcc-vdz
basis=yb.ano-rcc-vtzp
basis=h.ano-rcc-mb
amfi
rx2c

&seward
noguessorb
cholesky

&RASSI
  Spin  Orbit
  QIALL
  DIPR = 1.0D-10
  HEFF
  EJOB

&gateway
coord=thf3-fac.xyz
group=nosym
basis=ano-rcc-vdz
basis=yb.ano-rcc-vtzp
basis=h.ano-rcc-mb
amfi
rbss

&seward
noguessorb
cholesky

&RASSI
  Spin  Orbit
  QIALL
  DIPR = 1.0D-10
  HEFF
  EJOB

I also tried a different order of DKH:

&gateway
coord=thf3-fac.xyz
group=nosym
basis=ano-rcc-vdz
basis=yb.ano-rcc-vtzp
basis=h.ano-rcc-mb

&seward
noguessorb
cholesky
rela
R07E07

>>COPY $CurrDir/$Project.JobMix $WorkDir/$Project.JobIph

&RASSI
  Spin  Orbit
  QIALL
  DIPR = 1.0D-10
  HEFF
  EJOB

All of the RASSI calculations produced the same "Total energies including SO-coupling".

Please tell me what I am doing wrong.

Thanks!

Offline

#2 2024-11-05 09:31:39

Ignacio
Administrator
From: Uppsala
Registered: 2015-11-03
Posts: 1,085

Re: Different treatment of AMFI for RASSI

I'm no expert, but I believe the AMFI integrals are not affected by the "RELA" level, and they should be enabled by default (no need to write the keyword). On the other hand, "RELA" would affect the one-electron integrals, but if you're copying the same JobMix file and using "EJOB", those have no effect, because the energies stored in the JobMix file (and AMFI) are used.

If you want to see the effect of different DKH orders, you'll have to re-compute the JobMix for each. For example, for a simple SCF calculation this shows some energy difference:

&GATEWAY
  Coord = 1

    Yb 0.0 0.0 0.0
  Basis = ANO-RCC-VTZP
  Group = NoSym
  RICD

&SEWARD

&SCF

&SEWARD
  Rela = R07E07

&SCF
                  Relativistic Douglas-Kroll-Hess integrals:
                    - Parametrization         : EXP
                    - DKH order of Hamiltonian: 2
                    - DKH order of Properties : 0

::    Total SCF energy                             -14046.4523901555
                  Relativistic Douglas-Kroll-Hess integrals:
                    - Parametrization         : EXP
                    - DKH order of Hamiltonian: 7
                    - DKH order of Properties : 7

::    Total SCF energy                             -14055.8009740011

Offline

#3 2024-11-05 18:54:16

taras
Member
Registered: 2024-11-05
Posts: 3

Re: Different treatment of AMFI for RASSI

Thanks for the prompt reply, Ignacio!

So I see, the AMFI and DKH integrals are decoupled and the issue was indeed in the fact that I copy the .JobMix file from my old CASPT2 calculation, in which the AMFI's are already computed.
For example, this simple input produces differents SO-RASSI state energies, as expected:

&GATEWAY
Coord = 1

Yb 0.0 0.0 0.0
Basis = ANO-RCC-VTZP
Group = NoSym
amfi=rx2c

&SEWARD

&SCF
charge=3
uhf
spin=2

&RASSCF
nactel=13 0 0
spin=2
charge=3
ras2=7
stav=7

&CASPT2
multistate
7
1 2 3 4 5 6 7

&RASSI
spin
ejob
heff

&GATEWAY
Coord = 1

Yb 0.0 0.0 0.0
Basis = ANO-RCC-VTZP
Group = NoSym
amfi=rbss

&SEWARD

&SCF
charge=3
uhf
spin=2

&RASSCF
nactel=13 0 0
spin=2
charge=3
ras2=7
stav=7

&CASPT2
multistate
7
1 2 3 4 5 6 7

&RASSI
spin
ejob
heff

I guess this solves it, so thanks again!

However, I do still have one question. Is there a way to not rerun all of the lengthy GATEWAY/SEWARD/RASSCF/CASPT2 calculation in order to compare the different treatments of the AMFI integrals? Ideally, I would like to have a converged RASSCF wavefuntion and then just run SEWARD/GATEWAY 3 times for DKH2/X2C/BSS hamiltonians to get different SO-RASSI energies. Is this possible with OpenMOLCAS? If not, I will just recompute everything.

Offline

#4 2024-11-06 13:15:01

Ignacio
Administrator
From: Uppsala
Registered: 2015-11-03
Posts: 1,085

Re: Different treatment of AMFI for RASSI

taras wrote:

the issue was indeed in the fact that I copy the .JobMix file from my old CASPT2 calculation, in which the AMFI's are already computed.

I don't think that's the case. The AMFI integrals are computed in SEWARD, but they're not stored in the JobMix, and they should not affect it. What changes the JobMix are the DKH integrals. Also, the AMFI keyword does not take an argument, but the "=" sign is translated to a newline, so what you are doing is enabling BSS or X2C, but probably for DKH (or rather as a replacement for DKH), I don't think it affects AMFI either.

Ideally, I would like to have a converged RASSCF wavefuntion and then just run SEWARD/GATEWAY 3 times for DKH2/X2C/BSS hamiltonians to get different SO-RASSI energies.

The problem is the DKH/X2C/BSS will affect the converged RASSCF wavefunction. What is converged with one will not be with the other, and the energies will be different. You could ignore this and pretend they're all the same, but then you have the issue that the AMFI integrals are always the same, you'll get always the same result:

* With DKH

&GATEWAY
Coord = 1

Yb 0.0 0.0 0.0
Basis = ANO-RCC-VTZP
Group = NoSym

&SEWARD
oneonly

> copy $CurrDir/amfi.JobMix JOB001

&RASSI
spin
ejob

* With BSS

&GATEWAY
Coord = 1

Yb 0.0 0.0 0.0
Basis = ANO-RCC-VTZP
Group = NoSym
rbss

&SEWARD
oneonly

> copy $CurrDir/amfi.JobMix JOB001

&RASSI
spin
ejob

* With X2C

&GATEWAY
Coord = 1

Yb 0.0 0.0 0.0
Basis = ANO-RCC-VTZP
Group = NoSym
rx2c

&SEWARD
oneonly

> copy $CurrDir/amfi.JobMix JOB001

&RASSI
spin
ejob

What you could do is take advantage of the fact that the optimized orbitals will probably be very similar, and use one solution as initial guess for the next. But you'll still need to run the full RASSCF+CASPT2 in order to get the correct results:

* With DKH

&GATEWAY
Coord = 1

Yb 0.0 0.0 0.0
Basis = ANO-RCC-VTZP
Group = NoSym

&SEWARD

&SCF
charge=3
uhf
spin=2

&RASSCF
nactel=13 0 0
spin=2
charge=3
ras2=7
stav=7

> copy $Project.rasscf.h5 dkh.h5

&CASPT2
multistate
7
1 2 3 4 5 6 7

> copy $Project.JobMix JOB001

&RASSI
spin
ejob

* With BSS

&GATEWAY
Coord = 1

Yb 0.0 0.0 0.0
Basis = ANO-RCC-VTZP
Group = NoSym
rbss

&SEWARD

&RASSCF
File = dkh.h5
nactel=13 0 0
spin=2
charge=3
ras2=7
stav=7

> copy $Project.rasscf.h5 bss.h5

&CASPT2
multistate
7
1 2 3 4 5 6 7

> copy $Project.JobMix JOB001

&RASSI
spin
ejob

* With X2C

&GATEWAY
Coord = 1

Yb 0.0 0.0 0.0
Basis = ANO-RCC-VTZP
Group = NoSym
rx2c

&SEWARD

&RASSCF
File = dkh.h5
nactel=13 0 0
spin=2
charge=3
ras2=7
stav=7

> copy $Project.rasscf.h5 x2c.h5

&CASPT2
multistate
7
1 2 3 4 5 6 7

> copy $Project.JobMix JOB001

&RASSI
spin
ejob

But again, this will show differences in the relativistic Hamiltonian and the spin-free states, not the AMFI integrals which will be the same in all cases (to the best of my knowledge).

PS. You're using the JobIph, not the JobMix file, and you're giving coflicting keywords EJOB and HEFF.

Offline

#5 2024-11-07 22:20:47

taras
Member
Registered: 2024-11-05
Posts: 3

Re: Different treatment of AMFI for RASSI

I see. Then I will just reuse the old orbitals and re-run SEWARD/RASSCF/CASPT2 each time. Thanks for such detailed answers, really helps a lot!

Offline

Pages: 1

Links: OpenMolcas | Molcas | Manual | Facebook

Board footer

Powered by FluxBB 1.5.11

Last refresh: Today 21:27:55