Molcas Forum
Support and discussions for Molcas and OpenMolcas users and developers
You are not logged in.
- Topics: Active | Unanswered
Announcement
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".#1 2024-10-25 12:14:36
- luca
- Member
- Registered: 2016-09-29
- Posts: 2
Bug in RASSI with h5 files and states energies switching
Dear Everyone,
I'm experiencing a weird behaviour by RASSI in the following situation.
For my molecular system, I'm computing 5, state average, roots at the RASSCF level of theory.
Roots 3, 4, and 5 are relatively close in energy.
A subsequent MS-RASPT2 calculation re-orders the last three states since they have different contributions.
So far so good.
The problem starts with the following RASSI.
Case 1: how RASSI results should be.
If I provide a JobIph file as input for CASPT2 and follow that up using the obtained JobMix as JOB001 for RASSI, I obtain the following output:
++ Dipole transition strengths (spin-free states):
-----------------------------------------------
for osc. strength at least 1.00000000E-05
From To Osc. strength Einstein coefficients Ax, Ay, Az (sec-1) Total A (sec-1)
-----------------------------------------------------------------------------------------------
1 2 3.35374546E-01 9.56544879E+06 2.82390247E+07 5.13114623E+05 3.83175881E+07
1 3 7.42941270E-02 8.46607600E+04 4.75057328E+05 1.77422739E+07 1.83019920E+07
1 4 2.03399392E-02 5.30687134E+04 7.05544540E+03 5.81229696E+06 5.87242112E+06
1 5 1.70650859E-03 2.35173973E+05 8.60403776E+04 3.39604564E+05 6.60818914E+05
2 3 8.27511557E-03 5.62296001E+04 1.51083895E+05 9.91993141E+01 2.07412695E+05
2 4 2.07480119E-02 1.49900535E+05 4.51299789E+05 2.22982541E+05 8.24182865E+05
2 5 6.41504689E-02 1.18261237E+06 3.41227109E+06 5.89012348E+05 5.18389581E+06
3 4 2.89585496E-03 1.28623602E+03 3.45943982E+03 1.19723645E+02 4.86539948E+03
3 5 6.54862138E-03 2.57887812E+04 7.70238310E+04 1.07086068E+03 1.03883473E+05
4 5 2.70484209E-04 5.06652401E+00 1.40402395E+02 1.80697286E+03 1.95244177E+03
-----------------------------------------------------------------------------------------------
--
++ Dipole transition vectors (spin-free states):
---------------------------------------------
for vector sizes at least 1.00000000E-04
From To Dx Dy Dz Total D (a.u.)
-------------------------------------------------------------------------------
1 2 1.45119233E+00 2.49342981E+00 3.36108500E-01 2.90450001E+00
1 3 7.67284676E-02 1.81755873E-01 -1.11075962E+00 1.12814422E+00
1 4 5.39314347E-02 1.96645907E-02 -5.64412458E-01 5.67324174E-01
1 5 -9.10935210E-02 -5.50990378E-02 1.09466094E-01 1.52698262E-01
2 3 3.47104224E-01 5.68966226E-01 -1.45791343E-02 6.66645678E-01
2 4 -4.01226512E-01 -6.96178484E-01 4.89354360E-01 9.40805444E-01
2 5 6.61610270E-01 1.12383562E+00 4.66920809E-01 1.38518940E+00
3 4 -3.98499160E-01 -6.53536678E-01 -1.21578573E-01 7.75043947E-01
3 5 -3.31277031E-01 -5.72517116E-01 -6.75060044E-02 6.64888998E-01
4 5 8.38110479E-03 4.41197941E-02 1.58278539E-01 1.64526275E-01
-----------------------------------------------------------------------------------------------
--Case 2: the possible bug.
If I provided to CASPT2 as input a $Project.rasscf.h5 file, I obtain a $Project.caspt2.h5 file. If I copy the latter as JOB001, to pass on to RASSI, I get the following output:
++ Dipole transition strengths (spin-free states):
-----------------------------------------------
for osc. strength at least 1.00000000E-05
From To Osc. strength Einstein coefficients Ax, Ay, Az (sec-1) Total A (sec-1)
-----------------------------------------------------------------------------------------------
1 2 3.35374545E-01 9.56544878E+06 2.82390246E+07 5.13114621E+05 3.83175880E+07
1 5 7.42941269E-02 8.46607588E+04 4.75057323E+05 1.77422738E+07 1.83019919E+07
1 3 2.03399393E-02 5.30687139E+04 7.05544574E+03 5.81229699E+06 5.87242115E+06
1 4 1.70650859E-03 2.35173973E+05 8.60403776E+04 3.39604562E+05 6.60818912E+05
2 5 8.27511561E-03 5.62296004E+04 1.51083896E+05 9.91993272E+01 2.07412696E+05
2 3 2.07480119E-02 1.49900534E+05 4.51299786E+05 2.22982540E+05 8.24182860E+05
2 4 6.41504688E-02 1.18261237E+06 3.41227107E+06 5.89012347E+05 5.18389578E+06
5 3 2.89585492E-03 1.28623598E+03 3.45943972E+03 1.19723643E+02 4.86539935E+03
5 4 6.54862138E-03 2.57887811E+04 7.70238307E+04 1.07086069E+03 1.03883472E+05
3 4 2.70484207E-04 5.06652320E+00 1.40402387E+02 1.80697285E+03 1.95244176E+03
-----------------------------------------------------------------------------------------------
--
++ Dipole transition vectors (spin-free states):
---------------------------------------------
for vector sizes at least 1.00000000E-04
From To Dx Dy Dz Total D (a.u.)
-------------------------------------------------------------------------------
1 2 1.45119233E+00 2.49342981E+00 3.36108500E-01 2.90450001E+00
1 5 7.67284671E-02 1.81755873E-01 -1.11075962E+00 1.12814422E+00
1 3 -5.39314350E-02 -1.96645912E-02 5.64412461E-01 5.67324177E-01
1 4 -9.10935211E-02 -5.50990379E-02 1.09466093E-01 1.52698262E-01
2 5 3.47104225E-01 5.68966228E-01 -1.45791352E-02 6.66645680E-01
2 3 4.01226512E-01 6.96178483E-01 -4.89354359E-01 9.40805444E-01
2 4 6.61610269E-01 1.12383562E+00 4.66920809E-01 1.38518940E+00
3 3 -1.64207906E+04 3.14852703E+03 -1.25864134E+04 2.09278138E+04
3 4 -8.38110412E-03 -4.41197928E-02 -1.58278539E-01 1.64526274E-01
-----------------------------------------------------------------------------------------------
--where there is a clear difference with states numbering in the upper section, and a totally unheard-of self-excitation (root 3 to root 3).
Please note: I used the keyword EJOB for the first RASSI input (the one with JobMix) and no special keyword in the latter case. The rest of the RASSI input is as follows:
&RASSI
Nr of Job = 1 all
TRDI
TRD1These results were obtained using OpenMolcas 24.06, tag: 12-g13ccb99e8.
My system is huge, so I don't exactly know how to post here the relevant information to reproduce the error. I'm working on finding a minimal working example.
Thanks for any suggestions.
Cheers!
/Luca
Offline
#2 2024-10-25 14:42:20
- Ignacio
- Administrator
- From: Uppsala
- Registered: 2015-11-03
- Posts: 1,085
Re: Bug in RASSI with h5 files and states energies switching
Hmm... Since there has been some recent changes in the RASSI code, could you check with the latest master, or at least with 24.10?
Also, how does the top of the RASSI output (where the input states are described) look like for both calculations? The reindexing is probably consistent with the different input methods: JobMix loses all information about the original RASSCF states, and the states are simply indexed by MS-CASPT2 energy order; caspt2.h5 works on the input (RASSCF) states plus the full effective Hamiltonian, so if there is no significant mixing, it can still identify each final MS-CASPT2 state with its corresponding RASSCF state. The top of the output should help you clarify what is going on.
From the "Dipole transition strengths" it looks like a simple (3,4,5)->(4,5,3) reindexing, but the "3 3" row in the "Dipole transition vectors" looks worrying indeed.
Ah, looking at the code (eigctl.f), I think I got it:
DO K_=1,NSTATE-1
I=IndexE(K_)
DO L_=I+1,NSTATE
J=IndexE(L_)it should be DO L_=K_+1,NSTATE, or rather:
DO K_=1,IEND
I=IndexE(K_)
DO L_=JSTART,NSTATE
J=IndexE(L_)to take into account the SUBSET keyword, as in other blocks. I guess another patch will be coming.
Offline
#3 2024-10-25 15:23:44
- luca
- Member
- Registered: 2016-09-29
- Posts: 2
Re: Bug in RASSI with h5 files and states energies switching
Dear Ignaciao,
thanks for your answer. I'll try to compile as quickly as possible 24.10 and simply apply the suggested change in my code and see what happens.
In any case, I guess that it would make sense if RASSI were to keep the state numbering as re-ordered after MS-CASPT2 to not confuse the user.
Cheers!
/Luca
Offline