Is "ricc2 -fanal" working in TM-6.0?

[evgeniy]

Dear All,

I am wondering whether "ricc2 -fanal" (to perform NBO analysis)
is working in TM-6.0. So far I get:

...
Now analysing density file "cc2-xsdn-1a-001-total.cao"

Population analysis disabled for this density ... skipped


when trying to apply it. My input for NBO is:

$anadens
cc2-xsdn-1a-001-total.cao
$pop paboon

I am aware that there were problems with "ricc2 -fanal" in previous versions of TM (5.9. and 5.10).
This was not the case for TM-5.8 where "ricc2 -fanal" works fine. Howevere I am interested in
doing $pop paboon which is not available in TM-5.8 but in TM-6.0. So I  will very appreciate
if someone could instruct me how to "make" "ricc2 -fanal" work in TM-6.0. Many Thanks!

Best regards,
Evgeniy

[Arnim]

Hi,

for a populations analysis you have to run ricc2 without -fanal, because
the orbital informations can not be skipped in case of a correlated calculation.

Best,
Arnim

[Jonas]

Dear Arnim,

can you please be somehow more explicit on that?

I'm using TURBOMOLE v. 5.10 and basically try to do the same, i.e. to carry out population analysis for CIS and CIS(D) excited state densities.
ricc2 -fanal says

...
Now analysing density file "ccs-tm0f-1a-001.cao"
Population analysis disabled for this density ... skipped
...

From your reply this appears to be the wrong way, anyway. However, simply stating "$anadens ccs-tm0f-1a-001.cao" and "$pop" in control and running ricc2, as implied by your post, provides no density analysis at all; at least not from ricc2, whereas dscf does a ground state density population analysis as it should.
I'm afraid the v. 5.10+ manuals don't say a thing that helps me out.
Can you possibly say what I'm doing wrong?

Thank you very much in advance,
best regards, Jonas

[Arnim]

Hi Jonas,

actually I have overlooked, that Evgeniy did use an exited state density. My answer refered to ground state calculations. Sorry!
Population analysis in exited states can not be done with the ricc2 module, as far as I know. You could use egrad for an analysis at CIS level.

Cheers,
Arnim

[Jonas]

Arnim,

thank you for your helpful reply!    
I am using egrad at the CIS level now, as you suggested, for a number of toy systems at the moment.
For this, I run through define's EX sub-menu and set CISS to ON to add something like
$scfinstab ciss
$soes
 a            4
to my control file. I then add $pop to control and run dscf, and egrad -proper.

Now, egrad gives a Mulliken PA on StdOut for one density only, regardless of what number of states in irrep a I give in $soes. Trying to state $exopt n for different n equal-or-less-than 4 does not change the Mulliken PA, which, however, should be (slightly) different for different excited state densities?
So, which density PA do I get from this? The lowest excited state one? Or the highest, as suggested by the v. 5.10 manual, p. 235?

I'm sorry for asking that much, but things appear to be not too straightforward ...  
However,  

thank you in advance,
Best regards, Jonas

[Jonas]

Dear Arnim,

maybe a (hopefully last) question related to this topic: 
I notice small differences in singlet excitation energies of my He2H^+ toy system on a CIS level (using egrad, as described above) and a RI-CIS/RI-CCS level of theory (using the ricc2 module), i.e.:

0.3712574389, 0.4858293214, 1.0666292910, 1.1825491403 (a.u., CIS using egrad)
0.3711942201, 0.4857787660, 1.0664976991, 1.1824377396 (a.u., RI-CIS/RI-CCS using ricc2)

I am aware that there is a very small error introduced from the RI approximation. However, I'm unsure if it's showing up like this, or I'm understanding something completely wrong.
As far as I know, CIS and CCS should give identical excitation energies, as the v. 5.10 manual states CIS and CCS to be equivalent?

Again, thank you so much in advance,
Best regards, Jonas

[Arnim]

Hi Jonas,

that's right. The excitation energies between CIS with egrad and RI-CCS with ricc2 should differ only slightly
due to the RI error and this is what you a seeing here.
This difference should get smaller, if you expand the cbasis, and rerun the calculation.

Best,
Arnim

[Jonas]

Thanks so much, Arnim!  
Can you, moreover, or somebody else maybe, comment on my other post in a minute, if there's an answer? I really have no idea where to find any information on which state's density I get a Mulliken PA for using egrad ...

Best regards, Jonas

[christof.haettig]

Dear Evgeny and Jonas,

the properties which are available through the -proper flag and described in the "Molecular Properties,..." capter of the manual fall into two classes:
some of then are computed from one-electron densities, others are computed from orbitals.
The type of properties, which are computed from the one-electron densities can be computed straight forwardly also for excited states.
The second class of properties, which are computed from orbitals (and maybe assume that they have occupation number of 2, 1 or 0) can NOT
be applied to excited states, when these are computed using any kind of response method like TDDFT or Coupled Cluster response (as used in ricc2)
or any dynamically correlated multireference approach. All these methods do not produce "excited state orbitals", but describe the excited state by a
linear combination of (many) determinants.
So neither egrad nor ricc2 produce "excited state orbitals", and the properties which a calculated from orbitals (either because the require them for
theoretical reasons or because of technical issues in TM) can not be computed this way. These are
        *   orbitals localization
        *  the following population analysis: Mulliken, Loewdin, Wiberg
        *  generation of plot files for canonical or localized orbitals

Note: When you use ricc2, these functionalities are disabled. When you use egrad the ground state DFT or HF-SCF canonical orbitals are used and some of the population analysis might give ill-defined results.

For ricc2 there are two different options for density visualization/analysis:
      * -proper : it passes well-defined defined ground and/or excited state density matrices to the property routines
                         and gives access to all properties which can be computed from these. It should work for the paboon and nbo
                         analysis...
      * -fanal:     it is (exclusively) meant for the visualization of difference densities or linear combination of densities
                         for such a user defined linear combination of densities  a population analysis in usually not meanigful,
                         therefore all population anylsis functionalities are disabled for -fanal.
 

[Jonas]

Dear Christof,

thank you very much for your very comprehensive answer! I think you guys do a great job in this forum!
I apparently confused myself with a number of points, as I didn't know that I need orbitals for a Mulliken population analysis. I thought a density would be enough; clearly, there are no "excited state orbitals" I can get from e.g. CC2, ...
For my local TURBOMOLE v. 5.10, however, ricc2 (both with and without the -proper flag) does run in analysis mode, but won't give any population analysis. It does find all density files I specify in control at $anadens, but skips both paboon- and nbo-type PAs, stating

Now analysing density file "ccs-tmf0-1a-002.cao"
Population analysis disabled for this density ... skipped

If I didn't make a stupid mistake, things are Ok for me, as I would not be able to fix it anyway. However, I will stop trying then ...

Thank you again for your support, Christof and Arnim,
Best regards, Jonas

[christof.haettig]

in TM V5.10 the population analysis was for ricc2 (not yet) enabled.

To use it, you need V6.0 or V6.1. With these recent version you can either:
    - insert the $pop keyword in the control file before you start any ricc2 calculation and use the usual ricc2 input to request a gradient calculation
      for the excited state in which you are interested. The population analysis will be printed in the output as following the excited state property
      gradient section
   -  if you already the the (excited state) gradient calculation, but kept you input and the file "exdens", you can do the population analysis also
      "post mortem" by calling the ricc2 once again with the -proper flag (inserting before the $pop keyword). But this only works if you haven't
      deleted the density or any input file



Christof

[Jonas]

Thank you again, Christof!

[evgeniy]

Dear participants of the discussion,

Many thanks for your posts. I have just a short
remark. I tried to run ricc2 -proper, having inserted
$pop in the control file, but I didn't get any population
analisys. It seems that one should also add nbo or panbo,
i.e. one should insert

$pop nbo

or

$pop panbo

After this I got population analysis. However when I compared
the results of nbo and panbo they turned out to be same.
Should it be so?

Best regards,
Evgeniy

[christof.haettig]

Dear Evgeniy,

a) as written in the manual, just $pop without any options requests a Mulliken population analysis, which is disabled in ricc2.

b) there is no 'panbo'  option for $pop, the program will only and interprete 'nbo'  and ignore 'pa'  and do an NBO analysis.
     Probably you wanted to use '$pop paboon'...
    I suggest to you define to set up the input file. It will check for such misspellings.

Christof

[evgeniy]

Dear Christof,

Thanks for your instructions. 'panbo' was a typo and the program
apparently took it as 'nbo'. When I corrected it everything got fine.

Just one more short remark, 'ricc2 -proper' somehow gives an error
in the case of the ground state density (here is the last part of the output):

....
 Now analysing ground state density on file "gsdens"

 ========================
  internal module stack:
 ------------------------
    ricc2
 ========================

 Error in fanal (1)
 ricc2 ended abnormally
 ricc2 ended abnormally
program stopped.


However if the complete calculation of gradients is done
(I mean if I re-run the calcultion of gradients by starting ricc2 and
'$pop nbo' is present in the input file) the population anylysis (nbo)
does work.

Best regards,
Evgeniy