Orbitals from Molpro for standalone MRCC

2 years 6 months ago - 2 years 6 months ago #650 by negoty
Dear users and developers,
I have the following issue: I would like to use some of MRCC's capabilities on open shell compounds but I found (by comparison to ORCA or Molpro) that the HF or UHF solutions often converge to excited states. I tried to fiddle with convergence settings and different convergers (AugHess etc.) but so far nothing helped.

Then I had the idea to use ORCA or Molpro interfaces to MRCC to compute orbitals and pass them to MRCC for post-HF stuff but the respective manuals state that only closed-shell molecules may be computed this way.

Thus my question: Is it possible to generate orbitals (e.g. with Molpro) and feed them to MRCC in standalone mode to be used in post-HF calculations? I feel like this should be possible.
Also, if this is indeed possible, could someone maybe explain how to do it?

I would greatly appreciate any help!

All the best, Benedikt
Last edit: 2 years 6 months ago by negoty.

2 years 6 months ago #651
Dear Benedict,

the interfaces are written only for the mrcc binary, not for the uccsd that you were using recently. Precisely which open shell MRCC capabilities would you like to use?

It would be better to resolve your problem with the RHF/UHF references. Could you share the details? (input, results from your output, results you compare to from other programs, etc.)

Bests,
Peter

2 years 6 months ago #652 by negoty
Dear Peter,
thank you for your quick reply. Ultimately I am interested in spin state differences in Iron complexes (both complexes are separately optimized for high spin and low spin state). This is for benchmarking purposes.
More specifically I want to compute CCSD(T) energies for different spin states of complexes like [Fe(H2O)6]2+ and [Fe(NH3)6]2+. Another point of interest is the influence of higher order CC (CCSDT or even CCSDT(Q)) which I would maybe calculate with a small basis set for the water complexes.
Obviously for these reasons I need open shell capabilities (which the interfaces to ORCA and Molpro unfortunately do not provide). In principle using MRCC in standalone mode works for me, however I noticed that especially the UHF of the high spin complexes tends to not give me the global minimum. I inferred this from results like the following:

Input
calc=ccsd(t)
mem=30000MB
scfmaxit=100
scfext=15
symm=off
scfiguess=restart

qscf=AugHessG

basis=def2-SVP

mult=5
charge=2

unit=angs
geom=xyz
19

Fe       0.000000000      0.000000000      0.000000000
O        0.000000000      0.000000000      2.132470000
H       -0.633127000      0.454504000      2.712386000
H        0.633127000     -0.454504000      2.712386000
O        0.792100000      1.961200000      0.000000000
H        1.035508000      2.485723000      0.781097000
H        1.035508000      2.485723000     -0.781097000
O       -1.994703000      0.776644000      0.000000000
H       -2.240397000      1.716560000      0.000000000
H       -2.823416000      0.269539000      0.000000000
O        0.000000000      0.000000000     -2.132470000
H        0.633127000     -0.454504000     -2.712386000
H       -0.633127000      0.454504000     -2.712386000
O       -0.792100000     -1.961200000      0.000000000
H       -1.035508000     -2.485723000     -0.781097000
H       -1.035508000     -2.485723000      0.781097000
O        1.994703000     -0.776644000      0.000000000
H        2.240397000     -1.716560000      0.000000000
H        2.823416000     -0.269539000      0.000000000

This was not the first run so I restarted from previous orbitals and tried to converge with a second order method. I also turned off symmetry to exclude problems from that side.

Result:
***FINAL HARTREE-FOCK ENERGY: -1717.7207438301527418 [AU]

ORCA gives me:
Total Energy : -1717.75642671 Eh

With UHF being variational I judged that MRCC didn't converge to the state I wanted.

Another example input where this happened:
calc=CCSD(T)
scf=UHF
mem=10000MB
mult=5
charge=2
basis=cc-pVDZ

unit=angs
geom=xyz
25

Fe       0.000053000      0.002510000     -0.003458000
N       -0.000808000      2.265165000     -0.005096000
H        0.370540000      2.712669000     -0.850948000
H        0.560002000      2.663165000      0.757037000
N       -2.260767000     -0.044908000     -0.074260000
H       -2.678228000     -0.911832000     -0.431727000
H       -2.653801000      0.689766000     -0.673880000
N        2.259795000      0.106824000      0.011879000
H        2.722587000     -0.798208000     -0.129316000
H        2.683562000      0.482789000      0.867852000
N        0.048879000     -2.258343000      0.050912000
H       -0.779411000     -2.714113000      0.450716000
H        0.830628000     -2.624893000      0.606109000
N       -0.016020000     -0.047124000      2.257943000
H       -0.143924000      0.872071000      2.696039000
H        0.826085000     -0.430182000      2.702644000
N       -0.032058000     -0.003129000     -2.266047000
H       -0.595977000     -0.768497000     -2.653314000
H        0.891529000     -0.126640000     -2.696115000
H       -0.413109000      0.839969000     -2.710582000
H       -0.930243000      2.681826000      0.120479000
H       -0.781837000     -0.616313000      2.636474000
H       -2.711746000      0.102732000      0.835693000
H        2.633545000      0.701755000     -0.736538000
H        0.165494000     -2.696479000     -0.869766000

Result:
***FINAL HARTREE-FOCK ENERGY: -1599.3790571648860350 [AU]

ORCA:
Total Energy : -1599.37990641 Eh

This certainly isn't very much but it is noticeable in the final energy difference. Maybe it's just a feeling but I got the impression that starting from previous HF orbitals often does not work as well as starting from scratch. Possibly I've been doing it wrong all the time

All the best and thanks for helping me out!

2 years 3 months ago #677
Dear Benedict,

I am sorry for the very late reply.
It took some time to adjust the scf code and then the release was delayed...

To solve the above issues, you should update to the latest version released last weekend. The new code with qscf should give HF results for those two Fe-complexes in agreement with your posted energies.

I hope this helps. Sorry again for this late reply.
Please, let us know if you have any more issues.
Bests,
Peter