Recent Posts

Pages: 1 ... 8 9 [10]
91
Escf and Egrad / Re: Keywords for CVS-BSE in Turbomole 7.9
« Last post by keweiz on April 22, 2025, 04:55:56 AM »
Dear Yannick,

Yes, including relativistic effects does have a big impact on the calculated energy and oscillator strength. I have attached the simulated spectra - with and without x2c - for comparison. We are still working on improving the agreement with the experimental data.

Thanks,
Kewei
92
Escf and Egrad / Re: Keywords for CVS-BSE in Turbomole 7.9
« Last post by yannickf on April 18, 2025, 12:03:51 AM »
Dear Kewei,

Thanks for the reply. I do not see your control but based on your output from escf it looks like you did not include relativistic effects. For the core states, these will be rather pronounced. In Turbomole, we offer relativistic all-electron exact two-component (X2C) for this (please see sec. 6.4 of the manual for details). This is activated with $rx2c, $rlocal, and $finnuc. These keywords use scalar-relativistic X2C in its local version and apply the finite nucleus model. You also need a relativistic basis set such as x2c-TZVPall (no cbas is checked in yet, so you would have to use the jbas as cbas or use another program such as ERKALE to get a cbas auxiliary basis set or I will check if I can privately send you a preliminary cbas for the elements of interest). Even for light elements this leads to rather pronounced effects for the energetically low states: https://doi.org/10.1080/00268976.2020.1755064 (hope you got access to this journal). The X2C-GW-BSE framework was originally developed for X-ray absorption spectroscopy with damped response but it can be applied to the CVS treatment, too.

escf ignores the data groups $excitations and $core_excitations. So, ncore will have no impact. These data groups are only relevant for ricc2. Sorry for not mentioning this in my previous post.

Best wishes,

Yannick
93
Escf and Egrad / Re: Keywords for CVS-BSE in Turbomole 7.9
« Last post by keweiz on April 17, 2025, 10:32:26 PM »
Dear Yannick,

Thank you for the reply, and it works. By freezing all the occupied orbitals other than the 1s orbital of Cr, I was able to compute the core-excitation energies directly.
Code: [Select]
$freeze
a 2-54 # There are 54 occupied orbitals in total
For some reason escf thought the 1s orbital (which should be orbital 1) was orbital 54 (the last occupied orbital), but the excitation energy is correct.
Code: [Select]
      occ. orbital   energy / eV   virt. orbital     energy / eV   |coeff.|^2*100
       54 a           -5922.38          64 a               2.38       97.0
Also, it looks like the keyword "ncore" has no impact on the excitation energy, which is suppose to specify the number of core-hole. I guess this is because only the 1s core-orbital is not frozen, so any hole has to be a core-hole?

Overall the simulated spectrum is quite different from the experiment (both attached). The 1s-to-3d quadrupole transition ("pre-edge") was not captured, and the absorption above the ionization threshold ("continuum") was not simulated well. We will look into this in more detail and work on improving the simulation.

Best,
Kewei
94
Aoforce and Numforce / Re: Screwer vs Vibration
« Last post by avsfan92 on April 17, 2025, 03:42:11 PM »
Thank you, Uwe! I am grateful for all that you do.
95
Define / Re: Automatimg Define
« Last post by uwe on April 17, 2025, 12:41:10 PM »
Hello,

a good starting point is the Turbomole tutorial (not the documentation, tutorial is named tutorial_turbomole.pdf in the DOC directory), one of the first chapters is named "The s[ai]mple input".

As define is not used to generate start orbitals the molecular charge and optionally the number of
unpaired electrons (1 for doublet, 2 for triplet, etc.) can be added by using the $eht keyword:
     $eht charge=<n>
     $eht charge=<n> unpaired=<m>
Examples:
     $eht charge=-1
     $eht charge=2 unpaired=1

Please note that when starting a job directly from a short control file some tasks will be done automatically:
  • symmetry is checked and activated if found, similar to running desy in define. If you do not want to use symmetry by default, add
    $symmetry c1
    to the input control file
  • a closed shell occupation is taken by default for even number of electrons, also similar to what define is doing - to enforce unrestricted calculations, add
    $uhf
    to the control file
  • for geometry optimizations internal redundant coordinates are automatically generated
  • energy and gradient results are directly written to the control file, so when I script Turbomole, I always add $energy and $grad to avoid getting a long control file:
    $energy file=energy
    $grad file=gradient
  • to freeze orbitals add $freeze with the energy window to define which ones to freeze
    $freeze
      fpc=-3.0 fpv=50.0
    will freeze orbitals with energies below -3.0 Hartree and larger than +50.0 Hartree

Hope this helps. Please report any issues using this way to write input!

Best wishes
96
Define / Re: Automatimg Define
« Last post by Michael_Patzschke on April 17, 2025, 11:58:08 AM »
Thank you for the reply.

How would I account for different spin multiplicities in such an input?
97
Aoforce and Numforce / Re: Screwer vs Vibration
« Last post by uwe on April 17, 2025, 11:15:54 AM »
Hi,

your observation is absolutely correct: screwer is a script which simply calls vibration, it has no functionality other or different or in addition to the vibration tool - except the help text if you call screwer -help

Best Regards
98
Escf and Egrad / Re: Keywords for CVS-BSE in Turbomole 7.9
« Last post by yannickf on April 17, 2025, 01:32:27 AM »
Dear Kewei,

there is no dedicated CVS keyword for escf. There is a somewhat hidden sentence in the computational section of Christof's paper, which explains how CVS-BSE in Turbomole works: "For the CVS approximation, all non-core spinors with an energy higher than the spinor of interest were frozen and projected out". Freezing such non-core orbitals in1c or non-core spinors in 2c calculations is possible with the data group "$freeze". So, your freeze input would have to be inverted for escf, i.e. not freezing the 1s orbital of Cr but the occupied valence orbitals.

If you want to validate the CVS ansatz for your study, you could use the damped response ansatz without the need to freeze orbitals.

Hope this helps

Best,

Yannick
99
Escf and Egrad / Keywords for CVS-BSE in Turbomole 7.9
« Last post by keweiz on April 16, 2025, 09:10:16 PM »
Dear Turbomole community,

I'm trying to simulate K-edge X-ray absorption spectra of elements using Bethe-Salpeter equation with core-valence separation approximation (CVS-BSE), following the 2023 JCP paper https://doi.org/10.1063/5.0160265.

For a small molecule like Cr(CO)6, I could use BSE to compute all the possible excitations and see the 1s-to-valence excitations close to Cr K-edge (5989 eV, please see figure attached). The calculation was performed following section 14.2.3 of Turbomole 7.9 manual "General recipe for a BSE calculation" (ridft - rigw - bse).

I then tried to call the CVS approximation following section 10.2.1 of the manual "CVS approximation for Core Spectra" (for ricc2), and added the following lines to the control file of the BSE calculation:
Code: [Select]
$freeze
 a 1    # 1s orbital of Cr
$core_excitations
 a 1
$excitations
irrep=a multiplicity=1 nexc=8000 npre=8000 nstart=8000 ncore=1   # one core-hole
spectrum states=all operators=diplen,qudlenIn this case, a smaller number of excitations were calculated, and the 1s-to-valence excitations were not computed. It seems that when 1s electrons of Cr were frozen, they were not considered for excitation.

Are these keywords correct for CVS-BSE? I have attached the escf.out files for both calculations. Any suggestion would be greatly appreciated.

Best,
Kewei
100
Define / Re: Automatimg Define
« Last post by uwe on April 16, 2025, 04:44:36 PM »
Hi,

why do you want to use define if you want to create automated inputs?

Instead of creating a define input and piping it into define, it is much easier to write a simple control file which is independent of your molecular structure. Like this one:

Code: [Select]
$coord file=coord
$atoms
  basis = def2-TZVP
$dft
  functional pbe
$rij
$marij
$cosmo
$cosmo_out file=out.cosmo
$end
This is for a RI-DFT input using COSMO with default settings. Place the coord file to the same directory and call jobex or ridft for a geometry optimization or a single-point energy calculation.

Or do you plan to use some specialties which only define is able to do?

Best Regards, Uwe
Pages: 1 ... 8 9 [10]