Author Topic: Problem with mpsts-noa2 in aoforce  (Read 1529 times)

JakubV

  • Full Member
  • ***
  • Posts: 39
  • Karma: +0/-0
Problem with mpsts-noa2 in aoforce
« on: February 13, 2023, 01:42:42 AM »
Dear Turbomole experts,

I have tried optimisation of molecular geometry followed by vibrational frequency calculation via aoforce in TM V.7.6.
With ECP/DFT - functional mpsts-noa2, basis set def-TZVPP, molecule NUHFI.

I got an error message:
"PREPARING NUMERICAL INTEGRATION

 Remaining core memory for DFT =                   475 MB

 Memory needed per atom    =                   870 KiB

     i. e.                     5 atoms per loop

 Calculating ground state density on molecular grid
 Invalid value of nfun (                   223 ) in <mgga_r0>!

========================
 internal module stack:
------------------------
    force
========================

  abnormal termination
 force ended abnormally
"

What is wrong and how can I work that around, please?

Best regards,
JakubV

uwe

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 556
  • Karma: +0/-0
Re: Problem with mpsts-noa2 in aoforce
« Reply #1 on: February 13, 2023, 02:57:46 PM »
Hi,

local hybrid functionals cannot be used in analytic frequency calculations using aoforce. While it is mentioned in the manual, aoforce fails to find the routines for 2nd derivatives of the functional and thus stops with this rather technical error message...
So please use NumForce instead of aoforce to use numerical 2nd derivatives.


JakubV

  • Full Member
  • ***
  • Posts: 39
  • Karma: +0/-0
Re: Problem with mpsts-noa2 in aoforce
« Reply #2 on: February 18, 2023, 02:10:36 AM »
Thank you very much.

I have additional questions - how do I force Turbomole to break down total DFT energy into kinetic, nucleus-electron coulombic, electron-electron coulombic, exchange and correlation parts?

I am thinking about some very general "local double-hybrid" with MP2/MP2(F12) part.
So, I would compute local hybrid single point and then MP2 or MP2(F12) with mos from the output. And compute a convex combination of correlation energy from local hybrid and that MP2.

Is it possible to get the Local Mixing Function (LMF) printed in some cube file? As it would be interesting to plot for different molecules.