Starting mos for supermolecule

[Robin1991]

Hey all,

I want to check if I can improve the convergence of my supermolecule dscf by starting from the combination of monomer orbitals (instead of the eht).
Is there any possibility to do this in define (or by hand by adjusting mos file)?

thanks in advance,
Robin

[uwe]

Hey Robin,

combining already converged molecular orbitals from subsystems as input for the supersystem is one of the steps that are done in EDA (energy decomposition analysis) calculations. Search the manual for EDA to get more information.

You can use the same tool as for EDA: promowa. It's one of the many Turbomole tools (a binary in this case).

How to use:

• calculate all subsystems individually, up to 10 are supported
  
  
• create a new directory and copy all the subsystems as subdirectories, for example:
  
  supersystem> ls *
  subsystem1:
  basis  control  coord  mos
  
  subsystem2:
  basis  control  coord  mos
  
  
• In the supersystem folder, create a coord file with all the atoms of the subsystems, in the same order as they are defined there. That can also be scripted, for example on the shell:
  
  supersystem>for i in subsystem*
  do
  cd $i
  sdg -b coord
  cd ..
  done > coord
  then you have to add a $coord as first line and $end as last line to the newly created coord file. Or just copy and paste with your preferred editor.
  
  
• run define as usual. Make sure to use the same basis set as for the subsystems. You could skip the eht part (just enter * in that menu), because the start orbitals will be created later on by the promowa tool - but that will also skip writing $scfmo file=mos to the control file. In that case the later promowa run will add the orbitals directly to the control file instead of an own mos or alpha/beta file as define usually does it. Both works, it is just a question of personal preference - I do not like huge control files... so I'd run eht nevertheless.
  
  
• add the keywords as for EDA to the control file:
  
  $subsystems
    molecule#1 file=subsystem1/control
    molecule#2 file=subsystem2/control
  
  and so on if you have more than two subsystems.
  
  Hint: If the task is not to do a decomposition analysis but just to get start orbitals, ignore all that is written about $scfdenapproxl etc. in the EDA chapter of the manual. And add the option opro to the $subsystems keyword to let promowa generate orthogonal orbitals. SCF convergence will be faster.
  
  $subsystems opro
   ...
  
  
• Now just run promowa and it will generate new orbitals by combining the molecular orbitals of the subsystems.
  
  
• I'd remove the $subsystems keyword before continuing, otherwise ridft will try to do an EDA job after energy has converged: kdg subsystems