How to prepare input files for LANCZOS?

hakuna · Post by **hakuna** » Thu May 04, 2006 2:13 am

i THINK dimer and lanczos are good choice for finding saddle point. For dimer, the initial files can be prepared much like that NEB, i think Dr. graeme has presented that carefully on his website, but how to use the lanczos seems
to be indirect.
are the INCAR, KPOINTS, POTCAR, POSCAR(a good guess at the
transition state) and MODECAR(the lowest curvature mode) enough to activate the vasp run?

thanks

andri · Post by **andri** » Thu May 04, 2006 4:41 am

Yes, a saddle point run using the Lanczos method uses the same input file as usual, i.e. POTCAR, KPOINTS, POSCAR and INCAR. A MODECAR is not strictly necessary for it to work. Note that there are no 01, 02 etc. subdirectories as for NEB or dimer runs. This is more like a regular structural minimization or MD run in that sense. All that is needed is a initial guess for the saddle point (POSCAR) and the necessary flags in the INCAR to initiate the run, see http://theory.cm.utexas.edu/vtsttools/lanczos/index.php. Default values for the "Standard Parameters" should suffice to begin with.

If you have a converged MEP then the neb2lan.pl script can be used to generate your initial guess for the saddle point, along with a MODECAR, see directions on http://theory.cm.utexas.edu/vtsttools/scripts/

hakuna · Post by **hakuna** » Fri May 05, 2006 2:06 am

tahnks, andri;
then should we put the initial guess of the saddle point into POSCAR or MODECAR ? what is the difference between POSCAR and MODECAR in the LANCZOSE case?

thanks again

[quote="andri"]Yes, a saddle point run using the Lanczos method uses the same input file as usual, i.e. POTCAR, KPOINTS, POSCAR and INCAR. A MODECAR is not strictly necessary for it to work. Note that there are no 01, 02 etc. subdirectories as for NEB or dimer runs. This is more like a regular structural minimization or MD run in that sense. All that is needed is a initial guess for the saddle point (POSCAR) and the necessary flags in the INCAR to initiate the run, see http://theory.cm.utexas.edu/vtsttools/lanczos/index.php. Default values for the "Standard Parameters" should suffice to begin with.

If you have a converged MEP then the neb2lan.pl script can be used to generate your initial guess for the saddle point, along with a MODECAR, see directions on http://theory.cm.utexas.edu/vtsttools/scripts/[/quote]

andri · Post by **andri** » Fri May 05, 2006 2:23 am

The configuration is always in the POSCAR. The MODECAR contains the unstable mode, or the current guess for it. It will be generated at the end of each Lanczos step and if one is in place when a run is initiated then it will be read and the mode used as an initial guess for the Lanczos procedure instead of a random mode, see for instance JCP 121 p. 9776 (2004).