Package 2Doptimize for tetragonal or hexagonal structures

by **Algerien1970** Fri 29 May - 18:41

From the userguide :

This package performs a convenient 2D structure optimization (Volume and c/a, i.e. tetragonal or hexagonal spacegroups). After initialization of a case, one generates a set of structures and a job-file 2Doptimize.job using the command

set2D_lapw

This calls setup2D and you have to specify the ``order of the fit'' (stored in .fordfitcoa) and the changes in volume and c/a. The resulting 2Doptimize.job script should be adapted (eg. use min_lapw instead of run_lapw; insert switches,...) and executed. Finally

ana2D_lapw

can be executed and will analyze the results. It uses a set of case.Vconst* files (produced by 2Doptimize.job and stored also in subdirectory Vconst) and the numbvcoa file. You can modify the order of the fits in .fordfitcoa and rerun ana2D_lapw to check the sensitivity of the results with the fitting. Note: Fits of high order (and few ``data points'') may lead to artificial results due to unphysical oszillations of the fit.

You can see results for

- energy vs. c/a for each volume,
- energy vs. volume (with optimized c/a) and
- c/a vs. volume.

Optionally you can specify more cases by rerunning set2D_lapw. Specify also your ``old'' volume and c/a points again (or leave them out on purpose in case they were very bad (eg. very far from the minimum). The old results will then be taken automatically into account without recalculation (unless you modify 2Doptimize.job, see the comments at the top of this file). Thus a ``good'' strategy is to use only 3x3 points (order of fit = 3) and in a second step you add points where they are needed.

When you want to rerun such an optimization with different parameters (RKmax, k-mesh, XC-potentials) modify the top of 2Doptimize.job and set answscf=no and a new savename (eg. "_pbe_rk8_1000k").