w90 input variables¶
This document lists and provides the description of the name (keywords) of the w90 input variables to be used in the input file for the abinit executable.
Test list (click to open). Rarely used, [8/998] in all abinit tests, [2/117] in abinit tutorials
In order to find the Maximally Localized Wannier Functions, the user has to provide an initial guess. A set of localized trial orbitals is chosen corresponding to some rough initial guess at the Wannier Functions, and these are projected onto the Bloch eigenstates. See [Souza2002a]. These initial projections are stored in a file .amn and the variable w90iniprj is used to construct them:
w90iniprj =1: Random projections.
w90iniprj =2: The initial projections will be a linear combination of hydrogenic atomic orbitals. The user has to define the projections in the secondary input file wannier90.win. Information about how to define them can be found in the manual of Wannier90. See www.wannier.org
Mnemonics: Wannier90- PRINT UNKp.s file
Mentioned in topic(s): topic_Wannier
Variable type: integer
Default value: 0
Comment: The default is set to zero because UNKp.s files occupy a lot of memory.
Only relevant if: prtwant == 2 or prtwant == 3
Test list (click to open). Rarely used, [8/998] in all abinit tests, [4/117] in abinit tutorials
Defines whether or not the UNKp.s file will be printed.
w90prtunk = 0: Do not print the UNKp.s files
w90prtunk = 1: Print the UNKp.s files on a fine grid
w90prtunk>1: Print the UNKp.s files on a coarse grid
Instead of printing every record we will print every w90prtunk records. This is useful to reduce the size of the UNKp.s files, but, the quality is also reduced.
These files contain the periodic part of the bloch states represented on a regular real space grid. They are indexed by k-point p (from 1 to nkpt) and spin s (‘1’ for ‘up’,‘2’ for ‘down’).
The name of the wavefunction file is assumed to have the form:
write(wfnname,200) p, spin 200 format (‘UNK’,i5.5,’.’,i1)
These file are unformatted. The first line of each file contains 5 integers: the number of grid points in each direction ( n1, n2 and n3 ), the k-point number ikpt and the total number of bands mband in the file. The following rows contain the wavefunctions in real space.
These files are written in the following way for the coarse grid:
write(iun_plot) n1/w90prtunk,n2/w90prtunk,n3/w90prtunk,ikpt,nband write(iun_plot) (((fofr(1,jj1,jj2,jj3),fofr(2,jj1,jj2,jj3),& & jj1=1,n1,w90prtunk),jj2=1,n2,w90prtunk),jj3=1,n3,w90prtunk)
Where fofr is a double precision variable which contains the wavefunctions in real space. Note that in order to reduce the size of the UNK files we are just including records in the wavefunctions for 1/(w90prtunk^3) of the grid points. That is why we divide n1, n2 and n3 by w90prtunk. The output .xsf files for plotting with XCrysDen will also be on the coarse grid. When this does not produce an acceptable plot, w90prtunk can be set to 1 to output every grid point. (You should try spline interpolation in XCrysDen first.)