[Wien] K-mesh compatibility

[Gus Hart]

There are two parts to your question: (1) how one does equivalent k-point 
meshes _generally_ and (2) how does one do it in WIEN2K. The first part is 
pretty straight forward. E.g., see Froyen's 1989 paper: PRB 39, 3168. 
Unfortunately, the answer to (2) is tricky as each lattice type is handled 
differently in WIEN2K. When calculating the formation enthalpy of a binary 
structure A_mB_n, sometimes (when I couldn't figure out how to translate 
the straightforward approach of Froyen to WIEN format), I was forced to do 
three supercell calculations: A_mA_n, B_mB_n, and the alloy A_mB_n. This 
can be wasteful but at least it eliminates k-mesh inconsistencies in the 
total energies.


At 06:43 PM 7/13/2004 +0200, you wrote:
>Dear All,
>
>
>My question is about k-mesh compatibility. I am doing slab calculations of 
>H on Pd(111) surface using different unit cells. Since some H 
>configurations can be calculated in different cells, I can check that the 
>different unit cells lead to the same adsorption energies. On that 
>respect, it is very important to use equivalent k-meshes in all unit 
>cells. For example:
>
>1x1 hexagonal cell ..................  12x12x1 K-mesh
>2x2 hexagonal cell ..................  6x6x1 K-mesh
>3x3 hexagonal cell ..................  4x4x1 K-mesh
>2x3 hexagonal cell ................... 6x4x1 K-mesh
>
>The former meshes would give exactly equivalent irreducible k-points in 
>all cells, independently of the cell size. But what about if I want to use 
>a sqrt(3)xsqrt(3) unit cell,  or rectangular unit cells?
>They are no longer integer multiples of the 1x1 cell, so that the k-meshes 
>are not exactly compatible. When doing calculations on such cells 
>(sqrt(3)xsqrt(3) or rectangular) I get adsorption energy differences for 
>the same H configurations in the order of 10-15 meV. Such differences are 
>very small, but for system like mine involving H, they are unacceptable.
>
>Is there a way to produce exactly compatible k-meshes in different unit 
>cells (like 3x3 and sqrt(3)xsqrt(3) for example)?.
>
>Please, notice that this problem has nothing to do with k-point 
>convergence (as long as one is working in the convergence region). From 
>compatible k-meshes in different cells, I would expect adsorption energy 
>differences within 1 meV for the same configurations (not 10-15 meV).

++++++++++++++++++++++++++++
Gus Hart   (Rm. 312)
Materials Simulation Group
Department of Physics and Astronomy
PO Box 6010
Flagstaff AZ 86011-6010
tel. (928)523-0426
fax (928)523-1371
gus.hart at nau.edu
http://www.physics.nau.edu/~hart