[Wien] Re : Supercell problem of Al2O3

[B. Yanchitsky]

swati chaudhury wrote:
> I have generated ( 2 2 1 ) supercell with vacuum slab 4 bohr .At last, 
> scf of the supercell is converged nicely by reducing Rkmax to 5.00.Space 
> group of the supercell is !67 R-3c.
> But one more query,is the (2 2 1 )type supercell treatment (for Al2O3) 
> correct or not ?

I don't know whether that correct or not. It's somewhat strange, the original
space group of small cell (10 atoms) was 167, next you transformed the cell into
hex. setting with triple of the cells (30 atoms), next you created 2x2x1 supercell
(120 atoms) and added some vacuum, and the space group is still 167.
Do you have 40 atoms in your supercell that is being used for calculations?
40 comes from 120/3=40.

> The above same calculation with Rkmax 6.00,dstart is completed without 
> any problem.But scf is ended after lapw0 in 1st cycle with an error 
> message given below:
> forrtl:severe (174) : SIGSEGV,segmentation fault occured.
> Stack trace terminated abnormally.

This may be related to the stack size, Intel fortran compiler needs for a large stack.
Check for the stack size (for csh  shell type "limit stacksize" to know size of the stack,
and set in .cshrc
limit stacksize unlimited

for bash shell this is equivalent to "ulimit -s" and
ulimit -s unlimited
in .profile

> Added information: ( 2 2 2 ) supercell calculation is terminated in 
> dstart step becoz of insufficient virtual memory problem. Total number 
> of atoms in this supercell is 240.

With 100 plane waves per atom, size of hamiltonian matrix is ~24000x24000,
as double float is 8 bytes, this only matrix would occupy near 4.6Gb of memory.

Regards,

Bogdan Yanchitsky

Institute of Magnetism
Vernadsky Blvd., 36-b
03142  Kiev
UKRAINE

Tel. (+380-44) 444 34 20
Fax. (+380-44) 444 10 20