[Pw_forum] Questions on supercell calculation of MgB2

Paolo Giannozzi giannozz at nest.sns.it
Thu Sep 1 12:13:22 CEST 2005

On Monday 29 August 2005 11:40, 张 洪彬 wrote:

> I have done a calculation to check the convergence of supercell
> method by calculating a 3*3*1 supercell without atom loss, but 
> the band energy is very different with those of original unit cell

you mean "the band energies" ? and what does it mean that 
"they are different" ? of course they are: the Gamma point of 
the supercell contains several k-points of the original cell that 
are refolded into the Gamma point of the supercell.
In order to establish an exact mapping of the band energies for 
the supercell into those of the original cell, you need to compare
calculations that use the same grid to sample the respective 
Brillouin zones. Do this and you will find that the supercell has 
exactly 9 times the total energy of the original cell and that all
eigenvalues at Gamma of the supercell can be mapped into
eigenvalues at Gamma + refolded k-points of the original cell.

> Further more, when I want to plot the band structure using
> bands.x to get the .dat file, there is an error like:
> [...]
>      from punch_band : error #         5
>       increase maxdeg

the degeneracy of energy bands in a supercell may be quite high. 
If you really want to plot bands in a supercell, do what it says and

> And at the beggining of the scf calculation procedure, there are
> "additional transition found"

good: this means that the identity plus a fractional translation 
is a symmetry operation, i.e. that your cell is a supercell. In a
supercell all translation vectors that are lattice vectors of the 
original cell but not of the supercell will appear as fractional


Paolo Giannozzi             e-mail:  giannozz at nest.sns.it
Scuola Normale Superiore    Phone:   +39/050-509876, Fax:-563513 
Piazza dei Cavalieri 7      I-56126 Pisa, Italy

More information about the users mailing list