[Pw_forum] Na phonon issues

[Paolo Giannozzi]

On May 17, 2007, at 16:30 , Miguel Martínez Canales wrote:

> I've been doing some phonon calculations on Na lately, and I've got
> an issue with the Na.pw91-n-mt.UPF pseudopotential [...]
>      omega( 1) =      -0.102259 [THz] =      -3.411026 [cm-1]
>      omega( 2) =       0.766410 [THz] =      25.564857 [cm-1]
>      omega( 3) =       1.171147 [THz] =      39.065511 [cm-1]
> [...]
> If I use the LDA pseudo Na.pz-n-vbc.UPF near equilibrium  
> (a=7.738a_0) with ecut=30 Ry and 18 18 18 shifted M-P grid, I get  
> the following omegas:
>      omega( 1) =       0.273215 [THz] =       9.113532 [cm-1]
>      omega( 2) =       0.835358 [THz] =      27.864728 [cm-1]
>      omega( 3) =       1.264994 [THz] =      42.195940 [cm-1]
>
> While the Na.pbe-sp-van_ak.UPF pseudo, with the same grid as the  
> enclosed input (20 20 20 shifted M-P), ecut=35 and ecutrho=300 Ry,  
> I get also sensible results (with 7.932 a_0 as lattice parametre):
>      omega( 1) =       0.390200 [THz] =      13.015753 [cm-1]
>      omega( 2) =       0.857166 [THz] =      28.592182 [cm-1]
>      omega( 3) =       1.238921 [THz] =      41.326237 [cm-1]
>
> So I have the following question: Is the mechanical instability I  
> see a product of my input or a product of the pseudopotential?

it might be the usual problem of the loss of translational invariance
at q=0, whose influence extends to small q as well; or it might be
related to the rather nasty numerical behaviour of the PW91 functional.
I would try the pw91 pseudopotential with lda (the pp file must be  
edited),
then the lda pp with pw91 functional, just to see what happens...

Paolo
---
Paolo Giannozzi, Democritos and University of Udine, Italy