The ABINIT Users Mailing List ( CLOSED )

[Thomas Maxisch <Thomas.Maxisch@epfl.ch>]

Dear Michel,

Thanks for the quick answer.

> I recently did some calculations for GaAs and I had to generated a Ga 
> pseudo with 3d electrons. I copy below my notes on this problem. I can 
> send you the pseudo to your personal address, I don't want to fill 
> everyone mailbox unnecessary.

Thanks a lot. A stupid question: With 3d electrons as core or as valence 
electrons. This is not clear for me with your description. I need them as 
valence in order to have reasonable results for GaN calculations.

The ghost states of Ga are known and there are different ways to overcome 
them. Our way to do that is to include the empty f orbitals into the 
pseudopotential and make them local. The resulting potential is very good 
in the sense of electronic properties. Nevertheless the lattice constant is 
also not as good as it should be (10.435 a.u.) Unfortunately up to my 
knowledge the only program who converges for this configuration is the 
program of J.L. Martins. And I haven't had the time until now to write a 
routine to make them readable by abinit.

But I have not yet tested this configuration with the fhi98pp code. I thing 
that it is worth to try it.
I think that I will try something in this direction the next days with the 
fhi98pp.
Why have you also modified the As potential?

When the 3d are as valence electrons in your potential, it woul be very 
nice if you can send me the Ga potential to my personal address.

Thanks a lot,
Thomas

----------------

> lattice parameter for GaAs
> --------------------------
>
> 31ga.pspnc                TM pseudo (s,p, local=p) (from the Abinit web page)
> 31-Ga:tm.fhi              TM pseudo (s,p,d, local=p) (generated by me with 
> fhi98pp)
> 31-Ga:tm_with_core.fhi    TM pseudo (s,p,d, local=p, with partial core) 
> (generated by me with fhi98pp)
> 33as.pspnc                TM pseudo (s,p) (from the Abinit web page)
> 33-As:tm.fhi              TM pseudo (s,p,d, local=p) (generated by me with 
> fhi98pp)
>
> GaAs_tm_from_Abinit_web_page.out         Optimazation of the lattice 
> parameter with pseudo 31ga.pspnc and 33
> as.pspnc
> GaAs_fhi_tm_Ga_with_core_As_no_core.out  Optimazation of the lattice 
> parameter with pseudo 31-Ga:tm_with_cor
> e.fhi and 33-As:tm.fhi
>
> When I use the TM pseudopotentials of the Abinit web page for Ga and
> As (31ga.pspnc, 33as.pspnc), I had a large error in the lattice
> parameter (opt at 10.29726 a.u. compare with exp at 10.677 (Kittel),
> 3.6% error). I then notice that these peudo did not have a d
> component. I generated new pseudopotentials with the code fhi98pp
> including the d component (pseudo 31-Ga:tm.fhi and 33-As:tm.fhi). I
> had some problems (ghost state) when I use the local d for ga so i put
> both local at p. This improved the lattice to about 10.45. I then
> included partial core charge for the Ga (31-Ga:tm_with_core.fhi). With
> this pseudo and 33-As:tm.fhi for the As, the optizimed lattice is
> 10.56888 a.u. so the error now is 1.01% which is acceptable.


--
__________________________________________________________________________
Thomas Maxisch                          | Tel  : +41 (0) 21 693 45 16
EPFL FSB-ITP                            | Fax  : +41 (0) 21 693 44 44
PHB-Ecublens, CH-1015 Lausanne          | Email : Thomas.Maxisch@epfl.ch
__________________________________________________________________________