The ABINIT Users Mailing List ( CLOSED )

["Eric J. Walter" <ewalter@particle.physics.wm.edu>]

Hi, 

This seems like your ecut is not quite high enough.  How well
converged is ecut?  How much does the total energy & pressure change
(ecutsm=0.0) when you try ecut=55 or 60 Ry?  Keep in mind that the
stress is more sensitive to ecut being unconverged.

Here is some of my own data for fcc-Cu using a 6x6x6 grid and
different ecut and ecutsm: 

ecut(Ha)  ecutsm(Ha)     Etot(Ha)    Pressure(GPa)
15.0       0.0        -41,104168412  -19.071
15.0       0.5        -40.981481465   14.498

22.5       0.0        -42.161574462  -17.578
22.5       0.5        -42.157585590   24.134
                       109 meV error  41.7 GPa error

24.0       0.0        -42.167377345   12.237
24.0       0.5        -42.166998665   22.880
                        10 meV error   10.6 GPa error

25.0       0.0        -42.168472697   19.968
25.0       0.5        -42.168329071   23.113
                         4 meV error    3.1 GPa error

27.5       0.0        -42.168781587   23.357
27.5       0.5        -42.168781405   23.368

35.0       0.0        -42.168890359   23.338
35.0       0.5        -42.168889907   23.361


I think this shows that ecut needs to be well converged before
ecutsm=0.0 and 0.5 will yield close results.

Hope this helps, 

Eric 


On Tue, Oct 04, 2005 at 04:54:44PM -0400, Nichols A. Romero wrote:
> ---------- Forwarded message ----------
> From: Nichols A. Romero <naromero@gmail.com>
> Date: Oct 4, 2005 4:50 PM
> Subject: ecutsm parameter
> To: forum@abinit.org
> 
> Hi,
>  I have been doing some benchmark comparisons between PWSCF and ABINIT with
> identical pseudopotentials generated by OPIUM on a simple test case:
> diamond.
> I was doing simple SCF, with identical kpts, ecut, lattice constant, etc.
>  The two did not agree with one another until I set ecutsm = 0.0. I had
> accidentally left is set to 0.5 Ha from a previous relaxation calculation.
>  Now, I understand that ecutsm is primarily used for variable cell
> relaxation. But on these simple scf calculations, it caused the pressure to
> differ by 30 GPa (quite a bit).
>  I would not have expected such a difference. Is this normal? I am using
> relatively hard norm-conserving PP (r_c=1.1) of the Troullier-Martins type
> with Ecut = 50 Ry.
>  I would appreciate any comments.
>  Thanks,
> --
> Nichols A. Romero, PhD.
> 1613 Denise Dr. Apt. D
> Forest Hill, MD 21050
> (217) 417-5210
> 
> 
> --
> Nichols A. Romero, PhD.
> 1613 Denise Dr. Apt. D
> Forest Hill, MD 21050
> (217) 417-5210