[Wien] Once again about Ni non-magnetic bcc calculation

[Dominik Legut]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Once again about Ni non-magnetic bcc calculation

On October 26, we wrote about the discontinuity of calculated total energy
as a function of volume for the bcc structure of non-magnetic Ni within the
GGA.

We experienced a total energy jump (discontinuity of total energy) of about
4mRy/atom at V/Vexp=0.90, see  table below, in the energy-volume curve. The
following parameters were used in our calculation:



RMT       1.9 - 2.2

Rkmax     8 - 12

Gmax      10 - 30

k-points  2000 - 8000



The results are as follows (Vexp corresponds to the fcc lattice constant of
6.651835 a.u. for fcc):



Table of total energy vs. volume (RMT = 2.0, Rkmax = 9.0, Gmax = 16, 5000
k-points):

V/Vexp      Etot

0.70        0.094890

0.75        0.059043

0.80        0.034713

0.85        0.019109

0.88        0.011881

                        <<-- discontinuity

0.90        0.005793

0.93        0.002596

0.96        0.000733

0.99        0.000013

1.00        0.000000

1.01        0.000023

1.03        0.000901

1.04        0.000795

1.07        0.002354

1.10        0.004393





We performed tests where we changed the Gmax parameter to 14,16,18,24 and
30. However we have not succeeded to suppress the discontinuity by
increasing the Gmax, as proposed by A. Postnikov in his letter:



> Some discontinuity might occur as the number of vectors in the

> G-summation experiences jumps at some volumes, as you keep GMAX fixed,

> but vary the volume - and hence reciprocal lattice constant.

> The effect must decrease as you take larger GMAX.

> How large - depends.

> GMAX=12 is anyway too small for a reliable GGA calculation.

> Fine G-mesh is less essential for LDA, that's why you don't see

> problems there.



The position of discontinuity in the total energy occurs at certain volume
that depends  on the values of RMT and RKmax parameters, nevertheless the
value of Gmax and number of KPoints do not affect the position of
discontinuity.

Does anybody know more details about why is this happening? We do not think
increasing Gmax over the value of 30 will help.



Thanking you in advance for any constructive comments,

Sincerely,

Dominik Legut, Martin Zeleny



-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (FreeBSD)

iD8DBQFBpEE//UOrhtVGHWgRAo0JAJ9YtIJdE/YvRMWq/bNEkpjqQNZolgCeOtmW
Pv1oUWavk6fYsYYe4pDVGa0=
=QYET
-----END PGP SIGNATURE-----