No subject

L is 5.41 eV - this difference between the experiment and thoery can be
explained by GGA error. A bit improvement (about 0.1 eV) can be done by
including SO.

Now the difference between the bulk and slab is about 0.1 eV. I think its
reasonable. I was using GGA PRB98, and 4 Ry to separate core and valence
states, I guess its fine since:

TOTAL CORE-CHARGE:                   18.00000000224257
TOTAL CORE-CHARGE INSIDE SPHERE:     17.99989730342819

Other settings were default.

Before calculation c(2x2) surface was relaxed on smaller slab. I used
2x2x1 mesh in the slab calculation. Slab was really big (supercell
with 36 atoms, it means 19 atomic layers c(2x2) terminated on both sides).
There is about 17 A of vacuum between the slabs - I guess its more than
enough. The convergence of the SCF cycle was very good.

The results for surface band structure are almost the same as the
ones I obtained with fhi98md pseudopotential code. However, the features
implemented in WIEN (LAPW2 program) allow to make identification of
surface states much more reliable and effective.

In summary I conclude that the problem with too small width of the VB is
not a key problem in this case. My goal is to compare these calculations
to my ARPES experiment - I have good agreement for surface features if I
downshift theoretical bands by 0.4 eV (to compensate the width). I also
think that it is not the width that is the most important but the shape of
surface band dispersions - once experimental dispersions agree with
theoretical ones one can say much more about the physics of given surface
feature (e.g. from which atomic orbitals its originates etc.).

Any comments welcome!

> PS: I've changed your subscription.
>
>                                       P.Blaha

Thank you Prof. Blaha!

Regards,

Lukasz