[Wien] Emax !

[Peter Blaha]

> Recently I calculated total energy of simple cubic Bi. For specified
> parameters, I calculated two lattice parametres with increasing Emax in
> *.in2 (see table below). I do not understand why in scalar relativistic
> calculation the total energy so strongly depend on it? I thought it should
> be pretty constant ?
>
> Emax	Ax=5.9 a.u.	Ax=6.4a.u.	Diff5.9-6.4
>
> 1.5	-43163.107635	-43163.115133	0.007498
> 2	-43163.107869	-43163.115130	0.007261
> 2.5	-43163.107556	-43163.115129	0.007573
> 3	-43163.107855	-43163.115027	0.007172
> 4	Not conv.	-43163.114548
> 6	-43163.840817	-43163.115274	-0.725543
> 8	-43163.107565	-43163.115275	0.007710
> 10	-43163.107560	-43163.115267	0.007707
> 12	-43163.107551	-43163.115273	0.007722
> 15	-43163.107559	-43163.115271	0.007712
> 20	-43163.107553	-43163.115266	0.007713

E-tot of a scalar relativistic calc. must not depend on Emax.

Of course, Emax must be large enough, so that it does not interfere with
your Fermi energy (it should be sufficiently larger than EF, otherwise an
interpolation between occupied and unocupied states might not be possible.
Check if you have at least a few "unoccupied" bands, see eg. beginning of
case.output2)

As to your table: Did you converge the scf better than just the default ?
(-ec 0.0001; max 20 iterations) Some of those numbers are simply not converged
or not correct.

For those heavy elements EF is sometimes pretty large and readjustment of
the E-parameters (or -in1new) is strongly recommended.

Regards

                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------