[Wien] SECLR4 - Error in lapw1

[Peter Blaha]

> Just one more question. Do you think it is safe to do the volume and
> internal parameters minimization with a smaller RKMAX? Then with the
> optimized structure increase RKMAX and do one final calculation. If this
> is possible it would be great because the calculation will take much less
> time.

Such a question cannot be answered in general. It depends on the 
accuracy you want to reach.

Internal parameters: Very often even RKmax 4.5 (for sp-elements) gives 
you a reasonable internal parameters. Usually it is always good to relax 
first (min_lapw) with a rather small RKmax (depends on the types of 
atoms: 4.5 - 6) until forces are small. After this step, increase RKmax 
by one and run another min_lapw. In most cases, forces will be similar 
or only a few geometry steps with the larger RKmax are necessary.
Volume opt: This is more critical with RKMax and I'd usually start with 
RKmax 6-7 (sp vs. d-elements). Too small RKmax leads to "noisy" curves, 
and too large volume. Again, only a 2-step procedure (first with 
smaller, than with larger RKmax can give you insight and security about 
your results. Highest precision (3 significant digits after the comma in 
a) requires RKmax 9-11.

-- 
Peter Blaha
Inst.Materialchemie, TU Wien
Getreidemarkt 9
A-1060 Vienna
Austria