[Wien] lcore stops

Fred Nastos nastos at physics.utoronto.ca
Mon Nov 8 21:29:31 CET 2004


On November 7, 2004 05:36 am, Peter Blaha wrote:
> > The Si-H bond is roughly 2.8 bohr.
> >
> > We first set the RKM to 3 and, and the H muffin tin
> > to 0.6 bohr, and all the Si tin radii to 2.2.  The
> > run_lapw script ran ok.
> >
> > We increased RKM to 4, but this increased gmin too
> > much (as expected), so we tried increasing the H
>
> gmin ? You probably mean NMAT ? (the matrix size in lapw1c)
> which really determine the size (time and memory) of the calculation.
> GMAX should be large anyway and not close to GMIN.

Right.  I really should have said NMAT.  We did all the calculations
(for different RKM values) as different cases, in different directories
and I was using the GMIN as a measure for the basis size required.
I know GMIN isn't the best measure.  In short, I wanted to try a
calculation with RKM =4, so I had to increase GMAX a lot.  But this
wasn't working, since lapw1 was limited to nmatmax, so I tried changing
the muffin-tin radii.

> RKM=4 for your radii means that the effective RKM for Si is:
> 5 / 0.8 * 1.97 , which is close to RKM=12 !
> I would expect this comes close to linear dependency! leading to
> ghostbands,....

Right.  Sorry.  Of course.  The RKMax specified applies only to the
smallest atom.

> I'd expect your core-error is due to a problem already in the previous
> iteration or in mixing. (forgot save_lapw or old *broyd* files ?; how big
> is :DIS ?, Are the eigenstates and partial charges reasonable ?...

Well, DIS seemed well bahaved, and in the calculations that converged
the eigenstates (from the resulting bandstructures) were fine.  Just this
one case was strange.  We are rerunning it now.

> > This last atom that it stops at is the Si atom that
> > binds to H atom, so it has a smaller RMT (of 1.97) than
> > the other Si atoms (2.2 bohr).  Output was given for
>
> You should never run a case where identical atoms (Si) have different
> radii !

I don't understand why?  The Si atoms are inequivalent.  I only
chose a smaller one for the Si atom near the hydrogen-terminated
surface.  Is it because of some basis-function mismatching problem
at the muffin-tin surface between the two Si atoms?

Thank you for your time.




More information about the Wien mailing list