[Wien] sphere size bug? lm expansion problem?

[Peter Blaha]

> I have encountered a severe problem with increasing sphere sizes to fill
> more of the interstitial (as generally recommended to make calculations
> more efficient). I am running the latest version (WIEN2k_05.3). The problem
> 
> (1) When I run with Rmt(U) = 2.3, Rmt(He) = 1.1, with RKmax = 3.5, 4.0,
> and 4.5, and Gmax = 20 or 24 there are no problems. All runs well.  
> However, when I do *nothing more* than increase Rmt(He) = 1.2 (and run
> dstart, clean, and rm *.scf to start fresh), the code crashes as 

You forgot to run    x lstart  after changing RMT. dstart makes a 
superposition of atomic densities (generated in lstart), and of course 
these densities must be on the proper radial mesh.

> (2) Since increasing sphere size requires more of the lattice harmonic
> expansion to provide an adequate representation, I tried increasing the
> maximum L in the .in2c file (for all atoms and just for He). However, when
> I added terms up to L = 8, I got the following error:
> 
> LAPW0 - Error. Check file lapw0.error.
>  'LAPW0' - lmmax1 for jatom  1 of  125 gt 124                                 
> 
> This implies that NCOM is insufficient. But my NCOM is set to 121 (the

Your analysis is correct. Simply increase NCOM in param.inc and recompile
(This may happen also in other programs).
PS: The sigma is very small anyway, so I really don't think it is 
necessary to increase the LM list in your case.

> ATPAR - Error
>  'ATPAR' - more than NGAU gaunts
>  'ATPAR' -   NGAU,   L0,   LP,   LL,    M,   MP,   MM                         
>  'ATPAR' -   2350,    4,    4,    4,   -1,    3,   -4,                        
> 
> This says that NGAU (max number of Gaunt coefficients) is insufficient.  
> But, my NGAU is 2350 (the default), which is presumably consistent with at
> least LMAX2 = 8 (otherwise, why make NCOM sufficient for LMAX2 = 10?).

Similar as above: edit the param.inc_c file and increase the value of 
ngau. You have to play around a little bit until you find a proper value. 
The program is "save" in that sense that it will not allow calculations 
when the dimensions are too small. 
PS: This is good old f77, where dynamic allocation was not possible. And 
since this is a case which happens very rarely (The problem is not just 
the L>6, but in addition there is no symmetry left, so all LM 
combinations appear), it was not converted.


                                      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/
--------------------------------------------------------------------------