[Wien] suppressing WF calc in lapw1 and FER calc in lapw2

[Jorissen Kevin]

Hello Martin,
about your second problem : it's logical that lapw2 can't find enough states to accommodate all your electrons - it used to take all states from about -7 to 0.25 Ry, and now it only gets those from 0 to 0.5 Ry.  So go ahead and decrease ne.  I've never tried to set it to zero ; if this is a problem, just try a small value (ne is allowed to be fractional).
Part of the output of lapw2 may of course be meaningless now, especially charges, forces, ...  But I think case.qtl should still be okay.
 
I can't help you with the SUPWF issue.  The UG says you also suppress the w.f. calculation, and it should be faster.
 
Kevin.
 

	-----Oorspronkelijk bericht----- 
	Van: Martin Gelfand [mailto:gelfand at lamar.ColoState.EDU] 
	Verzonden: di 2/3/2004 7:38 
	Aan: WIEN2k usergroup 
	CC: 
	Onderwerp: [Wien] suppressing WF calc in lapw1 and FER calc in lapw2
	
	

	Greetings all.
	
	I have tried to do two things just slightly different from
	"routine" use of the Wien2k code and have been left puzzled.
	
	First, after doing a structure refinement, I wanted to obtain
	the band dispersions on a dense set of k points, which
	I did by appending the list of k points to case.in1, and
	running lapw1.  The calculation took several days, and then
	crashed just at the end, when it was trying to write a file
	larger than 2G.  Uggh.  Since I wasn't interested in the
	vectors, just the eigenvalues, I changed the output switch
	in case.in1 to SUPWF.  This time lapw1 didn't crash, but
	it was no faster than the previous run.  I had expected
	it to be considerably faster calculating eigenvalues only,
	but when I looked through the code (warning: this is version 2,
	but I don't believe there have been changes here) it appeared
	that SUPWF suppressed only the printing of the eigenvectors.
	Is this correct?
	
	Second, I wanted determine the character of just the bands
	near the Fermi energy.  So, I reran lapw1, but this time
	setting Emin=0 and Emax=0.5 (the Fermi energy was about 0.25)
	and with WFFIL.  (A confession, I goofed and used WFPRI, but
	this couldn't cause problems, could it?)  Now the output files
	with the wavefunctions are under 2G.  However, setting the
	switch in case.in2 to QTL and efmod,eval to ALL, 0.5 just
	gives a
	FERMI - Error
	with
	 'FERMI' -  NOT ENOUGH EIGENVALUES FOR304. ELECTRONS
	 'FERMI' - ELECN,ELN,EMIN,INDEX304.00000133.05539 10.5000022819
	 'FERMI' -     INCREASE ENERGY WINDOW IN CASE.IN1
	 'FERMI' -  OR INCREASE PARAMETER NUME
	 'FERMI' -  OR DECREASE NE IN CASE.IN2
	in lapw2.error.
	This seems clear enough, I guess--it looks like the next step
	should be to reduce ne in case.in2.  But it's not obvious why this
	should be necessary with eval set to ALL.  And is there an
	easy way to see what I ought to set ne to?  Can I just put ne=0
	without harm?  (I'm concerned that case.qtl will depend
	on ne somehow.)
	
	Thanks!
	
	Martin Gelfand
	Dept of Physics
	Colorado State University
	
	_______________________________________________
	Wien mailing list
	Wien at zeus.theochem.tuwien.ac.at
	http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
	
	

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/ms-tnef
Size: 6750 bytes
Desc: not available
Url : http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20040205/7fcd3079/attachment.bin