[Wien] ELNES (tetra error)

[Jorissen Kevin]

Hi Giovanni,
I'll send the script to the ML, so everybody can remind me if I forget about it again ;-)
 
I always use a really full k-mesh (so not including inversion symmetry).  For H formula, the cross section is calculated using 'cross-density of states'; ie, where normal DOS is like the square of a particular lm component of the wave function, this will be the product of an lm and a l'm' component.  Such a product does not necessarily have the full symmetry that the regular DOS has.  I suspect that in many cases one can actually retain some symmetry, but finding out which ones is probably more work than just running on a full k-mesh.  Remember, this is necessary only once after you have achieved SCF convergence (since these cross terms are not necessary to calculate the electron density).
I use a mesh without inversion symmetry, but am not sure that including inversion symmetry should necessarily yield wrong results.
If Peter has any remarks to this, I'd love to hear them.
 
The tetra error you get is a little strange.  Tetra *doesn't* work differently because some symmetry is not there (unlike eg lapw1), only some arrays will become twice as big now.  Did you check with a smaller full mesh?  (tetra can get quite big if you run it with lxdos=3 for large k-meshes, though this may surprise 'regular users').
 
Sorry I don't have any other ideas for your tetra error at this time.
 
Kevin.
 

	-----Original Message----- 
	From: Bertoni Giovanni [mailto:giovanni at cemes.fr] 
	Sent: Fri 5/7/2004 12:37 PM 
	To: wien at zeus.theochem.tuwien.ac.at 
	Cc: 
	Subject: [Wien] ELNES (tetra error)
	
	

	Dear users (in particular prof Blaha and Jorissen Kevin...),
	we have discussed this topic before but I have other questions...
	
	in calculating ELNES with q orientation (formula H as Kevin suggested),
	I've usually generated with kgen a full k-mesh (with only the identity
	simmetry in case.struct) BUT adding the inversion of k-points (that
	means considering +k and NOT considering -k), i.e. a "half-full" k-mesh.
	Then I put back all the symmetries of the system in case.struct,
	isplit=99 and go with "x lapw1" etc...
	is it correct ?
	I think that for para-magnetic systems -k is always obtained from +k
	with time-reversal.
	but I don't know how the telnes routine works with H Formula.
	
	The question is: Do I need also -k in the full k-mesh ?
	
	Problem: I tried on a simple-system, everything works with the
	"half-full" mesh (adding inversion in k-points in kgen), BUT does't work
	if I do not add the inversion in kgen. I have an error in tetra:
	
	lib-4205 : UNRECOVERABLE library error
	  Unable to allocate memory.
	IOT Trap
	Abort (core dumped)
	0.0u 0.0s 0:00 20% 0+0k 0+1io 0pf+0w
	
	normally I have this error when the case.qtl file is uncompleted or not
	corrected. that can mean that tetra doesn't want the mesh without the
	inversion in k-points...
	any suggestions ?
	Thank you
	
	P.S. for Kevin
	Please, can you give me the script for calculating the ELNES/DOS for
	every atoms of the cell automatically... I have asked for it before, you
	answered "yes", but then I didn't received it...
	Thank you and bye.
	giovanni
	--
	-------------------------------------------------------
	Giovanni BERTONI
	CEMES/CNRS -  Toulouse - FRANCE
	-------------------------------------------------------
	
	_______________________________________________
	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: 6482 bytes
Desc: not available
Url : http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20040505/a8578804/attachment.bin