Re: A weird error in FEFF 8.1

[Matt Newville <newville@cars.uchicago.edu>]

On Mon, 7 Oct 2002, Jeff Terry wrote:

> I believe that you should already have the LDOS written to disk by
> this point.

Alex responded:

> As Jeff Terry wrote, you should already have files with LDOS
> at this point. You can skip the rest of calculations
> by using smaller RMAX (e.g. RMAX 0.1).

That's an interesting point.  It might be interpreted to mean that
it is OK for feff to crash if the desired calculation is finished.
I hope that's not actually the case.

> > On Monday, October 7, 2002, at 03:52 AM, Angel Garcia-Adeva
> > wrote:
> >
> > I'm trying to calculate some local DOS with FEFF. I needed more
> > potential sites, so I modified nphx to take a value of 16.
> > Also, I had to modify the maximum number of atoms in the
> > pathfinder to 2000. The LDOS calculation goes well up to 10
> > unique potentials.  However, when the number of unique potentials
> > is 11, FEFF crashes after doing the FMS part of the code. The
> > text of the error is as follows:

AA> The error message you have is typical when the pathfinder module
AA> finds too many paths. To reduce number of found paths you can use
AA> RMAX, NLEG, and PCRITERIA cards.(see manual for details).

I believe the problem may NOT be that the pathfinder has too many
paths, but is instead related to the increase value of nphx.

Not only does the parameter 'nphx' need to be set greater than the
default value of 7, but that the parameter 'nphx1' also needs to be
set to the same value.  I don't know what version of the code you're
using, but in the broken apart source code, nphx is set in
HEADERS/dim.h and nphx1 is set in DEBYE/dwpar.h.  Both are set to '7'
by default, they are not tied together (as, perhaps, they should be).


It might help if you could send an input file that causes the crash.

Hope that helps,

--Matt