[Pw_forum] xspectra calculation error

[sreekar guddeti]

>
>
> Dear all,

Respected Mr. Ibrahim,

>
> ?? I am trying to run xspectra calculation on MgSe. Knowing that there is
> no gipaw pseudopotential for both contributing element on
> ? the pwscf-pseudo page, i tried to generate one for both Mg and Se. After
> running scf with this pseudo, i proceed to running the
> ? xspectra with xspectra.x but to my surprise, it keeps given this error ;
>
> ? ?? from allocate_fft : error #???????? 1
> ???? the nr"s are too small!
> Please, can anybody help me out with this?
>
>
A helpful tool(weapon :D) for tracking down errors is to use (learnt from
Prof. Gianozzi's replies on the forum)
$ grep -e '<the error>' */*.f90

In this case,,
$ grep -e 'are too small' */*.f90

Results gives  the relevant code file where it occurs.

$PW/allocate_fft.f90:     CALL errore ('allocate_fft', 'the nr"s are too
small!', 1)
$PW/allocate_fft.f90:     CALL errore ('allocate_fft', 'the nrs"s are too
small!', 1)

A look at the file PW/allocate_fft.f90 suggests that
---------------------------------
  IF (nrxx.lt.ngm) THEN
     WRITE( stdout, '(/,4x," nr1=",i4," nr2= ", i4, " nr3=",i4, &
          &" nrxx = ",i8," ngm=",i8)') nr1, nr2, nr3, nrxx, ngm
     CALL errore ('allocate_fft', 'the nr"s are too small!', 1)
---------------------------------

Here
nrxx = nr1*nr2*nr3 (Modules/griddim.f90), the minimal 3D real-space FFT
grid required to fit the G-vector sphere with G^2 <= gcut
ngm = local  number of G vectors (Modules/recvec.f90).

I have not generated PPs, but I can only infer that your FFT grid density is
not sufficient enough to accommodate the G-vectors.

Sincerely,

--
Sreekar Guddeti
Undergraduate (5th year)
Engineering Physics
IIT Bombay
India
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20110905/01befa26/attachment.html>