neighborlist issue

[Toon Verstraelen]

Hi,

I'm experiencing a problem with the simulation attached to this message.
I have reduced the input to the bare minimum that still reproduces the
problem. The force-field parameters are artificial and the model is one
of the smaller zeolites, i.e. sodalite. One can find a picture in the
zeolite atlas:

http://topaz.ethz.ch/IZA-SC/FMPro?-db=Atlas_main.fp5&-lay=web%20layout&-format=FWtopology.htm&STC=SOD&-find

The simulation is just an energy_force calculation. The error message is
this:

   SUBCELL GRID  INFO FOR THE NONBONDED NEIGHBOR LISTS

      NUMBER OF SUBCELLS             ::                   1         1
       1
      NUMBER OF PERIODIC      IMAGES ::                   2         2
       2
      NUMBER OF INTERACTING SUBCELLS ::                   2         2
       2
   WARNING| Particles:      35     19 at distance [au]:     3.03963678
less than:      3.71000000; increase EMAX_SPLINE.


   *************************************
   *** ERROR in build_neighbor_lists ***
   *************************************

   *** GEOMETRY wrong or EMAX_SPLINE too small! ***


   ===== Routine Calling Stack =====

              5 build_neighbor_lists
              4 build_fist_neighbor_lists
              3 list_control
              2 fist_force_control
              1 CP2K

   CP2K| Stopped by process number
          0
   CP2K| Abnormal program termination

Aborted


The problem is that particle 35 and 19 are actually bonded (see psf
file). I suppose this problem is caused by the fact that the bond graph
(or topology) of this system is also periodic. In most Fist regtests,
the models contain many disconnected molecules. While the systems are
periodic in these regtests, the bond graphs are not.

For as far as I can see, the part of the code that is responsible for
this issue can be found in the subroutine fist_neighbor_add in
fist_neighbor_list_types.F.

      IF (ANY(exclusion_list==atom_b).AND.(.NOT.is_tersoff_pot)) THEN
         ! Only atoms really connected (not the periodic images) need to
be excluded..
         rabc  = particle_set(atom_b)%r - particle_set(atom_a)%r
         IF (ALL(ABS(rab-rabc)<eps_default)) THEN
             added = .FALSE.
             RETURN
         END IF
      END IF

In the zeolite system, it happens that a particle is connected with an
atom in the periodic image and not in the same unit cell. When I make a
small change to the code, the simulation seems to work as expected, but
I'm not enterely sure that I'm not doing anything wrong. I'm not
familiar enough with the code to be sure about it. This is the change:

      IF (ANY(exclusion_list==atom_b).AND.(.NOT.is_tersoff_pot)) THEN
         ! Only atoms really connected (not the periodic images) need to
be excluded..
         rabc  = particle_set(atom_b)%r - particle_set(atom_a)%r
         IF (ALL(ABS(rab-rabc)<eps_default).OR.&
             (SUM(rab*rab)<SUM(rabc*rabc))) THEN
             added = .FALSE.
             RETURN
         END IF
      END IF

The idea is that exceptions on the exclusion list are reduced to the
cases where the periodic image of the atom is further away that then
atom in the same unit cell. Is this correct?


best regards,

Toon


Note: I have set R0_NB relatively high in the input file. This
guarantees that errors in the exclusion list will result in an "increase
EMAX_SPLINE" warning. Otherwise the problem would manifest itself as a
wrong energy_force calculation, which is harder to recognize.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: sod.inp
Type: chemical/x-gamess-input
Size: 2358 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20070604/a936ea3c/attachment.inp>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: sod.psf
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20070604/a936ea3c/attachment.ksh>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: sod.xyz
Type: chemical/x-xyz
Size: 1559 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20070604/a936ea3c/attachment.xyz>