Problem with periodic electric field

[Gabriele Tocci]

Dear All,

There have been some discussions about electric fields recently and a 
possible bug has been spotted on the calculation of the forces in the
qs_efield_utils.F routine concerning fields applied for isolated molecules. 
Here I post another issue that may be related to those already discussed 
concerning periodic fields.

While running a DFT-MD of a liquid water-solid interface under the presence 
of a periodic electric field of 0.5 V/A along +x, I noticed that the waters 
align their
dipoles in the opposite direction to the field (that is -x), whereas they 
should align their dipoles in the same verse of the field. Also the energy 
(i.e. the conserved quantity in the NVT ensemble using the CSVR thermostat) 
shows a large drift, of about 0.15 eV/(ps*atom), while the same simulation 
without the field gives 7*10^(-5) eV/(ps*atom).

Here I report some tests on the geometry optimizations of a water molecule 
with and without the electric field (using PERIODIC_EFIELD) and I found 
that in the presence of the field the water molecule in optimized 
configuration has its dipole pointing towards -x instead of towards +x. Not 
surprisingly when running a simulation with the field towards -x the 
molecule the reorients with the dipole pointing towards +x...

*water in no field:*
ENERGY= -17.222255518668135 a.u.

*water in periodic field using the section:*

    &PERIODIC_EFIELD
     INTENSITY 0.0097234 #in au 1 au=51.42207 V/A
     POLARISATION 1.0 0.0 0.0
    &END

ENERGY=-17.229857100831254 a.u.

I attach the input, coordinates during optimization and the output files 
for the cases above. 

I checked in the qs_efield_berry.F file in case the forces on the ions may 
have a wrong sign as suggested in the post https://groups.google.com/forum/#!searchin/cp2k/electric$20field|sort:date/cp2k/kH_a6k21FLQ/IZAphZhKtkMJ<https://groups.google.com/forum/#%21searchin/cp2k/electric$20field%7Csort:date/cp2k/kH_a6k21FLQ/IZAphZhKtkMJI>concerning the routines for the EFIELD section, but I could not identify a 
similar term in the periodic field case (the line where the forces are 
computed looks quite different in that file).

To sum up I have a few questions concerning the periodic field using the 
berry phase.
Why the water dipoles are oriented in the opposite direction of the field? 
Why is the energy drift so big for the liquid water-solid interface MD 
compared to the one where there is no field?
Is it possible that there is a bug in the forces computed in the case of 
the periodic field? I hope someone can shed some light on this.

Thanks a lot,
Gabriele
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20131109/bbf3f1af/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: H2O-nofield-pos-1.xyz
Type: chemical/x-xyz
Size: 9272 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20131109/bbf3f1af/attachment.xyz>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: inp_nofield.inp
Type: chemical/x-gamess-input
Size: 1889 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20131109/bbf3f1af/attachment.inp>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: out_nofield.out
Type: application/octet-stream
Size: 646230 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20131109/bbf3f1af/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: H2O-periodic_field-pos-1.xyz
Type: chemical/x-xyz
Size: 13176 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20131109/bbf3f1af/attachment-0001.xyz>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: inp_periodic_efield.inp
Type: chemical/x-gamess-input
Size: 1892 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20131109/bbf3f1af/attachment-0001.inp>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: out_periodic_efield.out
Type: application/octet-stream
Size: 1012454 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20131109/bbf3f1af/attachment-0001.obj>