reference position bug during dipole moment calculation?

[Matt W]

Hello,

not a complete answer to all your points (and I've not looked at the 
files), but a couple of ideas.

(i) to get a good result for the non-periodic direction I'd suggest using a 
non-periodic setup in that direction - life is much easier then. Personally 
using the WAVELET poisson solver and XZ periodic boundary conditions has 
been reliable (i.e.make the slab normal in Y direction). The slab need to 
be centred at Y/2 in that case.

(ii) to get a dipole moment in a periodic direction you need to use the 
berry phase dipole moment.

Matt

On Thursday, July 12, 2018 at 3:33:06 PM UTC+1, Phil G. wrote:
>
> Dear all,
>
> in order to get a bulk polarization of a ferroelectric slab system (2x2x1 
> slab with 3D periodic boundary condition) I let CP2K start to calculate the 
> dipole moments of the slab system (in fact the unit cell has the hexagonal 
> shape, but it does not matter if I cut off a piece of the large system of 
> the duplicates of the unit cell in xyz coordinate system and this cut piece 
> satisfy the 3D pbc in the xyz coordinate system?). In order to check the 
> influence of the choice of a reference point on the dipole moment 
> calculation, I let start several calculations with different reference 
> points, e.g. COAC, ZERO, user_defined -> reference_point x y z.
>
> If using the reference_point ZERO, then I notice an non-explainable 
> difference between the quantity of the dipole moments for up-polarized slab 
> and down-polarized slab (the slabs are the same, but are only 180 degree 
> rotated to each other) and it seems to be the false choice for the dipole 
> moment calculation.
> If I use COAC as the reference point, then the both dipole moment values 
> are almost of the same quantity (absolute value) in z direction, but of 
> opposite sign (which is correct). If using COM as the reference point, then 
> I am wondering why not really the center of the slab was used as expected 
> (consider the slab system with lengths a_x, a_y, and c_z along the three 
> x,y,z coordinate axes, so I expect the COM to be nearly at {a_x / 2, a_y / 
> 2, c_z / 2}, but in the dipole moment file the reference point lies outside 
> of the slab system. Why? 
> If using the reference_point USER_DEFINED, REFERENCE_POINT X Y Z , in 
> which X, Y, and Z are of an arbitrary choice, for example at the center of 
> the top slab surface (X=a_x / 2 , Y=a_y / 2, Z=0) and the z=0 is exactly 
> the surface and below it there is the 2x2x1 slab system (i.e. with negative 
> c_z lying in the bottom slab surface), then after calculation I get the 
> info from dipole moment file that the y value has the opposite sign. Is 
> there a bug in the cp2k during the dipole moment calculation?
>
> After obtaining the calculated dipole moment values, I am able to 
> calculate the bulk polarization if the z-dipole moment (which is much 
> larger than in x- and y- directions) is divided by the slab volume (2x2x1 
> unit cells in the x,y,z coordinate system). Is that correct? If yes, why 
> the value is much smaller compared to the experimental value at room 
> temperature? (e.g. the experimental value of the polarization is 0.71 
> C/m^2  [APL 12 (2009), 209]  and the calculated value is 0.042 C/m^2 , i.e. 
> it is at least one order of magnitude smaller).
> Is there any mistake I have calculated or should I take account anything 
> for correct polarization calculation?
>
> The input file, output and dipole moment file after 200 geometry 
> optimization steps are attached here below.
>
> Kind regards,
>
> Phil
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20180712/4149acad/attachment.htm>