Dear Ioan,

I didn't get an answer here, but yes it seems like the polarization is actually 
a dipole moment, so you need to divide by the volume. And I managed to get 
converged results after a long testing of all of the different inputs.

What you did sounds sensible. But with siesta it seems like there is no 
guarantee that you will get results on a particular polarization branch 
(somebody please correct me if I am wrong). This is probably why the 
polarization seems large. What you should do first is calculate the quantum of 
polarization manually using the atomic positions and formal charges (check the 
output of siesta or pseudopotential file if you are not sure), and take each 
polarization modulo the quantum. After that you can look at changes in 
polarization with respect to the cubic phase, and hopefully the numbers will 
seem more reasonable. But I don't know anything about BFO, so somebody else 
might be able to give some more specific advice.

From: <> on behalf of ionut 
ghitiu <>
Sent: 21 April 2022 09:38
To: <>
Subject: Re: [SIESTA-L] A few quesitons about macroscopic polarization

Hi Daniel,

I was wondering whether you received any response to your questions regarding 
the calculation of macroscopic polarisation in siesta, and if so, could you be 
so kind to share it. I'm currently trying to calculate the polarisation of some 
BiFeO3 systems but the values I obtain are about four times larger than 
expected. These were obtained by converting the output from Debye to C*m and 
then dividing each component by the volume of the unit cell to get the 
polarisation components. Following this I subtracted the values obtained for 
the symmetrical structure from those of the real system and calculated the 
resultant polarisation.

Kind regards,

On Monday, March 28, 2022, 11:04:25 PM GMT+3, Daniel Bennett 
<> wrote:

Hi all,

I have a few questions about calculating the macroscopic polarization in siesta:

  1.  I see that the units are given in Debye, which is a dipole moment. So is 
siesta really outputting the macroscopic polarization * unit cell volume?
  2.  From the manual, it seems like the formula used to calculate the 
polarization is only defined up to a quantum. Does anybody know if the output 
polarization are given modulo the quanta (like in abinit), or if it is not 
possible to tell which polarization branch they are on? I am doing a set of 
calculations for a particular system which does not have a well-defined 
reference state, so I can't use the Born charges to measure the change in 
polarization across my calculations. The best I can do is measure the change of 
the macroscopic polarization across my calculations, so I am wondering if it is 
guaranteed that they will be on the same branch.
  3.  Does anybody have advice for choosing a good polarization grid? I know I 
should converge my results with respect to this, but not sure what the best way 
to do this would be. Should I just converge the diagonal terms and then the 
off-diagonal terms? Any advice would be appreciated.


Daniel Bennett

SIESTA is supported by the Spanish Research Agency (AEI) and by the European 
H2020 MaX Centre of Excellence (
SIESTA is supported by the Spanish Research Agency (AEI) and by the European 
H2020 MaX Centre of Excellence (

Responder a