Calculating change in electronic polarization due LCALCPOL

[sophie_weber1]

Hello,

We are trying to apply calculate the electric polarization induced by an applied magnetic field in a magnetoelectric material, analogously to what is done in this work. https://journals.aps.org/prl/abstract/1 ... 106.107202

However, we are a bit confused about how to do this using LCALCPOL. The documentation seems to imply that only a single calculation is needed, but since it is the change in polarization that is relevant, it seems that we would have to do two calculations, one with no magnetic field applied, and one with magnetic field applied (via BEXT), where for the latter we read in the converged CHGCAR from the converged calculation with no magnetic field self-consistently, and both calculations are done with LCALCPOL=.TRUE. But it is not at all clear if this is the correct procedure based on the wiki page.

Right now, as a first step we have trying a single calculation using LCALCPOL with a finite applied magnetic field, and all three components of the ionic and electronic polarization are exactly zero.

Thank you in advance!

[marie-therese.huebsch]

Hi Sophie,

I am trying to reproduce the Cr2O3 calculation, but it is not clear to me which XC functional and which PAW potential they used. Do you know by any chance?

Regarding LCALCPOL, yes it is the difference of the polarization so you need the calculation at 0T, 10T and 20T. And I'd say restaring from WAVECAR is saver than CHGCAR to get the adiabatic conection between the unpolarized and the polarized calculation.

My preliminary observation is that the 10T run shows an induced electronic polarization of 0.00009 |e|·Å along x with net transverse magnetization of -0.0425 μ_B (field-induced canting). That's using PBE, PAW Cr_pv 02Aug2007 and PAW O 22Mar2012 plus the same specs as provided in the paper. Maybe my structure is off. I need to do some diligent testing, but wanted to let you know what I have thus far.

Best regards,
Marie-Therese