[Pw_forum] problem with tddfpt: This type of the supercell is not supported
iurii.timrov at epfl.ch
Mon Dec 19 10:14:30 CET 2016
> 2) as you can imagine no it is not a molecule. If the problem is due to the cell parameters being not orthorhombic I might consider to run on a cluster.
If the system is not a molecule, or more generally is not a *finite system* which can be placed in the center of the simulation cell such that it does not touch cell's boundaries (such that the charge density of the system is vanishing at the cell's boundaries), then the treatment of a dipole in real space (by setting d0psi_rs=.true.) is not allowed. Please see a discussion in Sec. 2.5 in X. Ge et al., Comput. Phys. Commun. 185, 2080 (2014). I think that in your case you should set d0psi_rs=.false., which would treat the dipole in reciprocal space by dealing with the commutator [H,r], which however currently does not contain a contribution from the exact-exchange potential [V_EXX,r] (which means that the relative (and absolute) intensities of the peaks in the absorption spectrum would not be correct due to the missing contribution from EXX in the evaluation of the matrix element of the dipole). You may also consider suggestions from Giuseppe in the previous post to this thread, which could be a workaround to this problem, if the use of the continuum dielectric effective medium is a satisfactory level of approximation for your studies. Moreover, since we are both at EPFL we can arrange a meeting and discuss in person about all these issues.
> On the other hand I am genuinely too busy for coding so just in case do you think comment that error and try to run the calculation could it be worthy?
I think this not a good idea because: 1) it seems to me that the routine "shift_d0psi" cannot be used in the case of cells with non-orthogonal lattice vectors (this routine has to be investigated more carefully), 2) you should not treat the dipole in real space if you do not have a finite system as it is discussed above.
Dr. Iurii Timrov
Swiss Federal Institute of Technology Lausanne (EPFL)
Laboratory of Theory and Simulation of Materials (THEOS)
CH-1015 Lausanne, Switzerland
Pw_forum mailing list
Pw_forum at pwscf.org
More information about the users