Dear Nick, Thank you for your clear and helpful explanation. Just to be sure, I am using "siesta-4.0b-485", so for tbtran run I need for example 100*100*1 k-point; am I right?
Thank you very much in advance. On Mon, Nov 14, 2016 at 12:48 PM, Nick Papior <[email protected]> wrote: > > > 2016-11-12 19:53 GMT+01:00 Zara Nosh <[email protected]>: > >> Dear siesta users >> >> I have seen several transiesta examples in siesta package (Test and >> Example directories); in all examples the value of k-point in z-direction, >> for the scattering and tbtran runs are not equal to one and it has high >> value for example A3~60. I think we have to have just one k-point in >> transport direction as we have open boundary condition in this direction. >> > Yes, you are essentially correct. > > When dealing with 2 electrodes which spans the full unit-cell in the > transverse directions and the semi-infinite directions are parallel one > need not have any k-points along the semi-infinite directions. > > However, remember that transiesta first performs a siesta calculation > which estimates the ground-state density which is the input for transiesta. > Thus the electronic structure comes from a fully periodic calculation and > then subsequently the open-boundary conditions are applied. > Thus, if the ground state electronic structure is more adequately > described using a couple, or more, k-points along the semi-infinite > directions, you may indeed add them. > But, then please note the output of transiesta which automatically reduces > the number of k-points to 1 along the semi-infinite direction. > > Lastly, one may envision a system where there is periodicity along the > semi-infinite directions (with 1D electrodes, for instance), in this case > k-points along the semi-infinite directions are required. > So finally the requirement is ONLY; if there is periodicity in your device > region after having removed your electrodes you SHOULD use k-points. > > >> >> If we have 3 steps as bellow to calculate the transmission >> 1-transiesta <electrod.fdf >> to calculate Hamiltonian and overlap matrix of electrods >> >> 2-transiesta <scat.fdf >> to calculate density matrix of scattering region ] >> >> 3-tbtran <scat.fdf >> to calculate transmission >> >> I think we should take A1*A2*A3 k-point for the electrode part as >> electrodes are periodic in 3 dimensions and A3 should be large for example >> A3=100 and A1 and A2 are like as standard DFT calculations. >> for the second part we should take A1*A2*1 k-point as the system is open >> in z-direction (A1 and A2 are same as electrode part). >> > Correct, with the additional considerations I have outlined above. > >> Finally for the tbtran run we need high k-point density in the periodic >> directions and just 1 kpoint in the z direction for example B1*B2*1 (B1, B2 >> > A1, A2). >> > > >> >> Would you please correct me if I misunderstood. >> > For general 2-electrode setups as in the transiesta implementation pre 4.1 > you are correct. > >> >> >> I really appreciate your answer. >> >> Best >> Zara >> >> > > > -- > Kind regards Nick >
