Dear Mengqiu: There might be systems (for example if you have a nanowire not periodic in XY) that they only require one K point in XY for the correct computation of the current.
But in general you need to specify a dense enough mesh in XY. As you have already seen, there is an obvious dependance on the k sampling (Kx and Ky) since the Transmission(E) is computed by summing up the individual probabilities of transmission of all electrons incident with that total energy E. Now for a given energy E, there are many "incidence angles" (or Kx,Ky points) possible. So the sum to all electrons with total energy E translates into a sampling to all possible incidence angles, i.e. a sampling to (Kx, Ky). The finer that sampling the better, but of course to save CPU time the correct thing to do is to test for convergence on that mesh, to make it just fine enough. In the examples you show probably the Kx,Ky mesh is not converged yet. (But it seems to me in any case that you are getting different expressions for a numerical "zero" for the current.) Yours -- Jose A. Torres, Ph.D. SIESTA Manager On Thu, 2010-06-17 at 13:04 +0000, CaiMeng-Qiu wrote: > Dear Jose A. Torres： > > Thank you very much again! > I also several questions for these. > > > > > > > > On Thu, 2010-06-17 at 11:54 +0000, CaiMeng-Qiu wrote: > > > Dear Jose A. Torres： > > > > > > Thank you much for your explanation in detail. > > > > > > > > > > > - Also choose a large enough number of points to > > > > describe the function T(E) properly! (it may > > > > have sharp peaks) > > > > > > Do you mean I should adopt the dense k-mesh and large value of > > > TSTBT.NPoints? > > > > Here I just mean a TSTBT.NPoints to be large enough. > > > In the TRANSIESTA-manual, it mentions:" In the case of tbtrans, the > k-point sampling has to be specified also using a kgrid_Monkhorst_Pack > block, and can differ from the sampling that was used in the > TRANSIESTA calculation. The convergence of the transmission function > with respect to the k sampling can be slower than the one for the > density matrix. This means that one may have to increase the number of > k-points used in tbtrans." Did it mean that we adopt the dense k-mesh > for the tbtrans calcualtion after the Transiesta calculation? > > I do find that, for the special TSTBT.NPoints, the varied k-mesh in > the xy-plane generate the changed current. > For examplefor the tbtrans calcultions: > v -1.0 eV > TSTBT.NPoints 1500 > k-mesh 6x6x50 > I=-3.17E-16 > > where: > v -1.0 eV > TSTBT.NPoints 1500 > k-mesh 6x6x70 > I=-3.17E-16 > > where: > > v -1.0 eV > TSTBT.NPoints 1500 > k-mesh 7x7x60 > I=-2.94E-16 > > where: > v -1.0 eV > TSTBT.NPoints 1500 > k-mesh 8x8x60 > I=-3.59E-16 > > If i should check both the k-mesh and TSTBT.NPoints for the > calculated current? > > Thank you in advance! > > Sincerely, > > Mengqiu > > > > > > > > > Generally, how much is it for the value of TS.TBT.NPoints? > > > > A few hundreds to a few thousand points. > > > > > > > > Thank you very much again! > > > > > > Sincerely, > > > > > > Mengqiu > > > > > > > > ______________________________________________________________________ > 使用新一代 Windows Live Messenger 轻松交流和共享！ 立刻下载！
