Although it's hard to compare all details, the main problem with your ATK calculation is the k-point sampling, which is plain wrong. You use 1 point in the C direction, it should be 100 or similarly large.
Also, it's generally a bad idea to use an even number of k-points in ATK since we don't have Gamma-centered even grids at the moment. So, you will have a lot better accuracy if you use 5x5 or 3x3. The same goes for the transmission - if you don't include the Gamma point you miss a large part of the most important contribution to the current.
To export the transmission data, just use the "nlprint" functionality instead of your own loop, or else make sure to query only spin=Spin.Sum, else you get all kinds of spin combinations like Up-Down etc, that's why you have so many columns.
Dear Anders,
Thank you very much for your reply to this thread. You mean the difference between the results from ATK and TranSIESTA probably come from the k-ponts setting in ATK? That is, in TranSIESTA, one can use even k-pionts in each direction, but it is not this case in ATK, where we will lose the Gamma k-point. However, if I use even k-points in one direction, I also do not sample the Gamma k-point in TranSIESTA calculation. In contrast, if I use even k-pionts in TranSIESTA, I could use a displacement of 0.5 to reduce the k-point number to by one half. That's why we usually use even k-points in TranSIESTA calculation.
Second, you say one should use more than 1 k-points in ATK in the device calculation, to be specific,
device_numerical_accuracy_parameters = NumericalAccuracyParameters(
k_point_sampling=(4, 4, 1),
density_mesh_cutoff=300.0*Rydberg,
here one should use more than 1 k-pionts in z (or c) direction.
As far as know, the k-point in z direction here only be used for generating a guess electron density for the subsequent NEGF calculation. Is this initial electron density for NEGF so important in ATK?
In the above test, the user use 6 k-point in the z direction for left and right electrodes calculations. I understand why you say one should use 100 k-points in z direction in electrode calculation, which is discussed many times before. It is just for getting a good results for electrode calculation while not bringing large burden for the NEGF calculations.
Last, the user just want to know how to get a total transmission spectrum as a function of energy (spin dependent or not dependent) so that one can analyze and plot the result at a more freely. Also, he want to know how to get the transmission spectrum as a function of energy for each k-point.
With best regards.
/Guangping Zhang
