Strange artefact/bug(?) since last update of QuantumATK

[F. Fuchs]

Dear QuantumATK developers,

we recently observed an artefact when switching from QuantumATK version 19.03 to 19.12. The transmission spectra, which we have calculated for two overlapping graphene sheets, shows a transmission of zero in a certain interval around the Fermi energy when using the newer version. You can find the comparison in the attachment including also the input scripts. There is no physical explanation for this artefact and we therefore assume a bug since the recent QuantumATK update (the results using QuantumATK 19.03 make perfect sense).

Thank you in advance.

[Anders Blom]

This is a feature! By avoiding to calculate the transmission in regions where it is known to be zero (i.e. where the electrodes have band gaps), a lot of time can be saved.  The reason the transmission actually was not zero in earlier version is due to numerical noise. From a practical perspective, to compute the current, both approaches should be completely equivalent to within many significant digits, however if you really want to old behavior back, you can change the parameter "enforce_zero_in_band_gap" (https://docs.quantumatk.com/manual/Types/TransmissionSpectrum/TransmissionSpectrum.html)

[Anders Blom]

Ok, it is not obvious that this is the issue - I thought the problem was zero vs very small transmission (sorry, didn't check the ZIP before answering, bad habit).

This new keyword could still be the underlying reason, so would you mind rerunning in 2019.12 with this keyword set to False, just as a verification? If that changes the transmission, then probably we need to more carefully assess how we determine the electrode band gap, in cases like this.

[F. Fuchs]

My student made a test calculation and set the keyword to False in 2019.12. The resulting transmission spectrum agrees with the result from the earlier QuantumATK version.

This gives us a viable work around. Thank you!

[Anders Blom]

Good, thanks for letting us know! We will make sure to look into this problem and fix the issue of course.