Broadening temperature

[Dipankar Saha]

Hi,

In basic settings/ numerical accuracy, the broadening temperature is set to 1000 K. However, considering the electrode parameters, the temp. value is 300 K.      How these two values are different?

Best_
Dipankar Saha

[Petr Khomyakov]

Because these are two different temperatures. The first one is an artificial temperature to achieve the convergence of device self-consistent calculations in a smooth and faster manner. The corresponding solution and derived physical quantities should not virtually depend on this temperature value within some user-set tolerance. If one thinks this dependence is significant, one should reduce this artificial temperature value. You can also reduce it to 300 K if your device calculations show no convergence issue and are also converged with respect to the k-point sampling of the lateral BZ if your device is periodic in the lateral directions. Typically, higher the temperature - less points are needed, but one cannot increase indefinitely for the reason described in the previous sentences.

The second temperature is real, physical one, entering the equation for the conductance (and current) calculation from the Transmission Spectrum. However, it only accounts for the temperature effect on the electron distribution, meaning that thermal disorder related to lattice vibration at a given temperature is not taken into account. For that, one may use special thermal displacement method, for instance, see https://docs.quantumwise.com/manual/Types/SpecialThermalDisplacement/SpecialThermalDisplacement.html and references therein.   

[Dipankar Saha]

Okay, I understand. / Thank you very much!

Best_
Dipankar

[Kim_W]

I want to calculate a MOFET in T=300K. The electrode temperature was set to T=300K for all calculations. I have noticed "Systems with a band-gap (semiconductors, insulators, molecules): Use either Fermi-Dirac or Gaussian smearing with a low broadening, e.g. around 0.01 eV." in https://docs.quantumwise.com/manual/technicalnotes/occupation_methods/occupation_methods.html#background. So I calculated two same devices with different broadening temperature, one is set to 300 K and the other is set to 0.01 eV. However, the results are largerly different. The first one is not satisfied with the requirements of the ITRS. Both calculations show no convergence issue. Which one is correct?

[Petr Khomyakov]

It is difficult to see which of the results are correct without doing further convergence tests. If you decrease the broadening parameter value, you should also check if the results are converged with respect to the k-point sampling. As a matter of fact, you should check this convergence for 300 K broadening as well. If the results (converged with respect to k-point sampling for each temperature!) are sensitive to broadening,  one should further decrease broadening, increasing the k-point sampling accordingly.  This a general procedure for any kind of calculations involving artificial broadening of the Fermi-Dirac distribution.