Mobility calculation on graphene

[paramjot]

Hi! I calculating doped graphene mobility using the tutorial in the website using dell precision 5720 of 1 tb RAM but it still takes 14 days to compute the result. is there any faster way to do so?

[Tue Gunst]

Hi,
So guess you are following this recently updated tutorial:
https://docs.quantumatk.com/tutorials/mobility/mobility.html

The time-limiting step is the full (k,q)-dependent electron-phonon coupling (EPC). If this is not the case then please state where you are waiting.
It should take approximately 10 hours on a 16-core node to do the EPC in the tutorial.
If that is not the case then maybe check again that all the settings are as in the tutorial.

In general, the EPC can be parallelized over many processes (as there are many combinations of k and q-points).
The following Mobility calculations can be done on a single core and only takes a few minutes.

[paramjot]

I use GGA for the calculation and the sampling format of EPC is not the same in my quantum atk software. so I use 3x3x1(i have attached the pics of my EPC sampling format and example's one) graph has 25 atoms. EPC and Hamiltonian matrix take the most of the time like around 13days. I use the isotropic scattering rate method

[Roc2019]

Dear Dr. Gunst ,

Is it possible to calculate the mobility of a spin-polarized structure, such as a 2D magnetic monolayer?

Thank you.

[Tue Gunst]

@paramjot:
It seems from your example that you only include 1 k and q point? Do I understand correctly that you try to do a calculation not for the primitive cell but a large graphene unitcell?
Parallelization will only help if you have k and q points.

@Roc2019:
Yes, it will work for polarized calculations. You can also do the spin life-time (check the ElectronPhononCoupling object description in the reference manual if interested).

[Roc2019]

Thank you Dr. Gunst .

[Roc2019]

Dear Dr. Gunst,

As I follow the new mobility example by 2019.12.sp1, I get the same mobility result.  But,  it prints the error infomation as analyzing the EPC.  So, what's wrong?  How to get the EPC picture in specified qx-qy palne?

Thank you.

[Tue Gunst]

I can reproduce the error.
It seems like the plotter fails for samplings with 1 q-point in the tetrahedron method.
For now I think the best option is to artificially set the number of q-points to 2 if the tetrahedron method is needed or use the gaussian method for 2D systems.
It is also possible to access the data directly through query-functions on the object.
Thanks for posting the issue, this is a new bug in 2019.12.sp1 and we are investigating the origin.

[Roc2019]

Thank you Dr. Gunst. 

However, I can not understand what you mean "For now I think the best option is to artificially set the number of q-points to 2 if the tetrahedron method is needed or use the gaussian method for 2D systems".  How to set this parameter and change the input file? 

Thanks again.

[Tue Gunst]

In the gui you can open the ElectronPhononCoupling widget and set the number of q-points under the section "q-point sampling".
Simply increase the sampling in the non-periodic direction from 1 to 2 q-points.
In the script instead you change the input parameter "qpoints_fractional".
The class is described in the reference manual:
https://docs.quantumatk.com/manual/Types/ElectronPhononCoupling/ElectronPhononCoupling.html
This also shows defines the query functions for the case you want to plot the result directly with pylab.
See for instance the query functions "couplingMatrixVsQ" and "unscaledCouplingMatrixVsQ" which gives the data used in the electron-phonon coupling plot.
The Gaussian method was only related to the plots in the Mobility analyzer, sorry if this was confusing.