Calculation of phonon bandstructure of 2D materials

[beauyy]

Dear sir,
     I am calculating the phonon bandstricture of 2D materials with ATK 13.8.1. It spends too much time for calculating it by ATK-DFT. Do you have any method to save the time? What is the accurancy difference between ATK-DFT and ATK-classical in calculating phonon bandstructure?

[Jess Wellendorff]

ATK-classical is a great alternative to ATK-DFT for phonon calculations, especially for large systems where the DFT calculations can become quite expensive. However, it is hard to give exact numbers that show how accurate ATK-classical is compared to ATK-DFT. It is most likely gonna differ from case to case. However, as you can see in our tutorials on phonon calculations (http://quantumwise.com/publications/tutorials/itemlist/category/94-phonons-and-thermal-properties) the results often come out quite nice, so I certainly recommend that you try it out with ATK-classical to supplement your DFT calculations.

[Anders Blom]

I actually don't remember when we introduced parallelization over displacements, but at least that's the most obvious way to get good performance for these calculations, and it works very nicely in ATK 2014 and 2015. If you have a 2-atom cell, you can calculate the phonon spectrum in the same time as a single calculation for that system, if you use 6 MPI processes, which is 6 times faster than if you just run it in serial (at least if we don't count the initial calculation for the reference system). Now, if you have 20 atoms, you can parallelize up to 60 MPIs for a speedup of 60x (that is, from 2 months to 1 day).