QuantumATK Forum
QuantumATK => General Questions and Answers => Topic started by: AsifShah on August 21, 2025, 18:30
-
Dear Admin,
I want to calculate phonon transmission across an interface of two 2D materials. The dimension are device are 7.83 A in periodic direction and a vacuum in out-of-plane direction and a total number of atoms as 153. In the dynamical matrix, it shows default repetitions as 3x1x1 however, which increases total atoms to 459.
This increases computational time very much beyond few weeks. Therefore, I used repetitions as 1x1x1.
I want to know is it the right approach and how would it affect my results if I don't use repetitions?
Also, would the approach be at-least useful for qualitative results?
Thanks
-
If you go to 1x1 then at least turn on the Wigner-Seitz scheme, else you are making a Gamma point phonon approximation. I would probably check if some of the machine-learned forcefields like MACE or M3GNet give at least decent phonon dispersions, and if so use them, for a much, much, much shorter calculation time, which hopefully is at least equally qualitative as a very approximate DFT simulation.
-
Thanks. I will take a look into universal forcefields.
-
Hi
I fine tuned MACE with additional data as per my requirement and when run on GPU, it is very very slow. It has been one day it shows only two progress cells in phonon transmission. The total size is 928 atoms
Is this normal?
-
No, does not sound normal, I'd expect the whole thing to take a few minutes only, even on CPU. OK maybe an hour on CPU, as MACE is slow on CPU.
Maybe do a simple test calculation first on CPU for MoS2 with a similar number of atoms (make some distortion in the center to make it interesting) using the built-in ReaxFF potential in QuantumATK, just to get a feeling for the calculation time.