Hello, I am running calculations (5x5 monolayer doped semiconductor, 75 atoms)
mpirun -np 80 ph.x -nk 4 -in ph.in > ph.out
However, it is taking a very long time. Do you have any recommendations on how to do this faster? more efficiently?
Is there any other faster way to calculate the electron mobility of 5x5 monolayer doped semiconductor, with 75 atoms?
Thank you for anyone who could assist.
Re: Length of calculations
Moderator: stiwari
-
jhosamelly
- Posts: 9
- Joined: Wed Jul 21, 2021 8:12 am
- Affiliation: DLSU
Re: Length of calculations
Hi jhosamelly,
The time required for a phonon calculation depends on several factors, especially the k-mesh used in the SCF calculation and the q-mesh for the phonon calculations. Since your supercell is sufficiently large, smaller meshes could improve computational efficiency.
For efficiency, and if you have enough computational resources, I recommend splitting the phonon calculations by q-points (using the start_q and last_q options in the ph.in file; refer to the phonon user guide for details). Once completed, gather the *.dyn, *dvscf, and pattern files in a single directory for the EPW calculation.
If you need further assistance, feel free to provide additional details.
Happy EPWing,
Hari
The time required for a phonon calculation depends on several factors, especially the k-mesh used in the SCF calculation and the q-mesh for the phonon calculations. Since your supercell is sufficiently large, smaller meshes could improve computational efficiency.
For efficiency, and if you have enough computational resources, I recommend splitting the phonon calculations by q-points (using the start_q and last_q options in the ph.in file; refer to the phonon user guide for details). Once completed, gather the *.dyn, *dvscf, and pattern files in a single directory for the EPW calculation.
If you need further assistance, feel free to provide additional details.
Happy EPWing,
Hari
-
jhosamelly
- Posts: 9
- Joined: Wed Jul 21, 2021 8:12 am
- Affiliation: DLSU
Re: Length of calculations
Good day, everyone! I am trying to calculate the electron mobility of doped
molybdenum disulfide. I am using epw.x as the latest version is now capable
of calculating the electron mobility of 2D materials. However, it is taking
a lot of time, especially the ph.x calculations.
May I know if there is another way to calculate the electron mobility using
the data I already have? I currently have the band structure, PDOS, and
optical properties calculations (imaginary
(𝜀2) and real (𝜀1) parts of the complex dielectric function of the
pristine and doped monolayer MoS2).
Thank you for your assistance.
molybdenum disulfide. I am using epw.x as the latest version is now capable
of calculating the electron mobility of 2D materials. However, it is taking
a lot of time, especially the ph.x calculations.
May I know if there is another way to calculate the electron mobility using
the data I already have? I currently have the band structure, PDOS, and
optical properties calculations (imaginary
(𝜀2) and real (𝜀1) parts of the complex dielectric function of the
pristine and doped monolayer MoS2).
Thank you for your assistance.
Re: Length of calculations
Hi Jhosamelly,
You cannot calculate mobility using EPW without including phonons. The phonon data is essential for this calculation.
As Hari mentioned, you can try to speed up your calculations by optimizing the k-mesh and energy cutoff, but you’ll need to perform a convergence test to check accuracy. Additionally, consider the following steps to improve efficiency:
1) Experiment with different -nk values to optimize the number of pools for parallelization. Compare SCF calculation times to find the best setting for your system.
1) Try adjusting mixing_beta in the SCF calculation to see if it improves convergence speed.
Once you find the optimized set of parameters, split the phonons among q-points and then collect into one folder.
Let me know if you need further clarification.
Best regards,
Shashi
You cannot calculate mobility using EPW without including phonons. The phonon data is essential for this calculation.
As Hari mentioned, you can try to speed up your calculations by optimizing the k-mesh and energy cutoff, but you’ll need to perform a convergence test to check accuracy. Additionally, consider the following steps to improve efficiency:
1) Experiment with different -nk values to optimize the number of pools for parallelization. Compare SCF calculation times to find the best setting for your system.
1) Try adjusting mixing_beta in the SCF calculation to see if it improves convergence speed.
Once you find the optimized set of parameters, split the phonons among q-points and then collect into one folder.
Let me know if you need further clarification.
Best regards,
Shashi