Hi,
I know the EP coupling code is still experimental, but I'm running an electron phonon coupling calculation on Gold, as I want to calculate the phonon limited resistivity at some point in the future.
I'm wondering how many iterations is reasonable in this sort of calculation?
The calculation is still running, some of the steps are done, some are taking a lot of iterations:
INFO_Q0000_0000_0000_S01_A001_P1.OUT - Done, 19 iterations
INFO_Q0000_0000_0000_S01_A001_P2.OUT - Done, 19 iterations
INFO_Q0000_0000_0000_S01_A001_P3.OUT - Done, 19 iterations
INFO_Q0102_0000_0000_S01_A001_P1.OUT - Done, 3 iterations
INFO_Q0102_0000_0000_S01_A001_P2.OUT - Done, 3 iterations
INFO_Q0102_0000_0000_S01_A001_P3.OUT - Done, 3 iterations
INFO_Q0104_0000_0000_S01_A001_P1.OUT - 12 iterations
INFO_Q0104_0000_0000_S01_A001_P2.OUT - 19 iterations
INFO_Q0104_0000_0000_S01_A001_P3.OUT - Done, 40 iterations
INFO_Q0108_0000_0000_S01_A001_P1.OUT - 88 iterations
INFO_Q0108_0000_0000_S01_A001_P2.OUT - 40 iterations
INFO_Q0108_0000_0000_S01_A001_P3.OUT - 41 iterations
INFO_Q0308_0000_0000_S01_A001_P1.OUT - 91 iterations
INFO_Q0308_0000_0000_S01_A001_P2.OUT - 39 iterations
INFO_Q0308_0000_0000_S01_A001_P3.OUT - 47 iterations
88 and 91 iterations seems like quite a lot. Am I missing some (other) convergence criterion? My input file is listed below, I'm also thinking that although the default Au.in may not be sufficient for accurate answers, will it cause this convergence difficulty?
Also, the calculations take so long that the timer loops around and becomes negative, maybe it should be made with more bits?
Anyway, any comments would be greatly appreciated, my input file is below, let me know if you want to see any outputs.
Cheers,
Marty
notes
!Au - EP Calc - ngridk 32
!+----------------------+
!| Unit Cell + Species |
!+----------------------+
avec
1.0 1.0 0.0
1.0 0.0 1.0
0.0 1.0 1.0
Hi,
I know the EP coupling code is still experimental, but I'm running an electron phonon coupling calculation on Gold, as I want to calculate the phonon limited resistivity at some point in the future.
I'm wondering how many iterations is reasonable in this sort of calculation?
The calculation is still running, some of the steps are done, some are taking a lot of iterations:
INFO_Q0000_0000_0000_S01_A001_P1.OUT - Done, 19 iterations
INFO_Q0000_0000_0000_S01_A001_P2.OUT - Done, 19 iterations
INFO_Q0000_0000_0000_S01_A001_P3.OUT - Done, 19 iterations
INFO_Q0102_0000_0000_S01_A001_P1.OUT - Done, 3 iterations
INFO_Q0102_0000_0000_S01_A001_P2.OUT - Done, 3 iterations
INFO_Q0102_0000_0000_S01_A001_P3.OUT - Done, 3 iterations
INFO_Q0104_0000_0000_S01_A001_P1.OUT - 12 iterations
INFO_Q0104_0000_0000_S01_A001_P2.OUT - 19 iterations
INFO_Q0104_0000_0000_S01_A001_P3.OUT - Done, 40 iterations
INFO_Q0108_0000_0000_S01_A001_P1.OUT - 88 iterations
INFO_Q0108_0000_0000_S01_A001_P2.OUT - 40 iterations
INFO_Q0108_0000_0000_S01_A001_P3.OUT - 41 iterations
INFO_Q0308_0000_0000_S01_A001_P1.OUT - 91 iterations
INFO_Q0308_0000_0000_S01_A001_P2.OUT - 39 iterations
INFO_Q0308_0000_0000_S01_A001_P3.OUT - 47 iterations
88 and 91 iterations seems like quite a lot. Am I missing some (other) convergence criterion? My input file is listed below, I'm also thinking that although the default Au.in may not be sufficient for accurate answers, will it cause this convergence difficulty?
Also, the calculations take so long that the timer loops around and becomes negative, maybe it should be made with more bits?
Anyway, any comments would be greatly appreciated, my input file is below, let me know if you want to see any outputs.
Cheers,
Marty
notes
!Au - EP Calc - ngridk 32
!+----------------------+
!| Unit Cell + Species |
!+----------------------+
avec
1.0 1.0 0.0
1.0 0.0 1.0
0.0 1.0 1.0
scale
3.85
atoms
1 : nspecies
'Au.in' : spfname
1 : natoms
0.0 0.0 0.0 0.0 0.0 0.0 : atposl, bfcmt
!PBE
xctype
20
!_______________________________________________________
! PART 1
!tasks
! 0
!ngridk
! 32 32 2
!vkloff
! 0.5 0.5 0.5
!gmaxvr
! 16.0
!lmaxapw
! 10
!lmaxvr
! 8
!_______________________________________________________
! PART 2
tasks
200
! good convergence required for accurate forces
epspot
1.d-7
! the q-point grid (a finer grid should be used for production)
ngridq
8 8 8
! set the scratch path to non-networked directory
scrpath
'/tmp/'
nempty
20
! this is an important parameter which fixes the k-point density irrespective of the size of the phonon supercell
radkpt
100.0