**NEGF: transmission, conductance and current** |
- Date: 2022/01/17 14:59
- Name:
**Anshu Gaur**
- Dear OpenMX developers and experts,
I have the following questions regarding NEGF calculations:
1. The transmission, the conductance, and the current are calculated for each k-point specified as "NEGF.tran.Kgrid" parameter. Is there any general strategy to define the k-points in non-transport directions? For example, for the case of graphene, does it matter if the Dirac point lies along the transport direction or otherwise included using dense k-point grid in the non-transport direction along which the Dirac point lies.
2. The reported values of conductance and current are average values (averaged over all k-points). Does this mean that these average values depend on the number and choice of k-points during calculations?
3. Is there any particular strategy to use a dense k-point grid without running out of memory? I have noticed that many times my calculations get stuck at
Parameters for the calculation of the current lower bound: -4.264846672279 (eV) upper bound: -3.846846672279 (eV) energy step: 0.010000000000 (eV) imaginary energy 0.001000000000 (eV) number of steps: 41
calculating...
myid0= 1 num_kloop0= 0 kloop0= 0 i2= 0 i3= 0 k2= -0.3333 k3= -0.4444 myid0= 1 num_kloop0= 0 kloop0= 1 i2= 0 i3= 0 k2= -0.3333 k3= -0.4444 myid0= 1 num_kloop0= 0 kloop0= 2 i2= 0 i3= 0 k2= -0.3333 k3= -0.4444 myid0= 1 num_kloop0= 0 kloop0= 3 i2= 0 i3= 0 k2= -0.3333 k3= -0.4444 myid0= 1 num_kloop0= 0 kloop0= 4 i2= 0 i3= 1 k2= -0.3333 k3= -0.3333 myid0= 1 num_kloop0= 0 kloop0= 5 i2= 0 i3= 1 k2= -0.3333 k3= -0.3333 myid0= 1 num_kloop0= 0 kloop0= 6 i2= 0 i3= 1 k2= -0.3333 k3= -0.3333 myid0= 1 num_kloop0= 0 kloop0= 7 i2= 0 i3= 1 k2= -0.3333 k3= -0.3333 myid0= 1 num_kloop0= 0 kloop0= 8 i2= 0 i3= 1 k2= -0.3333 k3= -0.3333 myid0= 1 num_kloop0= 0 kloop0= 9 i2= 0 i3= 2 k2= -0.3333 k3= -0.2222 myid0= 1 num_kloop0= 0 kloop0=10 i2= 0 i3= 2 k2= -0.3333 k3= -0.2222 myid0= 1 num_kloop0= 0 kloop0=11 i2= 0 i3= 2 k2= -0.3333 k3= -0.2222 .... .... ....
for several hours (6-12 hours) and get killed with "KILLED BY SIGNAL:9 (Killed)" error.
The device structure has about 1200 atoms (including left and right electrodes) and the k-point grid for transmission calculation is 3 x 9 for non-transport directions. The machine on which I am running this calculation has 32 cores (64 threads) with 256 GB RAM. Would it help if some of the values are written to temporary files on the hard disk to free up the memory during calculations?
Thanks and regards, Anshu
| |