Re: NEGF calculations. ( No.1 ) |
- Date: 2018/02/23 18:51
- Name: T. Ozaki
- Hi,
> 1- One is i have successfully converges the calculations upto 10-6 when the leads are FM. > But with the same settings i am unable to converge with AFM leads even up to 10-4. > Shouldn't same settings work on AFM leads??
I think that the threshold of 1.0E-6 may be reached even for the AFM leads by tuning the mixing parameters, since you have already reached to 1.0e-4.
> 2- Then i did a lot of effort (change parameters again and agian) and the calculation is > almost converged upto 10-4 but job killed and when i restart the calculations again it > didn't start from where it was finished and suddenly jumped to far away from that > converged value and then takes a lot of time to reach to that convergence again?? Please > explain it and how can i solve this problem. i have pasted screen shot of .DFTSCF file.
The restart files are saved after finishing the SCF loop. Therefore, if your calculation is terminated during the SCF iteration, your second calculation cannot find any restart files. This is what you encountered. If you can guess before calculation that your job might be terminated by wall time, you should set a rather smaller scf.maxIter which guarantees the normal end of the job, and restart again the job with the restart files.
Regards,
TO
|
Re: NEGF calculations. ( No.2 ) |
- Date: 2018/03/01 15:16
- Name: Hashmi <arqumhashmi@gmail.com>
- Dear Sir,
Thank you very much for your prompt reply and your guidance. I tried to calculate transport according to your suggestion. I have one question that i have set my criteria for the calculation is scf.criterion 1.0e-5 # default=1.0e-6 (Hartree)
I have pasted the screenshot of .DFTSCF file. As you can see criteria has already been achieved from SCF= 278 but calculation is still running...... Can you explain it.??
SCF= 270 NormRD= 0.000011603228 Uele= 0.000000000000 SCF= 271 NormRD= 0.000011578918 Uele= 0.000000000000 SCF= 272 NormRD= 0.000011554969 Uele= 0.000000000000 SCF= 273 NormRD= 0.000054770300 Uele= 0.000000000000 SCF= 274 NormRD= 0.000052510751 Uele= 0.000000000000 SCF= 275 NormRD= 0.000046014720 Uele= 0.000000000000 SCF= 276 NormRD= 0.000030358660 Uele= 0.000000000000 SCF= 277 NormRD= 0.000012490034 Uele= 0.000000000000 SCF= 278 NormRD= 0.000006822217 Uele= 0.000000000000 SCF= 279 NormRD= 0.000006506257 Uele= 0.000000000000 SCF= 280 NormRD= 0.000006375016 Uele= 0.000000000000 SCF= 281 NormRD= 0.000006252835 Uele= 0.000000000000 SCF= 282 NormRD= 0.000006139219 Uele= 0.000000000000 SCF= 283 NormRD= 0.000006033692 Uele= 0.000000000000 SCF= 284 NormRD= 0.000005935785 Uele= 0.000000000000 SCF= 285 NormRD= 0.000005845047 Uele= 0.000000000000 SCF= 286 NormRD= 0.000005761038 Uele= 0.000000000000 SCF= 287 NormRD= 0.000005683333 Uele= 0.000000000000 SCF= 288 NormRD= 0.000005611525 Uele= 0.000000000000 SCF= 289 NormRD= 0.000005545223 Uele= 0.000000000000 SCF= 290 NormRD= 0.000005484049 Uele= 0.000000000000 SCF= 291 NormRD= 0.000005427643 Uele= 0.000000000000
|
Re: NEGF calculations. ( No.3 ) |
- Date: 2018/03/02 14:45
- Name: T. Ozaki
- References: t-ozaki@issp.u-tokyo.ac.jp
- Hi,
For NEGF calculations, the termination condition is given by
sqrt(fabs(NormRD[0]))< scf.criterion
I see that your calculation seems to be convergent. In such a case, you can stop the calculation by on-the-fly control of SCF mixing parameters. Please take a look at http://www.openmx-square.org/openmx_man3.8/node44.html
Regards,
TO
|
Re: NEGF calculations. ( No.4 ) |
- Date: 2018/03/05 11:31
- Name: Hashmi <arqumhashmi@gmail.com>
- Dear Sir,
I have one more query. I have achieved convergence up to "scf.criterion 1.0e-5". But at the end of convergence i got strange results.
SCF= 485 NormRD= 0.000002560446 Uele= 0.000000000000 SCF= 486 NormRD= 0.000002155160 Uele= 0.000000000000 SCF= 487 NormRD= 0.000001818278 Uele= 0.000000000000 SCF= 488 NormRD= 0.000001538988 Uele= 0.000000000000 SCF= 489 NormRD= 0.000001308255 Uele= 0.000000000000 SCF= 490 NormRD= 0.000001118503 Uele= 0.000000000000 SCF= 491 NormRD= 0.000000963363 Uele= 0.000000000000
The leads are AFM (i have pasted my input before). The magnetic moment intrinsic bias and the net bias at the end should be zero. But what i got is
Chemical potential of left lead (Hartree) -0.165363818129 Chemical potential of right lead (Hartree) -0.123022191185
*********************************************************** *********************************************************** Informaions for NEGF calculation *********************************************************** ***********************************************************
Intrinsic chemical potential (eV) of the leads Left lead: -4.499778437490 Right lead: -3.347604146363 add voltage = 0.0000 (eV) to the left lead: new ChemP (eV): -4.4998 add voltage = 0.0000 (eV) to the right lead: new ChemP (eV): -3.3476
Intrinsic bias volatage: 1.152174291128 (eV) Net bias volatage: 1.152174291128 (eV)
Parameters for the integration of the non-equilibrium part lower bound: -5.453691291927 (eV) upper bound: -2.393691291927 (eV) energy step: 0.020000000000 (eV) number of steps: 153
*********************************************************** *********************************************************** Mulliken populations *********************************************************** ***********************************************************
Total spin moment (muB) -4.189208324
Please comment on it. What went wrong?? How to rectify it....
|
Re: NEGF calculations. ( No.5 ) |
- Date: 2018/03/05 13:13
- Name: T. Ozaki <t-ozaki@issp.u-tokyo.ac.jp>
- Hi,
When you calculated the following left and right leads by the band method,
left-lead.hks right-lead.hks
did you check the chemical potentials for both the cases? I wonder that you have the difference of chemical potentials in those calculations, which may happen due to improper parameter setting such as different k-grids and cutoff energy, etc.
Regards,
TO
|