Hi all,
I have a problem with a calculation to the doped TiO2 system.
Starting with the same input files (same INCAR,POTCAR,POSCAR and KPOINTS), but considering different number of processors or nodes, I have different results. Even, some default parameters are differents The principal results are sumarized in the following table:
Version 5.2.11_c1_48 5.2.11_c1_16 5.2.11_c1_48 4.6.34_c2_24 5.2.12_c2_48
Ionic step 1 -1195.9524 -1195.9524 -1195,9524 -1195.9522 -1197.7534
Ionic step 20 -1196.9382 -1196.9761 -1196,9382 -1196.9614 -1197.7887
Ionic step 40 -1197.1798 -1197.2251 -1197,1798 -1197.1319 -
Optimized E -1197.2084 -1197.5576 -1197,7903 -1197.7839 -1197.7887
Parameters
IRDMAX 5905 5905 5905 5905 5905
Maximum index for augmentation-charges 376 555 376 376 380
EENTRO -0.35715641 -0.31169744 -0.35715641 -0.32224586 -0.33830403
EBANDS -992.3551 -820.6070 -992.3551 -3391.3072 -3702.5572
Notation: we use in version line ci_k, where c1 and c2 mean cluster 1 and 2, and k the number of processors, respectively.
I will be glad for any help
Thanks a lot in advance
same calculations, different results
Moderators: Global Moderator, Moderator
-
nany
same calculations, different results
Last edited by nany on Tue Aug 14, 2012 5:38 pm, edited 1 time in total.
-
alex
- Hero Member
- Posts: 585
- Joined: Tue Nov 16, 2004 2:21 pm
- License Nr.: 5-67
- Location: Germany
same calculations, different results
Could you please post the inputs as well?
Thx
Thx
Last edited by alex on Wed Aug 15, 2012 12:25 pm, edited 1 time in total.
-
nany
same calculations, different results
INCAR
SYSTEM = anatase surface 001 1x1x3
LCHARG = .FALSE.
LWAVE = .FALSE.
LREAL = Auto
ALGO = N
LSCALU = .FALSE.
NSIM = 4
LPLANE = T
#Electron Wavefunction Optimization
NELMIN = 6
PREC = N
ENCUT = 430
#Ionic relaxation
NSW = 40
IBRION = 2
ISIF = 2
ISMEAR = 0
SIGMA = 0.1
ISYM = 0
ISPIN = 2
LORBIT = 10
SYSTEM = anatase surface 001 1x1x3
LCHARG = .FALSE.
LWAVE = .FALSE.
LREAL = Auto
ALGO = N
LSCALU = .FALSE.
NSIM = 4
LPLANE = T
#Electron Wavefunction Optimization
NELMIN = 6
PREC = N
ENCUT = 430
#Ionic relaxation
NSW = 40
IBRION = 2
ISIF = 2
ISMEAR = 0
SIGMA = 0.1
ISYM = 0
ISPIN = 2
LORBIT = 10
Last edited by nany on Thu Oct 04, 2012 5:29 pm, edited 1 time in total.
-
boris
- Full Member
- Posts: 106
- Joined: Wed Jan 16, 2008 1:27 pm
- License Nr.: 567
- Location: France
same calculations, different results
Hi
The difference between 5.2.11_c1_48 and 5.2.11_c1_16 is really surprising. You just changed the number of cpus and you get a different result.
Try first to impose the number of bands, so that it's exactly the same for the two calculations. I don't know how many electrons you have, but be sure to have enough bands to accomodate all electrons.
Check in the outcar file if the number of bands has changed (even if you specify it in the incar, vasp may change it to fit the parallelization), and if it's exactly the same in both calculations, check the total energy.
<span class='smallblacktext'>[ Edited Tue Oct 09 2012, 01:30PM ]</span>
The difference between 5.2.11_c1_48 and 5.2.11_c1_16 is really surprising. You just changed the number of cpus and you get a different result.
Try first to impose the number of bands, so that it's exactly the same for the two calculations. I don't know how many electrons you have, but be sure to have enough bands to accomodate all electrons.
Check in the outcar file if the number of bands has changed (even if you specify it in the incar, vasp may change it to fit the parallelization), and if it's exactly the same in both calculations, check the total energy.
<span class='smallblacktext'>[ Edited Tue Oct 09 2012, 01:30PM ]</span>
Last edited by boris on Tue Oct 09, 2012 1:27 pm, edited 1 time in total.
-
Dr_Nick
same calculations, different results
Try to set NPAR = 1 in all calculations. This might not be the optimal setting but it might give you consistent results.
I have similar issues when using different numbers of CPUS and different values of NPAR, however the discrepancies are small and fully tolerable for what I need.
By the way, when you don't set NPAR, its value is automatically set to then number of cores.
I have similar issues when using different numbers of CPUS and different values of NPAR, however the discrepancies are small and fully tolerable for what I need.
By the way, when you don't set NPAR, its value is automatically set to then number of cores.
Last edited by Dr_Nick on Wed Oct 10, 2012 12:36 pm, edited 1 time in total.
-
nany
same calculations, different results
Yes, the difference is large. It was changing the number of cpus, and using the same input files.
In these calculations the number of band is not impossed, but the value is tha same in all of them (720), even in those that have different total energies. I check line by line the OUTCAR files of the all tests and the differences are in IRDMAX, Maximum index for augmentation-charges, EENTRO and EBANDS parameters. The rest are the same.
I haven 1080 electrons.
I dont know which is the problem exactly.
I contacted the BSC to find a solution. They said that it could be the code. Right now, they are doing some tests with my input files.
Thanks to all
In these calculations the number of band is not impossed, but the value is tha same in all of them (720), even in those that have different total energies. I check line by line the OUTCAR files of the all tests and the differences are in IRDMAX, Maximum index for augmentation-charges, EENTRO and EBANDS parameters. The rest are the same.
I haven 1080 electrons.
I dont know which is the problem exactly.
I contacted the BSC to find a solution. They said that it could be the code. Right now, they are doing some tests with my input files.
Thanks to all
Last edited by nany on Wed Oct 10, 2012 7:12 pm, edited 1 time in total.
-
nany
same calculations, different results
Ok, I will try both alternatives. thanks a lot!!
Last edited by nany on Thu Oct 11, 2012 10:07 am, edited 1 time in total.
-
nany
same calculations, different results
Hi all,
I included in the INCAR file the NBANDS and NPAR, as you suggested in the previous comments and the results are:
VASP 5.2.11
np 16 48
Step 1 -1195,9524 -1195,9524
Step 20 -1197,1616 -1197,1617
Step 40 -1197.1616 -1197.1617
Parameters
IRDMAX 5905 5905
Maximum index for augmentation-charges 555 376
NBANDS 610 610
EENTRO -0.21908174 -0.21908174
EBANDS 618.34222516 618.34222516
So, we can say that the results are similar.
Then, is it necessary to specify NBANDS and NPAR parameters to reach "good" and coherent results?
I included in the INCAR file the NBANDS and NPAR, as you suggested in the previous comments and the results are:
VASP 5.2.11
np 16 48
Step 1 -1195,9524 -1195,9524
Step 20 -1197,1616 -1197,1617
Step 40 -1197.1616 -1197.1617
Parameters
IRDMAX 5905 5905
Maximum index for augmentation-charges 555 376
NBANDS 610 610
EENTRO -0.21908174 -0.21908174
EBANDS 618.34222516 618.34222516
So, we can say that the results are similar.
Then, is it necessary to specify NBANDS and NPAR parameters to reach "good" and coherent results?
Last edited by nany on Tue Oct 16, 2012 2:03 pm, edited 1 time in total.
-
danith
- Newbie
- Posts: 19
- Joined: Wed Sep 12, 2012 4:19 pm
same calculations, different results
Hi Nany,
Could you help me with my problem? Im a new vasp user, but I want to calculate the energy of a small molecule in different positions (150 atoms) on a surface of MoO3, Is it possible to do it in gasp?
Thank you in advance,
Danith..
Could you help me with my problem? Im a new vasp user, but I want to calculate the energy of a small molecule in different positions (150 atoms) on a surface of MoO3, Is it possible to do it in gasp?
Thank you in advance,
Danith..
Last edited by danith on Tue Oct 16, 2012 2:14 pm, edited 1 time in total.
-
nany
same calculations, different results
Last edited by nany on Wed Oct 17, 2012 2:02 pm, edited 1 time in total.