The questions about the optimization of SIGMA ?

Problems running VASP: crashes, internal errors, "wrong" results.


Moderators: Global Moderator, Moderator

Post Reply
Message
Author
User avatar
VASP001
Jr. Member
Jr. Member
Posts: 72
Joined: Tue Oct 27, 2009 2:56 pm
License Nr.: 176, upgraded to paw
Location: Hefei National Laboratory for Physical Sciences at the Microscale

The questions about the optimization of SIGMA ?

#1 Post by VASP001 » Sat Dec 26, 2009 4:42 pm

I have a question about the SIGMA in the manual in page 70.
When the ISMEAR =0 or 1 2 ,it's necessary to choose a proper SIGMA .And set SIGMA1 SIGMA2 SIGMA3 or other values ,and compare the values of entropy T*S, and choose as much value of the SIGMA ,which the total energy (ie the term 'entropy T*S') in the OUTCAR file neligible .
My question is :Is the term 'entropy T*S' is fond in the Relaxtion or in the static-computation ?
Last edited by VASP001 on Sat Dec 26, 2009 4:42 pm, edited 1 time in total.

forsdan
Sr. Member
Sr. Member
Posts: 339
Joined: Mon Apr 24, 2006 9:07 am
License Nr.: 173
Location: Gothenburg, Sweden

The questions about the optimization of SIGMA ?

#2 Post by forsdan » Sat Dec 26, 2009 6:20 pm

The entropy contribution from the fictitious temperature of the electrons is written during each electronic step in the self-consistency cycle. Look at the converged result after one cycle (i.e. the first ionic step). Then change your sigma in order to make this value sufficiently small. Once you have made sigma small enough for the first ionic step it will typically be sufficient for all following ionic relaxation steps as well.

Hope this answers your question.

Cheers,
/Dan

<span class='smallblacktext'>[ Edited Sat Dec 26 2009, 07:22PM ]</span>
Last edited by forsdan on Sat Dec 26, 2009 6:20 pm, edited 1 time in total.

User avatar
VASP001
Jr. Member
Jr. Member
Posts: 72
Joined: Tue Oct 27, 2009 2:56 pm
License Nr.: 176, upgraded to paw
Location: Hefei National Laboratory for Physical Sciences at the Microscale

The questions about the optimization of SIGMA ?

#3 Post by VASP001 » Sun Dec 27, 2009 12:21 am

And I face the question is as below :
entropy T*S EENTRO = -0.29053915
energy without entropy = 4371.52751002 energy(sigma->0) = 4371.38224045
entropy T*S EENTRO = -0.05748008
energy without entropy = -159.57666251 energy(sigma->0) = -159.60540255
entropy T*S EENTRO = -0.00001925
energy without entropy = -1264.44543375 energy(sigma->0) = -1264.44544338
entropy T*S EENTRO = -0.00001612
energy without entropy = -1313.23637981 energy(sigma->0) = -1313.23638787
entropy T*S EENTRO = -0.00001673
energy without entropy = -1314.33240761 energy(sigma->0) = -1314.33241598
entropy T*S EENTRO = -0.00000006
energy without entropy = -1207.15845486 energy(sigma->0) = -1207.15845489
entropy T*S EENTRO = -0.00000984
energy without entropy = -1195.66204538 energy(sigma->0) = -1195.66205031
entropy T*S EENTRO = -0.00000847
energy without entropy = -1204.21306236 energy(sigma->0) = -1204.21306659
entropy T*S EENTRO = -0.00004214
energy without entropy = -1211.61783367 energy(sigma->0) = -1211.61785474
entropy T*S EENTRO = -0.00006990
energy without entropy = -1220.22197021 energy(sigma->0) = -1220.22200516
entropy T*S EENTRO = -0.00012019
energy without entropy = -1222.59655496 energy(sigma->0) = -1222.59661506
entropy T*S EENTRO = -0.00022681
energy without entropy = -1225.48831308 energy(sigma->0) = -1225.48842649
entropy T*S EENTRO = -0.00034863
energy without entropy = -1228.09557720 energy(sigma->0) = -1228.09575152
entropy T*S EENTRO = -0.00037931
energy without entropy = -1230.27304838 energy(sigma->0) = -1230.27323804
entropy T*S EENTRO = -0.00038931
energy without entropy = -1231.55284878 energy(sigma->0) = -1231.55304343
entropy T*S EENTRO = -0.00038281
energy without entropy = -1231.92727594 energy(sigma->0) = -1231.92746735
entropy T*S EENTRO = -0.00037798
energy without entropy = -1232.08034396 energy(sigma->0) = -1232.08053295
entropy T*S EENTRO = -0.00037912
energy without entropy = -1232.10846066 energy(sigma->0) = -1232.10865022
entropy T*S EENTRO = -0.00037994
energy without entropy = -1232.11631088 energy(sigma->0) = -1232.11650085
entropy T*S EENTRO = -0.00038019
energy without entropy = -1232.12163677 energy(sigma->0) = -1232.12182687
entropy T*S EENTRO = -0.00038085
energy without entropy = -1232.12277070 energy(sigma->0) = -1232.12296113
entropy T*S EENTRO = -0.00038064
energy without entropy = -1232.12297508 energy(sigma->0) = -1232.12316540
entropy T*S EENTRO = -0.00038052
energy without entropy = -1232.12330121 energy(sigma->0) = -1232.12349147
entropy T*S EENTRO = -0.00038047
energy without entropy = -1232.12359821 energy(sigma->0) = -1232.12378845
entropy T*S EENTRO = -0.00038045
energy without entropy = -1232.12369706 energy(sigma->0) = -1232.12388728
entropy T*S EENTRO = -0.00038050
energy without entropy = -1232.12374948 energy(sigma->0) = -1232.12393973
entropy T*S EENTRO = -0.00038044
energy without entropy = -1232.12377006 energy(sigma->0) = -1232.12396029
entropy T*S EENTRO = -0.00038042
energy without entropy = -1232.12378838 energy(sigma->0) = -1232.12397860
entropy T*S EENTRO = -0.00038043
energy without entropy = -1232.12380200 energy(sigma->0) = -1232.12399221
entropy T*S EENTRO = -0.00038043
energy without entropy = -1232.12380777 energy(sigma->0) = -1232.12399798
energy without entropy= -1232.123808 energy(sigma->0) = -1232.123998
entropy T*S EENTRO = -0.00037411
energy without entropy = -1232.12350946 energy(sigma->0) = -1232.12369651
entropy T*S EENTRO = -0.00037304
energy without entropy = -1232.12329389 energy(sigma->0) = -1232.12348041
entropy T*S EENTRO = -0.00037365
energy without entropy = -1232.12342413 energy(sigma->0) = -1232.12361095
entropy T*S EENTRO = -0.00037395
energy without entropy = -1232.12346928 energy(sigma->0) = -1232.12365625
entropy T*S EENTRO = -0.00037386
energy without entropy = -1232.12353636 energy(sigma->0) = -1232.12372329
entropy T*S EENTRO = -0.00037386
energy without entropy = -1232.12355156 energy(sigma->0) = -1232.12373849
entropy T*S EENTRO = -0.00037389
energy without entropy = -1232.12356026 energy(sigma->0) = -1232.12374721
energy without entropy= -1232.123560 energy(sigma->0) = -1232.123747
entropy T*S EENTRO = -0.00037807
energy without entropy = -1232.12387195 energy(sigma->0) = -1232.12406098
entropy T*S EENTRO = -0.00037815
energy without entropy = -1232.12390281 energy(sigma->0) = -1232.12409189
entropy T*S EENTRO = -0.00037814
energy without entropy = -1232.12390615 energy(sigma->0) = -1232.12409522
entropy T*S EENTRO = -0.00037815
energy without entropy = -1232.12390701 energy(sigma->0) = -1232.12409609
entropy T*S EENTRO = -0.00037815
energy without entropy = -1232.12390717 energy(sigma->0) = -1232.12409625
entropy T*S EENTRO = -0.00037815
energy without entropy = -1232.12390745 energy(sigma->0) = -1232.12409653
energy without entropy= -1232.123907 energy(sigma->0) = -1232.124097
And the first electronic self-consistency cycles ,the values don's satisfy the rules ,in fact only the firt two self-consistency cycles don't obey the rule ,and the rest ones of the first ionic step obey the rules .Is anything wrong for this result ?By the way ,my sigma =0.1
Last edited by VASP001 on Sun Dec 27, 2009 12:21 am, edited 1 time in total.

forsdan
Sr. Member
Sr. Member
Posts: 339
Joined: Mon Apr 24, 2006 9:07 am
License Nr.: 173
Location: Gothenburg, Sweden

The questions about the optimization of SIGMA ?

#4 Post by forsdan » Sun Dec 27, 2009 8:42 am

The relevant values are the converged result from each self-consistency cycle. So from your results we have:

1st ionic step: entropy T*S EENTRO = -0.00038043
2nd ionic step: entropy T*S EENTRO = -0.00037389
3rd ionic step: entropy T*S EENTRO = -0.00037815

Judging from your total energy values, you seem to have around 100 atoms in your system. This gives less than 4e-6 eV/atom, which is sufficient small. However, my own general criteria is the minimum of 1e-3 eV/atom and 20e-3 eV/supercell. So based on the this criterion you can even consider to increase sigma in order to reduce the computational costs, since a small sigma value will require a dense k-point grid.

Please also notice that the entropy contribution will be approximately the same for all ionic steps provided that the geometry doesn't change too dramatically. So therefore it is sufficient to perform the convergence tests for sigma with only one ionic step.

Hope this helps.

Cheers,
/Dan

<span class='smallblacktext'>[ Edited Sun Dec 27 2009, 09:48AM ]</span>
Last edited by forsdan on Sun Dec 27, 2009 8:42 am, edited 1 time in total.

Post Reply