Negative entropy while relaxing metalic system ?!?

Message

lahaye · #1 Post by **lahaye** » Tue Jul 25, 2006 7:59 am

Hello,

I have a metalic carbon nanotube with Fe inside.
I relax the ions and also the cell shape & volume (using ISIF = 3).

As this is a metalic system, I use ISMEAR = 1 and SIGMA = 0.35,
which gave a small enough entropy with the pure carbon nanotube
(without the Fe atoms).

Now, with the Fe inside, I get negative entropies, like this:

Code: Select all

$ grep "entropy T" OUTCAR
  entropy T*S    EENTRO =        -0.03348333
  entropy T*S    EENTRO =        -0.04085538
  entropy T*S    EENTRO =        -0.02225508
  entropy T*S    EENTRO =         0.01793880
  entropy T*S    EENTRO =         0.03325612
  entropy T*S    EENTRO =        -0.03050861
  entropy T*S    EENTRO =         0.02235358
  entropy T*S    EENTRO =        -0.00616872
  entropy T*S    EENTRO =        -0.05104471
  entropy T*S    EENTRO =         0.00910373
  entropy T*S    EENTRO =        -0.05910159
  entropy T*S    EENTRO =         0.00575567
  entropy T*S    EENTRO =        -0.07575832
  entropy T*S    EENTRO =        -0.03117156
  entropy T*S    EENTRO =        -0.07769648
 ...

Should I ignore this, or am I doing something completely wrong?

Input files are here:
http://surfion.snu.ac.kr/~lahaye/vasp/
Thanks,
Rob.

bandy · #2 Post by **bandy** » Tue Jul 25, 2006 8:27 am

No. Gaussian smearing also gives negative entropy.
<span class='smallblacktext'>[ Edited Tue Jul 25 2006, 11:39AM ]</span>

#3 Post by **admin** » Tue Jul 25, 2006 10:10 am

please mind that the entropy term is the electronic contribution to the entropy. Due to the fit of the stepfunction at the Fermi edge by a Hermite polynomial of order N (ISMEAR) if you use the Methfessel-Paxton method, small negative T*S-terms may arise (for visualization, have a look at eg. p16 of k-points.pdf of the VASP-workshop tutorial handouts).
increase the order of the polynomial fit (ISMEAR=2 or higher)

lahaye · #4 Post by **lahaye** » Wed Sep 20, 2006 1:57 pm

Dear admin,

I again got a system that produces negative entropy values.

Step by step I increased the ISMEAR value, starting from 1, then 2, etc.
Up to ISMEAR = 5, I keep getting negative entropies.
Only for ISMEAR = 6, the entropy becomes positive.

Should I then stick to ISMEAR = 6 for this (metalic) system, when
performing ionic relaxation?

Thanks,
Rob.