Should we choose: Free energy TOTEN, energy without entropy or energy(sigma->0)?

[dyc_2008]

dear all

I am so confused by the three type of energy.
The physics meaning of them is clear, but i just don't know how to chose the suitable one .

is there any correlation between the suitable choice and the parameters in the INCAR , such as ISMERA and IBRION?

furthermore, is there any difference between energy without entropy and energy(sigma->0)? in my opinion, when sigma->0 ,the entropy is zero.
however, in many calculations , energy without entropy and energy(sigma->0) are not equal

Thanks
<span class='smallblacktext'>[ Edited ]</span>

[tlchan]

[quote="dyc_2008"]
furthermore, is there any difference between energy without entropy and energy(sigma->0)? in my opinion, when sigma->0 ,the entropy is zero. however, in many calculations , energy without entropy and energy(sigma->0) are not equal
[/quote]
For energy (sigma->0), I think there is an extrapolation formula to estimate the value of the total energy corresponding to sigma=0 by using a sigma>0 result. The total energy without entropy simply subtracts the electronic thermal energy away from the calculated total energy.

[dyc_2008]

that is a very brilliant interpretation
thank you so much

[KWH]

I could use some help with this. In the O atom example on the Wiki, it says to use the "energy without entropy" for atoms and molecules. Other sources say use the sigma>0 approx always. This becomes especially important in desorption calculations where we are taking differences between systems using fermi smearing and gaussian smearing.

Can someone enlighten me on this issue?

[fabien_tran1]

SIGMA should be primarily considered as a "technical" parameter for the integration into the Brillouin zone. In some cases, like the O atom, the entropy term can be much larger (than in bulk metals) and become unphysical (see wiki/index.php/O_atom). Note also that the default value of SIGMA is 0.2 eV, which corresponds to a temperature of about 2000 K.

[fabien_tran1]

You can also have a look at ISMEAR.