[DFTB-Plus-User] Electronic temperature

Reinaldo Pis Diez reinaldo.pisdiez at gmail.com
Sat Sep 8 17:52:20 CEST 2012


Dear Bálint,

Thanks for your answer.

Of course, I agree with you in that if you're looking for a ground 
state configuration at 0K then, using an electronic temperature to 
speed up convergence is something like a cheat at least and all the 
efforts should be devoted to reduce temperature up to 0K.

My point is that I noticed that in some cases, when the electronic 
temperature is about 20-50K, occupation numbers are still 
non-integer but E_{Mermin} differs from E_{KS} by a few tenths, 
hundredths or even thousandths of eV. This means that the entropy 
term that characterizes the Fermi-Dirac distribution at temperatures 
other than 0K is extremely small and then, E_{Mermin} could be taken 
for safe.

Well, I must admit that I'm not an expert at all in the Mermin free 
energy, I'm a theoretical chemist used to work with isolated 
molecules. I promised to deep into the Mermin free energy concept.

Thanks again for you interest. Best regards,

Reinaldo



On 09/07/2012 01:09 PM, Bálint Aradi wrote:
> Dear Reinaldo,
>
>> Following your advice, I've compared the total energy and the Mermin
>> free energy for those hard_to_converge cases I mentioned in my former
>> mail. I've noted that the difference between them is a few tenths of eV
>> for, say, 20-50 K. I guess this is a good result but now a new question
>> arises: Can the difference between the total energy and the Mermin free
>> energy be taken as a measure of the reliability of the results? In other
>> words, can I state that the smaller the difference between the total
>> energy and the Mermin free energy, the more reliable the result?
>>
> I have difficulties to interpret the phrase "more reliable results". At
> finite temperature you have non-integer occupancies. If you want to
> simulate a system at exactly 0 K, and are switching on temperature only
> in order to achieve convergence, that would indeed "falsify" your
> results to some extent. The difference between the two energies would
> then indicate you, how far you are from the case with electron
> distribution at 0 K. But otherwise I do not think, you can get a
> "reliability" information from that.
>
>    Cheers,
>
>     Bálint
>
>
>
>
> _______________________________________________
> DFTB-Plus-User mailing list
> DFTB-Plus-User at dftb-plus.info
> http://www.dftb-plus.info/mailman/listinfo/dftb-plus-user



More information about the DFTB-Plus-User mailing list