Thermal properties of metals with CP2K

[Alex]

Hi all,

Is there a way with CP2K to calculate things like thermal expansion coefficient of a metal, or the electronic part of heat capacity and/or conductance? From the theoretical point of view, at least the way I see it, this would require calculating the electronic band structure in combination with existing theory (e.g. http://arxiv.org/abs/cond-mat/0109020 ). It is my understanding that CP2K only calculates the electrons' energy at the gamma point of the Brillouin zone. Are there any other ways to approach this problem with CP2K? I am very new to all of this, so feel free to let me know if I am completely off base here.

Thank you,

Alex

[Ying Zhang]

Hey Alex,

I have the same question and I am brand new user. Do you figure out the way to calculate thermodynamic with CP2K now?

Thanks,

Ying

[Alex]

Hi Ying,

The short answer is no and I decided it was not a very well thought out idea. The longer answer is that you _may_ be able to get good data, depending on what satisfies you, how much you know about combining methods, what data you need, etc, etc. The naive way for, say, heat capacity, or the expansion coefficient, is to do a set of simulations at various temperatures and plot the average total energy and lattice spacing (respectively), and then calculate the local slopes. The lack of k-sampling aside, neither of these will give you what you want. Well, the heat capacity will be the "correct" classical 3R value, while the expansion coefficient is likely to be nonsensical. 

A way to get heat capacity is to split it into the phonon and the electron portions and then use the appropriate partition functions from QM. For phonons, for instance, this means the ability to obtain the ion core velocity autocorrelation to get PDOS, then apply Bose-Einstein statistics. For electrons, it can in principle be done similarly. Problem is, I think CP2K only outputs the total energy of the system, so please inquire with the developers and please let me know what they say. 

For any thermodynamic properties, the biggest issues are the ensemble size and the time scale limitation. With atom counts in the 100s and simulated times in tens of picoseconds, it will be a bit hard to justify the use of what I described above. So, using DFT directly is probably a bad idea. Indirect methods exist (e.g. obtain force constants from DFT, calculate phonon spectra analytically, etc), but that's a whole separate discussion.

Alex

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/SOQvHM4XLWA/unsubscribe.
To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.
To post to this group, send email to cp...@googlegroups.com.
Visit this group at http://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.