QuantumATK Forum
QuantumATK => General Questions and Answers => Topic started by: Dipankar Saha on September 25, 2015, 08:59
-
Hi,
Can I use the relaxation energy (to reach a thermodynamically stable state) information {e.g. 'x' eV} ...for setting the force tolerance limit?
Regards_
Dipankar
-
I think not. The default 0.05 eV/Å is in general a very reasonable tolerance. However, in special cases (perhaps for Phonon calculations), you might want to use 0.01 eV/Å or so. Going lower than that with DFT is questionable.
-
Okay....Thanks a lot.... :) / But, what if it's a pretty high value....say 1 eV ?!!
Regards_
Dipankar
-
"what if it's a pretty high value....say 1 eV ?!!"
The results obtained such calculation parameters may nonsense.
-
You got it completely wrong zh .... May be you did not follow the previous replies as well as the original question....
If you would have...you could find.... that the question was not at all _ "what if force tolerance 1 eV/Ang" / It's quite obvious... / On the contrary...it is very much sensible and practical ......to have result (experimental) which shows that_
under a certain condition.....the meta-stable state of a material is getting transformed into a stable state (with struct. diff.) ....; and the relaxation energy can easily be 1 eV or, so.... / Moreover, there may or may not be any direct connection... as such!! / As Jess said...he does not think that these two can some what be related.... That is perfectly fine.... :)
I hope... now, I made it quite easy for you to understand....
-
Dipankar: ZH is quite right, you cannot "pick out" a metastable state by using an unreasonably high force convergence criterion, such as 1 eV/Å.
Please have a look at this image: https://www.google.dk/search?q=dft+force+minimization&espv=2&biw=1871&bih=1089&source=lnms&tbm=isch&sa=X&ved=0CAYQ_AUoAWoVChMIqOfj3c2ZyAIVpfNyCh2-nAfg#tbm=isch&q=metastable+state&imgrc=6o9aAb6d1fLa4M%3A (https://www.google.dk/search?q=dft+force+minimization&espv=2&biw=1871&bih=1089&source=lnms&tbm=isch&sa=X&ved=0CAYQ_AUoAWoVChMIqOfj3c2ZyAIVpfNyCh2-nAfg#tbm=isch&q=metastable+state&imgrc=6o9aAb6d1fLa4M%3A)
The total force in the metastable state (red) is zero! (force = dE/dx). However, it is not the globally most stable state, because the green state is lower in energy.
Please do not confuse "relaxation energy" (whatever that is supposed to mean) with the force convergence criterion in a geometry optimization (force is a derivative of the energy, not an actual energy). The force criterion simply specifies the maximum force allowed in the state after optimization. As such, it specifies how close to zero force you want to get. It makes no sense to go arbitrarily close, because at some point numerical fluctuations and imprecisions will dominate the forces. 0.05 to 0.01 eV/Angstrom is a very reasonable range.
-
No I'm not confused.... I repeat...what I'm looking for is not at all a force tolerance of 1 eV/Å. / Rather....if the activation energy for transformation...from one meta-stable state to the stable state of the struct. is measured as 'x' eV/atom .... then can it be somewhat linked with the Hellmann–Feynman force tolerance (say, which is 'y' eV/Å) .....?
Anyways........ :) / Thanks a lot Jess...for sharing your point of view....
_________
Besides, can you please tell me that_
whether we can calculate the 'Raman Tensor' in ATK or not?
Regards_
Dipankar