[Pw_forum] stress or total energy

Katalin Gaal-Nagy katalin.gaal-nagy at physik.uni-regensburg.de
Thu Dec 28 11:18:55 CET 2006

Dear Javad,

one reason for the discrepancy could be that you are not converged enough 
with respect to the number of plane waves, since the stress is very 
sensitive to that quantity. If you have calculated also the stress during 
your convergence tests with respect to the ecut, you can estimate the 
error in the stress.
Of course one expects that the stress vanishes at the minimum of the total 

Another reason coul be that you are using the experimental lattice 
constant, which could yield a energy which is not at the minimum of the 
total energy due to the over/underestimation of the lattice constant using 
LDA/GGA. Thus, I suggest to relax also the lattice constant together with 
the c/a in order to obtain the values at the energy minimum of your 

However, 10 kbar is a rather small value which could be somehow 

But maybe someone else could give you further hints.

Best wishes,

On Thu, 28 Dec 2006, Javad hashemifar wrote:

> Dear PWscf users
> I am trying to relax the c lattice parameter of a tetragonal supercell
> that its a lattice parameter is fixed to some experimental value.
> Therefore I am calculating this supercell at different values of c
> (celldm(3)) and looking at the total energy and stress.
> My feeling is that these two parameter should have consistent behavior
> and both could be used to find the optimized c. I mean I expect that
> when total energy is minimized, then the stress should vanish but it
> is not the case for my calculation and I observe that when the total
> energy is minimum, there is considerable value of stress (~ -10 kbar)
> in the system.
> I would appreciate if some body explain what is the problem and
> if it is really true that stress and total energy have not the same
> behavior, which one should be used to find out relaxed lattice
> parameters. I expect that total energy is the main criteria for
> relaxation of the lattice parameters ??
> Best regards
> Javad Hashemifar

More information about the users mailing list