STRESS_TENSOR Convergence Issue

Matt W MattWa... at
Mon Oct 17 13:12:01 UTC 2016


at quick glance your EPS_DEFAULT setting is very large (1E-7), this might 
well hit the accuracy of forces more than energies...something more like 
1E-10 to -12 would be more normal.


On Monday, October 17, 2016 at 1:59:56 PM UTC+1, Sebastian Hütter wrote:
> Hi all,
> I am using cp2k to calculate elastic properties of metals (example: 
> Molybdenum, bcc). I believe my solution is well converged, as I can 
> calculate E(V) for different volumes around the relaxed state, fit 
> Birch-Murnaghan (i.e. using gibbs2) and can find the correct bulk modulus 
> (~300GPa) from this. 
> However, when I try to use analytical STRESS_TENSOR output to compute the 
> full stiffness tensor from 6 deformations (x,y,z,xy,xz,yz), there appears 
> to be a strong sensitivity on simulation volume. I have then tried 
> different MULTIPLE_UNIT_CELL values, and so far have not seen anything 
> looking like convergence. I have tried:
> 2x2x2 (16 Atoms), plausible result for c11&c22, but c44 is wrong,
> 3x3x3 (54 Atoms), plausible result,
> 4x4x4 (128 Atoms), plausible c11&c22, c44 wrong,
> 5x5x5 (250 Atoms), wrong result,
> 6x6x6 (432 Atoms), plausible result,
> 7x7x7 (686 Atoms), wrong result
> All converge to the same energy/cell with high precision, so I wouldn't 
> expect any problem there. It should be noted that the 'plausible' results 
> are very close to each other and the 'wrong' results are very close to each 
> other as well. The 'plausible' values are also very close to what I have 
> found going back from known E, B and G and also from published experimental 
> values, that is C11 ~550GPa, C12~150GPa, C44~100GPa.
> Please find attached the base input file. Deformation in X with e = 0.05 
> replaces the cell parameters with the following, and so on:
> A      3.27239      0.00000      0.00000
> B      0.00000      3.11656      0.00000
> C      0.00000      0.00000      3.11656
> Do you see something wrong with my input file? I am a bit confused by the 
> fact that the energies seem to be converged well but forces not so much...
> Thanks in advance,
> Sebastian Hütter
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the CP2K-user mailing list