[QE-users] Phonons on a flat potential energy surface

[Lorenzo Monacelli]

Dear Antonine Jay,

By looking at your frequencies, they seem not well converged. Consider 
that in a perfectly converged calculation the ASR should be respected 
and its application should affect only transnational modes. However, all 
the phonon frequencies change considerably there. Consider increasing 
the tr2_ph value to 10^-20 or higher in the ph.x input script and 
eventually also the wfc/density cutoff. Converging soft modes when a lot 
of atoms and few symmetries are present in the system could be really 
cumbersome with DFPT.

Bests,

Lorenzo


On 29/10/21 19:58, Tamas Karpati wrote:
> Dear Antoine Jay,
>
> I think that in case of multiple imaginary eigenvalues the lower
> energy of a conformer
> may be less indicative. On the other hand, frozen phonons might be
> both needed and to be avoided,
> unless you consider a sufficient number of surrounding atoms. Have you
> checked if inclusion
> of more atoms helps the imaginaries fade away? Did you try to toss O where the
> imaginary eigenvectors suggest? This way imaginaries can be sorted out
> one after the other
> (lucky ones get rid of them at once :)
>
> Hope this helps,
>    t
>
>
>
>
> On Fri, Oct 29, 2021 at 10:10 AM Antoine Jay <ajay at laas.fr> wrote:
>> Dear users,
>> I would like to perform a phonon calculation at Gamma on a 216-atoms crystalline silicon supercell in which has been added 1 single oxygen (PBE functional).
>> This interstitial O is located 'nearly' in the middle of two Si, and unfortunately, many local minima are possible for its position: C2, C1, C1h, D3d, each of them being separated by a very small energy barrier and a small distance as already shown in Countinho-2000:
>> https://sci-hub.mksa.top/10.1103/physrevb.62.10824
>>
>> When I calculate phonons of the C1h (that has the smallest energy) within DFPT, I obtain imaginary eigenvalues in the dynamical matrix.
>> Of course, when I look at the corresponding eigenvectors, I see that they are exactly located on the O.
>> Then I manually do a frozen phonon calculation using a displacement equal to the eigenvector and I see a positive curvature of the parabola meaning that the system is well at a minimum.
>> Increasing the size of the displacement in my Frozen phonon permits me to see the double minima switching between C2 and C1h.
>> I suppose that the size of the perturbation in DFPT is too high...? Where can I change it is the code?
>>
>> Best regards,
>>
>> Antoine Jay
>>
>>
>> Inputs in attachment
>> Results:
>> Forces are over converged:  Total force =  0.00049 Ry/au
>> Frequencies after DFPT are these
>>       freq (   1-   1) =       -118.6  [cm-1]   -->   ?
>>       freq (   2-   2) =       -114.5  [cm-1]   -->   ?
>>       freq (   3-   3) =        -99.2  [cm-1]   -->   ?
>>       freq (   4-   4) =        -38.0  [cm-1]   -->   ?
>>       freq (   5-   5) =        -37.7  [cm-1]   -->   ?
>>       freq (   6-   6) =        -37.7  [cm-1]   -->   ?
>>       freq (   7-   7) =         87.7  [cm-1]   --> A'              I+R
>>       freq (   8-   8) =         91.5  [cm-1]   --> A'              I+R
>>       freq (   9-   9) =         91.8  [cm-1]   --> A''             I+R
>>
>> and after simple ASR:
>> # mode   [cm-1]    [THz]      IR
>>      1   -107.94   -3.2359    0.0000
>>      2    -38.83   -1.1640    0.0000
>>      3    -10.69   -0.3206    0.0000
>>      4     -2.67   -0.0801    0.0000
>>      5      8.82    0.2643    0.0000
>>      6     29.09    0.8720    0.0000
>>      7     97.81    2.9324    0.0000
>>      8     98.94    2.9661    0.0000
>>      9     98.97    2.9670    0.0000 _______________________________________________
>> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
>> users mailing list users at lists.quantum-espresso.org
>> https://lists.quantum-espresso.org/mailman/listinfo/users
> _______________________________________________
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
> users mailing list users at lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users