[gpaw-users] Inaccuracies in phonon dispersion calculation

Jens Jørgen Mortensen jjmo at dtu.dk
Tue Nov 14 10:54:13 CET 2017


On 11/13/2017 04:09 PM, Holden Parks wrote:
> Hi Jens and Kristen,
>
> Thank you both for your suggestions. Kristen, though I believe my 
> structure does converge to ~1meV, I was not specifically setting this 
> as a convergence criterion but will do so from now on. I also agree 
> that using a cutoff for setting up the dynamical matrix is correct - I 
> had just ignored it recently because I believe my problems in getting 
> good dispersions were caused by a more serious problem in the original 
> force calculations.
>
> Jens, would you please send me your script and/or dispersion curve? I 
> would like to run this calculation as well. Thank you!

Attached.

Jens Jørgen

>
> -Holden
>
> On Fri, Nov 10, 2017 at 2:48 AM, Kristen Kaasbjerg 
> <kkaa at nanotech.dtu.dk <mailto:kkaa at nanotech.dtu.dk>> wrote:
>
>     Good point Jens, it is indeed crucial to crank up the convergence
>     criteria in phonon calculations. As the force is the derivative of
>     the total energy, I usually converge that one to ~< 1 meV. In this
>     respect, you should keep in mind that the criteria you specify for
>     the total energy in GPAW is per electron, i.e. you have to convert
>     to the total energy of all your supercell electrons.
>
>
>     Let me also emphasize that it is important to apply the cutoff to
>     the force constant in real space once setting up the dynamical
>     matrix. If you don't do that, the interatomic forces will
>     have longer range in some directions compared to others, which is
>     of course undesirable.
>
>
>     Best,
>     Kristen
>
>
>
>
>
>     ------------------------------------------------------------------------
>     *From:* Jens Jørgen Mortensen
>     *Sent:* 10 November 2017 07:18
>     *To:* Holden Parks
>     *Cc:* gpaw-users at listserv.fysik.dtu.dk
>     <mailto:gpaw-users at listserv.fysik.dtu.dk>; Kristen Kaasbjerg
>     *Subject:* Re: [gpaw-users] Inaccuracies in phonon dispersion
>     calculation
>     On 11/07/2017 03:20 AM, Holden Parks wrote:
>     > Hi Jens,
>     >
>     > Thank you for the reply. I've tried running the plane-wave
>     > calculations with your suggested cutoff but haven't seen an
>     > improvement. I've confirmed that inaccuracies in the forces are the
>     > cause of the problem in the dispersions, as I've run the same
>     > calculation with the Brennan potential (part of the ASAP
>     calculator)
>     > and the problematic features aren't present (I've attached it
>     here).
>     > Do you have any thoughts on how I should move forward?
>
>     I got a nice looking phonon dispersion with a 7x7x7 cell and these
>     parameters:
>
>     calc = GPAW(symmetry={'point_group': False},
>                  mode='lcao',
>                  basis='szp(dzp)',
>                  convergence={'density': 1e-7},
>                  xc='PBE',
>                  occupations=FermiDirac(0.01))
>
>     the important thing being a more tight convergence (I think).
>
>     Jens Jørgen
>
>     >
>     > -Holden
>     >
>     > On Fri, Nov 3, 2017 at 2:37 AM, Jens Jørgen Mortensen
>     <jjmo at dtu.dk <mailto:jjmo at dtu.dk>
>     > <mailto:jjmo at dtu.dk>> wrote:
>     >
>     >     On 10/31/2017 11:19 PM, Holden Parks via gpaw-users wrote:
>     >
>     >         Dear GPAW users,
>     >
>     >         I am having a problem calculating the phonon dispersion for
>     >         bulk diamond silicon. This calculation is performed in ase,
>     >         and I at first suspected that something was wrong with the
>     >         implementation. However, the example with aluminum using EMT
>     >         on the ase manual works perfectly, so I believe my
>     problem is
>     >         that I am incorrectly calculating the force constants
>     with my
>     >         calculator.
>     >
>     >         I've been calculating phonon dispersion curves for silicon
>     >         with a variety of supercell and k-point sizes, grid spacing
>     >         values, and displacement (delta) values. I've also tried a
>     >         using GGA xc functionals rather than LDA, and turned
>     >         point-group symmetry off. However, there are nonzero (and
>     >         imaginary frequencies) at the gamma point, and the acoustic
>     >         dispersion curves don't degenerate where they're expected
>     >         (between the gamma and W points and between the L and gamma
>     >         points). I've attached an example dispersion plot and
>     the code
>     >         I used to generate it.
>     >
>     >         Any suggestions at this point would be extremely helpful.
>     >         Thank you!
>     >
>     >
>     >     Try running your calculation in plane-wave mode.  Instead of
>     >     h=0.16, use something like:
>     >
>     >         calc = GPAW(..., mode=PW(300), ...)
>     >
>     >     and remember to do "from gpaw import PW".  This should give you
>     >     better forces.
>     >
>     >     Jens Jørgen
>     >
>     >         -Holden
>     >
>     >         ---------------------
>     >
>     >         from ase.build import bulk
>     >         from ase.dft.kpoints import ibz_points, bandpath
>     >         from ase.phonons import Phonons
>     >         from gpaw import GPAW, FermiDirac
>     >         import numpy as np
>     >
>     >         # Setup crystal and EMT calculator
>     >         atoms = bulk('Si', 'diamond', a=5.477)
>     >         calc = GPAW(kpts=(2, 2, 2),
>     > symmetry={'point_group': False},
>     >                         h=0.16,
>     >                         xc='PBE',
>     > occupations=FermiDirac(0.01))
>     >
>     >         ph = Phonons(atoms, calc, supercell=(4, 4, 4), delta=0.05)
>     >         ph.run()
>     >
>     >         # Read forces and assemble the dynamical matrix
>     >         ph.read(acoustic=True)
>     >
>     >         High-symmetry points in the Brillouin zone
>     >         G = [0, 0, 0]
>     >         X = [0.5, 0, 0.5]
>     >         W = [0.5,0.25,0.75]
>     >         L = [0.5, 0.5, 0.5]
>     >
>     >         point_names = ['$\Gamma$', 'X', 'W', 'L', '$\Gamma$']
>     >         path = [G, X, W, L, G]
>     >
>     >         # Band structure in meV
>     >         freq_conv = 2.41798902*(10**14)
>     >         path_kc, q, Q = bandpath(path, atoms.cell, 100)
>     >         omega_kn = freq_conv*ph.band_structure(path_kc)/(10**12)
>     >
>     >
>     >         import matplotlib.pyplot as plt
>     >         plt.figure(1, (8, 6))
>     >         plt.axes([.1, .07, .67, .85])
>     >
>     >         for n in range(len(omega_kn[0])):
>     >             omega_n = omega_kn[:, n]
>     >             plt.plot(q, omega_n, 'k-', lw=2)
>     >
>     >         plt.xticks(Q, point_names, fontsize=18)
>     >         plt.yticks(fontsize=18)
>     >         plt.xlim(q[0], q[-1])
>     >         plt.ylabel("Frequency ($\mathrm{THz}$)", fontsize=22)
>     >         plt.grid('on')
>     >
>     >         plt.show()
>     >
>     >
>     >         _______________________________________________
>     >         gpaw-users mailing list
>     > gpaw-users at listserv.fysik.dtu.dk
>     <mailto:gpaw-users at listserv.fysik.dtu.dk>
>     >         <mailto:gpaw-users at listserv.fysik.dtu.dk
>     <mailto:gpaw-users at listserv.fysik.dtu.dk>>
>     > https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users
>     <https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users>
>     >        
>     <https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users
>     <https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users>>
>     >
>     >
>     >
>
>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: phonon.py
Type: text/x-python
Size: 1249 bytes
Desc: not available
URL: <http://listserv.fysik.dtu.dk/pipermail/gpaw-users/attachments/20171114/c6c61b8a/attachment-0001.py>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: phonons.png
Type: image/png
Size: 56446 bytes
Desc: not available
URL: <http://listserv.fysik.dtu.dk/pipermail/gpaw-users/attachments/20171114/c6c61b8a/attachment-0001.png>


More information about the gpaw-users mailing list