[Pw_forum] phonon calc. for fcc at X
Stefano de Gironcoli
degironc at sissa.it
Sun Dec 10 23:34:55 CET 2006
If I remember correctly there used to be a term in the phonon
calculation in which the perturbing potential was written as DeltaV(r) =
exp(+iqr)*\DeltaVperiodic(r) and \DeltaVperiodic was expanded up to
the charge density cutoff ... this term was therefore slightly different
for two q wavevectors that would otherwise be equivalent. This never
gave any noticeable difference in real calculations and I would be
surprised if this would be the cause of your results but a way to check
this is to increase the charge density cutoff and see what happens as
in this case the problem should disappear when the potential description
become complete.
>From your input it looks to me that you are using the default value for
the selfconsistent threshold (conv_thr=1.d-6 in the &electrons namelist).
This is usually fine for total energy which is variational, but already
for structural relaxation it is often reduced so as to get more accurate
forces.
Phonon calculations are even more sensitive to the quality of
self-consistency of your unperturbed system. Try to reduce conv_thr to
1.d-8 in the scf (and nscf) steps and see if this improves things.
InN is a (wide band gap) semiconductor: there is no real need to use a
metallic setup which, by the way, is incompatible with the calculation
of dielectric constant and effective charges, that you will probably need
to get correct phonons frequencies at gamma.
I would expect that 2 or 10 Chadi-Cohen points should be enough for this
system (automatic grids 2 2 2 1 1 1 and 4 4 4 1 1 1, respectively).
hope this helps,
stefano
On Sun, 10 Dec 2006, e kb wrote:
> Dear Stefano , QE users,
> Thank you for your interest,
>
> Here are the scf, nscf and ph.x input files respectively, for X(0,0,1).
> For X(0,1,1) the scf file is the same, in nscf input I just changed the phonon namelist to (0.00 1.00 1.00) so not to make the mail longer and unreadable, I am skipping it and just adding ph.X input for X(0,1,1).
>
> I should note that this was a part of test calculations for the new pseudo for Indium that I recently generated using the 'atomic' code. The cut off energy that I used for these calculations (34 Ryd) is chosen to be enough to represent correct lattice constants, but still the pseudo and the value of cut off energy needs further examination. That is why I would not be surprised to see any discrepiencies in the frequencies. But again, I could not understand the differences that occured for the same point.
> With a verbose output I checked the phonon polarization vectors,
> As far as I understand, the calculation is done for the same modes for the two points.
>
> I do not want to bother you with the whole output files,
> if this is a very simple problem that an experienced/careful user can notice from the inputs,
> otherwise I can attach the output files.
>
> Buona domenica,
> Emine Kucukbenli
> ------------
> &control
> calculation = 'scf'
> restart_mode='from_scratch',
> prefix='34_vib',
> tprnfor = .true.
> pseudo_dir = '$PSEUDO_DIR' ,
> etot_conv_thr = 1.0D-5
> /
> &system
> ibrav= 2,
> celldm(1)=9.593140000
> nat= 2,
> ntyp= 2,
> ecutwfc =34.0,
> ecutrho =272.0,
> occupations = 'smearing'
> smearing = 'gaussian'
> degauss = 0.004,
> /
> &electrons
> mixing_mode = 'plain'
> mixing_beta = 0.7
> /
> ATOMIC_SPECIES
> In 114.818 In_us_2.upf
> N 14.0067 N.pbe-van_ak.UPF
> ATOMIC_POSITIONS { crystal }
> In 0.00000000000 0.00000000000 0.0000000000000
> N 0.25000000000 0.25000000000 0.2500000000000
> K_POINTS { automatic }
> 8 8 8 0 0 0
> ------------
> &control
> calculation = 'phonon'
> restart_mode='from_scratch',
> prefix='34_vib',
> tprnfor = .true.
> pseudo_dir = '$PSEUDO_DIR' ,
> etot_conv_thr = 1.0D-5
> /
> &system
> ibrav= 2,
> celldm(1)=9.593140000
> nat= 2,
> ntyp= 2,
> ecutwfc =34.0,
> ecutrho =272.0,
> occupations = 'smearing'
> smearing = 'gaussian'
> degauss = 0.004,
> /
> &electrons
> mixing_mode = 'plain'
> mixing_beta = 0.7
> /
> &phonon
> xqq(1) = .00, xqq(2) = .00, xqq(3) = 1.00
> /
> ATOMIC_SPECIES
> In 114.818 In_us_2.upf
> N 14.0067 N.pbe-van_ak.UPF
> ATOMIC_POSITIONS { crystal }
> In 0.00000000000 0.00000000000 0.0000000000000
> N 0.25000000000 0.25000000000 0.2500000000000
> K_POINTS { automatic }
> 8 8 8 0 0 0
> -------------------------
>
> &inputph
> tr2_ph=1.0d-14,
> amass(1)=114.818,
> amass(2)=14.0067,
> prefix='34_vib',
> fildyn='34_vib.dyn',
> /
> 0.00 0.00 1.00
> --------------------------------
>
>
> &inputph
> tr2_ph=1.0d-14,
> amass(1)=114.818,
> amass(2)=14.0067,
> prefix='34_vib',
> fildyn='34_vib.dyn',
> /
> 0.00 1.00 1.00
> ----------------------------------
> X(0,0,1)
> omega( 1) = 3.163928 [THz] = 105.537987 [cm-1]
> omega( 2) = 3.163928 [THz] = 105.537987 [cm-1]
> omega( 3) = 6.030058 [THz] = 201.142433 [cm-1]
> omega( 4) = 14.775784 [THz] = 492.870383 [cm-1]
> omega( 5) = 14.775784 [THz] = 492.870383 [cm-1]
> omega( 6) = 16.372030 [THz] = 546.115769 [cm-1]
> X(0,1,1)
> omega( 1) = 2.153190 [THz] = 71.823161 [cm-1]
> omega( 2) = 2.153190 [THz] = 71.823161 [cm-1]
> omega( 3) = 6.239891 [THz] = 208.141736 [cm-1]
> omega( 4) = 14.879198 [THz] = 496.319913 [cm-1]
> omega( 5) = 14.879198 [THz] = 496.319913 [cm-1]
> omega( 6) = 16.354833 [THz] = 545.542147 [cm-1]
> ----------------------------------
>
> Stefano Baroni <baroni at sissa.it> wrote:
> Dear Emin, dear all: all my apologies!
>
> Nicola was of course right, and I was WRONG! (011) IS indeed equivalent to (001), because the latter is equivalent to (-1,0,0) and the difference between (011) and (-100) is (111) which IS an FCC reciprocal-lattice vector. The coordinates of the K point are (3/4,3/4,0). With reference to the figure mentioned in the link below, K is the intersection among the planes x+y+z=3/2, z=0, and x=y, hence the above coordinates.
>
> To come to your original question. Are the symmetries/degeneracies of the results what you expect? Can you make the input/output of your calculation accessible to us?
>
>
> Stefano
>
> On Dec 10, 2006, at 2:46 PM, e kb wrote:
>
> Thank you for your answer Stefano,
> Of course this explains the different frequencies,
> and that also means that I misunderstand the kpoints given in example 5,
> about band structure for an fcc,
> and Dr. Marzari's explanation to another QE user,
> who was confused like me.
> http://www.democritos.it/pipermail/pw_forum/2006-June/004365.html
>
> Thank you again,
> Emine Kucukbenli
>
> Stefano Baroni <baroni at sissa.it> wrote:
> Emine: (001) and (011) are *NOT* two "X" points in the FCC Brillouin zone. The difference between the second and the first (010) is *NOT* a reciprocal-space vector. If it was, X would also be, which you will agree it is not. As a matter of fact, (011) has its own nickname (it is called "K"). No surprise, therefore, that frequencies @K are different from frequencies @X. Are these arguments clear and convincing enough? Stefano
>
> On Dec 9, 2006, at 11:10 PM, e kb wrote:
>
> Dear QE users,
> I am performing phonon calculations for a zinc blende structure,
> I calculated the frequencies at high symmetry points, evaluating a single point at each time.
> So I followed the procedure of scf, nscf and ph.x calculation one by one for L(0.5 0.5 0.5) - Gamma - X(0,0,1) - X(0,1,1) points.
> In these calculations every parameter (else than q-vector in &phonon list in nscf calculation and phonon card for ph.x input of course) is kept fixed.
>
> Although I expect the frequencies at two X points - (0,0,1) and (0,1,1) - to be the same,
> the difference in TA modes' frequencies are around 30 cm-1,
> in the other ones, around 5 cm-1.
>
> I would like to attribute this to unconverged lattice parameters, cut off energies or thresholds etc. but what I do not understand is why the program treats these two points differently, although they should be equivalent. So I expect to see the same value for frequency at these points, even if it is wrong, highly negative, or sth else that it should not be.
>
> I will appreciate any answers.
> Thank you.
>
> Emine Kucukbenli
>
>
>
>
>
> ---------------------------------
> Everyone is raving about the all-new Yahoo! Mail beta.
>
> ---
> Stefano Baroni - SISSA & DEMOCRITOS National Simulation Center - Trieste
> [+39] 040 3787 406 (tel) -528 (fax) / stefanobaroni (skype)
>
>
> Please, if possible, don't send me MS Word or PowerPoint attachments
> Why? See: http://www.gnu.org/philosophy/no-word-attachments.html
>
>
>
>
>
>
>
>
>
>
>
> ---------------------------------
> Check out the all-new Yahoo! Mail beta - Fire up a more powerful email and get things done faster.
>
> ---
> Stefano Baroni - SISSA & DEMOCRITOS National Simulation Center - Trieste
> [+39] 040 3787 406 (tel) -528 (fax) / stefanobaroni (skype)
>
>
> Please, if possible, don't send me MS Word or PowerPoint attachments
> Why? See: http://www.gnu.org/philosophy/no-word-attachments.html
>
>
>
>
>
>
>
>
>
>
> ---------------------------------
> Have a burning question? Go to Yahoo! Answers and get answers from real people who know.
More information about the users
mailing list