Negative frequency with denser q-points

Moderators: mverstra, aromero

Post Reply
lilymali03
Posts: 10
Joined: Mon Jul 13, 2020 11:30 pm

Negative frequency with denser q-points

Post by lilymali03 » Fri Aug 21, 2020 5:56 pm

Dear abinit users,

I am trying to calculate the phonon frequency for monolayer TaB2, which has entirely positive values in the paper "Predicting Novel 2D MB2 (M=Ti, Hf, V, Nb, Ta) Monolayers with Ultrafast Dirac Transport Channel and Electron-orbital Controlled Negative Poisson’s Ratio." My calculations are all positive when I am using 4*4*1 q-points, but it turns negative at the K point with 8*8*1 and 10*10*1. I've tried with higher ecut and ngkpt values, along with different tsmear values, but they didn't change my results. Any help would be greatly appreciated!

Here are my input files:

ndtset 11
#Set 1 : ground state self-consistency

occopt 7
tsmear 0.0055

getwfk1 0 # Cancel default
kptopt1 1 # Automatic generation of k points, taking
# into account the symmetry
nqpt1 0 # Cancel default
tolvrs1 1.0d-18 # SCF stopping criterion (modify default)
rfphon1 0 # Cancel default

#Q vectors for all datasets

#Complete set of symmetry-inequivalent qpt chosen to be commensurate
# with kpt mesh so that only one set of GS wave functions is needed.
#Generated automatically by running GS calculation with kptopt=1,
# nshift=0, shiftk=0 0 0 (to include gamma) and taking output kpt set
# file as qpt set. Set nstep=1 so only one iteration runs.

nqpt 1 # One qpt for each dataset (only 0 or 1 allowed)
# This is the default for all datasets and must
# be explicitly turned off for dataset 1.



qpt2 0.00000000E+00 0.00000000E+00 0.00000000E+00
qpt3 1.00000000E-01 0.00000000E+00 0.00000000E+00
qpt4 2.00000000E-01 0.00000000E+00 0.00000000E+00
qpt5 3.00000000E-01 0.00000000E+00 0.00000000E+00
qpt6 4.00000000E-01 0.00000000E+00 0.00000000E+00
qpt7 5.00000000E-01 0.00000000E+00 0.00000000E+00
qpt8 2.00000000E-01 1.00000000E-01 0.00000000E+00
qpt9 3.00000000E-01 1.00000000E-01 0.00000000E+00
qpt10 4.00000000E-01 1.00000000E-01 0.00000000E+00
qpt11 5.00000000E-01 1.00000000E-01 0.00000000E+00
qpt12 4.00000000E-01 2.00000000E-01 0.00000000E+00
qpt13 5.00000000E-01 2.00000000E-01 0.00000000E+00
qpt14 -4.00000000E-01 2.00000000E-01 0.00000000E+00
qpt15 -4.00000000E-01 3.00000000E-01 0.00000000E+00

#Sets 4-10 : Finite-wave-vector phonon calculations (defaults for all datasets)

getwfk 1 # Use GS wave functions from dataset1
kptopt 3 # Need full k-point set for finite-Q response
rfphon 1 # Do phonon response
rfatpol 1 3 # Treat displacements of all atoms
rfdir 1 1 1 # Do all directions (symmetry will be used)
tolvrs 1.0d-10 # This default is active for sets 3-10

#######################################################################
#Common input variables

acell 5.6907412415E+00 5.6907412415E+00 1.7324075862E+01
rprim 0.5 sqrt(3/4) 0
-0.5 sqrt(3/4) 0
0 0 1


#Definition of the atom types
ntypat 2 # There is only one type of atom
znucl 73 5 # The keyword "znucl" refers to the atomic number of the
# possible type(s) of atom. The pseudopotential(s)
# mentioned in the "files" file must correspond
# to the type(s) of atom. Here, the only type is Aluminum

#Definition of the atoms
natom 3 # There are two atoms
typat 1 2 2 # They both are of type 1, that is, Silicon.
xcart -3.2999532462E-27 4.0111926777E-28 1.2275477921E-01
-1.3658486347E-16 3.2855509877E+00 3.0705627007E+00
-6.2340394405E-16 6.5711019753E+00 3.0705627007E+00


#Gives the number of band, explicitely (do not take the default)
nband 14


#Definition of the planewave basis set

ecut 35.0 # Maximal kinetic energy cut-off, in Hartree
pawecutdg 70
pawovlp -1
pawxcdev 0

#Definition of the k-point grid
ngkpt 20 20 1

nshiftk 1
shiftk 0 0 0

#Definition of the SCF procedure
nstep 200 # Maximal number of SCF cycles


And my anaddb file:

!Flags
ifcflag 1 ! Interatomic force constant flag
ifcout 0

!Wavevector grid number 1 (coarse grid, from DDB)
brav 1 ! Bravais Lattice : 1-S.C., 2-F.C., 3-B.C., 4-Hex.)
ngqpt 10 10 1 ! Monkhorst-Pack indices
nqshft 1 ! number of q-points in repeated basic q-cell
q1shft 3*0.0

!Effective charges
chneut 1 ! Charge neutrality requirement for effective charges.
asr 2

!Interatomic force constant info
dipdip 1 ! Dipole-dipole interaction treatment

!Phonon band structure output for band2eps - See note near end for
! dealing with gamma LO-TO splitting issue.
eivec 4

!Wavevector list number 1 (Reduced coordinates and normalization factor)
nph1l 35 ! number of phonons in list 1

qph1l

1/2 0 0 1 #M point
0.45 0 0 1
0.4 0 0 1
0.35 0 0 1
0.3 0 0 1
0.25 0 0 1
0.2 0 0 1
0.15 0 0 1
0.1 0 0 1
0.05 0 0 1
0 0 0 1 #G point
1/24 1/48 0 1
1/12 1/24 0 1
3/24 3/48 0 1
2/12 2/24 0 1
5/24 5/48 0 1
3/12 3/24 0 1
7/24 7/48 0 1
4/12 4/24 0 1
9/24 8/48 0 1
5/12 5/24 0 1
11/24 11/48 0 1
6/12 6/24 0 1
13/24 13/48 0 1
7/12 7/24 0 1
15/24 15/48 0 1
2/3 1/3 0 1 #K point
31/48 7/24 0 1
15/24 3/12 0 1
29/48 5/24 0 1
14/24 2/12 0 1
27/48 3/24 0 1
13/24 1/12 0 1
25/48 1/24 0 1
1/2 0 0 1 #M point

!Wavevector list number 2 (Cartesian directions for non-analytic gamma phonons)

!The output for this calculation must be cut-and-pasted into the
! t59_out.freq file to be used as band2eps input to get proper LO-TO
! splitting at gamma. Note that gamma occurrs twice.

nph2l 1 ! number of directions in list 2

qph2l 1.0 0.0 0.0 0.0

ebousquet
Posts: 462
Joined: Tue Apr 19, 2011 11:13 am
Location: University of Liege, Belgium

Re: Negative frequency with denser q-points

Post by ebousquet » Sat Aug 29, 2020 5:36 pm

Dear lilymali03,
If you have tested the convergence w.r.t. k-points and ecut (and tsmear) and that you reproduce the same behaviour, i.e. appearance of an instability in the phonon dispersion curves, this might means that it can be a real instability. Did you check with increasing the number of bands (for metals it can matter)?
However, did the authors of the reference you compare with did such a q-grid convergence? If yes then there is indeed a disagreement...
A first step to check is:
- do you have the instability in the DFPT calculations or is it restricted in points between the DFPT calcaulted q-points?
- For all the DFPT calculation, could check in the output the lines that look like that:

Code: Select all

 Resulting in :
 2DEtotal=    0.9195153738E+01 Ha. Also 2DEtotal=    0.250212858005E+03 eV
    (2DErelax=   -8.3002708950E+04 Ha. 2DEnonrelax=    8.3011904104E+04 Ha)
    (  non-var. 2DEtotal :    9.1951537365E+00 Ha)
They are printed for each perturbation after the SCF convergence. In all cases the 2DEtotal should be equal (or very close to) non-var. 2DEtotal. Could you confirm this is the case?

Best wishes,
Eric

andyamygto
Posts: 42
Joined: Tue Nov 26, 2019 12:50 am

Re: Negative frequency with denser q-points

Post by andyamygto » Sun Aug 30, 2020 8:24 pm

Dear lilymali03,

Here is my suggestion.
in anaddb files, test the phonon dispersion with the variable rifcsph.
The variable decide the Radius of the Interatomic Force Constant sphere.
If you use the denser q-point, the short range force range is larger. Which might cause some sort of mistake.
But the test and the converge check is needed as well.
Use the variable as your own risk

BTW, you can use DFPT to calculate direct the q-point where you got the negative frequency to check if the imaginary frequency was coming from the DFPT itself or the Fourier transform in dynamic matrix.

Regards,
Andy Hsu

lilymali03
Posts: 10
Joined: Mon Jul 13, 2020 11:30 pm

Re: Negative frequency with denser q-points

Post by lilymali03 » Mon Aug 31, 2020 12:15 am

Dear Eric,

Thank you for your reply! I did check with a larger number of nbands (27), and the imaginary frequency was still there.

You mentioned:
ebousquet wrote:
Sat Aug 29, 2020 5:36 pm
- do you have the instability in the DFPT calculations or is it restricted in points between the DFPT calcaulted q-points?
Could you please elaborate on how I could check this?

For this:
ebousquet wrote:
Sat Aug 29, 2020 5:36 pm
- do you have the instability in the DFPT calculations or is it restricted in points between the DFPT calcaulted q-points?
- For all the DFPT calculation, could check in the output the lines that look like that:

Code: Select all

 Resulting in :
 2DEtotal=    0.9195153738E+01 Ha. Also 2DEtotal=    0.250212858005E+03 eV
    (2DErelax=   -8.3002708950E+04 Ha. 2DEnonrelax=    8.3011904104E+04 Ha)
    (  non-var. 2DEtotal :    9.1951537365E+00 Ha)
They are printed for each perturbation after the SCF convergence. In all cases the 2DEtotal should be equal (or very close to) non-var. 2DEtotal. Could you confirm this is the case?
the 2DEtotal is equal to the non-var. 2DEtotal.

Best,
Lily

lilymali03
Posts: 10
Joined: Mon Jul 13, 2020 11:30 pm

Re: Negative frequency with denser q-points

Post by lilymali03 » Mon Aug 31, 2020 12:18 am

Dear Andy,

Thank you for answering!

I tested my anaddb file with different values of rifcsph, and rifcsph=8 helped the negative frequency disappear at the K point. However, small negative frequencies do appear at the gamma point. How do I determine which value of rifcsph I should use/perform convergence for it?

You mentioned:
andyamygto wrote:
Sun Aug 30, 2020 8:24 pm

BTW, you can use DFPT to calculate direct the q-point where you got the negative frequency to check if the imaginary frequency was coming from the DFPT itself or the Fourier transform in dynamic matrix.
Could you please explain a bit more about how I can check where the imaginary frequency is coming from?

Thanks,
Lily

ebousquet
Posts: 462
Joined: Tue Apr 19, 2011 11:13 am
Location: University of Liege, Belgium

Re: Negative frequency with denser q-points

Post by ebousquet » Tue Sep 01, 2020 5:36 pm

Dear Lily,
Good if the 2DTE and non-var are the same, it means that the DFPT calculation is a priori OK.

Regarding qpt from DFPT vs qpt from interpolation is the following:
You do calculation with DFPT on a reduced number of q-points (from a given grid defined by ngqpt 10 10 1 in your case) and from these DFPT calculations, anaddb will interpolate the phonon frequencies on other points that you did not computed with DFPT. My question was if the unstable modes you obtain are at qpt that are coming from DFPT or are those in between these DFPT qpt points such that the frequencies comes from the interpolation only.
With a grid of ngqpt 10 10 1, it means you have the following DFPT calculated qpoints:
0.0 0.0 .0.0
0.1 0.0 0.0
0.2 0.0 0.0
...
0.5 0.0 0.0
0.5 0.1 0.0
0.5 0.2 0.0
...
0.5 0.5 0.0

My question is then, do you have the instabilities in those DFPT q-points or do they appear only way from these points? it is the same question as last Andy's question: If the instabilities appear only on q-points that are the DFPT ones, then we can conclude that they are indeed present (unless it is a problem of convergence w.r.t. k-points, ecut or tsmear, but you have tested them, so it is fine). Now, if the unstable modes appear only at q-points that are in between the DFPT calculation q-points, it is possible that only the interpolation is giving them, which could be wrong. To check this you have to do a DFPT calculation at some of these q-points where the interpolation is giving "negative" frequencies and check if for the "exact" DFPT calculation you still see the unstable mode (don't forget to impose the acoustic sum rule to compare). If the DFPT calculation also give unstable modes, then you can definitively conclude that you indeed have an instability in your calculation. If not, it means the interpolation failed to give the correct phonon frequencies outside the DFPT points and you need to add more DFPT q-points on your grid of interpolation (10x10 is already quite good but it depends on the system or as Andy say sometimes you can have spurious effects depending on the grids of interpolations).

In the paper you mentioned, which grid of DFPT q-point did they use to report their phonon band structures, is it the same as you?

Best wishes,
Eric

lilymali03
Posts: 10
Joined: Mon Jul 13, 2020 11:30 pm

Re: Negative frequency with denser q-points

Post by lilymali03 » Tue Sep 01, 2020 8:59 pm

Dear Eric,

Thank you for the response and explanation!

Part of my groundstate output file for dataset 3 says

Code: Select all


==> Compute Derivative Database <==
  
  2nd-order matrix (non-cartesian coordinates, masses not included,
   asr not included )
     j1       j2             matrix element
  dir pert dir pert     real part     imaginary part
  
   1    1   1    1         2.1120801869         0.0001953982
   1    1   2    1        -0.9069151766         0.0000000000
   1    1   3    1         0.0002684103         0.5822768366
   1    1   1    2        -0.5924786575         0.4693123197
   1    1   2    2         0.4533749036         0.1789820805
   1    1   3    2        -0.3233191887        -1.1962946537
   1    1   1    3        -0.7503522614         0.0867874570
   1    1   2    3         0.1936321849        -0.4479388192
   1    1   3    3        -0.6170186271        -1.0748821685
   
Does this mean that the instabilities appear within the DFPT 10*10*1 q-points? If not, which output files should I use to check where the instability is coming from?

Unfortunately in the paper, the only information I know about the q-points is that the particle swarm search method is used. I do know that the paper used a larger vacuum layer- can that lead to differences in stability?

Best,
Lily

andyamygto
Posts: 42
Joined: Tue Nov 26, 2019 12:50 am

Re: Negative frequency with denser q-points

Post by andyamygto » Wed Sep 02, 2020 11:39 am

Dear Lily,

You should grep the frequency from the DFPT calculation at each q-point (the abinit output, not the anaddb output)

If you are on the linux you can use command to get the frequency calculated form the DFPT at each point:
$grep -A 3 cm 'your_result'.out

The converge of the vacuum layer should be checked but it should not lead to the instability unless the vacuum is not enough at all.

Regard,
Andy

ebousquet
Posts: 462
Joined: Tue Apr 19, 2011 11:13 am
Location: University of Liege, Belgium

Re: Negative frequency with denser q-points

Post by ebousquet » Wed Sep 02, 2020 12:01 pm

Dear Lily,
As Andy said, could you return the phonon frequencies instead?
A pity that the authors of the paper you are comparing with do not mention the details on how they obtained their results, it is hard to know how to compare then... :(
Best wishes,
Eric

lilymali03
Posts: 10
Joined: Mon Jul 13, 2020 11:30 pm

Re: Negative frequency with denser q-points

Post by lilymali03 » Wed Sep 02, 2020 10:31 pm

Dear Andy and Eric,

Thank you for your responses! I used the grep command and got the following:

Phonon frequencies in cm-1 :
- 5.171502E+00 5.171502E+00 1.165883E+01 3.042359E+02 3.042359E+02
- 3.651107E+02 5.250094E+02 8.804881E+02 8.804881E+02

--
Phonon frequencies in cm-1 :
- 3.260848E+01 8.061505E+01 1.088342E+02 3.076453E+02 3.510386E+02
- 3.774866E+02 5.143386E+02 8.763915E+02 8.771009E+02

--
Phonon frequencies in cm-1 :
- 9.470749E+01 1.219964E+02 1.554897E+02 3.486361E+02 3.649353E+02
- 4.895418E+02 5.027326E+02 8.666628E+02 8.712611E+02

--
Phonon frequencies in cm-1 :
- 1.212077E+02 1.548803E+02 1.753208E+02 3.355376E+02 3.905816E+02
- 4.534831E+02 6.419212E+02 8.414367E+02 8.665279E+02

--
Phonon frequencies in cm-1 :
- 1.258512E+02 1.689430E+02 1.940943E+02 3.377814E+02 4.090744E+02
- 4.166469E+02 7.080554E+02 8.230405E+02 8.672963E+02

--
Phonon frequencies in cm-1 :
- 7.541465E+01 1.118701E+02 1.531357E+02 3.379679E+02 3.698527E+02
- 4.579896E+02 4.956587E+02 8.698876E+02 8.738445E+02

--
Phonon frequencies in cm-1 :
- 1.216694E+02 1.401708E+02 1.676065E+02 3.485624E+02 3.906711E+02
- 4.604479E+02 5.876613E+02 8.481835E+02 8.748657E+02

--
Phonon frequencies in cm-1 :
- 1.085218E+02 1.647846E+02 1.839831E+02 3.369065E+02 4.260230E+02
- 4.308333E+02 6.904883E+02 8.175169E+02 8.759990E+02

--
Phonon frequencies in cm-1 :
- 9.643757E+01 1.578852E+02 1.852651E+02 3.538510E+02 4.282023E+02
- 4.538210E+02 6.721982E+02 8.132348E+02 8.830847E+02

--
Phonon frequencies in cm-1 :
- 4.163840E+01 1.685014E+02 1.743244E+02 3.745825E+02 4.066336E+02
- 4.811379E+02 7.251715E+02 7.763612E+02 8.825645E+02


It looks as though all of the frequencies are positive? (This is for the GGA-PBE pseudopotential. There are just dashes before each line, and not negatives.)

Could the negative frequency that appears in the anaddb graph be related to pseudopotentials or the variable pawxcdev? I retested the material with the PAW LDA pseudopotential instead of the PAW GGA-PBE one and without setting pawxcdev to 0, and the resulting phonon dispersion graph was entirely positive, matching up with the sample paper pretty well.

Thanks,
Lily

ebousquet
Posts: 462
Joined: Tue Apr 19, 2011 11:13 am
Location: University of Liege, Belgium

Re: Negative frequency with denser q-points

Post by ebousquet » Fri Sep 04, 2020 12:13 pm

Dear Lily,
Very good, we are progressing and thank you for having tested the pawxcdev option, it might comes from this?
Before going further, I would like to check if is not an effect of the acoustic sum rule (ASR), could you resend the phonon frequencies of DFPT with and without imposition of the ASR (i.e. obtained with anaddb)? Unless you already imposed asr in your main DFPT abinit calculations?
Best wishes,
Eric

lilymali03
Posts: 10
Joined: Mon Jul 13, 2020 11:30 pm

Re: Negative frequency with denser q-points

Post by lilymali03 » Fri Sep 04, 2020 8:59 pm

Dear Eric,

Thanks for your reply! I have returned the phonon frequencies for the GGA-PBE pseudopotential (with the negative).

Here are the frequencies with asr=0

Phonon frequencies in cm-1 :
- 1.258512E+02 1.689430E+02 1.940943E+02 3.377814E+02 4.090744E+02
- 4.166469E+02 7.080554E+02 8.230405E+02 8.672963E+02

--
Phonon frequencies in cm-1 :
- 1.214913E+02 1.600192E+02 1.815485E+02 3.330796E+02 3.970222E+02
- 4.442422E+02 6.636315E+02 8.359091E+02 8.665076E+02

--
Phonon frequencies in cm-1 :
- 1.125348E+02 1.354878E+02 1.611795E+02 3.529435E+02 3.665408E+02
- 4.770267E+02 5.624854E+02 8.581806E+02 8.686041E+02

--
Phonon frequencies in cm-1 :
- 6.985622E+01 1.080934E+02 1.441224E+02 3.300319E+02 3.726123E+02
- 4.384981E+02 5.022249E+02 8.719217E+02 8.738970E+02
Speed of sound for this q and mode:
--
Phonon frequencies in cm-1 :
- 2.438143E+01 6.795538E+01 9.165766E+01 3.042030E+02 3.320048E+02
- 3.745458E+02 5.175401E+02 8.776147E+02 8.780769E+02
Speed of sound for this q and mode:
--
Phonon frequencies in cm-1 :
- 5.171502E+00 5.171502E+00 1.165883E+01 3.042359E+02 3.042359E+02
- 3.651107E+02 5.250094E+02 8.804881E+02 8.804881E+02

--
Phonon frequencies in cm-1 :
- 2.251980E+01 6.653816E+01 8.857909E+01 3.034168E+02 3.300597E+02
- 3.740019E+02 5.179701E+02 8.779260E+02 8.780736E+02
Speed of sound for this q and mode:
--
Phonon frequencies in cm-1 :
- 6.181333E+01 1.058462E+02 1.446110E+02 3.274505E+02 3.727636E+02
- 4.290211E+02 5.019461E+02 8.723674E+02 8.743050E+02
Speed of sound for this q and mode:
--
Phonon frequencies in cm-1 :
- 1.129898E+02 1.277584E+02 1.651774E+02 3.578334E+02 3.735479E+02
- 4.716817E+02 5.405194E+02 8.576656E+02 8.748765E+02

--
Phonon frequencies in cm-1 :
- 1.180570E+02 1.494336E+02 1.796088E+02 3.482703E+02 4.277434E+02
- 4.413951E+02 6.336763E+02 8.310398E+02 8.808806E+02

--
Phonon frequencies in cm-1 :
- 6.460076E+01 1.632967E+02 1.820689E+02 3.669422E+02 4.156665E+02
- 4.743201E+02 7.037007E+02 7.922540E+02 8.834528E+02

--
Phonon frequencies in cm-1 :
- -3.186430E+01 1.696794E+02 1.696794E+02 3.942322E+02 3.942322E+02
- 4.925033E+02 7.489140E+02 7.489140E+02 8.822390E+02

--
Phonon frequencies in cm-1 :
- 1.635307E+01 1.683392E+02 1.727923E+02 3.822032E+02 4.031154E+02
- 4.873835E+02 7.319554E+02 7.674511E+02 8.825194E+02

--
Phonon frequencies in cm-1 :
- 5.991658E+01 1.687891E+02 1.759237E+02 3.660295E+02 4.096808E+02
- 4.728434E+02 7.194985E+02 7.849571E+02 8.822168E+02

--
Phonon frequencies in cm-1 :
- 8.892646E+01 1.690725E+02 1.799921E+02 3.491082E+02 4.142554E+02
- 4.523380E+02 7.115234E+02 8.005000E+02 8.796383E+02

--
Phonon frequencies in cm-1 :
- 1.097047E+02 1.689618E+02 1.857521E+02 3.386621E+02 4.164419E+02
- 4.312087E+02 7.080257E+02 8.126224E+02 8.747068E+02

--
Phonon frequencies in cm-1 :
- 1.219353E+02 1.689124E+02 1.915698E+02 3.368700E+02 4.151644E+02
- 4.167732E+02 7.077117E+02 8.203454E+02 8.695516E+02

--
Phonon frequencies in cm-1 :
- 1.258512E+02 1.689430E+02 1.940943E+02 3.377814E+02 4.090744E+02
- 4.166469E+02 7.080554E+02 8.230405E+02 8.672963E+02

--
Phonon frequencies in cm-1 :
- 5.171502E+00 5.171502E+00 1.165883E+01 3.042359E+02 3.042359E+02
- 3.651107E+02 5.250094E+02 8.804881E+02 8.804881E+02




And here are the phonon frequencies with asr=2
Phonon frequencies in cm-1 :
- 1.251964E+02 1.688795E+02 1.940405E+02 3.377704E+02 4.090129E+02
- 4.166370E+02 7.080200E+02 8.230098E+02 8.672674E+02

--
Phonon frequencies in cm-1 :
- 1.208238E+02 1.599523E+02 1.814845E+02 3.330673E+02 3.969589E+02
- 4.442338E+02 6.635934E+02 8.358789E+02 8.664786E+02

--
Phonon frequencies in cm-1 :
- 1.119561E+02 1.354083E+02 1.610317E+02 3.529221E+02 3.664725E+02
- 4.770192E+02 5.624395E+02 8.581514E+02 8.685752E+02

--
Phonon frequencies in cm-1 :
- 6.902256E+01 1.079892E+02 1.439496E+02 3.299575E+02 3.725780E+02
- 4.384376E+02 5.022178E+02 8.718929E+02 8.738682E+02
Speed of sound for this q and mode:
--
Phonon frequencies in cm-1 :
- 2.157041E+01 6.777096E+01 9.148636E+01 3.041265E+02 3.319316E+02
- 3.744987E+02 5.175334E+02 8.775861E+02 8.780483E+02
Speed of sound for this q and mode:
--
Phonon frequencies in cm-1 :
- 0.000000E+00 0.000000E+00 0.000000E+00 3.041622E+02 3.041622E+02
- 3.650601E+02 5.250028E+02 8.804596E+02 8.804596E+02

--
Phonon frequencies in cm-1 :
- 1.942402E+01 6.634928E+01 8.840550E+01 3.033402E+02 3.299861E+02
- 3.739547E+02 5.179634E+02 8.778974E+02 8.780449E+02
Speed of sound for this q and mode:
--
Phonon frequencies in cm-1 :
- 6.085218E+01 1.057398E+02 1.444502E+02 3.273756E+02 3.727279E+02
- 4.289588E+02 5.019391E+02 8.723387E+02 8.742762E+02
Speed of sound for this q and mode:
--
Phonon frequencies in cm-1 :
- 1.124231E+02 1.276739E+02 1.650352E+02 3.578083E+02 3.734809E+02
- 4.716742E+02 5.404717E+02 8.576364E+02 8.748476E+02

--
Phonon frequencies in cm-1 :
- 1.173638E+02 1.493623E+02 1.795541E+02 3.482556E+02 4.276846E+02
- 4.413870E+02 6.336363E+02 8.310098E+02 8.808517E+02

--
Phonon frequencies in cm-1 :
- 6.331534E+01 1.632318E+02 1.820115E+02 3.669320E+02 4.156575E+02
- 4.742674E+02 7.036650E+02 7.922225E+02 8.834236E+02

--
Phonon frequencies in cm-1 :
- -3.434673E+01 1.696175E+02 1.696175E+02 3.942230E+02 3.942230E+02
- 4.924522E+02 7.488805E+02 7.488805E+02 8.822095E+02

--
Phonon frequencies in cm-1 :
- 1.014877E+01 1.682765E+02 1.727316E+02 3.821937E+02 4.031064E+02
- 4.873320E+02 7.319211E+02 7.674184E+02 8.824901E+02

--
Phonon frequencies in cm-1 :
- 5.852725E+01 1.687264E+02 1.758640E+02 3.660193E+02 4.096719E+02
- 4.727904E+02 7.194636E+02 7.849251E+02 8.821876E+02

--
Phonon frequencies in cm-1 :
- 8.799580E+01 1.690095E+02 1.799338E+02 3.490972E+02 4.142465E+02
- 4.522828E+02 7.114881E+02 8.004686E+02 8.796092E+02

--
Phonon frequencies in cm-1 :
- 1.089517E+02 1.688985E+02 1.856958E+02 3.386506E+02 4.164328E+02
- 4.311509E+02 7.079903E+02 8.125914E+02 8.746778E+02

--
Phonon frequencies in cm-1 :
- 1.212588E+02 1.688490E+02 1.915153E+02 3.368586E+02 4.151041E+02
- 4.167637E+02 7.076763E+02 8.203146E+02 8.695227E+02

--
Phonon frequencies in cm-1 :
- 1.251964E+02 1.688795E+02 1.940405E+02 3.377704E+02 4.090129E+02
- 4.166370E+02 7.080200E+02 8.230098E+02 8.672674E+02

--
Phonon frequencies in cm-1 :
- 0.000000E+00 0.000000E+00 0.000000E+00 3.041622E+02 3.041622E+02
- 3.650601E+02 5.250028E+02 8.804596E+02 8.804596E+02



The negative seems to occur both with and without the Acoustic Sum Rule.

Best,
Lily

andyamygto
Posts: 42
Joined: Tue Nov 26, 2019 12:50 am

Re: Negative frequency with denser q-points

Post by andyamygto » Fri Sep 04, 2020 11:20 pm

Dear Lily,

Here is few comment and question...

1, which q-point you get the negative frequency?
One thing you can do is to print out the phonon eigenvector to see which kind of vibration cause the negative frequency (with q-vector, it usually hard to distinguish the vibration symmetry, but still worth to give it a try.)
2, Have you do the full relax for each test before calculating the phonon?
If so, have you turn on the pawxcdev=0 when relaxing the atom? (also the variable ixc is important, but usually Abinit will read the psp files)
Last thing, can you can check the force in the first dataset of your phonon calculation? (the highly converged SCF)

Best wishes,
Andy

lilymali03
Posts: 10
Joined: Mon Jul 13, 2020 11:30 pm

Re: Negative frequency with denser q-points

Post by lilymali03 » Sat Sep 05, 2020 5:22 am

Dear Andy,

Thanks for your reply! I redid the relaxation with the variable pawxcdev, but this actually seemed to make the negative frequency worse with the GGA-PBE...

With the new relaxation, I got the negative frequencies with the first and last q-points (0.00000000E+00 0.00000000E+00 0.00000000E+00 and -3.75000000E-01 2.50000000E-01 0.00000000E+00).

At the end of the first dataset, residm=1.014E-04 and nres2=2.70E-11.

Best,
Lily

andyamygto
Posts: 42
Joined: Tue Nov 26, 2019 12:50 am

Re: Negative frequency with denser q-points

Post by andyamygto » Sat Sep 05, 2020 9:10 pm

Dear Lily,

If the negative frequency is happened at gamma point and you have tested most the converge variables, It might mean that the system is not stable under GGA.
The force I mean is that the force acting on each atom. In particular case, reading the force may find something interesting.
Have you check the vibration of unstable mode?

Regards,
Andy

lilymali03
Posts: 10
Joined: Mon Jul 13, 2020 11:30 pm

Re: Negative frequency with denser q-points

Post by lilymali03 » Wed Sep 09, 2020 8:38 pm

Dear Andy,

How can I check the vibration of the unstable mode and the forces of the atoms?

Just for clarification- I realized that originally both NbB2 and TaB2 had negative frequencies, but the paper said that both were positive. Transitioning from GGA-PBE to LDA helped NbB2 become positive, but TaB2 unfortunately remains negative. For the most part, the LDA matches up with the graph from the paper, but it is slightly different in height since it is higher at the top.

Thanks,
Lily

andyamygto
Posts: 42
Joined: Tue Nov 26, 2019 12:50 am

Re: Negative frequency with denser q-points

Post by andyamygto » Sun Sep 13, 2020 6:03 pm

Dear Lily

you can grep the eigenvector with the corresponding mode and use xcrysden or VESTA to check the vibration mode.

In my experiment, LDA usually perform better with DFPT method (I have no idea why is that a case...)
So it's not surprising that you said changing to LDA will fix the negative frequencies.
If you have time, I suggest you can try to use norm conserving psp to test if that makes different with PAW.

Regards,
Andy

Post Reply