ABINIT Discussion Forums

I try to perform a phonon response function calculation.

I have optimised an oxygen atom adsorbed in an fcc hollow site on Pd(111).

I have added ``rfphon 1`` to the original input file, that I have used for the geometry optimisation.

Upon running the job, I received the following error msg:

--- !ERROR
src_file: m_pawxc.F90
src_line: 3792
mpi_rank: 0
message: |
Kxc not implemented for GGA!

What is Kxc standing for?
How can I determine the curvature of the PES at the optimised state? Do I need to perform a so called frozen phonon calculation instead?

This actually has been implemented. A new release has come out recently with gga paw phonons properly activated. Which version are you using?

Please check the release notes for the last public releases eg
https://docs.abinit.org/about/release-notes/

Dear Willem,
Yes, as Matthieu said, you need to run with the latest production version 8.10.1 and do your calculation using pawxcdev=0 for all your calculations (even the relaxation and ground state wave function to be consistent with the DFPT run).
Best wishes,
Eric

I got the phonon response function calculation running. However the calculated energies are abnormal:

$ tailf pd.txt
ETOT 13 -1.10894776896575E+18 -8.517E+17 1.273E+13 3.928E+29
ETOT 14 -1.79594291525765E+17 9.294E+17 4.354E+12 9.097E+28
ETOT 15 -1.34645008754373E+17 4.495E+16 2.297E+12 5.171E+28
ETOT 16 -2.07852196185765E+17 -7.321E+16 2.574E+12 7.382E+28
ETOT 17 -3.03775697793423E+17 -9.592E+16 3.593E+12 1.053E+29
ETOT 18 -1.69399188349057E+22 -1.694E+22 2.890E+17 5.498E+33
ETOT 19 -4.49340469541107E+21 1.245E+22 5.239E+17 1.732E+33
ETOT 20 -3.21193388101716E+21 1.281E+21 6.804E+16 1.366E+33
ETOT 21 -7.29917957588542E+21 -4.087E+21 2.237E+17 2.393E+33
ETOT 22 -1.47941452784392E+21 5.820E+21 1.885E+17 7.605E+32

The energies of the optimised structure are in a normal range:
$ grep ETOT ../pd.txt | tail
ETOT 20 -2608.4655727803 -4.108E-07 6.692E-09 1.274E-05
ETOT 21 -2608.4655728298 -4.943E-08 8.062E-09 3.655E-05
ETOT 22 -2608.4655725701 2.597E-07 1.939E-08 3.099E-05
ETOT 23 -2608.4655725566 1.344E-08 2.023E-08 1.430E-05
ETOT 24 -2608.4655722645 2.922E-07 7.267E-08 1.567E-05
ETOT 25 -2608.4655722634 1.044E-09 2.277E-08 1.850E-05
ETOT 26 -2608.4655727850 -5.216E-07 9.554E-09 3.586E-06
ETOT 27 -2608.4655728824 -9.732E-08 4.839E-09 3.097E-06
ETOT 28 -2608.4655729131 -3.074E-08 5.139E-09 2.865E-06
ETOT 29 -2608.4655728996 1.352E-08 1.025E-08 1.525E-06

Is this a bug?

I will include my input files:

Dear Willem,
This looks more like a convergence problem, even the ground state does not look converging very nicely. You might have to first play with diemix (and diemixmag is magnetization is there) flags, nline>4 might helps too.
Best wishes,
Eric

Dear Eric and ABINIT friends,

Please keep in mind that I did not change other parameters than ``rfphon 1`` from ground state calculation (ion relaxation) to phonon response function calculation.
The total energies in the former case seem to be reasonable. This is in contrast to the latter case, where the total energies are huge.

Changing diemix to 0.5 or 0.3 does not lead to success either. The energy values remain huge. I will play with nline as well. However, considering the above, I don't have the impression it is related to the parameter settings.

Dear Eric and ABINIT friends,

setting nline = 6 did not improve the outcome.

Energy values remain unrealistically huge:

-ETOT 1 -723764921.44809 -7.237E+08 7.763E+03 2.282E+20
ETOT 2 -1.07565194714471E+14 -1.076E+14 1.053E+09 3.463E+25
ETOT 3 -9911228406409.3 9.765E+13 6.786E+08 3.775E+24
ETOT 4 -10276967357971. -3.657E+11 2.502E+08 3.476E+24
ETOT 5 -4898544636441.6 5.378E+12 2.955E+08 1.978E+24
ETOT 6 -2408109164706.8 2.490E+12 7.175E+07 1.137E+24
ETOT 7 -1070692092453.3 1.337E+12 1.728E+07 4.669E+23

Is there something wrong? Am I doing something wrong?

Dear Willem,
Could you define all the rf flags as follows to be sure everything is set up correctly?
kptopt 2
rfphon 1
rfatpol 1 22
iscf 7
tolvrs 1.0E-16
rfdir 1 1 1
nqpt 1
qpt 0.0 0.0 0.0

(and remove the other input flags that make conflict with these ones).
Best wishes,
Eric

Dear Eric and ABINIT friends,

I have tried the suggestions/additions/corrections by Eric, but unfortunately without success.

$ tail -f pd.txt.save

dfpt_looppert : total number of electrons, from k and k+q
fully or partially occupied states are 3.720000E+02 and 3.720000E+02.
Initialisation of the first-order wave-functions :
ireadwf= 0

iter 2DEtotal(Ha) deltaE(Ha) residm vres2
-ETOT 1 -720166411.92093 -7.201E+08 9.826E+03 2.270E+20
ETOT 2 -5.84942149240457E+14 -5.849E+14 9.214E+09 1.883E+26
ETOT 3 -55356105259480. 5.296E+14 3.846E+09 3.426E+25

The energy values remain out of order.

In the log files, I detect that the method has problems with minimisation:

<snip>
...
dfpt_cgwf: problem of minimisation (likely metallic), set resid to -1
...
</snip>

resid seems not to be a tuneable parameter. But specifying resid -1 doesn't help either.


If only someone would take this seriously and would be able to repeat this calculation, then we might come to a solution or conclusion.

I will upload the input files once more:

Dear Marc, Gregory and ABINIT friends,

At the ABINIT 2019 school, I got the advice to converge the TOLerance on WaveFunction squared Residual to less than 1.00E-20 for the SCF calculation.

So I tried this as well for the SCF calculation.

The following gives the convergence of the SCF calculation:


iter Etot(hartree) deltaE(h) residm nres2
ETOT 1 -2608.4705287159 -2.608E+03 5.029E-06 2.283E+00
ETOT 2 -2608.5487563296 -7.823E-02 1.400E-03 4.747E+01
ETOT 3 -2608.3320200691 2.167E-01 5.286E-03 9.052E+00
ETOT 4 -2608.4569402452 -1.249E-01 1.211E-04 2.795E+00
ETOT 5 -2608.4538799825 3.060E-03 3.720E-04 1.005E+00
ETOT 6 -2608.4638028707 -9.923E-03 5.276E-05 4.802E-01
ETOT 7 -2608.4641540612 -3.512E-04 6.815E-05 1.855E-01
ETOT 8 -2608.4653181086 -1.164E-03 4.048E-05 1.388E-01
ETOT 9 -2608.4652694828 4.863E-05 3.602E-05 4.793E-02
ETOT 10 -2608.4651690944 1.004E-04 4.902E-05 1.553E-02
ETOT 11 -2608.4651193789 4.972E-05 2.461E-05 1.306E-02
ETOT 12 -2608.4653796471 -2.603E-04 6.650E-06 5.935E-03
ETOT 13 -2608.4654827014 -1.031E-04 3.795E-07 1.576E-03
...
...
ETOT 257 -2608.4655740486 -1.319E-11 5.584E-20 3.752E-19
ETOT 258 -2608.4655740485 5.821E-11 7.263E-20 1.143E-19
ETOT 259 -2608.4655740485 4.275E-11 4.560E-20 7.897E-20
ETOT 260 -2608.4655740485 -1.637E-11 5.862E-20 1.549E-19
ETOT 261 -2608.4655740485 -1.182E-11 3.535E-20 1.769E-19
ETOT 262 -2608.4655740485 4.275E-11 4.582E-20 1.974E-19
ETOT 263 -2608.4655740485 -6.366E-12 2.778E-20 1.812E-19
ETOT 264 -2608.4655740486 -6.457E-11 3.567E-20 1.329E-19
ETOT 265 -2608.4655740485 5.002E-11 2.189E-20 9.698E-20
ETOT 266 -2608.4655740485 -3.638E-12 2.836E-20 9.087E-20
ETOT 267 -2608.4655740485 -2.410E-11 1.718E-20 7.865E-20
ETOT 268 -2608.4655740485 2.547E-11 2.221E-20 5.584E-20
ETOT 269 -2608.4655740485 5.775E-11 1.352E-20 3.027E-20
ETOT 270 -2608.4655740485 -4.775E-11 1.764E-20 2.537E-20
ETOT 271 -2608.4655740484 5.002E-11 1.061E-20 1.136E-20
ETOT 272 -2608.4655740485 -3.320E-11 1.393E-20 5.256E-21
ETOT 273 -2608.4655740486 -6.730E-11 8.341E-21 1.883E-21

At SCF step 273 max residual= 8.34E-21 < tolwfr= 1.00E-20 =>converged.


However, after I started the phonon calculation, I got huge energy values. The energy values remain out of order:

iter 2DEtotal(Ha) deltaE(Ha) residm vres2
-ETOT 1 -993045758.44554 -9.930E+08 9.055E+03 3.131E+20
ETOT 2 -1.02127820448525E+15 -1.021E+15 1.185E+10 3.275E+26
ETOT 3 -86884493884229. 9.344E+14 5.148E+09 5.781E+25
ETOT 4 -64412214900268. 2.247E+13 3.607E+09 2.509E+25
ETOT 5 -32852224002200. 3.156E+13 2.207E+09 1.711E+25
ETOT 6 -20661258030968. 1.219E+13 3.221E+08 7.897E+24
ETOT 7 -14125323951442. 6.536E+12 1.321E+08 3.527E+24
ETOT 8 -11651565598112. 2.474E+12 4.683E+07 2.746E+24
ETOT 9 -10674579539611. 9.770E+11 3.054E+07 2.596E+24
ETOT 10 -4.99135469269484E+16 -4.990E+16 6.725E+11 1.581E+28
ETOT 11 -5.06096314051697E+17 -4.562E+17 1.300E+13 1.619E+29
ETOT 12 -2.55189952388683E+17 2.509E+17 9.062E+12 8.634E+28
ETOT 13 -1.06812504078632E+18 -8.129E+17 1.268E+13 3.794E+29
ETOT 14 -1.72926059951867E+17 8.952E+17 4.116E+12 8.743E+28
ETOT 15 -1.45965043204564E+17 2.696E+16 2.253E+12 5.490E+28
ETOT 16 -2.18975725143766E+17 -7.301E+16 2.409E+12 7.771E+28
ETOT 17 -2.84679911310223E+17 -6.570E+16 3.858E+12 9.953E+28
ETOT 18 -1.71864403083607E+22 -1.719E+22 2.863E+17 5.576E+33
ETOT 19 -4.81483599895236E+21 1.237E+22 5.593E+17 1.817E+33
ETOT 20 -3.16144880483277E+21 1.653E+21 6.918E+16 1.369E+33
ETOT 21 -7.68522378335802E+21 -4.524E+21 2.358E+17 2.516E+33

The energy values of the phonon calculation are not close to the corresponding values of the SCF calculation. This cannot be!!!


I'm very willing to provide the input files, if someone wants to take serious look at it.

The unreasonable energy values seem to be due to the PAW data set.

Gregory is currently working on/testing it.

I hope to continue this soon.