[QE-users] sternheimer_kernel: root not converged in phonon calculations
pboulet
pascal.boulet at univ-amu.fr
Sun Aug 28 18:14:21 CEST 2022
> Le 28 août 2022 à 17:29, Lorenzo Paulatto <lorenzo.paulatto at cnrs.fr> a écrit :
>
> It seems to converge just fine on my desktop pc.
>
>
All right, it could be that Irene Rome is doing wrong…
> I see that your outdir starts with /ccc, are you by any chance running on the CEA Irene Rome AMD machine?
>
>
Yes. I also have been struggling a lot to make QE 6.x and 7.x work on Irene Rome, but it seems now to run without erratic/random crashes as I compiled it with intelMPI. One thing I did is changing -O3 optimization to -O2 one in the make.inc file. Anyhow, openMPI is definitely not working with QE on Irene, I don’t know the reason.
Yet, if you tell me that the job runs fine on your PC, then perhaps Irene Rome is still doing weird things…
> I had all kind of random bad results on it, compiling with eithe ifort or nvfort (gfortranwithout mkl was fine but slow), to the point that I just gave and only use Idris since.
>
>
I see, so I shall move to other machines, then.
Thank you for your help and time,
Best
Pascal
> hth
>
> On 28/08/2022 16:06, pboulet wrote:
>> Bonjour Lorenzo,
>>
>> Thank you for your help.
>> The problem appears right at q-point #2 and representation #1.
>>
>> I have put an archive of the most important files at this address (that hopefully help to reproduce/restart the calculations):
>> https://filesender.renater.fr/?s=download&token=759fb88f-1374-4029-bcb4-73d7c23d7975 <https://filesender.renater.fr/?s=download&token=759fb88f-1374-4029-bcb4-73d7c23d7975>
>>
>> Here is also a more detailed summary of the output file from ph.x.
>>
>> At the beginning of the output:
>> >>>>>
>> Reading input from Mg2Si.in <http://mg2si.in/>
>>
>> Reading xml data from directory:
>>
>> ./Mg2Si.save/
>>
>> IMPORTANT: XC functional enforced from input :
>> Exchange-correlation= PBE
>> ( 1 4 3 4 0 0 0)
>> Any further DFT definition will be discarded
>> Please, verify this is what you really want
>>
>>
>> Parallelization info
>> --------------------
>> sticks: dense smooth PW G-vecs: dense smooth PW
>> Min 41 41 12 1225 1225 192
>> Max 42 42 13 1229 1229 195
>> Sum 1993 1993 583 58887 58887 9257
>>
>> Using Slab Decomposition
>>
>>
>> negative rho (up, down): 1.928E+00 0.000E+00
>> Reading collected, re-writing distributed wavefunctions
>>
>> Saving dvscf to file. Distribute only q points, not irreducible represetations.
>> 7 / 8 q-points for this run, from 2 to 8:
>> N xq(1) xq(2) xq(3)
>> 1 0.000000000 0.000000000 0.000000000
>> 2 -0.250000000 0.250000000 -0.250000000
>> 3 0.500000000 -0.500000000 0.500000000
>> 4 0.000000000 0.500000000 0.000000000
>> 5 0.750000000 -0.250000000 0.750000000
>> 6 0.500000000 0.000000000 0.500000000
>> 7 0.000000000 -1.000000000 0.000000000
>> 8 -0.500000000 -1.000000000 0.000000000
>>
>>
>> Calculation of q = -0.2500000 0.2500000 -0.2500000
>>
>> Bands found: reading from ./_ph0/Mg2Si.q_2/
>>
>> Reading xml data from directory:
>>
>> ./_ph0/Mg2Si.q_2/Mg2Si.save/
>> Message from routine qexsd_readschema :
>> input info not found or not readable in xml file
>>
>> IMPORTANT: XC functional enforced from input :
>> Exchange-correlation= PBE
>> ( 1 4 3 4 0 0 0)
>> Any further DFT definition will be discarded
>> Please, verify this is what you really want
>>
>> <<<<<<<<<<
>>
>> Later in the output, that’s the q-point #2, where ph.x fails:
>> >>>>>
>> Computing dynamical matrix for
>> q = ( -0.2500000 0.2500000 -0.2500000 )
>>
>> 6 Sym.Ops. (no q -> -q+G )
>> <<<<<
>>
>> Then, note that there is a lot of k-points for this q-point…
>> >>>>>>>>
>> G cutoff = 1468.0362 ( 1226 G-vectors) FFT grid: ( 60, 60, 60)
>> number of k points= 816
>> <<<<<<<<
>>
>>
>> For the representations:
>>
>> >>>>>>>>
>> Representation 1 1 modes - To be done
>> Irreps are as follows:
>>
>> mode # 1
>> ( -0.57735 0.00000 )
>> ( 0.57735 -0.00000 )
>> ( -0.57735 -0.00000 )
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>>
>> Representation 2 2 modes - To be done
>> Irreps are as follows:
>>
>> mode # 2 mode # 3
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( -0.09481 0.00000 ) ( -0.81097 0.00000 )
>> ( 0.65492 0.00000 ) ( -0.48760 -0.00000 )
>> ( 0.74973 0.00000 ) ( 0.32338 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>>
>> Representation 3 1 modes - To be done
>> Irreps are as follows:
>>
>> mode # 4
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>> ( -0.57735 0.00000 )
>> ( 0.57735 0.00000 )
>> ( -0.57735 -0.00000 )
>>
>> Representation 4 2 modes - To be done
>> Irreps are as follows:
>>
>> mode # 5 mode # 6
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( -0.53501 0.00000 ) ( -0.61679 0.00000 )
>> ( -0.31285 0.53223 ) ( -0.26906 -0.46166 )
>> ( 0.22215 0.53223 ) ( 0.34773 -0.46166 )
>>
>> Representation 5 1 modes - To be done
>> Irreps are as follows:
>>
>> mode # 7
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>> ( -0.57735 0.00000 )
>> ( 0.57735 0.00000 )
>> ( -0.57735 -0.00000 )
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 )
>>
>> Representation 6 2 modes - To be done
>> Irreps are as follows:
>>
>> mode # 8 mode # 9
>> ( -0.40825 0.00000 ) ( -0.70711 0.00000 )
>> ( -0.81648 0.00399 ) ( -0.00001 -0.00230 )
>> ( -0.40824 0.00399 ) ( 0.70710 -0.00230 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>> ( 0.00000 0.00000 ) ( 0.00000 0.00000 )
>>
>>
>> PHONON : 7.65s CPU 11.31s WALL
>>
>>
>>
>> Representation # 1 mode # 1
>>
>> Self-consistent Calculation
>> kpoint 5 sternheimer_kernel: root not converged, thresh < 2.227E+03
>> kpoint 6 sternheimer_kernel: root not converged, thresh < 1.393E+11
>> kpoint 13 sternheimer_kernel: root not converged, thresh < 1.425E+04
>> kpoint 14 sternheimer_kernel: root not converged, thresh < 1.223E+11
>> kpoint 21 sternheimer_kernel: root not converged, thresh < 1.368E+10
>> …
>> kpoint 198 sternheimer_kernel: root not converged, thresh < 2.098E+11
>> kpoint 200 sternheimer_kernel: root not converged, thresh < 4.597E+03
>> kpoint 203 sternheimer_kernel: root not converged, thresh < 3.231E+11
>>
>> iter # 17 total cpu time : 77.1 secs av.it <http://av.it/>.: 35.1
>> thresh= 1.000E-02 alpha_mix = 0.100 |ddv_scf|^2 = 2.286E+37
>> kpoint 5 sternheimer_kernel: root not converged, thresh < 7.024E-01
>> kpoint 6 sternheimer_kernel: root not converged, thresh < 5.781E+13
>> kpoint 7 sternheimer_kernel: root not converged, thresh < 2.895E+07
>> kpoint 12 sternheimer_kernel: root not converged, thresh < 3.097E+06
>> …
>> iter # 18 total cpu time : 212.6 secs av.it <http://av.it/>.: 83.3
>> thresh= 1.000E-02 alpha_mix = 0.100 |ddv_scf|^2 = 4.537E+41
>> kpoint 5 sternheimer_kernel: root not converged, thresh < 2.236E+14
>> kpoint 6 sternheimer_kernel: root not converged, thresh < 3.750E+00
>> kpoint 7 sternheimer_kernel: root not converged, thresh < 7.790E+14
>> kpoint 13 sternheimer_kernel: root not converged, thresh < 3.059E+10
>> kpoint 14 sternheimer_kernel: root not converged, thresh < 1.581E+18
>> kpoint 21 sternheimer_kernel: root not converged, thresh < 3.527E+19
>> …
>> <<<<<<<<<<<
>>
>> If more information is needed, please ask me.
>>
>> Best,
>> Pascal
>>
>>
>>
>>> Le 28 août 2022 à 10:21, Lorenzo Paulatto <lorenzo.paulatto at cnrs.fr <mailto:lorenzo.paulatto at cnrs.fr>> a écrit :
>>>
>>> Bonjour Pascal,
>>>
>>> on which representation of which q-point does this happen?
>>>
>>> The first rep of q-point 2 seems to converge perfectly fine on my PC.
>>>
>>> I'd try to reproduce, but I would rather avoid repeating all the points.
>>>
>>> cheers
>>>
>>> On 27/08/2022 17:57, pboulet wrote:
>>>> Dear all,
>>>>
>>>> I am encountering the problem mentioned in the subject of this post.
>>>>
>>>> Do you know to solve this problem?
>>>>
>>>> I am running ph.x on Mg2Si (not a big deal!) with NC pseudopotentials from dojo website.
>>>> I have optimized the structure, then ran a SCF with 60 k-points.
>>>>
>>>> Here is the input for ph.x:
>>>> Mg2Si Phonon
>>>> &inputph
>>>> verbosity = 'high',
>>>> alpha_mix(1)= 0.1,
>>>> tr2_ph = 1.0d-12,
>>>> diagonalization = 'cg',
>>>> prefix = 'Mg2Si',
>>>> ldisp = .true.,
>>>> epsil = .true.,
>>>> lqdir = .true.,
>>>> search_sym = .false.
>>>> fildyn = 'Mg2Si.dyn.xml',
>>>> fildvscf = 'dvscf'
>>>> recover = .true.,
>>>> start_q = 2, last_q = 8,
>>>> nq1=4, nq2=4, nq3=4,
>>>> /
>>>>
>>>> and a sample of the output:
>>>> iter # 41 total cpu time : 5901.1 secs av.it <http://av.it/>.: 210.4
>>>> thresh= 1.000E-02 alpha_mix = 0.100 |ddv_scf|^2 = 1.350+100
>>>> kpoint 3 sternheimer_kernel: root not converged, thresh < 3.699E+06
>>>> kpoint 4 sternheimer_kernel: root not converged, thresh < 2.570E+03
>>>> kpoint 5 sternheimer_kernel: root not converged, thresh < 1.322E+43
>>>> kpoint 6 sternheimer_kernel: root not converged, thresh < 4.836E+01
>>>> kpoint 7 sternheimer_kernel: root not converged, thresh < 2.110E-01
>>>> kpoint 12 sternheimer_kernel: root not converged, thresh < 2.237E+25
>>>> kpoint 13 sternheimer_kernel: root not converged, thresh < 1.420E+01
>>>> kpoint 14 sternheimer_kernel: root not converged, thresh < 1.199E+21
>>>> kpoint 15 sternheimer_kernel: root not converged, thresh < 6.741E+07
>>>> kpoint 18 sternheimer_kernel: root not converged, thresh < 8.917E+00
>>>>
>>>> Note that the calculation ran fine for the Gamma point, so in the present run I ‘recovered’ for k-points 2 to 8.
>>>>
>>>> The input file for the SCF calculation is:
>>>> &CONTROL
>>>> title = 'Mg2Si',
>>>> calculation = 'scf',
>>>> etot_conv_thr = 1d-9,
>>>> tprnfor = .true.,
>>>> tstress = .true.,
>>>> restart_mode = 'restart',
>>>> verbosity = 'high',
>>>> prefix = 'Mg2Si',
>>>> wf_collect = .true.,
>>>> /
>>>> &SYSTEM
>>>> ibrav = 2,
>>>> celldm(1) = 12.037d0,
>>>> nat = 3,
>>>> ntyp = 2,
>>>> ecutwfc = 100.d0,
>>>> ecutrho = 400.d0,
>>>> occupations = 'fixed',
>>>> nbnd = 20,
>>>> /
>>>> &ELECTRONS
>>>> conv_thr = 1d-12,
>>>> mixing_beta = 0.2d0,
>>>> /
>>>> ATOMIC_SPECIES
>>>> Mg 24.3 Mg.upf
>>>> Si 28.1 Si.upf
>>>>
>>>> ATOMIC_POSITIONS crystal
>>>> Si 0.000000000 0.000000000 0.000000000
>>>> Mg 0.250000000 0.250000000 0.250000000
>>>> Mg 0.750000000 0.750000000 0.750000000
>>>>
>>>> K_POINTS automatic
>>>> 8 8 8 1 1 1
>>>>
>>>> Thank you for your help and time.
>>>> Best regards,
>>>> Pascal
>>>>
>>>> .
>>>> Pascal Boulet
>>>> —
>>>> Professor in computational materials chemistry - DEPARTMENT OF CHEMISTRY
>>>> University of Aix-Marseille - Avenue Escadrille Normandie Niemen - F-13013 Marseille - FRANCE
>>>> Tél: +33(0)4 13 55 18 10 - Fax : +33(0)4 13 55 18 50
>>>> Email : pascal.boulet at univ-amu.fr <mailto:pascal.boulet at univ-amu.fr>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> The Quantum ESPRESSO community stands by the Ukrainian
>>>> people and expresses its concerns about the devastating
>>>> effects that the Russian military offensive has on their
>>>> country and on the free and peaceful scientific, cultural,
>>>> and economic cooperation amongst peoples
>>>> _______________________________________________
>>>> Quantum ESPRESSO is supported by MaX (www.max-centre.eu <http://www.max-centre.eu/>)
>>>> users mailing list users at lists.quantum-espresso.org <mailto:users at lists.quantum-espresso.org>
>>>> https://lists.quantum-espresso.org/mailman/listinfo/users <https://lists.quantum-espresso.org/mailman/listinfo/users>--
>>> Dr. Lorenzo Paulatto
>>> IdR @ IMPMC - CNRS UMR 7590 & Sorbonne Université
>>> phone: +33 (0)1 442 79822 / skype: paulatz
>>> http://www.impmc.upmc.fr/~paulatto/ <http://www.impmc.upmc.fr/~paulatto/> - https://anharmonic.github.io/ <https://anharmonic.github.io/>
>>> 23-24/423 B115, 4 place Jussieu 75252 Paris CX 05
>>> _______________________________________________
>>> The Quantum ESPRESSO community stands by the Ukrainian
>>> people and expresses its concerns about the devastating
>>> effects that the Russian military offensive has on their
>>> country and on the free and peaceful scientific, cultural,
>>> and economic cooperation amongst peoples
>>> _______________________________________________
>>> Quantum ESPRESSO is supported by MaX (www.max-centre.eu <http://www.max-centre.eu/>)
>>> users mailing list users at lists.quantum-espresso.org <mailto:users at lists.quantum-espresso.org>
>>> https://lists.quantum-espresso.org/mailman/listinfo/users <https://lists.quantum-espresso.org/mailman/listinfo/users>
>>
>>
>>
>> _______________________________________________
>> The Quantum ESPRESSO community stands by the Ukrainian
>> people and expresses its concerns about the devastating
>> effects that the Russian military offensive has on their
>> country and on the free and peaceful scientific, cultural,
>> and economic cooperation amongst peoples
>> _______________________________________________
>> Quantum ESPRESSO is supported by MaX (www.max-centre.eu <http://www.max-centre.eu/>)
>> users mailing list users at lists.quantum-espresso.org <mailto:users at lists.quantum-espresso.org>
>> https://lists.quantum-espresso.org/mailman/listinfo/users <https://lists.quantum-espresso.org/mailman/listinfo/users>--
> Dr. Lorenzo Paulatto
> IdR @ IMPMC - CNRS UMR 7590 & Sorbonne Université
> phone: +33 (0)1 442 79822 / skype: paulatz
> http://www.impmc.upmc.fr/~paulatto/ <http://www.impmc.upmc.fr/~paulatto/> - https://anharmonic.github.io/ <https://anharmonic.github.io/>
> 23-24/423 B115, 4 place Jussieu 75252 Paris CX 05
> _______________________________________________
> The Quantum ESPRESSO community stands by the Ukrainian
> people and expresses its concerns about the devastating
> effects that the Russian military offensive has on their
> country and on the free and peaceful scientific, cultural,
> and economic cooperation amongst peoples
> _______________________________________________
> 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
Pascal Boulet
—
Professor in computational materials chemistry - DEPARTMENT OF CHEMISTRY
University of Aix-Marseille - Avenue Escadrille Normandie Niemen - F-13013 Marseille - FRANCE
Tél: +33(0)4 13 55 18 10 - Fax : +33(0)4 13 55 18 50
Email : pascal.boulet at univ-amu.fr <mailto:pascal.boulet at univ-amu.fr>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20220828/8afcdd07/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1269 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20220828/8afcdd07/attachment.p7s>
More information about the users
mailing list