[Pw_forum] phonon calculations do not converge
Hello Matt, Can I introduce some broadening, keeping in mind this structure is semiconductor no metallic? Thanks IYAD On Wed, Aug 15, 2012 at 1:38 PM, Matthijs Mentink wrote: > Hi Iyad, > > Have you considered introducing some broadening in the pwscf calculation? > It helps. Be careful not to go overboard though. You could try introducing > some broadening (like 1e-3 Ry or so) and see if it converges. If that works > than you can see how far you can reduce broadening while still having the > phonon calculations converge. > > cheers, > > Matthijs > > > On Wed, Aug 15, 2012 at 10:27 AM, Iyad AL-QASIR wrote: > >> Hi Wiliam, >> >> I used tr2_ph = 1.0d-10, it did not converge either after 100 steps. The >> |ddv_scf|^ values are about ~ E-07. >> >> However, when I used tr2_ph = 1.0d-14, the |ddv_scf|^ values are >> diverging strongly ~E+20 and even more. >> >> Kindest Regards, >> >> >> >> >> On Fri, Aug 10, 2012 at 5:16 PM, William Parker wrote: >> >>> Hi Iyad, >>> >>> How strong is the failure to converge at (0, 1/6, 1/6)? Perhaps a less >>> tight tr2_ph value of, say, 1.0d-10 would get you where you need to be. >>> >>> --William >>> >>> On Aug 10, 2012, at 12:29 PM, Iyad AL-QASIR wrote: >>> >>> > Dear PW_Forum, >>> > >>> > I am running phonon calculations for FeSi (a narrow band gap >>> semiconductor) using QE-5.0. I am using a 6x6x6 q-vectors mesh. The first 4 >>> q-points ran successfully and got good phonon values. However when I reach >>> the 5th point(0 1/6 1/6). The self-consistent calculations do not converge. >>> I tried to reduce alpha_mix(1) to 0.3 but still have the same problem. >>> > >>> > Below are my input files: >>> > >>> > >>> > calculation='scf', >>> > restart_mode='from_scratch', >>> > title='FeSi', >>> > outdir='./', >>> > prefix='FeSi', >>> > pseudo_dir='./', >>> > / >>> > >>> > ibrav=1, >>> > celldm(1)=8.4165, >>> > nat=8, ntyp=2, >>> > ecutwfc=60.0, >>> > ecutrho= 720.0, >>> > / >>> > >>> > conv_thr=1.0d-9, >>> > mixing_mode='plain', >>> > mixing_beta=0.7, >>> > diagonalization='david', >>> > / >>> > ATOMIC_SPECIES >>> > Fe 55.847 Fe.pw91-sp-van_ak.UPF >>> > Si 28.086 Si.pw91-n_van.UPF >>> > ATOMIC_POSITIONS >>> > Fe 0.134855265 0.134855265 0.134855265 >>> > Fe 0.634855265 0.365144735 0.865144735 >>> > Fe 0.865144735 0.634855265 0.365144735 >>> > Fe 0.365144735 0.865144735 0.634855265 >>> > Si 0.839976366 0.839976366 0.839976366 >>> > Si 0.339976366 0.660023634 0.160023634 >>> > Si 0.160023634 0.339976366 0.660023634 >>> > Si 0.660023634 0.160023634 0.339976366 >>> > K_POINTS {automatic} >>> > 10 10 10 0 0 0 >>> > *** >>> > phonons of FeSi >>> > >>> > tr2_ph=1.0d-14, >>> > prefix='FeSi', >>> > amass(1)=55.847, >>> > amass(2)=28.086, >>> > outdir='./' >>> > fildyn='FeSi.dyn', >>> > trans=.true., >>> > ldisp=.true., >>> > nq1=6, nq2=6, nq3=6 >>> > recover=.true., >>> > alpha_mix(1)=0.30 >>> > >>> > Any comments and suggestions are welcomed. >>> > >>> > Thanks >>> > IYAD >>> > >>> > >>> > ___ >>> > Pw_forum mailing list >>> > Pw_forum at pwscf.org >>> > http://www.democritos.it/mailman/listinfo/pw_forum >>> >>> * >>> William D. Parker phone: (630) 252-4834 >>> Computational Postdoctoral Fellow fax: (630) 252-4798 >>> MSD-212, Rm. C-215 >>> Argonne National Laboratory >>> 9700 S. Cass Ave. >>> Argonne, IL 60439 >>> ***** >>> >>> >>> >>> >>> >>> ___ >>> Pw_forum mailing list >>> Pw_forum at pwscf.org >>> http://www.democritos.it/mailman/listinfo/pw_forum >>> >> >> >> >> -- >> ___ >> IYAD I. AL-QASIR, PhD >> Postdoc Research Associate >> >> X-Ray and Neutron Scattering and Spectroscopy Group >> Materials Science and Technology Division >> Oak Ridge National Lab >> Oak Ridge, TN >> >> >> ___ >> Pw_forum mailing list >> Pw_forum at pwscf.org >> http://www.democritos.it/mailman/listinfo/pw_forum >> >> > > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > > -- ___ IYAD I. AL-QASIR, PhD Postdoc Research Associate X-Ray and Neutron Scattering and Spectroscopy Group Materials Science and Technology Division Oak Ridge National Lab Oak Ridge, TN -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20120815/1711e896/attachment-0001.htm
[Pw_forum] phonon calculations do not converge
Hi Wiliam, I used tr2_ph = 1.0d-10, it did not converge either after 100 steps. The |ddv_scf|^ values are about ~ E-07. However, when I used tr2_ph = 1.0d-14, the |ddv_scf|^ values are diverging strongly ~E+20 and even more. Kindest Regards, On Fri, Aug 10, 2012 at 5:16 PM, William Parker wrote: > Hi Iyad, > > How strong is the failure to converge at (0, 1/6, 1/6)? Perhaps a less > tight tr2_ph value of, say, 1.0d-10 would get you where you need to be. > > --William > > On Aug 10, 2012, at 12:29 PM, Iyad AL-QASIR wrote: > > > Dear PW_Forum, > > > > I am running phonon calculations for FeSi (a narrow band gap > semiconductor) using QE-5.0. I am using a 6x6x6 q-vectors mesh. The first 4 > q-points ran successfully and got good phonon values. However when I reach > the 5th point(0 1/6 1/6). The self-consistent calculations do not converge. > I tried to reduce alpha_mix(1) to 0.3 but still have the same problem. > > > > Below are my input files: > > > > > > calculation='scf', > > restart_mode='from_scratch', > > title='FeSi', > > outdir='./', > > prefix='FeSi', > > pseudo_dir='./', > > / > > > > ibrav=1, > > celldm(1)=8.4165, > > nat=8, ntyp=2, > > ecutwfc=60.0, > > ecutrho= 720.0, > > / > > > > conv_thr=1.0d-9, > > mixing_mode='plain', > > mixing_beta=0.7, > > diagonalization='david', > > / > > ATOMIC_SPECIES > > Fe 55.847 Fe.pw91-sp-van_ak.UPF > > Si 28.086 Si.pw91-n_van.UPF > > ATOMIC_POSITIONS > > Fe 0.134855265 0.134855265 0.134855265 > > Fe 0.634855265 0.365144735 0.865144735 > > Fe 0.865144735 0.634855265 0.365144735 > > Fe 0.365144735 0.865144735 0.634855265 > > Si 0.839976366 0.839976366 0.839976366 > > Si 0.339976366 0.660023634 0.160023634 > > Si 0.160023634 0.339976366 0.660023634 > > Si 0.660023634 0.160023634 0.339976366 > > K_POINTS {automatic} > > 10 10 10 0 0 0 > > *** > > phonons of FeSi > > > > tr2_ph=1.0d-14, > > prefix='FeSi', > > amass(1)=55.847, > > amass(2)=28.086, > > outdir='./' > > fildyn='FeSi.dyn', > > trans=.true., > > ldisp=.true., > > nq1=6, nq2=6, nq3=6 > > recover=.true., > > alpha_mix(1)=0.30 > > > > Any comments and suggestions are welcomed. > > > > Thanks > > IYAD > > > > > > ___ > > Pw_forum mailing list > > Pw_forum at pwscf.org > > http://www.democritos.it/mailman/listinfo/pw_forum > > * > William D. Parker phone: (630) 252-4834 > Computational Postdoctoral Fellow fax: (630) 252-4798 > MSD-212, Rm. C-215 > Argonne National Laboratory > 9700 S. Cass Ave. > Argonne, IL 60439 > * > > > > > > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > -- ___ IYAD I. AL-QASIR, PhD Postdoc Research Associate X-Ray and Neutron Scattering and Spectroscopy Group Materials Science and Technology Division Oak Ridge National Lab Oak Ridge, TN -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20120815/48abc622/attachment.htm
[Pw_forum] phonon calculations do not converge
Hi Iyad, Have you considered introducing some broadening in the pwscf calculation? It helps. Be careful not to go overboard though. You could try introducing some broadening (like 1e-3 Ry or so) and see if it converges. If that works than you can see how far you can reduce broadening while still having the phonon calculations converge. cheers, Matthijs On Wed, Aug 15, 2012 at 10:27 AM, Iyad AL-QASIR wrote: > Hi Wiliam, > > I used tr2_ph = 1.0d-10, it did not converge either after 100 steps. The > |ddv_scf|^ values are about ~ E-07. > > However, when I used tr2_ph = 1.0d-14, the |ddv_scf|^ values are diverging > strongly ~E+20 and even more. > > Kindest Regards, > > > > > On Fri, Aug 10, 2012 at 5:16 PM, William Parker wrote: > >> Hi Iyad, >> >> How strong is the failure to converge at (0, 1/6, 1/6)? Perhaps a less >> tight tr2_ph value of, say, 1.0d-10 would get you where you need to be. >> >> --William >> >> On Aug 10, 2012, at 12:29 PM, Iyad AL-QASIR wrote: >> >> > Dear PW_Forum, >> > >> > I am running phonon calculations for FeSi (a narrow band gap >> semiconductor) using QE-5.0. I am using a 6x6x6 q-vectors mesh. The first 4 >> q-points ran successfully and got good phonon values. However when I reach >> the 5th point(0 1/6 1/6). The self-consistent calculations do not converge. >> I tried to reduce alpha_mix(1) to 0.3 but still have the same problem. >> > >> > Below are my input files: >> > >> > >> > calculation='scf', >> > restart_mode='from_scratch', >> > title='FeSi', >> > outdir='./', >> > prefix='FeSi', >> > pseudo_dir='./', >> > / >> > >> > ibrav=1, >> > celldm(1)=8.4165, >> > nat=8, ntyp=2, >> > ecutwfc=60.0, >> > ecutrho= 720.0, >> > / >> > >> > conv_thr=1.0d-9, >> > mixing_mode='plain', >> > mixing_beta=0.7, >> > diagonalization='david', >> > / >> > ATOMIC_SPECIES >> > Fe 55.847 Fe.pw91-sp-van_ak.UPF >> > Si 28.086 Si.pw91-n_van.UPF >> > ATOMIC_POSITIONS >> > Fe 0.134855265 0.134855265 0.134855265 >> > Fe 0.634855265 0.365144735 0.865144735 >> > Fe 0.865144735 0.634855265 0.365144735 >> > Fe 0.365144735 0.865144735 0.634855265 >> > Si 0.839976366 0.839976366 0.839976366 >> > Si 0.339976366 0.660023634 0.160023634 >> > Si 0.160023634 0.339976366 0.660023634 >> > Si 0.660023634 0.160023634 0.339976366 >> > K_POINTS {automatic} >> > 10 10 10 0 0 0 >> > *** >> > phonons of FeSi >> > >> > tr2_ph=1.0d-14, >> > prefix='FeSi', >> > amass(1)=55.847, >> > amass(2)=28.086, >> > outdir='./' >> > fildyn='FeSi.dyn', >> > trans=.true., >> > ldisp=.true., >> > nq1=6, nq2=6, nq3=6 >> > recover=.true., >> > alpha_mix(1)=0.30 >> > >> > Any comments and suggestions are welcomed. >> > >> > Thanks >> > IYAD >> > >> > >> > ___ >> > Pw_forum mailing list >> > Pw_forum at pwscf.org >> > http://www.democritos.it/mailman/listinfo/pw_forum >> >> * >> William D. Parker phone: (630) 252-4834 >> Computational Postdoctoral Fellow fax: (630) 252-4798 >> MSD-212, Rm. C-215 >> Argonne National Laboratory >> 9700 S. Cass Ave. >> Argonne, IL 60439 >> * >> >> >> >> >> >> ___ >> Pw_forum mailing list >> Pw_forum at pwscf.org >> http://www.democritos.it/mailman/listinfo/pw_forum >> > > > > -- > ___ > IYAD I. AL-QASIR, PhD > Postdoc Research Associate > > X-Ray and Neutron Scattering and Spectroscopy Group > Materials Science and Technology Division > Oak Ridge National Lab > Oak Ridge, TN > > > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > > -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20120815/cae5d14f/attachment.htm
[Pw_forum] Problem with parallel phonon calculation using image method
actually, I typed wrong command in my earlier mail. My SCF calculation was indeed performed using 64 processors as: aprun -N 8 -n 64 ./ph.x -npool 4 < ph.in > ph.out Next, I tried the image method for phonons using: aprun -N 8 -n 640 ./ph.x -nimage 10 -npool 4 < ph.in > ph.out Thanks, David ___ Pw_forum mailing list Pw_forum at pwscf.org http://www.democritos.it/mailman/listinfo/pw_forum -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20120815/deb2995e/attachment.htm
[Pw_forum] 110-direction Vs 110-point calculation
Dear S. Chab: Definitely, I am not clear about your system. Which is the 4x4x4 M-P mesh? for k-points or q-points? If it is for k-points, then I think it is too small. But any way, I think the convergence test and well relaxation should be done before phonon calculation. For the [110] direction, which means the path from Gamma to X, in your system. X is just a point in FBZ and corresponds to (110) or (1/2, 1/2, 0) based on different cells. Of course, you can directly calculate frequencies by specifying a series of q-points along [110] direction in ph.x. But, running the whole phonon dispersion at converged q-mesh and only draw Gamma->X out is better choice. -- GAO Zhe CMC Lab, Materials Science & Engineering Department, Seoul National University, South Korea At 2012-08-14 21:18:39,"Satyananda Chab" wrote: Dear Users, I am trying to calculate the full phonon spectrum of a ferromagnetic alloy but the results that I found are not matching with literature. So I want to know (1) How to assign 110-direction for a grid calculation? (2)Whether calculating phonon spectrum from Gamma to X point along 110 direction is the same as that of a single point calculation at 110 point for this particular point. (3)Will both calculations give the same frequencies at 110 point? (4)My phonon spectrum do not match previous papers. Is there anything to do with grid size?(I used 4X4X4 Monkhorst pack ) If anybody throw me some light on these questions, i will be very happy. Thanking you in advance SATYANAND CH Junior Research Fellow Department of Physics Gauhati University Guwahati 781014 (+91)950-895-6853 email: csatya11 at gmail.com me as that of a single point calculation at 110 point. (3)Will both calculations give the same frequencies at 110 point? (4)My phonon spectrum do not match previous papers. Is there anything to do with grid size?(I used 4X4X4 Monkhorst pack ) If anybody of you throw me some light, i will be very happy. Thanking you in advance SATYA NANDA CH Junior Research Fellow Department of Physics Gauhati University Guwahati 781014 (+91)950-895-6853 email: csatya11 at gmail.com -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20120814/0ce1f11c/attachment.htm