Re: [QE-users] "Error in routine set_irr_sym_new" with phonon qe 6.2 to 6.4.1
For future reference, I report here the explanation and workaround by Lorenzo Paulatto that seems to work in this case: "The phonon code checks if modes are degenerate by symmetrizing a random matrix and checking if the eigenvalues are the same. For sufficiently large systems there are going to be accidental degeneracies. Try to uncomment the line !!#define __UNIFORM_DISTRIB in PHonon/PH/random_matrix.f90 and recompile" On Thu, May 7, 2020 at 8:51 AM Paolo Giannozzi wrote: > On Thu, May 7, 2020 at 12:22 AM Merlin Meheut > wrote: > > I am not sure how to implement you solution >> > > it is not a "solution", unfortunately: just an empirical observation, > based on many similar reports. Sometimes it is sufficient to make the > system closer to the expected symmetry to see the problem disappear. It may > not be the case here, though > > (and also, why is that not happening with version 5.3?) >> > > changes in the symmetry algorithm? > > As a first guess, I would be tempted to give more accurate cell >> parameters, or more accurate positions, but this seems weird, because those >> come from a relaxation (calculation='vc-relax') with pw.x, and further, >> pw.x (whatever the version) finds all the symetries of the structure, >> before and after relaxation. >> > > this is the correct and expected behavior, but there are thresholds that > decide whether a symmetry is there or not. In some unfortunate cases, these > thresholds may give inconsistencies between the self-consistent and the > phonon code. > > [...] I modified it to return the value of wrk that creates an error, and >> it gives me wrk=0.640333821809072, in the second case (jpert.eq.ipert). >> This large value seems contradictory with your suggestion, as this is quite >> far from 1. For a "not close enough" symmetry, I would have expected >> something smaller, dont you think? >> > > you are likely right > > Would you have some guidance on how to proceed further? >> > > unfortunately no: I have lost track of how symmetry works in the phonon > code since a long time. I suspect that these lines appearing in the output: > Eigenvectors exchange needed 23 235 > are in some way related to your problem, but it is impossible to say more > without an extensive investigation. > > Paolo > > >> Best regards, >> >> Merlin Méheut >> >> Le mar. 5 mai 2020 à 21:54, Paolo Giannozzi a >> écrit : >> >>> Typically, this error happens with cells that are very close to a given >>> symmetry, but not close enough. >>> >>> Paolo >>> >>> On Sun, May 3, 2020 at 1:16 AM Merlin Meheut >>> wrote: >>> Dear PWSCF users, I am trying to compute phonon properties of a garnet (spacegroup Ia-3d) structure, and the computation is working perfectly well with version 5.3.0, but I have an error message for versions 6.2, 6.3MaX, 6.4 and 6.4.1 (I did not compile 6.5, yet): %% Error in routine set_irr_sym_new (1622): wrong representation %% stopping ... Would you have an idea on the origin of this error? the scf input: calculation = 'scf', restart_mode = 'from_scratch' , prefix = 'GROSS', disk_io = 'default' , pseudo_dir = './', outdir ='./', tprnfor= .true., tstress= .true., disk_io= 'low', dt = 30 , nstep = 1300 , etot_conv_thr = 5.d-5, forc_conv_thr = 5.d-4, / ibrav = 0 , celldm(1)=22.36869, nat =80, ntyp = 4, ecutwfc = 80.0 ecutrho = 320.0, / electron_maxstep = 80, conv_thr = 1.d-11, mixing_mode = 'plain', startingwfc = 'atomic', mixing_beta = 0.5, / / wmass = 0.01, cell_dynamics = 'damp-w', press_conv_thr = 0.2 / ATOMIC_SPECIES Al 26.9815 Al.pbe2.UPF Ca 39.9626 Ca.pbe-spn-rrkjus_psl.1.0.0.UPF Si 27.9769 Si.pbe2.UPF O15.9949 O.pbe.UPF CELL_PARAMETERS (alat= 22.36869000) 0.509146517 0.509146517 0.509146517 -0.509146517 0.509146517 0.509146517 -0.509146517 -0.509146517 0.509146517 ATOMIC_POSITIONS (crystal) Al -0.0 0.0 -0.0 Al 0.0 0.5 0.5 Al 0.5 0.5 0.0 Al 0.0 0.5 0.0 Al 0.5 0.0 0.5 Al 0.5 0.0 0.0 Al 0.0 0.0 0.5 Al 0.5
Re: [QE-users] "Error in routine set_irr_sym_new" with phonon qe 6.2 to 6.4.1
On Thu, May 7, 2020 at 12:22 AM Merlin Meheut wrote: I am not sure how to implement you solution > it is not a "solution", unfortunately: just an empirical observation, based on many similar reports. Sometimes it is sufficient to make the system closer to the expected symmetry to see the problem disappear. It may not be the case here, though (and also, why is that not happening with version 5.3?) > changes in the symmetry algorithm? As a first guess, I would be tempted to give more accurate cell parameters, > or more accurate positions, but this seems weird, because those come from a > relaxation (calculation='vc-relax') with pw.x, and further, pw.x (whatever > the version) finds all the symetries of the structure, before and after > relaxation. > this is the correct and expected behavior, but there are thresholds that decide whether a symmetry is there or not. In some unfortunate cases, these thresholds may give inconsistencies between the self-consistent and the phonon code. [...] I modified it to return the value of wrk that creates an error, and > it gives me wrk=0.640333821809072, in the second case (jpert.eq.ipert). > This large value seems contradictory with your suggestion, as this is quite > far from 1. For a "not close enough" symmetry, I would have expected > something smaller, dont you think? > you are likely right Would you have some guidance on how to proceed further? > unfortunately no: I have lost track of how symmetry works in the phonon code since a long time. I suspect that these lines appearing in the output: Eigenvectors exchange needed 23 235 are in some way related to your problem, but it is impossible to say more without an extensive investigation. Paolo > Best regards, > > Merlin Méheut > > Le mar. 5 mai 2020 à 21:54, Paolo Giannozzi a > écrit : > >> Typically, this error happens with cells that are very close to a given >> symmetry, but not close enough. >> >> Paolo >> >> On Sun, May 3, 2020 at 1:16 AM Merlin Meheut >> wrote: >> >>> Dear PWSCF users, >>> >>> I am trying to compute phonon properties of a garnet (spacegroup Ia-3d) >>> structure, and the computation is working perfectly well with version >>> 5.3.0, but I have an error message for versions 6.2, 6.3MaX, 6.4 and 6.4.1 >>> (I did not compile 6.5, yet): >>> >>> >>> %% >>> Error in routine set_irr_sym_new (1622): >>> wrong representation >>> >>> >>> %% >>> >>> stopping ... >>> >>> Would you have an idea on the origin of this error? >>> >>> the scf input: >>> >>>calculation = 'scf', >>> restart_mode = 'from_scratch' , >>> prefix = 'GROSS', >>>disk_io = 'default' , >>> pseudo_dir = './', >>> outdir ='./', >>> tprnfor= .true., >>> tstress= .true., >>> disk_io= 'low', >>> dt = 30 , >>> nstep = 1300 , >>> etot_conv_thr = 5.d-5, >>> forc_conv_thr = 5.d-4, >>> / >>> >>> ibrav = 0 , >>> celldm(1)=22.36869, >>> nat =80, ntyp = 4, ecutwfc = 80.0 >>> ecutrho = 320.0, >>> / >>> >>> >>>electron_maxstep = 80, >>> conv_thr = 1.d-11, >>>mixing_mode = 'plain', >>>startingwfc = 'atomic', >>>mixing_beta = 0.5, >>> / >>> >>> >>> / >>> >>> >>> wmass = 0.01, >>> cell_dynamics = 'damp-w', >>> press_conv_thr = 0.2 >>> / >>> >>> ATOMIC_SPECIES >>> Al 26.9815 Al.pbe2.UPF >>> Ca 39.9626 Ca.pbe-spn-rrkjus_psl.1.0.0.UPF >>> Si 27.9769 Si.pbe2.UPF >>> O15.9949 O.pbe.UPF >>> >>> CELL_PARAMETERS (alat= 22.36869000) >>>0.509146517 0.509146517 0.509146517 >>> -0.509146517 0.509146517 0.509146517 >>> -0.509146517 -0.509146517 0.509146517 >>> ATOMIC_POSITIONS (crystal) >>> Al -0.0 0.0 -0.0 >>> Al 0.0 0.5 0.5 >>> Al 0.5 0.5 0.0 >>> Al 0.0 0.5 0.0 >>> Al 0.5 0.0 0.5 >>> Al 0.5 0.0 0.0 >>> Al 0.0 0.0 0.5 >>> Al 0.5 0.5 0.5 >>> Ca 0.12500 0.25000 0.87500 >>> Ca 0.37500 0.75000 0.62500 >>> Ca 0.37500 0.87500 0.25000 >>> Ca 0.87500 0.75000 0.12500 >>> Ca 0.12500 0.62500 0.75000 >>> Ca 0.62500 0.25000 0.37500 >>> Ca 0.25000 0.87500 0.87500 >>> Ca 0.75000 0.62500 0.62500 >>> Ca 0.62500 0.12500 0.75000 >>> Ca 0.87500 0.37500 0.25000 >>> Ca 0.75000 0.12500 0.12500 >>> Ca 0.25000 0.37500
Re: [QE-users] "Error in routine set_irr_sym_new" with phonon qe 6.2 to 6.4.1
Dear Paolo, Thank you very much for your help! I am not sure how to implement you solution, however (and also, why is that not happening with version 5.3?). As a first guess, I would be tempted to give more accurate cell parameters, or more accurate positions, but this seems weird, because those come from a relaxation (calculation='vc-relax') with pw.x, and further, pw.x (whatever the version) finds all the symetries of the structure, before and after relaxation. Further, the cell is cubic (I tried both with ibrav=3 and with ibrav=0, but the result is the same). Also, what you said encouraged me to take a look at the subroutine set_irr_sym_new and identified the line that was responsible for the error: if (jpert.ne.ipert .and. abs(wrk)> 1.d-6 ) & call errore('set_irr_sym_new','wrong representation',100*irr+10*jpert+ipert) if (jpert.eq.ipert .and. abs(wrk-1.d0)> 1.d-6 ) & call errore('set_irr_sym_new','wrong representation',100*irr+10*jpert+ipert) I modified it to return the value of wrk that creates an error, and it gives me wrk=0.640333821809072, in the second case (jpert.eq.ipert). This large value seems contradictory with your suggestion, as this is quite far from 1. For a "not close enough" symmetry, I would have expected something smaller, dont you think? Would you have some guidance on how to proceed further? Best regards, Merlin Méheut Le mar. 5 mai 2020 à 21:54, Paolo Giannozzi a écrit : > Typically, this error happens with cells that are very close to a given > symmetry, but not close enough. > > Paolo > > On Sun, May 3, 2020 at 1:16 AM Merlin Meheut > wrote: > >> Dear PWSCF users, >> >> I am trying to compute phonon properties of a garnet (spacegroup Ia-3d) >> structure, and the computation is working perfectly well with version >> 5.3.0, but I have an error message for versions 6.2, 6.3MaX, 6.4 and 6.4.1 >> (I did not compile 6.5, yet): >> >> >> %% >> Error in routine set_irr_sym_new (1622): >> wrong representation >> >> >> %% >> >> stopping ... >> >> Would you have an idea on the origin of this error? >> >> the scf input: >> >>calculation = 'scf', >> restart_mode = 'from_scratch' , >> prefix = 'GROSS', >>disk_io = 'default' , >> pseudo_dir = './', >> outdir ='./', >> tprnfor= .true., >> tstress= .true., >> disk_io= 'low', >> dt = 30 , >> nstep = 1300 , >> etot_conv_thr = 5.d-5, >> forc_conv_thr = 5.d-4, >> / >> >> ibrav = 0 , >> celldm(1)=22.36869, >> nat =80, ntyp = 4, ecutwfc = 80.0 >> ecutrho = 320.0, >> / >> >> >>electron_maxstep = 80, >> conv_thr = 1.d-11, >>mixing_mode = 'plain', >>startingwfc = 'atomic', >>mixing_beta = 0.5, >> / >> >> >> / >> >> >> wmass = 0.01, >> cell_dynamics = 'damp-w', >> press_conv_thr = 0.2 >> / >> >> ATOMIC_SPECIES >> Al 26.9815 Al.pbe2.UPF >> Ca 39.9626 Ca.pbe-spn-rrkjus_psl.1.0.0.UPF >> Si 27.9769 Si.pbe2.UPF >> O15.9949 O.pbe.UPF >> >> CELL_PARAMETERS (alat= 22.36869000) >>0.509146517 0.509146517 0.509146517 >> -0.509146517 0.509146517 0.509146517 >> -0.509146517 -0.509146517 0.509146517 >> ATOMIC_POSITIONS (crystal) >> Al -0.0 0.0 -0.0 >> Al 0.0 0.5 0.5 >> Al 0.5 0.5 0.0 >> Al 0.0 0.5 0.0 >> Al 0.5 0.0 0.5 >> Al 0.5 0.0 0.0 >> Al 0.0 0.0 0.5 >> Al 0.5 0.5 0.5 >> Ca 0.12500 0.25000 0.87500 >> Ca 0.37500 0.75000 0.62500 >> Ca 0.37500 0.87500 0.25000 >> Ca 0.87500 0.75000 0.12500 >> Ca 0.12500 0.62500 0.75000 >> Ca 0.62500 0.25000 0.37500 >> Ca 0.25000 0.87500 0.87500 >> Ca 0.75000 0.62500 0.62500 >> Ca 0.62500 0.12500 0.75000 >> Ca 0.87500 0.37500 0.25000 >> Ca 0.75000 0.12500 0.12500 >> Ca 0.25000 0.37500 0.37500 >> Si 0.62500 0.62500 0.25000 >> Si 0.87500 0.87500 0.75000 >> Si 0.25000 0.12500 0.62500 >> Si 0.75000 0.87500 0.37500 >> Si 0.37500 0.37500 0.75000 >> Si 0.75000 0.37500 0.87500 >> Si 0.25000 0.62500 0.12500 >> Si 0.12500 0.12500 0.25000 >>
Re: [QE-users] "Error in routine set_irr_sym_new" with phonon qe 6.2 to 6.4.1
Typically, this error happens with cells that are very close to a given symmetry, but not close enough. Paolo On Sun, May 3, 2020 at 1:16 AM Merlin Meheut wrote: > Dear PWSCF users, > > I am trying to compute phonon properties of a garnet (spacegroup Ia-3d) > structure, and the computation is working perfectly well with version > 5.3.0, but I have an error message for versions 6.2, 6.3MaX, 6.4 and 6.4.1 > (I did not compile 6.5, yet): > > > %% > Error in routine set_irr_sym_new (1622): > wrong representation > > > %% > > stopping ... > > Would you have an idea on the origin of this error? > > the scf input: > >calculation = 'scf', > restart_mode = 'from_scratch' , > prefix = 'GROSS', >disk_io = 'default' , > pseudo_dir = './', > outdir ='./', > tprnfor= .true., > tstress= .true., > disk_io= 'low', > dt = 30 , > nstep = 1300 , > etot_conv_thr = 5.d-5, > forc_conv_thr = 5.d-4, > / > > ibrav = 0 , > celldm(1)=22.36869, > nat =80, ntyp = 4, ecutwfc = 80.0 > ecutrho = 320.0, > / > > >electron_maxstep = 80, > conv_thr = 1.d-11, >mixing_mode = 'plain', >startingwfc = 'atomic', >mixing_beta = 0.5, > / > > > / > > > wmass = 0.01, > cell_dynamics = 'damp-w', > press_conv_thr = 0.2 > / > > ATOMIC_SPECIES > Al 26.9815 Al.pbe2.UPF > Ca 39.9626 Ca.pbe-spn-rrkjus_psl.1.0.0.UPF > Si 27.9769 Si.pbe2.UPF > O15.9949 O.pbe.UPF > > CELL_PARAMETERS (alat= 22.36869000) >0.509146517 0.509146517 0.509146517 > -0.509146517 0.509146517 0.509146517 > -0.509146517 -0.509146517 0.509146517 > ATOMIC_POSITIONS (crystal) > Al -0.0 0.0 -0.0 > Al 0.0 0.5 0.5 > Al 0.5 0.5 0.0 > Al 0.0 0.5 0.0 > Al 0.5 0.0 0.5 > Al 0.5 0.0 0.0 > Al 0.0 0.0 0.5 > Al 0.5 0.5 0.5 > Ca 0.12500 0.25000 0.87500 > Ca 0.37500 0.75000 0.62500 > Ca 0.37500 0.87500 0.25000 > Ca 0.87500 0.75000 0.12500 > Ca 0.12500 0.62500 0.75000 > Ca 0.62500 0.25000 0.37500 > Ca 0.25000 0.87500 0.87500 > Ca 0.75000 0.62500 0.62500 > Ca 0.62500 0.12500 0.75000 > Ca 0.87500 0.37500 0.25000 > Ca 0.75000 0.12500 0.12500 > Ca 0.25000 0.37500 0.37500 > Si 0.62500 0.62500 0.25000 > Si 0.87500 0.87500 0.75000 > Si 0.25000 0.12500 0.62500 > Si 0.75000 0.87500 0.37500 > Si 0.37500 0.37500 0.75000 > Si 0.75000 0.37500 0.87500 > Si 0.25000 0.62500 0.12500 > Si 0.12500 0.12500 0.25000 > Si 0.37500 0.25000 0.12500 > Si 0.62500 0.75000 0.87500 > Si 0.12500 0.75000 0.37500 > Si 0.87500 0.25000 0.62500 > O0.690298613 0.005517534 0.607713976 > O0.613231510 0.494482466 0.195816147 > O0.809701387 0.582584637 0.304183853 > O0.695816147 0.386768490 0.005517534 > O0.392286024 0.494482466 0.309701387 > O0.309701387 0.994482466 0.392286024 > O0.886768490 0.917415363 0.892286024 > O0.107713976 0.309701387 0.494482466 > O0.804183853 0.005517534 0.386768490 > O0.386768490 0.505517534 0.804183853 > O0.392286024 0.113231510 0.582584637 > O0.695816147 0.917415363 0.190298613 > O0.190298613 0.417415363 0.695816147 > O0.804183853 0.190298613 0.917415363 > O0.082584637 0.607713976 0.386768490 > O0.994482466 0.113231510 0.304183853 > O0.304183853 0.613231510 0.994482466 > O0.107713976 0.582584637 0.113231510 > O0.886768490 0.195816147 0.494482466 > O0.607713976 0.505517534 0.690298613 > O0.113231510 0.082584637 0.107713976 > O0.417415363 0.195816147 0.309701387 > O0.505517534 0.190298613 0.892286024 > O0.892286024 0.690298613 0.505517534 > O0.809701387 0.607713976 0.005517534 > O0.195816147 0.994482466 0.613231510 > O0.505517534 0.304183853 0.113231510 > O0.417415363