[QE-users] Problem using ph.x with DFPT+U in QE 6.4.(1)

Shenli Zhang shlzhang at uchicago.edu
Wed Nov 13 17:46:27 CET 2019


Dear Iurii and Paolo,



Thank you so much for all the advice! It works for me now. And I’ve tested a bit to see which factor matters:

  1.  I used Iurii’s input, it works well using QE 6.4.1, but would crash using QE 6.4 (not sure if this is because a compilation issue as I just used the module on the cluster)
  2.  I verified each different parameters using QE 6.4.1 (add/remove nbnd, change ecutwfc/ecutrho ratio, conv_thr and add ‘’ for outdir), and found ecutwfc/ecutrho is the reason for my previous crash.
  3.  I set the occupations to be “fixed” because I want to calculate dielectric constant, it seems the ph.x program will complain if I use “smearing”, even if it converges to an insulating state.



Thank you!

Shenli

--
Shenli Zhang
Postdoc
Pritzker School of Molecular Engineering
The University of Chicago
5640 South Ellis Avenue, Chicago, IL 60637

From: users <users-bounces at lists.quantum-espresso.org> on behalf of Timrov Iurii <iurii.timrov at epfl.ch>
Reply-To: Quantum ESPRESSO users Forum <users at lists.quantum-espresso.org>
Date: Wednesday, November 13, 2019 at 4:54 AM
To: Quantum ESPRESSO users Forum <users at lists.quantum-espresso.org>
Subject: Re: [QE-users] Problem using ph.x with DFPT+U in QE 6.4.(1)


Dear Shenli and Paolo,



I changed a bit the input for PW and PH and it did not crash for me (only at the very end, for some reason). I used QE 6.4.1, 8 cores, ifort 15.0.2 20150121.



The input and output files are here (available temporary):

https://drive.google.com/open?id=1Z4fSndJZPcK_A5PVa3hMsft8IGqXWHZB



For reference, I show here only the input files:
&control
    calculation='scf',
    restart_mode='from_scratch',
    prefix='LCO'
    pseudo_dir = '/scratch/timrov/q-e-qe-6.4.1/pseudo/',
    outdir='/scratch/timrov/q-e-qe-6.4.1/tempdir/'
 /
 &system
    ibrav = -5,
    celldm(1) = 10.1352,
    celldm(4) = 0.47876847,
    nat = 10,
    ntyp = 3,
    ecutwfc = 40.0
    ecutrho = 320.0
    occupations='fixed'
    lda_plus_u = .true.,
    lda_plus_u_kind = 0,
    U_projection_type = 'atomic',
    Hubbard_U(2) = 3.0
 /
 &electrons
    conv_thr =  1.d-14
    mixing_beta = 0.7
 /
ATOMIC_SPECIES
La 138.9055 La.pbe-spfn-kjpaw_psl.1.0.0.UPF
Co 58.9332  Co.pbe-spn-kjpaw_psl.0.3.1.UPF
O  16       O.pbe-n-kjpaw_psl.1.0.0.UPF
ATOMIC_POSITIONS {crystal}
O        0.312627155   0.750000000   0.187372845
O        0.750000000   0.187372845   0.312627155
O        0.187372845   0.312627155   0.750000000
O        0.687372845   0.250000000   0.812627155
O        0.812627155   0.687372845   0.250000000
O        0.250000000   0.812627155   0.687372845
Co       0.000000000   0.000000000   0.000000000
Co       0.500000000   0.500000000   0.500000000
La       0.250000000   0.250000000   0.250000000
La       0.750000000   0.750000000   0.750000000
K_POINTS {automatic}
 4 4 4 0 0 0


phonons of LCO at Gamma
 &inputph
  prefix = 'LCO',
  outdir='/scratch/timrov/q-e-qe-6.4.1/tempdir/',
  tr2_ph = 1.0d-14,
  epsil = .true.
  fildyn='LCO.dynG',
 /
0.000000000000000   0.000000000000000   0.000000000000000



Shenli: Could you try please to use exactly the same inputs as me with QE 6.4.1 and see what happens?



Paolo: I do not know what to say why you obtain NaN's... Do you use exactly the same input as Shenli?



Greetings,

Iurii


--
Dr. Iurii Timrov
Postdoctoral Researcher
STI - IMX - THEOS and NCCR - MARVEL
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: users <users-bounces at lists.quantum-espresso.org> on behalf of Paolo Giannozzi <p.giannozzi at gmail.com>
Sent: Wednesday, November 13, 2019 11:13:09 AM
To: Quantum ESPRESSO users Forum
Subject: Re: [QE-users] Problem using ph.x with DFPT+U in QE 6.4.(1)

It doesn't segfaults for me but it produces NaN's. Note that:
- SCF convergence with fixed occupancies is very bad. I started from the charge density and wavefunctions obtained with smearing.
-  outdir='rhomb_U3', between ' '
Paolo

On Tue, Nov 12, 2019 at 10:52 PM Shenli Zhang <shlzhang at uchicago.edu<mailto:shlzhang at uchicago.edu>> wrote:
Dear Iurii,

Thank you so much for your reply. The output error “forrtl: severe (174): SIGSEGV, segmentation fault occurred” is not in the output file of ph.x, but is in the output file generated by the slurm system, which is a job scheduler for the cluster I use. In the ph.x output, there is no error message, but the program just stopped in the line “Calculating the dnsorth_cart matrix…”
I double checked the memory setting, it doesn’t seem to be the problem here. As I’ve increased the number of CPUs used to the point where I could actually run DFPT+U on an orthorhombic supercell of the same system.

And thanks for the advice on the parameter settings. I reduced the ecutrho/ecutwfc here  in order to do a quick test. I tried the ultrasoft pseudopotentials, and the error still persists. And yes my system is a nonmagnetic insulator and the scf calculation converged successfully.

The reason I think this error is probably due to the unit cell shape is because I also created an orthorhombic supercell of the same system with the same parameter settings, and I could run the DFPT+U without any problem.

Regards,
Shenli

--
Shenli Zhang
Postdoc
Pritzker School of Molecular Engineering
The University of Chicago
5640 South Ellis Avenue, Chicago, IL 60637

From: users <users-bounces at lists.quantum-espresso.org<mailto:users-bounces at lists.quantum-espresso.org>> on behalf of Timrov Iurii <iurii.timrov at epfl.ch<mailto:iurii.timrov at epfl.ch>>
Reply-To: Quantum ESPRESSO users Forum <users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>>
Date: Tuesday, November 12, 2019 at 11:31 AM
To: "users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>" <users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>>
Subject: Re: [QE-users] Problem using ph.x with DFPT+U in QE 6.4.(1)


Dear Shenli Zhang,



> with an output error “forrtl: severe (174): SIGSEGV, segmentation fault occurred”



Was is printed in the output file of the PHonon calculation after the crash?



Check if you have enough RAM to run DFPT+U, because it is heavy.



Your input for a PW SCF calculation has some problem:

- ecutrho/ecutwfc = 4 - this is true for norm-conserving pseudopotentials. But you are using PAW, for which the ratio should be 8 or more. Try to use DFPT+U with ultrasoft pseudopotentials (the ratio should be also 8 or more).

- nbnd=50 - this is not needed: just remove it from the input file

- occupations='fixed' : is your system a nonmagnetic insulator? Did your PW SCF calculation converged successfully?

- conv_thr=1.0d-12 : I suggest to try conv_thr=1.0d-14 if you plan to compute phonons afterwards.



HTH



Regards,

Iurii


--
Dr. Iurii Timrov
Postdoctoral Researcher
STI - IMX - THEOS and NCCR - MARVEL
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: users <users-bounces at lists.quantum-espresso.org<mailto:users-bounces at lists.quantum-espresso.org>> on behalf of Shenli Zhang <shlzhang at uchicago.edu<mailto:shlzhang at uchicago.edu>>
Sent: Tuesday, November 12, 2019 5:57:47 PM
To: users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>
Subject: [QE-users] Problem using ph.x with DFPT+U in QE 6.4.(1)

Dear all,

I’m testing phonon calculations using ph.x with DFPT+U in QE 6.4 and QE 6.4.1, however the calculation always stops at the line “Calculating the dnsorth_cart matrix…”  with an output error “forrtl: severe (174): SIGSEGV, segmentation fault occurred”
I ran my system without U successfully. And I also ran the LiCoO2 +U example successfully. The only difference I can think of between my input file and the example file is the unit cell shape, where my LaCoO3 system has a rhombohedral unit cell and the example LiCoO2 has an orthorhombic one. Could this be a problem or did I make some mistakes?

Attached are the input files for my system LCO. Thank you!

SCF calculation input:
&control
    calculation='scf'
    prefix='LaCoO3p_relax'
    outdir=rhomb_U3
    pseudo_dir = '/home/shlzhang/QE_simulation/LaCoO3/pseudo',
/
&system
    nbnd=50
    ibrav=-5, celldm(1)=10.1352, celldm(4) =0.47876847, nat=10, ntyp= 3,
    ecutwfc =90, ecutrho =360, !for PAW ecutrho=4*ecutwfc, but need test
    occupations='fixed' !energy units in Ry
    lda_plus_u=.true.  Hubbard_U(2)=3
/
&electrons
conv_thr=1.0d-12
/
&ions
/
&cell
!cell_dofree='shape'
/
ATOMIC_SPECIES
La 138.9055 La.pbe-spfn-kjpaw_psl.1.0.0.UPF
Co 58.9332 Co.pbe-spn-kjpaw_psl.0.3.1.UPF
O  16 O.pbe-n-kjpaw_psl.1.0.0.UPF
ATOMIC_POSITIONS (crystal)
O        0.312627155   0.750000000   0.187372845
O        0.750000000   0.187372845   0.312627155
O        0.187372845   0.312627155   0.750000000
O        0.687372845   0.250000000   0.812627155
O        0.812627155   0.687372845   0.250000000
O        0.250000000   0.812627155   0.687372845
Co      -0.000000000   0.000000000  -0.000000000
Co       0.500000000   0.500000000   0.500000000
La       0.250000000   0.250000000   0.250000000
La       0.750000000   0.750000000   0.750000000
K_POINTS automatic
  4 4 4 0 0 0

Phonon calculation input:

phonons of bulk LaCoO3 at Gamma
&inputph
   prefix='LaCoO3p_relax'
   outdir=rhomb_U3
   fildyn='LaCoO3.dyn'
   max_seconds=82800
   tr2_ph=1.0d-14
   verbosity='debug'
   amass(1)=138.9055
   amass(2)=58.9332
   amass(3)=16
   epsil=.true.
   trans=.true.
/
0.00000000000 0.000000000000 0.000000000000000


--
Shenli Zhang
Postdoc
Pritzker School of Molecular Engineering
The University of Chicago
5640 South Ellis Avenue, Chicago, IL 60637
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso<http://www.max-centre.eu/quantum-espresso>)
users mailing list users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>
https://lists.quantum-espresso.org/mailman/listinfo/users


--
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20191113/4abc7fbd/attachment.html>


More information about the users mailing list