[QE-users] Relaxing magnetic structures
Timrov Iurii
iurii.timrov at epfl.ch
Thu Aug 13 20:22:16 CEST 2020
Dear Hien Vo,
Unfortunately I do not have access to your files (access denied).
> I was concerned about this too, but I’m trying to relax the structure for computing phonons by the finite displacement, so would larger values be ok in this case?
>From my experience, forc_conv_thr=1.0D-5 and etot_conv_thr=1.0D-6 should be a safe choice. Bum maybe others can also comment on this.
> The k-point mesh I used is 8x6x8 which (after symmetry), gives 130 k-points for the 20-atom unit cell. In the 2nd link (my results for convergence tests), I did check degauss against different k-point meshes. I’m using this value so that my results are comparable to past work done in the group. Do you think I should change it?
Ok. If the convergence w.r.t. k-mesh and smearing was checked, then it should be fine.
> I’m using this U since it was computed by a postdoc in the group and we’re trying to maintain consistency so that my results and hers are comparable.
Ok
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 Hien Vo <hvo at uchicago.edu>
Sent: Thursday, August 13, 2020 5:50:17 PM
To: Quantum ESPRESSO users Forum
Subject: Re: [QE-users] Relaxing magnetic structures
Hi Timrov Iurii,
Thank you so much for replying to me! I’ve included my affiliation at the bottom of this email.
>What is not clear to me is why you have a different starting_magnetization in the final SCF run. Why not using the same starting_magnetization? There might be different minima, and by using different starting_magnetization you can end up in different minima. Not sure that this is what happens in your case though.
I think I was being unclear here. The final scf was done as part of the vc-relax routine. I’m including a link to my output here : https://drive.google.com/drive/folders/1JgEJWZVYAbLJI3w_OsSxAUhZQDA5JuLR?usp=sharing . Since the relaxation took quite some time I had to split it into 2 runs, the second run is in the 0.7-2.out file. At the end when it runs the final scf run at the relaxed structure, the program picks starting_magnetization = -0.69 and 0.69, while originally my input is -0.7 and 0.7. I think it’s fine that the program picks another starting_magnetization, just wish it gave me a lower P.
> As far as I know, typically, the differences in the pressure between the final step of vc-relax and the final SCF run is due to the not-large-enough cutoff (check the pw_forum archive for more discussions about this). But in your case you have 100/800 Ry which are already seem to be large, but I do not know if these are "good" cutoffs for the pseudos that you use. So I am not sure that this is the source of the problem in your case, but maybe it is worth checking.
So before doing these calculations I checked the convergence on a 10-atom cell. The results are here: https://drive.google.com/file/d/1dZRt-5dzDAibgV49Z4nX3QQKt04TA3hR/view?usp=sharing . I think it looks fine?
>These are extremely small. I would use values not smaller than this:
>forc_conv_thr=1.0D-5,
>etot_conv_thr=1.0D-6,
>and in many cases even larger values should be OK, like this
>forc_conv_thr=1.0D-4,
>etot_conv_thr=1.0D-5
I was concerned about this too, but I’m trying to relax the structure for computing phonons by the finite displacement, so would larger values be ok in this case?
>> degauss=0.0036,
>This is also very small. Which k point mesh do you use? You did not specify.
The k-point mesh I used is 8x6x8 which (after symmetry), gives 130 k-points for the 20-atom unit cell. In the 2nd link (my results for convergence tests), I did check degauss against different k-point meshes. I’m using this value so that my results are comparable to past work done in the group. Do you think I should change it?
>> Hubbard_U(2)=3
>> Hubbard_U(3)=3
>It seems you are using an empirical U or maybe you took from literature. Please note that it is possible to compute U for your system from first principles using the HP code of QE. If you are >interested, have a look at this paper: https://journals.aps.org/prb/abstract/10.1103/PhysRevB.98.085127<https://urldefense.com/v3/__https:/journals.aps.org/prb/abstract/10.1103/PhysRevB.98.085127__;!!BpyFHLRN4TMTrA!uOBrZH-JZqDOB3sVmaaTpAd3eaugAUSRpX2LTF_Rn2aS51MmXHeNbanCM-0yoHP-$>
I’m using this U since it was computed by a postdoc in the group and we’re trying to maintain consistency so that my results and hers are comparable.
Best,
Hien Vo
Graduate Student @ The Chemistry Department
The University of Chicago
From: Timrov Iurii<mailto:iurii.timrov at epfl.ch>
Sent: Thursday, August 13, 2020 3:30 AM
To: users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>
Subject: Re: [QE-users] Relaxing magnetic structures
Dear Hien Vo,
First of all, do not forget to add your affiliation when posting to the pw_forum. Also, it is useful to specify which version of QE was used. Check the posting guidelines here: https://www.quantum-espresso.org/forum<https://urldefense.com/v3/__https:/www.quantum-espresso.org/forum__;!!BpyFHLRN4TMTrA!uOBrZH-JZqDOB3sVmaaTpAd3eaugAUSRpX2LTF_Rn2aS51MmXHeNbanCM4tbV-uy$>
> What I notice is the vc-relax would get P to 0 but the final scf calculation at the relaxed structure would give me a large P using a different starting_magnetization from the one I used for the input.
What is not clear to me is why you have a different starting_magnetization in the final SCF run. Why not using the same starting_magnetization? There might be different minima, and by using different starting_magnetization you can end up in different minima. Not sure that this is what happens in your case though.
As far as I know, typically, the differences in the pressure between the final step of vc-relax and the final SCF run is due to the not-large-enough cutoff (check the pw_forum archive for more discussions about this). But in your case you have 100/800 Ry which are already seem to be large, but I do not know if these are "good" cutoffs for the pseudos that you use. So I am not sure that this is the source of the problem in your case, but maybe it is worth checking.
Some comments about your input parameters:
> forc_conv_thr=1.0D-6,
> etot_conv_thr=1.4D-9,
These are extremely small. I would use values not smaller than this:
forc_conv_thr=1.0D-5,
etot_conv_thr=1.0D-6,
and in many cases even larger values should be OK, like this
forc_conv_thr=1.0D-4,
etot_conv_thr=1.0D-5
> degauss=0.0036,
This is also very small. Which k point mesh do you use? You did not specify.
> Hubbard_U(2)=3
> Hubbard_U(3)=3
It seems you are using an empirical U or maybe you took from literature. Please note that it is possible to compute U for your system from first principles using the HP code of QE. If you are interested, have a look at this paper: https://journals.aps.org/prb/abstract/10.1103/PhysRevB.98.085127<https://urldefense.com/v3/__https:/journals.aps.org/prb/abstract/10.1103/PhysRevB.98.085127__;!!BpyFHLRN4TMTrA!uOBrZH-JZqDOB3sVmaaTpAd3eaugAUSRpX2LTF_Rn2aS51MmXHeNbanCM-0yoHP-$>
> La 138.9055 La.pbe-spfn-kjpaw_psl.1.0.0.UPF
> Co1 58.9332 Co.pbe-spn-kjpaw_psl.0.3.1.UPF
> Co2 58.9332 Co.pbe-spn-kjpaw_psl.0.3.1.UPF
> O 16.00 O.pbe-n-kjpaw_psl.1.0.0.UPF
This is your choice to use these pseudos. Just for the reference, I suggest to have a look here: https://www.materialscloud.org/discover/sssp/table/efficiency<https://urldefense.com/v3/__https:/www.materialscloud.org/discover/sssp/table/efficiency__;!!BpyFHLRN4TMTrA!uOBrZH-JZqDOB3sVmaaTpAd3eaugAUSRpX2LTF_Rn2aS51MmXHeNbanCM-RgGvRX$>
Cheers,
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<https://urldefense.com/v3/__http:/people.epfl.ch/265334__;!!BpyFHLRN4TMTrA!uOBrZH-JZqDOB3sVmaaTpAd3eaugAUSRpX2LTF_Rn2aS51MmXHeNbanCM3NUdNoK$>
From: users <users-bounces at lists.quantum-espresso.org> on behalf of Hien Vo <hvo at uchicago.edu>
Sent: Wednesday, August 12, 2020 7:39:07 PM
To: users at lists.quantum-espresso.org
Subject: [QE-users] Relaxing magnetic structures
Hello QE community,
I’m trying to relax a-type afm LaCoO3 using the vc-relax option and I can’t seem to get P close to 0. What I notice is the vc-relax would get P to 0 but the final scf calculation at the relaxed structure would give me a large P using a different starting_magnetization from the one I used for the input. I’m including my input here (I’m calculating phonons with these so I’m trying to reduce the force and stress as much as possible) as well as relevant output from the run before final scf calculation and also the final scf calculation. Any tips would be greatly appreciated!
INPUT:
&CONTROL
calculation='vc-relax',
tprnfor=.TRUE.,
forc_conv_thr=1.0D-6,
etot_conv_thr=1.4D-9,
max_seconds=64800
/
&SYSTEM
ibrav=12,
celldm(1)=10.284016,celldm(2)=1.43711648,celldm(3)=1.011898,celldm(4)=-0.025604,
nat=20,ntyp=4,
occupations='smearing',degauss=0.0036,
ecutwfc=100,ecutrho=800,
nspin=2,starting_magnetization(2)=-0.7,starting_magnetization(3)=0.7,
lda_plus_u=.TRUE. Hubbard_U(2)=3
Hubbard_U(3)=3
/
&ELECTRONS
electron_maxstep=3000
mixing_beta=0.05D
conv_thr=1.4D-9
/
&IONS
trust_radius_ini=0.2
trust_radius_max=0.5
/
&CELL
/
ATOMIC_SPECIES
La 138.9055 La.pbe-spfn-kjpaw_psl.1.0.0.UPF
Co1 58.9332 Co.pbe-spn-kjpaw_psl.0.3.1.UPF
Co2 58.9332 Co.pbe-spn-kjpaw_psl.0.3.1.UPF
O 16.00 O.pbe-n-kjpaw_psl.1.0.0.UPF
ATOMIC_POSITIONS (crystal)
La 0.50000 0.25000 0.00000
La 0.50000 0.75000 0.00000
La 0.00000 0.75000 0.50000
La 0.00000 0.25000 0.50000
Co2 0.00000 0.00000 0.00000
Co1 0.00000 0.50000 0.00000
Co1 0.50000 0.50000 0.50000
Co2 0.50000 0.00000 0.50000
O 0.00000 0.25000 0.06296
O 0.00000 0.75000 0.93704
O 0.28148 0.96852 0.21852
O 0.71852 0.03148 0.78148
O 0.21852 0.03148 0.71852
O 0.78148 0.96852 0.28148
O 0.71852 0.53148 0.21852
O 0.28148 0.46852 0.78148
O 0.78148 0.46852 0.71852
O 0.21852 0.53148 0.28148
O 0.50000 0.75000 0.56296
O 0.50000 0.25000 0.43704
OUTPUT OF RUN BEFORE FINAL SCF:
Forces acting on atoms (cartesian axes, Ry/au):
atom 1 type 1 force = 0.00000012 -0.00000087 -0.00000173
atom 2 type 1 force = -0.00000012 0.00000087 0.00000173
atom 3 type 1 force = -0.00000012 0.00000087 -0.00000173
atom 4 type 1 force = 0.00000012 -0.00000087 0.00000173
atom 5 type 3 force = 0.00000000 0.00000000 0.00000000
atom 6 type 2 force = -0.00000000 0.00000000 0.00000000
atom 7 type 2 force = 0.00000000 -0.00000000 -0.00000000
atom 8 type 3 force = -0.00000000 -0.00000000 0.00000000
atom 9 type 4 force = 0.00000005 -0.00000028 0.00000287
atom 10 type 4 force = -0.00000005 0.00000028 -0.00000287
atom 11 type 4 force = 0.00000119 0.00000007 -0.00000115
atom 12 type 4 force = -0.00000119 -0.00000007 0.00000115
atom 13 type 4 force = -0.00000119 -0.00000007 -0.00000115
atom 14 type 4 force = 0.00000119 0.00000007 0.00000115
atom 15 type 4 force = -0.00000118 -0.00000098 -0.00000108
atom 16 type 4 force = 0.00000118 0.00000098 0.00000108
atom 17 type 4 force = 0.00000118 0.00000098 -0.00000108
atom 18 type 4 force = -0.00000118 -0.00000098 0.00000108
atom 19 type 4 force = -0.00000005 0.00000028 0.00000287
atom 20 type 4 force = 0.00000005 -0.00000028 -0.00000287
Computing stress (Cartesian axis) and pressure
total stress (Ry/bohr**3) (kbar) P= -0.00
0.00000000 0.00000000 0.00000000 0.00 0.00 0.00
0.00000000 -0.00000002 0.00000000 0.00 -0.00 0.00
0.00000000 0.00000000 -0.00000000 0.00 0.00 -0.00
Message from routine bfgs:
history already reset at previous step: stopping
bfgs converged in 30 scf cycles and 29 bfgs steps
(criteria: energy < 1.4E-09 Ry, force < 1.0E-06Ry/Bohr, cell < 5.0E-01kbar)
End of BFGS Geometry Optimization
Final enthalpy = -3840.8280670928 Ry
Begin final coordinates
new unit-cell volume = 1540.07698 a.u.^3 ( 228.21586 Ang^3 )
density = 7.15506 g/cm^3
CELL_PARAMETERS (alat= 10.28401600)
0.992999501 0.012644330 0.000000000
-0.018068355 1.404995305 0.000000000
0.000000000 0.000000000 1.014750287
ATOMIC_POSITIONS (crystal)
La 0.499998392 0.250000173 -0.003228188
La 0.500001608 0.749999827 0.003228188
La 0.000001608 0.749999827 0.496771812
La -0.000001608 0.250000173 0.503228188
Co2 0.000000000 0.000000000 0.000000000
Co1 -0.000000000 0.500000000 0.000000000
Co1 0.500000000 0.500000000 0.500000000
Co2 0.500000000 0.000000000 0.500000000
O 0.000000937 0.250000107 0.072799662
O -0.000000937 0.749999893 0.927200338
O 0.250055043 0.962702639 0.249942716
O 0.749944957 0.037297361 0.750057284
O 0.249944957 0.037297361 0.749942716
O 0.750055043 0.962702639 0.250057284
O 0.749947635 0.537297843 0.249945383
O 0.250052365 0.462702157 0.750054617
O 0.750052365 0.462702157 0.749945383
O 0.249947635 0.537297843 0.250054617
O 0.499999063 0.749999893 0.572799662
O 0.500000937 0.250000107 0.427200338
End final coordinates
OUTPUT FROM FINAL SCF RUN:
A final scf calculation at the relaxed structure.
The G-vectors are recalculated for the final unit cell
Results may differ from those at the preceding step.
Parallelization info
--------------------
sticks: dense smooth PW G-vecs: dense smooth PW
Min 938 469 129 58844 20811 2997
Max 939 471 130 58845 20812 2999
Sum 9385 4699 1291 588445 208113 29981
bravais-lattice index = 12
lattice parameter (alat) = 10.2840 a.u.
unit-cell volume = 1540.0770 (a.u.)^3
number of atoms/cell = 20
number of atomic types = 4
number of electrons = 184.00
number of Kohn-Sham states= 110
kinetic-energy cutoff = 100.0000 Ry
charge density cutoff = 800.0000 Ry
convergence threshold = 2.3E-11
mixing beta = 0.0500
number of iterations used = 8 plain mixing
Exchange-correlation = SLA PW PBX PBC ( 1 4 3 4 0 0)
celldm(1)= 10.213861 celldm(2)= 1.416338 celldm(3)= 1.019956
celldm(4)= -0.007299 celldm(5)= 0.000000 celldm(6)= 0.000000
…..
atomic species valence mass pseudopotential
La 11.00 138.90550 La( 1.00)
Co1 17.00 58.93320 Co( 1.00)
Co2 17.00 58.93320 Co( 1.00)
O 6.00 16.00000 O ( 1.00)
Starting magnetic structure
atomic species magnetization
La -0.000
Co1 -0.069
Co2 0.069
O 0.000
…..
Forces acting on atoms (cartesian axes, Ry/au):
atom 1 type 1 force = -0.00001228 0.00097281 0.00036384
atom 2 type 1 force = 0.00001228 -0.00097281 -0.00036384
atom 3 type 1 force = 0.00001228 -0.00097281 0.00036384
atom 4 type 1 force = -0.00001228 0.00097281 -0.00036384
atom 5 type 3 force = 0.00000000 0.00000000 0.00000000
atom 6 type 2 force = 0.00000000 0.00000000 0.00000000
atom 7 type 2 force = 0.00000000 0.00000000 -0.00000000
atom 8 type 3 force = 0.00000000 0.00000000 0.00000000
atom 9 type 4 force = 0.00000007 -0.00000556 0.00123004
atom 10 type 4 force = -0.00000007 0.00000556 -0.00123004
atom 11 type 4 force = -0.00000878 0.00068623 0.00000095
atom 12 type 4 force = 0.00000878 -0.00068623 -0.00000095
atom 13 type 4 force = 0.00000878 -0.00068623 0.00000095
atom 14 type 4 force = -0.00000878 0.00068623 -0.00000095
atom 15 type 4 force = 0.00000535 -0.00045735 0.00000063
atom 16 type 4 force = -0.00000535 0.00045735 -0.00000063
atom 17 type 4 force = -0.00000535 0.00045735 0.00000063
atom 18 type 4 force = 0.00000535 -0.00045735 -0.00000063
atom 19 type 4 force = -0.00000007 0.00000556 0.00123004
atom 20 type 4 force = 0.00000007 -0.00000556 -0.00123004
Total force = 0.003618 Total SCF correction = 0.000003
Computing stress (Cartesian axis) and pressure
total stress (Ry/bohr**3) (kbar) P= 5.83
0.00004353 0.00000056 0.00000000 6.40 0.08 0.00
0.00000056 0.00002911 0.00000000 0.08 4.28 0.00
0.00000000 0.00000000 0.00004631 0.00 0.00 6.81
Best,
Hien Vo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20200813/8d35e31a/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 52CC53ED739E4F75AF25E385C1A334B5.png
Type: image/png
Size: 144 bytes
Desc: 52CC53ED739E4F75AF25E385C1A334B5.png
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20200813/8d35e31a/attachment.png>
More information about the users
mailing list