# Spin polarized DFT calculation in Quantum ESPRESSO

I am trying to run spin polarized DFT calculation on system Ti2FeSn.

I am doing vc-relax calculation to relax the unit cell but having problem to converge total energy. Total magnetization is also sometimes increases and sometimes decreases on each iteration step and final total magnetization is also getting converged to '0'. This system is supposed to be half metallic system with magnetic moment of 2 Bohr mag/cell (as per the reference https://doi.org/10.1016/j.intermet.2013.01.005). I am trying to run DFT calculation to see if I can converge to the same results.
I have also attached the pwscf input file which I am using for the calculation in Quantum Espresso:

&CONTROL
title = 'Ti2FeSn' ,
calculation = 'vc-relax' ,
outdir = '.' ,
pseudo_dir = '.' ,
prefix = 'Ti2FeSn' ,
etot_conv_thr = 1.0D-6 ,
forc_conv_thr = 1.0D-6 ,
tstress = .true. ,
tprnfor = .true. ,
/
&SYSTEM
ibrav = 2,
celldm(1) = 10,
nat = 4,
ntyp = 3,
ecutwfc = 40 ,
ecutrho = 400 ,
occupations = 'smearing' ,
smearing = 'marzari-vanderbilt' ,
degauss = 0.01,
nspin = 2 ,
starting_magnetization(1) = 0.2,
starting_magnetization(2) = 0.4,
starting_magnetization(3) = 0.2,

/
&ELECTRONS
/
&IONS
/
&CELL
/
ATOMIC_SPECIES
Ti   55.84500  ti_pbe_v1.4.uspp.F.UPF
Fe  118.71000  Fe.pbe-spn-kjpaw_psl.0.2.1.UPF
Sn   47.86700  Sn_pbe_v1.uspp.F.UPF
ATOMIC_POSITIONS alat
Ti      0.000000000    0.000000000    0.000000000
Ti      0.250000000    0.250000000    0.250000
Fe      0.500000000    0.500000000    0.500000000
Sn     -0.250000000   -0.250000000   -0.250000000
K_POINTS automatic
6 6 6   0 0 0


Any suggestion on how can I resolve this issue and get same results as in the reference will be helpful.
Thanking you!

• I should began changing the convergence tolerance to a lower value, do an optimization, increase, do an optimization, etc. – Camps Oct 30 '20 at 21:05
• Why is your plane wave cutoff so low? 'ecutwfc=40' - This is where the problem likely stems from. It is too low for a system with elements like Fe, Ti and Sn which have atomic numbers >25. Convergence obviously varies from system to system, but this plane wave cutoff you are using is too low. You should try something like 90-120 - As a sidenote, you can also look up materialscloud pseudopotential convergence data (SSSP efficiency ) : materialscloud.org/discover/sssp – Xivi76 Oct 31 '20 at 6:09
• I agree with the cutoff being too low. Also your atomic masses are mixed up; Iron is 55.845, Ti is 47.867 and Sn is 118.71. This can affect the relaxation. Also, the reference you posted is an all-electron calculation, so keep that in mind. – Kevin J. May Oct 31 '20 at 6:35
• @Kevin J. M. what does it mean by mixed up masses. I didn't get that. Can you please explain more? – Ujjawal M. Oct 31 '20 at 7:40
• @UjjawalM. The masses for your atomic species are wrong! The numerical values of the three masses are correct but are assigned to the wrong atoms. Again, the masses are supposed to be: Fe = 55.845, Ti = 47.867 and Sn = 118.71. – Kevin J. May Nov 1 '20 at 3:00

• The calculated structure is the following:

• The self-consistent input file for QE calculation:

&CONTROL
calculation = 'scf'
etot_conv_thr =   4.0000000000d-05
forc_conv_thr =   1.0000000000d-04
outdir = './out/'
prefix = 'aiida'
pseudo_dir = './pseudo/'
tprnfor = .true.
tstress = .true.
verbosity = 'high'
/
&SYSTEM
degauss =   1.4699723600d-02
ecutrho =   1.0800000000d+03
ecutwfc =   9.0000000000d+01
ibrav = 0
nat = 4
nspin = 2
ntyp = 3
occupations = 'smearing'
smearing = 'cold'
starting_magnetization(1) =   3.1250000000d-01
starting_magnetization(2) =   1.0000000000d-01
starting_magnetization(3) =   4.1666666667d-01
/
&ELECTRONS
conv_thr =   8.0000000000d-10
electron_maxstep = 80
mixing_beta =   4.0000000000d-01
/
ATOMIC_SPECIES
Fe     55.847 Fe.pbe-spn-kjpaw_psl.0.2.1.UPF
Sn     118.71 Sn_pbe_v1.uspp.F.UPF
Ti     47.88 ti_pbe_v1.4.uspp.F.UPF
ATOMIC_POSITIONS crystal
Ti           0.0000000000       0.0000000000       0.0000000000
Ti           0.7500000000       0.7500000000       0.7500000000
Fe           0.5000000000       0.5000000000       0.5000000000
Sn           0.2500000000       0.2500000000       0.2500000000
K_POINTS automatic
11 11 11 0 0 0
CELL_PARAMETERS angstrom
3.7418479919       0.0000000000       0.0000000000
1.8709239960       3.2405354181       0.0000000000
1.8709239960       1.0801784727       3.0552060918

• Output magnetization:

total magnetization       =     3.64 Bohr mag/cell
total magnetization       =     1.70 Bohr mag/cell
total magnetization       =     2.00 Bohr mag/cell
total magnetization       =     1.94 Bohr mag/cell
total magnetization       =     1.96 Bohr mag/cell
total magnetization       =     1.55 Bohr mag/cell
total magnetization       =     1.24 Bohr mag/cell
total magnetization       =     1.07 Bohr mag/cell
total magnetization       =     1.07 Bohr mag/cell
total magnetization       =     1.04 Bohr mag/cell
total magnetization       =     1.04 Bohr mag/cell
total magnetization       =     1.05 Bohr mag/cell
total magnetization       =     1.04 Bohr mag/cell
total magnetization       =     1.04 Bohr mag/cell
total magnetization       =     1.03 Bohr mag/cell
total magnetization       =     1.03 Bohr mag/cell
total magnetization       =     1.02 Bohr mag/cell
total magnetization       =     1.02 Bohr mag/cell
total magnetization       =     1.02 Bohr mag/cell
total magnetization       =     1.01 Bohr mag/cell
total magnetization       =     1.01 Bohr mag/cell
total magnetization       =     1.01 Bohr mag/cell
total magnetization       =     1.01 Bohr mag/cell

• A useful website to generate the input file for QE calculation: https://www.materialscloud.org/work/tools/qeinputgenerator

• thanks for your answer, but space group of sample is 225. I think structure you have shown is not correct. Can you tell me how much time it took to complete the calculation in your machine? In my machine calculation is everlasting that it is not getting job done. – Ujjawal M. Nov 7 '20 at 6:57
• +1. For such a detailed answer. I can tell a lot of work went into it. Ujjawal would probably like an answer to your comment though. – Nike Dattani Nov 8 '20 at 0:12
• @UjjawalM. Five minutes. Can you upload your structure with XXX.vasp format? – Jack Nov 8 '20 at 0:31
• I have no knowledge about the VASP code and I am also finding it hard to upload a file on this site. But I have uploaded an image of the structure build in xcrysden software. You may open the structure in xcrysden using the pwscf input file which I have provided. Thanks! – Ujjawal M. Nov 8 '20 at 4:57
• @UjjawalM. Did you try after rectifying the wrong atomic masses? Since its a vc-relax calculation I guess it matters. – Anoop A Nair Nov 8 '20 at 17:32