Hands on Session IV
Martijn MARSMAN
Institut für Materialphysik and Center for Computational Material Science
Universität Wien, Sensengasse 8/12, A-1090 Wien, Austria
b-initio
ackage
ienna imulation
M. M ARSMAN , H ANDS ON (4): MAGNETISM Page 1
Overview
fcc Ni, an elementary ferromagnetic metal
NiO, antiferromagnetic coupling
LSDA+U (Dudarev’s approach)
SOI: freestanding fcc Fe and Ni (100) monolayers
Constraining magnetic moments
What to do about convergence problems?
M. M ARSMAN , H ANDS ON (4): MAGNETISM Page 2
fcc Ni
POSCAR
fcc:
-10.93 Volume set to 10.93 Å
0.5 0.5 0.0
0.0 0.5 0.5 fcc primitive cell
0.5 0.0 0.5
1 KPOINTS
Cartesian
11 11 11 Γ-centered
0 0 0
Monkhorst-Pack grid
POTCAR
k-points
0 makepaw_GGA Ni
Gamma (a PAW-GGA PW91 potential)
11 11 11
0 0 0
M. M ARSMAN , H ANDS ON (4): 4 1 N I Page 3
INCAR
SYSTEM = Ni fcc bulk
ISTART = 0 Spin polarized calculation (collinear)
ISPIN = 2
MAGMOM = 1.0 Initial magnetic moment: 1 µB
ISMEAR = -5 Interpolation of the correlation part of
VOSKOWN = 1
the exchange-correlation functional
LORBIT = 11
according to:
Or copy the files from: S. H. Vosko, L. Wilk and M. Nusair, Can.
J. Phys. 58, 1200 (1980).
˜vw/4_1_Ni
k-mesh integration:
tetrahedron method with Blöchl’s correc-
tions
Orbital resolved DOS
and calculation of local magnetic moment
M. M ARSMAN , H ANDS ON (4): 4 1 N I Page 4
The magnetic moment
In OSZICAR (total magnetic moment):
N E dE d eps ncg rms rms(c)
DAV: 1 0.139935173959E+02 0.13994E+02 -0.35801E+03 2338 0.828E+02
DAV: 2 -0.623612680591E+01 -0.20230E+02 -0.19281E+02 2282 0.123E+02
DAV: 3 -0.643764005251E+01 -0.20151E+00 -0.19906E+00 2536 0.140E+01
DAV: 4 -0.643786482872E+01 -0.22478E-03 -0.22442E-03 2344 0.459E-01
DAV: 5 -0.643786514671E+01 -0.31798E-06 -0.31687E-06 1832 0.173E-02 0.793E+00
...
DAV: 9 -0.545953126374E+01 0.48409E-02 -0.96206E-03 2946 0.839E-01 0.847E-02
DAV: 10 -0.545946513577E+01 0.66128E-04 -0.77007E-05 1364 0.126E-01
1 F= -.54594651E+01 E0= -.54594651E+01 d E =0.000000E+00 mag= 0.5781
in OUTCAR (integration of magnetic moment in the PAW sphere):
magnetization (x)
# of ion s p d tot
----------------------------------------
1 -0.007 -0.026 0.625 0.591
M. M ARSMAN , H ANDS ON (4): 4 1 N I Page 5
DOS
fcc Ni
3
n(E) (states/eV atom)
2
1
0
−1
−2
−3
−4 −2 0 2 4 6
E(eV)
Exchange splitting 0.5 eV
M. M ARSMAN , H ANDS ON (4): 4 1 N I Page 6
Proper initialization of magnetic moment
Too small initial moment will/may lead to a nonmagnetic solution
(the previous example with MAGMOM = 0.0)
...
DAV: 9 -0.540773198300E+01 0.31931E-03 -0.39421E-04 2091 0.339E-01 0.300E-02
DAV: 10 -0.540780668590E+01 -0.74703E-04 -0.46454E-05 1059 0.106E-01
1 F= -.54078067E+01 E0= -.54078067E+01 d E =0.000000E+00 mag= 0.0020
Badly initialized calculations take longer to converge
Coexistence of low- and high spin solutions
M. M ARSMAN , H ANDS ON (4): 4 1 N I Page 7
Noncollinear magnetism
Replace ISPIN = 2 and MAGMOM = 1.0 by:
LNONCOLLINEAR = .TRUE.
MAGMOM = 0.0 0.0 1.0
leads to
DAV: 9 -0.546480633680E+01 0.41628E-02 -0.49402E-04 7532 0.330E-01 0.695E-02
DAV: 10 -0.546475032360E+01 0.56013E-04 -0.52286E-05 4328 0.446E-02
1 F= -.54647503E+01 E0= -.54647503E+01 d E =0.000000E+00 mag= 0.0000 0.0000 0.5792
or with MAGMOM = 1.0 0.0 0.0
DAV: 9 -0.546481348871E+01 0.41496E-02 -0.50294E-04 7548 0.330E-01 0.692E-02
DAV: 10 -0.546474438319E+01 0.69106E-04 -0.51451E-05 4288 0.432E-02
1 F= -.54647444E+01 E0= -.54647444E+01 d E =0.000000E+00 mag= 0.5792 0.0000 0.0000
idem for MAGMOM = 0.0 1.0 0.0
DAV: 9 -0.546481179459E+01 0.41515E-02 -0.50430E-04 7552 0.330E-01 0.692E-02
DAV: 10 -0.546474640011E+01 0.65394E-04 -0.51658E-05 4292 0.434E-02
1 F= -.54647464E+01 E0= -.54647464E+01 d E =0.000000E+00 mag= 0.0000 0.5792 0.0000
M. M ARSMAN , H ANDS ON (4): 4 1 N I Page 8
NiO
Rocksalt structure
AFM ordering of Ni (111) planes
M. M ARSMAN , H ANDS ON (4): 4 2 N I O Page 9
POSCAR
NiO AFM
4.17 AFM coupling:
1.0 0.5 0.5 4 atoms in the basis (instead of 2)
0.5 1.0 0.5
0.5 0.5 1.0 KPOINTS
2 2
Cartesian 4 4 4 Γ-centered
0.0 0.0 0.0 Monkhorst-Pack grid
1.0 1.0 1.0
0.5 0.5 0.5 POTCAR
1.5 1.5 1.5
makepaw Ni O_s
k-points (PAW-LDA potentials)
0
Gamma
4 4 4
0 0 0
M. M ARSMAN , H ANDS ON (4): 4 2 N I O Page 10
INCAR
SYSTEM = NiO
Initial magnetic moment:
ISPIN = 2 2 µB (Ni), 0 µB (O)
MAGMOM = 2.0 -2.0 2*0
AMIX=0.2 and AMIX MAG=0.8 (default)
ENMAX = 250 BMIX and BMIX MAG practically zero,
EDIFF = 1E-3
i.e. linear mixing
ISMEAR = -5
Or copy the files from:
AMIX = 0.2
BMIX = 0.00001 ˜vw/4_2_NiO
AMIX_MAG = 0.8
BMIX_MAG = 0.00001
LORBIT = 11
M. M ARSMAN , H ANDS ON (4): 4 2 N I O Page 11
The magnetic moment
In OSZICAR (total magnetic moment = 0!):
N E dE d eps ncg rms rms(c)
...
DAV: 13 -0.267936242334E+02 0.12794E-03 -0.12638E-04 552 0.298E-01 0.169E-02
DAV: 14 -0.267936352231E+02 -0.10990E-04 -0.21775E-05 520 0.107E-01
1 F= -.26793635E+02 E0= -.26793635E+02 d E =0.000000E+00 mag= 0.0000
in OUTCAR (integration of magnetic moment in the PAW sphere):
magnetization (x)
# of ion s p d tot
----------------------------------------
1 -0.012 -0.014 1.245 1.219
2 0.012 0.014 -1.242 -1.216
3 0.000 -0.001 0.000 -0.001
4 0.000 -0.001 0.000 -0.001
------------------------------------------------
tot 0.000 -0.003 0.003 0.000
M. M ARSMAN , H ANDS ON (4): 4 2 N I O Page 12
Total DOS, and LDOS Ni d-orbitals
DOS Ni d−LDOS
8
4 t2g
6
n(E) (states/eV atom)
n(E) (states/eV atom)
eg
4
2
2
0 0
−2
−2
−4
−6 −4
−8
−4 −2 0 2 4 6 8 10 −4 −2 0 2 4 6 8 10
E(eV) E(eV)
mNi = 1.21 µB (exp. 1.70 µB ) Egap = 0.44 eV (exp. 4.0 eV)
M. M ARSMAN , H ANDS ON (4): 4 2 N I O Page 13
LSDA+U; Dudarev’s approach
... addition to INCAR of NiO calc.
LDAU = .TRUE.
Switch on L(S)DA+U
LDAUTYPE = 2
LDAUL = 2 -1 Select Dudarev’s approach
LDAUU = 8 00 0.00 (LSDA+U Type 2)
LDAUJ = 0.95 0.00
LDAUPRINT = 2 L quantum number for which on site interac-
tion is added
Or copy the files from: (-1 = no on site interaction)
˜vw/4_3_NiO_LSDA+U U parameter
J parameter
Print occupation matrices in OUTCAR
L,U, and J must be specified for all atomic types!
M. M ARSMAN , H ANDS ON (4): 4 3 N I O LSDA+U Page 14
On site occupancies (see OUTCAR)
atom = 1 type = 1 l = 2
onsite density matrix
...
...
occupancies and eigenvectors
o = 0.1696 v = 0.0000 0.0000 0.0000 0.0000 0.0000 0.0013 -0.0006 -0.9999 -0.0007 -0.0104
o = 0.1696 v = 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 -0.0011 -0.0104 0.0011 0.9999
o = 0.9770 v = 0.0000 0.0000 0.0000 0.0000 0.0000 0.7787 -0.1766 0.0015 -0.6020 0.0005
o = 0.9770 v = 0.0000 0.0000 0.0000 0.0000 0.0000 0.2456 -0.7972 0.0005 0.5516 -0.0015
o = 0.9770 v = 0.0000 0.0000 0.0000 0.0000 0.0000 0.5774 0.5774 0.0000 0.5774 0.0000
o = 0.9803 v = -0.0193 0.7166 0.0001 -0.6972 -0.0039 0.0000 0.0000 0.0000 0.0000 0.0000
o = 0.9803 v = 0.8163 -0.3914 -0.0039 -0.4249 -0.0001 0.0000 0.0000 0.0000 0.0000 0.0000
o = 0.9803 v = 0.5774 0.5774 0.0000 0.5774 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
o = 1.0248 v = -0.0032 0.0016 -1.0000 0.0016 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
o = 1.0248 v = 0.0000 0.0027 0.0000 -0.0027 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000
dxy dyz d z2 dxz dx2 dxy dyz dz2 dxz dx2
r2 y2 r2 y2
M. M ARSMAN , H ANDS ON (4): 4 3 N I O LSDA+U Page 15
For comparison:
when U=0 and J=0 (i.e. just LSDA) the on site occupancies are as follows:
o = 0.3462 v = 0.0000 0.0000 0.0000 0.0000 0.0000 -0.0048 0.0028 0.9951 0.0020 -0.0986
o = 0.3462 v = 0.0000 0.0000 0.0000 0.0000 0.0000 0.0005 0.0039 -0.0986 -0.0044 -0.9951
o = 0.9491 v = 0.0000 0.0000 0.0000 0.0000 0.0000 0.5774 0.5774 0.0000 0.5774 0.0000
o = 0.9495 v = 0.0000 0.0000 0.0000 0.0000 0.0000 -0.0588 0.7347 -0.0004 -0.6759 0.0059
o = 0.9495 v = 0.0000 0.0000 0.0000 0.0000 0.0000 0.8144 -0.3563 0.0059 -0.4581 0.0004
o = 0.9527 v = 0.0477 -0.0256 0.9974 -0.0221 -0.0420 0.0000 0.0000 0.0000 0.0000 0.0000
o = 0.9527 v = 0.0020 0.0403 0.0420 -0.0423 0.9974 0.0000 0.0000 0.0000 0.0000 0.0000
o = 0.9598 v = 0.5774 0.5774 0.0000 0.5774 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
o = 0.9599 v = -0.1186 0.7577 0.0085 -0.6391 -0.0579 0.0000 0.0000 0.0000 0.0000 0.0000
o = 0.9599 v = 0.8064 -0.3005 -0.0579 -0.5059 -0.0085 0.0000 0.0000 0.0000 0.0000 0.0000
dxy dyz d z2 dxz dx2 dxy dyz dz2 dxz dx2
r2 y2 r2 y2
M. M ARSMAN , H ANDS ON (4): 4 3 N I O LSDA+U Page 16
The Ni d-LDOS and local magnetic moment
LSDA Dudarev U=8 J=0.95
4 t2g 4 t2g
n(E) (states/eV atom)
n(E) (states/eV atom)
eg eg
2 2
0 0
−2 −2
−4 −4
−4 −2 0 2 4 6 8 10 −4 −2 0 2 4 6 8 10
E(eV) E(eV)
magnetization (x)
# of ion s p d tot
----------------------------------------
1 -0.003 -0.006 1.721 1.711
2 0.003 0.006 -1.719 -1.710
3 0.000 -0.001 0.000 -0.001
4 0.000 -0.001 0.000 -0.001
------------------------------------------------
tot 0.000 -0.002 0.002 0.000
M. M ARSMAN , H ANDS ON (4): 4 3 N I O LSDA+U Page 17
Total Energy
On site occupancy matrix is NOT idempotent
Total energy contains penalty contribution!
...
DAV: 15 -0.229633055256E+02 -0.11057E-03 -0.50020E-05 520 0.104E-01 0.118E-02
DAV: 16 -0.229633263321E+02 -0.20806E-04 -0.16650E-05 520 0.492E-02
1 F= -.22963326E+02 E0= -.22963326E+02 d E =0.000000E+00 mag= 0.0000
The total energy for U J 0 is in that case always higher than for U J 0
(just LSDA, see below):
...
DAV: 13 -0.267936242334E+02 0.12794E-03 -0.12638E-04 552 0.298E-01 0.169E-02
DAV: 14 -0.267936352231E+02 -0.10990E-04 -0.21775E-05 520 0.107E-01
1 F= -.26793635E+02 E0= -.26793635E+02 d E =0.000000E+00 mag= 0.0000
Comparing the total energies from calculations with different U J is meaningless!
M. M ARSMAN , H ANDS ON (4): 4 3 N I O LSDA+U Page 18
SOI: freestanding fcc Fe and Ni (100) monolayers
POSCAR
fcc Ni (100) monolayer
3.53 Lattice constant for bulk fcc Ni
.50000 .50000 .00000 (for Fe take a0 3 45 Å)
-.50000 .50000 .00000
.00000 .00000 5.00000 POTCAR
1
Cartesian makepaw_GGA Ni
.00000 .00000 .00000 or
makepaw_GGA Fe
K-Points
0
Monkhorst-Pack
9 9 1
0 0 0
M. M ARSMAN , H ANDS ON (4): 4 4 SOI N I AND 4 4 SOI F E Page 19
INCAR
ISTART = 0
ENCUT = 270.00 Initialize moment along z-direction
LNONCOLLINEAR = .TRUE.
(out of plane)
MAGMOM = 0.0 0.0 1.0
VOSKOWN = 1
For Fe: MAGMOM = 0.0 0.0 3.0
LSORBIT = .TRUE. Switch on Spin-Orbit Interaction
For the second calculation, switch to in-plane magnetization, setting
MAGMOM = 1.0 0.0 0.0 (for Fe: MAGMOM = 3.0 0.0 0.0)
Input files can be found in:
˜vw/4_4_SOI_Ni and ˜vw/4_4_SOI_Fe
M. M ARSMAN , H ANDS ON (4): 4 4 SOI N I AND 4 4 SOI F E Page 20
Results
fcc Ni (100) monolayer (out of plane magnetization)
...
DAV: 20 -0.371322930070E+01 0.15852E-03 -0.11632E-03 636 0.235E-01 0.225E-02
DAV: 21 -0.371323204989E+01 -0.27492E-05 -0.13047E-05 500 0.184E-02
1 F= -.37132320E+01 E0= -.37139803E+01 d E =0.224478E-02 mag= 0.0000 0.0000 0.9035
fcc Ni (100) monolayer (in plane magnetization)
...
DAV: 19 -0.371443443024E+01 -0.80757E-04 -0.35822E-03 1084 0.323E-01 0.119E-02
DAV: 20 -0.371446032472E+01 -0.25894E-04 -0.42423E-05 916 0.263E-02
1 F= -.37144603E+01 E0= -.37150300E+01 d E =0.170900E-02 mag= 0.9049 0.0000 0.0000
EMAE E m E m 1 2 meV (easy axis lies in plane)
For Fe the same procedure yields
EMAE E m E m 0 2 meV (easy axis lies out of plane)
M. M ARSMAN , H ANDS ON (4): 4 4 SOI N I AND 4 4 SOI F E Page 21
Constraining the direction of magnetic moments
POSCAR
Fe dimer
1.0 An iron dimer in a box
8.0 0.0 0.0
0.0 8.0 0.0
0.0 0.0 8.0 KPOINTS
2
Cartesian We only take the Γ point
3.0 0.0 0.0
5.0 0.0 0.0
POTCAR
k-points
makepaw_GGA Fe
0
Gamma
1 1 1 Or copy the files from:
0 0 0 ˜vw/4_5_Fe_dimer
M. M ARSMAN , H ANDS ON (4): 4 5 F E DIMER Page 22
INCAR
ISTART = 0
ISYM = 0 Switch of symmetry
LNONCOLLINEAR = .TRUE.
MAGMOM = 0 0 3 0 0 3 Initialize moments for
VOSKOWN = 1 ferromagnetic coupling
LORBIT = 11
...
DAV: 20 -0.929676054634E+01 -0.26101E-03 -0.16780E-03 60 0.102E-01 0.537E-02
DAV: 21 -0.929679955346E+01 -0.39007E-04 -0.30319E-04 60 0.590E-02
1 F= -.92967996E+01 E0= -.93047629E+01 d E =0.238900E-01 mag= -0.0006 -0.0003 6.0537
Now take MAGMOM = 0 0 3 0 2 2
magnetization (y) magnetization (z)
# of ion s p d tot # of ion s p d tot
---------------------------------------- ----------------------------------------
1 0.018 -0.001 1.071 1.087 1 0.045 -0.003 2.587 2.628
2 0.019 -0.001 1.069 1.087 2 0.045 -0.003 2.588 2.629
---------------------------------------- ----------------------------------------
tot 0.037 -0.003 2.140 2.174 tot 0.089 -0.007 5.175 5.257
System converges to FM solution
M. M ARSMAN , H ANDS ON (4): 4 5 F E DIMER Page 23
However when we add the following lines to the INCAR
I_CONSTRAINED_M = 1
RWIGS = 1.0 Switch on constraints
LAMBDA = 10 on magnetic moments
M_CONSTR = 0 0 1 0 1 1
Integration radius to determine
local moments
Weight in penalty functional
Target directions
a penalty functional is added to the system which drives the integrated local moments
into the desired directions.
Beware: The penalty functional contributes to the total energy
M. M ARSMAN , H ANDS ON (4): 4 5 F E DIMER Page 24
The necessary information is found in the OSZICAR:
E_p = 0.35424E-02 lambda = 0.100E+02
ion MW_int M_int
1 0.000 0.013 1.557 0.000 0.014 2.674
2 0.000 1.092 1.110 0.000 1.880 1.901
DAV: 35 -0.905322335169E+01 0.58398E-04 -0.60872E-04 60 0.734E-02
1 F= -.90532234E+01 E0= -.90355617E+01 d E =-.529849E-01 mag= -0.0005 2.1161 5.1088
E p is the energy arising from the penalty functional
It decreases with increasing LAMBDA!
By increasing LAMBDA stepwise one can bring E p down
(slowly so the solution remains stable from one run to another)
E_p = 0.22591E-03 lambda = 0.500E+02
ion MW_int M_int
1 0.000 0.002 1.545 0.001 -0.005 2.654
2 0.000 1.086 1.087 0.001 1.871 1.862
DAV: 33 -0.907152551238E+01 0.48186E-04 -0.33125E-04 60 0.163E-01
1 F= -.90715255E+01 E0= -.90541505E+01 d E =-.521251E-01 mag= 0.0042 2.0902 5.0659
This way one approaches the LSDA total energy for a given magnetic configuration
M. M ARSMAN , H ANDS ON (4): 4 5 F E DIMER Page 25
What can one do when convergence is bad?
Start from charge density of non-spin-polarized calculation, using
ISTART = 0 (or remove WAVECAR)
ICHARG = 1
Linear mixing
BMIX = 0.0001 ; BMIX MAG = 0.0001
Mix slowly, i.e., reduce AMIX and AMIX MAG
Reduce MAXMIX, the number of steps stored in the Broyden mixer
(default = 45)
Restart from partly converged results
(stop a calculation after say 20 steps and restart from the WAVECAR)
Use constraints to stabilize the magetic configuration
Pray
M. M ARSMAN , H ANDS ON (4): MAGNETISM Page 26
