Magnetism in DFT PDF

MAGNETISM IN DFT
FROM THEORY TO PRACTICE
(Paris, France)
Saclay
Cyrille Barreteau
CEA Saclay
(Copenhagen, Denmark)
Lyngby 2 JUILLET 2014 MASTANI, Pune, INDIA, July 2014 | PAGE 1
MAGNETISM
From the atomic nucleus to the inner core of Earth

time
Magnetic rotor
GMR Read Head
NMR
length
2 JUILLET 2014 MASTANI, Pune, INDIA, July 2014 | PAGE 2

MAGNETISM
Intense activity in which nanomagnetism plays a crucial role
Search for permanent hard magnets without rare earth
Replace Samarium-Cobalt, Neodyum by nanostructured « cheap » TM magnets
Spintronics From Fundamentals to applications
micromagnetics
DFT is the perfect tool to adress (at least partly) most of these problems
MASTANI, Pune, INDIA, July 2014 | PAGE 3
DENSITY FUNCTIONAL THEORY
Hohenberg & Kohn (1964) + Kohn Sham (1965)
n(r )n(r ') 3 3

E  n   T0  n    Vext (r )n(r )d 3r  d rd r ' EI  I  E xc  n 
1
2
 r r'
 2

   Veff (r )   (r )     (r )
2
n(r )      , (r )  2    (r )
2 2
 2m    occ  occ
H KS if 
Vext (r )  VHartree (r )  Vxc (r )
VHartree (r )  
n(r ') 3
d r'  Exc
r r' Vxc (r ) 
 n(r )
(pseudo)-potental describing the interaction of valence electrons

Vext (r ) with the ions (nucleus+core electrons)
Can also include a « true » external potential
Local density approximation (LDA)
Exc  n    n(r ) xc (n(r ))d 3r
 xc (n) : exchange correlation energy (per particle) of homogenous gas
 xc (n)   x (n)   c (n)

1
3  3n  3
 x ( n)     Hartree Fock in an homogenous jellium

4  
 c ( n)  F ( n) Parametrized from QMC
d 
Vxc (r )   (n  xc (n)) 
 dn  nn (r )

Generalized Gradient approximation (GGA)
Exc  n    n(r ) xc (n(r ),  n(r))d 3r
Exc  n    n(r ) xchom (n(r )) Fxc (n(r ),  n(r)) d 3r

 xchom (n) : exchange correlation energy (per particle) of homogenous gas
Fxc (n(r ),  n(r)) : dimensionless enhancement factor
Some important sum rules and other relevant conditions should be verified…
But still a large variety of functionals
Perdew and Wang (PW91)

Most popular
Perdew Burke and Enzerhof (PBE)
…………

Kohn Sham algorithm
DFT implementation diagram
Initial guess Combination of atomic densities
n(r )
Effective potential
Vext (r )  VHartree (r )  Vxc (r )
Solve Poisson equation
but not in Harris Functional scheme
Solve KS equation
 2

     (r )     (r )
2
 Veff (r )
charge mixing  2m 
Determine the Fermi level
electron density and total energy

n(r )  2    (r ) Etot  n(r )  T  Eband  Edc
2
 occ
Converged? Analayse results

Spin polarization
Spin moment operator (collinear case) spin moment
 s  g s  B
S
  B σ σ  z
 zspin   s , z    B    z  
gs  2  occ
       0 
 0 i  0 1 1 0 
y        z         or   
   0    
x
i 0  1 0  0 1
      
in collinear case
(r )     (r )     (r )  n  n
2 2
spin  
m z
 occ  occ
Orbital polarization
orbital moment operator orbital moment
 L   B
L
  B l m orb (r )    L 
 occ
Average orbital moment: usually small (quenched) in bulk

and strictly null if spin-orbit coupling (SOC) is ignored.
Local Spin density approximation (LSDA)
 xc (n , n )   x (n , n )   c ( n , n )
     
m
Alternative formulation nn n  
m  n  n 
n
 x (n,  )   x (n, 0)   x (n,1)   x (n, 0)  f x ( )
1 (1   ) 3  (1   ) 3  2  1 if   1
4 4
f x ( )  
2 3 1 0 if   0
1
2
 c (n,  )   c (n, 0)   c (n,1)   c (n, 0)  f c ( ) f c ( )  f x ( ) (Perdew Zunger)

Spin dependent potential
Veff  Vext (r )  VHartree (r )  Vxc (n(r ), m(r ))   Bxc (r )

  (n(r ), m(r))    (n(r), m(r)) 
Vxc (r )   xc (n  , n  )  n(r )  xc  Bxc (r )  n(r )  xc 
 n(r )   m(r ) 
Bxc (r ) exchange correlation magnetic field

Bext (r ) external magnetic field can be added
Spin polarized DFT implementation diagram
Initial guess Combination of atomic densities
  1 n (r )  (n(r )   m(r ))
1
2
+ initial moment (symmetry breaking)
Spin dependent potential

  1 Vext (r )  VHartree (r )  Vxc (r )
Solve the two KS equation

  1
 2
    
   2
 V ( r ) 
  ( r )      (r )
 2m 
eff
charge mixing
Determine the common Fermi level
EF  EF  EF
electron and spin density, total energy
n(r )  n (r )  n (r )
n (r )     (r ) Etot  n(r ), m(r ) 
  2
 
 occ
m(r )  n (r )  n (r )
Converged? Analayse results

ANALYSE RESULTS
What do we get out of spin-polarized DFT calculation

Pw(scf)
Etot  neq , meq  Total energy → find most stable structure! (not always easy..)
Pw(scf)
Total (and absolute) spin magnetic moment

M   n (r )  n  (r ) d 3r  M abs   n (r )  n  (r ) d 3r
Pw(nscf)
Spin polarized band structure (for up and down spins)
Pw(scf) Local analysis (many options…)
+projwf.x
ni ( E )      E   
2

+pp;x at
i ,  
DOS 
n( E , r )     (r )   E    
2
mi   iat,  
2 2
 iat,   Atomic moment
 occ
Local moment
m(r )     (r )    (r )
2 2
Spin density
2 JUILLET 2014  occ MASTANI, Pune, INDIA, July 2014 | PAGE 11
PHYSICAL INSIGHT
DFT is a very powerful tool but there is still a need for simpler phenomenological
models and local analysis to get a deeper physical understanding of the phenomena
Stoner model
Heisenberg/Ising model
Local analysis mi

LSDA and the Stoner model
 xc (n,  )   xc (n, 0)  12  xc" (n, 0) 2  o( 2 )

1 "
Bxc (r )    xc (n(r ), 0)m(r )
n(r )
  M : magnetic moment per atom

Veff  V  0
IM
1  xc" (n(r ), 0) 3
eff
2

M  m(r )d r I    d r0
3
  n(r )
Rigid shift of up and down eigenvalues
 

n (E)  n (E 0
IM )     0  IM
2 2
EF
N  (n 0 ( E  12 IM )  n 0 ( E  12 IM )) dE  EF ( M )
E EF
M    ( E  12 IM )) dE
0 1 0
(n ( E IM ) n
EF 2
 F (M )  M

n ( E )  n 0 ( E  IM )
2 JUILLET 2014 2 MASTANI, Pune, INDIA, July 2014 | PAGE 13
Stoner criterion
M max
F (M )
Non zero solution if F '(0)  1 M  F (M  )
M 0 M1
In0 ( EF )  1
 M max
I 1eV In most elements

 n0 ( EF ) plays a crucial role in magnetic
susceptibility and onset of magnetism
Criterion can be also derived by analyzing
M 0
• magnetic susceptibility  
H (1  In0 ( EF ))
• Total energy Etot      12  Imi2
 occ i
Nowadays Stoner model often used in parametrized Tight-Binding models


H ij  H ij0  IM i ij
2 JUILLET 2014 2 MASTANI, Pune, INDIA, July 2014 | PAGE 14
General trends
• Low coordination favors magnetism
n( E )
Z  n( EF )
Spin magnetic moment is generally enhanced on

W Z low coordinated atoms
E
Z : number of neigbors

General trends
• Lattice expansion generally favors magnetism and vice versa…
n( E )
R W  n( EF )
 q ( R / R0 ) 1 E
W e
M (d ) E(d )
Fe bcc Pd fcc Pd fcc
Cr bcc Rh6

Multiple magnetic solutions
Stoner criterium does not tell the whole story
unstable stable
stable
F ( M in )  M out
M in  M out ?
The iterative scheme converges towards a solution which depends on initial spin moment
How to be sure not to miss the most stable solution?
• Several starting magnetization

(fishing strategy)
• Fix spin moment (FSM) calculation

allow to explore E(V,M) energy surface
Fixed Spin moment
Conventional scheme FSM
EF  EF N  N  N M ini  M  EF  EF N  N  N M  N  N
one scf loop many calculations
E E
EF  0
M  M ini
M M
Emin EF  2 H ( M )
E
E EF  0
M  M ini
M M
Emin
Penalization technique
Add a penalization term to the total energy functional to impose a given condition
• DFT
E  n, m   Etot  n, m     (m(r )  m 0 (r )) 2 d 3r

Minimization → modified KS Hamiltonian
H   H  2 (m(r )  m0 (r ))  .
• TB
E ci   Etot ci     mi  m0 
2
   ci iat
i
H ij  H ij  2 (mi  m 0 ). ij

SOME IMPORTANT TECHNICAL POINTS
Before running (or trusting) a DFT calculation you should be aware of

several important technicalities
Pseudopotential (LDA pz /GGA pw91, pbe)
Energy cut-off (ecutwfc, ecutrho≥8ecutwfc for US pseudo)

K-points sampling: denser mesh for metals
note that in QE the #k points is doubled in spin-polarized system
Smearing (smearing, degausss) Marzari Vanderbilt recommended
Initial magnetization (starting_magnetization)
Number of computed eigenvalues (nbnd) often needs to be increased
Convergence threshold (conv_thr)

pseudopotential
As a rule of thumb: aLDA <aexp<aGGA

LDA bad for 3d but OK for 5d→ problematic for alloys (FePt, CoPt)
E(V)
bcc FM
fcc NM
bcc FM hcp NM

Initial magnetization
Cr bcc AF
Define the system as simple cubic with 2 types of atoms
per unit cell and opposite starting magnetization
ibrav=1
starting_magnetization(1)= 0.5
nat=2 starting_magnetization(2)=-0.5
ntyp=2
2
Cr SDW m  m0 cos(q.R ) q  0.953 0.95 in unit of
a0
1 2
L  20a0  q  (1  )
20 a0
2 JUILLET 2014 | PAGE 22

MASTANI, Pune, INDIA, July 2014
NON COLLINEAR MAGNETISM
Why should we care about non collinearity?
  Mn
Non-collinearity does exist!
The future is in skyrmions 

Spin excitation
Spin gymnastics
( x, y, z ) global axis ( x ", y ", z ") local axis
"  sin  cos   0

   
m  m  sin  sin   m"  m 0
 cos   1
   
Spin ½ rotation matrix
" "
 e  i 2 cos(  ) e  i 2 sin(  ) 
 
 i 2  z  i 2  y
U ( ,  )   2 2
 e e
 e 2 sin(  ) e 2 cos(  ) 
 i 
i 
 2 2 
σ i "  U ( ,  )σ iU † ( ,  )
σ   x ,  y ,  z  also transforms like a space vector
2 JUILLET 2014
σ i "  Rij σ j MASTANI, Pune, INDIA, July 2014 | PAGE 24
2x2 matrices algebra
M  M2( )
M  aI  b.σ a  12 Tr (M ) b  12 Tr (M.σ )
If M is hermitian then a and b are real numbers
diagonalization 2 eigenvalues a  b
 sin  cos  
 
b b n n   sin  sin  
 cos  
 
a  b 0 †
M U  U
 0 a b 
 e  i 2 cos(  ) e  i 2 sin(  ) 
 
U   2 2

 e  i 2 sin(  ) ei 2 cos(  ) 

 2 2 
From electron density/magnetization vector to the Density matrix
n, m
n(r )   f   (r )  (r ) m(r )   f Ψ† (r )σΨ (r )   f  ' (r )σ  
'  (r )
 ,   
, , '
   
Ψ  
Density matrix  
  
     
  '
(r )   f  '


(r ) (r )

ρ(r )  
  


   

n,m → 
1 1  n  mz mx  im y 
ρ(r )   n(r ) I σ.m(r )    
2 2  mx  im y n  mz 
 → n,m
n  Tr (ρ) m  Tr (ρσ )
The non-collinear Stoner Hamiltonian
Local spin axis

I 1 0  I
  m    m z "
"
H
2  0 1
Stoner
2
Global spin axis
I  cos  e  i sin  
H Stoner  UH "
U   m  i
†

2  e sin   cos  
Stoner
I  0 1  0 i   1 0 
H Stoner   m sin  cos     sin  sin     cos   
2   1 0   i 0   0 1 
I
H Stoner   m.σ
2
A lot of fuss for a straightforward result!

Kohn Sham Hamiltonian
2
 '
H KS    2  Veff ' (r )
2m
 n(r ') 3 
Veff (r )   Vext (r )  
 '
d r '   '  Vxc ' ρ(r ) 
 r r'  Exc ρ ( r )
 '
 n(r ') 3 E ρ(r )  
Veff ' (r )   Vext (r )   d r ' xc   '  σ.B xc
 r  r ' n ( r )  Exc ρ ( r )
m ( r )
In LDA  xc (n)  f (n)
  xc (n  ) 0 
diagonalization of ρ  diagonalization of  xc (ρ)   
 0  
 xc (n ) 

 (m(r ),  (r ),  (r )) at each position r in space
 rotate "back" with matrix U to get  xc (ρ(r ))
In GGA  xc (n)  f (n, n)

Additional complication since ρ and ρ cannot be diagonalized in the same basis
off diagonal terms of  are neglected
2 JUILLET 2014 MASTANI, Pune, INDIA, July 2014| PAGE 28
PHYSICAL INSIGHT
Classical Heisenberg Hamiltonian

ej
1
E    J ij ei e j
2 i, j J ij
J ij Can be obtained by various approaches from DFT ei
(not yet available in QE)
J ij  0 Favors FM order J ij  0 Favors AF order
Frustration
Fairly happy

Some examples
Fe/Cr(001) interface Fe/Cr(110) interface

Cr/Cu(111)
Fe
Cr
Fe bibyramidal cluster
Ecol  Encol Ecol  Encol

SPIN-ORBIT COUPLING
Why should we care about SOC?
Small quantity (at least in 3d) with huge physical consequences

• At the origin of magneto-crystalline anisotropy …..and therefore of the stability of magnets!
bccFe hcpCo fccNi
E  ,  
• At the origin Anisotropic Magneto-Resistance
R  
B
current
Relativistic effects
 (r , t )
Dirac equation i   cα.p  βmc 2  V   (r , t )
t
  1 (r , t )  
 0 i     (r , t ) Large component
I 0  
αi    β     2 ( r , t ) 

 i 0   0 I 
( r , t )   (r , t )  
 3
  (r , t )  
  (r , t ) Small component
 4 
v
Scalar relativistic Small v/c limit  (r , t )  (r , t )
c
 2
 
H    2  V (r )   B (L  2S).B Schrödinger + Zeeman
 2m 
p4 Contraction and stabilization
 3 2 Mass-velocity
of s and p shells + expansion
8m c and destabilization of d and f
1 1 dV
2 2
L.S   (r )l.s  12  ( r )l.σ Spin-Orbit Coupling Splitting of orbitals with
2m c r dr angular momentum.
2
 2 2
 2V Darwin
8m c
The good basis
•Without SOC L et S commute with H
l, m  
2 2
Basis diagonalizing L , Lz ,S ,Sz
2 3
S   
2 2
L l, m  l(l  1) 2
l, m
4
Lz l, m  m l, m Sz    
•With SOC L et S no longer commute with H  H so

we consider J  LS
2
Basis diagonalizing J , Jz j, m j
2
J j, m j  j(j 1) 2
j, m J ls  j ls
J z j, m j  m j j, m j m j   j,  j  1, ,j
2 j1
2
1 2 
L.S j, m j   J  L  S  j, m j   j(j 1)  l(l  1)  s(s  1)  j, m j
2 2
2  2
•From one basis to the other = Clebsh Gordan

1 l  12 (2l+2) 
s j  1  2(2 l 1)
2 l  2 (2l) 
 2 
j  l  s  j, m j   j(j 1)  l(l  1)  12 ( 12  1)  j, m j
2 2
l.s j, m j 
2   2
l  1 (p orbitals)
p
(4) Ep3/2  2
p
(6)
Ep1/2   p
l  2 (d orbitals) (2)
Ed5/2   d
(6)
d (10)
3
Ed3/2    d
(4) 2 MASTANI, Pune, INDIA, July 2014
2 JUILLET 2014 | PAGE 34
Relativistic pseudo-potentials
•Without SOC But including other (scalar) relativistic effects
Vpseudo  Vloc (r )   ElI lI Yl ,Iml lI Yl ,Im

l
I l , ml
 VNL
Fe.pbe-nd-rrkjus.UPF
Pseudopotential type: ULTRASOFT

Method: Rappe Rabe Kaxiras Joannopoulos
Functional type: Perdew-Burke-Ernzerhof (PBE) exch-corr
Nonlinear core correction
scalar relativistic
•With SOC For technical details please contact Andrea dal Corso….
Vpseudo  Vloc (r )    VNLj   Vl Id  Vl SO L.S  Ylm Ylm

1 lm
j l 
2
Fe.rel-pbe-spn-rrkjus_psl.0.2.1.UPF
Pseudopotential type: ULTRASOFT

Method: Rappe Rabe Kaxiras Joannopoulos
Functional type: Perdew-Burke-Ernzerhof (PBE) exch-corr
Semi-core state in valence
Nonlinear core correction
2 JUILLET 2014 full relativistic MASTANI, Pune, INDIA, July 2014| PAGE 35
SOME IMPORTANT TECHNICAL POINTS
Relativistic Pseudopotential (LDA pz /GGA pw91, pbe)
 L.S  '

Is not diagonal in spins → impossible to split up and down spins
= drastic increase of the computational cost
  0 
collinear   Non-collinear    
 0      
Very dense K-points mesh
Check very carefully the influence of degauss
Use penalization when necessary (constrained_magnetization)

E  ,  
Very strict convergence threshold (conv_thr)

A FEW THINGS ABOUT SOC
 (r ) Is short range (atomic-like) in TB (or LCAO) on site term   I  L.S  I

I
 (r ) Z4 Increases drastically with atomic number


ll '   Rlat (r )Rlat' (r ) (r) r 2 dr d 0.05eV for 3d
0 d 0.5eV for 5d
• The band structure depends on the magnetization axis
• SOC is at the origin of the magnetocrystalline anisotropy

MCA 102  103 meV / atom in bulk Fe, Co, Ni
E  ,   Much larger in nanostructures
MCA meV / atom in bulk FePt L10
• SOC is at the origin of the anisotropic magneto_resistance
R  
2 JUILLET 2014 AMR 1% in bulk
larger in atomic wires
• SOC is at the origin of the orbital moment

MASTANI, Pune, INDIA, July 2014 | PAGE 37
MAGNETOCRSYTALLINE ANISOTROPY
How to calculate MCA?
•Brute force method (self-consistent) EMCA  Etot

n1
 Etot
n2
where En1 and En2 are obtained from SCF calculation including SOC
In principle « exact » but very time consuming and hard to converge

One should use penalization techniques to obtain En for any direction
•Force Theorem method

E1F EF2
EMCA  E n1
band E
n2
band  En ( E )dE  
1
En 2 ( E )dE
n1 n2
Eband and Eband are band energies of NSCF calculations including SOC
Initial density is obtained from a SCF spin-collinear calculation and spin-moment

further rotated to appropriate spin direction
Very stable numerically but cannot be applied to systems with too large SOC.
MCA: the local picture
•Force Theorem in grand canonical ensemble
E1F EF2 EF0

E   En1 ( E )dE   En 2 ( E )dE   ( E  EF0 )n( E )dE
EF0
Egc   ( E  EF0 )n( E )dE
•Local decomposition
EF0
Projection onto atomic orbitals E   ( E  EF0 )ni ( E )dE
i
gc
EF0
Real space picture E (r )   ( E  EF0 )n( E , r )dE
gc
The local decommposition in « canonical » picture leads to spurious oscillations

due to long range Friedel-like charge oscillations.. But the total MCA value is
kept due to charge conservation
MCA: the local picture
MCA   MCAi
i
MCA of a slab orginates from surface atoms of the outermost layers
fccCo Fe pyramid

PHYSICAL INSIGHT
Extended Heisenberg Hamiltonian

1
E    J ij ei e j   Fi (ei )   Dij .ei  e j
2 i, j i ij
MCA Fi (ei )  eit K i ei

F (ei )   K (n.ei ) 2 Uniaxial anisotropy
DM Dzyaloshinskii-Moriya interaction D only between first neighbours

(only exists in the absence of inversion symmetry)
DM favors non-collinear configurations

At the origin of skyrmion structures
DM=0 DM=0 DM≠0 MASTANI, Pune, INDIA, July 2014 | PAGE 41

SHAPE ANISOTROPY
•Classical dipole-dipole interaction m2 •Breit interaction

(2 body relativistic term)
0 
Edip (r ) 
4 r 3 
m1 .m 2 
3

r2
m1 .r m
2 .r



r
Colllinear magnetism m1
0
Edip ( )  m1m2 1  3cos  
4 r 3
For <54.74° FM For >54.74° AF
In thin films: in-plane magnetization is always favored
For ultra-thin films MCA is generally dominant

EMCA Surface while the shape anisotropy always ends up by
Eshape Volume dominating for thicker films

IMPORTANT THINGS I DID NOT TALK ABOUT
•Orbital moment
m orb (r )    L  m iorb (r )    L 
i
 occ  occ
For modern theory of orbital magnetization see David Vanderbilt
•DFT+U and orbital polarization
When using the rotationally invariant scheme of Liechtenstein (lda_plus_u_kind=1) the

Racah B parameter plays a crucial role in orbital magnetization
•Spin-spirals and generalized Bloch theorem
•Mapping DFT on model Hamiltonian Jii
•Spin dynamics etc….
•Spin-polarized transport
ACKNOWLEDGEMENTS
TEAM
Alexander SMOGUNOV
Sylvain LATIL
Yannick DAPPE
Dongzhe LI (PhD)
COLLEAGUES COLLEAGUES (Julich)
Chu Chu FU Stefan BLUGEL

Romain SOULAIROL (PhD) Gustav BIHLMAYER
Pascal THIBAUDEAU Timo SCHENA (Master)
David BEAUJOUAN (PhD)

BIBLIOGRAPHY
Electronic Structure
Basic Theory and Practical Methods
Richard Martin
Cambridge University Press
Theory of itinerant electron magnetism

Jürgen Kübler
Oxford Science Publications
Nouvelles méthodes pour le calcul ab-initio des propriétés

statiques et dynamiques des matériaux magnétiques.
Ralph Gebauer PhD thesis 1999
(in english)
THANK YOU FOR YOUR ATTENTION
QUESTIONS?
COMMENTS?
| PAGE 46
MASTANI, Pune, INDIA, July 2014
2 JUILLET 2014

Magnetism in DFT PDF

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Magnetism in DFT PDF

Загружено:

Авторское право:

Доступные форматы

MAGNETISM IN DFT

FROM THEORY TO PRACTICE

From the atomic nucleus to the inner core of Earth

GMR Read Head

2 JUILLET 2014 MASTANI, Pune, INDIA, July 2014 | PAGE 2

Intense activity in which nanomagnetism plays a crucial role

Search for permanent hard magnets without rare earth

Replace Samarium-Cobalt, Neodyum by nanostructured « cheap » TM magnets

Spintronics From Fundamentals to applications

Hohenberg & Kohn (1964) + Kohn Sham (1965)

n(r )n(r ') 3 3

(pseudo)-potental describing the interaction of valence electrons

Exc  n    n(r ) xc (n(r ))d 3r

 xc (n) : exchange correlation energy (per particle) of homogenous gas

 xc (n)   x (n)   c (n)

 x ( n)     Hartree Fock in an homogenous jellium

2 JUILLET 2014 MASTANI, Pune, INDIA, July 2014 | PAGE 5

Exc  n    n(r ) xc (n(r ),  n(r))d 3r

Exc  n    n(r ) xchom (n(r )) Fxc (n(r ),  n(r)) d 3r

Perdew and Wang (PW91)

2 JUILLET 2014 MASTANI, Pune, INDIA, July 2014 | PAGE 6

electron density and total energy

Converged? Analayse results

Average orbital moment: usually small (quenched) in bulk

 c (n,  )   c (n, 0)   c (n,1)   c (n, 0)  f c ( ) f c ( )  f x ( ) (Perdew Zunger)

Veff  Vext (r )  VHartree (r )  Vxc (n(r ), m(r ))   Bxc (r )

Bxc (r ) exchange correlation magnetic field

Spin dependent potential

Solve the two KS equation

Converged? Analayse results

What do we get out of spin-polarized DFT calculation

2 JUILLET 2014 MASTANI, Pune, INDIA, July 2014 | PAGE 12

 xc (n,  )   xc (n, 0)  12  xc" (n, 0) 2  o( 2 )

  M : magnetic moment per atom

I 1eV In most elements

Criterion can be also derived by analyzing

Nowadays Stoner model often used in parametrized Tight-Binding models

• Low coordination favors magnetism

Spin magnetic moment is generally enhanced on

2 JUILLET 2014 MASTANI, Pune, INDIA, July 2014 | PAGE 15

• Lattice expansion generally favors magnetism and vice versa…

Fe bcc Pd fcc Pd fcc

2 JUILLET 2014 MASTANI, Pune, INDIA, July 2014 | PAGE 16

• Several starting magnetization

• Fix spin moment (FSM) calculation

EF  EF N  N  N M ini  M  EF  EF N  N  N M  N  N

one scf loop many calculations

Minimization → modified KS Hamiltonian

H ij  H ij  2 (mi  m 0 ). ij

2 JUILLET 2014 MASTANI, Pune, INDIA, July 2014 | PAGE 19

Before running (or trusting) a DFT calculation you should be aware of

Pseudopotential (LDA pz /GGA pw91, pbe)

Energy cut-off (ecutwfc, ecutrho≥8ecutwfc for US pseudo)

Smearing (smearing, degausss) Marzari Vanderbilt recommended

Initial magnetization (starting_magnetization)

Number of computed eigenvalues (nbnd) often needs to be increased

Convergence threshold (conv_thr)

2 JUILLET 2014 MASTANI, Pune, INDIA, July 2014 | PAGE 20

As a rule of thumb: aLDA <aexp<aGGA

2 JUILLET 2014 MASTANI, Pune, INDIA, July 2014 | PAGE 21

2 JUILLET 2014 | PAGE 22

Why should we care about non collinearity?

Non-collinearity does exist!