Topics in Magnetism II. Models of Ferromagnetism Anne Reilly Department of Physics College of William and Mary After reviewing this lecture, you should be familiar with: 1. General source of ferromagnetism 2. Curie temperature 3. Models of ferromagnetism: Weiss, Heisenberg and Band Material from this lecture is taken from Physics of Magnetism by Chikazumi In ferromagnetic solids, atomic magnetic moments naturally align with each other. However, strength of ferromagnetic fields not explained solely by dipole interactions! m1 m2 U D 3 r N S N S Estimating m ~ 10-29 Wb m and r ~ 1 Ǻ, UD~10-23 J (small, ~1.3K) (see Chikazumi, Chp. 1) In 1907, Weiss developed a theory of effective fields Magnetic moments (spins*) in ferromagnetic material aligned in an internal (Weiss) field: Hw HW = wM w=Weiss or molecular field coefficient Average total magnetization is: H (applied) M ( H wM ) exp 0 kT cos sin d M NM M ( H wM ) exp 0 kT sin d M = atomic magnetic dipole moment *Orbital angular momentum gives negligible contribution to magnetization in solids (quenching) Weiss Theory of Ferromagnetism M ( H wM ) exp 0 kT cos sin d M ( H wM ) M NM NML NML kT M ( H wM ) exp sin d 0 kT Langevin function Consider graphical solution: Tc is Curie temperature M/Ms 1 At Tc, spontaneous magnetization disappears and material become paramagnetic 0 T/Tc (see Chikazumi, Chp. 6) 1 Tc 2 NM eff w 3k Weiss Theory of Ferromagnetism Tc 2 NM eff w 3k For Iron (Fe), Tc=1063 K (experiment), M=2.2mB (experiment), And N=8.54 x 1028m-3 Find w=3.9 x 108 And Hw=0.85 x 109 A/m (107 Oe) Other materials: Cobalt (Co), Tc=1404 K Nickel (Ni), Tc= 631K Weiss theory is a good phenomenological theory of magnetism, But does not explain source of large Weiss field. Heisenberg and Dirac showed later that ferromagnetism is a quantum mechanical effect that fundamentally arises from Coulomb (electric) interaction. Key: The Exchange Interaction •Central for understanding magnetic interactions in solids •Arises from Coulomb electrostatic interaction and the Pauli exclusion principle Coulomb repulsion energy high UC e2 40 r 2 Coulomb repulsion energy lowered ~ 10 18 J (105 K !) The Exchange Interaction Consider two electrons in an atom: Hamiltonian: H H 1 H 2 H 12 2 Ze 2 H1 1 2me 40 r1 1 r12 er1 e- 2 r2 + Ze 2 Ze 2 H2 2 2me 40 r2 H 12 e 2 40 r12 2 2 2 j 2 2 2 x j y j z j Using one electron approximation: 1 1 (r1 )2 (r2 ) 2 (r1 )1 (r2 ) s (r1 , r2 ) 2 1 1 (r1 )2 (r2 ) 2 (r1 )1 (r2 ) A (r1 , r2 ) 2 1,2 singlet triplet are normalized spatial one-electron wavefunctions We can write energy as: H E 1 E 1* (r1 )2* (r2 ) 2* (r1 )1* (r2 ) (H 1 H 2 H 3 )1 (r1 )2 (r2 ) 2 (r1 )1 (r2 )d 3r1d 3r2 2 1* (r1 )H 11 (r1 )d 3r1 2* (r1 )H 12 (r1 )d 3r1 1* (r2 )H 21 (r2 )d 3r2 2* (r2 )H 22 (r2 )d 3r2 Individual energies (ionization) = 2I1 + 2I2 1* (r1 )2* (r2 )H 121 (r1 )2 (r2 )d 3r1d 3r2 Coulomb repulsion = 2K12 2* (r1 )1* (r2 )H 122 (r1 )1 (r2 )d 3r1d 3r2 1* (r1 )2* (r2 )H 122 (r1 )1 (r2 )d 3r1d 3r2 2* (r1 )1* (r2 )H 121 (r1 )2 (r2 )d 3r1d 3r2 Exchange terms =2 J12 We can write energy as: E I1 I 2 K12 J12 Lowest energy state is for triplet, with E I1 I 2 K12 J12 Parallel alignment of spins lowers energy by: e2 1 3 3 J12 ( r ) ( r ) ( r ) ( r ) d r d r2 2 1 1 2 1 2 1 40 r1 r2 * 1 * 2 (if J12 is positive) You can add spin wavefunctions explicitly into previous definitions: (singlet) 1 1 (r1 )2 (r2 ) 2 (r1 )1 (r2 ) (1) (2) (2) (1) s (r1 , r2 ) 2 (1) (2) 1 1 (r1 )2 (r2 ) 2 (r1 )1 (r2 ) (1) (2) (2) (1) A (r1 , r2 ) 2 (1) (2) (triplet) 1 Spin +1/2 0 0 1 Spin -1/2 You can add spin wavefunctions explicitly into previous defintions. (singlet) 1 1 (r1 )2 (r2 ) 2 (r1 )1 (r2 ) (1) (2) (2) (1) s (r1 , r2 ) 2 (1) (2) 1 1 (r1 )2 (r2 ) 2 (r1 )1 (r2 ) (1) (2) (2) (1) A (r1 , r2 ) 2 (1) (2) 1 Spin +1/2 (triplet) 0 0 1 Spin -1/2 Heisenberg and Dirac showed that the 4 spin states above are eigenstates of operator S1 S 2 Heisenberg Model Heisenberg and Dirac showed that the 4 spin states above are Eigenstates of operator S1 S 2 S, 2 , (Pauli spin matrices) Hamiltonian of interaction can be written as (called exchange energy or Hamiltonian): H ex 2 JSi S j J is the exchange parameter (integral) Assume a lattice of spins that can take on values +1/2 and -1/2 (Ising model) The energy considering only nearest-neighbor interactions: n n j 1 j 1 U 2 JSi S j 2m B H m S j average molecular field due to rest of spins Find, for a 3D bcc lattice: kTc 2.446 J For more on Ising model, see http://www.physics.cornell.edu/sss/ising/ising.html http://bartok.ucsc.edu/peter/java/ising/keep/ising.html Band (Stoner) Model Heisenberg model does not completely explain ferromagnetism in metals. A band model is needed. Assumes: I S n E (k ) E (k ) N I S n E (k ) E (k ) N Is is Stoner parameter and describes energy reduction due to electron spin correlation n , n is density of up, down spins Band (Stoner) Model n n (spin excess) Define R N ~ E ( k ) E ( k ) I R / 2 S Then ~ E (k ) E (k ) I S R / 2 note: M m B I s (n n ) ~ E (k ) E (k ) 2N Spin excess given by Fermi statistics: 1 R N f , f (k ) f (k ) k ~ exp E (k ) I s R / 2 E F / kT N R V 1 Band (Stoner) Model Let R be small, use Taylor expansion: 2 x g ( x x / 2) g ( x x / 2) g ' ( x)x g ' ' ' ( x) ... 3! 2 with x I s R 1 R N f (k ) 1 k E~(k ) ( I s R) 24 N 3 f (k ) 3 ( I R ) k E~(k )3 s ... 3 V V ~ f f k E~ (2 )3 N dk E~ 2 3 N dk ( ( E EF )) (at T=0) V f(E) D( EF ) 2 D.O.S.: density of states at Fermi level EF E Band (Stoner) Model V ~ Let DE F D( EF ) 2N Then Density of states per atom per spin ~ R DEF I s R O(3) Third order terms ~ ( 3) R(1 DEF I s ) O When is R> 0? ~ 1 DEF I s 0 or ~ DEF I s 1 Stoner Condition for Ferromagnetism For Fe, Co, Ni this condition is true Doesn’t work for rare earths, though Heisenberg versus Band (itinerant or free electron) model Both are extremes, but are needed in metals such as Fe,Ni,Co Band theory correctly describes magnetization because it assumes magnetic moment arises from mobile d-band electrons. Band theory, however, does not account for temperature dependence of magnetization: Heisenberg model is needed (collective spin-spin interactions, e.g., spin waves) To describe electron spin correlations and electron transport properties (predicted by band theory) with a unified theory is still an unsolved problem in solid state physics.