 Spin interaction

 Exchange interaction
Coupling of magnetic moments results from the interplay of the Pauli principle with Coulomb repulsion and
electron hopping.
Classically, magnetic moments originate from electric currents: A current density generates a magnetic
moment

These moments interact via the dipole-dipole interaction. The magnetostatic interaction energy between two dipoles
at a distance being the unit-vector from the position of the first to that of the second dipole,

depends on their distance and relative orientation. This can, however, not be the origin of the magnetism found in
actual materials: In a classical system charges cannot flow in thermodynamic equilibrium, and hence there are no
magnetic moments to begin with.
In quantum mechanics, however, non-vanishing charge currents in the ground state are not uncommon: An electron
in state corresponds to a current density

According to (1) it produces a magnetic moment proportional to the expectation value of the angular momentum

Tuhin Maity 76
  Spin interaction
 Exchange interaction
Coulomb exchange: The Coulomb repulsion between electrons,

is manifestly spin-independent. Nevertheless, because of the antisymmetry of the many-electron wave function, the
eigenenergies of HU depend on the spin. This is the basis of the multiplet structure in atoms and of Hund’s first two
rules.
To understand the mechanism of this Coulomb exchange we consider a simple two-electron model.

Let us consider two orbitals α= a; b. Then the two-electron Slater determinants with spins σ and σ’. Wave function of
the 2 electrons:

orbital part Spin part

Tuhin Maity 77
 Spin interaction
 Exchange interaction
Coulomb exchange:
When both electrons have the same spin (σ= σ’), we can factor out the spin functions

and obtain

where the direct terms are the Coulomb integral

while the cross terms give the exchange integral

Tuhin Maity 78
 Spin interaction
 Exchange interaction
Coulomb exchange:
For the states where the electrons have opposite spin (σ= - σ’),
diagonal matrix-elements

off-diagonal matrix-elements

Since HU does not change the spins, these are the only non-zero matrix elements. In the basis
of the states the Coulomb term is thus given by

Tuhin Maity 79
 Spin interaction
 Exchange interaction
Coulomb exchange:
eigenstates

total spin angular momentum 1

total spin angular moment 0:

Tuhin Maity 80
 Spin interaction
 Exchange interaction
Coulomb exchange:
Q. Zhang:
Calculations of Atomic Multiplets
across the Periodic Table
MSc thesis, RWTH Aachen 2014
www.cond-mat.de/sims/multiplet

Tuhin Maity 81
 Spin interaction
 Exchange interaction
Kinetic exchange:
When electron-hopping plays the main role in the exchange mechanism, we speak of kinetic exchange. In contrast
to Coulomb exchange the resulting interactions are usually antiferromagnetic, i.e., they prefer antiparallel spins. The
physical principle of kinetic exchange can be understood in a simple two-site system. We discuss this problem in
some detail and introduce two key concept along the way: down-folding and second quantization.

1. A toy model
As a toy model, we consider the minimal model of an H2 molecule. We restrict ourselves to two (orthogonal)
orbitals, and , separated by some distance. If we add an electron to the system, that electron will be able to
move between the two orbitals, with a matrix element -t. Because we allow the electron to only occupy two s-
orbitals, the Hamiltonian is a 2 x 2 matrix with the hopping matrix element as -t,

If there are two electrons in the system, i.e., one electron per orbital, we can again use basis states which just
specify, which orbitals the electrons occupy. For two electrons of opposite spin we then find two states where the
electrons are in different orbitals
“covalent states”
and two states where the electrons are in the same orbital
“ionic states”.

Tuhin Maity 82
 Spin interaction
 Exchange interaction
Kinetic exchange:
1. A toy model
In this basis the Hamiltonian matrix for our simple model of the H2 molecule has the form

As before, moving an electron to a neighboring orbital gives a matrix element -t, with an additional sign when the
order of the electrons is changed (Fermi statistics!). For the ionic states, where both electrons are in the same
orbital, we have the Coulomb matrix element U. Coulomb matrix elements involving electrons on different sites are,
for reasonably large distance between the sites, negligible. So there is no Coulomb exchange, just the local
Coulomb repulsion in our model. Diagonalizing H we find the energy spectrum and the corresponding eigenstates:

Tuhin Maity 83
 Spin interaction
 Exchange interaction
Kinetic exchange:
2. Direct exchange
The antiparallel alignment of the spins (left) is favored, since it allows the electrons to hop to the neighboring site.
For parallel spins (right) hopping is suppressed by the Pauli principle. These orbitals are typically localized d-orbitals
of transition-metals.

Tuhin Maity 84
 Spin interaction
 Exchange interaction
Kinetic exchange:
3. Super exchange
Direct exchange cannot explain the antiferromagnetism of most transition-metal compounds: Since the d-orbitals
are so localized, hopping can only occur between orbitals on different atoms that are very close to each other. But
most antiferromagnetic insulators are transition-metal oxides, so that the transition-metal cations are separated by
large oxygen anions. In such a situation, shown in figure 5, direct hopping between the d-orbitals is very unlikely.
The concept of direct exchange can, however, be extended to these cases by taking into account hopping via
the intermediate p-orbital. This mechanism is called superexchange.

&

pZ
MnO: Mn2+ are separated by O2-

Tuhin Maity 85
  Spin interaction
 Exchange interaction
Kinetic exchange:
3. Super exchange pZ
Cation A Cation B

If A and B antiparallel, pz↑ hybridize with A

pz↓ hybridize with B

If A and B parallel: pz↑ hybridize with A and B, but no

hybridization for pz↓ Tuhin Maity 86
 Spin interaction
 Exchange interaction
Kinetic exchange: When the d-orbitals interact via
3. Super exchange
an oxygen in-between (the
180° geometry shown on the
left), both d-orbitals couple to
the same p-orbital, while the
hopping to the two other p-
orbitals vanishes by symmetry.
The result is antiferromagnetic
superexchange. When the
angle of the M-O-M group is
90° (right), the d-orbitals couple
to orthogonal p-orbitals, making
it impossible for an electron on
one d-orbital to reach the d-
orbital on the other site. In this
case, superexchange is
mediated via the Coulomb
exchange on the connecting
oxygen.
Tuhin Maity 87
 Spin interaction
 Exchange interaction
Kinetic exchange:
4. Double exchange
Double exchange takes its name from the fact that it results from a combination of Coulomband kinetic-exchange.
In that sense the 90° superexchange mechanism discussed above is a double exchange mechanism. More
commonly, double exchange is encountered in mixed valence compounds. So far we have considered systems with
an integer number of electrons per site. When correlations are strong the lowest energy state will essentially have
the same number of electrons on every site and hopping will be strongly suppressed by the Coulomb repulsion
energy U as we have seen for the simple two-site model of kinetic exchange. In a mixed valence system the
number of electrons per site is non-integer, so even for large U some site will have more electrons than others. Thus
electrons can hop between such sites without incurring a cost U. Hence these compounds are usually metallic.

Metallic systems are often mixed

valence: example of manganites:
La1-xCaxMnO3: coexistence of
Mn4+ (3 electrons, S=3/2) and Mn3+
(4 electrons, S=2 , localized spin
3/2 + 1 conduction electron in eg
band)

Tuhin Maity 88
 Spin interaction
 Spin relaxation
Spin relaxation refers to processes that bring an unbalanced population of spin states into equilibrium. If, say, spin
up electrons are injected into a metal at t = 0 creating spin imbalance at a later time t = T1, the balance is restored
by a coupling between spin and orbital degrees of freedom. Spin relaxation is of paramount importance in
spintronics, since there is a serious concern for using spin polarization of either a single charge carrier or the net
spin polarization of an ensemble of charge carriers to encode and decode information. If the given spins were to
host the information reliably for considerable time, it must be protected against random or spontaneous
depolarization caused by various relaxation mechanisms. Therefore, one needs to understand all various possible
relaxation mechanisms that cause spin relaxation in a solid.

Hence, the next slides are intended to provide

• Basics on spin relaxation process, and

• Four major spin relaxation mechanisms in the solids.

Tuhin Maity 89
 Spin interaction
 Spin relaxation
When an electron is introduced in a solid, the interaction between the electron and its environment affects its spin
orientation, i.e., the environment may give rise to an effective magnetic field which interacts with the spin and
changes its orientation. The effective magnetic field in a solid arises from the following sources:

• the spin of other electrons and holes in the solid,

• nuclear spin,

• phonons (lattice vibrations) that gives time dependent magnetic field (in some circumstances), and

• spin-orbit interactions in the solid.

Any effective magnetic field interacting with the magnetic moment of the electron's spin can alter the electron's spin
state. In order to understand how this occurs, we shall first consider the time-independent Pauli equation describing
the electron:

Tuhin Maity 90
 Spin interaction
 Spin relaxation
The stable spin polarizations in the presence of the effective magnetic field are those that are parallel or anti-parallel
to the magnetic field. However, most of the time, the electron's initial spin will not be parallel or anti-parallel to the
new effective magnetic field and therefore, the initial spin will not exactly be stable and begin to change with time as
soon as the electron encounters the new effective magnetic field. This will result in spin precession about the new
effective magnetic field Beff with an angular frequency where g is Lande g-factor of the medium where
the electron resides. This phenomenon is known as Larmor precession.

It is well known that the magnetic flux density Beff caused by the spin-orbit interaction depends on the electron's
velocity or wavevector k. Therefore, if an electron's velocity or wavevector changes randomly owing to scattering,
then the axis about which the spin precesses changes direction because of the change in the effective magnetic
field direction and the frequency of Larmor precession. These changes occur randomly in time because scattering
events are random. Hence, the spin polarization of an electron changes randomly in time, which leads to the
relaxation of spins.

There are four major spin flip mechanisms in a solid:

(i) Elliott-Yafet mechanism,
(ii) D'yakonov Perel mechanism,
(iii) Bir-Aronov-Pikus mechanism, and
(iv) Hyperfine interactions with nuclear spins.

Tuhin Maity 91
 Spin interaction
 Spin relaxation
(i) Elliott-Yafet Mechanism;

The Elliott-Yafet mechanism is based on the fact that in real crystals Bloch states (the wavefunction of a particle,
usually, an electron, placed in a periodic potential) are not spin Eigen states. As a result, an electron's spin in the
crystal does not really have one of two fixed polarizations: up or down. However, the actual spin polarization
depends slightly on what the electron's wave vector is, which is schematically depicted in Figure.

Figure: (a) Energy dispersion relation showing the spin polarizations at different wavevector states. Each
wavevector has two possible mutually anti-parallel spin polarizations. A momentum changing collision event, that
changes the wavevector state from k1 to k2, also changes the spin orientation since the Eigenspin polarizations at
k1 and k2 are generally not parallel, (b) The periodic spin-orbit interaction makes the spin up Bloch states contain
small spin-down amplitude, and vice-versa.

Tuhin Maity 92
 Spin interaction
 Spin relaxation
(i) Elliott-Yafet Mechanism;

Each wavevector state still has two possible spin orientations that are mutually anti-parallel but spin orientations
associated with different wavevector states can have arbitrary angle between them. If a collision event with a non-
magnetic scatterer changes the momentum or wavevector of an electron, then it will change the spin as well since
the spin orientations associated with the initial and final wavevector states are never mutually parallel. The spin
state of a nearly up-spin electron at a wavevector state k1 and that of a nearly down-spin electron at a different
wavevector state k2 are not strictly orthogonal, since they are not strictly anti-parallel. Thus, any collision with a non-
magnetic scatterer that changes the wavevector of an electron from an initial state k1 to a state k2 can also couple
the nearly down spin state at k1 to the nearly up spin state at k2. This coupling can flip the electrons spin from nearly
up to nearly down. Such spin relaxation is always accompanied by some degree of momentum relaxation since the
wavevector must change in order to change the spin.

Tuhin Maity 93
 Spin interaction
 Spin relaxation
(ii) D'Yakonov Perel mechanism:

If any given solids lack inversion symmetry because of its crystalline structure or because of an externally applied
electric field, then an electron in the solid will experience strong spin-orbit interaction. Now, let us consider an
ensemble of electrons drifting and diffusing in a solid. If every electron's velocity is the same and does not change
with time, then the field is the same for all electrons and every electron precesses about this constant field with a
fixed frequency. This does not cause any spin relaxation at all, i.e., if all electrons started with the same spin
polarization, then after any arbitrary time they all have precessed by exactly the same angle and therefore they all
have again the same spin polarization. The direction of this spin polarization may be different from the initial one,
but that does not matter as the magnitude of the ensemble averaged spin does not change with time. Hence, there
is no spin relaxation.

On the other hand, if changes randomly with time

because of scattering, then different electrons would have
precessed by different angle after a given time because
they have different scattering histories. Thus, even if all
the electrons were injected with the same spin
polarization, their spin polarizations gradually go out of
phase with each other after a given time and the
magnitude of the ensemble averaged spin will decay to
zero (see Figure).
Tuhin Maity 94
 Spin interaction
 Spin relaxation
(iii) Bir-Aronov-Pikus mechanism:

The Bir-Aronov-Pikus (BAP) mechanism is a source of spin relaxation of electrons in semiconductors where there is
a significant concentration of both electrons and holes (see Figure). In such a case, an electron will be usually in
close proximity to a hole, so that their wavefunctions will overlap. This would cause an exchange interaction
between them. If the hole's spin flips, then the exchange interaction will cause the electron's spin to flip as well,
leading to spin relaxation.

Figure: Schematic drawing of BAP mechanism. The exchange interaction between electrons (filled circles) and
holes (open circles) causes the electron spins to precess along some effective magnetic field determined by hole
spins. In the limit of strong hole spin relaxation, this effective field randomly changes before the full precession is
completed, reducing the electron spin relaxation.

Tuhin Maity 95
 Spin interaction
 Spin relaxation
(iv) Hyperfine interactions with nuclear spins:

In solids, the nuclear spins generate an effective magnetic field which interacts with the electron spins via hyperfine
interactions, resulting in spin relaxation. The Hamiltonian describing this interaction is given by

where |n> is the coupled electron and nuclear site, Sz(0) and Sz(t) are the initial spin eigenstate at time t=0 and the
Eigenstate at some later time t, respectively, and Sz(t) is given by . From the behavior of Cn(t),
one can deduce the temporal behavior of Sz(t). The decay is much faster for the fully polarized case, but the steady
state value <Sz(∞)> is much closer to Sz(0) in the fully polarized case that in the unpolarized case. In any case, the
hyperfine interaction does not cause complete loss of electron spin polarization.
Tuhin Maity 96
 Spin interaction
 Spin Transport
The magnetism of materials is carried by electron spin, while electrical transport is caused by the motion of electron
charge. Although these two fundamental properties of solids have been well known for many centuries, the fields of
magnetism and electrical transport have been developed almost independently. As the fabrication techniques in
micro and nanoscale samples have shown tremendous progress, the field of spintronics has been developed,
where the coupling of electron spin and charge plays an important role. Since the fundamental properties of
spintronics are closely related to the length scale L characteristic of samples and to the motion of the electrons in
metal, it is very much important to understand the electron transport properties, i.e., the control of spins by an
electric field and the control of electrical current by a magnetic field in metals. Hence, the primary motivation of the
discussion is to

1. Take you through a simple introduction to transport theory covering Boltzmann equation,

2. Understand the transport properties in thin films and the Fuchs-Sondheimer model for thin films

3. Discuss the effect of magnetic field on the electron transport properties and to understand the normal
magnetoresistance,

4. Study quantum interference effects in metals with strong electron scattering.

Tuhin Maity 97
 Spin interaction
 Spin Transport
(1) Electron transport theory covering Boltzmann equation

The Boltzmann equation is a semiclassical approach to the calculation of electrical conductivity which assumes that
the electric field (Ɛ) is sufficiently small and the response to the current, j, is linear, so that j=σ.Ɛ, where σ is the
electrical conductivity. In addition, it assumes that the momentum is well defined and hence k is a good quantum
number. This assumption is true if the wavelength of the electron is small compared to its mean free path. In such
cases, the electron appears as a plane wave before each scattering event, as shown in Figure. In these
circumstances, the condition, kFλ>>1 is satisfied, where kF is the Fermi wavevector and λ is the mean free path of
the electrons. The electrons can be described by their distribution function f(k). At equilibrium this is simply a Fermi-
Dirac distribution is given by

In the presence of an electric field the distribution deviates from the equilibrium such that . In the
steady state condition, the current (and hence the conductivity) can be calculated from a simple average of the
electron velocity using this distribution function.

Tuhin Maity 98
 Spin interaction
 Spin Transport
(1) Electron transport theory covering Boltzmann equation

Substituting the distribution function in eqn.(2) results

Where . Note that the second term in eqn.(3) is zero, since this distribution function describes the
equilibrium state where there is no net flow of electrons.

Therefore, the problem of calculating the conductivity is now reduced to calculating the distribution function g. In its
simplest form, the Boltzmann equation is a simple expression of the steady state condition that the distribution
function is not changing with time, i.e.,

The first term in eqn.(4) describes the fact that the function f is being driven away from equilibrium due to the
acceleration of electrons by the electric field. On the other hand, the second term describes the relaxation back into
equilibrium due to various scattering processes. Now let us discuss how this simple expression can be used to
calculate the conductivity in a simple free electron model, in more complex models and in thin films.

Tuhin Maity 99
  Spin interaction
 Spin Transport
Diffusive scattering:

If we assume diffusive scattering at surfaces as shown in Figure, then we have to postulate that all electrons starts
with their equilibrium distribution after the collision at the surface. It means that they lose their field drift completely.

Figure: Schematic representation of diffusive

scattering of electrons at the film surfaces.

On the other hand, elastic collision of the electrons at the surface will not alter the bulk conductivity since the drift is
the same before and after scattering (see Figure).

Tuhin Maity 100

 Spin interaction
 Magnetoresistance
The Hall effect

Tuhin Maity 101

 Spin interaction
 Magnetoresistance
Basic transport measurement

Tuhin Maity 102

 Spin interaction
 Magnetoresistance
•Family of the Hall effect
–Ordinary Hall effect (metals, 1879)
–Anomalous Hall effect (magnetic metals, 1880)
–(Extraordinary Hall effect)

•Modern family of the Hall effect

–Quantum Hall effect (2DEG, 1985 Nobel prize in physics)
–Fractional quantum Hall effect (2DEG, 1998 Nobel prize in physics)
–Quantum anomalous Hall effect (topological insulators, 2010s)
–Spin Hall effect (metals, 2000s)
–Quantum spin Hall effect (topological insulators, 2010s)

Tuhin Maity 103

 Magnetoresistance
 Magnetoresistance
The resistance of some of the metal and the semiconductor material varies in the presence of the magnetic field,
this effect is called the magnetoresistance. The element which has these effects is known as the magnetoresistor. In
other words, the magnetoresistor is a type of resistor whose resistance varies with the magnetic field.

The resistance of the magneto resistor is directly proportional to the magnetic field, i.e., their resistance raises with
the increase of the magnetic field. The variation in resistance occurs because of the magneto effect.

The magnetoresistor operates without physical contacts which is their major advantage. The magnetoresistor has
various applications like it is used in the hard disk of the computer, an electronic compass, for measuring the current
etc.

William Thomson (Lord Kelvin) first discovered ordinary magnetoresistance in 1856. He experimented with pieces of
iron and discovered that the resistance increases when the current is in the same direction as the magnetic force
and decreases when the current is at 90° to the magnetic force. He then did the same experiment with nickel and
found that it was affected in the same way but the magnitude of the effect was greater. This effect is referred to as
anisotropic magnetoresistance (AMR).

In 2007, Albert Fert and Peter Grünberg were jointly awarded the Nobel Prize for the discovery of giant
magnetoresistance.
Tuhin Maity 104
 Spin interaction
 Magnetoresistance

Working Principle of Magnetoresistor

It works on the principle of electrodynamics, which states that the force acting on the current place in the magnetic
field changes their direction. In the unavailability of the magnetic field, the charge carriers of the magneto resistor
move in the straight path. In the presence of the magnetic field, the direction of the current becomes changes, and it
flows in the opposite direction. The indirect path of the current increases the mobility of their charge carrier which
causes the collision.

The collision increases the loss of energy in the form of heat. This heat increases the resistance of the magneto
resistor. The current of very small magnitude flows in the magnetoresistor because of few free electrons. The
deflection of the magnetoresistor electrons depends on their mobility. It is more in the semiconductor material as
compared to the metals. The mobility of the indium arsenides or indium antimonides is approximately 2.4m2/Vs.

The Effect of Magnetic Fields on Resistance

Resistance is caused by collisions between charge carriers (like electrons) and other carriers or atoms. An electron
moving through a perfect crystal of metal at a temperature of absolute zero will experience no collisions, so the
crystal would have zero resistance. Imperfections, however, do exist, and temperatures above absolute zero cause
the atoms to vibrate out of their lattice locations. These vibrations and imperfections cause collisions, increasing the
resistance of the crystal.

Tuhin Maity 105

 Spin interaction
 Magnetoresistance
Applying a magnetic field can also increase the resistance of a material, since the magnetic force on the moving
charges will tend to increase the number of collisions between charges. This dependence of resistance on magnetic
field is called magnetoresistance. Magnetoresistance is proportional to the strength of the magnetic field, with a
larger field producing a higher resistance. This property is used in computers to read magnetic data. A potential
difference is applied to a wire that is placed close to the magnetic material on disk or tape. As the magnetic fields
representing data on the material pass by the wire, the resistance of the wire changes with the magnetic field of the
data. This change in resistance changes the current through the wire. Monitoring this current provides a reading of
the magnetic field on the tape or disk.

Charges moving through a magnetic field experience a magnetic force given by

|FB| = q|v| x B,

where the bold face on F, v, and B indicates that these quantities are vectors: they have both direction and
magnitude. The x is a cross-product, which means that the magnitude (size) of the force is found from

| FB | = q |v| |B| sinθ,

and the direction of the force is found from a right-hand rule: put your thumb in the direction of (qv), your second
finger in the direction of B, and your middle finger (or palm) will point in the direction of the force. θ represents the
angle between B and v. Be careful when applying this rule to a negatively-charged particle like the electron: qv
points in the opposite direction of v, since q is negative.
Tuhin Maity 106
 Spin interaction
 Magnetoresistance
You should understand the following consequences of this force:
1. A charge at rest experiences no magnetic force, since v is zero.
2. A charge moving parallel to the magnetic field experiences no magnetic force: since the angle q between B and v
is zero, sinθ is zero.
3. A charge moving un-parallel to the magnetic field experiences a magnetic force perpendicular to both the
charge's motion and the magnetic field.
4. A negatively charged particle experiences a magnetic force opposite the direction of the magnetic force on a
positive charge.
Let us apply this magnetic force to a charge moving through a wire (Figure). The teal conduction electrons are
moving to the left (red arrow) through a wire. Since the charge of the moving charges is negative, qv points
opposite the direction of v, or to the right (brown arrow).
A uniform magnetic field is present throughout the
wire, pointing into the page (lavender x's). Using the
right-hand rule, with the thumb to the right
representing qv, the second finger into the page
representing B, we find that the palm representing F
points toward the top of the page. So the conduction
electrons feel a force toward the top of the wire. This
magnetic force disrupts the normal flow of electrons,
causing more collisions with atoms and other
electrons, which increases the resistance of the wire.
Tuhin Maity 107
 Magnetoresistance
 Magnetoresistance
• Family of MR
– Ordinary magnetoresistance (OMR)
– Anisotropic magnetoresistance (AMR)

• Modern family of MR
– Giant magnetoresistance (GMR)
– Tunneling magnetoresistance (TMR)
– Colossal magnetoresistance (CMR)
– Spin Hall magnetoresistance (SMR)
– Unidirectional spin Hall magnetoresistance (USMR)
– Magnetoconductance (MC)
– Ballistic magnetoresistance (BMR)

Tuhin Maity 108

 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
Novel magnetotransport phenomena such as giant magnetoresistance (GMR) (in magnetic multilayers) and tunnel
magnetoresistance (TMR) (in ferromagnetic tunnel junctions) appear when the size scale of the magnetic materials
becomes nanolevel. In particular, the subject of spintronics received wide attention with the discovery that the
electric current in a multilayer film, consisting of a sequence of thin magnetic layers separated by a thin non-
magnetic metallic layer, is strongly influenced by the relative orientation of the magnetization of the magnetic layers.
This suggests that the internal moment of the electrons associated with their spin plays an important role in the
transport of electric charge. Therefore, it is required to understand the spin-dependent scattering process in
magnetic multilayer structure and now we will

1. Introduce the resistance change in multilayer structure,

2. Discuss the physical origin of GMR,
3. Demonstrate spin-dependent scattering of electrons in multilayer films, and
4. To derive equation for determining the GMR using resistor network theory.

Tuhin Maity 109

 Magnetoresistance
 Giant Magnetoresistance
•(Perhaps) The beginning of modern magnetism or spintronics

Tuhin Maity 110

 Magnetoresistance
 Giant Magnetoresistance
•(Perhaps) The beginning of modern magnetism or spintronics

Tuhin Maity 111

 Magnetoresistance
 Giant Magnetoresistance
•(Perhaps) The beginning of modern magnetism or spintronics

Tuhin Maity 112

 Magnetoresistance
 Giant Magnetoresistance
•(Perhaps) The beginning of modern magnetism or spintronics

Tuhin Maity 113

 Magnetoresistance
 Giant Magnetoresistance
•(Perhaps) The beginning of modern magnetism or spintronics

Tuhin Maity 114

 Magnetoresistance
 Giant Magnetoresistance
•(Perhaps) The beginning of modern magnetism or spintronics

Tuhin Maity 115

 Magnetoresistance
 Giant Magnetoresistance
•(Perhaps) The beginning of modern magnetism or spintronics

Tuhin Maity 116

 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
(1) Resistance change in multilayer structure:
Figure shows the typical multilayer structure consisting of a sequence of thin ferromagnetic (FM) layers separated
by equally thin non-magnetic (NM) metallic layer. The resistance of the magnetic multilayer is low when the
magnetizations of all the FM layers are parallel (see Figure a) and the resistance becomes larger when the
magnetizations of the neighbouring FM layers are ordered antiparallel (Figure b).

Figure: (a) Ferromagnetic and (b) antiferromagnetic configurations of magnetic multilayers film.

In the above scenario, the optimistic magnetoresistance (MR) ratio is defined by

where R↑↓ and R↑↑ are the resistance of the multilayer film in antiferromagnetic (AFM) and FM configurations,
respectively. The most commonly used combinations of magnetic and non-magnetic layers are Cobalt - Copper and
Iron - Chromium. But multilayers based on permalloy (Ni80Fe20) as the magnetic component are also used
frequently.

Tuhin Maity 117

 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
(2) Physical Origin of GMR:
There are two principal geometries of the GMR effect, which are schematically shown in Figure. In the first case
(Figure a), the current flows perpendicular to the multilayers (CPP geometry), whereas in the Figure b the current
flows in the plane of the layers (CIP). The underlying physical mechanism is the same for both the CPP and CIP
geometries. For a simple demonstration, let us consider a trilayer magnetic film with two magnetic layers separated
by a non-magnetic metallic spacer layer. The electron spin is conserved over distances of up to several tens of
nanometers, which is much larger than the thickness of a typical multilayer. Therefore, one can assume that the
electric current flows in two channels: one corresponding to electrons with spin projection ↑ and the other to
electrons with spin projection ↓ .

Figure: (a) Current perpendicular to plane (CPP) and (b) current in plane (CIP) GMR geometries.

Tuhin Maity 118

 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
(2) Physical Origin of GMR:

Figure: Schematic representations of simple resistor model for describing GMR.

Tuhin Maity 119

 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
(2) Physical Origin of GMR:

Since the ↑ and ↓ spin channels are independent (spin is conserved) they can be regarded as two wires connected
in parallel. In addition, the electrons with spin projections parallel and antiparallel to the magnetization of the FM
layer are scattered at different rates when they enter into FM. This is called spin-dependent scattering. Let us
assume that electrons with spin antiparallel to the magnetization of the FM layer are scattered more strongly. The
GMR effect in a trilayer can be now explained qualitatively using a simple resistor model as shown in Figure.

Figure: Schematic representations of simple resistor model for describing GMR.

Tuhin Maity 120
 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
(2) Physical Origin of GMR:

In the FM configuration, the electrons with ↑ spin are weakly scattered both in the first and second FM layers
whereas the electrons with ↓ spin are strongly scattered in both FM layers. This can be simulated by two small
resistors in the ↑ spin channel and by two large resistors in the ↓ spin channel in the equivalent resistor network
shown in Figure a. Therefore, the resistance in FM configuration is determined by the low-resistance ↑ spin channel
which shorts the high-resistance ↓ spin channel. On the other hand, in the AFM configuration, ↓ spin electrons are
strongly scattered in the first FM layer but weakly scattered in the second FM layer. Similarly, the ↑ spin electrons
are weakly scattered in the first FM layer and strongly scattered in the second. This is modeled in Figure b by one
large and one small resistor in each spin channel. There is no shorting feasibility now and the total resistance in the
AFM configuration is much higher than that in the FM configuration. This simple resistor model is believed to be
correct but needs to be converted into a quantitative theory that can explain the differences between the CIP and
CPP geometries, and the observed dependence of the GMR on the layer thicknesses and also the material.

Tuhin Maity 121

 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
(3) Resistor Network Theory of GMR:
Periodic Superlattice:
Consider a periodic superlattice of alternating FM and NM layers as shown in Figure with spin-dependent scattering
in the bulk of the FM layers. Since the whole superlattice is made up of identical building blocks, superlattice unit
cells, it is sufficient to calculate the resistances of a unit cell. Consider each FM and NM layers consists of M and N
atomic planes, respectively. Electrons of given spin travelling in a superlattice sees regions of different local
resistivities. The resistivity is high in those regions where there is a high density of states at EF available for
scattering.

Figure: Schematic illustration of Magnetic Superlattice.

There are three different local resistivities in the superlattice unit cell:
the resistivity of the non-magnetic spacer layer, ρNM
- Low resistivity for the same (parallel) spin orientation configuration. This is almost same as resistivity of NM space
layer.
- High resistivity for the anti-parallel spin orientation configuration.
Therefore, the distribution of such regions with different resistivities in a superlattice magnetic cell with FM and AFM
configurations can be schematically drawn as:
Tuhin Maity 122
 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
(3) Resistor Network Theory of GMR:
Periodic Superlattice:

Figure : Schematic illustration of distribution of local resistivities in a magnetic unit cell.

The schematic diagram (Figure) clearly shows that the unit cell of a magnetic superlattice is equivalent to a system
of eight resistors, with four resistors in each spin channel. Therefore, to determine the MR, we need a rule to add up
four resistors in the same spin channel and extend it for both spin channels. Then, resistors of the both spin
channels can simply be added as resistors in parallel to give a total resistance of the magnetic unit cell. This needs
to be done for the FM (↑↑) and AFM (↑↓) configurations. Following the above prescription, the resistance R↑↑ and R↑↓
can be defined as

(5)

Tuhin Maity 123

 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
(3) Resistor Network Theory of GMR:
Periodic Superlattice:
For the FM configuration, the problem reduces to the calculation of the resistance of a two component superlattice
with alternating regions of thickness a and b having resistance ρa and ρb. However, for the AFM configuration a four-
component superlattice needs to be considered. To clarify the underlying physics, let us assume for simplicity that
the resistivity ρa is higher than ρb because there is a higher density of scatterers in the layer a (see Figure).

Figure: Schematic of scattering in Magnetic superlattice.

As there are no impenetrable partitions between the neighboring regions in a superlattice, electrons can go easily
through the interface and undergo scattering in both the layers. In such case, the system is not two independent
resistors, i.e. the two layers do not behave as ordinary resistors in parallel. There are two important physical limits in
which the total resistance of a superlattice can be easily evaluated.

Tuhin Maity 124

 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
(3) Resistor Network Theory of GMR:
Periodic Superlattice:
Case A: If the mean free path in each layer of a superlattice is much shorter than the thickness of the layer, then
very few electrons starting in one layer reach the neighbouring layer. In such case, electrons from different layers do
not mix and flow in their respective separate resistor channel as if the layers were separated by barrier. Hence, all
resistors behave as resistors in a conventional resistor network and should be added in parallel. Therefore, from the
resistivity distribution figure (Figure), there are exactly the same numbers of resistors of each type in the FM and
AFM configurations, and there is no MR in this limit.

Case B: The mean free path in each layer is much longer than the thickness of the layer. For a metallic superlattice
with a small number of atomic planes in each layer, the typical mean free paths are of the order of tens or even
hundreds of interatomic distances. Therefore, one would expect GMR.

Conduction electrons now sample equally layers with low and high resistivity and experience an average resistivity.
For a two-component superlattice, the average resistivity is given by

Tuhin Maity 125

 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
(3) Resistor Network Theory of GMR:
Periodic Superlattice:
Now, using the above generalization, we can go for evaluating the MR:

Adding both part in eqn.(7) results the two-component superlattice resistance for FM configuration

For four component superlattice, eqn.(8) turns out to be

Similarly, for the AFM configuration

Tuhin Maity 126

 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
(3) Resistor Network Theory of GMR:
Periodic Superlattice:
Substituting eqn.(9) and eqn.(10) in eqn.(1), we get

The above equation is now easy to pinpoint the main factors that determine the GMR.

Tuhin Maity 127

 Magnetoresistance
 Spin dependent scattering, Giant Magnetoresistance
(3) Resistor Network Theory of GMR:
Periodic Superlattice:
Eqn.(12) clearly suggests that the

1. (ΔR/R) is a function of two variables, β and (Mμ/N).

2. For large GMR, the spin asymmetry ratio β should be large.

3. For a given value of β, GMR increases with increasing (Mμ/N) values and saturates eventually.

4. On the other hand, the GMR value decreases with increasing the spacer layer thickness, i.e., the GMR falls as
(1/N2) as shown in the Figure below for the Fe/Cr multilayers on the thickness of the non-magnetic chromium layer, .

Tuhin Maity 128

 Magnetoresistance
 Spin–dependent Tunneling and transport
Magnetic Junctions:

At first, we shall focus on simple classification of magnetic junctions. Subsequently, we go into how the basic
models of tunneling need to be modified in the presence of a spontaneous magnetization of the electrodes. Further,
we will discuss (i) the changes that the interface can induce in the basic elastic processes, (ii) inelastic tunneling
mediated by magnetic excitations at the electrodes or at the surface, (iii) the changes in the magnetic structure of
the surface, and (iv) charging effects. In general, the transport between two bulk metallic electrodes can be
classified into two types as shown in Figure:

Figure: Schematic picture of a junction: (a) Tunnel type junction, and (b) Contact type junction.

Tuhin Maity 129

 Magnetoresistance
 Spin–dependent Tunneling and transport
Magnetic Junctions:
1) When the separation between the electrodes exceeds a few angstroms (see Figure a), electrons transfer
between the electrodes by tunneling process. The probability that any one electron tunnels through a barrier of
height V and length l is given by:

where c is a constant of order unity, which depends on the detailed shape of the barrier and on the electronic
wavefunctions. The barrier can be either (i) the vacuum created between the two electrodes, in which case the
height of the barrier is given by the work function of the electrodes, or (ii) by inserting an insulating layer between
the two electrodes. The barrier height in the latter case depends on the position of the edges of the gap of the
insulating material with respect to the Fermi level of the electrodes. It is clear from the eqn.(1) that the tunneling
depends exponentially on the distance between the electrodes, and we may expect that in a junction of
macroscopic size, the current will be due to tunneling events at bulging of the interface. A change of a few
angstroms even can greatly modify the tunneling probability. The conductance at any of these points is given by:

Tuhin Maity 130

 Magnetoresistance
 Spin–dependent Tunneling and transport
Magnetic Junctions:
(2) On the other hand, the two electrodes can be in contact in some points (see Figure b). Then, the conductance of
each contact is given by e2/h times the number of electron channels through the contact. It is roughly given by the
cross section of the contact expressed in units of inverse square Fermi wavevector ( ). Therefore:

(3)

where A is the area of the junction.

Tuhin Maity 131

 Magnetoresistance
 Spin–dependent Tunneling and transport
Magnetic Properties of the electrodes:

In the earlier part of the discussion, we ignored the magnetic properties of the electrodes. If the number of electrons
of the two spin polarizations is not equal, we must define a spin-dependent conductance, at each of the points
where electrons move from one electrode to the other. The expressions defined earlier are modulated by the density
of states (DOS) of each type of electrons. For a given bias voltage V, the electrons which participate in the
conduction come from the levels located at a distance eV from the Fermi energy, EF.

Therefore, to understand the transport at small bias voltages we need to know DOS at EF, D↑(EF) and D↓(EF). As
mentioned earlier, the tunneling magnitude will be spin dependent, as shown in Figure.

Figure: Schematic illustration of DOS at the two sides of a magnetic junction.

Tuhin Maity 132
 Magnetoresistance
 Spin–dependent Tunneling and transport
Magnetic Properties of the electrodes:

The relative resistance change—or effect amplitude—is defined as

where is the electrical resistance in the anti-parallel state, whereas is the resistance in the parallel state.

The TMR effect was explained by Jullière with the spin polarizations of the ferromagnetic electrodes. Tunneling of
spin state from first film into second film is determined by unfilled states of same spin in second. The spin
polarization P is calculated from the spin dependent density of states (DOS) at the Fermi energy:

The spin-up electrons are those with spin orientation parallel to the external magnetic field, whereas the spin-down
electrons have anti-parallel alignment with the external field. The relative resistance change is now given by the spin
polarizations of the two ferromagnets, P1 and P2:

Tuhin Maity 133

 Magnetoresistance
 Spin–dependent Tunneling and transport
Magnetic Properties of the electrodes:

Large spin-polarization is needed for large TMR

Tuhin Maity 134

 Magnetoresistance
 Spin–dependent Tunneling and transport

Tuhin Maity 135

 Magnetoresistance
 Spin–dependent Tunneling and transport
Type of magnetic tunnel junction:

Standard junction Exchange Biased junction Double barrier junction

Tuhin Maity 136

 Magnetoresistance
 Spin–dependent Tunneling and transport
Type of magnetic tunnel junction:

Standard junction

Tuhin Maity 137

 Magnetoresistance
 Spin–dependent Tunneling and transport
Type of magnetic tunnel junction:

Standard junction

Tuhin Maity 138

 Magnetoresistance
 Spin–dependent Tunneling and transport
Type of magnetic tunnel junction:

Exchange Biased junction

Tuhin Maity 139

 Magnetoresistance
 Spin–dependent Tunneling and transport

Tuhin Maity 140

 Magnetoresistance
 Spin–dependent Tunneling and transport

Tuhin Maity 141