Tunneling Magnetoresistance

Chapter 8
Tunneling-Magnetoresistance
The final topic covered in this thesis concerns the current prodigy in cutting edge elec-
tronic engineering. Tunneling magnetoresistance (TMR) may be regarded as the new
candidate for mesoscopic scale magnetic sensors and magnetic random-access memory
elements [76]. Since TMR is relatively new (at least in industrial form), relatively little
is known about these systems. In particular, little (theoretical) research is devoted to
the influence of impurities on the behavior of the MR. It is the purpose of this chapter
to make a contribution to the understanding of TMR and the structures which by their
definition combine aspects from both GMR and TMR structures (referred to as hybrid
formats, see subsection 8.2.1). In particular, we are interested in the effect of disorder on
the MR since quantitatively very little is known about this feature although being quite
relevant to TMR.
8.1 Introduction
The most elementary TMR system is obtained simply by replacing the non-magnetic
spacer layer of a GMR sandwich configuration by an insulating (IS) layer. We write
this system in short-hand symbols as: FM-IS-FM. A principal difference between TMR
and GMR MMs systems is that the former type often maintains a ferromagnetic cou-
pling between the FM electrodes whereas the latter type maintains anti-ferromagnetic
coupling1 .
At the heart of TMR, as with GMR, lies the spin-polarization of the transport current
due to spin-dependent scattering (see section 6.3). The fact that a ferromagnetic metal is
magnetic spontaneously is due to the spin imbalance in the density of d−states. A frac-
tion of these polarized electrons, the itinerant d−electrons, contribute to the (tunneling)
current, as do the unpolarized s− and p−electrons. As a consequence the current is par-
tially polarized. Moreover, unlike with GMR systems where the current is predominantly
carried by s− and p−electrons, the current in TMR systems mainly is constituted by
itinerant d−electrons [77]. It was argued by Stearns [77] that the effective mass of these
1 We recall that the sign of the coupling may even become irrelevant in spin-valves (chapter 7).
136 CHAPTER 8. TUNNELING-MAGNETORESISTANCE
FM IS FM FM IS FM
a mIS = mFM b mIS < mFM
FM IS FM
Figure 8.1: The mapping problem in a FM-IS-FM tunnel junction due to the different effective
masses of the electron inside the FM layers (mFM ) and IS layer (mIS ). a. The situation where
the effective masses are equal, no mapping problem. b. The masses are not-equal, in effect the
dimensions of the barrier representing the IS layer (see section 8.2) shrink and induces a lattice
mismatch at the interfaces. c. A partial way out to this problem is provided by only rescaling the
longitudinal dimension, the transverse dimension remains untampered with, thus maintaining a
proper matching at the interfaces. A drawback of course, is that the actual physical situation
is compromised.
electrons inside Fe(-layers) roughly equals that of a free-electron with a mobility equal to
that of s − p electrons. Inside the mediating IS layer, Stearns deduced that the effective
mass is significantly less than me . Calculations show a mass ranging from 0.2 me to
0.4 me , see Refs. [82] and [83] respectively. On the basis of this rather wide parameter
span it seems that the actual value of the effective mass is not that all important and
suggests a certain stability of the results under ‘variation’ of the value of the electron
effective mass inside the IS layer. As was indicated in subsection 3.2.2 the effective mass
determines the length-scale, being proportional to m∗ −1/2 . Indeed, by varying the effec-
tive mass, one effectively re-scales the system. In 1D, there is a natural matching between
regions supporting different effective masses. In higher dimensions, however, we run into
a mapping problem as illustrated by Fig.8.1. For the dynamic method we partly remedy
this difficulty by re-scaling only the longitudinal dimension. The difficulty for a complete
cure arises from the implementation of the dynamic method which was not designed to
handle a spatially dependent mass-parameter. On adopting different values of effective
masses we may expect to find quantitative changes. However,it seems unlikely, simply by
minor rescaling, to expect qualitative discrepancies to arise.
8.1.1 Discovery of TMR and Early Models

The discovery of TMR dates back to the work of Julliére in 1975 [81]. The TMR sandwich
system Julliére studied consisted of a thin insulating (IS) germanium layer separating
a thin iron layer from a thin cobalt layer. The thickness of the Ge-layer was then
chosen to minimize the magnetic coupling between the two FM electrodes. The two
electrodes have a common easy axis of magnetization but have different coercivities.
8.1. INTRODUCTION 137
This property allows one to have an anti-parallel configuration of the magnetizations of

the two electrodes when the strength of an applied magnetic field (H) is chosen to lie in
between the two coercive fields. The application of higher fields will then result into an
alignment of the magnetizations once more. The mechanism of this basic type of TMR
is analogous to that governing the GMR spin-valve structures. Intermediate magnetic
fields force the magnetizations into anti-parallel mode, thus reducing conductivity. High
positive or negative fields result into a parallel mode, thus increasing conductivity. The
(optimistic) TMR is then found through application of Eq. (1.3) or Eq. (1.5). Similar
to the spin-valve structures, the benefit of this kind of TMR set-up is that the strengths
of H involved are actually of at least an order of magnitude smaller than that required
for magnetic multilayer GMR systems (see section 6.1 and 7.1).
Julliére Model versus Slonczewski Model

There are several theoretical models available to explain the TMR effect which may be
classified into two approaches according to the treatment of the wave function describing
the tunneling electrons. The first and earliest, the Julliére model [81], is based on the
tunnel Hamiltonian theory, in which the tunneling system is artificially separated into
two uncorrelated subsystems by the insulating layer, each having its own Hamiltonian.
This model is based on an assumption due to Tedrow and Meservey [79], stating that
the tunneling current in each spin channel is proportional to the product of the “effective
tunneling density of states” at the Fermi level for the given spin channel for the metals
on either side of the insulator. If the system is in F-mode, Julliére’s model predicts a
conductance (proportional to current in the linear response regime) of
gF ∝ D1↑ D2↑ + D1↓ D2↓ , (8.1)
where Dnσ denotes the density of states in layer n and spin σ. Alternatively, when in
AF-mode we have
gAF ∝ D1↑ D2↓ + D1↓ D2↑ . (8.2)
The pessimistic MR ratio (also see section 1.2) follows from
gF − gAF D ↑ D ↑ + D1↓ D2↓ − D1↑ D2↓ − D1↓ D2↑

γpes ≡ = 1 2 , (8.3)
gF D1↑ D2↑ + D1↓ D2↓
which can be casted into the familiar form on using the definitions of the spin-polarized
currents
D ↑ − D1↓ D2↑ − D2↓
P ≡ 1↑ and P ≡ , (8.4)
D1 + D1↓ D2↑ + D2↓
giving
2PP
γpes = . (8.5)
1 + PP
Alternatively, the optimistic MR is given by
gF
γopt ≡ − 1. (8.6)
gAF
From Eq. (8.5) we obtain gF /gAF = (1 − PP )/(1 + PP ) which, substituted in Eq. (8.6),
yields
2PP
γopt = , (8.7)
1 − PP
which differs from Eq. (8.5) by a minus-sign in the denominator. Indeed, the adjective
‘optimistic’ serves its right well since γopt explodes if PP → 1. Using the definitions
(8.4) this corresponds to either the condition
D1↑ = 0 and D2↑ = 0, (8.8)
or
D1↓ = 0 and D2↓ = 0. (8.9)
We see from Eqs. (8.1) and (8.2) that this corresponds to a perfectly current blocking
AF-mode and one blocked spin-channel in F-mode. This is the pursued ideal case of
a 100% spin-polarized material. This half-metallic feature is explored theoretically in
Ref. [83].
Though the Julliére model is admittedly simple, it is nonetheless quite successful in
predicting TMR values based on current spin-polarization values. A big deficit of the
Julliére model however, is that it does not account for the finiteness of the IS layer, the
model implicitly assumes the FM electrodes to be completely isolated from one another.
The finite width and height of the barrier representing the IS layer1 imply that the elec-
trodes are not at all separate physical systems. The wave-functions of the carriers in
both electrodes overlap, which mandates rather than merely justifies the treatment of
the electrodes and IS layer as one physical system. Indeed, the implementation of the
dynamic and static method originally designed to handle CPP-GMR systems also sup-
ports the mentioned tunneling scenario. In fact, both methods can handle TMR systems
without any principal modifications. The wave-numbers inside the NM layers automat-
ically turn into attenuation constants inside IS layers. The conceptual similarity and
analytical uniformity between CPP-GMR and TMR was also recognized and exploited
by Mathon [84].
The earliest paper, by Slonczewski [85], which acknowledged the just mentioned
proper conception of the barrier only came some fifteen years later after the advent
of the Julliére model. This new type of tunneling model for the FM-IS-FM TMR format
is based on the textbook solution of the free single-electron 1D Schrödinger equation for
a rectangular potential barrier without disorder 2 . The approach Slonczewski adopted
shows similarities with the static method as outlined in chapter 4. He imagined infinitely
wide iron layers sandwiching the IS layer by introducing a spin-dependency into the
wave-number of the transmitting plane waves. He then studied the dependency of the
conductance on besides the barrier height and width also the angle between the magne-
tizations of the iron layers. The part of his result important to us, can be summarized
1 See
section 8.2.
2 This
is a sufficient condition to deploy the 1D SE representing a physical system of higher di-
mensionality also. Since the potential is dependent of the longitudinal (perpendicular) direction only,
the multi-dimensional problem may be separated off into the 1D longitudinal direction and 1D or 2D
transverse dimensions.
8.2. MODEL 139
by the following expression for conductance (in a form adopted from Ref. [83])
g(θ) = gfbf (1 + Pfb

2
cos θ), (8.10)
and
2
e2 κ κ(k↑ + k↓ )(κ2 + m2b k↑ k↓ ) k↑ − k↓ κ2 − m2b k↑ k↓
gfbf ≡ e−2κw , Pfb ≡ ,
πh̄ πw (κ2 + m2b k2↑ )(κ2 + m2b k2↓ ) k↑ + k↓ κ2 + m2b k↑ k↓
(8.11)
where g is the surface

conductance per unit area, P fb is the effective polarization of
the current, κ ≡ 2mb (U0 − E⊥ )/h̄: the attenuation constant in the barrier of a plane
wave with perpendicular energy E⊥ ; U0 − E⊥ is the barrier height as seen by the same
plane wave; mb is the electron effective mass inside the barrier of width w. From Eqs.
(8.10) and (8.11) we see the sensitivity of the conductance on the barrier parameters, as
opposed to the rigidity of the Julliére model. For an interesting comparison on FM-IS-
FM systems of the Julliére model with the Slonczewski model and a validity discussion
of both models with a more rigorous model see Ref. [86].
8.2 Model
The general model employed in this thesis (chapters 3 and 4) can also be applied to TMR
systems. Following convention, we model the IS-layer by a square potential barrier with
a height, denoted by VIS , with respect to the Fermi level. Though the actual form of the
barrier potential is likely to deviate from the ideal rectangular shape, one may substitute
the real form by a suitable (i.e. more-or-less equivalent) rectangular potential conform
the method of Simmons [87].
8.2.1 Hybrid Formats

The introduction of the IS layer, into systems composed of FM and NM layers (GMR
systems) opens the doors to a great diversity of new possible structures. These systems,
composed of FM, NM and IS layers, will exhibit properties of both GMR systems and
TMR systems, and are henceforth referred to as hybrid formats in this thesis. Indeed,
as TMR is a relatively new phenomenon, many hybrid formats still remain unexplored,
the justification for exploring them is then self-evident.
Since the layer make-up of the formats might get rather extensive, we introduce the
following further short-hand notation: FM→F, NM→N, IS→I. Also the minus signs in
between the layer-symbols will be omitted. Example: the format NM-FM-NM-IS-NM-
FM-NM shrinks to NFNINFN. Note that the only role of the outer NM layers is to signify
the termination of the thickness the FM layers, in effect they take over the semi-infinite
property of FM layers in the FIF format. The (hybrid-) formats which have been studied
to date are the following:
FIF This is the first TMR format, pioneered by Julliére [81]. The single insulating
barrier (Ge) is sandwiched by two FM electrodes of different material (Fe and
Co). All aspects of this most basic of all formats were exhaustively studied both
experimentally and theoretically, see e.g. Refs. [81]- [94].
FNIF The insulating layer is separated from one of the FM electrodes by introducing
an NM layer, it is argued in Ref. [96] that this measure increases MR and magnetic
field sensitivity.
FIFIF The most elementary double barrier system, studied in e.g. Ref. [95]. The main
conclusion of this paper is that in this double barrier structure the absolute value
of MR is substantially enhanced as compared to the FIF single barrier system.
Negative and positive MR can be achieved by adjusting the bias voltage.
FIF1 IF2 IF This triple-barrier system is argued to be viable for very large MRs (2000%)
due to the formation of two spin-dependent quantum wells inside F1 and F2 [100].
NFIFN The single barrier sandwiched by two FM layers each connected to NM elec-
trodes. This is a predecessor to the hybrid format. An implication of the inclusion
of NM bordering electrodes into the analysis is that the calculated MR shows de-
pendency on the thickness of the FM layers. In Ref. [97] it is argued that the MR
oscillates with the thickness of the FM layers and may reach a very large value
under suitable conditions. The mechanism responsible for the oscillations in MR
turns out to be a very basic quantum property and will be revealed later on in this
chapter.
FINIF The double barrier sandwiching an NM layer. In turn, the double barrier is
sandwiched by two FM electrodes. This is the first example of an hybrid format,
studied in Ref. [98]. The main finding given in this paper is that with increasing
the NM thickness, the MR exhibits (analogous to the previous format) oscillating
behavior. The MR can allegedly be enhanced by selecting FM layers with high spin
polarization and by controlling the NM layer thickness finely.
In this thesis we will study four types of hybrid formats of which three types were
not previously studied. A vulnerable point in most of the previously discussed hybrid
formats is that the outer FM layers are implicitly assumed to be semi-infinite. In real
structures however, this cannot be valid since the systems are of finite dimensions. The
FM layers are of finite dimensions and are connected to external NM electrodes. The
only format which properly acknowledges this property is the NFIFN format given above.
A consequence of the ‘inclusion’ of the FIF system by NM electrodes is that, through
additional (spin-dependent) scattering at two new exterior interfaces, this generally leads
to an increase in MR as compared to the FIF format. Secondly, the MR now becomes
dependent on the thicknesses of the (outer) FM layers. In fact, the MR turns out to
be oscillatory in the thicknesses of these layers, as was already mentioned above (also
see Ref. [97]). Analogous to GMR systems, the distance between the FM layers in
hybrid formats is assumed to be chosen such that the coupling of the magnetizations
in the FM layers be naturally anti-parallel. This condition forbids the variation of the
distance between the FM layers to arbitrary extent, instead it is reduced to relatively
narrow intervals. Thence, we will mainly vary the thicknesses of the FM layers. Note
8.2. MODEL 141
that the hybrid formats are different in character from the historic FIF format in that
with the latter format the magnetic coupling is naturally ferromagnetic rather than anti-
ferromagnetic. In these FIF systems the two FM layers have different coercivities which
enables one to pin the magnetization of one layer and rotate freely the magnetization
of the other layer (see subsection 8.1.1). In hybrid formats however the FM layers are
assumed to be constructed out of one material only, in this thesis we confine our attention
to Fe and Co only. The representation of the NM layers appearing in the hybrid formats
have either an exclusive Cr or Cu character in case of having Fe or Co FM layers,
respectively.
The next hybrid formats were studied in this thesis:
NFIFN The most elementary hybrid format. Previously studied by Zhang et al. [97],
also see the fourth of the previously formats.
NFNINFN This format is obtained if NM layers are inserted between the FM-IS and
IS-FM bilayers. The thickness of the NM-IS-NM layer sequence must be chosen
such as to maintain an anti-ferromagnetic coupling between the two FM layers. The
idea behind introducing the extra NM layers is to create additional spin-dependent
band-bottom potential wells serving to ‘trap’ the wave-function in between the FM
and IS layers. If the system is in F-mode, the levels of the band-bottom potentials
cause only one of both spin-channels to be trapped effectively. When the system
is in AF-mode however, each spin faces a trap though. The trap for one spin
is located at a different side of the barrier as compared to the other spin. As a
consequence one of the F-mode spin-channels and both the spin-channels in the
AF-mode system experience a heavy reflection, which then leads to high values of
MR.
Due to the typical location of these potential-levels the effect is expected to be larger
for Co layers, where the levels tower above the level of the NM layers, than for Fe
layers, where the levels are ‘submerged’ (also see Fig.8.2). A second motivation
for this measure is to facilitate extra room for impurities. In case of having Fe
FM layers and Cr NM layers such measure is expected to have a positive effect of
impurities on the MR since annealing FeCr GMR multilayers increases MR (see
chapter 6). We may, at least, hope to find a lessening of the reducing effect on the
MR of impurities.
NFNFIFNFN If FM layers are added to the exterior of the first format mentioned in
this list, we obtain the present format1 . The motivation for taking such a step is
again to facilitate extra space for impurities.
NFNFNINFNFN The last format treated in this thesis arises if extra NM layers are
inserted adjacent to the IS barrier in the previous format. The motivation is two-
fold and identical to that given in the discussion of the second format.
1 The format is thought to be enclosed by NM electrodes.

IS IS
Co Co
Cu Cu Cu Cu Cr Fe Cr Cr Fe Cr
spin−dependent ‘traps’ less efficient

‘traps’
a b
Figure 8.2: The band-bottom potential profiles of NFNINFN hybrid formats illustrating the
‘trapping’-mechanism responsible for an increase in MR in particular for Co FM layers (a)
compared to Fe FM layers (b). Both formats are in F-mode.
8.2.2 Basic Format

Aside from the previously discussed hybrid formats we will also briefly focus on the basic
FIF format using the static 1D method of chapter 4 and calculate the MR as a function
of barrier height and thickness. The acquired data then serves as a means for comparison
with the more complex hybrid formats.
The system of interest describes the original Fe-Ge-Co tunnel junction pioneered by
Julliére [81]. The ‘incoming’ static waves enters the Fe layer and by the FM nature of the
electrode are given a spin-dependent momentum. The ‘outgoing’ or transmitted waves are
defined on the Co layer. Since the Fe and Co layers are modeled by different band-bottom
potentials the momentum of the wave changes in going from Fe to Co. The transmission
coefficients are defined as, following convention, the outgoing current divided by the
incoming current. In case of having NM terminals the momenta of the incoming wave
cancels out with that of the transmitted wave to give the usual transmission coefficients of
the form |t|2 (see chapter 4). On having FM terminals, however, we obtain transmission
coefficients of the form |t|p2 /p1 , with p1 and p2 the momentum of the outgoing and
incoming wave, respectively. Also note that these momenta are spin-dependent and that
the outgoing momentum also depends, by set-up, on the mode of system (AF-mode or
F-mode) (also see Fig.8.3).
The results of the calculations are given in Fig.8.4. The effective mass of the electrons
in the Ge barrier was set at m∗ = 0.082 me [99]. Fig.8.4 basically exhibits two features:
the MR is a decreasing function of the width of the insulating layer and increasing
function of the barrier height.
The former feature, is a general property not specifically owed to quantum mechanics.
To prove this, it suffices to advance a more simple non-quantum model which features
the same property. Consider a modification of the resistor model as given in section 6.2.
To represent TMR systems, we insert an additional resistor with value rIS , representing
the insulating layer, in between the resistors that represent the FM layers. We assume
resistance of the resistor to be proportional to its length (the width of the IS layer).
8.2. MODEL 143
FM1 IS FM2
eik↑x
re−ik↑x teip↓x
p
T = |t|2 k↑↓
Figure 8.3: Illustration of the redefinition of the transmission coefficient in case the terminals
are of FM nature. The F1 IF2 format is in AF-mode. A spin-up wave enters from the left
with momentum k↑ and is partially transmitted with an altered momentum p↓ , yielding the
expression for the transmission coefficient T as given in the figure.
0.07 0.12
0.06 0.1
γ 0.05 VIS = 0.2 EF γ VIS = 0.2 EF
VIS = 0.5 EF 0.08 VIS = 0.5 EF
0.04 VIS = 1.0 EF VIS = 1.0 EF
0.06
0.03
0.04
0.02
0.01 0.02
0 0
a 2 4 6 8 10 12 14 16
Th. IS (# atoms)
18 20
b 2 4 6 8 10 12 14 16 18
Th. IS (# atoms)
20
Figure 8.4: The original Fe-Ge-Co tunnel junction given a traditional analysis. The band
parameters of the Fe layer were taken from Expr. (6.18). a. The results on MR with Co band-
parameters of Expr. (6.20). b. The MR obtained with Co band-parameters of Expr. (6.21).
Then repeating the same steps of section 6.2, we end up with a modified expression for
the MR
β̃2 β
γ=− where β̃ ≡ . (8.12)
1 − β̃ 2 1 + 2 rIS /r̄
1
As the layer-width increases, then so does rIS which, in return, reduces β̃ and γ. Hence
we have shown that the reduction of MR in a TMR system with the width of its insulating
layer is not exclusively a quantum property.
The second property, the increase of MR with the barrier height, however is a true
quantum (tunneling) property of the barrier. Consider the system to be in F-mode.
Recall from elementary quantum mechanics that the amplitude √ of the wave function
decays, in the process of tunneling through a barrier, as e− VIS −E⊥ d , with d being the
barrier thickness and E⊥ the perpendicular (longitudinal) energy. If the barrier is low
(but still above Fermi level), the channels with the highest E⊥ will suffer the least of
VIS > 0
VIS = 0
EF
2th 2th
2tl 2tl
a FM IS FM b FM IS FM
Figure 8.5: Illustration of the physical mechanism responsible for the increase in polarization of
the current (conductance) and hence MR, as a reaction to a boost in VIS . Consider the tunneling
of two plane-waves, one with high and one with low (perpendicular) energy, through a barrier.
a. The barrier is low, meaning that the discrepancy in rates of transmission between high and
low energy channels, relatively given by |th |2 and |tl |2 , is greater compared to the high barrier
case, given in sub-figure b with the rates of transmissions given by |th |2 and |tl |2 . In summary,
|th |2 /|tl |2 > |th |2 /|tl |2 .
decay through the barrier and will contribute most significantly to the tunneling current.
The channels with the lowest E⊥ will rapidly decay inside the barrier and will contribute
least significantly to the current. The effect of the barrier, on increasing its height VIS ,
is to reduce the differences in transmission of the tunneling current between the various
channels. That is, the channels with highest E⊥ loose significance while the channels
with low E⊥ relatively gain significance. In due process, the total number of channels
gradually gains relevance over the privileged highest E⊥ channels. A graphical support for
this line of argument is presented by Fig.8.5. Since the system is in F-mode, this means
that the number of majority spin-channels exceeds that of the minority spins. Therefore,
increasing VIS results in an increasing discrepancy between the transmitted currents due
to the majority and minority spins. Hence, the current and thus conductance gets more
polarized (in favor of majority spin).1 When the system is in AF-mode, only the channels
with perpendicular energy above the minority spin level contribute to the tunneling
current. The number of channels contributing to the current then is spin-independent
and the tunneling current is roughly equal for both spins. The increase of barrier height
then does not lead to an increase in spin-polarization of conductance. In summary,
increasing the barrier height increases the polarization of the current (conductance) in
F-mode which, in turn leads, by increasing the difference between F-mode conductance
and AF-conductance, to an increase of MR.
1 Note that this consequence is relative, the actual magnitudes of current and conductance drop.
8.2. MODEL 145
8.2.3 Sub-ensembles
The way the 2D dynamic method is set-up here is identical to the set-up of the GMR
systems except for one subtle but significant modification. Consider the definition of
the multi-Gaussian Eq. (3.45), it turns out that in tunneling formats the transmission
is strongly dependent on the set of random phases {φn }. Each set of random phases
gives a significantly different transmission than the next, the standard deviation from
the average transmission is around 20-30 percent.
The solution to this difficulty is provided by introducing sub-ensembles. A sub-
ensemble consists of a set of equivalent multi-Gaussians, with each member having its
own unique set of random phases {φn }. The multi-Gaussians are launched in on the
system1 at hand separately. The accumulated transmission coefficients then serve as the
transmission coefficient representative to that system. The above steps are then repeated
for the entire ensemble of systems in the usual manner (chapters 5 and 6). The average
of the accumulated transmission coefficients is then used to obtain the magnetoresis-
tance. The criterion on the size of the sub-ensembles is reasonable convergence of the
magnetoresistance. In practice, a size between 8 to 16 seems to suffice. A big drawback
of having to invoke the concept of sub-ensembles is that the required computing time is
proportional to the size of the sub-ensembles, in practice this means that this amount of
time easily increases by a factor of ten as compared to the single-member subgroup case.
We stress though, that this effect is exclusively pronounced for tunneling systems, the
fluctuations in transmission of GMR seem to be of an order of a magnitude small. We
therefore did not consider it to be worth the effort to apply the concept of sub-ensembles
to GMR systems too.
8.2.4 Disorder
Since the role of disorder in TMR systems has not yet received much attention, we con-
sider it worth while to study its effect on TMR systems. Disorder enters the model in the
two distinct ways as mentioned in section 1.3: By introducing finite-sized spin-dependent
potential bars (representing impurities) and a non-uniform variation of the layer thick-
nesses. The impurities reflect the occurrence of foreign stray atoms in, the otherwise,
pure system and the procedure of introduction was already discussed in subsection 6.3.1.
It is argued (see e.g. [94]) that disorder in the form of impurities has a detrimental ef-
fect on the value of TMR, this may indeed be the case for simple TMR formats (such
as FM-IS-FM) but for GMR-TMR hybrid structures this may very well be different as
e.g. the GMR is seen to increase with the degree of disorder for FeCr MMs (see chapter
6). Therefore it might be rewarding to explore the influence of disorder on these hybrid
structures.
The CPP character and their modest size make impure TMR systems very well suited
for treatment by the dynamic model. In addition, since the 1D static model with impu-
rities proved to be ill-suited for this matter2 the value of the dynamic model increases
even further. However, the validity of the 1D static model in the clean limit still remains.
1 With ‘system’ is meant here a system with an, in principle, unique impurity configuration.
2 The rather obtrusive character of the impurity modeling in the 1D static model is held responsible
NM FM NM IS NM FM NM
FM atom IS atom NM atom

Figure 8.6: Illustration of a 2D disordered hybrid format (NFNINFN) system.
As with GMR systems, disorder, in the form of interface roughness (see subsections
6.1.3 and 6.3.1), is introduced by annealing the samples. It is interesting to note that
annealing serves a two-staged role in TMR systems. It is found experimentally [94] that
on exposing the TMR samples to short annealing up the MR ratio improved considerably.
The temperature leading to an optimum MR was inferred to be around 230o C beyond
which the interfaces began to deteriorate. The improvement in the interface properties
upon annealing has been attributed to barrier homogenization [94] and better magnetic
properties of the FM films near the interfaces. Other factors may include reduction of
the defect density in the barrier and sharpening of the FM-IS interfaces. Some of these
suggestions are supported by the slight (inferred) increase of the barrier height as well
as the large increase in MR. Beyond 230o C, annealing will cause degradation of the
interfaces in the sense of inducing the familiar mechanism of interlayer atomic diffusion
(see subsection 6.3.1).
Since our clean TMR systems already correspond to the optimum anneal result by
construction, we will be restricted to implement the second stage of annealing only. Just
as with GMR systems interface impurities are conceived as foreign stray atoms originating
from a neighboring layer migrated into the host layer. Since we now have three types
of layers, FM, NM and IS, we now also have three kinds of (interface) impurities, see
Fig.8.6 for a typical depiction of a disordered hybrid system.
Disorder within the insulator was the subject of recent theoretical investigation by
Tsymbal and Pettifor (T&P), see Refs. [101], [102] and the review article Ref. [94]. In
the first article T&P introduced disorder through a random fluctuation in the on-site
potentials within the insulator. Then, through calculating the conductances of both spin
channels the polarization of the current was found, from which they calculated the MR
for the unrealistic results obtained for the impure 1D TMR systems. It was observed that the strongly
fluctuating TMR rose far beyond the “clean” value with standard deviations greater than the actual
TMR ratios. For this reason we abandon the 1D impurity implementation of the 1D static model and
confine our attention strictly to clean 1D systems.
8.2. MODEL 147
by the Julliére formula (8.5). Their main finding was that disorder induces the formation
of highly conducting resonant states connecting the two metals through the insulator.
This effectively reduces the height of the tunneling barrier which then reduces the spin-
polarization of conductance and hence the MR. They further argued that the effective
height of the barrier decreased still as the width of the disordered insulator was further
increased.
In their second article [102], T&P introduced impurities next to the random on-site
potential fluctuations. The on-site potential of the impurities was fixed but distributed
randomly across the insulator. By calculating the DOS at both sides next to the insulator
they again estimated the MR through invoking the Julliére formula. They found that
the introduction of impurities lead to the formation of an impurity miniband within the
band-gap of the insulator. The latter mechanism provides an additional contribution
to the conductance next to the resonant state formation due to disorder. Aside from
one special case, they found that the polarization of the conductance and hence MR
decreased with impurity concentration. In light of the barrier potential picture, this
scenario is equivalent to the formation of corridor like states through the conductor in
which tunneling can proceed with greater ease. The formation of these corridors strongly
increase conductance and, in fact, will dominate over the impurity-free contribution to
the tunneling conductance. As the impurity level increases, then so does the formation of
corridors. Since corridors in effect lower the effective barrier height, the MR will decrease
with increasing impurity levels.
From the findings of T&P we may infer what the effect will be if, the otherwise pure,
tunneling system is provided with impurities. Firstly, impurities that model diffused
atoms from the insulator into the metal layers adjacent to the insulator will feature
insulating properties. Hence their presence will effectively widen the barrier, and have a
decreasing impact on the MR. Secondly, impurities that model diffused metallic atoms
into the barrier will help create relatively high conductive corridors with a much lower
effective barrier height. There is however a subtle discrepancy in character of metallic
impurities inside the insulator T&P used and the ones we use. T&P parameterize these
impurities with a spin-independent constant on-site potential. In our model, whether or
not these impurities are represented by spin-dependent or spin-independent potentials is
dependent on the character of the layers next to the IS layer. If the neighboring layers
of the IS layer are ferromagnetic we define the on-site impurity potential for majority
spins by an amount of two times the magnitude of the magnetization lower than that for
minority spins (Zeeman splitting). On the other hand, if these layers are non-magnetic
we define the on-site potential spin-independent (analogous to the paper of T&P). In the
former case, this measure introduces a subtle spin-dependency into the corridor tunneling
mechanism, giving a slight increase in MR as compared to the latter case, producing spin-
independent corridors.
In summary, the effect of impurities modeled here, will have a negative impact on
the MR by effectively widening the layer and lowering the barrier height. The effect of
impurities within the insulator will depend somewhat on the type of these impurities,
the strong decreasing effect of NM impurities will be milder with FM impurities.
8.2.5 Algorithms of Study

In the types of hybrid formats taken up here, the character of the FM layers uniformly
is either iron-like or cobalt-like. The NM layers mimic either chromium or copper. To
maintain contact with the GMR systems, the FM-NM combination remains either FeCr
or CoCu. For the usual band parameters that go with them, see section 6.3 (we have
chosen the band-parameters of Expr. (6.20) for the Co layers).
The FM layers appearing in the hybrid formats modeled here, are thought to be of
identical make-up analogous to the MMs GMR systems treated in chapter 6 and 7. By
the same mechanism as in MMs, this will lead to a negative MR. We stick to F/AF-mode
type of MR in this thesis and quantify MR by utilizing Def. (6.1).1
We are interested in two matters of investigation concerning hybrid formats. Firstly,
the dependence of the MR on the thicknesses of the layers that constitute the format,
especially the FM layers. Furthermore, we will try to determine the conditions which
lead to the highest values of MR. The method to study this, is an application of the
1D static method as outlined in chapter 4. Secondly, the dependence of the MR on the
impurity concentration. For this purpose we use the 2D dynamic method of chapter 3.
We will vary the height of the potential barrier VIS leveled from 0 to 1 EF above Fermi
level. We assume this range to be sufficiently big to cover all possible and actual barrier
heights.
After the layer-configuration has been defined, each format is subjected to the follow-
ing two algorithms of study. In either program, each layer in every format is allowed to
vary in some predetermined thickness-interval, we call this set of ranges the thickness
spectrum. To respect the demand for having an anti-ferromagnetic coupling between
the magnetizations of consecutive FM layers we will primarily vary the FM layer thick-
nesses and keep variation of the thicknesses in the other layers to a minimum.
In the first program the thicknesses of the NM layers and IS layers are set fixed. The
MR as a function of the thickness of the FM layers is then calculated according to the
following scheme:
1. The height of the IS barriers is set at Fermi level (VIS = 0).
2. The thicknesses of all NM and IS layers assume initial values from the thickness
spectrum.
3. The thicknesses of all FM layers assume initial values from the thickness spectrum.
4. The MR is calculated and recorded.
5. The thicknesses of two of the FM layers are increased simultaneously until the
spectrum has been ran through, in the latter case we proceed with the next step
otherwise we return to step 4.
6. The thicknesses of the NM and/or IS layers are changed as permitted by the thick-
ness spectrum, we return to step 3 if the spectrum has not been completed or
proceed with step 7 otherwise.
1 If we would consider the angular variation of MR, we would prefer Def. (7.2) over Def. (7.1).
8.3. RESULTS 149
7. The height of the IS barriers is increased by some fraction of EF until the height
has reached 1 EF above Fermi level, we then terminate or continue by returning to
step 2 otherwise.
The impurity aspect of hybrid formats is studied according to the following optimiza-
tion algorithm:
1. The height of the IS barriers is set at Fermi level.
2. All layer thicknesses assume initial values as prescribed by the thickness spectrum.
3. The MR is calculated and recorded.
4. The thicknesses of the layers are changed allowed by the thickness spectrum. Until
the spectrum has been ran through we return to step 3 or proceed with step 5
otherwise.
5. The resulting spectrum of MRs is sorted from high to low.
6. The height of the IS barriers is increased by some fraction of EF until the height
has reached 1 EF above Fermi level, we then proceed with step 7 or return to step
2 otherwise.
7. The average1 spectrum of sorted MR is calculated and yields the average MR ‘top
10’.
8. The configurations of the MR top 10 qualify for further investigation using the
dynamic model with the sub-ensemble facility to study the influence of impurities.
A criterion for a configuration to proceed to 2D analysis is that it has to offer room
to an extensive number of impurities, i.e. the system should not be too small.
8.3 Results
In short, our objective was to study the dependence of the MR of hybrid formats on
layer thickness variation. Secondly we were interested in the influence of impurities on
the MR.
The thicknesses of the layers appearing in some format are united in string form.
Each string then corresponds to a unique system, and is referred to as a configuration.
Consider for example the configuration: {a, b, c}, meaning that the format consists of
three layers (terminated by NM layers), with the layers having respective thicknesses
a,b and c in units of one atomic diameter. If some of the characters in the string are
replaced by dots, e.g. {., b, .}, this means that the thicknesses of the corresponding layers
are varied simultaneously within a specified range. If the string features the subscript,
2×, as in {a, b, c}2× it means that the results of the configuration at hand also apply
to its mirror image (the mirror image of {a, b, c} is {c, b, a}). The thickness spectrum of
1 For each member of the thickness spectrum: every barrier height produces an MR value, the average
is taken over all such values.

some format is written also in string form. Consider the example [a − b, c − d, e − f],
meaning that the thickness of layer one varies from a to b, of layer two from c to d and
of layer three from e to f.
The following estimations for the Fermi energies were used. In case we had Fe FM
layers we set EF = 5.8 eV (Fermi energy of Cr). In case of Co FM layers we set EF = 9.4 eV
(Fermi energy of Cu). Stearns [77] argued that inside Fe layers of tunneling junctions
the effective mass of itinerant electrons equals the free electron mass: m∗ = me . We will
extend this assumption to Co FM layers and both types of NM layers. Inside barriers we
will assume a lower mass, m∗ = 0.3 me . Together with the values of the Fermi energy
we obtain from Eq. (3.31) the following length-scales (atomic diameter): 0.4 λF inside
Co and Cu layers; 0.3 λF inside Fe and Cr layers; 0.2 λF inside barriers.
8.3.1 1D Systems
The results on MR using the 1D static method of chapter 4 are presented in Figs.8.9-8.15.
The values of the barrier potential VIS were chosen to lie in an interval that covers all
possible values extracted by inference from experiment. In FM=Fe systems we set the
maximum barrier potential height at 1 EF . In FM=Co systems we set the maximum
barrier height 0.5 EF .
The results on the MR as a function of the thickness of two of all FM layers are
summarized in Figs.8.9-8.14. If in the figures, the configurations associated with some
format are given exclusively in the top sub-figures (a and e) it means that the results in
the sub-figures below them we obtained with the same configurations.
On comparing the results with that of the simple FIF format (Fig.8.3), we notice at
once the more complicated evolution of the MR with the thickness of the FM layers.
Since the thickness of the FM layers in the FIF were effectively infinite, this discrepancy
has to be owed to the finiteness of the thicknesses of the FM layers in the hybrid formats.
Another striking difference with the FIF format is that hybrid formats do not seem to
show an unconditional increase in MR with the height of the barrier. Instead, hybrid
formats seem to feature a contrast increasing property between the graphs of several
configurations. Increasing the barrier height seems to either pull the graphs apart or
make the peaks more pronounced. In general though, increasing the barrier height seems
to have a positive effect of the MR. The oscillatory behavior in MR as a function of the
thickness of the FM layers calculated by other authors [97] on the NFIFN format, is also
found through our calculations, as Fig.8.9 shows.
The oscillatory behavior in MR with the thicknesses of its layer components however
is not restricted to the NFIFN format but is observed in all formats. The cause of
this very general feature is a very fundamental quantum mechanical one. Consider the
format NFN, an FM layer of thickness a enclosed by two NM terminals. The 1D SE
is solved by the standard procedure (see chapter 4). The situations with all the usual
definition is depicted in Fig.8.7a. At the RHS the wave is transmitted into positive
infinity. The coefficient of transmission |t|2 is dependent on the position of the FN
interface due to the matching requirement of the wave, as Fig.8.7b illustrates. It is
precisely due to the oscillatory behavior of the wave inside the FM layer and the matching
of the wave that the transmission amplitude t oscillates too. This is the essence of
8.3. RESULTS 151
NM FM NM FM NM FM NM
2
eikx + re−ikx teikx |t| < |t |
2
aeipx + be−ipx t
t
0 x0 x0 x0 x0 + δ
a b
Figure 8.7: Illustration of the quantum mechanism ultimately responsible for the inherent
oscillatory behavior of the MR in the thicknesses of its (non-insulating) layers. a. The SE-
type of set-up of a NFN format, the amplitudes are fixed by the usual matching of function
and derivative. b. A close-up at the FM-NM interface with coordinate x0 : the transmission
amplitude t follows the oscillating wave within the FM layer. After increasing the size of the
FM layer, by an amount δ0 , the transmission amplitude changes also.
the mechanism which ultimately is responsible for the observed oscillations in MR (the
collective transmission gives conductance and conductance gives MR). The complexity
of the nature of the oscillation in MR increases with the complexity of the configuration
of the format. Note also that the layers defined in the format (only the FM-layer in this
case) must account for at least one k−channel which does not have an imaginary wave-
number (attenuation constant). If the layer becomes insulating the wave-numbers become
attenuation constants and the transmission-coefficient loses its oscillatory quality. Thence
from the above argument we might infer that the MR is oscillatory in the thicknesses of
all conductive layers.
Let us have a closer look at Fig.8.9. First of all, regarding the LHS figures (FM=Co),
the MR of the configurations {., 3, .} remains more or less immune to barrier height in-
crement. Secondly, the configurations with larger barrier widths respond with a decrease
in MR on increasing the VIS . From an application point of view, the plots given in e-h
(FM=Fe) are perhaps more interesting. The configurations responsible for these figures
allow for the formation of peaks in MR, which become ever more pronounced as the
height of the barrier increases. A feature of both the Co and Fe systems is the phe-
nomenon that the MR generally drops with increasing barrier width (see Fig.8.4). This
feature is more clearly pronounced though for high barriers. In conclusion, in order to
achieve maximal MR values one should stick to Fe layers, minimize the barrier width
and maximize the barrier height (by finding suitable barrier-material, and by sharpening
the interfaces). Then fine-tune the thickness of the FM layer to peak-value in MR. Since
the first peak occurs for low FM thicknesses this suggests that the size of the systems
may be held small.
Consider the format NFNFIFNFN, which might be conceived as the NFIFN format
sandwiched by additional FM layers. We now simultaneously vary the thicknesses of
either the outer FM layers or the inner FM layers. The results are given by Fig.8.10
(FM=Co) and Fig.8.11 (FM=Fe). The oscillating behavior of the MR is again well-
represented for both the Fe and Co formats. First consider the variation of the thick-
nesses of the outer FM layers a-d of both figures. The figures suggest that the MR
increases significantly on simultaneously increasing the width of the inner FM layers.
This however is only apparent, we recall that the MR is oscillatory in the thicknesses
of all conductive layers. Additional calculations, not presented here, indeed verify this
behavior, by revealing a subsequent drop in MR on further increasing the widths of the
inner FM layers. The variation of the thicknesses of the inner FM layers are given in
e-h. Both FM=Fe as FM=Co formats show, aside from the oscillatory behavior, a pro-
nounced peak development in MR. It is remarkable that the MR due to the FM=Co
formats maintains a near-insensitive response to the increase in VIS . In the FM=Fe for-
mats on the other hand, we see ever more pronounced peaks emerging as VIS increases.
Apart from the peak-behavior, both for FM=Fe and FM=Co systems there is no real
noteworthy discrepancy between the evolutions of the MR from the several configurations
(‘spaghetti’). The results suggest that the MR drops with the barrier width. Through
additional calculations, not presented here, it has been verified that indeed the MR in-
creases as the width of the barrier decreases. In order to find the largest values of MR
the directions given at the end of the discussion of the previous format also apply to this
one.
The format NFNINFN can be thought to be constructed out of the NFIFN format by
insertion of two NM layers adjacent to the IS barrier. The reason for doing so is explained
in the text, see subsection 8.2.1 and Fig.8.2. The results on this format are given by
Fig.8.12. The discrepancies between the FM=Co and FM=Fe formats are striking. First
of all, while the Co formats show a typical single-peak1 behavior the Fe formats maintain
oscillatory behavior in MR. Also the MR again seems to be insensitive to increases in VIS
for FM=Fe formats, while the MR in FM=Co formats increase strongly as a function of
VIS . Also the MR strongly increases with barrier width of FM=Co formats while slightly
decreasing for the FM=Fe formats. Through additional calculations we have confirmed
that this behavior perseveres for wider barriers and leads to spectacular gains in MR in
FM=Co configurations. The MR in FM=Fe continues to decrease mildly as the barrier
width increases. To achieve the highest MR values, one should choose FM=Co systems,
make the barrier as wide as permitted (by the AF-coupling requirement) and fine tune
the widths of the FM layer simultaneously until peak-value.
The large discrepancy between the MR values due to NFNIFN formats with either
FM=Co or FM=Fe urges a closer analysis of the involved mechanisms. In subsection
8.2.1 we casually anticipated on the results to come by mentioning the so-called trapping
effect. It was on the brink of completion of this thesis that we stumbled on a typical
counter-intuitive manifestation of quantum mechanics. Consider Fig.8.2a. Intuitively
we expect that the total transmission of the spin-up channel T↑ exceeds that of the
spin-down channel T↓ 2 since the potential barrier spin-up waves face is less than that
of the spin-down waves. Secondly, we expect the polarization factor T↑ /T↓ to be larger
for the NFNINFN FM=Co format than the NFIFN FM=Co format to account for the
1 The graphs have been cut-off at a thickness of 20 atomic diameters, extension to 30 add nothing
new since the plots remain flat.

2 The system is assumed to be in F-mode.
8.3. RESULTS 153
1.8 2
1.7
T̄↑ 1.8
T̄↑ 1.6 NFIFN FM=Co T̄↓1.6 NFNINFN FM=Co
1.5
T̄↓1.4 1.4
1.2
1.3
1
1.2
0.8
1.1
1 0.6
0.9 0.4
0.8 0.2
-1 -0.5 0 0.5 1 1.5 2 -1 -0.5 0 0.5 1 1.5 2
a VIS (EF) b VIS (EF)
1.06 1.25
1.04 1.2
1.02 1.15
T̄↑ 1 T̄↑
T̄↓0.98 T̄↓1.1
1.05
0.96 1
0.94
NFIFN FM=Fe 0.95
0.92 0.9
NFNINFN FM=Fe
0.9 0.85
c -1 -0.5 0 0.5 1
VIS (EF)
1.5 2
d
-1 -0.5 0 0.5
VIS (EF)
1 1.5 2 2.5
Figure 8.8: The average polarization factor T̄↑ /T̄↓ as a function of the barrier height for the
NFIFN formats at the LHS (a and c) and the NFNINFN formats (b and d). The thickness
spectra averaged over are [5-15, 3-4, 5-15] and [5-15, 3, 3-4, 3, 5-15] for the NFIFN and NFNINFN
formats, respectively.
tremendous elevations in MR for NFNINFN configurations as compared with the NFIFN.

However, additional calculations reveal an equally unexpected and significant twist to this
scenario. We have calculated the average of T↑ /T↓ as a function of the barrier height for
the NFIFN and NFNINFN formats, the results are given in Fig.8.8.
First we consider the FM=Co systems, Fig.8.8a and b. Intuitively, we expect the
average of T↑ /T↓ to be unconditionally greater than one. For low barrier heights this
expectation is indeed verified, but on increasing VIS the polarization becomes significantly
less than unity. That is, the polarization inverts. On comparing the sub-figures a and b,
we notice that the inverting of polarization is more pronounced for the NFNINFN formats
than the NFIFN formats. Though less severe, in Fig.8.8c and d a similar behavior is
witnessed for the FM=Fe formats (with the role of spin-up and spin-dn interchanged).
From this observation we infer that the NM layers surrounding the IS layer, has a decisive
role in the inverting of the polarization. Regrettably, the reason for this remains obscure,
and we do not hesitate to mention that further analysis on this peculiar quantum effect
is needed to clarify its origin.1
The last format treated in this thesis arises when the previous format NFNINFN is
sandwiched in between two FM layers (in turn attached to NM electrodes, of course).
The results are presented in Fig.8.13 and Fig.8.14. We treat Fig.8.13 first. At the LHS
of the figure, we notice, as VIS increases, the gradual appearing of a peak-like behavior
1 We note that the inverting of the polarization factor is not an artifact due to our 1D implementation.
Calculations on 2D systems reveal similar behavior.

in MR as a function of the thickness of the outer FM layers, yet preserving its oscillatory
character (though of secondary role). This phenomenon is most pronounced with the
{., 6, 8, 2, 5, 2, 8, 6, .} configurations (pluses). This is an example of a format that shows
an actual increase in MR with the barrier width. In general, the MR increases with
the barrier height significantly and consistently. Notice however, that these gains in MR
with the barrier height are not as spectacular as seen with the NFNINFN format. The
addition of extra FM layers clearly reduced the spin-trapping mechanism to significant
extent.
At the RHS of Fig.8.13 and Fig.8.14, where the thicknesses of the inner two FM layers
are varied, we again notice the development of peaks as VIS increases. The configura-
tions responsible for this behavior, once more, reveal a positive effect of increasing the
barrier width. Apart from the peak-behavior, the evolutions of the MR again tend to
spaghetti-like clustering. On regarding Fig.8.13, we at once notice the tendency of the
MR evolutions to remain insensitive to an increasing VIS .
In the last figure on 1D systems: Fig.8.15 we list the evolution of the MR with the
height of the barrier VIS . For each format, we give the seven configurations responsible
for the highest average MR. For this purpose we have used the optimization algorithm of
subsection 8.2.5. The general observation is that although often the MR is an increasing
function of the barrier height, more than sometimes it is not. This is contrary to what
one might expect based on the results for the FIF format (subsection 8.2.2) where the
MR increases unconditionally with the barrier height. This seems to imply that in ‘real’
systems, i.e. systems that do not have infinitely wide layers, the behavior of the MR as
a function of VIS is more complicated than simply non-decreasing.
We conclude the discussion on 1D systems by making some household remarks on
the number of kx −points needed to achieve proper convergence of the MR1 . It turned
out that tunneling systems require more points than GMR systems. In most formats
a number of O(1000) sufficed for both Fe and Co FM layers. In the NFNINFN and
NFNFNINFNFN formats with FM=Co, however, the number of required points is an
increasing function of the barrier thickness and height. The number of required points
ranged from O(104 ) to O(105 ).
8.3.2 2D Systems
The following symbols were used in the presentation of the results on 2D systems. The
transverse dimension of the system Ly is given in units of the atomic diameter. The
spreading in the impurity acceptance criterion σ as appearing in Eq. (6.22) basically
determines the concentration of interface impurities. The smaller σ is chosen the more
the impurities will be centered around interfaces.
The presence of impurities is quantified by using the concentration-concept similar to
that used in the domain wall problem (chapter 5). The concentration or percentage is
defined as the total area covered by impurities divided by the system size. In the results,
the average values (ensemble size: 8 members, sub-ensemble size: 8-16 members) of the
MR is given by small diamonds interconnected by solid lines (guides to the eye), the MR
1 The longitudinal component of the momentum k ranges from zero to one (the Fermi momentum,
x
see chapter 4).
8.3. RESULTS 155
due to individual ensemble members form the dotted lines (spaghetti) appearing at the
background.
The results on 2D systems are condensed into Figs.8.16-8.19. We have performed
the calculations for the two outer values of the allowed barrier-heights. In case the MR
evolution behaves similar for both edge values, one might assume this also to be the case
for intermediate values of VIS . If there is a discrepancy, educated guesswork seems to be
in place in the form of interpolation.
First consider Fig.8.16. At VIS = 0 we clearly see an average negative consequence
of increasing the impurity concentration. However, at VIS = 1.0 EF FM=Fe we notice a
tendency of the MR to actually increase with the impurity density, although with large
standard-deviations. The large standard-deviation in MR is due to the great fluctuations
in conductance as a result to the great sensitivity of the conductance to the precise
distribution of the impurities across the system.
Regarding Fig.8.17, we notice the following features. First, at VIS = 0.0 EF the
average MR drops with the impurity concentration. At the higher barrier potential, we
notice the formations of minima in the evolution of MR. This result is a consequence
of the competition between the impurity MR-enhancing GMR mechanism (impurities in
the FM-NM regions) and the usual TMR mechanism (impurities in the IS vicinity).
The general behavior displayed in Fig.8.18 is a uniform reduction in MR as a function
of the impurity density. This response may be due to the spin-independent NM nature
of the impurities within the barrier. It was stated in the text that the MR decreasing
effect of NM impurities inside the insulating layer is stronger than that of FM impurities
residing in the insulator.
The results given in Fig.8.19 are more promising. The addition of two FM layers
to the sides of the NFNINFN format has a positive effect on the evolution of the MR
(compare Fig.8.19 with Fig.8.18). In case of FM=Co at VIS = 0.5 EF we even observe the
presence of a maximum, though with large standard deviations, at low impurity densities.
In the last part of the analysis of the 2D systems we have tried to find correlations
between properties of the impurity distributions across the systems and the MR values.
For example, one may hope to find a relation between high MR values and the preferred
occurrence somewhere in the systems of one of the impurity species (FM, NM or IS). Be-
sides the distribution of impurities the value of VIS proved to be rather influential in the
evolution of MR with the impurity concentration. This is seen on noticing the discrepan-
cies between sub-figures a and b for FM=Co and c and d for FM=Fe in Figs.8.16-8.19.
The sensitive dependency of the MR on VIS makes it hard to find any correlations be-
tween favored impurity distributions and MR. Moreover, the lack of available data makes
it even harder. It then comes as no surprise that we failed to reveal such correlations.
Simply more data is needed.
8.3.3 1D vs. 2D
On comparing clean 1D systems with clean 2D systems, with identical configurations, we
notice the occurrence of discrepancies in the values of MR. Table 8.1 shows a comparison
of the MR values obtained with the static method (1D) with those obtained with the
dynamic method (2D). The discrepancies are due to the differences in implementation
MR values (%), FM=Co

Format, Config. S1 D1 S1 / D1 S2 D2 S2 / D2
NFIFN
{3, 5, 3} 8.3 5.7 1.5 4.6 6.1 0.75
NFNFIFNFN
{6, 5, 3, 5, 3, 5, 6} 47 8.6 5.5 51 7.2 7.1
NFNINFN
{5, 3, 4, 3, 5} 13 3.1 4.2 74 4.9 15.1
NFNFNINFNFN
{3, 6, 4, 2, 5, 2, 4, 6, 3} 30 4.9 6.1 34 11 3.1
FM=Fe
NFIFN
{5, 5, 5} 7.5 4.9 1.5 4.8 2.0 2.4
NFNFIFNFN
{6, 5, 3, 5, 3, 5, 6} 24 11 2.2 160 11 14.5
NFNINFN
{5, 3, 3, 3, 5} 5.6 5.5 1.0 8.1 6.4 1.3
NFNFNINFNFN
{4, 6, 4, 2, 3, 2, 4, 6, 4} 17 11 1.5 25 16 1.6
Table 8.1: The MR values of the 1D static (‘S’) method compared with the 2D dynamic (‘D’)
method. The addition of ‘1’ and ‘2’ to ‘S’ and ‘D’ in the heading means that the potential is
VIS = 0 in the former and either VIS = 0.5 EF (FM=Co) or VIS = 1.0 EF (FM=Fe) in the latter
case.
of the static method and the dynamic method. In general, the discrepancy in FM=Co
systems is more pronounced than the FM=Fe systems.
The first source of discrepancy is interference, or better, lack of interference due to
the way the static method is implemented. In this method the solutions are calculated
separately, solutions arising due to superposing a multiple of these separate solutions
is ignored. Hence, the particular interference caused by all of the solutions treated as
a whole is absent. This multiple wave-solution interference is however included in the
dynamic method in a natural way.
The second source of discrepancy lies in the way the dynamic method differs in set-up
from the static method. In the dynamic method, a wave-packet is set up at the LHS
of the system and given momentum to propagate to the RHS. Due to this asymmetry
the wave-packet is mostly reflected back and only a small (tunneling) fraction reaches
the region beyond the barrier. Consider the NFNINFN FM=Co format showing the
largest discrepancy between 1D and 2D results. The NM layers encompassing the IS
layer, serving as spin-dependent wave traps, are held responsible for the great MR values
seen in 1D systems. The reason that large MR values show up in 1D systems is that now
both traps have a more symmetric effect. This lack of symmetry in 2D systems we hold
responsible for the large calculated discrepancy.
To increase the trapping mechanism to full potential, with the dynamic method,
would possibly be to maintain a continuous flow of incoming wave-function flux of which
the transmitted and reflected fractions are continued to be absorbed at the terminals of
8.3. RESULTS 157
the system. In this way the traps will be able to, metaphorically speaking, fill themselves
up with probability-intensity. Then the waves within the trap will ricochet back and
forth at the exteriors of the trap. The mentioned modification in the dynamic method
is, in fact, a step closer to the actual situation. In reality, electrons are continuously
injected from external electrodes into the system and after a brief stay in the conductor
they are either transmitted or reflected into the same electrodes.
To determine the relative contribution of both sources to the observed discrepancy
would require to, ideally, ceiling off one source while opening the remaining source com-
pletely. The format which lends itself to that purpose is the NFIFN format, and to
lesser extent the NFNFIFNFN format, where the trapping source is absent in the former
case and hardly present in the latter. Unfortunately, a lack of available data prevents us
from making substantial estimations. From the limited data, the least we can state is
that interference seems to have a decreasing impact on the MR. Needless to say though,
simply more data on more formats is needed.
8.3.4 Conclusions and Open Issues

The results seem to imply the following conclusions.
Concerning 1D systems:
1. For three out of four formats, we have verified the oscillatory behavior in the
evolution of the MR with its layer widths, provided the layers have a non-insulating
character. The one format which does not develop oscillatory behavior within the
considered FM thickness range is the NFNINFN FM=Co format1 . This format is
responsible for the formation of well pronounced high-valued maxima in MR, this
makes the format very eligible for application ends.
2. The effect of increasing the barrier width has a generally reducing effect on the
MR, though it is less clear-cut than naively expected from the elementary FIF
format. The exception again is provided by the NFNINFN, FM=Co format where
the systems react with a strong increase in MR on increasing the width of the IS
layer.
3. Increasing the barrier potential VIS turns out to have a strong format dependent
and FM-type dependent effect on the MR. The response of the MR on increasing
VIS is not at all unconditionally non-decreasing as the results on the FIF format
suggest.
Concerning 2D systems:
1. The general tendency of hybrid formats, in agreement with expectations based on

the FIF format, is to show a decay of MR with the impurity density. A possible
exception to this general tendency is perhaps best given by the NFNFNINFNFN
FM=Co format, where, for the given configuration, the average MR even devel-
ops a maximum as a function of impurity concentration. From the point of view
1 It is, of course, possible that the periods of oscillation fall well out of the considered range.
of technological applications, where the MR levels need to be conserved, this for-

mat might be the best candidate. We stress, however, that these are preliminary
findings, simply more data is needed before drawing any general conclusions.
2. The erratic character of the evolution of the MR seen in individual systems and
its strong dependence on VIS makes it hard, to say the least, to find correlations
between certain properties of impurity distributions and the value of MR.
3. The differences in the calculated MR values obtained with the 1D static method
and the 2D dynamic method suggest two sources of discrepancy. The dynamic
method includes multi-wave interference-effects but does not properly reflect the
mentioned spin trapping effect (see subsection 8.2.1 and the discussion in subsection
8.3.3). On this matter, the static method is precisely its antagonist.
We end this chapter, and hence this thesis, by giving a list of issues which have not
been addressed yet.
1. Due to the diversity of the concept of hybrid formats many formats, such as double
(triple etc.) barrier systems, and their properties remain unexplored.
2. The results obtained on impure 2D systems is insufficient to expose any general
behavior, except in a confirming way. The general behavior may be extracted only
after many equivalent systems have been analyzed.
3. The angular behavior of the MR was not treated here. This, however, can be done
without further modification of the model in a way similar to the angular behavior
of GMR systems.
4. A feature not included in the model is the provision of the FM impurities with
a paramagnetic flavor. As was shown in chapter 6 this has a destructive effect
on the MR. Applied to the hybrid formats, the GMR effect on the MR will shift
from constructive to destructive and hence the MR will drop even faster with the
impurity concentration.
5. We are aware of the necessity to have an anti-ferromagnetic coupling between the
FM electrodes in the here assumed multi-layer type of TMR, but was disregarded
nonetheless. Respecting this condition will put severe restrictions on the allowed
distances between the FM layers, and hence possible configurations. A way out to
this problem is provided by the alternative of spin-valve type of MR (see chapter
7) where the FM coupling is less important and can even be irrelevant. A technical
prerequisite (which might be difficult?) then is to properly parameterize the band
parameters of the different FM layers.
6. In order to give a better representation of the spin-trapping effect we should modify
the dynamic model. To acknowledge the continuous flow of emitted current into the
system we might continuously initialize new waves at the emitting side of the system
throughout the simulation time. This modified model might also successfully be
applied to CIP-GMR systems (chapter 6).
8.3. RESULTS 159
0.1 0.11
−γ −γ0.09
0.09 0.1
{., 3, .} {., 3, .}
0.08 VIS = 0.1 EF {., 6, .} VIS = 0.25 EF {., 6, .}
0.07 {., 9, .} 0.08
{., 9, .}
0.07
0.06
0.06
0.05
0.05
0.04
0.04
0.03
0.03
0.02 0.02
0.01 0.01
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
a Th. FM (# atoms) e Th. FM (# atoms)
0.08 0.11
{., 3, .}
0.07
VIS = 0.2 EF {., 6, .} −γ0.09
0.1
VIS = 0.5 EF
−γ {., 9, .}
0.06 0.08 {., 3, .}
{., 4, .}
0.05 0.07 {., 6, .}
{., 9, .}
0.06
0.04
0.05
0.03 0.04
0.02 0.03
0.02
0.01
0.01
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
b Th. FM (# atoms) f Th. FM (# atoms)
0.08 0.12
{., 3, .}
VIS = 0.3 EF {., 4, .} VIS = 0.75 EF
0.07
−γ0.06 {., 6, .} 0.1
{., 3, .}
{., 9, .} −γ {., 4, .}
{., 6, .}
0.05
0.08
{., 9, .}
0.04 0.06
0.03
0.04
0.02
0.02
0.01
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
c Th. FM (# atoms) g Th. FM (# atoms)
0.07 0.25
{., 3, .}
0.06 VIS = 1.0 EF
−γ
0.2
{., 4, .} VIS = 0.5 EF −γ {., 3, .}

{., 4, .}
0.05
0.15
{., 6, .}
0.04 {., 9, .}
{., 6, .}
0.1
0.03
0.05
0.02
{., 9, .}
0.01 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
d Th. FM (# atoms) h Th. FM (# atoms)
Figure 8.9: The MR, of the format NFIFN, as a function of thicknesses of the outer FM layers
(Th. FM) for several barrier-potentials (the thicknesses are varied simultaneously). In a-d the
FM layers are Co-like, in e-h Fe-like.
0.18 0.6
−γ
0.16
0.5 VIS = 0.1 EF
0.14 −γ {8, 5, ., 4, ., 5, 8}
{8, 5, ., 5, ., 5, 8}
0.12
0.4
{9, 5, ., 4, ., 5, 9}
{9, 5, ., 5, ., 5, 9}
0.1 0.3
0.08 0.2
0.06
0.1
0.04 {., 5, 8, 4, 8, 5, .}
0.02 VIS = 0.1 EF {., 5, 8, 5, 8, 5, .}
{., 5, 9, 4, 9, 5, .}
0
{., 5, 9, 5, 9, 5, .}
0 -0.1
a 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20
e 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20
0.18 0.6
0.16 VIS = 0.2 EF

−γ 0.5
0.14 −γ
0.4
0.12
0.1 0.3
0.08 0.2
0.06
0.1
0.04
VIS = 0.2 EF
0
0.02
0 -0.1
b 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20
f 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20
0.18 0.6
0.16
−γ 0.14 0.5 VIS = 0.3 EF
0.12
−γ
0.4
0.1
0.08 0.3
0.06
0.2
0.04
0.02
0.1
0 VIS = 0.3 EF
-0.02 0
c 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20
g 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20
0.2 0.6
0.18
−γ0.16 0.5 VIS = 0.5 EF
0.14
−γ
0.4
0.12 VIS = 0.5 EF
0.1 0.3
0.08
0.2
0.06
0.04
0.1
0.02
0 0
d 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20
h 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20
Figure 8.10: The MR, of the format NFNFIFNFN, as a function of thicknesses of two of the
four FM layers for several barrier-potentials. The FM layers are all Co-like.
8.3. RESULTS 161
0.22 0.4
0.2 VIS = 0.25 EF VIS = 0.25 EF
−γ0.18 0.35
−γ {8, 5, ., 4, ., 5, 8}
0.16 0.3 {8, 5, ., 5, ., 5, 8}
{9, 5, ., 4, ., 5, 9}
0.14 0.25 {9, 5, ., 5, ., 5, 9}
0.12
0.2
0.1
0.08 0.15
0.06
{., 5, 8, 4, 8, 5, .} 0.1
0.04 {., 5, 8, 5, 8, 5, .}
{., 5, 9, 4, 9, 5, .} 0.05
0.02 {., 5, 9, 5, 9, 5, .}
0 0
a 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20 e 0 2 4 6 8 10 12 14 16 18
Th. FM (# atoms)
20
0.25 0.45
VIS = 0.5 EF
−γ
0.4 VIS = 0.5 EF
0.2
0.35
−γ
0.15 0.3
0.25
0.1 0.2
0.15
0.05
0.1
0 0.05
b 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20 f 0 2 4 6 8 10 12 14 16 18
Th. FM (# atoms)
20
0.3 1
VIS = 0.75 EF 0.9
VIS = 0.75 EF
0.25 −γ0.8
−γ 0.7
0.2
0.6
0.15 0.5
0.4
0.1
0.3
0.2
0.05
0.1
0 0
c 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20 g 0 2 4 6 8 10 12 14 16 18
Th. FM (# atoms)
20
0.3 1.4
VIS = 1.0 EF
0.25 1.2 VIS = 1.0 EF
−γ
−γ 1
0.2
0.8
0.15
0.6
0.1
0.4
0.05 0.2
0 0
d 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20 h 0 2 4 6 8 10 12 14 16 18
Th. FM (# atoms)
20
Figure 8.11: The MR, of the format NFNFIFNFN, as a function of thicknesses of two of the
four FM layers for several barrier-potentials. The FM layers are all Fe-like.
0.25 0.075
0.07
VIS = 0.1 EF 0.065 VIS = 0.25 EF

0.2
−γ 0.06
−γ {., 2, 2, 2, .}
0.055 {., 2, 3, 2, .}
0.15
{., 3, 3, 3, .} 0.05
{., 2, 4, 2, .}
{., 3, 4, 3, .}
{., 3, 5, 3, .} 0.045
0.1 0.04
0.035
0.05 0.03
0.025
0 0.02
a 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20 e 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20
0.45 0.08
0.4
VIS = 0.2 EF 0.07 VIS = 0.5 EF
0.35 −γ
−γ 0.06
0.3
0.25
0.05
0.2
0.15 0.04
0.1
0.03
0.05
0 0.02
b 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20 f 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20
0.7 0.09
0.6 VIS = 0.3 EF −γ

0.08
VIS = 0.75 EF
−γ 0.07
0.5
0.4 0.06
0.3 0.05
0.2 0.04
0.1 0.03
0 0.02
c 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20 g 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20
2 0.09
1.8
VIS = 0.5 EF 0.08
VIS = 1.0 EF
1.6 −γ
−γ1.4 0.07
1.2 0.06
1
0.05
0.8
0.6 0.04
0.4
0.03
0.2
0 0.02
d 0 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20 h 2 4 6 8 10 12 14 16
Th. FM (# atoms)
18 20
Figure 8.12: The MR, of the format NFNINFN, as a function of thicknesses of the outer FM
layers for several barrier-potentials. In a-d the FM layers are Co-like, in e-h Fe-like.
8.3. RESULTS 163
0.22 0.3
{., 6, 8, 2, 4, 2, 8, 6, .}
−γ 0.2
VIS = 0.1 EF {., 6, 8, 2, 5, 2, 8, 6, .}
{., 6, 9, 2, 4, 2, 9, 6, .} −γ VIS = 0.1 EF {8, 6, ., 2, 4, 2, ., 6, 8}
{8, 6, ., 2, 5, 2, ., 6, 8}
0.18 {., 6, 9, 2, 5, 2, 9, 6, .} 0.25 {9, 6, ., 2, 4, 2, ., 6, 9}
{9, 6, ., 2, 5, 2, ., 6, 9}
0.16
0.2
0.14
0.12
0.15
0.1
0.08
0.1
0.06
0.04 0.05
0.02
0 -0.05
0
a 0 5 10 15 20
Th. FM (# atoms)
25 30 e 0 2 4 6 8 10 12 14
Th. FM (# atoms)
16 18 20
0.22 0.3
VIS = 0.2 EF −γ
−γ
0.2
0.25
0.18
0.16
0.2
0.14
0.12 0.15
0.1
0.1
0.08
0.06
0.05
0.04 VIS = 0.2 EF
b 0.02 0 5 10 15 20
Th. FM (# atoms)
25 30 f 0
0 2 4 6 8 10 12 14
Th. FM (# atoms)
16 18 20
0.24 0.6
0.22 VIS = 0.3 EF −γ
−γ 0.2 0.5
VIS = 0.3 EF
0.18
0.4
0.16
0.14 0.3
0.12
0.2
0.1
0.08
0.1
0.06
-0.05
c 0.04 0 5 10 15 20
Th. FM (# atoms)
25 30 g 0
0 2 4 6 8 10 12 14
Th. FM (# atoms)
16 18 20
0.45 1
VIS = 0.5 EF 0.9
VIS = 0.5 EF
−γ 0.4
−γ 0.8
0.35 0.7
0.3 0.6
0.5
0.25
0.4
0.2 0.3
0.15 0.2
0.1
0.1
0
0.05 -0.1
d 0 5 10 15 20
Th. FM (# atoms)
25 30 h 0 2 4 6 8 10 12 14
Th. FM (# atoms)
16 18 20
Figure 8.13: The MR, of the format NFNFNINFNFN, as a function of thicknesses of two of
the four FM layers for several barrier-potentials. The FM layers are all Co-like.
0.16 0.25
{., 6, 8, 2, 4, 2, 8, 6, .}
0.14 {., 6, 8, 2, 5, 2, 8, 6, .} VIS = 0.25 EF VIS = 0.25 EF
{., 6, 9, 2, 4, 2, 9, 6, .}
0.12 {., 6, 9, 2, 5, 2, 9, 6, .} 0.2
−γ 0.1 −γ
0.15
0.08
0.06
0.1
0.04
0.02
0.05
0 {8, 6, ., 2, 4, 2, ., 6, 8}
{8, 6, ., 2, 5, 2, ., 6, 8}
-0.02 0 {9, 6, ., 2, 4, 2, ., 6, 9}
-0.04 {9, 6, ., 2, 5, 2, ., 6, 9}
a -0.06 0 2 4 6 8 10 12 14 16 18
Th. FM (# atoms)
20 e-0.05 0 2 4 6 8 10 12 14 16 18
Th. FM (# atoms)
20
0.16 0.25
0.14 VIS = 0.5 EF 0.2

VIS = 0.5 EF
0.12 −γ
−γ
0.1 0.15
0.08
0.1
0.06
0.04 0.05
0.02
0
0
b -0.02 0 2 4 6 8 10 12 14 16 18
Th. FM (# atoms)
20 f -0.05 0 2 4 6 8 10 12 14 16 18
Th. FM (# atoms)
20
0.16 0.25
0.14 VIS = 0.75 EF VIS = 0.75 EF

0.12 0.2
−γ −γ
0.1
0.15
0.08
0.06
0.1
0.04
0.02 0.05
0
c -0.02 0 2 4 6 8 10 12 14 16 18
Th. FM (# atoms)
20 g 0
0 2 4 6 8 10 12 14 16 18
Th. FM (# atoms)
20
0.14 0.3
0.12
VIS = 1.0 EF VIS = 1.0 EF
−γ
0.25
0.1 −γ
0.2
0.08
0.15
0.06
0.1
0.04
0.02 0.05
0 0
d 0 2 4 6 8 10 12 14 16 18
Th. FM (# atoms)
20 h 0 2 4 6 8 10 12 14 16 18
Th. FM (# atoms)
20
Figure 8.14: The MR, of the format NFNFNINFNFN, as a function of thicknesses of two of
the four FM layers for several barrier-potentials. The FM layers are all Fe-like.
8.3. RESULTS 165
0.09 0.22
{3, 3, 3}
0.085 NFIFN {3, 4, 3} 0.2 {3, 3, 3}
{3, 6, 3} {14, 3, 14}
{3, 5, 3}
−γ0.08 [2-30, 3-6, 2-30] {10, 3, 10}
{19, 3, 19} −γ0.16
0.18 {23, 3, 23}
{25, 3, 25}
0.075
{5, 3, 5} {14, 4, 14}
{4, 3, 4}
0.07 0.14 NFIFN {13, 3, 13}
0.065 0.12 [2-30, 3-6, 2-30]
0.06 0.1
0.055 0.08
0.05 0.06
0.045 0.04
0.04 0.02
a 0 0.1 0.2
VIS (EF)0.4
0.3 0.5 e 1 1.2 1.4 1.6
VIS (EF)1.8 2
0.8 0.085
0.7
−γ {5, 3, 4, 3, 5}
{6, 3, 5, 3, 6}
0.08
−γ0.075
0.6 {3, 4, 4, 4, 3}
{3, 5, 4, 5, 3}
0.5 {3, 5, 5, 5, 3} {4, 2, 2, 2, 4}
{10, 5, 5, 5, 10} {4, 2, 3, 2, 4}
{8, 5, 4, 5, 8} 0.07 {5, 3, 3, 3, 5}
0.4 {5, 4, 2, 4, 5}
{5, 3, 2, 3, 5}
0.065 {4, 4, 2, 4, 4}
0.3 {5, 4, 3, 4, 5}
0.2 NFNINFN 0.06
NFNINFN
[2-20, 3-5, 3-5, 3-5, 2-20] 0.055
0.1 [2-20, 2-4, 2-4, 2-4, 2-20]
0 0.05
b 0 0.1 0.2 0.3
VIS (EF)0.4 0.5 f 0 0.2 0.4 0.6
VIS (EF0.8
) 1
0.6 2.2
0.55 −γ 2 NFNFIFNFN
−γ 1.8
[3-7, 4-6, 3-7, 4-5, 3-7, 4-6, 3-7]
1.6
0.5 {5, 6, 3, 5, 3, 6, 5}
1.4 {6, 6, 3, 5, 3, 6, 6}
{5, 4, 3, 5, 3, 4, 5}
0.45 1.2 {5, 4, 3, 4, 3, 4, 5}
NFNFIFNFN {5, 6, 3, 4, 3, 6, 5}
1 {3, 4, 5, 5, 5, 4, 3}
0.4
[3-7, 4-6, 3-7, 4-5, 3-7, 4-6, 3-7] {3, 4, 5, 4, 5, 4, 3}
0.8
{6, 5, 3, 5, 3, 5, 6}
{6, 5, 3, 5, 3, 5, 4}2× 0.6
0.35 {6, 5, 3, 4, 3, 5, 6}
{3, 5, 3, 5, 3, 5, 3}
{3, 5, 3, 5, 3, 5, 6}2× 0.4
0.3 0.2
c 0 0.1 0.2 0.3
VIS (EF)0.4 0.5 g 0 0.2 0.4 0.6
VIS (EF)0.8 1
0.36 0.4
{6, 6, 3, 2, 5, 2, 3, 6, 6}
{6, 6, 3, 2, 4, 2, 3, 6, 6}
0.34 {6, 6, 3, 2, 3, 2, 3, 6, 6}
−γ 0.35 {6, 6, 3, 2, 5, 2, 3, 6, 2}2×
0.32 −γ {4, 6, 4, 2, 3, 2, 4, 6, 4}
{5, 6, 4, 2, 5, 2, 4, 6, 5}
0.3
0.3
0.28 NFNFNINFNFN 0.25

[2-6, 6, 2-6, 2, 2-5, 2, 2-6, 6, 2-6]
0.26
{3, 6, 4, 2, 5, 2, 4, 6, 3} 0.2
{3, 6, 4, 2, 4, 2, 4, 6, 3}
0.24 {3, 6, 4, 2, 3, 2, 4, 6, 3}
{6, 6, 3, 2, 3, 2, 3, 6, 6} NFNFNINFNFN
{6, 6, 4, 2, 3, 2, 4, 6, 6} 0.15
0.22 {6, 6, 3, 2, 2, 2, 3, 6, 6} [2-6, 6, 2-6, 2, 2-5, 2, 2-6, 6, 2-6]
{6, 6, 4, 2, 4, 2, 4, 6, 6}
0.2 0.1
d 0 0.1 0.2 0.3
VIS (EF)0.4 0.5 h 0 0.2 0.4 0.6
VIS (EF)0.8 1
Figure 8.15: The MR as a function of the barrier height VIS . Shown here are the top-7 (highest
average MR) members from each given thickness spectrum of the corresponding format. In a-d
we have FM=Co, in e-h FM=Fe.
0.12 0.06
0.1
VIS = 0.0 EF
−γ 0.04
0.08 −γ
0.02
0.06
0.04 0
0.02
-0.02
0
VIS = 0.0 EF
-0.04
-0.02
a -0.04 0 2 4 6
c (%) 8 10 c -0.06 0 2 4 6
c (%) 8 10
0.08 0.5
−γ
0.06
0.4 VIS = 1.0 EF
0.04 −γ
0.3
0.02
0 0.2
-0.02
0.1
-0.04 VIS = 0.5 EF
0
-0.06
b -0.08 0 2 4 6
c (%) 8 10 d -0.1 0 2 4 6
c (%) 8 10
Figure 8.16: The calculated annealing MR behavior for 2D systems of the format NFIFN.
In a&b the FM layers are Co-like, configuration: {3, 5, 3}. In c&d the FM layers are Fe-like,
configuration:{5, 5, 5}. Transverse dimension: Ly = 40 atoms, spreading: σ = 3 atoms.
8.3. RESULTS 167
0.15 0.2
0.15
0.1 −γ0.1
−γ
0.05 0.05
0
0 -0.05
-0.05
-0.1
VIS = 0.0 EF
VIS = 0.0 EF -0.15
-0.1 -0.2
a 0 2 4 6
c (%) 8 10 c 0 2 4 6
c (%) 8 10
0.25 0.4
0.2 0.35
−γ VIS = 0.5 EF
−γ0.25
0.3
0.15
0.1 VIS = 1.0 EF

0.2
0.05
0.15
0
0.1
-0.05
0.05
-0.1 0
-0.15 -0.05
b 0 2 4 6
c (%) 8 10 d 0 2 4 6
c (%) 8 10
Figure 8.17: The calculated annealing MR behavior for 2D systems of the format NFNFIFNFN.
In a&b the FM layers are Co-like, configuration: {5, 6, 3, 5, 3, 6, 5}. In c&d the FM layers are
Fe-like, configuration:{5, 6, 3, 5, 3, 6, 5}. Transverse dimension: Ly = 20 atoms, spreading: σ = 3
atoms.
0.04 0.08
VIS = 0.0 EF
−γ
0.03
−γ
0.06
0.02 0.04
0.01 0.02
0 0
VIS = 0.0 EF
-0.01 -0.02
a-0.02 0 2 4 6
c (%) 8 10 c -0.04 0 2 4 6
c (%)8 10
0.07 0.08
0.06 0.07
−γ0.05 VIS = 0.5 EF VIS = 1.0 EF

−γ0.05
0.06
0.04 0.04
0.03 0.03
0.02 0.02
0.01
0.01
0
0 -0.01
b-0.01 0 2 4 6
c (%) 8 10 d -0.02 0 2 4 6
c (%)8 10
Figure 8.18: The calculated annealing MR behavior for 2D systems of the format NFNINFN.
In a&b the FM layers are Co-like, configuration: {5, 3, 4, 3, 5}. In c&d the FM layers are Fe-like,
configuration:{5, 3, 3, 3, 5}. Transverse dimension: Ly = 20 atoms, spreading: σ = 1 atoms.
8.3. RESULTS 169
0.12 0.12
−γ0.08
0.1
−γ0.08
0.1
0.06
0.06
0.04
0.02 0.04
0 0.02
-0.02
0
-0.04
VIS = 0.0 EF -0.02 VIS = 0.0 EF
-0.06
-0.08 -0.04
-0.1
a 0 2 4 6
c (%) 8 10 c -0.06 0 2 4 6
c (%) 8 10
0.6 0.2
0.5 −γ
−γ VIS = 0.5 EF 0.15 VIS = 1.0 EF
0.4
0.1
0.3
0.2
0.05
0.1
0
0
b -0.1 0 2 4 6
c (%) 8 10 d -0.05 0 2 4 6
c (%) 8 10
Figure 8.19: The calculated annealing MR behavior for 2D systems of the format NFNFN-
INFNFN. In a&b the FM layers are Co-like, configuration: {3, 6, 4, 2, 5, 2, 4, 6, 3}. In c&d the
FM layers are Fe-like, configuration:{4, 6, 4, 2, 3, 2, 4, 6, 4}. Transverse dimension: Ly = 20 atoms,
spreading: σ = 1 atoms.

Tunneling Magnetoresistance

Uploaded by

Copyright:

Available Formats

Tunneling Magnetoresistance

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Tunneling Magnetoresistance

Uploaded by

Copyright:

Available Formats

Chapter 8

a mIS = mFM b mIS < mFM

8.1.1 Discovery of TMR and Early Models

This property allows one to have an anti-parallel conﬁguration of the magnetizations of

Julliére Model versus Slonczewski Model

gF ∝ D1↑ D2↑ + D1↓ D2↓ , (8.1)

gF − gAF D ↑ D ↑ + D1↓ D2↓ − D1↑ D2↓ − D1↓ D2↑

D1↑ = 0 and D2↑ = 0, (8.8)

g(θ) = gfbf (1 + Pfb

8.2.1 Hybrid Formats

1 The format is thought to be enclosed by NM electrodes.

spin−dependent ‘traps’ less efficient

8.2.2 Basic Format

FM atom IS atom NM atom

8.2.5 Algorithms of Study

1. The height of the IS barriers is set at Fermi level (VIS = 0).

4. The MR is calculated and recorded.

1. The height of the IS barriers is set at Fermi level.

3. The MR is calculated and recorded.

5. The resulting spectrum of MRs is sorted from high to low.

is taken over all such values.

new since the plots remain ﬂat.

tremendous elevations in MR for NFNINFN conﬁgurations as compared with the NFIFN.

Calculations on 2D systems reveal similar behavior.

MR values (%), FM=Co

8.3.4 Conclusions and Open Issues

1. The general tendency of hybrid formats, in agreement with expectations based on

of technological applications, where the MR levels need to be conserved, this for-

{., 4, .} VIS = 0.5 EF −γ {., 3, .}

0.16 VIS = 0.2 EF

VIS = 0.1 EF 0.065 VIS = 0.25 EF

0.6 VIS = 0.3 EF −γ

0.14 VIS = 0.5 EF 0.2

0.14 VIS = 0.75 EF VIS = 0.75 EF

0.28 NFNFNINFNFN 0.25

0.1 VIS = 1.0 EF

−γ0.05 VIS = 0.5 EF VIS = 1.0 EF

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.