. MLIBRAEYI
, ic igan ta e
L008 University

This is to certify that the
dissertation entitled

STUDIES OF CURRENT INDUCED MAGNETIZATION REVERSAL
AND GENERATION OF GHz RADIATION IN MAGNETIC
NANOPILLARS.

presented by

Mustafa Yousef AlHajDanNish

has been accepted towards fulfillment
of the requirements for the

PhD. degree in Physics

 

 

L Ma 43%,

-Major P f'é’ssjor’s Signature

Date

MSU is an affirmative-action, equal-opportunity employer

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K‘lProyAcc8PresICIRCIDateDue indd

 

STUDIES OF CURRENT INDUCED MAGNETIZATION REVERSAL AND
GENERATION OF GHz RADIATION IN MAGNETIC NANOPILLARS.

By

Mustafa Yousef AlHajDarwish

A DISSERTATION
Submitted to
Michigan State University
in Partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY

Department of Physics and Astronomy

2008

ABSTRACT

STUDIES OF CURRENT INDUCED MAGNETIZATION REVERSAL AND
GENERATION OF GHZ RADIATION IN MAGNETIC NANOPILLARS.

By

Mustafa Yousef AlHajDarwish

This thesis describes studies of two phenomena: Current-Induced Magnetization
Switching (CIMS), and Current-Induced Generation of GHz Radiation.

The CIMS part contains results of measurements of current-perpendicular-to-plane
(CPP) magnetoresistance (MR) and CIMS behavior on Ferromagnetic/Non-
metal/Ferromagnetic (F l/N/FZ) nanopillars. Judicious combinations of F1 and F2 metals
with different bulk scattering asymmetries, and with F UN and N/F 2 interfaces having
different interfacial scattering asymmetries, are shown to be able to controllably, and
independently, ‘invert’ both the CPP-MR and the CIMS. In ‘normal’ CPP-MR, R(AP) >
R(P), where R(AP) and R(P) are the nanopillar resistances for the anti-parallel (AP) and
parallel (P) orientations of the F1 and F2 magnetic moments. In ‘inverse’ CPP-MR, R(P)
> R(AP). In ‘normal’ CIMS, positive current switches the nanopillar from the P to the
AP state. In ‘inverse’ CIMS, positive current switches the nanopillar from AP to P. All
four possible combinations of CPP-MR and CIMS—(a) ‘normal’-‘normal’, (b) ‘normal’-
‘inverse’, ‘inverse’-‘normal’, and (d) ‘inverse’-‘inverse’ are shown and explained.
These results rule out the self-Oersted field as the switching source, since the direction of
that field is independent of the bulk or interfacial scattering asymmetries. Successful use

of impurities to reverse the bulk scattering asymmetry shows the importance of scattering

off of impurities within the bulk F l and F2 metals—Le. that the transport must be treated
as ‘diffiisive’ rather than ‘ballistic’.

The GHz studies consist of five parts: (1) designing a sample geometry that allows
reliable measurements; (2) making nanopillar samples with this geometry; (3)
constructing a system for measuring frequencies up to 12 GHz and measuring current-
driven GHz radiation data with it; (4) showing ‘scaling’ behavior of GHz data with the
critical fields and currents for nominally identical (but actually slightly different)
samples, and justifying such scaling; and (5) designing and constructing a system for

frequency domain studies up to 40 GHz and for time domain studies.

DEDICATION

To the family of Mustafa AlHajDarwish.

iv

ACKNOWLEDGMENTS

While completing the research program for his Ph.D. degree, Mustafa AlHajDarwish
was stricken with cancer of the knee. This cancer led to replacement of his knee, then
removal of his leg, and, finally, his untimely death. The continuing pain, plus debilitating
effects of the operations, and extended chemotherapy before, between, and after the
operations, precluded his writing his PhD thesis before his death.

His PhD guidance committee, considering his outstanding performance in formal
classwork (grades all 4.0/4.0) and on subject examinations (three 4.03 and one 3.5), as
well as his high quality research production, concluded that he merited a PhD degree.
This research had led to 8 publications, two as first author, and included also design and
construction of a new sample geometry and new measuring systems for studying dc
current-induced microwave radiation, as well as obtaining new data with such samples.

Mustafa’s advisor, Jack Bass, agreed to construct a PhD thesis by tying together what
Mustafa had produced, with minimal connective material, and writing a brief abstract and
a summary and conclusions. He was assisted by William P. Pratt Jr., who helped to
supervise Mustafa’s work, and by Amit Sharma, a PhD student in the process of writing
his own thesis. Helpful comments were received from Norman 0. Birge, S.D. (Bhanu)
Mahanti, and members of Mustafa’s guidance committee.

Mustafa was such an exceptional student, productive researcher, and wonderful
person, that all who worked with him and assisted in the awarding of this PhD degree feel
honored to have done so.

Thanks are given to the US National Science Foundation and to Seagate Technology

Corporation for funding Mustafa and his research.

TABLE OF CONTENTS

List of Tables ....................................................................... vii
List of Figures ..................................................................... viii
1. Introduction ....................................................................... l
1.1. Background and Overview” l
1.2 Publications ofMustafa AlHaj Darwish 5

2. Inverted current-driven switching in Fe(Cr)/Cr/Fe(Cr)
nanopillars ......................................................................... 6
2.1 Background ................................................................... 6
2.2. Inverted current-driven switching in Fe(Cr)/Cr/Fe(Cr) nanopillars. . . . . .....10

3. Controlled normal and inverse current-induced magnetization
switching and magnetoresistance in magnetic nanopillars. . .....l7

4. Spin-Transfer—Torque Generation of High
Frequency (GHz) Radiation in Nanopillars: Sample Geometry,

Measuring Systems, and Sample Data .................................. 31
4.1 . Background .................................................................... 31
4.2. Sample Geometry .............................................................. 31
4.3. Measuring System to 12 GHz, Expected Behaviors, and Data ................ 34

4.3.1. The Measuring System to 12 GHz. .. ............................................ 34
4.3.2. Expected Dynamical Motions. . . .................................................. 35
4.3.3. Experimental Data and an Example of Scaling. .. ........................... 37

4.4. System for measuring GHz radiation up to 40 GHz and time
domain studies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5. Summary and Conclusions. .......................................................... 47

Bibliography. . . ..................................................................................... 49

vi

List of Tables

Table 3.1: F 1/N/F2 for each figure, giving: the spin anisotropies (+ = positive,
- = negative) of F1, FUN, and their net anisotropy F1(Net), those of F2,
N/F 2, and their net anisotropy F2(Net), and the signs of the observed MR

and CIMS (+ = normal and - = inverse) .................................. 25

vii

Fig. 2.1.

Fig. 2.2.

Fig. 2.3.

Fig. 2.4.

Fig. 3.1.

Fig. 3.2.

Fig. 3.3.

Fig. 3.4.

Fig. 3.5.

Fig. 4.1

Fig. 4.2.

List of Figures.

As an addition for this thesis, we show data sets for Co/Au/Co at both
295K and 4.2K, measured by Mustafa AlHajDarwish, that could have been
used as examples of ‘normal’ switching in place of Fig. 2.2 for Py/Cu/Py ........ 9

Py/Cu/Py data at 295K and 4.2K showing normal switching for dV/dI
vs I at H = 50 Oe (main figures) and also for dV/dI vs H at I = O
(insets). (From Urazhdin et al.26) ...................................................................... 15

Fe(Cr)/Cr/Fe(Cr) data at 295K and 4.2K showing inverted switching for
dV/dI vs I at H = 0 (main figures) but normal dV/dI vs H at I = O (insets) ..... 15

Fe(Cr)/Cr/Fe(Cr) data at 4.2K showing: (a) normal MR for dV/dI vs H
at various I, but (b) inverted switching for dV/dI vs I at various H. Curves
in (a) for I at 0 and in (b) for H at 0 are shified vertically for clarity .................. 16

Py(24)/Cu(10)/Py(6) data at 295K (top) and 4.2K (bottom) showing normal
MR (dV/dI vs H at I = 0) in the insets and normal CIMS for dV/dI vs I

in the main figures at H = 0 0e for 295K and at H = 20 Oe for 4.2K.

In all figures, I > 0 is always from F1 to F2 ...................................................... 26

Fe(Cr)(30)/Cr(6)/Fe(Cr)(3.5) data at 295K (top) and 4.2K (bottom) showing
normal MR (dV/dI vs H at I = O) in the insets but inverse CIMS for dV/dl vs
I at H = O in the main figures ............................................................................ 27

Py(20)/Cu(7)/Cr(3)/Fe(Cr)(3) data at 295K (top) and 4.2K (bottom)
showing inverse MR (dV/dI vs H at I = O) in the insets but normal
CIMS for dV/dI vs I at H = O in the main figures ............................................ 28

Ni(Cr)(20)/Cu(20)/Py(10) data at 295K (top) and 4.2K (bottom)
showing inverse MR (dV/dI vs H at I = O) in the insets and inverse
CIMS for dV/dI vs I at H = 0 in the main figures ............................................ 29

Py(24)/Cu(lO)/Ni(Cr)(4) data at 4.2K showing inverse MR (dV/dI vs H
at I = 0) in the inset and normal CIMS for dV/dI at H = 0 0e in the
main figure ....................................................................................................... 30

Sample configuration for GHz measurements. Each Si substrate contains

6 samples (lower left). Each sample has four Au contacts (lower right).

Two of these Au contacts are grounds and two constitute 50 Q strip lines
(upper figure). Each sample can be contacted by a picoprobe coaxial cable
pair in two ways as illustrated at the top and bottom of the upper figure ........ 33

Schematic drawings of a nanopillar (left), a sample with pico-probes
(middle), and the system for measuring GHz radiation up to 12 GHz ........... 34

viii

Fig. 4.3. The x,y.z. scale, and the direction of application of the magnetic field
(top left). Three different monodomain motions of the free layer
magnetization (assumed monodomain) (top middle and right).
Frequencies expected for different applied magnetic fields (lower right) ...... 35

Fig. 4.4. Phase diagram as functions of applied magnetic field H (ordinate) and
current I (abscissa) expected from monodomain analysis ............................... 36

Fig. 4.5. GHz radiation data from sample #2b5. Top: Examples of experimental
‘phase diagrams’ showing MR (upper left) and integrated power (upper
right) as functions of H and 1. Bottom: Examples of GHz peaks for fixed
H = - 700 Oe and variable I (lower lefi) and for fixed I = 17 mA and
variable H (lower right) .................................................................................. 38

Fig. 4.6. GHz radiation data from sample #2b3. Top: Examples of experimental
‘phase diagrams’ showing MR (upper left) and integrated power (upper
right) as functions of H and 1. Bottom: Examples of GHz peaks for
fixed H = - 560 Oe and variable I (lower left) and for fixed I = 14 mA

and variable H (lower right) ........................................................................... 39
Fig. 4.7. Power versus frequency plots comparing spectra for sample #1 and #2

at normalized values of H/Hc and I/AIC ......................................................... 40
Fig. 4.8. Schematic of the system for studying current-generated signals to

40 GHz ........................................................................................................... 41
Fig. 4.9. Details of the system for both frequency studies to 40 GHz and time

domain studies ............................................................................................... 43
Fig. 4.10. DC pulse generator ...................................................................................... 44
Fig. 4.11. Microwave pulse generator ......................................................................... 45

Images in this dissertation are presented in color.

ix

Chapter 1. Introduction

1.1. Background and Overview.

Mustafa AlHajDarwish came to Michigan State University (MSU) in 2000 after
obtaining a Masters degree in Physics at Yarmouk University in Jordan. He came highly
recommended, and his academic performance at MSU was outstanding: straight 40/40
in his coursework, and three 405 and one 3.5 on his comprehensive examinations. He
began his Ph.D. research in 2001, initially contributing to two projects involving
measurements of Current-Perpendicular-to-Plane (CPP) magnetoresistance (MR) of
magnetic multilayers containing combinations of ferromagnetic (F) and Non-magnetic
(N) metals. The first project involved a search for claimed mean-free-path effects in
magnetic multilayers, and concluded that any such effects were not nearly as significant
as had been claimed. The second project, a study of CPP-MR properties of Ru and
Co/Ru interfaces, provided the first quantitative information about CPP-MR properties of
a metal, Ru, that was beginning to be used to make artificial antiferromagnets for use in
CPP-devices, and of the interfaces of Ru with the most widely used ‘pure F-metal’, Co.
During these studies, Mustafa learned to prepare multilayers using a state-of-the art
sputtering system, and to measure very small resistances at 4.2K using a unique
superconducting-quantum device (SQUID) based bridge circuit.

After these preliminary studies, Mustafa started work on his Ph.D. thesis, which was
planned to contain two parts, both involving magnetic multilayer nanopillars. Here, he
learned to use a combination of optical and electron beam lithographies to make F/N/F

nanopillars of typical dimensions ~ 70 nm x 130 nm, and to measure both current-

induced magnetization switching (CIMS) and current induced generation of GHz
radiation.

(1) In the first part, he succeeded in producing nanopillar multilayers with inverted
current-induced magnetization switching (CIMS)—where positive current drives the
magnetic order of the sample from anti-parallel (AP) to parallel (P), and negative current
does the opposite. Mustafa’s work here led to two publications, one in the Journal of
Applied Physics and one in Physical Review Letters, which are included as chapters 2
and 3 of this PhD thesis. Further information about them, and the student who helped
him with them, will be given at the start of chapter 2.

(2) The second part, the topic of Chapter 4, involved several tasks: (a) designing a
new sample geometry for studying current-induced generation of GHz radiation, and
making samples with this geometry; (b) assembling a system for measuring GHz
radiation at frequencies up to 12 GHz using a borrowed spectrometer; (c) writing
computer programs to control the massive data taking as functions of applied magnetic
field H and applied DC current I; (d) obtaining and analyzing good GHz data; and (e)
designing and constructing a more flexible system for extending GHz measurements up
to 40 GHz and for time domain studies. Having obtained experience in the topic by
working for a summer at Seagate Technology, he succeeded in all parts. Unfortunately,
however, by the time he obtained his first few data sets, he had been diagnosed with a
bone cancer of the knee area, the effects of which progressively worsened, causing pain
that quickly reduced his ability to carry out further research. Since there was no evidence
that the cancer had spread beyond the leg, after a period of chemotherapy, he decided

with his doctors to have the knee and part of the thigh-bone removed in hopes of

eliminating the cancer. The operation appeared to be successful, and was followed by
months of additional chemotherapy during which he was mostly incapacitated. Again,
unfortunately, by the end of the treatment it became clear that the cancer had not been
completely eliminated and he and his doctors decided that the best hope was to remove
the leg up to the thigh. He underwent another operation and further chemotherapy. For
a while it looked as if this treatment might be successful. But in the end the cancer
spread further, and finally took his life.

The thesis that is presented includes both parts, however with the second part not as
extensive as it would have been absent his illness. The guidance committee judges that
this totality, combined with his outstanding academic performance, are more than
sufficient to justify the awarding of a Ph.D. degree.

Given the unusual circumstances of this thesis, it was decided to describe the work
that he did, specifying clearly and completely his contributions, but adding only the
minimum of connective material to tie together his work.

The thesis consists of the following four chapters.

Chapter 1. Background and Overview. This chapter contains some background and
an overview of the thesis, plus a list of Mustafa’s publications

Chapter 2. Inverted current-driven switching in Fe(Cr)/Cr/Fe(Cr) nanopillars.
This chapter contains a brief rational for the study of Inverted Current-driven Switching,

and then the text of the first publication, a study of Fe/Cr/Fe multilayers.

Chapter 3. Controlled normal and inverse current-induced magnetization
switching and magnetoresistance in magnetic nanopillars. This chapter contains the

second publication on inverting CIMS, which extended data to several additional F/N/F

structures, and was designed to test the importance of contributions from the bulk F-

metals and the F/N interfaces. .

Chapter 4. Spin-Transfer-Torque Generation of High Frequency (GHz)
Radiation in Nanopillars: Sample Geometry, Measuring Systems, Data Taking, and
Sample Data. This chapter contains descriptions of: (4.1) The sample geometry that
Mustafa helped to develop; (4.2) The measuring system for taking data up to 12 GHz,
examples (prepared by Mustafa) of the expected phenomena, and some data (taken by
Mustafa), including an illustration of his discovery that the data could be ‘scaled’ to make
manifest similarities of data from nominally identical, but actually slightly different,
samples; (4.3) the system that Mustafa helped to design and construct for extending
measurements to 40 GHz and for time domain studies. Mustafa also wrote programs for
data taking and analysis. Since those programs are complex, and inappropriate for

inclusion, we simply note here that he wrote them, and omit them from the thesis.

1.2. Publications of Mustafa AlHaj Darwish (MAHD).

Bold = experiments that MAHD led. Unbold = experiments to which he
contributed.

. Absence of Mean-Free-Path Effects in CPP Magnetoresistance of Magnetic
Multilayers. (K. Eid, D. Portner, J. Borchers, R. Loloee, M. AlHaj Darwish, M. Tsoi,
H. Kurt, K.V. O’Donavan, W.P Pratt Jr., and J. Bass), Phys. Rev. B65, 054424
(2002).

. Current-perpendicular—to-plane Magnetoresistance Properties of Ru and Co/Ru
Interfaces, (K. Eid, R. Fonck, M. AlHajDarwish, W.P. Pratt Jr., and J. Bass), J. Appl.
Phys. 91, 8102 (2002).

. Inverted Current-Driven Switching in Fe(Cr)/Cr/Fe(Cr) Nanopillars,
(M. AlHajDarwish, A. Fert, W.P. Pratt Jr., and J. Bass), J. Appl. Phys. 95, 6771
(2004).

. Current-induced magnetization Dynamics in Current Perpendicular to the Plane Spin
Valves, (M. Covington, M. AlHajDarwish, Y, Ding, et al.), Phys. Rev. B69, 184406
(2004)

Controlled Normal and Inverse Current-Induced Magnetization Switching and
Magnetoresistance in Magnetic Nanopillars, (M. AlHajDarwish, H.Kurt, S.
Urazhdin, A. F ert, R. Loloee, W.P. Pratt Jr., and J. Bass), Phys. Rev. Lett. 93,
157203 (2004).

. Spin Transfer Effects in Current perpendicular to the Plane Spin Valves, (M.
Covington, M. AlHajDarwish, Y. Ding, et al.), J. Magn. Magn. Mat. 287, 325 (2005).

. Manipulating Current-Induced Magnetization Switching, (S. Urazhdin, H. Kurt, M.
AlHajDarwish, N.O. Birge, W.P. Pratt Jr., and J. Bass), J. Appl. Phys. 97, 10C701
(2005)

. Effect of asymmetric leads on critical switching current in magnetic nanopillars. (H.
Kurt, M. AlHajDarwish, W.P. Pratt Jr., and J. Bass), Appl. Phys. Lett. 89, 082513
(2006).

Chapter 2. Inverted current-driven switching in Fe(Cr)/Cr/Fe(Cr)
nanopillars.

2.1.Background

The work in both Chapters 2 and 3 involved use of a lock-in amplifier for ac
measurements of dV/dI on F/N/F nanopillars as functions of applied magnetic field H for
applied DC current I = 0 (CPP-MR), and of applied DC I for applied magnetic field H = 0
(CIMS). Fig. 2.1 contains examples of ‘normal’ CPP-MR and ‘normal’ CIMS data for a
Co(30nm)/Au(10nm)/Co(4nm) nanopillar that Mustafa both prepared and measured. The
‘top’ part of the nanopillar, consisting of the thinner (here 4 nm thick Co) layer and a
nominal half of the separating layer (here Au), was patterned to nanopillar size (typically
~ 70 nm x 130 nm) and the ‘bottom’ part, composed of the rest of the separating layer
and the thicker (here 30 nm thick Co) were left unpattemed. Application of a large in-
plane magnetic field H, with DC current I z 0, will put the sample into a parallel (P)
orientation of the magnetizations M of the two F-layers, both pointing along the field.
When the field H is reversed, the thicker layer will reverse at smaller H than the thinner
layer, giving a temporary antiparallel (AP) orientation of the two magnetizations, before

the thinner layer reverses in larger H. In contrast, for H z 0, application of a large DC I

will affect only the thinner F-layer, due to its smaller volume and the much larger current
density that it experiences. The top part of Fig. 2.1 (CPP-MR) shows dV/dI vs H for Idc =
0. Normal CPP-MR means that dV/dI(AP) is larger in the state where the magnetic
moments of the two F-layers are aligned anti-parallel (AP) to each other than dV/dI(P) in

the state where the moments are aligned parallel (P) to each other. Note that dV/dI vs H

is symmetric about H = 0. The bottom part of Fig. 2.1 (CIMS) shows dV/dI vs Idc for H
= 0. Normal CIMS means that positive I (defined as electrons flowing from the thinner
‘switching’ F-Iayer to the thicker ‘fixed’ F-layer) drives the sample magnetic order to AP,
and negative I (electrons flowing from the thicker F-layer to the thinner one) drives the
system to the P-state. Thus dV/dI vs [do is asymmetric about I = 0. Note that the values
of dV/dI(AP) and dV/dI(P) are closely the same for the CPP-MR and CIMS, showing
that the nanopillar switches in both cases between fully AP and fully P states.

By the time these experiments were undertaken, most researchers accepted that the
asymmetry of the CIMS switching in Fig. 2.1, meant that the CIMS switching could not
be due primarily to the self-Oersted field of the large DC current I, because should a field
should produce effective ‘P-like’ states at both large +H and large —H (we say ‘effective’
P-like, because the circular field of a linear I should produce stable vortex-like final
states). However, some were still not convinced, and one of the purposes of these
experiments was to address this issue in a new way.

The purposes of the studies in the two papers below were, thus, two-fold. One was to
test unequivocally for dominance of CIMS, in that the self-Oersted field cannot produce
inverted CIMS, since its direction is independent of the specific scattering asymmetries
of the F metal and F/N interfaces. The other was to see if samples containing F and N
metal pairs with different bulk and interfacial scattering anisotropies would behave as
predicted by the original simple Slonczewski ‘ballistic’ model of CIMS, or if one could
distinguish effects of ‘spin-accumulation’ in the nanopillars. Said another way, to see if
transport in the samples was better treated as ‘ballistic’ (original Slonczewski) or

‘diffusive’ (later models). Showing that CIMS could be qualitatively changed by

alloying the F -metal would be strong evidence in favor of diffusive transport. Details of
the samples and their behaviors, and the conclusions reached, are given in the two papers.

All of the samples involving Fe and Cr in both Chapters 2 and 3 were made and
measured by Mustafa. As shown in Fig. 1.1, Mustafa also measured simple ‘norrnal’
CIMS samples comparable to the Py/Cu/Py samples included in Chapters 2 and 3. The
Py/Cu/Py data were chosen because we had already published a complete study of such
samples in ref. [26] below, and we used Mustafa’s new samples of Py/Cu/Cr/Fe(Cr)
samples in Chapter 3. The remaining samples in Chapter 3 were made by Huseyin Kurt
and studied jointly by Mustafa and Kurt. These samples were included in the earlier

Ph.D. thesis of Huseyin Kurt.

Normal switching in Co30nm/Au10nm/Co4nm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Nanopillars
E . . . . . 3 0.95? . . . . . ‘
mi KI E 2
gm: Q E I
2- s 4
3...; % s 3
I 295 K: I , 4.2.3
1.22’ * * ‘ ‘ 4 I 0.85' _‘2 _‘1 b :' ‘2 7
H(kOe)
0.96, ' '
Q - g
15 . g
B 0.9
' . 295K; .
1.22"“;i6 ““““““““““ 1'0 ~‘ 0.35"

 

 

 

 

 

 

 

Fig. 2.1. As an addition for this thesis, we show CPP-MR (top) and CIMS (bottom) data
for Co/Au/Co at both 295K (left) and 4.2K (right), measured by Mustafa AlHajDarwish,
that could have been used as examples of ‘normal’ switching in place of Fig. 2.2 for
Py/Cu/Py in the following paper.

2.2. Inverted current-driven switching in Fe(Cr)/Cr/Fe(Cr) nanopillars.

M. AlHajDarwish,l A. Fert,2 W.P. Pratt Jr.,I and J. Bass1
[1] Department of Physics and Astronomy, Center for Sensor Materials, Center for
Fundamental Materials Research, Michigan State University, East Lansing, MI, USA

48824-2320.

[2] Unite Mixte de Physique, CNRS/THALES, Orsay, France 91404.

From both theory and experiment, scattering of minority electrons is expected to be
weaker than scattering of majority electrons in both dilute Fe(Cr) alloys and at Fe(Cr)/Cr
interfaces. We show that Fe(Cr)/Cr/Fe(Cr) trilayer nanopillars display a normal
magnetoresistance—Le, largest resistance at low magnetic fields and smallest at high
fields, but an inverted current-driven switching—Le, positive current flowing from the
fixed to the reversing layer switches the trilayer from higher to lower resistance, and

negative current switches it from lower to higher.

Reprinted with permission from M. AlHajDarwish, A. F ert, W.P. Pratt Jr.,
and J. Bass, J. Appl. Phys. 95, 6771 (2004), Copyright 2004, American Institute

of Physics.

10

There is great interest in current-driven magnetic switching in nanofabricated
ferromagnetic/non-magnetic/ferromagnetic (F/N/F) trilayers, both to understand the
underlying physics and for device potential."27 For simplest switching, one F-layer
(pinned) is made much thicker (and sometimes with larger area) than the other. An
applied dc current then reverses only the thinner (and sometimes smaller) F -layer (free).
In all nanopillars studied so far, minority electrons were scattered more strongly than
majority ones both within the F-layers and at the F/N interfaces. In such a case, a
positive current flowing in the spacer layer from the thick pinned layer to the thin free
one is positively polarized in the frame of the thick layer (i.e. its magnetic moment is
parallel to the magnetization of the pinned layer). Large enough positive current was
then found to cause the free layer’s magnetic moment to rotate anti-parallel (AP) to that
of the pinned layer, and reversed (negative) current caused it to rotate parallel (P).
Because the scattering was the same (strongest for minority electrons) within both F-
metals and at all F/N interfaces, the resistance of the trilayer was largest in the AP state at
low magnetic fields H and smallest in the P state at high H, corresponding to a normal
magnetoresistance (MR).28 Sufficiently large positive current then produced a step
increase in resistance and sufficiently large negative current a step decrease, which,

together, we call normal switching.

If, however, majority electrons are scattered more strongly both within the pinned
F -metal layer and at its F/N interface, the direction of polarization of the exiting current
should reverse—Le. positive current should be negatively spin-polarized. Published
models imply that negatively-polarized positive current impinging upon the free layer

should cause its moment to rotate P to that of the pinned layer,"2 and negative current

11

should cause the moment of the free layer to rotate AP. If the two F layers are identical,
the field-driven magnetoresistance (MR) should remain normal, with larger resistance in
the low field AP-state than in the high field P-state.28 However, the current-driven
switching should ‘invert’, in that positive current should drive the system from the higher
resistance AP state to the lower resistance P state, and negative current the opposite.

”'33 indicate that both a dilute Fe(Cr) alloy and an

Together, theory and experiment
Fe(Cr)/Cr interface should scatter majority electrons more strongly. In this paper we
show that an Fe(Cr)/Cr/Fe(Cr) trilayer displays the two behaviors described in the
previous paragraph, namely a normal MR—resistance smallest in the high field P-state,

but an inverted current-driven switching—positive current drives the system to the lower

resistance P state and negative to the higher resistance AP state

Magnetic nanopillars of approximately elliptical shape and dimensions ~ 70 nm x
130 nm were prepared by triode sputtering onto Si substrates.2"’27 The Fe(Cr) alloy
contained ~ 5 at.% Cr. The multilayers consisted of a thick Cu lower contact, a 30 nm
pinned Fe(Cr) layer, a 6 nm thick Cr layer, a 3.5 nm thick ‘free’ Fe(Cr) layer, and a thick
Au top contact. To minimize dipolar coupling between the Fe(Cr) layers, the sample was
ion-milled through part of the Cr layer, so as to leave the bottom (pinned) Fe(Cr) layer
wide. With this geometry, the wide layer ‘switches’ upon application of a relatively
small magnetic field, but does not switch upon application of a current large enough to
switch the patterned top (free) Fe(Cr) layer. Differential resistances, dV/dI, were
measured with four probes and lock-in detection, adding an ac current of ~ 20 pA at 8

kHz to the dc current I. H is directed along the easy axis of the nanopillar.

12

As a comparison standard for our Fe(Cr)/Cr/Fe(Cr) data, Fig. 2.2 shows previously
published data [26] for deI vs I at H = 50 Oe and dV/dI vs H at I = 0 for nanopillars of
Py/Cu/Py (Py = Permalloy = N134F616), where positive current was defined the same as in
the present paper. The dV/dI vs I data are shown full size, and the dV/dI vs H (i.e., MR)
data as insets in the top center. The upper curves are for room temperature (295K) and
the lower ones for 4.2K. Both the MR and dV/dI vs I curves for Py/Cu/Py are
‘normal’—i.e. dV/dI is smallest in the high field P state, and positive current switches the
sample from its low resistance (P) state to its high resistance (AP) state, and vice versa
for negative current. In the MR curves at both temperatures, the transitions from the P to
AP states occur only after the magnetic field passes through zero, consistent with weak

magnetic coupling.

Fig. 2.3 shows our new data for Fe(Cr)/Cr/Fe(Cr) trilayers. The changes in dV/dl
vs I or H for Fe(Cr)/Cr/Fe(Cr) are considerably smaller than those for Py/Cu/Py, but still
visible. As expected, the MR curves at I = 0 are still ‘normal’—i.e., smallest in the P
state. And, similarly to Py/Cu/Py, the MR transitions from the P to AP states appear only
after the magnetic field passes through zero. In contrast to the similar MR curves in Figs.
2.2 and 2.3, the dV/dI vs I curves for the Fe(Cr)/Cr/Fe(Cr) samples behave oppositely to
those for Py/Cu/Py. Positive I switches the Fe(Cr) free layer from the high resistance AP
to the low resistance P state, and negative I switches it from the P to AP state. The dV/dI
vs I curves for Fe(Cr)/Cr/Fe(Cr) are ‘inverted’ from the ‘normal’ behavior for Py/Cu/Py.
In all cases in both Fig. 2.2 and Fig. 2.3, the agreement between the minimum and

maximum values of dV/dl, together with almost single step switching (e. g., at -11 and +

13

6 ma at 4.2K), show that the current-driven switching is complete. Normal MR5 and

inverted current-driven switching were also seen in other Fe(Cr)/Cr/Fe(Cr) samples.

We also checked that similar results were obtained for the MR with I ¢ 0 and for
dV/dI vs I for H at 0. Fig. 2.4 shows how dV/dl vs H and dV/dI vs I change at 4.2K for
the same Fe(Cr)/Cr/Fe(Cr) trilayer when we vary H at fixed I (Fig. 2.4a) and I at fixed H
(Fig. 2.4b). In part because the jumps in dV/dI are small, the switching is irregular and
often partial. But all dV/dI vs H switching seen is normal, and all dV/dI vs I switching
seen (both hysteretic and reversible) is inverted. Asymmetry in H is presumably due to a

combination of the self-Oersted field and sample shape asymmetry.

To summarize, we have shown that F e(Cr)/Cr/F e(Cr) trilayers give normal MR—
smallest dV/dI vs H at high fields when the magnetizations of the two Fe(Cr) layers are
aligned parallel (P) to each other, but inverted dV/dI vs I—large positive current causes
switching from the AP to the parallel (P) state and large negative current causes

switching from the P to the AP state.

Acknowledgments: The authors thank S. Urazhdin for helpful suggestions and for
permission to use his published Py/Cu/Py data in this paper. This research was supported
in part by the MSU CFMR, CSM, the MSU Keck Microfabrication Facility, NSF grants
DMR 02-02476, 98-09688, NSF-EU collaborative grant 00-98803, and Seagate

Technology.

14

 

Pleu/Py
é 1.60I-
1.7 _. 5
g 3 1.55-
i
>
'0

 

 

 

-
I

 

 

 

 

 

 

"1.2-
1.3 - g m
2 .
r- I- >
9 131.1.
a 1-2' -o.2 oio‘oiz‘
S H(kOel
.5
1.1 -
4.2 K
I . l 4 . 1
-1O -5 10
l(mA)
Fig. 2.2. Py/Cu/Py data at 295K and

4.2K showing normal switching for
dV/dI vs I at H = 50 Oe (main figures)
and also for dV/dI vs H at I = 0 (insets).
(From Urazhdin et a1. [26]).

15

.N
w
to

dV/dl (Q)

2.33 -

Fe(Cr)/CrlFe(Cr)

 

pp
cow
GIN

 

 

 

 

 

 

 

 

 

 

 

 

 

-1'0 1 b ‘ 1o
|(mA)

Fig. 2.3. Fe(Cr)/Cr/Fe(Cr) data at
295K and 4.2K showing inverted
switching for dV/dI vs I at H = 0
(main figures) but normal dV/dI vs H
at I = 0 (insets).

N
N

dV/dKQ)
P
a

N
_s

2.05

 

 

vvv'v V‘V'Vv'vvvvlvvvv'fv

Fe(Cr)/CIIFeICr) (b)
2.4 4.2 K

10009
800e

   

 

  
 

2.3

 

dV/dI(Q)

 

 

 

1 (mA)

Fig. 2.4. Fe(Cr)/Cr/Fe(Cr) data at 4.2K
showing: (a) normal MR for dV/dI vs H at
various I, but (b) inverted switching for
dV/dI vs I at various H. Curves in (a) for I
at 0 and in (b) for H at O are shifted
vertically for clarity.

16

Chapter 3. Controlled normal and inverse current-induced
magnetization switching and magnetoresistance in magnetic

nanopillars.

M. AlHajDarwish,‘ H. Kurt,‘ s. Urazhdin,‘ A. Fert,‘ R. Loloee, W.P. Pratt Jr.,' and J.
BassI
[ 1] Department of Physics and Astronomy, C enter for Sensor Materials, Center for
Fundamental Materials Research, Michigan State University, East Lansing, MI, USA

48824-2320.

[2] Université Paris-Sud and Unite Mixte de Physique, CNRS/THALES, Orsay, France

91404.

By combining pairs of ferromagnetic metals with the same or different signs of
scattering anisotropies in ferromagnetic/non-magnetic/ferromagnetic metal nanopillars,
we independently invert just the magnetoresistance, just the direction of current-induced
magnetization switching, or both together, at room temperature (295K) and at 4.2K. In
all cases studied, the switching direction is correctly predicted from the net scattering

anisotropy of the fixed ferromagnet, including both bulk and interfacial contributions.

Reprinted with permission from M. AlHajDarwish, H.Kurt, S. Urazhdin, A. Fert, R.

Loloee, W.P. Pratt Jr., and J. Bass, Phys. Rev. Lett. 93, 157203 (2004), Copyright 2004,

American Physical Society.

17

The magnetization of a ferromagnetic (F) metal can be reversed by spin transfer from
a spin-polarized current, i.e., without applying a magnetic field [1-4]. Such Current-
Induced Magnetization Switching (CIMS) [1-7,9,12,13,15,23,26,27,34-45] has been seen
in (F1/N/F2) trilayer nanopillars in which the spin polarized current prepared with a
thicker layer F1 passes through a non-magnetic metal (N) and switches the moment of a
thinner layer F2. CIMS is promising for switching small magnetic devices (e. g., magnetic

random-access-memory, MRAM), and also raises subtle fundamental issues.

Although CIMS is expected to result from spin polarization of the current, it has yet
to be shown that CIMS can be manipulated (e.g., inverted) by changing that polarization.
In prior studies, minority electrons were scattered more strongly in F1, F2 and at FUN
and N/F 2 interfaces (positive spin anisotropy). The current is then always positively spin-
polarized in the F-layers, i.e., carried mainly by majority spin electrons. In these
“standard” conditions, electrons flowing from F1 to F2 (negative charge current, I < 0)
switch the moment M2 of F2 from antiparallel (AP) to M1 (high resistance R) to parallel
(P) to M1 (low R). Conversely, positive I (> 0) switches F2 from P to AP. We call these
behaviors ‘normal’ CIMS and ‘normal’ current-perpendicular-to-plane (CPP)

magnetoresistance (MR).

We present CIMS experiments exploiting the possibility of inverting the spin
anisotropy by doping F 1, F2, or both together, with an impurity (Cr) that scatters
majority spin electrons more strongly [28,29,31,32,33,46-48]. We thus show, for the first
time, that inversion of the spin anisotropy can invert the CIMS direction, i.e., invert the
signs of I for AP to P and P to AP transitions. We also find inversions of the MR (larger

R for P state) with appropriately doped samples, as expected from prior CPP-MR results

18

at low temperature [28,29]. Measurements at 295K and 4.2K show that the qualitative

behaviors of both CIMS and MR are independent of temperature over this range.

Analysis of the switching behaviors lets us discriminate between different models of
CIMS. We divide the standard models of the spin-transfer-torque (STT) used to describe
CIMS into two classes, ballistic [1,3,37] and diffusive [2,4,12,13,41,45,49]. Both predict
that changing scattering anisotropies can invert the MR and/or CIMS. However,
expectations for the two need not agree. In ballistic transport, the spin anisotropy comes
only from reflections at the F/N interfaces. Inverting CIMS is predicted to require
negative anisotropy at Fl/N [37]. Inverting the MR should require opposite scattering
anisotropies at FUN and N/F2 [28]. In diffusive transport, the spin anisotropy of
scattering within the F -layers is also important, so that one must consider the net
anisotropy of a given F -layer (e.g., the resultant effect of the bulk of F and its F/N
interface). An additional effect, spin accumulation, can also either support or compete
with the effect of polarized current [2,4,12,13,41,45,49]. In addition to determining the
relation between spin anisotropies and CIMS direction, we will answer the following
questions relevant to understanding CIMS. (a) Is the CIMS direction set only by interface
scattering anisotropy? No. (b) Can impurity scattering within the layers be important?
Yes. (0) Do the anisotropies of F1 and F2 play different roles for the CIMS direction of

F2? Yes. (d) Can spin-accumulation be important? Yes.

To determine how changing spin anisotropies changes CIMS directions, we combine
in different ways three pairs of materials: Py/Cu (Py = Ni34Fe16) where both bulk and
interface anisotropies are positive [28]; Fe(Cr)/Cr (Fe(Cr) = FCgsCI‘s), where both are

negative [29,31,32,33,46-48]; and Ni(Cr)/Cu (Ni(Cr) = Ni97Cr3), where the Ni(Cr) is

19

thick enough so its negative anisotropy dominates the positive anisotropy of the interface
[50]. The net anisotropy is determined using the MR. Comparing Fe(Cr)/Cr with
Ni(Cr)/Cu for F1 or F2 explicitly tests the importance of interface anisotropies. Table 3.1
lists the signs for F1, F2 bulk, Fl/N, N/F 2 interfaces, and F1(net), F2(net), as well as for

the observed MR and CIMS.

Our sample preparation and measurement techniques are described elsewhere [26,27].
Our multilayers were triode sputtered onto Si substrates, and patterned into nanopillars of
approximately elliptical shape and dimensions ~ 70 nm x 130 nm. The samples consisted
of a thick Cu lower contact, the multilayer, and a thick Au top contact. The N-layer was
made thick (6-20 nm) to minimize exchange coupling between F1 and F2. To simplify
switching, samples were ion-milled only through F2 and part of N, leaving F1 (fixed
polarizer) to have much larger area (~ umz) and to be thicker than F2. Dipolar coupling
between F1 and F2 is then minimal, and H (along the nanopillar easy axis) reverses M1
and M2 sequentially, but I reverses only M2 of F2 (free switcher). For each combination
of MR and CIMS, the switching direction is the same at 295K and 4.2K. Each switching

behavior was also independently reproduced, and no inconsistent switching was seen.

Py and Py/Cu interfaces both have positive scattering anisotropy [28]. In agreement
with prior data [26,27], Fig. 3.1 shows that Py(24)/Cu(10)/Py(6) nanopillars (layer
thicknesses are in nm) give normal MR and normal CIMS. At both 295K and 4.2K, the
MR transitions from P to AP occur after H passes through zero, consistent with little or
no magnetic coupling. The agreement between the minimum and maximum values of
dV/dI for the MR and current—driven curves also shows that the switching is complete.

Figs. 3.2-3.5 show similarly weak coupling and complete switching.

2O

In contrast to Py and Py/Cu, Fe(Cr) and F e/Cr interfaces both have negative scattering
anisotropies [29,31,32,33,46-48]. Since F1 and F2 are the same alloy,
Fe(Cr)(30)/Cr(6)/Fe(Cr)(3.5) nanopillars should give normal MR [28,29]. Fig. 3.2 shows
that they do, and also give inverse CIMS. The changes in dV/dI vs I or H are smaller
than for Py/Cu/Py, due to spin-memory-loss in the Cr(6) layer [47] and smaller scattering

anisotropy of F e(Cr) [29]. More Fe(Cr)/Cr/Fe(Cr) data are given in [46].

Fig. 3.3 shows data for the four component system Py(20)/Cu(7)/Cr(3)/Fe(Cr)(3).
Combining the net positive scattering anisotropy for F1 with the net negative anisotropy
for F2 gives the expected inverse MR [28,29]. But the CIMS is normal—I > 0 switches

from P to AP— remembering that inverse MR means largest resistance in the P state.

Fig. 3.4 shows the fourth case, Ni(Cr)(20)/Cu(20)/Py(10). Combining net negative
anisotropy for Ni(Cr) with net positive anisotropy for Py gives the expected inverse MR,

and now inverse CIMS.

Fig. 3.5 shows another way to achieve inverse MR with normal CIMS, using
Py(24)/Cu(10)/NiCr(4). In Figs. 3.3 and 3.5, this same combination of MR and CIMS
occurs with opposite F2 interface anisotropies—negative in Fig. 3.3 but positive in Fig.

3.5.

Lastly, in Figs 3.2 and 3.4, opposite interface anisotropies for F1 don’t change the

CIMS direction.

Before comparing our data with theory, we summarize the results in Figs. 3.1-3.5. As
expected for the MR [28,29], when the net scattering anisotropies for F1 and F2 are the

same (Figs. 3.1, 3.2), the MR is normal, and when they are opposite (Figs. 3.3-3.5), the

21

MR is inverse. New for CIMS, when the net scattering anisotropy for F1 is positive,
CIMS is normal (Figs. 3.1,3.3,3.5), and when it is negative (Figs. 3.2,3.4), CIMS is
inverse. For these samples, the direction of CIMS is set by the net scattering anisotropy
of F1 and is independent of that of F2. Comparing Figs. 3.3 and 3.5, and Figs. 3.2 and
3.4, shows that, when bulk scattering predominates, the CIMS direction is independent of
the scattering anisotropy of F UN and N/F 2. Finally, dominance of the bulk contribution
of scattering anisotropy in either F1 (e.g., Ni(Cr) in Fig. 3.4) or F2 (Ni(Cr) in Fig. 3.5) is
inconsistent with ballistic transport through the nanopillar, where the interfaces must

dominate the scattering.

As just noted, ballistic STT models cannot describe our data involving Ni(Cr).
Intriguingly, however, the CIMS directions in Figs. 3.1-3.5 accord with the ballistic
prediction of [37], if the scattering anisotropy at the F1/N interface is simply replaced by
the net anisotropy for F1. This last result also follows directly from the non-STT model

of ref. [44].

For diffusive transport, the current polarization in N depends upon the net scattering
anisotropies of both F1 and F2, and CIMS depends upon both the spin polarized charge
current and spin-accumulation effects [2,4,12,41,45]. Eq. 3.1 reproduces Eq. 5 of ref.
[45] for the torque FP at a small angle from the P state (for 1“”, replace P by AP). Note

that our notations for F1 and F2 are reversed from [45].

rpm = [{(vF mfivs +(j,’,’,,N>/2}<1 - e"’N/1‘V)+{(VF min/4) +

15.,“ }e"~’t~ 1(sz(szM.)) (3.1)

22

Eq. 3.1 comes from an extension of the Valet-Fert [51] model of CPP-MR to non-
colinear states. vF is the Fermi velocity in N, m; and j; N are the spin accumulation
density [15,16,19,30] and spin current density in N just outside the F2/N interface
calculated for the P state, mil and jg,“ are the same quantities in F1 just inside the

FI/N interface, and tN and 2N are the thickness and mean-free-path of N. The second { }
parenthesis dominates in the usual case where tN << AN and the first dominates if tN >>
2N. In both cases, the signs of spin—current and spin-accumulation can support each other

or compete.

Using the best available parameters from CPP-MR experiments [28,47,50], we
calculate spin currents and spin accumulations [28,51], and then insert these into F P or
F "P . In all cases except Fig. 3.4, the signs of spin current and accumulation agree and
are the same as those expected from the sign of the net spin anisotropy of F1 seen by MR.
The situation for Fig. 3.4 is more complex. For P to AP, the spin accumulation
dominates F P and gives the observed inverse CIMS at I <0; because of the particular
parameters of Ni(Cr) and Py, for this case the spin current alone would predict normal
CIMS. For AP to P, the spin current dominates F AP and gives the observed inverse

CIMS at I > 0. Thus, we reproduce the behaviors in Fig. 3.4.

To summarize, we have shown that judiciously chosen pairs of ferromagnetic (F)
metals or ferromagnetic alloys can produce all four combinations of normal and inverse
MR and current-induced magnetization switching (CIMS) at both 4.2K and 295K. The
MR is normal if the net scattering anisotropies of F1 and F2 have the same sign, and

inverse if they don’t. For the samples studied, the CIMS direction is determined solely

23

by the net anisotropy for F1, although in Fig. 3.4 this result requires dominance of spin-
accumulation for the P to AP transition. This latter result, as well as the inverted MRS in
Figs. 3.4 and 3.5, show that the interpretation of MR and CIMS must generally take
account not only of the interface scattering assumed in ballistic models, but also
scattering (and diffusion) within the F-layers. As the widely accepted mechanism of
CIMS is a quasi-interfacial absorption of the transverse component of the spin current
[1,12,37,41,45], the importance of scattering within the F-layers might seem surprising.
However, in a non-collinear magnetic configuration, the transverse spin current in the
frame of F2 is related to the longitudinal one in F1, and a global treatment [12,41,45] of
the longitudinal and transverse components of the spin current and spin accumulation

then requires the diffusive aspects of CPP-MR theory [28,51].

Acknowledgments: The authors thank Henri Jaffrés for spin-accumulation
calculations, and NO. Birge, P.M. Levy and MD. Stiles for helpfill suggestions. This
research was supported by the MSU CFMR, CSM, Keck Microfabrication Facility, NSF
grants DMR 02-02476, 98-09688, NSF-EU collaborative grant 00-98803 and Seagate

Technology.

24

Table 3.1

 

 

 

 

 

 

Fig. F1/N/F2 F1 Fl/N F1(Net) F2 N/F2 F2(Net) MR CIMS
#

1 Py/Cu/Py + + + + + + + +

2 Fe(Cr)/Cr/Fe(Cr) - - - - - - + -

3 Py/Cu/Cr/Fe(Cr) + + + - - - - +

4 Ni(Cr)/Cu/Py + + + + - -

5 Py/Cu/Ni(Cr) + + - + - - +

 

 

 

 

 

 

 

 

 

 

 

 

Table 3.1: F1/N/F2 for each figure, giving: the spin anisotropies (+ = positive, - = negative)
of F1, FUN, and their net anisotropy F 1(Net), those of F2, WM, and their net anisotropy
F2(Net), and the signs of the observed MR and CIMS (+ = normal and - = inverse).

25

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

g ' AP
160 £1.45 .
g - P
51.55 1-4°~_________'r :
i
3 1.50»
1:
1.45.
1.40 1
1.4 511 AP.
6
1.3. 31.0. P.
a 1: > '4. ,
:' 1,2 -0.4 0 0.4
E H(kOe)
>
'5 1.1
1.0 pf j
42K
-6“"—“'_ .4 -2
l(mA)2

Fig. 3.1. Py(24)/Cu(10)/Py(6) data at 295K (top) and 4.2K (bottom) showing normal MR
(dV/dI vs H at I = 0) in the insets and normal CIMS for dV/dI vs I in the main figures at
H = 0 0e for 295K and at H = 20 Oe for 4.2K. In all figures, I > 0 is always from F1 to
F2.

26

Fe(Cr)/CrlFe(Cr)

 

4.02

 

 

3.98 .

dVIdI (Q)

3.94 .

 

3.35

 

 

3.25

dVIdI (fl)

 

 

3.15 .

 

 

IImAl

Fig. 3.2. Fe(Cr)(30)/Cr(6)/Fe(Cr)(3.5) data at 295K (top) and 4.2K (bottom) showing
normal MR (dV/dI vs II at I = 0) in the insets but inverse CIMS for dV/dI vs I at H = 0 in

the main figures.

27

Pleu/CrlFe(Cr)

 

 

 

 

 

012.34
2.38 - a P
,. $32.32 WA...
8236- \ .i ' i 'i
2 H(k0e)
>
'5 2.34 . :\ J/
\ /
Am: ,7
2-32 - ._~'/ 295K‘
518.1 ' '
1.92 g P
>1.80
A '5 AP
g 1.88.
3
E 1.84.
1.80-

 

 

 

-8 .21 6 4 8
I (mA)
Fig. 3.3. Py(20)/Cu(7)/Cr(3)/Fe(Cr)(3) data at 295K (top) and 4.2K (bottom) showing

inverse MR (dV/dI vs H at I = 0) in the insets but normal CIMS for dV/dI vs I at H = 0 in
the main figures.

28

Ni(Cr)ICuI

 

 

 

 

 

 

 

 

3.14 F , .
3.12 ~
3.10
9. 3.08 .
§ 3.06 ‘1\ p
'o \’,\
3.04- AP 295K-
3.02 ~ >< T’.?‘:\
3.00 . . A '
\ 9.2.
2.45, a
fix g 2.
E 2.40» ‘
S 2.35 ~ <3. p
'3 ‘\
AN
2.30 x; 42K
\r: f - 7 f /
2-25 - - - e ’ -
-12-10-8-6 -4-2 0 2
I(mA)

Fig. 3.4. Ni(Cr)(20)/Cu(20)/Py(10) data at 295K (top) and 4.2K (bottom) showing
inverse MR (dV/dI vs H at I = 0) in the insets and inverse CIMS for dV/dI vs I at H = 0 in
the main figures.

29

PleulNi(Cr)

 

       
 

 

 

 

 

 

 

1.12 11
i I. -

a 1.11 ~ :10943 -
75 ' -0.5 0 0.5
2 H(k0e)
% 1.1-

1'09-10 -5 0 5

I(mA)

Fig. 3.5. Py(24)/Cu(10)/Ni(Cr)(4) data at 4.2K showing inverse MR (dV/dI vs H at I = 0)
in the inset and normal CIMS for dV/dI at H = 0 0e in the main figure.

30

Chapter 4. Spin-Transfer—Torque Generation of High Frequency
(GHz) Radiation in Nanopillars: Sample Geometry, Measuring Systems,

Data Taking, and Sample Data.

4.1. Background.

The first experimental evidence for generation of high frequency oscillations in F/N
multilayers by a large current density was given in [5,6]. The first spectrosc0pic studies,
such as those described in this chapter, were [52] for nanopillars and [53] for injection
from a point contact into extended multilayers. When Mustafa began his PhD thesis
research, our group had no experience with GHz measurements. We were, thus, pleased
when Seagate invited him to work with them for one summer on early GHz studies. He
was supervised there by Mark Covington, who was very pleased with Mustafa’s work,

which led to two papers [54,55].

After returning to MSU, Mustafa contributed to: (1) developing a sample geometry
for GHz studies, (2) assembling a system for measuring such radiation at frequencies up
to 12 GHz and using this system for taking some data, and (3) developing a system for
measuring to frequencies up to 40 GHz. Unfortunately, he became incapacitated before
he could use this latter system. We describe here each of these items, using figures
prepared by Mustafa (with only occasional minor modifications for clarity), and
providing only minimal connective material.

4.2 Sample Geometry.

Jointly with Irinel Chiorescu and Bill Pratt, Mustafa designed a sample geometry for

GHz measurements. Together they decided that we could put six samples on a chip, with

31

each sample part of a coplanar waveguide to give a 50 0 effective impedance as seen by

a ‘picoprobe’ pair (50 Q miniature coaxial cable contact probes). Details of the sample
structure are shown in Fig. 4.1. As the lithographic techniques for preparing nanopillar
samples are similar to those described in the MSU PhD thesis of Huseyin Kurt (2005), we
do not repeat them here.

Fig. 4.1 shows the final chosen sample geometry. Each chip contains six nanopillars
as shown at the bottom left. Each nanopillar is at the center of four contacts (two
grounded) as shown at the bottom right. This structure is in the center of a larger co-axial
strip geometry shown in the upper middle. The picoprobe contacts can be connected

either at the top or the bottom (both cases shown).

32

MSU Au-electrode configuration for making
connections to nanopillars & nanocontacts

SO-Q miniature coax with
contact probes (“picoprobe")

Jr.
200nm
T

 

        
  
   
 

Au central electrode
50-!) strip line

 

.1

 

  
   

Au ground plane

 

A—
C

V
p.
A

contacts to nanopillars
in this region

r“
I \
L-

Fig. 4.1 Sample configuration for GHz measurements. Each Si substrate contains 6
samples (lower left). Each sample has four Au contacts (lower right). Two of these Au
contacts are grounds and two constitute 50 Q strip lines (upper figure). Each sample can
be contacted by a picoprobe coaxial cable pair in two ways as illustrated at the top and
bottom of the upper figure.

 
 
  
  

_

        
   

-l

 

Si substrate

 

 

 

33

4.3 Measuring System to 12 GHz, Expected Behaviors, and Data.

4.3.1. The Measuring System to 12 GHz.

The measuring system for up to 12 GHz consists of the sample, RL, a Bias T to allow
simultaneous handling of input of a large DC current into the sample and emission of
GHz radiation by the sample, and a pre-amplifier and a spectrometer to measure the
emitted GHz radiation.

Fig. 4.2 shows schematic drawings of: a nanopillar (left), a sample with picoprobe
attachment (center), and the system for measuring frequencies to 12 GHz (right). Since
we have already discussed the nanopillar geometry in sections 2 and 3, and the sample
geometry for GHz measurements in section 4.2, we focus in this section of chapter 4 on

the measuring system and on data taken with it.

   

Pico probe

 

50 n
Coplanar Waveguide

Fig. 4.2. Schematic drawings of a nanopillar (left), a sample with pico-probes (middle),
and the system for measuring GHz radiation up to 12 GHz.

34

4.3.2 Expected Dynamical Motions.

The next issues are the dynamical motions and frequencies expected from the spin-
torque model for assumed monodomain motion of the free (i.e., precessing) layer. Fig.
4.3 shows examples of trajectories of the magnetization of the ‘free’ layer, calculated by
Mustafa, using a model by J .Z. Sun [56], of the dynamics expected for an F/N/F
nanopillar subject to spin-transfer torque, assuming that the excited free layer remains

monodomain.

Dynamics in MonoDolnain Spin Torque Model

(I) Small Angle (II) Large Angle (III) Out of Plane

     

 

 

 

 

Mx , - . - 3
4‘ 1 g II Ill 1
Frequency Vs Current: 5
‘ A 3‘ .9 . o
-SmallAngle. (£9 2‘ ...o. o l
a: . 5 . .° .. ° 1
- Higherl -) frequency 1‘ K, 1.,» . . .
Harmonrcs. , 5-..oi,.--""
0 :"'"~-:§ ‘
4 ' s ' é ' i V 8 V 8
I(mA)

Fig. 4.3. The x,y.z. scale, and the direction of application of the magnetic field (top left).
Three different monodomain motions of the flee layer magnetization (assumed
monodomain) (top middle and right). Frequencies expected for different applied
magnetic fields (lower right).

35

The motions such as those shown in Fig. 4.3 give rise to a predicted phase diagram vs H and I

for monodomain precession of the fiee layer (including reversals at low enough H and I). Fig. 4.4
shows the phase diagram as functions of H (ordinate) and I (abscissa) expected from
monodomain analysis such as in Fig. 4.3. Between 1c. and 10+, the sample switches irreversible
between parallel (P) and anti-parallel (AP) states. Examples of such behavior were shown in
sections 2 and 3. For large enough positive I, and small H, the sample is driven into the AP state.
For large enough magnitude negative I, and all H, the sample is driven into the P state. For large

enough I and large enough H the sample is driven into dynamical states with motions such as

shown in Fig. 4.3.
Phase Diagram R vs I&H
H ’A I, = 1,..- I,

H

II III

 

%
co

dy Dynamics

AP

 

   

 

 

O—
+

lc_

Fig. 4.4. Phase diagram as functions of applied magnetic field H (ordinate) and current I

(abscissa) expected from monodomain analysis. I, H, and III correspond to the different
motions shown in Fig. 4.3.

36

4.3.3. Experimental Data and an Example of Scaling.

Finally, we turn to examples of data, shown for two samples (#3 and #5) from chip
#2b in Figs. 4.5 and 4.6. Comparison of the various parts of these figures shows general
similarities for the two samples, but differences in details. Mustafa discovered that the
similarities could be made clearer by comparing data at closely similar normalized values
of H/Hc and I/AIC, where AIC is the difference between the critical switching currents Ic+
and Ic-, and HC is the critical switching field, both indicated in Fig. 4.4. Normalized
values for the two samples from chip #2b are shown in Fig. 4.7. He argued that such a
result was consistent with the Landau-Lifshitz-Gilbert (LLG) equation including the
Slonczewski (S) torque term as follows.

First, define m as the unit vector along the direction of magnetization M of the free
(switching) F-layer, and m’ as the unit vector along the direction of M’ of the polarizing
(fixed) F-layer. Also, define an effective magnetic field HF = HApp + HAN + Bo , where
HApp is the applied magnetic field, HAN is the in-plane anisotropy field, and HD is the
demagnetizing field. Then the LLG-S equation can be written as [57,58]

dm/dt = - yom x Hf+ 01(m x (dm/dt)) + IK(m x m x m’), (4.1)
where I is the applied current and K is a constant that depends on known constants (such
as the electron charge and Planck’s constant) and a polarizing factor derived from the
angular dependence of the Slonczewski spin-torque. If the F- and N-layers of two
samples are the same metals, with the same nominal thicknesses, but with modestly
different values of Hc and Ic due to subtle differences between the samples, then one

might hope that normalizing HF by He, and I by AIC, where Hc and Ic are defined in Fig.

37

4.4, would give a normalized Eq. 4.1 that corrects for these differences, resulting in

‘scaling’ behavior . Fig. 4.7 shows an example of such scaling.

  
  
  
   
      
 

  
  
        

 

 

 

 

 

 

 

 

 

Py30nm/Cu10/Py6
I(mA) I(mA)
10 15 20 1o 15 20W
“‘03. men
A450 °5°°° 8450 1530904
Egg-100 ”"23 I 100 9779
0.02062 _50 2989
0.000 -3800
20 __ 20
I3. )) 12:: ———(A) 450005
15 . A15 —(B) -7200e
A ——(0) 17mA m
m —(D’) 20mA B —(C) ~7°°°¢
B10 1.10 —(D) -5000e
g 5 H=—700 Oe § 5 I=17mA
a
0- 0
° 2 4 e
. . . o
0 2 4 e 8
f (GHZ) f IGHZI

Fig. 4.5. GHz radiation data fiom sample #2b5. Top: Examples of experimental ‘phase
diagrams’ showing MR (upper lefi) and integrated power (upper right) as functions of H
and 1. Bottom: Examples of GHz peaks for fixed H = - 700 Oe and variable I (lower
left) and for fixed I = 17 mA and variable H (lower right).

38

Py30nm/Cul0/Py6

I(mA)

10 15 20

   
  
     
 

I(mA)
-5 0 5 1015 20

   
  
   
 
    

R
-1 50 ((0)0620!)

’3’
O
'100 0.04408 {-100 13935-04

Integrated

Power (dB)
-1 50 2.120904

H (De)

0.02204
4990

0.000

 

 

 

 

 

 

 

-3950
20 .
—I0'l :gflAA 20 ——(A) 450009
A” _—(C')14mA ——(B) -6000e
% D°)20mA Q15 —(c) -5600e
:10 — :10 —(D) -2000e
g H=-5600e g
a? 5 o- 5 I=14mA
0. . , . o 2 4 6
o 2 4 6 8 f GHZ
f (G Hzl ( )

Fig. 4.6. GHz radiation data from sample #2b3. Top: Examples of experimental ‘phase
diagrams’ showing MR (upper left) and integrated power (upper right) as functions of H
and 1. Bottom: Examples of GHz peaks for fixed H = - 560 Oe and variable I (lower
left) and for fixed I = 14 mA and variable H (lower right).

39

Comparison

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1o: I-17mA [—-19200e]
Sample #1: 2.03 - —-.--.,. - .
m1o.‘ _.
H/H,~700/270~2.60 330; , , 21, .5 72°81]
£13{ I [—-7000e]
l/Al ~17/7.1~2.39 ‘1 1 "’
c 10: |——-2600—e]
0 ~ :~- . 2~4
o 4 8
flGHzl
j 1-14mA _
Sample #2: 13‘ g m I _2-1500C.)e.]

 

 

 

$101 I I |-—-5600e]
l/Alc ~ 1445.8 ~ 2.41 O 0.

 

 

 

 

 

 

 

n- I
1°? [—-2000e]
017' 2:84 is? ”8‘
fIGHZl

Fig. 4.7. Scaling Plots: Power versus frequency plots comparing spectra for sample #1
and #2 at normalized values of H/Hc and I/AIc.

40

4.4. System for measuring GHz radiation up to 40 GHz and Time Domain
studies.

We conclude this thesis by describing the system for measuring GHz radiation up to
40 GHz and for time domain studies, that Mustafa designed and constructed jointly with
Irinel Chiorescu. As illustrated schematically in Fig. 4.8, for simple GHz studies, the
system consists of: a microwave (MW) generator up to 40 GHz, a mixer, low-bandwidth
amplifier, and a detector. The input signal from the sample is mixed with that from the
MW generator, and the frequency of the latter is adjusted to obtain a difference signal at
low enough frequency to be processed by the low bandwidth amplifier (~ 100 MHz) and
a detector. As shown in Fig. 4.9, the detector could be an Acqiris oscilloscope operating

in a Fast Fourier Transform (FF T) mode.

 

MW Generator up to
4OGHz

+40 68

Input signal > Q Mixer
Low-bandwidth
amplifier

Detector

 

 

 

 

 

 

 

 

 

 

 

Fig. 4.8. Schematic of the system for studying current-generated signals to 40 GHz.

41

Fig. 4.9 shows details of the system of Fig. 4.8, along with the additional components
needed for time domain studies. For simplicity, we have omitted GPIB connections that
allow the computer to control the current to the electromagnet and the Anritsu 0.01-40
GHz oscillator, and also to the dc pulser and the Acqiris oscilloscope. Fig. 4.8, the
frequency-spectrum mode, is obtained by setting the switches in Fig. 4.9 to position (1)
on the right hand side and to position (2a) on the left hand side. The right-hand
picoprobe is used to measure either the sample resistance (V/I) using a dc current, or
dV/dI using a low frequency ac current and a lock-in amplifier, and also to insert the
large dc current needed to generate the spin-torque and to carry out to the mixer the
excited GHz radiation.

Contacting the sample with the left hand picoprobe, allows injection of a do or
microwave current pulse for time domain studies. In this case, the switches in Fig. 4.9
are set to position 2 on the right hand side and to either 2a or 2b on the left hand side.
Such studies start with the computer sending a trigger signal to the Acqiris oscilloscope
that in turn sends a trigger signal to the dc pulser (shown in Fig. 4.10). The output of the
dc pulser is either sent through the bias-T (lefi switches in the ‘2a’ position) as a current
pulse to the left picoprobe, or directly (left switches in the ‘2b’ position) to the
microwave (MW) pulser (shown in Fig. 4.11). With the right switches in the ‘2’ position,
the transient output of the high bandwidth amplifier goes directly to the Acqiris
oscilloscope and the continuous output of the Anritsu oscillator is connected to the MW
pulser. The use of the Acqiris oscillator for both triggering and transient-signal

acquisition generates less time ‘jitter’ between the trigger start and the display time-

42

sweep of the oscilloscope. Together, the Acqiris oscilloscope bandwith of ZGHz, and

sampling rate of 2G-samp1es/sec, are adequate for nanosecond time scale experiments.

 

low-frequency ac

Voltmeter

sampl' [I _

I I “I " ,

I a
e. [Ix I' .. '

dc currents

 

' do
I
pulser Q ,
L - - . - _ ' \ J/ .
picoprobe -

trigger In
~ 3

 

 
 

 

       

 

 

 

 

trigger out
Acqlrle
. 0861314011! mum, IRmeer RF
trloeorgin - (lowbandwldth) ‘_ _____ . 1F L0
. ‘ FFpultO out J .puleer r—Q)
TL ‘

 

 

Fig. 4.9. Details of the system for both frequency studies to 40 GHz and time domain
studies.

43

Stanford Research Systems (SR5)

 

SRS 06535 Pulsar

 
 
     
      
  

v”...
Mini-Circuits _
SR8 04A “3 a“ of .
mam sw-sswz
comprosso.’

  
   
  
     
 

Mini-Circuits
attenuuors of X db
Mini-Circtits

2FRSC-42 splitter
6 dB insertion loss
1

    

dc pulse out I V,,,..,.' attenuued by 6+X+6+Y dB
attenuator of Y db

Fig. 4.10. DC Pulse Generator.

Fig. 4.10 shows the circuit for generating dc current-pulses. A voltage pulse
(“Vpulse”) of desired duration leaves the AB-channel output and enters the leading-edge
compressor, giving 0.1-ns rise time and ~ 2-ns fall time. The signal passes through an
attenuator of X dB and is combined at the splitter/combiner with the signal from the CD-
channel output, which has been attenuated by the same amount as the AB signal. The
edge compressor requires the AB channel output to be offset by ~ -1.65 V. The

amplitude of the pulse on channel AB is 0.9V < Vpuls, < 2.1 V. To compensate the effect

of the offset on the final pulse, the offset of channel CD is tuned to + 1.6 V. Channel CD
does not emit a pulse. Thus the splitter/combiner puts out a dc pulse with zero-offset that

can then be attenuated by YdB.

Markl-Mlcrowave Mlxers

   
  

attenuators

  
 

1.65 (3112
low -pass filters

   

Mini-Circuits
A

 

Fig. 4.11. Microwave (MW) Pulse Generator.

Fig. 4.11 shows the diagram for generating MW pulses. The output from the dc pulse
circuit is divided by the splitter and fed via the attenuators and 1.65-GHz low-pass filters
to two mixers. The mixers amplitude-modulate the continuous MW signal coming from
the Anritzu oscillatorg. These particular mixers operate at < 26 GHz, but operating

frequencies up to 50 GHz are available from this manufacturer. In the off state, a mixer

45

will still pass some MW signal—the on/off-ratio (ION/OFF) is not infinite. To reduce this
problem, two mixers are placed in series so that the overall on/off ratio is now (ION/OFF)2.
The low-pass filters have a cutoff frequency fc = 1.65 GHz, and the value of fc obeys the
following inequality: (21t'c)'l < fc < wa where T is the desired rise time of the dc pulse
and flew is the frequency of the MW signal. The attenuators should be set as high as

possible, consistent with the mixers being able to open.

46

 

5. Summary and Conclusions.

This thesis contains results of studies of two related phenomena: (A) Current-Induced
Magnetization Switching (CIMS), and (B) Current-Induced Generation of GHz
Radiation.

(A) The CIMS part focuses upon two papers that report results of measurements of
current-perpendicular-to-plane (CPP) magnetoresistance (MR) and CIMS behavior in
Ferromagnetic/Non-metal/Ferromagnetic (F1/N/F2) nanopillars. It was shown that
judicious combinations of F 1 and F2 metals with different bulk scattering asymmetries,
and with FUN and N/F 2 interfaces having different interfacial scattering asymmetries,
could ‘invert’ both the CPP-MR and the CIMS, controllably and independently. In
‘normal’ CPP-MR, R(AP) > R(P), where R(AP) and R(P) are the resistances in the anti-
parallel (AP) and parallel (P) orientations of the F1 and F2 magnetic moments. In
‘norrnal’ CIMS, positive current switches the nanopillar from the P to the AP state. In
‘inverse’ CPP-MR, R(P) > R(AP). In ‘inverse’ CIMS, positive current switches the
nanopillar from the AP to the P state. All four possible combinations of CPP-MR and
CIMS—(a) ‘normal’-‘normal’, (b) ‘normaI’-‘inverse’, ‘inverse’-‘normal’, and (d)
‘inverse’-‘inverse’ were demonstrated and explained with carefully chosen combinations
of F1, N, and F2. These results ruled out the self-Oersted field as a potential source of
the switching, since the direction of the self-Oersted field does not depend upon the bulk
or interfacial scattering asymmetries. By using carefully chosen impurities to reverse the
bulk scattering asymmetry, the data showed also the importance of scattering off of
impurities within the bulk F 1 and F2 metals—i.e. that the transport must be treated as

‘diffusive’ rather than ‘ballistic’.

47

(B) The GHz studies consisted of five parts: (a) designing an appropriate sample
geometry; (b) fabricating nanopillar samples in this geometry; (c) constructing a system
for measurements up to 12 GHz and using it to obtain current-driven GHz radiation data;
((1) providing a justification for ‘scaling’ behavior of GHz data for two nominally
identical (but in practice subtlely different) samples, and showing an example of such
scaling; and (e) designing and constructing a system for measurements up to 40 GHz and

for time domain studies.

48

BIBLIOGRAPHY

49

 

[1] J. Slonczewski, J. Magn. Magn. Mater. 159, L1 (1996).
[2] J. Slonczewski, J. Magn. Magn. Mat.. 247, 324 (2002).
[3] L. Berger, Phys. Rev. B54, 9353 (1996).

[4] L. Berger, J. Appl. Phys. 89, 5521 (2001).

[5] M. Tsoi, A.G.M. Jansen, J. Bass, W.-C. Chiang, M. Seck, V. Tsoi, P.W. Wyder,
Phys. Rev. Lett. 80, 4281 (1998); 81 493 (E) (1998)

[6] M. Tsoi, A.G.M. Jansen, J. Bass, W.-C. Chiang, V. Tsoi, P.W. Wyder, Nature
(London) 406, 46 (2000.

[7] EB. Myers, D.C. Ralph, J .A. Katine, R.N. Louie, R.A. Buhrman,., Science 285, 867
(1999)

[8] Ya. B. Bazaliy, B.A. Jones S.C. Zhang, J. Appl. Phys. 89, 6793 (2001).

[9] A. Brataas, Y. V. Nazarov, G.E.W. Bauer, Phys. Rev. Lett. 84, 2481 (2000).
[10] X. Waintal and PW. Brouwer, Phys. Rev. B63, 220407 (2001).

[11] C. Heide, Phys. Rev. B65, 054401 (2001).

[12] MD. Stiles and A. Zangwill; Phys. Rev. B66, 014407 (2002).

[13] S. Zhang, P.M. Levy, A. Fert, Phys. Rev. Lett. 88, 236601 (2002).

[14] J .-E. Wegrowe, Phys. Rev. B62, 1067 (2000).

[15] J.A. Katine, F .J. Albert, R.A. Buhrman, E.B. Myers, D.C. Ralph, Phys. Rev. Lett.
84, 3149 (2000).

[16] F.J. Albert, J.A. Katine, R.A. Buhrman, D.C. Ralph, Appl. Phys. Lett. 77, 3809
(2000)

[17] RB. Mancoff and SE. Russek, IEEE Trans. Magn. 38, 2853 (2002).

[18] J.Z. Sun, D.J. Monsma, D.W. Abraham, M.J. Rooks, R.H. Koch, Appl. Phys. Lett.
81, 2202 (2002).

[19] EB. Myers, F.J. Albert, J.C. Sankey, E. Bonet, R.A. Buhrman, D.C. Ralph, Phys.
Rev. Lett. 89, 196801 (2002).

[20] F .J. Albert, N.C. Emley, E.B. Myers, D.C. Ralph, R.A. Buhrman, Phys. Rev. Lett.
89, 226802 (2002).

50

 

[21] Y. Ji, C.L. Chien, and MD. Stiles, Phys. Rev. Lett. 90, 106601 (2003).
[22] W.H. Rippard, M.R. Pufall, T.J. Silva, Appl. Phys. Lett. 82, 1260 (2003).

[23] J. Grollier, V. Cros, A. Hamzic, J.M. George, H. Jaffres, A. Fert, G. Faini, J. Ben
Youssef, H. Le Gall, Appl. Phys. Lett 78, 3663 (2001).

[24] MR. Pufall, W.H. Rippard, T.J. Silva, Appl. Phys. Lett. 83, 323 (2003).

[25] B. Oezyilmaz, A.D. Kent, D. Monsma, J .Z. Sun, M.J. Rooks, R.H. Koch, Phys. Rev.
Lett. 91, 067203 (2003).

[26] S. Urazhdin, N.O. Birge, W.P. Pratt Jr., J. Bass, Phys. Rev. Lett. 91, 146803 (2003)
[27] S. Urazhdin, H. Kurt, W.P. Pratt Jr., J. Bass, Appl. Phys. Lett. 83, 114 (2003).

[28] J. Bass and W.P. Pratt Jr., J. Magn. Magn. Mat. 200, 274 (1999); Erratum 296, 65
(2006).

[29] C. Vouille, A. Barthelemy, F. Elokan Mpondo, A. F ert, P.A. Schroeder, S.Y. Hsu,
A. Reilly, R. Loloee, Phys. Rev. B60, 6710 (1999)

[30] I. Mertig, R. Zeller, and PH. Dederichs, Phys. Rev. B47, 16178 (1993).

[31] LA. Campbell and A. Fert, Ferromagnetic Materials, 3, ER Wolforth, Ed., North-
Holland, 1982, Pg. 747.

[32] MD. Stiles and DR. Penn, Phys. Rev. B61, 3200 (2000).

[33] K. Xia, P.J. Kelly, G.E.W. Bauer, 1. Turek, J. Kudmovsky, V. Drchal, Phys. Rev.
B63, 064407 (2001).

[34] J. Grollier, V. Cros, H. Jaffres, M.A. Hamzic, J .M. George, G. Faini, J. Ben Youssef,
H. Le Gall, A. F ert, Phys. Rev. B67, 174402 (2003).

[3 5] For additional experimental references see J. Bass, S. Urazhdin, N.O. Birge, W.P.
Pratt Jr. Phys. Stat Sol. (a) 201, No. 7, 1379 (2004).

[36] Ya. B. Bazaliy, B.A. Jones, Shou-Cheng Zhang, Phys. Rev. B57, R3213 (1998).
[37] X. Waintal, E.B. Myers, P.W. Brouwer, D.C. Ralph, Phys. Rev. B62, 12317 (2000).
[3 8] C. Heide, P.E. Zilberman, R.J. Elliott, Phys. Rev. B63, 064424 (2001)

[39] C. Heide, Phys. Rev. Lett. 87, 197201 (2001).

[40] Y. Tserkovnyak, A. Brataas, G.E.W. Bauer., Phys. Rev. B66, 224403 (2002).

[41] MD. Stiles and A. Zangwill, J. Appl. Phys. 91, 6812 (2002)

51

 

[42] J .-E. Wegrowe, D. Kelly, Y. Jaccard, Ph. Guittienne, and J. Ph. Ansermet, Europhys.
Lett. 45, 626 (1999);

[43] J .-E. Wegrowe, X. Hoffer, Ph. Guittienne, A. Fabian, L. Gravier, T. Wade, J. Ph.
Ansermet, J. Appl. Phys. 91, 6806 (2002).

[44] S. Urazhdin, Phys. Rev. B69 134430 (2004).

[45] A. Fert, V. Cros, J .-M. George, J. Grollier, H. Jaffres, A. Hamzic, A. Vaures, G.
Faini, J. Ban Youssef, H. Le Gall, J. Mag. Mag. Mat. 272-276, 1706 (2004).

[46] M. AlHajDarwish, A. Fert, W.P. Pratt Jr., J. Bass, J. Appl. Phys. 95, 6771 (2004).

[47] A. Zambano, E. Eid. R. Loloee, W.P. Pratt Jr., J. Bass, J. Magn. Magn. Mat. 253, 51
(2002)

[48] I. Mertig, P. Zahn, M. Richter, H. Eschrig, R. Zeller, P.H. Dederichs, J. Magn.
Magn. Mat.. 151, 363 (1995).

[49] J. Bamas, A. Fert, M. Gmitra, I. Weymann, V.K. Dugaev, Phys. Rev. B72, 024426
(2005).

[50] W. Park, R. Loloee, J .A. Caballero, W.P. Pratt Jr., P.A. Schroeder, J. Bass, A. Fert,
C. Vouille, J. Appl. Phys. 85, 4542 (1999).

[51] T. Valet and A. Fert, Phys. Rev. B48, 7099 (1993).

[52] 8.1. Kiselev, J .C. Sankey, I. N. Krivorotov, N.DC. Emley, R.J. Schoelkopf, R.A.
Buhrman, D.C. Ralph, Nature 425, 380 (2003).

[53] W.H. Rippard, M.R. Pufall, S. Kaka, S.E. Russek, T.J. Silva, Phys. Rev. Lett. 92,
027201(2004)

[54] M. Covington, M. AlHajDarwish, Y, Ding, NJ. Gokemeijer, and MA. Seigler, Phys.
Rev. B69, 184406 (2004).

[55] Spin Transfer Effects in Current perpendicular to the Plane Spin Valves, M.
Covington, M. AlHajDarwish, Y. Ding, A. Rebei, G.J. Parker, N.J. Gokemeijer, and
MA. Seigler, J. Magn. Magn. Mat. 287, 325 (2005).

[56] J .Z. Sun, Phys. Rev. B62, 570 (2000).

[57] O. Boulle, V. Cros, J. Grollier, L.G. Pereira, C. Deranlot, F. Petroff, G. Faini, J.
Bamas, A. Fert, Phys. Rev. B77, 174403 (2008).

[58] DC. Ralph and MD. Stiles, J. Magn. Magn. Mat. 320, 1190 (2008).

52

 

 

   

ll[lulljjflllflljj[[1]]I