Vous êtes sur la page 1sur 51

Center for Superconductivity Research

Department of Physics
University of Maryland
Samir Garzon
Richard Webb
Spin injection and detection in
Cu spin valve structures
Motivation-Devices
Spin transport in Cu wires
Spintronic applications:
Magnetic reading heads
Spin valve/GMR
Fujitsu
(CIP: Current in plane)
Mainstream technology which
allows Gbyte hard drive storage
MTJ
F
1

F
2

tunnel barrier
IBM
MRAM (nonvolatile)
(CPP: Current perpendicular to plane)
Under design
Motivation-Science
Spin transport in Cu wires
Spectroscopic tool to study:
symmetry properties of unconventional superconductors
excitations in the quantum Hall regime
spin-charge separation in non-Fermi liquids
[1,2]
[3,4]
[5,6]
[1] Vasko et. al, PRL 78, 1134 (1997)
[2] Ngai et. al, APL 84, 1907 (2004)
[3] MacDonald et. al, PRL 83, 3262 (1999)
[4] Chan et. al, PRL 83, 3258 (1999)
[5] Si et. al, PRL 81, 3191 (1998)
[6] Balents et. al, PRL 85, 3464 (2000)
Johnson and Silsbee, PRL 55, 1790 (1985) pioneered field spin injection
Produce and detect spin-polarized currents
Our research:
Spin transport in Cu wires
Find spin relaxation length and understand the
mechanisms that are responsible
Study a nonlocal geometry different from that used for
GMR and MTJ applications
Gain understanding of interfacial spin transport
Understand the high temperature behavior of spin
injection and detection
Outline
Spin transport in Cu wires
Theory
Relaxation mechanisms
Spin relaxation time scales
Spin transport
Spin transport in Cu wires
Theory
Measurements
Switching of ferromagnetic contacts
Spin precession: Hanle effect
Temperature dependence of spin signal
Outline
Sample fabrication
Outline
Spin transport in Cu wires
Measurements
Cross checks
Discussion
Analysis of temperature dependence of spin signal
Evidence for interfacial spin-flip scattering
Theory
Sample fabrication
Outline
Spin transport in Cu wires
Measurements
Discussion
Conclusions (overview)
Further work
Theory
Sample fabrication
Time scales
Spin Transport in Cu wires
e s
t t <
Interactions that
conserve total spin
s
t
Single electron spin coherence time
(quantum computing)
e
t
Electron ensemble spin coherence time
(spintronics)
p
t
Momentum relaxation time
t e
t we will represent simply as
In metals:
*
2 2 1
T T T
e
= = = =t
t
Spin relaxation mechanisms
Spin Transport in Cu wires
Spin dependent Magnetic impurity scattering
(nonmagnetic metals)
magnetic materials: magnons
interfaces: surface magnons
enhanced magnetic scattering
Spin relaxation mechanisms
Spin Transport in Cu wires
Spin dependent Magnetic impurity scattering
high T
(293K)
low T
(4.2K)
(nonmagnetic metals)
Spin independent
due to spin orbit
coupling
+ + | | c
so
| + + + c
so
(Elliot-Yafet)
phonons
impurities
boundaries
so
contribute to spin relaxation hence
p
p
a =
t
t
+ + + | | = + | H H H
so so
*
c c
hence
Transition
amplitude to
=
Spin transport: F-N junction
Spin Transport in Cu wires
No current, (M
N
=0)
Current ON, (finite M
N
)
Spin valves
Spin Transport in Cu wires
F
1:
Ferromagnet 1 F
2:
Ferromagnet 2
1
m

2
m

Insulator or
nonmagnetic
Aligned magnetizations
or
Anti-aligned magnetizations
or
Spin valves
Spin Transport in Cu wires
Aligned magnetizations
or
Anti-aligned magnetizations
or
CPP
Current perp. to plane
I I
CHANGE IN RESISTANCE BETWEEN ALIGNED
AND ANTI-ALIGNED CONFIGURATIONS
I
I
CIP
Current in-plane
I is the current
TMR vs Nonlocal Geometry
Spin Transport in Cu wires
Tunnelling Magneto Resistance
I
V
R
TMR
=
F
1
F
2

insulator
I
N
1
N
2

V
No charge current at F
2
!
Charge current at both F1 and F2
Nonlocal configuration
I
V
I
V
R
NL
=
F1 and F2 are equivalent F1 and F2 are not identical
In the absence of spin effects, 0 =
TMR
R
In the absence of spin effects, 0 =
NL
R
Electrochemical potential
Continuous electrochemical potential
(transparent interfaces)
Nonlocal Geometry
Spin Transport in Cu wires
Nonlocal configuration
I
V
1 1
2 2
N T
N T
NL
I
V V
R

=
||
2
T
eV
|+
2
T
eV
N
eV
As opposed to TMR,
in the absence of spin effects
0 =
NL
R
Nonlocal Geometry
Spin Transport in Cu wires
Nonlocal configuration
I
V
1 1
2 2
N T
N T
NL
I
V V
R

=
Electrochemical potential
||
2
T
eV
|+
2
T
eV
N
eV
As opposed to TMR,
in the absence of spin effects
0 =
NL
R
spin dependent
interface resistance
+ =|, o
( ) e I R
F N
o o o o
/ =
Spin transport in Cu wires
Sample fabrication
2 levels of standard
e-beam lithography
ion-milling
thermal evaporation
lift-off
1
st
level: Co
Al
2
O
3
tunnel barrier
2
nd
level: Cu
Spin transport in Cu wires
Samples
Bonding pads
Alignment and
focusing spot

Cu
F
1
(Co)
F
2
(Co)
Al
2
O
3

Injector
Detector
100 nm
Spin transport
Spin Transport in Cu wires
2 current model
I
o
,


,+


for o =


current electrochemical potential
Spin transport
Spin Transport in Cu wires
2 current model
I
o
,


,+


for o =


linearized T=0
Boltzmann equation
current
density
electric
field
conductivity
diffusion
constant
0
) (
o o
n r n

electron
density
Spin transport
Spin Transport in Cu wires
linearized T=0
Boltzmann equation
2 current model
I
o
,


,+


for o =


diffusion eqns.
o
t
scattering
time

relaxation
(diffusion)
length
modified continuity
equation
.
conserved I
o


boundary conds.
( ) e I R
F N
o o o o
/ =
Spin transport in Cu wires
Spin transport in Cu wires
I
V
In the limit of R
o
>>
,
resistivity
relaxation length
A cross sectional area
resistance over a
spin relaxation length
Spin transport in Cu wires
In the limit of R
o
>>
,
Spin transport in Cu wires
I
V
where
N
is the spin diffusion length
can be interpreted as the single
junction spin current polarization
at interface i defined as:
+ |
+ |
+

=
I I
I I
P
i
Outline
Spin transport in Cu wires
Measurements
Switching of ferromagnetic contacts
Spin precession: Hanle effect
Temperature dependence of spin signal
Theory
Sample fabrication
Spin transport in Cu wires
I
V
T=4.2 K
Characteristic switching
Spin transport in Cu wires
I
V

Larmor frequency
Hanle effect (spin precession)
T=4.2 K
Spin transport in Cu wires
Temperature dependence
should have equal magnitude
but different sign for and
|| |+
NL
R
Spin transport in Cu wires
Temperature dependence
should have equal magnitude
but different sign for and
|| |+
NL
R
Outline
Spin transport in Cu wires
Measurements
Cross checks
Discussion
Analysis of temperature dependence of spin signal
Evidence for interfacial spin-flip scattering
Theory
Sample fabrication
Spin transport in Cu wires
T dependence: analysis
( ) 2
,
|+ ||
=
NL NL A S
R R R
2 1
2
) exp(
P P
A
L
R
N
N N N
NL

=
|+ ||
=
1 1
P P
A S
N
N N N
A S
P P
A
L
R
, 1 ,
2
) exp(
=
( ) 2
2 2 ,
|+ ||
= P P P
A S
where
Previous experiments could only measure R
A
since R
TMR

included a large offset not related to spin injection
???
Spin transport in Cu wires
T dependence: analysis
( ) 2
,
|+ ||
=
NL NL A S
R R R
A S
N
N N N
A S
P P
A
L
R
, 1 ,
2
) exp(
=
N
p
p
p
ne a
D m
a
D
D

t t
1
2
*
= =
is unknown (T>>4.2K)
p
a
( )
2 / 3
0
1 T P q
Magnons
(q unknown)
) 1 (
2 / 3
0
T P P
A
q =
Since R
A
is the previously observed
signal,
Fit R
A
with parameters q and
p
a
samir:
what about temp
dep of D?
Spin transport in Cu wires
T dependence: fit

A S
N
N N N
A S
P P
A
L
R
, 1 ,
2
) exp(
=
* Nature 410, 345 (2001)
p
K
p
a a
2 . 4
4
06 . 1 10 6 . 6 = =

4
2 . 4
10 9 . 6

=
p
K
a
Jedema et al.*:


2 / 3 5
10 4 . 8

= K q
larger than bulk, as expected due to
interface effects (Shang et al.**)
** PRB 58, R2917 (1998)
known T dep.
Spin transport in Cu wires
T dependence: P
A
, P
S
4
10 6 . 6

=
p
a
2 / 3 5
10 4 . 8

= K q
,
A S
N
N N N
A S
P P
A
L
R
, 1 ,
2
) exp(
=
known T dep.
1
,
,
2
) exp(
P
A
L
R
P
N
N N N
A S
A S

=
Spin transport in Cu wires
T dependence: P
A
, P
S
A S
N
N N N
A S
P P
A
L
R
, 1 ,
2
) exp(
=
4
10 6 . 6

=
p
a
2 / 3 5
10 4 . 8

= K q
,
1
,
,
2
) exp(
P
A
L
R
P
N
N N N
A S
A S

=
known T dep.
) 1 (
2 / 3
0
T P P
A
q =
T K
S
e
B
P
/ 1227
1+
= Thermal activation:
Spin transport in Cu wires
Can we explain this?

2 1
2
) exp(
P P
A
L
R
N
N N N
NL

=
( )
( )
+ |
+ |
+

=
2 2
2 2
2
/ 1 / 1
/ 1 / 1
R R
R R
P
N F2
| 2
R
+ 2
R
|
+
( )
| | | |
=
2
2
/ R e I
F N

similarly for spin
+
A S
N
N N N
A S
P P
A
L
R
, 1 ,
2
) exp(
=
( ) 2
2 2 ,
|+ ||
= P P P
A S
where
Spin transport in Cu wires
Can we explain this?

A S
N
N N N
A S
P P
A
L
R
, 1 ,
2
) exp(
=
( ) 2
2 2 ,
|+ ||
= P P P
A S
where
interface spin-flip
scattering
( )
( )
+ |
+ |
+ +
+
2 2
2 2
' / 1 ' / 1
' / 1 ' / 1
R R
R R
N F2
| 2
R
+ 2
R
| 2
' R
+ 2
' R
|
+
( )
| + |
+
2
2
' / R
F N

similarly for spin
+
( )
| | | |
=
2
2
/ R e I
F N

2 1
2
) exp(
P P
A
L
R
N
N N N
NL

=
( )
( )
+ |
+ |
+

=
2 2
2 2
2
/ 1 / 1
/ 1 / 1
R R
R R
P
Spin transport in Cu wires
Physical meaning

( )
( ) ( )
+ | + |
+ |
+ + +

=
2 2 2 2
2 2
' / 1 ' / 1 / 1 / 1
/ 1 / 1
R R R R
R R
P
A
( )
( ) ( )
+ | + |
+ |
+ + +

=
2 2 2 2
2 2
' / 1 ' / 1 / 1 / 1
' / 1 ' / 1
R R R R
R R
P
S
What do P
A
and P
S
mean?
P
A
: ratio of spin-polarized to
total transport into F
2
P
S
: ratio of spin-dependent
spin-flip to total transport
into F
2

Spin transport in Cu wires
Temperature dependence conclusions:

Temperature dependent measurements reveal a new
component of the nonlocal resistance.
Data analysis shows that the new component can be fit
well with a model related to temperature activation with
T=1227 K (surface magnons?, enhanced magnetic
scattering?).
The existence of the new signal is explained by extending
the previous model to include interface spin-flip scattering.
A physical interpration of the new defined quantities P
S

and P
A
is given. P
s
gives information on the differential
spin-flip scattering at the detector, while P
A
describes the
differential spin conserving transport.
The different character of injector and detector is clear in
the nonlocal geometry.
Spin transport in Cu wires
Cross checks: temperature dependence
Co-Cu-Co
Ti-Cu-Ti
Spin transport in Cu wires
Cross checks: length dependence
Py-Cu-Py
Py-Cu-Py
Cu-Cu-Cu
Py-Cu-Cu
Cu-Cu-Py
Spin transport in Cu wires
Conclusions (overview)
Performed comprehensive spin injection and detection
experiments in Cu-Co spin valves.
Measured spin precession in Cu and extracted the spin
diffusion length and the current spin polarization P.
Spin diffusion length measurements are consistent with each
other and with previous measurements.
Found a temperature dependent symmetric component in the
nonlocal resistance R
S
that is consistent with the hypothesis of
interface spin-flip scattering.
Made various cross checks to make sure R
S
did not come
from capacitive leakage, electrostatic geometric effects, and
heating combined with thermoelectric effect.
Spin transport in Cu wires
Further work
Study the microscopic origin of interfacial spin-flip
scattering.

Nonlocal measurements with MgO tunnel barriers, which has
been shown to enhance the magnetoresistance in MTJ, should
be used for comparison.
For device applications, high frequency measurements of
spin injection and detection might be of importance.
A direct measurement of the spin polarization and relaxation
lengths, not requiring a transport model for data interpration
can be useful (MFM).
Use electron statistics such as shot noise to further study
interfacial spin transport, even in the absence of charge current.
Spin transport in Cu wires
Experimental setup
Lock-in
amplifier
Lock-in
amplifier
Ramp
generator
Current
amplifier
osc. out
ref. in
sig. out
in
osc. out
out
Spin transport in Cu wires
Cross checks: Geometric effects
Spin transport in Cu wires
Cross checks: heating and Seebeck effect

S
V
DC rejection of lock-in
<1nV
2e rejection<1nV
DC
DC
( ) O m
I
V
AC
S
Spin transport in Cu wires
Why heating and Seebeck effect?
Spin transport in Cu wires
Current leakage?
Time scales
Spin Transport in Cu wires
s
t
Single electron spin coherence time
e
t
Electron ensemble spin coherence time
*
2
T
Reversible spin
dephasing time
e s
t t <
Interactions that
conserve total spin
1
T
Spin relaxation time
spin-lattice
longitudinal
2
T
Irreversible spin
decoherence time
transverse
1
0
T
M M
dt
dM
z z

=
2
T
M
dt
dM
x x
=
2
T
M
dt
dM
y y
=
Blochs phenomenological equations
x
y
z
B
2 1
T T
e
= = t
B~0
isotropic system
*
2 2
T T =
conduction electrons:
p
t
Momentum relaxation time
What does small B mean?
Spin Transport in Cu wires
1 <<
c L
t e
where
/

= B g
B L
e is the Larmor frequency
and
c
t
is the correlation time
(typical time interval between spin changing events)
Conductivity mismatch
Spin Transport in Cu wires
I I
|
I
+
I
|
R
+
R
F N
N N
o / 2
N N
o / 2
|
o /
F
+
o /
F
T

Vous aimerez peut-être aussi