Exchange Bias: Interface vs. Bulk
Magnetism
Miyeon Cheon
Hongtao Shi
Zhongyuan Liu
Jorge Espinosa
David Lederman
Hendrik Ohldag
Joachim Stöhr
Elke Arenholz
Department of Physics
Optical and Vibrational Spectroscopies Symposium:
A Tribute to Manuel Cardona
August 20, 2010
Exchange Bias
6
MR: “Remanent” magnetization
- Maximum value of M
- Depends on FM
2
-4
m (10 emu)
4
0
HC
HC: Coercivity
- Depends on FM magnetic
anisotropy
- Represents energy
required to reverse
magnetic domain
MR
-2
HE
-4
-6
-1.0
-0.5
0.0
H (kOe)
FM
AF
0.5
1.0
HE: Exchange Bias
-Absent in pure FM,
results from AF-FM
interaction
Application: Magnetic
Tunnel Junction /GMR
Sensors
Free magnetic layer
(analyzes electron spin)
Ferromagnetic layers
~1.0-5.0 nm thick
Pinned magnetic layer
(polarizes tunneling
electrons)
Pinning Antiferromagnet 1 - 100 mm

Insulator/NM Metal
~1.0-2.5 nm
Antiferromagnet
~10 - 50 nm
R  Ro  R cos 
Albert Fert & Peter Grünberg
2007 Nobel Prize in Physics
“for the discovery of Giant Magnetoresistance”
How does the pinning of bottom FM layer work?
(www.research.ibm.com)
Key Questions
• Given that:
– All EB models require presence of uncompensated magnetization in
the antiferromagnet (interface)
– Details of EB behavior (e.g. temperature dependence, magnitude)
depend strongly on AF anisotropy (bulk)
• Some key questions are:
– Can uncompensated moments in the AF be detected?
– Can the effects of uncompensated moments in the AF be studied
systematically?
– Can the magnetic anisotropy be studied systematically?
MF2 Antiferromagnets
NiF2
• Rutile structure (a = 0.4651 nm, c = 0.3084 nm)
• Antiferromagnetic, TN= 73 K
• Weak ferromagnetic
[001]
• Magnetization lies in the a-b plane
weak anisotropy antiferromagnet
FeF2
• Rutile structure (a = 0.4704 nm, c = 0.3306 nm)
• Antiferromagnetic, TN=78 K
[001]
• Magnetization along the c-axis
dilute antiferromagnet
ZnF2
• Rutile structure (a = 0.4711 nm, c = 0.3132 nm)
• non-magnetic
[001]
So… where does Manuel Cardona
fit in?
Naïve graduate student asks: can
antiferromagnetic superlattice
magnons be observed?
Two-magnon Raman line for 1.3 mm FeF2 thin film
Growth and Characterization
• MBE co-deposition of FeF2 (e-beam) and ZnF2, NiF2 (Kcell), Pbase = 7 x 10-10 Torr, Pgrowth < 4 x 10-8 Torr
• TS (AF) = 297 0C, poly-Co @125 0C, poly-MgF2 @RT
• Growth along (110)
• Twin sample holder – simultaneous growth of
underlayer, different overlayers
• In-situ RHEED, AFM
• X-ray diffraction and reflectivity
• Cooling field (HCF = 2 kOe) in the film
plane along the c-axis of FexZn1-xF2
• M vs H via SQUID magnetometer,
horizontal sample rotator
Key Questions
• Can uncompensated moments in the AF be detected?
• Can the effects of uncompensated moments in the AF be
studied systematically?
• Can the magnetic anisotropy be studied systematically?
Magnetic Dichroism in X-ray Absorption
6
X-ray magnetic circular dichroism
 sensitive to FM order.
4
e
-
e
-
2
0
700
710
720
730
Photon Energy (eV)
-
e
-
e
Antiferromagnetic Domains
X-ray magnetic linear dichroism
 sensitive to AF order.
NiO L2a, L2b
300
Electron Yield (a.u.)
Electron Yield (a.u.)
Fe L3, L2
200
100
876
879
882
Photon Energ y (eV)
Element specific technique sensitive to antiferromagnetic as well as ferromagnetic order.
Antiferromagnetic Order of FeF2(110)
FeF2 L2 absorption edge
Electron Yield [a.u.]
Einc || [001]
Co/FeF2(110)
E || [001]
E || [110]
bare FeF2(110)
E || [001]
E || [110]
3
2
Einc || [110]
Fe
1
F
F
Fe
F
Fe
Fe
Fe
0
718
721
F
724
Photon Energy [eV]
Stronger XMLD signal for Co/FeF2(110) compared to bare FeF2(110) indicates an
increase in antiferromagnetic order caused by exchange to the FM Co layer.
Interface Coupling and Exchange Bias
2 nm Pd cap
2.5 nm Co
68 nm FeF2
10
0.1
RT
MgF2(110) sub.
0.0
-10
-0.1
0.5
10
15K
0
Ferromagnet
Fe XMCD [%]
Co XMCD [%]
0
Measy
0.0
-10
-0.5
-3
-2
-1
0
1
2
Mpinned
3
Applied Field [kOe]
Interface
Room T: “Free” uncompensated moments follow FM
Low T: Additional “pinned” uncompensated moments antiparallel to easy direction.
Results
Fe XMCD
Fe M XMCD
Fe XMLD
1.0
0.5
0.5
0.0
0.0
Field [Oe]
Co XMCD HC
Co XMCD HE
400
XMLD [arb. u.]
XMCD [%]
1.0
• Fe in FeF2/Co interface,
despite being non-metallic,
has
– Unpinned magnetization to RT
– Pinned magnetization to TB
– AF order verified to TN via
XMLD
• Co at interface
– TB~TN
– HC peak near TB
200
0
0
40
80
120
300
Temperature [K]
Ohldag et al., PRL 96, 027203 (2006)
1.0
1.0
0.5
0.5
0.0
0.0
0
20
40
60
80
100
120
300
XMLD [a.u.]
XMCD [%]
Parallel Interface Coupling and Exchange Bias
2.) XMCD is
indication of
interfacial magnetic
order at RT.
Temperature [K]
1.) XMLD and long
range AF order
vanish at TN.
Also, see Roy et al, PRL 2006
Related to enhancement of
coercivity for T >> TN
(Grimsditch et al, PRL 2003)
Key Questions
• Can uncompensated moments in the AF be detected?
– Uncompensated moments exist in AF, not due to “metallization”
– Pinned uncompensated moments in AF vanish near TN
– Unpinned uncompensated moments exist up to RT, well above TN
• Can the effects of uncompensated moments in the AF be
studied systematically?
• Can the magnetic anisotropy be studied systematically?
Systems
FexNi1-xF2
FexZn1-xF2
[001]
Dilute antiferromagnet
Systematic study of uncompensated M
[001]
Random anisotropy antiferromagnet
Effects of Dilution
• Domain state model: dilute AF should
make small domain creation easier due
to nonmagnetic impurities (Malozemoff
model)
• Net magnetization of AF domains
should increase effective interface
interaction
Previous Results
Co1-xMgxO/ CoO (0.4 nm) /Co
P. Miltényi, et al., Phys. Rev. Lett., 84, 4224 (2000)
Sample Profile
5 nm MgF2 Cap
5 nm MgF2 Cap
18 nm Cobalt (F)
18 nm Cobalt (F)
65 nm (110)
FexZn1-xF2 (AF)
(110)-MgF2 Sub
1.0 nm FeF2
Pure interface
layer (PIL)
65 nm (110)
FexZn1-xF2 (AF)
(110)-MgF2 Sub
Magnetic interface changes with x in FexZn1-xF2
0
0
-100
-100
With PIL
Without PIL
-200
x = 0.34
HCF = 2 kOe
-300
HE (Oe)
HE (Oe)
HE, HC Dependence on T
TB
-400
-300
-400
400
TB
-500
300
Without PIL
With PIL
300
HC (Oe)
HC (Oe)
x = 0.57
HCF = 2 kOe
-200
200
200
100
100
0
0
10
20
T (K)
30
40
0
0
10
20
30
40
50
T (K)
PIL affects HE, HC; no effect on TB
60
70
80
90
HE, HC vs. Temperature for x = 0.75
300
150
HE (Oe)
0
-150
x = 0.75
HCF = 2 kOe
-300
TB
-450
HC (Oe)
300
200
Without PIL
With PIL
100
• HE changes sign as T increases to TB.
• HC has two peaks corresponding to HE = 0.
• Therefore AF ground state is not unique
0
1.0
MR/MS
0.8
0.6
0.4
0
10
20
30
40
50
T (K)
60
70
80
90
100
TB vs. x in FexZn1-xF2
90
With PIL
Without PIL
Bulk TN
80
70
TB agrees
reasonably well
with bulk TN data
TB (K)
60
50
40
30
20
10
0.2
0.4
0.6
x in FexZn1-xF2
0.8
1.0
Interface Energy Dependence on x
T = 5K
2
E, E/x (erg/cm )
1.2
ΔE = -tCo*HE*MS
1.0
0.8
0.6
E with PIL
E/x no PIL
E no PIL
0.4
0.2
0.2
0.4
0.6
0.8
1.0
x in FexZn1-xF2
• No large HE enhancement observed
• Small AF domains not formed at large x ?
Net AF Magnetization
T = 5K
0.16
1.0
0.20
0.12
x=0.75
0.16
M/MS
0.5
0.12
0.08
0.0
0.08
0.04
-0.5
0.04
x=0.34
x=0.57
0.00
0
10
20
30
40
0
10
20
30
T (K)
40
50
60
70
80
90
T (K)
0.04
M/MS
0.02
0.02
-1.0
M / MS
0.00
1.5
1.0
x=0.34
0.00
0.5
0.00
0
10
20
0.0
x=0.82
x=0.75
-0.02
30
40
50
T (K)
60
70
80
90
-0.02
100
0
10
20
30
40
50
T (K)
60
70
80
90
100
-0.5
M -1.0
-2000
-1000
0
H (Oe)
1000
2000
Key Questions
• Can uncompensated moments in the AF be detected?
– Uncompensated moments exist in AF, not due to “metallization”
– Pinned uncompensated moments in AF vanish near TN
– Unpinned uncompensated moments exist up to RT, well above TN
• Can the effects of uncompensated moments in the AF be
studied systematically?
– Uncompensated M does not necessarily lead to HE enhancement;
critical concentration of impurities must be achieved
– However, uncompensated M dependent on defect concentration
• Can the magnetic anisotropy be studied systematically?
Systems
FexNi1-xF2
FexZn1-xF2
[001]
Dilute antiferromagnet
[001]
Random anisotropy antiferromagnet
Systematic study of AF anisotropy
Magnetic Order
FeF2
 Rutile structure (a = 0.4704 nm, c = 0.3306 nm)
 Antiferromagnet, TN=78 K
 Magnetization along the c-axis
[001]
NiF2
 Rutile structure (a = 0.4651 nm, c = 0.3084 nm)
 Antiferromagnetic, TN= 73 K (80 K in films)
 Weak ferromagnet
[001]
 Magnetization lies in the a-b plane
Growth and measurements
5 nm Al,Pd cap
18 nm Co
MBE Growth
MgF2 (110) substrate
 Growth temperature 210 ˚C
 Fe concentration: 0.0, 0.05, 0.21, 0.49, 0.55 1.0

x=0.0
x=1.0
[001]
[001]
50 nm FexNi(1-x)F2
MgF2(110) sub.
magnetic anisotropy
changes with x.
FexNi1-xF2
Expectations
For nearest neighbor interactions
[001]
2
2
E  J FeFe zx 2 S Fe
cos   J NiNi z (1  x) 2 S Ni
cos   J FeNi zx (1  x) S Ni S Fe cos 
2
2
 DFe xS Fe
cos 2   DNi (1  x) S Ni
cos 2 (   )
For small , there is a critical Fe concentration
xc beyond which spins will lie along the c-axis:


2
DNi S Ni
xc 
2
2
DNi S Ni
 DFe S Fe
For FeF2 and NiF2 xc = 0.14
FeF2/Co
NiF2/Co
49 nm NiF2 / 16 nm Co
3
HCF = 2 kOe
_
H, HCF || NiF2 [110]
5K
0.5
1
MR/MS
-4
m (10 emu)
2
1.0
0
H┴ c
0.0
-1
-0.5
-2
-3
-2
-1
0
1
-1.0
-0.6
2
H (kOe)
T=5K
T = 90 K
-0.4
-0.2
0.0
0.2
0.4
0.6
H (kOe)
9
0
m(10-4emu)
6
HE (Oe)
-100
-200
-300
-400
0
30
60
90
120
T (K)
• Exchange bias along c-axis
• TB ~ 81 K
150
3
5K
75 K
95 K
H || c
0
-3
-6
-9
-2
-1
0
H(kOe)
1
2
• No exchange bias along c-axis
H. Shi et al., Phys. Rev. B 69, 214416 (2004).
Fe0.05Ni0.95F2/Co
6
6
2
4
m(10-4emu)
-4
m(10 emu)
4
5K
20 K
35 K
0
-2
-4
-6
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
H(kOe)
For T ≤ 45 K
• Negative exchange bias along the c-axis
• Asymmetric saturation magnetization
50 K
80 K
2
0
-2
-4
-6
-10 -8 -6 -4 -2 0 2 4 6 8 10
H(kOe)
For 50 K ≤ T ≤ 70 K
• No exchange bias
• Wide hysteresis loop
For 75 K ≤ T
• No exchange bias
6
Large coercivity loops of Fe0.05Ni0.95F2/Co
50 K
55 K
60 K
65 K
70 K
2
10
0
-2
0
-5
-4
-6
-H'
+H'
5
H(kOe)
m(10-4 emu)
4
-10
-10
-5
0
H (kOe)
5
10
50
55
60
65
T(K)
• For 50 K ≤ T ≤ 70 K, large coercivity loops appear for the scanning
field range -10 kOe to 10 kOe.
• Negative exchange bias (HE ~ -500 Oe) for T = 50 K and 55 K
70
Fe0.21Ni0.79F2/Co
4
6
5K
20 K
35 K
m(10 emu)
2
0
-4
-4
m(10 emu)
6
-2
-4
-6
-1.5
50 K
75 K
4
2
0
-2
-4
-6
-1.0
-0.5
0.0
0.5
1.0
1.5
H(kOe)
• Similar behavior to Fe0.05Ni0.95F2/Co
• Negative HE along the c-axis at T≤ 40 K
• Asymmetric saturation magnetization
-10 -8
-6
-4
-2
0
2
4
6
8
H(kOe)
For 45 K ≤ T ≤ 70 K
• No exchange bias effect
• Wide hysteresis loop
For 75 K ≤ T
• HE = 0
10
Large HC loops of Fe0.21Ni0.49F2/Co
6
40 K
50 K
60 K
70 K
-H'
+H'
HE
8
6
2
H (kOe)
-4
m(10 emu)
4
10
0
-2
4
2
0
-2
-4
-6
-4
-8
-6
-10
40
-10
-5
0
5
10
45
50
55
60
65
70
75
T (K)
H(kOe)
• For 40 K ≤ T ≤ 70 K, large HC loops appear for the scanning field range ±10 kOe
• Negative exchange bias effect (HE ~ - 1000 Oe) for 40 K ≤ T ≤55 K
Fe0.49Ni0.51F2/Co
2
0
-2
-4
-6
-1.5
55 K
75 K
6
m(10 emu)
4
5K
10 K
30 K
35 K
-4
-4
m(10 emu)
6
4
2
0
-2
-4
-6
-1.0
-0.5
0.0
0.5
H(kOe)
1.0
1.5
For T ≤ 15 K
• Negative exchange bias
• Asymmetric saturation magnetization
For 25 K ≤ T ≤ 50 K
• Positive exchange bias
• Asymmetric saturation magnetization
-10 -8
-6
-4
-2
0
2
4
H(kOe)
For 50 K ≤ T ≤ 65 K
• No exchange bias
• Wide hysteresis loop
For 70 K ≤ T
• No exchange bias
6
8
10
Large HC loops of Fe0.49Ni0.51F2/Co
5K
10 K
15 K
20 K
4
60
-H'
+H'
HE
40
2
H (kOe)
-4
m(10 emu)
6
0
-2
-4
20
0
+H'
from 10 kOe
-20
-40
-6
0
-60
-40
-20
0
20
40
10
20
60
H(kOe)
• For 5 K ≤ T ≤ 55 K, large HC loops appear for H=± 70 kOe
• Positive exchange bias effect with HE ≥10 kOe
• For 55 K ≤ T ≤ 70 K, large HC loops appear for H = ±10 kOe
30
40
T(K)
50
60
70
Is it Possible to Control the Sign of HE?
Magnetization measurement
Exchange bias studies after field cooling with 2000 Oe from
95 K with SQUID
 Measurement direction: c-axis
 Measurement sequence: 70 kOe → -70 kOe → 70 kOe, ( )
70 kOe → -20 kOe → 70 kOe, ( )
-70 kOe, 20 kOe → -70 kOe → 20 kOe ( )

M
-70 kOe
-20 kOe
20 kOe
70 kOe
H
Fe0.49Ni0.51F2/Co
•Tunable exchange bias (reversal of wide hysteresis loop)
6
-4
m (10 emu)
4
70 kOe, -70 kOe
70 kOe, -20 kOe
-60 kOe, 20 kOe
2
0
-2
5K
-4
-6
-2.0
-1.5
-1.0
-0.5
0.0
0.5
H(kOe)
1.0
1.5
2.0
Reversible Exchange Bias
• MCo favors parallel exchange coupling with Muncompensated
MCo
Muncompensated
1.0
(a)
0.5
0.0
1.0
M/MS
0.5
-0.5
0.0
M/Ms
-0.5
-1.0
-1.0
1.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
H (kOe)
(b)
0.5
0.0
1.0
0.5
M/Ms
-0.5
0.0
-0.5
-1.0
-1.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
H (kOe)
-60
-40
-20
0
20
40
60
H (kOe)
Consistent with micromagnetic
modeling
M. Cheon, Z. Liu, and D. Lederman, Appl. Phys. Lett. 90, 012511 (2007)
Summary for FexNi(1-x)F2/Co bilayers
0.10
0.15
0.05
Ms/Ms
0.20
0.00
0.10
-0.05
0.05
-0.10
400
0
HE(Oe)
-100
0.00
0.05
0.21
1.00
-200
-300
-400
0
20
40
60
T(K)
Note low TB
80
100
200
HE(Oe)
Ms/Ms
TN
0
-200
-400
0
0.49
20
40
60
80
100
T(K)
Note sign change of HE correlated with M
(same as in FeZnF2 samples)
What about FeZnF2?
Can HE be Reversed at Low T?
Fe0.05Ni0.95F2/Co
Fe0.21Ni0.79F2/Co
8
4
6
m(10-4 emu)
m(10-4 emu)
4
2
0
-2
-1.5
-1.0
-0.5
0.0
0.5
1.0
H(kOe)
0
-2
-4
30 K
-6
30 K
-4
2
-8
-1.5
1.5
-1.0
-0.5
0.0
no effect at 5K
m(10-4 emu)
m(10-4 emu)
4
2
0
-2
-4
-6
-8
-1.5
-1.0
-0.5
0.0
H(kOe)
0.5
1.0
4
2
0
-2
-4
20 K
1.5
1.5
1 nm FeF2
6
6
1.0
H(kOe)
Fe0.36Zn0.64F2/Co
8
0.5
-6
-1.5
20 K
-1.0
-0.5
0.0
H(kOe)
0.5
1.0
1.5
Key Questions
• Can uncompensated moments in the AF be detected?
– Uncompensated moments exist in AF, not due to “metallization”
– Pinned uncompensated moments in AF vanish near TN
– Unpinned uncompensated moments exist up to RT, well above TN
• Can the effects of uncompensated moments in the AF be
studied systematically?
– Uncompensated M does not necessarily lead to HE enhancement;
critical concentration of impurities must be achieved
– However, uncompensated M dependent on defect concentration
• Can the magnetic anisotropy be studied systematically?
– Low magnetic anisotropy leads to reversible HE, in addition to low TB,
as a result of reversal of “pinned” uncompensated M in the AF
– Low TB ≠ low TN
– Reversible HE requires uncompensated M in the AF
– Dilute AF system can also be reversed, but only at higher temperatures
due to coupling of H to uncompensated magnetization
Remaining Questions
• How universal is the effect of uncompensated
moments in the AF?
– Can it explain, e.g., low TB , in other AFs?
– Is it possible to engineer desirable interface exchange
properties by manipulating AF anisotropy?
• What is the size of the AF domains? And does
their size really matter?
– If they don’t matter, what is the coupling mechanism
and where does the uncompensated magnetization
come from?
• Strain (piezomagnetism)?
• Defects?
– Update: surprisingly, domain size does not
seem to matter much – see Fitzsimmons et al., PRB
77, 22406 (2008).
Group
Areas of Interest
5K
10 K
15 K
20 K
4
2
0
-2
5 nm Al,Pd cap
18 nm Co
-4
-6
-60
-40
-20
0
T = 565 °C
50 40
nm Fe
Ni F
60 x (1-x) 2
20
H(kOe)
MgF2(110) sub.
Exchange bias
GMR in anisotropic structures
Self-assembly and surface dynamics
Magnetic Nanostructures and Interfaces
Myoglobin
T~5.7K-5.8K
40
20
0
Bias Voltage (mV)
-4
m(10 emu)
6
-20
-40
-60
-80
-100
YMnO3/GaN
-120
-140
-2
-1
0
1
2
3
Gate Voltage (V)
Myoglobin Single Electron Transistor
Biomolecular Electronics
Hybrid Multifunctional Heterostructures
Areas of Interest
5K
10 K
15 K
20 K
4
2
0
-2
5 nm Al,Pd cap
18 nm Co
-4
-6
-60
-40
-20
0
T = 565 °C
50 40
nm Fe
Ni F
60 x (1-x) 2
20
H(kOe)
MgF2(110) sub.
Exchange bias
GMR in anisotropic structures
Self-assembly and surface dynamics
Magnetic Nanostructures and Interfaces
Myoglobin
T~5.7K-5.8K
40
20
0
Bias Voltage (mV)
-4
m(10 emu)
6
-20
-40
-60
-80
-100
YMnO3/GaN
-120
-140
-2
-1
0
1
2
3
Gate Voltage (V)
Myoglobin Single Electron Transistor
Biomolecular Electronics
Hybrid Multifunctional Heterostructures
Uncompensated M, x=0.75
0.04
400
0.02
0.00
0
HE (Oe)
M/MS
200
-200
-0.02
-400
-0.04
0
10
20
30
40
50
60
70
80
90
T(K)
Sign change of HE due to reversal of AF structure
H. Shi and D. Lederman, Phys. Rev. B 66, 094426 (2002)
Measurement Procedure
6
2
-4
m (10 emu)
1. Cool in HCF from above T = TN
4
2. Measure M vs. H at T < TN
0
-2
-4
-6
-1.0
Conventional view: Eint   J int  Si , A  S j ,F
-0.5
0.0
0.5
H (kOe)
Interface exchange interaction sets low T antiferromagnet configuration
F
Jint
AF
HCF
H
Jint
1.0
Direct Exchange Mechanism
• Direct exchange mechanism
(Meiklejohn and Bean, 1956) predicts
– a) wrong magnitude (~100 times too large)
– b) no exchange bias in compensated or
disordered surfaces
H E  J int / a 2 M F t F
HE = 0
F
Jint
AF
Ideal Uncompensated
Compensated
Roughness
Random Exchange at Interface
• Due to interface roughness, defects, etc.
• Antiferromagnetic domains created with
local exchange satisfied during cooling
H E  2 J int / LaM F tF
L = domain size in AF
Malozemoff, 1987
AF Domain Wall Formation
• AF or F domain walls created during
cool-down procedure
H
H
Jint
H E  2 AK / a 2 M F t F
Correct order of magnitude
Exchange stiffness A  J AF ,F / a
Magnetic anisotropy energy K
Lattice parameter a
Malozemoff, 1987; Mauri et al. 1987