Magneto-Seebeck effect in magnetic tunnel junctions

Magneto-Seebeck effect in magnetic tunnel junctions

Jakob Walowski, Andreas Mann, Marvin Walter, Valdislav Zbarsky, Anissa Zeghuzi, Markus Münzenberg, I. Physikalisches Institut, Universität Göttingen, Germany

Michael Czerner, Michael Bachmann, Christian Heiliger, I. Physikalisches Institut, Justus-Liebig-Universität Gießen

J.S. Moodera, MIT Cambridge, USA

Markus Schäfers, Daniel Ebke, Günter Reiss, Andy Thomas*, Department of Physics, Universität Bielefeld, Germany

Department of Physics, * Institut für angewandte Physik, Hamburg, Germany

„chargeaccumulation“

„spinaccumulation“

TÑ

V+ -

mS

+

TÑ

-

TSV D=D

Charge and spin-dependent Seebeck effect

TSSs D=Dm

Theory: M. Johnson and R. H. Silsbee, Phys. Rev. B 35, 4959 (1987). Experiment: L. Gravier Phys. Rev. B 73, 024419 (2006).K. Uchida Nature 455, 778 (2008).A. Slachter Nature Phys. 6, 879 (2010).G.E.W. Bauer, Spin Caloritronics, Solid State Comm. 150, 459 (2010).

FeCo

Fe/ MgO/ FeCo/ MgO/ Co

+ -

TÑ

+ -

TÑ

V V

… in bulk and tunnel junctions

M. Czerner et al. Phys. Rev. B 83, 132405 (2011)

Spin-scattering mechanisms

G. M. Müller et al., Nature Mater. 8, 56 (2009)

Elliott-Yafet process: femtosecond demagnetization and FMR

R. J. Elliott, Phys. Rev. 96 (1954).G. M. Müller et al., Nature Mater. 8, 56 (2009)

)()(~ 2FhFe EncEnW ¯­

­¯

exchSO Ec DV~ Spin-orbit interaction

Spin mixing:

Fs pump-probe:

m

)(En­ )(En¯

E

­¯W

¯­

¯­

+-

=ee

ee

nnnnP0

Spin polarization:

Spin-scattering mechanisms

Elliott-Yafet process: femtosecond demagnetization and FMR

R. J. Elliott, Phys. Rev. 96 (1954).G. M. Müller et al., Nature Mater. 8, 56 (2009)Fs pump-probe:

Spin polarization P dependent relaxation

For : • Spin-flip processes are prohibited• Electron and spin system are

isolated

1®P

­¯W

)(En­ )(En¯

E

m

)()(~ 2FhFe EncEnW ¯­

­¯

exchSO Ec DV~ Spin-orbit interaction

Spin mixing:

Spin-scattering mechanisms

¯­

¯­

+-

=ee

ee

nnnnP0

Test of the effect of the coupling parameter:

• gel-sp

• 2nd: suppression of the spin scattering

Spin-scattering mechanisms

spins (Ts)

latsp-tspel-t

lattice (Tl)

latel-t

electrons (Te)

e-

Coupling parameters• gel-lat > gsp -lat > gel-sp

0 5 10 15

Ni

Co2MnSi

DqKe

rr [ar

b. u

nits

]

CrO2

LSMO

Dt [ps]

Fe3O4

• Spin-flip processes are prohibited• Electron and spin system are

isolated

Three temperature model: half metal

spins (Ts)

latsp-tspel-t

lattice (Tl)

latel-t

electrons (Te)e-

Spin relaxation in half metals

G. M. Müller et al., Nature Mater. 8, 56 (2009)

Collaboration with: Bielefeld University and Hitachi GST

Spin relaxation in half metals

0.4 0.6 0.8 1.0

0.1

1

10

100

t m,

fit [p

s]

P

Ni CoFeCo2FeAl

Fe, Ni, Py

CrO2

(CoFe)1-xGex

Co2MnSi/V/Si

CoFeSb*

Co2MnSi/MgO Co2MnGe

TP (E)

TAP (E)

• “Conventional” Seebeck effect: Voltage generation due to a temperature gradient in the material:

• In a magnetic tunnel junction depence on magnetization orientation:

“Magneto-Seebeck effect”

TSV D=D

SMS =SP - SAP

min(SP,SAP)

Magneto-Seebeck effect?

Jakob WalowskiBenjamin Lenk

Andreas Mann (now EPFL), Henning Ulrichs (now U Münster), Fabian Garbs Marvin Walter

…. + Valdislav ZbarskyBachelor students:

Martin Lüttich, Christian Leutenantsmeyer, Jelena Panke, Nils Abeling, Mirco Marahrens, Anissa Zeghuzi

Priority program SpinCaT

People who do the work

• Introduction into coherent tunneling

• Magneto-Seebeck experiment:

•Experimental setup

•Simulations of the temperature gradients

•Ab initio calculations

•Comparison to the experiment

• Summary

Outline

• Introduction into coherent tunneling

• Magneto-Seebeck experiment:

•Experimental setup

•Simulations of the temperature gradients

•Ab initio calculations

•Comparison to the experiment

• Summary

Outline

V

Parallel magnetization

V

P

PAP

RRR -

=TMR

Antiparallel magnetization

IAPFM 1BarrierFM 2

FM 1BarrierFM 2

IP

Giant tunneling magnetoresistance

Resistance change is tunnelingmagnetoresistance (TMR):

+

-

+

-

Experimentaly achieved TMR604%

S. Ikeda et al.APL 93,082508(2008)

220%

Giant tunneling magnetoresistance

S. Yuasa und D. D. Djayaprawira, J. Phys. D: Appl. Phys. 40, R337-R354 (2007)

Difference between Al2O3 and MgO barriers

S. Yuasa und D. D. Djayaprawira, J. Phys. D: Appl. Phys. 40, R337-R354 (2007)

• Transmission probability is the same for different electronic bands

• Transmission probability changes for different electronic bands

Giant tunneling magnetoresistance

Fe|MgO|Fe coherent tunneling

Giant tunneling magnetoresistance

How thin can we get?

040

-220

-400

220

-2-20

400

2-20

0-40

-200

200

020

21-1

0-20

200 2-11

01-1 0-11

-21-1 -200 -2-11

02-2

-11-1

-400

11-1

-1-11

400

1-110-22

-200

200110

200

1-10

020 0-20

-110-200

-1-10

MgO(001) CoFeB(011)

MgO(011) CoFeB(001)

-2-20

220

(100)

(100)MgO(0-10)Fe(0-11)

MgO(001)Fe(011)

MgO(100)Fe(100)

MgO(0-11)Fe(0-10)

MgO(011)Fe(001)

MgO(100)Fe(100)

Atomic structure – columnar growth

TMR = 320%

Sample annealed 400°CIn collaboration with: M. Seibt U GöttingenA. Thomas, G. Reiss U Bielefeld

V

Parallel magnetization

V

P

PAP

RRR -

=TMR

-5.0 -2.5 0.0 2.5 5.0

200

400

600

800

R (O

hm)

µ0H (mT)

0

50

100

150

200

250

300

TM

R (%

)

Antiparallel magnetization

CoFeB/ MgO/ CoFeB junctions are half metallic with highest TMR (~600% at RT)

Appl. Phys. Lett. 95, 232119 (2009)J. Appl. Phys. 105, 073701 (2009)

IAPFM 1BarrierFM 2

FM 1BarrierFM 2

IP

Giant tunneling magnetoresistance

+

-

+

-

• Introduction into coherent tunneling

• Magneto-Seebeck experiment:

•Experimental setup

•Simulations of the temperature gradients

•Ab initio calculations

•Comparison to the experiment

• Summary

Outline

)(En

E

Semiconductor

Magneto-Seebeck effect

Semiconductors have a large Seebeck effect

Orgin is the• Large asymmetrie of the electrons at

around the Fermi energy

Conduction• Determined by the density of states n(E)

Seebeck effect S=ΔV/ΔT

)(En

E

Semiconductor

Magneto-Seebeck effect

Semiconductors have a large Seebeck effect

Orgin is the• Large asymmetrie of the electrons at

around the Fermi energy

EEnE D- )()( m

Conduction• Determined by the density of states n(E)

Seebeck coefficient S• Determined by density of states• weigthed by (E-m) • times derivative

Geometric center

Seebeck effect

dEEdf )(

TSV D=D

E

m

m

)(ETAP)(ETP

Magneto-Seebeck effect

Tunnel junction For tunnel junctions:

Conduction• Determined by the transmission T(E)• High TMR

M. Walter et al. Nature. Mater. 10, 742 (2011).

E

m

m

Magneto-Seebeck effect

Tunnel junction For tunnel junctions:

Definition of the magneto-Seebeck effect

Conduction• Determined by the transmission T(E)

Seebeck coefficient• Determined transmission T(E)• weigthed by (E-m) • times derivative

EETE D- )()( m

Geometric center

dEEdf )(

dEEdf )(

)(ETAP)(ETPM. Walter et al. Nature. Mater. 10, 742 (2011).

• Introduction into coherent tunneling

• Magneto-Seebeck experiment:

•Experimental setup

•Simulations of the temperature gradients

•Ab initio calculations

•Comparison to the experiment

• Summary

Outline

Camera

B=+/- 80 mTLaser diode5 -100 mW

Chopper(frequency 0.8, 1.5 3 kHz)

Experimental setup

• Toptica stabilized laser diode (784 nm)• Most experiments: 30 mW laser power• Modulation 800 Hz, 1.5 kHz, and 3 kHz

1 mm

Junction array

Camera

B=+/- 80 mTLaser diode5 -100 mW

250 mm

Junction array Au bond pad

Chopper(frequency 0.8, 1.5 3 kHz)

Experimental setup

• Focus15-20 μm diameter• High input impedance (100 GΩ) (LT1113 Linear Technology)• Resistance change < 1 nV

B=+/- 80 mTLaser diode5 -50 mW

Seebeck effect:

Magneto -Seebeck effect:

Experimental setup

TSV APPAPP D=D ,,

TÑAPS

TÑ

+-

500 nm

V

PS +-

V

0.0 0.5 1.0 1.5 2.0 2.50

50

100

150

200

250

300 3 kHz 1.5 kHz 800 Hz reference

Seeb

eck v

oltag

e [mV

]

Time [ms]

Experimental setupModulation chopper and laser intensity Seebeck voltage

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

0.0 0.2 0.4 0.6 0.8 1.0-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

closed open

Ampl

itude

(arb

. uni

ts)

Chopper Signal Diode Signal

3000 Hz Asymmetric Chopperop

enclosed

Ampl

itude

(arb

. uni

ts)

Time (ms)

3000 Hz Normal Chopper

10ms

-5.0

-5.2

-5.4

-5.6

-5.8

-40 -20 0 20 4050

100

150

Volta

ge [m

V]

3 kHz 1.5 kHz

R [k

W]

Magnetic field [mT]

0

50

100

150

TMR

[%]

See

beck

vol

tage

[mV

]

3 kHz1.5 kHz

R [k

W]

Magneto-Seebeck effectSeebeck voltage 0.45 mV, DS= -8.7 mV/KMagneto-Seebeck effect: -8.8%

TÑAPS

TÑ

+-

500 nm

V

PS +-

V

• Introduction into coherent tunneling

• Magneto-Seebeck experiment:

•Experimental setup

•Simulations of the temperature gradients

•Ab initio calculations

•Comparison to the experiment

• Summary

Outline

-40 -20 0 20 400

150

300

450

600

T [K]

z [n

m]

x [mm]

293

296

298

300302

Au contact

SiO2

SiO2

Ta contact

Au contact

SiO2

250 mm

COMSOL finite element modeling

• COMSOL uses finite elements methods• Solving of heat conduction equation

QTktTc +Ñ×Ñ=

¶¶ )(pr

Au contact

SiO2

Co-Fe

Co-FeMgO

Ta

200 ps

1ms

Temperature gradients: 1 to 50 mK

COMSOL finite element modeling

x 40

Device structure by HRTEM

3nm2 nm

• COMSOL uses finite elements methods• solving of heat conduction equation

• 3d or 2d cylindrical model

COMSOL finite element modeling

QTktTc +Ñ×Ñ=

¶¶ )(pr

COMSOL finite element modeling

Au contact

SiO2

200 ps

Co-Fe

Co-FeMgO

Ta

1ms

Cross section of the finite element model Temperature gradients: 1 to 50 mK

COMSOL finite element modeling

TVS DD=

M. Walter et al. Nature. Mater. 10, 742 (2011).

COMSOL finite element modeling

• Introduction into coherent tunneling

• Magneto-Seebeck experiment:

•Experimental setup

•Simulations of the temperature gradients

•Ab initio calculations

•Comparison to the experiment

• Summary

Outline

For tunnel junctions:

Conduction

Seebeck coefficient

Definition of the magneto-Seebeck effect

Magneto-Seebeck effect

Calculations by C. Heiliger U Gießen

Co0.5Fe0.5 Co0.5Fe0.5MgO

200 400 600 800

-80

-40

0

S (parallel) S (antiparallel)S P,

AP [

mV/K

]

Temperature [K]

200 400 600 800

-100

0

100

200

300

M

agne

to S

eebe

ck [%

]

Temperature [K]

SP [μV/K] SAP [μV/K] SP - SAP

[μV/K]SMS [%]

CoFe -19.7 -32.4 12.7 64.1FeCo 45.9 -50.0 95.9 209.0CFFC 9.4 -44.6 54.0 573.2Co0.5Fe0.5 -34.0 -21.9 -12.1 -55.2Experimental value

-107.9 (-1300) -99.2 (-1195) -8.7 (-105) -8.8 (-8.8)

Device structure by HRTEM Seebeck voltage 0.45 mV, DS= -8.7 mV/K

Co0.5Fe0.5 MgO

Magneto-Seebeck effect

M. Walter et al. Nature. Mater. 10, 742 (2011).

10 20 30 40

-10

-20

-30

-40

AP P , model

Seeb

eck

volta

ge [m

V]

Pump fluence [mW]

-40 -20 0 20 40

-9.0-10.5-16.5-18.0-19.5

-30-32

-37-38-39

-42

-43

-44

Seeb

eck

volta

ge [m

V]

5 mW

m0H [mT]

10 mW 20 mW

30 mW

40 mW

Magneto-Seebeck effect

M. Walter et al. Nature. Mater. 10, 742 (2011).

Higher T and gradient expected for higher laser powers• nonlinear behavior TTSV D=D )(

-100-50

050

100150

-100-50050100150

-100-50

050

100150

-100-50050100150

0 0.5 1.0 1.5-100-50

050

100150

0 0.5 1.0 1.5 2.0-100-50050100150

10 mW AP 10mW P

30 mW AP 30 mW P

Mon

itor O

ut (m

V)

Mon

itor O

ut (m

V) 60 mW AP 60 mW P

90 mW AP 90 mW P

Time (ms)

120 mW AP 120 mW P

open

closed

Time (ms)

150 mW AP 150 mW P

High fluence data: femtosecond laser10 mW

150 mW

30 mW

60 mW 90 mW

120 mW

0 2 4 6 8 100

5

10

15

20

25 laser pulse

temperature difference within the MgO layer

te

mpe

ratu

re [K

]

time [ns]

0.00 0.05 0.10 0.150

5

10

15

20

tem

pera

ture

[K]

time [ns]

High fluence data: femtosecond laser

•Large temperature gradinets are achievable: 10K per nanometer

-10

0

10

20

30

40

0 50 100 150

Ti:Sa laserlaser diode

Mag

neto

-See

beck

effe

ct [%

]

300 350 400 450 500

P [mW]

Temperature [K]-40 -20 0 20 40

-30

-40

-40

-40

-40

-40

-50-40

-50

10 mW T= 300K

30 mW T=330K

60 mW T=365K

Seeb

eck

volta

ge [m

V]

90 mW T=400K

120 mW T=440 K

m0H [mT]

150 mW T=475 K

Sign change as predicted:

M. Walter et al. Nature. Mater. 10, 742 (2011).

Magneto-Seebeck effect

• Magneto-Seebeck experiment:

•Experimental verification

•Simulations of the temperature gradients

•Switching of Seebeck coefficients demonstrated

•Ab initio calculations

•Typical features are reproduced by the experiment

Summary

Increasing temporal resolutionSeebeck voltage autocorrelation:

• Femtosecond laser excitation (40 fs temporal resolution)

• Delay t in between the pulses

• Probes non-linearity of the Seebeck voltage

• ~80 ps time scale is the expected timescale when the maximum temperature difference is reached (from simulation)