A New Spin on Electronics -Spintronics-

A New Spin on Electronics-Spintronics-

Stuart WolfStuart WolfUniversity of VirginiaUniversity of Virginia

Presented atPresented atSPIN 08 October 11, 2008SPIN 08 October 11, 2008

Charlottesville, VACharlottesville, VA

SPIN 08 October 11, 2008

Beyond Conventional Electronics: Spintronics

Conventional Electronics Charge• Based on number of charges and their energy• Performance limited in speed and dissipation

Spintronics Spin• Based on direction of spin and spin coupling• Capable of much higher speed at very low power


Outline of talk

Spin Transport Spintronic sensors for Magnetic Recording Magnetic Random Access Memory (MRAM) Spin Transfer Torque Random Access

Memory (STTRAM) Spin Torque Nano-Oscillators


Spin Dependent Transport

- in all ferro and ferri-magnetic systems current is carried independently in two spin-channels

- conductivity in two channels can be very different

can be described by spin-dependent mean free paths or scattering times

current is spin-polarized

manipulate flow of spin polarized current useful sensors and memories

Energy

4s 3d

Co

0.35 0.35

3.35

Density of states

Spin-down

Spin-up

Neville Mott (1934)


Two main types of digital data storage

Random access memory Hierarchy of memories SRAM- fast but expensive DRAM- less fast and less

expensive Highly reliable but volatile Flash: non-volatile, less

expensive, very slow, limited endurance

Hard disk drives Massive storage Non-volatile Very cheap Very slow Less reliable!

Digital data storage

ma s s s t o r a g e


In a non-magnetic conductor, electrons scatter the same amount regardless of spin as current flows.

How much they scatter determines the resistance of the device.

Current in a metallic conductor


In a Ferromagnetic conductor, however, electrons scatter differently depending on whether they are spin up or spin down.

In this case, the spin up electrons are scattered strongly while the spin down electrons are scattered only weakly.

Current in a ferromagnetic conductor


If a non-magnetic conductor is sandwiched between two oppositely magnetized ferromagnetic layers, a number of electrons will scatter strongly when they try to cross between layers.

this gives higher resistance.

Spin-Dependent Scattering

If the ferromagnetic layers are magnetized in the same direction, far fewer electrons are strongly scattered and more current flows

This is measured as lower resistance

Useful for sensing magnetic fields or as a magnetic memory element


To make a technologically useful device, a “pinning” layer is added to make it harder to change the magnetization of one layer than the other.

The pinning layer can be a simple layer of an antiferromagnetic material.

Spin-valve

Antiferromagnet


400 H (Oe)-40

400

110

H (kOe)-40 H // [ 0 11]

spin-valve

multi-layer

Co95Fe5/Cu[110]

R/R~110% at RTField ~10,000 Oe

Py/Co/Cu/Co/Py

R/R~8-17% at RTField ~1 Oe NiFe + Co

nanolayer

NiFeCo nanolayerCuCo nanolayerNiFeFeMn

H(Oe)

H(kOe)[011]

10

MR(%)Giant Magnetoresistance (GMR)

NOBEL PRIZE !

Fert and Gruenberg


Magnetic engineering at the atomic scale

Spin ValveGMR sensor

+ interface engineering

Spin ValveMagnetic Tunnel Junction

Ferromagnet

Ferromagnet

Spacer layerMetal or insulator

Anti- Ferromagnet

+ interface layer

+ interface layer

+ ArtificialAntiferromagnet


Hard Disk Drive

N 5 8

3 m m

2 0 0 0 Å

C op p er

P erm allo y

A l O2 3

B ak ed p h o to -resist

W rite H ead

R ea d H ead


Year

25% CAGR

~200%

60%

IBM Disk products

Lab demos

1st Thin Film Head

1st MR Head

1st GMR Head

IBM RAMAC (1st Hard Disk Drive)

100 Gb/in2

l

wtra ck w id th

b it len g th

d

A n iso tro p y k eep s th e m o m en t p o in tin g in th e d irec tio n o f

th e track

T h e tran s itio n w id th is a ffec ted b y b o th th e an iso tro p y

an d th e m ag n e tiza tio n

d

1 9 9 61 G b /in 2

1 9 9 86 G b /in 2

2 0 0 02 0 G b /in 2

.0 1 m

2 0 0 31 0 0 G b /in 2

l

wtra ck w id th

b it len g th

d

A n iso tro p y k eep s th e m o m en t p o in tin g in th e d irec tio n o f

th e track

T h e tran s itio n w id th is a ffec ted b y b o th th e an iso tro p y

an d th e m ag n e tiza tio n

d

1 9 9 61 G b /in 2

1 9 9 86 G b /in 2

2 0 0 02 0 G b /in 2

.0 1 m

2 0 0 31 0 0 G b /in 2

~30%

Hard Disk Drive areal density evolution


SPIN 08 October 11, 2008 Seagate 2006


Hard Disk Drive capacity shipped per year

100 Exabytesin ~2005

Year

Byte

s

Sh

ipp

ed

/ Y

ear

~100% CGR


Spintronics Spin valve sensor Major impact on hard disk drive storage enabled >400x increase in storage capacity

since 1998 makes possible minaturization of hard disk

drives cell phones, PDA, MPEG players

makes possible access to all information

Spintronics Magnetic Tunnel Junction Major impact on random access memory? Just introduced to hard disk drive storage


Spin Polarized Electron Tunneling: FM-I-FM

E E

NS NS

E F E F

E F E F

eV eV

FM FMFM FMI I

M MM M

Im a j Im a j

Im in Im in

1 2 1 2PI NN N N 21 1 2

AP N NI N N

Pinned FM

Free FM

1 2

1 2

2

1

with

AP P

P

R R PPMR

R PP

PN

N

N

N

Juliere (1975)


CORRECTION: The legend in Figure 1(d) should read T = 295 K.

T=295 K

T=295 K

1975 1982

1995 1995



Exchange-Biased Magnetic Tunnel Junction (MTJ)

Non-Volatile Memory!Non-Volatile Memory!

FieldH=0

R/R

Ti, Ti/Pd or Ta/ Pt

Si, quartz, N58

Underlayer

Antiferromag

net

CoFe or NiFe/CoFe

Al2O3 Al2O3 CoFe/NiFe

Top lead

Substrate

Free ferromagnet

Pinned ferromagnet

Tunnel

barrier

Tunnel

barrier

Antiferromag

net Bottom electrode

MR

(%

)

Field (Oe)

-100-80 -60 -40 -20 0 20 40 60 80 1000

20

40

60

80

“”“”

“”“”

Magnetization


History of development of MTJs

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 20040

50

100

150

200

250

T

unne

ling

Mag

neto

resi

stan

ce (

%)

Year

Al2O

3 results

MgO results in public domain IBM internal results with MgO

Record TMR –500%


Conventional MRAM (1T-1MTJ)

u

B

K V

k T Thermal Stability Factor

SCALING PROBLEM Beyond 65 nm node!!!

Freescale


Net change in S per e

Spin Torque Transfer Switching

STorque

t

Absorbed Angular Momentum Torque

Polarizing “fixed” layer (thick)

Active “free” layer (thin)

Spin polarized current generates torque on magnetization of free layer

2

S I

t e

2 2

IS N t t

e

Torque

MR ratio 0.5-5% Jc~107A/cm2

Katine et al, Phys. Rev. Lett. 84, (2000) 3149 .


Switching current scales down with cell size

~ 6mA

~ 0.5mA

Albert et al, Appl Phys. Lett., 77 3809 (2000).

Grandis Inc


“State of the Art” in STT-MTJ’s

Reductions in Jc ~ 9×105 A/cm2 and TMR ~ 73%

MgO increases

The improvement is over amorphous AlOx tunnel barriers that were initially studied and gave Jc ~ 8×106

A/cm2 and TMR ~ 42%J. Hayakawa, JJAP 41 (2005) L1267

Thermal Stability Factor Not Satisfied!!


Current Scaling – MRAM vs STTRAM


Challenges for STTRAM Switching Current Density Too High!

Small current needed to decrease size of MOSFET in series with MTJ cell (1T-1MTJ)

Small voltage across device needed to reduce probability of tunneling barrier breakdown

Need to reduce current density required to switch cell while achieving high MR%

Jc needs to be lowered to ~105 A/cm2


New Materials Lower Switching Current Density

Spin Transfer Model

• Ms Saturation Magnetization Decrease• Gilbert damping parameter Decrease• Spin Transfer Efficiency Increase

0

2 2s F K sc

e M t H H MJ

[J.C. Slonczewski J. Magn Mater. 159 (1996) L1]

Also require: Anisotropy Energy / kT > 60 for 10 year retention


New Materials Can we do better?

Co70Fe20B10P ~ 53 %~ 0.032

P measured using Superconducting Tunneling Spectroscopy (STS) with an AlOx tunnel barrier and was determined with FMR characterization post anneal

CrO2P ~ 94 %~ 0.0023

[C. Bilzer et al, JAP 100 (2006) 053903]

[P. Lubitz et al, JAP 89 (2001) 6695]

[Parker et al, PRL 88 (2002) 196601]

[P.V. Paluskar et al JAP 99 (2006) 08E503]

M1-xCrxO2 Newly Discovered RT Ferromagnetic Oxides! M=V and RuVO2 ×10 with charge injection Jc~ 104 A/cm2


Excellent write selectivity ~ localized spin-injection within cell Highly Scalable ~ write current scales down with cell size Low power ~ low write current Simpler Architecture ~ no write lines, no bypass line and no cladding High Speed ~ Few nanoseconds

Key Advantages and Potential of STTRAM


9.6 9.7 9.8 9.9 10.0

0.0

0.1

0.2

0.3

0.4

9 mA

8.5 mA 8 mA

7.5 mA

7 mA

6.5 mA

6 mA

5.5 mA

Pow

er (

pW)

Frequency (GHz)

Spin-Current Switched MRAM

Spin Transfer Nano-Oscillators

50 nm

1 m

0 50 100 150 200

-1.0

-0.5

0.0

0.5

1.0Switching in response to a 10 mA current pulse

Ea

sy A

xis

Ma

gn

etiz

atio

n

Time (ps)

Spin Torque Nano-Oscillators

Simulations: OOMMF math.nist.gov/oommf/

I

Tunnel junction

High-speed switching

Tunable High Q oscillator (2 GHz – 100 GHz)

Au

Cu

0.7 T, = 10o

CoFe

NiFe

I

simulation

data


17.025 17.050 17.075

0

2000

4000

6000

8000

10000

Pow

er (

nV2 /

Hz)

Frequency (GHz)

f = 17.052 GHzf = 3.00 MHz

6 8 10

16

17

18

19

Fre

quen

cy (

GH

z)

Current (mA)

0.5 GHz/mA

Summary of Present Status Summary of Present Status

0.0 0.2 0.4 0.6 0.8 1.00

10

20

30

40

Field (T)

Freq

uenc

y (G

Hz)

Field TunableCurrent Tunable

Narrow Band

•Oscillators are tunable over a wide range of frequencies via applied field or current•Output is narrow band with Q values > 10,000•Voltage outputs in the mV regime

28 GHz/T


Fundamental Frequency Limits

The gyromagnetic precession frequency of spins has no upper bound!

0 effH

For ultra-small contacts of diameter 3 nm < d < 8 nm, intralayer exchange dominates the energetics:

2

0

exeff

DH

d

SMT oscillators could fill the “THZ gap.”

“THz gap”


0.1 0.2 0.3 0.4 0.5 0.6 0.7

-200

-100

0

100

200

VO

sc (V

)

Time (ns)

Idc= 7.4 mA

Idc= 7.6 mA

Idc= 7.8 mA

Idc= 7.85 mA

7.4 7.5 7.6 7.7 7.8

-100

-50

0

50

100

Rel

. Pha

se S

hift

(deg

rees

)Idc

(mA)

locked

Electronic Phase Control of Oscillations

The relative phase can be varied using the DC current!

W. H. Rippard et al, Phys. Rev. Lett. 95, 067203 (2005).

A

B

Locked

Phase Locking in Closely Spaced Spin Transfer Nano-Oscillators


A biased at 11.5 mA; B swept 0 – 15 mA

nV)2/Hz

3.7

13.3

A

B

Locked

13.8 14.0 14.2 14.4 14.6 14.8

0

20

40

60

80

PS

D (

nV)2 /H

z

f (GHz)

13.8 14.0 14.2 14.4 14.6 14.8

0

2

4

6

8

PS

D

(nV

)2 /Hz

f (GHz)

13.8 14.0 14.2 14.4 14.6 14.8

0

10

20

30

PS

D (

nV)2 /H

z

f (GHz)

AB Locked

A

B

Phase Locking 500 nm Spaced ContactsPhase Locking 500 nm Spaced Contacts

Kaka et al, Nature, Sept. 2005

Spin valve

A

B500 nm

IB

IA

When phase locked power increases & linewidth decreases

)cos(2 BABAT PPPPP

0 90 180 2700

2

4

6

8

10

Po

we

r (p

W)

Phase Shift (deg)


Applications of Spin Transfer Nano-OscillatorsApplications of Spin Transfer Nano-Oscillators

0 1 2 3 4

Frequency

Ampl

itude

Signal S(t) in

Inducedspin waves

Component signals out

Point contact STNOs

200 nm

i1i2i

3

)(~

1fS)(~

2fS)(~

3fS

STNO

I2

sin(t+2)

STNO

I3

sin(t+3)

STNO

I4

sin(t+4)

STNO

I1

sin(t+1)

SMT device/GMR sensor

Spin-transferoscillator

Near-field antenna

Chip-to-chip microwireless

Nanoscale Phased Array

High-speed parallel signal processing

Referenceoscillator

Not going to replace existing VCOs!

Target new applications requiring nanoscale high frequency

components!

Documents

A New Spin on Electronics -Spintronics-