You are on page 1of 27

Novel spin FETs

J. Carlos Egues
Department of Physics and Informatics,
Institute of Physics of Sao Carlos
University of Sao Paulo, Brazil,
and
Department of physics and Astronomy,
University of Basel, Switzerland

Collaborators:
Guido Burkard (IBM)
Daniel Loss
Patrik Recher
Daniel Saraga
John Schliemann
Funding:
NCCR Nanoscience, Swiss
NSF, DARPA and ARO.

UNIVERSIDADE
DE SO PAULO
Instituto de Fsica de So Carlos

egues@if.sc.usp.br

Motivation
Spin FETs:
Most popular proposed spintronic device (Datta and Das 90);
Highlights coherent electric control of magnetic degrees of freedom;
Low energy to rotate/flip spins; much faster (?);
More control and various modes of operation. Logic gates with
dynamic functions - reprogrammable on the fly; also multifunctional:
logic, storage, processing, communication, etc, in a single chip?
Design flexibility (hybrid devices, geometry, etc. etc.)
Quantum effects: coherent microchips?
Novel physical phenomena: interplay of nanomagnetism and quantum
confinement; spin-flip mechanisms, spin injection and spin polarized
transport (ballistic and diffusive), non-local quantum correlations, etc...
Awschalom, Flatte, and Samarth, Scientific American, 2002.

Outline
Brief review of spin-orbit (s-o) interaction
Two dimensional electrons + s-o Rashba interaction
Datta-Das transistor (spin FETs): basics
Novel spin FETs:
Enhanced control (two channels)
Nonballistic spin FET (Robustness against spin flip)
[Coherent control of entangled electron pairs (noise)]
Summary/conclusions

Rashba spin-orbit interaction


(Back-of-the-envelope derivation or poor mans)

Heterojunction:

Atomic case:

Potencial profile: V(z)

v
r

-e
-e

+Ze

+Ze
-v
In its rest frame, the electron feels:

B eff

1
= v E nucleus
c

ge
=
S ( v E)
2
2mc

Corresponding Hamiltonian:

H s o

geh
=
(p V )
2

2mc 2
or, H

dV
H s o
r ( p),
dr

Two-dimensional
electron gas (2DEG)

In analogy with the atomic case:V


Hence, the Rashba s-o term is:

Spin-orbit interaction:

Es o = spin B eff

Can control
slope of V(z)
via gate!

Note structural
inversion asymmetry
(SIA)

s o

Ze 2
L S
2 2 3
2m c r

dV
V = r
dr

dV
= z
dz

Electric
Field.

dV
HsRo
z ( p)
dz

Defining a s-o coupling constant :

R
s o

( p) z =

p y y px )

H sRo = i ( y x x y )

Rashba (free) electrons in a 2DEG


Full Hamiltonian

H=

1
2
2
p x + p y + ( x p y y p x )
h
2m

Exact solution:

H =

m
kR =
h

h 2 k R2
R =
2m

h 2k 2
h 2 (k + sk R ) 2
s (k ) =
+ sk =
R, s =
2m
2m

s ,k =

k-dependent spinor!

rr
ik r

k
1 1 i ( k ) k y
i ( k ) , e
= i x
k
k
A 2 se

Spin expectation values:

+ ,k x + ,k
+ ,k y + ,k
+ ,k z + ,k

ky
0 1 1
i ( k ) = cos (k ) =

1 0 e
k
ky
= sin (k ) = x

r
1
k
,k ,k = m k x
k

=0
0

1
= 1 e i ( k )
2

10

Effective magnetic field:

h
2
H = (p z) 2
2
h
B eff
h
R
H so = gB Beff
2
2
B eff =
(k y x k x y )
2
kx
g B h

[1

ky
[010]

(k x , k y = 0)
+
+

+
+

+
+

+
+

Rashba bands & spin directions

k F+ k F k x
[100]
Paraboloids
Spin degeneracy at a given k (consequence of spatial
and rtime reversal
r symmetry) is lifted by SIA; however,
(k ) = (k ) since s-o preserves time reversal.

R
s o

(k dependent)

py
H p x
Velocity operator: v x =
=
y , vy =
+ x
p x m h
m h
kx
kx
r
r
h
1
2( + R ) 1
v ,k ,k v ,k = (k k R ) k y =
ky
m
k
m
k
0
0
r
Absolute value of v , k depends only on the energy.

Rashba+Dresselhaus s-o terms


Linearized version (QW).

Bulk inversion asymmetry (BIA):

1
2
2
H=
p x + p y + ( x p y y p x ) + ( x p x y p y )
h
h
2m
(ballistic case)

SIA

BIA

Rashba

Dresselhaus

Can still solve exactly for any and ; spin directions differ qualitatively:
Winkler
(cond-mat/0305315)

Interesting feature: can tune

B eff

Ting et al.
PRB (2003)

= ,

so that B eff has a fixed direction, i.e.,

[1, 1 ,0]
2
( k x + k y )(x m y )
=
2
g B h
[1,1,0]

In addition, the spinor is now k independent! (Nonballistic spin FET)

Datta-Das transistor: spin FET

Electric manipulation of intrinsic magnetic degrees of freedom.

Datta-Das spin transistor


S. Datta and B. Das, Appl. Phys. Lett. 56, 655 (1990).

polarizer
z

analyzer

y
x

L
(Phys. Today, adapted)

Main ingredients: 1DEG + spin orbit + FM emitter/collector


Hamiltonian:
(lowest subband)
1

2
1

2
2
H1DEG =
px + y px
H 2 DEG =
px + p y + y px x p y
2m
h
h
2m
Time evolution: rotation about the y axis

UR = e

i k y / h

) (

cos( R / 2)
=
sin( R / 2)

R 2mL / h
L = vF = (hk F / m)
2

sin( R / 2)

cos( R / 2)

( z basis)

s o
R

1 cos( R / 2)


0 sin( R / 2)

I , 1 cos R

Rashba Semiconductor: quasi-1D case


Full Hamiltonian+ confinement

H 2 DEG + Rashba + confinement

1
2
2
=
p x + p y + V ( y ) + i ( y x x y )
2m

Zeroth-order solution (no Rashba): quantum wire subbands

H 2 DEG + confinement =

1
2
2
px + p y + V ( y)
2m

h 2k 2
n , (k ) =
+ n , n = a,b,...
2m

k ,n , eikxn ( y ) , =,

Effective quasi-1D Rashba Hamiltonian: two lowest subbands a and b

H eff

id
ik
0
a , (k )
ik (k ) id
d d / dy
0
a
b
a ,

=
0
id
b , (k ) ik (interband mixing)
& Barnes, 99

Moroz
& Kirczenow, 01
ik b , (k ) Mireles
0
id
Governale & Zelicke, 02

Strictly 1D Case: d=0


H

1D
eff

Eigenvalues:

Neglects interband mixing


a (k ) ik
=
(lowest subband)

ik a (k ) Eigenvectors (spin):

h2
h 2 k R2
2
(k sk R )
s (k ) =
,
2m
2m

s=

m
kR 2
h

"Interface"
No Rashba

[
[

1
+i
2
1
=
i
2
=

]
]

Rashba (gate induced)

Small offset:

R << F

(k )

(k )

group velocity

kF

k1

k2

+ (k 2 ) = (k1 ) k 2 k1 = 2k R

Boundary conditions: strictly 1D Case


Continuity of the wave function at x=0:
eik F x

x 0

1
+ e ik 2 x + eik1x
2

x 0

Velocity operator:

hk F ik F x
e
m

=
x 0

1
2

assuming unity transmission


+

hk / m i / h
vR =

i
k
m

/
h
h
/

h
h
ik 2 x
ik1 x
(
)
(
)
k
k
e
k
k
e
+

+
1
R
+
m 2 R
m
x 0

h 1 h 1
1 vR (k 2 ) + vR (k1 )
1
kF =
(k k ) + (k + k ) =

m 2 m i 2 R i 1 R 2
0

Note that injection is possible because vF (k F ) =


with

h 2m
( + R )
vR (k 2 ) = vR (k1 ) =
2
m h
(group velocity)

1
[vR (k2 ) + vR (k1 )] ,
2

At x=L:

cos( R / 2) ikL
e
=
sin( R / 2)

Novel spin FETs:


1) enhanced spin control, and
2) Robustness against scattering

A new "spin" on the Datta-Das transistor?


Vg

(a)

or

L
L0
(b)

Novelty: two sets of s-o coupled Rashba bands.


J. C. Egues, G. Burkard, and D. Loss, Appl. Phys. Lett., 82, 2658 (2003)

Generalized spin-rotated state


(additional modulation due to interband mixing)

Upon traversing the Rashba region L, a spin-up electron - with


energy near the band crossing - is left in the state:
cos( d / 2)e i R / 2 + e i R / 2
1


i R / 2
i R / 2
+ ie

d
2m d
0 1 i cos( d / 2)e
UR
, where d = L = h 2 k = R k
0
2
i sin( d / 2)
c
c

sin( d / 2)
Wave vector

d a d / dy b

at band crossing.

("interband mixing")- side gates)

Note the additional modulation d due to s-o interband coupling.


For d = 0 , the above state reduces to the usual Datta-Das
state describing a single spin rotation R :
1 cos( R / 2)

U R
0 sin( R / 2)

Spin-resolved current
Upon traversing the Rashba region L, a spin up electron is left in the state:

i / 2
i / 2
1 e R cos( d / 2) + e R
1 i sin( d / 2)
=

+
y
(
)

b ( y )
i R
i R / 2 a
2 i cos( d / 2)e + ie
2 sin( d / 2)

From the state

above, we can easily find (electrons impinging near the crossing)


Spin-down conductance

I ,

e
= eV [1 cos( d / 2) cos R ].
h

Note the additional rotation d = R


due to s-o interband coupling.

2
G (e 2 / h)

d
kc

2
(Usual spin-transistor geometry: single modulation R = 2mL / h )

Weak s-o interband coupling enhances spin control in the Datta-Das transistor.

Two-subband case: details


Two sets of
R2
Rashba bands
R1

Relevant crossing point:

R1, (kc ) = R 2, + (kc ) kc =

kc k2 k

In the Rashba basis:

H eff

2md
= 2
h kc

b a
2

Wave vector
splitting

d
0
0
R1, + (k )

(
k
)
d
0
R1,

=
0
d
R 2, + (k )
0

0
0
R 2, (k )
d

Simplicity (PT): Reduced Hamiltonian (nearly-free electron model)

s-o induced interband coupling


Near the crossing: use PT (N.F.E.)

(k)

d
R1, (k )
~
H red =

(
k
)
R 2, +

h 2k 2 1
(k ) =
+ ( a + b ) d
2m 2
1
=
a + b
2

a ,b

a ,b

y
a ,b
y
a ,b

a ,b

a ,b

+i
i

a ,b

a ,b

Lead 1

kc1 kc kc2 k2

kc 2 = kc + / 2
k c1 = k c / 2

2md
Thin lines: no interband mixing (d=0)
Use PT bands to find: = 2
h kc
[ + (kc1 ) = (kc 2 )]
Dotted curves: PT
8 Blue/pink curves: exact
For infinite transverse confinement: d =
3w (4x4 diagonalization)
(w: channel width)

Spin injection: two bands


strictly 1D case:
(one channel)

eik F x

x 0

1
" a e ikc x "+ + a eik 2 x
2

x 0 +

(k)

s-o interband coupled states

F
kc2 = kc + / 2
k c1 = k c / 2

kc1 kc kc2 k2

Quasi 1D case:
(two channels)

eik F x

x 0

1 1
i ( k c + / 2 ) x
i (kc / 2 ) x
ik 2 x

e
e
+
+
+

a
2 2
x 0 +

"Include only three intersection points" (unity transmission)

Realistic parameters
Infinite transverse confinement:

3h 2 2
,
b a =
2
2mw
Let us choose

8
d=
, and
3w

b a
3h 2 2
kc
=
2
4mw2

b a = 16 R . This choice implies:

(i) (kc ) = 24 R ( F ), (ii) kc = 8 R / , and


(iii) = ( 3 / 4)h2 / mw = 3.451011 eVm(and R 0.39 meV),
( w = 60 nm, m = 0.05m0 )
Assuming

L = 69 nm , we find

(Nitta et al., Engels et al. Sato et al.)

d
d
R = and d = R , since 0.5.
kc 2
kc
In addition,

/ 4k F 0.05 (velocity boundary condition)

Ballistic spin-FET using Rashba coupling


(Datta&Das 1990)
gate

or

FM

HR =

(
p
h
x

p y

FM

Rashba spin-orbit coupling acts on electron spin as an effective


momentum-dependent magnetic field (Beff perpendicular to p).
Problem: Even spin-independent scatterers will randomly change the
momentum and therefore randomize the spin (e.g. Dyakonov-Perel).

Thus: nearly ballistic electron transport is required.

Randomization of spin
2DEG

Beff
HR =

spin

(p
h

p y x

Rashba, 1960

HR =

(
p
h
x

p y

r r r
r r r
r
r
r
r
= E ( p ) = ( E p) g B Beff , with Beff p = 0

disorder/boundaries random walk of electron in real space


random direction of momentum p random direction of Rashba field
Beff random change of spin precession axis random walk in spin
space spin direction randomizes within time T
(Dyakonov-Perel mechanism 71)
2
spin diffusion time: T = 1 / , = g B Beff = k F , = l / vF
T = 0.1 ns , for = 1 ps ( l = 100 nm) , k F = 0.1meV ( Beff = 1T )

Rashba & Dresselhaus spin-orbit in 2DEG

(
p
h

HR =

HD =

(
p
h

p y x

p y

Structure-inversion asymmetry tunable by


gates: ~ E
(Rashba, 1960)
o z

[ ~ 0 . 1 eV A , Nitta et al.97, Engels et al 97 ]

Bulk-inversion asymmetry (Dresselhaus, 1955)

[ pz2 <pz2> & non-parabolicity


(p4) neglected;
o
= < p z 2 > ~ 0 . 1 eV A , Lommer et al. 88 ]
(Pikus&Pikus 95, Averkiev&Golub 99)
x

tune Rashba term s.t. =


2
HR + HD =
( px m p y ) , where := ( x y )
h

is a new conserved quantity (new basis) which makes spin

states of electrons independent of their momentum! (it fixes


the direction of Beff)

J. Schliemann, J. C. Egues, and D. Loss, Phys. Rev. Lett. 90, 146801 (2003)

Tuned Rashba & Dresselhaus coupling:

r2
p
2
r
( p x p y ) + V (r ) , = x + y
H=
+
2m
h
r
V (r ) : arbitrary scalar potential (confinement, impurities,)

1 1 r i
r
i / 4 (r )e
Eigenstates: (r ) =
2 e

2m ( x y ) / h 2

, = 1

spinor is independent of
orbital state k!

h2 2
r r
2 2 m r

+ V (r ) (r ) = + 2 (r )
h
2m

Scalar Schroedinger eq., independent of spin

A robust two-dimensional spin-FET

Vgate
scource
FM

=
2DEG

drain
FM

off state : , spin randomization (e.g. Dyakonov-Perel)


unpolarized electron current
on state : = , spin state protected due to additional conserved
quantity polarized electron current
Several operation modes: source and drain with parallel or antiparallel
spin-polarization; also device switching between = possible.

Beam-splitter geometry
l
ca a
Lo h b
as

Egues, Burkard, and Loss, Phys. Rev. Lett. 89, 176401 (2002)

Band structure

r
Beam
splitter

Entangler

(k )

F
(Liu et al. 98)

Incoming electrons:
(i) entangled pairs:
(ii) spin polarized:

1 2 m 1 2
1 2

Lead 1

kc k

Features:
Manipulate entanglement
Probe/control: shot noise and
spin-polarized transport
Measure s-o coupling
New spin on the Datta-Das

Gate control of Rashba coupling (2D) : Nitta et al. 97; Engels et al. 97;.

Summary and conclusions


Rasba s-o (interband) coupling strongly modulates current (and shot noise)
thus allowing the design of spin FETs with additional control ( R , d );
s-o provides a means of controllably rotating spinors (single qubit rotation).
For pairwise electron states, the Rashba rotation can convert a singlet into an
entangled triplet (and vice versa), relevant for two-qubit manipulations.
The interplay of the gate-tunable Rashba and the Dresselhaus couplings
allows for a partial cancellation of these two effects when their strenghts are
equal. This gives rise to s-o eigenstates with spinors that are independent of
the k vector (suppression of EY spin flip mechanism).
The effective magnetic field Beff is also fixed in space: this suppresses the
DP-type spin relaxation.
Tuned Rashba and Dresselhaus coupling: Nonballistic spin FET.
(Too) Many device proposals; need desperately an experimental realization.
Review article:
Egues, Recher, Saraga, Burkard, Golovach, Sukhorukov, and Loss, in Quantum noise in
mesoscopic Physics, vol. 97, p 241274, Kluwer 2003 (NATO SERIES), cond-mat/0210498.

You might also like