Quasi-Two-Dimensional Extraordinary Hall Ee

t
N. Ryzhanova1,2 , A. Vedyayev1,2 , A. Pertsova2 , B. Dieny1
1

SPINTEC, URA 2512 CEA/CNRS, 38054 Grenoble Cedex 9, Fran e

arXiv:0803.3649v2 [cond-mat.mtrl-sci] 9 Apr 2008

Physi s Department, M.V. Lomonosov Mos ow State University,

Leninskie Gory, 1199991 Mos ow, Russia

(Dated: April 9, 2008)

Abstra t

Quasi-two-dimensional transport is investigated in a system onsisting of one ferromagneti layer

pla ed between two insulating layers. Using the me hanism of skew-s attering to des ribe the
Extraordinary Hall Ee t (EHE) and al ulating the ondu tivity tensor, we ompare the quasitwo-dimensional Hall resistan e with the resistan e of a massive sample. In this study a new
me hanism of EHE (geometri me hanism of EHE) due to non-ideal interfa es and volume defe ts
is also proposed.
PACS numbers: 75.47.-m, 75.70.-i, 73.50.-b

I. INTRODUCTION
Re ently there has been an in reased fo us on the fabri ation of new spintroni devi es
based on the tunnel magnetoresistan e ee t (TMR)1 and the re ently dis overed spin torque
ee t (ST)2 in multilayered stru tures onsisting of several ferromagneti nanolayers separated by thin insulating barriers. The geometri parameters of su h stru tures also give us
the possibility to investigate some of their quasi-two-dimensional transport properties and
parti ularly to ompare the relationship between diagonal and o-diagonal (responsible for
EHE) ondu tivities in massive and above mentioned samples. This study is of great interest to the eld of spintroni s sin e quasi-two-dimensional EHE may provide an additional
me hanism for re ording and storing information in logi devi es. Fo using on this aspe t of
EHE we study both diagonal and o-diagonal (Hall) ondu tivities in a system onsisting of
one ferromagneti layer of thi kness a pla ed between two insulating layers and magnetized
in the dire tion perpendi ular to the interfa es (zdire tion).
The remainder of the paper is organized as follows. In Se tion II EHE is attributed to the
me hanism of skew-s attering3 on the bulk and interfa e impurities. We report our results
for the ondu tivity tensor in luding size ee t terms al ulated within the framework of the
Kubo formalism and ompare the quasi-two-dimensional Hall resistan e H
2D with that of a
massive sample H
bulk . Se tion III dis usses the nonideality of the interfa es and the existen e
of volume defe ts that inuen e the form of urrent lines. We propose a new me hanism
of EHE (we will refer to it as the geometri me hanism of EHE ) due to these defe ts using
the diusion equation4 and taking into a ount that the diusion oe ient has o-diagonal
omponent proportional to the spin-orbit intera tion. We summarize our results and oer
some on lusions in Se tion IV.

II. SKEW-SCATTERING MECHANISM OF EHE IN A THREE-LAYERED

STRUCTURE
We will onsider the geometry of urrent parallel to the ydire tion resulting in the
appearan e of the x omponent of the Hall eld. In this ase:
jy = yy Ey + yx ExH
xy
ExH =
Ey
xx

jx = xy Ey + xx ExH = 0
yx
H
2D,bulk =
2
yy xx + yx

(1)
(2)

where are diagonal and o-diagonal ondu tivities, ExH is the Hall eld. For the al ulation of H
2D,bulk we will use the Kubo formula with vertex orre tions responsible for the
transverse omponent of the urrent:
=

~e2 X

G+
(zz )G (z z)
a40
z

(3)

where ~ is the velo ity ve tor along the interfa e, G+0 (zz ), G0 (zz ) are advan ed and
retarded Green's fun tions in mixed oordinate-momentum representation. To al ulate the
y x omponent of the ondu tivity tensor we will use the perturbation theory and will
take into a ount only the orre tions in linear order of the spin-orbit intera tion:
+

G+
(zz ) = G0 (zz ) + G0 (zz )

+
so
z ) G+
T
z ) + H
(~
(~
0 (z z )

so
z ) = iso (~z ) a20 mz [ ]z = iso (~
z ) a20 mz x y x y
H
(~

(4)
(5)

+
so
z) and H
z) are the s attering matrix and the spin-orbit intera tion, orwhere T
(~
(~
respondently, dependant on the type of atom in (~z) position ; mz is the unit ve tor along
the magnetization, so
m is the spin-orbit parameter and a0 is the latti e onstant. It follows
so
from Eq. (5) that H (~z ) = Hso (~z ).
For the T matrix in one-site approximation we an write down:

T (~z) =

n (z)
1 (n (z)) G (~n z, ~n z)
n
X

z) =
ei( )(~~m ) so
so
m
(~

ei( )(~~n )

(6)
(7)

where n is the one-site energy. For the binary system AB n and so

m take values A,B and
so
~n z, ~n z) =
A,B , respe tively; G (

a20

2
a0

dG (zz). is the oherent potential and an

be found from the system of self- onsistent equations. The rst equation of the system is
valid for s attering on both bulk and interfa e impurities:
cA

B (z)
A (z)
+ cB
=0
1 (A (z)) G0 (~n z, ~n z)
1 (B (z)) G0 (~n z, ~n z)

(8)

while the se ond one is written in the form orresponding to the interfa e s attring sin e we
are interested in al ulation of the interfa e oherent potential:
G (~n z, ~n z) =

G0 (~n z, ~n z)
1 (z) G0 (~n z, ~n z)

(9)

For the al ulation of xx , yy we will use Eq. (3) with Green's fun tions diagonal on
and renormalized on the oherent potential. For o-diagonal omponent averaging on the
impurities distribution gives:
yx =


 


~3 e2 Mz X 2 2  +
+
2
2  +
Im T
~n z )so
~n z )
G
(z
z
)
G
(zz
)

(
(

y x
2 2
a0 m

(10)

~n

where ~n z is the impurity position. We keep in Eq. (10) only the main term with n = m.
For the binary AB stru ture and purely random distribution of A, B summing over ~n
will give . It is onvenient to divide so into average and s attering parts:
so
so
so
so
so

so
A = cA A + cB B + cB A cB B = + cB
cA so
so = cA so + cB so + cA so cA so =
B

(11)
(12)

The average in Eq. (10) is:

1 X
hTn (z)so
n (z)i
N n


A (z)
B (z)

cA
+ cB
1 (A (z)) G (~n z, ~n z)
1 (B (z)) G (~n z, ~n z)


B (z)
A (z)
so

+ cA cB
1 (A (z)) G (~n z, ~n z) 1 (B (z)) G (~n z, ~n z)

(13)

A ording to Eq. (8) and Eq. (9) we only take the imaginary part of the last term in Eq.
(13). We rewrite Eq. (13) with renormalized Green's fun tion:

(A (z)) (1 (z) G0 (~n z, ~n z))
1 X
so
so
hTn (z)n (z)i cA cB Im
Im
N n
1 A G0 (~n z, ~n z)

(B (z)) (1 (z) G0 (~n z, ~n z))

1 B G0 (~n z, ~n z)

G0 (~n z, ~n z) F0 (z)

(14)
(15)

Eq. (6) for oherent potential is:

(E, z) = +
= cA A + cB B ,

cA cB 2 F (E, z)
1 (
(E, z)) F (E, z)

(16)

= cA B + cB A ,

(17)

= A B

Usually it is onvenient to hoose = 0. In this ase = (cA cB ) , A = cB ,

B = cA .
Next we assume that s attering parameters for bulk impurities su h as s attering potential
and on entration are small enough to keep only the main terms for all values. In this ase
bulk
the imaginary part of is of order 2 so for yx
we will use:
1 X
so
2
hTn (z)so
n (z)i bulk bulk cAbulk cBbulk (cAbulk cBbulk )
N n

1
ImF
2
(1 cBbulk bulk ReF ) (1 + cAbulk bulk ReF )2

(18)

For the interfa e values the full self- onsistent s heme is ne essary. For both bulk and
interfa e we suppose that the real part of the oherent potential just represents the renormalization of ele tron spe trum so an be onsidered as purely imaginary.
The zero order Green's fun tion found from S hrdinger equation in z representation
is (we will further use the units with energy dimension [L] = A ):

G+
0 (0 < z < z < z1 ) =

1
2

2ik1 eik1 a (q + ik1 ) eik1 a (q ik1 )2

eik1 (zz1 ) (q ik1 ) eik1 (zz1) (q + ik1 )




eik1 z (q + ik1 ) eik1 z (q ik1 )

(19)

k1 =

kF

2

2kF
+i
c1 + id1
l1

(20)

s
1
 
2
 
 2

2
2
1
4k
c1 =
kF 2 + F +
kF 2
l1
2

s
1
 
2
 
 2

2
2
1
4k
d1 =
kF 2 + F
kF 2
l1
2
q
2m
q = q02 + 2 ,
q02 = 2 (U EF )
~

(21)
(22)
(23)

where 0 and z1 are the oordinates of the left and right interfa es, a is the layer thi kness; kF ,

l1 are the Fermi momentum and the mean free path for spin "up", respe tively, c1 d1 = l1F
(for spin "down" we will use index 2 ); U is the height of the potential barrier.
The poles of the Green's fun tion in Eq. (19) dene the quantized energy spe trum of
the thin ferromagneti layer.

A. Cal ulation of the bulk quasi-two-dimensional diagonal ondu tivity

For xx = yy in Eq. (3) we will take into a ount s attering on the interfa e responsible for size ee t as well as on the bulk of the sample by using the Dyson equation with
renormalized Green's fun tion:
G (zz ) = G0 (zz ) + G0 (z0) G (0z ) = G0 (zz ) +

G0 (z0) G0 (0z )
1 G0 (00)

(24)

Integrating over z from 0 to z for z < z and from z to a for z > z gives the ondu tivity
in the units ohm1 cm1 :
Z 3
dNom
0 l1 108

xx =
(25)

c1 Den

2kF


q 2 + c21 + ||2 + 2qRe sinh 2d1 a + 2c1 |Im| cosh 2d1 a




+ 2c1 Im q 2 + c21 cosh 2d1 (z a) + 2 sinh2 d1 z (cos 2c1 (z a) + 2qc1 sin 2c1 (z a))




2c1 (q + Re) q 2 + c21 sinh 2d1 (z a) sin 2c1 z + q 2 + c21 + ||2 + 2qRe







q 2 + c21 sinh 2d1 (z a) cos 2c1 z sinh 2d1 z q 2 c21 cos 2c1 (z a) + 2qc1 sin 2c1 (z a)

Nom = q 2 + c21



(26)

Den = q 2 + c21
q 2 + c21 + ||2 + 2qRe cosh 2d1 a + 2c1 |Im| sinh 2d1 a








q 4 6q 2 c21 + c41 + ||2 + 2qRe q 2 c21 4qc21 Re cos 2c1 a




2qc1 2 q 2 c21 + qRe + ||2 sin 2c1 a

(27)

4x104

-1

-1

(ohm . cm )

5x104

xx

3x104

2x104

10
a (

15

20

Figure 1: Averaged diagonal ondu tivity as a fun tion of a (thi kness): < xx > (a) for kF =
, l2 = 60(A)
, c = 0.3 (see Eq. (28))
1.1(A 1 ), kF = 0.6(A 1 ), l1 = 100(A)
e
= 10
(ohm1 ) is the elementary ondu tivity of one hannel.
where 0 = 2~
13.6
For large enough layer thi kness we an average Eq. (25) over os illations so that the
averaged ondu tivity is:
3

 0 l1 108
xx =
2kF

#
"
3 d
l1
|Im| c1 sinh 2d1 a


1
c1
a q 2 + c21 + ||2 + 2qRe sinh 2d1 a + 2c1 |Im| cosh 2d1 a

(28)
The rst term in Eq. (28) is the ondu tivity of the massive sample and the se ond one
is due to the quasi- lassi al size ee t. The full ondu tivity representing the sum of two
spin hannels is shown on Fig. (1) as a fun tion of the layer thi kness.

B. Cal ulation of the o-diagonal ondu tivity due to the spin-orbit interfa e s attering

(z) using Eq. (10) with Green's fun tion dened by Eq. (24)
Now we will al ulate xy

and z = 0. After integration over z the ondu tivity is:

xy

=
Z

0 l1 a40 108
2 2 kF
3 d

Im hT (~n 0) so (~n 0)i (q 2 + c21 ) sinh 2d1a

3 dc1 



 ik1 a

ik1 a (q ic )2 1 +
(q + ic1 )2 1 + q+ic

e
e
1
1

qic1

2



(q 2 + c21 ) cosh 2d1 (z a) (q 2 c21 ) cos 2c1 (z a) 2c1 q sin 2c1 (z a)

(29)



2


 ik1 a

ik1 a (q ic )2 1 +

e
(q + ic1 )2 1 + q+ic

e
1
qic1
1

15

xy

-1

-1

(ohm . cm )

20

10

5
5

10
a (

15

20

Figure 2: O-diagonal ondu tivity as a fun tion of a (thi kness): xy (a) for kF = 1.1(A 1 ),
, l2 = 60(A)
, c = 0.3, so = 0.05(A 1 ) (see Eq. (29))
kF = 0.6(A 1 ), l1 = 100(A)

This ondu tivity os illates with the thi kness and the distan e from the interfa e z = 0.
Its behavior be omes more lear after averaging over os illations:
Z

 0 l1 a40 108
Im hT (~n 0) so (~n 0)i sinh 2d1 a
3

dc
xy =


1
q 2 + c21 + | |2 + 2qRe sinh 2d1 a + 2c1 Im cosh 2d1 a
8 2 kF
Z
cosh 2d1 (z a)


3 d
2
2
q 2 + c1 + | | + 2qRe sinh 2d1 a + 2c1 Im cosh 2d1 a

(30)

The same is done for spin "down". The sum of these two terms is shown on Fig. (2).

de reases with z a sin e the fun tions cosh 2d1 (za) have maximum
It is lear that xy

over these fun tions gives the fa tor la1 so for innite a this
values at z = 0. Averaging xy
term tends to zero.

+ xy
(see Fig. (3)) using Eq. (10) with
We also al ulate the bulk ondu tivity xy

additional integration over z and bulk s attering parameters with the bulk oherent poten2
tial in Bohrn approximation bulk = icbulk (1 cbulk )bulk
ImF bulk (zz). In the absen e of the
interfa ial s attering this approa h gives us:

200

xy

bulk

-1

-1

(ohm . cm )

300

100

10
a (

15

20

bulk (a) for k = 1.1(A 1 ),

Figure 3: Bulk o-diagonal ondu tivity as a fun tion of a (thi kness): xy
F
, cbulk = 0.01, bulk = 1(A 1 ) (see Eq.
, l2 = 60(A)
, so = 0.03(A)
kF = 0.6(A 1 ), l1 = 100(A)
bulk
(31))

bulk
xy

 8Z
Z 3
2


0 l12 a30 Im T bulk so
3 d (q 2 + c21 ) sinh 2d1 a
d 1 n 2
bulk 10
2 2
=

q
+
c
1
2
c1
Den
c1 Den
8 2 kF


sinh 2d1 a q 4 c41 (sinh 2d1 z cos 2c1 (z a) sinh 2d1 (z a) cos 2c1 z)

+2qc1 (sinh 2d1 z sin 2c1 (z a) + sinh 2d1 (z a) sin 2c1 z)}

(31)

Den = cosh 2d1 a q 2 + c21

2

q 4 6q 2 c21 + c41 cos 2c1 a + 4c1 q q 2 c21 sin 2c1 a

(32)

or after averaging over os illations:

bulk
xy

 8 Z
2
0 l12 a30 Im T bulk so
3 d
bulk 10
=
c1
8 2 kF2

(33)

This ondu tivity has an os illating behavior for the thin layer but tends to the onstant
value when a whi h oin ides with its value for the massive sample. If we take into
bulk
a ount the interfa ial s attering the expression for xy
be omes too ompli ated so we
H
xy
bulk
al ulated using
don't show it here. But the thi kness dependen es of xy and = xx
the full formula with renormalized Green's fun tion are presented at Fig. (4) and Fig. (5),
1 bulk = 0.01, bulk = 1(A 1 ).
orrespondently. The bulk parameters are: so
bulk = 0.03(A ), c

250

xy

bulk

-1

-1

(ohm . cm )

300

200

150

10

15

a (

20

Figure 4: Bulk o-diagonal ondu tivity with interfa ial s attering as a fun tion of a (thi kness):
bulk (A) for k = 1.1(A 1 ), k = 0.6(A 1 ), l = 100(A)
, l2 = 60(A)

xy
1
F
F

-2

-2

1.0x10

xy

xx

1.5x10

-3

5.0x10

10

a (

15

20

1
Figure 5: Hall angle = xx as a fun tion of a (thi kness); bulk parameters: so
bulk = 0.03(A ),
cbulk = 0.01, bulk = 1(A 1 ); interfa e parameters: so = 0.05(A 1 ), c = 0.3, = 1.5(A 1 )
H

xy

III. GEOMETRIC MECHANISM OF EHE

Let us onsider an ele tri urrent through a thin ferromagneti layer of thi kness a
lo ated between two thin insulating barriers and magnetized in zdire tion perpendi ular
to the interfa es. The interfa es are not ideal and besides the impurities have topologi al
9

defe ts whi h will be modeled as ylinders of radius R so that the urrent lines in the vi inity
of these defe ts follow their shape.
Diusion equations for harge and spin urrents in the absen e of pre ession are:
n jex jey
+
+
=0
t
x
y
m
~
j x
j y
m
~
+ m + m =
t
x
y
sf

For the stable state solution

n
t

m
~
t

(34)
(35)

= 0. For urrents we have the system of equations:

n
m
m
n
Dxx
Dxy
Dxy
x
x
y
y
n
m
n
m
jey = yy 0 + yx E Dyy
Dyy
Dyx
Dyx
y
y
x
x
m
m
n
n
x
Dxx
Dxy
Dxy
jm
= xx E Dxx
x
x
y
y
n
m
n
m
y
jm
= yy 0 + yx E Dyy
Dyy
Dyx
Dyx
y
y
x
x
jex = xx E Dxx

(36)
(37)
(38)
(39)

~ = {E, 0, 0}, m
Here we take into a ount that E
~ = {0, 0, m} are the ele tri eld and
spin a umulation orrespondently, D are the omponents of diusion oe ient tensors.
O-diagonal omponents Dxy and Dyx of these tensors are proportional to the spin-orbit
intera tion and they are antisymmetri al in the x y transposition. For a metal with the
ubi symmetry xx = yy , Dxx = Dyy D0 . Then we insert Eq. (36)-(39) into Eq. (34)
and Eq. (35) and after some manipulations we obtain two equations:
n = m

m
n + m =
sf

And for sf D0 (1 2 ) = 2sf :

m
=0
2sf

(40)
(41)
(42)

For the ylindri al defe t shape it is onvenient to sear h for solution in polar oordinates
so we an rewrite Eq. (42):
1
r r

m
r
r

1 2m
m
1 2m
m
2 m 1 m
+ 2
2
+
+ 2
2 =0
2
2
2
r
sf
r
r r
r
sf

(43)

where is the angle between xaxe and the radius-ve tor ~r with the oordinates (x, y).
The solution of Eq. (43) is:
m = m1 (r) m2 ()
m2 () = A1n cos n + A2n sin n

10

(44)
(45)

As

2 m2
2

= m2 n2 , Eq. (43) an be transformed:

! #
"
n2
1
2 m1 1 m1
+

+ 2 m1 = 0
m2 ()
r 2
r r
2sf
r

(46)

The solution of Eq. (46) is5 :

m1 (r) = Bk Kk
m=

r
sf

(A1n cos n + A2n sin n) Kk

(47)


r
sf

(48)

where Kk ( rsf ) is the solution of the modied Bessel equation5 .

From Eq. (40) it follows that:
(49)
(50)

n = m + n0
n0 = 0

For n0 the solution is:

n0 =

(C1n cos n + C2n sin n)

1
rn

(51)

Taking into a ount Eq. (40) we an rewrite Eq. (36) and Eq. (37):
n0
n0
Dxy
x
y
n0
n0
+ Dyx
jey = yx E Dyy
y
x

jex = xx E Dxx

(52)
(53)

Now it is onvenient to use the polar oordinate system and to write down r and
proje tions of the urrents. Then we an use the boundary onditions to nd unknown
oe ients. These proje tion are:
0
jre
= xx E cos + xy E sin

(54)

(usual term)
jre = Dxx

n0
n0
n0
n0
cos Dxy
cos Dyy
sin + Dxy
sin
x
y
y
x

(55)

(additional diusion term)

Now we will make some transformations:
n r
n
n
=
+
x
r x x
n r n
n
=
+
y
r y y

11

(56)
(57)

Using the expressions for derivatives of r, over x, y , whi h is not too di ult to obtain,
we write down the harge and spin urrents in polar oordinates:



n r n
n
n r
+ sin
+
+
jre = D0 cos
r x x
r y y





n r
n r n
n
Dxy cos
sin
+
+
r x x
r y y
n
n
Dxy
= D0
r
r


n
n
Dxy
r
r
m
m
Dxy
= D0
r
r

(58)

n
jrm
= D0

(59)

m
jrm

(60)

0
jrm
= xx E cos + yx E sin

(61)

To nd the unknown oe ients in Eq. (48) and Eq. (51) we will use the boundary
onditions on the surfa e of the ylinder representing that rproje tion of urrents are
equal to zero:
X
n0
n
n0
|r=R + Dxy
|r=R = D0
(C1n cos n + C2n sin n) n+1
r
R
R
n
X
n
Dxy
(C1n sin n C2n cos n) n+1
R
n

n
jR0
+ jRn = 0 D0

(62)

It gives us the system

xx E = R2 (D0 C11 Dxy C21 )
xy E = R2 (Dxy C11 + D0 C21 )

(63)
(64)

with solution
C11 =
n0 =

D0 xy + Dxy xx
D0 xx Dxy xy
, C21 =
2
2
2
D0 + Dxy
D02 + Dxy

R2 E (D0 xx Dxy xy ) cos (D0 xy + Dxy xx ) sin

2
r
D02 + Dxy

(65)
(66)

Spin urrent rproje tion is also zero, and we an write down:

m
m
jR0m + jRm = 0 E (xx cos xy sin ) = D0
|r=R + Dxy
|r=R
r
R




1
r
R

|r=R + Dxy (A11 sin + A21 cos ) K1

= D0 (A11 cos + A21 sin ) K1
r
sf
R
sf

(67)

12

Then we use some properties of Bessel fun tions5 :

I0 (x)
2

(I1 I+1 ) = lim

I (x) =
(68)
0 sin x
0 2 sin ( + 1)
x
K1 (x)

I1 (x)

2
= lim
(I1 I+1 ) = lim
I (x) =
(69)
0 2 sin ( + 1) x
0 sin x x
x
x

K1 (x) = lim

where x rsf , x
= 1sf .
r
Inserting Eq. (68)-(69) in Eq. (67) we get:
I1
R
I0 sf
+ Dxy (A11 sin A21 cos )
R2

E (xx cos xy sin ) = D0 (A11 cos + A21 sin )

(70)

and it follows that:

I0 sf
I1
Dxy A21 2
R
R
I1
I0 sf
= Dxy A11 2 D0 A21
R
R

Exx = D0 A11

(71)

Exy

(72)


 
 

r
r
ER xx D0 I1 sf R + xy Dxy I0 sf sf
=

  2 
2
 
D0 I1 rsf R + Dxy I0 rsf sf
  

 
r
2
ER xx Dxy I0 sf sf xy D0 I1 rsf R
=

  2 
2
 
D0 I1 rsf R + Dxy I0 rsf sf
2

A11

A21

 
ER2 sf I0 rsf
m = 
2 
 
  2 
r
r
r
D0 I1 sf R + Dxy I0 sf sf







r
r
R + xy Dxy I0
sf
cos xx D0 I1
sf
sf


 


r
r
sf xy D0 I1
R
+ sin xx Dxy I0
sf
sf

(73)

(74)

(75)

Now we an dene the additional Hall eld due to this ylindri al interfa e defe t onsidering that Hall ele trodes are the surfa es with oordinates y = a and y = a. This eld
is proportional to n(a) n(a), n = m + n0 , r = |sina | . After integrating over from 0
to for the left surfa e and from to 2 for the right one we will have:
ER2
n0 (a) n0 (a) = 2
a

ER2 (Dxy xx + D0 xy )
sin2 (Dxy xx + D0 xy )


=
(76)
2
2
D02 + Dxy
a D02 + Dxy

13

The se ond term due to m is:

nm (a) nm (a) =

2 ER
a

  

 
sin2 xx Dxy I0 rsf sf xy D0 I1 rsf R
(77)
d

  2 
2
 
r
r
D0 I1 sf R + Dxy I0 sf sf

At last, we have to multiply Eq. (76) and Eq . (77) by the on entrations of defe ts and
ele tron harge.

IV. CONCLUSION
It was shown that due to the additional s attering of ele trons on the defe ts of the
metal-insulator interfa es the total ondu tan e de reases. From Eq. (28) it follows that
for small values of the ratio al the bulk ondu tivity is ompletely suppressed and ee tive
ondu tivity is proportional to the ee tive s attering length on the interfa es instead of the
bulk mean free path. Hall ondu tivity, if we don't take into a ount the additional s ews attering on the interfa e, de reases with de reasing the thi kness of the ferromagneti
metalli layer. However the ontribution to the Hall ondu tivity due to the additional
s ew-s attering on the interfa e in reases. So the important hara teristi of the onsidered
xy
xy
devi e, Hall angle xx
= xx
, is larger for the thin ferromagneti layer ompared to the bulk
layer. Besides that, the inuen e of insulator olumns penetrating into the metalli layer
may further in rease the value of the Hall ee t.

A knowledgments
N. Ryzhanova and A. Vedyayev are grateful to SPINTEC for hospitality. This work was
partially supported by the Russian Foundation for Basi Resear h.

1
2
3
4
5

J. S. Moodera, L. R. Kinder, T. M. Wong and R. Meservey, Phys. Rev. Lett. 74, 3273 (1995).
J. C. Slon zewski, J. Magn. Magn. Mat. L1, 159 (1996).
R. Karplus and J. M. Luttinger, Phys. Rev. 95, 1154 (1954).
S. Zhang, P. M. Levy and A. Fert, Phys. Rev. Lett. 88, 236601 (2002).
A. N. Tikhonov and A. A. Samarskii, Equations of Mathemati al Physi s (PERGAMON PRESS,
Oxford-London, New-York, Paris, 1963), p. 656.

14