a

r
X
i
v
:
0
9
0
3
.
2
4
6
8
v
1


[
c
o
n
d
-
m
a
t
.
m
e
s
-
h
a
l
l
]


1
3

M
a
r

2
0
0
9
Mach Zehnder spin interferometer for Rashba and Dresselhaus media: Exact solutions
for perfect spin ltering
Alexander L opez,
1
Ernesto Medina,
1, 2, 3
Nelson Bolvar,
2
and Bertrand Berche
3
1
Centro de Fsica, Instituto Venezolano de Investigaciones Cientcas, Apartado 21874, Caracas 1020-A, Venezuela
2
Departamento de Fsica, Universidad Central de Venezuela, Caracas, Venezuela
3
Statistical Physics Group, P2M, Institut Jean Lamour,
Nancy Universite, BP70239, F- 54506 Vanduvre les Nancy, France
Spin ltering through quantum spin interference is addressed exactly, in two dimensions, in a
medium that has both Rashba and Dresselhaus spin-orbit couplings and an applied external magnetic
eld. The problem is addressed from the fully non-Abelian Gauge formulation. We propose an
experimentally feasible electronic Mach Zehnder Interferometer and solve for the perfect spin ltering
conditions. We nd two broad solutions, one where ltering is achieved in the original incoming
basis, that is purely a non-Abelian solution, and the other where one needs a tilted axis to observe the
polarized output spinor. The latter solution is well approximated by an Abelianized approximation.
Both solutions apply for arbitrary incoming polarization, and are only limited by the randomness
of the incoming spinor state.
PACS numbers: 03.65.Vf
I. INTRODUCTION
The Rashba and Dresselhaus SO interactions arise in
materials which lack either structural or bulk inversion
symmetry, respectively[1, 2, 3]. These two kinds of in-
teractions have recently been given a great deal of at-
tention due to their potential role in the generation and
manipulation of spin polarized currents, spin lters[4],
spin accumulation[5], and spin optics[6].
A reformulation of the spin-orbit coupling Hamilto-
nian in terms of non-Abelian gauge elds[7] was explic-
itly given in ref. [8, 9, 10, 11] where the SO interaction
is presented as a SU(2) U(1) gauge theory. As the
Yang-Mills gauge theory is well understood and is the
underpinning of well established theory, enormous insight
can be brought upon new problems. Such gauge point of
view, in more general terms, has been known for some
time[12, 13, 14]. This formulation is very revealing, since
the consistent gauge structure of the theory becomes ob-
vious and the physics of spin currents, persistent currents
and color diamagnetism[15] can be understood in a man-
ner analogous to the well known U(1) gauge theories. A
consistent SU(2) U(1) gauge approach was presented
in reference [10, 11] where it was found that for the Pauli
type Hamiltonians (including Rashba and 2 dimensional
reductions of the Dresselhaus Hamiltonian), Gauge Sym-
metry Breaking (GSB) is necessarily built into the the-
ory and leads to vanishing of the spin conductivity in
constant electric elds[11]. In addition, the Yang Mills
interpretation of the Rashba and Dresselhaus SO interac-
tions renders the associated gauge elds real, with topo-
logical consequences analogous to the Aharonov Casher
eect[10, 11].
Recent proposals have been made for the construction
of perfect spin lters based on active Rashba spin orbit
media[16], ballistic spin interferometers[17] and the anal-
ysis of the persistent spin helix[18, 19], where the Yang
Mills gauge point of view is advantageous. Here we read-
dress the problem of spin ltering by interferometry in
a quasi two dimensional system and make connection to
an experimentally feasible test of these ideas through an
electronic Mach Zehnder[20] interferometer (MZI) within
Rashba and Dresselhaus media. Our analysis, within this
setup, enables us to obtain exact conditions for spin l-
tering which can be achieved by tuning appropriate ex-
perimental parameters. Such conditions for spin ltering
greatly generalize previous special situations where the
spin polarization is a conserved quantity[21], and show
new possibilities for spin ltering beyond previous ap-
proximate treatments.
The structure of the paper is as follows. First we con-
sider the Hamiltonian with both Rashba and Dresselhaus
interactions for a two dimensional electron gas (2DEG)
including a magnetic ux described by a U(1) gauge eld,
following the approach given in ref. [19]. We rewrite the
Rashba and Dresselhaus contributions in terms of a Yang
Mills gauge eld and review how this approach leads to
the introduction of a GSB term analogous to the Proca
term for the Maxwell eld. Then, we propose an inter-
ference setup in the form of an electronic MZI where the
electrons spin transport is modulated due to the pres-
ence of Rashba and Dresselhaus active media. We derive
the conditions for perfect spin ltering and discuss and
compare our results with those found by Hatano, Shi-
rasaki and Nakamura[16], Chen[19] and Ting[21].
II. YANG MILLS FORMULATION, GSB AND
SPIN CURRENT
We consider a two dimensional system consisting of
non interacting electrons subject to both Rashba and
Dresselhaus spin orbit interactions. In addition, there is
an external transverse magnetic ux
B
described by a
U(1) gauge vector potential A. Addressing a single par-
2
M
L
QPC
2
QPC
1
1
D
2
2
M
1
D

I
II
FIG. 1: Sketch of the electronic Mach Zenhder interferometer
setup. The arms of the square are made of active SO Rashba
and Dresselhaus media. The beam splitters are implemented
through two Quantum Point Contacts (QPCs). There is a
magnetic ux B through the square.
ticle Hamiltonian including such couplings we can write
H =

2
2m

+V +(
x

x
)+(
y

x
), (1)
where = p eA, m

is the electrons eective mass,

V a substrate lattice potential that can be assumed pe-
riodic, = (
x
,
y
,
z
) a vector of Pauli matrices, and
and are material-dependent parameters character-
izing the Rashba and Dresselhaus interactions, respec-
tively. The linear in k term describing the Dresselhaus
interaction results from averaging a cubic in k contribu-
tion (for the bulk) in the conning direction and neglect-
ing other cubic terms in the strong lateral connement
situation[22].
Following [19], we introduce a non-Abelian (spin de-
pendent) gauge eld WW whose components are given by
gWW
a

a
= m

_


__

y
_
(2)
with
a
=
a
/2, and g/h is the SU(2) coupling constant.
Using this gauge eld we can rewrite equation (1) in the
form
H =
(p eAgWW
a

a
)
2
2m

+eA
0

g
2
WW
a
WW
a
8m

. (3)
The rst term describes the total kinetic energy taking
into account the contribution of the non Abelian gauge
eld; the second term is the background lattice potential
whereas the third term represents a mandatory gauge
symmetry breaking contribution similar to the Proca
eld for the U(1) electromagnetic gauge eld originally
discussed in references [10, 11]. Such a term is a
neccessary ingredient of the gauge theory in order to
obtain the correct physical currents and ground state
properties[15, 23].
III. ELECTRONIC MACH ZEHNDER SPIN
INTERFEROMETER AND PERFECT SPIN
FILTERING
A device conguration that permits us addressing the
problem of spin ltering in a gauge independent[11] man-
ner is the Mach Zehnder Interferometer (MZI). The setup
for an electronic MZI is sketched in (Fig.1). Here we are
interested in determining the resulting amplitude
Di
at
detector D
i
, with i = 1, 2 and to nd the conditions for
perfect spin ltering [16] at either detector. There is an
interesting issue that must be discussed regarding spin
1/2 ltering. If the state at the input is a pure state
spinor of spin 1/2 the electron is polarized on some in-
determinate axis, in principle random, coming from the
Fermi sea of the input conductor. If one could nd this
axis for every electron extracted then one would have a
perfect spin lter for each electron, nevertheless the re-
sulting current is unpolarized. We thus dene the spin
lter as one acting on any entering (pure state) polar-
ization and returning a polarized state along a denite
axis. This approach will serve to build a polarized spin
current.
The relevant processes within the interferometer are
described as follows (see Fig.1): Single electrons are as-
sumed to be extracted from the Fermi sea as pure states

0
=
_

+
0

0
_
. Experimentally, this can be accomplished
by applying an external voltage. The electrons then
pass through a beam splitter that can be implemented
by a Quantum Point Contact (QPC
1
) described by a
2 2 scattering matrix S
1
. The resulting beams follow
path I (II) that consists of a rst horizontal L
I
(verti-
cal L
II
) arm made of Rashba-Dresselhaus media whose
length is L
I
(L
II
). The electrons are then specularly
reected from an ideal mirror M
1
(M
2
) followed by a
vertical L

I
(horizontal L

II
) arm of length L
I
(L
II
) of
the same material. Then the electrons pass through a
second QPC (QPC
2
) described by the corresponding S-
Matrix S
2
. Finally, two electron beams are collected at
detector D
i
(i = 1, 2), and we have
Di
=
I,i
+
II,i
,
where,
I,i
(
II,i
) is the corresponding transfered spinor
through the ith-arm. These amplitudes can be written in
terms of the injected spinor
0
as
Di
= U
Di

0
, where
the 2 2 matrices U
Di
are given by
U
D1
= (t
2
) exp
_
i
h
_
L

I
dl(gWW
a

a
)
_
(t
1
) exp
_
i
h
_
LI
dl(gWW
a

a
)
_
+
3
(r
2
) exp
_
i
h
_
L

II
dl(gWW
a

a
)
_
(r
1
) exp
_
i
h
_
LII
dl(gWW
a

a
)
_
,
U
D2
= (r
2
) exp
_
i
h
_
L

I
dl(gWW
a

a
)
_
(t
1
) exp
_
i
h
_
LI
dl(gWW
a

a
)
_
+
(t
2
) exp
_
i
h
_
L

II
dl(gWW
a

a
)
_
(r
1
) exp
_
i
h
_
LII
dl(gWW
a

a
)
_
. (4)
Note that U
Di
is not a unitary operator. The normaliza-
tion condition |
D1
|
2
+|
D2
|
2
= 1 for the total probabil-
ity at the detectors require that U

D1
U
D1
+U

D2
U
D2
= 1l,
the unit matrix. This simply means that the amplitudes
received at the detectors do not interfere. The U(1)
gauge potential A describes a uniform and perpendic-
ular magnetic ux. At this point we must stress that the
external magnetic eld B, does not act on the ring itself,
like in the Aharonov-Bohm experiment, thus it does not
interact with the spin or charge of the electrons. This
assumption is made clear from Hamiltonian (eq.1) be-
cause there is no Zeeman contribution. In addition, if
the electric eld E is uniform and static, the operators
= p eA and gWW
a

a
commute. Thus, we can sepa-
rate the orbital from the internal translation operators.
For simplicity we will assume a square interferometer,
thus L
1
= L
2
= L. Otherwise there are no restrictions or
approximations related to the dimensions of the arms of
the interferometer[16]. As in Chen and Chang [19] we will
make the discussion general by treating both the Rashba
and Dresselhaus spin-orbit coupling on equal footing.
Concerning the orbital contribution it is easy to see
that this will consist of a global phase exp[p (L
1
+L
2
)]
which we can drop, and a relative U(1) phase
B
which
arises from the noncommutation of p and A. Using
the denition for the magnetic ux
B
= BL
2
and
that for the ux quantum
0
= h/e, the nontrivial or-
bital phase is written as 2
B
= 2
B
/
0
. On the
other hand, the internal part gives rise to the SU(2)
spin-dependent phase contribution. In order to simplify
the resulting expressions, we introduce the adimensional
variables = (m

L/h)
_

2
+
2
, and = tan
1
(/)
along with the matrices
1
= cos
x
sin
y
and

2
= sin
x
cos
y
, such that (
i
)
2
= 1l, with 1l the
identity matrix. After the previous considerations we can
rewrite Eq. 4 in the form
U
D1
= t
1
t
2
exp(i
1
) exp(i
2
) +
r
1
r
2
exp(2i
B
) exp(i
2
) exp(i
1
),
U
D2
= t
1
r
2
exp(i
1
) exp(i
2
) +
r
1
t
2
exp(2i
B
) exp(i
2
) exp(i
1
).
Due to the symmetry of these expressions (U
D2
is ob-
tained from U
D1
by the substitutions r
2
t
2
) we can
focus on the rst process, and obtain the second by
making the necessary substitutions. Using the identity
exp(i
n
) = cos 1l i
n
sin , valid also for our redef-
inition , the matrix U
D1
takes the form
U
D1
= t
1
t
2
[cos
2
1l i sincos (
1
+
2
)
1

2
sin
2
] +
r
1
r
2
e
2iB
[cos
2
1l i sincos (
1
+
2
)
2

1
sin
2
]
Now, we can easily determine that
1

2
= sin21l
i
z
cos 2 thus
2

1
= sin 21l +i
z
cos 2 and
1
+
2
=
(cos + sin)(
x

y
). Substituting these results and
rearranging the obtained expressions leads to
U
D1
= A
+
[cos
2
sin
2
sin2]1l +i sinIM,
where we have introduced the traceless matrix IM =
A

sin cos 2
z
A
+
cos (cos + sin)(
x

y
) and
A

= t
1
t
2
r
1
r
2
e
2iB
. The traceless condition sim-
plies the diagonalization of IM, and the eigenvalues for
U
D1
are easily found to be

D1

= A
+
[cos
2
sin
2
sin2] i sin
_
A
2

sin
2
cos
2
2 + 2A
2
+
cos
2
(1 + sin 2). (5)
If we now dene B

= t
1
r
2
r
1
t
2
e
2iB
, the eigenvalues
of the matrix U
D2
are obtained from the previous result
by making the substitution A

D2

= B
+
[cos
2
sin
2
sin2] i sin
_
B
2

sin
2
cos
2
2 + 2B
2
+
cos
2
(1 + sin 2). (6)
In order to get more insight into the nature of the
conditions for perfect spin ltering we will specialize
the previous expression to symmetric beam splitters i.e.
r
1
= r
2
= r, and t
1
= t
2
= t. Within this case we
have A

= t
2
r
2
e
2iB
. Since we are interested in
ltering one spin component, say the up component, we
now proceed to determine the vanishing conditions of the
corresponding eigenvalue
D1
+
.
From expressions (5, 6), these vanishing conditions can
be found by either having cos = 0 or cos = 0. In the
rst case, the corresponding U
D1
becomes diagonal with
respect to the original quatization axis, and we speak of
a ltering along a non-tilted axis. The other situation
corresponds to nding a new axis where the up spin is
ltered and we call such axis the tilted quantization axis.
Note that both these ltering conditions (non-tilted and
tilted) are independent of the polarization axis of the in-
coming state. We will comment further on this below.
4
A. Non-tilted ltering
Let us rst analyze the non-tilted situation. In this
case the ltering condition requiring
D1
+
= 0 (see Eq. 5)
leads to the relation
tan 2 =
i(t
2
r
2
e
2iB
)
(t
2
+r
2
e
2iB
)
.
Two 50 50 beam splitters for which r = i/

2, t =
1/

2, will then lead to the relation sin

B
sin 2 =
cos
B
cos 2, equivalent to the simple expression
cos(
B
+ 2) = 0, satised by the condition

B
+ 2 = (2n + 1)

2
. (7)
Figure 2 depicts the relation between the spin-orbit pa-
rameters and the magnetic ux, for n = 0, necessary for
perfect ltering the up component in the original quan-
tization axis. The spin-orbit parameters are in a reason-
able range as depicted since for a GaAs heterostructure
h 3.9 10
12
eV m[24] and h 2.4 10
12
eV m
and h
2
/m

L 1.7 10
12
eV m assuming the arm of
the interferometer 10
6
m[20] and an eective mass of
m

= 0.046m
0
. These parameters yield ||, || < 6 in
units of h/(m

L). Note that our denition of , diers

by a factor h to the standard denition, see Eq.1.
The solutions are on a helix as can be shown from the
previous relations where
=
h
m

L
_
(2l + 1)/2 cos[/4(2n + 1 2
B
)],
=
h
m

L
_
(2l + 1)/2 sin[/4(2n + 1 2
B
)]. (8)
The integer n was dened in eq.7 while the second integer
l is dened by the condition cos = 0.
The previous conditions depicted in Fig.2 do not tell us
about the intensity of the signal received in detector D
1
i.e. the eciency of the lter given an incident intensity.
For this, one has to look back at the eigenvalues. While

D1
+
= 0 the amplitudes of the outgoing polarized spinor
is given by

D1
=
_
0

D1

0
_
=
_
0
ie
iB
cos(
B
2)

0
_
, (9)
whose modulus squared is cos
2
(
B
2)|

0
|
2
. Fig.3
shows a polar plot for the amplitude of the ltered signal
(radius vector) as a function of the parameter designating
the eld ux
B
and the , combination given by Eq.8
for n, l = 0. The gure shows that while ltering occurs
for all the uxes (given the appropriate values of , )
the amplitude can be zero or very small for some ux
values. On the other hand, for some values of the ux,
ltering can be very strong since the probability for a
polarized spin can approach unity. The behavior of the
second detector D
2
while the rst detector sees a ltered
signal can be seen through the resulting condition for
-1
0
2
4
6
8
-1
0
1
-1
0
1

FIG. 2: Perfect ltering by interference for the non-tilted axis

(original incoming basis). The plot shows the relation be-
tween , in units of h/(m

L), and B that yields perfect

polarization of the spin from an unpolarized input. The gure
corresponds the the values n, l = 0 according to Eqs.8.
-1 -0.5 0.5 1
-1
-0.5
0.5
Spin Flip on
detector D
1
FIG. 3: Filtering probability for the non-tilted solution of de-
tector D1 assuming the entering component |

0
| = 1. The
radius vector depicted shows the ltered probability for the
output spinor for one whole period in the parameters , as
given in the Fig.2. The position of the dashed vector cor-
responds to B = 0.25. The grey points represents spin
ipping or opposite ltering solutions for detector D2.
the eigenvalues of that detector having substitued the
condition
D1
+
= 0

D2
+
= ie
iB
,

D2

= ie
iB
sin(
B
2). (10)
It is obvious that the second detector D
2
does not l-
5
-6 -4 -2 0 2 4 6
-6
-4
-2
0
2
4
6

FIG. 4: Perfect ltering by interference for the tilted axis.

The plot shows the relation between , in units of h/(m

L),
and sin B in a contourplot, the darker regions indicate
larger values for the magnetic ux needed to yield perfect
ltering, from an unpolarized input. Highlighted circle and
diagonal lines depict the zero ux solutions that yield perfect
ltering.
ter concomitantly with the D
1
, in general. Furthermore
one can only nd conditions for the second component
to be zero (opposite ltering to detector D
1
) since the
rst component has modulus one. This takes us to the
non-tilting spin ipped or opposite ltering solution at
detector D
2
, only occurring while detector D
1
is ltering
with maximal eciency i.e. maximal polar radii in Fig.3.
The ltering amplitude is proportional to the projec-
tion of the incoming spinor (which has arbitrary weights
onto the chosen quantization axis) to the surviving com-
ponent at the output (see Eq.9). This means that for each
arbitrary incident spinor from the Fermi sea one gets a
ltering probability that depends on this projection. The
resulting polarized current will thus have a random noise
associated with this eect besides the contribution from
shot noise.
It is important to note that this solution does not ap-
pear in the Abelian approximation in references [16, 19],
and thus is intrinsically a non-Abelian scenario for spin
ltering.
B. Tilted ltering axis
For the tilted axis solution (i.e. cos = 0), we have
that
D1
+
= 0 implies
A
+
[cos
2
sin
2
sin2] = i sin
-6 -4 -2 0 2 4 6
-6
-4
-2
0
2
4
6

FIG. 5: Perfect ltering probability for the tilted axis. The

plot shows the relation between , in units of h/(m

L),
and the ltered intensity in a contourplot assuming |

0
| =
1. The lighter regions indicate larger values for the intensity
of ltering for the relation between parameters depicted in
Fig.4. Note that circles and diagonals evident from gure 4
correspond to zero output amplitude.
_
A
2

sin
2
cos
2
2 + 2A
2
+
cos
2
(1 + sin 2).
Squaring both sides, we get
A
2
+
[cos
2
sin
2
sin2]
2
= sin
2

[A
2

sin
2
cos
2
2 + 2A
2
+
cos
2
(1 + sin 2)].
Collecting to the left (right) hand side the A
2
+
(A
2

), one
gets
A
2
+
[cos
4
+ sin
4
sin
2
2 + 2 sin
2
cos
2
] =
A
2

sin
4
cos
2
2.
Using the identity sin
2
2 = 1 cos
2
2 in the second
term and after some algebra one nds
A
2
+
= sin
4
cos
2
2(A
2
+
A
2

). (11)
Using the denitions for A

, and taking the square root,

we reduce Eq. 11 to
t
2
+r
2
e
2iB
= 2rte
iB
sin
2
cos 2.
Employing the 50 50 mirror condition, we get after
substitution
sin
B
= sin
2
cos 2. (12)
This is the relation between the spin-orbit parameters
6
-6 -4 -2 0 2 4 6
-6
-4
-2
0
2
4
6

- +
+
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
FIG. 6: Detector D2 output while D1 lters out the up spin
component (spin down polarization). The plus (minus) zones
represent the regions where only the up spin (down) survives
at the D2 detector. Note that either one or the other is l-
tered. The white regions represent no output in the detector
and correspond to the localized phase. On can have either up
or down spin ltering in D2 while up spin is ltered out in
D1.
and the magnetic ux that leads to perfect ltering in the
tilted axis. The solution is depicted in a contour in Fig.4
where the value of sin
B
is represented in shades of gray
as a function of and . Each contour corresponds to
a constant magnetic ux value and runs over the perfect
ltering values of and . The circular contour, depicted
in the gure, corresponds to a
B
= 0 solution to Eq.12
that leads to (m

L/h)
_

2
+
2
= p for p integer. The
gure depicts the solution for p = 1, 2, i.e. circles in units
of h/(m

L).
In order to see if the lter is actually working we must
address the ltered amplitudes by looking to the second
eigenvalue at detector D
1
. For the ltering condition

D1

= 2ie
iB
sin
B
_
cos
2
sin
2
sin2

. (13)
Substituting Eq.12 in this expression and computing the
modulus squared of the eigenvalue we determine the
strength of the ltered output, as was done in Eq.9. We
have depicted the analytical solution for a range of values
of , in the contour plot of Fig.5. The darkest shade
corresponds to zero amplitude, and as the shade light-
ens the probability is higher for the ltered output. We
note that the ltering solutions for the circular contours
in Fig.4 and the lines = have zero amplitude. Such
zero amplitude solutions correspond to those of local-
ized solutions of Cheng and Chang[19] where there is
no ltered output. In our setup the whole amplitude is
redirected to the second detector.
Behavior of detector D
2
, while D
1
is ltering out the
spin up component (down spin polarization), is shown in
Fig.6. Regions with plus (minus) signs depict up (down)
spin phases for detector D
2
. Note that the two regions
are mutually exclusive so that while pure spin down is
being detected in D
1
one can have either spin up or spin
down in D
2
depending on the range of , . The white
regions correspond to no output at D
2
. Comparing with
Fig.5 we see that no-output region are not identical for
both detectors, these being larger for D
1
, i.e. one can
have zero output at D
1
while having non-zero output at
D
2
. As discussed before the outputs depicted in Fig.6
are also modulated by the modulus of the corresponding
component at the input, so the probability of the out-
put exhibits noise coming from the random input spin
orientation.
IV. SUMMARY
We have solved exactly, for the spin ltering condi-
tions for a Mach-Zehnder type spin interferometer. The
solutions are subject to no limitations on the spin-orbit
strengths and interferometer dimensions as in previous
work. The treatment can be easily extended to unequal
arm lengths that are likely to occur in the actual imple-
mentation of the interferometer. Such a generalization
would provide an additional parameter to manipulate l-
tering conditions. We nd both a non-tilted axis and
tilted axis spin ltering solutions referred to the axis of
quantization in which one writes the input states. The
non-tilted case is not found in the approximate scenario
where the SU(2) can be approximated by a U(1), and
is peculiar to the non-Abelian theory. It has the advan-
tage of simplicity and can be found even at zero magnetic
elds. Perfect ltering means that all spins in one of the
detectors are polarized always in the same axis and ori-
entation. This has the drawback that the current is not
steady since the probability of producing a completely
polarized electron varies with the initial projection, of
the input spinor, onto the chosen quantization axis. This
projection is random as electrons are injected from the
Fermi sea. It should be said that the intereference setup
does not produce a pure spin current, since polarization
is accompanied by a charge current.
An interesting insight, exploiting the analogy with the
Aharonov-Bohm eect in the Abelian case, comes from
observing the role of in the non-Abelian case. and the
voltage V = EL essentially play the same role as the pair
2
B
and magnetic ux. Indeed, for a purely Pauli type
SO interaction, as = (mL/h) and = heE/(m
2
c
2
),
then can be rewritten as 2EL/(2mc
2
/e) = 2V
E
/V
0
,
where V
E
= EL, the voltage along the arm of length L
in an electric eld of strength E. V
0
is a quantum of
voltage[11]. Although V
0
is very large for this calcula-
tion, the material Rashba coecient would lower it to
the order of 1 eV/e.
7
The interference setup dispels a very important prob-
lem associated with the denition of spin currents. A
gauge symmetric theory has gauge dependent spin cur-
rents. The gauge symmetry breaking term in Eq.3
chooses a preferred non-Abelian gauge where currents are
no longer ambiguous. In the spin ltering mechanism de-
scribed here the gauge dependence is no longer an issue,
since by construction it is gauge independent[25].
Acknowledgments
We acknowledge fruitful discussions with C. Chatelain,
J. C. Egues and R. Winkler. This work was supported
by CNRS-Fonacit grant PI-2008000272.
[1] R. E. Rashba, Sov. Phys. Solid State 2, 1109 (1960).
[2] G. Dresselhaus, Phys. Rev. 100, 580 (1955).
[3] R. Winkler, Spin-Orbit Coupling Eects in Two Dimen-
sional Electron and Hole Systems, (Springer) 2003.
[4] J. Nitta, and T. Koga, J. Supercond. 16, 689 (2003).
[5] I. Zutic, J. Fabian, and S. Das Sarma, Rev. Mod. Phys.
76, 323 (2004).
[6] G. Usaj and C. A. Balseiro, Europhys. Lett. 72, 631
(2005).
[7] L. H. Ryder, Quantum Field Theory, (Cambridge Uni-
versity Press) 1985.
[8] A. Rebei and O. Heinonen, Phys. Rev. B 73, 153306
(2006).
[9] P. Q. Jin, Y. Q. Li,and F. C. Zhang, J. Phys. A: Math.
Gen 39, 7115 (2006).
[10] B. W. A Leurs et al, Ann. Phys. 323, 907 (2008).
[11] E. Medina, A. Lopez, and B. Berche, Europhys. Lett. 83,
47005 (2008).
[12] A. S. Goldhaber, Phys. Rev. Lett. 62, 482 (1989).
[13] V.P. Mineev, G.E. Volovik, J. Low Temp. Phys. 89, 823
(1992).
[14] J. Frohlich and U. M. Studer, Rev. Mod. Phys. 65, 733
(1993).
[15] I. V. Tokatly, Phys. Rev. Lett. 101, 106601 (2008).
[16] N. Hatano, R. Shirasaki, and H. Nakamura, Phys. Rev.
A, 75, 032107 (2007).
[17] T. Koga, J. Nitta, and M. van Veenhuizen, Phys. Rev.
B, 70, 161302(R) (2004).
[18] B. A. Bernevig, J. Orenstein, and S. C. Zhang, Phys.
Rev. Lett. 97, 236601 (2006).
[19] S. -H. Chen and C. -R. Chang, Phys. Rev. B. 77, 045324
(2008).
[20] Y. Ji, Y. Chung, D. Sprinzak, M. Heiblum, D. Mahalu,
and H. Shtrikman, Nature 422, 415 (2003).
[21] David Z.-Y. Ting, and X. Cartoixa, Phys. Rev. B. 68,
235320 (2003).
[22] H. A. Engel, E. I Rashba, and B. I. Halperin,
arXiv:cond-mat/0603306v3.
[23] A. Lopez, E. Medina, N. Bolvar, and B. Berche,
arXiv:cond-mat/0902.4635.
[24] S. Datta and B. Das, App. Phys. Lett. 56, 665 (1990).
[25] M. E. Peskin, and D. V. Schroeder, An Introduction to
Quantum Field Theory, (ABP) 1995.