Crossed Andreev reflection in superconducting graphene spin-valves: Spin-switch effect

Jacob Linder,1 Malek Zareyan,2 and Asle Sudbø1

1
Department of Physics, Norwegian University of Science and Technology, N-7491 Trondheim, Norway
2
Institute for Advanced Studies in Basic Sciences, 45195-1159, Zanjan, Iran
(Dated: Received June 23, 2009)
We consider the non-local quantum transport properties of a graphene superconducting spin-valve. It is shown
that one may create a spin-switch effect between perfect elastic co-tunneling (CT) and perfect crossed Andreev-
reflection (CAR) for all bias voltages in the low-energy regime by reversing the magnetization direction in
one of the ferromagnetic layers. This opportunity arises due the possibility of tuning the local Fermi-level
in graphene to values equivalent to a weak, magnetic exchange splitting, thus reducing the Fermi surface for
minority spins to a single point and rendering graphene to be half-metallic. Such an effect is not attainable in
a conventional metallic spin-valve setup, where the contributions from CT and CAR tend to cancel each other
arXiv:0906.3738v1 [cond-mat.supr-con] 19 Jun 2009

and noise-measurements are necessary to distinguish these processes.

PACS numbers: 74.25.Fy,74.45.+c,74.50.+r,74.62.-c

I. INTRODUCTION Ferromagnet F1 Superconductor S Ferromagnet F2

Quantum entanglement1 describes a scenario where the

quantum states of two objects separated in space are strongly
correlated. These correlations can be exploited in emerging
technologies such as quantum computing, should one be able
to spatially separate the entangled objects without destroy-
Fermi level CT
ing the correlations. In a broader context, quantum entangle- R
AR (P configuration) CAR
ment could prove to be of practical importance in the fields
of spintronics2 and information cryptography3. It also holds a
considerable interest from a purely fundamental physics point
of view, prompting some of the more philosophically inclined
discussions related to quantum theory and causality.
Superconductors have been proposed as natural sources
Fermi level CT
for entangled electrons4,5 , as Cooper pairs consist of two R
AR (AP configuration)
electrons that are both spin and momentum-entangled. The CAR
Cooper pair can be spatially deformed by means of the crossed
Andreev reflection (CAR) process in superconducting het-
erostructures. In this scenario, an electron and hole excitation
are two separate metallic leads are coupled by means of An-
dreev scattering processes at two spatially distinct interfaces. FIG. 1: (color online) Proposed experimental setup for the spin
Unfortunately, the signatures of CAR are often completely switch effect between crossed Andreev reflection and elastic co-
masked by a competing process known as elastic co-tunneling tunneling. Ferromagnetism and superconductivity are induced by the
(CT) which occur in the same type of heterostructures. In fact, proximity effect to a host material. The induced exchange fields in
the conductances stemming from CT and CAR may cancel the non-superconducting graphene regions are oriented either paral-
each other completely6 , thus necessitating the usage of noise- lel or antiparallel with respect to each other. In the parallell align-
ment, the density of states vanishes for both normal Andreev reflec-
measurements to find fingerprints of the CAR process in such
tion and crossed Andreev reflection processes, such that only elastic-
superconducting heterostructures. cotunneling contributes to non-local transport. In the anti-parallel
Recently, graphene7 has been studied as a possible arena for alignment, the density of states vanishes for both normal Andreev
CAR-processes. In Ref.8 , it was shown how a three-terminal reflection and elastic co-tunneling, leaving only crossed Andreev re-
graphene sheet containing n-doped, p-doped, and supercon- flection as the non-local transport channel.
ducting regions could be constructed to produce perfect CAR
for one particular resonant bias voltage. Also, the signatures
of the CAR process in the noise-correlations of a similar de- In this paper, we show that precisely such an opportunity
vice were studied in Ref.9 . However, the role played by the exists – it is possible to obtain a spin-switch effect between
spin degree of freedom in graphene devices probing non-local virtually perfect CAR and perfect CT in a superconducting
transport has not been addressed so far. This is a crucial point graphene spin valve. In contrast to Ref.8 , this effect is seen
since it might be possible to manipulate the spin-properties of for all bias voltages in the low-energy regime rather than just
the system to interact with the spin-singlet symmetry of the at one particular applied voltage difference. The key obser-
Cooper pair in a fashion favoring CAR. vation is that the possibility of tuning the local Fermi-level
 2

to values equivalent to a weak, magnetic exchange splitting and in this case reads
in graphene renders both the usual Andreev reflection process
and CT impossible. In contrast, this opportunity does not ex- uσ = (ψA,+
σ σ
, ψB,+ )T (4)
ist in conventional conductors where the Fermi energy is large
while v −σ = T uσ . Here, T denotes the transpose while T is
and of order O(eV). We show that graphene spin valves pro-
the time-reversal operator.
vide a possibility for a unique combination of non-local An-
From Eq. (1), one may now construct the quasiparticle
dreev reflection and spin-dependent Klein tunneling10. Our
wavefunctions that participate in the scattering processes20 .
model is shown in Fig. 1, where ferromagnetism and super-
We consider positive excitation energies ε ≥ 0 with in-
conductivity are assumed to be induced by means of the prox-
coming electrons of n-type, i.e. from the conduction band
imity effect11,12 to leads with the desired properties. A similar
ε = vF |p| − µF (we set vF = 1 from now on). The incom-
setup was considered in Ref.13 , where the magnetoresistance
ing electron from the left ferromagnet may either be reflected
of the system was studied.
normally or Andreev-reflection (AR). In the latter process, it
We organize this work as follows. In Sec. II, we establish
tunnels into the superconductor with another electron situated
the theoretical framework which will be used to obtain the
at (−ε), leaving behind a hole excitation with energy ε. The
results. In Sec. III, we present our main findings for the non-
scattering coefficients for these two processes are re and rh ,
local conductance in the graphene superconducting spin-valve
respectively, and the total wavefunction may thus be written
with a belonging discussion of them. Finally, we summarize
as:
in Sec. IV.
1 1
   
eıθ  σ −e−ıθ  −ıpσe cos θx
ψL =  eıpe cos θx + re  e
II. THEORY 0 0 
0 0
 
We consider a ballistic, two-dimensional graphene structure 0
as shown in Fig. 1. In the left ferromagnetic region x < 0,  0  −ıpσ cos θσ x
the exchange field is h = h0 z, while it is h = ±h0 z in the + rh  1 e
 h A , (5)
σ
right ferromagnetic region x > L. In the superconducting e −ıθA
region 0 < x < L, the order parameter is taken to be con-
stant with a real gauge ∆ = ∆0 . To proceed analytically, we where we have defined the wavevectors
make the usual approximation of a step-function behavior at
pσe = ε + µF + σh0 , pσh = ε − µF + σh0 . (6)
the interfaces for all energy scales, i.e. the chemical poten-
tials {µF , µS }, the exchange field h0 , and superconducting We have omitted a common factor eıpy y for all wavefunctions.
gap ∆0 . This assumption is expected to be good when there Similarly, assuming that the charge carriers in the right ferro-
is a substantial Fermi-vector mismatch between the F and S magnetic region are also of the n-type, we obtain:
regions, as in the present case. To make contact with the ex-
1
 
perimentally relevant situation, we assume a heavily doped S
region satisfying µS ≫ µF . eıθ  ıp±σ ±σ
ψR = te   e e cos θN x
We use the Dirac-Bogoliubov de Gennes equations first em- 0
ployed in Ref.14 . For quasiparticles with spin σ, one obtains 0
in an F|S graphene junction:15,16,17,18,19 
0

 0  −ıp±σ cos θ±σ x
uσ uσ + th 
 1 e . (7)
     h A
Ĥσ (x) σ∆(x)1̂ 
=ε , (1)
σ∆∗ (x)1̂ −Ĥ−σ (x) v −σ v −σ e−ıθA
±σ

where It should be noted that the AR hole is generated in the con-

duction band if ε − µF − σh0 > 0 (retro-AR), whereas it is
Ĥσ (x) = vF p · σ̂ − [µ(x) + σh(x)]1̂ (2) generated in the valence band otherwise (specular-AR). The
± sign above refers to parallell/antiparallell (P/AP) magneti-
and . ˆ. . denotes a 2 × 2 matrix. Here, we have made use zation configuration.
of the valley degeneracy and p is the momentum vector in We assume that the superconducting region is heavily
the graphene plane while σ is the vector of Pauli matrices doped, µS ≫ µF + h0 , which causes the propagating quasi-
in the pseudospin space representing the two A, B sublat- particles to travel along the x-axis since the scattering angle in
tices of graphene hexagonal structure. The superconducting the superconductor satisfies θS → 0. We obtain the following
order parameter ∆(x) couples electron- and hole-excitations wavefunction (λ = ±1):
in the two valleys (∓) located at the two inequivalent corners  
of the hexagonal Brillouin zone. The uσ spinor describes the eıλβ
X ±eıλβ 
electron-like part of the total wavefunction ΨS = lλ±   ±(ıµS −λκ)x ,
 1 e (8)
λ,±
ψ σ = (uσ , v −σ )T , (3) ±1
 3
p
where κ = ∆20 − ε2 while 1.5

0.8
β = acos(ε/∆0 ) (9)
1.2
1

GCT /GF
for subgap energies |ε| < ∆0 and

β = −ıacosh(ε/∆0 ) (10) L/ξ = 0.4

0.5

for supergap energies |ε| > ∆0 . 2.0

1.6
It is important to consider carefully the scattering angles in 0
the problem. Since we assume translational invariance in the 0 0.5 1 1.5 2
y-direction, the y-component of the momentum is conserved. eV /∆
1.5
This gives us
eV/∆ = 1
pσe sin θ = pσh σ
sin θA = p±σ
e
±σ
sin θN . (11)
1

GCT /GF
It is clear that the angle of transmission for the electrons in
the right ferromagnet is equal to the angle of incidence when
σ
the magnetizations are P, i.e. θN = θ. Also, one infers that 0.5
there exists a critical angle above which the scattered waves
become evanescent, i.e. decaying exponentially. This may eV/∆ = 0
be seen by observing that the scattering angles exceed π/2 0
0 0.5 1 1.5 2 2.5
(thus becoming imaginary) above a certain angle of incidence L/ξ
θ. For instance, the AR wave in the left ferromagnetic region
σ
becomes evanescent for angles of incidence θ > θAR , where
σ σ FIG. 2: (color online) Plot of the conductance for CT processes
the critical angle θ = θAR is obtained by setting θA = π/2 in GCT /GF versus bias voltage in the upper panel and versus length
the equation of the S region in the lower panel. Here, we consider the P alignment
and µF = h0 such that GCAR → 0.
pσe sin θ = pσh sin θA
σ
, (12)

expressing conservation of momentum perpendicular to the

interface. One finds that:
specular. In contrast, the situation becomes quite fascinating
σ
θAR ≡ |asin[(ε − µF + σh0 )/(ε + µF + σh0 )]|. (13) when we consider the case µF = h0 under the assumption
of a doped situation µF ≫ (ε, ∆0 ). First of all, the incom-
Thus, AR waves in the regime θ > |θcσ | do not contribute ing quasiparticles from the left ferromagnetic region are com-
to any transport of charge. A similar argument can be made pletely dominated by the majority spin carriers σ =↑, since
for the transmitted electron wave-function in the right ferro- the density of states (DOS) for σ =↓ electrons vanishes at the
magnetic region, corresponding to the CT process, where the Fermi level. Since µF = h0 , the AR process is suppressed for
critical angle for this process becomes ↑
all incoming waves as θAR → 0. We now show how the fate
σ
θCT ≡ |asin[(ε + µF ± σh0 )/(ε + µF + σh0 )]|. (14) of the cross-conductance in the right ferromagnetic region de-
pends crucially on whether the magnetization configuration is
↑
In the P configuration, the CT process thus always contributes P or AP. In the P configuration, we see that θCAR → 0, which
to the transport of charge. Finally, the contribution to transport means that the transport is purely governed by the CT process.
↑
of charge from CAR comes from the hole-wave function in In the AP configuration, we see that θCT → 0, which means
the right ferromagnetic region, which becomes evanescent for that the transport is mediated purely by the CAR process. This
angles of incidence above the critical angle suggests a remarkable spin-switch effect – by reversing the
σ
direction of the field in the right ferromagnet, one obtains an
θCAR > |asin[(ε − µF ± σh0 )/(ε + µF + σh0 )]|. (15) abrupt change from pure CT to pure CAR processes mediat-
ing the transport of charge. In each case, there is no local AR
In the P configuration, this criteria is the same as the vanishing in the left ferromagnetic region. In the standard metallic case,
of local AR expressed by Eq. (13). the distinct signatures for the CT and CAR contributions are
masked by each other, and it becomes necessary to resort to
noise-measurements in order to say something about the con-
III. RESULTS AND DISCUSSION
tribution from each process. In the present scenario, we have
showed how it is possible to separate the two contributions
Intuitively, one might expect that the most interesting phe- directly by a simple spin-switch effect which is commonly
nomena occur when the exchange field h0 is comparable in employed in experimental work on F|S heterostructures.
magnitude to the chemical potential µF . If µF ≫ h0 , the ef-
fect of the exchange field should be minor and the AR is never Let us now evaluate the conductance in the P and AP con-
 4

figuration quantitatively by using 1.5

Z π/2
X 0.8
GCAR /GF = (Gσ /GF ) dθ cos θ|th |2 , (16) 1

GCAR /GF
σ −π/2

where we have introduced

0.5 1.6 2.0
Gσ = e2 N σ (eV )/π (17) L/ξ = 0.4

as the spin-σ normal-state conductance that takes into account 1.2

0
the valley degeneracy, in addition to 0 0.5 1 1.5 2
eV /∆
1.5
GF = G+ + G− . (18)

The density of states is determined by

eV/∆ = 0 eV/∆ = 1

GCAR /GF
1
σ
N (ε) = |ε + µF + σh|W/(πvF ), (19)

where W is the width of the junction. The expression for 0.5

GCT is obtained by replacing th with te in Eq. (16). Since
we here consider the case µF = h0 and h0 ≫ (ε, ∆0 ),
the formulas for the GCAR and GCT may be simplified since 0
0 0.5 1 1.5 2 2.5
G− ≪ G+ . Also, since the DOS vanishes for minority spins L/ξ
for the injected electrons, only σ =↑ contributes for incoming
electrons. The crucial point here is that in the P alignment, FIG. 3: (color online) Plot of the conductance for CAR processes
GCAR → 0 and GCT 6= 0 such that GCAR /GF versus bias voltage in the upper panel and versus length of
the S region in the lower panel. Here, we consider the AP alignment
|re |2 + |te |2 = 1, (20) and µF = h0 such that GCT → 0.

while in the AP alignment GCAR 6= 0 and GCT → 0 such that

sheets22 . Moreover, it has been speculated that such charge
|re |2 + |th |2 = 1. (21)
inhomogeneities may play an important role with regard to
In the actual numerical calculations, we use h0 /∆0 = 50 and limiting the transport characteristics of graphene23 near the
µS /∆0 = 500. Assuming a value of ∆0 = 0.1 meV for Dirac points. However, for our purposes this is actually bene-
the proximity-induced gap, this corresponds to an exchange ficial – it is precisely the suppression of charge and spin trans-
splitting of h0 = 5 meV in the F regions and a doping level port at Fermi level for the Andreev reflection and co-tunneling
µS = 50 meV in the S region, which should be experimen- process which renders possible the spin-switch effect. There-
tally feasible21 and well within the range of the validity for fore, we do not expect that the inclusion of charge inhomo-
the linear dispersion relation in graphene. In Fig. 2, we plot geneities should alter our results qualitatively. Finally, we
the cross-conductance GCT /GF in the P alignment both as a note that since the spin of the charge-carriers in each of the
function of bias voltage and width of the S region. The same non-superconducting graphene sheets are practically speaking
thing is done for GCAR /GF in the AP alignment in Fig. 3. In fixed due to the vanishing DOS for minority spins, the spin-
both cases, the magnitude of the conductance varies strongly switch effect for CAR and EC predicted in this paper can not
when considering different widths L due to the fast oscilla- be directly related to entanglement. Nevertheless, it consti-
tions which pertain to the formation of resonant transmission tutes a clear non-local signal for quantum transport which can
levels inside the superconductor. Also, it is seen that while the be probed experimentally, and should be helpful in identifying
CT process is favored for short junctions L/ξ ≪ 1, the CAR clear-signatures of the mesoscopic CAR phenomenon.
process is suppressed in this regime in favor of normal reflec-
tion. Upon increasing the junction width, the CT conductance
drops while the CAR conductance peaks at widths L ∼ ξ. IV. SUMMARY
The remarkable aspect is that it is possible to switch between
these two scenarios of exclusive CT and exclusive CAR sim- To summarize, we have considered non-local quantum
ply by reversing the direction of magnetization in one of the transport in a graphene superconducting spin-valve. We have
ferromagnetic layers. shown how one may create a spin-switch effect between
In order to obtain analytical results, we have assumed that perfect elastic co-tunneling and perfect crossed Andreev-
the Coulomb interaction and charge inhomogeneities may be reflection for all applied bias voltages by reversing the magne-
neglected. It would be challenging to obtain a truly homoge- tization direction in one of the ferromagnetic layers. The basic
neous chemical potential in a graphene sheet, and electron- mechanism behind this effect is that the local Fermi-level in
hole puddles appear to be an intrinsic feature of graphene graphene may be tuned so that the Fermi surface for minority
 5

spins reduces to a single point in the presence of a weak, mag- and 167498/V30 (STORFORSK).
netic exchange splitting. This is very distinct from the equiv-
alent spin valve structures in conventional metallic systems,
where noise-measurements are required to clearly distinguish
between these processes.

Acknowledgments

J.L. and A.S. were supported by the Norwegian Research

Council Grant Nos. 158518/431, 158547/431, (NANOMAT),

1
J. M. Raimond, M. Brune, and S. Haroche, Rev. Mod. Phys. 73, A. Shailos, W. Nativel, A. Kasumov, C. Collet, M. Ferrier, S.
565 (2001); L. Amico, R. Fazio, A. Osterloh, and V. Vedral, Rev. Gueron, R. Deblock, and H. Bouchiat, Europhys. Lett. 79, 57008
Mod. Phys. 80, 517 (2008). (2007).
2 13
I. Zutic, J. Fabian, and S. Das Sarma, Rev. Mod. Phys. 76, 323 C. Bai, Y. Yang, and X. Zhang, Appl. Phys. Lett. 92, 102513
(2004). (2008).
3 14
A. Galindo and M. A. Martin-Delgado, Rev. Mod. Phys. 74, 347 C. W. J. Beenakker, Phys. Rev. Lett. 97, 067007 (2006).
15
(2002). J. Linder, T. Yokoyama, D. Huertas-Hernando, and A. Sudbø,
4
G. Burkard, D. Loss, and E. V. Sukhorukov, Phys. Rev. B 61, R16 Phys. Rev. Lett. 100, 187004 (2008).
16
303 (2000). A. G. Moghaddam and M. Zareyan, Phys. Rev. B 78, 115413
5
P. Recher, E. V. Sukhorukov, and D. Loss, Phys. Rev. B 63, (2008).
17
165314 (2001). M. Zareyan, H. Mohammadpour, and A. G. Moghaddam, Phys.
6
G. Falci, D. Feinberg, F.W.J. Hekking, Europhys. Lett. 54, 255 Rev. B 78, 193406 (2008).
18
(2001). Y. Asano, T. Yoshida, Y. Tanaka, and A. A. Golubov, Phys. Rev.
7
K. S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, Y. Zhang, B 78, 014514 (2008).
19
S.V. Dubonos, I.V. Grigorieva, and A.A. Firsov, Science 306, 666 Q. Zhang, D. Fu, B. Wang, R. Zhang, and D. Y. Xing, Phys. Rev.
(2004). Lett. 101, 047005 (2008).
8 20
J. Cayssol, Phys. Rev. Lett. 1100, 147001 (2008). J. Linder and A. Sudbø, Phys. Rev. B 77, 064507 (2008); J. Linder
9
C. Benjamin and J. K. Pachos, arXiv:0802.3181. and A. Sudbø, Phys. Rev. Lett. 99, 147001 (2007).
10 21
M. I. Katsnelson, K. S. Novoselov, and A. K. Geim, Nature Phys. H. Haugen, D. Huertas-Hernando, and A. Brataas, Phys. Rev. B
2, 620 (2006). 77, 115406 (2008).
11 22
E. W. Hill et al., IEEE Trans. Magn. 42, 2694 (2006); N. Tombros J. Martin, N. Akerman, G. Ulbricht, T. Lohmann, J. H. Smet, K.
et al., Nature 448, 571 (2007); M. Ohishi et al., Jpn. J. Appl. Phys. von Klitzing, and A. Yacoby, Nature Physics 4, 144 (2008).
23
46 L605 (2007). E.-A. Kim and A. H. Castro Neto, arXiv:0702.562.
12
H. B. Heersche, P. Jarillo-Herrero, J. B. Oostinga, L. M. K. Van-
dersypen, and A. F. Morpurgo, Nature (London) 446, 56 (2006);