See discussions, stats, and author profiles for this publication at: [Link]

net/publication/45914149

Triplet proximity effect and odd-frequency pairing in graphene

Article  in  Physical review. B, Condensed matter · April 2010

DOI: 10.1103/PhysRevB.82.041409 · Source: arXiv

CITATIONS READS
13 16

3 authors, including:

Asle Sudbø
Norwegian University of Science and Technology
234 PUBLICATIONS   5,574 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Interfacial Effects in Spin Transport in Magnetic Tunnel Junctions View project

All content following this page was uploaded by Asle Sudbø on 19 May 2014.

The user has requested enhancement of the downloaded file.

 Triplet proximity effect and odd-frequency pairing in graphene

Jacob Linder,1 Annica M. Black-Schaffer,2 and Asle Sudbø1

1
Department of Physics, Norwegian University of Science and Technology, N-7491 Trondheim, Norway
2
NORDITA, Roslagstullsbacken 23, SE-106 91 Stockholm, Sweden
(Dated: Received April 28, 2010)
We study the interplay between proximity-induced superconductivity and ferromagnetism in graphene by
self-consistently solving the Bogoliubov-de Gennes equations on the honeycomb lattice. We find that a strong
triplet proximity effect is generated in graphene, leading to odd-frequency pairing correlations. These odd-
frequency correlations are clearly manifested in the local density of states of the graphene sheet, which can be
probed via STM-measurements. Motivated by recent experiments on S|N|S graphene Josephson junctions, we
arXiv:1004.4629v1 [[Link]-con] 26 Apr 2010

also study the spectrum of Andreev-bound states formed in the normal region due to the proximity effect. Our
results may be useful for interpreting spectroscopic data and can also serve as a guideline for future experiments.

PACS numbers: 74.45.+c, [Link]

Introduction. Graphene1 constitutes a new exciting setting should provide a guideline for spectroscopic measurements in
for studying the interplay between different types of long- graphene when superconductivity and/or ferromagnetism are
range orders, such as ferromagnetism and superconductivity. induced by means of the proximity effect.
Although the intrinsic appearance of both of these phenom- Theory. We start out with the following tight-binding lattice
ena only occurs under special circumstances in graphene (see Hamiltonian for graphene
e.g. Refs. 2–4), they can always be induced via proximity X † † † †
X
to host materials with the desired properties. The study of H = −t (fiσ gjσ + giσ fjσ ) − µiσ (fiσ fiσ + giσ giσ )
how the peculiar electronic properties5 of graphene interact hi,ji,σ iσ
with superconducting correlations has recently attracted much X † † † †
attention both theoretically6–8 and experimentally.9–11 Such − Ui (fi↑ fi↑ fi↓ fi↓ + gi↑ gi↑ gi↓ gi↓ ), (1)
i
studies are done by depositing two superconducting leads on
graphene in order to create a graphene S|N|S Josephson junc- and largely employ the notation and methods of Ref. 17. Here
tion. By also exposing the N region to a ferromagnetic host, †
fiσ †
(giσ ) is the creation operator on the A (B) site of the hon-
a hybrid S|F|S junction is constructed, which then will of- eycomb lattice, t ∼ 2.5 eV is the nearest neighbor hopping
fer an excellent platform in which to study the interplay be- parameter, hi, ji denotes summation over nearest neighbors,
tween ferromagnetism and superconductivity. Previous work and σ is the spin index. Moreover, µiσ = µi + σhi is the spin-
on such hybrid structures have reported on interesting effects dependent chemical potential. We will assume that the native
by studying its transport properties via a scattering matrix chemical potential µi is a constant within each region of the
approach.12,13 However, this formalism does not include the junction (S or N/F). Experimentally, an overall chemical po-
full extent of the superconducting proximity effect, as it does tential can be set in the whole sample by applying a back gate
not self-consistently solve for the superconducting order pa- voltage. In addition, it is expected that some charge transfer
rameter inside the junction. A self-consistent solution, on takes place between graphene and the superconducting leads
the other hand, will explicitly include the Cooper pair deple- and, therefore, µS can sometimes be higher than µN/F . hi
tion in S, and the corresponding leakage into N, near the in- is the site-dependent exchange field, which regulates the fer-
terfaces. In addition, the scattering matrix approach cannot romagnetic order induced by proximity to the ferromagnetic
deduce the symmetry of the induced correlations in the non- host material. Thus, hi is only non-zero in the region between
superconducting region, nor calculate their manifestation in the two superconducting leads. The last term in Eq. (1) mod-
the local density of states (DOS). In particular, it remains be els the influence of the superconducting leads on graphene.
clarified how odd-frequency correlations14,15 adapt to the un- The attractive on-site interaction Ui gives rise to s-wave su-
usual electronic environment of graphene and what their sig- perconductivity and this parameter is only non-zero in the S
nature is in experimentally accessible quantities. The issue of regions of the junctions.
odd-frequency pairing has recently generated much activity in We perform a mean-field approximation on Eq. (1), with
the field of conventional F|S junctions,16 but has not yet been the superconducting order parameter defined by
addressed in the context of graphene.
In this paper, we present a self-consistent lattice-study of ∆i = −Ui [hfi↓ fi↑ i + hgi↓ gi↑ i]/2. (2)
the interplay between ferromagnetism and superconductivity
in graphene, with focus on the behavior of the local DOS to We have ignored spatial variations within one unit cell and
probe this interaction. Specifically, we investigate an S|F|S used ∆f = ∆g . We further consider a geometry with trans-
junction and the concomitant manifestation of odd-frequency lational invariance along the interfaces, i.e. orthogonal to the
pairing. To make contact with the current experimental status, direction of the junction. For concreteness, we focus on a zig-
we also consider an S|N|S junction and identify the appear- zag interface and Fourier-transform the eigenvectors in the y-
ance of subgap Andreev-bound states in the DOS. Our results direction. Please note that for an s-wave symmetry, the spe-
 2

FIG. 1: (Color online). Local DOS on a large energy scale (left panel) and near the gapped region (right panels) in an S|N|S graphene junction
for a variety of lengths and FVMs. Large values of the DOS are indicated by a bright color, while small values are given by a dark color. We
have fixed LS = 45 sites and µS /t = 0.6, as well as set the superconducting phase difference to zero, φ = 0. (a): µN = µS (no FVM), and
LN = 8 sites (ξ ∼ 25 sites). (b): µN = µS and LN = 24 sites. (c): µN /t = 0.2 and LN = 8 sites. (d): µN /t = 0.2 and LN = 24 sites. As
seen, in-gap bound states, below the gap edge, are formed and persist even in the long-junction regime.

cific direction of the interface will not matter. After diagonal- ky = 2πl/(Ny a), where l is an integer such that ky ∈
izing the problem, we arrive at the tight-binding Bogoliubov- ] − π/a, π/a], and a is the lattice constant. Moreover, ψnν =
de Gennes (BdG) equations for graphene [uνn (ky ), ynν (ky ), vnν (ky ), znν (ky )]T is generated by two copies
(one for each sublattice) of the standard canonical Bogoliubov
X Ĥ (n, m) ∆(n, ˆ

↑ m) ν ν ν transformation. We have also introduced the following matri-
ˆ † (n, m) −Ĥ↓ (n, m) ψn = E (ky )ψn . (3)
∆ ces
m

Here n is the lattice site index along the junction and

   
−(µn + σhn )δnm −t[δnm + 2δn+1,m cos(ky a/2)] ˆ ∆n δnm 0
Ĥσ (n, m) = , ∆(n, p) = . (4)
−t[δnm + 2δn−1,m cos(ky a/2)] −(µn + σhn )δnm 0 ∆n δnm

For a self-consistent solution of the above equations, we first by Fourier-transforming the relative time-coordinate (τ − τ 0 ),
guess an initial ∆n , then find the corresponding eigenvalues and it thus follows that such an odd triplet amplitude must
and eigenvectors to Eq. (3), followed by a recalculation of ∆n be antisymmetric with respect to (τ − τ 0 ). In effect, this
using the self-consistency criteria in Eq. (2). This process is amounts to a strong retardation effect since the correlator van-
iterated until ∆n no longer changes between subsequent iter- ishes at equal times. The study of proximity-induced odd-
ations. We note in passing that for the case of a zero exchange frequency pairing has recently generated much interest in con-
field, Eq. (3) is particle-hole symmetric and it is then enough ventional metallic S|F systems, but has not yet been explored
to solve for only the negative eigenvalues. However, this is in graphene. To investigate the presence of odd-frequency
not the case for a non-zero h, and we are therefore forced to pairing, we introduce18
calculate all the eigenvalues.
We are particularly interested in investigating the appear-
t
ance of spin-triplet correlations in the system when the ex- Ff,i = hfi↑ (τ )fi↓ (0) + fi↓ (τ )fi↑ (0)i. (5)
change field hi is non-zero. These triplet correlations are
necessarily of an unusual nature, since the spatial symme-
try of the superconducting order parameter is isotropic (s- As with the order parameter in Eq. (2), we define the effec-
wave). According to the Pauli-principle, it is possible to tive odd-frequency correlator Fit as the average between the
have superconducting correlations which are both isotropic expectation values on the two sublattices described by the f -
and spin-triplet simultaneously as long as these have an odd- and g-fermion operators. We will also be concerned with the
frequency symmetry. The frequency dependence is obtained local DOS Ni (E) which in the low-temperature limit is ob-
 3

tained via the charge density ρ as follows: Spatial profile of the order parameter
0.05
Z 0 (a)
† †
X
ρi = Ni (E)dE = hfiσ fiσ + giσ giσ i. (6) 0.04
−∞ σ −3
x 10
0.03 3

∆/t
Results and Discussion. In what follows, we present a nu- 2

Im{F t }
merical and self-consistent solution of the above equations. 0.02
1
Let us first consider the S|N|S case, shown in Fig. 1. To model
experimentally relevant scenarios, we consider both short and 0.01 0
40 60 80
long junctions in addition to the presence or absence of a Lattice site
Fermi-vector mismatch (FVM) at the interface. It may be in- 0
20 30 40 50 60 70 80 90
structive to start with a reminder of the analytical result for Lattice site
the Andreev bound-state energy inside a junction, obtainable
0.1
in the ultrashort-junction regime where L  ξ, with ξ be-
3
ing the superconducting coherence length. This relation reads
2 (b) 0
E = ∆0 [1 − D sin2 (φ/2)]1/2 , where D denotes the inter-
1 2
face transparency and φ the superconducting phase difference −0.1
0

E/t
between the two leads. For φ = 0, it is seen that the bound- 1 20 40 60 80

state lies right at the gap edge, independent of the the inter- −1
0
face transparency. As a consistency-check, we have verified −2 0
that we also obtain this result numerically when using a non- −3
50
selfconsistent step-function profile of the superconducting or- 20 40 60 80 100
−0.1 −0.05 0 0.05 0.1
der parameter. Let us now turn to the self-consistent treatment Lattice site
in Fig. 1, where we consider four scenarios for a junction. In
(a), we have no FVM (i.e. µS = µN ) and a short junction,
LN = 8 sites to be compared with ξ ∼ 25 sites. The su- FIG. 2: (Color online). Local DOS in an S|F|S graphene junction
with LS = 50 sites, LF = 8 sites, h/t = 0.05, and µS /t =
perconducting regions are chosen to be large, LS = 45 sites,
µF /t = 0.6 (no FVM). (a): Spatial profile for the superconduct-
so that they act as superconducting reservoirs. As seen, the ing order parameter. The position of the interfaces are marked with
self-consistent solution for the order parameter only slightly black vertical lines. The inset shows the imaginary part of the in-
shifts the bound-states inside the gap. This can be understood duced odd-frequency correlations, which from bottom to top corre-
as a consequence of the proximity effect suppression of the spond to times tτ = 1, 2, 3, 4, 5. (b): Local DOS on a large energy
order parameter near the interface. In (b), we increase the scale (left panel) and near the gapped region (right panels). Large
length of the normal region to LN = 24 sites. In this case the values of the DOS are indicated by a bright color, while small values
bound-states now reside well within the superconducting gap, are given by a dark color. As seen, odd-frequency correlations give
in contrast to the analytical prediction for ultrashort junctions. rise to a zero-energy peak in the DOS.
Note that the large value of LN ensures that these states do
not pertain to some surface-effect, but that they penetrate into
the entire N region. When turning on a FVM (µS > µN ) in superconducting order parameter and the generation of odd-
(c) and (d), it is seen that the magnitude of the in-gap bound- frequency correlations in the F region (inset). The magnitude
states in the N region is strongly reduced, due to the reduced of the corresponding anomalous Green’s function peaks in the
normal-state DOS in the N region, although the general en- middle of the ferromagnetic region, and penetrates a short dis-
ergy dependence is seen to be similar as in (a) and (b). tance into the S regions. We have here plotted the imaginary
We now turn to the case where the region between the su- part of F t since it couples directly to the DOS.19 The experi-
perconducting leads is ferromagnetic via a proximity to a fer- mental manifestation of such odd-frequency correlations have
romagnetic host material. The ground-state of an S|F|S junc- previously been discussed in conventional metals, in which
tion can occur for a superconducting phase difference φ of case one expects an enhancement of the DOS at the Fermi
either 0 or π, so it becomes necessary to consider the free level.19 In Fig. 2(b) we demonstrate that the same signature
energy of the junction to correctly identify the ground-state. applies for graphene: a zero-energy peak emerges and it is
In Fig. 2 we consider a short junction with LF = 8 sites also flanked by additional in-gap bound-states. The ground-
and no FVM. Our results remain qualitatively the same also state for the parameters used in Fig. 2 was numerically found
when including a moderate FVM (e.g. µF /µS = 0.8). How- to be the 0-phase.
ever, for a sufficiently low chemical potential in the F region, It is well-known that the superconducting phase-difference
the DOS becomes too small to sustain any appreciable in-gap φ can be tuned actively via an external flux or current flowing
electron density, as was also seen in Fig. 1. In the experi- through the system. We investigate in Fig. 3 how the local
mentally relevant scenario, it is reasonable to expect that µF DOS in a S|F|S junction changes when the length of the F re-
is doped away from the Dirac point to support the presence gion is increased (LF = 24 sites) and compare specifically
of ferromagnetism, which is exactly the case considered here. the 0- and π-phases. As seen in Fig. 3(a), the odd-frequency
In Fig. 2(a), we show the spatial self-consistent profile of the pairing correlations are clearly manifested for the 0-phase as a
 4

(a) (b) odd-frequency correlations are still visible in (c) and (d).
0.1 0.1
The presence of significant in-gap DOS in the whole F re-
0.05 0.05
gion, even for long junctions (LF = 24 sites), rules out the
0 0 possibility of these being caused by surface-states at the S|F
−0.05 −0.05 interfaces. We have performed numerical calculations for sev-
−0.1 −0.1
eral sets of parameters to investigate the robustness of the
20 40 60 80 100 20 40 60 80 100 odd-frequency peak, and find that it in general competes with
E/t

(c) (d) the singlet correlations which instead induce a standard mini-
0.1 0.1
gap in the electronic spectrum inside the junction. In spite of
0.05 0.05 this coexistence, our results above demonstrate that the odd-
0 0 frequency amplitude can be read out from spectroscopic in-
formation in a feasible parameter regime.
−0.05 −0.05
Summary. In summary, we have investigated in a self-
−0.1 −0.1
20 40 60 80 100 20 40 60 80 100 consistent manner the proximity effect and its implications for
Lattice site the local DOS in both magnetic and non-magnetic graphene
Josephson junctions. We have considered several experimen-
FIG. 3: (Color online). Local DOS in an S|F|S graphene junction tally relevant ranges of doping levels and junction lengths. It
with LS = 50 sites, LF = 24 sites, µS /t = 0.6, and h/t = 0.06. is found that a considerable triplet proximity effect can be in-
(a): 0-phase, µF /t = 0.6 (no FVM). (b): π-phase, µF /t = 0.6 duced in an S|F|S graphene junction, giving rise to so-called
(no FVM). (c): 0-phase, µF /t = 0.2. (d): π-phase, µF /t = 0.2. odd-frequency correlations. These are manifested clearly as
Large values of the DOS are indicated by a bright color, while small zero-energy peaks in the DOS, which may be probed by STM-
values are given by a dark color. As seen, the odd-frequency pairing measurements. We have also identified the appearance of
amplitude is most pronounced in the 0-phase, both with and without
Andreev-bound states in a S|N|S graphene Josephson junc-
a FVM.
tion. They appear as in-gap resonances in the DOS, in con-
trast to non-self-consistent results positioning them at the gap
strong zero-energy peak when there is no FVM. Going to the edge. Our results should be helpful for the interpretation of
π-phase in (b), it is seen that this peak is suppressed whereas spectroscopic data and will hopefully serve as a guideline for
the non-zero bound-states inside the gap are instead more pro- future experimental activity.
nounced. Upon introducing a strong FVM in (c) and (d), we Acknowledgments. M. Cuoco is thanked for helpful discus-
see again how the magnitude of the proximity effect in the sions. J.L. and A.S. were supported by the Norwegian Re-
F region is severely suppressed, although some signs of the search Council Grant No. 167498/V30 (STORFORSK).

1
K. S. Novoselov et al. , Science 304, 666 (2004). (2008); Y. Asano et al. , Phys. Rev. B 78, 014514 (2008).
2 14
H. Ohldag et al., Phys. Rev. Lett. 98, 187204 (2007). V. L. Berezinskii, JETP Lett. 20, 287 (1974)
3 15
J. Červenka et al., Nature Phys. 5, 840 (2009). A. Balatsky and E. Abrahams, Phys. Rev B 45 13125 (1992); A.
4
P. Esquinazi et al., Phys. Rev. B 78, 134516 (2008). V. Balatsky and J. Bonca, Phys. Rev. B 48, 7445, (1993).
5 16
A. H. Castro Neto et al. , Rev. Mod. Phys. 81, 109 (2009). F. S. Bergeret et al. , Rev. Mod. Phys. 77, 1321 (2005).
6 17
C. W. J. Beenakker, Phys. Rev. Lett. 97, 067007 (2006). A. M. Black-Schaffer and S. Doniach, Phys. Rev. B 78, 024504
7
M. Titov and C. W. J. Beenakker, Phys. Rev. B 74, 041401 (2006). (2008); ibid Phys. Rev. B 79, 064502 (2009).
8 18
C. W. Beenakker, Rev. Mod. Phys. 80, 1337 (2008). K. Halterman, P. H. Barsic, and O. T. Valls, Phys. Rev. Lett. 99,
9
H. B. Heersche et al. , Nature 446, 56 (2007). 127002 (2007); K. Halterman et al. , Phys. Rev. B 77, 174511
10
X. Du et al. , Phys. Rev. B 77, 184507 (2008). (2008).
11 19
C. Ojeda-Aristizabal et al. , Phys. Rev. B 79, 165436 (2009). T. Yokoyama et al. , Phys. Rev. B 75, 134510 (2007).
12
J. Linder et al. , Phys. Rev. Lett. 100, 187004 (2008).
13
A. G. Moghaddam and M. Zareyan, Phys. Rev. B 78, 115413

View publication stats