RAPID COMMUNICATIONS

PHYSICAL REVIEW B 82, 161408共R兲 共2010兲

Transmission coefficient through a saddle-point electrostatic potential for graphene

in the quantum Hall regime
Martina Flöser and Serge Florens
Institut Néel, CNRS–Université Joseph Fourier, BP 166, 25 Avenue des Martyrs, 38042 Grenoble Cedex 9, France

Thierry Champel
Laboratoire de Physique et Modélisation des Milieux Condensés, CNRS–Université Joseph Fourier, BP 166, 25 Avenue des Martyrs,
38042 Grenoble Cedex 9, France
共Received 9 September 2010; published 14 October 2010兲
From the scattering of semicoherent-state wave packets at high magnetic field, we derive analytically the
transmission coefficient of electrons in graphene in the quantum Hall regime through a smooth constriction
described by a quadratic saddle-point electrostatic potential. We find anomalous half-quantized conductance
steps that are rounded by a backscattering amplitude related to the curvature of the potential. Furthermore, the
conductance in graphene breaks particle-hole symmetry in cases where the saddle-point potential is itself
asymmetric in space. These results have implications both for the interpretation of split-gate transport experi-
ments and for the derivation of quantum-percolation models for graphene.

DOI: 10.1103/PhysRevB.82.161408 PACS number共s兲: 73.43.Jn, 73.43.Cd, 71.70.Di, 73.22.Pr

The quantum point-contact geometry formed by metallic
split gates in the quantum Hall regime is a cornerstone of
many experiments in two-dimensional electron gases
共2DEGs兲 based on semiconducting heterostructures. For ex-
ample, in recent interferometry experiments made with elec-
trons, the quantum point contact plays the role of an elec-
tronic beam splitter.1 At the theoretical side, the
consideration of a smooth constriction represents a simple
toy model of elaborated quantum transport theories. For in-
stance, the transmission coefficient through a saddle-point
potential2 is a central piece of the percolation network
models,3,4 which have been introduced to describe the inter-
plateaus dissipative transport in the quantum Hall effect.
Smooth constrictions are usually modeled by the local
potential profile at the bottleneck of the constriction
V共r兲 = by 2 − ax2 ,
where a and b are real positive coefficients characterizing the
potential 共for convenience, we chose the electrostatic poten-
tial value at the saddle as the origin for the energies兲. Impor-
tant insights on the tunneling processes determining the
transport properties through quantum point contacts can be
gained from the quantum-mechanical motion of electrons in
such a simple potential as given by Eq. 共1兲. In the standard
2DEGs usually described by Schrödinger’s equation, the en-
ergy dependence of the transmission coefficient through the
quadratic potential in Eq. 共1兲 is well known and given under
high magnetic fields 共i.e., neglecting Landau-level mixing
while keeping the magnetic length lB finite兲 by the
expression5
on the other side of the constriction while an electron at
E ⬎ 0 goes weakly backscattered through the available chan-
nels.
With monolayer graphene consisting of carbon atoms
packed in a two-dimensional honeycomb lattice appears a
new host material,6 where electrons are confined to two di-
mensions, yet with some exotic properties. The observation
of an anomalous quantum Hall effect in graphene7,8 under-
stood in terms of a relativisticlike spectrum of low-energy
electrons9 has been followed by numerous experimental and
theoretical contributions10 aiming at exhibiting specific sig-
natures of the 2D massless Dirac fermions in a nonuniform
potential. The studies of simple analytical problems, such as
the quantum-mechanical motion of massless particles in the
quadratic saddle-point potential 关Eq. 共1兲兴, are of valuable in-
terest for identifying such unusual properties. However, it
共1兲 turns out that quadratic potentials, which are exactly solvable
in ordinary 2DEGs at any magnetic field, become generally
not analytically solvable within the Dirac equation. Here, we
consider the regime of large magnetic fields, in which the
Landau-level spacing in graphene is large enough so that one
can work in the single Landau-level limit. We have recently
shown11 using a semicoherent-state Green’s-function formal-
ism that quadratic problems then become soluble in this re-
gime.
The determination of the transmission coefficient through
a smooth constriction for graphene in the high magnetic field
limit 关i.e., the counterpart of Eq. 共2兲兴 is the main result of this
Rapid Communication, that we start by discussing in relation
to conductance quantization in graphene, with a detailed
derivation using semicoherent-state Green’s functions mak-

再 冋 册冎
ing the rest of the Rapid Communication. The exact trans-
E − 共n + 1/2兲共ប␻c + ␨兲 −1 mission coefficient for graphene in the absence of Landau-
Tn共E兲 = 1 + exp − ␲ 共2兲 level mixing reads in the nth Landau level 共with n ⱖ 1兲
lB2 冑ab
for the nth Landau level, where lB = 冑បc / 共兩e兩B兲 is the mag-
netic length, ប␻c is the Landau-level spacing in the 2DEG,
冋 冉
Tn,⑀共E兲 = 1 + exp − ⑀␲
E − En,⑀
lB2 冑ab
冊册 −1
, 共3兲

and ␨ = lB2 共b − a兲. Tunneling processes give rise to a nonzero where ⑀ = ⫾ 1 is a band index characterizing the electron and
probability for the electron at energy E ⬍ 0 to be transmitted holelike contributions, and

1098-0121/2010/82共16兲/161408共4兲 161408-1 ©2010 The American Physical Society

 RAPID COMMUNICATIONS

FLÖSER, FLORENS, AND CHAMPEL PHYSICAL REVIEW B 82, 161408共R兲 共2010兲

in an experiment. In graphene, the effect of the potential

14 asymmetric saddle point
asymmetry should be more easily seen in experiments since
symmetric saddle point
it yields an asymmetry between the positive- and negative-
10 energy dependences of the conductance.


Now, we focus on the derivation of Eq. 共3兲. We consider a

e2
h

single-particle Hamiltonian model for an electron of charge

G

6
e = −兩e兩 and of Fermi velocity vF confined to a two-
2
dimensional graphene sheet in the plane 共x , y兲 in the pres-
ence of both a perpendicular uniform magnetic field B = Bẑ
0
−2 −1 0 1 2
and an electrostatic 共scalar兲 potential term V共r兲 关given by Eq.
共1兲兴
µ [h̄Ωc ]

FIG. 1. 共Color online兲 Zero-temperature conductance for a

saddle-point electrostatic potential with l2Ba = l2Bb = 0.05⍀c 共symmet-
Ĥ = vF 冉 0
⌸x + i⌸y
⌸x − i⌸y
0
冊 + V共r兲1̂ 共6兲

ric case兲 and l2Ba = 4l2Bb = 0.1⍀c 共asymmetric case兲. Asymmetries of
the electrostatic saddle-point potential reflect in asymmetries be-
tween the electron and hole sectors for the conductance.
En,⑀ = n␨ + ⑀冑n共ប⍀c兲2 + ␨2/4 共4兲
with ⍀c = 冑2vF / lB 共vF is the graphene Fermi velocity兲. The
transmission probability for the lowest Landau level 共n = 0兲 is
given by T0共E兲 = 1 / 2.
The zero-temperature conductance at chemical potential ␮
is given in terms of the transmission probabilities by the
with ⌸ = −iបⵜr − eA共r兲 / c, where A共r兲 is the vector potential
defined by the equation ⵜr ⫻ A共r兲 = B. Here 1̂ corresponds to
the unity matrix in the pseudospin space 共representing elec-
tron and hole degrees of freedom兲 and c is the speed of light.
For convenience, we will omit both physical spin and valley
indices, thus assuming that the two valleys of graphene re-
main completely decoupled form each other and can be stud-
ied separately. In addition, we shall not consider the effects
of ripples or a mass potential but these could be studied
following Ref. 11.
To describe the electron dynamics at high magnetic field,
it is useful11 to introduce the graphene vortex states

冉 冊
Landauer-Büttiker formula
1 ␭⌿n−1,R共r兲
⌿̃n,R,␭共r兲 = , 共7兲
冑1 + 兩␭兩
冋 册
i⌿n,R共r兲
+⬁
4e2
T0共␮兲 + 兺 兺 Tn,⑀共␮兲 ,
冉冑 冊
G共␮兲 = 共5兲 2+兩Z兩2−2Zzⴱ兲/共4l2 兲
h n=1 ⑀=⫾ e−共兩z兩 B z−Z n
⌿n,R共r兲 = 共8兲
冑2␲lB2 n! 2lB
where we have accounted for the spin and valley degenera- with z = x + iy and Z = X + iY. Here R = 共X , Y兲 is a doubly con-
cies in graphene with the overall prefactor 4. tinuous quantum number corresponding to the guiding center
Then, as found in Ref. 12 for the 2DEG case, the conduc- position in the plane, n is a positive integer, ␭ is a band index
tance quantization in the graphene case shown in Fig. 1 di- 共defined for a given n兲, which is equal to ⫾1 if n ⱖ 1, and 0
rectly follows from the transmission probabilities. The first for n = 0. States in Eq. 共7兲, which can be written as
obvious observation is the half-integer quantization of the ⌿̃n,R,␭共r兲 = 具r 兩 n , R , ␭典 within the Dirac bracket notation, are
conductance in terms of the conductance quantum 共here eigenstates of Hamiltonian 共6兲 in absence of the electrostatic
4e2 / h兲 with plateaus at values 共n + 1 / 2兲4e2 / h, reminiscent of potential 共V = 0兲 with the energy quantization En,␭ = ␭冑nប⍀c.
the half-integer quantization of the Hall conductance. Two Despite being nonorthogonal with respect to the degeneracy
different configurations, symmetric and asymmetric with re- quantum number R
spect to ␲ / 2 rotation of the saddle point potential in Eq. 共1兲,
have been considered, which yield to two different curves. A 具n1,R1,␭1兩n2,R2,␭2典 = ␦n1,n2具R1兩R2典␦␭1,␭2 , 共9兲
symmetric saddle-point potential is characterized by a = b,
thus ␨ = 0, and the conductance shows clear particle-hole
symmetry with respect to change in the energy sign. In the
case of an asymmetric saddle-point potential 共a ⫽ b兲, ␨ be-
具R1兩R2典 = exp − 冋 共R1 − R2兲2 − 2iẑ · 共R1 ⫻ R2兲
4lB2
册 共10兲

comes nonzero, signaling a breaking of particle-hole symme- the set of quantum numbers 兩n , R , ␭典 obeys a completeness
try in the energy levels Eq. 共4兲, and resulting in a nonuniform relation
shift of the conductance steps, which is more pronounced for

冕
+⬁
the highest Landau levels. It is worth noting that the asym- d 2R
metry of the potential in Eq. 共1兲 has a different consequence 2␲lB2
兺 兺 兩n,R,␭典具n,R,␭兩 = 1̂.
n=0 ␭
共11兲
in the case of the standard 2DEG, where it just leads to a
redefinition of the Landau-level spacing, see Eq. 共2兲, and In fact, the states in Eq. 共7兲 form an overcomplete basis of
such a small quantitative modification appears difficult to states, which have the coherent states character with respect
perceive to the quantum number R. Relation 共11兲 allows one to

161408-2
 RAPID COMMUNICATIONS

TRANSMISSION COEFFICIENT THROUGH A SADDLE-… PHYSICAL REVIEW B 82, 161408共R兲 共2010兲

project the electron dynamics onto the vortex representation. 冑nប⍀c ␨

␣n = ␤n = 共17兲
We can then introduce the vortex Green’s function 冑n共ប⍀c兲2 + ␨2/4 , 冑n共ប⍀c兲2 + ␨2/4
G共n1 , R1 , ␭1 , t1 ; n2 , R2 , ␭2 , t2兲, which gives the probability
amplitude for a vortex with circulation 共or Landau-level in- and h0;0;0共t兲 = e−it␨/2 for the lowest Landau level n = 0. It is
dex兲 n1 and band index ␭1 that is initially at position R1 at worth stressing that the previous expressions are valid for
time t1 to be at point R2 at time t2 with a new circulation n2 Landau-level indices n not too high, for which the inequality
and a band index ␭2. In the following, we consider the dy- 兩␨兩 , 冑兩␥兩 Ⰶ 共冑n + 1 − 冑n兲ប⍀c holds.
namics projected onto a single Landau level, meaning that The action of the differential operator exp关共lB2 / 4兲⌬R兴 on
the vortex circulation is conserved 共n1 = n2 = n兲. Formally, this the function g̃共R兲 in Eq. 共12兲 is evaluated as13
corresponds to taking the limit vF → +⬁. Within a single Lan-
dau level n, the retarded Green’s function takes the form11
2
Gn;␭1;␭2共R1,R2兲 = 具R1兩R2典 冕 d 2u
␲lB2 g̃n;␭1;␭2共u兲e
2
−共u − R12兲2/lB
.
Gn;␭1;␭2共R1,R2兲 = 具R1兩R2典e共lB/4兲⌬R12关g̃n;␭1;␭2共R12兲兴, 共12兲
共18兲

1 Then, inserting Eq. 共14兲 into Eq. 共18兲, we can perform the
R12 = 关R1 + R2 + i共R2 − R1兲 ⫻ ẑ兴, 共13兲 Gaussian integrals over u to get
2

冕 +⬁ +兲
− ieit共E+i0
where ⌬R is the Laplacian operator. In the absence of Gn;␭1;␭2共R1,R2兲 = 具R1兩R2典 dt hn;␭1;␭2共t兲
Landau-level mixing and for a quadratic saddle-point elec- 0 cosh共冑兩␥兩t兲

⫻ 冑 f共t兲e−if共t兲␶共t兲V共R12兲e␥ f共t兲␶
trostatic potential with the spatial dependence given by 2共t兲共R2 /l2 兲
12 B
Eq. 共1兲, the function g̃ can be calculated exactly11 with the
result in the energy representation 共i.e., after Fourier trans- 共19兲
formation with respect to the time difference t1 − t2 = t兲
with f共t兲 = 关1 + i␨␶共t兲 − ␥␶ 共t兲兴 .2 −1

g̃n;␭1;␭2共R兲 = 冕0
+⬁
dt
− ie −i␶共t兲V共R兲

cosh共冑兩␥兩t兲
hn;␭1;␭2共t兲e it共E+i0+兲
, 共14兲
We note from Eq. 共16兲 that for n ⱖ 1 and ␨ ⫽ 0 the func-
tion hn;␭1;␭2共t兲 is not diagonal in the ␭ space, indicating that ␭
is generically no more a good quantum number in the pres-
where ␶共t兲 = 共1 / 冑兩␥兩兲tanh共冑兩␥兩t兲. Here E is the energy and 0+ ence of asymmetric saddle-point potentials. A straightfor-
an infinitesimal positive quantity. The parameters ␥ and ␨ are ward diagonalization shows that ⑀ = ⫾ appears instead as a
geometric coefficients characterizing V共r兲 good number with hn;⑀共t兲 = e−itEn,⑀. According to the above
formula 共4兲 for En,⑀, the quantum number ⑀ clearly labels the
lB4 2 lB2 electronlike and holelike energy bands. We shall henceforth
␥= 关共⳵ V兲共⳵2y V兲 − 共⳵x⳵yV兲2兴, ␨= ⌬rV共r兲. 共15兲 represent the Green’s function in this ⑀ representation, where
4 x 2
the latter takes a diagonal form.
The coefficient ␥ is directly proportional to the Gaussian In order to determine the transmission coefficient with a
curvature of the electrostatic potential, which turns out to be given energy channel 共i.e., at Landau-level index n and band
negative for a saddle-shaped 共or hyperbolic兲 quadratic func- index ⑀ fixed兲, we only need the Green’s function
tion V共r兲. Finally, the functions hn;␭1;␭2共t兲, which contain the Gn;⑀共R1 , R2兲 when the states at vortex positions R1 and R2
full dependences on the Landau-level index n and on the correspond to the same energy and are asymptotically far
band indices ␭1 and ␭2 read for n ⱖ 1 from the saddle-point located at the origin. For a saddle-
point potential of the form 共1兲, this means taking the limits
e−itEn,⑀
hn;␭1;␭2共t兲 = 兺 关共1 + ⑀␭1␣n兲␦␭ ,␭
⑀=⫾
1 2
+ ⑀␤n␦−␭1,␭2兴
2
, 兩X1兩 → ⬁ and 兩X2兩 → ⬁ while V共X1 , Y 1兲 = V共X2 , Y 2兲 = const.
Making the change in variable s = d关1 − 兩␥兩␶2共t兲兴 / 关1
共16兲 + 兩␥兩␶2共t兲兴, where d = 兩X1Y 2 − X2Y 1兩 / lB2 , we can easily take the
limit d → +⬁ in the integral in Eq. 共19兲, and obtain the
where En,⑀ is defined in Eq. 共4兲 expression

⬁
Gn; ⑀共R1,R2兲 = ⌫ 冉 2
−i
2冑兩␥兩
e 冊
1 E − En,⑀ −␴共␲/4兲关共E−E 兲/冑兩␥兩+i兴
n,⑀
冑 冑 冑 冑 冑 冑 冑 冑 2
共− i兲ei关共 aX1− bY 1兲共 bX1+ aY 1兲+共 aX2+ bY 2兲共 bX2− aY 2兲兴/2lB共a+b兲

冑兩␥兩兲
d−1/2+i共E−En,⑀兲/共2 冑aX1 − 冑bY 1兲2+共冑aX2 + 冑bY 2兲2兴/2lB2 共a+b兲
⫻ e−关共 , 共20兲
冑4兩␥兩 + 2i␨冑兩␥兩

161408-3

FLÖSER, FLORENS, AND CHAMPEL
refl
refl
Y
trans
in
in trans
X tn,−
FIG. 2. 共Color online兲 Schematic to identify the Green’s func-
tions for reflection and transmission. Semiclassically the trajectories
of vortex states lie on contours of constant potential. Solid 共dashed兲
lines correspond to an equipotential line V ⬎ 0 共V ⬍ 0兲 and therefore
for E ⬎ 0 to a vortex in the conduction 共valence兲 band, respectively.
Dotted lines denote the asymptotes of the potential in Eq. 共1兲.
where ␴ = sgn共X1X2兲 and ⌫共z兲 is the Gamma function. At
infinity the vortex is close to the asymptotes of the saddle-
point potential V共R兲, i.e., Y 1 ⬃ X1冑a / b and Y 2 ⬃ −X2冑a / b.
The two modes ⑀ = ⫾ circulate in the same direction and are
not mixed since they are well separated in energy.
Within the present saddle-point geometry, the transmis-
sion coefficient can be extracted from the asymptotic form of
the retarded vortex Green’s function, thus performing scat-
tering theory in terms of coherent-state wave packets instead
of the more standard plane waves. In the conduction band
共⑀ = +兲, the transmission of a vortex from the left half-plane
to the right half-plane is described by Green’s function ex-
pression 关Eq. 共20兲兴 with the sign function ␴ = − while the
vortex reflection, where the vortex remains in the left half-
plane, is characterized by ␴ = + 共see Fig. 2兲. Therefore, the
ratio of the transmission amplitude to the reflection ampli-
tude for a given Landau level n in the conduction band is
given by the ratio of these two Green’s functions
1 I. Neder, N. Ofek, Y. Chung, M. Heiblum, D. Mahalu, and V.
Umansky, Nature 共London兲 448, 333 共2007兲.
2 H. A. Fertig and B. I. Halperin, Phys. Rev. B 36, 7969 共1987兲.
3
J. T. Chalker and P. D. Coddington, J. Phys. C 21, 2665 共1988兲.
4 B. Kramer, T. Ohtsuki, and S. Kettemann, Phys. Rep. 417, 211
共2005兲.
5 This result for the 2DEG can be derived from Eqs. 共1.2兲–共1.6兲 of
Ref. 2 in the limit ␻c → ⬁ by neglecting the Landau-level mixing
correction to the curvature energy E1. Such corrections lead to
additional yet small contributions to the energy dependence of
the transmission coefficient in the large cyclotron frequency
limit, without affecting its qualitative behavior.
161408-4
RAPID COMMUNICATIONS
PHYSICAL REVIEW B 82, 161408共R兲 共2010兲
tn,+
rn,+
= i exp冋␲ E − En,+
2 冑兩␥兩
.册 共21兲
In the valence band the vortices carry a positive charge.
Therefore, the correspondence between the transmission 共or
reflection兲 process of a negative charge carrier and the sign
of ␴ is now reversed so that we have
冋 册
rn,− ␲ E − En,−
= i exp . 共22兲
2 冑兩␥兩
The relations between the transmission probabilities Tn,⑀ and
the transmission and reflection amplitudes, i.e., 兩tn,⑀兩2 = 1
− 兩rn,⑀兩2 = Tn,⑀, finally provide result in Eq. 共3兲.
For the peculiar case of the lowest Landau level, contri-
butions both from the original conduction and valence bands
arise. We get for the electronlike excitations T+0 共E兲 = 兵1
+ exp关−␲共E − ␨ / 2兲 / 冑兩␥兩兴其−1, and for the holelike excitations
T−0 共E兲 = 兵1 + exp关␲共E − ␨ / 2兲 / 冑兩␥兩兴其−1. Summing up these two
contributions and considering the equipartition of the current
between the two types of excitations yielding a 1/2 prefactor,
we finally get the already quoted lowest Landau-level con-
tribution
1 1
T0共E兲 = 关T+0 共E兲 + T−0 共E兲兴 = . 共23兲
2 2
In conclusion, we have calculated the transmission coef-
ficient for a quadratic saddle-point electrostatic potential in
graphene, and found that shape asymmetries generic to quan-
tum point contacts break particle-hole symmetry in the con-
ductance. Our results should be relevant for future split gate
experiments in graphene as well as for the formulation of
quantum network models. Theoretically, we have also pre-
sented an alternative way of deriving transmission coeffi-
cients from the scattering of coherent-state wave packets.
6 A. K. Geim and K. S. Novoselov, Nature Mater. 6, 183 共2007兲.
7
K. S. Novoselov et al., Nature 共London兲 438, 197 共2005兲.
8 Y. Zhang, Y. W. Tan, H. L. Stormer, and P. Kim, Nature 共Lon-
don兲 438, 201 共2005兲.
9 V. P. Gusynin and S. G. Sharapov, Phys. Rev. Lett. 95, 146801
共2005兲.
10 A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov,
and A. K. Geim, Rev. Mod. Phys. 81, 109 共2009兲.
11 T. Champel and S. Florens, Phys. Rev. B 82, 045421 共2010兲.
12
M. Büttiker, Phys. Rev. B 41, 7906 共1990兲.
13 T. Champel and S. Florens, Phys. Rev. B 80, 125322 共2009兲; 80,
161311共R兲 共2009兲.