ARTICLES

Chiral tunnelling and the Klein paradox

in graphene
M. I. KATSNELSON1 *, K. S. NOVOSELOV2 AND A. K. GEIM2 *
1
Institute for Molecules and Materials, Radboud University Nijmegen, 6525 ED Nijmegen, The Netherlands
2
Manchester Centre for Mesoscience and Nanotechnology, University of Manchester, Manchester M13 9PL, UK
* e-mail: [email protected]; [email protected]

Published online: 20 August 2006; doi:10.1038/nphys384

he term Klein paradox1–7 refers to a counterintuitive relativistic

T
The so-called Klein paradox—unimpeded penetration
process in which an incoming electron starts penetrating
of relativistic particles through high and wide potential through a potential barrier if its height, V0 , exceeds the
electron’s rest energy, mc 2 (where m is the electron mass and c
barriers—is one of the most exotic and counterintuitive
is the speed of light). In this case, the transmission probability,
consequences of quantum electrodynamics. The T , depends only weakly on the barrier height, approaching the
perfect transparency for very high barriers, in stark contrast to
phenomenon is discussed in many contexts in particle, the conventional, non-relativistic tunnelling where T exponentially
decays with increasing V0 . This relativistic effect can be attributed
nuclear and astro-physics but direct tests of the Klein to the fact that a sufficiently strong potential, being repulsive
paradox using elementary particles have so far proved for electrons, is attractive for positrons and results in positron
states inside the barrier, which align in energy with the electron
impossible. Here we show that the effect can be tested in continuum outside4–6 . Matching between electron and positron
wavefunctions across the barrier leads to the high-probability
a conceptually simple condensed-matter experiment using tunnelling described by the Klein paradox7 . The essential feature
of quantum electrodynamics (QED) responsible for the effect is
electrostatic barriers in single- and bi-layer graphene.
the fact that states at positive and negative energies (electrons
Owing to the chiral nature of their quasiparticles, quantum and positrons) are intimately linked (conjugated), being described
by different components of the same spinor wavefunction. This
tunnelling in these materials becomes highly anisotropic, fundamental property of the Dirac equation is often referred to
as the charge-conjugation symmetry. Although Klein’s gedanken
qualitatively different from the case of normal, non- experiment is now well understood, the notion of paradox is still
relativistic electrons. Massless Dirac fermions in graphene widely used2–7 , perhaps because the effect has never been observed
experimentally. Indeed, its observation requires a potential drop
allow a close realization of Klein’s gedanken experiment, of ≈mc 2 over the Compton length h̄/mc , which yields enormous
electric fields2,3 (ε > 1016 V cm−1 ) and makes the effect relevant
whereas massive chiral fermions in bilayer graphene offer only for such exotic situations as, for example, positron production
around super-heavy nuclei2,3 with charge Z ≥ 170 or evaporation
an interesting complementary system that elucidates the
of black holes through generation of particle–antiparticle pairs
basic physics involved. near the event horizon8 . The purpose of this paper is to show
that graphene—a recently found allotrope of carbon9 —provides
an effective medium (‘vacuum’) where relativistic quantum
tunnelling described by the Klein paradox and other relevant QED
phenomena can be tested experimentally.
DIRAC-LIKE QUASIPARTICLES IN GRAPHENE

Graphene is a single layer of carbon atoms densely packed in a

honeycomb lattice, or it can be viewed as an individual atomic
plane pulled out of bulk graphite. From the point of view of

620 nature physics VOL 2 SEPTEMBER 2006 www.nature.com/naturephysics

©2006 Nature Publishing Group

 ARTICLES

a intersection near the edges of the Brillouin zone (shown in red

and green in Fig. 1a) yields the conical energy spectrum. As a
result, quasiparticles in graphene exhibit the linear dispersion
σ relation E = h̄kvF , as if they were massless relativistic particles with
–k k momentum k (for example, photons) but the role of the speed of
–σ σ
q
V0 E light is played here by the Fermi velocity vF ≈ c/300. Owing to
the linear spectrum, it is expected that graphene’s quasiparticles
will behave differently from those in conventional metals and
semiconductors where the energy spectrum can be approximated
by a parabolic (free-electron-like) dispersion relation.
Although the linear spectrum is important, it is not the
b only essential feature that underpins the description of quantum
k
V0 transport in graphene by the Dirac equation. Above zero
E D
energy, the current carrying states in graphene are, as usual,
electron-like and negatively charged. At negative energies, if
the valence band is not full, its unoccupied electronic states
c behave as positively charged quasiparticles (holes), which are
often viewed as a condensed-matter equivalent of positrons. Note,
however, that electrons and holes in condensed-matter physics
are normally described by separate Schrödinger equations, which
are not in any way connected (as a consequence of the Seitz
E V0 sum rule15 , the equations should also involve different effective
masses). In contrast, electron and hole states in graphene are
interconnected, exhibiting properties analogous to the charge-
conjugation symmetry in QED10–12 . For the case of graphene, the
latter symmetry is a consequence of its crystal symmetry because
graphene’s quasiparticles have to be described by two-component
wavefunctions, which are needed to define relative contributions of
Figure 1 Tunnelling through a potential barrier in graphene. a, Schematic sublattices A and B in quasiparticles’ make-up. The two-component
diagrams of the spectrum of quasiparticles in single-layer graphene. The spectrum description for graphene is very similar to the one by spinor
is linear at low Fermi energies (<1 eV). The red and green curves emphasize the wavefunctions in QED, but the ‘spin’ index for graphene indicates
origin of the linear spectrum, which is the crossing between the energy bands sublattices rather than the real spin of electrons and is usually
associated with crystal sublattices A and B. b, Potential barrier of height V0 and referred to as pseudospin σ .
width D. The three diagrams in a schematically show the positions of the Fermi There are further analogies with QED. The conical spectrum
energy E across such a barrier. The Fermi level (dotted lines) lies in the conduction of graphene is the result of intersection of the energy bands
band outside the barrier and the valence band inside it. The blue filled areas indicate originating from sublattices A and B (see Fig. 1a) and, accordingly,
occupied states. The pseudospin denoted by vector σ is parallel (antiparallel) to the an electron with energy E propagating in the positive direction
direction of motion of electrons (holes), which also means that σ keeps a fixed originates from the same branch of the electronic spectrum (shown
direction along the red and green branches of the electronic spectrum. in red) as the hole with energy −E propagating in the opposite
c, Low-energy spectrum for quasiparticles in bilayer graphene. The spectrum is direction. This yields that electrons and holes belonging to the
isotropic and, despite its parabolicity, also originates from the intersection of energy same branch have pseudospin σ pointing in the same direction,
bands formed by equivalent sublattices, which ensures charge conjugation, similar which is parallel to the momentum for electrons and antiparallel for
to the case of single-layer graphene. holes (see Fig. 1a). This allows the introduction of chirality12 , that
is formally a projection of pseudospin on the direction of motion,
which is positive and negative for electrons and holes, respectively.
its electronic properties, graphene is a two-dimensional zero-gap The term chirality is often used to refer to the additional built-in
semiconductor with the energy spectrum shown in Fig. 1a, and its symmetry between electron and hole parts of graphene’s spectrum
low-energy quasiparticles are formally described by the Dirac-like (as indicated by colour in Fig. 1) and is analogous (although not
hamiltonian10–12 completely identical11,16 ) to the chirality in three-dimensional QED.
0 = −i h̄vF σ∇,
H (1)
KLEIN PARADOX REFORMULATED FOR SINGLE-LAYER GRAPHENE
where vF ≈ 106 m s−1 is the Fermi velocity and σ = (σ x , σ y )
are the Pauli matrices. Neglecting many-body effects, this Because quasiparticles in graphene accurately mimic Dirac
description is accurate theoretically10–12 and has also been proved fermions in QED, this condensed-matter system makes it possible
experimentally13,14 by measuring the energy-dependent cyclotron to set up a tunnelling experiment similar to that analysed by Klein.
mass in graphene (which yields its linear energy spectrum) and, The general scheme of such an experiment is shown in Fig. 1, where
most clearly, by the observation of a relativistic analogue of the we consider the potential barrier that has a rectangular shape and is
integer quantum Hall effect. infinite along the y axis:
The fact that charge carriers in graphene are described by 
the Dirac-like equation (1), rather than the usual Schrödinger V0 , 0 < x < D,
V (x) = (2)
equation, can be seen as a consequence of graphene’s crystal 0 otherwise.
structure, which consists of two equivalent carbon sublattices10–12 ,
A and B. Quantum mechanical hopping between the sublattices This local potential barrier of width D inverts charge carriers
leads to the formation of two cosine-like energy bands, and their underneath it, creating holes playing the role of positrons, or

nature physics VOL 2 SEPTEMBER 2006 www.nature.com/naturephysics 621

©2006 Nature Publishing Group

ARTICLES

a 90° b 1.0 90°

1.0
60° 60°
0.8 0.8

0.6 0.6
30° 30°
0.4 0.4

0.2 0.2

0 0° 0 0°

0.2 0.2

0.4 0.4
–30° –30°
0.6 0.6

0.8 0.8
–60° –60°
1.0 1.0
–90° –90°

Figure 2 Klein-like quantum tunnelling in graphene systems. a,b, Transmission probability T through a 100-nm-wide barrier as a function of the incident angle for
single- (a) and bi-layer (b) graphene. The electron concentration n outside the barrier is chosen as 0.5× 1012 cm−2 for all cases. Inside the barrier, hole concentrations p are
1× 1012 and 3× 1012 cm−2 for red and blue curves, respectively (such concentrations are most typical in experiments with graphene). This corresponds to the Fermi energy
E of incident electrons ≈80 and 17 meV for single- and bi-layer graphene, respectively, and l ≈ 50 nm. The barrier heights V0 are (a) 200 and (b) 50 meV (red curves) and
(a) 285 and (b) 100 meV (blue curves).

vice versa. For simplicity, we assume in (2) infinitely sharp edges, angle, s = sgn E and s = sgn(E − V0 ). Requiring the continuity of
which allows a direct link to the case usually considered in QED1–7 . the wavefunction by matching up coefficients a, b, t , r , we find the
The sharp-edge assumption is justified if the Fermi wavelength, following expression for the reflection coefficient r
l, of quasiparticles is much larger than the characteristic width
of the edge smearing, which in turn should be larger than the r = 2ieiφ sin(q x D)
lattice constant (to disallow Umklapp scattering between different sin φ − ss sin θ
valleys in graphene)17 . Such a barrier can be created by the × .
electric field effect using a thin insulator or by local chemical ss [e−iqx D cos(φ + θ) + eiqx D cos(φ − θ)] − 2i sin(q x D)
doping9,13,14 . Importantly, Dirac fermions in graphene are massless (3)
and, therefore, there is no formal theoretical requirement for
the minimal electric field, ε, to form positron-like states under Figure 2a shows examples of the angular dependence of
the barrier. To create a well-defined barrier in realistic graphene transmission probability T = |t|2 = 1 − |r|2 calculated using the
samples with a disorder, fields ε ≈ 105 V cm−1 routinely used in above expression. In the limit of high barriers |V0 |  |E|, the
experiments9,14 should be sufficient, which is eleven orders of expression for T can be simplified to
magnitude lower than the fields necessary for the observation of
the Klein paradox for elementary particles. cos2 φ
T= . (4)
It is straightforward to solve the tunnelling problem shown in 1 − cos2 (q x D) sin2 φ
Fig. 1b. We assume that the incident electron wave propagates at an
angle φ with respect to the x axis and then try the components of Equations (3) and (4) yield that under resonance conditions
the Dirac spinor ψ1 and ψ2 for the hamiltonian H = H0 + V (x) in q x D = πN , N = 0,±1,. . . the barrier becomes transparent (T = 1).
the following form: More significantly, however, the barrier always remains perfectly
transparent for angles close to the normal incidence φ = 0. The
⎧
⎨ (eikx x + r e−ikx x )eiky y , x < 0, latter is the feature unique to massless Dirac fermions and is
ψ1 (x, y) = (aeiqx x + be−iqx x )eiky y , 0 < x < D, directly related to the Klein paradox in QED. This perfect tunnelling
⎩ t eikx x+iky y , x > D, can be understood in terms of the conservation of pseudospin.
⎧ Indeed, in the absence of pseudospin-flip processes (such processes
⎨ s(eikx x+iφ − r e−ikx x−iφ )eiky y , x < 0, are rare as they require a short-range potential, which would act
ψ2 (x, y) = s (aeiqx x+iθ − be−iqx x−iθ )eiky y , 0 < x < D, differently on A and B sites of the graphene lattice), an electron
⎩ st eikx x+iky y+iφ , x > D, moving to the right can be scattered only to a right-moving electron
state or left-moving hole state. This is shown in Fig. 1a, where
charge carriers from the ‘red’ branch of the band diagram can
where kF = 2π/l is the Fermi wavevector, k x = kF cos φ and
be scattered into states within the same ‘red’ branch but cannot
k y = k
F sin φ are the wavevector components outside the barrier,
be transformed into any state on the ‘green’ branch. The latter
qx = (E − V0 )2 / h̄2 vF2 − k2y , θ = tan−1 (k y /q x ) is the refraction scattering event would require the pseudospin to be flipped. The

622 nature physics VOL 2 SEPTEMBER 2006 www.nature.com/naturephysics

©2006 Nature Publishing Group

 ARTICLES

matching between directions of pseudospin σ for quasiparticles

inside and outside the barrier results in perfect tunnelling. In the 1.0
strictly one-dimensional case, such perfect transmission of Dirac
fermions has been discussed in the context of electron transport in
carbon nanotubes17,18 (see also ref. 19). Our analysis extends this
0.8
tunnelling problem to the two-dimensional (2D) case of graphene.

CHIRAL TUNNELLING IN BILAYER GRAPHENE

0.6

T
To elucidate which features of the anomalous tunnelling in
graphene are related to the linear dispersion and which features
are related to the pseudospin and chirality of the Dirac spectrum, 0.4
it is instructive to consider the same problem for bilayer
graphene. There are differences and similarities between the two
graphene systems. Indeed, charge carriers in bilayer graphene 0.2
have a parabolic energy spectrum as shown in Fig. 1c, which
means they are massive quasiparticles with a finite density of
states at zero energy, similar to conventional non-relativistic 0
electrons. On the other hand, these quasiparticles are also chiral 0 10 20 30 40 50
and described by spinor wavefunctions20,21 , similar to relativistic D (nm)
particles or quasiparticles in single-layer graphene. Again, the
origin of the unusual energy spectrum can be traced to the
crystal lattice of bilayer graphene with four equivalent sublattices21 . Figure 3 Chiral versus non-chiral tunnelling. Transmission probability T for
Although ‘massive chiral fermions’ do not exist in the field normally incident electrons in single- and bi-layer graphene (red and blue curves,
theory, their existence in condensed-matter physics (confirmed respectively) and in a non-chiral zero-gap semiconductor (green curve) as a function
experimentally20 ) offers a unique opportunity to clarify the of width D of the tunnel barrier. Concentrations of charge carriers are chosen as
importance of chirality in the relativistic tunnelling problem n = 0.5× 1012 cm−2 and p = 1× 1013 cm−2 outside and inside the barrier,
described by the Klein paradox. In addition, the relevant QED-like respectively, for all three cases. This yields barrier heights of ∼450 meV for
effects seem to be more pronounced in bilayer graphene and easier graphene and ∼240 meV for the other two materials. Note that the transmission
to test experimentally, as discussed below. probability for bilayer graphene decays exponentially with the barrier width, even
Charge carriers in bilayer graphene are described by an off- though there are plenty of electronic states inside the barrier.
diagonal hamiltonian20,21
2 
0 = − h̄ 0 (k x − ik y )2
H (5)
2m (k x + ik y )2 0 manifest itself. In this case, scattering at the barrier (2) is the same
as for electrons described by the Schrödinger equation. However,
which yields a gapless semiconductor with chiral electrons and for any finite φ (even in the case V0 < E ), waves localized at the
holes with a finite mass m. An important formal difference between barrier interfaces are essential to satisfy the boundary conditions.
the tunnelling problems for single- and bi-layer graphene is that in The most intriguing behaviour is found for V0 > E , where
the latter case there are four possible solutions for a given energy electrons outside the barrier transform into holes inside it, or
E = ±h̄2 kF2 /2m. Two of them correspond to propagating waves vice versa. Examples of the angular dependence of T in bilayer
and the other two to evanescent waves. Accordingly, for constant graphene are plotted in Fig. 2b. They show a dramatic difference
potential V i , eigenstates of hamiltonian (5) should be written as compared with the case of massless Dirac fermions. There are
again pronounced transmission resonances at some incident angles,
ψ1 (x, y) = (a i eikix x + b i e−ikix x + c i eκix x + d i e−κix x )eiky x where T approaches unity. However, instead of the perfect
transmission found for normally incident Dirac fermions (see
 Fig. 2a), our numerical analysis has yielded the opposite effect:
di
ψ2 (x, y) = s i a i eikix x+2iφi + b i e−ikix x−2iφi − c i h i eκix x − e−κix x eiky y massive chiral fermions are always perfectly reflected for angles
hi close to φ = 0.
Accordingly, we have analysed this case in more detail and
where found the following analytical solution for the transmission

coefficient t :
s i = sgn (V i − E); h̄k ix = 2m|E − V i | cos φ i ;

h̄k iy = 2m|E − V i | sin φ i 4ik1 k2

 2 t= , (6)
 (k2 + ik1 ) e − (k2 − ik1 )2 ek2 D
2 −k2 D

κ ix = k2ix + 2k2iy ; h i = 1 + sin2 φ i − sin φ i .

where subscripts 1 and 2 label the regions outside and inside
To find the transmission coefficient through barrier (2), we the barrier, respectively. The case of a potential step, which
should set d1 = 0 for x < 0, b3 = c3 = 0 for x > D and satisfy the corresponds to a single p–n junction, is particularly interesting.
continuity conditions for both components of the wavefunction Equation (6) shows that such a junction should completely reflect
and their derivatives. For the case of an electron beam that is a normally incident beam (T = 0). This is highly unusual because
incident normally (φ = 0) and low barriers V0 < E (over-barrier the continuum of electronic states at the other side of the step is
transmission), we obtain ψ1 = −ψ2 both outside and inside the normally expected to allow some tunnelling. Furthermore, for a
barrier, and the chirality of fermions in bilayer graphene does not single p–n junction with V0  E , the following analytical solution

nature physics VOL 2 SEPTEMBER 2006 www.nature.com/naturephysics 623

©2006 Nature Publishing Group

ARTICLES

for any φ has been found: a

E
T= sin2 (2φ) (7)
V0

which again yields T = 0 for φ = 0. This behaviour is in

obvious contrast to single-layer graphene, where normally incident
electrons are always perfectly transmitted.
The perfect reflection (instead of the perfect transmission) can
be viewed as another incarnation of the Klein paradox, because the
effect is again due to the charge-conjugation symmetry (fermions
in single- and bi-layer graphene exhibit chiralities that resemble
those associated with spin 1/2 and 1, respectively)20,21 . For single- b
layer graphene, an electron wavefunction at the barrier interface
perfectly matches the corresponding wavefunction for a hole with
the same direction of pseudospin (see Fig. 1a), yielding T = 1. In D
φ
contrast, for bilayer graphene, the charge conjugation requires a
propagating electron with wavevector k to transform into a hole
with wavevector ik (rather than −k), which is an evanescent wave
inside a barrier.
COMPARISON WITH TUNNELLING OF NON-CHIRAL PARTICLES

For completeness, we compare the results obtained with the case Figure 4 The chiral nature of quasiparticles in graphene strongly affects its
of normal electrons. If a tunnel barrier contains no electronic transport properties. a, A diffusive conductor of a size smaller than the
states, the difference is obvious: the transmission probability in this phase-coherence length is connected to two parallel one-dimensional leads. For
case is known to decay exponentially with increasing barrier width normal electrons, transmission probability T through such a system depends
and height22 so that the tunnel barriers discussed above would strongly on the distribution of scatterers. In contrast, for massless Dirac fermions,
reflect electrons completely. However, both graphene systems are T is always equal to unity due to the additional memory about the initial direction of
gapless, and it is more appropriate to compare them with gapless pseudospin (see text). b, Schematic diagram of one of the possible tunnelling
semiconductors with non-chiral charge carriers (such a situation experiments in graphene. Graphene (light blue) has two local gates (dark blue) that
can be realized in certain heterostructures23,24 ). In this case, we find create potential barriers of a variable height. The voltage drop across the barriers is
measured by using potential contacts shown in orange.
4k x q x
t= ,
(q x + k x )2 e−iqx D − (q x − k x )2 eiqx D
consideration can be important for the understanding of the
where k x and q x are x -components of the wavevector outside minimal conductivity ≈e2 /h observed experimentally in both
and inside the barrier, respectively. Again, similar to the case single-layer13 and bilayer20 graphene.
of single- and bi-layer graphene, there are resonance conditions To further elucidate the dramatic difference between quantum
q x D = πN , N = 0,±1, . . . at which the barrier is transparent. For transport of Dirac fermions in graphene and normal 2D
the case of normal incidence (φ = 0), the tunnelling coefficient electrons, Fig. 4a suggests a gedanken experiment where a diffusive
is then an oscillating function of tunnelling parameters and can conductor is attached to ballistic one-dimensional leads, as in the
exhibit any value from 0 to 1 (see Fig. 3). This is in contrast Landauer formalism. For conventional 2D systems, transmission
to graphene, where T is always 1, and bilayer graphene, where and reflection coefficients through such a conductor are sensitive to
T = 0 for sufficiently wide barriers D > l. This makes it clear that detailed distribution of impurities and a shift of a single impurity by
the drastic difference between the three cases is essentially due to a distance of the order of l can completely change the coefficients27 .
different chiralities or pseudospins of the quasiparticles involved In contrast, the conservation of pseudospin in graphene strictly
rather than any other feature of their energy spectra. forbids backscattering and makes the disordered region in Fig. 4a
always completely transparent, independent of disorder (as long
IMPLICATIONS FOR EXPERIMENT as it is smooth on the scale of the lattice constant17 ). This
extension of the Klein problem to the case of a random scalar
The tunnelling anomalies found in the two graphene systems are potential has been proved by using the Lippmann–Schwinger
expected to play an important role in their transport properties, equation (see the Supplementary Information). Unfortunately, this
especially in the regime of low carrier concentrations, where particular experiment is probably impossible to realize in practice
disorder induces significant potential barriers and the systems because scattering at graphene’s edges does not conserve the
are likely to split into a random distribution of p–n junctions. pseudospin17,28 . Nevertheless, the above consideration shows that
In conventional 2D systems, strong enough disorder results in impurity scattering in the bulk of graphene should be suppressed
electronic states that are separated by barriers with exponentially compared with that of normal conductors.
small transparency25,26 . This is known to lead to the Anderson The above analysis shows that the Klein paradox and associated
localization. In contrast, in both graphene materials all potential relativistic-like phenomena can be tested experimentally using
barriers are relatively transparent (T ≈ 1 at least for some angles) graphene devices. The basic principle behind such experiments
which does not allow charge carriers to be confined by potential would be to use local gates and collimators similar to those used
barriers that are smooth on the atomic scale. Therefore, different in electron optics in 2D gases29,30 . One possible experimental setup
electron and hole ‘puddles’ induced by disorder are not isolated is shown schematically in Fig. 4b. Here, local gates simply cross
but effectively percolate, thereby suppressing localization. This the whole graphene sample at different angles (for example, 90◦

624 nature physics VOL 2 SEPTEMBER 2006 www.nature.com/naturephysics

©2006 Nature Publishing Group

 ARTICLES

and 45◦ ). Intrinsic concentrations of charge carriers are usually 7. Krekora, P., Su, Q. & Grobe, R. Klein paradox in spatial and temporal resolution. Phys. Rev. Lett. 92,
low (∼1011 cm−2 ), whereas concentrations up to 1 × 1013 cm−2 can 040406 (2004).
8. Page, D. N. Hawking radiation and black hole thermodynamics. New J. Phys. 7, 203 (2005).
be induced under the gated regions by the bipolar electric field 9. Novoselov, K. S. et al. Electric field effect in atomically thin carbon films. Science 306, 666–669 (2004).
effect9 . This allows potential barriers with heights up to V0 ≈ 0.4 eV 10. Slonczewski, J. C. & Weiss, P. R. Band structure of graphite. Phys. Rev. 109, 272 (1958).
11. Semenoff, G. W. Condensed-matter simulation of a three-dimensional anomaly. Phys. Rev. Lett. 53,
and ≈0.23 eV for single- and double-layer samples, respectively. 2449–2452 (1984).
By measuring the voltage drop across the barriers as a function 12. Haldane, F. D. M. Model for a quantum Hall effect without Landau levels: Condensed-matter
realization of the ‘parity anomaly’. Phys. Rev. Lett. 61, 2015–2018 (1988).
of applied gate voltage, their transparency for different V0 can be 13. Novoselov, K. S. et al. Two-dimensional gas of massless Dirac fermions in graphene. Nature 438,
analysed. Figure 2 shows that for graphene the 90◦ barrier should 197–200 (2005).
14. Zhang, Y., Tan, Y. W., Stormer, H. L. & Kim, P. Experimental observation of the quantum Hall effect
exhibit low resistance and no significant change in resistance with and Berry’s phase in graphene. Nature 438, 201–204 (2005).
changing gate voltage. In comparison, the 45◦ barrier is expected 15. Vonsovsky, S. V. & Katsnelson, M. I. Quatum Solid State Physics (Springer, Berlin, 1989) Sect. 4.6.6.
16. Boyanovsky, D., Blankenbecler, R. & Yahalom, R. Physical origin of topological mass in 2 + 1
to have much higher resistance and show a number of tunnelling dimensions. Nucl. Phys. B 270, 483–505 (1986).
resonances as a function of gate voltage. The situation should be 17. Ando, T., Nakanishi, T. & Saito, R. Berry’s phase and absence of back scattering in carbon nanotubes.
qualitatively different for bilayer graphene, where local barriers J. Phys. Soc. Japan 67, 2857–2862 (1998).
18. McEuen, P. L., Bockrath, M., Cobden, D. H., Yoon, Y. G. & Louie, S. G. Disorder, pseudospins, and
should result in a high resistance for the perpendicular barrier and backscattering in carbon nanotubes. Phys. Rev. Lett. 83, 5098–5101 (1999).
pronounced resonances for the 45◦ barrier. 19. Tworzydlo, J., Trauzettel, B., Titov, M., Rycerz, A. & Beenakker, C. W. J. Quantum-limited shot noise
in graphene. Phys. Rev. Lett. 96, 246802 (2006).
Furthermore, the fact that a barrier (or even a single p–n 20. Novoselov, K. S. et al. Unconventional quantum Hall effect and Berry’s phase of 2π in bilayer
junction) incorporated in a bilayer graphene device should lead graphene. Nature Phys. 2, 177–180 (2006).
21. McCann, E. & Falko, V. I. Landau-level degeneracy and quantum Hall effect in a graphite bilayer.
to exponentially small tunnelling current can be exploited in Phys. Rev. Lett. 96, 086805 (2006).
developing graphene-based field effect transistors (FET). Such 22. Esaki, L. New phenomenon in narrow germanium para-normal-junctions. Phys. Rev. 109,
603–604 (1958).
transistors are particularly promising because of their high mobility 23. Meyer, J. R., Hoffman, C. A., Bartoli, F. J. & Rammohan, L. R. Type-II quantum-well lasers for the
and ballistic transport at submicron distances9,13,14 . However, the midwavelength infrared. Appl. Phys. Lett. 67, 757–759 (1995).
24. Teissier, R. et al. Experimental determination of gamma-X intervalley transfer mechanisms in
fundamental problem along this route is that the conducting GaAs/AlAs heterostructures. Phys. Rev. B 54, 8329–8332 (1996).
channel in single-layer graphene cannot be pinched off (because of 25. Ziman, J. M. Models of Disorder (Cambridge Univ. Press, Cambridge, 1979).
the minimal conductivity), which severely limits achievable on–off 26. Lifshitz, I. M., Gredeskul, S. A. & Pastur, L. A. Introduction to the Theory of Disordered Systems
(Wiley, New York, 1988).
ratios for such FETs (ref. 9) and, therefore, the scope for their 27. Lee, P. A., Altshuler, B. L. & Webb, R. A. (eds) Mesoscopic Phenomena in Solids (North-Holland,
applications. A bilayer FET with a local gate inverting the sign of Amsterdam, 1991).
28. Berry, M. V. & Mondragon, R. J. Neutrino billiards—time reversal symmetry-breaking without
charge carriers should yield much higher on–off ratios. magnetic fields. Proc. R. Soc. London A 412, 53–74 (1987).
29. Spector, J., Stormer, H. L., Baldwin, K. W., Pfeiffer, L. N. & West, K. W. Electron focusing in
2-dimensional systems by means of an electrostatic lens. Appl. Phys. Lett. 56, 1290–1292 (1990).
Received 18 April 2006; accepted 20 June 2006; published 20 August 2006. 30. Dragoman, D. & Dragoman, M. Optical analogue structures to mesoscopic devices. Prog. Quantum
Electron. 23, 131–188 (1999).
References
1. Klein, O. Die reflexion von elektronen an einem potentialsprung nach der relativistischen dynamik Acknowledgements
von Dirac. Z. Phys. 53, 157–165 (1929). We are grateful to A. C. Neto, V. Fal’ko, P. Guinea and D. Khveshchenko for illuminating discussions.
2. Greiner, W., Mueller, B. & Rafelski, J. Quantum Electrodynamics of Strong Fields (Springer, This work was supported by EPSRC (UK) and FOM (Netherlands).
Berlin, 1985). Correspondence and requests for materials should be addressed to M.I.K.
3. Grib, A. A., Mamayev, S. G. & Mostepanenko, V. M. Vacuum Effects in Strong Fields (Friedmann, Supplementary Information accompanies this paper on www.nature.com/naturephysics.
St-Petersburg, 1994).
4. Su, R. K., Siu, G. C. & Chou, X. Barrier penetration and Klein paradox. J. Phys. A 26,
1001–1005 (1993). Competing financial interests
5. Dombey, N. & Calogeracos, A. Seventy years of the Klein paradox. Phys. Rep. 315, 41–58 (1999). The authors declare that they have no competing financial interests.
6. Calogeracos, A. & Dombey, N. History and physics of the Klein paradox. Contemp. Phys. 40,
313–321 (1999). Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

nature physics VOL 2 SEPTEMBER 2006 www.nature.com/naturephysics 625

©2006 Nature Publishing Group