Imprint of topological degeneracy in quasi-one-dimensional fractional quantum Hall

states
Eran Sagi, Yuval Oreg, and Ady Stern
Department of Condensed Matter Physics, Weizmann Institute of Science, Rehovot, Israel 76100

Bertrand I. Halperin

arXiv:1502.01665v1 [cond-mat.mes-hall] 5 Feb 2015

Department of Physics, Harvard University, Cambridge MA 02138, USA

We consider an annular superconductor-insulator-superconductor Josephson-junction, with the
insulator being a double layer of electron and holes at Abelian fractional quantum Hall states of
identical fillings. When the two superconductors gap out the edge modes, the system has a topological ground state degeneracy in the thermodynamic limit akin to the fractional quantum Hall
degeneracy on a torus. In the quasi-one-dimensional limit, where the width of the insulator becomes
small, the ground state energies are split. We discuss several implications of the topological degeneracy that survive the crossover to the quasi-one-dimensional limit. In particular, the Josephson effect
shows a 2d-periodicity, where d is the ground state degeneracy in the 2 dimensional limit. We find
that at special values of the relative phase between the two superconductors there are protected
crossing points in which the degeneracy is not completely lifted. These features occur also if the
insulator is a time-reversal-invariant fractional topological insulator. We describe the latter using
a construction based on coupled wires. Furthermore, when the superconductors are replaced by
systems with an appropriate magnetic order that gap the edges via a spin-flipping backscattering,
the Josephson effect is replaced by a spin Josephson effect.
PACS numbers: 73.43.-f,73.21.Hb,03.65.Vf, 74.78.Fk

I.

INTRODUCTION

One of the hallmarks of the fractional quantum Hall

effect (FQHE) is that if the two-dimensional electron system resides on a manifold with a nontrivial topology, it
will have a ground state degeneracy which depends on the
topology [1]. For a fractional quantum Hall state on an
infinite torus, the degeneracy of the ground state equals
the number of topologically distinct fractionalized quasiparticles allowed in that state. Since this degeneracy
is topological, it does not originate from any symmetry,
and in particular does not require the absence of disorder. Furthermore, no local measurement may distinguish
between the degenerate ground states.
When the torus is of large but finite size, the degeneracy is split, but the splitting is exponentially small in
L, where L = min {Lx , Ly } and Lx , Ly are the two circumferences of the torus. In the thin torus regime, where
one circumference of the torus is infinite and the other is
smaller or comparable to the magnetic length, the fractional quantum Hall state crosses over into a charge density wave (CDW), and the degenerate ground states correspond to different possible phases of the CDW [24]. In
that regime a local impurity may pin the charge density
wave and lift the degeneracy between the ground states.
Equivalently, a local measurement is able to identify the
phase of the CDW, and hence the ground state.
In this work we consider two systems that are topologically equivalent to a torus, and - unlike the torus are within experimental reach. The first is that of an
annular shaped electron-hole double-layer in which the

electron and hole densities are equal, and are both tuned
to the same FQHE state (see Fig. (1a)). In the absence of any coupling between the layers, both the interior edge and the exterior edge of the annulus carry pairs
of counter-propagating edge modes of the electrons and
the holes. These pairs may be gapped by means of interlayer back-scattering, resulting in a fully gapped system
with the effective topology of the torus. In fact, this system is richer than a seamless torus, since the interior and
exterior edges may be gapped in different ways. In particular, gapping the counter-propagating edge modes by
coupling them to a superconductor may have interesting
consequences. Some of these consequences are central to
the current paper.
The second realization we consider is that of a two
dimensional time-reversal-invariant fractional topological
insulator [5]. To be concrete, we assume that it is constructed of wires subjected to spin-orbit coupling and
electron-electron interaction (see Fig. (1b)). In this realization, electrons of spin-up form a FQHE state of filling factor , and electrons of spin-down form a FQHE
of filling factor . Similar to the particle-hole case,
the edges carry pairs of counter-propagating edge modes
with opposite spins that may be gapped in different ways.
Remarkably, when the edge modes are gapped by being
coupled to superconductors, the system is invariant under time-reversal, yet topologically equivalent to a FQHE
torus.
We use these realizations of a toroidal geometry and
their inter-relations to investigate the transition of a frac-

(a)

(b)

(c)

FIG. 1. (a) The first realization we consider is that of an electron annulus (blue) and a hole annulus (red) under the action of
a uniform magnetic field. It is evident that coupling the annulis edges forms the topology of a torus. The second realization
we suggest is that of a fractional topological insulator. Fig. (b) shows a possible model for a fractional topological insulator.
We have an array of N wires, with a strong spin-orbit coupling. The spin orbit coupling is linear with the wire index n. The
similarity of the resulting spectrum (see Fig. (3a) below) to the one corresponding to the wires construction of quantum Hall
states suggests an equivalence to two quantum Hall annuli subjected to opposite magnetic fields (each annulus corresponds to a
specific spin). The use of the wires construction enables us to include interaction effects using a bosonized Tomonaga-Luttinger
liquid theory for the description of the wires. (c) The edge modes of the two above models can be gapped out by proximity
coupling to superconductors. In the case of a thin (quasi-1D) system, the phase difference between the inner and the outer
superconductors leads to a Josephson effect mediated by tunneling across the region of a fractional quantum Hall double layer
or a fractional topological insulator. The spectrum as a function of the phase difference is depicted in Fig. (2) below. The
edge modes can also be gapped using proximity to magnets, in which case one can measure the spin-Josephson effect.

tional quantum Hall system from the thermodynamic

two-dimensional to the quasi-one dimensional regime of a
few wires. In particular, we find signatures of the topological ground state degeneracy of the two-dimensional
(2D) limit (akin to that of fractional quantum Hall states
on a torus) that survive the transition to the quasi onedimensional (1D) regime and propose experiments in
which these signatures may be probed. For example, for
an Abelian fractional quantum Hall state, we find a 2dperiodic Josephson effect, where d is the degeneracy in
the 2D thermodynamic limit.
The structure of the paper is as follows: in Sec. II we
summarize the physical ideas and the main results of the
paper. In Sec. III we define the systems in more detail
and identify the topological degeneracy in the thermodynamic limit. In Sec. IV, we discuss the quasi onedimensional regime, and point out observable signatures
of the topological degeneracy in that regime. Our discussions in these sections focus on the = 1/3 case. In Sec.
V we discuss how the results of the previous sections are
generalized to other Abelian QHE states.

II.

THE MAIN RESULTS AND THE PHYSICAL

PICTURE
A.

The systems considered

The electron-hole double-layer system is conceptually

simple to visualize (see Fig. (1a)). We consider an
electron-hole double-layer shaped as an annulus with

equal densities of electrons and holes, and a magnetic

field that forms FQHE states of in the two layers.
The system breaks time reversal symmetry, but its low
energy physics satisfies a particle-hole symmetry. For
most of our discussion we focus on the case = 1/3. In
that case each edge carries a pair of counter-propagating
= 1/3 edge modes. The edge modes may be gapped
by means of normal back-scattering (possibly involving
spin-flip, induced by a magnet) or by means of coupling
to a superconductor. In line with common notation, we
refer to these two ways as F and S respectively.
To model the fractional topological insulator we consider an array of N coupled quantum wires of length Lx ,
each satisfying periodic boundary conditions (Fig. (1b)).
The wires are subjected to a Rashba spin-orbit coupling,
and we consider a case in which the spin-orbit coupling
constant in the nth wire is proportional to 2n 1 (similar to the model considered by Ref. [6]). Effectively, this
form of spin-orbit coupling subjects electrons of opposite
spins to opposite magnetic fields. While this particular
coupled-wire model of a time reversal invariant topological insulator does not naturally allow for the regime of
a large N , other realizations, such as those proposed in
Ref. [6, 7], allow for such a regime. These realizations
require more wires in a unit cell, and are therefore more
complicated that the one considered here. Most of the
results of our analysis are independent of the specific realization of the fractional topological insulator, and we
present the analysis for the realization that is simplest to
consider.
For non-interacting electrons, the spectrum of the array we consider takes the form shown in Fig. (3a). Single-

electron tunneling processes (which conserve spin) gap

out the spectrum in all but the first and last wires, which
carry helical modes (Fig. (3b)). If the chemical potential is tuned to this gap, then in the limit of large N the
system is a topological insulator (TI), and therefore the
gapless edge modes are protected by time-reversal symmetry and charge conservation [8]. This construction is
then equivalent to two electron QH annuli with opposite
magnetic fields.
The edge modes may be gapped by coupling the two
external wires (n = 1 and n = N ) to a superconductor or
to a system with appropriate magnetic order. A Zeeman
field that is not collinear with the spin-orbit coupling
direction is necessary to couple the different spin directions. Moreover, in our coupled-wires model the spin-up
and the spin-down electrons at the n = N edge have different Fermi-momenta, so that edge would not be gapped
by a simple ferromagnet. In order to conserve momentum one would need to introduce a periodic potential
that could modulate the coupling to the ferromagnet at
the appropriate wave vector, or one would need to use
a spiral magnet with the appropriate pitch. In more sophisticated wire models, such as those discussed by Refs.
[6, 7], or in actual realizations of topological insulators,
the two edge modes can have the same Fermi momenta,
so a simple ferromagnet can be used.
In order to construct a fractional topological insulator,
we first tune the chemical potential such that the density is reduced by a factor of three, to = 1/3. For an
array of wires in a magnetic field and spinless electrons,
Kane et al. [9] have introduced an interaction that leads
to a ground state of a FQHE = 1/3. Furthermore,
they argued that there is a range of interactions that will
flow to the topological phase described by this state [9
11]. Here we show that the same interaction, if operative
between electrons of the same spin only, leads to a formation of a fractional topological insulator, i.e., to the
spin-up electrons forming a = 1/3 state and the spindown electrons forming a = 1/3 state. Note that the
same type of interaction terms were used by several authors to construct various 2D fractional topological states
[6, 7, 12], and 1D fractional states [11, 1315].
Our analysis is based on bosonization of the wires degrees of freedom, and a transformation to a set of composite chiral fields, that may be interpreted as describing
fermions at filling = 1. In terms of the composite fields,
one can repeat the process which led to a gapping of the
non-interacting case either by normal or by superconducting mechanisms. In terms of the original electrons,
these mechanisms involve multi-electron processes, which
either conserve the number of electrons or change it by a
Cooper pair.
Both the electron-hole double-layer and spin-orbit wire
system have counter propagating edge modes. They
are distinct, however, in a few technical details. An
electron-hole double layer system has been realized before in several materials, such as GaAs quantum wells

and graphene. The requirements we have here - no bulk

tunneling, sample quality that is sufficient for the observation of the fractional quantum Hall effect, and a good
coupling to a superconductor or a magnet - are not easy
to realize, but are not far from experimental reach [16
18]. In addition, we assume that the two layers are far
enough such that inter-layer interactions do not play an
important role, but close compared to the superconducting coherence length to enable pairing on the edges.
The array of wires we describe can in principle be
formed using semi-conducting wires such as InAs and
InSb [1921], where variable Rashba spin-orbit coupling
could be achieved by applying different voltages to gates
above the wires. We stress that the wires construction is nothing but a specific example of a fractional
topological insulator, and that any fractional topological insulator is expected to present the effects we discuss. Two-dimensional topological insulators were conclusively observed [2228], and more recently proximity
effects to a superconductor were demonstrated on their
edges [2931]. However, fractionalization effects due to
strong electron-electron interaction were not observed yet
in these systems and are less founded theoretically.

B.

Ground state degeneracy and its fate in the

transition to one dimension

In Sec. III we investigate the topological degeneracy of

the ground state in the 2D thermodynamic limit. Using
general arguments, we find that the degeneracy depends
on the gapping mechanism of the edges: when both edges
are gapped by the same mechanism, be it proximity coupling to a superconductor or to a magnet, the topological
degeneracy is three, as expected. However, if one edge is
gapped using a superconductor and the other is gapped
using a magnet the ground state of the system is not
degenerate.
Physically, the degeneracy is most simply understood
in terms of the charge on the edge modes. For an annular
geometry there are two edges, in the interior and the exterior of the annulus, and therefore four edge modes with
four charges, q1 , q2 , q3 , and q4 (here we use the subscript
1,2 to denote the two counter-propagating edge modes
on the interior edge, and 3,4 to denote the modes on the
exterior edge. Edges 1 and 4 belong to one layer (or
one spin direction) and edges 2 and 3 belong to the other
layer (other spin direction); see Fig. (1a)). It will be useful below to distinguish between the integer part of qi ,
which we denote by ni , and the fractional part denoted
by fi , to which we assign the values fi = 1/3, 0, 1/3,
such that qi = ni + fi .
When a pair of counter-propagating edge modes, say
with charges q1 , q2 , is gapped by normal back-scattering
of single electrons, their total charge q1 + q2 is conserved.
Since there is an energy cost associated with the total
charge, it assumes a fixed value for all ground states.
(The tunneling between the edges gaps the system and

makes it incompressible, leading to an energy cost associated with a change of the total charge.) For simplicity,
we fix this value to be zero, making q1 = q2 . A strong
back-scattering term makes n1 n2 strongly fluctuating
but leaves the fractional part f1 = f2 fixed. As a consequence, there are three topological sectors of states that
are not coupled by electron tunneling, characterized by
f1 being 0, 1/3 or -1/3.
Since each of the layers (in the double-layer system)
or each spin direction (in the spin-orbit-coupling system)
must have an integer number of electrons, the sums q1 +q4
and q2 +q3 must both be integers. This condition couples
the fractional parts of the charges on all edges. Combining all constraints, we find that when both edges are
gapped by a normal backscattering, the following conditions should be fulfilled
f1 = f2 ,

f3 = f4 ,

(1)

f1 = f4 ,

f2 = f3 .

(2)

There are three solutions for these equations describing

l
three ground states, with fl = (1) p3 , where p may
take the values 0, 1, 1 and l = 1, 2, 3, and 4. When
both edges are gapped by a superconductor, f2 and f4
change sign in Eq. (1) and the fractional parts satisfy
f1 = f2 = f3 = f4 = p/3, Finally, when one edge
is gapped by a superconductor and the other by normal
back-scattering, only one of the two equations labeled
(1) change sign and the only possible solution is fl = 0
so that all qs must be integers, and the ground state is
unique.
Formally, the degeneracy of the ground state may be
shown by an explicit construction of two unitary operators, Ux and Uy , that commute with the low-energy effective Hamiltonian and satisfy the operator relation
Ux Uy = Uy Ux e

2
3 i

(3)

The existence of a matrix representation of this relation,

acting within the ground state manifold, requires a degenerate subspace of minimal dimension 3.
We construct such operators for the electron-hole system under the assumption that the only active degrees of
freedom are those of the edge, and for the coupled wire
system when we confine ourselves to an effective Hamiltonian. For both cases, one of these operators, say Ux ,
measures the fl s and the other operator, Uy , changes the
fl s by 1/3 (the sign depends on l and on the type of
gapping mechanism). We choose to work with a representation of Ux , Uy in which both operators, projected to
subspace of ground states, are independent of position.
Even when Lx is infinite, a finite Ly splits the degeneracy. The source of lifting of the degeneracy is tunneling
of quasi-particles between the two edges of the annulus,
i.e., tunneling of quasi-particles from the first to the last
wire. More precisely, we find that as long as the bulk gap

does not close, the only term that may be added to the
low-energy Hamiltonian is of the form
Uy + Uy

(4)

This term is generated by high orders of perturbation

theory that lead to a transfer of quasi-particles between
edges. The amplitude decays exponentially with the
width of the system. For the wires realization this translates to an exponential decay with N , the number of
wires. Other factors that determine the magnitude and
phase of are elaborated on in the next subsection.
If Lx is also finite, there will be additional terms in the
Hamiltonian proportional to Ux and Ux , with coefficients
that fall off exponentially in Lx . The physical explanation of these terms is that when Lx is finite, root-meansquare fluctuations in the total charge in an edge mode
1/2
are not infinite, but are proportional to Lx . This leads
to energy differences between states with different values
of the fractional charge fl that decrease exponentially
with increasing Lx

C.

Remnants of the degeneracy in the quasi-one

dimensional regime

The topological degeneracy is lifted in the transition

from a two-dimensional system to a quasi-one dimensional one, but it leaves behind an imprint which can
in principle be measured. This is seen when we add another parameter to the Hamiltonian. For a torus, this
parameter may be the flux within the torus. For the systems we consider here, when gapped by one superconductor at the interior edge and one superconductor at
the exterior edge, this parameter may be the phase difference between the two superconductors. In this case
the fractional quantum Hall torus forms the insulator
in a superconductor-insulator-superconductor Josephson
junction.
The dependence of the spectrum on these parameters
is encoded in the amplitude of Eq. (4). In particular,
since the tunneling charge is 2/3 of an electron charge,
which is 1/3 of a Cooper pair, we find that the tunneling amplitude at the point x along the junction is proportional to the phase factor ei(x)/3 , where (x) is the
phase difference between the two superconductors at the
point x. For the fractional topological insulator, no magnetic flux is enclosed between the superconductors, and
the equilibrium phase difference does not depend on x.
In contrast, for the electron-hole quantum Hall realization the magnetic flux threading the electron-hole double
layer makes (x) vary linearly with x, such that the phase
of the tunneling amplitude winds as a function of the position of the tunneling. The amplitude of Eq. (4) is
an integral of contributions from all points at which the

FIG. 2. The spectrum of the three low energy states as a

function of the phase difference between the two superconductors (see text for elaboration). The amplitude of oscillations falls exponentially with the number of wires N . For a
finite N , each eigenstate has a periodicity of 6. At the special points = n the spectrum remains 2-fold degenerate.
If the system is of finite length Lx , the degeneracy at these
points is lifted by a term that is exponentially small in Lx .

superconductors are tunnel-coupled,

Z
= dxT (x)

(5)

where T (x) is the local tunneling amplitude.

When the superconductors are tunnel-coupled only at
a single point (say x = 0), such that T (x) (x), the
spectrum of the three ground states as a function of ,
which is now the argument of T (x = 0) can be written
in the explicit form


2
,
(6)
E = 2t0 cos
3
where = 0, 1, 1 enumerates the ground states. This
is shown in Fig. (2).
While the amplitude t0 is exponentially small in the
width Ly , or in the number of wires N , we find that
the spectrum as a function of the phase difference across
the junction has points of avoided crossing in which the
scale of the splitting between the two crossing states is
proportional to eLx /x , i.e., is exponentially small in
Lx (here x is a characteristic scale which depends on
the microscopics). Thus, in the quasi-one-dimensional
regime, where Ly or N are small but Lx is infinite, the
three states are split, but cross one another at particular
values of .
Remarkably, this crossing cannot be lifted by any perturbation that does not close the gap between the three
degeneracy-split ground states and the rest of the spec-

trum. This lack of coupling between these states result from the macroscopically different Josephson current (from the inner edge to the outer edge) that they
carry. The Josephson junction formed between the two
superconductors will show a 6- periodic DC Josephson
effect for as long as the time variation of the phase is slow
compared to the bulk energy gap, but fast compared to
a time scale that grows as eLx /x . This Josephson current distinguishes between the three ground states. This
current oscillates as a function of the position of the tunneling point for an electron-hole quantum Hall system
and is position-independent for the fractional topological
insulator.
When tunneling between edges takes place in more
than one point, T (x) in (43) is non-zero at all these
points, and has to be integrated. A particularly interesting case is that of a uniform junction. In that case T (x)
and the Josephson current are constant for the fractional
topological insulator, while in the electron-hole doublelayer the phase of T (x) winds an integer number of times
due to the magnetic flux between the superconductors,
and the Josephson current averages to zero.
A magnetic coupling between the electron and hole
layers, or between electrons of the two spin directions
may lead to a (fractional) spin Josephson effect, in
which spin current takes the place of charge current in
the Josephson effect [3234]. In this case, assuming that
the spin up and down electrons are polarized in the z
direction, coupling between the edge modes occurs by a
magnet that exerts a Zeeman field in the xy plane. The
role of the phase difference in the superconducting case is
played here by the relative angle between the magnetization at the interior and exterior edge, but an interesting
switch between the two systems we consider takes place.
In the electron-hole quantum Hall case the direction of
the magnetization is uniform along the edges and a uniform and opposite electric current flows in the two layers.
For the fractional topological insulators the edges are
gapped only when for one of the edges the direction of
the magnetization in the x y plane winds as a function
of position. As a consequence, in our coupled-wire model
the spin current oscillates an integer number of oscillations along the junction, and thus averages to zero.
Our discussion may be extended beyond the case of
= 1/3. For Abelian states, we find that the periodicity
of the Josephshon effect is 2/e , where e is the smallest
fractional charge allowed in the state. In any Abelian
state, this is also 2 times the degeneracy of the ground
state in the thermodynamical limit.

III.

GROUND STATE DEGENERACY IN THE

THERMODYNAMIC 2D LIMIT

In this section we derive in detail the degeneracy of

the ground state in the thermodynamic two-dimensional
limit of the two systems we consider.

6
energetically favorable to form singlets, such that

A.

q1 = q2 ,
q3 = q4 .

Description in term of edge modes only

The systems we consider have two edges, each of which

carrying a pair of counter-propagating edge modes.
In the absence of coupling between the layers, the
bosonic Hamiltonian of the edges is composed of the kinetic term

H0 =

v
2

Z
dx

(x l ) .

(7)

(11)

Notice that if Eq. (11) is not satisfied, the edge carries

a non-zero spin which cannot be screened by the superconductor. This configuration is therefore energetically
costly.
In the case where both edges are gapped by normal
back-scattering processes, which we refer to as the FF
case, it is energetically favorable to preserve total charge
neutrality because an insulating magnet cannot screen
charge. This gives us the conditions

l=1,2,3,4

Here we assumed all edge velocities to be the same

and neglected small-momentum interaction between the
edges, for simplicity.
The fields i satisfy the commutation relation
1
1
[l (xl ), j (xj )] = i (1)l lj sign(xl xj )+i sign(lj).
3
9
(8)
Coupling between the edge modes has the form
Z
H1 = dx cos 3 (l j ) ,
(9)
where l, j = 1, 2 for the interior edge and l, j = 3, 4 for
the exterior edge. The plus sign refers to superconducting
coupling and the minus sign to normal back-scattering.
The edge is gapped when the coupling constant is large,
which we assume to be the case.
The charge on the lth edge modes
R is related to the
1
dxx l (x), where
winding of l , namely ql = (1)l 2
ql is the charge in units of the electron charge. For uncoupled edge modes, the charges ql are quantized in units of
the quasi-particle charge, 1/3. When two edge-modes are
coupled through a normal or superconducting coupling,
the charge on each edge heavily fluctuates. However,
due to the fact that only whole electrons may be transferred between edge modes on different layers, or between
edge modes and an adjacent superconductor, the operators ei2ql commute with both parts of the Hamiltonian
Eqs. (7) and (9). We therefore characterize the different
states according to these operators, i.e., according to the
fractional part of the charge on the various edges. The
fact that the total charge on each layer is an integer gives
the two general constraints
exp [i2 (q1 + q4 )] = 1,
exp [i2 (q2 + q3 )] = 1,

(10)

regardless of the mechanism for coupling the edges. Two

other relations come from energy considerations, which
depend on the gapping mechanism. For the case where
the two edges are gapped using a superconductor it is

q1 = q2 ,
q3 = q4 .

(12)

Altogether, then, for the SS and FF gapping mechanisms, there are three possible values for ei2q1 , namely
1, ei2/3 , ei4/3 , and the eigenvalue of this operator fixes
the values of all operators ei2ql (for l = 2, 3, 4). These
operators are of course equal to the ei2fl introduced
above. In fact, the operators ei2ql may all serve as the
unitary operators Ux from Eq. (3). To establish a ground
state degeneracy, we need to find an operator that commutes with the Hamiltonian and varies Ux . This operator
is the one that transfers a charge of 1/3 in each layer (for
the SS case), or charges of 31 , 13 (for the FF case) from
the interior to the exterior. For example, if we choose
Ux = e2iq1 then,
Uy = exp [i (1 2 3 4 )] .

(13)

Here the upper sign refers to superconducting coupling

and the lower sign to coupling to a magnet. The fields
i in (13) are all to be evaluated at the same point x.
It is easy to see that this assignment of Ux , Uy satisfies Eq. (3), thus establishing the ground state degeneracy of the Hamiltonian in Eqs. (7) and (9) for the
cases of SS and FF gapping mechanisms. In the case
where the two edges are gapped using different mechanisms (FS or SF), the only solution is the one where
ei2ql =1 (for l = 1, 2, 3, 4), and the ground state is therefore non-degenerate.
For a finite system the three-fold degeneracy is split.
In particular, in the quasi-1D regime in which Lx is infinite and Ly is finite, the splitting is a consequence of
tunnel coupling between the interior and the exterior.
This regime will be explored below.
Before doing that, however, we introduce the coupled
wires system and study its ground state degeneracy directly.

B.

The coupled wires construction for a Fractional

Topological Insulator

In this Section we explain how a fractional topological

insulator may be constructed from a set of coupled wires,
as a result of a combination of spin-orbit coupling and
electron-electron interaction. We start with the case of
non-interacting electrons, in which case a 2D topological
insulator is formed, and then introduce interactions that
lead to the fractionalized phase.

1.

The integer case - a non-interacting quantum spin Hall

state

We consider an array of N quantum wires, with a

Rashba spin-orbit coupling (see Fig. (1b)). Each wire
is of length Lx and has periodic boundary conditions.
We tune the Rashba electric field (which we set to be in
the y direction, for simplicity) such that the spin-orbit
coupling of wire number n is linear with n. The resulting
term in the Hamiltonian takes the form
Hso,n = (2n 1)ukx z ,

(kx + (2n 1)kso z )

,
2m

(15)

kF0
,
kso

(16)

where kF0 is the Fermi momentum without a spin-orbit

coupling (see Fig. (3a)).
In the integer case, = 1, the chemical potential is
tuned to the crossing points of two adjacent parabolas.
We linearize the spectrum around the Fermi points,
and use the usual bosonization technique to define two
R/L
chiral bosonic fields n, , where n is the wire index, is
the spin index, and R (L) represents right (left) movers.
In terms of these bosonic fields, the fermion operators
take the form
R/L

R/L
n,
ei(n,

R/L
+kn,
x)

kn,
= ((2n 1)kso + kF0 )

is the appropriate Fermi-momenta in the absence of interactions and tunneling between the wires, with = 1
(1) corresponding to spin up (down), and = 1 (1)
corresponding to right (left) movers. The chiral fields
satisfy the commutation relations

i


0
0
0
n (x), n0 0 (x0 ) = i,0 ,0 n,n0 sign(x x0 ) + isign(n n0 ) + n,n0 y, + ,0 y, .

Eq. (18) guarantees that the fermion fields defined in Eq.

(17) satisfy Fermi-statistics.

(18)

type
Ht = t

Once we linearize the spectrum, it becomes convenient to present it diagrammatically by plotting only the
Fermi-momenta as a function of the wire index. Fig. (4)
shows the diagram corresponding to = 1 , where a right
(left) mover is represented by the symbol  ().

(17)

where

(14)

where z is the spin in the z direction, and u is the spinorbit coupling. The spectrum of wire number n is therefore
En (k) =

u
where m is the effective mass, and kso = m
. The energy
of the different wires as a function of kx is shown in Fig.
(3a).
The similarity of the spectrum to the starting point of
the wires construction of the QHE [9, 10, 35] is evident.
This system is then analogous to two annuli of electrons
of opposite spins subjected to opposite magnetic fields or
to the electron-hole double-layer we discussed above (see
Fig. (1a)).
Following the analogy with the wires construction of
the QHE, we define the filling factor as

N
1 Z
X



L
R
dx n+1,
n,
+ h.c. =

n=1

N 1 Z
kF0 X

Ht = t

R
dx cos L
n+1, n, ,

n=1
N
1 Z
X



R
L
dx n+1,
n,
+ h.c. =

n=1

One sees that single electron tunneling operators of the

N 1 Z
kF0 X

n=1

L
dx cos R
n+1, n, ,

(19)

(a)

(b)

FIG. 3. (a) The spectrum of a system consisting of three wires (see Fig. (1b)) with non-interacting electrons subjected to
spin orbit coupling whose magnitude depends on the wire index according to Eq. (14), when tunneling between the wires is
switched off. The spectra in blue, red, and green correspond to wires number 1,2, and 3. Solid lines correspond to spin-down,
and dashed lines correspond to spin-up. (b) The resulting spectrum when a weak spin-conserving tunneling amplitude is
switched on between the wires. The bulk is now gapped, with helical modes localized on the edges.

are allowed by momentum conservation (these operators are represented by the arrows in Fig. (4)). Noting that these operators commute with one another, the
fields within the cosines may be pinned, and therefore
the bulk is gapped. These terms, however, leave 4 gapL
L
R
less modes on wires 1 and N : R
1, , 1, , N, , N, . In
fact, the above model is a topological insulator, and the
gapless helical modes are the corresponding edge modes,
protected by time-reversal symmetry and charge conservation. Although our model also has a conservation of Sz ,
this is not actually necessary to preserve the gapless edge
modes. To completely gap out the spectrum, we have to
gap out the two edges separately. This can be done using
two mechanisms: proximity coupling of wire 1 and N to
a superconductor which breaks charge conservation, or
to a magnet which breaks time-reversal symmetry. The
terms in the Hamiltonian that correspond to these cases
are
Z

H1S = 1
H1F = B1

S
HN
= N
F
HN
= BN

Z
Z

L
dx cos R
1, + 1, + 1 ,


L
dx cos R
1, 1, + 1 ,

i.e., the Zeeman field acting on the N th wire must rotate

in the x y plane at a period of 2/(kso N ). This field
then breaks translational invariance.

2.

The fractional case - a Fractional Topological Insulator

We now consider the case = 1/3, depicted diagrammatically in Fig. (5). Single electron tunneling processes
of the type we considered above do not conserve momentum (see Fig. (5)) for this filling factor, and one has to
consider multi-electron processes in order to gap out the
bulk. The problem is simplified if one defines new chiral
fermion fields in each wire according to the transformation

2 

R/L
R/L
R/L
R/L
L/R
= n,
n,
n,
ei(pn, x+n, ) ,
(21)
with
R/L
L/R
n,
= 2R/L
n, n, ,

R
dx cos L
N, + N, + N ,

R
dx cos L
N, N, + N + 4kso N x .
(20)

The phases 1 , N are the phases of the superconducting

order parameter of the superconductors that couple to

the wires 1, N respectively. The phases 1 , N are the

angles of the Zeeman fields (which lie in the x y plane)
coupling to the wires 1, N respectively, with respect to
the x-axis. As the last equation shows, for the magnetic
field coupled to the nth wire to allow for a momentumconserving back-scattering, we must have N = kso N x,

R/L
L/R
pR/L
n, = 2kn, kn, .

(22)

Strictly speaking, the operators in (21) should operate at

separated yet close points in space, due to the fermionic
R/L
nature of n, .
It is simple to check that

i


0
0
0

n
(x), n0 0 (x0 ) = 3i,0 ,0 n,n0 sign(x x0 ) + isign(n n0 ) + n,n0 y, + 3,0 y, .

(23)

FIG. 4. A diagrammatic representation of the spectrum in the case = 1. Once we linearize the spectrum around the Fermipoints, it becomes convenient to plot only the Fermi-momenta as a function of the wire index (n). The symbol  () represents
a right (left) mover. Blue (red) symbols represent the spin-down (spin-up) component. One can observe that single electron
spin-conserving tunneling operators conserve momentum, and can therefore easily gap out the bulk in this case.

Eq. (23) implies that satisfies Fermi statistics. In

addition, if one draws the diagram that corresponds to
the ps, the effective Fermi-momenta of the fields, one
gets the same diagram as in the = 1 case (Fig. (4)).
The linear transformation defined in Eq. (22) can therefore be interpreted as a mapping from = 31 for the
The mapping
electrons, to = 1 for the fermions .
from = 1/3 to = 1 suggests a relation between the
local transformation defined in Eq. (22) and the ChernSimons transformation that attaches two flux quanta to
each electron, making it a composite fermion. This relation will be explored in a future work [36]. Single-
tunneling operators conserve momentum, and one can
repeat the process that led to a gapped spectrum in the
integer case. First, we switch on single- tunneling operators of the form
t = t
H

N
1 Z
X



L
R
dx n+1,
n,
+ h.c. =

n=1

t
4

kF0

3 NX
1 Z

L
R
dx cos n+1,
n,
,

n=1

t = t
H

N
1 Z
X



R
L
dx n+1,
n,
+ h.c. =

weak coupling limit. However, they may be made relevant if one introduces strong repulsive interactions [911],
or a sufficiently strong t.
For N wires, Eqs. (24) introduces 2N 2 tunneling
terms, which gap out 4N 4 modes, and leave 4 gapless
chiral -modes on the edges. Two counter-propagating
modes are at the j = 1 wire, and two are at the j =
N wire. Notice that the gapless -fields on the edges
are related to the corresponding -fields defined in Sec.
III A by = /3. Once again, these may be gapped
by proximity coupling to a superconductor or a magnet.
Operators of the type shown in Eq. (20), however, do not
commute with the operators defined in Eq. (24). The
arguments of the cosines in (20) cannot then be pinned
by Eq. (24). The lowest order terms that commute with
the operators in Eq. (24) are
Z


S =
1 dx cos R + L + 1 ,
H
1
1,
1,
Z


F
R
L

H1 = B1 dx cos 1,
1,
+ 1 ,
Z


S =
N dx cos L + R + N ,
H
N
N,
N,
Z


F
L
R

HN = BN dx cos N,
N,
+ N + 4kso N x .
(25)

n=1

t
4

kF0

3 NX
1 Z

R
L
dx cos n+1,
n,
.

(24)

n=1

While these operators are simple tunneling operators in

terms of the -fields,

they represent the multi-electron
processes described by the arrows in Fig. (5). In terms

of the -fields,
it is clear that one cannot write analogous interactions between electrons of opposite spins,
and therefore the dominating terms are those that couple electrons with the same spins. Notice that as opposed
to the integer case, these operators are irrelevant in the

Again, for the magnetic coupling to gap the edge modes

on the nth wire, it must wind in the x y plane with
a period of 2/(kso N ). The electronic density is three
times smaller than in the previous case, so on average
there is 1/3 of an electron per period. Guided by the
analogy between the above construction and the = 1/3
FQH state on a torus, we expect the ground state to have
a 3-fold degeneracy.
Using the present formalism, will be able to see how
this degeneracy is lifted as one goes from an infinite array
to the limiting case of a few wires.

10

FIG. 5. A diagrammatic representations of the fractional case = 1/3. Now, we find that only multi-electron processes can

gap out the bulk. The processes we consider are represented by colored arrows. In terms of the composite -fields,
however,
the diagram corresponding the fractional case is identical to the one corresponding to the integer case = 1 (Fig. (4)). In this
case, the complicated multi-electron processes are transformed into single- tunneling operators. The transformation from
to therefore proves very useful in analyzing the fractional case.

where
XXX
K = 1
H
2 n
0
0



,0 

0
dx x n
V,0 x n
(27)
0

, ,

3.

Ground state degeneracy in the wire construction

For simplicity, we focus first on the FF case, where the

analogy to the FQHE on a torus is explicit. In this case,
we define the idealized Hamiltonian as
K + H
t + H
t + H
F + H
F,
HI = H
1
N

PN

Uy (x) = ei 3 (
1

Ux = e

i 31

n=1

(26)

R
L
R
L
R
L
R
L
n,
+n,
n,
N,
+1,
1,
(n,
)) = ei(x) ei 31 (N,
),

RL

R
x 1,
dx
0

"N 1
X
n=1

R
n+1,

L
n,

N
1
X

#
L
n+1,

(28)
(29)

All the fields are functions of position x. The phase

(x) in Eq. (28) is given by

1
(x) =
3

is the quadratic term that contains the non-interacting

part of the Hamiltonian, and small momentum interactions (for simplicity, we consider only intra-wire small
momentum interactions). We assume that all the interwire terms become relevant and acquire an expectation
value. To investigate the properties of the ground state
manifold, we define the two unitary operators

R
n,

Notice that the second equality in Eq. (28) shows that

Uy (x) defined in terms of the wires degrees of freedom is
identical to Eq. (13) (up to a phase). The form of Uy (x)
shown in the first equality of Eq. (28) is useful because
it allows us to express Uy (x) as a product of electronic
operators:

Uy = ei(

n=1

(30)
Since all the operators in the sum are pinned by the
bulk Hamiltonian, they may be treated as classical fields,
and their value becomes x-independent in any one of the
ground states. Similarly, the combination of operators
R
L
R
L
(N,
N,
+ 1,
1,
) which appears on the right side
of Eq. (28) is pinned by the coupling to the boundary,
and becomes independent of x. Therefore, the operators
Uy (x) may be considered to be independent of x within
the manifold of ground states.

PN

n=1

L
R
L
(R
n, n, +n, n, )).

(31)

where the x-dependence of the operators is omitted for

brevity. It can be verified that
[Uy (x), Uy (x0 )] = 0

(32)

[Ux , HI ] = [Uy , HI ] = 0,

(33)

and that

11
imum as
2

so that operating Uy (x) or Ux on a ground state leaves

the system in the ground state manifold. Using Eq. (23),
it can also be checked directly that
Ux Uy (x) = Uy (x)Ux e

2
3 i

(34)

independent of x. The smallest representation of this

algebra requires 3 3 matrices [37], which shows that
the ground state of the idealized Hamiltonian (26) must
be at least 3-fold degenerate.
The operators Uy (Ux ) can be interpreted as the creation of a quasiparticle-quasihole pair, tunneling of the
quasiparticle across the y (x) direction of the torus and
annihilating the pair at the end of the process. In fact,
if we adopt this interpretation, Eq. (34) is a direct consequence of the fractional statistics of the quasiparticles
[37].
A similar analysis can be carried out for the SS case.
Ux is identical to the operator used in the FF case, but
now Uy takes the form
Uy = ei 3 (
1

PN

n=1

R
L
L
R
n,
+n,
n,
(n,
)),

The coupled wires construction of an

electron-hole double layer

In this Section we explain how one can model a quantum Hall electron-hole double layer at a fractional filling
factor = 1/3 using a set of coupled wires. Most of
the analysis is very similar to the analysis presented for
the fractional topological insulator, but some technical
differences are worth pointing out.
We examine a system with two layers, each containing
an array of wires. In one layer, the electron layer, we
tune the system such that only states near the bottom
of the electronic band are filled. In this case, we can approximate the spectra of the various wires as parabolas.
If we add a constant magnetic field B perpendicular to
the layers, and use the Landau gauge to write the electromagnetic potential as A = By
x, the entire band
structure of wire number n will be shifted by an amount
2k n, where k is defined as k = eBa
2~ . The energy of
wire number n is therefore written in the form (if we
choose the position of wire number 1 to be at y = a/2)
2

En (k) =

(kx ( 2n 1 )k )
+ Ue ,
2m

(kx ( 2n 1 )k )
+ Uh .
2m

(37)

In the above, we assumed that the effective masses of the

electron and the hole layers have the same magnitude
and opposite signs. We assume that Uh > Ue , and tune
h
. Defining  =
the chemical potential to be = Ue +U
2
Uh Ue
,
we
get
the
spectra
2
"
#
2
(kx ( 2n 1 )k )
En (k) =
 ,
(38)
2m
where = 1(1) for the electron (hole) layer. This way
the system has a built-in particle-hole symmetry in its
low energy Hamiltonian. Notice that as a result of the
magnetic field, the spectra of the two layers are shifted in
the same direction. This is a consequence of the common
origin of the electron and hole spectra from a Bloch band
whose shift is determined by the direction of the magnetic
field.

We define kF0 = 2m, and the filling factor is now

k0

(35)

and the entire analysis can be repeated.

C.

En (k) =

given by = kF . In the case = 1, the corresponding

spectrum is given by Fig. (6a). As before, if we apply
tunneling between adjacent wires in the same layer, we
get the gapped spectrum in Fig. (6b). Furthermore, we
see that each edge carries a pair of counter propagating
edge modes (one for each layer).
It is straightforward to generalize this to the case of
filling = 1/3, shown in Fig. (6c). To treat this case,
we follow exactly the same steps as in Sec. III B: we first
linearize the spectrum, and write the problem in terms
R/L
of the chiral bosonic degrees of freedom n, , where now
= e, h represents the layer number, and n represents
the wire index. To treat the fractional case, we define
L/R
R/L
R/L
new chiral fields n, = 2n, n, . Like before, it
can be checked that these modes behave like modes at
filling 1, so we can repeat the analysis performed in this
case.
This process leaves us with two counter propagating R
R
L
L
, 1,h
, N,e
, N,h
. These modes
modes on each edge: 1,e
can be gapped out by terms analogous to the terms in
Eq. (25):


L
R
1S =
1 cos 1,e
H
+ 1,h
+ 1 ,


L
R
1F = B
1 cos 1,e
H
1,h
+ 1 ,


S
R
L
N
N cos N,e
H
=
+ N,h
+ N + 4k N x ,


F
R
L
N
N cos N,e
H
=B
N,h
+ N .
(39)

(36)

where Ue is a constant term, and m is the effective mass.

In the hole layer the bands of the various wires are nearly
filled, such that we can expand the energy near the max-

In contrast to the case of the fractional topological insulator, here the backscattering terms conserve momentum,
i.e., do not include phases that are linear in x. Rather,
S appears not to conserve
the superconducting term H
N

12

(a)

(b)

(c)

FIG. 6. (a) The spectrum of the wires model for an electron-hole double layer at filling = 1 when all the inter-wire terms are
switched off. The spectra in blue correspond to wires in the electron layer, and the spectra in red correspond wires in the hole
layer. (b) The spectrum when tunneling between wires in the same layer is switched on. A gap is formed in the bulk, and we
get achiral edge modes. (c) The spectrum in the fractional case = 1/3.

momentum. However, the flux between the two superconductors will lead to a winding of the phase difference
between them, which can cancel the x-dependent phase
S
of HN
.
Let us first consider the situation where the bounding
superconductor wires are thin enough that there are no
vortices inside them. The energy of a superconducting
ring is minimized when , the change in the superconducting phase around the ring is equal to 2e, where
is the magnetic flux enclosed by a circle embedded at
the center of the wire. The value of is quantized
in multiples of 2, and in practice there may exist a
number of metastable states where it differs from 2e
by a finite amount and the wire carries a supercurrent
around its circumference. Let us consider a model where
there is a distance a0 between the center of the inner
most superconductor and the center of our first electronhole nanowire and a similar separation between the N th
nanowire and the outer superconductor. If the centers
of the nanowires are separated from each other by a distance a, then the flux is equal to BaLx (N 1+2(a0 /a)).
In this case, if the superconductors

 are in their ground
a0

states, we get 1 = 2 + 4 a k x + 10 and N =



0
0
0
4N 2 + 4 aa k x + N
, where 1(N
) do not depend
on x. If a0 is tuned to a0 = a/2, the oscillating phases
are eliminated from Eq. (39).
If a0 differs from a/2, it may be still possible to gap out
the edges. If the phase mismatch is small, and if coupling
to the superconductor is not too weak, then there can be
an adjustment of the electron and hole occupations in
the nanowires nearest the two edges, which allows the
phase change around the nanowires to match the phase
change in the superconductors. The energy gain due to
formation of a gap can exceed the energy cost of altering
the charge densities in the nanowires.
If the difference between a0 and a/2 is too large, then
carrier densities in the inner and outer nanowires will not
change enough to satisfy the phase matching condition.
In this case, a variation of the magnetic field of order

1/N would eliminate the x-dependence of the phases at

the cost of introducing quantum Hall quasiparticles in
the bulk of the system. For large N , the density of these
quasiparticles will be small. Presumably they will become localized and not take the system out of the quantum Hall plateau.
We note that the separation a0 can be engineered, and,
in principle can even be made negative. Consider, for example, a situation where the superconducting wire sits
above the plane of the nanowires, so that depending on
the shape of a cross-section of the wire, its center of gravity may sit inside or outside of the line of contact to the
outermost nanowire.
The situation is more complicated if the superconductors are thick enough that they contain vortices in the
presence of the applied magnetic field. If the vortices
are effectively pinned, however, it should be possible
to achieve conditions where the electron-hole system is
gapped and experiments such as Josephson current measurements can be performed.
The degeneracy of the ground states in both the SS and
FF cases may be shown by defining the two operators Ux
and Uy in exactly in the same form as we did in Sec.
III B 3 (with e and h), and following the same
analysis.

IV. MEASURABLE IMPRINT OF THE

TOPOLOGICAL DEGENERACY IN QUASI-ONE
DIMENSIONAL SYSTEMS

We now look at the quantum Hall double-layer system with = 1/3. As long as the bulk gap does not
close, in the limit of infinite Lx and infinite N (or Ly )
we expect deviations from the idealized Hamiltonian not
to couple the three ground states. When N and Ly are
finite and Lx is still infinite, coupling does occur, and the
degeneracy is lifted.
Generally, hermitian matrices operating within the
3 3 subspace of ground states of the idealized Hamiltonian may all be written as combination of nine unitary

13

matrices Oj,k
(1,1)

H =

j,k Oj,k .

(40)

(j,k)=(1,1)

where
Ojk = Uxj Uyk .

(41)

i2jk

and jk = j,k e 3 . Note that a direct consequence

of Eq. (34) is that Ux3 = Uy3 = 1 (this can most easily
be understood by recalling that the operators transport
quasiparticles across the torus. Acting three times with
each of them is equivalent to transporting an electron
around the torus, which cannot take us from one ground
state to another). However, in the limit of infinite Lx local operators cannot distinguish between states of different fractional charges, and therefore cannot contain the
operator Ux . Thus, up to an unimportant constant originating from 00 , deviations from the idealized Hamiltonian (projected to the ground state manifold) take the
form of Eq. (4):
H = Uy + Uy ,

(42)

The coefficient may be expressed as an integral,

Z
= dxT (x),
(43)
and we expect that the absolute value of the amplitude
T (x) should fall off exponentially with N , as discussed
in Section II B. One can see this explicitly in the various models we have constructed from coupled wires. For
example, in the case of a fractional topological insulator
with magnetic boundaries, the operator Uy , according to
(32 ) and (17) involves a product of factors involving four
electronic creation and annihilation operators on each of
the N wires. The bare Hamiltonian contains only fourfermion operators on a single wire, and two-fermion operators that connect adjacent wires, with an amplitude t
that we consider to be small. The operator Uy can only
be generated by higher orders of perturbation theory, in
which the microscopic tunneling amplitude t occurs at
least 2N times. In our analysis, we have assumed that interaction strengths on a single wire are comparable to the
Fermi energy EF , so we expect T to be of order |t/EF |2N
or smaller. Similar arguments apply to the other cases of
superconducting boundaries or electron-hole wires. We
also note that if the system is time-reversal invariant, we
must have T = T .
The phase of T (x) depends on the realization electron-hole quantum Hall vs. fractional topological insulator - and on the gapping mechanism - two superconductors or two magnets. We start from the case of the
fractional topological insulator gapped by two supercon-

ductors. Eqs. (25) shows that for the edges to be gapped,

the superconductors on the two edges should have uniform phases 1 , N . We choose a gauge where 1 = 0 and
denote = N to be the phase difference.
In the case of a fractional topological insulator, the
proximity gapping terms are
Z


S =
1 dx cos L + R + ,
H
1
1,
1,
Z


S
R
L

(44)
HN = N dx cos N,
+ N,
(note that these terms involve coupling to the supercon 1(N ) |1(N ) |, where
ductor, and we therefore have
1(N ) are the corresponding superconducting order parameters). We define new bosonic fields through the additional transformation

L
R
1,
= L
, R
,
(45)
1, +
1, = 1, +
2
2
and n, = n, for all the other values of n, , . If
we rewrite the Hamiltonian in terms of the new fields,
the phase is eliminated from the idealized Hamiltonian. However, this modifies the operator Uy (defined in
Eq.(28)), which now takes the form
Uy = ei 3 (
1

PN

n=1

R
L
L
R

n,
+
n,

n,
(n,
)) ei 3 .

(46)

Thus, a non-zero phase difference shifts the argument

of in Eq. (42) by 3 . In the time reversal symmetric
case is real, and we find, by diagonalizing H, that

E1 = 2Lx cos
,

3
2
,
E2 = 2Lx cos
3


+ 2
E3 = 2Lx cos
.
(47)
3
The resulting spectrum as a function of is depicted
in Fig. (2).
At = n the degeneracy is not completely lifted, as
two states remain 2-fold degenerate. These states are not
coupled by the low energy Hamiltonian (42) and the lifting of their degeneracy requires terms of j 6= 0 in (41).
Such terms distinguish between states of different edge
charges fi and originate from tunneling between the three
physically distinct minima of the potential (9). The amplitude for tunneling, and hence the splitting, is proportional to eS , with S the imaginary action corresponding
to the tunneling trajectory. Due to the integration over x
in the Hamiltonian, this action is linear in Lx , and hence
the tunneling amplitude scales as e(Lx /x ) . Neglecting
this splitting, Eq. (47) shows that all eigenstates have a
6 periodicity. A measurement of the Josephson current,
given by the derivative of the energy with respect to ,
can detect the 6-periodicity. Due to the exponentially
small splitting at the crossing points, this property can

14

be observed by changing the flux at a rate that is not

slow enough to follow this splitting.
Note that the 6-periodic component of the spectrum
is completely determined by Eq. (34). This part of
the spectrum is therefore highly insensitive to the microscopic details, and can serve as a directly measurable imprint of the topological degeneracy with only a few wires.
There will also be a contribution from ordinary Cooper
pair tunneling between the superconductors, which does
not distinguish between the ground states and has 2
periodicity. This term will alter the detailed shapes of
the three spectra but not their splitting or periodicity.
In the case where time reversal symmetry does not hold,
is not necessarily real. Consequently, the spectrum in
Eq. (47) is shifted according to + Arg (), and
the crossing points are not constrained to be at = n.
Similar results arise in the FF case for a quantum Hall
electron-hole double layer. Now, the angle is the relative orientation angle of the Zeeman fields (which lies in
the x y plane). To be precise, if we fix the Zeeman field
at wire number N to point at the x direction, and the
field at wire number 1 to have an angle relative to the
x direction, we get the proximity terms
Z


F
L
R

H1 = 1F dx cos 1,
1,
+ ,
Z


F
N F dx cos R L .
N
H
=
(48)
N,
N,
Similar to Eq. (45), we define new bosonic fields through
the transformation

R
L
, R
,
(49)
1,
= L
1, = 1,
1, +
2
2
and n, = n, for the other fields. Again, the gapping
term acting on the N th wire returns to
its original form

(with = 0), but Uy becomes Uy ei 3 . Therefore, the

spectrum as a function of is identical to the spectrum
found in the SS case.
In the other two cases, the situation is more complicated, since depends on x. For the quantum Hall
electron-hole double layer gapped by superconductors
increases linearly with x, due to the flux penetrating the
junction between the two superconductors. For the fractional topological insulator gapped by magnets, Eq. (25)
requires that N increases linearly
with x. In both cases,
R
this winding leads to = dx|t(x)|ei2nx/L+i , with n
an integer. A uniform tunneling amplitude |t(x)| then
leads to a vanishing , while non-uniformity allows for a
non-vanishing .

V.

EXTENSIONS TO OTHER ABELIAN

STATES

We have shown above that it is possible effectively realize experimentally the = 31 FQHE state on a torus,
and that by measurement of the Josephson effect in the

resulting construction we can directly measure the corresponding topological degeneracy. In this section we
extend the above results to other Abelian FQHE states.
For a FQHE state described by a M M K-matrix,
there is a ground state degeneracy of d = det K on a
torus, and d topologically distinct quasiparticles. Each
quasiparticle is a multiple of the minimally charged quasiparticle, whose charge is e = de .
Repeating the analysis we carried out in Sec. III,
we consider an electron-hole double layer system or a
fractional topological insulator, and couple the counterpropagating edge modes. Since there are now M pairs
of counter-propagating modes on each edge, we need m
scattering terms. We assume that these terms are all
mutually commuting, that they are either all chargeconserving or all superconducting, and that the M edge
modes of each layer (spin-direction) are mutually coupled. Under these assumptions, each of the four edges
is characterized by one quantum number - the fractional
part of the total charge fi (with i = 1, , 4), which may
d3
d1
take the values d1
2d , 2d , , 2d . Similar to the case
where = 1/3, the requirements of a total integer charge
for each layer or spin direction, together with the mechanism of gapping and the requirement to minimize the
energy of the edge Hamiltonians, relate all values of fi to
one another.
We work in a basis |f i where the fractional charges
fi are well defined. We define the unitary operator Uy
which transfers a single minimally charged quasiparticle, analogously to the operator defined before, such that
Uy |f i = |(f + e /e) mod(1)i. It follows that Uyl |f i =
|(f + le /e) mod(1)i, and that Uyd = 1. We therefore
have in general
Uyl

= Uydl .

(50)

Again, in the quasi-1D limit where Lx is infinite and

N is finite, Hermitian combinations of the operators Uyl
are the only operators capable of lifting the degeneracy.
The amplitude of these terms falls exponentially with N .
In order to analyze the effects of these perturbations we
consider terms of the form
(d1)/2

H =

l Uyl eil + h.c. ,

(51)

l=1

where l eN/l is a real coefficient (note that we expect terms with l > 1 to result from higher orders in eN .
More specifically, we expect l 1l ). The summation
was terminated at (d 1)/2 because of Eq. (50) and the
requirement that the Hamiltonian is hermitian. Again,
the resulting spectrum depends on the realization, the
gapping mechanism, and the uniformity of the tunneling
amplitude. This dependence is similar to the one discussed for = 1/3. For example, for uniform tunneling
between two superconductors separated by a fractional
topological insulator, a relative phase between the two

15

superconductors translates to l = ee l.
The spectrum of this Hamiltonian for the time reversal
symmetric case is then
(d1)/2

Ep = 2

X
l=1


l cos


l
( + 2p) ,
d

(52)

with p = 1 . . . d. Each eigenstate has a 2d-periodicity,

and like the = 1/3 case we find that the overall periodicity is 2 times the degeneracy of the system in the
thermodynamic limit. In addition, similar to the = 1/3
case, at the time-reversal invariant points = n, we
have degeneracy points protected by the length of the
wires. For example, at = 0, we have d1
pairs of
2
states |pi ,|d pi (p = 1, . . . d1
)
which
have
the
same
2
energy. It can easily be checked from Eq. (52) that
the same number of crossings occurs for any = n.
Hence if the spectrum is measured, the degeneracy d can
found by simply counting the number of crossing points
at = n. Note that due to the terms with l > 1, we can
have additional crossing points at 6= n. Again, if time
reversal symmetry does not hold the crossing points can
be shifted. One can still show that in the most general
case there must be at least the same number of crossing
points as the number of crossing points at = n in
the time reversal invariant case. The smallest number of
degeneracy points occurs when the functions Ep have
a single maximum and a single minimum between 0 and
2d. In that case, the energies that correspond to two different values of p must cross at two points between 0 and
2d. We therefore have 2 crossing points for each pair
p1 , p2 , The total number of degeneracy
summed

 points,
d
= d(d 1),
over all the pairs p1 , p2 is therefore 2
2
which is the number of crossing points at all the values
= n in the time reversal invariant case. Depending
on the values of l , we may have more than a single minimum and a single maximum, in which case we can get
additional crossing points.
As an example we examine the case = 2/5, which
can be characterized by the K-matrix


3 2
K=
.
(53)
2 3

The degeneracy on a torus in this case is d = 5 and

the spectrum (in the time reversal invariant case) is




1
2
Ep = 21 cos
( + 2p) + 22 cos
( + 2p) ,
5
5
(54)
with p = 1 . . . 5. If we take for example 2 /1 = 0.2, the
resulting spectrum is shown in Fig. (7).

FIG. 7.
The spectrum corresponding to = 2/5 with
2 /1 = 0.2 as a function of the relative phase difference
. The periodicity of each eigenstate is 10. At the points
= n, we find two crossing points whose splitting falls exponentially with Lx .

VI.

CONCLUSIONS

The topological degeneracy on a torus is perhaps the

defining property of a fractionalized phase, and the most
prominent signature of a topological order. As such, it
is unfortunate that for the most accessible fractionalized
phase - the Fractional Quantum Hall effect - it is impossible to to directly create a toroidal geometry, that
requires magnetic monopoles. In this work we study two
annular geometries that are topologically equivalent to
that of a torus. One geometry is based on an electronhole double layer where the electrons and the holes are
at fractional quantum Hall states of opposite filling fractions. The other is based on a fractional topological insulator at which the two spin directions of the electrons are
at fractional quantum Hall states of opposite filling fractions. Both geometries carry counter-propagating edge
modes on the interior and the exterior edges of the annuli, and these edge modes may be coupled and gapped
in two mechanisms - back-scattering and proximity coupling to superconductors.
Considering the two dimensional regime where the annuli are too wide to have a significant coupling between
the interior and the exterior edges, we established here
the topological degeneracy that characterizes each of the
geometries we consider, and their dependence on the gapping mechanism on each of the edges. Furthermore, we
used the quantum number of the fractional charge or
dipole on each of the edges to characterize the ground
states.
In the regime where the annuli are narrow such that
the interior and the exterior are coupled, the degenerate
ground states split in energy. Searching for remnants
of the topological order that survive the transition to the
quasi-one dimensional regime, we studied the dependence
of the spectrum of split ground states as a function of the

16

phase difference between the two superconductors or the

relative angle between the direction of magnetization of
the two magnets. We find that the spectrum includes
points in which the splitting is exponentially small in the
circumference of the annulus, and thus is not split when
the width becomes small.
At finite temperature there will be thermally excited
pairs of quasiparticles and quasiholes in the bulk. When
reaching the edge, these excitations carry the potential
of introducing transitions between the states that cross
at Figs. (2) and (7). The density of these quasiparticles and the resulting transition rates are expected to be
exponentially small at low temperatures.
The spectra of Figs. (2) and (7) give rise to a remarkable experimental consequence. As long as experiments
are done on timescales at which the exponentially small
transitions between states at the crossing points may be
neglected, the Josephson effects give a 2d-periodicity,
where d is degeneracy in the 2D thermodynamic limit.
Despite the fact that the degeneracy was lifted in the
quasi-1D regime, it leaves an imprint in the Josephson
effect.

ACKNOWLEDGMENTS

We are indebted to Erez Berg and Eran Sela for insightful and important conversations. This work was
supported by Microsofts Station Q, the US-Israel Binational Science Foundation, the Israeli Science Foundation (ISF), the Minerva foundation, and the European
Research Council under the European Communitys Seventh Framework Program (FP7/2007-2013)/ERC Grant
agreement No. 340210.

Appendix A: Projection of Hermitian matrices onto

the ground state manifold

In section IV we stated that hermitian matrices operating within the 3 3 subspace of ground states of the
idealized Hamiltonian may all be written as a combination of nine unitary matrices as in Eq. (40). In this appendix we prove this statement. To do so, we show that

[1] X. G. Wen and Q. Niu, Phys. Rev. B 41, 9377 (1990).

[2] E. J. Bergholtz and A. Karlhede, Phys. Rev. Lett. 94,
026802 (2005).
[3] E. J. Bergholtz and A. Karlhede, J. Stat. Mech. Theor.
Exp. 2006, L04001 (2006).
[4] A. Seidel, H. Fu, D.-H. Lee, J. M. Leinaas, and J. Moore,
Phys. Rev. Lett. 95, 266405 (2005).
[5] M. Levin and A. Stern, Phys. Rev. Lett. 103, 196803
(2009).

any operator acting on this subspace can be written as a

combination of Oj,k . It follows that in particular any hermitian operator can be written in the form shown in Eq.
i2jk
(40), if the constraint jk = j,k e 3 is imposed.
The operator Ux defined in Eq. (29) measures the
charge on the edge modes. We expect that it will have
three eigenvalues corresponding to edge charges equal to
0, 1/3, and 1/3. We denote the eigenstate with zero
charge on the edge by |0i. In addition we introduce the
notation: |1i = Uy |0i and |1i = Uy |1i = Uy2 |0i.
By definition Ux |0i = |0i and it is easy to check, using
the identity Ux Uy = ei Uy Ux , = 2/3, that:
Ux |1i = Ux Uy |0i = ei Uy Ux |0i = ei Uy |0i = ei |1i
(A1)
and similarly we find
Ux |1i = e2i |1i = ei |1i .

(A2)

The set |1i , |0i , |1i forms a complete basis for the 3 3
prosubspace of ground states so that any operator O,
jected onto this subspace, can be written in this basis
as:
X
=
O
|ji hj| O |li hl| .
(A3)
j,l=1,0,1

Since
Ux = ei |1i h1| + |0i h0| + e+i |1i h1|

(A4)

we find that (notice that cos = cos 2 for = 2/3 )

|0i h0| = (Ux + Ux 2 cos 1)/(2(1 cos ))

(A5)

with 1 = |1i h1| + |0i h0| + |1i h1| being a unit matrix
in the 3 3 subspace. All the other |ji hl| operators in
the expansion of Eq. (A3) can be obtained by multiplying
the presentation of |0i h0| in Eq. (A5) by Uy or Uy from
left or right. For example:
|1i h0| = Uy |0i h0| = Uy Ux + Uy Ux 2 cos Uy . (A6)
Hence the expansion of Eq. (40) follows.

[6] J. Klinovaja and Y. Tserkovnyak, Phy. Rev. B 90, 115426

(2014).
[7] E. Sagi and Y. Oreg, Phy. Rev. B 90, 201102(R) (2014).
[8] C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 226801
(2005).
[9] C. L. Kane, R. Mukhopadhyay, and T. C. Lubensky,
Phys. Rev. Lett. 88, 036401 (2002).
[10] J. C. Y. Teo and C. L. Kane, Phy. Rev. B 89, 085101
(2014).

17

[11] Y. Oreg, E. Sela, and A. Stern, Phy. Rev. B 89, 115402

(2014).
[12] T. Neupert, C. Chamon, C. Mudry, and R. Thomale,
Phy. Rev. B 90, 205101 (2014).
[13] J. Klinovaja and D. Loss, Phy. Rev. B 90, 045118 (2014).
[14] J. Klinovaja and D. Loss, Phy. Rev. Lett. 112, 246403
(2014).
[15] C. P. Orth, R. P. Tiwari, T. Meng, and T. L. Schmidt,
(2014), arXiv:1405.4353.
[16] R. V. Gorbachev, A. K. Geim, M. I. Katsnelson, K. S.
Novoselov, T. Tudorovskiy, I. V. Grigorieva, A. H. MacDonald, S. V. Morozov, K. Watanabe, T. Taniguchi, and
L. A. Ponomarenko, Nat. Phys. 8, 896 (2012).
[17] A. Kou, B. E. Feldman, A. J. Levin, B. I. Halperin,
K. Watanabe, T. Taniguchi, and A. Yacoby, Science
(New York, N.Y.) 345, 55 (2014).
[18] J. Eisenstein, Annu. Rev. Condens. Matter Phys. 5, 159
(2014).
[19] V. Mourik, K. Zuo, S. M. Frolov, S. R. Plissard, E. P.
A. M. Bakkers, and L. P. Kouwenhoven, Science 336,
1003 (2012).
[20] A. Das, Y. Ronen, Y. Most, Y. Oreg, M. Heiblum, and
H. Shtrikman, Nat. Phys. 8, 887 (2012).
[21] M. T. Deng, C. L. Yu, G. Y. Huang, M. Larsson,
P. Caroff, and H. Q. Xu, Nano lett. 12, 6414 (2012).
[22] B. A. Bernevig, T. L. Hughes, and S.-C. Zhang, Science
314, 1757 (2006).
[23] M. K
onig, S. Wiedmann, C. Br
une, A. Roth, H. Buhmann, L. W. Molenkamp, X.-L. Qi, and S.-C. Zhang,
Science 318, 766 (2007).
[24] C. Liu, T. L. Hughes, X.-L. Qi, K. Wang, and S.-C.
Zhang, Phys. Rev. Lett. 100, 236601 (2008).
[25] A. Roth, C. Br
une, H. Buhmann, L. W. Molenkamp,

[26]

[27]
[28]

[29]
[30]

[31]

[32]
[33]

[34]

[35]
[36]

J. Maciejko, X.-L. Qi, and S.-C. Zhang, Science 325,

294 (2009).
K. C. Nowack, E. M. Spanton, M. Baenninger, M. K
onig,
J. R. Kirtley, B. Kalisky, C. Ames, P. Leubner, C. Br
une,
H. Buhmann, L. W. Molenkamp, D. Goldhaber-Gordon,
and K. A. Moler, Nat. Mater. 12, 787 (2013).
L. Du, I. Knez, G. Sullivan, and R.-R. Du, (2013),
arXiv:1306.1925.
E. M. Spanton, K. C. Nowack, L. Du, G. Sullivan, R.R. Du, and K. A. Moler, Phys. Rev. Lett. 113, 026804
(2014).
I. Knez, R.-R. Du, and G. Sullivan, Phys. Rev. Lett.
109, 186603 (2012).
S. Hart, H. Ren, T. Wagner, P. Leubner, M. M
uhlbauer,
C. Br
une, H. Buhmann, L. W. Molenkamp, and A. Yacoby, , 20 (2013), arXiv:1312.2559.
V. S. Pribiag, A. J. A. Beukman, F. Qu, M. C. Cassidy,
C. Charpentier, W. Wegscheider, and L. P. Kouwenhoven, (2014), arXiv:1408.1701.
J. Nilsson, A. R. Akhmerov, and C. W. J. Beenakker,
Phys. Rev. Lett. 101, 120403 (2008).
L. Jiang, D. Pekker, J. Alicea, G. Refael, Y. Oreg,
A. Brataas, and F. von Oppen, Phys. Rev. B 87, 075438
(2013).
F. Pientka, L. Jiang, D. Pekker, J. Alicea, G. Refael,
Y. Oreg, and F. von Oppen, New J. Phys. 15, 115001
(2013).
S. L. Sondhi and K. Yang, Phys. Rev. B 63, 054430
(2001).
E. Sagi, Y. Oreg, and A. Stern (unpublished).

[37] C. Nayak, A. Stern, M. Freedman, and S. Das Sarma,

Rev. Mod. Phys. 80, 1083 (2008).