Majorana Zero Modes in Synthetic Dimensions

Or Golan and Eran Sela

Raymond and Beverly Sackler School of Physics and Astronomy, Tel-Aviv University, IL-69978 Tel Aviv, Israel

Kirill Shtengel
Department of Physics and Astronomy, University of California, Riverside CA 92511, USA

Recent experimental advances in the field of cold atoms led to the development of novel techniques for
producing synthetic dimensions and synthetic magnetic fields, thus greatly expanding the utility of cold atomic
arXiv:1711.11064v2 [cond-mat.quant-gas] 27 Mar 2018

systems for exploring exotic states of matter. In this paper we investigate the possibility of using experimentally
tunable interactions in such systems to mimic the physics of Majorana chains, currently a subject of intense
research. Crucially to our proposal, the interactions, which are local in space, appear non-local in the synthetic
dimension. We use this fact to induce coupling between counter-propagating edge modes in the quantum Hall
regime. For the case of attractive interactions in a system composed of two tunneling-coupled chains, we find
a gapless quasi-topological phase with a doubly-degenerate ground state. While the total number of particles
in the system is kept fixed, this phase is characterized by strong fluctuations of the pair number in each chain.
Each ground state is characterized by the parity of the total particle number in each chain, similar to Majorana
wires. However, in our system this degeneracy persists for periodic boundary conditions. For open boundary
conditions there is a small splitting of this degeneracy due to the single-particle hopping at the edges. We show
how subjecting the system to additional synthetic flux or asymmetric potentials on the two chains can be used
to control this nonlocal qubit. We propose experimental probes for testing the nonlocal nature of such a qubit
and measuring its state.

I. INTRODUCTION states “separated” in the synthetic dimension. Our proposal

builds on an earlier proposal for inducing superconducting
Physical systems exhibiting topological order, interesting proximity in the helical edges of 2D topological insulators16 ,
in their own right, have become a subject of intense attention an influential idea which led to subsequent proposals for Ma-
recently due to their potential utility for quantum information jorana zero modes in semiconductor nanowire settings17,18
processing1 . Much of the recent experimental effort has been as well as more recent fractional generalizations19–22 . Cru-
focused on one- and quasi-one-dimensional systems hosting cially, our setup is different from the usual condensed matter
Majorana zero modes2,3 , in part due the recent advances in schemes in that it utilizes a closed system with particle con-
fabricating these systems using semiconducting nanowires, servation23–30 .
chains of atoms deposited on the surface of a superconduc- The idea of additional “synthetic” dimensions is schemati-
tor and other similar systems. Meantime, a steady progress in cally illustrated in Fig. 1, with the role of an extra “dimension”
the field of cold atoms led to the creation a new experimental played by an internal atomic degree of freedom, such as nu-
toolbox4 , allowing new avenues for testing similar ideas out- clear spin. This extra dimension is intrinsically both discrete
side of the realm of condensed matter systems. Interactions and finite, nonetheless it allows one to effectively turn a phys-
between atoms in cold atomic systems can be custom-tailored ical 1D atomic chain into a 2D strip/ladder.
by coupling individual atomic states to light. At the same A key feature of the synthetic dimension approach is that in-
time, recent advances led to the dual possibility of creating teractions become non-local in the synthetic dimension5 ; see
synthetic gauge fields and endowing these systems with an Fig. 1(a,b). This opens a new possibility, which we exploit
extra synthetic dimension5 . Using this approach, two recent in our proposal: namely, it allows coupling between the de-
milestone experiments demonstrated realizations of quantum grees of freedom which are normally spatially separated in the
Hall-like states and their associated chiral edge modes in syn- usual condensed matter setting. In particular, this enables us
thetic ribbons with artificial gauge fields, one using fermions6 to create attractive interactions between counter-propagating
and the other one – bosons7 . quantum Hall edge states31 in order to induce superconduct-
In addition, this approach opens interesting new possibil- ing instabilities; see Fig. 1(c).
ities for quantum engineering of topological states, e.g. 4D In what follows, we will focus on a system consisting of two
quantum Hall states8 , or for inducing strong correlation ef- identical chains, forming two synthetic ribbons in the quan-
fects in low dimensions, e.g. in magnetic crystals9 where one tum Hall regime. Using the renormalization group analysis23 ,
could observe fractional charge pumping10,11 and probe sig- we show that this closed system forms a many-body phase
natures of chiral Laughlin-like edge states12–15 . with strong pair-tunneling coupling between the two chains.
In this paper, we discuss the possibility of realizing topo- Its ground state is doubly degenerate ground state, resembling
logical states with Majorana-like zero modes within the that of the Kitaev chain32 . However, the total number of par-
aforementioned approach which relies on synthetic dimen- ticles is fixed in our case. This is the crucial difference be-
sions/synthetic gauge fields. Specifically, we demonstrate the tween our approach and a related earlier proposal presented in
appearance of such modes in a cold atomic system where Ref. [31] where a single Hall ribbon has been treated in a BCS
a pairing interaction is induced between quantum Hall-like mean-field approximation. By its nature, such a mean-field
 2

approximation breaks particle number conservation, leading Here, σ = P {a, b} labels the legs (strips) of the ladder,
†
the authors of Ref. [31] to a physically tenuous conclusion and nx,σ = m cx,m,σ cx,m,σ . The inter-chain tunneling
about the existence of Majorana zero modes in their setup. In term (1b) also also allows for a position-dependent tunneling
contrast, our approach does not rely on the mean-field approx- phase Φx , which will be discussed later.
imation; we show that the prerequisite pairing instability is
triggered by arbitrarily weak attractive interactions. Neverthe- In what follows, we focus on the case of strongly
less, the presence of interactions is crucial here; zero modes anisotropic tunneling, specifically t⊥  t. We envision each
found in a related, but non-interacting setup in Ref. [33] are of the synthetic strips to be in the quantum Hall regime, so that
of the Su–Schriefer–Heeger34 , not Majorana type. (In partic- the system forms two weakly coupled quantum Hall strips,
ular, those zero modes can be individually occupied or empty, which requires the following relation between the model pa-
implying e.g. the wrong quantum dimension that is inconsis- rameters:
tent with the claimed non-Abelian braiding statistics of the |U |, t⊥  Ω  t. (2)
Ising type.) Meantime the particle conservation constraint is
circumvented in our case by considering a double synthetic
ribbon. The ground state degeneracy is no longer associated Let us first consider the case of decoupled strips, t⊥ = 0. A
with the overall fermion parity of the closed system and is single particle dispersion relation in the absence of the inter-
instead encoded in the parity in each chain. action term (1d) is shown in Fig. 2. When there is no tunneling
in the synthetic dimension as well (Ω = 0), the dispersion re-
The crucial advantage of using the quantum Hall regime lation for each strip (σ = a, b) consists of M cosines (dashed
is that it naturally allows for generalizations to the frac- lines in in Fig. 2), shifted with respect to one another hori-
tional case, where we expect that with small modifications zontally in the momentum space due to the presence of the
the present setup will allow an experimental realization of synthetic magnetic flux Φ. For finite Ω, m → m ± 1 tran-
fractional topological superconductor phases containing ex- sitions lead to avoided crossings and open gaps (solid lines
otic anyons, e.g. of the parafermion type. In this paper we in Fig. 2). Each synthetic strip thus realises a coupled-wire
shall concentrate on the proof of concept for the simplest pos- construction of Ref. [35]. We assume that the temperature is
sible case, leaving such generalization to fractional state to the lower than the quantum Hall gap Ω, allowing one to reach the
future. quantum Hall regime. Specifically we consider the filling fac-
tor ν = 1, meaning that the Fermi level lies within the first
gap, as shown in Fig. 2. This corresponds to the number of
atoms per (synthetic) site being hnx,m,σ i = Φ/2π.
II. MODEL
In the quantum Hall regime all modes in the bulk (1 < m <
M ) are gapped and the low energy physics is governed solely
We consider a double chain, or two-leg ladder, of atoms by the two chiral edge states crossing the Fermi level. Specif-
with an internal quantum number m = 1, .., M , as shown in ically, for each physical chain of atoms σ = a, b, the first syn-
Fig. 1(a). Atoms can hop along the chain or between the two thetic chain m = 1 hosts a left-moving mode, and the M -th
chains with hopping amplitudes t and t⊥ respectively. These synthetic chain hosts a right-moving edge mode,
hopping processes conserve the internal quantum number m. σ σ
In addition, our model allows internal transitions m → m ± 1 cx,M,σ = eikF x Rσ (x), cx,1,σ = e−ikF x Lσ (x). (3)
with amplitude Ω and an imprinted phase Φ, which can be
Here Rσ (x), Lσ (x) are slowly varying fermionic fields (we
achieved in practice by illuminating the system by additional
have set the short distanceR cut-off a0 = 1). The effective edge
lasers at judiciously chosen angles5 . These transitions can
Hamiltonian is Hedge = dx (H0 + H⊥ ), where
be regarded as hopping in the transverse ‘synthetic’ direction.
I.e., each leg of the physical ladder is now effectively a strip, X
ivσ Rσ† ∂x Rσ − L†σ ∂x Lσ + U Rσ† Rσ L†σ Lσ
 

as shown in Figure 1(b). Finally, an on-site interaction U – see H0 =
σ
in Fig. 1(a) – becomes in effect a non-local interaction within
(4a)
each synthetic strip. The resulting tight binding Hamiltonian  
for our system is Hlattice = Ht + H⊥ + HΩ + Hint where H⊥ = t⊥ e−iδkx eiΦx /2 Ra† Rb + eiδkx e−iΦx /2 L†a Lb + h.c.
X (4b)
Ht = −t (c†x,m,σ cx+1,m,σ + h.c.), (1a)
x,σ,m The first term contains both the kinetic energy and density–
m
X
eiΦx M c†x,m,a cx,m,b + h.c. , density interaction within each physical chain. Here vσ =


H⊥ = t ⊥ (1b)
x,m 2t sin(kFσ ) where kFσ = nσ π with nσ = hn̂x,σ,m i being the
−1 average density in each synthetic chain. We shall ignore band
XM
curvature and hence assume va = vb = vF . The second term
X
eixΦ c†x,m,σ cx,m+1,σ + h.c. ,


HΩ = Ω (1c)
x,σ m=1
describes tunneling between the two physical chains; using
X Eq. (3) we can see that for a finite density difference between
Hint = U n2x,σ . (1d) the two chains this term oscillates as eiδkx where δk = kFa −
x,σ kFb and is therefore irrelevant as long as kFa 6= kFb .
 3

(a) (b) (c)

Φ
𝑚 𝑈 𝐿𝑎
𝑚=
𝑈 Φ𝑥 𝑈
1,2 … , 𝑀 𝑅𝑎
𝜎=𝑎 Ω
𝑡⊥ 𝑡⊥ 𝑡⊥
𝑅𝑏
𝜎= 𝑏
𝑡 𝑡

FIG. 1. (a) Two coupled chains with an on-site interaction U and internal atomic quantum number m = 1, ..., M . m is conserved upon either
intrachain (t) or interchain (t⊥ ) hopping. (b) Equivalent picture where quantum number m, 1 ≤ m ≤ M , is interpreted as an additional
“dimension”. In addition a finite transition amplitude m → m + 1 is included (Ω) with imprinted phase Φ. (c) In the quantum Hall regime the
bulk modes m = 1, M are gapped and at low energy the system consists of edge modes. Counter propagating edge modes interact non-locally
via the initially on-site Hubbard interaction U .

specific to our proposed setup based on synthetic quantum

Hall ribbons, particularly the role of the non-locality of inter-
actions in the synthetic dimension and the flux Φx inserted
within a loop in the plane containing the synthetic dimen-
𝐿𝑎 𝑅𝑎 sion. We begin by characterizing the resulting superconduct-
Ω ing phase from the point of view of a Luttinger liquid instabil-
−𝑘𝐹𝑎 𝑘𝐹𝑎
… M ity to pair tunneling, see Fig. 2, by following the analysis and
1 2
notations of Cheng and Hao23 who studied an identical effec-
tive model. We will address its ground state degeneracy using
bosonization. In close analogy to the analysis of Ref. 29, we
will construct Majorana operators and show that the model
has an emergent Z2 symmetry obtained at a special value of
𝐸 the flux Φx ; when projected to the low energy subspace, this
𝐿𝑏 𝑅𝑏 Z2 symmetry coincides with the parity symmetry generated
−𝑘𝐹𝑏 𝑘𝐹𝑏 by the Majorana operators and hence protects the ground state
1 2 … M degeneracy from local perturbations.
𝑘

FIG. 2. Energy-momentum diagram and interaction processes of the

model Eq. (1). The cosine dispersions for each m are shifted in mo- A. Renormalization group analysis
mentum space due to the flux Φ (dashed lines). The m → m±1 pro-
cesses Ω create avoided crossings and open energy gaps separating In the presence of interactions, the inter-chain tunneling
Landau levels (full lines). Two counter propagating edge modes Rσ , [see Eq. (4b)] results in the generation of two more quartic
Lσ remain at the Fermi level (dashed line) for each chain σ = a, b.
terms in the effective low-energy Hamiltonian: simultane-
Among interaction processes, the pair tunneling term transfers two
particles between the two chains. It is facilitated by the non-local
ous backscattering within the two physical chains, and pair
interaction U between the edge states. tunneling
 between them.  These terms are proportional to
e La Ra Rb† Lb + h.c. and Ra† L†a Lb Rb respectively. As
2iδkx †

can be seen from Fig. 2, at energies much smaller than Ω these

III. SUPERCONDUCTING PHASE FOR ATTRACTIVE processes involve just the edge modes. Under normal circum-
INTERACTIONS stances the amplitudes of these processes would be suppressed
due to the spatial separation of the counter-propagating edge
With the goal of realising topological superconducting modes. However, in our setting this separation occurs in the
states, we now turn to the analysis of our model in the case synthetic dimension, resulting in no such suppression: the in-
of attractive interactions, U < 0, which is a realistic possi- teraction U is nonlocal in this dimension.
bility in the realm of cold atoms. For our analysis to be self- Notice that for a finite density imbalance between the
contained, we shall review some previous results on Majorana chains, δk = kFa − kFb 6= 0, while both the single particle tun-
fermions in setups with particle number conservation23,29,36 , neling described by Eq. (4b) as well as the aforementioned si-
whose effective description closely matches that of Eq. (4). multaneous backscattering operator do not conserve momen-
We will emphasize features and experimental knobs that are tum and, as a result, oscillate with wavenumbers δk and 2δk
 4
p
respectively. Meantime, the induced pair tunneling term does π/2 nθ with integer-valued operator nθ .
not oscillate. This property can be used to selectively promote Generally, the ground state degeneracy associated with a
the pair tunneling process. topological nontrivial phase depends on the boundary con-
It is convenient to bosonize this model, ditions. Our 1D system dominated by pair tunneling is sur-
rounded by trivial regions both on the left (x < 0) and the
1 √
R(L)σ = √ ei π(θσ +rϕσ ) , (5) right (x > L). We can mimic
P R these boundaries by introduc-
2πa0 ing a large mass term M σ dxRσ† Lσ + h.c. to the trivial
regions36 . Equivalently, after bosonization, this term can be
where r = +/− for R/L and σ = a, b; a0 is a short distance written as
cutoff, and the two bosonic fields satisfy [∂x ϕσ (x), θσ (x0 )] =
iδ(x − x0 ). The field ϕσ is related to the charge density in XZ M √
chain σ via ρσ = √1π ∂x ϕσ (x), and its conjugate field θσ may HM = cos(2 πϕσ )
σ
πa0
be interpreted as the phase of the pair field Lσ Rσ . Finally we Z √ √
define even and odd combinations ϕ± = √12 (ϕa ± ϕb ) and 2M
= cos( 2πϕ+ ) cos( 2πϕ− ). (7)
πa0
θ± = √12 (θa ± θb ).
In the bosonized language the edge Hamiltonian Hedge , In the strong coupling limit,√it pins the ϕ± fields on the left
which includes pair tunneling and backscattering, becomes and right sides, ϕ− (x) = 2πnL ϕ , (x < 0) and ϕ− (x) =
√ L/R
H = H0 + H⊥ + H2 where R
2πnϕ , (x > L), with nϕ being integer-valued opera-
X vµ  tors.37
Kµ (∂x θµ )2 + Kµ−1 (∂x ϕµ )2 ,

H0 = (6a) These integer eigenvalues have a transparent physical
µ=±
2
2t⊥ √  √
meaning. First consider the individual chains √ σ = a, b. The
H⊥ = cos 2π ϕ̃− − Φx /2 cos 2πθ− , (6b) difference ϕσ (x > L) − ϕσ (x < 0) = πnσ according to
πa0 Eq. (7). Using ρσ = √1π ∂x ϕσ , we see that nσ is the number of
g1 √ g2 √
H2 = cos 8π ϕ̃− + cos 8πθ− , (6c) particles in chain σ. Conservation of the total particle number
(πa0 ) 2 (πa0 ) 2
constraints na + nb = const. Let us focus on the case when
this number is even. Since both na and nb are integer, the dif-
with K+ = K− = 1 − O(U ) being the Luttinger parameter, ference na − nb must be even as well. The individual particle
2
√ and g1 , g2 ∝ t⊥ .
v+ = v− being the Luttinger liquid velocity, numbers na and nb are not conserved by the Hamiltonian due
In the last two lines ϕ̃− ≡ ϕ− + δkx/ 2π, which accounts to both the single particle and pair tunneling terms. The pair
for the density imbalance between the two chains. tunneling operator changes this number by 2 in each chain,
This form of the Hamiltonian allows us to discuss insta- thus preserving their individual parities. Hence it commutes
bilities of the Luttinger liquid. Notice that the even sector with parity P̂ = eiπδQ , where δQ = (na − nb )/2 = nR L
ϕ −nϕ .
(µ = +) remains gapless. The odd sector (µ = −) con- Therefore, if we ignore the single particle tunneling term
tains both the single particle tunneling [Eq. (6b)] and the two- (we shall return to this point later), the entire low energy spec-
particle processes [Eq. (6c)]. Following the RG analysis of trum (and not just the ground state!) is two-fold degenerate:
Ref. 23, we find that the single particle tunneling has scaling the degeneracy corresponds to P̂ = ±1. We denote the corre-
−1
dimension x⊥ = (K− + K− )/2 and should therefore should sponding states as |ei and |oi respectively.
be relevant (x⊥ < 2) even for U = 0. However, for a fi- In the bulk, 0 <px < L, the pair tunneling operator pins the
nite density mismatch between the chains (which we assume θ− field to θ− = π2 nθ with integer-valued nθ . Thus one can
here), it becomes oscillating and hence irrelevant. The cor- √
related backscattering and pair tunneling terms have scaling consider the operator ei 2πθ− ≡ eiπnθ , and try to distinguish
dimensions xb = 2K− and xp = 2K− −1
respectively; for ar- states within each parity sector, with different locking of θ− .
bitrarily small attraction, U < 0, we have K− > 1 so that pair However, despite of this additional pinning, the states |ei and
tunneling becomes relevant (xp < 2) while backscattering – |oi with well defined δQ = nR L
ϕ − nϕ can not be further dis-
irrelevant (xb > 2). tinguished by the value of θ since the integer-valued operators
R/L
Hence even weak attractive interactions in the presence of nϕ and nθ do not all commute. Using the non-local com-
density imbalance between the chains would stabilize a quan- mutation relation between ϕ− (x) and θ− (x), and projecting
tum phase dominated by pair tunneling. We now turn to a them to the low energy subspace, one arrives at
discussion of the properties of this phase.
i
nR
   L 
ϕ , nθ = , nϕ , nθ = 0. (8)
π
B. Ground state degeneracy This implies that P̂ = eiπδQ and Θ̂ = eiπnθ anti-commute
rather than commute, P̂ Θ̂ = −Θ̂P̂. Consequently, working
We consider a system of linear extent L (i.e. 0 < x < L), in the P̂ basis one finds that hP|eiπnθ |Pi = 0. Therefore, the
whose behavior
√ is dominated by the relevant pair tunneling low-energy spectrum of the system is indeed only two-fold
term cos 8πθ− . It locks the θ− field, i.e., the difference degenerate.
in the phases of the pair fields La Ra and Lb Rb , to θ− = The above algebra suggests a Pauli matrix representation of
 5

the aforementioned operators, discussed here does not correspond to any microscopic sym-
metry. For instance, it is different from the time reversal sym-
P̂ = σ z , Θ̂ = σ x . (9) metry discussed in a similar context in [38].
It is instructive to bosonize the single particle tunneling op-
One can change basis from the parity basis to the eigenbasis erator, replace ϕ− and θ− by their expressions in terms of
of Θ̂, as |Θi = √12 (|ei ± |oi). integer-valued operators, and compare the resulting expres-
sion with that of the Majorana operators γ R/L in Eq. (10):
L/R
L†a Lb (x = 0) ∼ eiπ(2nϕ +nθ )
. (14)
C. Majorana operators, single particle tunneling, and Z2
symmetry
Note that the single Majorana operators in our strongly inter-
acting state are nonlocal in terms of the original particles.
Following Chen et. al.29 it is natural to define Majorana
operators
L/R
D. Finite splitting for Φx 6= π
γL/R = eiπ(nϕ +nθ )
, (10)
We shall now address the single-particle tunneling process
which are Hermitian, square to one, and anti-commute. In
H⊥ near the edges (i.e., the zero-dimensional boundary) in
terms of these operators, the parity operator P̂ = iγL γR . We more detail, focusing on its effect on the approximate two-
note, however, that no symmetry protects the ground state de- fold degeneracy. Although no local operator can measure the
generacy associated with this parity in the presence of a finite total parity and hence distinguish the |ei and |oi states, it is
single particle hopping term, t⊥ 6= 0. Even if irrelevant, it can possible to mix these states by a local process – tunneling of
be effective near the edges, i.e., a local perturbation such as a single particle.
H⊥ near the left or right edge can couple to one of the Majo-
Let us express the single particle tunneling operator of
rana operators and change the parity. However, this coupling
Eq. (6b) near the left edge, x = 0, or the right edge, x = L, in
can be eliminated by tuning flux Φx to a specific value. A
terms of Pauli matrices σ i . H⊥ (x) is off-diagonal in the par-
special symmetry emerges in this case and prohibits coupling
ity basis since it transfers one particle between the chains, and
to the individual Majorana operators.
therefore it has to be a combination of σ x and σ y . In addition,
Consider the bosonized Hamiltonian given by Eq. (6) at L/R L/R
as it contains e2πinϕ rather than eπinϕ , it is diagonal in
Φx = π. One can identify the special symmetry
the |Θi basis. Thus H⊥ (x) acts in the ground state manifold
as σ x . To compute its x dependence near the interfaces we ap-
p
U : θ− → θ− + π/2,
proximate the cosine potential by a mass term in the appendix,
ϕ̃− → −ϕ̃− . (11) and obtain
− → −ϕ− and x → −x.
The latter transformation implies ϕp 2t⊥
The transformation θ− → θ− + π/2 shifts θ− between hΘ|H⊥ (x)|Θi = cos(Φx /2)eiπnθ e−πx/(2ξ) , (15)
πξ
subsequent minima√ of the pair-tunneling term, but changes
the sign of cos( 2πθ− ) in the single particle hopping term, with correlation length ξ = v/∆ and energy gap ∆ ∝
which however gets compensated by ϕ̃− → −ϕ̃− exactly at 1
(g2 ) 2−xp (here we have assumed K− = 1; for the dependence
Φx = π. on K− see the appendix). Integrating over x one obtains
In addition, one can see how this transformation acts on
the Majorana operators. Since it shifts nθ by unity, and takes 4t⊥
hΘ|H⊥ |Θi = cos(Φx /2)eiπnθ . (16)
nφ → −nφ = nφ mod 2, we may conclude that this symmetry π2
U acts on the Majorana fermions as The exponential decay of these matrix elements could be
−1 tested numerically by computing the x dependence of the op-
U γL/R U = −γL/R . (12) P †
erator the operator H⊥ (x) = m cx,m,1 cx,m,2 + h.c.. In
Thus within the low-energy subspace we can identify U with addition, such inter-site hopping process may be directly mea-
the parity operator, surable in experiment using the techniques for detecting par-
ticle currents39 .
U = P̂ ≡ iγL γR , (13) In full agreement with the symmetry argument of the pre-
vious section, the synthetic magnetic flux Φx between the
which acts in the same way on the Majorana operators. There- two synthetic ladders can completely cancel the effect of sin-
fore, by fine tuning the parameter Φx we can reach a point gle particle tunneling within the low energy subspace. This
where a coupling to a single Majorana fermion, which could cancellation has a simple physical interpretation in terms of
change the parity quantum number P , becomes forbidden by a two-path interference. The total result for the matrix el-
symmetry. We shall see this mechanism in action in the cal- ement of H⊥ in Eq. (15) is in fact a sum of two indepen-
culation of matrix elements of the single particle tunneling dent hopping terms which turn out to have equal weight:
term presented in the next section. We note that the symmetry hnθ |Ra† Rb |nθ i = hnθ |L†a Lb |nθ i - one due to the hopping be-
 6

tween the right moving modes at one end in the synthetic di- energy gaps. We should recall, however, that our system is in
mension, and the other due to hopping between the left mov- fact gapless due to the extra even sector associated with the
ing modes at the other end. In the presence of the flux Φx , the total charge. Its effective description is the Luttinger liquid
phase difference between the two paths can result in a destruc- with fields θ+ , ϕ+ . As discussed in the previous sections and
tive interference. This once again emphasizes the fact that the earlier works23,29,36 , the topological properties in the odd sec-
Majorana operators are non-local in the synthetic dimension. tor (i.e. the spin sector), remain approximately impervious to
the charge sector, which therefore acts merely as a spectator.
However, as we shall see, the charge sector plays a crucial
IV. FUSION OF MAJORANA FERMIONS AND role in determining the adiabatic condition for our “fusion”
NON-LOCAL ENTANGLEMENT protocol.
While the total charge N is fixed, as the barrier is raised,
the total number of particles in the left side NL or in the right
So far we have been considering a single region where su-
side NR = N − NL can fluctuate. Yet, it is precisely the
perconductivity emerges intrinsically and leads to an approxi-
Luttinger liquid of the charge sector which provides a finite
mate two-fold degeneracy associated with Majorana edge op-
charge compressibility. Assuming for concreteness that N is
erators. We now turn to the setting with multiple supercon-
even, as well as a symmetric left-right bipartition, the lowest
ducting regions separated by trivial regions. Specifically, we
energy charge state will have (NL , NR ) = (N/2, N/2). The
will focus on the process of nucleating such an extra trivial
excited charge states (NL , NR ) = (N/2 ± 1, N/2 ∓ 1) will
region within a superconducting region in real time, which is
have an energy cost of order ~v L originating from the Luttinger
a possibility in cold atom systems.
liquid Hamiltonian of the charge sector. This energy scale sets
By analogy with the setup of Ref. [40], consider the pro- the adiabatic condition implying that the system size should
cess whereby a uniform system is initialized in a well-defined not be too large.
parity state and then divided it into two separate parts by e.g. As an alternative to this very restrictive adiabatic condition
ramping up the potential energy term in the central region, one may in fact choose the opposite, namely perform a sudden
as depicted by the step (b)→(c) in Fig. 3. This process re- quench of the barrier. After a sudden ramp-up of the potential
sults in creating a pair of new Majorana operators, γ2 and energy term in the central region, we end up with the vacuum
γ3 . Consequently, each new superconducting region can be state described by Eq. (7). The field ϕ+ is pinned in the
either in an even or odd parity state, described by the eigen- p central
region, but can take different values ϕ+ (L/2) = π/2NL
values of the operators iγ1 γ2 and iγ3 γ4 . We should remind
with NL being an integer-valued operator describing the to-
the reader that, owing to the particle conservation constraint
tal (summed over σ = a, b) number of particles in the left
in each segment, the notion of parity in our case is different
side. Imagine now performing a strong charge measurement
from that of Ref. [40]; for us the overall parity of each seg-
of NL . After such a measurement the quantum state collapses
ment is fixed while the degeneracy is associated with the in-
onto one with a fixed NL . Independently, the value of ϕ− in
dividual parity of its two constituent chains. Let us assume
the barrier region is described by the integer valued operator
for concreteness that the system is initialized in an even parity
nbarrier
ϕ , taking two physically distinct values corresponding to
state iγ1 γ4 = 1, denoted as |e14 i. Upon adiabatic creation of
the relative parity of the chains in the left side. The latter is
the middle barrier, two new Majorana operatorsγ2,3 are cre-
linked with the parity in the right side: for even NL both par-
ated from the vacuum and hence are also in an even parity
ities are equal, and for odd NL the two parities are opposite.
state |e23 i. By changing to the basis associated with the fu-
Since before the sudden quench the wave function in the cen-
sion outcomes of Majorana pairs in the right and left halves,
tral region was dominated by pair tunneling, with pinned θ−
i.e. {γ1,2 } and {γ3,4 }, one obtains
field, it follows that this state is a superposition in terms of
1 nbarrier
ϕ ; Thus, we conclude that via a sudden quench of the
|e14 e23 i = √ (|e12 e34 i + |o12 o34 i) . (17) barrier, and after a strong measurement of the charge in the
2
left side, one effectively obtains the entangled state Eq. (17)
The parities associated with the of the left and right sides are for any outcome of the measured charge NL .
maximally entangled. This is a topological effect, as it does In Appendix B we complement this discussion of the fu-
not depend on the details such as the precise asymmetry of the sion protocol in a number conserving system by studying an
bipartition of the system. exactly solvable toy model introduced by Iemini et. al.27 . This
The resulting state contains maximal parity fluctuations in allows for an explicit treatment of the subtleties associated
each segment. Detecting them can be done first preparing the with the gapless charge sector.
system in the state with no overall parity fluctuations and then
implementing a protocol similar to that of Ref. 40 and de-
picted in Fig. 3. This test requires measuring parity states of V. CONCLUSIONS
each segment of the system, using, for example, time of flight
measurements as discussed e.g. in Ref. [41]; In this work we described a closed, particle-conserving
The validity of the low-energy effective description of our setup in a cold atom system endowed with both a synthetic
system in terms of Majorana zero modes requires adiabatic- dimension and a synthetic gauge field with the goal of real-
ity of the time dependent process, with respect to the relevant izing close analogues of Majorana zero modes. In contrast
 7

(a) 𝑡⊥ = 0
𝑛𝑎 =fixed

𝑛𝑏 =fixed

𝑡, 𝑡⊥ ≠ 0 𝑉𝐿/2 → ∞ 𝑡⊥ = 0
(b) (d)

𝛾1 𝛾4

|014 >

𝑉𝐿/2 → ∞ 𝑡⊥ ≠ 0 𝑡⊥ ≠ 0
(c) (e)

𝛾1 𝛾2 𝛾3 𝛾4 𝛾1 𝛾2 𝛾3 𝛾4

1
|𝑒14 𝑒23 > = |𝑒12 𝑒34 > +|𝑜12 𝑜34 > |𝑒12 𝑒34 >
2

FIG. 3. Implementation of a fusion-testing protocol analogous to Ref. [40] in a cold atom setting with particle number conservation. (a) Two
disconnected chains with t⊥ = 0 and a fixed number of particles in each chain na , nb . (b) By turning on the inter-chain coupling t⊥ the
system is driven to the doubly degenerate state with well defined parity state of Majorana operators iγ1 γ4 = +1 dented |e14 i. (c) A strong
potential barrier is created in the center. The resulting central region has a trivial gap, similar to the regions at x < 0 and x > L. Thus an
additional degeneracy emerges described by two additional Majorana operators γ2 , γ3 which have been created from the vacuum in an even
parity state |e23 i. As described in Eq. (17) this is a maximally entangled state in terms of the parity of the left and right sides. (d,e) To test
that parity fluctuations in each subregion are intrinsic to the system, we can prepare a system described by the same four Majorana operators
but with a well defined parity in each half, by first breaking the decoupled chains into two separate segments each (d), and then turning on the
inter-chain coupling t⊥ (e).

to other approaches relying on the superconducting proxim- The fact that Majorana operators are non-local in the syn-
ity effect, the two-particle hopping process, which is respon- thetic dimension allows us to manipulate the associated de-
sible for the topological superconducting phase, is generated generacy by adding a synthetic flux Φx . This in turn can pro-
in our setup from the single particle hopping in the presence vide us with a useful knob for combining “topological” and
of small attractive interaction. We emphasize, that our model “non-topological” qubit operations. An even more intriguing
builds upon an experimentally tested implementation of quan- possibility is to utilize a similar setup for producing and ma-
tum Hall edge states in the cold atom setting. nipulating fractionalized zero modes19–22 , a topic that we in-
The synthetic dimension approach has a number of advan- tend to cover elsewhere.
tages. Local particle-particle interactions in real space be- VI. ACKNOWLEDGMENTS
come non-local in the synthetic dimension. This is the key
feature which effectively turns a small local attractive interac-
tion into an attractive interaction between “distant” edge states
in a synthetic quantum Hall ribbon, thus resulting in the for- The authors would like to thank Hans Peter Büchler, Se-
mation of analogues of Majorana zero modes. The wavefunc- bastian Diehl, Yuval Gefen, Leonardo Mazza, and Christophe
tion of these modes, while local in the real space, remains Mora for useful discussion. This project was supported in
non-local in the synthetic dimension. The prerequisite non- part by the ISF Grant No. 1243/13 and by the BSF Grant
local interactions can not be induced in a conventional con- No. 2016255; KS was supported in part by the NSF DMR-
densed matter setup. 1411359 grant.

1 3
C. Nayak, S. H. Simon, A. Stern, M. Freedman, and C. W. J. Beenakker, Annu. Rev. Condens. Matter Phys. 4, 113
S. Das Sarma, Rev. Mod. Phys. 80, 1083 (2008), (2013), arXiv:1112.1950.
4
arXiv:0707.1889. N. Goldman, J. C. Budich, and P. Zoller, Nat. Phys. 12, 639
2
J. Alicea, Rep. Prog. Phys. 75, 076501 (2012), arXiv:1202.1293. (2016), arXiv:1607.03902.
 8

5 37
A. Celi, P. Massignan, J. Ruseckas, N. Goldman, I. B. Spielman, For finite density mismatch δk 6= 0 the integer-valued operators
G. Juzeliūnas, and M. Lewenstein, Phys. Rev. Lett. 112, 043001 nϕ are defined in the same way in terms of ϕ̃− rather than ϕ− .
38
(2014), arXiv:1307.8349. A. Haim, K. Wölms, E. Berg, Y. Oreg, and K. Flensberg, Phys.
6
M. Mancini, G. Pagano, G. Cappellini, L. Livi, M. Rider, Rev. B 94, 115124 (2016), arXiv:1605.09385.
39
J. Catani, C. Sias, P. Zoller, M. Inguscio, M. Dalmonte, and M. Atala, M. Aidelsburger, M. Lohse, J. T. Barreiro, B. Paredes,
L. Fallani, Science 349, 1510 (2015), arXiv:1502.02495. and I. Bloch, Nat. Phys. 10, 588 (2014), arXiv:1402.0819.
7 40
B. K. Stuhl, H.-I. Lu, L. M. Aycock, D. Genkina, and I. B. Spiel- D. Aasen, M. Hell, R. V. Mishmash, A. Higginbotham, J. Danon,
man, Science 349, 1514 (2015), arXiv:1502.02496. M. Leijnse, T. S. Jespersen, J. A. Folk, C. M. Marcus,
8
H. M. Price, O. Zilberberg, T. Ozawa, I. Carusotto, and N. Gold- K. Flensberg, and J. Alicea, Phys. Rev. X 6, 031016 (2016),
man, Phys. Rev. Lett. 115, 195303 (2015), arXiv:1505.04387. arXiv:1511.05153.
9 41
S. Barbarino, L. Taddia, D. Rossini, L. Mazza, and R. Fazio, Nat. C. V. Kraus, S. Diehl, P. Zoller, and M. A. Baranov, New J. Phys.
Commun. 6, 8134 (2015), arXiv:1504.00164. 14, 113036 (2012), arXiv:1201.3253.
10
T.-S. Zeng, C. Wang, and H. Zhai, Phys. Rev. Lett. 115, 095302
(2015), arXiv:1504.02263.
11
L. Taddia, E. Cornfeld, D. Rossini, L. Mazza, E. Sela, and Appendix A: Matrix elements of H⊥
R. Fazio, Phys. Rev. Lett. 118, 230402 (2017), arXiv:1607.07842.
12
A. Petrescu and K. Le Hur, Phys. Rev. B 91, 054520 (2015),
arXiv:1410.6105. Our goal is to compute matrix elements of H⊥ (x) in the
13
E. Cornfeld and E. Sela, Phys. Rev. B 92, 115446 (2015), ground state manifold. To set up the notation,pwe write the
arXiv:1506.08461. mode expansion (in this appendix we set θ = K− θ− , ϕ =
14 p
M. Calvanese Strinati, E. Cornfeld, D. Rossini, S. Barbarino, ϕ− / K− , v− = v)
M. Dalmonte, R. Fazio, E. Sela, and L. Mazza, Phys. Rev. X
7, 021033 (2017), arXiv:1612.06682. X 1  
15
A. Petrescu, M. Piraud, G. Roux, I. P. McCulloch, and K. Le Hur, ϕ(t, x) = ϕ0 + p βn eikn x−iv|kn |t + h.c. (A1)
,
k6=0
4π|n|
Phys. Rev. B 96, 014524 (2017), arXiv:1612.05134.
16
L. Fu and C. L. Kane, Phys. Rev. Lett. 100, 096407 (2008),
arXiv:0707.1692. where kn = 2πn LR and [βn , βn ] = δn,n . The free Hamilto-
0 0

17 v L 2 2
R. M. Lutchyn, J. D. Sau, and S. Das Sarma, Phys. Rev. Lett. 105, nian is H0 = 2 0 dx[(∂x ϕ) + Π ] with the conjugate field
077001 (2010), arXiv:1002.4033. Π(t, x) = v1 ∂t ϕ(t, x) satisfying [ϕ(t, x), Π(t, y)] = iδ(x−y).
18
Y. Oreg, G. Refael, and F. von Oppen, Phys. Rev. Lett. 105, We define also Π(t, x) = −∂x θ(x). One can check that
177002 (2010), arXiv:1003.1145.
19 these canonical commutation relations are recovered using the
D. J. Clarke, J. Alicea, and K. Shtengel, Nat. Commun. 4, 1348
(2013), arXiv:1204.5479.
mode expansion and that the Hamiltonian becomes H0 =
2πv
P∞ † †
n=1 n[βn βn + β−n β−n ].
20
N. H. Lindner, E. Berg, G. Refael, and A. Stern, Phys. Rev. X 2, L
041002 (2012), arXiv:1204.5733. We consider the interface
p x = 0 between the region where
21
M. Cheng, Phys. Rev. B 86, 195126 (2012), arXiv:1204.6084. θ is pinned
√ θ(x) = π/2 n θ for x > 0, and ϕ is pinned to
22
J. Alicea and P. Fendley, Annu. Rev. Condens. Matter Phys. 7, 119 ϕ(x) = 2πnL for x < 0. The latter is accounted for by the
ϕ √
(2016), arXiv:1504.02476.
23
M. Cheng and H.-H. Tu, Phys. Rev. B 84, 094503 (2011). boundary condition ϕ(x = 0, t) = 2πnL ϕ . Using the mode
24
J. D. Sau, B. I. Halperin, K. Flensberg, and S. Das Sarma, Phys. expansion, this boundary condition implies βn + β−n = 0,
Rev. B 84, 144509 (2011), arXiv:1106.4014. hence
25
C. V. Kraus, M. Dalmonte, M. A. Baranov, A. M. Läuchli, and √ X 1
ϕ(t, x) = 2πnL √ sin(kn x) βn − βn† ,


P. Zoller, Phys. Rev. Lett. 111, 173004 (2013), arXiv:1302.0701. ϕ +i
26
J. Ruhman, E. Berg, and E. Altman, Phys. Rev. Lett. 114, 100401 πn
k6=0
(2015), arXiv:1412.3444. X 1
√ cos(kn x) βn + βn† .
27


F. Iemini, L. Mazza, D. Rossini, R. Fazio, and S. Diehl, Phys. θ(t, x) = (A2)
Rev. Lett. 115, 1 (2015), arXiv:1504.04230. πn
k6=0
28
L. Mazza, M. Aidelsburger, H.-H. Tu, N. Goldman, and M. Bur- p
rello, New J. Phys. 17, 105001 (2015), arXiv:1506.08057. Ignoring tunneling between minima θ(x) = π/2 nθ of the
29
C. Chen, W. Yan, C. Ting, Y. Chen, and F. Burnell, the region x > 0, we replace the latter by
cosine potential in p
arXiv:1701.01794 (2017). a mass term g(θ − π/2 nθ )2 (where p g → g̃ = g/K− ). This
30
K. Guther, N. Lang, and H. P. Büchler, Phys. Rev. B 96, 121109
(2017), arXiv:1705.01786. gives a finite average value hθ(x)i = π/2 nθ . To account
31
Z. Yan, S. Wan, and Z. Wang, Sci. Rep. 5, 15927 (2015), for fluctuations around this value, we rewrite the Hamiltonian
arXiv:1504.03223. using the mode expansion:
32
A. Y. Kitaev, Phys.-Usp. 44, 131 (2001), arXiv:cond-
∞
mat/0010440. X
33
J. Klinovaja and D. Loss, Phys. Rev. Lett. 110, 126402 (2013). Hnθ = [An (βn† βn + βn βn† ) + Bn (βn2 + (βn† )2 )], (A3)
34 n=1
W. P. Su, J. R. Schrieffer, and A. J. Heeger, Phys. Rev. B 22, 2099
(1980).
35
C. L. Kane, R. Mukhopadhyay, and T. C. Lubensky, Phys. Rev. where An = vkn + Bn , Bn = g/k P n . After a Bogoliubov
†
Lett. 88, 036401 (2002), cond-mat/0108445. transformation one obtains
r  H = n E n bn br where
n βn =
36
A. Keselman and E. Berg, Phys. Rev. B 91, 235309 (2015),
  
arXiv:1502.02037.
fn bn + gn b†n , fn = 12 AEn + 1 , gn = −
n 1 An
2 En − 1 ,
 9

En2 = A2n − Bn2 . The Hamiltonian Hnθ describes excitations Integrating over x one obtains
above the |nθ i ground state.
hnθ |H⊥ |nθ i
 1/2K− +K− /2−1
4t⊥ a0
= cos(Φx /2)eiπnθ . (A11)
Having argued that H⊥ is diagonal in this basis, we will π 2 K− ξ
evaluate

hnθ |H⊥ (x)|nθ i = t⊥ eiΦx /2 hnθ |Ra† Rb |nθ i Appendix B: Fusion protocol within an exactly solvable
−iΦx /2
particle-number conserving model
+ t⊥ e hnθ |L†a Lb |nθ + c.c.i (A4)

This requires calculating quantities such as In order to better understand the physics of the fusion pro-
cess in our model, we turn to an exactly-solvable toy model27 ,
1 which is particle-number conserving as well. We begin by re-
q
2π
−i (θ(x)+K− ϕ(x))
hnθ |Ra† Rb |nθ i = hnθ |e K−
|nθ i. viewing the construction of the model27 , and then we describe
πa0
the fusion process.
The well known Kitaev model32 is a non particle-number
conserving model whose Hamiltonian is

Introducing a shorthand notation h...i ≡ hnθ |...|nθ i and us- X † 1


1 2 HK = −Jaj aj+1 − ∆aj aj+1 + h.c. − µ(nj − ) .
ing heA i = ehAi e 2 h(A−hAi) i , we have j
2
q
2π
(B1)
i (θ+K− ϕ) 2
i−(hK− ϕ+θi)2
he K−
i = eiπnθ e−π/K− (h(K− ϕ+θ) . It is exactly solvable and at the “sweet spot” in the parameter
(A5) space where µ = 0, and ∆ = J > 0 (we set ∆ = J = 1),
it is diagonalized as HK = j `†j `j , where `j = Cj† + Aj ,
P
Computing the fluctuations using Eq. (A2) we obtain
q
2π
Cj† = a†j + a†j+1 , and Aj = aj − aj+1 . In order to construct a
i K− (θ(x)+K− ϕ(x)) iπnθ −πvar(θ− (x))−πvar(ϕ− (x))
he i=e e , number conserving model out of the Kitaev model, we recall
(A6) that in BCS theory we begin with a number-conserving, quar-
and tic Hamiltonian, and then by employing a mean field descrip-
∞ tion, we turn into a quadratic Hamiltonian and lose number
X cos2 (kn x) vkn conservation. Analogously, one can write the quartic Hamil-
var(θ(x)) = p , P †
n=1
πn (vkn )2 + 2gv tonian Ha = Lj Lj , where Lj = Cj† Aj . It is immedi-
∞ j
X sin2 (kn x) vkn + k2gn ate to see that this model is number conserving. Also the
var(ϕ(x)) = . (A7)
ground state of the Kitaev model satisfying `j |GSK i = 0,
p
n=1
πn (vkn )2 + 2gv
is the ground state of this model as well since Lj |GSK i =
Cj† Aj |GSK i = Cj† Cj† |GSK i = 0. In fact, one can project
q
vK−
One can see a length scale ξ = 2g . Evaluating these sums
from the model states with well defined number of particles,
for ξ  x  L gives
each of which is a ground state of Ha .
1 The next step is to create an interacting two-chain ver-
var(θ− (x)) ∼ log(ξ/a0 ),
2πK− sion of this model. Denoting the chains by an extra sub-
K− 1 xK− script a, b, we do this by adding an interaction term - LI,j =
var(ϕ− (x)) ∼ log(ξ/a0 ) + . (A8) †
Ca,j †
Ab,j + Cb,j Aa,j . This interaction term is selected since it
2π 2 ξ
annihilates the Kitaev ground state as well, while conserving
One can see then that the total particle number.
We can use this model to better understand the fusion
hnθ |Ra† Rb |nθ i = hnθ |L†a Lb |nθ i process in our setup. The fusion/splitting processes can be
1/2K +K− /2−1 modeled by making coupling constants for the central links
eiπnθ e−πxK− /(2ξ) a0 −
= (A9) (α = a, b) as well as central plaquette (α = I) time depen-
πξ 1/2K− +K− /2 dent:
The result of the calculation is then `−1
L̂†α,j L̂α,j + g(t) L̂†α,`/2 L̂α,`/2 ,
X X X
Ĥ =
hnθ |H⊥ (x)|nθ i α=a,b,I j6=`/2 α=a,b,I
1/2K +K /2−1 (B2)
2t⊥ eiπnθ e−πxK− /(2ξ) a0 − − cos(Φx /2)
= . where a, b label different chains and ` is the chain lenth (as-
πξ 1/2K− +K− /2 sumed to be even). Initially g(0) = 1, corresponding to a
(A10) uniform system, and eventually g(t → ∞) = 0, correspond-
 10

ing to two decoupled halves. It is important to notice that sition is given by

each single term in this Hamiltonian annihilates the ground
state. Hence, if we begin in a ground state, the time depen- N
X X
L
dent Schrödinger equation trivially guarantees that we remain |ψ` (N )iee = ANL ,µµ0 |ψ`/2 (NL )iµµ0
in the same ground state. Therefore for this specific model µ=e,o NL =0
we need not assume anything about the functional form of the R
⊗ |ψ`/2 (N − NL )iµµ0 , (B5)
time dependence g(t).
Here NL is the charge of the left part of the system. Writing

N = na + nb , N L = nL L R R
a + nb , N − NL = na + nb , (B6)
L
since we assumed na and nb are even, the parity of na equals
The ground states of this Hamiltonian at g = 1 are charac-
that of nR L
a and is denoted µ; then the equal parities of nb and
terized by the total number of particles N , and by the parity of 0
nRb which are denoted µ equal
the two chains, being both even or both odd for an even num-
ber of particles, or one even and one odd if N is odd. They µ0 = µ(for NL even), µ0 = µ̄ for NL odd, (B7)
are given by
N/2
with ō = eqand ē = o. The coefficients are given by
Nµµ0 ,`/2,N Nµµ0 ,`/2,N −N
−1/2
X X
|~j2n i ⊗ |~qN −2n i, ANL ,µµ0 = . In fact the depen-
L L
|ψ` (N )iee = Nee,`,N Nee,`,N
0
n=0 {~j2n } dence of these coefficients on µ, µ becomes exponentially
{~
qN −2n } small with increasing system size. Ignoring these exponen-
N/2−1 tially small effects we have ANL ,µµ0 = ANL and can there-
−1/2
X X
|ψ` (N )ioo = Noo,`,N |~j2n+1 i ⊗ |~qN −2n−1 i. fore write
n=0 {~j2n+1 }
{~
qN −2n−1 } |ψ` (N )iee
(B3) X h
L R
where ~jm , ~qm are m-sized vectors denoting the ordered loca- = ANL |ψ`/2 (NL )iee ⊗ |ψ`/2 (N − NL )iee
tions of particles on the a and b chains respectively, and the NL =even
i
sum is over all possible configurations, L
+ |ψ`/2 R
(NL )ioo ⊗ |ψ`/2 (N − NL )ioo
N/2    X h
` ` L R
Nee,`,N =
X
, (N even) + ANL |ψ`/2 (NL )ieo ⊗ |ψ`/2 (N − NL )ieo
n=0
2n N − 2n NL =odd
i
L R
N/2−1 
X `

`
 + |ψ`/2 (NL )ioe ⊗ |ψ`/2 (N − NL )ioe (B8)
Noo,`,N = . (B4)
2n + 1 N − 2n − 1
n=0 We can now explicitly see that for each particle number NL in
Similarly for N odd the number of configurations with well the left segment, the parity of each individual chain is not well
defined parities in each chain is defined and is in fact entangled between the left and right sides
in accordance to the fusion rules for Majorana zero modes
(N −1)/2    γ3 , γ4 .
X ` `
Neo,`,N = Noe,`,N = . (N odd) However one should note a peculiarity of this model. In this
n=0
2n N − 2n equation all charge states are equally likely. The degeneracy
of the system is extensive as all these states differing by NL
are degenerate, in addition to the degeneracy associated with
the parity of each chain in each side, attributed to Majorana
operators. In any generic system like the one considered in the
In the end of the fusion process g = 0 and the system sep- main text, these states differing by NL are non-degenerate and
arates into two halves. Since the wave function in this model actually have an energy difference which scales with the sys-
does not change at all, in order to describe the final state we tem size as power law. This can be understood using bosoniza-
simply perform a left-right decomposition of the ground state tion. While the odd sector is gapped everywhere, either by
wave function. We assume that the total number of particles pair tunneling in the superconductor regions or by the mass
N as well as well as the number of particles in each individ- term in the trivial regions, the even sector, i.e. the charge sec-
ual chain is even. Since the system remains in the zero-energy tor, is only gapped in the trivial regions. The superconductor
ground state of the decoupled left-right Hamiltonians, it is regions act as Luttinger liquids for the charge sector. Hence
guaranteed that we can use the same ground states Eq. (B3) they provide a finite stiffness, which favors a state with a fixed
of the two half-systems ` → `/2. We find that the decompo- charge.