Detecting topological edge states with the dynamics of a qubit

Meri Zaimi,1, ∗ Christian Boudreault,2, † Nouédyn Baspin,3, ‡ Nicolas

Delnour,4, § Hichem Eleuch,5, ¶ Richard MacKenzie,4, ∗∗ and Michael Hilke3, ††
1
Centre de Recherches Mathématiques, Université de Montréal,
C.P. 6128, succursale Centre-ville, Montréal, QC H3C 3J7, Canada
2
Département des sciences de la nature, Collège militaire royal de Saint-Jean
15 Jacques-Cartier Nord, Saint-Jean-sur-Richelieu, QC Canada, J3B 8R8
3
Department of Physics, McGill University, Montréal, QC, Canada, H3A 2T8
4
Département de physique, Université de Montréal, Complexe des Sciences,
C.P. 6128, succursale Centre-ville, Montréal, QC, Canada, H3C 3J7
5
Department of Applied Sciences and Mathematics,
College of Arts and Sciences, Abu Dhabi University, Abu Dhabi, UAE
Institute for Quantum Science and Engineering, Texas A&M University, College Station, Texas 77843, USA
We consider the Su-Schrieffer-Heeger (SSH) chain, which has 0, 1, or 2 topological edge states
depending on the ratio of the hopping parameters and the parity of the chain length. We couple a
qubit to one edge of the SSH chain and a semi-infinite undimerized chain to the other, and evaluate
the dynamics of the qubit. By evaluating the decoherence rate of the qubit we can probe the edge
states of the SSH chain. The rate shows strong even-odd oscillations with the number of sites
reflecting the presence or absence of edge states. Hence, the qubit acts as an efficient detector of
the topological edge states of the SSH model. This can be generalized to other topological systems.

I. INTRODUCTION with which it interacts to improve the qubit’s decoher-

ence, the qubit is viewed as a measuring device capable
Qubits are the building blocks of any quantum infor- of determining properties of its environment. In partic-
mation processing device. Two of the most challenging ular, we explore the dynamics of a qubit attached to a
problems for quantum computing and other applications Su-Schrieffer-Heeger (SSH) chain which is then attached
are decoherence due to the interaction with environment to a third system modeling the environment; the third
and perturbations due to manufacturing imperfections system is a standard tight-binding hopping Hamiltonian
[1–3]. These effects limit the effective performance of (without dimerization).
quantum devices, such as the speed of an eventual quan- The SSH model, described in detail below, is one of the
tum computer. Thus, evaluating the decoherence rate for simplest systems exhibiting interesting topology such as
the qubit or for an ensemble of coupled qubits is of great solitons and, of interest here, edge states [5–10]. In spite
importance. of the model’s inherent simplicity, it manages to cap-
In previous work [4], the decay rate of a qubit coupled ture many interesting and important physical effects in
to another system with or without disorder was studied. topological systems. The model has also been extended
The main objective was to investigate under which cir- to study topological insulators of higher dimensions [11].
cumstances the interaction of a qubit with its surround- Normally, almost-zero-energy edge states have exponen-
ings can be designed to improve the qubit’s performance tially localized wavefunctions at the edges. These states
in a quantum device by increasing the decoherence time. are a particular type of topological edge states. Topologi-
It was shown that the decoherence rate of the qubit is cal edge states have captured the interest of researchers in
related to transport properties of the coupled system. in several fields of physics due to their diverse surprising
Furthermore, it was proven that disorder lowers the deco- proprieties. To name but a few examples, they can en-
herence rate on average. This suggests potential applica- hance the sound intensity at phononic crystal interfaces
tions to increase the performance of qubits in a quantum [12], allow a robust one-way propagation [13] or protect
device by adding impurities to the system. light transport in nanophotonics systems [14]. For fur-
In this work, a similar composite system is studied ther applications and references, see [15, 16].
from a different perspective. Rather than viewing the We evaluate the decoherence rate of the qubit and how
qubit as the system of interest and tailoring the system it depends on the properties of the SSH system in order
to probe the edge states of the SSH chain. As we will
see, it is strongly affected by such states at the qubit end
of the coupled system.
∗ Electronic address: meri.zaimi@umontreal.ca In the next section, we review the isolated SSH model,
† Electronic address: Christian.Boudreault@cmrsj-rmcsj.ca mainly to establish notation but also to highlight the con-
‡ Electronic address: nouedyn.baspin@mail.mcgill.ca
§ Electronic address: nicolas.delnour@umontreal.ca
ditions for the existence of edge states and their proper-
¶ Electronic address: heleuch@fulbrightmail.org ties. In Section III, we explore the double dot coupled
∗∗ Electronic address: richard.mackenzie@umontreal.ca to an SSH chain which is itself coupled to semi-infinite
†† Electronic address: hilke@physics.mcgill.ca chain. An expression for the decoherence rate is derived
 2

using a semi-analytic approximation which is in excel-

lent agreement with numerical simulations of the same
system. We will see a strong effect of edge states on the
decoherence rate. We conclude with a discussion of our
results and avenues for future work in Section IV.

II. SU-SCHRIEFFER-HEEGER MODEL

The SSH model [17] is a one-dimensional tight-binding

model with alternating coupling strengths due to the
Peierls instability [18, 19], leading to a parity effect in
the chain length. The Hamiltonian for a chain of N sites
is
 
0 t1
 
 t1 0 t2 
 
  FIG. 1: (color online) The 4 possible configurations of the
 t2 0 t1  finite SSH model: (a) and (b) correspond to even chain length
HSSH = , (1)
 ..  (N = 20), (a) with no edge states (t1 > t2 ), (b) with 2 edge
 t1 0 . 
  states (t1 < t2 ). (c) and (d) correspond to odd chain length
 .. .. 
(N = 21), (c) with 1 edge state on the right (t1 > t2 ) and (d)
 . . t 
with 1 edge state on the left (t1 < t2 ). The dispersion curves
t 0
for the infinite chain N = ∞ are shown as lines together with
the discrete energy levels of the finite chain as dots. We used
where t = t1 or t2 for N even or odd, respectively. We t1 = 1.2 and t2 = 1/t1 on the left panels and the reverse for
will assume t1 , t2 > 0 for simplicity. the right panels.

B. Analytical solution of the SSH model

A. Overview of the SSH model solutions
Many features of the SSH hamiltonian can be deter-
There are several noteworthy features of the SSH mined analytically; we do so in some detail here for the
model that have attracted a substantial amount of in- case N even (writing N = 2M ) mainly as a review
terest in the literature. For instance, the infinite chain and also to establish notation to be used in what fol-
has a gap given by 2|t1 − t2 | [17] and illustrated in Fig. 1. lows. Many of the results are known (see, for instance,
The existence of this gap constitutes the cornerstone for [21, 23, 24]). At the end of this section, the case N odd
the topological features of the SSH model. For periodic will be discussed in much less detail.
dimers coupled by t1 , the topological winding number The Schroedinger equation for a solution of energy E
(1D Berry phase or Zak phase [20]) is 1 if t1 > t2 but (HSSH − E) |ψi = 0 (2)
zero for t1 < t2 . The reverse is true when considering
dimers coupled by t2 . Hence, the infinite SSH model couples any site with its immediate neighbours. Trans-
with t1 6= t2 exhibits two topologically distinct insulat- lational invariance (by an even number of sites) suggests
ing phases depending on how the dimers are constructed the following ansatz for an eigenstate of HSSH :
[21]. In the presence of kink defects the model exhibits
M −1
soliton solutions at zero energy with charge fractionaliza- X
tion [22]. These zero-energy solutions are exponentially |ψi = (A |2n + 1i + B |2n + 2i) ein2k . (3)
n=0
localized at the defects.
Similar exponentially localized solutions exist also in The middle components (all but the first and last) of (2)
the finite SSH chain, but this time at the edge. The reduce to a pair of equations:
structure of these solutions is summarized in Fig. 1 and
t1 + t2 e−i2k
  
will be discussed in more detail in the next section. In −E A
= 0. (4)
short, for energies near the band center there exists 0, 1 t1 + t2 ei2k −E B
or 2 edge states depending on the parity of the number A nontrivial solution requires that the determinant of the
of sites and the values of the coupling strengths. Outside matrix vanish, giving the following dispersion relation:
the band center, one recovers typical Bloch solutions as
expected for periodic systems (see Fig. 1). E 2 = t21 + t22 + 2 t1 t2 cos 2k. (5)
 3

If (t1 − t2 )2 < E 2 < (t1 + t2 )2 , the wave number k is sign, sin[(N − 2n)k]. This sign turns out to be that of
real (corresponding to bulk states; this range gives the sin(N k) giving, finally,
energy bands in the continuum limit), whereas outside
−1 n
this range it is complex (any such solution being an edge  bulk  MX
state, as we will see). For the moment, we will assume k ψ± = sgn(sN ) sN −2n |2n + 1i
is real. n=0
o
Since k → k + π has no effect on the ansatz, k can ± s2n+2 |2n + 2i . (11)
be taken to be in the range (−π/2, π/2]. For any energy
other than E 2 = (t1 ± t2 )2 (a case which can safely be
Let us return now to the solution of (9) for k. Recall
ignored), there are two solutions to (5), which we will
that, if real, k is in the range [0, π/2]. It can be shown
write ±k, where 0 < k < π/2.
that the endpoints (which correspond to E 2 = (t1 ± t2 )2 )
It is useful to write t1 + t2 e±i2k = |E|e±i2ϕ ; like k, can be excluded. Eq. (9) cannot be solved analytically
0 < ϕ < π/2. Then the solution to (4) can be written for k, but it can be solved graphically. The number of
   −iϕ  solutions in the range (0, π/2) depends, naturally, on the
A e
= , (6) number of sites. More surprisingly, it also depends on the
B ±eiϕ
parameter r. This parameter has a critical value given
by [23]
where here and in what follows the upper (lower) sign is
for the solution of positive (negative) energy. The most N
general solution to the middle components of (2) is given rC ≡ . (12)
N +2
by a sum of (3) with (6) and the same expression with
(k, ϕ) → (−k, −ϕ): If r > rC , there are N/2 solutions of (9) for k in
(0, π/2). Each pair k, −k corresponds to two energy
M −1 n
X eigenstates of the form (11) with equal and opposite en-
C+ e−iϕ ei2nk + C− eiϕ e−i2nk |2n + 1i


|ψ± i = ergies. Thus, there are a total of N solutions, forming a
n=0
o complete set of solutions of (2). Since all these solutions
± C+ eiϕ ei2nk + C− e−iϕ e−i2nk |2n + 2i


(7) have oscillatory behavior as a function of the site index,
they are bulk states, as was mentioned earlier.
where C± are constants. If, however, r < rC , there is one fewer solution of (9)
The first and last components of (2) give in (0, π/2), giving a total of N − 2 solutions. But HSSH
clearly has N eigenvalues and eigenvectors, so two have
yet to be found. The missing solutions have complex
 iϕ −i2k −iϕ i2k   
e e e e C+
= 0. (8) wave numbers. If we substitute k = π/2 + iκ into (9), we
e−iϕ eiN k eiϕ e−iN k C−
find
As above, a nontrivial solution requires that the determi-
sinh(N κ)
nant vanish, giving the following equation for k, written = r. (13)
in terms of the hopping-parameter ratio r ≡ t1 /t2 and sinh((N + 2)κ)
sj ≡ sin(jk):
This equation cannot be solved analytically, but it is easy
r sN +2 + sN = 0. (9) to see that there are two real, equal and opposite solu-
tions for κ if r < rC and none if r > rC , which is exactly
We will return to the solution of this equation, and the what is needed to make up for the two missing solutions
energy spectrum which follows from (5), shortly. of (2) for k real. Defining κ to be the positive solution,
Solving (8) for C± and substituting in (7) gives the two corresponding solutions of (2), identified with
edge states since they have an exponential nature, turn
 bulk  MX−1 n out to be:
 
ψ± = sin (2n + 2)k − 2ϕ |2n + 1i −1
 E MX 
n=0  edge n
  o ψ± = (−) shN −2n |2n + 1i
± sin (2n + 2)k |2n + 2i , (10) n=0

± sh2n+2 |2n + 2i , (14)
where we have noted explicitly that these states, being
oscillatory, are bulk states. The coefficient of |2n + 1i
can be simplified
 slightly as follows. First, note that (9) where we have written shj = sinh(jκ).
implies sin (N + 2)k − 2ϕ = 0 which, given the range of The wave numbers are determined by (9) and, if r <
ϕ, then implies 2ϕ = (N + 2)k mod π. Now, modding rC , (13). The energies are then determined by (5). The
by π either has no effect on the coefficient or changes it energy spectrum for N = 20 is displayed as a function of r
by a sign, depending on whether the subtraction is an in Fig. 2. (The energy spectrum has appeared in various
even or odd multiple of π, so the coefficient is, up to a forms in the literature; see for instance [21, 23, 25–29].)
 4

Energies (N=20) the band gap (defined in the thermodynamic limit) for
E/t2 r < rC . These are the edge states given by (14).
Describing the states given by (14) as edge states mer-
3
r < rc r > rc its some discussion. On the one hand, the wave number
is complex, so for a sufficiently large system the states
are confined to the edges with a penetration length into
2
the bulk given by l ≡ 1/κ. On the other hand, as r → rC
from below, κ goes to zero and the penetration length
goes to infinity (see Fig. 3). Thus for any finite-size sys-
1
tem and r sufficiently close to rC , the penetration length
is longer than the system size and the state is for all in-
tents and purposes no longer confined to the edges, ren-
r
0.5 1 1.5 2 dering it relatively indistinguishable from the rest of the
states.
rc An approximate analytic solution to (13) can be given
-1
if N is large and/or r is small. One finds

| log r| rN (1 − r2 )
-2 κ= − + O(r2N ). (15)
2 2
The first term becomes dominant rapidly as either N gets
-3 0.8 0.9 1 large or r gets small; in Fig. 3 only that term is included
in the analytic curves.
FIG. 2: Energy spectrum of the SSH hamiltonian as a func- We can derive an approximate analytic expression for
tion of the hopping-parameter ratio r = t1 /t2 for N = 20. the edge-state energies by substituting k = π/2 + iκ into
The solid lines are bulk states for all r; the dashed lines are (5), with κ given by (15). The dominant term in (15)
states which are bulk (edge) states for r > rC (r < rC ). The gives E = 0 (and indeed these states are often erro-
grey areas are the energy bands in the thermodynamic limit neously described as zero-energy states); the energies are
(N → ∞). Notice (see inset) that the edge states are exactly dominated by the second term, giving
at the band edges for r = rC .
E = ±t2 rN/2 (1 − r2 ) + O(rN ), (16)

showing the well-known exponentially decreasing behav-

κ l ior of the energies as a function of N [21, 23].
1.5 N=4 30
Bulk state (N=20)
N=8 |ψ n >

1 N=20 20 0.3

Analytic 0.2 (r,E/t2 )

(1.4,0.712)
κ 0.1
(0.91,0.415)
0.5 10 n (0.4,0.636)
5 10 15 20
-0.1
l
-0.2
0 0 -0.3
0 0.2 0.4 0.6 0.8 1
r
FIG. 4: (color online) Example of a bulk state (given by (11)),
for N = 20 and three values of the hopping-parameter ratio
FIG. 3: (color online) Decay parameter κ and penetration
r, the middle of which is rC . The state chosen is the low-
length l = 1/κ for edge states as determined by (13), dis-
est positive-energy solid line in Fig. 2. Key observation: all
played as a function of the hopping-parameter ratio r = t1 /t2
states are large throughout the system, as expected given the
for several values of N . There is no solution (and therefore
oscillatory nature of the solution (11).
there are no edge states) if r > rC = N/(N + 2) (vertical
broken lines). Also shown (dashed lines) is the analytic ap-
proximation given in (15); only the N -independent first term The distinction between bulk and edge states, and just
is included. This is also the penetration length for any N odd how “edgy” the edge states are, are illustrated Figs. 4 and
(see (18)). 5, showing respectively a bulk state for all r and a state
whose nature (edge vs. bulk) changes at r = rC .
We conclude this section with a brief discussion of the
The most striking feature is the existence of states in case N odd (writing N = 2M + 1). Much of the above
 5

Edge state (N=20) (unnormalized) zero-energy state is

|ψ n >
M
0.6
X
|ψ0 i = (−r)n |2n + 1i . (17)
0.4 (r,E/t2 ) n=0
(1.4,0.49)
0.2 Clearly if r > 1 the state grows exponentially from left
(0.91,0.091)
n (0.4,0.00009) to right with the opposite conclusion if r < 1, with pen-
5 10 15 20
-0.2 etration length
-0.4 l = 2/| log r| (N odd) (18)
-0.6
Unlike the case N even, for which the penetration
length is given by the inverse of (15), for N odd the pen-
FIG. 5: (color online) Example of a state that transitions to etration length is independent of N . The two agree in
an edge state (given by (14) below rC ) as r is decreased, for the large-N limit, as expected. The energy spectrum as
N = 20 and three values of the hopping-parameter ratio r, a function of r and the edge state for two representative
the middle of which is rC . The state chosen is the positive- values of r are displayed in Figs. 6 and 7.
energy dashed line in Fig. 2. Key observation: the first two
states are described by (11) and are large throughout the Edge state (N=21)
system, whereas the third is described by (14) and decreases |ψ n >
exponentially in the bulk. 1.0
r=2.0
Energies (N=21) r=0.5
E/t2 0.5

3
r<1 r>1
n
5 10 15 20
2
-0.5

1
FIG. 7: (color online) Zero-energy edge state (right for r > 1,
left for r < 1) for N = 21. The dotted lines are the expo-
nential envelope functions ± exp[(n/2) log r] (up to normal-
r ization).
0.5 1.0 1.5 2.0

Having established the main properties of the SSH

-1 chain, we now consider the coupling of a qubit to the
SSH chain.
-2
III. SSH CHAIN COUPLED TO SEMI-INFINITE
LEAD AND QUBIT
-3
In this section we will consider the full system consist-
FIG. 6: Energy spectrum of the SSH hamiltonian as a func- ing of the SSH chain coupled on the left to a semi-infinite
tion of the hopping-parameter ratio r = t1 /t2 for N = 21. lead and on the right to a qubit. Since we will ultimately
The solid lines are bulk states for all r; the dashed line is the incorporate the effects of the SSH chain and semi-infinite
zero-mode which is a right (left) edge state for r > 1 (r < 1). lead into an effective qubit Hamiltonian, in the next sub-
The grey areas are the energy bands in the thermodynamic
section we will quickly review the relevant features of the
limit (N → ∞).
qubit. We will then study the full system (depicted in
the top panel of Fig. 10) in the following subsection.
analysis applies with only slight modification. It turns
out that for all values of r there are N −1 bulk states with A. Isolated qubit
equal and opposite energies and one zero-energy state.
The latter is localized on the left (right) edge for r < 1
The isolated qubit, or double dot (see Fig. 8), is de-
(r > 1). This state is easily constructed by noting that
scribed by the Hamiltonian
(2) with E = 0 decouples into two sets of equations: one  
for the coefficients of odd sites and one for those of even DD 2 τ
H = . (19)
sites. The latter are zero, and it is easy to show that the τ 1
 6

SSH chain which is connected at the other end to a semi-

infinite lead. The Hamiltonian is
H∞ W 0
 

H =  W † HSSH VN  (23)
FIG. 8: Isolated Qubit
0 VN† H DD

where H DD and HSSH are defined in (19) and (1), VN is a

The (energy-dependent) Green’s function, defined by
N ×2 matrix whose only nonzero element is (VN )N 1 = tC ,
GDD = (E − H DD )−1 , is easily calculated; for instance,
W is an ∞ × N matrix whose only nonzero element is
its (1,2) component is
W∞1 = tL , and H∞ is
 
DD τ 1 1  
G12 (E) = − , (20) 0 1
δ E − λ+ + i0+ E − λ− + i0+  
1 0 1 
where H∞ =  , (24)
 
.. 
q

 1 0 .

.. ..
p
δ = (1 − 2 )2 + 4τ 2 = δ0 2 + 4τ 2 , . .
1 δ
λ± = (1 + 2 ± δ) = 0 ± (21) The dynamics of the entire system can be determined
2 2
by evaluating its Green’s function. But since we are in-
are the energy splitting and the energies of the full Hamil- terested only in that of the double dot, we can incorpo-
tonian, respectively, and we have defined the uncoupled rate the effect of the SSH chain and lead in a self-energy
(τ = 0) energy splitting δ0 = 1 − 2 and the average using standard techniques (see for instance [30] for a gen-
energy 0 = (1 + 2 )/2. eral discussion; we will adapt the analysis of [4], following
The infinitesimal positive quantity 0+ in (20) gives the the notation introduced there, to the current system).
pole prescription necessary to compute the retarded time- It is useful to do this in two steps. First, the effect
dependent Green’s function. This is obtained by Fourier of the lead on the SSH chain can be incorporated by
transformation, giving zero for t < 0 while for t > 0 replacing the lead by a modification of HSSH :
GDD (t) consists of two terms oscillating at frequencies  
λ± ; for instance, Σ∞ t1 0 · · · 0
.
 t1 0 t2 . . . .. 
Z ∞ 
!
GDD (t) = dE e−iEt GDD
12 (E) H∞ W
 
12 →
 .. .. 
−∞ †
 0 t2 . . 0 
W HSSH 
 . .

2πiτ −iλ+ t  .. .
. . . . 0 t

− e−iλ− t


=− e
δ
4πτ −i0 t 0 ··· 0 t 0
=− e sin(δt/2). (22)
δ ≡ HSSH,∞ . (25)
For an isolated qubit the time dependent solutions are Thus, HSSH,∞ is identical to HSSH except for the upper-
simply periodic. However, if the qubit is coupled to an ex- left element, which is a self-energy proportional to the
ternal infinite system (for example, a semi-infinte lead), surface Green’s function of the lead [4]:
the oscillations will decay. The decoherence of a qubit
can then be evaluated by evaluating the off-diagonal el- tL 2  p 
Σ∞ = tL 2 GS∞ (E) = E − i 4 − E2 . (26)
ement of the Green’s function [4]. The coupling between 2
the qubit and external system can take on different forms.
Here we restrict ourselves to the simplest case, where the The second step repeats the above, incorporating the
qubit is a double dot connected to an external system via effect of HSSH,∞ on the double dot by an appropriate
a small tunneling coupling. modification of H DD :
! !
HSSH,∞ VN 2 + ΣSSH,∞ τ
→
B. Double dot coupled to SSH chain and VN† H DD τ 1
semi-infinite lead DD
≡ HSSH,∞ . (27)

As was mentioned in the introduction, our main inter- As above, the only modification of H DD needed is a term
est is to investigate how the dynamics of a double dot added to the upper-left element. This added term is pro-
is affected by the presence of edge states in an adjoin- portional to the surface Green’s function of HSSH,∞ :
ing system. We therefore consider a system composed
of three parts: the double dot coupled to one end of an ΣSSH,∞ = tC 2 GSSSH,∞ (E) (28)
 7

The surface Green’s function GSSSH,∞ is the (N, N ) com- It is useful to define
ponent of the Green’s function of HSSH,∞ defined by q
δ 0 = δ|2 →0 = (1 − 02 )2 + 4τ 2 (33)
2
GSSH,∞ (E − HSSH,∞ ) = 1. (29) 1
λ0± = λ± |2 →0 = 0 + (ΣSSH,∞ ± δ 0 ). (34)
2 2
To determine GSSSH,∞ ,
we write the first row of GSSH,∞
in a form identical to (3) and follow the steps used to Since HSSH,∞ is not Hermitian (reflecting the fact
derive (10). (The current calculation is somewhat eas- that the double dot is not a closed system), the time-
ier because we only need the last component.) Here as dependent Green’s function will have decaying behavior
above, the cases N even and odd must be handled differ- (in contrast with the oscillatory behavior exhibited in
ently. We find (22)) from which the decoherence rate can be extracted
by determining the slowest decay. This determines the
Et2 sN − Σ∞ (t1 sN −2 + t2 sN ) long-term behavior of the double dot.
GSSSH,∞ = (30)
2
t2 (t1 sN +2 + t2 sN ) − Et2 Σ∞ sN However, the time-dependent Green’s function cannot
be evaluated exactly since the very complicated depen-
if N is even and dence of ΣSSH,∞ on E precludes an exact evaluation of
the Fourier transform of GSSSH,∞ .
t2 (t2 sN −1 + t1 sN +1 ) − EΣ∞ sN −1 An analytic approximation can be obtained by noting
GSSSH,∞ = (31)
t1 t2 EsN +1 − t1 Σ∞ (t2 sN +1 + t1 sN −1 ) that the frequencies in the isolated double dot Green’s
function GDD are λ± . Eq. (34) suggests using λ0± instead.
if N is odd. This is not quite correct, since λ0± are energy-dependent.
The surface Green’s functions is shown in Fig. 9 for However, for small coupling between the double dot and
the case where N is odd, where there is an edge state the rest of the system, one can show [4] that to a good
on the right of the SSH chain. This corresponds to case approximation the (now complex) frequencies should be
(d) shown in Fig. 1. The striking feature is the large evaluated at the corresponding poles of the isolated dou-
negative imaginary part of the Green’s function at zero ble dot Green’s function, λ0+ (λ+ ) and λ0− (λ− ). According
energy, corresponding to the large local density of states to this analytic approximation, the decay rates are given
of the right edge state. This large negative imaginary by the imaginary part of the frequencies, and we conclude
part of the surface Green’s function is the main source of that the decoherence time τφ is given by
decoherence for the double dot coupled to the rightmost  
site of the SSH chain. −1 1 0
(τφ ) ≈ min − = {ΣSSH,∞ (λ± ) ± δ (λ± )} . (35)
2
This expression makes it clear that there are two contrib-
utors to the decay rate, corresponding to the coupling of
each double dot state to the SSH chain. The slower of
the two rates dominates at long times, so it is this that
−1
determines (τφ ) .
Alternatively, one can determine the decoherence time
by numerically evaluating the Fourier transform of the
Green’s function and extracting the decay constant of the
long-time behavior. All matrix elements will decay with
the same rate. Here we use the off-diagonal element to
compute it. For the Fourier transform, it is important to
use a very small discretization of the energy and we used
∆E ∼ 10−5 (for a bandwidth of 4). The time dependence
FIG. 9: (color online) Line plots of the surface Green’s func- is then evaluated using a fast Fourier transform, which
tion GSSSH,∞ (31) of the right most site as a function of energy, is fitted to multiple exponential decay functions, from
while the dots are obtained by evaluating numerically the full which the slowest decay is extracted at long times.
matrix (25). Here N = 11 and t1 > t2 were used, which
Both methods will be used in what follows; the excel-
corresponds to case (c) in Fig. 9.
lent agreement between the analytic approximation and
numerical evaluation of the decoherence time is a con-
The energy-dependent Green’s function for the dou- vincing post hoc justification of the analytic approxima-
ble dot including the effect of the SSH chain and lead is tion (see Figs. 10 and 11). The parameters in these fig-
obtained in a straightforward manner by the substitution ures were chosen so that one double dot energy is zero
(so that it couples to any SSH edge states, whose ener-
2 → 2 + ΣSSH,∞ ≡ 02 (32) gies are also zero or exponentially small), while the other
double dot energy lies in the SSH continuum. The latter
in (20) (and similar equations for the other components). gives rise to rapid decoherence, so the overall decoherence
 8

is determined by whether or not the zero-energy state right edge of the SSH chain (with coupling tC ), the qubit
also decoheres rapidly. In the discussion that follows, will decohere much faster when there is an edge state (N
we therefore focus on the decoherence of the zero-energy odd), while the decoherence rate will be exponentially
state; we will see that this depends strongly on whether suppressed for the gapped N even case. This is what we
or not there is a same-side edge state. see in Fig. 10. Already for N = 80, with the choice of
There are some small deviations between the numerics parameters given in Fig. 10, we see approximately a five-
and analytical solution (35). Most of the small differences decade difference in the decoherence rate between the
can be attributed to extracting numerically the decay odd (one edge state) and the even (no edge states) cases.
rates from a finite time interval of a strongly oscillating Hence, the qubit can act as a very sensitive detector of
function. the edge state.
While we have shown that the coupling of the qubit
to the SSH chain affects its dynamics, the reverse is also
true. The topological nature of the edge states in the
SSH chain is also perturbed. However, the coupling of
the qubit induces a perturbation of the SSH states of
order t2c , which can be neglected for tc  1, which is the
situation we are considering here.

FIG. 11: (color online) The decoherence rate for even/odd

number of sites. The analytical expression uses (30,31) and
(35), while the numerical result evaluates the chain numeri-
cally and performs the Fourier transform numerically and fits
the time decay. The parameters used here are the same as in
Fig. 10 except for t2 = 1.1 and t1 = 1/t2 .

Rather different behavior occurs when r < rC . In this

case, when N is odd there is an edge state on the left
edge of the SSH chain, while for N even there are two
edge states, each confined to both edges. If we look at the
FIG. 10: (color online) Top: the schematics of the model, decoherence rate of the qubit attached to the rightmost
where the SSH chain is coupled to an infinite lead on one site (see Fig. 11), we observe an exponentially decaying
side and to a qubit on the other side. (The example shown rate for N odd, but not for N even. The even case is
has N odd, so the rightmost SSH hopping parameter is t2 ; quite intuitive, since the existence of edge states on both
if N is even, it would be t1 .) Bottom: The decoherence rate
sides of the SSH chain will naturally lead to an increased
for site number N even (red) vs. odd (blue). The analyti-
cal expression uses (30,31) and (35), while the numerical re- decoherence of the qubit. The odd case is different since
sult evaluates the chain numerically and performs the Fourier there is no edge state to the right. However, despite
transform numerically and fits the time decay. The param- having an edge state next to the lead, the decoherence
eters used here are: tC = .035, τ = .03, tL = .65, t1 = 1.1 rate is exponentially suppressed with N . This is because
and t2 = 1/t1 . The double dot energies are 1 = 0.4022 and the local density of states at zero energy is close to zero
2 = 0.0022, chosen to give one zero eigenvalue for the iso- at the right edge where the qubit is located, and therefore
lated double dot: λ− = 0. The nonzero double dot energy is it does not lead to an increase in decoherence.
λ+ = 0.4044, lying in the SSH energy band. For all cases, the system displays oscillatory behavior
of the decoherence rate as a function of N at small enough
We have seen above that there is an edge state of en- N . This is a consequence of resonances within the SSH
ergy zero at the right edge of the SSH chain whenever chain: for values of N for which an SSH energy is close
r > 1 for an odd number of sites, while if r > rC (a to λ+ , the corresponding double dot state couples more
weaker condition) there are no states near E = 0 (a gap) strongly to the lead via the SSH chain, causing a higher
for an even number of sites. Both conditions are satis- decoherence rate compared to other values of N . When
fied if r > 1, in which case for a qubit coupled to the reducing tL , the coupling to the lead, these resonances
 9

become more pronounced. For larger N this behavior

is then taken over by exponential decay as a function of
N when there is no edge state (r > rC and N even) or
when the edge state is on the opposite side of the qubit
(r < rC and N odd). This change of behavior can be
understood mathematically from expression (35). For N
small, the minimum is given by the expression evaluated
at the double dot eigenvalue within the SSH band (λ+ ),
while for N large enough and when there are no edge
states (N even) or an edge state opposite to the qubit (N
odd) the minimum is given by the expression evaluated FIG. 12: (color online) The transmission through a SSH chain
at the double dot eigenvalue at zero energy (λ− = 0). coupled to 2 semi-infinite leads on both sides. Left: Trans-
Alternatively, if there are two edge states (N even) or if mission at E = 0 as a function of N . Right: Transmission as
a function of energy. The parameters used here are: tc = .035
there is a single edge state next to the qubit (N odd), the
(the coupling between the SSH chain and the leads). t2 = 1.1
decoherence rate is determined by (35) evaluated at λ+ . and t1 = 1/t2 for the open squares on the left and broken line
Evaluating (35) at λ− = 0 would lead to a divergence on the right. t1 = 1.1 and t2 = 1/t1 for the filled squares on
with large N . Hence the minimum (and the physical the left and solid blue line on the right. For N odd, the cases
decoherence rate) is determined by λ+ . Physically, this t2 = 1.1, t1 = 1/t2 and t1 = 1.1, t2 = 1/t1 have the same
can be understood by identifying the decoherence rate of transmissions (shown in red open and closed circles).
the qubit with the smallest escape rate from the double
dot into the SSH chain.
mission first increases with N , because the edge state
energies converge to zero, before the transmission decays
IV. DISCUSSION AND CONCLUSIONS again (for N & 70 in Fig. 12) due to the localized nature
of the edge states. It is worthwhile noting that the trans-
Interestingly, the exponential decrease in the rate of mission in this case (2 edge states) is still five orders of
decoherence is similar for the case where we have an magnitude larger than in the odd case, which only has
edge state on the opposite side of the qubit (N odd in one edge state. This is a consequence of the symmetry
Fig. 11) and when there are no edge states at all (N of each of the 2 edge states in the even case, which live
even in Fig. 10). In both cases, the decoherence rate at both edges (see figs. 1 and 5). Indeed, the edge state
is exponentially suppressed. This is quite similar to the solutions are either antisymmetric (E > 0) or symmetric
case of localization with a random potential along the (E < 0) in site position, while the edge state in the odd
chain, which also leads to an exponential suppression of case is restricted to either the left or right side of the SSH
the decoherence rate with chain length [4]. From this chain. The takeaway message here is that with a simple
perspective, the case of a localized edge state close to transmission probe it is not possible to distinguish the
the qubit is quite different, since despite it being a lo- left from the right edge state, in stark contrast to our
calized state, the decoherence rate is enhanced by the decoherence probe.
existence of a localized edge state, even though trans- In variants of the SSH model, it is also possible for the
mission [31, 32] through the SSH chain will eventually be localized zero-energy mode to exist at a local topological
suppressed (see Fig. 12). defect (or soliton) when breaking the translational sym-
It is instructive to look at the case of transmission metry of the dimers. For example, instead of alternating
through the SSH chain as a reference. To this end, we t1 and t2 throughout the chain there could be two con-
have determined the transmission numerically using the secutive t1 ’s before the alternation continues. This would
standard non-equilibrium Green’s function method [30]. lead to a similar exponentially localized mode at the de-
As mentioned above, the transmission at the band center fect. If the qubit could be scanned over such a chain, one
will eventually decay with N because of the localized na- could identify the position and properties of the topolog-
ture of the edge solution. This is clearly seen for the N ical defect. This assumes a small local coupling between
odd case shown in Fig. 12. However, in notable contrast the qubit an the SSH chain, as would be the case in a typ-
to the decoherence rate discussed earlier, there is no dif- ical scanning probe experiment. Hence, using a qubit as
ference in the transmission with respect to the location a probe for topological edge states would be very interest-
of the edge solution (left or right). The N even case is ing and could be extended to other toplogical excitations
more interesting, in that there is a significant difference such as Majorana fermions with implications in quantum
between the case where r > rC (no edge states) and that computing [33].
where r < rC (2 edge states), as seen in Fig. 12. In Summarizing, we have shown that a qubit can be used
the former case (no edge states), the transmission sim- as sensitive local detector of topological edge states by
ply decays exponentially with N due to the gap at zero looking at its dynamics, without affecting the topological
energy. This is similar to the decoherence rate shown in nature of the states. This has important implications on
Fig. 10. In the latter case (two edge states), the trans- the experimental detection of topological states as well
 10

as, more generally, for the implementation of topological Québec via the INTRIQ strategic cluster grant. RM is
quantum computation. grateful for the hospitality of Perimeter Institute where
part of this work was carried out. Research at Perime-
ter Institute is supported by the Government of Canada
Acknowledgments through the Department of Innovation, Science and Eco-
nomic Development and by the Province of Ontario
This work was supported in part by the Natural Sci- through the Ministry of Research, Innovation and Sci-
ence and Engineering Research Council of Canada and ence.
by the Fonds de Recherche Nature et Technologies du

[1] A. Steane, Reports on Progress in Physics 61, 117 (1998). [17] W. P. Su, J. R. Schrieffer, and A. J. Heeger, Phys. Rev.
[2] M. A. Nielsen and I. L. Chuang, Quantum Computation Lett. 42, 1698 (1979).
and Quantum Information (Cambridge University Press, [18] R. Peierls, Annalen der Physik 396, 121 (1930).
2000). [19] R. Peierls, More surprises in Theoretical Physics, Prince-
[3] Z. Xu, L. P. Garcı́a-Pintos, A. Chenu, and A. del Campo, ton Series in Physics, Vol. 19 (Princeton University Press,
Phys. Rev. Lett. 122, 014103 (2019). 1991).
[4] H. Eleuch, M. Hilke, and R. MacKenzie, Physical Review [20] J. Zak, Phys. Rev. Lett. 62, 2747 (1989).
A 95, 062114 (2017). [21] J. K. Asbóth, L. Oroszlány, and A. Pályi, A short course
[5] A. J. Heeger, S. Kivelson, J. R. Schrieffer, and W. P. Su, on topological insulators, Lecture Notes in Physics, Vol.
Rev. Mod. Phys. 60, 781 (1988). 919 (Springer International Publishing, 2016).
[6] M. Atala, M. Aidelsburger, J. T. Barreiro, D. Abanin, [22] R. Jackiw and C. Rebbi, Physical Review D 13, 3398
T. Kitagawa, E. Demler, and I. Bloch, Nature Physics (1976).
9, 795 (2013). [23] P. Delplace, D. Ullmo, and G. Montambaux, Phys. Rev.
[7] L. Wang, M. Troyer, and X. Dai, Physical review letters B 84, 195452 (2011).
111, 026802 (2013). [24] C. W. Duncan, P. Öhberg, and M. Valiente, Phys. Rev.
[8] M. Lohse, C. Schweizer, O. Zilberberg, M. Aidelsburger, B 97, 195439 (2018).
and I. Bloch, Nature Physics 12, 350 (2016). [25] L. Fu and C. L. Kane, Phys. Rev. Lett. 100, 096407
[9] S. Nakajima, T. Tomita, S. Taie, T. Ichinose, H. Ozawa, (2008).
L. Wang, M. Troyer, and Y. Takahashi, Nature Physics [26] L. Li, Z. Xu, and S. Chen, Physical Review B 89, 085111
12, 296 (2016). (2014).
[10] M. Leder, C. Grossert, L. Sitta, M. Genske, A. Rosch, [27] X. Gu, S. Chen, and Y.-x. Liu, arXiv:1711.06829 [quant-
and M. Weitz, Nature communications 7, 13112 (2016). ph] (2017).
[11] D. Xie, W. Gou, T. Xiao, B. Gadway, and B. Yan, npj [28] Y. Yao, M. Sato, T. Nakamura, N. Furukawa, and M. Os-
Quantum Information 5, 1 (2019). hikawa, Phys. Rev. B 96, 205424 (2017).
[12] M. Xiao, G. Ma, Z. Yang, P. Sheng, Z. Q. Zhang, and [29] N. Batra and G. Sheet, arXiv:1906.08435 [cond-mat.mes-
C. T. Chan, Nature Physics 11, 240 (2015). hall] (2019).
[13] Y. Peng, C. Qin, D. Zhao, Y. Shen, X. Xu, M. Bao, [30] S. Datta, Quantum transport: atom to transistor (Cam-
H. Jia, and X. Zhu, Nature Communications 7, 13368 bridge University Press, 2005).
(2016). [31] L. Ruocco and A. Gómez-León, Phys. Rev. B 95, 064302
[14] X.-T. He, E.-T. Liang, J.-J. Yuan, H.-Y. Qiu, X.-D. (2017).
Chen, F.-L. Zhao, and J.-W. Dong, Nature Commu- [32] S. Böhling, G. Engelhardt, G. Platero, and G. Schaller,
nications 10, 872 (2019). Phys. Rev. B 98, 035132 (2018).
[15] M. Z. Hasan and C. L. Kane, Rev. Mod. Phys. 82, 3045 [33] M. Freedman, A. Kitaev, M. Larsen, and Z. Wang,
(2010). Bulletin of the American Mathematical Society 40, 31
[16] Y. Tokura, K. Yasuda, and A. Tsukazaki, Nature Re- (2003).
views Physics 1, 126 (2019).