Leggett-Garg Inequality for a Two-Level System under Decoherence: A Broader

Range of Violation
Nasim Shahmansoori1, and Afshin Shafiee1, 2,
1

Research Group on Foundations of Quantum Theory and Information,

Department of Chemistry, Sharif University of Technology P.O.Box 11365-9516, Tehran, Iran
2
School of Physics, Institute for Research in Fundamental Sciences (IPM), P.O.Box 19395-5531 Tehran, Iran

arXiv:1605.00394v2 [quant-ph] 12 May 2016

We consider a macroscopic quantum system in a tilted double-well potential. By solving Hamiltonian equation, we obtain tunneling probabilities which contain oscillation effects. To show how one
can decide between quantum mechanics and the implications of macrorealism assumption, a given
form of Leggett-Garg inequality is used. The violation of this inequality occurs for a broader range
of decoherence effects, compared to previous results obtained for two-level systems.
PACS numbers: 03.65.Xp, 03.65.Ta, 03.65.Yz

INTRODUCTION

Extrapolating the laws of quantum mechanics QM , up to the scale of everyday objects, means that objects composed
of many atoms exist in quantum superpositions of macroscopically distinct states. In 1935, Schrodinger attempted to
demonstrate the limitations of QM using a thought experiment in which a cat is put in a quantum superposition of alive
and dead states [1]. The idea remained theoretical until 1980s, when much progress has been made in demonstrating
the macroscopic quantum behavior of various systems such as superconductors [25], nanoscale magnets [6, 7],
laser-cooled trapped ions [8], photons in a microwave cavity [9] and C60 molecules [10].
A typical double-well potential system provides a unique opportunity to study the fundamental behavior of a
macroscopic quantum system (MQS), such as macroscopic quantum tunneling and quantum coherence. In the context
of a double-well potential, Schrodingers cat describes a state in which one system simultaneously occupies both
wells. There are also studies focused on decoherence effects in double-well potentials. Huang et al. [11] showed that
decoherence due to the interactions of atoms with the electromagnetic vacuum can cause the collapse of Schrodinger
cat-like states. Thermal effects [12] and dissipation [13] constitute other sources of decoherence and can suppress
tunneling between wells [14, 15]. In addition, double-well potential is used to describe some special phenomena like
ammonia filipping. The resulting quantum tunneling have been extensively applied in many branches of physics. For
example, it appears in the dynamics of Bose-Einstein condensates, the recent developments of ion trap technology,
the ultracold trapped atoms theory and its applications [1620].
Such a situation brings in mind the question of how the everyday macroscopic world works. The Leggett-Garg
inequality (LGI) provides a method to investigate the existence of macroscopic coherence, to test the applicability
of QM as we scale from the micro- to the macro-world [21, 22]. In this fashion, we can test the correlations of a
single system measured at different times. Violation of LGI implies either the absence of a realistic description of
the system or the impossibility of measuring the system without disturbing it. QM violates different forms of LGIs
on both accounts. A number of experimental tests and violations of these inequalities have been demonstrated in
recent years [23, 24]. Leggett and Garg initially proposed an rf-SQUID flux qubit as a promising system to test
their inequalities [22], which was later improved by Tesche [25]. The first measured violation of a type of LGI was
reported by Palacious-Laloy and coworkers [26]. Palacios-Laloy et al. found that LGI was violated by their qubit,
albeit with a single data point, with the conclusion that their system could not admit a realistic, non-invasivelymeasurable description. Recently, several experimental tests of LGIs were implemented, all of which confirm the
predicted violations in accordance with the fundamental laws of QM [2633]. Most of these experiments were weak
measurements, where the effects of the measured back-action in a sequential set up are minimized [34].
In this article we calculate the so-called time correlations in a tilted double-well potential. To do this, we consider
the effect of the environment as a perturbation on the system. Then, we use the time correlations to test a given
type of LGI. For a symmetric double-well potential considered as a two-level quantum system undergoing coherent
oscillations between the two states, it has been shown that QM violates different forms of LGI [35]. Morover, no
violation occurs, when strong decoherence is at work. According to our calculations for a tilted double-well potential,
however, it is possible to see the violation, even for significant effects of decoherence.
The structure of our paper is as follows. In section 2, we focus on a tilted double-well model, to introduce its
Hamiltonian, considering the effects of the environment on it. Then we calculate the tunneling probabilities to obtain
time correlations. In section 3, we show the violation of a given LGI under decoherence. Finally, in section 4, we

2
conclude the results.
TILTED DOUBLE-WELL POTENTIAL

We consider a typical tilted double-well, where its asymmetric form is measured by the amount of the parameter
. Here |i (|+i) denotes the state in which the macrosystem is localized in the left (right) well. They also describe
the ground states in the left and right wells, respectively.
The macroscopic feature of the system is identified by dimensionless equations. We define dimensionless variables
S , the potential V (q) and the time t as the following relations
for the momentum p, the position q, the Hamiltonian H
([35], p.18):
q :=

R
;
R0

t :=

tconv
0

S := HSconv ;
H
U0

V (q) :=

U (R)
U0

(1)

where,
0 :=

R0
1 ;
(U0 /M ) 2

U0 = M (R0 /0 )2

(2)

Here R0 and U0 are the characteristic length and the characteristic energy of the system, respectively. Also, 0 is the
characteristic time, estimated by the time needed for a particle of mass M to pass the distance R0 with kinetic energy
instead of
of the order of U0 . Regarding the relations (1) and (2), one can define a new dimensionless parameter h
planks constant in units of action U0 0 :
=
h

~
P0 R 0

(3)

quantitatively shows the macroscopic behavior of the system. So that, for smaller values of h,
the
The new constant h

situation is more quasi-classical. Yet, to detect the quantum tunneling effect, h shouldnt be too small. Considering
is about 0.1 to support the macroscopic quantum trait
a tilted double-well potential, we assume that the value of h
of the system, in a quasi-classical situation.
When the macrosystem is isolated from its environment, it can be described effectively by the following Hamiltonian:
 
h
(4)
H=
2
is a measure of the tilt , and is a measure of the strength of the tunneling between the
where = (E E+ )/h
1

two wells. The eigenvalues of the Hamiltonian (4) are h2 ( 2 + 2 ) 2 and the eigenstates of this Hamiltonian are
obtained as:
1
1
1
|0i = ( 2
) 2 (|+i + B|i) B = (2 + 2 ) 2 +
(5a)
+ B2
1
1
1
|1i = ( 2
) 2 (|+i A|i) A = (2 + 2 ) 2
(5b)
+ A2
For our next purposes, we refine the eigenstates (5a) and (5b) as the following:

r
where sin =

B
and cos =
A+B

|0i = cos |+i + sin |i

(6a)

|1i = sin |+i + cos |i

(6b)

A
. Also we define
A+B
|+i = a|0i + b|1i
0

|i = a |0i + b |1i

(7a)
(7b)

where a = cos , b = sin , a0 = sin and b0 = cos . One can show that the probability of the tunneling from the
left to the right well is
P+ =

1 t
2
sin2 [(2 + 2 ) 2 ]
2
2
+
2

(8)

3
and contains oscillation effects. Nevertheless, to deal with real systems, the inevitable effects
which is independent of h
of the environment should be considered. So, in order to retain oscillation effects and therefore the macroscopic
quantum coherence, we consider the effects of the environment as a kind of perturbation on the system. For the
Hamiltonian of the system, Hs we have
Hs |ni = En |ni;

E0 < E1 < E2 < ...

(9)

where {|ni} denotes the complete set of eigenstates of Hs with n = 0, 1, 2, ... and En is the energy of the system. We
also define |i and as the energy eigenstates and the energy eigenvalues of the environment, respectivley. Here,
H |i = |i

H |vaci = 0

(10)

and |vaci is the vacuum state. The state of the entire system could be written as |n, ii |ni|i. Apparently, the
is |vaci and |i = b |vaci is the state with a
environment is assumed to be a bosonic field. The ground state of H

single boson . The state |n, vacii is an eigenstate of H0 = Hs + H with energy En .

s between the system
We define En as the related shift due to the perturbation of the interaction Hamiltonian H
and its environment. Then, the stationary perturbation theory gives:
En ' En(1) +

0
X
vacii|2
|hhm, |H|n,
m,

En (Em + )

(11)

(1)

s |n, vacii is the contribution of the first order perturbation. Considering the frequency
where En = hhn, vac|H
distribution of the environmental oscillators J(), one can define (11) as ([35], Ch.6):
Z
d J()
1X
2
|fmn | mn P
En =
(12)
m
+ mn
0
Similarly, we have:
n =

2X
|fmn |2 J(nm )(nm)

(13)

where 1
n is the life time of the shifted energy En + En . The symbol P in relation (12) denotes the principal
value and (nm ) is a step function which indicates that the macrosystem initially in an excited state is allowed to
make transition to the lower states only. We also define fmn = hm|f (q)|ni where f (q) is an arbitrary function of q,
Here,
depending on how the macrosystem exerts force on the environmental oscillators and nm = (En Em )/h.
10 = . Supposing that the macrosystem is initially in the state |, vacii, we investigate the time evolution of the
entire system with perturbation theory, which indicates the preservation of the macroscopic quantum coherence.
To do so, we are going to calculate the probability of finding the macrosystem in each well. This could be defined
as
P+ = |h+|(t)ii|2
where |(t)ii, is the quantum state of the entire system at time t:
X

|(t)ii =
|nihn|eiH t/h U
I |(0)i

(14)

(15)

Here, UI is the time-evolution operator in the interaction picture, given by

I (t) = exp(iH
0 t/h)exp(i
h)
U
Ht/
The relation (15) could be written in the following form:
X

|(t)ii =
eiEn t/h |ni|n (t)i

(16)

(17)

I |(0)i. Hence, we have

where n = and |n (t)i = hn|eiH t/h U
P+ = h+ (t)|+ (t)i

(18)

4
The time evolution operator UI could be expanded up to the second order with respect to the interaction Hamiltonian
Hs as:
Z t
Z t
Z t2
s (t1 ) 1
s (t2 )H
s (t1 )
I (t) ' 1 i
dt1 H
dt
dt1 H
(19)
U
2
0
2 0
h
h
0
s (t) = eiH 0 t/h H
s eiH 0 t/h . In (19), U
I (t) contains the following terms:
where H
vac (t) = i
U

Z
0

1 X
V (t1 )dt1

2h

t2

dt2
0

dt1 f (t2 )ei(t2 t1 ) f (t1 )

(20)

Z t
i

dt ei t f (t1 )
U (t) =
0 1
2h
Z t
Z t2
1

dt
dt1 f (t2 )ei t2 +i t1 f (t1 )
U (t) =
0 2 0
2h

(21)
(22)

1P 2
(f (
q (t)))2 , is the frequency of the particle in the environment. Assuming that the
2
environmental oscillator is displaced by f (q), we use the separable model in which f (q) is independent of ,
f (q) = f (q), where is a positive constant. All time-operators V (t), f (t), ... in (20) to (22) are also defined in
the interaction picture.
Using the relations (20)-(22) one can show that:
where V (t) =

I (t)|(0)ii
|+ (t)i = h+|eiH t/h U
X
vac |s (0)i +
|s (0)i
= |vacih+|U
ei t |ih+|U

i( + )t

(t)|s (0)i
|ih+|U

(23)

If s (0) = |i, we get;

2

vac |0i|2 + a0 b0 |h1|U

vac |1i|2
P+ = h+ |+ i = a2 b2 |h0|U
vac |0i h1|U
vac |1i + a0 2 b2 |h1|U
|0i|2
+2aa0 bb0 <h0|U

(24)

where < denotes the real part. We have also used the relations (7a) and (7b) for the states |i.
In the symmetric double-well the following two assumptions could be considered:
vac |ni = 0, when m n is odd.
A1: The potential term V (q) and so V (q) are even. Then, all elements hm|U

A2: The function f (q) is odd, so all elements hm|U |ni = 0, when m n is even.
vac |1i, h0|U
|0i and h1|U
|1i should be
In the tilted double-well, also, our calculations show that the elements h0|U
zero again, similar to what is resulted from A1 and A2 for a symmetric model. The detailed results are given in
Appendix A. There are also some other assumptions, appropriate in our case:
' 0.
A3. The higher orders of f2 can be neglected, so U
A4. The frequency distribution of the environment can be supposed as ohmic. This means that J() = where
and 10 = .
is a measure of the strength of the interaction between the macrosystem and the environment (  h)
2
2

As a cosequence, in (13) 1 = (2/h)|f01 | J() ' (2/h)|f01 |  . Also 0 = 0. So all the terms with 1 / are
negligible, in our calculations.
A5. The distribution J(mn ) is always positive. Thus J(0) = 0 and J() = 0 where 10 = .
With all these assumptions in mind, the tunneling probability can be obtained as (see Appendix B):
10 t)e1 t/2
P+ = sin2 + (sin2 cos 2)e1 t 2 sin2 cos2 cos(

(25)

This result shows that there is a decay factor e1 t/2 in the term containing cos(
10 = (E1 E0 )h.
10 t)
where
which reduces the strength of the oscillation due to the decoherence (dissipation) effects. In order to diminish the
effect of e1 t/2 , we consider the principal time domain, which requires that 1 t  1 [35].
With the same appraoch, we can calculate other probabilities. For example, when the macrosystem is in the state

5
|+i initially, the probability that it will be found in the state |i at time t is denoted by P+ . Taking into account
the other probabilities P++ and P , one can show that:
10 t)e1 t/2
P+ = cos2 cos2 cos 2e1 t 2 sin2 cos2 cos(
10 t)e1 t/2
P++ = cos2 sin2 cos 2e1 t + 2 sin2 cos2 cos(
2

P = sin + (cos cos 2)e

1 t

10 t)e1 t/2
+ 2 sin cos cos(
2

(26)
(27)
(28)

VIOLATION OF LEGGETT-GARG INEQUALITY UNDER DECOHERENCE

In order to show the detection of a cat state, the experimental results in question should be compatible with QM
but incompatible with macrorealism (M R). The assumption of M R demands that, first, one can assign definite
states to a macrosystem, so that it could be actually in one of these states independent of any observation. Second,
it requires the non-invasive measurability of such macrostates which should not be affected, when they are measured.
LGI serves to examine quantitatively whether the theories satisfying M R are compatible with QM or not. For this,
we use the following LGI:
K1 |C32 C31 | + C21 1

(29)

where the time-correlation function for the two-value variables r and q (r, q = 1) at three moments of time t3 > t2 > t1
is defined as the following for the time sequences (i, j) = (3, 2), (3, 1), (2, 1):
X
Cij =
rqPrti ,qtj
(30)
r,q=1

For the symmetric double well potential and any other two-level system studied, these calculations show a maximum
violation of K = 3/2, when the effect of decoherence is negligible [22]. Now, let us assume that:
t3 t2 = t2 t1 =

10

1
= ,

10

z = e

(31)

Then, the estimation of a maximum value of that violates LGI gives = 0.31 [35, Ch.9]. We also choose = , so
3
10 (t2 t1 ) = 1 . Then, we have:
that cos
2
K1 = |P+t3 |+t2 P+t2 + Pt3 |t2 Pt2 P+t3 |t2 Pt2
Pt3 |+t2 P+t2 (P+t3 |+t1 P+t1 + Pt3 |t1 Pt1
P+t3 |t1 Pt1 Pt3 |+t1 P+t1 )| + P+t2 |+t1 P+t1
+Pt2 |t1 Pt1 P+t2 |t1 Pt1 Pt2 |+t1 P+t1

(32)

where, e.g., P+t3 |+t2 = P+t2 +t3 is the conditional probability that when the macrosystem is in the state |+i at t2 ,
it can be found in |+i at t3 (t3 > t2 ). Generally, we have Prti ,qtj = Pqtj |rti Prti due to Bayesian rule and Prti is the
single variable probability for the system being in the state |ri at ti (i = 1, 2, 3). Conditional probabilities are given in
relations (25) to (28), albeit without time labling. Let us suppose that the macrosystem is initially in the state |i,
so that P+t1 = 0. Accordingly P+t2 is obtained from the following relation:
P+t2 = P+t2 |+t1 P+t1 + P+t2 |t1 Pt1 = P+t2 |t1

(33)

Having into account the above considerations and using the relations (25) to (28), one can find that:
1

K1 = |(sin2 cos2 + 2 cos2 cos 2z + 2 sin2 cos2 z 2 )

1

(sin2 + cos2 cos 2z + sin2 cos2 z 2 )

1

+(cos2 sin2 2 sin2 cos 2z + 2 sin2 cos2 z 2 )

1

(cos2 cos2 cos 2z sin2 cos2 z 2 )

(sin2 cos2 + 2 cos2 cos 2z 2 2 sin2 cos2 z)|
1

+ sin2 cos2 + 2 cos2 cos 2z + 2 sin2 cos2 z 2

(34)

6
If we consider sin2 = 0.2 and cos2 = 0.8, at z = 1 the inequality is violated, maximally. This situation is analogous
to negligible decoherence. Generally, the important result is that for 0.5 < z < 1, the inequality will be violated. This
yields 0 < < 0.66 which shows a broader range of violation compared to = 0.31 for the symmetric double well
potential and/or other proposed two-level systems [3537]. In Fig.1. K1 in (34) is plotted against z for = 26.6 . It
is obvious that K1 increases as z increases from 0 to 1. In Fig.2 K1 is plotted against sin2 for z = 0.6 (upper curve),
z = 0.5 (middle curve) and z = 0.4 (lower curve).

CONCLUSION

Considering a macrosystem prepared in a quasi-classical situation described in a tilted double-well potential, we

studied the effect of the environment as a perturbation source. In this regime, the decoherence (dissipative) effects
are reduced according to the so-called principal time domain in which t  1
1 in (13) for n = 1. Calculations of the
tunneling probabilities show that the coherence effects are present, in spite of the interaction with the environment. To
decide between the predictions of QM and the requirements of M R, a type of LGI (K1 1) is examined in (29), when
decoherence is assumed to be present, but not so dominant. The violation of this inequality shows that the quantum
behavior of a macrosystem could be present in more realistic situations. So, the key parameter (characterizing the
effect of dissipation) in (31) is improved from = 0.31 in previous works to = 0.66. This improvement is crucial
for showing the violation of LGIs in proposed future experiments. When the classical trait of the system is increased,
which is illustrated by the larger values of in (31), the assumption of non-invasive measurement becomes more
determinate. This means that time-correlations could be assumed to be achieved by higher time-ordered probabilities
at the macro-level [35]. Due to the quantum calculations, however, this should be denied, since no three-varibale
joint probability could be defined for our model in quantum formalism from which one can obtain two-variable time 1 ) which shows more
correlations. So, for broader ranges of violation due to increased values of , ( 1 (h)
classicality of the system, the violation of Leggett-Garg inequality features the violation of non-invasive measurability
of the system in a more concrete fashion. Yet, it is an open queston, if in practice the same violation could be
approved. If so, a consistency of M R with an invasive-measurement account should be envisaged more seriously.
K1

1.2

violation line
1.0

0.8

0.6

0.4
0.2

0.4

0.6

0.8

FIG. 1: The amount of K1 vs. z for sin2 = 0.2

1.0

7
K1

1.05

violation line
1.00

0.95

0.90

0.85
0.2

0.4

0.6

0.8

1.0

sin2 HL

FIG. 2: The amount of K1 vs. sin2 for three different values of z = 0.6 (upper curve), z = 0.5 (middle curve) and
z = 0.4 (lower curve).

Appendix A

vac |1i and h0|U

|0i here to show that they are approximately equal to zero, even for non-symmetric
We calculate h0|U
vac |1i, we have:
double-well potentials. First, for h|U
vac |1i = i h0|V (t)|1i 1 h0|g|1i
h0|U

2h
h

(A-1)

where g is defined as:

1 X
g=

2h

Z
dt2

t2

dt1 f (t2 )ei(t2 t1 ) f (t1 )

(A-2)

For the first term, one can show that it is equal to:
1 X
1X 2
h0|f2 |1i =
2 h0|f |mihm|f |1i
2
2 ,m=0,1
1X 2
=
(h0|f |0ih0|f |1i + h0|f |1ih1|f |1i)
2
Z
t
1 X 2
d
(f00 .f01 + f01 .f11 ) = (f00 .f01 + f01 .f11 )
J()
=
2

0
Z
10
f01
t f00 + f11 2
d
t f00 + f11

=
f01 10
J() =
.1
10
f01
f01 10

f01 10
0

(A-3)

which is negligible, because 1 t/10  1. The second term is zero, because the following integrals have meaningful
values, only when the terms in denominator are equal to zero (i.e., 10 + = 0), which is impossible since > 0
and 10 > 0, so the entire term vanishes. To show this, we have

1 X

2h
,m

1 X

2h

t2

dt1 h0|f (t2 )|mihm|f (t1 )|1iei (t2 t1 )

dt2
0

0
t

t2

dt1 ei(10 )t1 h0|f |0ih0|f |1iei t2

dt2
0

+ei t1 h0|f |1ih1|f |1iei(10 + )t2

Z
1 X t
1
=
dt2 (
(ei(10 )t2 1)h0|f |0ih0|f |1iei t2

i(10 )
2h
0

1 i t2
1)h0|f |1ih1|f |1iei(10 + )t2
(e
+
i
1 X
1
1
1
=
{
(ei10 t 1) +
(ei t 1)}
{h0|f |0i.h0|f |1i

i(
+

)
i
i
2h
10

10

+h0|f |1ih1|f |1i

1
1
1
{
(ei10 t 1) +
(ei(10 + )t 1)}} ' 0
i i10
i(10 + )

(A-4)

|0i, one can show that

For h0|U
|0i = i
h0|U

2h

Z
0

2
dt1 h0|f (t1 )|0iei t1 = i fmn D1 ( )ei t/2

2h

(A-5)

where
D1 (; t) =

1 sin(t/2)
2 /2

(A-6)

Then
2
|0i|2 = 2t f00
|h0|U

dJ()D2 ()

(A-7)

where
D2 (; t) =

1 sin(t/2) 2
{
}
2t
/2

(A-8)

We work in the principal time domain for which D2 (, t) (). So the relation (A-7) is equal to zero, since J(0) = 0.
|0i could be neglected. The same situation holds true for the element h1|U
|1i with relations similar
So, the term h0|U

to h0|U |0i.
Appendix B

Here, we calculate the term P|+ as an instance. Other probabilities can be obtained in the same way. We need to
calculate some terms at first and then put them in the main formula. To show this, we have:
vac |0i|2 + a02 b02 |h1|U
vac |1i|2
P|+ = a2 b2 |h0|U
2

vac |0i h1|U

vac |1i + a0 b2 |h1|U
|0i|2
+2aa0 bb0 <h0|U

(B-1)

vac |0i, h1|U

vac |1i, h1|U
|0i and h0|U
|1i are calculated in [35]. So, we have
The terms h0|U
vac |0i|2 = 1,
|h0|U

vac |1i|2 = e1 t
|h1|U

|ni = i 2 fmn D1 ( + mn ; t)ei(mn + )t/2

hm|U

2h

(B-2)

(B-3)

9
where D1 (; t) =

1 sin(t/2)
|0i|2 , we have:
. For |h1|U
2 /2
Z
t 2
dJ()D2 ( 10 )
f10

h
0
t 2
= f10
J(10 ) = 1 t ' 1 e1 t

|0i|2 =
|h1|U

(B-4)

where D2 ( 10 ) ( 10 ).
are zero, because there is 1 ratio in all of
All the terms that produced by multiplying the terms containing U
10
them.
There is also one non-zero multiplying term as the following:
10 t)
vac |0i h1|U
vac |1i = e1 t/2 cos(
<h0|U

(B-5)

10 t)e1 t/2
P+ = cos2 cos2 cos 2e1 t 2 sin2 cos2 cos(

(B-6)

Finally, we obtain:

[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]

shahmansoori@ch.sharif.edu
Corresponding Author: shafiee@sharif.edu
E. Schrodinger, Naturwissenschaften 23, 807 (1935).
R. Rouse, S. Han and J.E. Lukens, Phys. Rev. Lett. 75, 1614 (1995).
J. Clarke, A.N. Cleland, M.H. Devoret, D. Esteve and J.M. Martinis, Science 239, 992 (1988).
P. Silvestrini, V.G. Palmieri, B. Ruggiero and M. Russo, Phys. Rev. Lett. 79, 3046 (1997).
Y. Nakamura, Y.A. Pashkin and J.S. Tsai, Nature 398, 786 (1999).
J.R. Friedman, M.P. Sarachik, J. Tejada and R. Ziolo, Phys. Rev. Lett. 76, 3830 (1996).
E. del Barco, J.M. Hernandez, J. Tejada, N. Biskup, R. Achey, I. Rutel, N. Dalal and J. Brooks, Europhys. Lett. 47, 722
(1999).
C. Monroe, D.M. Meekhof, B.E. King and D.J.A. Wineland, Science 272, 1131 (1996).
M. Brune, E. Hagley, J. Dreyer, X. Maitre, A. Maali, C. Wunderlich, J.M. Raimond and S. Haroche, Phys. Rev. Lett. 77,
4887 (1996).
M. Arndt, O. Nairz, J.V. Andreae, C. Keler, G.V.D. Zouw and A. Zeilinger, Nature 401, 680 (1999).
Y. P. Huang and M. G. Moore, Phys. Rev. A 73, 023606 (2006).
L. Pitaevski and S. Stringari, Phys. Rev. Lett. 87, 180402 (2001).
P.J.Y. Louis, P.M.R. Brydon and C.M. Savage, Phys. Rev. A 64, 053613 (2001).
I. Zapata, F. Sols and A. J. Legget, Phys. Rev. A 67, 021603 (2003)
J.P. Paz, S. Habib and W.H. Zurek, Phys. Rev. D 47, 488 (1993).
P. Kumar, M. Ruiz-Altaba and B. Thomas, Phys. Rev. Lett. 24, 2749 (1986)
G.Theocharis, P.G. Kevrekidis, D.J. Frantzeskakis and P. Schmelcher, Phys. Rev. E 74, 056608 (2006)
F. Grossmann, T. Dittrich, P. Jung and P. Hanggi, Phys. Rev. Lett. 67, 516 (1991)
G. Della Valle, M. Omigotti, C. Cianci, V. Foglietti, P. Laporta and S. Longhi, Phys. Rev. Lett. 98, 263601 (2007)
H. Lignier, C. Sias, D. Ciampini, Y. Singh, A. Zenesini, O.Morsch and E. Arimondo, Phys. Rev. Lett. 99, 220403 (2007)
A.J. Leggett, J. Phys.: Condens. Matter 14, R415 (2002).
A.J. Leggett and A.Garg, Phys. Rev. Lett. 54, 857 (1985).
M.E. Goggin, M.P. Almeida, M. Barbieri, B.P. Lanyon, J.L. OBrien, A.G. White and G.J. Pryde, Proc. Natl. Acad.
Sci. USA 108, 1256 (2011).
V. Athalye, S.S. Roy and T.S. Mahesh, Phys. Rev. Lett. 107, 130402 (2011)
C.D. Tesche, Phys. Rev. Lett. 64, 2358 (1990).
A. Palacios-Laloy, F. Mallet, F. Nguyen, P. Bertet, D. Vion, D. Esteve and A.N. Korotkov, Nat. Phys. 6, 442 (2010).
C.G. Knee, S. Simmons, E.M. Gauger, J.J.L. Morton, H. Riemann, N.V. Abrosimov, P. Becker, H.J. Pohl, K.M. Itoh,
M.L.W. Thewalt, G. Andrew, D. Briggs and S.C. Benjamin, Nature Commun. 3, 606 (2012).
A.N. Jordan, A.N. Korotkov and M. Buttiker, Phys. Rev. Lett. 97, 026805 (2006)
N.S. Williams and A.N. Jordan, Phys. Rev. Lett. 100, 026804 (2008)
M.E. Goggin, M.P. Almeida, M. Barbieri, B.P. Lanyon, J.L. OBrien, A.G. White and G.L. Pryde, Acad. Sci. USA 108,
1256 (2011).
A. Fedrizzi, M.P. Almeida, M.A. Broome, A.G. White and M. Barbieri, Phys. Rev. Lett. 106, 200402 (2011).
J. Dressel, C.J. Broadbent, J.C. Howell and A.N. Jordan, Phys. Rev. Lett. 106, 040402 (2011).

10
[33]
[34]
[35]
[36]
[37]

J.S. Xu, C.F. Li, X.B. Zou and G.C. Guo, New Journal of Physics 14, 103022 (2012).
Y. Aharonov, D.Z. Albert and L. Vaidman, Phys. Rev. Lett. 60, 1351 (1988).
S. Takagi, Macroscopic Quantum Tunneling, Cambridge university press, New York (2005).
C. Emary, N. Lambert, F. Nori, Rep. Prog. Phys. 77, 016001 (2014).
C. Emary, Phys. Rev. A 87, 032106 (2013).