Design of coupling for targeting synchronization in time-delayed systems

Dibakar Ghosh1 , Ioan Grosu2 , and Syamal K. Dana3

1
Department of Mathematics, Dinabandhu Andrews College, Garia, Kolkata 700084, India
2
Faculty of Bioengineering, University of Medicine and Pharmacy, “Gr. T. Popa,” Iasi, Romania
3
Indian Institute of Chemical Biology, Council of Scientific and Industrial Research, Jadavpur, Kolkata 700032, India.
(Dated: June 28, 2011)
We report a general method of designing delay coupling for targeting a desired synchronization
state in delay dynamical systems. We are able to target synchronization, antisynchronization, lag-,
anti-lag synchronization, amplitude death and generalized synchronization in mismatched oscilla-
tors. We apply the theory for targeting a type of mixed synchronization where synchronization,
antisynchronization and amplitude death coexist in different pairs of state variables of the coupled
system. By this method, a scaling of the size of an attractor is possible in different synchronization
arXiv:1106.5298v1 [nlin.CD] 27 Jun 2011

regimes. We derive the stability conditions of synchronization based on Kravoskii-Lyapunov function

theory and Hurwitz matrix criterion. We present numerical demonstration using the Mackey-Glass
system and a delay Rössler system.

PACS numbers: 05.45.Xt, 42.65.Sf, 05.45.Pq

I. INTRODUCTION diffusive coupling or a special type of delay coupling [16].

In conventional studies [1] of chaos synchronization,
the coupling configuration is always known a priori, as
examples, diffusive coupling, synaptic coupling. The sta-
bility of synchronization is established above a critical
coupling based on the negative measure of transverse
Lyapunov exponent [2]. As a general condition, a master
stability function [3] is used for determining the critical
coupling strength for synchronization in identical oscilla-
tors. Alternatively, an engineering approach to synchro-
nization is recently attempted [4-9] for known systems
where the coupling is unknown a priori. The coupling is
rather designed that can ensure the asymptotic stability
of the error dynamics of the coupled oscillators. So far a
few specific issues were investigated such as engineering
coherent patterns [5], a desired amplitude dynamics with
a target frequency [6], synchronization in cellular systems
[7] but each uses a separate approach to define the cou-
pling. These studies were mainly restricted to non-delay
systems and made in absence of coupling delay. How-
ever, a design of coupling in the form of delay feedback
[8] was proposed for effective control of synchronization
and a desired desynchronization in non-delay system. A
design of coupling was also proposed [9] for enhancing
anticipation synchronization in presence of coupling de-
lay in non-delay systems. A coupling delay [10] appears
during the transmission of signals or information in many
natural systems such as long axons of neurons [11] and,
also in practical systems like communication channel [12].
This induces additional instability in the process of an
emergent synchronization state. On the other hand, in
physical systems like optoelectronic devices [13], even in
electronic systems, time delay is an inherent property of
the system. As example, consider a time-delayed system
i.e., ẋ = f (x(t), x(t − τ )) where τ is a delay time. This
intrinsic time delay makes a system infinite dimensional
and more complex. Synchronization in delay systems was
thus given a special attention [10, 12,14-16] using delay
Stability of synchronization in delay systems was usually
established using the Krasovskii-Lyapunov function the-
ory [17] that showed dependence of system parameters
and coupling strength. In all these studies, the coupling
configuration was always pre-defined. An adaptive cou-
pling was proposed recently for synchronization [18] in
time-delayed optoelectronic delay feedback loops, where
the coupling configuration was assumed known but the
coupling strength was a priori unknown. They dynam-
ically tracked and estimated the time-varying coupling
strength for compensating against environmental fluctu-
ation to keep it within the critical limit of stable synchro-
nization. The critical coupling was estimated using the
master stability function [3] which was based on assump-
tion of identical systems. But a parameter mismatch is
unaviodable in practical systems. A method of intermit-
tent linear feedback coupling is found [19] to take care of
the instability due to mismatch in delay systems, but the
coupling did not consider the presence of a conduction de-
lay. A delay dynamical system with conduction delay is
of practical utility, particularly, in digital communication
via optical-link [12]. Therefore, a unified approach for a
robust design of delay coupling in the sense of stability
under parameter mismatch and other external perturba-
tion is necessary to enhance the practical use of delay
systems. Here, we tried to address this issue starting
with a known delay system but the coupling is a priori
unknown. We propose a general framework of designing
unidirectional delay coupling that can maintain stability
of synchronization under parameter perturbation and,
the stability is independent of the system parameters.
We target several synchronization scenarios, complete
synchronization (CS), antisynchronization (AS), lag syn-
chronization (LS), anti-lag synchronization (anti-LS) and
generalized synchronization (GS) by simply controlling
an external parameter conveniently inserted in the defi-
nition of the coupling. An open-plus-closed-loop (OPCL)
coupling scheme [20] was used recently that can engi-
 2

neer different desired synchronization states, CS, AS, LS (αij )n×n is a constant matrix called as the scaling matrix.
and GS in nondelay systems. This coupling scheme has The response system after coupling is,
an advantage that it overrules many of the restrictions
valid for other conventional coupling configurations: an- ẋ = f (x, xτ ) + D(x, xτ , y, yτ ) (2)
tiphase is observed in inversion symmetric systems [21]
only, LS in mismatched systems [22], GS in nonidenti- where D is a coupling term to be defined or designed.
cal or largely mismatched systems [23]. We utilize these The error signal of the coupled system is defined by e =
common benefits of the OPCL scheme and extend it to x − αyτc where τc ≥ 0 is the coupling delay. The f (x, xτ )
delay systems in presence of coupling delay. We derive can be written as
the stability condition of synchronization based on the
f (x, xτ ) = f (αyτc + e, αyτ +τc + eτ ) = f (αyτc , αyτ +τc )
Krasovskii-Lyapunov function theory supported by the
Hurwitz matrix criterion [20] when the stability of syn-
chronization becomes independent of the system parame- ∂f ∂f
ters and is valid for mismatched oscillators too. Stability + e+ eτ + · · · (3)
∂(αyτc ) ∂(αyτ +τc )
being free from the system parameters makes the control
of synchronization more viable. By control of synchro- Keeping the first order terms in (3), the error dynamics
nization, we mean a strategy to change one to the other is approximated,
form of synchronization, either smoothly or abruptly,
without loss of synchronization during a transition. This ė = ẋ−αẏτc
control of synchronization is really implemented by in-
serting a scaling parameter in the design of the coupling.
∂f ∂f
Using this parameter, a scaling of the size of a driver at- = f (αyτc , αyτ +τc ) + e+ eτ + D − αẏτc
tractor (amplification or attenuation) is also possible at ∂(αyτc ) ∂(αyτ +τc )
a response system. Additionally, we can target a type
of mixed synchronization (MS) [20] in a response oscilla- = H1 e+H2 eτ
tor when different pairs of state variables of the systems (4)
attain separate form of synchronization, CS and AS. A where
partial amplitude death [24] can also be induced in one of
the state variables of the response system. Interesting to ∂f
D = αẏτc − f (αyτc , αyτ +τc ) + (H1 − )e
note that an amplitude death (AD) [25] can be induced ∂(αyτc )
in the response system by stabilizing its equilibrium ori-
gin. We support the theory of the delay coupling design
∂f
with numerical simulations of Mackey-Glass system [26] +(H2 − )eτ (5)
and a time-delayed Rössler system [27]. ∂(αyτ +τc )
The rest of the paper is organized as follows: the gen-
is now defined as the OPCL delay coupling and H1 , H2
eral theory of the OPCL delay coupling in time-delayed
are two arbitrary constant matrices.
systems is discussed in Section II. The stability condi-
To derive the stability condition of synchronization, we
tion of the desired state of synchrony is derived for mis-
first apply the Krasovskii-Lyapunov function [17],
matched delay systems using Krasovskii-Lyapunov func-
tion theory and Hurwitz matrix criterion. In Section 1
Z 0
III, numerical examples of the Mackey-Glass system and V (t) = eT e + γ eT (θ)e(θ)dθ (6)
the time-delayed Rössler system are described for both 2 −τ

instantaneous and delay coupling. In Section IV, the where γ is a constant. Taking time derivative of (6) and
targeting of the generalized synchronization state is dis- using the error equation (4),
cussed with the example of time-delayed Rössler system.
Results are concluded in Section V. 1 T T T T
V̇ = [(e H1 +eτ H2 )e+(H1 e+H2 eτ )eT ]+γ(eT e−eTτ eτ )
2
II. GENERAL THEORY OF OPCL DELAY
T 
COUPLING 1 T 1
  
=
e 2 (H1 + H1 ) + γIn×n 2 H2 e
1 T
eτ H
2 2 −γIn×n eτ
We consider a drive-response system where the cou-
pling is undirectional. The driver is,
= E T QE
n
ẏ = f (y, yτ ) + ∆f (y, yτ ), y, yτ ∈ R (1) (7)
where
where ∆f (y, yτ ) is the mismatch parameter and yτ =  1 T 1
  
y(t−τ ). It drives a response system ẋ = f (x, xτ ), x, xτ ∈ e 2 (H1 + H1 ) + γIn×n 2 H2
E= , Q= 1 T
Rn , to induce a desired goal dynamics g(t) = αy(t); α = eτ 2 H2 −γIn×n
 3

Now we provide a proof that V̇ < 0 if Q < 0: (a) (b) (c)

Let Q and R be two symmetric matrices and a matrix 0 2 2

y,x,x−y

y,x,x+y
S has a suitable dimension, then 1

2
−2 0

H
 
Q S 0
<0
ST R −4 −1 −2
−4 −2 0 0 25 50 0 25 50
T H1 t t
if and only if both R < 0 and Q − SR S < 0.
−1

From the above results we see that V̇ < 0 if (d) (e) (f)
4 2 2
H1T ) γH2 H2T

y,x,x−2y

y,x,x+2y
H = 2(H1 + + 4γIn×n + <0 (8) 2 0

y,x
0
The asymptotic stability of synchronization is ensured 0 −2
provided the elements of the H1 and H2 matrices are so −2 −4 −2
appropriately chosen that the condition (8) is satisfied. 0 25
t
50 0 25
t
50 0 25
t
50

We construct H1 and H2 from the Jacobian matrices of

∂f
the system, ∂(αy τc )
and ∂(αy∂f
τ +τc )
. If the elements of
the Jacobian matrices contain state variables, we replace FIG. 1: (Color online) Synchronization region in the H1 − H2
them by constants hi keeping other elements (constants plane in black in (a) for Mackey-Glass system. For τc = 0, a
or zeros) unchanged. On the other hand if there is no pair of time series of driver (online blue) and response (online
state variable in the Jacobian, we replace some of the red) and, the synchronization error in black lines are shown,
zero elements by constants pi , keeping other constants as (b) CS for α = 1, (c) AS for α = −1, (d) CS with amplification
it is. For a 3 × 3 Jacobian, the characteristic equation of for α = 2, and (e) AS with amplification for α = −2, (f)
a derived H matrix is λ3 + a1 λ2 + a2 λ + a3 = 0 where its response at amplitude death for α = 0.
coefficients ai (i = 1, 2, 3) are defined by the elements of
the matrix. The corresponding Routh-Hurwitz (RH) cri-
terion [20] is given by a1 > 0, a3 > 0, a1 a2 > a3 when the
H matrix has eigenvalues all with negative real parts. (b + ∆b)yτ
In fact, the parameter values hi and pi in H1 and H2 ẏ = −(a + ∆a)y + (9a)
1 + yτ10
are so selected that the RH criterion is satisfied and the
condition (8) is thereby fulfilled. It ensures asymptotic
stability of synchronization. To complete the design of bxτ
ẋ = −ax + +D (9b)
the OPCL coupling D, the α matrix and the coupling 1 + x10
τ
delay τc remains to be decided. The stability condition
(8) clearly depends upon the choices of hi and pi but which are chaotic for a set of parameters a = 1, b =
independent of the system parameters and the choice of 2, τ = 2 and mismatches ∆a = 0.2, ∆b = 0.3. The re-
α matrix and τc . This incorporates a great flexibility in sponse system (9b) after adding the OPCL delay cou-
this coupling design in realization of several synchroniza- pling (5) is
tion scenarios, CS, AS, LS, anti-LS, GS and MS in mis-
α(b + ∆b)yτ +τc bαyτ +τc
matched systems. The coupling scheme is not difficult to ẋ = −α∆ayτc + 10 −
implement in electronic circuits and laser experiments. 1 + yτ +τc 1 + α10 yτ10+τc
As in many practical delay dynamical systems, x, y ∈
R1 , when α is a scalar parameter, one can smoothly con-
1 − 9α10 yτ10+τc
trol synchronization using this parameter and induce flip- +(H1 +a)(x−αyτc )+[H2 −b ](xτ −αyτ +τc )
flopping between CS (α = 1) and AS (α = −1) states, (1 + α10 yτ10+τc )2
which is of practical use in digital encoding. A scaling (10)
of the size (amplification or attenuation) of the driver For this example, H1 and H2 are scalars, the condition
attractor at the response oscillator is possible by taking for asymptotic stability is derived from (8),
| α |> 1 or | α |< 1. In higher dimensional systems too,
x, y ∈ Rn , we can realize similar kind of scenarios and H22
H1 + 1 + <0 (11)
in addition, we can observe some new synchronization 4
regimes. We elaborate using numerical examples in the
next section. The stability region for synchronization in the H1 −H2
plane is numerically determined as shown in the black
region in Fig.1(a). We then make a choice of H1 =
III. TARGETING SYNCHRONIZATION: −2, H2 = −1 for our simulations. We first discuss the
NUMERICAL EXAMPLES case with zero coupling delay, τc = 0. Figure 1(b) shows
that the pair of time series of the driver y(t) (online blue
We first elaborate the design of delay coupling using color) and the response x(t) (online red color) are iden-
the mismatched Mackey-Glass system, tical in amplitude and phase for α = 1 and in a CS state
 4

(a) (b)
Figure 2(c) shows the similarity measure S for any delay
1.5 1.4 τs . S has a global minimum at zero for a τs =τs0 = 1.0
which estimates the lag time between the pair of x(t) and
1 1 y(t) in Fig.2(a) as τc = 1.0 = τs0 (dimensionless). A sim-
y, x

y, x
0.5 0.6
ilar result is shown in Fig.2(d) to confirm the LS with a
larger time lag τc = 3.0 = τs0 in Fig.2(b). Thus it is pos-
0 0.2 sible to induce any arbitrarily desired lag time τc in the
0 20 40 0 20 40
t t drive-response system. We choose τc values one larger
and another lower than the intrinsic time lag τ and, LS
(c) (d)
0.4 0.4 is not affected by this choice. The efficacy of a coupling
0.3
design for larger coupling delay is, particularly, impor-
0.3
tant in regards to synchronization in distant regions of
S 0.2
S

0.2 the brain with large conduction delays [28]. This knowl-
0.1 0.1 edge can be utilized in practical situations.
0 0 As a second example, we consider a time-delayed
1 2 2 3 4
τs τ Rössler system [27]. The driving system is
s
ẏ1 = −y2 − ey3 (t−τ ) (12a)
FIG. 2: (Color online) Lag synchronization in Mackey-Glass
system: time series of driver (blue) and response (red) for (a)
ẏ2 = y1 + ay2 (12b)
α = 1.0, τc = 1.0, (b) α = 1.0, τc = 3.0. Similarity measure S
has a global minimum at (c) τs0 = 1.0, (d) τs0 = 3.0.
ẏ3 = y1 + (b + ∆b)e−y3 (t−τ ) − c (12c)

since the synchronization error e(t)=x(t) − y(t) is at zero where ∆b is a mismatch. The driver system is chaotic
in black line. For α = −1, the AS state between x(t) and for a = 0.4, b = 2.0, c = 4.0 and 0 ≤ τ ≤ 0.35. In our
y(t) is clear in Fig. 1(c) where the synchronization error simulation we take τ = 0.1 and ∆b = 0.05.
e(t)=x(t) + y(t) is found at zero in black line. Figures After adding the OPCL coupling (5), the response sys-
1(d) and 1(e) show the CS (α > 1) and AS (|α| > 1) tem becomes,
respectively with amplification of the driver at the re-
sponse in both the cases. Their corresponding errors at ẋ1 = −x2 − ex3 (t−τ ) − α11 {y2 (t − τc ) + ey3 (t−τ −τc ) }
zero are shown in black lines. Similarly, attenuation can
be observed by simply taking 0 <| α |< 1 but we do not
+α22 y2 (t − τc ) + eα33 y3 (t−τ −τc) + p1 {x3 − α33 y3 (t − τc )}
elaborate here. It is also interesting that an amplitude
death (AD) can be induced in the response system by
taking α = 0 as shown in Fig.1(f). The driver y(t) is os- +{h1 +eα33 y3 (t−τ −τc) }{x3 (t−τ )−α33 y3 (t−τ −τc )} (13a)
cillatory while the response x(t) ceases to oscillate and is
stable at zero. The coupling is able to stabilize the equi-
librium origin and thereby induces AD in the response ẋ2 = x1 +ax2 +(α22 −α11 )y1 (t−τc )
system. In fact, the coupling has an inherent property to (13b)
induce AD in a response system by stabilizing the equi-
librium origin and even create a new equilibrium at origin ẋ3 = x1 +be−x3 (t−τ ) −α33 c+(α33 −α11 )y1 (t−τc )
if not present in the uncoupled system and then stabilize
it. We like to elaborate this case in the future.
+α33 (b + ∆b)e−y3 (t−τ −τc ) − be−α33 y3 (t−τ −τc )
For nonzero coupling delays τc , systems (9) and (10)
develop a LS state instead of a CS state. In a LS state, all
the state variables of the oscillators maintain amplitude +{h2 + be−α33 y3 (t−τ −τc ) }{x3 (t − τ ) − α33 y3 (t − τ − τc )}
correlation but shifted by the desired coupling delay τc . (13c)
Figure 2 plots the pair of time series of the response x(t) where we choose the constant matrix α =
(online red color) and the driver y(t) (online blue color) [α11 0 0 ; 0 α22 0 ; 0 0 α33 ] as a simple
in LS for two different choices of time lag τc = 1.0, 3.0. diagonal matrix. In this case, x, y ∈ R3 , and hence
Figures 2(a) and 2(b) in upper panels show LS with (a) H1 and H2 matrices are 3 × 3 matrices. The Jacobian
τc = 1.0, and (b) τc = 3.0. To confirm LS, we calculate matrices of the system (12) are
a similarity measure (S)[1, 19, 21] using the pair of time    
series of y(t) and x(t), y3 (t−τ )
∂f 0 −1 0 ∂f 0 0 −e
=1 a 0 ; =0 0 0 
< [x(t) − y(t − τs )]2 > ∂y ∂y
1 0 0 τ
0 0 −be−y3 (t−τ )
S= p
< x2 (t) >< y 2 (t) >
 5

10 ters in the h1 − p1 plane for stable synchronization are

obtained, using numerical simulation, as shown in the
dark region in Fig.3 where we consider h2 = h1 . By
5
this equality, we find a symmetric non-synchronization

1
p
region in white. However, other choices of parameters
0 are always possible for h1 6= h2 . Clearly, a wide choices
of p1 − h1 parameters is available. Once the asymp-
−5 totic stability of synchronization is thereby established,
−10 0 10 we can easily target several synchronization scenarios,
h1
similar to the first example, by making different choices
of the elements of the matrix α and coupling delay τc . If
FIG. 3: Synchronization region in h1 − p1 plane in dark black α11 = α22 = α33 = 1.0, with no coupling delay (τc = 0),
for delay Rössler system. h1 = h2 systems (12) and (13) are in CS. It means all similar pairs
of state variables of the driver and the response maintain
a CS state. Similarly, for α11 = α22 = α33 = −1.0, all
5
(a) (b) the pairs of state variables of the systems are in AS state.
4 For a numerical example, we make a different choice of
α11 = 1, α22 = −1, α33 = 0 without a coupling delay
y1(t), x1(t)

2
y1(t)

0
0 to reveal an uncommon synchronization scenario. As a
−2
result, we find that different pairs of state variables of
the driver and the response develop different types of co-
−5 −4
0 20 40 −5 0 5 herence: y1 (t) and x1 (t) are in CS in Fig.4(a), y2 (t) and
t x (t)
(c) (d)
1
x2 (t) in AS in Fig.4(c). The y(t) vs. x(t) plots in Fig.4(b)
5 4
and 4(d) confirms CS, AS respectively in similar pairs of
state variables. The third variable y3 (t) of the driver is in
y2(t), x2(t)

2
chaotic state in Fig.4(e) while the corresponding response
y2(t)

0
0 variable x3 (t) is in a steady state or at partial amplitude
−2 death [24] in Fig.4(f). This typical scenario is defined
−5
0 20 40 −5 0 5
as a MS state where three different pairs of state vari-
t x (t)
2 ables of the systems (12) and (13) exist simultaneously
2
(e)
0.4
(f) in different coherent states. Note that the system (12)
is not inversion symmetric, even then AS is realized in
1 0.2
this system. The coupling thus overrules the restriction
y3(t)

x3(t)

0 0
−1 −0.2
−2 −0.4
0 20 40 0 20
t t
FIG. 4: (Color online) Mixed synchronization in delay Rössler
system: CS between y1 and x1 for α11 = 1 in (a) confirmed by
y1 vs. x1 plot in (b), AS between y2 and x2 for α22 = −1 in
(c) confirmed by y2 vs. x2 plot in (d) and, (e) chaotic driver
y3 and corresponding response x3 at amplitude death in (f)
for α33 = 0.
We construct the matrices H1 and H2 from the above
Jacobian matrices by following the rules stated above:
H1 = [0 − 1 p1 ; 1 a 0 ; 1 0 0] and
H2 = [0 0 h1 ; 0 0 0 ; 0 0 h2 ]. By a suitable
choice of the parameters p1 = −1 and h1 = −2, h2 = −2,
the eigenvalues of the H matrix are determined as -12,
-4, -2.4 which satisfies the Hurwitz condition(8) and en-
sures asymptotic stability of synchronization. This choice
is certainly not unique, other choices of parameters are
available for which H satisfies (8) and can be easily de-
termined as shown in the first example. The parame-
of inversion symmetry as valid for diffusive coupling.
In presence of a coupling delay, systems (12) and (13)
40 will obviously develop a LS state. Instead, as a numer-
ical example, we present a more general scenario of MS
state with LS for a coupling delay τc = 1.0. Let us
choose α11 = 1.0, α22 = −1.0 and α33 = 1.0 : three
different pairs of state variables simultaneously develop
LS, anti-LS and LS but with a common delay τc = 1.0.
In Fig.5(a), the time series of y1 (t) (online blue color)
and x1 (t) (online red color) are shown in LS. The corre-
sponding x1 (t) vs. y1 (t−τc ) plot confirms LS in Fig.5(b).
Figure 5(c) shows anti-LS in the pair of time series for
α22 = −1.0 and a coupling delay τc = 1.0, which is con-
firmed by the x2 (t) vs. y2 (t − τc ) plot in Fig.5(d). In
case of anti-LS, a pair of state variables are in opposite
phase and amplitude correlated but shifted by the desired
time lag τc . The x3 (t) and y3 (t) in Fig.5(e) also shows
LS since we choose α33 = 1.0. Figure 5(f) plots x3 (t)
vs. y3 (t − τc ) to confirm LS between y3 (t) and x3 (t).
A scaling of the size of the driver attractor at response
is also possible by making the elements of the α matrix
larger/lower than unity. It is demonstrated that a choice
of the elements of the α matrix is so flexible that one
can easily target different desired synchronization sce-
narios. The choice of the α matrix does not interfere
 6

(a) (a) (b) (c)

(b)
5 20 20
4
4
10 2 10

1
2

y ,x
y1(t−τc)

1
1
y1, x1

g
y
0

1
0 0 0
0
−2
−2 −10 −4 −10
0 50 100 −10 0 10 20 −10 0 10 20
−5 −4 t x x
1 1
60 80 100 −4 −2 0 2 4 (d) (e) (f)
t x (t) 5 4
1
2 2
(c) (d)
5 0
0 0

2
2

y ,x

g2
2
−2

y
2
−2 −4
y2(t−τc) 0 −5
y2, x2

−4 −6
0
−2 −8
0 50 100 −6 −4 −2 0 2 4 −8 −6 −4 −2 0 2 4
t x2 x2
−4 (g) (h) (i)
−5 2 2 2
60 80 100 −2 0 2 4
t x (t) 1
2 0 0

3
(e) (f)

y ,x

g3
3
0

y
2 2

3
−2 −2
−1
1 1
y3(t−τc)

−4 −2 −4
y3, x3

0 50 100 −4 −2 0 2 −4 −2 0 2
0 t x x
0 3 3

−1
−1
−2
60 80 100 −1 0 1 2
t x (t)
3
FIG. 6: (Color online) Generalized synchronization in delay
Rössler system: (a) time series of driver y1 (blue color) and
response x1 (red color), (b) shows y1 vs x1 and (c) trans-
form driver g1 and response variable x1 in one-to-one relation.
FIG. 5: (Color online) Lag synchronization in delay Rössler Similar plots of the pair (y2 , x2 ), (y2 vs.x2 ), and g2 vs x2 in
system: (a) LS for α11 = 1, corresponding x1 (t) vs. y1 (t − (d)-(f). For the third response g3 , all similar plots are shown
τc ) plot in (b), (c) anti-lag synchronization for α22 = −1, in (g)-(i)
corresponding x2 (t) vs. y2 (t − τc ) plot in (d0, (e) LS for .
α33 = 1, and x3 (t) vs. y3 (t − τc ) plot in (f).

in the stability of synchronization. By varying one or
more elements of this α matrix, a practical scheme can
be implemented for smooth control of synchronization
from one regime to another without loss of synchroniza-
tion during this transition. One can even externally force
an abrupt change in one or more elements of the α matrix
to induce a flip-flopping of synchronization which can be
utilized for phase-shift keying [29] in digital communica-
tion. Presently, we use this property of the α matrix to
demonstrate a GS scenario in the next section.
IV. TARGETING GENERALIZED
SYNCHRONIZATION
A GS scenario was reported in nonidentical systems or
largely mismatched systems under linear diffusive cou-
pling [23]. In a GS state the response system develop a
functional relationship with the driver. GS in delay sys-
tems under delay coupling is also investigated [24], how-
ever, the coupling is assumed a priori known. Instead,
we follow the design strategy of the OPCL delay coupling
for engineering a GS state even in slightly mismatched or
identical delay systems. We appropriately choose the el-
ements of the α matrix for targeting a desired functional
relationship between the driver and the response. As a
simple example, we choose some of the off-diagonal ele-
ments of the scaling matrix α as constants in addition
to the principal diagonal elements, the driver-response
system enter into an interesting GS regime. For numer-
ical demonstration, we use the coupled delayed Rössler
systems (12) and (13) with an arbitrary choice of the
scaling matrix: α = [1 − 3 − 2 ; 0 1 2 ; 0 0.5 1].
The goal dynamics g(t) = [g1 (t) g2 (t) g3 (t)]T is de-
fined by [g1 (t) = y1 (t) − 3y2 (t) − 2y3 (t), g2 (t) = y2 (t) +
2y3 (t), g3 (t) = 0.5y2 (t) + y3 (t). The state variables of
the goal dynamics or the response variables are linear
combinations of the driver variables since the elements
of the α matrix are chosen constant or zero. Figure 6(a)
shows time series of the driver y1 (t) (online blue color)
and the response x1 (t) (online red color) which appar-
ently shows no correlation between them by visual check.
The y1 (t) vs. x1 (t) plot appeared to show no correlation
in Fig.6(b). However, the x1 (t) vs. g1 (t) plot in Fig.6(c)
clearly shows 1:1 correlation and confirms that the tar-
geted GS state is actually realized at the response system.
Other pairs of state variables similarly show their corre-
sponding targeted GS states defined by g2 (t) and g3 (t) in
Fig.6(d)-(f) and Fig.6(g)-(i) respectively. Since the scal-
ing matrix is only involved in the definition of a GS state
and it does not disturb the stability of synchronization,
in general, one can easily make a wide variety of choices

of the scaling matrix to define a desired functional rela-
tionship between the driver and response. A nonlinear
functional relationship can also be adopted to target a
more complex GS scenario by inserting state variables of
the driver or a third dynamical system into the α matrix
as shown, in detail, in case of instantaneous OPCL cou-
pling for GS [30]. The proposed design scheme is thereby
able to engineer a varieties of desired GS state in delay
system too in presence of delay coupling.
V. CONCLUSION
In conclusion, we started with a known delay system
and a priori unknown coupling and, then proposed a
design of unidirectional delay coupling for engineering
various synchronization scenarios. We made necessary
modifications in the theory of the instantaneous OPCL
coupling for implementing in delay systems with a con-
duction delay. We improved the stability condition of
synchronization by supporting the Krasovskii-Lyapunov
function theory with a Hurwitz matrix criterion and
thereby we achieve an important benefit over the con-
ventional methods that the stability of synchronization
became independent of the system parameters. Another
immediate benefit is that the OPCL delay coupling al-
lows a flexible control of the synchronization. In fact, for
the purpose of controlling synchronization, a scaling ma-
trix is introduced in the coupling definition. The choice
of the elements of the scaling matrix does not interfere
in the stability of synchronization. One can thus switch
or smoothly change from one regime to another with-
out loss of stability during a transition. This is a spe-
[1] A. Pikovsky, M. Rosenblum and J. Kurths, Synchroniza-
tion: A Universal Concept in Nonlinear Sciences (Cam-
bridge University Press, Cambridge, England, 2001), S.
Boccaletti, J. Kurths, G. Osipov, D. L. Valladares, C. S.
Zhou, Phys. Rep. 366, 1(2002).
[2] L. M. Pecora, T. L. Carroll, Phy. Rev. Lett. 64,
821(1990), H. Fujisaka, T. Yamada, Prog. Theor. Phy.
69, 32 (1983).
[3] L. M. Pecora, T. L. Carroll, Phy. Rev. Lett. 80, 2109
(1998).
[4] J. Sun, E. M. Bolt, T. Nishikawa, SIAM J. Applied Dy-
namical Systems 8 (1), 202 (2009).
[5] I. Z. Kiss, C. G. Rusin, H. Kori, J. L. Hudson, Science
316 (5853), 1886 (1995).
[6] A. Prasad, M. Dhamala, B. M. Adhikari, R. Ramaswamy,
Phy. Rev.E 82, 027201 (2010).
[7] J-R. Kim, D. Shin, S. H. Jung, P-H. Harrison, K-H. Cho,
J. Cell Sciences 123 (4), 537 (2010).
[8] P. A. Popovych, P. A. Tass, Phy. Rev. E 82, 026204
(2010).
[9] K. Pyragas, T. Pyragine, Phy. Rev. E 78, 046217 (2008).
[10] M. Lakshmanan, D. V. Senthilkumar, Nonlinear dynam-
ics of time delay systems, Springer, Berlin, 2010.
7
cial property [20, 30] of the OPCL coupling which can
be used for practical applications. The scaling matrix is
also used for enlarging or attenuating a driver attractor
at a response system. This method has added advan-
tages over the conventional methods of synchronization
in delay system under delay coupling: it can engineer
a desired synchronization regime in the chaotic oscilla-
tors without restrictions of the nature of the system and
in presence of parameter mismatch. An AS state can
also be induced in systems without inversion symme-
try as usually found restricted for other coupling con-
figurations. Further, contrary to conventional coupling
schemes, a variety of GS regimes (functional relationship)
can be engineered in slightly mismatched oscillators by a
wider range of choices in the elements of the scaling ma-
trix. GS can be induced even in identical oscillators. We
supported the theory with numerical simulations of the
Mackey-Glass system and a delayed Rössler model. The
OPCL scheme is thus successfully extended to delay sys-
tems with delay coupling for targeting any of the desired
synchronization states, CS, AS, LS, anti-LS, AD and a
MS state. This delayed version of the OPCL coupling is
physically implementable in time-delayed optoelectronic
feedback loops and electronic circuits which is our future
objective. A delay dynamical system represents many
natural systems and since it has practical utility in laser
based communication, a robust design of coupling, in the
sense of stability to internal and external perturbations,
such as the OPCL delay coupling is always important in
search of future applications.
S.K.D. acknowledges partial support by the BRNS, In-
dia under grant 2009/34/26/BRNS.
[11] H. Hakens, Brain dynamics: synchronization and activity
pattern in pulse coupled nueral net and noise,(Springer,
Berlin, Germnay, 2006), E. Scholl, G. Hiller, P. Hovel,
M. A. Delhem, Phil. Trans. R. Soc. A 367, 1079 (2009).
[12] A. Argyris, D. Syvridis, L. Larger, V. Annovazzi-Lodi, P.
Colet, I. Fischer, J. Garcia-Ojalvo, C.R. Mirasso, L. Pes-
quera and K.A. Shore Nature 438, 343 (2005), J. Zamora-
Munt, C. Masoller, and J. Garcia-Ojalvo, R.Roy, Phy.
Rev. Lett 105, 264101 (2010).
[13] G. D. VanWiggeren, R. Roy, Phys. Rev. Lett. 81, 3547
(1998), G. D. VanWiggeren and R. Roy, Science 279,
1198 (1998).
[14] K. Pyragas, Phy. Rev. E 58, 3067 (1998), M. J. Binner,
W. Just, Phy. Rev. E 58, 4072R (1998), I. Fischer, R.
Vicente, J. M. Baldu, M. Peil, C. R. Mirasso, M. C. Tor-
rent, J. Garcia-Ojalvo, Phy.Rev.Lett. 97, 123902 (2006).
[15] D. Ghosh and A. Roy Chowdhury, Phys. Lett. A 374,
3425 (2010), S. Banerjee, D. Ghosh, A. Ray and A. Roy
Chowdhury, Europhys. Lett. 81, 20006 (2008), D. Ghosh,
S. Banerjee and A. Roy Chowdhury, Europhys. Lett. 80,
30006 (2007), K. Srinivasan, D. V. Senthilkumar, K. Mu-
rali, M. Lakshmanan, and J. Kurths, Chaos 21, 023119
(2011).

[16] D. V. Senthilkumar, M. Lakshmanan, and J. Kurths,
Phys. Rev. E 79, 066208(R) (2009), D. Ghosh, Chaos
19, 013102 (2009).
[17] N. N. Krasovskii, Stability of Motion ( Stanford Univer-
sity Press, Stanford, 1963).
[18] F. Sorrentino, E. Ott, Chaos, 19, 033108 (2009), B.
Ravoori, A. B. Cohen, A. V. Setty, F. Sorrentino, T. E.
Murphy, E. Ott, R. Roy, Phys. Rev. E 80, 056205 (2009).
[19] T.Huang, C.Li, W.Wu. G.Chen, Nonlinearity 22, 569
(2009).
[20] I. Grosu, E. Padmanaban, P. K. Roy and S. K. Dana,
Phys. Rev. Lett. 100, 234102(2008), I. Grosu, R. Baner-
jee, P. K. Roy and S. K. Dana, Phys. Rev. E 80, 016212
(2009).
[21] L-Y cao, Y-C. Lai, Phy. Rev. E 58, 382 (1998), V. As-
takov, A. Shabun, V. Anischencko, Int. J. Bifurcation
Chaos Appl. Sci. Engg. 10, 849 (2000), W. Liu et al.,
Phy. Rev. E 73, 057203 (2006), C-M. Kim et al. Phy.
Lett. A 320, 39 (2003), V. Belykh, I. Belykh, M. Hasler,
Phy. Rev. E 62, 6332 (2000), S. K. Dana, B. Blasius, J.
Kurths, Chaos 16, 023111,(2006).
[22] S. Taherion, Y-C. Lai, Phy. Rev. E 59, 6247R (1999),
M. G. Rosenblum, A. S. Pikovsky, J. Kurths, Phy. Rev.
Lett. 78, 4193 (1997); P. K. Roy, S. Chakraborty, S. K.
Dana Chaos 13, 342 (2003).
8
[23] N. F. Rulkov, M. M. Sushchik, L. S. Tsimring, H. D. I.
Abarbanel, Phy. Rev. E 51, 980 (1995), L. Kocarev, U.
Parlitz, Phy. Rev. Lett. 76 (11), 1816 (1996), D. He, L.
Stone, Z. Zheng, Phy. Rev. E 66, 036208 (2002).
[24] M. Zhan, X. Wang, X. Gong, G. W. Wei and C. H. Lai
Phys. Rev. E 68 036208 (2003), W. Liu, J. Xiao, J. Yang,
Phy. Rev. E 72, 057201 (2005).
[25] D. V. R. Reddy, A. Sen, G. L. Johnston, Phy. Rev. Lett.
80, 5109 (1998).
[26] M. C. Mackey and L. Glass, Science 197, 287 (1977),
M.-Y. Kim, C. Sramek, A. Uchida, R. Roy, Phys. Rev.
E 74, 016211 (2006), S. Sano, A. Uchida, S. Yoshimori,
and R. Roy, Phys. Rev. E 75, 016207 (2007).
[27] H. Huijberts, H. Nijmeijer, T. Oguchi, Chaos 17, 013117
(2007).
[28] A. K. Engel, P. Koenig, A. K. Kreiter, W. Singer, Science
252, 1177 (1991), R.Vicente, L.L.Gollo, C.R.Mirasso,
I.Fischer, G.Pipa, Proc.Nat.Acad.Sci 105, 17157 (2008).
[29] B.Sklar, Digital Communications: Fundamentals and
Applications (Addison Wesley Longman (Singapore),
Delhi, India 2001).
[30] P. K. Roy, C. Hens, I. Grosu and S. K. Dana, Chaos 21,
013106 (2011).