You are on page 1of 20

Journal of Taibah University for Science

ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/tusc20

Codimension-one bifurcation analysis and chaos


of a discrete prey–predator system

Abdul Qadeer Khan & Syeda Noor-ul-Huda Naqvi

To cite this article: Abdul Qadeer Khan & Syeda Noor-ul-Huda Naqvi (2024) Codimension-one
bifurcation analysis and chaos of a discrete prey–predator system, Journal of Taibah University
for Science, 18:1, 2317505, DOI: 10.1080/16583655.2024.2317505

To link to this article: https://doi.org/10.1080/16583655.2024.2317505

© 2024 The Author(s). Published by Informa


UK Limited, trading as Taylor & Francis
Group.

Published online: 28 Feb 2024.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tusc20
JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE
2024, VOL. 18, NO. 1, 2317505
https://doi.org/10.1080/16583655.2024.2317505

Codimension-one bifurcation analysis and chaos of a discrete prey–predator


system
Abdul Qadeer Khan and Syeda Noor-ul-Huda Naqvi
Department of Mathematics, University of Azad Jammu and Kashmir, Muzaffarabad, Pakistan

ABSTRACT ARTICLE HISTORY


In this paper, we explore local dynamics at equilibrium points, the existence of bifurcation sets Received 12 May 2023
and codimension-one bifurcation analysis of a discrete prey–predator system with Holling type- Revised 15 October 2023
II functional response. Further, OGY and Hybrid control strategies are utilized to control chaos in Accepted 7 February 2024
the under study discrete model due to the occurrence of Neimark–Sacker and flip bifurcations. KEYWORDS
Finally, numerical simulations are given to verify theoretical results. Predator–prey model;
codimension-one
bifurcations; chaos;
non-standard scheme;
numerical simulation
MSC CLASSIFICATIONS
70K50; 92D25; 40A05

1. Introduction response [1]:


⎧  x cxy
1.1. Motivation and mathematical formulation ⎪
⎨ẋ = ax 1 − K − m + x ,
The Lotka–Volterra model, often known as the prey–  (1)
⎪ fx
predator model, is a mathematical representation of the ⎩ẏ = y −d ,
m+x
dynamics of two species, one being the predator and
the other being the prey in a given population. The where x is prey’s density and y is predator’s density,
model explains the long-term interconnections among while parameters c, a, d, f, K, and m show the captur-
the populace of the predator and prey species. It con- ing rate, intrinsic growth rate of prey, predator death
siders variables like the rate of birth and death of the rate, maximal predator growth rate, carrying capacity
predator and prey, as well as the rate at which the and half saturation constant, respectively. The so-called
predator hunts and kills the prey. The prey–predator ratio-dependent theory, which holds that the rate of per
model has been broadly used to explore the dynam- capita predator growth should rely on the ratio of prey
ics of population interactions in various fields such as to predator abundance, should be the foundation of a
designing conservation procedure to safeguard endan- more suitable general theory of predator–prey relation-
gered species and understanding the interconnections ships. This is especially true when the predator has to
between several species in an ecosystem. Although look for food, and it is supported by numerous field,
these dynamical concerns resulting from the mathe- lab studies and observations [2–5]. The basic form of
matical simulation of prey–predator systems may ini- ratio-dependent prey–predator system is:
tially appear to be straightforward, a deep study of ⎧ 
⎪ x
these models frequently give rise to incredibly chal- ⎪
⎨ ẋ = xf (x) − yp ,
y
lenging problems. The main goal of modelling a pop-   (2)

⎪ x
ulation ecosystem is to ensure that the mathematical ⎩ẏ = cq − d y,
y
model in question can represent the noteworthy system
behaviours for the system being observed. Dynamic where c is conversion rate, predator functional
modelling of ecological systems is often an evolv- response is represented by p(x), q(x) is supplan-
ing technique. Prey–predator relationships modelled ted by p(x). Moreover, Kuang and Beretta [6] have
mathematically using the famous Lotka–Volterra type extended the prey–predator model (2) by the
prey–predator model with Holling type-II functional effect of Holling type-II functional response as

CONTACT Abdul Qadeer Khan abdulqadeerkhan1@gmail.com Department of Mathematics, University of Azad Jammu and Kashmir,
Muzaffarabad 13100, Pakistan
© 2024 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (http://creativecommons.org/licenses/by-nc/4.0/), which
permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. The terms on which this article has been
published allow the posting of the Accepted Manuscript in a repository by the author(s) or with their consent.
2 A. Q. KHAN AND S. NOOR-UL-HUDA NAQVI

follows: where u, b, c, m are positive constants. In contrast to


⎧ 
⎪ x cxy discrete models, many investigators investigated the

⎨ẋ = ax 1 − K − my + x , dynamics of continuous-time systems designated by
 (3)

⎪ fx differential equations. For instance, Tunç and Tunç [12]
⎩ẏ = y −d + .
my + x have explored dynamics of the following system of
second-order differential equations with delay:
It is anticipated that model (3) becomes the form [6, 7]:

⎪ sxy ẍ + F (x, ẋ) ẋ + H (x(t − τ (t))) = P (t, x, ẋ) . (11)

⎨ẋ = x (1 − x) − x + y ,
 (4) Tunç [13] has investigated the dynamics of the follow-

⎪ x
⎩ ẏ = δy −r + , ing vector lienard equation:
x+y
by ẍ + F (x, ẋ) ẋ + H (x(t − τ )) = P (t) . (12)

⎪ x

⎪ x= , Tunç [14] has investigated the boundedness of solu-
⎨ K
tions of a class of non-autonomous differential equa-
t = at, (5)

⎪ my tions.

⎩y = ,
K
where 1.3. The novelty of the proposed work
⎧ c

⎪ s= , It is important here to note that the earlier work is about

⎪ ma

⎨ the global qualitative study on system (3) where the dis-
f
δ= , (6) cussion of authors focuses on the qualitative study of

⎪ a

⎪ the ratio-dependent predator–prey systems. Their work

⎩r = d .
f builds upon previous work but aims to address the open
questions regarding the global qualitative behaviour
1.2. Literature survey of the model. By transforming the Michaelis-Menten-
type ratio-dependent model into a Gause-type preda-
Many investigators investigated the dynamics of prey– tor–prey system, the authors obtain a complete classi-
predator models. For instance, Singh and Malik [8] have fication of the asymptotic behaviour of the solutions.
investigated the dynamics at equilibrium points and They resolve open questions regarding the global sta-
exploration the bifurcation of the following model: bility of equilibria and the uniqueness of limit cycles.
⎧ 
⎪ pyt α Their work provides insights into how the outcomes of

⎨xt+1 = xt + xt 1 − xt − − ,
γ + xt δ + xt the model depend on initial conditions. Furthermore,
 (7) they present the biological implications of the findings,

⎪ qyt
⎩yt+1 = yt + βyt 1 − , showcasing the practical relevance of the study. Over-
γ + xt
all, Kuang and Beretta [6] discussed the dynamics of
where yt and xt denote the predator and prey populace continuous-time model (3), and demonstrate that if the
densities. Ma et al. [9] have studied the local dynamics system’s positive steady state (3) is locally asymptot-
and bifurcation of following model: ically stable, then the system cannot have non-trivial

⎪ xt (1 − m)xt yt positive periodic solutions. It also includes some results
⎪xt+1 = xt +
⎨ (r − axt ) − ,
1 + kyt b + (1 − m)xt regarding the global stability of the positive steady


⎪ cyt state. Moreover, they non-dimensionalize the system (3)
⎩yt+1 = yt 1 + μ − ,
b + (1 − m)xt into (4), then, Hsu et al. [7] investigated the full classifica-
(8) tion for the asymptotic behaviour of (4) by transforming
it into a Gause-type predator–prey system.
where r, a, m, b, c, r, μ, k are positive constants. Santra
In contrast to the continuous-time model, our aim in
et al. [10] have explored behaviour of the following
this paper is to explore the dynamical characteristics of
model involving prey refuge:
the following prey–predator system:
xt+1 = axt (1 − xt ) − c (xt − byt ) yt , ⎧
(9) ⎪ xt (1 + h) hsxt yt
yt+1 = d (xt − byt ) yt , ⎪
⎨xt+1 = − ,
1 + hxt (xt + yt )(1 + hxt )
 (13)

⎪ xt
where a, c, b, r, d are positive parameters. Chen et al. [11] ⎩yt+1 = yt + hδyt −r + ,
have studied local stability and bifurcation at fixed xt + yt
points of the following model: which is discrete version of (4) by non-standard finite

⎨xt+1 = xt + xt (1 − xt ) xt − bxt yt , difference scheme, where h is a step size. Furthermore,
u + xt (10) s denotes the ratio of capturing rate to the half satura-
⎩y
t+1 = yt + cxt yt + myt − yt ,
2
tion constant times intrinsic growth rate of prey, δ is the
JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE 3

ratio of maximal predator growth rate to the intrinsic understanding the stability and dynamics of the sys-
growth rate of prey, and finally r is the ratio of preda- tem under different conditions. Additionally, discrete
tor death rate to the maximal predator growth rate. models are useful for studying chaos in ecological
The reason for studying the dynamics of discrete-time and biological systems. Techniques like the OGY (Ott-
system (13), instead of of continuous-time system (4), Grebogi-Yorke) method can be employed to analyse
is that the discrete model simplifies the complexity of the local behaviour of fixed points and identify chaotic
the system, making it more accessible for analysis and regions in the parameter space. This study of chaos can
interpretation. Continuous models often involve differ- provide insights into the underlying mechanisms driv-
ential equations, which can be challenging to solve ana- ing chaotic behaviour. Furthermore, discrete-time mod-
lytically or numerically. By converting the model into els facilitate the development and analysis of hybrid
a set of difference equations, the computational com- control strategies. Hybrid control combines continu-
plexity is reduced, making it easier to simulate and ous and discrete dynamics to design control strategies
analyse the dynamics of the system. Additionally, the that effectively manage complex systems. By calculat-
discretization of the model allows for direct compari- ing the local behaviour for fixed points, researchers
son with experimental data. Experimental data is often can determine control strategies that stabilize the sys-
collected at discrete-time points, and by converting tem, suppress undesirable behaviour, or guide the sys-
the model into a discrete version, the model’s predic- tem towards desired states. Lastly, discrete-time models
tions can be directly compared with the observed data. offer opportunities for numerical validation of theoreti-
This facilitates experimental validation and enhances cal results. Researchers can simulate the discrete model
the credibility of the model. Furthermore, stability anal- and compare the results with theoretical predictions,
ysis becomes more feasible with discrete models. By validating the accuracy and reliability of their theoret-
studying the behaviour of the system over discrete-time ical findings. This numerical validation helps build con-
steps, it becomes possible to analyse the stability of fidence in the model and its ability to capture the essen-
the fixed points and determine if the system will con- tial dynamics of the system. In summary, calculating
verge to an equilibrium or exhibit oscillatory behaviour. the local behaviour for fixed points in discrete mod-
This stability analysis can provide valuable insights for els provides advantages in identifying bifurcation sets,
system design and control. Lastly, parameter estima- conducting bifurcation analysis, studying chaos, devel-
tion is often easier with discrete models. Estimating oping hybrid control strategies, and numerically vali-
parameters in continuous models can be challenging dating theoretical results. These advantages enhance
due to the complexity of the equations. Discrete mod- our understanding of ecological and biological systems
els, with their simpler forms, allow for more straightfor- and enable more effective decision-making and control
ward parameter estimation techniques, enhancing the strategies. Due to aforementioned fact, in this paper we
accuracy and reliability of the model’s predictions. In explore dynamics of discrete system model (13) where
summary, converting a continuous-time model into a our key investigations include:
discrete version offers advantages in terms of compu-
tational efficiency, simplicity, compatibility with exper- • Local behaviour for fixed points.
imental data, stability analysis, and parameter esti- • Identification of bifurcation sets for fixed points.
mation. These advantages make the discrete model • Bifurcation analysis of discrete model (13).
a valuable tool in the literature of ecology and biol- • Study of chaos by OGY and Hybrid control strategies.
ogy for studying and analysing the local behaviour • Numerical validation of theoretical results.
of fixed points. In the context of ecology and biol-
ogy, calculating the local behaviour for fixed points
1.4. The advantages of the proposed work
in discrete models offers several advantages. Firstly, it
allows for the identification of bifurcation sets, which In the setting of a discrete prey–predator system,
are critical points where the qualitative behaviour of codimension-one bifurcation analysis has numerous
the system changes. By analysing the local behaviour advantages for understanding the dynamics, stability,
of fixed points, researchers can identify bifurcations and potential for chaos in ecological systems:
such as the emergence of multiple stable states or
the occurrence of oscillatory behaviour. Understanding • Codimension-one bifurcation analysis aids in iden-
these bifurcation sets provides insights into the under- tifying key parameter value at which qualitative
lying mechanisms driving system dynamics. Secondly, changes in the system dynamics occur. It enables
discrete-time models enable bifurcation analysis, which researchers to identify the precise parameter value
involves studying how the system’s behaviour changes that cause bifurcations or the formation of chaotic
as model parameters are varied. This analysis can attractors.
reveal the existence of different types of bifurcations, • This type of analysis gives information about the sta-
such as saddle-node, transcritical, Neimark–Sacker, flip, bility of the system’s equilibrium points and periodic
or pitchfork bifurcations. Bifurcation analysis helps in orbits. It helps identify whether prey and predator
4 A. Q. KHAN AND S. NOOR-UL-HUDA NAQVI

populations tend to stabilize or oscillate under varied Equation (15) can also be rewritten as
parametric value.
• Researchers can forecast system transitions by exam- x2 − x
y= , (16)
ining codimension-one bifurcations. They can, for 1−x−s
example, forecast when a stable equilibrium will and
become unstable, resulting in the formation of
chaotic behaviour. (1 − r)x
y= . (17)
• Codimension-one bifurcation analysis can aid in r
detecting the emergence of chaos in a discrete Using (17) into (16), one gets
prey–predator system.
• Codimension-one bifurcation analysis can be used as x = 1 − s + sr. (18)
a teaching tool for students and researchers to learn
about complex ecological dynamics. Further, using (18) into (17), one gets
• The calculation of local behaviour for fixed points in
(1 − r) (1 − s + sr)
discrete-time models offers advantages in identify- y= . (19)
r
ing bifurcation sets, conducting bifurcation analysis,
studying chaos, developing hybrid control strate- So, from Equations (18) and (19), one can obtained that
gies, and numerically validating theoretical results. if s−1
s < r < 1 then in R+ model’s interior equilibrium
2

These advantages enhance our understanding of point is ϕ2 = (1 − s + sr, (1−r)(1−s+sr)


r ). Now the varia-
ecological systems, aid in predicting and managing tion matrix V|ϕ of the linearized system of (13) under the
ecological dynamics, and inform conservation and map (f1 , f2 ) → (xt+1 , yt+1 ) is
ecosystem management practices. ⎛ ⎞
(1 + h)(x + y)2
⎜ +h2 sx 2 y − hsy2 −hsx 2 ⎟
⎜ ⎟
1.5. Paper layout ⎜ ⎟
V|ϕ := ⎜ (x + y)2 (1 + hx)2 (x + y)2 (1 + hx) ⎟ ,
⎜ ⎟
The organization of rest of the paper is as follows: stabil- ⎝ hδy2 hδx 2 ⎠
1 − hδr +
ity analysis and bifurcation sets are studied in Section 2. (x + y)2 (x + y)2
In Section 3, we studied flip bifurcation for boundary (20)
fixed point, whereas Neimark–Sacker and flip bifurca-
tions at interior fixed points are briefly investigated in where
Sections 4 and 5, respectively. The chaos control by OGY ⎧
⎪ x(1 + h) hsxy
and Hybrid control strategies is presented in Section 6. ⎪
⎨f1 := − ,
1 + hx (x + y)(1 + hx)
Finally, numerical simulations and conclusion are given  (21)

⎪ x
in Sections 7 and 8, respectively. ⎩f2 := y + hδy −r + .
x+y

Hereafter, we will give local dynamics for the equilibria


2. Bifurcation sets and stability analysis ϕ1,2 by stability theory [15–17]. So for ϕ1 , (20) becomes
In the present section, we explore the stability anal- ⎛ ⎞
1 −hs
ysis at equilibrium points and bifurcation sets for the
discrete prey–predator model (13). If ϕ = (x, y) is an V|ϕ1 := ⎝ 1 + h 1+h ⎠, (22)
0 1 − hδr + hδ
equilibrium point of (13), then one has
⎧ with
⎪ x(1 + h) hsxy

⎨x = − ,
1 + hx (x + y)(1 + hx) 1
 (14) λ1 = , λ2 = 1 − hδr + hδ. (23)

⎪ x 1+h
⎩y = y + hδy −r + .
x+y
Theorem 2.1: ϕ1 of model (13) is
Since ϕ1 = (1, 0) satisfied system (14) obviously, and 2
therefore for all h, s, δ and r discrete model (13) has (i) a sink if h > δ(r−1) ;
semitrivial equilibrium point ϕ1 = (1, 0). In order to get (ii) never source;
2
the model’s interior equilibrium point, one need to (iii) a saddle if 0 < h < δ(r−1) ;
2
solve the following system (iv) non-hyperbolic if h = δ(r−1) .


⎪ 1+h hsy Proof: By stability theory, ϕ1 of discrete model (13) is

⎨ − = 1, 1
1 + hx (x + y)(1 + hx) a sink if |λ1,2 | < 1. Therefore, from (23) if |λ1 | = 1+h <
 (15) 2

⎪ x 1 and |λ2 | = |1 − hδr + hδ| < 1, that is, h > δ(r−1) then
⎩δ −r + = 0.
x+y ϕ1 is a sink. Similarly, it is easy to obtain that ϕ1 is never
JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE 5

source, saddle if 0 < h < 2 Theorem 2.3: If < 0, then the model’s interior equilib-
δ(r−1) , and non-hyperbolic if
2
h = δ(r−1) .  rium point ϕ2 is

1−s+sr2
Now in the following, one has the following result (i) a stable focus if 0 < δ < r2 −r
with r > max

2
regarding bifurcation set at ϕ1 if h = δ(r−1) holds. {± s−1s , 1};
1−s+sr2
Theorem 2.2: For the model’s semitrivial fixed point ϕ1 , (ii) an unstable focus if δ > r2 −r
with r > max

the flip bifurcation set is written as F |ϕ1 := {(h, s, δ, r), h = {± s−1
s , 1};
2
δ(r−1) }. (iii) non-hyperbolic if δ = 1−s+sr2
.
r2 −r
2
Proof: Since ϕ1 is non-hyperbolic if h = δ(r−1) . There- Proof: If < 0 then from (28) one gets |λ1,2 | =
2 
fore, if h = δ(r−1) holds then λ2 |h= 2 = −1 but 1−hrs(r−1) h2 r2 sδ(r−1)2
δ(r−1) ( 1+h+hs(r−1) )(1 + hrδ(r − 1)) + 1+h+hs(r−1) < 1. This
λ1 |h= 2 = δ(r−1)
= 1 or − 1 which implies that at 
δ(r−1)+2 1−s+sr2
δ(r−1) implies that if 0 < δ < r2 −r with r > max{± s−1 s , 1}
ϕ1 eigenvalues criterion for the existence of flip bifur-
then ϕ2 is a stable focus. Similarly, easy calculation yields
cation holds if (h, s, δ, r) passes F |ϕ1 := {(h, s, δ, r), h = 2
the fact that it an unstable focus if δ > 1−s+sr r2 −r
with r >
2
δ(r−1) }.  
1−s+sr2
max{± s−1 s , 1}, and non-hyperbolic if δ = r2 −r . 
Now for ϕ2 , (20) becomes
⎛ ⎞ Theorem 2.4: If ≥ 0 then the model’s interior equilib-
1 − hrs(r − 1) −hr2 s rium point ϕ2 is
⎜ ⎟
V|ϕ2 := ⎝ 1 + h + hs(r − 1) 1 + h + hs(r − 1) ⎠ ,
hδ(r − 1)2 1 + hrδ(r − 1) −4−2h(1−s(r−1)2 )
(i) a stable node if δ > hr(r−1)(2+h+hs(r−1)) with s >
2+h 2+h
2 , h(1−r) };
(24) max{ h(r−1)
−4−2h(1−s(r−1)2 )
with the following characteristic equation: (ii) an unstable node if 0 < δ < hr(r−1)(2+h+hs(r−1)) and
2+h 2+h
s > max{ h(r−1)2 , h(1−r) };
⎛ ⎞
1 − hrs(r − 1) −hr2 s −4−2h(1−s(r−1)2 )
−λ (iii) non-hyperbolic if δ = hr(r−1)(2+h+hs(r−1)) .
det ⎝ 1 + h + hs(r − 1) 1 + h + hs(r − 1) ⎠ = 0,
hδ(r − 1)2 1 + hrδ(r − 1) − λ
(25) Proof: It is noted that ϕ2 is a stable node if real roots
of (28) satisfying |λ1,2 | < 1, that is, if δ >
−4−2h(1−s(r−1)2 ) 2+h 2+h
that is, hr(r−1)(2+h+hs(r−1)) with s > max{ h(r−1) 2 , h(1−r) }
then ϕ2 is a stable node. Similarly, easy calculation also
λ2 − 1λ + 2 = 0, (26) implies that it is an unstable node if 0 < δ <
−4−2h(1−s(r−1)2 )
where hr(r−1)(2+h+hs(r−1)) , and non-hyperbolic if δ =
−4−2h(1−s(r−1)2 )
1 − hrs(r − 1) hr(r−1)(2+h+hs(r−1)) . 
1 = + 1 + hrδ(r − 1),
1 + h + hs(r − 1) Now in the following, one has the following result

1 − hrs(r − 1) regarding bifurcation set at ϕ2 if δ = 1−s+sr
2
and δ =
2 = (1 + hrδ(r − 1)) r2 −r
1 + h + hs(r − 1) −4−2h(1−s(r−1)2 )
hold, respectively.
hr(r−1)(2+h+hs(r−1))
h2 r2 sδ(r − 1)2
+ . (27) Theorem 2.5: For the model’s interior equilibrium point
1 + h + hs(r − 1)
ϕ2 , the Neimark–Sacker and flip bifurcation sets, respec-
From (26), one has tively are

1 ± 1−s+sr2
λ1,2 = , (28) (i) N |ϕ2 := {(h, s, δ, r), δ = r2 −r
};
2
−4−2h(1−s(r−1)2 )
(ii) F |ϕ2 := {(h, s, δ, r), δ = hr(r−1)(2+h+hs(r−1)) }.
where
2 2
= 1 −4 2, Proof: (i). It is noted that if δ = 1−s+sr
r2 −r
holds, then V|ϕ2
 2 has complex eigenvalues with |λ1,2 | 1−s+sr2 = 1 which
1 − hrs(r − 1) δ=
r2 −r
= + 1 + hrδ(r − 1)
1 + h + hs(r − 1) implies that at ϕ2 eigenvalues
 criterion for the existence of N-S bifurcation holds if
1 − hrs(r − 1) 2
−4 (1 + hrδ(r − 1)) (h, s, δ, r) passes N |ϕ2 := {(h, s, δ, r), δ = 1−s+sr }.
1 + h + hs(r − 1) r2 −r
−4−2h(1−s(r−1) ) 2
(ii). On the other hand, if δ = hr(r−1)(2+h+hs(r−1))
h2 r2 sδ(r − 1)2
+ . (29) holds then one has λ1 | −4−2h(1−s(r−1)2 ) = −1 but
1 + h + hs(r − 1) δ= hr(r−1)(2+h+hs(r−1))
6 A. Q. KHAN AND S. NOOR-UL-HUDA NAQVI

λ2 | −4−2h(1−s(r−1) ) 2 = 2+h(1+s(r−1))(1+hrs(r−1))
= 1 or where δ = δ ∗ + and 0 < |δ ∗ | 1. For the perturba-
δ= hr(r−1)(2+h+hs(r−1)) (1+h+hs(r−1))(2+h+hs(r−1))
tion system (35), the roots of V|ϕ2 are
− 1 which implies that at ϕ2 eigenvalues criterion for

the existence of flip bifurcation holds if (h, s, δ, r) passes
1( ) ± ι 4 2( ) − 1( )
2
through the curve F |ϕ2 := {(h, s, δ, r), δ = λ1,2 := , (36)
−4−2h(1−s(r−1)2 ) 2
hr(r−1)(2+h+hs(r−1)) }. 
1−hrs(r−1) ∗
where 1( )= 1+h+hs(r−1) + 1 + hr(δ + )(r − 1)
and 2 ( ) = ( 1+h+hs(r1−hrs(r−1) ∗
3. Analysis of a flip bifurcation for semitrivial − 1) )(1 + hr(δ + )(r − 1)) +
fixed point h2 r2 s(δ ∗ + )(r−1)2
1+h+hs(r−1) . Now due to the fact that
Based on the local dynamic study for semitrivial fixed (h, s, δ, r) ∈ N |ϕ2 , the characteristic equation (36)
point ϕ1 in Section 2, here we investigate flip bifurcation of V|ϕ2 has two conjugate complex root with |λ1,2 | =

 1−hrs(r−1) 
by bifurcation theory [18–27].  1+h+hs(r−1) (1 + hr(δ ∗ + )(r − 1))
 , and subseq-
2 2 s(δ ∗ + )(r−1)2
Theorem 3.1: If (h, s, δ, r) ∈ F |ϕ1 then for ϕ1 the flip + h r1+h+hs(r−1)
bifurcation does not take place in the discrete-time model uently one gets the following quantity by noticing that
2
(13). δ = 1−s+sr
r2 −r
:

Proof: Since model (13) is invariant under y = 0, and so d|λ1,2 | hr(r − 1)


| =0 = = 0. (37)
one restrict it on the line y = 0 where it becomes d 2 (1 + h + hs(r − 1))
xt (1 + h) Moreover for = 0, the characteristics roots of (36)
xt+1 = . (30)
1 + hxt must not be found at the intersections of unit cir-
From (30), one has cle with coordinate axes. Therefore, one need to sup-
pose that λ1 = 1 and λ2 = 1 ( = 1, 2, 3, 4) for =
x(1 + h) 1−hrs(r−1)
f (h, x) := . (31) 0, which is equivalent to 1 (0) = 1+h+hs(r−1) +1+
1 + hx 2
hrδ(r − 1) = −2, 0, 1, 2. But if δ = 1−s+sr
r2 −r
holds, then
Finally, if x = x ∗ = 1 and h = h∗ = δ(r−1)
2
, from (31), one (0) = (0) =
 −2,
2 1, and so, 1 2. Therefore, 1 (0) =
has 1−hrs(r−1)
 1+h+hs(r−1) + 1 + hrδ(r − 1) = 0, 1, and so, by straight-
∂f  δ(r − 1) forward calculation one obtains
 := = −1, (32)
∂x ∗ 2 ∗
h = δ(r−1) , δ(r − 1) + 2
x =1 √
−2 − h(2 + r) ± −4 − 8r + hr(−4 + hr)
 s = ,
∂ 2 f  −4δ(r − 1) 2h(r2 − 1)
:= = 0, (33) √
∂x 2  h∗ δ(r−1)
2
, x ∗ =1 (δ(r − 1) + 2)2 −1 − h(2 + r) ± −3 − 4r + hr(−2 + hr)
. (38)
and 2h(r2 − 1)

∂f  Now to transform ϕ2 one use the following transforma-
:= 0. (34)
∂h h∗ = 2 ∗ tions:
δ(r−1) ,x =1

From (32) and (34), one can be concluded that the flip ut = xt − x ∗ , vt = yt − y∗ , (39)
bifurcation does not take place if (h, s, δ, r) ∈ F |ϕ1 . 
with x ∗ = 1 − s + sr and y∗ = (1−r)(1−s+sr)
r . From (39)
and (35), we get
4. Analysis of a Neimark–Sacker bifurcation ⎧
for interior fixed point ⎪ (ut + x ∗ )(1 + h)

⎪ut+1 =

⎪ 1 + h(ut + x ∗ )
Based on the local dynamic study for the equilibrium ϕ2 ⎪



⎨− hs(ut + x ∗ )(vt + y∗ )
in Section 2, we investigate the Neimark–Sacker bifur- − x∗ ,
((ut + x ∗ ) + (vt + y∗ ))(1 + h(ut + x ∗ ))
cation if (h, s, δ, r) ∈ N |ϕ2 where δ is regarded as a bifur- ⎪


⎪vt+1 = (vt + y∗ ) + h(δ ∗ + )(vt + y∗ )
cation parameter. ⎪
⎪ 

⎪ (ut + x ∗ )

⎩ −r + − y∗ .
Theorem 4.1: At ϕ2 , discrete model (13) undergoes N–S (ut + x ∗ ) + (vt + y∗ )
bifurcation if (h, s, δ, r) ∈ N |ϕ2 . (40)

Proof: Recall that if δ is a bifurcation parameter then If = 0, then we will derive normal form of (40). On
perturbation system of discrete model (13) takes the expanding (40) by Taylor series up to second-order at
form origin, one gets

⎪ xt (1 + h) hsxt yt

⎨xt+1 = − , ut+1 = 111 ut + 112 vt + 113 u2t + 114 ut vt + 115 vt2 ,
1 + hxt (xt + yt )(1 + hxt )
 (35) vt+1 = 211 ut + 212 vt + 213 u2t + 214 ut vt + 215 vt2 ,

⎪ xt
⎩yt+1 = yt + h(δ ∗ + )yt −r + , (41)
xt + yt
JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE 7

where its coefficients are mentioned in (A1). Now (41) 1


11 = (r11 + r13 + ι(r21 + r23 )) ,
takes the form: 2
1
xt+1 = ηxt − ζ yt + P̄(xt , yt ), 20 = (r11 − r13 + r22 + ι(r21 − r23 − r12 )) ,
(42) 4
yt+1 = ζ xt + ηyt + Q̄(xt , yt ),
21 = 0. (47)
by following matrix transformation
  From (45) the condition  = 0 holds as (h, s, δ, r) ∈ N |ϕ2

ut 112 0 xt then Neimark–Sacker bifurcation takes place, and addi-
:= , (43)
vt η − 11 −ζ
1 yt tionally supercritical (subcritical) Neimark–Sacker bifur-
cation take place if  < 0 ( > 0). 
where
P̄(xt , yt ) = r11 xt2 + r12 xt yt + r13 yt2 , 5. Analysis of a flip bifurcation for interior
(44)
Q̄(xt , yt ) = r21 xt2 + r22 xt yt + r23 yt2 , fixed point
and the quantities η, ζ and rij (i, j = 1, 2, 3) are depicted Based on the local dynamic study for the equilibrium
in (A2). Furthermore, from (44) we get ∂∂xP̄2 |ϕ2 =(0,0) =
2
ϕ2 in Section 2, we investigate the flip bifurcation if
t (h, s, δ, r) ∈ F |ϕ2 where δ is regarded as a bifurcation
2r11 , ∂x∂t ∂y |ϕ =(0,0) = r12 , ∂∂yP̄2 |ϕ2 =(0,0) = 2r13 , ∂∂xP̄3 |ϕ2 =(0,0)
2 P̄ 2 3
t 2 t t
parameter.
= 0, ∂x∂2 ∂y |ϕ2 =(0,0) = 0, ∂x∂ ∂yP̄ 2 |ϕ2 =(0,0) = 0, ∂∂yP̄3 |ϕ2 =(0,0) =
3
P̄ 3 3

t
∂ 2 Q̄
t t t t Theorem 5.1: At ϕ2 , discrete model (13) undergoes flip
= 2r21 , ∂x∂ t ∂y |ϕ =(0,0) = r22 , ∂∂yQ̄2 |ϕ2 =(0,0)
2 Q̄ 2
0, |
∂xt2 ϕ2 =(0,0) t 2 bifurcation if (h, s, δ, r) ∈ F |ϕ2 .
t
= 2r23 , ∂∂xQ̄3 |ϕ2 =(0,0) = ∂x∂2 ∂y |ϕ2 =(0,0) = ∂x∂ ∂y
3 3 Q̄ 3 Q̄
2 |ϕ2 =(0,0) =

∂ 3 Q̄
t t t t t
Proof: Recall that if δ is a bifurcation parameter then
|
∂yt3 ϕ2 =(0,0)
= 0. Now the following discriminatory quan- the discrete model (13) takes the form (35), and by (39)
tity should not be zero in order to guarantee the occur- it further becomes
rence of Neimark–Sacker bifurcation at ϕ2 of system
(42): ut+1 = 111 ut + 112 vt + 113 u2t + 114 ut vt + 115 vt2 ,

(1 − 2λ̄)λ̄2 1 vt+1 = 211 ut + 212 vt + 213 u2t + 214 ut vt + 215 vt2
 := − 11 20 − 11 2
1−λ 2 + k1 ut + k2 vt + k3 u2t
2
 
− 02 + λ̄21 , (45)
+ k4 ut vt + k5 vt2 , (48)
where
 where it coefficients are depicted in (A3). Now using
1 ∂ 2 P̄ ∂ 2 P̄ ∂ 2 Q̄ following transformation:
02 = − 2 +2
8 ∂xt 2 ∂yt ∂xt ∂yt ⎛
 1
  ut
∂ 2 Q̄ ∂ 2 Q̄ ∂ 2 P̄  := ⎝ 2 − hs(−1 + r)2 + h

+ι − + 2  , vt
∂xt2 ∂yt2 ∂xt ∂yt  hsr2
ϕ2 =(0,0) ⎞
   1 
1 ∂ P̄ ∂ P̄
2 2 ∂ Q̄ ∂ Q̄
2 2  xt
 2(1 − r)(1 + h(1 + sr − s)) ⎠ , (49)
11 = + + ι +  , yt
4 ∂xt 2 ∂yt2 ∂xt 2 ∂yt2  r(2 + h − hs + hsr)
ϕ2 =(0,0)

1 ∂ 2 P̄ ∂ 2 P̄ ∂ 2 Q̄ Equation (48) gives
20 = − 2 +2
8 ∂xt 2 ∂yt ∂xt ∂yt    
xt+1 −1 0 xt P(u , v , )
 
 = +  t t , (50)
∂ 2 Q̄ ∂ 2 Q̄ ∂ 2 P̄  yt+1 0 λ2 yt Q(ut , vt , )
+ι − − 2  ,
∂xt2 ∂yt2 ∂xt ∂yt 
ϕ2 =(0,0) where
 ⎧ ⎫
1 ∂ 3 P̄ ∂ 3 P̄ ∂ 3 Q̄ ∂ 3 Q̄ ⎪ 2hrs(r − 1)(1 + h + hsr − hs) ⎪
21 = + + + ⎪
⎪ ⎪
16 ∂xt3 ⎪
⎪ + + − + + 2 − hs) ⎪ ⎪

∂yt3 ∂xt2 ∂yt ∂yt3 ⎪ 4 h(1 s(r 1))(4 h hsr ⎪

⎪  1 2  ⎪

  ⎪
⎪  + 1
+  1 2 ⎪

∂ 3 Q̄ ∂ 3 Q̄ ∂ 3 P̄ ∂ 3 P̄  ⎪
⎪ 13 u t 14 ut v t 15 v t ⎪

 ⎪
⎪ ⎪

+ι + − −  . ⎪
⎨ ⎪

∂xt 3 ∂xt ∂yt 2 ∂xt ∂yt
2 ∂yt 3  hr s (2 + h(1 + sr − s))
2
ϕ2 =(0,0) 
P= + ,
(46) ⎪
⎪ 4 + h(1 + s(−1 + r)) ⎪


⎪ ⎪


⎪ (4 + h + hs(r2 − 1)) ⎪


⎪ ⎪

Using partial derivatives, (46) becomes ⎪
⎪ ⎪


⎪ × ( 2
u 2
+  2
u v +  2
v 2
+ k u ⎪


⎪ 13 t 14 t t 15 t 1 t ⎪

1 ⎪
⎩ ⎪

02 = (r11 − r13 + r22 + ι(r21 − r23 + r12 )) , 2
+ k2 vt + k3 ut + k4 ut vt + k5 vt ) 2
4
8 A. Q. KHAN AND S. NOOR-UL-HUDA NAQVI

⎧ ⎫
⎪  − (2 + h(1 + sr − s))  ⎪

⎪ ⎪


⎪ −2 + h(−1 + s(r − 1) 2) ⎪
⎪ 

⎪ ⎪
⎪ 2 − hs(−1 + r)2 + h
2

⎪ + + − + + − ⎪
⎪ vt2 = xt2

⎪ 4 h(1 s(r 1))(4 h hs(r 2 1)) ⎪

⎪ 
⎪  ⎪
⎪ hsr2

⎪  1 2
+  1
+  1 2 ⎪
⎪ 


u
13 t u v
14 t t v
15 t ⎪

⎨ ⎬ 2(1 − r)(1 + h(1 + sr − s)) 2 2
 hr s (2 + h(1 + sr − s))
2 + yt
Q= , r(2 + h − hs + hsr)

⎪ − ⎪


⎪ 4 + h(1 + s(−1 + r)) ⎪
⎪ 

⎪ ⎪
⎪ 2 − hs(−1 + r)2 + h

⎪ (4 + h + hs(r 2 − 1)) ⎪
⎪ +2

⎪ ⎪
⎪ hsr2

⎪ ⎪
⎪ × (2 u2 + 2 ut vt + 2 v 2 + k1 ut ⎪
⎪ ⎪
⎪ 

⎪ 13 t 14 15 t ⎪
⎪ 2(1 − r)(1 + h(1 + sr − s))

⎩ ⎪
⎭ × xt yt . (52)
2
+ k2 vt + k3 ut + k4 ut vt + k5 vt ) 2 r(2 + h − hs + hsr)
(51)
Now we calculate the centre manifold F C O at O in a small
neighborhood of = 0 where mathematically it can be
ut = xt + yt , expressed as
2 − hs(−1 + r)2 + h
vt = xt F C O = {(xt , yt ) : yt = v0 + v1 xt2 + v2 xt + v3 3
hsr2  
2(1 − r)(1 + h(1 + sr − s)) + O (|xt | + | |)3 }, (53)
+ yt ,
r(2 + h − hs + hsr)
with
u2t = xt2 + yt2 + 2xt yt , ⎧ ⎧ ⎛ ⎫

⎪ ⎪
⎪ − (2 + h(1 + sr − s)) ⎪

2 − hs(−1 + r)2 + h 2 ⎪
⎪ ⎪
⎪ ⎜   ⎪

ut vt = ⎪
⎪ ⎪
⎪ ⎜ ⎪

xt ⎪ ⎪
⎪ 1 ⎜ −2 + h(−1 + s(r − 1) ) 2


hsr2 ⎪
⎪ ⎪
⎪ ⎜ ⎪


⎪ ⎪
⎪ − λ ⎜ ⎪

2(1 − r)(1 + h(1 + sr − s)) ⎪
⎪ ⎪

1 2
⎝ 4 + h(1 + s(r − 1)) ⎪

+ xt yt ⎪
⎪ ⎪
⎪ ⎪

r(2 + h − hs + hsr) ⎪
⎪ ⎪
⎪ ⎪


⎪ ⎪ (4 + h + hs(r 2 − 1)) ⎪

⎪ ⎪
⎪  ⎪

2 − hs(−1 + r)2 + h ⎪
⎪ ⎪
⎪ 2 − hs(−1 + r) 2+h ⎪

+ ⎪
⎪ ⎪
⎪ ×  1
+  1 ⎪

xt yt ⎪
⎪ ⎪
⎪ 13 2 14 ⎪

hsr2 ⎪
⎪ ⎪
⎪ 
hsr  ⎪


⎪ ⎪
⎪ 2 ⎪

2(1 − r)(1 + h(1 + sr − s)) 2 ⎪ ⎨ 2 − hs(−1 + r)2 + h ⎬
+ yt , ⎪
⎪ +  1
r(2 + h − hs + hsr) ⎪
⎪ v = hsr 2 15 ,


1

⎪ ⎪

 ⎪
⎪ ⎪
⎪ hr s (2 + h(1 + sr − s))
2 ⎪

2 − hs(−1 + r)2 + h
2 ⎪
⎪ ⎪
⎪ ⎪

2 ⎪
⎪ ⎪
⎪ − ⎪

vt = xt2 ⎪
⎪ ⎪
⎪ 4 + h(1 + s(r − 1)) ⎪

hsr2 ⎪
⎪ ⎪
⎪ ⎪


⎪ ⎪
⎪   ⎪

 ⎪
⎪ ⎪
⎪ 4 + h + hs(r 2 − 1) ⎪

2(1 − r)(1 + h(1 + sr − s)) 2 2 ⎨ ⎪
⎪  ⎪

+ yt ⎪
⎪ 2 − hs(−1 + r) 2 + h ⎪

r(2 + h − hs + hsr) ⎪ ⎪
⎪ × 13 + 2
 2 ⎪


⎪ ⎪
⎪ 2 14 ⎪

 ⎪ ⎪
⎪ 
hsr  ⎪

2 − hs(−1 + r)2 + h ⎪
⎪ ⎪
⎪ 2 ⎪

+2 ⎪
⎪ ⎪
⎪ 2 − hs(−1 + h) 2 + h ⎪

hsr2 ⎪
⎪ ⎪
⎩ +  2



⎪ hsr 2 15
 ⎪
⎪ ⎧ ⎫
2(1 − r)(1 + h(1 + sr − s)) ⎪
⎪ ⎪ 1 ⎪
× xt yt , ⎪
⎪ ⎪
⎪ ⎪

r(2 + h − hs + hsr) ⎪
⎪ ⎪
⎪ − λ ⎪


⎪ ⎪
⎪ ⎛ 1 2 ⎪


⎪ ⎪
⎪ ⎪

ut = xt + yt , ⎪
⎪ ⎪
⎪ ⎪


⎪ ⎪
⎪ ⎜ ⎪


⎪ ⎨ ⎜ −hr 2 (2 + + − ⎬
u2t = xt2 + yt2 + 2xt yt , ⎪
⎪ ⎜ s h(1 sr s))

⎪ v = ⎜ ,
⎪ 2 ⎪
⎪ ⎪ ⎜ 4 + h(1 + s(r − 1)) ⎪ ⎪
2 − hs(−1 + r)2 + h ⎪
⎪ ⎪
⎪ ⎝   ⎪ ⎪

⎪ ⎪
⎪ ⎪

vt = xt ⎪
⎪ ⎪
⎪ 4 + h + hs(r 2 − 1)


hsr2 ⎪
⎪ ⎪
⎪  ⎪


⎪ ⎪
⎪ 2 − hs(−1 + r) 2 + h ⎪

2(1 − r)(1 + h(1 + sr − s)) ⎪
⎪ ⎪
⎩ × k + k ⎪

+ yt , ⎪
⎪ 1 2 2
r(2 + h − hs + hsr) ⎪
⎪ hsr

v0 = v3 = 0.
2 − hs(−1 + r)2 + h 2
ut vt = xt (54)
hsr2
2(1 − r)(1 + h(1 + sr − s)) Finally, one write (50) restrict to F c O as
+ xt yt
r(2 + h − hs + hsr)
f1 (xt ) = −xt + m1 xt2 + m2 xt + m3 xt2 + m4 xt 2
2 − hs(−1 + r)2 + h  
+ xt yt
hsr2 + m5 xt3 + O (|xt | + | |)4 , (55)
2(1 − r)(1 + h(1 + sr − s)) 2
+ yt ,
r(2 + h − hs + hsr)
JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE 9

where

⎧ ⎧ ⎫
⎪ ⎪
⎪ 2hrs(r − 1)(1 + h + hs(r − 1)) ⎪


⎪ ⎪ ⎪

⎪ ⎪
⎪ 4 + h(1 + s(r − 1))(4 + h + hs(r 2 − 1)) ⎪


⎪ ⎪
⎪     ⎪


⎪ ⎪
⎪ 2 − hs(−1 + r) + h 2 2 − hs(−1 + r) + h 2 2 ⎪


⎪ ⎪
⎪ ×  1
+  1
 1 ⎪


⎪ ⎪
⎨ 13 14 15 ⎪


⎪ hsr 2 hsr 2

⎪ 1
m = ,

⎪ ⎪ hr2 s (2 + h(1 + sr − s)) ⎪

⎪ ⎪
⎪ + ⎪

⎪ ⎪
⎪ ⎪


⎪ ⎪
⎪ 4 + h(1 + s(−1 + r))(4 + h + hs(r2 − 1))
  ⎪⎪


⎪ ⎪
⎪ 2 − hs(−1 + r) + h 2 2 − hs(−1 + r) + h 2 2 ⎪


⎪ ⎪
⎪ 2 2 2 ⎪


⎪ ⎪
⎩ ×  13 +  14  15 ⎪ ⎭

⎪ hsr 2 hsr 2

⎪ 

⎪ hr s (2 + h(1 + sr − s))
2 2 − hs(−1 + r) + h 2

⎪m2 = k1 + k2 ,

⎪ 4 + h(1 + s(−1 + r))(4 + h + hs(r − 1)) 2 hsr2

⎪ ⎧ ⎫

⎪ ⎪ 2hrs(r − 1)(1 + h + hs(r − 1)) ⎪

⎪ ⎪
⎪ ⎪


⎪ ⎪ 4 + h(1 ⎪
⎪  + s(r − 1))(4 +h + hs(r − 1)) ⎪
2

⎪ ⎪
⎪ ⎪


⎪ ⎪
⎪ 2(1 − r)(1 + h(1 + sr − s)) 2 − hs(−1 + r)2 + h ⎪


⎪ ⎪
⎪ × 2 1
v +  1
v + ⎪


⎪ ⎪
⎪ 13 2 14 2
r(2 + h − hs + hsr) hsr2 ⎪


⎪ ⎪
⎪    ⎪


⎪ ⎪
⎪ 2 − hs(−1 + r) + h 2 2(1 − r)(1 + h(1 + sr − s)) ⎪


⎪ ⎪
⎪ + 115 v2 2 ⎪


⎪ ⎪
⎪ ⎪


⎪ ⎪
⎪ hsr 2 r(2+ h − hs + hsr)  ⎪


⎪ ⎪
⎪ 2 s (2 + h(1 + sr − s)) 2(1 − r)(1 + h(1 + sr − s)) ⎪


⎪ ⎪

hr ⎪


⎪ ⎪
⎪ + 2213 v2 + 214 v2 ⎪


⎪ ⎪
⎪ 4 + h(1 + s(−1 + r))(4 + h + hs(r 2 − 1)) r(2 + h − hs + hsr) ⎪


⎪ ⎪
⎪ ⎪


⎪ ⎪
⎪ ⎪


⎪ ⎪
⎪ ⎪


⎪ ⎨ + 2 − hs(−1 + r) + h
⎪ 2





⎪ m3 = 2

⎪ ⎪  hsr ⎪
,

⎪ ⎪
⎪ 2 − hs(−1 + r)2 + h ⎪


⎪ ⎪
⎪ +  2
v 2 ⎪

⎨ ⎪
⎪ 15 2
hsr2 ⎪


⎪  ⎪


⎪ 2(1 − r)(1 + h(1 + sr − s)) ⎪


⎪ ⎪
⎪ ⎪


⎪ ⎪ r(2 + ⎪




⎪  h − hs + hsr) ⎪



⎪ ⎪
⎪ 2(1 − r)(1 + h(1 + sr − s)) ⎪


⎪ ⎪
⎪ + k1 v 1 + k2 v 1 ⎪


⎪ ⎪
⎪ r(2 + h − hs + hsr) ⎪


⎪ ⎪
⎪  ⎪


⎪ ⎪
⎪ 2 − hs(−1 + r)2 + h ⎪


⎪ ⎪
⎪ + + ⎪


⎪ ⎪

k 3 k4 ⎪


⎪ ⎪
⎪ 
hsr2  ⎪


⎪ ⎪
⎪ 2 ⎪


⎪ ⎪
⎪ 2 − hs(−1 + r)2 + h ⎪


⎪ ⎪ +
⎩ k5 ⎪


⎪ hsr2

⎪ ⎧ ⎫

⎪ ⎪ hr2 s (2 + h(1 + sr − s)) ⎪

⎪ ⎪
⎨ ⎪




⎪ = 4 + h(1 + s(−1 + r))(4 + h + hs(r 2 − 1))

⎪m 4
⎪ 2(1 − r)(1 + h(1 + sr − s)) ⎪
,

⎪ ⎪
⎩ + ⎪


⎪ k v
1 2 k v
2 2

⎪ ⎧ r(2 + h − hs + hsr) ⎫

⎪ 2hrs(r − 1)(1 + h + hs(r − 1))

⎪ ⎪
⎪ ⎪


⎪ ⎪
⎪ ⎪


⎪ ⎪ 4 + h(1
 + s(r − 1))(4 +h + hs(r2 − 1)) ⎪

⎪ ⎪
⎪ ⎪

⎪ ⎪
⎪ 2(1 − r)(1 + h(1 + sr − s)) 2 − hs(−1 + r)2 + h ⎪


⎪ ⎪
⎪ × 2 1
v +  1
v + ⎪


⎪ ⎪

13 1 14 1
r(2 + h − hs + hsr) hsr 2 ⎪


⎪ ⎪
⎪    ⎪


⎪ ⎪
⎪ 2 − hs(−1 + r) 2 + h 2(1 − r)(1 + h(1 + sr − s)) ⎪


⎪ ⎪
⎨ +  1
v 2 ⎪


⎪ 15 1 2 + − +

⎪ =
hsr r(2  h hs hsr)


m 5
⎪ hr s (2 + h(1 + sr − s))
2

.

⎪ ⎪
⎪ + 2 ⎪


⎪ ⎪
⎪ 4 + h(1
2 13 v 1 ⎪

⎪  + s(−1 + r))(4 + h + hs(r − 1))
2

⎪ ⎪
⎪ ⎪


⎪ ⎪
⎪ 2(1 − r)(1 + h(1 + sr − s)) 2 − hs(−1 + r)2 + h ⎪


⎪ ⎪
⎪ +  2
+ ⎪


⎪ ⎪
⎪ 14 v 1
+ − + 2 ⎪


⎪ ⎪
⎪   r(2 h hs hsr)  hsr ⎪


⎪ ⎪
⎪ − + 2 + − + + − ⎪


⎩ ⎪
⎩ 2
+ 15 v1 2
2 hs(−1 r) h 2(1 r)(1 h(1 sr s)) ⎪

hsr2 r(2 + h − hs + hsr)
(56)

Finally, in order for the existence of flip bifurcation at  


 2
ϕ2 , the following two discriminatory quantities 1 and 1 ∂ 3 f1 1 ∂ 2 f1
2 should not be zero [18, 19]: 2 := + |(0,0) = m5 + m21 . (57)
6 ∂xt3 2 ∂xt2

∂ 2 f1 1 ∂f1 ∂ 2 f1 Consequently, based on above calculation, we can
1 := + |(0,0) = m2 ,
∂xt ∂ 2 ∂ ∂xt2 say that if (h, s, δ, r) ∈ F |ϕ2 and 2 = 0 then at ϕ2 flip
10 A. Q. KHAN AND S. NOOR-UL-HUDA NAQVI

bifurcation takes place. Besides, the period-2 points Moreover, the model (58) is controlled provided that the
bifurcate from ϕ2 of model (13) are stable (unstable) if following matrix
2 > 0 (2 < 0). 


h(−1 + r)(1 − s0 + s0 r)
6. Chaos control  
C = P : VP = ⎝ 1 + h + hs0 (−1 + r)
In the present section, we will apply the following two 0
feedback control strategies to the model (13) to get a ⎞
h(−1 + r)(1 − hrs0 (−1 + r))(1 − s0 + s0 r)
stable trajectory: ⎟
(1 + h + hs0 (−1 + r))2 ⎟,
h δ(−1 + r) (1 − s0 + s0 r)
2 3 ⎠
6.1. By OGY method 1 + h + hs0 (−1 + r)
(62)
In order to study chaos, we fist use OGY method,
which was proposed by Ott et al. [28], for the discrete
model (13). By existing chaos theory [29, 30], one can
is of rank 2. Additionally, it must be understood that
write the model (13) as follows:
δ cannot be used in this situation as a control param-
⎧ eter to utilize the OGY method to the model (13).
⎪ xt (1 + h) hsxt yt t (1+h)

⎨xt+1 = − = f1 (xt , yt , s), As, if we take f1 (xt , yt , δ) = x1+hx − (xt +yhsx t yt
t )(1+hxt )
and
1 + hxt (xt + yt )(1 + hxt ) t
 f2 (xt , yt , δ) = yt + hδyt (−r + xt +yxt
), then Rank(C)

⎪ xt t
⎩yt+1 = yt + hδyt −r + = f2 (xt , yt , s), becomes zero. So, to apply the OGY method to the
xt + yt
model (13), s is taken as a control parameter. Now, if
(58)
one take s − s0 = −T( xytt −x −y ), where T = (k1 k2 ), then the
model (59) can be written as
where s represents the control parameter under which
one can acquire the desired chaos control by small per-
turbations. To do this, one restrict it where s ∈ (s0 −  
xt+1 − x x −x
σ , s0 + σ ) with σ > 0, and s0 indicates nominal value ≈ (V − PT) t . (63)
yt+1 − y yt − y
corresponds to chaotic region. The trajectory is moved
in the direction of the target orbit using the stabilizing
feedback control strategy. Assuming that ϕ2 be unsta-
Furthermore, the corresponding controlled model of
ble equilibrium point of the model (13), which is in the
(13) is
chaotic region created by the occurrence of N-S bifur-
cation, then the given below linear map can be used to
approximate the model (58): ⎧
⎪ xt (1 + h)

⎪xt+1 =
  ⎪
⎪ 1 + hxt
xt+1 − x x −x ⎪
⎨ h(s − k (x − x) − k (y − y))x y
≈ V(x, y, s0 ) t + P(s − s0 ), (59) 0 1 t 2 t t t
yt+1 − y yt − y − , (64)

⎪ (x + y )(1 + hx )


t 
t t

⎪ xt
where ⎩yt+1 = yt + hδyt −r + .
xt + yt
⎛ ⎞
∂f1 (x, y, s0 ) ∂f1 (x, y, s0 )
⎜ ∂x ∂y ⎟
V|(x,y,s0 ) =⎜ ⎟
⎝ ∂f2 (x, y, s0 ) ∂f2 (x, y, s0 ) ⎠ By stability theory ϕ2 is a sink iff roots of V −PT satisfying
|λ1,2 | < 1 where its variational matrix is
∂x ∂y
⎛ ⎞
1 − hrs0 (r − 1) −hr2 s0
⎜ ⎟ ⎛ 1 − hrs0 (r − 1)
= ⎝ 1 + h + hs0 (r − 1) 1 + h + hs0 (r − 1) ⎠ ,
hδ(r − 1)2 1 + hrδ(r − 1) ⎜ 1 + h + hs0 (r − 1)

(60) V − PT = ⎜
⎜−
h(r − 1)(1 − s0 + s0 r)
⎝ k1
1 + h + hs0 (r − 1)
and hδ(r − 1)2

⎛ ⎞ −hr2 s0
∂f1 (x, y, s0 ) ⎛ ⎞ ⎟
h(r − 1)(1 − s0 + s0 r) 1 + h + hs0 (r − 1) ⎟
⎜ ∂s ⎟ ⎝ h(r − 1)(1 − s0 + s0 r) ⎟
P=⎝ ⎠= 1 + h + hs0 (r − 1) ⎠ . ⎟. (65)
∂f2 (x, y, s0 ) − k2 ⎟
0 1 + h + hs0 (r − 1) ⎠
∂s
(61) 1 + hrδ(r − 1)
JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE 11

The characteristic equation of V −PT is 6.2. By hybrid control feedback


⎧ 

⎪ 2 1 − hrs0 (r − 1) h(r − 1)(1 − s0 + s0 r) We use a hybrid control feedback method to control

⎪ λ − − k1

⎪ 1 + h + hs 0 (r − 1) 1 + h + hs0 (r − 1) the chaos due to the emergence of flip bifurcation in



⎪ + 1 + hrδ(r − 1)) λ the model (13) by existing theory [31]. If the model (13)

⎪ 

⎪ 1 − hrs0 (r − 1) undergoes bifurcation at ϕ2 then one can write corre-



⎪ +
⎨ 1 + h + hs0 (r − 1) sponding controlled model as
h(r − 1)(1 − s0 + s0 r) ⎧ 

⎪ − k1 (1 + hrδ(r − 1)) ⎪ xt (1 + h) hsxt yt

⎪ 1 + h + hs0 (r − 1) ⎪
⎪xt+1 = α −

⎪ ⎪
⎪ + (x +

⎪ 2
hr s0 ⎨ 1 hxt t y t )(1 + hxt )

⎪ +hδ(r − 1)2

⎪ 1 + h + hs0 (r − 1) +(1 − α)xt ,

⎪ ⎪
⎪  

⎪ − − s0 + s0 r) ⎪
⎪ xt

⎪ h(r 1)(1 ⎪
⎩yt+1 = α yt + hδyt −r + + (1 − α)yt ,
⎩ + k2 = 0.
xt + yt
1 + h + hs0 (r − 1)
(66) (72)

If λ1,2 denotes the roots of (66) then where 0 < α < 1 and controlled strategy (72) is a com-
1 − hrs0 (r − 1) bination of feedback control as well as parameter per-
λ1 + λ2 = turbation. The V|ϕ2 is
1 + h + hs0 (r − 1)
h(r − 1)(1 − s0 + s0 r) ⎛ ⎞
− k1 + 1 + hrδ(r − 1), αh(−1 + s − sr2 ) −αhr2 s
1 + h + hs0 (r − 1) 1 +
V|ϕ2 =⎝ 1 + h + hs(r − 1) 1 + h + hs(r − 1) ⎠ .
(67)
αhδ(r − 1)2 1 + αhrδ(r − 1)
and (73)

1 − hrs0 (r − 1)
λ1 λ2 = The characteristics equation of V|ϕ2 evaluated at ϕ2 is
1 + h + hs0 (r − 1)
h(r − 1)(1 − s0 + s0 r) λ2 − 1λ + =0 (74)
− k1 (1 + hrδ(r − 1)) 2
1 + h + hs0 (r − 1)
 where
2 hr2 s0
+ hδ(r − 1)
1 + h + hs0 (r − 1) αh(−1 + s − sr2 )
=2+ + αhδr(r − 1),
h(r − 1)(1 − s0 + s0 r) 1
1 + h + hs(r − 1)
+ k2 . (68) 
1 + h + hs0 (r − 1) αh(−1 + s − sr2 )
2 = 1+ (1 + αhrδ(r − 1))
Next we take λ1 = ±1 and λ1 λ2 = 1 to get lines of 1 + h(1 + s(r − 1))
marginal stability for (64). Also, these restriction give α 2 h2 r2 sδ(r − 1)2
the fact that characteristics roots satisfying |λ1,2 | < 1. + . (75)
1 + h(1 + hs(r − 1))
Assuming that λ1 λ2 = 1 then from (68) one has
So, based on stability theory one has the result:
L1 : h(r − 1)(1 − s0 + s0 r)(1 + hrδ(r − 1))k1
− h2 δ(r − 1)3 (1 − s0 + s0 r)k2 Lemma 1: Equilibrium ϕ2 of model (72) is a sink iff |2 +
αh(−1+s−sr2 ) αh(−1+s−sr2 )
1+h+hs(r−1) + αhδr(r − 1)| < 1 + (1 + 1+h+hs(r−1) )
2
+ h − hrδ(r − 1) + hs0 (r − 1) = 0. (69)
α 2 h2 r2 sδ(r−1)2
If λ1 = 1 then from (67) and (68), one gets (1 + αhrδ(r − 1)) + 1+h+hs(r−1) < 2.

L2 : h2 rδ(r − 1)2 (1 − s0 + s0 r)k1 Proof: The necessary and sufficient condition for roots
2 3
− h δ(r − 1) (1 − s0 + s0 r)k2 of (74) satisfying |λ1,2 | imply that ϕ2 of model (72)
is a sink iff |2 + αh(−1+s−sr
2)

+ h2 rδ(r − 1) (1 − s0 + s0 r)) = 0. (70) 1+h+hs(r−1) + αhδr(r − 1)| < 1 + (1 +


αh(−1+s−sr2 ) α 2 h2 r2 sδ(r−1)2
1+h+hs(r−1) )(1 + αhrδ(r − 1)) + 1+h+hs(r−1) < 2. 
Finally, if λ1 = −1 then from (67) and (68), one gets
L3 : h(r − 1)(1 − s0 + s0 r)(2 + hrδ(r − 1))k1 7. Numerical simulations
2 3
− h δ(r − 1) (1 − s0 + s0 r)k2 − 4 Example 7.1: If h = 0.6, s = 1.37, r = 0.5, δ = [0.1, 1.1]
− 2h + 2hs0 (r − 1) 2 with (x0 , y0 ) = (0.03, 0.04) then at δ = 0.1100000000
0000032 discrete model (13) undergoes the N-S bifurca-
− hrδ(−1 + r) (2 + h(1 + s0 r − s0 )) = 0. (71)
tion. The bifurcation diagrams and Maximum Lyapunov
Then, stable eigenvalues lie within the triangular region exponent are drawn in Figure 1. Further, at (h, s, r, δ) =
(k1 k2 )-plane bounded by the straight lines L1,2,3 for (0.6, 1.37, 0.5, 0.11000000000000032) model (13) has
parametric values h, s, r and δ. interior equilibrium point ϕ2 = (0.31499999999999995,
12 A. Q. KHAN AND S. NOOR-UL-HUDA NAQVI

Figure 1. Bifurcation diagrams (B.Ds) with MLE of discrete model (13). (a) B.D for xt . (b) B.D for yt . (c) B.D for xt and yt . (d) B.D for s
and xt . (e) B.D for s and yt . (f) MLEs.

0.31499999999999995) and moreover from (24) one ∈ N |ϕ2 =(0.31499999999999995,0.31499999999999995) . Further-


gets: more, by fixing h = 0.6, s = 1.37, r = 0.5, and varying
the value of bifurcation parameter δ < 0.11000000000
V|ϕ2 = (0.31499999999999995, 0.31499999999999995) 000032 then one can obtain that respective interior
 equilibrium solution is a stable focus. To verify this if δ =
1.0138772077375944
= 0.10000 < 0.11000000000000032 then Figure 2(a) indi-
0.016500000000000046
cates that ϕ2 = (0.31499999999999995, 0.314999999
−0.17283431455004208 99999995) is a stable focus. Similarly if one varies δ =
, (76)
0.9834999999999999 0.10299, 0.10500, 0.10880, 0.10987, 0.10999 < 0.11000
000000000032 then Figure 2(b–f) also show that, for
with
each variation of δ, the equilibrium solution ϕ2 =
λ2 − 1.9973772077375944λ + 1 = 0. (77) (0.31499999999999995, 0.31499999999999995) is also
stable focus. On the other hand, if one varies the
The roots of (77) are λ1,2 = 0.9986886038687972 ± bifurcation parameter δ > 0.11000000000000032 then
0.05119641103234161ι with |λ1,2 | = 1, and so from (i) we can concluded that respective interior equilibrium
of Theorem 2.5 one has the Neimark–Sacker bifurcation solution is an unstable focus, and as a result supercrit-
set (h, s, r, δ) = (0.6, 1.37, 0.5, 0.11000000000000032) ical Neimark–Sacker bifurcation must take place. For
JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE 13

Figure 2. Phase portrait of discrete model (13) with (0.03, 0.04). (a) If δ = 0.10000. (b) If δ = 0.10299. (c) If δ = 0.10500. (d) If δ =
0.10880. (e) If δ = 0.10987. (f) If δ = 0.10999.

instance, if h = 0.6, s = 1.37, r = 0.5 then from (37) the and


d|λ |
non-degenerate conditions, i.e. d1,2 | =0 = −0.06307
821698906643 = 0 holds, and moreover if δ = 0.11000
000000000032 then from (A1) and (A2) one gets: η = 0.9986886038687973,
(79)
⎧ ζ = 0.05119641103234052.

⎪111 = 1.0138772077375944,



⎪1 = −0.17283431455004208,




12
Utilizing (78) and (79) into (A2), one gets:

⎪  1 = −0.2372883502540626,

⎪ 13



⎪  1 = −0.46146372048017426,



14
⎨1 = 0.2743401818254637, ⎪
⎪ r11 = 3.2139366880763243,
15 ⎪

(78) ⎪
⎪ = 0.021156708511249955,

⎪  2 = 0.016500000000000046, ⎪r12


⎪ 11 ⎪
⎨r

⎪ = −0.0041604325437973955,
⎪12 = 0.9834999999999999,
2
⎪ 13
(80)

⎪ ⎪

⎪213 = −0.02619047619047627, ⎪
⎪r21 = −0.0014241879052949461,

⎪ ⎪


⎪2 = 0.05238095238095254, ⎪
⎪r22 = −0.0634844121741236,

⎪ ⎪



14 ⎩
⎩2 = −0.02619047619047627, r23 = 0.0025751472836634173.
15
14 A. Q. KHAN AND S. NOOR-UL-HUDA NAQVI

Figure 3. Invariant closed curves of discrete model (13) with (0.03, 0.04). (a) If δ = 0.110999. (b) If δ = 0.125670. (c) If δ = 0.130000.
(d) If δ = 0.139999. (e) If δ = 0.141100. (f) If δ = 0.149999. (g) If δ = 0.158896. (h) If δ = 0.161231. (i) If δ = 0.167890.

Now in view of (80) and (47), one obtains 0.161231, 0.167890 > 0.11000000000000032 then sta-
ble invariant curves also appears which are depicted

⎪ in Figure 3(b–i). Therefore, in conclusion we can say
⎪02 = 0.7886531771114995


⎪ that numerical simulations in Example 7.1 agree with

⎪+0.004289343330572898ι,

⎪ theoretical results obtained in Theorem 2.3, (i) of



⎨11 = 1.6048881277662634
⎪ Theorem 2.5 and Theorem 4.1.

⎪+0.0005754796891842356ι, (81)

⎪ Example 7.2: If h = 1.28, s = 2.5, r = 0.798, δ = [10.1,

⎪ 20 = 0.7886531771114995

⎪ 16.99] with (x0 , y0 ) = (0.495, 0.125301) then at δ =



⎪−0.0062890109250520795ι,

⎪ 11.591710484782386 discrete model (13) undergoes

21 = 0. the flip bifurcation. The flip bifurcation diagrams and
Maximum Lyapunov exponent are drawn in Figure 4.
Now finally, using (81) along with λ, λ̄ = 0.9986886038 Further at (h, s, r, δ) = (1.28, 2.5, 0.798, 11.5917104847
687973 ± 0.05119641103234052ι into (45) one gets: 82386) discrete model (13) has equilibrium solution
 = 3.9314493377802 > 0 which give the fact that ϕ2 = (0.4950000000000001, 0.12530075187969925) and
closed invariant curve must exists, and hence discrete moreover from (24) one gets:
model undergoes subcritical Neimark–Sacker at indi- 
0.9279059745347696
cated interior fixed point ϕ2 = (0.31499999999999995, V|ϕ2 =
0.6054248379149572
0.31499999999999995) (see Figure 3(a)). In similar man-
ner, we can also obtain that if one varies δ = 0.125670, −1.2474123408423115
, (82)
0.130000, 0.139999, 0.141100, 0.149999, 0.158896, −1.3917278250303777
JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE 15

with λ1 = −1 and λ2 = 0.5361781495043931 = 1 or equations of V|ϕ2 =(0.21999999999999997,0.32999999999999996)


− 1, and hence based on these simulations one are λ1,2 = 0.9977207207207206 ± 0.0674786147199
can obtain that for parametric values (h, s, r, δ) = 6602ι with |λ1,2 | = 1. In order to apply OGY control
(1.28, 2.5, 0.798, 11.591710484782386) ∈ method for model (13), if s0 = 0.3 then it has equilib-
F |ϕ2 =(0.4950000000000001,0.12530075187969925) . Moreover in rium ϕ2 = (0.82, 1.2299999999999998). So, model (64)
this parametric domain, from (A3), (54) and (56) one becomes
gets: ⎧

⎪ xt+1 = xt + 0.5xt (1 − xt )
⎧ ⎪

⎪ ⎪


⎪ 111 = 0.9279059745347696, ⎪
⎪ 0.5(0.3 − k1 (xt − 0.82)

⎪ ⎪


⎪ 112 = −1.2474123408423115, ⎪
⎨ −k2 (yt − 1.2299999999999998))xt yt



⎪ − , (86)

⎪ 113 = −0.5982005795778922, ⎪ xt + yt

⎪ ⎪


⎪ 114 = −0.040685691442227376, ⎪
⎪ yt+1 = yt + 0.5(0.38333333333333336)

⎪ ⎪


⎪ ⎪ 


⎪ 115 = 2.0109798949336652, ⎪
⎪ xt

⎪ ⎩yt −0.4 + ,

⎪ xt + yt

⎪ 211 = 0.6054248379149572,



⎨12 = −1.3917278250303777,
2
⎪ where its variation matrix V −BK is
213 = −0.976018223547749, (83) ⎛

⎪ 0.7347517730496455

⎪ 14 = 7.711510320703999,
2

⎪ V − BK = ⎝+0.17446808510638295k1



⎪ 215 = −15.232141673073743, 0.0689994



⎪ k1 = 0.05222911999999998, ⎞

⎪ −0.017021276595744685



⎪ k2 = −0.20633088, +0.17446808510638295k1 ⎠ . (87)




⎪k3 = −0.0841996722424242,
⎪ 0.9540004



⎪ k4 = 0.6652607767272725,

⎪ Now for marginal stability lines L1 , L2 and L3 are

⎩k = −1.3140547025454543,
5
⎧ L1 : 3.340151253503999k1 + 16.523026880255998k2

⎪ v0 = 0,

⎪ + 13.10398752 = 0, (88)
⎨v = 12.929077600120031,
1

(84) L2 : 11.015351253503999k1 + 16.523026880255998k2

⎪ v2 = 0.10826715884189564,

⎩ − 18.358918755839998 = 0, (89)
v3 = 0,
and
and

⎪ L3 : −4.335048746496k1 + 16.523026880255998k2

⎪ m1 = −21.71571925204092,


⎪ − 10.60110620416 = 0. (90)
⎨m2 = −0.2165343176837912,

m3 = −2.143746543978618, (85) So, lines (88), (89) and (90) determine triangular region



⎪m4 = −0.0011046972084443997, that gives |λ1,2 | < 1(see Figure 6).



⎩m = −27.41483556845985.
5
Example 7.4: Finally, if h = 1.28, s = 2.5, r = 0.798, δ =
Utilizing (85) into (57) one gets: 1 = −0.216534317 11.591710484782386 with (x0 , y0 ) = (1.3, 1.12) then
6837912 = 0 and 2 = 444.1576270650008 > 0. Since model (13) undergoes flip bifurcation. For this, by
2 = 444.1576270650008 > 0 and so it can be con- applying hybrid strategy to get stable orbit at ϕ2 =
cluded that stable period-2 points bifurcate from ϕ2 = (0.4950000000000001, 0.12530075187969925). For this
(0.4950000000000001, 0.12530075187969925). There- model (72) takes the form
fore, in conclusion we can say that numerical simu- ⎧ 

⎪ (1.28)(2.5)xt yt
lations in Example 7.2 agree with theoretical results ⎪
⎪ x t+1 = α xt + 1.28xt (1 − xt ) −

⎪ xt + yt
obtained in Theorem 2.4, (ii) of Theorem 2.5 and ⎪

Theorem 5.1. +(1 − α)xt ,

⎪ yt+1 = α (yt + (1.28)(11.591710484782386)yt



⎪  
Example 7.3: If h = 0.5, r = 0.4, δ = 0.38333333333 ⎪
⎩ −0.798 + xt
xt +yt + (1 − α)yt ,
333336, s ∈ [0.1, 1.33] with (x0 , y0 ) = (0.02, 0.03) then
model (13) undergoes Neimark–Sacker bifurcation. The (91)
maximum Lyapunov exponents with bifurcation dia- where Vϕ2 =(0.4950000000000001,0.12530075187969925) is
grams are drawn in Figure 5. Further if s = 1.3 then

model (13) has equilibrium ϕ2 = (0.21999999999999997, 1 − 0.07209402546523036α
Vϕ 2 =
0.32999999999999996) where roots of characteristics 0.6054248379149572α
16 A. Q. KHAN AND S. NOOR-UL-HUDA NAQVI

Figure 4. Flip B.Ds with MLE of discrete model (13). (a) B.D for xt . (b) B.D for yt . (c) B.D for xt and yt . (d) B.D for h and xt . (e) B.D for h
and yt . (f) B.D for s and xt . (g) B.D for s and yt . (h) MLEs.
JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE 17

Figure 5. B.Ds with MLE of discrete model (13). (a) B.D for xt . (b) B.D for yt . (c) B.D for xt and yt . (d) MLEs.

Figure 7. Plot of t vs. xt and yt for controlled system (58).

+ 0.9276437009912135α 2 = 0. (93)

Furthermore, roots of (93) satisfying |λ1,2 | < 1 if 0 <


α < 1. So, for the allowed interval of control parame-
ter α the flip bifurcation is completely eliminated. If α =
0.999 then for controlled model (91), plots of t vs. xt and
yt are drawn in Figure 7.

8. Conclusion
This work explored the local dynamics at fixed points,
Figure 6. Region of stability where |λ1,2 | < 1. chaos, bifurcations of a discrete prey–predator model
(13) in the region: R2+ = {(x, y) : x, y ≥ 0}. We studied
−1.2474123408423115α local dynamics at semitrivial and interior fixed points
1 − 2.3917278250303777α
, (92) ϕ1,2 of model (13), and explored that ϕ1 of model (13) is a
2 2
sink if h > δ(r−1) , never source, saddle if 0 < h < δ(r−1) ,
with 2
and non-hyperbolic if h = δ(r−1) . Further it proved that
2
λ2 − (2 − 2.463821850495606α)λ + 1 ϕ2 of model (13) is a stable focus if 0 < δ < 1−s+sr
r2 −r
with

s−1 1−s+sr2
− 2.463821850495606α r > max{± s , 1}, an unstable focus if δ > r2 −r
18 A. Q. KHAN AND S. NOOR-UL-HUDA NAQVI


s−1 [8] Singh A, Malik P. Bifurcations in a modified Leslie–Gower
with r > max{± s , 1}, non-hyperbolic condition if
predator–prey discrete model with Michaelis–Menten
1−s+sr2 −4−2h(1−s(r−1)2 )
δ= r2 −r
, stable node if δ > hr(r−1)(2+h+hs(r−1)) with prey harvesting. J Appl Math Comput. 2021;67(1-2):
2+h 2+h 143–174. doi: 10.1007/s12190-020-01491-9
s > max{ h(r−1)2 , h(1−r) }, an unstable node if 0 < δ < [9] Ma R, Bai Y, Wang F. Dynamical behavior analysis of
−4−2h(1−s(r−1) ) 2
2+h 2+h
hr(r−1)(2+h+hs(r−1)) with s > max{ h(r−1)2 , h(1−r) }, and
a two-dimensional discrete predator–prey model with
−4−2h(1−s(r−1) ) 2 prey refuge and fear factor. J Appl Anal Comput.
non-hyperbolic if δ = hr(r−1)(2+h+hs(r−1)) . In order to 2020;10(4):1683–1697.
studied the bifurcation analysis at ϕ1,2 , we first iden- [10] Santra PK, Panigoro HS, Mahapatra GS. Complexity of
tified bifurcation sets for understudied model (i) flip a discrete-time predator–prey model involving prey
2
bifurcation set F |ϕ1 := {(h, s, δ, r), h = δ(r−1) } at ϕ1 , (ii) refuge proportional to predator. Jambura J Math. 2022;4
(1):50–63. doi: 10.34312/jjom.v4i1
Neimark–Sacker bifurcation set N |ϕ2 := {(h, s, δ, r), δ =
[11] Chen J, Chen Y, Zhu Z, et al. Stability and bifurcation of
1−s+sr2
r2 −r
} at ϕ2 (iii) flip bifurcation set whose mathemati- a discrete predator–prey system with Allee effect and
cal expression is F |ϕ2 := {(h, s, δ, r), δ = other food resource for the predators. J Appl Math Com-
−4−2h(1−s(r−1)2 ) put. 2023;69(1):529–548.
hr(r−1)(2+h+hs(r−1)) } at ϕ2 , and then proved that at ϕ1
[12] Tunç C, Tunç O. Qualitative analysis for a variable delay
flip bifurcation did not take place if (h, s, δ, r) ∈ F |ϕ1 but system of differential equations of second order. J Taibah
at ϕ2 model subjected Neimark–Sacker and flip bifur- Univ Sci. 2019;13(1):468–477. doi: 10.1080/16583655.20
cations if (h, s, δ, r) ∈ N |ϕ2 and (h, s, δ, r) ∈ F |ϕ2 , respec- 19.1595359
tively. Furthermore, OGY and Hybrid control strategies [13] Tunç C. Stability to vector Lienard equation with constant
are utilized to control chaos in the under study discrete deviating argument. Nonlinear Dyn. 2013;73(3):1245–
1251. doi: 10.1007/s11071-012-0704-8
model due to occurrence of Neimark–Sacker and flip
[14] Tunç C. A note on boundedness of solutions to a
bifurcations. Finally numerical simulations depicted to class of non-autonomous differential equations of sec-
verify theoretical results. ond order. Appl Anal Discret Math. 20104:361–372.
doi: 10.2298/AADM100601026T
[15] Wikan A. Discrete dynamical systems with an introduc-
Acknowledgments tion to discrete optimization problems. London (UK):
Authors declare that they got no funding on any part of this Bookboon. com; 2013.
research. [16] Kulenović MRS, Ladas G. Dynamics of second order ratio-
nal difference equations: with open problems and con-
jectures. New York: Chapman and Hall/CRC; 2001.
Disclosure statement [17] Camouzis E, Ladas G. Dynamics of third-order rational dif-
ference equations with open problems and conjectures.
No potential conflict of interest was reported by the author(s). New York: CRC Press; 2007.
[18] Guckenheimer J, Holmes P. Nonlinear oscillations, dyna-
ORCID mical systems, and bifurcations of vector fields. New
York: Springer Science & Business Media; 2013.
Abdul Qadeer Khan http://orcid.org/0000-0002-0278-1352 [19] Kuznetsov YA, Kuznetsov IA, Kuznetsov Y. Elements of
applied bifurcation theory. New York: Springer; 1998.
[20] Rana SM. Chaotic dynamics and control of discrete ratio-
References dependent predator–prey system. Discrete Dyn Nat Soc.
[1] Freedman HI. Deterministic mathematical models in 2017;2017:1–13.
population ecology. New York: Marcel Dekker Incorpo- [21] Al-Basyouni KS, Khan AQ. Discrete-time predator–prey
rated; 1980. model with bifurcations and chaos. Math Probl Eng.
[2] Arditi R, Ginzburg LR, Akcakaya HR. Variation in plankton 2020;2020:1–14. doi: 10.1155/2020/8845926
densities among lakes: a case for ratio-dependent pre- [22] Chakraborty P, Ghosh U, Sarkar S. Stability and bifur-
dation models. Am Nat. 1991;138(5):1287–1296. doi: 10. cation analysis of a discrete prey–predator model with
1086/285286 square-root functional response and optimal harvesting.
[3] Akcakaya HR, Arditi R, Ginzburg LR. Ratio-dependent pre- J Biol Syst. 2020;28(01):91–110. doi: 10.1142/S02183390
dation: an abstraction that works. Ecology. 1995;76(3): 20500047
995–1004. doi: 10.2307/1939362 [23] Liu W, Cai D. Bifurcation, chaos analysis and control in
[4] Cosner C, DeAngelis DL, Ault JS, et al. Effects of spatial a discrete-time predator–prey system. Adv Differ Equ.
grouping on the functional response of predators. Theor 2019;2019(1):1–22. doi: 10.1186/1687-1847-2012-1
Popul Biol. 1999;56(1):65–75. doi: 10.1006/tpbi.1999. [24] Agiza HN, Elabbasy EM, El-Metwally H, et al. Chaotic
1414 dynamics of a discrete prey–predator model with Holling
[5] Arditi R, Ginzburg LR. Coupling in predator–prey dynam- type II. Nonlinear Anal Real World Appl. 2009;10(1):116–
ics: ratio-dependence. J Theor Biol. 1989;139(3):311–326. 129. doi: 10.1016/j.nonrwa.2007.08.029
doi: 10.1016/S0022-5193(89)80211-5 [25] Liu X, Xiao D. Complex dynamic behaviors of a discrete-
[6] Kuang Y, Beretta E. Global qualitative analysis of a time predator–prey system. Chaos Solit Fractals.
ratio-dependent predator–prey system. J Math Biol. 2007;32
1998;36(4):389–406. doi: 10.1007/s002850050105 (1):80–94. doi: 10.1016/j.chaos.2005.10.081
[7] Hsu SB, Hwang TW, Kuang Y. Global analysis of the [26] Khan AQ, Ma J, Xiao D. Bifurcations of a two-dimensional
Michaelis–Menten-type ratio-dependent predator–prey discrete time plant-herbivore system. Commun Nonlin-
system. J Math Biol. 2001;42(6):489–506. doi: 10.1007/s00 ear Sci Numer Simul. 2016;39:185–198. doi: 10.1016/j.cn
2850100079 sns.2016.02.037
JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE 19

[27] Khan AQ, Ma J, Xiao D. Global dynamics and bifurca- 2(η − 111 )
tion analysis of a host-parasitoid model with strong Allee r22 = −(η − 111 )114 − 115 + 214 112
112
effect. J Biol Dyn. 2017;11(1):121–146. doi: 10.1080/1751
3758.2016.1254287 + 2215 (η − 111 ),
[28] Ott E, Grebogi C, Yorke JA. Controlling chaos. Phys Rev 115 (η − 111 )
Lett. 1990;64(11):1196–1199. doi: 10.1103/PhysRevLett. r23 = ζ − ζ 215 . (A2)
64.1196 112
[29] Elaydi S. An introduction to difference equations. New
York: Springer-Verlag; 1996. Appendix 3. Expression for the coefficients
[30] Lynch S. Dynamical systems with applications using quantities of system (48)
mathematica. Boston: Birkhäuser; 2007.
[31] Luo XS, Chen G, Wang BH, et al. Hybrid control of period-
(1 + h) (x ∗ + y∗ )2 + h2 sx ∗2 y∗ − hsy∗2
doubling bifurcation and chaos in discrete nonlinear 111 = ,
dynamical systems. Chaos Solit Fractals. 2003;18(4):775– (1 + hx ∗ )2 (x ∗ + y∗ )2
783. doi: 10.1016/S0960-0779(03)00028-6 −hsx ∗2
112 = ,
(1 + hx ∗ ) (x ∗ + y∗ )2

Appendices −h (x ∗ + y∗ )3 − h2 (x ∗ + y∗ )3 − h3 sx ∗3 y∗ + hsy∗2
+3h2 sx ∗ y∗2 + h2 sy∗3
Appendix 1. Expression for the coefficients 113 = ,
(1 + hx ∗ )3 (x ∗ + y∗ )3
quantities of system (41)
h2 sx 2 (x ∗ − y∗ ) − 2hsx ∗ y∗
114 = ,
(1 + hx ∗ )2 (x ∗ + y∗ )3
(1 + h) (x ∗ + y∗ )2 + h2 sx ∗2 y∗ − hsy∗2
111 = , hsx ∗2
(1 + hx ∗ )2 (x ∗ + y∗ )2 115 = ,
(1 + hx ∗ ) (x ∗ + y∗ )3
−hsx ∗2
112 = , hδ ∗ y∗2 hδ ∗ x ∗2
(1 + hx ∗ ) (x ∗ + y∗ )2 211 = , 212 = 1 − hrδ ∗ + ,
(x ∗ + y∗ )2 (x ∗ + y∗ )2
−h (x ∗ + y∗ )3 − h2 (x ∗ + y∗ )3 − h3 sx ∗3 y∗
+hsy∗2 + 3h2 sx ∗ y∗2 + h2 sy∗3 hδ ∗ y∗2
213 = − ,
113 = , (x ∗ + y∗ )3
(1 + hx ∗ )3 (x ∗ + y∗ )3
2hδ ∗ x ∗ y∗ hδ ∗ x ∗2 hy∗2
h2 sx 2 (x ∗ − y∗ ) − 2hsx ∗ y∗ 214 = , 215 = − , k1 = ,
114 = , (x ∗ + y∗ )3 (x ∗ + y∗ )3 (x ∗ + y∗ )2
(1 + hx ∗ )2 (x ∗ + y∗ )3
hx ∗2 −hy∗2
hsx ∗2 k2 = −hr + , k3 = ,
115 = , (x ∗ + y∗ )2 (x ∗ + y∗ )3
(1 + hx ∗ ) (x ∗ + y∗ )3
2hx ∗ y∗ −hx ∗2
hδ ∗ y∗2 hδ ∗ x ∗2 k4 = , k5 = . (A3)
211 = , 212 = 1 − hrδ ∗ + , (x ∗ + y∗ )3 (x ∗ + y∗ )3
(x ∗ + y∗ )2 (x ∗ + y∗ )2
hδ ∗ y∗2
213 = − ,
(x ∗ + y∗ )3
2hδ ∗ x ∗ y∗ hδ ∗ x ∗2
214 = , 215 = − . (A1)
(x ∗ + y∗ )3 (x ∗ + y∗ )3

Appendix 2. Expression for the coefficients


quantities of system (44)

1 1 − hrs(r − 1)
η= + 1 + hrδ (r − 1) ,
1 + h + hs(r − 1)
2
   
 
1 1 − hrs(r − 1) h2 r2 sδ (r − 1)2 1 − hrs(r − 1) 2
ζ = 4 (1 + hrδ (r − 1)) + − + 1 + hrδ (r − 1) ,
2 1 + h + hs(r − 1) 1 + h + hs(r − 1) 1 + h + hs(r − 1)
 2
1 111 115 2η111 115
r11 = 113 112 + η114 − 111 114 − η2 15 + − ,
112 112 112
2ζ η115 2ζ 111 115
r12 = −ζ 114 − + ,
112 112
115
r13 = ζ 2 ,
112
   
1 3 1
1  1
 1 1  
1 2 1 η − 11 15 2
 1 2 2 1 1 2
 1 2

r21 = η − 11 13 12 + η − 11 14 − 13 12 14 12 (η − 11 ) − 15 η − 11 ,
ζ 112

You might also like