Hindawi Publishing Corporation

Abstract and Applied Analysis

Volume 2014, Article ID 838396, 18 pages
http://dx.doi.org/10.1155/2014/838396

Research Article
Hopf Bifurcation and Stability of Periodic Solutions for Delay
Differential Model of HIV Infection of CD4+ T-cells

P. Balasubramaniam,1 M. Prakash,1 Fathalla A. Rihan,2,3 and S. Lakshmanan2

1
Department of Mathematics, Gandhigram Rural Institute-Deemed University, Gandhigram, Tamil Nadu 624 302, India
2
Department of Mathematical Sciences, College of Science, UAE University, P.O. Box 15551, Al-Ain, UAE
3
Department of Mathematics, Faculty of Science, Helwan University, Cairo 11795, Egypt

Correspondence should be addressed to Fathalla A. Rihan; frihan@uaeu.ac.ae

Received 17 February 2014; Revised 13 June 2014; Accepted 19 June 2014; Published 31 August 2014

Academic Editor: Cemil Tunç

Copyright © 2014 P. Balasubramaniam et al. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.

This paper deals with stability and Hopf bifurcation analyses of a mathematical model of HIV infection of CD4+ T-cells. The model
is based on a system of delay differential equations with logistic growth term and antiretroviral treatment with a discrete time delay,
which plays a main role in changing the stability of each steady state. By fixing the time delay as a bifurcation parameter, we get a
limit cycle bifurcation about the infected steady state. We study the effect of the time delay on the stability of the endemically infected
equilibrium. We derive explicit formulae to determine the stability and direction of the limit cycles by using center manifold theory
and normal form method. Numerical simulations are presented to illustrate the results.

1. Introduction have been introduced to the models to describe the time

Since 1980, the human immunodeficiency virus (HIV) or
the associated syndrome of opportunistic infections that
causes acquired immunodeficiency syndrome (AIDS) has
been considered as one of the most serious global public
health menaces. When HIV enters the body, its main target
is the CD4 lymphocytes, also called CD4 T-cells (including
CD4+ T-cells). When a CD4 cell is infected with HIV, the
virus goes through multiple steps to reproduce itself and
create many more virus particles. The AIDS term, which
is known as the late stage of HIV, covers the range of
infections and illnesses which can result from a weakened
immune system caused by HIV. Based on the clinical studies,
it is known that, for a normal person, the CD4+ T-cells
count is around 1000 mm−3 and for HIV infected patient it
gradually decreases to 200 mm−3 or below, which leads to
AIDS. However, this may take several years for the number
of CD4 T-cells to reduce to a level where the immune system
is weakened [1–6].
Mathematical models, usingdelay differential equations
(DDEs), have provided insights in understanding the dynam-
ics of HIV infection. Discrete or continuous time delays
between infection of a CD4+ T-cell and the emission of
viral particles on a cellular level [7–13]. In general, DDEs
exhibit much more complicated dynamics than ODEs since
the time delay could cause a stable equilibrium to become
unstable and cause the populations to fluctuate [14–16]. In
studying the viral clearance rates, Perelson et al. [17] assumed
that there are two types of delays that occur between the
administration of drug and the observed decline in viral load:
a pharmacological delay that occurs between the ingestion of
drug and its appearance within cells and an intracellular delay
that is between initial infection of a cell by HIV and the release
of new virion. In this paper, we incorporate an intracellular
delay to the model to describe the time between infection of
a CD4+ T-cell and the emission of viral particles on a cellular
level [18]. We study the impact of the presence of such time
delay on the dynamics of the model.
The outline of the present paper is as follows. In Section 2,
we describe the model. In Section 3, we study the qualitative
behavior of the model via stability of the steady states
and Hopf bifurcation when time delay is considered as a
bifurcation parameter. In Section 4, we provide an explicit
formula to determine the direction of bifurcating periodic
2 Abstract and Applied Analysis
solution by applying center manifold theory and normal
form method. We provide some numerical simulations to
demonstrate the effectiveness of the analysis in Section 5 and
we conclude in Section 6.
2. Description of the Model
Let us start the analysis with some basic models of the
dynamics of target (uninfected) cells and infected CD4+ T-
cells by HIV. As a first approximation, the dynamics between
HIV and the macrophage population was described by the
simplest model of infection dynamics presented in [19–21].
Denoting uninfected cells by 𝑥(𝑡) and infected cells by 𝑦(𝑡)
and assuming that viruses are transmitted mainly by cell to
cell contact, the model is given by
𝑥̇ (𝑡) = Λ − 𝛿1 𝑥 (𝑡) − 𝛽𝑥 (𝑡) 𝑦 (𝑡) ,
(1)
𝑦̇ (𝑡) = 𝛽𝑥 (𝑡) 𝑦 (𝑡) − 𝛿2 𝑦 (𝑡) .
The target (uninfected) CD4+ T-cells are produced at a rate
Λ, die at a rate 𝛿1 , and become infected by virus at a rate 𝛽.
The infected host cells die at a rate 𝛿2 . The basic reproductive
ratio of the virus is then given by R0 = Λ𝛽/𝛿1 𝛿2 . If there is
no infection or if R0 < 1, there is only trivial equilibrium
(E0 = (Λ/𝛿1 , 0)) with no virus-producing cells. Whereas if
R0 > 1, the virus can establish an infection and the system
converges to the equilibrium with both uninfected cells and
infected cells, E1 = (𝛿2 /𝛽, Λ/𝛿2 − 𝛿1 /𝛽).
However, in most viral infections, the CTL response plays
a crucial part in antiviral defence by attacking viral infected
cells [22, 23]. As the the cytotoxic T-lymphocyte (CTL)
immune response is necessary to eliminate or control the
viral infection, we incorporated the antiviral CTL immune
response into the basic model (1). Therefore, if we add CTL
response, which is denoted by 𝑧(𝑡), into model (1) (see [19]),
then the extended model is
𝑥̇ (𝑡) = Λ − 𝛿1 𝑥 (𝑡) − 𝛽𝑥 (𝑡) 𝑦 (𝑡) ,
𝑦̇ (𝑡) = 𝛽𝑥 (𝑡) 𝑦 (𝑡) − 𝛿2 𝑦 (𝑡) − 𝑝𝑦 (𝑡) 𝑧 (𝑡) , (2)
𝑧̇ (𝑡) = 𝑐𝑞𝑦 (𝑡) 𝑧 (𝑡) − ℎ𝑧 (𝑡) .
Thus, CTLs proliferate in response to antigen at a rate 𝑐, die
at a rate ℎ, and lyse infected cells at a rate 𝑝. We assume that
the CTL pool consists of two populations: the precursors 𝑤(𝑡)
and the effectors 𝑧(𝑡). In other words, we assume that there are
primary and secondary responses to viral infections. Then,
the model (2) becomes
𝑥̇ (𝑡) = Λ − 𝛿1 𝑥 (𝑡) − 𝛽𝑥 (𝑡) 𝑦 (𝑡) ,
𝑦̇ (𝑡) = 𝛽𝑥 (𝑡) 𝑦 (𝑡) − 𝛿2 𝑦 (𝑡) − 𝑝𝑦 (𝑡) 𝑧 (𝑡) ,
(3)
𝑤̇ (𝑡) = 𝑐 (1 − 𝑞) 𝑦 (𝑡) 𝑤 (𝑡) − 𝑏𝑤 (𝑡) ,
𝑧̇ (𝑡) = 𝑐𝑞𝑦 (𝑡) 𝑤 (𝑡) − ℎ𝑧 (𝑡) .
The infected cells are killed by CTL effector cells at a rate
𝑝𝑦𝑧. Upon contact with antigen, CTLp proliferate at a rate
𝑐𝑦(𝑡)𝑤(𝑡) and differentiate into effector cells CTLe at a rate
Λ
𝛽
+
Uninfected cell (x) Free virus Infected cell (y)
𝛿1 𝛿2
CTL-mediated
lysis (p)
CTLp proliferate to generate
the memory population
cyw
Contact with antigen cqwy
Naive Activated
Differentiation
Effector CTL (z)
CTL CTLp (w)
h
b
Figure 1: A simplified model of virus-CTL interaction. The virus
dynamics is described by the basic model of Nowak and Bangham
[19]. The uninfected target cells are produced at a rate Λ and die
at a rate 𝛿1 𝑥. They become infected by the virus at a rate 𝛽𝑥𝑦.
The infected cells produce new virus particle and die at a rate 𝛿2 𝑦.
When CTL𝑝 recognize antigen on the surface of infected cells, they
become activated and expand at a rate 𝑐𝑦𝑤, decay at a rate 𝑏𝑤, and
differentaite into efector cells at a rate 𝑐𝑞𝑤𝑦. The effector cells lyse
the infected cells at a rate 𝑝𝑦𝑧.
𝑐𝑞𝑦(𝑡)𝑤(𝑡). CTL precursors die at a rate 𝑏𝑤, and effectors die
at a rate ℎ𝑧(𝑡); see Figure 1.
Since the proliferation of CD4+ T-cells is density depen-
dent, that is, the rate of proliferation decreases as T-cells
increase and reach the carrying capacity, we then extend
the above basic viral infection model to include the density
dependent growth of the CD4+ T-cell population (see [24–
26]). It is also known that HIV infection leads to low levels of
CD4+ T-cells via three main mechanisms: direct viral killing
of infected cells, increased rates of apoptosis in infected
cells, and killing of infected CD4+ T-cells by cytotoxic T-
lymphocytes [26]. Hence, it is reasonable to include apoptosis
of infected cells. An average of 1010 viral particles is produced
by infected cells per day. The treatment with single antiviral
drug is considered to be failed, so that the combination
of antiviral drugs is needed for the better treatment [25].
Therefore, in the below revised model, we combine the
antiretroviral drugs, namely, reverse transcriptase inhibitor
(RTI) and protease inhibitor (PI) to make the model realistic
(see [27–29]). RTIs can block the infection of target T-cells
by infectious virus, and PIs cause infected cells to produce
noninfectious virus particles. The modified model takes the
form
𝑥 (𝑡) + 𝑦 (𝑡)
𝑥̇ (𝑡) = Λ − 𝛿1 𝑥 (𝑡) + 𝑟 (1 − ) 𝑥 (𝑡)
𝑇max
− (1 − 𝜖) (1 − 𝜂) 𝛽𝑥 (𝑡) 𝑦 (𝑡) ,
𝑦̇ (𝑡) = (1 − 𝜖) (1 − 𝜂) 𝛽𝑥 (𝑡) 𝑦 (𝑡)
− 𝛿2 𝑦 (𝑡) − 𝑒1 𝑦 (𝑡) − 𝑝𝑦 (𝑡) 𝑧 (𝑡) ,
Abstract and Applied Analysis 3

Table 1: Parameter definitions and estimations used in the underlying model.

Parameter Notes Estimated Value Range Source

Λ Source of uninfected CD4+ T-cells 10 0–10 [26]
𝛽 Rate of infection 0.1 0.00001–0.5 [26]
𝑇max Total carrying capacity 1500 1500 [26]
𝑟 Logistic growth term 0.03 0.03–3 [26]
𝛿1 Mortality rate of CD4+ T-cells 0.06 0.007–0.1 [26]
𝜖 Antiretroviral (RTI) therapy 0.9 0-1 see text
𝛿2 Infected cells died out naturally 0.3 0.2–1.4 [26]
𝑒1 Apoptosis rate of infected cells 0.2 0.2 [26]
𝑝 Clearance rate of infected cells 1 0.001–1 [26]
𝜂 Protease inhibitor therapy 0.9 [0, 1] see text
𝑞 Rate of differentiation of CTLs 0.02 Assumed —
𝑏 Death rate of CTL precursors 0.02 0.005–0.15 [26]
𝑐 Proliferation of CTLs responsiveness 0.1 0.001–1 [26]
ℎ Mortality rate or CTL effectors 0.1 0.005–0.15 [26]

𝑤̇ (𝑡) = 𝑐𝑦 (𝑡) 𝑤 (𝑡) − 𝑐𝑞𝑦 (𝑡) 𝑤 (𝑡) − 𝑏𝑤 (𝑡) , 𝑦̇ (𝑡) = (1 − 𝜖) (1 − 𝜂) 𝛽𝑥 (𝑡) 𝑦 (𝑡)

𝑧̇ (𝑡) = 𝑐𝑞𝑦 (𝑡) 𝑤 (𝑡) − ℎ𝑧 (𝑡) . − (𝛿2 + 𝑒1 ) 𝑦 (𝑡) − 𝑝𝑦 (𝑡) 𝑧 (𝑡) ,

(4)
𝑤̇ (𝑡) = 𝑐 (1 − 𝑞) 𝑦 (𝑡 − 𝜏) 𝑤 (𝑡 − 𝜏) − 𝑏𝑤 (𝑡)
The first equation of model (4) represents the rate of change in
𝑧̇ (𝑡) = 𝑐𝑞𝑦 (𝑡 − 𝜏) 𝑤 (𝑡 − 𝜏) − ℎ𝑧 (𝑡) .
the count of healthy CD4+ T-cells that produced at rate Λ and
become infected at rate 𝛽, with the mortality 𝛿1 . We assume (5)
that the uninfected CD4+ T-cells proliferate logistically, thus
the growth rate 𝑟 is multiplied by the term (1 − (𝑥 + 𝑦)/𝑇max ) The range of parameter values of the model are given in
and this term approaches zero when the total number of T- Table 1.
cells approaches the carrying capacity 𝑇max . The effects of We start our analysis by presenting some notations that
combination of RTI and PI antiviral drugs are represented will be used in the sequel. Let 𝐶 = 𝐶([−𝜏, 0], R4+ ) be the
by the term (1 − 𝜖)(1 − 𝜂)𝛽𝑥𝑦, where (1 − 𝜖), 0 < 𝜖 < 1, Banach space of continuous functions mapping the interval
represents the effects of RTI and (1 − 𝜂), 0 < 𝜂 < 1, represents [−𝜏, 0] into R4+ , where R4+ = (𝑥, 𝑦, 𝑤, 𝑧); the initial conditions
the effects of PI. The second equation of model (4) denotes are given by
the rate of change in the count of infected CD4+ T-cells.
The infected CD4+ T-cells decay at a rate 𝛿2 and 𝑒1 denotes
apoptosis rate of infected cell; infected cells are killed by CTL 𝑥 (𝜃) = 𝜑1 (𝜃) ≥ 0, 𝑦 (𝜃) = 𝜑2 (𝜃) ≥ 0,
effectors at a rate 𝑝. The third equation of the model denotes 𝑤 (𝜃) = 𝜑3 (𝜃) ≥ 0, 𝑧 (𝜃) = 𝜑4 (𝜃) ≥ 0, (6)
the rate of change in the CTLp population; proliferation
rate of the CTLp is given by 𝑐 and is proportional to the 𝜃 ∈ [−𝜏, 0] ,
infected cells 𝑦; CTLp die at a rate 𝑏 and differentiate into CTL
effectors at a rate 𝑐𝑞. The last equation of the model represents
where 𝜑𝑖 (𝜃) ∈ C1 are smooth functions, for all 𝑖 =
the concentration of CTL effectors, which die at a rate ℎ.
1, 2, 3, 4. From the fundamental theory of functional dif-
In reality, the specific immune system is not immediately
ferential equations (see [32, 33]), it is easy to see that the
effective following invasion by a novel pathogen. There may
solutions (𝑥(𝑡), 𝑦(𝑡), 𝑤(𝑡), 𝑧(𝑡)) of system (5) with the initial
be an explicit time delay between infection and immune
conditions as stated above exist for all 𝑡 ≥ 0 and are unique. It
initiation and there may be a gradual build-up in immune
can be shown that these solutions exist for all 𝑡 > 0 and stay
efficacy during which the immune response develops, before
nonnegative. In fact, if 𝑥(0) > 0, then 𝑥(𝑡) > 0 for all 𝑡 > 0.
reaching maximal specificity to the pathogen ([8, 30, 31]). In
The same argument is true for the 𝑦, 𝑤, and 𝑧 components.
order to make model (4) more realistic, time delay in the
immune response should be included in the following model: Hence, the interior R4+ is invariant for system (5).

𝑥̇ (𝑡) = Λ − (1 − 𝜖) (1 − 𝜂) 𝛽𝑥 (𝑡) 𝑦 (𝑡) 3. Steady States

𝑥 (𝑡) + 𝑦 (𝑡) We can obtain the steady state values by setting 𝑥̇ = 𝑦̇ =
+ 𝑟 (1 − ) 𝑥 (𝑡) − 𝛿1 𝑥 (𝑡) ,
𝑇max 𝑤̇ = 𝑧̇ = 0. The steady state value of the infection-free
4 Abstract and Applied Analysis

steady sate E0 is given by E0 = ((𝑇max /2𝑟)(𝑟 − 𝛿1 + where 𝑐1 = 𝑐(1 − 𝑞)𝑟, 𝑐2 = 𝑇max 𝑏𝛽(1 − 𝜖)(1 − 𝜂) + 𝑏𝑟 − 𝑐(1 −
√(𝑟 − 𝛿1 )2 + 4𝑟Λ/𝑇max ), 0, 0, 0), while the infected steady 𝑞)𝑇max (𝑟 − 𝛿1 ), 𝑐3 = 𝑐(1 − 𝑞)Λ𝑇max .
state E+ = (𝑥∗ , 𝑦∗ , 𝑤∗ , 𝑧∗ ) is given by

𝑏 ℎ (1 − 𝑞) 𝑧∗
𝑦∗ = , 𝑤∗ = ,
𝑐 (1 − 𝑞) 𝑞𝑏 3.1. Stability and Hopf Bifurcation Analysis of Infected Steady
(1 − 𝜖) (1 − 𝜂) 𝛽𝑥∗ − (𝛿2 + 𝑒1 ) State E+ . In order to study full dynamics of model (4) by
𝑧∗ = ,
𝑝 using time delay as a bifurcation parameter, we need to
(7) linearize the model around the steady state E+ and determine
the characteristic equation of the Jacobian matrix. The roots
and 𝑥∗ is given by the following quadratic equation: of the characteristic equation determine the asymptotic
stability and existence of Hopf bifurcation for the model. The
𝑐1 𝑥2 + 𝑐2 𝑥 − 𝑐3 = 0, (8) characteristic equation of the linearized system is given by

󵄨󵄨 󵄨󵄨
󵄨󵄨−𝐴 1 𝑦∗ + 𝑟 − 2𝑟 𝑥∗ − 𝑟 𝑦∗ − 𝛿1 − 𝜆 −𝐴 1 𝑥∗ −
𝑟 ∗
𝑥 0 0 󵄨󵄨󵄨
󵄨󵄨 𝑇 𝑇 𝑇 󵄨󵄨
󵄨󵄨 max max max 󵄨
󵄨󵄨 𝐴 1 𝑦∗ 𝐴 1 𝑥∗ − (𝛿2 + 𝑒1 ) − 𝑝𝑧∗ − 𝜆 0 −𝑝𝑦∗ 󵄨󵄨󵄨󵄨 = 0,
󵄨󵄨 (9)
󵄨󵄨 󵄨
󵄨󵄨 0 𝑐 (1 − 𝑞) 𝑒−𝜆𝜏 𝑤∗ 𝑐 (1 − 𝑞) 𝑒−𝜆𝜏 𝑦∗ − 𝑏 − 𝜆 0 󵄨󵄨󵄨
󵄨󵄨 󵄨
󵄨󵄨 0 −𝜆𝜏
𝑐𝑞𝑒 𝑤 ∗ −𝜆𝜏
𝑐𝑞𝑒 𝑦 ∗
−ℎ − 𝜆󵄨󵄨󵄨

which is equivalent to the equation 𝑎6 = 𝑐 (1 − 𝑞) 𝑤∗ ,

𝜆4 + 𝑝1 𝜆3 + 𝑝2 𝜆2 + 𝑝3 𝜆 + 𝑝4 𝑎7 = 𝑐 (1 − 𝑞) 𝑦∗ ,
(10)
+ 𝑒−𝜆𝜏 (𝑞1 𝜆3 + 𝑞2 𝜆2 + 𝑞3 𝜆 + 𝑞4 ) = 0, 𝑎8 = − 𝑏,

where 𝐴 1 = (1 − 𝜖)(1 − 𝜂)𝛽 and 𝑎9 = 𝑐𝑞𝑤∗ ,

𝑝1 = − 𝑎1 − 𝑎4 − 𝑎8 − 𝑎11 , 𝑎10 = 𝑐𝑞𝑦∗ ,

𝑝2 = 𝑎1 𝑎8 + 𝑎8 𝑎11 + 𝑎1 𝑎11 + 𝑎4 𝑎8 + 𝑎4 𝑎11 + 𝑎1 𝑎4 − 𝑎2 𝑎3 , 𝑎11 = − ℎ.

(11)
𝑝3 = 𝑎2 𝑎3 𝑎8 + 𝑎2 𝑎3 𝑎11 − 𝑎1 𝑎8 𝑎11
Let us consider the following equation:
− 𝑎4 𝑎8 𝑎11 − 𝑎1 𝑎4 𝑎8 − 𝑎1 𝑎4 𝑎11 ,
𝜑 (𝜆, 𝜏) = 𝜆4 + 𝑝1 𝜆3 + 𝑝2 𝜆2 + 𝑝3 𝜆 + 𝑝4
𝑝4 = 𝑎1 𝑎4 𝑎8 𝑎11 − 𝑎2 𝑎3 𝑎8 𝑎11 ,
(12)
𝑞1 = − 𝑎7 , + (𝑞1 𝜆3 + 𝑞2 𝜆2 + 𝑞3 𝜆 + 𝑞4 ) 𝑒−𝜆𝜏 .

𝑞2 = 𝑎1 𝑎7 + 𝑎7 𝑎11 + 𝑎4 𝑎7 − 𝑎5 𝑎9 , For the nondelayed model (say 𝜏 = 0), from (10), we have

𝑞3 = 𝑎5 𝑎8 𝑎9 + 𝑎1 𝑎5 𝑎9 + 𝑎2 𝑎3 𝑎7 − 𝑎1 𝑎7 𝑎11 𝜆4 + 𝐷1 𝜆3 + 𝐷2 𝜆2 + 𝐷3 𝜆 + 𝐷4 = 0, (13)

− 𝑎4 𝑎7 𝑎11 − 𝑎1 𝑎4 𝑎7 , where
𝑞4 = 𝑎1 𝑎4 𝑎7 𝑎11 − 𝑎1 𝑎5 𝑎8 𝑎9 − 𝑎2 𝑎3 𝑎7 𝑎11 , 𝐷1 = 𝑝1 + 𝑞1 , 𝐷2 = 𝑝2 + 𝑞2 ,
(14)
2𝑟𝑥∗ 𝑟𝑦∗ 𝐷3 = 𝑝3 + 𝑞3 , 𝐷4 = 𝑝4 + 𝑞4 .
∗
𝑎1 = − (1 − 𝜖) (1 − 𝜂) 𝛽𝑦 + 𝑟 − − − 𝛿1 ,
𝑇max 𝑇max Lemma 1. For 𝜏 = 0, the unique nontrivial equilibrium is
∗ locally asymptotically stable if the real parts of all the roots of
𝑟𝑥
𝑎2 = − (1 − 𝜖) (1 − 𝜂) 𝛽𝑥∗ − , (13) are negative.
𝑇max
Proof. The proof of the above lemma is based on holding
𝑎3 = (1 − 𝜖) (1 − 𝜂) 𝛽𝑦∗ ,
the following conditions: 𝐷1 > 0, 𝐷3 > 0, 𝐷4 > 0, and
𝑎4 = (1 − 𝜖) (1 − 𝜂) 𝛽𝑥∗ − (𝛿2 + 𝑒1 ) − 𝑝𝑧∗ , 𝐷1 𝐷2 𝐷3 > 𝐷12 𝐷4 + 𝐷32 , as proposed by Routh-Hurwitz
criterion. We conclude that equilibrium E+ is locally asymp-
𝑎5 = − 𝑝𝑦∗ , totically stable if and only if all the roots of the characteristic
Abstract and Applied Analysis 5

equation (13) have negative real parts which depends on Let

the numerical values of parameters that are shown in the
𝑏1 = 𝑞2 − 𝑝1 𝑞1 , 𝑏2 = 𝑝3 𝑞1 − 𝑞4 − 𝑝2 𝑞2 + 𝑝1 𝑞3 ,
numerical exploration.
𝑏3 = 𝑝2 𝑞4 + 𝑝4 𝑞2 − 𝑝3 𝑞3 , 𝑏4 = −𝑝4 𝑞4 ,
3.2. Existence of Hopf Bifurcation. We here study the impact
𝑏5 = 𝑞12 , 𝑏6 = 𝑞22 − 2𝑞1 𝑞3 ,
of the time-delay parameter on the stability of HIV infection
of CD4+ T-cells. We deduce criteria that ensure the asymp- 𝑏7 = 𝑞32 − 2𝑞2 𝑞4 , 𝑏8 = 𝑞42 ,
totic stability of infected steady state E+ , for all 𝜏 > 0. We
arrive at the following theorem. 𝑏9 = 𝑞1 , 𝑏10 = 𝑝1 𝑞2 − 𝑞3 − 𝑝2 𝑞1 ,

Theorem 2. Necessary and sufficient conditions for the 𝑏11 = 𝑝2 𝑞3 + 𝑝4 𝑞1 − 𝑝3 𝑞2 − 𝑝1 𝑞4 , 𝑏12 = 𝑝3 𝑞4 − 𝑝4 𝑞3 .

infected equilibrium E+ to be asymptotically stable for all delay (18)
𝜏 ≥ 0 are as follows
From (16), we have
(i) the real parts of all the roots of 𝜑(𝜆, 𝜏) = 0 are negative; 𝜔8 + 𝑐1 𝜔6 + 𝑐2 𝜔4 + 𝑐3 𝜔2 + 𝑐4 = 0, (19)

where
(ii) for all 𝜔 and 𝜏 ≥ 0, 𝜑(𝑖𝜔, 𝜏) ≠ 0, where 𝑖 = √−1.
𝑐1 = 𝑝12 − 2𝑝2 − 𝑞12 , 𝑐2 = 𝑝22 − 2𝑝1 𝑝3 + 2𝑞1 𝑞3 + 2𝑝4 −𝑞22 ,
Proof. Assume that Lemma 1 is true. Now, for 𝜔 = 0, we have
𝑐3 = 𝑝32 − 2𝑝2 𝑝4 + 2𝑞2 𝑞4 − 𝑞32 , 𝑐4 = 𝑝42 − 𝑞42 .
𝜑 (0, 𝜏) = 𝐷4 = 𝑝4 + 𝑞4 ≠ 0. (15) (20)

The conditions (i) and (ii) of Theorem 2 hold if and only if

Substituting 𝜆 = 𝑖𝜔 (𝜔 > 0) into (5) and separating the real (19) has no real positive root.
and imaginary parts of the equations yields Let 𝑚 = 𝜔2 ; then (19) takes the form

𝑚4 + 𝑐1 𝑚3 + 𝑐2 𝑚2 + 𝑐3 𝑚 + 𝑐4 = 0. (21)
(𝜔4 − 𝑝2 𝜔2 + 𝑝4 ) + (−𝑞2 𝜔2 + 𝑞4 ) cos (𝜔𝜏)
If 𝑐4 < 0, then (19) has at least one positive root. In the case
+ (−𝑞1 𝜔3 + 𝑞3 𝜔) sin (𝜔𝜏) = 0, when (19) has four positive roots, we have
(16)
(−𝑝1 𝜔3 + 𝑝3 𝜔) + (−𝑞1 𝜔3 + 𝑞3 𝜔) cos (𝜔𝜏) 𝜔1 = √𝑚1 , 𝜔2 = √𝑚2 ,
(22)
2 𝜔3 = √𝑚3 , 𝜔4 = √𝑚4 .
− (−𝑞2 𝜔 + 𝑞4 ) sin (𝜔𝜏) = 0.
From (16), we have
After some mathematical manipulations, we obtain the fol- (𝑗) 1 𝑏 𝜔7 + 𝑏 𝜔5 + 𝑏 𝜔3 + 𝑏 𝜔
lowing equations 𝜏𝑘 = {arcsin 9 𝑘 6 10 𝑘 4 11 𝑘2 12 𝑘 + 2𝑗𝜋} ,
𝜔𝑘 𝑏5 𝜔𝑘 + 𝑏6 𝜔𝑘 + 𝑏7 𝜔𝑘 + 𝑏8
(23)
cos (𝜔𝜏)
where 𝑘 = 1, 2, 3, 4 and 𝑗 = 0, 1, 2, . . .; we choose 𝜏0 =
(𝑗)
= ((𝑞2 − 𝑝1 𝑞1 ) 𝜔6 + (𝑝3 𝑞1 − 𝑞4 − 𝑝2 𝑞2 + 𝑝1 𝑞3 ) 𝜔4 min(𝜏𝑘 ).
To establish Hopf bifurcation at 𝜏 = 𝜏0 , we need to show
+ (𝑝2 𝑞4 + 𝑝4 𝑞2 − 𝑝3 𝑞3 ) 𝜔2 − 𝑝4 𝑞4 ) that
−1 𝑑𝜆
× (𝑞12 𝜔6 + (𝑞22 − 2𝑞1 𝑞3 ) 𝜔4 + (𝑞32 − 2𝑞2 𝑞4 ) 𝜔2 + 𝑞42 ) , R( ) ≠ 0. (24)
𝑑𝜏 𝜏=𝜏0
sin (𝜔𝜏) By differentiating (10) with respect to 𝜏, we can get
7 5
= (𝑞1 𝜔 + (𝑝1 𝑞2 − 𝑞3 − 𝑝2 𝑞1 ) 𝜔 𝑑𝜆
= 𝜆𝑒−𝜆𝜏 (𝑞1 𝜆3 + 𝑞2 𝜆2 + 𝑞3 𝜆 + 𝑞4 )
𝑑𝜏
+ (𝑝2 𝑞3 + 𝑝4 𝑞1 − 𝑝3 𝑞2 − 𝑝1 𝑞4 ) 𝜔3
× ( (4𝜆3 + 3𝑝1 𝜆2 + 2𝑝2 𝜆 + 𝑝3 ) + 𝑒−𝜆𝜏
+ (𝑝3 𝑞4 − 𝑝4 𝑞3 ) 𝜔) (25)
× [(3𝑞1 𝜆2 + 2𝑞2 𝜆 + 𝑞3 )
−1
× (𝑞12 𝜔6 + (𝑞22 − 2𝑞1 𝑞3 ) 𝜔4 + (𝑞32 − 2𝑞2 𝑞4 ) 𝜔2 + 𝑞42 ) .
−1
(17) − 𝜏 (𝑞1 𝜆3 + 𝑞2 𝜆2 + 𝑞3 𝜆 + 𝑞4 )] ) .
6 Abstract and Applied Analysis

It follows that 4. Direction and Stability of Bifurcating

Periodic Solutions
𝑑𝜆 −1
( ) = ((4𝜆3 + 3𝑝1 𝜆2 + 2𝑝2 𝜆 + 𝑝3 ) + 𝑒−𝜆𝜏
𝑑𝜏 In the previous section, we obtained conditions for Hopf
(𝑗)
bifurcation to occur when 𝜏0 = 𝜏𝑘 , 𝑗 = 0, 1, 2, . . .. It is
× [(3𝑞1 𝜆2 + 2𝑞2 𝜆 + 𝑞3 ) also important to derive explicit formulae from which we
(26)
3 2 can determine the direction, stability, and period of periodic
−𝜏 (𝑞1 𝜆 + 𝑞2 𝜆 + 𝑞3 𝜆 + 𝑞4 )]) solutions bifurcating around the infected equilibrium E+ at
−1 the critical value 𝜏0 . We use the cafeteria of normal forms
× (𝜆𝑒−𝜆𝜏 (𝑞1 𝜆3 + 𝑞2 𝜆2 + 𝑞3 𝜆 + 𝑞4 )) . and center manifold proposed by Hassard [34]. We assume
that the model (5) undergoes Hopf bifurcation at the infected
Then, by combining (10), we get (𝑗)
equilibrium E+ when 𝜏0 = 𝜏𝑘 , 𝑗 = 0, 1, 2, . . ., and
then ±𝑖𝜔0 are the corresponding purely imaginary roots of
𝑑𝜆 −1 the characteristic equation at the infected equilibrium E+ .
( ) = ((4𝜆3 + 3𝑝1 𝜆2 + 2𝑝2 𝜆 + 𝑝3 )
𝑑𝜏 Assume also that
𝑇
+ 𝑒−𝜆𝜏 (3𝑞1 𝜆2 + 2𝑞2 𝜆 + 𝑞3 )) (𝑋1 (𝑡) , 𝑋2 (𝑡) , 𝑋3 (𝑡) , 𝑋4 (𝑡))
−1 𝜏 = (𝑥 (𝑡) − 𝑥∗ , 𝑦 (𝑡) − 𝑦∗ (𝑡) , (32)
× (𝜆𝑒−𝜆𝜏 (𝑞1 𝜆3 + 𝑞2 𝜆2 + 𝑞3 𝜆 + 𝑞4 )) − .
𝜆
𝑇
(27) 𝑤 (𝑡) −𝑤∗ (𝑡) , 𝑧 (𝑡) − 𝑧∗ (𝑡)) ;

Substituting 𝜆 = 𝑖𝜔0 in (27) (where 𝜔0 > 0 and 𝑖 = √−1) then using Taylors expansion for system (3) at the equilibrium
yields point yields

󵄨 𝑋̇ 1 = 𝑘11 𝑋1 (𝑡) + 𝑘12 𝑋2 (𝑡)

𝑑𝜆 −1 󵄨󵄨 𝑑 + 𝑖𝑑2 𝜏
( ) 󵄨󵄨󵄨󵄨 = 1 − , (28)
𝑑𝜏 󵄨󵄨𝜏=𝜏0 𝑑3 + 𝑖𝑑4 𝜆 + 𝑘13 𝑋1 (𝑡) 𝑋1 (𝑡) + 𝑘14 𝑋1 (𝑡) 𝑋2 (𝑡) ,

where 𝑋̇ 2 = 𝑘21 𝑋1 (𝑡) + 𝑘22 𝑋2 (𝑡) + 𝑘23 𝑋4 (𝑡)

𝑑1 = (𝑝3 − 3𝑝1 𝜔02 ) + (𝑞3 − 3𝑞1 𝜔02 ) cos (𝜔0 𝜏0 ) + 𝑘24 𝑋1 (𝑡) 𝑋2 (𝑡) + 𝑘25 𝑋2 (𝑡) 𝑋4 (𝑡) ,
(33)
+ 2𝑞2 𝜔0 sin (𝜔0 𝜏0 ) , 𝑋̇ 3 = 𝑘31 𝑋3 (𝑡) + 𝑘32 𝑋2 (𝑡 − 𝜏)
+ 𝑘33 𝑋3 (𝑡 − 𝜏) + 𝑘34 𝑋2 (𝑡 − 𝜏) 𝑋3 (𝑡 − 𝜏) ,
𝑑2 = (2𝑝2 𝜔0 − 4𝜔3 ) + 2𝑞2 𝜔0 cos (𝜔0 𝜏0 )
𝑋̇ 4 = 𝑘41 𝑋4 (𝑡) + 𝑘42 𝑋2 (𝑡 − 𝜏)
− (𝑞3 − 3𝑞1 𝜔02 ) sin (𝜔0 𝜏0 ) ,
+ 𝑘43 𝑋3 (𝑡 − 𝜏) + 𝑘44 𝑋2 (𝑡 − 𝜏) 𝑋3 (𝑡 − 𝜏) .
𝑑3 = (𝑞1 𝜔04 − 𝑞3 𝜔02 ) cos (𝜔0 𝜏0 ) + (𝑞4 𝜔0 − 𝑞2 𝜔03 ) sin (𝜔0 𝜏0 ) ,
Here,
𝑑4 = (𝑞4 𝜔0 − 𝑞2 𝜔03 ) cos (𝜔0 𝜏0 ) − (𝑞1 𝜔04 − 𝑞3 𝜔02 ) sin (𝜔0 𝜏0 ) . 2𝑟𝑥∗ 𝑟𝑦∗
𝑘11 = − 𝐴 1 𝑦∗ + 𝑟 − − − 𝛿1 ,
(29) 𝑇max 𝑇max

Thus, 𝑟𝑥∗
𝑘12 = − 𝐴 1 𝑥∗ − ,
󵄨 𝑇max
𝑑𝜆 −1 󵄨󵄨󵄨 𝑑 𝑑 + 𝑑2 𝑑4
R( ) 󵄨󵄨󵄨 = 1 32 . (30) 2𝑟
𝑑𝜏 󵄨󵄨𝜏=𝜏0 𝑑3 + 𝑑42 𝑘13 = − ,
𝑇max
Notice that 𝑟
𝑘14 = − − 𝐴 1,
󵄨 𝑇max
𝑑𝜆(𝑡) 󵄨󵄨󵄨󵄨 𝑑𝜆 −1 󵄨󵄨
sign (R )󵄨󵄨 = sign (R( ) )󵄨󵄨󵄨󵄨 . (31)
𝑑𝜏 󵄨󵄨𝜏=𝜏0 𝑑𝜏 󵄨󵄨𝜏=𝜏 𝑘21 = 𝐴 1 𝑦∗ ,
0

By summarizing the above analysis, we arrive at the following 𝑘22 = 𝐴 1 𝑥∗ − 𝐴 2 − 𝑝𝑧∗ ,

theorem.
𝑘23 = − 𝑝𝑦∗ ,
Theorem 3. The infected equilibrium E+ of the system (5) is 𝑘24 = 𝐴 1 ,
asymptotically stable for 𝜏 ∈ [0, 𝜏0 ) and it undergoes Hopf
bifurcation at 𝜏 = 𝜏0 . 𝑘25 = − 𝑝,
Abstract and Applied Analysis 7

𝑘31 = − 𝑏, ±𝑖𝜏0 𝜔0 . By Reisz representation, there exists a function 𝜂(𝜃, 𝜇)

of bounded variation for 𝜃 ∈ [−𝜏, 0] such that
𝑘32 = 𝑐 (1 − 𝑞) 𝑤∗ ,
0
𝑘33 = 𝑐 (1 − 𝑞) 𝑦∗ , 𝐿 𝜇 𝜙 = ∫ 𝑑𝜂 (𝜃, 𝜇) 𝜙 (𝜃) . (38)
−𝜏

𝑘34 = 𝑐 (1 − 𝑞) ,
In fact, we can choose
𝑘41 = − ℎ,
𝜂 (𝜃, 𝜇) = (𝜏0 + 𝜇) 𝐺1 𝛿 (𝜃) + (𝜏0 + 𝜇) 𝐺2 𝛿 (𝜃 + 𝜏) , (39)
∗
𝑘42 = 𝑐𝑞𝑤 ,
where 𝛿(𝜃) is Dirac delta function. Next, for 𝜙 ∈ 𝐶1 ([−𝜏,
𝑘43 = 𝑐𝑞𝑦∗ ,
0], R4 ), define
𝑘44 = 𝑐𝑞.
𝑑𝜙
(34) {
{ , 𝜃 ∈ [−𝜏, 0)
{
{ 𝑑𝜃
For convenience, let 𝜏 = 𝜏0 + 𝜇 and 𝑢𝑡 (𝜃) = 𝑢(𝑡 + 𝜃) for 𝐴 (𝜇) 𝜙 = { (40)
{
{ 0
{
𝜃 ∈ [−𝜏, 0]. Denote 𝐶𝑘 ([−𝜏, 0], R4 ) = {𝜙 | 𝜙 : [−𝜏, 0] → R4 }; ∫ 𝑑𝜂 (𝜃, 𝜇) 𝜙 (𝜃) , 𝜃 = 0,
{ −𝜏
𝜙 has 𝑘-order continuous derivative. For initial conditions
𝜙(𝜃) = (𝜙1 (𝜃), 𝜙2 (𝜃), 𝜙3 (𝜃), 𝜙4 (𝜃))𝑇 ∈ 𝐶([−𝜏, 0], R4 ), (33) can {0, 𝜃 = [−𝜏, 0)
be rewritten as 𝑅 (𝜇) 𝜙 = { (41)
{𝐹 (𝜙, 𝜇) , 𝜃 = 0.
𝑢̇ (𝑡) = 𝐿 𝜇 (𝑢𝑡 ) + 𝐹 (𝑢𝑡 , 𝜇) , (35)
̇ = 𝑢̇𝑡 (𝜃), (35) can be written as
Since 𝑢(𝑡)
where 𝑢(𝑡) = (𝑢1 (𝑡), 𝑢2 (𝑡), 𝑢3 (𝑡), 𝑢4 (𝑡))𝑇 ∈ 𝐶, 𝐿 𝜇 : 𝐶 → R4 ,
and 𝐹 : 𝐶 → R4 are given, respectively, by 𝑢̇𝑡 = 𝐴 (𝜇) 𝑢𝑡 + 𝑅 (𝜇) 𝑢𝑡 , (42)
𝐿 𝜇 𝜙 = (𝜏0 + 𝜇) 𝐺1 𝜙 (0) + (𝜏0 + 𝜇) 𝐺2 𝜙 (−𝜏) ,
(36) where 𝑢𝑡 = 𝑢(𝑡 + 𝜃), 𝜃 ∈ [−𝜏, 0]. For 𝜓 ∈ 𝐶1 ([0, 𝜏], R4 ), the
𝐹 (𝜙, 𝜇) = (𝜏0 + 𝜇) (𝐹1 , 𝐹2 , 𝐹3 , 𝐹4 ) .
𝑇 adjoint operator 𝐴∗ of 𝐴 can be defined as

𝐿 𝜇 is one parameter family of bounded linear operators in 𝐶 { 𝑑𝜓 (𝑠)

{
{ − , 𝑠 ∈ (−𝜏, 0]
and { 𝑑𝑠
𝐴∗ 𝜓 (𝑠) 𝜙 = { 0 (43)
𝑘11 𝑘12 0 0 {
{
{∫ 𝑑𝜂 (𝜃, 𝜇) 𝜙 (𝜃) , 𝑠 = 0.
𝑘21 𝑘22 0 𝑘24 { −𝜏
𝐺1 = ( ),
0 0 𝑘31 0 For 𝜙 ∈ 𝐶1 ([−𝜏, 0], R4 ) and 𝜓 ∈ 𝐶1 ([0, 𝜏], R4 ), in order to
0 0 0 𝑘41 normalize the eigenvalues of operator 𝐴 and adjoint operator
𝐴∗ , the following bilinear form is defined by
0 0 0 0
0 0 0 0 ⟨𝜓, 𝜙⟩ = 𝜓 (0) 𝜙 (0)
𝐺2 = ( ),
0 𝑘32 𝑘33 0 (37) 0 𝜃 (44)
−∫ ∫ 𝜓 (𝜉 − 𝜃) [𝑑𝜂 (𝜃)] 𝜙 (𝜉) 𝑑𝜉,
0 𝑘42 𝑘43 0 𝜃=−𝜏 𝜉=0

𝑘13 𝜙1 (0) 𝜙1 (0) + 𝑘14 𝜙1 (0) 𝜙2 (0)
𝑘24 𝜙1 (0) 𝜙2 (0) + 𝑘25 𝜙2 (0) 𝜙4 (0)
( )
𝐹= ( ).
𝑘34 𝜙2 (−𝜏) 𝜙3 (−𝜏)
𝑘44 𝜙2 (−𝜏) 𝜙3 (−𝜏)
( )
From the discussion in the above section, we know that if
𝜇 = 0, then model (5) undergoes a Hopf bifurcation at the
infected equilibrium E+ , and the associated characteristic
equation of model (5) has a pair of purely imaginary roots
where 𝜂(𝜃) = 𝜂(𝜃, 0) and 𝜓 is complex conjugate of 𝜓. It can
verify that 𝐴∗ and 𝐴(0) are adjoint operators with respect to
this bilinear form.
We assume that ±𝑖𝜔0 are eigenvalues of 𝐴(0) and the other
eigenvalues have strictly negative real parts. Thus, they are
also eigenvalues of 𝐴∗ . Now we compute the eigenvector 𝑞
of 𝐴 corresponding to the eigenvalue 𝑖𝜔0 and the eigenvector
𝑞∗ of 𝐴∗ corresponding to the eigenvalue −𝑖𝜔0 . Suppose that
𝑞(𝜃) = (1, 𝑝1 , 𝑝2 , 𝑝3 )𝑇 𝑒𝑖𝜔0 𝜃 is eigenvector of 𝐴(0) associated
with 𝑖𝜔0 ; then, 𝐴(0)𝑞(𝜃) = 𝑖𝜔0 𝑞(𝜃). It follows from the
definition of 𝐴(0) and (36), (38), and (40) that
8 Abstract and Applied Analysis

𝑘11 − 𝑖𝜔0 𝑘12 0 0 0

𝑘21 𝑘22 − 𝑖𝜔0 0 𝑘23 0
( −𝑖𝜔0 𝜏0 −𝑖𝜔0 𝜏0
) 𝑞 (0) = ( ) . (45)
0 𝑘32 𝑒 𝑘31 + 𝑘33 𝑒 − 𝑖𝜔0 0 0
0 𝑘42 𝑒−𝑖𝜔0 𝜏0 𝑘43 𝑒−𝑖𝜔0 𝜏0 𝑘41 − 𝑖𝜔0 0

Solving (45), we can easily obtain 𝑞(0) = (1, 𝑝1 , 𝑝2 , 𝑝3 )𝑇 , (𝑘11 − 𝑖𝜔0 ) (𝑘22 − 𝑖𝜔0 ) − 𝑘12 𝑘21
𝑝3 = .
where 𝑘12 𝑘23
𝑖𝜔0 − 𝑘11 (46)
𝑝1 = ,
𝑘12 Similarly, suppose that the eigenvector 𝑞∗ of 𝐴∗ correspond-
𝑘32 (𝑘11 − 𝑖𝜔0 ) 𝑒−𝑖𝜔0 𝜏0 ing to −𝑖𝜔0 is 𝑞∗ (𝑠) = (1/𝐷)(1, 𝑝1∗ , 𝑝2∗ , 𝑝3∗ )𝑇 𝑒𝑖𝜔0 𝑠 , 𝑠 ∈ [0, 𝜏]. By
𝑝2 = , the definition of 𝐴∗ and (36), (38), and (40), one gets
𝑘12 (𝑘31 + 𝑘33 𝑒−𝑖𝜔0 𝜏0 − 𝑖𝜔0 )

𝑘11 + 𝑖𝜔0 𝑘21 0 0 0

−𝑖𝜔0 𝜏0 −𝑖𝜔0 𝜏0
𝑘12 𝑘22 + 𝑖𝜔0 𝑘32 𝑒 𝑘42 𝑒 ∗
0
( ) 𝑞 (0) = ( ). (47)
0 0 𝑘31 + 𝑘33 𝑒−𝑖𝜔0 𝜏0 + 𝑖𝜔0 𝑘43 𝑒−𝑖𝜔0 𝜏0 0
0 𝑘23 0 𝑘41 + 𝑖𝜔0 0

0
Solving (47), we easily obtain 𝑞∗ (0) = (1/𝐷)(1, 𝑝1∗ , 𝑝2∗ , 𝑝3∗ )𝑇 , −∫
1 ∗ ∗ ∗
(1, 𝑝1 , 𝑝2 , 𝑝3 ) 𝜃𝑒𝑖𝜔0 𝜃
where −𝜏0 𝐷
𝑇
𝑘 + 𝑖𝜔0 × [𝑑𝜂 (𝜃)] (1, 𝑝1 , 𝑝2 , 𝑝3 )
𝑝1∗ = − 11 ,
𝑘21 1 ∗ ∗ ∗
= ( (1 + 𝑝1 𝑝1 + 𝑝2 𝑝2 + 𝑝3 𝑝3 )
𝑘23 𝑘43 (𝑘11 + 𝑖𝜔0 ) 𝑒−𝑖𝜔0 𝜏0 𝐷
𝑝2∗ = − , (48) ∗ ∗ ∗
𝑘21 (𝑘41 + 𝑖𝜔0 ) (𝑘31 + 𝑘33 𝑒−𝑖𝜔0 𝜏0 + 𝑖𝜔0 ) + 𝜏0 𝑒−𝑖𝜔0 𝜏0 (1, 𝑝1 , 𝑝2 , 𝑝3 )
𝑇
𝑘23 (𝑘11 + 𝑖𝜔0 ) × 𝐺2 (1, 𝑝1 , 𝑝2 , 𝑝3 ) )
𝑝3∗ = .
𝑘21 (𝑘41 + 𝑖𝜔0 )
1 ∗ ∗ ∗
= ( (1 + 𝑝1 𝑝1 + 𝑝2 𝑝2 + 𝑝3 𝑝3 ) + 𝜏0 𝑒−𝑖𝜔0 𝜏0
𝐷
In order to assure that ⟨𝑞∗ , 𝑞⟩ = 1, we need to determine the
∗ ∗
value of 𝐷. From (44), one gets × ((𝑘32 𝑝2 + 𝑘42 𝑝3 ) 𝑝1
∗ ∗
∗ 𝑇 + (𝑘33 𝑝2 + 𝑘43 𝑝3 ) 𝑝2 ) ) ;
⟨𝑞 , 𝑞⟩ = 𝑞∗ (0) 𝑞 (0)
∗ ∗ ∗
𝐷 = (1 + 𝑝1 𝑝1 + 𝑝2 𝑝2 + 𝑝3 𝑝3 )
0 𝜃
𝑇
−∫ ∫ 𝑞∗ (𝜉 − 𝜃) [𝑑𝜂 (𝜃)] 𝑞 (𝜉) 𝑑 (𝜉) ∗
+ 𝜏0 𝑒−𝑖𝜔0 𝜏0 ((𝑘32 𝑝2 + 𝑘42 𝑝3 ) 𝑝1
∗
𝜃=−𝜏0 𝜉=0
∗ ∗
1 ∗ ∗ ∗ + (𝑘33 𝑝2 + 𝑘43 𝑝3 ) 𝑝2 ) .
= (1 + 𝑝1 𝑝1 + 𝑝2 𝑝2 + 𝑝3 𝑝3 )
𝐷 (49)
0 𝜃
1 Let
∗ ∗ ∗
−∫ ∫ (1, 𝑝1 , 𝑝2 𝑝3 ) 𝑒−𝑖𝜔0 (𝜉−𝜃) V (𝑡) = ⟨𝑞∗ , 𝑢𝑡 ⟩ ,
−𝜏0 𝜉=0 𝐷
(50)
𝑇 𝑖𝜔0 𝜉 𝑊 (𝑡, 𝜃) = 𝑢𝑡 − V𝑞 − V𝑞 = 𝑢𝑡 − 2 Re (V (𝑡) 𝑞 (𝜃)) .
× [𝑑𝜂 (𝜃)] (1, 𝑝1 , 𝑝2 , 𝑝3 ) 𝑒 𝑑𝜉
On the center manifold Ω0 , we have
1 ∗ ∗ ∗
= (1 + 𝑝1 𝑝1 + 𝑝2 𝑝2 + 𝑝3 𝑝3 ) 𝑊 (𝑡, 𝜃) = 𝑊 (V (𝑡) , V (𝑡) , 𝜃) , (51)
𝐷
Abstract and Applied Analysis 9

where where
V2 VV2 V2 V2
𝑊 (V, V, 𝜃) = 𝑊20 (𝜃) + 𝑊11 (𝜃) VV + 𝑊02 (𝜃) + ⋅⋅⋅ . 𝐻 (V, V, 𝜃) = 𝐻20 (𝜃) + 𝐻11 (𝜃) VV + 𝐻02 (𝜃) + ⋅ ⋅ ⋅ .
2 2 2 2
(52) (60)

V and V are local coordinates of the center manifold Ω0 in the On the center manifold Ω0 , we have
direction of 𝑞∗ and 𝑞∗ , respectively. Note that 𝑊 is real if 𝑢𝑡 is
real. So we only consider real solutions. From (50), we obtain 𝑊̇ = 𝑊V V̇ + 𝑊V V.̇ (61)

⟨𝑞∗ , 𝑊⟩ = ⟨𝑞∗ , 𝑢𝑡 − V𝑞 − V𝑞⟩ Substituting (52) and (55) into (61), one obtains
(53)
= ⟨𝑞∗ , 𝑢𝑡 ⟩ − V (𝑡) ⟨𝑞∗ , 𝑞⟩ − V (𝑡) ⟨𝑞∗ , 𝑞⟩ . 𝑊̇ = (𝑊20 V + 𝑊11 V + ⋅ ⋅ ⋅ ) (𝑖𝜔0 V + 𝑔)
(62)
For the solution 𝑢𝑡 ∈ Ω0 of (35), from (41) and (44), since + (𝑊11 V + 𝑊02 V + ⋅ ⋅ ⋅ ) (−𝑖𝜔0 V + 𝑔) .
𝜇 = 0, we have
Substituting (52) and (60) into (59) yields
V̇ (𝑡) = ⟨𝑞∗ , 𝑢̇𝑡 ⟩
2
V
= ⟨𝑞∗ , 𝐴 (0) 𝑢𝑡 + 𝑅 (0) 𝑢𝑡 ⟩ 𝑊̇ = (𝐴𝑊20 + 𝐻20 ) + (𝐴𝑊11 + 𝐻11 ) VV
2
(63)
= ⟨𝑞∗ , 𝐴 (0) 𝑢𝑡 ⟩ + ⟨𝑞∗ , 𝑅 (0) 𝑢𝑡 ⟩ (54) V2
+ (𝐴𝑊02 + 𝐻02 ) + ⋅ ⋅ ⋅ .
𝑇 2
= ⟨𝐴∗ 𝑞∗ , 𝑢𝑡 ⟩ + 𝑞∗ (0) 𝐹 (𝑢𝑡 , 0)
𝑇
Comparing the coefficients of (62) and (63), one gets
= 𝑖𝜔0 V (𝑡) + 𝑞∗ (0) 𝑓0 (V, V) .
(𝐴 − 𝑖2𝜔0 ) 𝑊20 (𝜃) = −𝐻20 (𝜃) ,
Rewrite (54) as
𝐴𝑊11 (𝜃) = −𝐻11 (𝜃) , (64)
V̇ (𝑡) = 𝑖𝜔0 V (𝑡) + 𝑔 (V, V) , (55)
(𝐴 + 𝑖2𝜔0 ) 𝑊02 (𝜃) = −𝐻02 (𝜃) .
where
Since 𝑢𝑡 = 𝑢(𝑡 + 𝜃) = 𝑊(V, V, 𝜃) + V𝑞 + V𝑞, then we have
𝑇
𝑔 (V, V) = 𝑞∗ (0) 𝑓0 (V, V)
𝑢1 (𝑡 + 𝜃)
𝑇
= 𝑞∗ (0) 𝐹 (𝑊 (V, V, 𝜃) + 2 Re {V (𝑡) 𝑞 (𝜃) , 0}) (56) 𝑢2 (𝑡 + 𝜃)
𝑢𝑡 = ( )
V2 V2 V2 V 𝑢3 (𝑡 + 𝜃)
= 𝑔20 + 𝑔11 VV + 𝑔02 + 𝑔21 ⋅⋅⋅ .
2 2 2 𝑢4 (𝑡 + 𝜃)
Substituting (42) and (54) into (50) yields 𝑊(1) (V, V, 𝜃) 1
̇ − V̇ 𝑞
(2)
𝑊̇ = 𝑢̇ (𝑡) − V𝑞 𝑊 (V, V, 𝜃) 𝑝1
=( (3)
) + V ( ) 𝑒𝑖𝜔0 𝜃 (65)
∗𝑇 𝑊 (V, V, 𝜃) 𝑝2
= 𝐴𝑢𝑡 + 𝑅𝑢𝑡 − (𝑖𝜔0 V + 𝑞 (0) 𝑓0 (V, V)) 𝑞
𝑊(4) (V, V, 𝜃) 𝑝3
− (𝑖𝜔0 V + 𝑞 ∗𝑇 (0) 𝑓0 (V, V)) 𝑞 (57)
1
= 𝐴𝑢𝑡 + 𝑅𝑢𝑡 − 𝐴V𝑞 − 𝐴V 𝑞 𝑝1
+ V ( ) 𝑒−𝑖𝜔0 𝜃 .
𝑇 𝑝2
− 2 Re (𝑞∗ (0) 𝑓0 (V, V) 𝑞) ,
𝑝3
∗𝑇
{𝐴𝑊−2 Re (𝑞 (0) 𝑓0
{ (V, V) 𝑞) , 𝜃 ∈ [−𝜏, 0) Thus, we obtain
𝑊̇ = {
{ ∗𝑇
{𝐴𝑊−2 Re (𝑞 (0) 𝑓0 (V, V) 𝑞)+𝑓0 (V, V) , 𝜃 = 0, 𝑢1 (𝑡 + 𝜃) = 𝑊(1) (V, V, 𝜃) + V𝑒𝑖𝜔0 𝜃 + V𝑒−𝑖𝜔0 𝜃
(58)
(1) V2 (1) (1) V2
= (𝑊20 (𝜃) +𝑊11 (𝜃) VV+𝑊02 (𝜃) + ⋅ ⋅ ⋅)
which can be written as 2 2

𝑊̇ = 𝐴𝑊 + 𝐻 (V, V, 𝜃) , (59) + V𝑒𝑖𝜔0 𝜃 + V𝑒−𝑖𝜔0 𝜃 ,

10 Abstract and Applied Analysis

𝑢2 (𝑡 + 𝜃) = 𝑊(2) (V, V, 𝜃) + V𝑝1 𝑒𝑖𝜔0 𝜃 + V𝑝1 𝑒−𝑖𝜔0 𝜃 also

(2) V2 (2) (2) V2 𝜙2 (−𝜏) = V𝑝1 𝑒−𝑖𝜔0 𝜏 + V 𝑝1 𝑒𝑖𝜔0 𝜏

= (𝑊20 (𝜃) + 𝑊11 (𝜃) VV + 𝑊02 (𝜃) + ⋅ ⋅ ⋅)
2 2
(2) V2 (2) (2) V2
+ 𝑊20 (−𝜏) + 𝑊11 (−𝜏) VV + 𝑊02 (−𝜏) +⋅ ⋅ ⋅ ,
+ V𝑝1 𝑒𝑖𝜔0 𝜃 + V𝑝1 𝑒−𝑖𝜔0 𝜃 , 2 2

𝑢3 (𝑡 + 𝜃) = 𝑊(3) (V, V, 𝜃) + V𝑝2 𝑒𝑖𝜔0 𝜃 + V𝑝2 𝑒−𝑖𝜔0 𝜃 𝜙3 (−𝜏) = V𝑝2 𝑒−𝑖𝜔0 𝜏 + V 𝑝2 𝑒𝑖𝜔0 𝜏

V2 V2 (3) V2 (3) (3) V2

= (3)
(𝑊20 (3)
(𝜃) + 𝑊11 (3)
(𝜃) VV + 𝑊02 (𝜃) + ⋅ ⋅ ⋅) + 𝑊20 (−𝜏) + 𝑊11 (−𝜏) VV + 𝑊02 (−𝜏) + ⋅ ⋅ ⋅
2 2 2 2
(69)
+ V𝑝2 𝑒𝑖𝜔0 𝜃 + V𝑝2 𝑒−𝑖𝜔0 𝜃 , and hence
(4) 𝑖𝜔0 𝜃 −𝑖𝜔0 𝜃
𝑢4 (𝑡 + 𝜃) = 𝑊 (V, V, 𝜃) + V𝑝3 𝑒 + V𝑝3 𝑒 𝜙2 (−𝜏) 𝜙3 (−𝜏) = 𝑝1 𝑝2 𝑒−2𝑖𝜔0 𝜏0 V2

(4) V2 (4) (4) V2 + 𝑝1 𝑝2 𝑒2𝑖𝜔0 𝜏0 V2 + (𝑝1 𝑝2 + 𝑝1 𝑝2 ) VV

= (𝑊20 (𝜃) + 𝑊11 (𝜃) VV + 𝑊02 (𝜃) + ⋅ ⋅ ⋅)
2 2
1
+ (2𝑝1 𝑒−𝑖𝜔0 𝜏0 𝑊11
(3)
(−𝜏)+𝑝1 𝑒𝑖𝜔0 𝜏 𝑊20
(3)
(−𝜏)
+ V𝑝3 𝑒𝑖𝜔0 𝜃 + V𝑝3 𝑒−𝑖𝜔0 𝜃 . 2
(66) + 2𝑝2 𝑒−𝑖𝜔0 𝜏0 𝑊11
(2)
(−𝜏)) V2 V + ⋅ ⋅ ⋅ .
It is obvious that (70)
(1) V2 (1)
𝜙1 (0) = V + V + 𝑊20 (0) + 𝑊11 (0) VV It follows from (54) that
2
(1) V2 𝑘13 𝜙1 (0) 𝜙1 (0) + 𝑘14 𝜙1 (0) 𝜙2 (0)
+ 𝑊02 (0) + ⋅⋅⋅ ,
2
𝑘24 𝜙1 (0) 𝜙2 (0) + 𝑘25 𝜙2 (0) 𝜙4 (0)
(2) V2 (2) (
𝑓0 (V, V) = (
)
𝜙2 (0) = V𝑝1 + V𝑝1 + 𝑊20 (0) + 𝑊11 (0) VV )
2 𝑘34 𝜙2 (−𝜏) 𝜙3 (−𝜏)
(67)
(2) V2
+ 𝑊02 (0) + ⋅⋅⋅ , 𝑘44 𝜙2 (−𝜏) 𝜙3 (−𝜏)
2 ( )
2 2 2
(71)
(4) V2 (4) 𝐹11 V + 𝐹12 V + 𝐹13 VV + 𝐹14 V V
𝜙4 (0) = V𝑝3 + V 𝑝3 + 𝑊20 (0) + 𝑊11 (0) VV
2
𝐹21 V2 + 𝐹22 V2 + 𝐹23 VV + 𝐹24 V2 V
(4) V2 (
=(
)
+ 𝑊02 (0) + ⋅ ⋅ ⋅ . ),
2 2 2
𝐹31 V + 𝐹32 V + 𝐹33 VV + 𝐹34 V V 2

So
2 2 2
(𝐹41 V + 𝐹42 V + 𝐹43 VV + 𝐹44 V V)
𝜙1 (0) 𝜙1 (0) = V2 + V2 + 2VV
1 where
(1) (1)
+ (4𝑊11 (0) + 2𝑊20 (0)) V2 V + ⋅ ⋅ ⋅ ,
2 𝐹11 = 𝑘13 + 𝑘14 𝑝1 ,
𝜙1 (0) 𝜙2 (0) = 𝑝1 V2 + 𝑝1 V2 + (𝑝1 + 𝑝1 ) VV
𝐹12 = 𝑘13 + 𝑘14 𝑝1 ,
1 (2) (2) (1)
+ (2𝑊11 (0) + 𝑊20 (0) + 𝑊20 (0) 𝑝1 𝐹13 = 2𝑘13 + 𝑘14 (𝑝1 + 𝑝1 ) ,
2
(1) (1)
(1)
+2𝑊11 (0) 𝑝1 ) V2 V + ⋅ ⋅ ⋅ , 𝐹14 = 𝑘13 (2𝑊11 (0) + 𝑊20 (0))
1
𝜙2 (0) 𝜙4 (0) = 𝑝1 𝑝3 V2 + 𝑝1 𝑝3 V2 (2)
+ 𝑘14 (2𝑊11 (2)
(0) + 𝑊20 (0)
2
+ [𝑝1 𝑝3 + 𝑝1 𝑝3 ] VV (1) (1)
+ 𝑊20 (0) 𝑝1 + 2𝑊11 (0) 𝑝1 ) ,
1 (4) (4)
+ (2𝑊11 (0) 𝑝1 + 𝑊20 (0) 𝑝1 𝐹21 = 𝑘24 𝑝1 + 𝑘25 𝑝1 𝑝3 ,
2
(0)
+𝑊20 (2)
(0) 𝑝3 + 2𝑊11 (0) 𝑝3 ) V2 V ⋅ ⋅ ⋅ ; 𝐹22 = 𝑘24 𝑝1 + 𝑘25 𝑝1 𝑝3 ,

(68) 𝐹23 = 𝑘24 (𝑝1 + 𝑝1 ) + 𝑘25 (𝑝1 𝑝3 + 𝑝1 𝑝3 ) ,

Abstract and Applied Analysis 11

1 (2) (2) (1) Comparing the coefficients of the above equation with those
𝐹24 = 𝑘24 (2𝑊11 (0) + 𝑊20 (0) + 𝑊20 (0) 𝑝1
2 in (61), we have
(2)
+ 2𝑊11 (0) 𝑝1 )
2 ∗ ∗ ∗
𝑔20 = (𝐹11 + 𝐹21 𝑝1 + 𝐹31 𝑝2 + 𝐹41 𝑝3 ) ,
1 (4) (4) (2) 𝐷
+ 𝑘25 (2𝑊11 (0) 𝑝1 + 𝑊20 (0) 𝑝1 + 𝑊20 (0) 𝑝3
2 1 ∗ ∗ ∗
𝑔11 = (𝐹13 + 𝐹23 𝑝1 + 𝐹33 𝑝2 + 𝐹43 𝑝3 ) ,
(2) 𝐷
+ 2𝑊11 (0) 𝑝3 ) , (74)
2 ∗ ∗ ∗
𝑔02 = (𝐹12 + 𝐹22 𝑝1 + 𝐹32 𝑝2 + 𝐹42 𝑝3 ) ,
𝐹31 = 𝑘34 (𝑝1 𝑝2 𝑒−2𝑖𝜔0 𝜏0 ) , 𝐷
2 ∗ ∗ ∗
𝐹32 = 𝑘34 (𝑝1 𝑝2 𝑒2𝑖𝜔0 𝜏0 ) , 𝑔21 = (𝐹14 + 𝐹24 𝑝1 + 𝐹34 𝑝2 + 𝐹44 𝑝3 ) .
𝐷

𝐹33 = 𝑘34 (𝑝1 𝑝2 + 𝑝1 𝑝2 ) , We need to compute 𝑊20 (𝜃) and 𝑊11 (𝜃) for 𝜃 ∈ [−𝜏, 0).
Equations (62) and (63) imply that
1
𝐹34 = 𝑘34 (2𝑝1 𝑒−𝑖𝜔0 𝜏0 𝑊11
(3)
(−𝜏) + 𝑝1 𝑒𝑖𝜔0 𝜏0 𝑊20
(3)
(−𝜏)
2 𝑇
𝐻 (V, V, 𝜃) = −2 Re {𝑞∗ (0) 𝑓0 (V, V) 𝑞 (𝜃)}
−𝑖𝜔0 𝜏0 (2)
+ 2𝑝2 𝑒 𝑊11 (−𝜏)) ,
= −2 Re {𝑔 (V, V) 𝑞 (𝜃)}
𝐹41 = 𝑘44 (𝑝1 𝑝2 𝑒−2𝑖𝜔0 𝜏0 ) ,
= −𝑔 (V, V) 𝑞 (𝜃) − 𝑔 (V, V) 𝑞 (𝜃) ,
2𝑖𝜔0 𝜏0
𝐹42 = 𝑘44 (𝑝1 𝑝2 𝑒 ), V2 V2 V2 V
𝐻 (V, V, 𝜃) = − (𝑔20 + 𝑔11 VV + 𝑔02 + 𝑔21 ⋅ ⋅ ⋅ ) 𝑞 (𝜃)
2 2 2
𝐹43 = 𝑘44 (𝑝1 𝑝2 + 𝑝1 𝑝2 ) ,
V2 V2 V2 V
1 −(𝑔20 + 𝑔11 VV + 𝑔02 + 𝑔21 ⋅ ⋅ ⋅) 𝑞 (𝜃) .
𝐹44 = 𝑘44 (2𝑝1 𝑒−𝑖𝜔0 𝜏0 𝑊11
(3)
(−𝜏) + 𝑝1 𝑒𝑖𝜔0 𝜏0 𝑊20
(3)
(−𝜏) 2 2 2
2
(75)
+ 2𝑝2 𝑒−𝑖𝜔0 𝜏0 𝑊11
(2)
(−𝜏)) .
(72) Comparing the coefficients of the above equation with (60),
we have

∗ ∗ ∗ 𝑇
Since 𝑞∗ (0) = (1/𝐷)(1, 𝑝1 , 𝑝2 , 𝑝3 ) , we have 𝐻20 (𝜃) = − 𝑔20 𝑞 (𝜃) − 𝑔02 𝑞 (𝜃) ,
𝐻11 (𝜃) = − 𝑔11 𝑞 (𝜃) − 𝑔11 𝑞 (𝜃) , (76)
∗ 𝑇
𝑔 (V, V) = 𝑞 (0) 𝑓0 (V, V) 𝐻02 (𝜃) = − 𝑔02 𝑞 (𝜃) − 𝑔20 𝑞 (𝜃) .
1 ∗ ∗ ∗
= (1, 𝑝1 , 𝑝2 , 𝑝3 ) It follows from (40) and (64) that
𝐷

𝐹11 V2 + 𝐹12 V2 + 𝐹13 VV + 𝐹14 V2 V 𝑊̇ (𝜃) = 𝐴𝑊20 = 2𝑖𝜔0 𝑊20 (𝜃) − 𝐻20 (𝜃)
(77)
𝐹21 V2 + 𝐹22 V2 + 𝐹23 VV + 𝐹24 V2 V = 2𝑖𝜔0 𝑊20 (𝜃) + 𝑔20 𝑞 (0) 𝑒𝑖𝜔0 𝜃 + 𝑔02 𝑞 (0) 𝑒−𝑖𝜔0 𝜃 .
( )
×( )
2 2 2
𝐹31 V + 𝐹32 V + 𝐹33 VV + 𝐹34 V V
By solving the above equation for 𝑊20 (𝜃) and for 𝑊11 (𝜃), one
2 2 2
(73)
(𝐹41 V + 𝐹42 V + 𝐹43 VV + 𝐹44 V V) obtains

1 ∗ ∗ ∗ 𝑖𝑔20 𝑖𝑔
= ( (𝐹11 + 𝐹21 𝑝1 + 𝐹31 𝑝2 + 𝐹41 𝑝3 ) V2 𝑊20 (𝜃) = 𝑞 (0) 𝑒𝑖𝜔0 𝜃 + 02 𝑞 (0) 𝑒−𝑖𝜔0 𝜃 + 𝐸1 𝑒2𝑖𝜔0 𝜃 ,
𝐷 𝜔0 3𝜔0
(78)
+ (𝐹12 +
∗
𝐹22 𝑝1 +
∗
𝐹32 𝑝2 +
∗
𝐹42 𝑝3 ) V2 𝑖𝑔 𝑖𝑔
𝑊11 (𝜃) = − 11 𝑞 (0) 𝑒𝑖𝜔0 𝜃 + 11 𝑞 (0) 𝑒−𝑖𝜔0 𝜃 + 𝐸2 ,
𝜔0 𝜔0
∗ ∗ ∗
+ (𝐹13 + 𝐹23 𝑝1 + 𝐹33 𝑝2 + 𝐹43 𝑝3 ) VV
where 𝐸1 and 𝐸2 can be determined by setting 𝜃 = 0 in
∗ ∗ ∗
+ (𝐹14 + 𝐹24 𝑝1 + 𝐹34 𝑝2 + 𝐹44 𝑝3 ) V2 V) . 𝐻(V, V, 𝜃).
12 Abstract and Applied Analysis

In fact, we have By the definition of 𝐴 and (40) and (64), we get

0
𝑇
𝐻 (V, V, 0) = −2 Re {𝑞∗ (0) 𝑓0 (V, V𝑞)} + 𝑓0 (V, V) ∫ 𝑑𝜂 (𝜃) 𝑊20 (𝜃) = 𝐴𝑊20 (0) = 2𝑖𝜔0 𝑊20 (0) − 𝐻20 (0) ,
−𝜏0
0
V2 V2 V2 V
= − (𝑔20 + 𝑔11 VV + 𝑔02 + 𝑔21 ⋅ ⋅ ⋅ ) 𝑞 (0) ∫ 𝑑𝜂 (𝜃) 𝑊11 (𝜃) = 𝐴𝑊11 (0) = −𝐻11 (0) .
2 2 2 −𝜏0
(81)
V2 V2 V2 V
−(𝑔20 + 𝑔11 VV + 𝑔02 + 𝑔20 + ⋅ ⋅ ⋅)𝑞 (0)
2 2 2 One can notice that
0
𝐹11 V2 + 𝐹12 V2 + 𝐹13 VV + 𝐹14 V2 V (𝑖𝜔0 𝐼 − ∫ 𝑒𝑖𝜔0 𝜃 𝑑𝜂 (𝜃)) 𝑞 (0) = 0,
−𝜏0
𝐹21 V2 + 𝐹22 V2 + 𝐹23 VV + 𝐹24 V2 V (82)
( ) 0
+( ); (−𝑖𝜔0 𝐼 − ∫ 𝑒−𝑖𝜔0 𝜃
𝑑𝜂 (𝜃)) 𝑞 (0) = 0.
𝐹31 V2 + 𝐹32 V2 + 𝐹33 VV + 𝐹34 V2 V −𝜏0

2 2 2 Thus, we obtain
(𝐹41 V + 𝐹42 V + 𝐹43 VV + 𝐹44 V V)
(79) 0
𝑇
(2𝑖𝜔0 𝐼 − ∫ 𝑒2𝑖𝑤0 𝜃 𝑑𝜂 (𝜃)) 𝐸1 = (𝐹11 , 𝐹21 , 𝐹31 , 𝐹41 )
−𝜏0
comparing the coefficients of the above equations with those (83)
in (61), it follows that 0
𝑇
(∫ 𝑑𝜂 (𝜃)) 𝐸2 = −(𝐹13 , 𝐹23 , 𝐹33 , 𝐹43 ) ,
−𝜏0
𝑇
𝐻20 (0) = −𝑔20 𝑞 (0) − 𝑔02 𝑞 (0) + (𝐹11 , 𝐹21 , 𝐹31 , 𝐹41 ) ,
(80) where 𝐸1 = (𝐸1(1) , 𝐸1(2) , 𝐸1(3) , 𝐸1(4) )𝑇 , 𝐸2 = (𝐸2(1) , 𝐸2(2) , 𝐸2(3) ,
𝑇
𝐻11 (0) = −𝑔11 𝑞 (0) − 𝑔11 𝑞 (0) + (𝐹13 , 𝐹23 , 𝐹33 , 𝐹43 ) . 𝐸2(4) )𝑇 ; the above equation can be written as

2𝑖𝜔0 − 𝑘11 −𝑘12 0 0 𝐹11

−𝑘21 2𝑖𝜔0 − 𝑘22 0 −𝑘23 𝐹21
( −𝑖𝑤0 𝜏0 −𝑖𝜔0 𝜏0
) 𝐸1 = ( ),
0 −𝑘32 𝑒 2𝑖𝜔0 − 𝑘31 − 𝑘33 𝑒 0 𝐹31
0 −𝑘42 𝑒−𝑖𝜔0 𝜏0 −𝑘43 𝑒−𝑖𝜔0 𝜏0 2𝑖𝜔0 − 𝑘41 𝐹41
(84)
𝑘11 𝑘12 0 0 𝐹13
𝑘21 𝑘22 0 𝑘23 𝐹23
( ) 𝐸2 = ( ) .
0 𝑘32 𝑘31 0 𝐹33
0 𝑘42 𝑘43 𝑘41 𝐹43

From (78), (84), we can calculate 𝑔21 , and we can derive the orbitally asymptotically stable with asymptotical phase (unsta-
following parameters: ble) if 𝛽2 < 0 (𝛽2 > 0); the period of the bifurcating periodic
solution increases (decreases) if 𝑇2 > 0 (𝑇2 < 0).
𝑖 󵄨 󵄨2 1 󵄨 󵄨2 𝑔
𝐶1 (0) = (𝑔20 𝑔11 − 2󵄨󵄨󵄨𝑔11 󵄨󵄨󵄨 − 󵄨󵄨󵄨𝑔02 󵄨󵄨󵄨 ) + 21 ,
2𝜔0 3 2
Re (𝐶1 (0))
5. Numerical Simulations
𝜇2 = − ,
Re (𝜆󸀠 (𝜏0 )) (85) In this section, we provide some simulations of model (4)
𝛽2 = 2 Re 𝐶1 (0) , to exhibit the impact of discrete time delay in the model.
We consider the parameters values: Λ = 10, 𝛿1 = 0.06,
Im {𝐶1 (0)} + 𝜇2 Im 𝜆 (𝜏0 ) 𝛿2 = 0.3, 𝑒1 = 0.2, 𝛽 = 0.1, 𝑝 = 1, 𝑐 = 0.1, 𝑏 = 0.02,
𝑇2 = − . 𝑞 = 0.02, 𝜂 ∈ [0, 1], ℎ = 0.1, 𝑟 = 0.03, 𝜖 ∈ [0, 1], and
𝜔0
𝑇max = 1500. According to the given parameters’ values, the
We arrive at the following theorem. threshold critical value 𝜏0 = 0.4957 from the formula (21)
exists. The steady state E+ exists and is asymptotically stable
Theorem 4. The periodic solution is supercritical (subcritical) (see Figure 1). We may notice that the solution converges to
if 𝜇2 > 0 (𝜇2 < 0); the bifurcating periodic solutions are the equilibrium E+ with damping oscillations as the value of 𝜏
Abstract and Applied Analysis 13

300 15

250

200 10

x 150 y

100 5

50

0 0
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
t t
(a) (b)
500 5

400 4

300 3
w z
200 2

100 1

0 0
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
t t
(c) (d)
6 600

4 400
z w
2 200

0 0
15 15
10 300 10 300
200 200
y 5 100 y 5 100
0 0 x 0 0 x

(e) (f)
600 600

400 400
w w
200 200

0 0
6 6
4 300 4 15
200 10
2 2
z 100
x z 5 y
0 0 0 0
(g) (h)

Figure 2: Each panel (from (a) to (h)) shows the time evolution and trajectory of model (4) when 𝜏(= 0.4) < 𝜏0 (critical value) and the effect
of therapies is considered to be 𝜖 = 0.9 and 𝜂 = 0.2. It shows that the endemic steady state E+ of model is asymptotically stable.

increases. Once the delay 𝜏 crosses the critical value 𝜏0 , then 𝜖 = 0.9, 𝜂 = 0.9, and time delay 𝜏 = 15), is shown
the model shows the existence of Hopf bifurcation which is in Figure 4. According to Theorem 4, the parameters 𝐶1 =
depicted in the Figure 2. In Figure 3, we consider the efficacy −2.1108𝑒+004+1.1224𝑒+005𝑖, 𝜆󸀠 = −12.1371−0.6438𝑖, 𝜇2 =
of antiretroviral value is 0.9, which may be responsible for −1.7391𝑒+003, 𝛽2 = −4.2215𝑒+004, and 𝑇2 = −2.8052𝑒+005
the loss of stability. The asymptotic behavior to the infection- are estimated. Based on these values one can conclude that
free steady state, when we consider antiviral treatment (with bifurcating periodic solutions are unstable and decreases in
14 Abstract and Applied Analysis

300 15

250

200 10

x 150 y

100 5

50

0 0
0 100 200 300 400 500 600 0 100 200 300 400 500 600
t t
(a) (b)
500 5

400 4

300 3
w z
200 2

100 1

0 0
0 100 200 300 400 500 600 0 100 200 300 400 500 600
t t
(c) (d)
6 600

4 400
z w
2 200

0 0
15 15
10 300 10 300
200 200
y 5 100 y 5 100
0 0 x 0 0 x

(e) (f)
600 600

400 400
w w
200 200

0 0
6 6
4 300 4 15
200 10
2 2 5
z 100 z y
0 0 x 0 0
(g) (h)

Figure 3: It shows the numerical simulations of model (4), when the time delay of immune activation exceeds the critical value, 𝜏 = 0.5 > 𝜏0 .
The endemic steady state E+ of the model undergoes Hopf bifurcation; stability switch and periodic solutions appear.

the period of bifurcating periodic solutions. The existence of
periodic solution is subcritical. For numerical treatment of
DDEs and related issues; we refer the readers to [35, 36].
Several packages and types of software are available for
the numerical integration and/or the study of bifurcations in
delay differential equations (see, e.g., [37, 38]. In this paper
we utilize MIDDE code [39]) which is suitable to simulate
stiff and nonstiff delay differential equations and Volterra
delay integrodifferential equations, using monoimplicit RK
methods.
Abstract and Applied Analysis 15

300 150

250

200 100

x 150 y

100 50

50

0 0
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
t t
(a) (b)
6000 100

5000
80

4000
60
w 3000 z
40
2000

1000 20

0 0
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
t t
(c) (d)
100 6000

4000
z 50 w
2000

0 0
150 150
100 300 100 300
200 200
y 50 100 y 50 100
0 0 x 0 0 x

(e) (f)
6000 6000

4000 4000
w w
2000 2000

0 0
100 100
300 150
50 200 50 100
z 100 z 50
0 0 x 0 0 y

(g) (h)

Figure 4: It shows the numerical simulations of model (4), when the efficacy rate of antiretroviral treatments is considered to be low; that is,
𝜖 = 0.2 and 𝜂 = 0.2. It shows that the equilibrium E+ of the model undergoes Hopf bifurcation with oscillatory behavior in solutions even
though the delay value is less than the critical value (𝜏 = 0.4 < 𝜏0 ).
16 Abstract and Applied Analysis

300 15
250

200 10

x 150 y

100 5

50

0 0
0 100 200 300 400 500 0 100 200 300 400 500
t t
(a) (b)
1200 15

1000

800 10

w 600 z

400 5

200

0 0
0 100 200 300 400 500 0 100 200 300 400 500
t t
(c) (d)

Figure 5: It shows the numerical simulations model (4) when the efficacy rate of antiretroviral treatment is at expected level, 𝜖 = 0.9 and
𝜂 = 0.9, and the delay value exceeds the critical value 𝜏 = 15 > 𝜏0 . The solution always lies within the feasible region and the infection-free
steady state E0 is asymptotically stable.

6. Concluding Remarks Conflict of Interests

In this manuscript, we provided a conceptual CD4+ T-cell
infection model which includes the logistic growth term
along with two different types of antiretroviral drug therapies.
The model includes a discrete time delay in the immune
activation response, which plays an important role in the
dynamics of the model. The infection-free and endemic
steady states of the model are determined (Figure 5). The
stability of steady states is analyzed. We deduced a formula
that determines the critical value (branch value) 𝜏0 . Necessary
and sufficient conditions for the equilibrium to be asymp-
totically stable for all positive delay values are proved. We
have seen that if the time delay exceeds the critical value
𝜏0 , model (4) undergoes a Hopf bifurcation. The direction
and stability of bifurcating periodic solutions are deduced
in explicit formulae, using center manifold and normal
forms. We also presented some numerical simulations to
the underlying model to investigate the obtained results and
theory. We have seen also that the antiretroviral treatments
help to increase the level of uninfected CD4+ T-cells. The
theoretical results that were confirmed by the numerical
simulations show that the delayed CTL response can lead
to complex bifurcations, and, in particular, the coexistence
of multiple stable periodic solutions. When the time delay
exceeds the critical (threshold) value, we may get subcritical
behaviour that leads to a loss of uninfected CD4+ T-cells.
The authors declare that they have no competing interests for
this paper.
Acknowledgments
This work is supported by United Arab Emirates University,
UAE, (Grant no. NRF-7-20886) and the National Board
for Higher Mathematics, Mumbai (Grant no. 2/48(3)/2012/
NBHM/R&D-II/11020). The authors would like to thank Pro-
fessor Cemil Tunc and referees for their valuable comments.
References
[1] D. Liu and B. Wang, “A novel time delayed HIV/AIDS model
with vaccination & antiretroviral therapy and its stability
analysis,” Applied Mathematical Modelling, vol. 37, no. 7, pp.
4608–4625, 2013.
[2] I. Ncube, “Absolute stability and Hopf bifurcation in a Plasmod-
ium falciparum malaria model incorporating discrete immune
response delay,” Mathematical Biosciences, vol. 243, no. 1, pp.
131–135, 2013.
[3] X. Song, S. Wang, and J. Dong, “Stability properties and Hopf
bifurcation of a delayed viral infection model with lytic immune
Abstract and Applied Analysis
response,” Journal of Mathematical Analysis and Applications,
vol. 373, no. 2, pp. 345–355, 2011.
[4] X. Song, S. Wang, and X. Zhou, “Stability and Hopf bifurcation
for a viral infection model with delayed non-lytic immune
response,” Journal of Applied Mathematics and Computing, vol.
33, no. 1-2, pp. 251–265, 2010.
[5] T. Wang, Z. Hu, and F. Liao, “Stability and Hopf bifurcation
for a virus infection model with delayed humoral immunity
response,” Journal of Mathematical Analysis and Applications,
vol. 411, no. 1, pp. 63–74, 2014.
[6] T. Wang, Z. Hu, F. Liao, and W. Ma, “Global stability analysis for
delayed virus infection model with general incidence rate and
humoral immunity,” Mathematics and Computers in Simulation,
vol. 89, pp. 13–22, 2013.
[7] Z. Bai and Y. Zhou, “Dynamics of a viral infection model with
delayed CTL response and immune circadian rhythm,” Chaos,
Solitons & Fractals, vol. 45, no. 9-10, pp. 1133–1139, 2012.
[8] E. Beretta, M. Carletti, D. E. Kirschner, and S. Marino, “Stability
analysis of a mathematical model of the immune response with
delays,” in Mathematics for Life Science and Medicine, pp. 177–
206, Springer, Berlin, Germany, 2007.
[9] Z. Hu, J. Zhang, H. Wang, W. Ma, and F. Liao, “Dynamics
analysis of a delayed viral infection model with logistic growth
and immune impairment,” Applied Mathematical Modelling,
vol. 38, no. 2, pp. 524–534, 2014.
[10] G. Huang, H. Yokoi, Y. Takeuchi, and T. Sasaki, “Impact of
intracellular delay, immune activation delay and nonlinear
incidence on viral dynamics,” Japan Journal of Industrial and
Applied Mathematics, vol. 28, no. 3, pp. 383–411, 2011.
[11] F. A. Rihan and D. H. Abdel Rahman, “Delay differential model
for tumor-immune dynamics with HIV infection of CD4+ T-
cells,” International Journal of Computer Mathematics, vol. 90,
no. 3, pp. 594–614, 2013.
[12] J. Tam, “Delay effect in a model for virus replication,” Journal of
Mathemathics Applied in Medicine and Biology, vol. 16, no. 1, pp.
29–37, 1999.
[13] Z. Wang and R. Xu, “Stability and Hopf bifurcation in a viral
infection model with nonlinear incidence rate and delayed
immune response,” Communications in Nonlinear Science and
Numerical Simulation, vol. 17, no. 2, pp. 964–978, 2012.
[14] J. E. Mittler, B. Sulzer, A. U. Neumann, and A. S. Perelson,
“Influence of delayed viral production on viral dynamics in
HIV-1 infected patients,” Mathematical Biosciences, vol. 152, no.
2, pp. 143–163, 1998.
[15] F. A. Rihan, D. H. Abdel Rahman, and S. Lakshmanan, “A time
delay model of tumour-immune system interactions: global
dynamics, parameter estimation, sensitivity analysis,” Applied
Mathematics and Computation, vol. 232, pp. 606–623, 2014.
[16] L. Zhang, C. Zhang, and D. Zhao, “Hopf bifurcation analysis
of integro-differential equation with unbounded delay,” Applied
Mathematics and Computation, vol. 217, no. 10, pp. 4972–4979,
2011.
[17] A. S. Perelson, A. U. Neumann, M. Markowitz, J. M. Leonard,
and D. D. Ho, “HIV-1 dynamics in vivo: virion clearance rate,
infected cell life-span, and viral generation time,” Science, vol.
271, no. 5255, pp. 1582–1586, 1996.
[18] A. V. M. Herz, S. Bonhoeffer, R. M. Anderson, R. M. May,
and M. A. Nowak, “Viral dynamics in vivo: limitations on
estimates of intracellular delay and virus decay,” Proceedings of
the National Academy of Sciences of the United States of America,
vol. 93, no. 14, pp. 7247–7251, 1996.
17
[19] M. A. Nowak and C. R. M. Bangham, “Population dynamics of
immune responses to persistent viruses,” Science, vol. 272, no.
5258, pp. 74–79, 1996.
[20] R. M. Anderson and R. M. May, “Population biology of
infectious diseases: part I,” Nature, vol. 280, no. 5721, pp. 361–
367, 1979.
[21] R. J. de Boer and A. S. Perelson, “Target cell limited and immune
control models of HIV infection: a comparison,” Journal of
Theoretical Biology, vol. 190, no. 3, pp. 201–214, 1998.
[22] D. Wodarz and M. A. Nowak, “Specific therapy regimes could
lead to long-term immunological control of HIV,” Proceedings of
the National Academy of Sciences of the United States of America,
vol. 96, no. 25, pp. 14464–14469, 1999.
[23] D. Wodarz, K. M. Page, R. A. Arnaout, A. R. Thomsen, J.
D. Lifson, and M. A. Nowak, “A new theory of cytotoxic T-
lymphocyte memory: implications for HIV treatment,” Philo-
sophical Transactions of the Royal Society B: Biological Sciences,
vol. 355, no. 1395, pp. 329–343, 2000.
[24] C. Lv, L. Huang, and Z. Yuan, “Global stability for an HIV-1
infection model with Beddington-DeAngelis incidence rate and
CTL immune response,” Communications in Nonlinear Science
and Numerical Simulation, vol. 19, no. 1, pp. 121–127, 2014.
[25] A. S. Perelson and P. W. Nelson, “Mathematical analysis of HIV-
1 dynamics in vivo,” SIAM Review, vol. 41, no. 1, pp. 3–44, 1999.
[26] Y. Wang, Y. Zhou, F. Brauer, and J. M. Heffernan, “Viral
dynamics model with CTL immune response incorporating
antiretroviral therapy,” Journal of Mathematical Biology, vol. 67,
no. 4, pp. 901–934, 2013.
[27] M. Pitchaimani, C. Monica, and M. Divya, “Stability analysis for
hiv infection delay model with protease inhibitor,” Biosystems,
vol. 114, no. 2, pp. 118–124, 2013.
[28] H. Shu and L. Wang, “Role of cd4+ t-cell proliferation in HIV
infection u nder antiretroviral therapy,” Journal of Mathematical
Analysis and Applications, vol. 394, no. 2, pp. 529–544, 2012.
[29] S. Wang and Y. Zhou, “Global dynamics of an in-host HIV
1-infection model with the long-lived infected cells and four
intracellular delays,” International Journal of Biomathematics,
vol. 5, no. 6, Article ID 1250058, 2012.
[30] A. Fenton, J. Lello, and M. B. Bonsall, “Pathogen responses
to host immunity: the impact of time delays and memory on
the evolution of virulence,” Proceedings of the Royal Society B:
Biological Sciences, vol. 273, no. 1597, pp. 2083–2090, 2006.
[31] M. A. Nowak, R. M. May, and K. Sigmund, “Immune responses
against multiple epitopes,” Journal of Theoretical Biology, vol.
175, no. 3, pp. 325–353, 1995.
[32] J. Hale, Theory of Functional Differential Equations, Springer,
New York, NY, USA, 1997.
[33] Y. Kuang, Delay Differential Equations with Applications in
Population Dynamics, vol. 191 of Mathematics in Science and
Engineering, Academic Press, San Diego, Calif, USA, 1993.
[34] B. D. Hassard, Theory and Applications of Hopf Bifurcation, vol.
41, CUP Archive, 1981.
[35] C. T. H. Baker, G. Bocharov, A. Filiz et al., Numerical modelling
by delay and volterra functional differential equations, Topics
in Computer Mathematics and its Applications, LEA Athens,
Hellas, 1999.
[36] F. A. Rihan, Numerical treatment of delay differential equations
in bioscience [Ph.D. thesis], The University of Manchester,
Manchester, UK, 2000.
[37] L. F. Shampine and S. Thompson, “Solving DDEs in Matlab,”
Applied Numerical Mathematics, vol. 37, no. 4, pp. 441–458, 2001.
18 Abstract and Applied Analysis

[38] C. Paul, “A user-guide to archi: an explicit runge-kutta code for

so lving delay and neutral differential equations and parameter
estimation problems,” MCCM Techical Report 283, University
of Manchester, 1997.
[39] F. A. Rihan, E. H. Doha, M. I. Hassan, and N. M. Kamel,
“Numerical treatments for Volterra delay integro-differential
equations,” Computational Methods in Applied Mathematics,
vol. 9, no. 3, pp. 292–308, 2009.
Advances in Advances in Journal of Journal of
Operations Research
Hindawi Publishing Corporation
Decision Sciences
Hindawi Publishing Corporation
Applied Mathematics
Hindawi Publishing Corporation
Algebra
Hindawi Publishing Corporation
Probability and Statistics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

The Scientific International Journal of

World Journal
Hindawi Publishing Corporation
Differential Equations
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Submit your manuscripts at

http://www.hindawi.com

International Journal of Advances in

Combinatorics
Hindawi Publishing Corporation
Mathematical Physics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Journal of Journal of Mathematical Problems Abstract and Discrete Dynamics in

Complex Analysis
Hindawi Publishing Corporation
Mathematics
Hindawi Publishing Corporation
in Engineering
Hindawi Publishing Corporation
Applied Analysis
Hindawi Publishing Corporation
Nature and Society
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

International
Journal of Journal of
Mathematics and
Mathematical
Discrete Mathematics
Sciences

Journal of International Journal of Journal of

Hindawi Publishing Corporation Hindawi Publishing Corporation Volume 2014

http://www.hindawi.com
Function Spaces
Hindawi Publishing Corporation
Volume 2014 http://www.hindawi.com http://www.hindawi.com
Stochastic Analysis
Hindawi Publishing Corporation
Optimization
Hindawi Publishing Corporation
Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014