Você está na página 1de 36

Accepted Manuscript

Modeling and mathematical analysis of an initial boundary value problem for the
hepatitis B virus infection

Calvin Tadmon, Severin Foko

PII: S0022-247X(19)30073-3
DOI: https://doi.org/10.1016/j.jmaa.2019.01.047
Reference: YJMAA 22898

To appear in: Journal of Mathematical Analysis and Applications

Received date: 24 September 2018

Please cite this article in press as: C. Tadmon, S. Foko, Modeling and mathematical analysis of an initial boundary value
problem for the hepatitis B virus infection, J. Math. Anal. Appl. (2019), https://doi.org/10.1016/j.jmaa.2019.01.047

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are
providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting
proof before it is published in its final form. Please note that during the production process errors may be discovered which could
affect the content, and all legal disclaimers that apply to the journal pertain.
MODELING AND MATHEMATICAL ANALYSIS OF AN INITIAL
BOUNDARY VALUE PROBLEM FOR THE HEPATITIS B VIRUS
INFECTION

CALVIN TADMON1,2,∗ AND SEVERIN FOKO1

Abstract. In this work, we investigate the hepatitis B virus infection. We first derive a nonlinear PDE
model for the studied biological phenomenon. The obtained initial boundary value problem is completely
analyzed. To begin with the analysis of the model, we use Lou and Zhao Lemma to prove boundedness of
potential solutions. Then we prove global existence, uniqueness and positivity of the solution by a variational
method combined with semigroups theory and some other useful tools from functional analysis. Moreover,
the basic reproduction number R0 determining the extinction or the persistence of the HBV infection is
thoroughly computed via a new method based on spectral properties of differential operators, the residue
Theorem and some arguments from numerical analysis. Also, the global asymptotical properties of the
HBV-free equilibrium of the model are derived via a skilful construction of a suitable Lyapunov function.
Local stability of the endemic equilibrium of the model is studied as well. Finally, numerical simulations are
performed to support the theoretical results obtained.

Keywords: PDE models; HBV infection; Diffusion; Well-posedness; Stability.

1. Introduction

The liver is the human organ that performs the greatest number of chemical transformations necessary
for the smooth functioning of the body. The functional unit of the liver is the hepatic lobule, which consists
mainly of hepatocytes. Hepatitis B virus (HBV) infection is a serious threat to the liver, and its resolution
relies mainly on the immune system. It is called acute when it lasts less than six months, and chronic when
it lasts more than six months. Chronic HBV infection can lead to death due to cirrhosis, liver failure or
hepatocellular carcinoma (liver cancer) [17]. A recent report by the World Health Organization (WHO)
indicates that HBV infection is a major public health problem at global level, and its eradication is a goal
to reach by the year 2030 [31]. As a matter of fact, more than 350 million people are infected by chronic
HBV, and 600 thousands die annually due to its complications mentioned above. Asia and Africa are the
most affected continents, but the other parts of the world are concerned as well [21]. In view of the vital
importance of the liver and the aforementioned facts, any contribution to a better understanding of HBV
infection process is of great interest. Mathematical models have been developed to help understand the
dynamics of HBV within an infected host [20, 11, 15, 23, 26]. In the basic mathematical model proposed
by Nowak et al. [20] the spatial mobility of cells and viruses is ignored. The relevance of spatial motion in
many biological phenomena has been emphasized in [6, 32]. In the study of spatial virus-immune dynamics,

Date: January 23, 2019.


1 Department of Mathematics and Computer Science, Faculty of Science, University of Dschang, P.O. Box: 67,

Dschang – Cameroon.
2 The Abdus Salaam International centre for Theoretical physics, Strada Costiera 11, 34151 Trieste, Italy.
∗ Corresponding author.
E-mail address: ctadmon@ictp.it (C. Tadmon).
Declarations of interest: none.
1
2 CALVIN TADMON-SEVERIN FOKO

Funk et al. [12] adopted a patchy model to simulate the diffusion of virus. In [28, 29] the authors introduced
the random mobility for viruses into Nowak et al. [20] model, and assumed that the motion of virus follows
the Fickian diffusion, that is to say, the population flux of virus is proportional to the concentration gradient
and the proportionality coefficient is negative [14]. In [7] the stability of two degenerate reaction-diffusion
systems was investigated. In this paper, motivated by the works in [28, 29], we formulate and analyze a PDE
model for HBV infection by using the standard incidence function and assuming logistic proliferation for both
healthy and infected hepatocytes as proposed in [24] and some references therein. It is worth mentioning
that in [28] the authors used the mass-action kinetics for viral infection and neglected the proliferation of
hepatocytes, whereas the diffusion of free virions is ignored in [24]. So the obtained model is an extension of
the one in [28].
The paper is organized as follows. In section 2 we present the motivation and assumptions relevant for the
construction of the PDE model for HBV infection. We also propose a mathematical model for the within-
host dynamics of HBV infection based on the understanding of biological interactions between virions and
hepatocytes. Section 3 is devoted to the mathematical analysis of the obtained model. We prove existence,
uniqueness positivity and boundedness of the solution to the established model. Section 4 deals with the
stability analysis of equilibria and numerical simulations. We draw a conclusion and provide a discussion in
section 5.

2. Relevant assumptions and the model

Let Ω ⊂ R3 be the domain representing the liver. Let t ≥ 0 be a given time and x = (x1 , x2 , x3 ) ∈ Ω.
Denote respectively by H(x, t), I(x, t) and V (x, t) the concentrations of healthy hepatocytes, HBV infected
hepatocytes, and free HBV virions at time t and location x. The dynamics of HBV infection is the result
of the dynamics of each compartment H, I, and V , and the various interactions between them. We now
describe the dynamics of each compartment.

2.0.1. Dynamics of healthy hepatocytes. Let V be an elementary volume in Ω. The variation of the
quantity of healthy hepatocytes in V is described under the following assumptions. Healthy hepatocytes
are produced at rate λ from the bone marrow and die at rate d1 H. Uninfected hepatocytes population is
assumed to maintain itself logistically, with homeostatic
 carrying
 capacity KH as proposed in [9, 16, 24, 26].
So, the healthy hepatocytes recruitment is r1 H 1 − KH , where r1 is the maximal per capita growth rate
H+I

or the proliferation rate. Virions infect the healthy hepatocytes at the rate − αHV
H+I , where α is the rate of
transmission of the infection. This standard incidence function replaces the mass-action function which has
been shown to cause unrealistic conditions for successful chronic HBV infection. Thus, the variation of H is
expressed by the following equation:
 
∂H H +I αHV
= λ − d1 H + r1 H 1 − − . (2.1)
∂t KH H +I

2.0.2. Dynamics of HBV infected hepatocytes. The assumptions are as follows. The HBV infected
hepatocytes die at rate δ per day so that 1/δ is the life-expectancy of HBV infected hepatocytes. Healthy
hepatocytes become infected at the rate αHV . As in [9, 24], we assume that HBV infected hepatocytes
 H+I 
proliferate by a complete logistics r2 I 1 − KH , where r2 is the proliferation rate or maximal rate of
H+I

growth per capita of HBV infected hepatocytes. Thus, the variation of I is expressed by the following
3

equation:
 
∂I αHV H +I
= − δI + r2 I 1 − . (2.2)
∂t H +I KH

2.0.3. Dynamics of free HBV virions. We assume the following hypotheses. The infected hepatocytes
produce virus at rate kI, and virus is cleared at the rate −μV . Also, the population of virions decreases
H+I , where u ∈ [0, 1] is a modification parameter. The spatial motion of
due to the infection at the rate − αuHV
virions follows the Fickian diffusion law. Thus, the variation of V is expressed by the following equation:
∂V αuHV
= DΔV + kI − μV − , (2.3)
∂t H +I
where D represents the free HBV virions diffusion coefficient and
∂2 ∂2 ∂2
Δ= 2 + 2 + ,
∂x1 ∂x2 ∂x23
is the usual Laplacian operator in three-dimensional space.

2.1. Forming partial differential equations (PDE) System. In this section we use the above equations
describing variables variations to set a global PDE system modeling biological dynamics for HBV infection.
Let T > 0 be a fixed time and define
ΩT = Ω × (0, T ). (2.4)

Therefore, in ΩT the whole system of PDE governing the HBV infection becomes :
⎧  


∂H
= λ − d1 H + r1 H 1 − H+I − αHV
H+I ,
⎨ ∂t  K H 

∂t = H+I − δI + r2 I 1 − KH
∂I αHV H+I (2.5)
,


⎩ ∂V
∂t = DΔV + kI − μV − H+I .
αuHV

The parameters of the model (2.5), their biological significance and their unit are summed up in Table 1
below.
We use the Neumann homogeneous boundary conditions:
∂V
=0 on ∂Ω × [0, T ], (2.6)
∂η

where ∂η denotes the outward normal derivative on ∂Ω. The initial conditions are the following:

H(x, 0) = H0 (x), I(x, 0) = I0 (x), V (x, 0) = V0 (x), x ∈ Ω. (2.7)

The boundary conditions in (2.6) imply that the free HBV virions do not move across the boundary ∂Ω. For
an epidemiological significance, we assume that the initial conditions are positive and Hölder continuous, and
∂V0
satisfy ∂η = 0 on ∂Ω.
We then obtain the following initial boundary value problem (IBVP):
⎧  


∂H
= λ − d1 H + r1 H 1 − KH  − H+I
H+I αHV
in ΩT ,


∂t 



⎨ ∂t = H+I − δI + r2 I 1 − KH
∂I αHV H+I
in ΩT ,
(2.8)
⎪ ∂t = DΔV + kI − μV − H+I in ΩT ,
∂V αuHV



⎪ ∂V

⎪ ∂η = 0 on ∂Ω × [0, T ],

H(x, 0) = H0 (x), I(x, 0) = I0 (x), V (x, 0) = V0 (x), x ∈ Ω.
4 CALVIN TADMON-SEVERIN FOKO

Table 1. Biological significance of the parameters of the model (2.5).

Parameters Biological significance of the parameters of the model (2.5) Units


λ Production rate of healthy hepatocytes HBV. day−1
d1 Decay rate of healthy hepatocytes HBV. day−1
α Rate of transmission of infection. no unit
δ Decay rate of HBV infected cells. day−1
μ Decay rate of HBV virions. cell/day
r1 Maximal rate of growth per capita of healthy hepatocytes. day−1
r2 Maximal rate of growth per capita of infected hepatocytes. day−1
KH Homeostatic carrying capacity of healthy HBV cells. day−1
k Rate of replication of HBV virions produced by HBV infected cells. day−1
u Modification parameter. no unit
D Free HBV virion diffusion coefficient. no unit

3. Mathematical analysis

In this section, we provide a thorough study of the dynamics of IBVP (2.8) which yields various outcomes.
Precisely, we prove existence, uniqueness, positivity and boundedness of solution for IBVP (2.8). This is done
by combining variational method and semigroups techniques to some useful functional analysis arguments.

3.1. Local existence and uniqueness for the IBVP (2.8). Set

F (H, I, V ) = (F1 (H, I, V ), F2 (H, I, V ), F3 (H, I, V ))T , (3.1)

where  
H +I αHV
F1 (H, I, V ) = λ − d1 H + r1 H 1 − − ,
KH H +I
 
αHV H +I
F2 (H, I, V ) = − δI + r2 I 1 − ,
H +I KH
and
αuHV
F3 (H, I, V ) = kI − μV −.
H +I
We have the following result which insures that the right-hand side, without diffusion, of the PDE-model
system (2.8) is Lipschitz.

Proposition 3.1. Let T > 0 and (H, I, V ) ∈ (CB


0
(Ω × [0, T ]))3 , where CB
0
(Ω × [0, T ]) is the space of bounded
and continuous functions on Ω × [0, T ]. We assume that F in (3.1) is defined on L2 (Ω × (0, T )). Then, F1 ,
F2 and F3 are uniformly Lipschitz continuous on L2 (Ω × (0, T )) with respect to H, I and V.

Proof. Let (H1 , I1 , V1 ), (H2 , I2 , V2 ). First, by direct calculation, one has:

F1 (H1 , I1 , V1 ) − F1 (H2 , I2 , V2 )2 ≤ L11 H1 − H2 2 + L12 I1 − I2 2 + L13 V1 − V2 2 , (3.2)

where
3r1 Tm r1 Tm
L11 = d1 + r1 + + αTm Vm , L12 = + αTm Vm and L13 = 2αTm 2
. (3.3)
KH KH
F2 (H1 , I1 , V1 ) − F2 (H2 , I2 , V2 )2 ≤ L21 H1 − H2 2 + L22 I1 − I2 2 + L23 V1 − V2 2 , (3.4)
where
r2 Tm 3r2 Tm
L21 = + αTm Vm , L22 = δ + r2 + + αTm Vm and L23 = 2αTm
2
. (3.5)
KH KH
5

and
F3 (H1 , I1 , V1 ) − F3 (H2 , I2 , V2 )2 ≤ L31 H1 − H2 2 + L32 I1 − I2 2 + L33 V1 − V2 2 , (3.6)
where
L31 = αuTm Vm , L32 = k + r2 + αuTm Vm and L33 = μ + 2αuTm
2
. (3.7)
The proof is complete. 
Now, consider the following IBVP


⎪ ∂t H = f (t, H, I, V ),





⎨ ∂t I = g(t, H, I, V ), in Ω × (0, T )
∂t V − DΔV = h(t, H, I, V ), (3.8)



⎪ ∂V
∂Ω × (0, T )

⎪ ∂η = 0 on

⎩ H=H , I=I , V =V
0 0 0 on Ω × {t = 0}.
In the sequel, we will need the following definition and results.

Definition 3.2. (Sectorial operator, [15]). Let A be a linear operator in a Banach space X and suppose A
is closed and densely defined. If there exist real numbers a, ω ∈ (0, π), M ≥ 1 such that

ρ(A) ⊃ Σ = {λ0 ∈ C : ω ≤ arg(λ0 − a) ≤ π, λ0 = a} (3.9)

and
M
Rλ0 (A) ≤ for all λ0 ∈ Σ, (3.10)
|λ0 − a|
then we say that A is sectorial.

The Neumann realization of the Laplacian A = −Δ, with domain




∂w
D(A) = w ∈ H 2 (Ω) : =0
∂η
is a sectorial operator in L2 (Ω). But since C0∞ (Ω) ⊂ D(A), it is densely defined in L2 (Ω). For β ≥ 0 large
enough, we define the fractional powers of the Helmholtz operator, Hβ = −Δ + βI, with domain D(Hβ )
equipped with graph norm  · D(Hβ ) =  · 2 + Hβ · 2 . One has the following general results.

Lemma 3.3. ([1]). Let 1 ≤ p < ∞. Then D(Hβ ) ⊂ C0∞ (Ω) with continuous injection for β > n/2p.

Lemma 3.4. ([15]). D(Hβ ) ⊂ CB


0
(Ω) with continuous injection for β > n/4.

Theorem 3.5. ([15]). If A is sectorial, then −A is the infinitesimal generator of an analytic semigroup,
G(t). If Reλ0 > a, a ∈ R whenever λ0 ∈ σ, then for any t > 0,
C −at
G(t) ≤ Ce−at , AG(t) ≤ e , (3.11)
t
and
d
G(t) = −AG(t), t > 0. (3.12)
dt
Corollary 3.6. ([15]). Let G be the analytic semigroup generated by −A. The following properties hold for
the semigroup G and the fractional powers of the Helmholtz operator Hβ :
(1) G(t) : L2 (Ω) → D(Hβ ) for all t > 0,
−β
(2) G(t)wD(Hβ ) ≤ Cβ,2 t w2 for all t > 0, w ∈ L2 (Ω),
(3) G(t)Hβ w = Hβ G(t)w for all t > 0, w ∈ D(Hβ ).
6 CALVIN TADMON-SEVERIN FOKO

The following basic hypotheses are assumed to hold:


(H1) D > 0,
(H2) H0 ≥ 0, I0 ≥ 0 and V0 ≥ 0 are continuous on Ω, H0 , I0 , V0 ∈ CB
0
(Ω),
4
(H3) f, g and h are continuously differentiable functions from R+ into R with f (t, 0, s, z) ≥ 0, g(t, r, 0, z) ≥
0 and h(t, r, s, 0) ≥ 0 for all t, r, s, z ≥ 0.
0 3
For x ∈ Ω, t ≥ 0, H, I, V ∈ CB
0
(Ω), define F, G and Q on R+ × CB (Ω) by

[F(t, H, I, V )](x) = f (t, H(x), I(x), V (x)),


[G(t, H, I, V )](x) = g(t, H(x), I(x), V (x)),
[Q(t, H, I, V )](x) = h(t, H(x), I(x), V (x)).

Moreover, we let G1 , G2 and G3 be the analytical semigroup generated by 0×Δ, 0×Δ and D×Δ respectively.
In the sequel, we will need the following results.

Lemma 3.7. ([15]). If H, I and V are continuous from [0, T ] to L2 (Ω), then the integrals:
 t
I1 (t) = G1 (t − τ )F(τ, H(τ ), I(τ ), V (τ )) dτ,
0
 t
I2 (t) = G2 (t − τ )G(τ, H(τ ), I(τ ), V (τ )) dτ,
0
 t
I3 (t) = G3 (t − τ )Q(τ, H(τ ), I(τ ), V (τ )) dτ,
0
exist and I1 (t), I2 (t) and I3 (t) are continuous on [0, T ) with I3 (t) ∈ D(A) and I3 (t) → 0+ in L2 as t → 0+
and I1 (t) → 0+ in L2 as t → 0+ , I2 (t) → 0+ in L2 as t → 0+ .

Lemma 3.8. If the IBVP (3.8) has a classical solution, then H, I and V satisfy
 t
H(t) = G1 (t)H0 + G1 (t − τ )F(τ, H(τ ), I(τ ), V (τ )) dτ, (3.13)
0
 t
I(t) = G2 (t)I0 + G2 (t − τ )G(τ, H(τ ), I(τ ), V (τ )) dτ, (3.14)
0
 t
V (t) = G3 (t)V0 + G3 (t − τ )Q(τ, H(τ ), I(τ ), V (τ )) dτ. (3.15)
0

Proof. Consider the L2 -valued functions θj (τ ) = Gj (t − τ )wj (τ ), j = 1, 2, 3. Then, θj is differentiable


since Gj is analytic and wj is differentiable. Then by Theorem 3.5, one has
dθ1
= G1 (t − τ )H  (τ ),

= G1 (t − τ )F(τ, H(τ ), I(τ ), V (τ )), (3.16)

dθ2
= G2 (t − τ )I  (τ ),

= G2 (t − τ )G(τ, H(τ ), I(τ ), V (τ )), (3.17)

dθ3
= ΔG3 (t − τ )V (τ ) + G3 (t − τ )V  (τ ),

= ΔG3 (t − τ )V (τ ) − G3 (t − τ )ΔV (τ ) + G3 (t − τ )Q(τ, H(τ ), I(τ ), V (τ )),
= G3 (t − τ )Q(τ, H(τ ), I(τ ), V (τ )). (3.18)
7

Integrating equations (3.16), (3.17) and (3.18) with respect to time, we obtain equations (3.13), (3.14) and
(3.15) respectively. 
Since in our work, n = 3, we take p = 2, so that β > 3/4 and therefore the domain D(Hβ ) is continuously
embedded in C0∞ (Ω) by Lemma 3.3. Now, let H, I and V be continuous functions from [0, T ] to D(Hβ ) → CB
0

satisfying (3.13), (3.14) and (3.15) respectively. One can then claime that H, I and V verify system (3.8).
The continuity of H, I and V implies continuity of t → F(t, H(t), I(t), V (t)), t → G(t, H(t), I(t), V (t)) and
t → Q(t, H(t), I(t), V (t)).
One can then conclude that, the linear Cauchy problem


⎪ ∂t y1 = F(t, H(t), I(t), V (t)),


⎨ ∂ y = G(t, H(t), I(t), V (t)),
t 2

⎪ ∂t y3 − DΔy3 = Q(t, H(t), I(t), V (t)),



y1 (0) = H0 , y2 (0) = I0 , y3 (0) = V0 ,

has a unique solution, with y1 , y2 and y3 given by (3.13), (3.14) and (3.15) respectively.
As in [2, 3, 4, 15, 18], we have the following main result for the local existence of (2.8), based on L2 -theory.

Proposition 3.9. If hypotheses (H1) − (H3) are satisfied, then the initial value problem (3.8) has a unique
0 3
solution (H, I, V ) ∈ CB ((0, T ], D(Hβ )) , with H(0) = H0 ∈ CB
0
(Ω), I(0) = I0 ∈ CB0
(Ω) and V (0) = V0 ∈
0
CB (Ω).

Proof. The proof is established by using Banach’s Fixed Point Theorem.


Choose β such that 3/4 < β < 1, then the injection I : D(Hβ ) → CB
0
is continuous by Lemma 3.4. For
r0 > 0 and T > 0 the following closed ball is considered:
 0 3 
Br0 (H0 , I0 , V0 ) = (H, I, V ) ∈ CB ((0, T ], D(Hβ )) : H − H0 β , I − I0 β , V − V0 β ≤ r0 .
(3.19)
By Proposition 3.1, we have the local Lipschitz properties:

F(t, H1 , I1 , V1 ) − F(t, H2 , I2 , V2 )2 ≤ L11 H1 − H2 D(Hβ ) + L12 I1 − I2 D(Hβ ) + L13 V1 − V2 D(Hβ ) ,
G(t, H1 , I1 , V1 ) − G(t, H2 , I2 , V2 )2 ≤ L21 H1 − H2 D(Hβ ) + L22 I1 − I2 D(Hβ ) + L23 V1 − V2 D(Hβ ) ,
Q(t, H1 , I1 , V1 ) − Q(t, H2 , I2 , V2 )2 ≤ L31 H1 − H2 D(Hβ ) + L32 I1 − I2 D(Hβ ) + L33 V1 − V2 D(Hβ ) ,
(3.20)
for 0 ≤ t ≤ T, (H1 , I1 , V1 ), (H2 , I2 , V2 ) ∈ Br0 (H0 , I0 , V0 ) and Lipschitz-constants L11 , L12 , L13 , L21 , L22 , L23 ,
L31 , L32 , L33 > 0.
Furthermore, for (y1 , y2 , y3 ) ∈ Br0 (H0 , I0 , V0 ), define P : [0, T ] → L2 (Ω), Q : [0, T ] → L2 (Ω) and R :
[0, T ] → L2 (Ω) by

 t
P y1 (t) = G1 (t)H0 + G1 (t − τ )F(t, y1 (τ ), y2 (τ ), y3 (τ )) dτ,
0

 t
Qy2 (t) = G2 (t)I0 + G2 (t − τ )G(t, y1 (τ ), y2 (τ ), y3 (τ )) dτ,
0

 t
Ry3 (t) = G3 (t)V0 + G3 (t − τ )Q(t, y1 (τ ), y2 (τ ), y3 (τ )) dτ.
0
8 CALVIN TADMON-SEVERIN FOKO

Finally, set M1 = sup F(t, y1 (0), y2 (0), y3 (0))2 , M2 = sup G(t, y1 (0), y2 (0), y3 (0))2 ,
t∈[0,T ] t∈[0,T ]
M3 = sup Q(t, y1 (0), y2 (0), y3 (0))2 and choose T so that,
t∈[0,T ]

3r0
G1 (h)H0 − H0 β , G2 (h)I0 − I0 β , G3 (h)V0 − V0 β ≤ , 0 ≤ h ≤ T, (3.21)
4
 T
r0
i
Cβ,2 (Mi + Li1 r0 + Li2 r0 + Li3 r0 ) s−β ds ≤ , i = 1, 2, 3, (3.22)
0 4
i
where Cβ,2 is the constant in property (2) of Corollary 3.6 for A (A is zero for i = 1, 2). Note that such T
exists since G1 (h), G2 (h) and G3 (h) → Id as h → 0+ by definition of an analytic semigroup and
 h
1
s−β ds = h1−β → 0, h → 0+ for β < 1.
0 1−β

The proof then goes as follows:


(a) It is shown that (P, Q, R) maps Br0 (H0 , I0 , V0 ) into itself,
(b) It is shown that (P, Q, R) is a strict contraction on Br0 (H0 , I0 , V0 ), allowing the use of Banach’s Fixed
Point Theorem to get the existence of a unique fixed point in Br0 (H0 , I0 , V0 ).
Let (H, I, V ) ∈ Br0 (H0 , I0 , V0 ) then, using (3.22) and property (2) of Corollary 3.6, one has

P H(t) − H0 D(Hβ )
  t 
 

= G1 (t)H0 − H0 + G1 (t − τ )F(τ, H(τ ), I(τ ), V (τ )) ,

0 D(Hβ )
 t
≤ G1 (t)H0 − H0 D(Hβ ) + G1 (t − τ )F(τ, H(τ ), I(τ ), V (τ ))D(Hβ ) dτ,
 t  0

3r0
≤ + G1 (t − τ ) F(τ, H(τ ), I(τ ), V (τ )) − F(τ, H0 , I0 , V0 )2 + F(τ, H0 , I0 , V0 )2 dτ,
4 0  t
3r0
≤ 1 1 1 1
+ Cβ,2 (M1 + L1 r0 + L2 r0 + L3 r0 ) (t − s)−β ds,
4 0
≤ r0 for 0 ≤ t ≤ T,

and similarly

QI(t) − I0 D(Hβ ) ≤ r0 and RV (t) − V0 D(Hβ ) ≤ r0 .

Showing that (P, Q, R) maps Br0 (H0 , I0 , V0 ) into itself. Furthermore, from property (1) in Corollary 3.6 and
Lemma 3.7 we compute:

P H(t + h) − P H(t)D(Hβ ) = G1 (t)(G1 (h) − Id)H0


 t
+ (G1 (h) − Id)G1 (t − τ )F(τ, H(τ ), I(τ ), V (τ )) dτ
0 t+h
+ G1 (t + h − τ )F(τ, H(τ ), I(τ ), V (τ )) dτ D(Hβ ) .
t

Using Lemma 3.7 and the strong continuity of the semigroup, we obtain that

P H(t + h) − P H(t)D(Hβ ) → 0 as h → 0+ .

Thus P is continuous from [0, T ] into D(Hβ ).


9

A similar computation shows that Q and R have the same properties and item (a) is proved. Now, let
(y1 , y2 , y3 ), (z1 , z2 , z3 ) ∈ Br0 (H0 , I0 , V0 ). Then
 t
P y1 (t) − P z1 (t)D(Hβ ) ≤ G1 (t − τ )(F(τ, y1 (τ ), y2 (τ ), y3 (τ )) − F(τ, z1 (τ ), z2 (τ ), z3 (τ )))D(Hβ ) dτ
0
 t 
−β
≤ Cβ,2 1
(t − τ ) dτ × L11 sup y1 − z1 D(Hβ )
τ ∈[0,t]
0

+L12 sup y2 − z2 D(Hβ ) + L13 sup y3 − z3 D(Hβ ) ,
 τ ∈[0,t] τ ∈[0,t]

1
≤ sup y1 − z1 D(Hβ ) + sup y2 − z2 D(Hβ ) + sup y3 − z3 D(Hβ ) ,
4 τ ∈[0,t] τ ∈[0,t] τ ∈[0,t]

for every t ∈ [0, T ]. Hence


 
1
sup P y1 (t) − P z1 (t)D(Hβ ) ≤ sup y1 − z1 D(Hβ ) + sup y2 − z2 D(Hβ ) + sup y3 − z3 D(Hβ ) .
t∈[0,T ] 4 τ ∈[0,t] τ ∈[0,t] τ ∈[0,t]

Similarly
 
1
sup Qy2 (t) − Qz2 (t)D(Hβ ) ≤ sup y1 − z1 D(Hβ ) + sup y2 − z2 D(Hβ ) + sup y3 − z3 D(Hβ ) ,
t∈[0,T ] 4 τ ∈[0,t] τ ∈[0,t] τ ∈[0,t]

and
 
1
sup Ry3 (t) − Rz3 (t)D(Hβ ) ≤ sup y1 − z1 D(Hβ ) + sup y2 − z2 D(Hβ ) + sup y3 − z3 D(Hβ ) .
t∈[0,T ] 4 τ ∈[0,t] τ ∈[0,t] τ ∈[0,t]

Therefore
sup (P, Q, R)(y1 (t), y2 (t), y3 (t)) − (P, Q, R)(z1 (t), z2 (t), z3 (t))D(Hβ )3
t∈[0,T ]
3
≤ sup (y1 (t), y2 (t), y3 (t)) − (z1 (t), z2 (t), z3 (t))D(Hβ )3 .
4 t∈[0,T ]
Hence (P, Q, R) is a strict contraction on Br0 (H0 , I0 , V0 ), and this proves Part (b). By Banach’s Fixed Point
Theorem, (P, Q, R) has a unique fixed point in Br0 (H0 , I0 , V0 ). This is the solution of (2.8) on [0, T ] with
initial value (H(0), I(0), V (0)) = (H0 , I0 , V0 ) ∈ (D(Hβ ))3 . This completes the proof of Proposition 3.9. 

3.2. Boundedness of the solutions for IBVP (2.8).


2 2,1
Proposition 3.10. Let (H, I, V ) ∈ C 0 (Ω × [0, T )) ∩ C 2,1 (Ω × [0, T )) × C 0 (Ω × [0, T )) ∩ CB (Ω × [0, T ))
be solutions of (2.8) with bounded initial conditions i.e. 0 < H0 (x) ≤ Tm , 0 < I0 (x) ≤ Tm , 0 < V0 (x) ≤ Vm
for all x ∈ Ω, and satisfying the boundary condition ∂V0
∂η = 0 on ∂Ω. We assume d1 < δ, r2 < r1 and r1 > d1 .
Then,
∀(x, t) ∈ Ω × [0, T ], H(x, t) ≤ Tm , I(x, t) ≤ Tm and V (x, t) ≤ Vm

with  
KH (r1 − d1 )  4r2 λ 1/2
Tm = 1+ 1+ and Vm is a constant.
2r1 (r1 − d1 )2 KH

Proof. Adding the first two equations in (2.8), yields


   
∂(H + I) H +I H +I
= λ − d1 H − δI + r1 H 1 − + r2 I 1 − . (3.23)
∂t KH KH
10 CALVIN TADMON-SEVERIN FOKO

Since d1 < δ and r2 < r1 , it follows that


 
∂(H + I) H +I
≤ λ − d1 (H + I) + r1 (H + I) 1 − . (3.24)
∂t KH
Setting Q = H + I, inequation (3.24) becomes
∂Q r1 2
≤ λ + (r1 − d1 )Q − Q . (3.25)
∂t KH
A particular solution of this inequation is
  1/2 
KH (r1 − d1 ) 4r1 λ
0 = 1+ 1+ . (3.26)
2r1 (r1 − d1 )2 KH

Since r1 > d1 , then 0 > 0. Set


1
Q(x, t) = + 0 . (3.27)
Z(x, t)
∂Q(x, t) 1 ∂Z(x, t)
From equation (3.27), one has =− 2
, and inequation (3.25) becomes
∂t z(x, t) ∂t
∂Z(x, t)  4r1 λ 1/2 r1
≥ (r1 − d1 ) 1 + z(x, t) + .
∂t (r1 − d1 ) KH
2 KH
The resolution gives
 
r1 r1
Z(x, t) ≥ Z(x, 0) + ep 0 t − , (3.28)
K H p0 K H p0
where
 4r1 λ 1/2
p0 = (r1 − d1 ) 1 + .
(r1 − d1 )2 KH
Thus, from (3.27) and (3.28), one has
 
1 KH (r1 − d1 )  4r1 λ 1/2
Q(x, t) ≤   + 1+ 1+ , (3.29)
Z(x, 0) + r1
ep0 t − r1 2r1 (r1 − d1 )2 KH
K H p0 KH p0

for t ∈ [0, T ), T > 0 and x ∈ Ω.


Set
1
Y (x, t) = 0 +   , (3.30)
Z(x, 0) + r1
KH p0 ep0 t − r1
KH p0

where 0 is given in equation (3.26). But since Z(x, 0) = 1


Q(x,0)−0 , for all x ∈ Ω, then,

KH p0 (Q(x, 0) − 0 )
Y (x, t) = 0 + .
(KH p0 + r1 (Q(x, 0) − 0 ))ep0 t − r1 (Q(x, 0) − 0 )
If Q(x, 0) > 0 for all x ∈ Ω, then T is given by
 
1 r1 r1
+ ep 0 T − = 0, T > 0, x∈Ω
Q(x, 0) − 0 K H p0 K H p0
that is  
1 r1 (Q(x, 0) − 0 )
T = ln > 0, for all x ∈ Ω.
p0 KH p0 + r1 (Q(x, 0) − 0 )
If Q(x, 0) < 0 , then for all x ∈ Ω, Q is increasing monotonically and Y (x, t) → 0 as t → ∞ and hence we
may take T = ∞.
For Q(x, 0) = 0 , Y (x, t) = 0 , for all t > 0 and hence T = ∞.
11

From this survey, one can conclude that as t → ∞ and x stays in Ω without pointing outwards, one has
 
KH (r1 − d1 )  4r1 λ 1/2
Q(x, t) ≤ 1+ 1+ , ∀(x, t) ∈ Ω × [0, +∞[,
2r1 (r1 − d1 )2 KH

which implies that


H(x, t) ≤ Tm and I(x, t) ≤ Tm , ∀(x, t) ∈ Ω × [0, +∞[, (3.31)
with  
KH (r1 − d1 )  4r1 λ 1/2
Tm = 1+ 1+ .
2r1 (r1 − d1 )2 KH
This prove that H and I are bounded.
Now, consider the last equation of PDE-model system (2.8):
∂V αuHV
= DΔV + kI − μV − . (3.32)
∂t H +I
From the second inequation in (3.31) and equation (3.32), one has
∂V
≤ DΔV + kTm − μV. (3.33)
∂t
By Lemma 1 in Lou and Zhao [19], kTm /μ is the globally attractive steady state for the scalar parabolic
equations
∂V (x, t)
= DΔV (x, t) + kTm − μV (x, t), x ∈ Ω, t ∈ [0, T ]
∂t
(3.34)
∂V (x, t)
= 0, x ∈ ∂Ω, t ∈ [0, T ].
∂η
Since the last equation of IBVP (2.8) is dominated by equation (3.34), the standard parabolic comparison
Theorem 7.3.4 of [22] implies that V is bounded. 

3.3. Global existence, uniqueness and positivity for the IBVP (2.8). We recast the IBVP (2.8) as
follows: ⎧

∂t − DΔw in Ω × [0, T ),
∂w

⎨ + q(w)w = f (w)
∂w3
∂η = 0 on ∂Ω × [0, T ), (3.35)


⎩ w(x, 0) = w (x) in Ω,
0

where w = (w1 , w2 , w3 )T = (H, I, V )T , D = diag(D1 , D2 , D3 ), q(w) = diag(q1 (w), q2 (w), q3 (w)), f (w) =
(f1 (w), f2 (w), f3 (w))T with
 
w 1 + w2 αw3
q1 (w) = d1 − r1 1 − + ,
KH w1 + w2
 
w 1 + w2 αuw1
q2 (w) = δ − r2 1 − , q3 (w) = μ + ,
KH w1 + w2
αw1 w3
f1 (w) = λ, f2 (w) = , f3 (w) = kw2 .
w1 + w2
Note that D1 = D2 = 0, D3 = D > 0. Denote

H = L2 (Ω) and E = H 1 (Ω),

and define as in [13] the Hilbert space




∂w
W (0, T, E, E  ) = w ∈ L2 ((0, T ), E) : ∈ L2 ((0, T ), E  ) , (3.36)
∂t
12 CALVIN TADMON-SEVERIN FOKO

equipped with the norm


 
 ∂w 2
w2W = w2L2 ((0,T ),E) 
+  . (3.37)
∂t L2 ((0,T ),E  )
and the following hypothesis for initial conditions:

w01 , w02 ∈ L∞ (Ω), w03 ∈ H and w0i ≥ 0 for i ∈ {1, 2, 3}. (3.38)

Here, we apply Theorem 2.7 of [13].


So, one approaches the solution by a sequence of solutions of linear equations.
For n = 0, w0 denotes the solution of


⎪ ∂w0

⎪ − DΔw0 = 0 in ΩT ,
⎨ ∂t
w0 (0) = w0 in Ω, (3.39)



⎪ ∂w 0
⎩ 3
= 0. on ∂Ω
∂η
This equation admits a strong solution and w0 ≥ 0.
By induction, wn denotes the solution of
⎧ ∂wn

⎪ − DΔwn + q(wn−1 )wn = f (wn−1 ) in ΩT ,

⎨ ∂t
wn (0) = w0 in Ω, (3.40)


⎪ ∂w3n
⎩ = 0. on ∂Ω
∂η
Since (3.40) is a linear equation, q(wn−1 ) and f (wn−1 ) can replace a0 and f (t) of Corollary 2.10 in [13]
respectively. Suppose that there exists a unique nonnegative solution wn−1 . Assuming by induction that
wi ≥ 0 and that by Proposition 3.10 wi is bounded for 0 ≤ i ≤ n − 1, we have
αuw1n−1
0≤ ≤ αu,
w1n−1 + w2n−1
which implies that
μ ≤ q3 (wn−1 ) ≤ μ + αu. (3.41)

Thanks to the boundedness of wij , one has

δ − r2 ≤ q2 (wn−1 ) ≤ r2 Tm + δ (3.42)

and
d1 − r1 ≤ q1 (wn−1 ) ≤ r1 Tm + d1 . (3.43)

It then follows that q1 (wn−1 ), q2 (wn−1 ), q3 (wn−1 ) ∈ L∞ (Ω×(0, T )). We also have f (wn−1 ) ≥ 0 and f (wn−1 ) ∈
L2 ((0, T ), E  ). Then, by Corollary 2.10 of [13], there exists a unique solution wn ∈ W (0, T, E, E  ) with wn ≥ 0.
αw1 w3
Since f1 (w) = λ, we deduce that w1n remains bounded in C 0 ([0, T ], H) and L2 ((0, T ), E). f2 (w) = ≤
w1 + w2
αw3 and f3 (w) = kw2 , thus f2 (w n−1
) ≤ αw3 n−1
and f3 (w n−1 n−1 2
) = kw2 , remain bounded in L ((0, T ), E);
therefore, w2n and w3n have the same property as w1n .
Now, we deduce that the sequence (wn )n≥0 (one can extract subsequence (wm )m≥0 ) converges weakly in
L2 ((0, T ), E) to wi and weakly star in L∞ ((0, T ), H) to w. Applying Proposition 2.11 of [13], it holds that
 t
n
w (t) = G(t)w0 + G(t − s)gn (s) ds, (3.44)
0
13

where G(t) is the semigroup generated by the unbounded operator −DAH , and

gn (s) = −q(wn−1 (s))wn (s) + f (wn−1 (s)). (3.45)

Then gn ∈ L2 ((0, T ), E). Since the sequence (wn )n≥0 is bounded in C 0 ((0, T ), H), the sequence (gn )n≥0 is
bounded in C 0 ((0, T ), H). Now, consider the operator G from C 0 ((0, T ), H) into C 0 ((0, T ), H) defined by
 t
G(f ) = G(t − s)f (s)ds. (3.46)
0

Let us prove that G is compact. Consider the triple (L2 (Ω), H 1 (Ω), a) with
n 
∂w ∂v
a(w, v) = dx, (3.47)
j=1 Ω
xj xj

where Ω is regular and bounded. As in [13], the unbounded variational operator AH associated to a is
a positive symmetric operator with compact resolvent. It admits a sequence (λk )k of positive eigenvalues
with limk→+∞ λk = +∞ and a Hilbert basis (ek )k of H consisting of eigenvectors of AH . If (G(t))t>0 is the
semigroup generated by AH , then for all w0 ∈ H,

+∞
G(t)w0 = e−tDλk (w0 , ek )ek . (3.48)
k=0

This proves that the operator is compact for all t > 0 since

lim e−tλk = 0. (3.49)


k→+∞

Setting

N
GN (t)w = e−tλk (w, ek )ek , (3.50)
k=0
one sees that GN (t) is an operator with finite rank which converges to G(t). The following Theorem is
relevant in the sequel.

Theorem 3.11. ([13]). Let t → G(t) be an application from [0, ∞) into L(H). One assumes that there exists
a sequence of operators (GN (t))N ≥0 on H verifying the following properties:
(1) for all N and all t > 0, GN (t) has finite rank independent of t,
(2) t → GN (t) is continuous from [0, ∞) into L(H) for all N,
(3) for N → ∞, GN (t) converges to G(t) in L1 ((0, T ), L(H)) for all T > 0.
then the operator G is compact from C([0, T ], H) to C([0, T ], H) for all T > 0.

We are now in the position to prove the global existence, uniqueness and positivity of the solution to the
IBVP (2.8).

Theorem 3.12. If the initial condition satisfies (3.38), then the IBVP (3.35) admits a unique nonnegative
solution w ∈ (W (0, T, E, E  ))3 .

Proof. We first prove the existence and positivity of the solution. From Theorem 3.11, there exist a
sequence (wm ) and a function w such that

wm → w in C([0, T ], H), (3.51)

with w ≥ 0 and w(0) = w0 .


14 CALVIN TADMON-SEVERIN FOKO

We check that

q wm−1 → q(w) and q wm−1 wn → q(w)w in C([0, T ], H). (3.52)

We also have

f wm−1 → f (w) in C([0, T ], H) ∩ L2 ((0, T ), E  ). (3.53)
But, wm is solution of
 m 
∂w  
, v + Awm , v + q wm−1 wm , v = f wm−1 , v ∀v ∈ E. (3.54)
∂t
We take φ ∈ D((0, T )), so that φv ∈ L2 ((0, T ), E),
 T m   T  T  T
∂w m−1 m  m−1 
, φv dt + Aw , φv dt +
m
q w w , φv dt = f w , φv dt. (3.55)
0 ∂t 0 0 0

The second term in the left side and the right side of the equality converges due to the weak convergence in
L2 ((0, T ), E  ). The third term in the left-hand side also converges, due to the convergence in C([0, T ], H).
∂wm
We deduce that converges weakly in L2 ((0, T ), E  ).
∂t
But we have
wm → w in C([0, T ], H). (3.56)
Then
∂wm ∂w
→ in D ((0, T ), H). (3.57)
∂t ∂t
Therefore, we obtain
∂wm ∂w
→ weakly in L2 ((0, T ), E  ),
∂t ∂t
 T   T  T  T
∂w
, φv dt + Aw, φv dt + (q (w) w, φv)H dt = f (w) , φv dt. (3.58)
0 ∂t 0 0 0
This being true for all φ, we have
 
∂w
, v + Aw, v + (q (w) w, v)H = f (w) , v ∀v ∈ E. (3.59)
∂t
That is to say,
d
(w, v)H + a(w, v) + (q (w) w, v)H = f (w) , v ∀v ∈ E,
dt (3.60)
∂w
= f (w) − Aw − q(w)w ∈ L2 ((0, T ), E  )
∂t
One has  t

wm (t) = G(t)w0 + G(t − s) −q wm−1 wm + f wm−1 (s) ds, (3.61)
0

and as q wm−1 wm and f wm−1 converge in C 0 ([0, T ], H), we have
 t
w(t) = G(t)w0 + G(t − s) (−q (w) w + f (w)) (s) ds. (3.62)
0

It remains to prove uniqueness.


Let v be another solution of IBVP (3.35). Then

v ∈ W (0, T, E, E  ) =⇒ v ∈ C([0, T ], H) and v ≥ 0.

Consequently, we obtain
q(v)v + f (v) ∈ L2 ((0, T ), E  ).
15

Thus, by Proposition 2.11 of [13], one has


 t  
v(t) = G(t)w0 + G(t − s) − q(v)v + f (v) (s) ds.
0

Substracting, we have
 t  
w(t) − v(t) = G(t − s) − q(w)w + q(v)v + f (w) − f (v) (s) ds. (3.63)
0

with
q(w)w − q(v)v = q(w)(w − v) + (q(w) − q(v))v.

Since w is positive, one has


 
 wj  1
 
 wk + wi  ≤ K wj ∞ ,
where
wj ∞ = wj L∞ (]0,T [,H) .

If we define

3
w∞ = wj ∞ ,
j=1

there is M1 > 0 such that


q(w)∞ ≤ M1 w∞ .

So, for i = 1, 2, 3, the numerator of qi (w) − qi (v) is the sum of terms of the form (wk − vk )vj or (wj − vj )wk ,
and we can find M2 > 0 such that
⎛ ⎞
3
|qi (w) − qi (v)|H (s) ≤ M2 ⎝ |wj (s) − vj (s)|H ⎠ .
j=1

Also we can find M3 > 0 such that


⎛ ⎞
3
|fi (w) − fi (v)|H (s) ≤ M3 ⎝ |wj (s) − vj (s)|H ⎠ .
j=1

Summing up and noting that Gi (t − s) ≤ Ni eΩi T with Ni , Ωi > 0, we can find M such that

3
|wj (s) − vj (s)|H ≤ M w − v∞ .
j=1

Replacing in (3.63), we obtain



3  t
t2
|wi (s) − vi (s)|H ≤ M 2 w − v∞ s ds = M 2 w − v∞ .
i=1 0 2

By induction, we have

3
Mn n
|wj (s) − vj (s)|H ≤ T w − v∞ ,
j=1
n!

Mn n
lim T w − v∞ = 0.
n→+∞ n!

This ends the proof of Theorem 3.12. 


16 CALVIN TADMON-SEVERIN FOKO

Remark 3.13. It is worth noting that positivity of the solution may be proved by applying the maximum
principle. Moreover, from the above results and the boundedness of the solution, one has that the solution of
IBVP (2.8) enters the region:

Σ = {(H, I, V ) ∈ Ω3 × R3+ : 0 < H(x, t) ≤ Tm , 0 < I(x, t) ≤ Tm , 0 < V (x, t) ≤ Vm },

where  
KH (r1 − d1 )  4r2 λ 1/2
Tm = 1+ 1+ , and Vm is a constant. (3.64)
2r1 (r1 − d1 )2 KH
Hence the region Σ, of biological interest, is positively-invariant under the flow induced by IBVP (2.8).

4. Stability analysis of equilibria

4.1. HBV-free equilibrium. HBV-free equilibrium arises when there is no virus within a host i.e, V = 0
in PDE-model system from (2.8). Simple computation shows that the HBV-free equilibrium for PDE-model
system from (2.8) is
E0 = (Λ, 0, 0), (4.1)
where   1/2 
KH (r1 − d1 ) 4r1 λ
Λ= 1+ 1+ .
2r1 KH (r1 − d1 )2

4.2. Basic reproduction number R0 . One of the main tools in epidemic models is the basic reproduction
number R0 which is an important threshold parameter to discuss the dynamic behavior of the epidemic
model. It quantifies the infection risk. It measures the expected average number of new infected hepatocytes
generated by a single virion in a completely healthy hepatocyte.
Let X := C(Ω, R3 ) be the Banach space of continuous functions with supremum norm  · X . Define
X+ := C(Ω, R3+ ). Then (X, X+ ) is a couple of strongly ordered spaces. Let Y = C(Ω, R) and Y+ = C(Ω, R+ ).
Let φ = (φ1 , φ2 ) ∈ X be the initial value functions, then for every φ, define

(G1 (t)φ1 )(x) = e−δt φ1 (x). (4.2)

Let G2 (t) : Y → Y be the C0 -semigroup associated with the operator DΔ − μ subject to the Neumann
boundary condition, that is,

−μt
(G2 (t)φ2 )(x) = e Γ(x, y, t, D)φ2 (y) dy, t ≥ 0, (4.3)
Ω

where Γ is the Green function associated with DΔ − μ and the Neumann boundary condition. It then follows
from [22, Section 7.1 and Corollary 7.2.3] that G2 (t) : C(Ω, R) → C(Ω, R) is compact and strongly positive,
∀t > 0.
Linearizing (2.8) at the HBV-free equilibrium E0 , we obtain the linearized system
⎧ !   "


∂w1
= r 1 − 2Λ
− d w1 − K
r1 Λ
w2 − αw3 ,
⎨ ∂t 1
!  KH  "
1 H

∂t = r2 1 − KH − δ w2 + αw3 ,
∂w2 Λ (4.4)


⎩ ∂w3
∂t = DΔw3 + kw2 − (μ + αu)w3 ,

subject to the boundary conditions


∂w3
= 0, ∀x ∈ ∂Ω, t > 0.
∂η
17

One can observe that equations for w2 and w3 , which correspond to the infectious compartments, are de-
coupled from w1 , and these two equations constitute a cooperative system, but its solution semiflows are
not compact since the two first equations in (4.4) has no diffusion term. From there, we get the following
cooperative system for the infected host cells and free virus particle:
⎧ !   "

⎪ ∂t2 = r2 1 − KH − δ w2 + αw3 ,
∂w Λ


∂w3
= DΔw3 + kw2 − (μ + αu)w3 , (4.5)


∂t
∂w3
∂η = 0, ∀x ∈ ∂Ω, t > 0,
supplemented by initial conditions. For every initial value functions φ = (φ1 , φ2 ) ∈ X, the solution semiflows
Πt : X → X associated with the linear system (4.5) is defined by

Πt (φ) = (w2 (·, t, φ), w3 (·, t, φ)) ∀φ ∈ X t ≥ 0.

Then, one can observe that Πt is a positive C0 -semigroup on C(Ω, R3 ), and its generator B can be written as
⎛   ⎞
r2 1 − KΛH − δ 0
B=⎝ ⎠.
k DΔ − (μ + αu)
Further, B is a closed and resolvent positive operator.
Setting w2 (x, t) = eλ0 t φ1 (x) and w3 (x, t) = eλ0 t φ2 (x), and substituting them into equations for w2 and
w3 , we obtain the following eigenvalue problem
⎧ !   "
⎨ λ φ (x) = r 1 − Λ − δ φ (x) + αφ (x),
0 1 2 KH 1 2
(4.6)
⎩ λ0 φ2 (x) = DΔφ2 (x) + kφ1 (x) − (μ + αu)φ2 (x),

∂φ2 (x)
= 0, ∀x ∈ ∂Ω, t > 0,
∂η
where φ = (φ1 , φ2 ) ∈ X × X.
From [25, Lemma 2.7] one has the following result concerning the existence of the principal eigenvalue of
(4.6).

Lemma 4.1. ([25]). Suppose s(B) is the spectral bound of B. Since all the parameters are constant, then λ01 =
s(B) is the principal eigenvalue of the eigenvalue problem (4.6) which has a strongly positive eigenfunction.

This means that λ01 is a real eigenvalue with algebraic multiplicity one, and Re(λ) < λ01 for any other
eigenvalue λ of (4.6). Furthermore, λ01 has a corresponding eigenvector φ0 (x) = (φ01 , φ02 ) satisfying φ0 (x) 
0, and any other nonnegative eigenvector of (4.6) is a positive multiple of φ0 (x).
In a recent article by Wang and Zhao [30], the concept of the basic reproduction number is extended to
reaction-diffusion epidemic systems with Neumann boundary conditions. Based on the theory of principle
eigenvalues, the authors defined the basic reproduction number R0 for a reaction-diffusion epidemic system
as the spectral radius of the operator
 ∞  ∞
L(φ(x)) = F(x)G(t)φ dt = F(x) G(t)φ dt. (4.7)
0 0
Consequently, they showed that  ∞
G(t)φ dt = −B −1 φ, (4.8)
0
where B = ∇ · (dI ∇) − VT . It then follows that the next generation operator is

L = −FB −1 . (4.9)
18 CALVIN TADMON-SEVERIN FOKO

Here F is the matrix characterizing the generation of secondary infectious cases/agents, and VT is the matrix
of transition rates between compartments. Both are analogues to the next-generation matrices associated
with the corresponding ODE system (i.e. without diffusion terms). G(t) = (G1 (t), G2 (t)) is the solution
semigroup for the linearized reaction-diffusion system; it denotes the distribution of the initial infection, and
dI = diag[0, D] is the diffusion coefficient vector.
The basic reproduction number of PDE-model system from (2.8) is then

R0 = ρ(L). (4.10)

Since we are in three dimension in space, at the HBV-free equilibrium E0 , we have


  ⎛   ⎞
Λ
0 α −r2 1 − +δ 0
F= , VT = ⎝ KH ⎠
0 0 −k μ + αu
and ⎛   ⎞
Λ
⎜ r2 1− −δ ⎟ 0
KH
B=⎜
⎝ 2 2
⎟.

∂ ∂ ∂2
k D 2 + D 2 + D 2 − (μ + αu)
∂x ∂y ∂z
In order to analyze the basic reproduction number, we proceed to calculate B −1 by solving

B(φ1 , φ2 )T = (y1 , y2 )T (4.11)

subject to homogeneous Neumann boundary conditions. The following result gives the explicit formula for
the basic reproduction number R0 .

Theorem 4.2. Suppose that D is a positive constant and δ > r2 . Then, one has
αk
R0 = !  " , (4.12)
(μ + αu) δ + r2 Λ
KH −1

To prove this theorem, we need the following lemma.

Lemma 4.3. ([27]). Let us consider the following n × n square matrix, n ≥ 1:


⎛ ⎞
N1 P1 N 1 P2 · · · N 1 Pn
⎜ ⎟
⎜ N 2 P1 N 2 P2 · · · N 2 Pn ⎟
⎜ ⎟
X1 = ⎜ .. .. .. .. ⎟, (4.13)
⎜ . . . . ⎟
⎝ ⎠
Nn P1 N n P2 · · · N n Pn
where Ni and Pi (1 ≤ i ≤ n) are arbitrary numbers. Then, the characteristic polynomial of the matrix (4.13)
is  

n
det(λIn − X1 ) = λ n−1
λ− N i Pi . (4.14)
i=1

Proof of Theorem 4.2. Equation (4.11) yields


⎛   ⎞   
r2 1 − KΛH − δ 0 φ1 y1
⎝ ⎠ = (4.15)
∂2 ∂2 ∂2
2 + D ∂y 2 + D ∂z 2 − (μ + αu)
k D ∂x φ2 y2

From equation (4.15), one has


y1 (x)
φ1 (x) =   , for all x ∈ Ω. (4.16)
r2 1 − Λ
KH −δ
19

and  
∂2 ∂2 ∂2 μ + αu ky1 1
− + + φ2 + φ2 = − !  " − y2 . (4.17)
∂x2 ∂y 2 ∂z 2 D D δ + r2 KΛH − 1 D

%2 (ξ), y&1 (ξ) and y&2 (ξ), respectively,


Denote the Fourier transform of φ2 (x, y, z), y1 (x, y, z) and y2 (x, y, z) by φ
with ξ = (ξ1 , ξ2 , ξ3 ) ∈ R3 . Then obtain

%2 (ξ) = − y1 (ξ)
k& y& (ξ)
φ !  " − 2 μ+αu ,
D |ξ|2 + μ+αu
δ + r2 KH − 1
Λ D |ξ|2 + D
D

where |ξ|2 = ξ12 + ξ22 + ξ32 . The inverse Fourier transform then yields
k(2π)−3 (2π)−3
φ(x, y, z) = − !  " K  y1 (x, y, z) − K  y2 (x, y, z), (4.18)
D δ + r2 KΛH − 1 D

where   
1
K=F = eix.ξ &
g(ξ) dξ,
|ξ| + μ+αu
2
D R3

with
1
x = (x, y, z)T and &
g(ξ) = .
|ξ|2 + μ+αu
D
Now, we calculate K. Set

g(x, y, z) = G(r) with r = (x, y, z) = (x2 + y 2 + z 2 )1/2 ,

and
&
g(ξ) = Φ(ρ) with ρ = (ξ12 + ξ22 + ξ32 )1/2 .
Denoting (e1 , e2 , e3 ) the canonical basis of R3 , the value of Φ in ρ is the value of &
g at the point ρe1 . Hence,
one has 
Φ(ρ) = e−iρe1 .x g(x) dx. (4.19)
R3
Using the polar coordinates x = rΘ, where Θ is on the unit sphere S2 of R3 , then it follows from [5, Theorem
B.4.7] that
 ∞ 
Φ(ρ) = e−iρr cos(e1 ,Θ) G(r)r2 drdσΘ , (4.20)
0 S2
where dσΘ is the measure of surface.
On the unit sphere, at each value β of the angle between e1 and Θ, corresponds a circle of length 2π sin β
along which the function to integrate is constant. One has
  π
sin t
e−iρr cos(e1 ,Θ) dσΘ = e−it cos β 2π sin β dβ = 4π ,
S2 0 t
and hence  ∞

Φ(ρ) = r sin(ρr)G(r) dr. (4.21)
ρ 0
Now, we write K as 
1
K = lim eix.ξ dξ, a.e for ξ.
R→∞ |ξ|≤R |ξ|2 + μ+αu
D
Then, from equation (4.20), one has
 R  R
4π ρ sin(rρ) 2π ρ sin(rρ)
K= lim μ+αu dρ = lim μ+αu dρ.
r R→∞ 0 D +ρ 2 r R→∞ −R D + ρ2
20 CALVIN TADMON-SEVERIN FOKO

It then follows that  +∞


2π ρ sin(rρ)
K= μ+αu dρ.
r −∞ D + ρ2
By the residue theorem, it follows that
2π 2 −r√ μ+αu
e
K= D . (4.22)
r
Reporting the value of K in (4.18), one finally gets
 √ μ+αu
k e− D x−τ 
φ(x, y, z) = − !  " y1 (τ ) dτ
4πD δ + r2 KΛH − 1 Δ1 x − τ 
 √ μ+αu (4.23)
1 e− D x−τ 
− y2 (τ ) dτ,
4πD Δ1 x − τ 
where
'
Δ1 = [0, x] × [0, y] × [0, z] and x − τ  = (x − τ1 )2 + (y − τ2 )2 + (z − τ3 )2 .
For consistency in notations, below we will switch φ1 and y1 , and φ2 and y2 in equation (4.23). We look at
the eigenvalue problem L[φ] = λ0 φ, that is,

− FB −1 φ = λ0 φ. (4.24)

A simple computation yields


 √ μ+αu  √ μ+αu
e− D x−τ 
e− D x−τ 
k11 φ1 (τ ) dτ + k12 φ2 (τ ) dτ = λ0 φi , (i = 1, 2), (4.25)
Δ1 x − τ  Δ1 x − τ 
with
αk α
k11 = ! " and k12 = .
4πD δ + r2 KΛH − 1 4πD
(
Using the fact that cosh x − sinh x = e−x and setting q = μ+αu D , (4.25) becomes

cosh qx − τ  − sinh qx − τ 
k11 φ1 (τ ) dτ
1
Δ x − τ 
(4.26)
cosh qx − τ  − sinh qx − τ 
+k12 φ2 (τ ) dτ = λ0 φi , (i = 1, 2).
Δ1 x − τ 
Here, we take Ω =]0, 1[×]0, 1[×]0, 1[, and use the discrete form to analyze the eigenvalue problem associated
with the triple integral equation (4.25). The partition of Ω =]0, 1[×]0, 1[×]0, 1[ uniformly with the different
nodes is
xi = iΔx 0 ≤ i ≤ M, with 0 = x0 < x1 < · · · < xM = M Δx = 1,
yj = jΔy 0 ≤ j ≤ M, with 0 = y0 < y1 < · · · < yM = M Δy = 1,
zk = kΔz 0 ≤ k ≤ M, with 0 = z0 < z1 < · · · < zM = M Δz = 1.
Set
pijk = [xi−1 , xi ] × [yj−1 , yj ] × [zk−1 , zk ].
But, for a function g defined and continuous on Δ1 , one has
 M  M  M
g(x, y, z)dxdydz = lim g(xi , yj , zk )ΔxΔyΔz,
Δ1 M →∞
M →∞ i=1 j=1 k=1
M →∞
pijk ⊂Δ1 ⎧ ⎫ (4.27)
M ⎨
 
 
M 
M ⎬
= lim lim lim g(xi , yj , zk )Δx Δy Δz.
M →∞ ⎩M →∞ M →∞ ⎭
pijk ⊂Δ1 i=1 j=1 k=1
21

Simply approximating each integral by using the right endpoint of each subinterval evaluated at xi , yj and
zk for i = 1, 2, · · · , M, j = 1, 2, · · · , M, and k = 1, 2, · · · , M, respectively, we obtain a matrix equation

AY = λ0 Y, (4.28)

with

Y [φ1 (x1 , y1 , z1 ), φ1 (x2 , y2 , z2 ), · · · , φ1 (xM , yM , zM ), φ2 (x1 , y1 , z1 ), φ2 (x2 , y2 , z2 ), · · · , φ2 (xM , yM , zM )]T .

The problem is now reduced to analyzing the spectral radius, ρ(A), of the coefficient matrix A; in particular,
we have
lim ρ(A) = ρ(L) = R0 .
Δx→0
Δy→0
Δz→0
Using equation (4.28), the coefficient matrix A can be written as:

A = A1 + A2 + A3 + A4 ,

where each matrix Ai0 (1 ≤ i0 ≤ 4), of dimension M × M, results from the discretization of the ith integral
in equation (4.26). Specifically, A1 can be represented by a block form
 
k11 Ã1 0M
A1 = ΔxΔyΔz , (4.29)
0M 0M
,1 = (a1,ij ) is an M ×M lower-triangular
where 0M denotes the zero square matrix of dimension M ×M, and A
matrix given by ⎧ ( 

⎪ μ+αu

⎨ sinh D xi − xj 
a1,ij = − if i ≥ j, (4.30)

⎪ xi − xj 

⎩ 0, otherwise.
In a similar way, one has  
0M ,3
k12 A
A3 = ΔxΔyΔz , (4.31)
0M 0M
,3 = (a1,ij ).
where A
The matrix A2 is given by  
,2
k11 A 0M
A2 = ΔxΔyΔz , (4.32)
0M 0M
,2 = (a2,ij ) given by
with the M × M block A
( 
μ+αu
cosh D xi − xj 
a2,ij = , for 1 ≤ i, j ≤ M, (4.33)
xi − xj 
Finally, the matrix A4 takes the form
 
0M ,4
k12 A
A4 = ΔxΔyΔz , (4.34)
0M 0M
,4 = (a2,ij ). One can observe that the spectral radius for each of A1 , A3 and A4 is 0, that is,
where A

ρ(A1 ) = ρ(A3 ) = ρ(A4 ) = 0. (4.35)

Thus, we are sure that the result will follow only by looking for spectral radius of A2 .
22 CALVIN TADMON-SEVERIN FOKO

,2 is
Then, by Lemma 4.3, it follows that the characteristic polynomial of A
 
det(λIM − A ,2 ,
,2 ) = λM −1 λ − λ (4.36)

with ( 
μ+αu

M  M cosh
M  D xi − xj 
,2 =
λ , (4.37)
i=1 i=1 i=1
xi − xj 
j=1 j=1 j=1

,2 is a positive matrix, i.e., all the entry element of A


Since, cosh is a positive function, A ,2 are positive and
,2 is the unique positive dominant eigenvalue of A
λ ,2 . The form of matrix A2 permit us to arrive at

,2 .
ρ(A2 ) = ΔxΔyΔzk11 λ (4.38)

Now, as presented in [5, p.24-25], we can observe that when Δx → 0, Δy → 0, Δz → 0,


 1 - 
μ + αu
lim ρ(A2 ) = 4πk11 r cosh r dr,
Δx→0 0 D
Δy→0
Δz→0 . -  - 
D μ + αu D μ + αu 4πk11 D
= 4πk11 sinh − 4πk11 cosh + ,
μ + αu D μ + αu D μ + αu
. -  - 
D μ + αu D μ + αu
= 4πk11 sinh − 4πk11 cosh
μ + αu D μ + αu D
αk
+ !  " ,
(μ + αu) δ + r2 KΛH − 1
-  - 
D μ + αu D μ + αu
≤ 4πk11 sinh − 4πk11 cosh
μ + αu D μ + αu D
αk
+ !  " ,
(μ + αu) δ + r2 KΛH − 1
αk
≤ !  " .
(μ + αu) δ + r2 KΛH − 1
(4.39)
This follows from the fact that

-  -  μ + αu
D μ + αu D μ + αu D −
4πk11 sinh − 4πk11 cosh = −4πk11 e D < 0.
μ + αu D μ + αu D μ + αu

Thus, considering R0 as a threshold, it follows that


αk
R0 = !  " .
(μ + αu) δ + r2 Λ
KH −1

We are done with te proof of Theorem 4.2. 

4.3. Local stability of HBV-free equilibrium. Here we deal with the local stability of HBV-free equi-
librium of PDE-model system (2.8) by analyzing the characteristic equations. We have the following result.

Lemma 4.4. The HBV-free equilibrium E0 of PDE-model system (2.8) is locally asymptotically stable when-
ever R0 ≤ 1 and unstable if R0 > 1.
23

Proof. Consider the Laplace operator −Δ and let 0 = μ1 < μ2 < μ3 < · · · be its eigenvalues on Ω with
the homogeneous Neumann boundary condition, and Eμi be the eigenspace corresponding to μi in C 1 (Ω).
Let X = [C 1 (Ω)]3 , {ϕij , j = 1, · · · , dim Eμi } be an orthogonal basis of Eμi and Xij = {ϕij c, / c ∈ R3 }.
Then,
∞ dim Eμi
/ /
X= Xi and Xi = Xij .
i=1 j=1

Using the linearization in (4.4) we obtain

Wt = LW = DΔW + G(E0 )W,

where D = diag(0, 0, D) and



⎛ !  " ⎞
r1 1 − K2ΛH − d1 w1 − Kr1 Λ
w2 − αw3
⎜ !   " H


G(E0 )W = ⎝ r2 1 − KH − δ w2 + αw3
Λ ⎟. (4.40)

kw2 − (μ + αu)w3
For each i ≥ 1, Xi is invariant under the operator L, and λ is an eigenvalue of L if and only if it is an
eigenvalue of the matrix −μi D + G(E0 ) for some i ≥ 1, in which case, there is an eigenvector in Xi . So, we
have
0   0
0 r 1− 2Λ
− d1 − λ −K
r1 Λ
−α 0
0 1 KH   0
0 H
0
det(−μi D + G(E0 ) − λId) = 00 0 r2 1 − Λ
KH −δ−λ α 0 = 0.
0
0 0
0 0 k −μi D − (μ + αu) − λ 0

From there, it follows that the characteristic equation of −μi D + G(E0 ) is


   1    
2Λ Λ
r1 1 − − d1 − λ λ + δ − r2 1 −
2
+ μi D + μ + αu λ
KH KH
      2 (4.41)
Λ Λ
+μi D δ − r2 1 − + (μ + αu) δ − r2 1 − (1 − R0 ) = 0.
KH KH

From (4.41), we get -


4r1 λ
λ0 = − (r1 − d1 )2 + < 0, (4.42)
KH
and       
λ2 + δ − r 2 1 − Λ
KH + μi D + μ + αu λ + μi D δ − r2 1 − KΛH
   (4.43)
Λ
+(μ + αu) δ − r2 1 − (1 − R0 ) = 0.
KH
The discriminant of (4.43) is
   2
Λ
Δ1 = δ − r2 1− − (μi D + μ + αu) + 4αk > 0, (4.44)
KH
and its roots satisfy    
Λ
λ 1 + λ2 = − δ − r 2 1− + μi D + μ + αu < 0, (4.45)
KH
and      
Λ Λ
λ1 · λ2 = μi D δ − r2 1− + (μ + αu) δ − r2 1− (1 − R0 ). (4.46)
KH KH
24 CALVIN TADMON-SEVERIN FOKO

From (4.44), (4.45) and (4.46), we see that for any i ≥ 1, if R0 ≤ 1, all the roots λ0 , λ1 and λ2 of the
equation (4.41) are negative real numbers. Hence, if R0 ≤ 1, the HBV-free equilibrium E0 (Λ, 0, 0) is locally
asymptotically stable.
Now, set
      
h1 (λ) = λ + δ − r2 1 −
2 Λ
KH + μi D + μ + αu λ + μi D δ − r2 1 − Λ
KH
  
Λ
+(μ + αu) δ − r2 1− (1 − R0 ).
KH

If R0 > 1, it is easy to show that for λ real and i = 1 (in this case, μ1 = 0),
  
Λ
h1 (0)|i=1 = (μ + αu) δ − r2 1 − − kα < 0 and lim h1 (λ) = +∞.
KH λ→+∞

Also, If R0 > 1, then from equation (4.46), it holds that

λ1 · λ2 |i=1 = h1 (0)|i=1 < 0.

Hence, equation (4.41) has a positive real root. Therefore, there is a complex root λ with positive real part in
the spectrum of G. Accordingly, if R0 > 1, the HBV-free equilibrium E0 (Λ, 0, 0) is unstable. This completes
the proof of Lemma 4.4. 

4.4. Global stability of the HBV-free equilibrium. Here, we are interested in the global stability of
the HBV-free equilibrium E0 of PDE system from (2.8) by using the method of construction of Lyapunov
function. For this purpose, we start by setting

αk
T0 = .
(μ + αu)(δ − r2 )

Then, it is clear that


αk αk
≤ ,
(μ + αu)(δ − r2 + r2 Λ
KH )
(μ + αu)(δ − r2 )

that is

R 0 ≤ T0 . (4.47)

The result stands as follows. following result.

Theorem 4.5. The HBV-free equilibrium E0 of PDE-model system (2.8) is globally asymptotically stable in
the positively-invariant region Σ if R0 ≤ T0 < 1.

Proof. We first define the function

G1 (t) = kI(t) + (δ − r2 )V (t).

Then, the derivative of G1 (t) with respect to t is

dG1 HV HV r2 k
= αk − μ(δ − r2 )V − αu(δ − r2 ) − I(H + I).
dt H +I H +I KH
25

HV HV
Since ≤ V, we have −V ≤ − , and
H +I H +I
dG1 HV HV HV r2 k
≤ αk − μ(δ − r2 ) − αu(δ − r2 ) − I(H + I),
dt H +I H +I H +I KH
HV   r k
2
= αk − (μ + αu)(δ − r2 ) − I(H + I),
H +I KH
HV  αk  r k
2
= (μ + αu)(δ − r2 ) −1 − I(H + I),
H + I (μ + αu)(δ − r2 ) KH
HV r2 k
= (μ + αu)(δ − r2 ) (T0 − 1) − I(H + I).
H +I KH
Now, we define the Lyapunov function as follows

L1 = G1 dx.
Ω

Calculating the time derivative of L along the positive solutions of model (2.8), we obtain
  
dL1 HV r2 k
≤ (μ + αu)(δ − r2 ) (T0 − 1) − I(H + I) dx.
dt Ω H +I KH
It is clear that the condition T0 ≤ 1 gives dL1
dt ≤ 0 for all H, I, V > 0. We note that the solutions of system
(2.8) are limited by Υ, the greatest invariant subset of E = {(H, I, V ) ∈ Σ | dL1
dt = 0}. We see that dL1
dt = 0 if
and only if V = 0 and I = 0. Each element of Υ satisfies V = 0 and I = 0. By Lyapunov-LaSalle invariance
Theorem, E0 is globally asymptotically stable if T0 < 1. So, we obtain a sufficient condition R0 ≤ T0 which
ensures that the HBV-free equilibrium E0 of PDE-model system (2.8) is globally asymptotically stable if
T0 < 1. 

4.5. Existence of endemic equilibrium. In this section, we investigate existence of a constant endemic
equilibrium of PDE system from (2.8). For this purpose, let E ∗ = (H ∗ , I ∗ , V ∗ ) be an equilibrium point with
infection of PDE system from (2.8), with H ∗ = 0, I ∗ = 0 and V ∗ = 0 satisfying the following system:
⎧  
⎪ ∗ ∗ H ∗ +I ∗ ∗ ∗
⎪ λ − d H + r H 1 − − αH V
H ∗ +I ∗ = 0,
⎨ 1 1
 KH 
αH ∗ V ∗ ∗ ∗ ∗ ∗

H ∗ +I ∗ − δI + r2 I 1 − HK+I = 0, (4.48)

⎪ H
⎩ ∗ ∗
kI ∗ − μV ∗ − αuH V
H ∗ +I ∗ = 0,

H
Set X = . Then 0 < X ≤ 1, and the system (4.48) becomes:
H +I
⎧  ∗


⎪ λ − d1 H ∗ − αV ∗ X + r1 H ∗ 1 − KHH X = 0,
⎨  

αV ∗ X − δI ∗ + r2 I ∗ 1 − KHH X = 0, (4.49)



kI ∗ − μV ∗ − αuV ∗ X = 0,
From the last equation of the system (4.49), one has
k
V∗ = I ∗. (4.50)
μ + αuX
Replacing this expression of V ∗ in the second equation of the system (4.49) yields
 
kαX ∗ ∗ ∗ H∗
I − δI + r2 I 1 − = 0. (4.51)
μ + αuX KH X
Equation (4.51) gives  
kαX H∗
− δ + r2 1 − = 0, (4.52)
μ + αuX KH X
26 CALVIN TADMON-SEVERIN FOKO

Table 2. Number of positive solutions of the equation (4.56) according to the sign of the
coefficients bi i = 0, 1, 2, 3.

b3 b2 b1 Number of positive solutions of the equation (4.56)


+ + + 0
+ + − 0 or 2
+ − + 0 or 2
+ − − 0 or 2
− + + 1
− + − 1 or 3
− − + 1
− − − 1

since I ∗ = 0. From equation (4.52), one deduces that


!    "
KH X (μ + αu) δ + r2 KΛH − 1 (R0 − 1)X + μ(r2 − δ)(1 − X) + r2 Λ
KH (μ + αu)X
H∗ = . (4.53)
r2 (μ + αuX)

Setting
 f (X)  = H  , it is easy to see that f  (X) > 0, ∀X ∈ (0, 1]. Furthermore, f (0) = 0 and f (1) =
KH
r2 δ + r2 KH − 1 (R0 − 1) + Λ > 0 if R0 > 1. Thus ∀X ∈ (0, 1], H ∗ > 0 if R0 > 1. Using the fact that
Λ

H 1−X ∗
X= , we get I ∗ = H and from equations (4.50) and (4.53), we have
H +I X
!    "
(1 − X)KH (μ + αu) δ + r2 KΛH − 1 (R0 − 1)X + μ(r2 − δ)(1 − X) + K r2 Λ
(μ + αu)X
I∗ =
H
, (4.54)
r2 (μ + αuX)
and
!    "
(1 − X)kKH (μ + αu) δ + r2 KΛH − 1 (R0 − 1)X + μ(r2 − δ)(1 − X) + r2 Λ
KH (μ + αu)X
V∗ = . (4.55)
r2 (μ + αuX)2
Now, to explicitly find X, we use the first equation of (4.49), (4.53) and (4.55), to obtain

b3 X 3 + b2 X 2 + b1 X + b0 = 0, (4.56)

where
b0 = λ(μr2 )2 > 0,
b1 = 2λαuμr22 + μKH (r2 − δ)[μ(r1 δ − r2 d1 ) − r2 αk],
b2 = λ(αur2 )2 + α[μr2 KH (r1 − d1 ) − r2 αkKH − 2μr1 KH ][k + u(r2 − δ)] + μr2 αKH (r2 − δ)(k + u(r1 − d1 )),
b3 = αKH [r2 α(k + u(r1 − d1 )) − r1 ][k + u(r2 − δ)].
Existence and number of positive solutions of the equation (4.56) depend on the sign of the coefficients b3 , b2 , b1
and b0 . For this, we used the rule of the signs of Descartes. The obtained results are summarized in Table 2.
We have the following result.

Proposition 4.6. If R0 > 1 and b3 < 0, b2 < 0, b1 > 0 or if R0 > 1 and bi < 0 for all i ∈ {1, 2, 3}, then the
PDE system from (2.8) has a unique positive constant endemic equilibrium.

Indeed, equation (4.56) reads


b 2 2 b1 b0
X3 + X + X+ = 0. (4.57)
b3 b3 b3
27

Setting
b2
X=Y − , (4.58)
3b3
we obtain
Y 3 + pY + q = 0, (4.59)

where
 
b1 b2 b2 2b2 9b1 b0
p= − 22 and q= − + .
b3 3b3 27b3 b3 b3 b3
Using the Cardan formula, we gain a particular solution of (4.59) as follows
. - . -
3 q 3
1 4p + 27q 2 3 q 1 4p3 + 27q 2
Y = − − + − + , (4.60)
2 2 27 2 2 27
where
4b31 27b20 4b0 b42 b21 b22 18b0 b1 b2
4p3 + 27q 2 = + + − − .
b33 b23 b43 b43 b33
From equations (4.58) and (4.60), we have
. - . -
3 q 1 4p3 + 27q 2 3 q 1 4p3 + 27q 2 b2
X= − − + − + − . (4.61)
2 2 27 2 2 27 3b3
Therefore we get H ∗ , I ∗ and V ∗ from the equations (4.53), (4.54) and (4.55) respectively. From Table 2, we
also have the following result.

Proposition 4.7. The PDE system from (2.8) has:


• No positive constant equilibrium points with infection when bi > 0 for all i ∈ {1, 2, 3}.
• Zero or two positive constant equilibrium points with infection if R0 > 1, b3 > 0, b2 > 0, b1 < 0 ;
b3 > 0, b1 > 0, b2 < 0 or when R0 > 1, b3 > 0, b2 < 0, b1 < 0. In this case, thanks to equation (4.56),
there is possibility that the PDE system from (2.8) exhibits the phenomenon of forks bifurcation, that
is, there is a coexistence between a HBV-free equilibrium which is stable and one or more equilibrium
points with infection which is stable when R0 < 1.
• One or three positive constant equilibrium points with infection if R0 > 1, b3 < 0, b1 < 0 and b2 > 0.
• One positive constant equilibrium point with infection if R0 > 1, b3 < 0, b2 < 0, b1 > 0 or if R0 > 1,
bi < 0 for all i ∈ {1, 2, 3}.

4.6. Local stability of the equilibrium with infection. Let us study the local stability of the unique
endemic equilibrium whose existence has been proven analytically. Consider the Laplace operator −Δ and
let 0 = μ1 < μ2 < μ3 < · · · be its eigenvalue on Ω with the homogeneous Neumann boundary condition,
and Eμi be the eigenspace corresponding to μi in C 1 (Ω). Let X = [C 1 (Ω)]3 , {ϕij , j = 1, · · · , dim Eμi } be an
orthogonal basis of Eμi and Xij = {ϕij c, / c ∈ R3 }. Then,

∞ dim Eμi
/ /
X= Xi and Xi = Xij .
i=1 j=1

Now, set w1 = H, w2 = I, w3 = V . Further, we use the vector notation W = (w1 , w2 , w3 )T , and D =


diag(0, 0, D). Then the linearization of the PDE system from (2.8) at E ∗ is of the form Wt = L(W ) =
28 CALVIN TADMON-SEVERIN FOKO

DΔW + G(E ∗ )W , where


⎛     ⎞
∗ r1 ∗ αI ∗ V ∗ ∗ αH ∗ V ∗ ∗
r1 − d1 − K 2r1
H − I − ∗ ∗ 2 w 1 + − r1
H + ∗ ∗ 2 w2 − HαH∗ +I ∗ w3
⎜  H K H  (H +I ) K H (H +I 
) ⎟
⎜ αI ∗ V ∗ αH ∗ V ∗ ∗ ⎟
G(E ∗ )W = ⎜ ∗ +I ∗ )2 − K
r2 ∗
I w 1 + r 2 − δ − r2
H ∗
− 2r2 ∗
I − ∗ +I ∗ )2 w2 + HαH
∗ +I ∗ w3 ⎟
⎝ (H H  K H  K H  (H  ⎠
αuI ∗ V ∗ αuH ∗ V ∗ ∗
− (H ∗ +I ∗ )2 w1 + k + (H ∗ +I ∗ )2 w2 − μ + HαuH ∗ +I ∗ w3
(4.62)
For each i ≥ 1, Xi is invariant under the operator L, and λ is an eigenvalue of L if and only if it is an
eigenvalue of the matrix −μi D + G(E0 ) for some i ≥ 1, in which case, there is an eigenvector in Xi .
The characteristic equation of −μi D + G(E ∗ ) is of the form

λ3 + a2 λ2 + a1 λ + a0 = 0, (4.63)
1
where setting c∗ = ∗ ,
 H + I∗  
2r1 ∗ r1 ∗ r2 ∗ 2r2 ∗
a0 = (μi D + μ) d1 − r1 + H + I + αc2∗ I ∗ V ∗ δ − r2 + H + I
 KH KH   KH  KH
2r 1 r 1 r 2 ∗ 2r 2 ∗
+αuc∗ H ∗ d1 − r1 + H∗ + I ∗∗ δ − r2 + H + I
KH  KH KH 2 KH
r2 ∗ r1 ∗ α ur1 2 ∗ ∗ ∗
+ I (μi D + μ + αuc∗ H ∗ ) αc2∗ H ∗ V ∗ − H − c H I V
KH  K H  KH ∗
2r1 ∗ r1 ∗ αr1 2 ∗ ∗ ∗
+αc2∗ H ∗ V ∗ (μi D + μ) d1 − r1 + H + I + c H I V (μi D + μ)
KH KH KH ∗
αr1
+α3 uc5∗ I ∗ (H ∗ V ∗ )2 (H ∗ + I ∗ − 1) + kαc∗ H ∗ (αc2∗ I ∗ V ∗ − 1) − kc∗ H ∗ I ∗ > 0,
   K H
2r1 ∗ r1 ∗
a1 = (μi D + μ) d1 − r1 + H + I + αc2∗ I ∗ V ∗ + μi D + μ + αuc∗ H ∗ + d1 − r1
 K H K H   
2r1 ∗ r2 ∗ 2r2 ∗ 2r1 ∗ r1 ∗
+ H + αc2∗ I ∗ V ∗ δ − r2 + H + I + αuc∗ H ∗ d1 − r1 + H + I
KH   KH KH KH KH
r1 ∗ 2r2 ∗ αr2 2 ∗ ∗ ∗
+ I δ − r2 + I + c H I V
KH  KH KH ∗ 
2r1 ∗ r1 ∗ r1 α 2 ∗ ∗ ∗
+αc2∗ H ∗ V ∗ μi D + μ + d1 − r1 + H + I + c H I V − kαc∗ H ∗ > 0,
KH KH KH ∗
2r1 + r2 ∗ r1 +2r2 ∗
a2 = μi D + μ + αuc∗ H ∗ + αc2∗ I ∗ V ∗ + (δ − r2 ) + (d1 − r1 ) + H + KH I + αc2∗ H ∗ V ∗ > 0,
 K H 
2r1 ∗ 
∗ 2 ∗ ∗ r1 ∗
a1 a2 − a0 = αuc∗ H + αc∗ I V + d1 − r1 + H + KH I a1 + (μi D + μ) (μi D + μ) d1 − r1
K
 H 
2r1 ∗ r1 ∗ 2r1 ∗
+ H + I + αc2∗ I ∗ V ∗ + μi D + μ + αuc∗ H ∗ + d1 − r1 + H + αc2∗ I ∗ V ∗
K
H KH   KH
r2 ∗ 2r2 ∗ ∗ 2r1 ∗ r1 ∗
× δ − r2 + H + I + αuc∗ H d1 − r1 + H + I
 KH KH KH KH
r1 ∗ 2r2 ∗ αr2 2 ∗ ∗ ∗
+ I δ − r2 + I + c H I V − kαc∗ H ∗
KH KH KH ∗
 
2 ∗ ∗ 2r1 ∗ r1 ∗
+αc∗ H V μ i D + μ + d 1 − r1 + H + I + α2 c4∗ H ∗ I ∗ V ∗ (μi D + μ)
KH KH
  
r2 ∗ r2 ∗ αr2 2 ∗ ∗ ∗
+αc∗ H V (μi D + μ + αuc∗ H ∗ ) δ − r2 +
2 ∗ ∗
H + I + c H I V
KH KH KH ∗
    
r2 ∗ 2r2 ∗ 2r1 ∗ 2 ∗ ∗ r1 ∗ 2r2 ∗
+ δ − r2 + H + I d 1 − r1 + H + αc∗ I V + I δ − r2 + I
 KH KH  KH KH KH
2r1 ∗ r1 ∗ αr1 2 ∗ ∗ ∗
+αuc∗ H ∗ d1 − r1 + H + I + c H I V − kαc∗ H ∗
KH KH KH ∗
 
2r 1 r 1
+αc2∗ H ∗ V ∗ μi D + μ + d1 − r1 + H∗ + I∗
KH KH
29

  
r2 ∗ 2r2 ∗ 2r1 ∗
+ δ − r2 + H + I μi D + μ + αuc∗ H ∗ + d1 − r1 + H + αc2∗ I ∗ V ∗
KH KH KH
   
r2 ∗ 2r2 ∗ r1 ∗ 2r2 ∗ αr2 2 ∗ ∗ ∗
× δ − r2 + H + I + I δ − r2 + I + c H I V
KH KH KH KH KH ∗ 
 
2 ∗ ∗ 2r1 ∗ r1 ∗ αr1 2 ∗ ∗ ∗ ∗
+αc∗ H V + μi D + μ + d1 − r1 + H + I + c H I V − kαc∗ H
KH KH KH ∗
r1 r2 α2 ur1 2 ∗ ∗ ∗ kαr1
+ 2 H ∗ I ∗ (μi D + μ + αuc∗ H ∗ ) + c H I V + c∗ H ∗ I ∗
KH KH ∗ KH
+α3 uc4∗ I ∗ (H ∗ V ∗ )2 (1 − c∗ ) + kαc∗ H ∗ (1 − αc2∗ I ∗ V ∗ ) > 0.

From the above result, it then follows from Routh-Hurwitz criterion that all roots of (4.63) have negative
real parts and hence we have the following proposition.

Proposition 4.8. If R0 > 1, then the equilibrium point with infection E ∗ of the PDE system from (2.8) is
locally asymptotically stable.

Now, we use another result to prove the local stability of the equilibrium point with infection. Proposition
4.6 establishes that R0 = 1 is a bifurcation parameter. In fact, in presence of the diffusion, across R0 = 1 the
HBV-free equilibrium, E0 changes its stability property from local stability to unstable (see Lemma 4.4). We
investigate the appearance of the transcritical bifurcation at R0 = 1 where the stable HBV-free equilibrium
E0 becomes unstable when R0 crosses one from below and gives rise to the stable endemic equilibrium E ∗ .
Since we are in presence of the diffusion, we will complete that bifurcation parameter. For this purpose, the
PDE-model system (2.8) can be written in the form Wt − DΔW = F (W ), with F = (F1 , F2 , F3 )T , as follows:
⎧  
⎪ ∂w1 w 1 + w2 αw1 w3

⎪ = F = λ − d w + r w 1 − − ,


1 1 1 1 1
⎨ ∂t  KH w1 + w2
∂w2 αw1 w3 w 1 + w2
= F2 = − δw2 + r2 w2 1 − , (4.64)

⎪ ∂t w1 + w2 KH

⎪ ∂w3

⎩ αuw1 w3
− DΔw3 = F3 = kw2 − μw3 − .
∂t w1 + w2
Note that system (4.64) can also be written as an ODE system of the form

dw
= X(w) = Aw + F (w), (4.65)
dt
where A = DΔ. System (4.65) has a HBV-free equilibrium given by E0 = (w10 , 0, 0), i.e., X(E0 ) = 0, with
  1/2 
0 KH (r1 − d1 ) 4r1 λ
w1 = Λ = 1+ 1+ .
2r1 KH (r1 − d1 )2

Also, using the linearization in (4.4) we obtain

Wt = DΔW + G(E0 )W, (4.66)

where

⎛ !  " ⎞
r1 1 − K2ΛH − d1 w1 − Kr1 Λ
w2 − αw3
⎜ !   " H


G(E0 )W = ⎝ r2 1 − KH − δ w2 + αw3
Λ ⎟. (4.67)

kw2 − (μ + αu)w3
30 CALVIN TADMON-SEVERIN FOKO

Let 0 = μ1 < μ2 < μ3 < · · · be the eigenvalues of the Laplacian operator −Δ. Then the characteristic
equation of −μi D + G(E0 ) at E0 is
  
  
2Λ Λ
λ + d1 + r1 −1 λ + δ + r2 −1 (λ + μ + αu + μi D) − (λ + kα) + λ = 0. (4.68)
KH KH
Now, we distinguish two cases.
  
(μ+αu+μi D) δ+r2 KΛ −1
(1) If k = α
H
, for all i ∈ N∗ , then zero is not the root of equation (4.68). So, the
second member of equation (4.68) can be written as
 
1 2 μ + αu + μ i D + δ + r 2
Λ
KH − 1 1
λ + λ+ − 1 = 0. (4.69)
kα kα R0
1
The discriminant of equation (4.69) is positive if − 1 < 0, i.e., R0 > 1. Then by Lemma 4.4, E0
R0
is unstable and thus gives rise to the stable endemic equilibrium E ∗ , i.e., the infection persist within
a host.   
(μ+αu+μi D) δ+r2 KΛ −1
∗ H
(2) If k = k = α , for some i, then zero is the root of equation (4.68). Especially,
if i = 1, one has μ1 = 0, which implies R0 = 1. This corresponds to bifurcation points of the HBV-free
equilibrium. From there, k becomes the bifurcation parameter and we get
   
μ Λ
k∗ = + u δ + r2 −1 . (4.70)
α KH

From equation (4.66), taking k = k ∗ given in (2), it follows that the characteristic equation of −μi D + G(E0 )
at E0 , denoted by −μi D + Gk∗ , is
   1     2
2Λ Λ
r1 1 − − d1 − λ λ2 + δ + r 2 − 1 + μi D + μ + αu λ = 0. (4.71)
KH KH

From equation (4.71) it follows that −μi D + Gk∗ has zero as a simple eigenvalue and all other eigenvalues
have negative real parts. Hence, the Center Manifold theory can be used to analyze the dynamics of system
(4.65).
Now, we compute the left and right eigenvectors of −μi D + Gk∗ .
The right eigenvector of −μi D + Gk∗ , denoted U = (u1 , u2 , u3 )T , verifies the following equation:

(−μi D + Gk∗ − λI3 )U = 0R3 ,

where I3 is the 3 × 3 identity matrix and 0R3 is the null vector of dimension 3. For λ = 0, we have
⎛   ⎞⎛ ⎞ ⎛ ⎞
r1 1 − K2ΛH − d1 −Kr1 Λ
−α u1 0
⎜  H  ⎟⎜ ⎟ ⎜ ⎟
⎜ 0 r2 1 − KΛH − δ α ⎟ ⎜ u2 ⎟ = ⎜ 0 ⎟ ,
⎝ ⎠⎝ ⎠ ⎝ ⎠
0 k −μi D − μ − αu u3 0

that is readily solved to give


⎧ !   "

⎪ r1 Λ(μi D + μ + αu) + kαKH r1 1 − K2ΛH − d1

⎪ u1 = !   "

⎪ u2 ,
⎨ KH (μi D + μ + αu) r1 1 − 2Λ − d1 KH

⎪ u2 = u2 > 0,



⎪ k
⎩ u3 = u2 .
μi D + μ + αu
31

The left eigenvector of −μi D + Gk∗ , denoted V = (v1 , v2 , v3 )T , verifies the following equation:

V (−μi D + Gk∗ − λI3 ) = 0R3 .

For λ = 0, one has


⎛   ⎞
r1 1 − 2Λ
− d1 −K
r1 Λ
−α
 ⎜ KH  H  ⎟  
v1 v2 v3 ⎜ 0 r2 1 − Λ
−δ α ⎟= 0 0 0 ,
⎝ KH ⎠
0 k −μi D − μ − αu

that is easily solved to give




⎪ v = 0,
⎨ 1
v2 = v2 > 0,

⎪ α
⎩ v3 = v2 .
μi D + μ + αu
Now, we compute some coefficients which can help us gain the local stability of equilibrium point with
infection. We call them a and b, defined as follows:


3
∂ 2 Fk 
3
∂ 2 Fk
a= v k ui uj (E0 ), b= v k ui (E0 ),
∂wi ∂wj ∂wi ∂α
k,i,j=1 k,i=1

where ui , i = 1, 2, 3 and vk , k = 1, 2, 3 are components of left and right eigenvector of −μi D +Gk∗ , respectively
and Fk , k = 1, 2, 3 are define in the system (4.64).
We first compute b. For the system (4.64), the associated non-vanishing second partial derivatives of F in
terms of wj , j = 1, 2, 3 and in terms of α at E0 are:

∂ 2 F1 ∂ 2 F2 ∂ 2 F3
(E0 ) = −1, (E0 ) = 1 and (E0 ) = −u.
∂w3 ∂α ∂w3 ∂α ∂w3 ∂α

Thus we get


3
∂ 2 Fk
b = v k ui (E0 ),
∂wi ∂α
k,i=1

∂ 2 F1 ∂ 2 F2 ∂ 2 F3
= v1 u 3 (E0 ) + v2 u3 (E0 ) + v3 u3 (E0 ),
∂w3 ∂α ∂w3 ∂α ∂w3 ∂α
∂ 2 F2 ∂ 2 F3
= v2 u3 (E0 ) + v3 u3 (E0 ), since v1 = 0,
∂w3 ∂α ∂w3 ∂α
= v2 u3 − uv3 u3 ,
!  "
(μi D + μ) δ + r2 KΛH − 1
= v2 u 2 > 0.
α(μi D + μ + αu)

We now compute a. For the system (4.64), the associated non-vanishing second partial derivatives of F in
terms of wj , j = 1, 2, 3 at E0 are:

∂ 2 F2 2r2 ∂ 2 F2 r2 ∂ 2 F2 α ∂ 2 F3 αu
(E0 ) = − , (E0 ) = − , (E0 ) = − and (E0 ) = .
∂w22 KH ∂w1 ∂w2 KH ∂w2 ∂w3 Λ ∂w2 ∂w3 Λ
32 CALVIN TADMON-SEVERIN FOKO

Thus

3
∂ 2 Fk
a = v k ui uj (E0 ),
∂wi ∂wj
k,i,j=1
 
∂ 2 F2 2
2 ∂ F2 ∂ 2 F2 ∂ 2 F3
= v2 2u1 u2 (E0 ) + u2 (E 0 ) + 2u 2 u 3 (E 0 ) + 2v 3 u 2 u 3 (E0 ),
∂w1 ∂w2 ∂w22 ∂w2 ∂w3 ∂w2 ∂w3
 
r2 r 2 α αu
= v2 − 2u1 u2 − 2u22 − 2u2 u3 + 2v3 u2 u3 ,
KH KH Λ Λ
⎧ !  " !  "
⎨ r2 KH δ + r2 KΛ − 1 d 1 + r 1

KH − 1 − r1 r2 Λ r2
H
= −2v2 u22 !  " +
⎩ K H d 1 + r1 K H − 1
2 2Λ K H
  !  " ⎫
δ + r2 KΛH − 1 δ + r2 KΛH − 1 αu ⎬
+ −
Λ Λ(μ + αu)R0 ⎭
⎧ !  " !  "
⎨ r2 KH δ + r2 KΛ − 1 d1 + r1 K2ΛH − 1 − r1 r2 Λ r2
H
= −2v2 u22 !  " +
⎩ K2 d + r 2Λ
−1 KH
H 1 1 KH
  ⎫
δ + r2 Λ
KH − 1  μR + αu(R − 1) ⎬
0 0
+
Λ (μ + αu)R0 ⎭

From there, we have that, if R0 > 1 and close to 1, a < 0. Hence, summarizing the above discussions, we
arrive at the following result.

Proposition 4.9. If R0 > 1 and close to 1, then the unique infected steady state E ∗ of PDE system from
(2.8) is locally asymptotically stable.

Remark 4.10. From the computation of the coefficients a and b, we see that even if one neglects the diffusion,
we would always get the local asymptotic stability of the unique endemic equilibrium of PDE system from (2.8),
guaranteed analytically. This means that the free diffusion of the virus has no effect on the local stability of
the endemic equilibrium.

It follows from this Remark 4.10 that, when diffusion is neglected, zero is a simple eigenvalue of Gk∗ and
the other eigenvalues of Gk∗ have negative real parts. The matrix Gk∗ has one right eigenvector U and one
left eigenvector V (each one corresponding to the zero eigenvalue). Moreover, we have b > 0 and a < 0 if
R0 ≥ 1. Then, according to Castillo-Chavez and Song [8], the endemic equilibrium is locally asymptotically
stable with the basic reproduction number near 1.
Now, since we have used the HBV-free equilibrium to prove it, we have the following remark.

Remark 4.11. Proposition 4.9 implies that HBV can persist within the body of an infected host (when
R0 > 1 but close to 1) if the initial sizes of infected cells and virions are in the basin of attraction of the
unique endemic equilibrium guaranteed analytically.

4.7. Numerical simulations. In this section, we present some numerical simulations to support the-
oretical results obtained in the previous sections. The numerical values of the parameters that we use for
the simulations are given in Table 3. Figure 1 illustrates the case R0 ≤ T < 1. From this figure, it clearly
appears that all trajectories converge to HBV-free equilibrium. It follows that the HBV-free equilibrium is
33

Table 3. Numerical values of the parameters of PDE-model system (2.8).

Parameters Values Parameters Values Parameters Values


4 −3
λ 3 × 10 d1 2 × 10 α 5.44 × 10−4
−2 −3
r1 3 × 10 r2 10 μ 2.15 × 10−2
δ 0.0987 KH 106 k 0.8898
−2
u 8 × 10 D 10

globally asymptotically stable. In this case, the infection could disappear within-host. Figure 2 illustrates
the case corresponding to R0 > 1 but close to 1. From this figure, one clearly sees that the trajectories
converge to one equilibrium with infection. In this case, the infection could persist within-host.

Figure 1. Simulations of IBVP (2.8) using various initial conditions when α = 5.44 × 10−4 , δ = 0.0987
(so that R0 = T0 = 0.0022 < 1). All other parameter values are as in Table 3.

5. Concluding remarks

Based on new statistics from the World Health Organization and to understand the effects of spatial
diffusion of free virions at the level of human liver, we began this work by using the standard incidence
function and assuming logistic proliferation for healthy and infected hepatocytes, as proposed and justified
in [9, 16, 24], to formulate a reaction-diffusion mathematical model for HBV infection within a host. Then
we completely analyzed the obtained model, which is an extension of the one proposed in [28], where the
mass-action function was used for viral infection and the proliferation of hepatocytes was ignored. We
started the analysis of the model by relying mainly on Lou and Zhao Lemma and the standard comparison
Theorem to show that potential solutions of the obtained PDE model are bounded. Then, after transforming
the established model to an abstract IBVP problem, we used an approach based on variational method,
semigroup theory, and some other useful tools from functional analysis to prove existence, uniqueness and
positivity of the solution. We computed the basic reproduction number R0 by developing a new method based
on spectral properties of differential operators, the residue Theorem and some arguments from numerical
34 CALVIN TADMON-SEVERIN FOKO

Figure 2. Simulations of IBVP (2.8) using various initial conditions when α = 0.3500, δ = 0.16 (so that
R0 = 1.0444 > 1 but near to 1). All other parameter values are as in Table 3.

analysis. We also constructed a suitable Lyapunov function to prove that there exists a threshold T0 such
that, if R0 ≤ T0 < 1, the HBV-free equilibrium is globally asymptotically stable. When R0 > 1, we used
Routh-Hurwitz criterion to obtain the local asymptotical stability of unique endemic equilibrium. Also,
relying on Center Manifold Theory, we proved that the unique endemic equilibrium is locally asymptotically
stable when R0 is near unity. From there, noticing that R0 has no relation to the diffusion coefficient D, we
inferred that the free diffusion of the virus has no effect on the local stability of such HBV infection problem
with Neumann homogeneous boundary conditions. Finally, we performed numerical simulations to illustrate
theoretical results obtained. It would be interesting to incorporate time delay into the current model. Also,
using a nonstandard difference scheme could be a serious issue. These two challenges will be the concerns of
future investigation.

Acknowledgment

C. Tadmon acknowledges support from the Abdus Salaam International Centre for Theoretical Physics,
where this work was initiated.

References
[1] R. A. Adams, Sobolev spaces. Academic Press, New York (1975).
[2] H. Amann, Dynamic theory of quasilinear parabolic equations-I. Abstract evolution equations, Nonlinear Analysis-Theory,
Methods and Applications 12 (1988), 895-919.
[3] H. Amann, Dynamic theory of quasilinear parabolic equations-II. Reaction-diffusion systems, Differential Integral Equations
3 (1990), 13-75.
[4] H. Amann, Dynamic theory of quasilinear parabolic systems-III. Global existence, Math. Z. 202 (1989), 219-250.
[5] J. M. Bony, I. Gallagher, Analyse de Fourier et theorie spectrale, Cours de l’ecole Polytechnique, (2012).
[6] N. F. Britton, Essential Mathematical Biology, Springer-Verlag, London, (2003).
[7] R. G. Casten, C. J. Holland, Stability properties of solutions to systems of reaction-diffusion equations, SIAM J. Appl.
Math. 33 (1977), 353-364.
[8] C. Castillo-Chavez, B. Song, Dynamical models of tuberculosis and their applications. Math. Biosci. Eng. 1 (2004), 361-404.
35

[9] S. M. Ciupe, R. M. Ribeiro, P. W. Nelson, A. S. Perelson, Modeling the mechanisms of acute hepatitis B virus infection, J.
Theor. Biol. 247 (2007), 23-35.
[10] R. Dautray, J. L. Lions, Analyse Mathematique et Calcul Numerique Pour les Sciences et les Techniques, Semi-Groupe
Variationnel, Masson, Paris, 8 (1988).
[11] W. J. Edmunds, G. F. Medley, D. J. Nokes, Evaluating the cost-effectiveness of vaccination programmes: a dynamic
perspective. Statistics in medicine 18 (1999), 3263-3282.
[12] G .A. Funk, V. A. A. Jansen, S. Bonhoeffer, T. Killingback, Spatial models of virus-immune dynamics, J. Theor. Biol. 233
(2005), 221-236.
[13] C. Goudjo, B. Leye, M. Sy, Weak Solution to a Parabolic Nonlinear System Arising in Biological Dynamic in the Soil. 2011
(2011), 1-25.
[14] S. A. Gourley, J. W. H. So, Dynamics of a food-limited population model incorporating nonlocal delays on a finite domain,
J. Math. Biol. 44 (2002), 49-78.
[15] Henry, Daniel, Geometric Theory of Semilinear Parabolic Equations, 1st edition, Springer-Verlag, (1981).
[16] S. Hews, S. Einkenberry, J. D. Nagy, Y. Kuang, Rich dynamics of Hepatitis B viral infection model with logistic hepatocyte
growth, J. Math. Biol. 60 (2010), 573-590.
[17] F. B. Hollinger, T. J. Liang, Hepatitis B virus, in: D. M. Knipe, P. M. Howley (Eds.), Fields Virology, Lippincott Williams
and Wilkins, Philadelphia, PA, 2 (2001), 29-71.
[18] S. L. Hollis, R. H. Martin, M. Pierre, Global existence and boundedness in reaction-diffusion systems, Siam J. Math. Anal.
18 (1987), 744-761.
[19] Y. Lou, X. Q. Zhao, A reaction-diffusion malaria model with incubation period in the vector population. J Math Biol 62
(2011), 543-568.
[20] Nowak M. A et al., Viral dynamics in hepatitis B virus infection, Proc. Natl. Acad. Sci. USA 93 (1996), 4398-4402.
[21] C. Seeger, W. Mason, Hepatitis B virus biology. Microbiol. Mol. Biol. Rev. 64 (2000), 51-68.
[22] H. L. Smith, Monotone dynamical systems: an introduction to the theory of competitive and cooperative systems. Providence
(RI): American Mathematical Society Providence, 41 (1995).
[23] X. Song, A. U. Neumann, Global stability and periodic solution of the viral dynamics. J. Math. Anal. Appl. 329 (2007),
281-297.
[24] A. Tridane, K. Hattaf, R. Yafia, F. A. Rihan, Mathematical modeling of HBV with antiviral therapy for the immunocom-
promised patients, Commun. Math. Biol. Neurosci. 2016 (2016), 20.
[25] F. B. Wang, Y. Huang, X. Zou, Global dynamics of a PDE in-host viral model, Applicable Analysis: An International
Journal 93 (2014), 2312-2329.
[26] L. Wang, M.Y. Li, Mathematical analysis of the global dynamics of a model for HIV infection of CD4+ T cells. Math.
Biosci. 200 (2006), 44-57.
[27] X. Wang, J. Wang, Analysis of cholera epidemics with bacterial growth and spatial movement, Journal of Biological
Dynamics 9 (2015), 233-261.
[28] K. Wang, W. Wang, Propagation of HBV with spatial dependence, Math. Biosci. 210 (2007), 78-95.
[29] K. Wang, W. Wang, S. Song, Dynamics of an HBV model with diffusion and delay. J. Theor. Biol. 253 ( 2008), 36–44.
[30] W. Wang, X. Q. Zhao, Basic reproduction numbers for reaction-diffusion epidemic models, SIAM J. Appl. Dyn. Syst. 11
(2012), 1652-1673.
[31] WHO, Global hepatitis report, 2017, ISBN: 978-92-4-156545-5
[32] J. Wu, Theory and Applications of Partial Functional Differential Equations. Springer, New York, (1996).

Você também pode gostar