JID:YJDEQ AID:8522 /FLA [m1+; v1.236; Prn:26/09/2016; 15:27] P.

1 (1-11)
Available online at www.sciencedirect.com

ScienceDirect
J. Differential Equations ••• (••••) •••–•••
www.elsevier.com/locate/jde

Existence of traveling waves with the critical speed

for a discrete diffusive epidemic model ✩
Chin-Chin Wu
Department of Applied Mathematics, National Chung Hsing University, 145 Xingda Road, Taichung 402, Taiwan
Received 9 June 2016

Abstract
We study the traveling wave solutions for a discrete diffusive epidemic model of classical Kermack–
McKendrik type. The existence of traveling waves with super-critical speeds are well-known. By a delicate
analysis of traveling waves with super-critical speeds, we derive the existence of traveling waves with the
critical speed.
© 2016 Elsevier Inc. All rights reserved.

MSC: primary 34K10, 34C37; secondary 92D25, 46N60

Keywords: Epidemic model; Lattice dynamical system; Traveling wave

1. Introduction

In this paper, we consider the following lattice dynamical system

⎧
⎪ dSj
⎨ S˙j :=
⎪ = d(Sj +1 − 2Sj + Sj −1 ) − βSj Ij , j ∈ Z,
dt
(1.1)
⎪
⎪ dI
⎩ I˙j := j = (Ij +1 − 2Ij + Ij −1 ) + (βSj − γ )Ij , j ∈ Z,
dt

✩
This work was partially supported by the Ministry of Science and Technology of Taiwan under the grant
104-2115-M-005-002.
E-mail address: chin@email.nchu.edu.tw.

http://dx.doi.org/10.1016/j.jde.2016.09.022
0022-0396/© 2016 Elsevier Inc. All rights reserved.
JID:YJDEQ AID:8522 /FLA [m1+; v1.236; Prn:26/09/2016; 15:27] P.2 (1-11)
2 C.-C. Wu / J. Differential Equations ••• (••••) •••–•••

where d, β, γ are positive constants. System (1.1) is the well-known classical Kermack–
McKendrik epidemic model ([7]) describing an infectious disease outbreak in a closed population
in which the environment is of one-dimensional and is divided into countably discrete niches.
Here Sj (t) is the susceptible population and Ij (t) is the infected population at niches j at time t ,
respectively, β is the transmission coefficient and γ is the recovery/remove rate per unit time.
We also assume the migration coefficients of susceptible and infective populations are d and 1,
respectively.
To see whether a disease can propagate spatially with a constant speed, one usually look for
the so-called traveling wave solutions defined as follows.
For a given s ∗ > 0, by a traveling wave solution of system (1.1), we mean a solution of system
(1.1) in the form

(Sj (t), Ij (t)) = (S(ξ ), I (ξ )), ξ = j + ct,

such that 0 < S < s ∗ and I > 0 in R, S(−∞) = s ∗ and I (±∞) = 0. Here the constant c (the
wave speed) and the function (S, I ) (the wave profile) are to be determined. It is easy to see that
(c, S, I ) satisfies the system

cS  = dD[S] − βSI in R, (1.2)

cI  = D[I ] + (βS − γ )I in R, (1.3)

where D[φ](ξ ) := φ(ξ + 1) + φ(ξ − 1) − 2φ(ξ ). Note that we only require the boundary condi-
tions

(S, I )(−∞) = (s ∗ , 0), I (+∞) = 0,

and leave the value of S at +∞ to be free.

There are tremendous works devoted to the study of traveling waves for different variants of
Kermack–McKendrik model and other epidemic models (see the nice survey paper [6]). We also
refer the reader to, for example, [4,11,10,12,9,8] for the studies of continuous reaction–diffusion
models and [5,3] for the discrete models. One should notice that the system (1.2)–(1.3) is nonlo-
cal and non-monotone.
For a given s ∗ > γ /β, we let

(eλ + e−λ − 2) + (βs ∗ − γ )

c∗ := inf . (1.4)
λ>0 λ

It was proved in [5] that, under certain restrictions on s ∗ , there exists a traveling wave solution of
(1.1) for any speed c > c∗ . Also, the non-existence of traveling waves for c < c∗ when s ∗ > γ /β
and any c > 0 when 0 < s ∗ ≤ γ /β was also derived. The question for the existence of traveling
waves for c = c∗ was left open in [5].
The main purpose of this work is to derive the existence of traveling wave solutions for the
critical speed c∗ . Moreover, we shall remove the restrictions on s ∗ for the existence of traveling
waves for c > c∗ . Our method is motivated by the recent work [3] in which the following lattice
dynamical system for the classical endemic model was studied
JID:YJDEQ AID:8522 /FLA [m1+; v1.236; Prn:26/09/2016; 15:27] P.3 (1-11)
C.-C. Wu / J. Differential Equations ••• (••••) •••–••• 3

⎧
⎪ dsn
⎪
⎨ dt = (sn+1 + sn−1 − 2sn ) + μ − μ sn − β sn in , n ∈ Z,
⎪
⎪ din
⎩ = d(in+1 + in−1 − 2in ) − μ in + β sn in − γ in , n ∈ Z,
dt
where the traveling waves are of front type for each component. However, our traveling wave
here is a mixed of front type (S-component) and pulse type (I -component). We now state our
main theorem of this paper as follows.

Theorem 1. Given s ∗ > γ /β and let c∗ be defined by (1.4). Then (1.1) has a traveling wave
solution with speed c for any c ≥ c∗ .

In order to derive the existence of traveling wave with critical speed, it is very natural to con-
sider a decreasing sequence of super-critical speeds and pass to the limit. However, the limit may
be trivial. We are able to overcome this difficulty by some suitable shifts with the helps of [2,
Remark 3.1] and [1, Theorem 4]. The main idea for deriving the non-triviality here is very differ-
ent from the one given in [3] due to the pulse type of I -component. Finally, the derivation of the
left-hand boundary condition for the S-component relies on the observation of the integrability
of the product SI over R.
The rest of this paper is organized as follows. In §2, we shall provide some lemmas which are
to be used in the next section. As a by-product, we also remove the restriction on s ∗ (see, e.g.,
(3.3) in [5]) for the existence of traveling waves with super-critical speeds. Finally, we derive the
existence of traveling waves with the critical speed in §3.

2. Preliminaries

In this section, let (c, S, I ) be a traveling wave solution of (1.1) with c > 0 and we shall prove
some lemmas to be used in the next section.
The following lemma can be proved as that in [2] (see also [3]). We provide a proof here for
reader’s convenience.

Lemma 2.1. The functions

I  (ξ ) I (ξ ± 1)
,
I (ξ ) I (ξ )

are bounded in R.

Proof. First, dividing (1.3) by I , we have

I  (ξ ) I (ξ + 1) I (ξ − 1)
c = + − 2 + βS(ξ ) − γ .
I (ξ ) I (ξ ) I (ξ )

Then the boundedness of the ratio I  /I over R follows from the boundedness of I (· ± 1)/I (·)
over R, since 0 < S < s ∗ .
To see that I (· − 1)/I (·) is bounded over R, we observe from (1.3) that

cI  (y) + (2 + γ )I (y) = I (y + 1) + I (y − 1) + βS(y)I (y) > 0 for all y ∈ R.

JID:YJDEQ AID:8522 /FLA [m1+; v1.236; Prn:26/09/2016; 15:27] P.4 (1-11)
4 C.-C. Wu / J. Differential Equations ••• (••••) •••–•••

It follows that eμy I (y) is strictly increasing and so

I (y − s) < I (y)eμs , s > 0, y ∈ R,

where μ := (2 + γ )/c. Hence I (· − 1)/I (·) is bounded over R.

For the boundedness of I (· + 1)/I (·), (1.3) implies

1
[eμy I (y)] > eμy I (y + 1) for all y ∈ R. (2.1)
c
By integrating the inequality (2.1) over [ξ, ξ + 1], using the fact that eμy I (y) is strictly increas-
ing, we obtain

1
I (ξ + 1) ≥ [I (ξ ) + I (ξ + 1)]e−μ . (2.2)
c
Plugging (2.2) into (2.1), we have

1 1 e−2μ
[eμy I (y)] > eμy [I (y) + I (y + 1)]e−μ > 2 eμ(y+1) I (y + 1). (2.3)
c c c
Now, integrating (2.3) over [ξ − 1/2, ξ ] gives

ξ
μ(ξ −1/2) e−2μ
e I (ξ ) ≥ e
μξ
I (ξ − 1/2) + 2 eμ(y+1) I (y + 1)dy
c
ξ −1/2

ξ
e−2μ
≥ 2 eμ(ξ +1/2) I (ξ + 1/2)dy
c
ξ −1/2

e−2μ
= eμ(ξ +1/2) I (ξ + 1/2)
2c2
for all ξ ∈ R, where the fact that eμy I (y) is strictly increasing is used. It follows that

I (ξ + 1/2)
≤ 2c2 e3μ/2 for all ξ ∈ R.
I (ξ )

Therefore, the function

I (ξ + 1) I (ξ + 1) I (ξ + 1/2)
=
I (ξ ) I (ξ + 1/2) I (ξ )

is bounded in R and the lemma is proved. 2

Remark 2.1. As we can easily see from the proof of Lemma 2.1, the sequence {Ik /Ik } is uni-
formly bounded in R for a family of traveling wave solutions {(ck , Sk , Ik )} of (1.1), if the wave
speeds {ck } lies in a compact sub-interval of (0, ∞).
JID:YJDEQ AID:8522 /FLA [m1+; v1.236; Prn:26/09/2016; 15:27] P.5 (1-11)
C.-C. Wu / J. Differential Equations ••• (••••) •••–••• 5

To prove the boundedness of I , we prepare the following two lemmas.

Lemma 2.2. Let {(ck , Sk , Ik )} be a sequence of traveling wave solutions of (1.1) with speeds {ck }
lies in a compact sub-interval of (0, ∞). If there is a sequence {yk } such that Ik (yk ) → ∞ as
k → ∞, then Sk (yk ) → 0 as k → ∞.

Proof. For contradiction, we suppose that there is a subsequence of {yk } (still denoted by {yk })
such that Sk (yk ) ≥  for all k ∈ N for some positive constant . By (1.2), Sk ≤ 2ds ∗ /ĉ in R, where
ĉ is a positive lower bound of {ck }. Then there is a positive constant δ such that Sk (ξ ) ≥ /2 for
all ξ ∈ [yk − δ, yk ] for all k ∈ N.
On the other hand, due to Remark 2.1, there is a positive constant K such that |Ik /Ik | ≤ K in
R for all k. It follows that
⎧y ⎫
Ik (yk ) ⎨ k I  (ξ ) ⎬
= exp k
dξ ≤ eδK , ∀ y ∈ [yk − δ, yk ], ∀ k.
Ik (y) ⎩ Ik (ξ ) ⎭
y

Hence

mk := min Ik (y) ≥ e−δK Ik (yk ) → ∞ as k → ∞.

y∈[yk −δ,yk ]

Then, by (1.2),

d 
max Sk (y) ≤ (2s ∗ ) − mk → −∞ as k → ∞,
y∈[yk −δ,yk ] ĉ 2

a contradiction to the fact that Sk < s ∗ in R. This proves the lemma. 2

Lemma 2.3. If lim supξ →∞ I (ξ ) = ∞, then limξ →∞ I (ξ ) = ∞.

Proof. We argue by contradiction and assume that m := lim infξ →∞ I (ξ ) < ∞. Then there is a
sequence {xk } tending to ∞ such that I (xk ) → m as k → ∞. Without loss of generality, we may
assume that I (xk ) ≤ m + 1 for all k ∈ N.
For each k ∈ N, we choose a point yk ∈ [xk , xk+1 ] such that I (yk ) = maxx∈[xk ,xk+1 ] I (x).
Then I (yk ) → ∞ as k → ∞, due to the assumption lim supξ →∞ I (ξ ) = ∞. Let K be a bound
of I  /I over R given in the proof of Lemma 2.1. Without loss of generality, we may assume that
I (yk ) > (m + 1)eK for all k ∈ N. Then, for each k, it follows from
⎧y ⎫
I (yk ) ⎨ k I  (ξ ) ⎬
= exp dξ ≤ eK , if |y − yk | ≤ 1,
I (y) ⎩ I (ξ ) ⎭
y

that [yk − 1, yk + 1] ⊂ (xk , xk+1 ). Hence I  (yk ) = 0 and D[I ](yk ) ≤ 0 for all k ∈ N.
Now, by (1.3), we have

[βS(yk ) − γ ]I (yk ) ≥ 0, ∀ k ∈ N.
JID:YJDEQ AID:8522 /FLA [m1+; v1.236; Prn:26/09/2016; 15:27] P.6 (1-11)
6 C.-C. Wu / J. Differential Equations ••• (••••) •••–•••

This contradicts Lemma 2.2, since I (yk ) → ∞ and S(yk ) → 0 as k → ∞. Therefore, the lemma
is proved. 2

Now, we are ready to prove the boundedness of I as follows.

Lemma 2.4. Let (c, S, I ) be a traveling wave solution of (1.1) with c > c∗ . Then I is bounded
in R.

Proof. Assume for contradiction that lim supξ →∞ I (ξ ) = ∞. Then, by Lemma 2.3, I (∞) = ∞.
Moreover, it follows from Lemma 2.2 that S(∞) = 0.
Next, dividing (1.3) by I and setting z(x) := I  (x)/I (x), z satisfies

x+1
x
cz(x) = e x z(s)ds
+ e− x−1 z(s)ds
− 2 + βS(x) − γ , x ∈ R.

It follows from a fundamental theorem of [2,1] (cf. [1, Theorem 4]) that the limit ω :=
limξ →∞ I  (ξ )/I (ξ ) exists and ω satisfies

cω = eω + e−ω − 2 − γ . (2.4)

Note that ω ≥ 0, since I (∞) = ∞. Hence ω is the unique positive root of (2.4).
Now, recalling from [5] that I (ξ ) ≤ eλ1 ξ for all ξ ∈ R, where λ1 is the smaller positive root
of

cλ = eλ + e−λ − 2 + βs ∗ − γ . (2.5)

Call the larger root of (2.5) by λ2 . Then it is easy to see that λ1 < λ2 < ω, due to βs ∗ > 0.
Furthermore, due to z(∞) = ω, there is a positive constant C such that

I (ξ ) ≥ Ce(ω+λ2 )ξ/2 for all ξ  1.

This is a contradiction to I (ξ ) ≤ eλ1 ξ for all ξ ∈ R. Hence the lemma is proved. 2

Finally, we give the following more general result than [5, Lemma 3.4]. The proof is similar
to that of [5, Lemma 3.4]. For reader’s convenience, we give a proof here.

Lemma 2.5. Let (c, S, I ) be a solution of (1.2)–(1.3) with c > 0, I ≥ 0 and 0 ≤ S ≤ s ∗ . If I is

bounded in R, then I (±∞) = 0.

Proof. First, we recall from (3.2) of [5] that SI is integrable over R, i.e.,

∞
(SI )(x)dx < ∞, (2.6)
−∞

using the boundedness of S. Indeed, this can be seen by integrating (1.2) from y to ξ > y and
letting y → −∞ and ξ → ∞.
Next, by integrating (1.3) from y to ξ > y, we have
JID:YJDEQ AID:8522 /FLA [m1+; v1.236; Prn:26/09/2016; 15:27] P.7 (1-11)
C.-C. Wu / J. Differential Equations ••• (••••) •••–••• 7

ξ +1 ξ y 
y+1

cI (ξ ) − cI (y) = I (x)dx − I (x)dx + I (x)dx − I (x)dx

ξ ξ −1 y−1 y

ξ ξ
+β S(x)I (x)dx − γ I (x)dx.
y y

Using (2.6) and setting K := supR I < ∞ (due to Lemma 2.4), it follows that

ξ ∞
γ I (x)dx ≤ (2 + c)K + β S(x)I (x)dx.
y −∞

Hence I is integrable over R, i.e.,

∞
I (x)dx < ∞. (2.7)
−∞

Now, if lim supξ →−∞ I (ξ ) > 0, then there are constant σ > 0 and a sequence {yk } tending to
−∞ such that I (yk ) ≥ σ for all k ∈ N. Since I  is bounded in R, there is a positive constant δ such
that I (y) ≥ σ/2 for all y ∈ [yk − δ, yk ] for all k ∈ N. This contradicts (2.7). Hence I (−∞) = 0.
Similarly, we have I (∞) = 0. Therefore, the lemma is proved. 2

Therefore, when s ∗ > γ /β, the existence of traveling wave solutions of (1.1) for any c > c∗
is proved. The extra restriction on s ∗ , e.g. (3.3) in [5], is not needed.

3. The existence of traveling waves for c = c∗

This section is devoted to the proof of the existence of traveling wave solution of (1.1) with
c = c∗ . To derive the existence of traveling wave with c = c∗ , it is natural to consider a strictly
decreasing sequence {ck } of wave speeds such that ck ↓ c∗ . Without loss of generality, we may
assume that ck ∈ (c∗ , c∗ + 1] for all k. For each k ∈ N, let (ck , Sk , Ik ) be a traveling wave solution
of (1.1). Then we have

Lemma 3.1. The sequence {Ik } is uniformly bounded in R.

Proof. For contradiction, we suppose that there is a sequence {yk } such that Ik (yk ) → ∞ as
k → ∞. Without loss of generality, due to Ik (±∞) = 0 and Ik > 0, we may assume that Ik (yk ) =
maxy∈R Ik (y) for all k ∈ N. Then, by Lemma 2.2, Sk (yk ) → 0 as k → ∞.
On the other hand, using (1.3), we have

0 = ck Ik (yk ) = D[Ik ](yk ) + [βSk (yk ) − γ ]Ik (yk ) ≤ [βSk (yk ) − γ ]Ik (yk ).

This implies that Sk (yk ) ≥ γ /β for all k, a contradiction. The lemma is proved. 2
JID:YJDEQ AID:8522 /FLA [m1+; v1.236; Prn:26/09/2016; 15:27] P.8 (1-11)
8 C.-C. Wu / J. Differential Equations ••• (••••) •••–•••

To obtain a non-trivial limit, we need to make a suitable shift for each k.

Recall that Sk (−∞) = s ∗ , Ik (±∞) = 0 and Ik > 0 in R. Integrating (1.3) for (ck , Sk , Ik )
over R, we obtain that

∞ ∞ ∞
0 = ck Ik (x)dx = D[Ik ](x)dx + [βSk (x) − γ ]Ik (x)dx
−∞ −∞ −∞
∞
= [βSk (x) − γ ]Ik (x)dx.
−∞

This implies that infx∈R Sk (x) < γ /β for all k. Hence for each k there is a point ξk such that
Sk (ξk ) = γ /β. By shifting ξk to the origin, we may assume without loss of generality that Sk (0) =
γ /β for each k. Note that 0 < Sk < s ∗ and {Ik } is uniformly bounded in R, due to Lemma 3.1.
Hence all of their derivatives are uniformly bounded in R. It follows from Arzela–Ascoli theorem
that, up to extracting a subsequence, Sk → S∗ and Ik → I∗ as k → ∞ in Cloc 1 (R) for some

nonnegative C 1 functions S∗ , I∗ defined in R. Moreover, (S∗ , I∗ ) satisfies (1.2) and (1.3) with
c = c∗ , i.e.,

c∗ S∗ = dD[S∗ ] − βS∗ I∗ in R, (3.1)

c∗ I∗ = D[I∗ ] + (βS∗ − γ )I∗ in R, (3.2)

such that S∗ (0) = γ /β, 0 ≤ S∗ ≤ s ∗ in R and 0 ≤ I∗ ≤ K in R for some positive constant K.

Note that, by Lemma 2.5, I∗ (±∞) = 0.
We claim that S∗ > 0 in R. For contradiction, suppose that S∗ (y) = 0 for some y ∈ R. Then
(S∗ ) (y) = 0 and, by (3.1), S∗ (y ± 1) = 0. An induction gives that S∗ (y + l) = 0 for all l ∈ Z. On
the other hand, by (3.1), it follows from I ∗ ≤ K that

c∗ S∗ (x) + (2d + γ K)S∗ (x) ≥ d[S∗ (x + 1) + S∗ (x − 1)] ≥ 0.

∗
Hence the function S∗ (x)e(2d+γ K)x/c is non-decreasing. This implies that S∗ ≡ 0 in R, a con-
tradiction to S∗ (0) = γ /β. Therefore, we conclude that S∗ > 0 in R.
Now, for each k ∈ N, due to 0 < Sk < s ∗ and Sk (−∞) = s ∗ > γ /β, we may choose ηk < 0
such that

Sk (x) > (s ∗ + γ /β)/2, ∀ x < ηk , Sk (ηk ) = (s ∗ + γ /β)/2. (3.3)

Then we have

Lemma 3.2. Suppose that maxR Ik → 0 as k → ∞. Then {ηk } is bounded below.

Proof. To show that {ηk } is bounded below, we argue it by contradiction. We may assume with-
out loss of generality that ηk → −∞ as k → ∞.
First, we introduce

Ik (x + ηk )
φk (x) := Sk (x + ηk ), ψk (x) := , x ∈ R.
Ik (ηk )
JID:YJDEQ AID:8522 /FLA [m1+; v1.236; Prn:26/09/2016; 15:27] P.9 (1-11)
C.-C. Wu / J. Differential Equations ••• (••••) •••–••• 9

Due to 0 < φk < s ∗ in R for all k and Remark 2.1, up to extracting a subsequence, we may
assume that φk → φ∞ and ψk → ψ∞ as k → ∞ locally uniformly in R for some C 1 functions
φ∞ , ψ∞ . Since (φk , ψk ) satisfies

ck φk = dD[φk ] − βφk ψk Ik (ηk ),

ck ψk = D[ψk ] + (βφk − γ )ψk ,

by letting k → ∞ and using the assumption maxR Ik → 0 as k → ∞, (φ∞ , ψ∞ ) satisfies

c∗ φ∞

= dD[φ∞ ], (3.4)
c ∗ ψ∞

= D[ψ∞ ] + (βφ∞ − γ )ψ∞ . (3.5)

It follows from (3.4), [2, Remark 3.1] and the boundedness of φ∞ that φ∞ ≡ C for some
constant C ∈ [0, s ∗ ]. Hence C = φ∞ (0) = (s ∗ + γ /β)/2. On the other hand, by (3.5) and [2,
Remark 3.1], we have

ψ∞ (x) = C1 eω1 x + C2 eω2 x , x ∈ R, (3.6)

for some constants C1 , C2 , where ω1 < ω2 , are two positive roots of

c∗ ω = (eω + e−ω − 2) + (βs ∗ − γ )/2.

Now, due to Sk (−∞) = s ∗ , it follows from (1.3) and [1, Theorem 4] that the limit

Ik (x)
λk := lim
x→−∞ Ik (x)

exists and λk is one of the positive roots {λ± } of

ck λ = (eλ + e−λ − 2) + (βs ∗ − γ ).

Note that λ± → λ∗ as k → ∞, where λ∗ is the unique positive root of

c∗ λ = (eλ + e−λ − 2) + (βs ∗ − γ ).

Hence for each x ∈ R it follows from ηk → −∞ as k → ∞ that

⎧ x+η ⎫
I (x + ηk ) ⎨  k I  (x) ⎬ ∗
ψk (x) = = exp k
dx → eλ x
Ik (ηk ) ⎩ Ik (x) ⎭
ηk

∗
as k → ∞. This gives that ψ∞ (x) = eλ x for all x ∈ R, a contradiction to (3.6), since ω1 < λ∗ <
ω2 . Therefore, we conclude that {ηk } is bounded below and the lemma is proved. 2
JID:YJDEQ AID:8522 /FLA [m1+; v1.236; Prn:26/09/2016; 15:27] P.10 (1-11)
10 C.-C. Wu / J. Differential Equations ••• (••••) •••–•••

With Lemma 3.2, we now claim that I∗ > 0 in R. Suppose that I∗ (y) = 0 for some y ∈ R.
Then the same argument as above leads that I∗ ≡ 0 in R. Therefore, either I∗ ≡ 0 in R or I∗ > 0
in R. Suppose that I∗ ≡ 0 in R. Then (3.1) is reduced to

c∗ S∗ (x) = dD[S∗ ](x), ∀x ∈ R.

It follows from [2, Remark 3.1] that S∗ (x) = C1 + C2 eωx for all x ∈ R for some constants C1 , C2 ,
where ω is the unique positive root of

c∗ ω = d(eω + e−ω − 2).

Since S∗ is bounded in R, we must have C2 = 0 and so S∗ ≡ S∗ (0) = γ /β in R. However,

since I∗ ≡ 0, Lemma 3.2 can be applied to obtain that ηk ≥ −M for some positive constant M.
It follows from (3.3) that Sk (−M) ≥ (s ∗ + γ /β)/2 for all k. Hence S∗ (−M) ≥ (s ∗ + γ /β)/2,
a contradiction. We conclude that I∗ > 0 in R.
For S∗ < s ∗ in R, we argue by a contradiction. Suppose that S∗ (y) = s ∗ for some y ∈ R.
Then S∗ (y) = 0 and D[S∗ ](y) ≤ 0. This leads to a contradiction, by using (3.1) and the fact that
I∗ > 0. Hence we also have S∗ < s ∗ in R.
Finally, we need to prove S∗ (−∞) = s ∗ . To show S∗ (−∞) = s ∗ , it suffices to prove that m :=
lim infx→−∞ S∗ (x) = s ∗ . For contradiction, we suppose that m < s ∗ . Then there is a sequence
{yk } tending to −∞ such that S∗ (yk ) → m as k → ∞. Consider the functions

φk (x) := S∗ (x + yk ), ψk (x) := I∗ (x + yk ), x ∈ R.

Then ψk (x) → 0 as k → ∞ locally uniformly in R. Up to extracting a subsequence, we may

assume that φk → φ∞ as k → ∞ locally uniformly in R for some C 1 function φ∞ . Moreover,
φ∞ satisfies c∗ φ∞
 = dD[φ ] in R. Hence, as before, φ ≡ m in R. That is,
∞ ∞

lim S∗ (x + yk ) = m locally uniformly for x ∈ R. (3.7)

k→∞

Integrating (1.2) for (c, S, I ) = (cj , Sj , Ij ) from −∞ to yk (with k fixed), we obtain

⎧ ⎫
⎪
⎨ yk +1 yk ⎪
⎬ yk
∗
cj [Sj (yk ) − s ] = d Sj (x)dx − Sj (x)dx − β Sj (x)Ij (x)dx,
⎪
⎩ ⎪
⎭
yk yk −1 −∞

using Sj (−∞) = s ∗ for all j . Sending j → ∞, we obtain

⎧ ⎫
⎪
⎨ yk +1 yk ⎪
⎬ yk
∗ ∗
c [S∗ (yk ) − s ] = d S∗ (x)dx − S∗ (x)dx − β S∗ (x)I∗ (x)dx. (3.8)
⎪
⎩ ⎪
⎭
yk yk −1 −∞

Here the following integrability property (see (2.6)) is used

JID:YJDEQ AID:8522 /FLA [m1+; v1.236; Prn:26/09/2016; 15:27] P.11 (1-11)
C.-C. Wu / J. Differential Equations ••• (••••) •••–••• 11

∞
(S∗ I∗ )(x)dx < ∞,
−∞

since 0 < S∗ < s ∗ . Now, using (3.7) and letting k → ∞ in (3.8), we obtain that

0 < c∗ (m − s ∗ ) = d(m − m) − 0 = 0,

a contradiction. Therefore, we conclude that S∗ (−∞) = s ∗ and this completes the proof of The-
orem 1.

References

[1] X. Chen, S.-C. Fu, J.-S. Guo, Uniqueness and asymptotics of traveling waves of monostable dynamics on lattices,
SIAM J. Math. Anal. 38 (2006) 233–258.
[2] X. Chen, J.-S. Guo, Uniqueness and existence of travelling waves for discrete quasilinear monostable dynamics,
Math. Ann. 326 (2003) 123–146.
[3] Y.-Y. Chen, J.-S. Guo, F. Hamel, Traveling waves for a lattice dynamical system arising in a diffusive endemic
model, Preprint, arXiv:1603.03390.
[4] O. Diekmann, Thresholds and traveling waves for the geographical spread of infection, J. Math. Biol. 6 (1978)
109–130.
[5] S.-C. Fu, J.-S. Guo, C.-C. Wu, Traveling wave solutions for a discrete diffusive epidemic model, J. Nonlinear
Convex Anal. (2016), in press.
[6] H.W. Hethcote, The mathematics of infectious diseases, SIAM Rev. 42 (2000) 599–653.
[7] M. Kermack, A. McKendrick, Contributions to the mathematical theory of epidemics, Proc. Roy. Soc. A 115 (1927)
700–721.
[8] Y. Li, W.T. Li, G. Lin, Traveling waves of a delayed diffusive SIR epidemic model, Commun. Pure Appl. Anal. 14
(2015) 1001–1022.
[9] W.T. Li, G. Lin, C. Ma, F.Y. Yang, Traveling waves of a nonlocal delayed SIR epidemic model without outbreak
threshold, Discrete Contin. Dyn. Syst. Ser. B 19 (2014) 467–484.
[10] X. Wang, H. Wang, J. Wu, Traveling waves of diffusive predator-prey systems: disease outbreak propagation, Dis-
crete Contin. Dyn. Syst. 32 (2012) 3303–3324.
[11] Z.C. Wang, J. Wu, Travelling waves of a diffusive Kermack–McKendrick epidemic model with non-local delayed
transmission, Proc. R. Soc. Lond. Ser. A Math. Phys. Eng. Sci. 466 (2010) 237–261.
[12] F.Y. Yang, Y. Li, W.T. Li, Z.C. Wang, Traveling waves in a nonlocal dispersal Kermack–McKendrik epidemic
model, Discrete Contin. Dyn. Syst. Ser. B 18 (2013) 1969–1993.