Abstract
This paper is concerned with the spatial propagation of bistable nonlocal dispersal equations in exterior domains. We first obtain the existence and uniqueness of an entire solution which behaves like a planar traveling wave front for large negative time. Then, when the entire solution comes to the interior domain, the profile of the front will be disturbed. However, the disturbance is local in space for finite time, which means the disturbance disappears as its location is far away from the interior domain. Furthermore, we prove that the solution can gradually recover its planar wave profile uniformly in space and continue to propagate in the same direction for large positive time provided that the interior domain is compact and convex. Our work generalizes the local (Laplace) diffusion results obtained by Berestycki et al. (2009) to the nonlocal dispersal setting by using new known Liouville results and Lipschitz continuity of entire solutions due to Li et al. (2010).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
This paper is concerned with the following bistable nonlocal dispersal equation
where \(\Omega ={\mathbb {R}}^N\backslash K\) and K is a compact subset of \({\mathbb {R}}^N\) and the dispersal kernel function J is nonnegative. Throughout the paper, we make the following assumptions.
-
(J)
The kernel function \(J\in C^1({\mathbb {R}}^N)\) is radially symmetric and compactly supported such that
$$\begin{aligned}J(x)\ge 0~\text {for}~x\in {\mathbb {R}}^N,~J(0)>0~\text {and}~\int _{{\mathbb {R}}^N}J(y)dy=1.\end{aligned}$$ -
(F)
\(f\in C^{1,1}([0,1])\) and there exists \(\theta \in (0,1)\) such that
$$\begin{aligned}&f(0)=f(1)=f(\theta )=0,\,\,f(s)<0 \text { in }(0,\theta ),\,\,\,f(s)>0 \text { in }(\theta ,1), \\&\int _0^1f(s)ds>0,\,\,f'(0)<0,\,\,f'(1)<0,\,\,f'(\theta )>0, \end{aligned}$$and
$$\begin{aligned} f'(s)<\inf \limits _{x\in \Omega }\int _{\Omega }J(x-y)dy<1 \end{aligned}$$(1.2)for \(s\in [0,1]\).
It follows from the assumption (F) that there exists \(L_f>0\) such that
On the other hand, we know that \(\inf \limits _{x\in \Omega }\int _{\Omega }J(x-y)dy\ge 1/2\) when K is convex. In this case, we can see that (1.2) is automatically satisfied if \(f'(s)\le 1/2\) for all \(s\in [0,1]\). Moreover, under the assumption (1.2), some nonlocal Liouville type results were established by Brasseur et al. [7]. Besides, note that K is compact, without loss of generality, we may assume that
It is well-known that the classical diffusion problem in exterior domain is established by the seminal works of Berestycki et al. [4] and Bouhours [6]. In order to study how a planar wave front propagates around an obstacle, they considered the following semi-linear parabolic problem
where \(\nu \) denotes the outward unit normal to the smooth exterior domain \(\Omega \). More precisely, they proved how a planar traveling front can eventually recover its profile after disturbed by an obstacle K, leaving the obstacle behind. Bouhours [6] further obtained the robustness for the Liouville type results in [4]. Later on, Guo et al. [21] showed that the global mean speed of the entire solution constructed in [4] is the speed of traveling waves in homogeneous environment. More recently, Guo and Monobe [22] extended the results in [4] to V-shaped front. Hoffman et al. [27] considered a similar problem for two dimensional lattice differential equations with directionally convex obstacles.
The nonlocal dispersal equation has got numerous scholars interested, in view of its extensive use to describe the long range effects of spatial structure in biology, physics and chemistry [1, 2, 11, 16,17,18,19]. Moreover, the problems of nonlocal dispersal equations in exterior domains have attracted much attention recently. In particular, Cortázar et al. [12,13,14] considered the asymptotic behaviors of the solutions to linear equations. Brasseur et al. [7, 8] have established some Liouville results for such nonlocal obstacle problems and found that the stationary solutions of (1.1) converging to 1 as \(|x|\rightarrow +\infty \) is indeed 1 for compact convex obstacle K.
When the obstacle K is empty, there have been many works devoted to the traveling wave solutions and entire solutions for (1.1) and its local dispersal counterpart in recent decades. In particular, the authors of [2, 17, 34, 40] have obtained the monotone traveling wave and its asymptotic behaviors for (1.1) with bistable type nonlinearities f, and Chen [17] showed the uniqueness and stability of the traveling wave. For other types of nonlinear terms in (1.1), one can refer to [11, 18, 19, 29, 36, 37, 39] and references therein. For the results on local dispersal equations, readers can consult [15, 20, 23, 24, 28, 35, 38, 42, 43] and references therein.
If \(\Omega ={\mathbb {R}}^N\) in (1.1), let \(u(x,t)=\phi (x_1+ct)\) and \(z=x_1\), we have
where \(c>0\) and \(J_1(x_1)=\int _{{\mathbb {R}}^{N-1}}J(x_1,y_2,y_3,\cdots ,y_{N})dy\). Then it follows from [2, 34] that there exists a unique real number \(c>0\) such that (1.4) admits a solution \(\phi \). In fact, the solution \(\phi \) is the unique monotone planar traveling wave solution of (1.1). Besides, \(\phi (z)\) satisfies
where \(\alpha _0,~\alpha _1,~\beta _0\) and \(\beta _1\) are some positive constants, \(\lambda \) and \(\mu \) are the positive roots of
and
for some constants \(\gamma _0,~\gamma _1,~\delta _0\) and \(\delta _1>0\). However, if the domain is not the whole space (such as (1.1), (1.3)), there is no classical traveling wave front. Therefore, it is naturally to consider the generalization of traveling fronts. In fact, such extensions have been introduced in [3, 5, 30]. In particular, the transition wave front, as a fully general notion of traveling front, has been widely established in many works [10, 25, 26, 31,32,33, 44, 45]. It is interesting to point out that the entire solution constructed in [4, 27] is indeed a generalized transition front.
In the present paper, we are interested to consider the nonlocal dispersal problem (1.1) in exterior domains. The main ingredient of this paper is to obtain a unique entire solution of (1.1) which behaves as planar wave fronts for large negative and positive time. More precisely, we first prove the existence and uniqueness of the entire solution like a planar wave front for large negative time by sub- and super-solutions method. Moreover, we find that the entire solution also approaches planar wave fronts as x is far away from K. Finally, we shall investigate the procedure how the front goes through K and eventually recovers its shape. Due to the lack of compactness of nonlocal operators, which is necessary to establish the uniqueness and asymptotic behaviors as \(|x|\rightarrow +\infty \) of the entire solution, the methods and techniques adopted here are different from that in [4, 27], and additional difficulties appear when the entire solution is constructed. So motivated by the recent work of Li et al. [29], we establish the Lipschitz continuity in space variable x of entire solutions to the nonlocal problem (1.1) in exterior domains. Then we can discuss the uniqueness and asymptotic behaviors of entire solutions of (1.1). In addition, the appearance of convolution term and interior domain leads to the fact that planar wave fronts are not the solutions of (1.1), which causes much trouble in verifying the sub- and super-solutions when we construct the entire solution. Particularly, nonlocal dispersal equations admit no explicit fundamental solutions as Laplacian dispersal equations. Therefore, the sub- and super-solutions in [4, 27] are not suitable here to study the asymptotic behaviors of such an entire solution for large positive time. Consequently, we have to construct new sub- and super-solutions inspired by [4, 27] to investigate the asymptotic behaviors of the entire solution to (1.1). At last, our results show that the geometric shape of the interior domain affects the propagation of planar wave fronts.
Now we are ready to state the main result of this paper.
Theorem 1.1
Assume that (F) and (J) hold. Let \((\phi ,c)\) be the unique solution of (1.4), and K be a compact subset of \({\mathbb {R}}^N\). Then there exists an entire solution u(x, t) to (1.1) with \(0<u(x,t)<1\) and \(u_t(x,t)>0\) for all \((x,t)\in {{\overline{\Omega }}}\times {\mathbb {R}}\), and satisfying that
and as \(|x|\rightarrow +\infty \) uniformly in \(t\in {\mathbb {R}}\). Moreover, if K is convex, then we have
In particular, the condition (1.7) determines a unique entire solution of (1.1).
Remark 1.2
It follows from Brasseur and Coville [9, Thorem 10] that the entire solution constructed in Theorem 1.1 is a generalized transition almost-planar front with global mean speed c.
Remark 1.3
The techniques and ideas developed in this paper can be modified to treat a much more general case for deformations of K (see [7, Definition 1.2]). Let \(K\subset {\mathbb {R}}^N\) be a compact convex set with non-empty interior and let \(\{K_\epsilon \}_{0<\epsilon \le 1}\) be a family of \(C^{0,\alpha }\) (\(\alpha \in (0,1]\)) deformations of K. Assume that
Then there exists \(\epsilon _0\in (0,1]\) such that the conclusions in Theorem 1.1 also hold true with K replaced by \(K_\epsilon \) for \(\epsilon \in (0,\epsilon _0]\).
In this paper, under assumptions that K is a compact convex set and the nonlocal dispersal kernel is compactly supported and radially symmetric, we establish the existence of an entire solution for the nonlocal dispersal equation (1.1) in the exterior domain \(\Omega \), which behaves like a planar traveling front for large negative and positive time. It is naturally to ask if the kernel function is not compactly supported, whether the entire solution we construct in this paper exists. In addition, we conjecture that when the obstacle K just being a compact subset of \({\mathbb {R}}^N\) is not convex, the entire solution of (1.1) can not recover its shape uniformly in space, but it converges to the nonconstant stationary solution in any bounded subset of \({\mathbb {R}}^N\) containing K. We shall study these in a future work.
This paper is organized as follows. In Sect. 2, we consider the Cauchy problem and establish the comparison principle for (1.1). Then the entire solution is constructed in Sect. 3. In Sect. 4, we study the behaviors of the entire solution far away from K in space for finite time. Section 5 is devoted to discussing the asymptotic behavior of the entire solution as time goes to positive infinity.
2 Preliminaries
2.1 The Nonlocal Cauchy Problem
We first consider the nonlocal Cauchy problem
where \(\Omega \) is a subset of \({\mathbb {R}}^N\). We call u(x, t) a solution of (2.1), if it satisfies
for \(x\in \Omega \) and \(t\ge 0\). Then we have the following theorem.
Theorem 2.1
Suppose that (J) holds and \(f\in C^{1,1}({\mathbb {R}})\). Then, for any \(u_0\in L^1(\Omega )\), there exists a unique solution \(u\in C([0,t_0],L^1(\Omega ))\) to (2.2) for some \(t_0>0\).
Proof
For every \(\omega \in C([0,t_0],L^1(\Omega ))\), we define the norm
and the operator
It is easily seen that
here
which means \({\mathcal {T}}\) maps \(C([0,t_0],L^1(\Omega ))\) into \(C([0,t_0],L^1(\Omega ))\). On the other hand, note that
It follows that
In fact, let \(t_0\) be sufficiently small such that \((2+L)t_0<1\). Then one can obtain that \({\mathcal {T}}\) is a strict contraction mapping in \(C([0,t_0],L^1(\Omega ))\).
To extend the solution to \([0,+\infty )\), we can take \(u(x,t_0)\in L^1(\Omega )\) as the initial datum and further obtain a solution in \([t_0,2t_0]\). Then by iterating this procedure, we get a solution in \([0,+\infty )\). \(\square \)
2.2 Comparison Principle
Theorem 2.2
Suppose that the assumptions of Theorem 1.1 hold and \(u(x,0),v(x,0),u_0(x)\in L^{\infty }(\Omega )\). Furthermore, if u(x, t), \(v(x,t)\in C^1([0,+\infty ),L^\infty (\Omega ))\) are uniformly bounded and satisfy
respectively, then
Proof
Define \(W(x,t)=u(x,t)-v(x,t)\), it follows that \(W(x,0)\ge 0\) and
where
Suppose that there exist \(t_*>0\) and \(x_*\in \Omega \) such that \(W(x_*,t_*)<0\). Denote \(\theta _*=-W(x_*,t_*)\), we can take \(\epsilon >0\) and \(K'>0\) such that \(\theta _*=\epsilon e^{2K't_*}\). Let
then we have \(0<T_*\le t_*\) since the facts \(W(x,\cdot )\in C^1(0,\infty )\) and \(W(x,0)\ge 0\). Moreover, it follows that
Without loss of generality, we may assume that \(0\in \Omega \) and \(W(0,T_*)<-\frac{7}{8}\epsilon e^{2K'T_*}\).
Consider now the function
in which \(\beta >0\) is a parameter and \(Z\in L^{\infty }\left( {\mathbb {R}}^N\right) \) with \(Z(0)=1\), \(\lim \limits _{|x|\rightarrow +\infty }Z(x)=3,~1\le Z(x)\le 3\). Take \(\beta _*\in \left( \frac{1}{8},\frac{1}{4}\right] \) as the minimal value of \(\beta \) for which \(W(x,t)\ge W^-(x,t)\) holds for all \((x,t)\in \Omega \times [0,T_*]\). Since
there exist \(x^*\in \Omega \) and \(0<t_0<T_*\) such that \(W(x^*,t_0)=W^-(x^*,t_0,\beta _*)\). The definition of \(\beta _*\) now implies that
In addition, by the fact that \(W(x,t)\ge W^-(x,t,\beta _*)\) for all \((x,t)\in \Omega \times [0,T_*]\), we have
It follows from (2.3) that
In particular, it follows from the assumptions of Theorem 2.2 that u(x, t) and v(x, t) are uniformly bounded and \(f'(v(x,t)+\theta W(x,t))\) is also bounded, which means there is some \({\mathfrak {M}}>0\) such that \(|f'(v(x,t)+\theta W(x,t))|<{\mathfrak {M}}\) and \(|F(x,t)|<{\mathfrak {M}}\) for \(x\in \Omega \) and \(t>0\). Then we obtain
This leads to a contradiction upon choosing \(K'\) to be sufficiently large. \(\square \)
Corollary 2.3
Under the assumptions of Theorem 2.2, let u(x, t) and v(x, t) be solutions of (1.1) with initial values u(x, 0) and v(x, 0), respectively. If \(u(x,0)\ge v(x,0)\) and \(u(x,0)\not \equiv v(x,0)\), then \(u(x,t)>v(x,t)\) for all \(x\in \Omega \) and \(t>0\).
Proof
It is sufficient to show that \(u(x,t)>v(x,t)\) for all \(x\in \Omega \) and \(t\in (0,t_0]\) for some \(t_0>0\). In fact, if \(u(x,t)>v(x,t)\) for all \(x\in \Omega \) and \(t\in (0,t_0]\), we can similarly have \(u(x,t)>v(x,t)\) for all \(x\in \Omega \) and \(t\in [t_0,2t_0]\). Then by repeating this process, we can obtain the results of this lemma. Now, denote
where \(\epsilon ,~p>0\) are real numbers and p is sufficiently large such that
with F(x, t) being defined as that in the proof of Theorem 2.2. Suppose that, by contradiction, there is \((x_*,t_*)\in \Omega \times (0,t_0]\) such that \(w(x_*,t_*)=0\). It then follows that \(\inf _{x\in {\overline{\Omega }},t\in (0,t_0]}{\tilde{w}}(x,t)<0\). Furthermore, one can find a sequence \((x_n,t_n)\) such that \(t_n\rightarrow {\tilde{t}}_*\) and
Observe that
Then we have that
Letting \(n\rightarrow \infty \), it follows that
Choose \(t_0>0\) being sufficiently small such that
which implies that \(\inf _{x\in {\overline{\Omega }},t\in (0,t_0]}{\tilde{w}}(x,t)>\inf _{x\in {\overline{\Omega }},t\in (0,t_0]}{\tilde{w}}(x,t)\), since \(\inf _{x\in {\overline{\Omega }},t\in (0,t_0]}<0\). Thus we have finished the proof. \(\square \)
3 Existence and Uniqueness of the Entire Solution
This section is devoted to establishing the existence and uniqueness of an entire solution to (1.1) which behaves as a planar traveling front until it approaches the interior domain K. Since the profile \(\phi \) in (1.4) is monotone increasing and unique in the translation sense, without loss of generality, we further assume that \(\phi (0)\le \theta \) and \(\phi ''(\xi )\ge 0\) for \(\xi \le 0\). The main result of this section is stated as follows.
Theorem 3.1
Assume that (F) and (J) hold and let \((\phi ,c)\) be the unique solution of (1.4). If \(K\subset \{x\in {\mathbb {R}}^N:~x_1\le 0\}\cap {\mathbb {R}}^N\setminus \)supp(J), then there exists an entire solution U(x, t) of (1.1) satisfying
and
Moreover, condition (3.1) determines a unique entire solution of (1.1).
In this section, the radial symmetry of \(J(\cdot )\) can be released to \(J(x)=J(-x)\). Moreover, the convexity and compactness of the obstacle are not required, while the boundedness of K is necessary. We prove Theorem 3.1 by constructing sub- and super-solutions.
3.1 Construction of the Entire Solution
To establish the entire solution, we shall construct some suitable sub- and super-solutions. Inspired by [4], we define the sub-solution
and the super-solution
here \(\xi (t)\) is the solution of the following equation
where \(M,~\lambda _0\) and T are positive constants to be specified later. A direct calculation yields that
For the function \(\xi (t)\) to be defined, one must have \(1-c^{-1}Me^{\lambda _0 ct}>0\). In addition, we suppose that
Thus set \(T:=\frac{1}{\lambda _0 c}\ln \frac{c}{c+M}\). Moreover, it follows from (1.5) that there exist two positive numbers \(K_\phi \) and \(k_\phi \) such that
Then the following proposition holds.
Proposition 3.2
Assume that \(\lambda _0<\min \{\lambda ,k_\phi \}\) and \(K\subset \{x\in {\mathbb {R}}^N:~x_1\le 0\}\cap {\mathbb {R}}^N\setminus \)supp(J). Then there exists a sufficiently large number \(M>0\) such that \(W^-(x,t)\) and \(W^+(x,t)\) are sub- and super-solutions of (1.1) in the time range \(-\infty <t\le T_1\) for some \(T_1\in (-\infty , T]\).
The proof of this lemma will be given in Appendix for the coherence of this paper.
Now we are in a position to construct the entire solution. Let \(u_n(x,t)\) be the unique solution of (1.1) for \(t\ge -n\) with initial data
Since \(W^-(x,t)\) is a sub-solution, it is not difficult to show that the sequence \(\{u_n(x,t)\}_{n=1}^\infty \) is nondecreasing in n. Choose some constant \(T^*>0\) such that \(c>{\dot{\xi }}(t)\) for \(t\le -\max \{T^*,T_1\}\). In the following discussing, without loss of generality, we assume that \(n^*\ge T^*\). Then, we have
for \(n\ge n^*\), \(t\ge -n\) and \(x\in \Omega \). Since \(\frac{\partial W^-(x,t)}{\partial t}=0\) for \(x_1\le 0\) and
for \(0<x_1\le |ct-\xi (t)|\) and \(t<-T^*\), it follows that
for all \(x_1\le |cn+\xi (-n)|\). Furthermore, by Corollary 2.3, \(u_n(x,t)\) satisfies
and
In particular, since the sequence \(\{u_n(x,t)\}_{n=n^*}^{\infty }\) being uniformly bounded in n satisfies (3.2) with \(f\in C^{1,1}([0,1])\), we have that the sequence \(\left\{ \frac{\partial u_n(x,t)}{\partial t}\right\} _{n=n^*}^{\infty }\) is uniformly bounded. Then it follows that \(\{u_n(x,t)\}_{n=n^*}^{\infty }\) is well-defined for each n and equicontinuous in t. Similarly, \(\left\{ \frac{\partial u_n(x,t)}{\partial t}\right\} _{n=n^*}^{\infty }\) is equicontinuous in t. Therefore, by Arzela-Ascolit theorem, for each fixed \(x\in \Omega \), there exists a subsequence, still denoted by \(\left\{ u_n(x,t),\frac{\partial u_n(x,t)}{\partial t}\right\} _{n=n^*}^{\infty }\) such that
where the convergence is locally uniform in \(t\in {\mathbb {R}}\). Moreover, via diagonalization, take a subsequence of \(\left\{ u_n(x,t),\frac{\partial u_n(x,t)}{\partial t}\right\} _{n=n^*}^{\infty }\), which converges to some function U(x, t). Since that \(u_n(x,t)\) is the solution to (3.3) with initial value \(u_n(x,-n)=W^-(x,-n)\) and that U(x, t) is the limit of \(u_n(x,t)\) as \(n\rightarrow \infty \), we have that U(x, t) is well defined for \(t\in {\mathbb {R}}\). Then it follows from Lebesgue’s dominated convergence theorem that
and
Besides, it follows from the definition of \(W^-(x,t)\) and \(W^+(x,t)\) that
Note that U(x, t) is not a constant, by applying Corollary 2.3 to \(U_t(x,t)\) and U(x, t), one have that
In addition, inspired by [29], we can show that U(x, t) satisfies the following proposition.
Proposition 3.3
Let U(x, t) be the entire solution in Theorem 3.1. Then U(x, t) satisfies
and
with \(M'>0\) and \(M''>0\) being two real numbers.
Proof
Since \(\int _{{\mathbb {R}}^N}J(x)dx=1,~J(x)\ge 0\) and J(x) is compactly supported, we have \(J'\in L^1({\mathbb {R}}^N)\). Furthermore, we get
Let
and v(t) be a solution of the equation
for some \(M\ge 2\sup \limits _{\xi \in {\mathbb {R}}}|\phi '(\xi )|\). In addition, denote \(V(x,t)=u_n(x+\eta ,t)-u_n(x,t)\), where \(u_n(x,t)\) is the solution of (1.1) with initial value \(u_n(x,-n)=W^-(x,-n)\). Then
where \({\overline{V}}\) is between \(u_n(x,t)\) and \(u_n(x+\eta ,t)\). Consequently, V(x, t) satisfies
Moreover, \(|V(x,t)|\le v(t)\le M^*|\eta |\) for any \(x\in \Omega ,~t\ge -n\) and \(M^*=M+\frac{L_1}{m}\). Indeed,
for any \(x\in {\mathbb {R}}^N\) and \(t\ge -n\). In particular, in view of \(f'(s)<1\) for \(s\in [0,1]\), there holds
At last, since \(u_n(x,t)\rightarrow U(x,t)\) locally uniformly in \(t\in {\mathbb {R}}\) as \(n\rightarrow +\infty \), we have
Now take \(M'=M^*+2\), we can show that (3.4) and (3.5) hold by taking \(M''=L_1+2+2M^*\). \(\square \)
3.2 Uniqueness of the Entire Solution
Now, we show the uniqueness of the entire solution constructed in Theorem 3.1. First, the following lemma is valid.
Lemma 3.4
Assume that the settings of Theorem 1.1 hold. Then for any \(\varphi \in (0,\frac{1}{2}]\), there exist constants \(T_\varphi =T_\varphi (\varphi )>1\) and \(K_\varphi =K_\varphi (\varphi )>0\) such that
where
Proof
It is easy to choose \(T_\varphi \) and \(M_\varphi \) such that \(\Omega _{\varphi }(t)\subset \left\{ x\in \Omega :~|x_1+ct|\le M_\varphi \right\} \subset \{x\in {\mathbb {R}}^N:~x_1\ge 1\}\). Now suppose there exist sequences \(t_k\in (-\infty ,-T_\varphi ]\) and \(x^k:=(x^k_1,x^k_2,...x^k_N)\in \Omega _{\varphi }(t)\) such that
There are only two cases that can happen: \(t_k\rightarrow -\infty \) or \(t_k\rightarrow t_*\) for some \(t_*\in (-\infty ,-T_\varphi ]\) as \(k\rightarrow +\infty \). For the former case, denote
By Lemma 3.3, \(\{U_k(x,t)\}_{k=1}^\infty \) is equicontinuous in \(x\in \Omega \) and \(t\in {\mathbb {R}}\). It follows that there exists a subsequence still denoted by \(\{U_k(x,t)\}_{k=1}^\infty \) such that
by Arzela-Ascolit theorem. Furthermore, \(U_*\) satisfies \(\frac{\partial U_*(0,0)}{\partial t}=0\). Applying strong maximum principle theorem to \(\frac{\partial U_*(x,t)}{\partial t}\), We further have
However, this is impossible because
For the second case, \(x^k_1\) remains bounded by the definition of \(\Omega _{\varphi }(t)\). Therefore, we assume that \(x^k_1\rightarrow x^*_1\) as \(k\rightarrow +\infty \) and let
Then, each \(U_k(x,t)\) is defined for all \((x,t)\in (-\infty ,-T_\varphi ]\times \{x\in {\mathbb {R}}^N\mid x_1\ge -1\}\) by the definition of \(\Omega _\varphi (t)\). Similarly, there exists a subsequence, again denoted by \(\{U_k\}^{\infty }_{k=1}\), such that
for some function \(U^*\) satisfying (1.1) on \(\{x\in {\mathbb {R}}^N\mid x_1\ge -1\}\times (-\infty ,-T_\varphi ]\). Note that \(U_t(x,t)>0\), we have
Then we obtain \(\frac{\partial U^*}{\partial t}(x,t)\equiv 0\) for \(t\le t_*\) by strong maximum principle, but this is impossible since
This ends the proof. \(\square \)
Now we are ready to show the uniqueness of the entire solution. Suppose that there exists another entire solution V(x, t) of (1.1) satisfying (3.1). Extend the function f as
and choose \(\eta >0\) being sufficiently small such that
Then for any \(\epsilon \in (0,\eta )\), we can find \(t_0\in {\mathbb {R}}\) such that
For each \(t_0\in (-\infty ,T_\varphi -\sigma \epsilon ]\), define
and
where the constant \(\sigma >0\) is specified later. Then by (3.6),
Next we show that \({\tilde{U}}^-(x,t)\) and \({\tilde{U}}^+(x,t)\) are sub- and super-solutions in \(t\in [0,T_\varphi -t_0-\sigma \epsilon ]\), respectively. Let
A straightforward calculation implies that
where \(0<\theta <1\). For any \(x\not \in \Omega _\eta \left( x,t_0+t+\sigma \epsilon \left( 1-e^{-\omega t}\right) \right) \), we can see
Consequently, \(f'\left( U+\theta \epsilon e^{-\omega t}\right) \le -\omega \), which implies that
For \(x\in \Omega _\eta \left( x,t_0+t+\sigma \epsilon \left( 1-e^{-\omega t}\right) \right) \), by Lemma 3.4, there holds
As a consequence, \({\mathcal {L}}{\tilde{U}}^+(x,t)\ge 0\) provided that \(\sigma \) is a sufficiently large number.
Similarly, we can show \({\mathcal {L}}{\tilde{U}}^-(x,t)\le 0\) in \(\Omega \times [0,T_\varphi -t_0-\sigma \epsilon ]\). In view of this and (3.7), we see that
Letting \(t+t_0\) be replaced by t, the inequality above can be rewritten as
for all \((x,t)\in \Omega \times [t_0,T_\varphi -\sigma \epsilon ]\) and \(t_0\in (-\infty ,T_\varphi -\sigma \epsilon ]\). As \(t_0\rightarrow -\infty \), we obtain that
By the comparison principle, the inequality holds for \(t\in {\mathbb {R}}\) and \(x\in \Omega \). Letting \(\epsilon \rightarrow 0\), since \(\sigma \) is independent of the choice of \(\epsilon \), we have \(V(x,t)\equiv U(x,t)\).
4 Behaviors Far Away from the Interior Domain
In this section, we are going to figure out what the entire solution, constructed in Theorem 3.1, is like far away from the interior domain.
Theorem 4.1
Assume that (F) and (J) hold. Let \((\phi ,c)\) be the unique solution of (1.4) and u(x, t) be a solution of
such that
Then, for any sequence \((x'_n)_{n\in {\mathbb {N}}}\in {\mathbb {R}}^{N-1}\) such that \(|x'_n|\rightarrow +\infty \) as \(n\rightarrow +\infty \), there holds
locally uniformly with respect to \((x,t)=(x_1,x',t)\in {\mathbb {R}}^N\times {\mathbb {R}}\).
Proof
In order to prove this theorem we need to extend the function f as that in Sect. 3, and let \(A>0\) be sufficiently large such that
Denote \(\delta =\min \limits _{[-A,A]}\phi '(\xi )>0\) and take \(T_\phi >0\) such that for any \(\epsilon \in (0,\frac{\eta }{2})\), we have that
Now we are in a position to show this theorem. Under the assumptions of Theorem 4.1, let \((x'_n)_{n\in {\mathbb {N}}}\in {\mathbb {R}}^{N-1}\) be a sequence such that \(|x'_n|\rightarrow +\infty \) as \(n\rightarrow +\infty \). And denote \(u_n(x,t)=u(x_1,x'+x'_n,t)\) for each \(t\in {\mathbb {R}}\) and \(x=(x_1,x')\in \Omega -(0,x'_n)\). Since \(0\le u\le 1\), K is compact and \(\left\{ u_n(x,t),\frac{\partial u_n(x,t)}{\partial t}\right\} _{n=1}^{\infty }\) is equicontinuous in \(x\in \Omega \) and \(t\in {\mathbb {R}}\) by Proposition 3.3, then there exists a subsequence, still denoted by \(\left\{ u_n(x,t),\frac{\partial u_n(x,t)}{\partial t}\right\} _{n=1}^\infty \), such that
locally uniformly in \((x,t)\in {\mathbb {R}}^N\times {\mathbb {R}}\). In addition, for all \((x,t)\in {\mathbb {R}}^N\times {\mathbb {R}}\), we further have that \(0\le {\mathfrak {u}}(x,t)\le 1\) and
since J is compactly supported, K is compact and \(|x'_n|\rightarrow +\infty \) as \(n\rightarrow \infty \). In addition, recall that
uniformly in \(x\in {\overline{\Omega }}\), the function \({\mathfrak {u}}(x,t)\) satisfies
Now define two functions \({\underline{u}}(x,t)\) and \({\overline{u}}(x,t)\) as follows
where
Then the following lemma holds, whose proof is left to Appendix as a regular argument. \(\square \)
Lemma 4.2
The functions \({\underline{u}}(x,t)\) and \({\overline{u}}(x,t)\) are sub- and super-solutions to (4.2) for \(t\ge t_0\), respectively.
By the comparison theorem and let \(\epsilon \rightarrow 0\), we have \({\mathfrak {u}}(x,t)\equiv \phi (x_1+ct)\). Since the limit is uniquely determined, the sequence \(\{u_n(x,t)\}_{n=1}^{\infty }\) converges to \(\phi (x_1+ct)\) locally uniformly in \((x,t)\in {\mathbb {R}}^N\times {\mathbb {R}}\) as \(n\rightarrow +\infty \). Then the proof of Theorem 4.1 is completed. \(\square \)
Theorem 4.3
Suppose that all the assumptions in Theorem 4.1 hold. Then the solution u(x, t) of (4.1) in Theorem 4.1 satisfies
Proof
Extend f as that in Sect. 3. Then define \(f_\delta (u)=f(u-\delta )\), \((c_\delta ,\phi _\delta )\) satisfies
and
where \(c_\delta >0\) and \(\delta >0\) is sufficiently small. Now we are going to show the theorem in three steps.
Step 1. For \(x_1\gg 1\), since
there exists a sufficiently large number \(T^*_1\ge 0\) such that
In particular, for any \(x\in \Omega \), let \(N_1\gg 1\) such that
Since \(\phi '>0,~u_t(x,t)>0\), we have
Step 2. For \(x_1\ll -1\), let \(\delta =\frac{1}{2}\epsilon \). Similarly as the first step, choose \(T_2>0\) and \(N_2\gg 1\) such that
Obviously, there exists \(x_0\in {\mathbb {R}}\) such that \(\phi _\delta (x_0)=1\). Then
Since that \(\phi _\delta (x,t)\) is increasing, we have
Then, by applying the comparison principle (see [41]) for \(t\ge -T_2\) and \(x_1\le -N_2\), one have that
In particular, since \(\phi _\delta (-\infty )=\delta =\frac{1}{2}\epsilon \), for any \(\tau \ge 0\), there exists a \(N_3\gg 1\) such that
which again shows
Step 3. For \(|x'|\gg 1\), it follows from the Theorem 4.1 that one can choose \(N_4>0\) being sufficiently large such that
whenever \(|x'|>N_4\) and \(x_1\in [-N_3,N_1]\). This finishes the proof. \(\square \)
5 The Behavior for the Large Time
In this section we intend to investigate the behavior of the solution constructed in Theorem 3.1 for large positive time. For this goal, we establish the following result.
Theorem 5.1
Suppose that (F) and (J) hold. Let \((\phi ,c)\) be the unique solution of (1.4), \(t_0\in {\mathbb {R}}\) and u(x, t) be a solution of
And assume that, for any \(\epsilon >0\), there is a number \(t_\epsilon \ge t_0\) and a compact set \(K_\epsilon \subset {\overline{\Omega }}\) such that
and
Then
The most important ingredient of the proof is to construct suitable sub- and super-solutions. This process is such cumbersome that will be divided several parts. Enlightened by Hoffman [27], we first construct a function z(t) in the following lemma, which plays an important role in the construction of sub- and super-solutions.
Lemma 5.2
For any \(0<\eta _z<\ln 2\), there are two constants \({\mathcal {I}}={\mathcal {I}}(\eta _z)>0\) and \(K_0=K_0(\eta _z)>0\) such that for any \(t_1\ge 0\), there exists a \(C^1\)-smooth function \({\tilde{z}}(t):[0,+\infty )\rightarrow {\mathbb {R}}\) that satisfies the following properties.
-
(i)
For all \(t\ge 0\), the inequalities \({\tilde{z}}'(t)\ge -\eta _z{\tilde{z}}(t)\) and \(0<{\tilde{z}}(t)\le {\tilde{z}}(0)=1\) hold.
-
(ii)
In addition, \({\tilde{z}}(t)\ge K_0(1+t-t_1)^{-\frac{3}{2}}\) for all \(t\ge t_1\) and \(\int _0^{+\infty }{\tilde{z}}(t)dt<{\mathcal {I}}\).
Proof
First we define
Let \(l_P(\eta _z)=1/\eta _z\). Then, by a direct calculation, it is easy to show that for any fixed \(0<\eta _z<\ln 2\) and \(0<\nu \le \eta _z\), \(P_-(x)\) satisfies that
with
and \(P_+(x)\) satisfies that
\(P_+(l_P)\ge \frac{1}{3}\) with
In addition, denote
Now if \(0<t_1<3\eta _z^{-1}\), then let \({\tilde{z}}(t)=z_1(t)\), otherwise we define the function \({\tilde{z}}(t)\) on five different intervals. In particular, define \(\nu =\frac{-z_1'(t_1-3\eta _z^{-1})}{z_1(t_1-3\eta _z^{-1})}\), which implies \(0<\nu \le \eta _z\). Then, let
It remains to specify \({\tilde{z}}(t)\) in \(\left[ t_1-2\eta _z^{-1}, t_1-\eta _z^{-1}\right] \). This can be done by choosing an arbitrary \(C^1\)-smooth function, under the constraints
and
Then we finish the proof since the properties (i) and (ii) are valid by a direct calculation. \(\square \)
Remark 5.3
It is not difficult to see that \(z(1)\ge \frac{1}{2}\) from \(0<\eta _z<\ln 2\) and the first statement in Lemma 5.2.
Now, we are in the position to construct the sub- and super-solutions.
5.1 Sub-solution
In this part, we construct a sub-solution to (5.1). Now define
where \(\theta (x',t)=\beta t^{-\alpha }e^{\frac{-|x'|^2}{\gamma t}}\) with \(\alpha ,~\gamma >1\) being two real numbers, \(z(t)=\epsilon _1{\tilde{z}}(t)\) with \(\epsilon _1=2\epsilon \) and \(Z(t)=K_z\int _0^tz(\tau )d\tau \) with \(K_z>0\) being a large number. z(t) is defined in Lemma 5.2. It follows from the definition of \(u^-(x,t)\) that
Thanks to \(\phi (-\infty )=0,~\phi '>0\), \(0<u(x,t)<1\) and \(\min \limits _{x\in K_\epsilon }u(x,t)>0\), the last inequality holds provided that \(\beta \) is sufficiently large. If \(x\in {\mathbb {R}}^N\backslash K_\epsilon \), then by \(z(1)\ge \frac{1}{2}\epsilon _1=\epsilon \) (from Remark 5.3), we have that
As a consequence, there holds \(u^-(x,1)\le {\tilde{u}}(x,1)\) for any \(x\in {\overline{\Omega }}\).
Lemma 5.4
The inequality \({\mathcal {L}}u^-(x,t)\le 0\) holds for all \(x\in \Omega \) and \(t\ge 1\), where
Proof
Since \(u^-(x,t)=\phi (\xi (x,t))-z(t)\), where \(\xi (x,t)=x_1+c(t-1+t_\epsilon ) -\beta t^{-\alpha }e^{\frac{-|x'|^2}{t\gamma }}-Z(t)\), we have
and
Denote \({\mathcal {D}}\phi =\int _{{\mathbb {R}}^N}J(x-y)[\phi (\xi (y,t))-\phi (\xi (x,t))]dy\). Then, applying mean value theorem, we get that
where \(0<{\tilde{\theta }}<1\). In particular,
Therefore, we have
Similarly as previous, one can get
It follows that
Now we go further to show \({\mathcal {L}}u^-(x,t)\le 0\) in two cases.
Case 1. We assume that \(|\xi (x,t)|\gg 1\) such that \(\phi (\xi (x,t))\in [0,\eta ]\cup [1-\eta ,1]\), where \(\eta \) is sufficiently small to ensure that \(f'(s)\le -\sigma <0\) for any \(s\in [0,\eta ]\cup [1-\eta ,1]\) and \(\sigma >2\eta _z\) with \(\eta _z\) defined as that in Lemma 5.2. Then, since the function z(t) constructed in Lemma 5.2 satisfies
there holds
Indeed, since \(\alpha >1\) and \(|\xi (x,t)|\) is sufficiently large such that \((\alpha \beta +C^K+C')\phi '(\xi )\le \frac{1}{2}\eta _zK_0\), the last inequality above holds obviously.
Case 2. Let \(\phi (\xi (x,t))\in [\eta ,1-\eta ]\) in this case. Since \(\phi '(\xi )>0\) for all \(\xi \in {\mathbb {R}}\), we may choose \(\tau _0>0\) being sufficiently small such that \(\phi '(\xi )\ge \tau _0>0\). Denote \(\max \limits _{[\eta ,1-\eta ]}f'(s)=\delta ^0>0\). Then
Let \(K_z\) be sufficiently large such that \(-K_z\tau _0+\delta ^0+\eta _z+\left( \alpha \beta +C^K+C'\right) \frac{1}{K_0}\Vert \phi '\Vert _\infty \le -\frac{1}{2}\eta _z\). One hence have that \({\mathcal {L}}u^-(x,t)\le -\frac{1}{2}\eta _zz(t)\). The proof is finished. \(\square \)
5.2 Super-solution
This part is devoted to verifying the super-solution defined as
where
with \(\alpha ^+>1\) and \(\beta ^+>0\) being a large number. It follows from the definition of \(u^+(x,t)\) that
for \(x\in K_\epsilon .\) Since \(\phi (-\infty )=0,~\phi '>0\), \(0<u(x,t)<1\) and \(\max \limits _{x\in K_\epsilon }u(x,t)<1\), the last inequality holds provided that \(\beta ^+\) is sufficiently large. If \(x\in {\mathbb {R}}^N\backslash K_\epsilon \), then by \(z(1)\ge \frac{1}{2}\epsilon _1=\epsilon \) (from Remark 5.3), we have
As a consequence, there holds \(u^+(x,1)\ge {\tilde{u}}(x,1)\) for any \(x\in {\overline{\Omega }}\).
Lemma 5.5
The inequality \({\mathcal {L}}u^+(x,t)\ge 0\) holds for all \(x\in \Omega \) and \(t\ge 1\), where
Proof
It follows from a direct calculation that
and
Note that
we have
Then it follows that
and we obtain
Now we focus on all the integral items above denoted by
By the same progress as the calculation of \(u^-(x,t)\), we have
Therefore,
Next we are going to show \({\mathcal {L}}u^+(x,t)\ge 0\) in two cases.
Case 1. Let \(|\psi (x,t)|\gg 1\) such that \(\phi (\psi (x,t))\in [0,\eta ]\cup [1-\eta ,1]\), where \(\eta >0\) is sufficiently small to ensure that \(f'(s)\le -\sigma <0\) for any \(s\in [0,\eta ]\cup [1-\eta ,1]\). Since
and \(|\psi (x,t)|\gg 1\), we obtain that
which implies
since \(\eta _z<\frac{1}{2}\sigma \) and \(\alpha ^+>1\).
Case 2. Since \(\phi (\psi (x,t))\in [\eta ,1-\eta ]\) we have \(\phi '(\psi (x,t))\ge \tau _0>0\). Denote \(\min \limits _{s\in [0,1]}f'(s)=-\delta '<0.\) Then, we obtain
Recall the fact \(\alpha ^+>1\), it follows that \(K_0t^{-\alpha ^+-\frac{1}{2}}<z(t)\). Moreover, one can take \(K_z>0\) being sufficiently large such that \(\left( K_z-\frac{\alpha ^+\beta ^++M'}{K_0}\right) \tau _0-\delta '-\eta _z\ge 0\). Then we have that \({\mathcal {L}}u^+(x,t)\ge 0\). The proof is finished. \(\square \)
The proof of Theorem 5.1
From the Lemmas 5.4 and 5.5, for \(t\ge t_{\epsilon }\), we have that
and
For any sufficiently small \(\eta _{s}>0\), take \(0<\epsilon _1<\frac{\eta _s}{{\mathcal {I}}\Vert \phi '\Vert _{L^{\infty }}}\). Then by the construction of z(t), we see that
and
Since \(\eta _s\) is arbitrary, we have \(\lim \limits _{t\rightarrow \infty }\sup \limits _{x\in {\bar{\Omega }}}|u(x,t)-\phi (x_1+ct)|=0\). Thus, we finish the proof of Theorem 5.1. \(\square \)
5.3 Proofs of Theorem 1.1
It follows from [7, Theorems 2.4 and 2.6] that under the conditions (F) and (J), the unique solution of the stationary problem of (1.1)
is \(v(x)\equiv 1 ~\text {in}~\overline{{\mathbb {R}}^N \backslash K}(~\text {or}~\overline{{\mathbb {R}}^N \backslash K_{\epsilon }}).\)
Now we are in position to show the main theorem. It is obvious that if the solution U(x, t) of (1.1) constructed in Theorem 3.1 satisfies the conditions in Theorem 5.1 together with Theorem 4.3, then the conclusions in Theorem 1.1 hold. Indeed, since \(U_t(x,t)>0\) and U(x, t) is Lipschitz continuous in \(x\in {\mathbb {R}}^N\backslash K\), we know that as time tends to positive infinity, U(x, t) locally uniformly converges to some continuous function V(x), without loss of generality, which can be deemed as a uniformly continuous function by [7, Lemma 3.2] because of the assumption (F). Furthermore, we claim that V(x) satisfies (5.2). Therefore, we have \(V(x)\equiv 1\) for all \(x\in {\mathbb {R}}^N\backslash K\). In fact, it is sufficient to show \(\sup _{{\mathbb {R}}^N\setminus K}V(x)=1\). In view of \(u(x,t)-\phi (x_1+ct)\rightarrow 0\) as \(t\rightarrow -\infty \) in Theorem 3.1, for any small \(\epsilon '>0\), there exist \(t_{\epsilon '}<0\) and \(X_1>0\) being sufficiently large such that
and
Thus
which implies that \(V(x)\ge 1-\epsilon '\) due to \(u_t(x,t)>0\). Since that \(\epsilon '\) is actually arbitrary, one has that
Therefore, \(V(x)\equiv 1\) for all \(x\in {\mathbb {R}}^N\backslash K\). Hence, \(u(x,t)\rightarrow 1\) as \(t\rightarrow +\infty \) for all \(x\in {\mathbb {R}}^N\backslash K\). It follows that for any \(\epsilon >0\), there are some \(t_\epsilon >0\) being sufficiently large and \(K_\epsilon \subset {\overline{\Omega }}\) with \(K\subset K_\epsilon \) such that
In addition, it follows from Theorem 4.3 that
Thus, from Theorem 5.1, we have
Moreover, since \(u(x,t)-\phi (x_1+ct)\rightarrow 0\) as \(t\rightarrow \pm \infty \) uniformly in \(x\in \Omega \) and Theorem 4.3, we have that \(u(x,t)-\phi (x_1+ct)\rightarrow 0\) as \(|x|\rightarrow +\infty \) uniformly in \(t\in {\mathbb {R}}.\) The proof of Theorem 1.1 is finished. Similarly, we know the results of Remark 1.2 hold.
References
Andreu-Vaillo, F., Mazón, J.M., Rossi, J.D., Toledo-Melero, J.: Nonlocal Diffusion Problems. Mathematical Surveys and Monographs, AMS, Providence, Rhode Island (2010)
Bates, P.W., Fife, P.C., Ren, X.F., Wang, X.F.: Traveling waves in a convolution model for phase transitions. Arch. Rational Mech. Anal. 138, 105–136 (1997)
Berestycki, H., Hamel, F.: Generalized travelling waves for reaction-diffusion equations, Perspectives in nonlinear partial differential equations. Amer. Math. Soc, Contemp. Math. 446, 101–123 (2007)
Berestycki, H., Hamel, F., Matano, H.: Bistable traveling waves around an obstacle. Comm. Pure Appl. Math. 62, 729–788 (2009)
Berestycki, H., Hamel, F.: Generalized transition waves and their properties. Comm. Pure Appl. Math. 65, 592–648 (2012)
Bouhours, J.: Robustness for a Liouville type theorem in exterior domains. J. Dynam. Differential Equations 27, 297–306 (2015)
Brasseur, J., Coville, J., Hamel, F., Valdinoci, E.: Liouville type results for a nonlocal obstacle problem. Proc. London Math. Soc. 119, 291–328 (2019)
Brasseur, J., Coville, J.: A counterexample to the Liouville property of some nonlocal problems. Ann. Inst. H. Poincaré Anal. Non Linéaire 37, 549–579 (2020)
Brasseur, J., Coville, J.: Propagation phenomena with nonlocal diffusion in presence of an obstacle. J. Dynam. Differential Equations (2021). https://doi.org/10.1007/s10884-021-09988-y
Bu, Z.-H., Guo, H., Wang, Z.-C.: Transition fronts of combustion reaction diffusion equations in \({\mathbb{R}}^N\). J. Dynam. Differential Equations 31, 1987–2015 (2019)
Carr, J., Chmaj, A.: Uniqueness of travelling waves for nonlocal monostable equations. Proc. Amer. Math. Soc. 132, 2433–2439 (2004)
Cortázar, C., Elgueta, M., Quirós, F., Wolanski, N.: Asymptotic behavior for a nonlocal diffusion equation in domains with holes. Arch. Rational Mech. Anal. 205, 673–697 (2012)
Cortázar, C., Elgueta, M., Quirós, F., Wolanski, N.: Asymptotic behavior for a one-dimensional nonlocal diffusion equation in exterior domains. SIAM. J. Math. Anal. 48, 1549–1574 (2016)
Cortázar, C., Elgueta, M., Quirós, F., Wolanski, N.: Asymptotic behavior for a nonlocal diffusion equation in exterior domains : The critical two-dimensional case. J. Math. Anal. Appl. 436, 586–610 (2016)
Chapuisat, G., Grenier, E.: Existence and nonexistence of traveling wave solutions for a bistable reaction-diffusion equation in an infinite cylinder whose diameter is suddenly increased. Comm. Partial Differential Equations 30, 1805–1816 (2005)
Chasseigne, E., Chavesb, M., Rossi, J.D.: Asymptotic behavior for nonlocal diffusion equations. J. Math Pures Appl. 86, 271–291 (2006)
Chen, X.: Existence, uniqueness and asymptotic stablility of traveling waves in nonlocal evolution equations. Adv. Differential Equations 2, 125–160 (1997)
Coville, J., Dupaigne, L.: On a nonlocal reaction diffusion equation arising in population dynamics. Proc. Roy. Soc. Edinburgh Sect. 137, 727–755 (2007)
Coville, J., Dávila, J., Martínez, S.: Nonlocal anisotropic dispersal with monostable nonlinearity. J. Differential Equations 244, 3080–3118 (2008)
Guo, J.-S., Morita, Y.: Entire solutions of reaction-diffusion equations and an application to discrete diffusive equations. Discrete Contin. Dyn. Syst. 12, 193–212 (2005)
Guo, H., Hamel, F., Sheng, W.-J.: On the mean speed of bistable transition fronts in unbounded domains. J. Math. Pures Appl. 136, 92–157 (2020)
Guo, H., Monobe, H.: V-shaped fronts around an obstacle. Math. Ann. 379, 661–689 (2021)
Hamel, F., Nadirashvili, N.: Entire solution of the KPP eqution. Comm. Pure Appl. Math. 52, 1255–1276 (1999)
Hamel, F., Nadirashvili, N.: Travelling fronts and entire solutions of the Fisher-KPP equation in \(R^{N}\). Arch. Rational Mech. Anal. 157, 91–163 (2001)
Hamel, F., Rossi, L.: Transition fronts for the Fisher-KPP equation. Trans. Amer. Math. Soc. 368, 8675–8713 (2016)
Hamel, F.: Bistable transition fronts in \({\mathbb{R}}^N\). Adv. Math. 289, 279–344 (2016)
Hoffman, A., Hupkes, H.J., Van Vleck, E.S.: Entire solutions for bistable lattice differential equations with obstacles. Mem. Amer. Math. Soc. 250, (2017)
Kolmogorov, A.N., Petrovsky, I.G., Piskunov, N.S.: Étude de l’équation de la diffusion avec croissance de la quantité de la matière et son application à un problème biologique. Bull. Univ. Etat Mosc. Sr. Int. A 1, 1-26 (1937)
Li, W.-T., Sun, Y.-J., Wang, Z.-C.: Entire solutions in the Fisher-KPP equation with nonlocal dispersal. Nonlinear Anal. Real World Appl. 11, 2302–2313 (2010)
Shen, W.: Traveling waves in diffusive random media. J. Dynam. Differential Equations 16, 1011–1060 (2004)
Shen, W., Shen, Z.: Transition fronts in time heterogeneous and random media of ignition type. J. Differential Equations 262, 454–485 (2017)
Shen, W.: Stability of transition waves and positive entire solutions of Fisher-KPP equations with time and space dependence. Nonlinearity 30, 3466–3491 (2017)
Sheng, W.-J., Guo, H.: Transition fronts of time periodic bistable reaction-diffusion equations in \({\mathbb{R}}^N\). J. Differential Equations 265, 2191–2242 (2018)
Sun, Y.-J., Li, W.-T., Wang, Z.-C.: Entire solutions for nonlocal dispersal equations with bistable nonlineartiy. J. Differential Equations 251, 551–581 (2011)
Sun, Y.-J., Zhang, L., Li, W.-T., Wang, Z.-C.: Entire solutions in nonlocal monostable equations: asymmetric case. Commun. Pure Appl. Anal. 18, 1049–1072 (2019)
Wang, J.-B., Li, W.-T., Dong, F.-D., Qiao, S.-X.: Recent developments on spatial propagation for diffusion equations in shifting environments. Discrete Contin. Dyn. Syst. Ser. B 27, 5101–5127 (2022)
Wang, Z.-C., Li, W.-T., Ruan, S.: Travelling wave fronts in reaction-diffusion systems with spatio-temporal delays. J. Differential Equations 222, 185–232 (2006)
Weinberger, H.F.: Long-time behavior of a class of biological models. SIAM J. Math. Anal. 13, 353–396 (1982)
Yagisita, H.: Existence and nonexistence of traveling waves for a nonlocal monostable equation. Publ. Res. Inst. Math. Sci. 45, 925–953 (2009)
Yagisita, H.: Existence of traveling wave solutions for a nonlocal bistable equation:an abstract approach. Publ. Res. Inst. Math. Sci. 45, 955–979 (2009)
Zhang, G.-B., Zhao, X.-Q.: Propagation phenomena for a two-species Lotka-Volterra strong competition system with nonlocal dispersal. Calc. Var. Partial Differential Equations 59, (2020) Paper No.10, 34pp
Zhang, L., Li, W.-T., Wang, Z.-C.: Entire solution in an ignition nonlocal dispersal equation: asymmetric kernel. Sci. China Math. 60, 1791–1804 (2017)
Zhang, L., Li, W.-T., Wang, Z.-C., Sun, Y.-J.: Entire solutions in nonlocal bistable equations: asymmetric case. Acta Math. Sin. 35, 1771–1794 (2019)
Zlatoš, A.: Generalized traveling waves in disordered media: existence, uniqueness, and stability. Arch. Ration. Mech. Anal. 208, 447–480 (2013)
Zlatoš, A.: Transition fronts in inhomogeneous Fisher-KPP reaction-diffusion equations. J. Math. Pures Appl. 98, 89–102 (2012)
Acknowledgements
We are grateful to the anonymous referee for his/her careful reading and valuable suggestions which led to an improvement of our original manuscript. When this paper is revised, we find that Brasseur and Coville [9] also considered the obstacle problem and they have cited our work. Research of W.-T. Li was partially supported by NSF of China (11731005) and NSF of Gansu (21JR7RA537). Research of J.-W. Sun was partially supported by FRFCU (lzujbky-2021-52) and NSF of Gansu (21JR7RA535).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Submitted to JDDE on August 7, 2019.
Appendix
Appendix
1.1 Proof of Proposition 3.2
In this subsection we intend to show the results of Proposition 3.2. For convenience we define the operator \({\mathcal {L}}\) as follows
We further show that \(W^-(x,t)\) is a sub-solution. A straightforward computation shows that
For \(x_1<0\), since \(J(x)\ge 0\) and \(W^-\ge 0\), we have
For \(x_1\ge 0\), in view of that
we have
where \(z_+(t)=x_1+ct-\xi (t),~z_-(t)=-x_1+ct-\xi (t)\). Recall that \(K\subset \{x\in {\mathbb {R}}^N\mid x_1\le 0\}\), it follows that
which implies that
Now we go further to show \({\mathcal {L}}W^-\le 0\) in two subcases.
Case A: \(0<x_1<-ct+\xi (t)\).
In this case the following lemma holds.
Lemma 6.1
Suppose that (F) holds. Let \((\phi ,c)\) be the unique solution of (1.4) and \(\phi ''(\xi )\ge 0\) for \(\xi \le 0\). Then there exists \(k_3>0\) such that
for \(\xi _2<\xi _1<0\).
Proof
It follows from (1.6) that there exists some \({\mathfrak {c}}>0\) such that
Then one can choose \(M'>\frac{\ln 2{\mathfrak {c}}}{\lambda }\) and \((\xi _1,\xi _2)\in {\mathbb {R}}^2\) with \(\xi _1-M'<\xi _2<\xi _1<0\) such that
If \(\xi _1-M'<\xi _2<\xi _1<0\), then we have
for some \((\theta ^1,\theta ^2)\in [\xi _2,\xi _1]^2\) with \(|\theta ^1-\theta ^2|<M'.\) This and the facts \(\phi ''(\xi )\ge 0\) for \(\xi \le 0\) and \(\phi '(\psi )>0\) for all \(\psi \in {\mathbb {R}}\) imply that (6.1) holds true for \(\xi _1-M'<\xi _2<\xi _1<0\).
When \(\xi _2+M'<\xi _1<0\), it follows that
This and the inequalities in (1.5) and (1.6) yield
Thus we finish the proof. \(\square \)
Now we are ready to show \({\mathcal {L}}W^-(x,t)\le 0\). By Lemma 6.1, we have
The last inequality holds provided that \(M\ge \frac{L_f\beta _0}{k_3}\).
Case B: \(x_1\ge -ct+\xi (t)\).
A direct calculation gives that
If \(\lambda \ge \mu \), then
for \(ct-\xi (t)\ll -1\) and \(M>1\) is sufficiently large.
When \(\lambda <\mu \), which means \(|f'(1)-f'(0)|>0\), there holds
for \(x_1+ct-\xi (t)>L_2>0\) with \(L_2\) being large enough, where \(0<k_4<\frac{1}{2}|f'(1)-f'(0)|\). The inequality above follows from that \(f'(\phi (z_+(t)))\rightarrow f'(1)\) and \(f'(\phi (z_-(t)))\rightarrow f'(0)\) as \(L_2\rightarrow +\infty \). Then
provided that \(ct+\xi (t)\ll -1\).
In addition, for \(0<x_1+ct-\xi (t)<L_2\), there holds
Since \(ct-\xi (t)\ll -1\) and \(M\gg 1\), we have \({\mathcal {L}}W^-(x,t)\le 0\).
Next, we show the \(W^+(x,t)\) is a super-solution. A straightforward computation shows that
When \(x_1\ge 0\), denote
In view of that \(K\subset {\mathbb {R}}^N\setminus \)supp\((J)\cap \{x\in {\mathbb {R}}^N:~x_1\le 0\}\), one gets
Observe that, if \(x_1>|ct+\xi (t)|>L\), where L is the diameter of the compact support of J, then the integral item of the last equality is equal to 0. Therefore, we obtain
If \(\mu \le \lambda \), by choosing \(M\gamma _1\ge L_f\alpha _0\), it is obvious that \({\mathcal {L}}W^+(x,t)\ge 0\) with \(x_1>|ct+\xi (t)|\) being sufficiently large.
For \(\mu >\lambda \), we have \(f'(1)<f'(0)\). Consider the case \(x_1+ct+\xi (t)\ge L_0\gg 1\). Then \(\phi (x_1+ct+\xi (t))\approx 1\) while \(\phi (-x_1+ct+\xi (t))\approx 0\). Furthermore,
which implies that \({\mathcal {L}}W^+(x,t)\ge 0\). For the other case \(x_1+ct+\xi (t)\le L_0\), we know
Since \(\lambda _0<\lambda \), we obtain \({\mathcal {L}}W^+(x,t)\ge 0\) holds provided that \(M\ge \frac{L_f\alpha _0}{\gamma _1}e^{\mu L_0}\).
For the case \(0<x_1<|ct+\xi (t)|\), one can see that
Since \(\phi (ct+\xi (t))\le \frac{\theta }{2}\) for \(ct+\xi (t)\ll -1\), \(\phi (0)\le \theta ,\) and \(\phi '>0\), we get that \(\phi (-y_1+ct+\xi (t))>\theta \ge 2\phi (ct+\xi (t))\) for \(y_1<ct+\xi (t)\). It follows that \(I_1\le 0\). We know that
The first inequality is follows from that there exist two numbers \(K_\phi >0\) and \(k_\phi >0\) such that \(\left| \phi (x_1)-C_{\phi }e^{\lambda x_1}\right| \le K_{\phi }e^{(k_{\phi }+\lambda )x_1}\) for \(x_1\le 0\) which is easy to obtain by (1.5). Then we have
This gives that \({\mathcal {L}}W^+\ge 0\), provided \(2M\alpha _0>L_f\beta _0+C_0\) and \(\lambda _0<\min \{k_{\phi },\lambda \}\).
For \(x_1<0\), we just deal with the case \(-L<x_1<0\) because that for \(x_1\le -L\),
Since \(\phi ''(x)\ge 0\) for \(x\le 0\), we have that
Observe that, if \(x_1<0\) then \(\left| \phi (x_1)-C_{\phi }e^{\lambda x_1}\right| \le K_{\phi }e^{(k_{\phi }+\lambda )x_1}\). Thus there is \(C_0>0\) such that
The last inequality above holds true, since \(x_1<0\) and \(f(\upsilon )=\upsilon +\frac{1}{\upsilon }\) is monotonically increasing in \(\upsilon \in (1,\infty )\). Then it follows that for \(M\ge \frac{C_0}{\gamma _0}\),
The second inequality follows from that \(f'(s)<0\) in \([\phi (ct+\xi (t)),2\phi (ct+\xi (t))]\) for \(ct+\xi (t)\ll -1\). The proof of Proposition 3.2 has been finished.
1.2 Proof of Lemma 4.2
We know that
When \(\xi _-(x,t)\in [-A, A]\), there holds \(\phi '(\xi _-(x,t))\ge \delta \). Therefore,
For \(|\xi _-(x,t)|\ge A\), we have
Then \(f'(s)\le -\omega \) for \(s\in [\phi (\xi _-(xt))-\epsilon e^{-\omega (t-t_0)},\phi (\xi _-(xt))]\). Hence,
For \(t_0\le -T\), one get
Until now, we have show the function \({\underline{u}}\) is a sub-solution to (4.2). Similarly one can show \({\overline{u}}\) is a super-solution to (4.2).
Rights and permissions
About this article
Cite this article
Qiao, SX., Li, WT. & Sun, JW. Propagation Phenomena for Nonlocal Dispersal Equations in Exterior Domains. J Dyn Diff Equat 35, 1099–1131 (2023). https://doi.org/10.1007/s10884-022-10194-7
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10884-022-10194-7