Abstract
We show the nonexistence of traveling wave solutions in the combustion model with fractional Laplacian \(\displaystyle (-\Delta )^s\) when \(\displaystyle s\in (0,1/2]\). Our method can be used to give a direct and simple proof of the nonexistence of traveling fronts for the usual Fisher-KPP nonlinearity. Also we prove the existence and nonexistence of traveling wave solutions for different ranges of the fractional power \(s\) for the generalized Fisher–KPP type model.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
The paper is concerned with the traveling fronts of the reaction diffusion equation:
for \(f\in C^1(\mathbb {R})\), namely the solution to the following equation:
where \(\mu \) is the speed of propagation of the front and the operator \((-\Delta )^s\) denotes the fractional power of the Laplacian in one dimension with \(0<s<1\). Recall the fractional Laplacian is defined as follows:
where (P.V.) stands for Cauchy principal value and \(\displaystyle C_{1,s}=\frac{2^{2s}s\Gamma ((1+2s)/2)}{\pi ^{1/2}\Gamma (1-s)}\), see for example [8]. The original models with the standard Laplacian \(\displaystyle (-\Delta )\) arise in applied sciences such as population genetics, combustion, and nerve pulse propagation, etc. The detailed formulations of the models were discussed by Fisher in [5], Kolmogorov et al. in [7] and Aronson and Weinberger in [1], etc. The classical results of the existence and nonexistence of traveling fronts for the models can be found therein.
By a compactness argument, we know that if (1.1) has a solution \(u(x)\) then
Multiplying \(u'(x)\) on both sides in (1.1) and integrating over \(\mathbb {R}\), we can get the Hamiltonian identity as in [6]:
Roquejoffre et al. [9] studied the combustion model, i.e, there exists some \(\theta \in (0,1)\) such that \(f\in C^1(\mathbb {R})\) satisfies
They have shown that when \(s\in (1/2, 1)\) and \(f\) satisfies (1.4), there exists an unique \((\mu , u)\) with \(\mu >0\) to (1.1).
In this paper, we will show that when \(s\in (0,1/2]\) and \(f\) satisfies (1.4), there is no traveling wave solution for the combustion model, i.e., (1.1) has no solution. In fact, we shall show the following:
Theorem 1.1
Suppose that there exists some \(\theta \in (0,1)\) such that \(f\in C^1(\mathbb {R})\) satisfies
Then there is no solution to (1.1) if \(0<s\le \frac{1}{2}\).
Obviously this theorem applies to the combustion model. For the Fisher-KPP model, i.e, \(f\in C^1(\mathbb {R})\) satisfies
Theorem 1.1 implies that if \(0<s\le 1/2\), (1.1) has no solution.
We shall also study the generalized Fisher-KPP model and prove nonexistence and existence of solutions to (1.1) for different ranges of \(s\in (0,1)\). We shall point out that there is a delicate balance between the diffusion factor \(s\) and the reaction power \(p\) in order to have a traveling wave solution. In fact, we shall prove the following theorem.
Theorem 1.2
Assume there exist some \(\theta \in (0,1)\), \(0<p<\infty \), \(A_1>0\) and \(A_2>0\) such that
Then (1.1) has a solution if and only if \(p>2\) and \(s\ge \frac{p}{2(p-1)}\).
We note that the condition \( A_1 u^p\le f(u)\le A_2u^p,\quad \forall u\in [0, \theta ]\) in the above theorem is not needed for the nonexistence result. We include it in (1.7) for the simplicity of the statement.
We also obtain the asymptotics of solutions as \(x\rightarrow \pm \infty \) when they exist. Indeed we show the following asymptotic behaviors.
Theorem 1.3
Assume that \(f\) satisfies (1.7), let \((\mu ,u)\) be a solution to (1.1) with \(\mu >0\). Then there exists some constant \(C>0\) such that
Note in Theroem 1.3, \(s\) is always bigger than \(1/2\) by Theorem 1.2.
We would also like to point out that Cabré and Roquejoffre in [4] already proved that when \(0<s<1\), there is no traveling wave solution for the Fisher-KPP model by studying the exponential speed of the front propagation. Theorem 1.3, in particular, shows directly that (1.1) has no solution for the Fisher-KPP model.
The plan of the paper is the following. In Sect. 2, we prove Theorem 1.1 by considering the \(s\)-harmonic extension of the fractional Laplacian given in [3]. The key ingredient is to show certain asymptotic rates of solutions at \(-\infty \). Section 3 is devoted to prove Theorem 1.2. The proof of nonexistence follows the same idea as in Sect. 2 by using \(s\)-harmonic extension of the fractional Laplacian. We use an iterative argument to obtain an accurate asymptotic behavior of possible solutions. The proof of existence relies on the truncation of domain, asymptotic behavior of solutions and a sliding argument as in [9]. In Sect. 4, detailed asymptotical behavior of solutions will be given.
2 Nonexistence in the combustion and Fisher–KPP models when \(0<s\le 1/2\)
In this section, we assume \(0<s\le 1/2\) and \(f\in C^1(\mathbb {R})\) satisfies condition (1.5). We prove Theorem 1.1 by contradiction. Assume that \((\mu , u)\) is a solution to (1.1). By (1.3) and (1.5), we have \(\mu >0\). Let \(\overline{u}\) be the \(s\)-harmonic extension of \(u\) on \(\mathbb {R}^2_+\), that is,
with
Let \(v(x,y)=\overline{u}_x(x,y)=P_y*u'(x)\) for all \((x,y)\in \mathbb {R}^2_+\), Caffarelli and Silvestre [3] proved that \(v\) satisfies
where \(\displaystyle d_s=\frac{2^{1-2s}\Gamma (1-s)}{\Gamma (s)}\).
By the standard maximal principle arguments, it is easy to see that \(u'(x)>0\) for all \(x\in \mathbb {R}\) and \(\displaystyle \lim _{|x|\rightarrow \infty }\ u'(x)=0\) (see, e.g., [6, 9]). Then we know that
By (1.1), without loss of generality, we can assume \(u(-1)=\theta \). Since \(u'(x)>0\) for all \(x\in \mathbb {R}\), we have \(f'(u(x))\ge 0\) for all \(x\le -1\). In summary, we know that \(v\) satisfies
Define the auxiliary function
Direct computations tell us that for all \(x\le -1\) and all \(y\ge 0\), we have
Since \(0<s\le \frac{1}{2}\), we have \(\frac{1}{|x|^2}\le \frac{1}{|x|^{1+2s}}\) for all \(x\le -1\). Hence for all \(x\le -1\), we have
For any \(\delta >0\), let \(w_\delta (x,y)=v(x,y)-\delta \varphi (x,y)\) for all \(x\le -1\) and all \(y\ge 0\), then \(w_\delta \) satisfies
Lemma 2.1
There exists some \(\delta _0>0\) such that \(w_{\delta _0}(-1,y)\ge 0\) for all \(y\ge 0\).
Proof
First we see that
Since \(u'(x)>0\) for all \(x\in \mathbb {R}\), then \(u(0)>u(-1)\), which implies that there exists some constant \(B_1>0\) such that
Since \(v(x,y)=P_y*u'(x)\) for all \((x,y)\in \mathbb {R}^2_+\), by (2.2), for all \(y\ge 1\) we have
On the other hand, since \(v(x,y)>0\) for all \((x,y)\in \overline{\mathbb {R}^2_+}\), there exists some \(B_2>0\) such that
Let \(\delta _0=\min \{B_1,B_2\}>0\), we know that
\(\square \)
Lemma 2.2
For the above \(\delta _0\) in Lemma 2.1, there holds
Proof
Assume \(w_{\delta _0}(x_0,y_0)<0\) for some \(x_0\le -1\) and some \(y_0\ge 0\). Since \(w_{\delta _0}(x,y)\rightarrow 0\), as \(|(x,y)|\rightarrow \infty \), by Lemma 2.1, we know that there exists some \(x_1<-1\) and some \(y_1\ge 0\) such that
By the strong maximum principle for uniformly elliptic equations, we know that \(y_1=0\). Applying Hopf lemma as in [2], we have
Since \(x_1\) is an interior minimum of \(w_{\delta _0}(x,0)\) in \(x<-1\), then we have \(D_xw_{\delta _0}(x_1,0)=0\). By (2.4), we get
We get a contradiction. Therefore
\(\square \)
Proof of Theorem 1.1
Assume \((\mu ,u)\) is a solution to (1.1). By Lemma 2.2, we know that
Since \(\displaystyle \varphi (x,0)\ge \frac{sd_s}{\mu }\cdot \frac{1}{|x|}\) for all \(x\le -1\), we know that
On the other hand, we know that \(\displaystyle \int _\mathbb {R}u'(x)\ dx=1\). This is a contradiction which implies that there is no solution to (1.1). \(\square \)
3 Generalized Fisher-KPP model when \(1/2<s<1\)
In this section, we assume that \(\frac{1}{2}<s<1\) and \(f\in C^1(\mathbb {R})\) satisfies condition (1.7). One example for (1.7) is the following:
where \(p>0\) is the reaction power.
Our goal is to find the critical exponent \(s=s(p)\) such that a solution of (1.1) exists if and only if \(1> s\ge s(p)\). In this section, we provide the proof of nonexistence of solutions for (1.1) when \(s<s(p)\) by studying the asymptotics of solutions related to (1.1). By Theorem 1.1, it is readily seen that the solution to (1.1) does not exist when \(\displaystyle 0< s\le 1/2\). Later, we shall discuss the existence of solutions to (1.1) by a similar argument as in [9].
3.1 Nonexistence results
The following lemma is important in the proof of nonexistence, and has already been proven in [9]. For completeness, we list the proof here.
Lemma 3.1
Let \(\frac{1}{2}<s<1\) and \(u\in C^2(\mathbb {R})\) such that \(\displaystyle \lim _{|x|\rightarrow \infty }u'(x)=0\) and \(\displaystyle \lim _{x\rightarrow \pm \infty }u(x)=L^{\pm }\) for some \(L^-,L^+\in \mathbb {R}\), then we have
Proof
For any \(R>0\), we have
For \(\displaystyle \int _{-R}^R\int _{|w|\ge 1}\frac{u(y)-u(y+w)}{|w|^{1+2s}}\ dwdy\), since \(\frac{1}{2}<s\), by Fubini–Tonelli’s theorem and the dominated convergence theorem, we know that
For \(\displaystyle \int _{-R}^R\int _{|w|<1}\frac{u(y+w)-u(y)-u'(y)w}{|w|^{1+2s}}\ dwdy\), since \(s<1\), by Fubini–Tonelli’s theorem and the dominated convergence theorem, we know that
Therefore, we can conclude that \(\displaystyle \int _{-R}^R(-\Delta )^su(y)\ dy\rightarrow 0\), as \(R\rightarrow \infty \). \(\square \)
Remark 3.1
If \((\mu ,u)\) is a solution to (1.1), since \(u'\in L^1(\mathbb {R})\), by Lemma 3.1 and \(f(u)\ge 0\) for all \(u\in [0,1]\) we know that \(f(u)\in L^1(\mathbb {R})\). In particular, if we know that there exists some constants \(C>0\) and \(r>0\) such that
then we have \(r>1\) by the integrability of \(u'\). On the other hand, by (1.7), we know that \(f(u(x))\ge A\left( \frac{C}{r-1}\cdot \frac{1}{|x|^{r-1}}\right) ^p\) for all \(x\le -1\). Hence it necessarily holds that \((r-1)p>1\), i.e., \(r>\frac{p+1}{p}\).
In the following, we assume that \((\mu ,u)\) is a solution to (1.1) with \(\mu >0\) and \(u(-1)=\theta \). Let \(\overline{u}\) be the \(s\)-harmonic extension of \(u\) on \(\mathbb {R}^2_+\) and \(v(x,y)=\overline{u}_x(x,y)=P_y*u'(x)\) for all \((x,y)\in \mathbb {R}^2_+\), by the same discussion as in Sect. 2, we know that \(v\) satisfies
For any \(\alpha \in [1,2s]\) and \(\beta >0\), we consider the auxiliary functions
By direct computations, for all \(x\le -1\) and all \(y\ge 0\) we know that
Hence for all \(x\le -1\), we have
For any \(\delta \in (0,1)\), let
Then \(w_{\delta ,\alpha ,\beta }\) satisfies
Lemma 3.2
For any fixed \(\alpha \in [1,2s]\) and \(\beta >0\), for all \(\delta \in (0,1]\), if we have
then there exists some constant \(C>0\) such that
Proof
Since \(\alpha \ge 1\), we know that \(\displaystyle \frac{1}{[1+y^2]^{\frac{\alpha }{2}}}\le \frac{1}{[1+y^2]^{\frac{1}{2}}}\) for all \(y\ge 0\). By taking the limit of the ratio, one can get
By the same arguments as in Lemma 2.1 and Lemma 2.2, we know that there exists some small \(\delta _0>0\) such that
Since \(\displaystyle \varphi _{\alpha ,\beta }(x,0)\ge \frac{2\beta sd_s}{\mu }\cdot \frac{1}{|x|^\alpha }\) for all \(x\le -1\), we have
\(\square \)
Lemma 3.3
(Initial Asymptotic Rate) There exists some constant \(C_0>0\) such that
Proof
Let \(\alpha =2s\) and \(\beta =1\) in Lemma 3.2. Observe that
Then Lemma 3.2 leads to the conclusion. \(\square \)
Remark 3.2
Lemma 3.3 provides an alternative proof of Proposition 4.2 in [9].
As an immediate consequence of Lemma 3.3 and Remark 3.1, we have the following
Theorem 3.1
Let \(\displaystyle \frac{1}{2}<s\le \frac{p+1}{2p}\), then there is no solution to (1.1). In particular, for all \(0<p\le 1\) and \(\displaystyle \frac{1}{2}<s<1\), there is no solution to (1.1).
Lemma 3.4
(Asymptotic Rate Lifting) Let \(\frac{p+1}{2p}<s<1\) and \(r\in (\frac{p+1}{p},2s]\), we assume there exists some constant \(B_0>0\) such that
Let \(\alpha \in [1,2s]\) be such that \( \alpha \ge p(r-1)\), then there exists some constant \(C>0\) such that
Proof
By the assumption and (1.7), for all \(\beta >0\), all \(\delta \in (0,1]\) and all \(x\le -1\), we know that
Let \(\displaystyle \beta =\frac{A_1B_0^p}{2\delta sd_s}>0\), by Lemma 3.2, we have completed the proof. \(\square \)
Remark 3.3
If \(\frac{p+1}{p}<r<\frac{p}{p-1}\), by letting \(\rho (r):=p(r-1)\), we know that
We shall show the following theorem.
Theorem 3.1
Let \(p>1\) and \(\frac{1}{2}<s<\min \left\{ 1,\frac{p}{2(p-1)}\right\} \), then (1.1) has no solution.
Proof
By Lemma 3.4, we have the following
Claim : if
for some \( r \in \left( \frac{p+1}{p}, 2s\right] , \) then
with \(\alpha \in \left( \frac{p+1}{p}, p(r-1)\right) \). This is a consequence of the fact that the function \(\rho (r)=p(r-1)\) has a unique fixed point at \(r = \frac{p}{p-1}\) and \(\rho (r) <r\) for \(r <\frac{p}{p-1}\), which implies that, for \(\alpha \) and \(r\) as above, there holds \(r<2s<\frac{p}{p-1}\) and then \(\alpha < r \le 2s\). The claim then follows from Lemma 3.4. Now, one can apply recursively the claim, starting with \(r = 2s < \frac{p}{p-1}\) and after a finite number of steps, get \(\alpha =\frac{p+1}{p}\), because \(\rho ^{(n)}(r):=\rho \circ \rho \cdots \rho (r) = \frac{ p^n[p(r-1)-r]+p}{p-1} \rightarrow -\infty \) as \(n \rightarrow \infty \). This is a contradiction to Remark 3.1. \(\square \)
Note that \(\frac{p}{2(p-1)}\ge 1\) if \(1<p\le 2\), and \(\frac{p}{2(p-1)}<1\) if \(2<p\). Therefore, there is no solution to (1.1) for all \( s \in (0, 1)\) if \(p \le 2\).
3.2 Existence results
In this subsection, we assume that \(f\) satisfies (1.7), \(p>2\) and \(\frac{p}{2(p-1)}\le s<1\), we will show that a solution to (1.1) exists. Mellet et al. [9] have shown the existence of traveling fronts for the non local combustion model when \(\frac{1}{2}<s<1\). The proof for the generalized Fisher-KPP model follows a similar argument to that in [9]. For any \(\mu \in \mathbb {R}\) and \(b>0\), we first consider the following truncated problem:
Proposition 3.1
Assume \(\displaystyle s\ge \frac{p}{2(p-1)}\) and \(f\) satisfies (1.7). Then there exists a constant \(M\) such that if \(b>M\) the truncated problem 3.3 has a solution \((u_b, \mu _b)\). Furthermore, the following properties hold:
-
(1)
There exists \(K\) independent of \(b\) such that \(-K\le \mu _b\le K\);
-
(2)
\(u_b\) is non-decreasing with respect to \(x\) and satisfies \(0<u_b(x)<1\) for all \(x\in (-b, b)\).
To prove this Proposition, we need the construction of sub- and super-solutions. The construction is based on the following lemmas, same as in [9]. We would like to present the proof of the following second lemma, and especially elaborate on the sliding method mentioned in [9].
Lemma 3.5
For any \(\mu \in \mathbb {R}\) and \(b>0\), (3.3) has a solution \(u_{\mu ,b}\) such that \(0\le u_{\mu ,b}(x)\le 1\) in \(\mathbb {R}\), \(u_{\mu ,b}\) is non-decreasing in \(\mathbb {R}\) and \(\mu \rightarrow u_{\mu ,b}\) is continuous.
Proof
The proof is the same as the proof of Lemma 2.4 in [9]. \(\square \)
Lemma 3.6
There exists some constants \(M,K>0\) such that for all \(b>M\), we have
-
a.
If \(\mu >K\), then \(u_{\mu ,b}(0)<\theta \);
-
b.
If \(\mu <-K\), then \(u_{\mu ,b}(0)>\theta \).
Together with Lemma 3.5, Lemma 3.6 implies that there exists \(\mu _b\in [-K, K]\) such that \(u_{\mu ,b}(0)=\theta \).
Proof
Consider the function
Since \(2s>1\), by Lemma 2.2 in [9], we have
Moreover, by (1.7), we get
Since \(\frac{p}{p-1}\le 2s\), we have \((2s-1)p\ge 2s\), which implies that for all \(\displaystyle \mu \ge \frac{C_{1,s}}{2s(2s-1)}+\frac{A_2+1}{2s-1}\),
Since \(4s-1>2s\), we know that there exists some large \(A>0\) , which is independent of \(\mu \), such that for all \(\displaystyle \mu \ge \frac{C_{1,s}}{2s(2s-1)}+\frac{A_2+1}{2s-1}\), we have
For \(-A<x<-1\), we know that \((-\Delta )^s\varphi (x)\) is bounded, but \(\displaystyle \varphi '(x)=\frac{2s-1}{|x|^{2s}}\ge \frac{2s-1}{A^{2s}}\). So there exists some \(K>0\) such that for all \(\mu \ge K\),
Hence for all \(\mu \ge K\), we have
On the other hand, by the definition of \(\varphi (x)\) and (1.2), we know that for all \(x\ge -1\), \((-\Delta )^s\varphi (x)>0\), \(\varphi '(x)=0\) and \(f(\varphi (x))=0\). In summary, for all \(\mu \ge K\), we have \(\varphi (x)\) is a super-solution for (3.3). Now fix some large \(M>0\) such that \(\varphi (-M)=\frac{1}{M^{2s-1}}<\theta \).
Claim: For all \(\mu \ge K\) and all \(b\ge M\), we have \(u_{\mu ,b}(x)\le \varphi (x-M)\) for all \(x\in \mathbb {R}\), in particular, \(u_{\mu ,b}(0)<\theta \).
Let \(\phi (x)=\varphi (x-M)\) and define
Let
then \(\mathcal {O}\) is nonempty since \(\{t\ge 2b\}\subset \mathcal {O}\). \(\mathcal {O}\) is clearly closed. Take a convergent sequence \(\{t_n\}\subset \mathcal {O}\), \(t_n\rightarrow t\) as \(n\rightarrow \infty \), then
Therefore \(t\in \mathcal {O}\).
Next we show that for any \(t\in \mathcal {O}\),
In fact, if there exists \(x_0\in (-b, b)\) such that \(\Psi _t(x_0)=\phi (x_0+t)-u_{\mu }(x_0)= 0\), then
This is a contradiction.
It follows that \(\mathcal {O}\) is open. Together with the fact that \(\mathcal {O}\) is closed, we get \(\mathcal {O}=[0, \infty )\). By the above sliding argument we know
Similarly, for a lower bound we define \(\varphi _1(x)=1-\varphi (-x)\). Then if \(\mu \le -K\), \(x>1\),
Moreover \(\varphi _1(x)=0\) for \(x\le 1\). Take \(M\) so that \(\varphi _1(-M)=1-t_0\), then \(\varphi _1(x)>\theta \) for \(x\ge M\). Define \(\varphi _{1,M}(x)=\varphi _1(x+M)\), then \(\varphi _{1,M}\) is a sub-solution to 3.3. Therefore by the same argument as above \(u_{\mu }(0)\ge \varphi _{1,M}(0)>\theta \) for \(\mu <-K\). \(\square \)
Theorem 3.2
Under the conditions of Proposition 3.1, there exists a subsequence \(b_n\rightarrow \infty \) such that \(\displaystyle u_{b_n}\rightarrow u_0\) and \(\mu _{b_n}\rightarrow \mu _0\). Furthermore, \(\mu _0\in (0,K]\) and \(u_0\) is a monotone increasing solution of (1.1).
Proof
By Lemma 3.6, \(\mu _b\in [-K, K]\) we have the elliptic estimate for \(u_b\):
for some \(\alpha \in (0,1)\). Thus there exists a subsequence \(b_n\rightarrow \infty \) such that
Thus \(u_0\) satisfies \((-\Delta )^s u_0+\mu _0u'_0=f(u_0)\). Also we know \(u_0\) is monotone increasing, \(u_0(0)=\theta \) and \(u_0\) is bounded. By a compactness argument, there exist \(\gamma _0\), \(\gamma _1\) such that \(\displaystyle \lim _{x\rightarrow -\infty }u_0(x)=\gamma _0\) and \(\displaystyle \lim _{x\rightarrow \infty }u_0(x)=\gamma _1\) with
We know both \(\gamma _0\) and \(\gamma _1\) satisfy \(f(\gamma _0)=0\) and \(f(\gamma _1)=0\) which implies \(\gamma _0=0,\ \gamma _1=1\). Moreover, by integrating \((-\Delta )^{s}u_0+\mu _0 u'_0=f(u_0)\) over \(\mathbb {R}\), together with Lemma 3.1, we know
\(\square \)
4 Asymptotic rate at \(\pm \infty \)
In this section, we will study asymptotic behaviors of solutions to (1.1) when \(x\rightarrow \pm \infty \). Let \(f\in C^1(\mathbb {R})\) satisfy (1.7) and \((\mu ,u)\) be a solution to (1.1). First we investigate the asymptotic behavior of \(u\) when \(x\rightarrow \infty \). Let \(M=\Vert f\Vert _{C^1([0,1])}>0\), by (3.1), we know that
We consider the auxiliary function
By direct computations, for all \(x\ge 1\) and all \(y\ge 0\), we know that
Hence for all \(x\ge 1\), we have
For any \(\delta >0\), let
Then \(w_{\delta }\) satisfies
We have the following
Proposition 4.1
There exists some constant \(C>0\) such that
Proof
By the same argument as in Lemma 2.2, we know that there is a positive constant \(\delta _0\) such that
In particular, we know that
\(\square \)
Lemma 4.1
Let \(\beta >0\), we consider the function
Then
-
a.
If \(0<\beta <1\), we have
$$\begin{aligned} (-\Delta )^s\psi _\beta (x)=-\frac{C_{1,s}\cdot B(2s+\beta ,1-\beta )}{x^{2s+\beta }}+o\left( \frac{1}{x^{2s+\beta }}\right) ,\quad \text {as } x\rightarrow \infty . \end{aligned}$$ -
b.
If \(\beta >1\), we have
$$\begin{aligned} (-\Delta )^s\psi _\beta (x)=-\frac{C_{1,s}}{\beta -1}\cdot \frac{1}{x^{1+2s}}+o\left( \frac{1}{x^{1+2s}}\right) ,\quad \text {as } x\rightarrow \infty . \end{aligned}$$ -
c.
If \(\beta =1\), we have
$$\begin{aligned} (-\Delta )^s\psi _1(x)=-\frac{C_{1,s}\ln x}{x^{2s+1}}+o\left( \frac{\ln x}{x^{2s+1}}\right) ,\quad \text {as } x\rightarrow \infty . \end{aligned}$$
Proof
In fact, for all \(x\ge 2\), by changing of variables, we know that
-
a.
When \(0<\beta <1\), we have
$$\begin{aligned} \displaystyle \int _{-2}^{-1}\frac{1}{|z+1|^\beta }\ dz<\infty . \end{aligned}$$By the dominated convergence theorem, we know that
$$\begin{aligned} \displaystyle \int _{-\infty }^{-1-\frac{1}{x}}\frac{1}{|z+1|^{\beta }|z|^{1+2s}}\ dz\rightarrow \int _{-\infty }^{-1}\frac{1}{|z+1|^{\beta }|z|^{1+2s}}\ dz, \quad \text {as } x\rightarrow \infty . \end{aligned}$$On the other hand, we know that
$$\begin{aligned} \int _{-\infty }^{-1}\frac{1}{|z+1|^{\beta }|z|^{1+2s}}\ dz&= \int _0^1 \frac{y^{1+2s}}{\left| -\frac{1}{y}+1\right| ^\beta }\cdot \frac{1}{y^2}\ dy \quad \left( \hbox {by letting } z=-\frac{1}{y}\right) \\&= \int _0^1 y^{2s+\beta -1}(1-y)^{-\beta }\ dy=B(2s+\beta ,1-\beta )>0. \end{aligned}$$So we know that
$$\begin{aligned} (-\Delta )^s\psi _\beta (x)=-\frac{C_{1,s}\cdot B(2s+\beta ,1-\beta )}{x^{2s+\beta }}+o\left( \frac{1}{x^{2s+\beta }}\right) ,\quad \text {as } x\rightarrow \infty . \end{aligned}$$ -
b.
When \(\beta >1\), we know that \(\displaystyle \int _{-2}^{-1}\frac{1}{|z+1|^{\beta }}\ dz=\infty \), which implies that
$$\begin{aligned} \displaystyle \int _{-\infty }^{-1-\frac{1}{x}}\frac{1}{|z+1|^{\beta }|z|^{1+2s}}\ dz\rightarrow \infty , \quad \text { as } x\rightarrow \infty . \end{aligned}$$By L’Hospital rule, we have
$$\begin{aligned} \lim _{x\rightarrow \infty }\ \frac{\int _{-\infty }^{-1-\frac{1}{x}}\frac{1}{|z+1|^{\beta }|z|^{1+2s}}\ dz}{x^{\beta -1}}&= \lim _{x\rightarrow \infty }\frac{x^{\beta }\cdot \left| 1+\frac{1}{x}\right| ^{-1-2s}\cdot \frac{1}{x^2}}{(\beta -1) x^{\beta -2}}=\frac{1}{\beta -1}. \end{aligned}$$So we derive
$$\begin{aligned} (-\Delta )^s\psi _\beta (x)=-\frac{C_{1,s}}{\beta -1}\cdot \frac{1}{x^{1+2s}}+o\left( \frac{1}{x^{1+2s}}\right) ,\quad \text {as } x\rightarrow \infty . \end{aligned}$$ -
c.
When \(\beta =1\), we know that \(\displaystyle \int _{-2}^{-1}\frac{1}{|z+1|}\ dz=\infty \), which implies that
$$\begin{aligned} \displaystyle \int _{-\infty }^{-1-\frac{1}{x}}\frac{1}{|z+1||z|^{1+2s}}\ dz\rightarrow \infty , \quad \text {as } x\rightarrow \infty . \end{aligned}$$By L’Hospital rule, we know that
$$\begin{aligned} \lim _{x\rightarrow \infty }\ \frac{\int _{-\infty }^{-1-\frac{1}{x}}\frac{1}{|z+1||z|^{1+2s}}\ dz}{\ln x}&= \lim _{x\rightarrow \infty }\ \frac{|x|\cdot \left| 1+\frac{1}{x}\right| ^{-1-2s}\cdot \frac{1}{x^2}}{\frac{1}{x}}=1. \end{aligned}$$Therefore we have
$$\begin{aligned} (-\Delta )^s\psi _1(x)=-\frac{C_{1,s}\ln x}{x^{2s+1}}+o\left( \frac{\ln x}{x^{2s+1}}\right) ,\quad \text {as } x\rightarrow \infty . \end{aligned}$$
\(\square \)
Lemma 4.2
Let \(\beta >0\), \(\psi _\beta (x)\) be defined as in Lemma 4.1, then we have the following estimates:
-
a.
If \(0<\beta <1\), there holds that
$$\begin{aligned} (-\Delta )^s\psi _\beta (x)=-\frac{C_{1,s}\cdot A(s,\beta )}{|x|^{2s+\beta }}+o\left( \frac{1}{|x|^{2s+\beta }}\right) ,\quad \text {as } x\rightarrow -\infty ; \end{aligned}$$where
$$\begin{aligned} A(s,\beta )=\int _1^\infty \frac{1}{|z|^{1+2s}|z+1|^\beta }\ dz-\frac{1}{s}+\int _{0}^{1} \frac{\frac{1}{|z-1|^\beta }+\frac{1}{|z+1|^\beta }-2}{|z|^{1+2s}}\ dz; \end{aligned}$$ -
b.
If \(\beta >1\), we have
$$\begin{aligned} (-\Delta )^s\psi _\beta (x)=-\frac{C_{1,s}}{\beta -1}\cdot \frac{1}{|x|^{1+2s}}+o\left( \frac{1}{|x|^{2s+1}}\right) ,\quad \text {as } x\rightarrow -\infty ; \end{aligned}$$ -
c.
If \(\beta =1\), we have
$$\begin{aligned} (-\Delta )^s \psi _1(x)=-\frac{C_{1,s}\ln |x|}{|x|^{2s+1}}+o\left( \frac{\ln |x|}{|x|^{2s+1}}\right) ,\quad \text {as } x\rightarrow -\infty . \end{aligned}$$
Proof
For all \(x<-2\), we know that \(x+1<-x-1\) and
For the first term inside the bracket, we know that
-
a.
Since \(\beta \in (0,1)\), we know that \(\displaystyle \int _0^1 \frac{1}{|z-1|^\beta }\ dz<\infty \), which implies that
$$\begin{aligned} \lim _{x\rightarrow -\infty }\ \int _{0}^{1+\frac{1}{x}} \frac{\frac{1}{|z-1|^\beta }+\frac{1}{|z+1|^\beta }-2}{|z|^{1+2s}}\ dz=\int _{0}^{1} \frac{\frac{1}{|z-1|^\beta }+\frac{1}{|z+1|^\beta }-2}{|z|^{1+2s}}\ dz. \end{aligned}$$Let
$$\begin{aligned} \displaystyle A(s,\beta )=\int _1^\infty \frac{1}{|z|^{1+2s}|z+1|^\beta }\ dz-\frac{1}{s}+\int _{0}^{1} \frac{\frac{1}{|z-1|^\beta }+\frac{1}{|z+1|^\beta }-2}{|z|^{1+2s}}\ dz, \end{aligned}$$then we have
$$\begin{aligned} (-\Delta )^s\psi _\beta (x)=-\frac{C_{1,s}\cdot A(s,\beta )}{|x|^{2s+\beta }}+o\left( \frac{1}{|x|^{2s+\beta }}\right) ,\quad \text {as } x\rightarrow -\infty . \end{aligned}$$ -
b.
Since \(\beta >1\), we know that \(\displaystyle \int _0^1\frac{1}{|z-1|^\beta }\ dz=\infty \), which implies that
$$\begin{aligned} \displaystyle \int _{0}^{1+\frac{1}{x}} \frac{\frac{1}{|z-1|^\beta }+\frac{1}{|z+1|^\beta }-2}{|z|^{1+2s}}\ dz\rightarrow \infty , \quad \text {as } x\rightarrow -\infty . \end{aligned}$$By L’Hospital rule, we know that
$$\begin{aligned} \lim _{x\rightarrow -\infty }\ \frac{\int _{0}^{1+\frac{1}{x}} \frac{\frac{1}{|z-1|^\beta }+\frac{1}{|z+1|^\beta }-2}{|z|^{1+2s}}\ dz}{(-x)^{\beta -1}}&= \lim _{x\rightarrow -\infty }\frac{\left[ |x|^\beta +\frac{1}{2^\beta }-2\right] \cdot \left( -\frac{1}{x^2}\right) }{-(\beta -1)(-x)^{\beta -2}}=\frac{1}{\beta -1}. \end{aligned}$$Hence we have
$$\begin{aligned} (-\Delta )^s\psi _\beta (x)=-\frac{C_{1,s}}{\beta -1}\cdot \frac{1}{|x|^{2s+1}}+o\left( \frac{1}{|x|^{1+2s}}\right) ,\quad \text {as } x\rightarrow -\infty . \end{aligned}$$ -
c.
Since \(\beta =1\), we know that \(\displaystyle \int _0^1\frac{1}{|z-1|}\ dz=\infty \), which implies that
$$\begin{aligned} \displaystyle \int _{0}^{1+\frac{1}{x}} \frac{\frac{1}{|z-1|^\beta }+\frac{1}{|z+1|^\beta }-2}{|z|^{1+2s}}\ dz\rightarrow \infty , \quad \text { as } x\rightarrow -\infty . \end{aligned}$$By L’Hospital rule, we know that
$$\begin{aligned} \lim _{x\rightarrow -\infty }\ \frac{\int _{0}^{1+\frac{1}{x}} \frac{\frac{1}{|z-1|}+\frac{1}{|z+1|}-2}{|z|^{1+2s}}\ dz}{\ln (-x)}&= \lim _{x\rightarrow -\infty }\frac{\left[ |x|+\frac{1}{2}-2\right] \cdot \left( -\frac{1}{x^2}\right) }{\frac{1}{x}}=1. \end{aligned}$$Hence we have
$$\begin{aligned} (-\Delta )^s \psi _1(x)=-\frac{C_{1,s}\ln |x|}{|x|^{2s+1}}+o\left( \frac{\ln |x|}{|x|^{2s+1}}\right) ,\quad \text {as } x\rightarrow -\infty . \end{aligned}$$
\(\square \)
Lemma 4.3
Consider the function
Then
Proof
-
a.
In fact, for all \(x\ge 2\), we have
$$\begin{aligned} (-\Delta )^s\phi (x)&= C_{1,s}\left[ \,\int _{-\infty }^{-x-1}\frac{\phi (x)-\phi (x\!+\!y)}{|y|^{1+2s}}\ dy\!+\!\mathrm{(P.V.)}\!\int _{-x-1}^\infty \frac{\phi (x)-\phi (x\!+\!y)}{|y|^{1+2s}}\ dy\!\!\right] \\&= -C_{1,s}\int ^{-x-1}_{-\infty } \frac{1}{|y|^{1+2s}}\ dy\\&= -\frac{C_{1,s}}{2s}\cdot \frac{1}{|x+1|^{2s}}\\&= -\frac{C_{1,s}}{2s}\cdot \frac{1}{|x|^{2s}}+o\left( \frac{1}{|x|^{2s}}\right) ,\quad \text {as } x\rightarrow \infty . \end{aligned}$$ -
b.
If \(x\le -2\), we have
$$\begin{aligned} (-\Delta )^s\phi (x)&= C_{1,s}\left[ \!\mathrm{(P.V.)}\!\int _{-\infty }^{-x-1}\frac{\phi (x)-\phi (x+y)}{|y|^{1+2s}}\ dy\!+\!\!\int _{-x-1}^\infty \frac{\phi (x)-\phi (x+y)}{|y|^{1+2s}}\ dy\!\!\right] \\&= C_{1,s}\int _{-x-1}^\infty \frac{1}{|y|^{1+2s}}\ dy\\&= \frac{C_{1,s}}{2s}\cdot \frac{1}{|x+1|^{2s}}\\&= \frac{C_{1,s}}{2s}\cdot \frac{1}{|x|^{2s}}+o\left( \frac{1}{|x|^{2s}}\right) ,\quad \text {as } x\rightarrow -\infty . \end{aligned}$$
\(\square \)
Below we show a form of the maximal principle which is a slight variation of those in [2, 6].
Lemma 4.4
(The Maximum Principle) Let \(H\) be a nonempty open subset of \(\mathbb {R}\), assume \(d(x)\ge 0\) for all \(x\in H\) and \(w\in C^1(\overline{H})\) satisfies
Then \(w(x)\ge 0\) for all \(x\) in \(\mathbb {R}\).
Proof
Assume \(w(x_0)<0\) for some \(x_0\in \mathbb {R}\), since \(w(x)\ge 0\) for all \(x\notin H\), \(\displaystyle \lim _{|x|\rightarrow \infty }\ w(x)=0\), and \(w\in C^1(\overline{H})\), then there exists some \(x_1\in H\) such that
Since \(x_1\) is a global minimum of \(w\) in \(\mathbb {R}\), \(x_1\in H\) and \(w\in C^1(H)\), then
Since \(d(x)\ge 0\) for all \(x\in H\), and \(x_1\in H\), so we have
which contradicts with the assumption. \(\square \)
The following two propositions give suitable lower and upper bounds of the asymptotic decay rates of \(u'\) and \(1-u\) at \(\infty \), which are expected to be a power of \(1+2s\) and \(2s\), respectively.
Proposition 4.2
Let \(\displaystyle \frac{1}{2}<s<1\) and \((\mu ,u)\) be a solution to (1.1) with \(\mu >0\). Assume that \(f'(1)<0\), then there exists some constant \(C>0\) such that
Proof
Since \(f'(1)<0\), there exists some \(m>0\) and \(\theta _0\in (0,1)\) such that \(f'(u)\le -m\) for all \(u\in [\theta _0,1]\). Let \(\epsilon >0\) be such that \(\displaystyle -\frac{C_{1,s}}{2s}+m\epsilon ^{-2s}=\frac{m}{2}\epsilon ^{-2s}\), that is, \(\displaystyle \epsilon =\left( \frac{sm}{C_{1,s}}\right) ^{\frac{1}{2s}}\). Consider
we know that
By Lemma 4.3 and Lemma 4.2, we know that
Hence we have
So there exists some large \(R>0\) such that
Up to a translation, without loss of generality, we assume \(u(0)=\theta _0\). Notice that \(v(x)=u'(x)>0\) in \(\mathbb {R}\) satisfies
Since \(\Psi (x)>0\) for all \(x\in \mathbb {R}\), there exists some \(C>0\) such that
Since \(\displaystyle \Psi (x)=\phi \left( x-\epsilon ^{-1}-1\right) =1\) for all \(\displaystyle x\le \epsilon ^{-1}\), we get \(\displaystyle C\Psi (x)=C\ge \Vert v\Vert _{C(\mathbb {R})}\) for all \(x\le \epsilon ^{-1}\). In summary, we know that
Let \(w(x)=C\Psi (x)-v(x)\) for all \(x\in \mathbb {R}\), we have
By Lemma 4.4, we have \(w(x)\ge 0\) in \(\mathbb {R}\), which implies that
\(\square \)
Proposition 4.3
Let \(\displaystyle \frac{1}{2}<s<1\), assume that \(f'(1)<0\), let \((\mu ,u)\) be a solution to (1.1) with \(\mu >0\). Then there exists some constant \(C>0\) such that
Proof
Since \(f'(1)<0\), there exists some \(m>0\) and \(\theta _0\in (0,1)\) such that \(f'(u)\le -m\) for all \(u\in [\theta _0,1]\). Let \(\epsilon >0\) be such that
That is, we have
By considering the function \(\Psi (x)=\psi _{2s}(\epsilon x-2)+\psi _{1+2s}(-\epsilon x)\) for all \(x\in \mathbb {R}\), we know that
By Lemma 4.1 and Lemma 4.2, we know that
So we get
Hence there exists some large \(R>0\) such that
Without loss of generality, we assume \(\displaystyle u\left( \epsilon ^{-1}\right) =\theta _0\), we know that \(v=u'\) satisfies
For all \(x\le \epsilon ^{-1}\), we have \(\epsilon x-2\le -1\) and \(-\epsilon x\ge -1\), which implies that
By Proposition (4.2), we know that there exists some constant \(C_1>0\) such that
Notice that for all \(\displaystyle x\ge \epsilon ^{-1}\), \(\displaystyle \Psi (x)\ge \psi _{1+2s}(-\epsilon x)>0\), which implies that there exists some \(C_2>0\) such that
Let \(C=\max \{C_1,C_2\}>0\) and \(w(x)=C\Psi (x)-v(x)\) for all \(x\in \mathbb {R}\), then
By Lemma 4.4, we know that \(w(x)\ge 0\) in \(\mathbb {R}\), which implies
The inequality for \(1-u(x)\) follows immediately. \(\square \)
Proposition 4.4
Let \(\displaystyle \frac{1}{2}<s<1\),let \((\mu ,u)\) be a solution to (1.1) with \(\mu >0\) in Theorem 3.2. Then there exists some constant \(C>0\) such that
Proof
We have shown in the proof of Theorem 3.2 that there exists some constant \(C>0\) such that
Now it suffices to show that there exists some constant \(C>0\) such that
Let \(\epsilon >0\) be such that \(\displaystyle -\frac{C_{1,s}}{2s-1}+2s\mu \epsilon ^{1-2s}=-\frac{C_{1,s}}{2(2s-1)}\), that is,
Let \(\Phi (x)=\psi _{2s}(\epsilon x)\) in \(\mathbb {R}\), then
By Lemma 4.2, we have
So we get
Therefore there exists some large \(R>0\) such that
Since \(f'(t)\ge 0\) for all \(t\in [0,\theta _0]\), without loss of generality, we may assume \(u(-\epsilon ^{-1})=\theta _0\). Notice that \(v(x)=u'(x)>0\) in \(\mathbb {R}\) satisfies
Since \(\Phi (x)=0\) for all \(x\ge -\epsilon ^{-1}\), we get \(\Phi (x)\le v(x)\) for all \(x\ge -\epsilon ^{-1}\). Since \(v(x)>0\) in \(\mathbb {R}\), there exists some \(C>1\) such that
Let \(w(x)=Cv(x)-\Phi (x)\) for all \(x\in \mathbb {R}\), we have
By Lemma 4.4, we have \(w(x)\ge 0\) in \(\mathbb {R}\), which implies that
\(\square \)
References
Aronson, D., Weinberger, H.: Nonlinear diffusion in population genetics, combustion, and nerve pulse propagation, partial differential equations and related topics, lecture notes in mathematics, pp. 4–49. Springer, Berlin (1975)
Cabré, X., Sire, Y.: Nonlinear equations for fractional Laplacians I: regularity, maximum principles, and hamiltonian estimates. ArXiv (2010)
Caffarelli, L., Silvestre, L.: An extension problem related to the fractional Laplacian. Commun. Partial Differ. Equ. 32, 1245–1260 (2007)
Cabré, X., Roquejoffre, J.: The influence of fractional diffusion in Fisher–KPP equations. Comm. Math. Phys. 320(3), 679–722 (2013)
Fisher, R.A.: The wave of advance of advantageous genes. Ann. Eugen. 7, 353–369 (1937)
Gui, C., Zhao, M.: Traveling Wave Solutions of Allen–Cahn Equation with a Fractional Laplacian, Ann. I. H. Poincaré-AN (2014). doi:10.1016/j.anihpc.2014.03.005
Kolmogorov, A., Petrovskii, I., Piskunov, N.: A study of the diffusion equation with increase in the amount of substance, and its application to a biological problem. Bull. Moscow Univ. Math. Ser. A 1, 1–25 (1937)
Landkof, N.S.: Foundations of Modern Potential Theory. In: Doohovskoy, A.P., (ed) Die Grundlehren der mathematischen Wissenschaften, Band 180. Translated from the Russian. Springer, New York, (1972)
Mellet, A., Roquejoffre, J., Sire, Y.: Existence and asymptotics of fronts in non local combustion models. Arxiv (2010)
Acknowledgments
This work was partially supported by a grant from the Simons Foundation (Award # 199305) and a NSF IPA award. The authors would also like to thank the anonymous referee for helpful suggesions for the revision of the manuscript.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by A. Malchiodi.
Rights and permissions
About this article
Cite this article
Gui, C., Huan, T. Traveling wave solutions to some reaction diffusion equations with fractional Laplacians. Calc. Var. 54, 251–273 (2015). https://doi.org/10.1007/s00526-014-0785-y
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00526-014-0785-y