This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Global boundedness and asymptotic behavior of time-space fractional nonlocal reaction-diffusion equation

Hui Zhan 2432593867@qq.com Fei Gao gaof@whut.edu.cn Liujie Guo 2252987468@qq.com Department of Mathematics and Center for Mathematical Sciences, Wuhan University of Technology, Wuhan, 430070, China
Abstract

The global boundedness and asymptotic behavior are investigate for the solution of time-space fractional non-local reaction-diffusion equation (TSFNRDE)

αutα=(Δ)su+μu2(1kJu)γu,(x,t)N×(0,+),\frac{\partial^{\alpha}u}{\partial t^{\alpha}}=-(-\Delta)^{s}u+\mu u^{2}(1-kJ*u)-\gamma u,\qquad(x,t)\in\mathbb{R}^{N}\times(0,+\infty),

where s(0,1),α(0,1),N2s\in(0,1),\alpha\in(0,1),N\leq 2. The operator tα\partial_{t}^{\alpha} is the Caputo fractional derivative, which (Δ)s-(-\Delta)^{s} is the fractional Laplacian operator. For appropriate assumptions on JJ, it is proved that for homogeneous Dirichlet boundary condition, this problem admits a global bounded weak solution for N=1N=1, while for N=2N=2, global bounded weak solution exists for large kk values by Gagliardo-Nirenberg inequality and fractional differential inequality. With further assumptions on the initial datum, for small μ\mu values, the solution is shown to converge to 0 exponentially or locally uniformly as tt\rightarrow\infty. Furthermore, under the condition of J1J\equiv 1, it is proved that the nonlinear TSFNRDE has a unique weak solution which is global bounded in fractional Sobolev space with the nonlinear fractional diffusion terms (Δ)sum(22N<m<1)-(-\Delta)^{s}u^{m}\,(2-\frac{2}{N}<m<1).

keywords:
Time-space fractional reaction-diffusion equation , Global boundedness , Asymptotic Behavior , Non-local , Nonlinear

1 Introduction

In this work we study the time-space fractional non-local reaction-diffusion equation

αutα\displaystyle\frac{\partial^{\alpha}u}{\partial t^{\alpha}} =\displaystyle= (Δ)su+μu2(1kJu)γu,\displaystyle-(-\Delta)^{s}u+\mu u^{2}(1-kJ*u)-\gamma u, (1.1)
u(x,0)\displaystyle u(x,0) =\displaystyle= u0(x),xN,\displaystyle u_{0}(x),\quad x\in\mathbb{R}^{N}, (1.2)

with (x,t)N×(0,+),0<α<1,N2,μ,k>0,γ>1(x,t)\in\mathbb{R}^{N}\times(0,+\infty),0<\alpha<1,N\leq 2,\mu,k>0,\gamma>1. According to [1], The nonlocal operator (Δ)s,\left(-\Delta\right)^{s}, known as the Laplacian of order ss, is defined for any function gg in the Schwartz class through the Fourier transform: if (Δ)sg=h,\left(-\Delta\right)^{s}g=h, then

h^(ξ)=|ξ|2sg^(ξ).\hat{h}(\xi)=\left|\xi\right|^{2s}\hat{g}(\xi). (1.3)

If 0<s<10<s<1, by [2], we can also use the representation by means of a hypersingular kernel,

(Δ)su=CN,sP.V.Nu(x)u(y)|xy|N+2s𝑑y,\displaystyle(-\Delta)^{s}u=C_{N,s}P.V.\int_{\mathbb{R}^{N}}\frac{u(x)-u(y)}{\left|x-y\right|^{N+2s}}dy, (1.4)

where CN,s=4ssΓ(N2+s)Γ(1s)πN2,C_{N,s}=\frac{4^{s}s\Gamma(\frac{N}{2}+s)}{\Gamma(1-s)\pi^{\frac{N}{2}}}, and P.V.P.V. is the principal value of Cauchy. considering the Sobolev space

Hs(N)={uL2(N):NN|u(x)u(y)|(|xy|)N+2s𝑑x𝑑y<}.H^{s}(\mathbb{R}^{N})=\left\{u\in L^{2}(\mathbb{R}^{N}):\int_{\mathbb{R}^{N}}\int_{\mathbb{R}^{N}}\frac{\left|u(x)-u(y)\right|}{(\left|x-y\right|)^{N+2s}}dxdy<\infty\right\}.

And tα\partial_{t}^{\alpha} denote the left Caputo fractional derivative that is usually defined by the formula

tαu(x,t)=1Γ(1α)0tus(x,s)(ts)α𝑑s,0<α<1\partial_{t}^{\alpha}u(x,t)=\frac{1}{\Gamma(1-\alpha)}\int_{0}^{t}\frac{\partial u}{\partial s}(x,s)(t-s)^{-\alpha}ds,\quad 0<\alpha<1\quad (1.5)

in the formula 1.5, the left Caputo fractional derivative tαu\partial_{t}^{\alpha}u is a derivative of the order in [3]. By [4] , Here J(x)J(x) is a competition kernel with

0JL1(N),NJ(x)𝑑x=1,infNJ>η0\leq J\in L^{1}(\mathbb{R}^{N}),\quad\int_{\mathbb{R}^{N}}J(x)dx=1,\quad\inf\limits_{\mathbb{R}^{N}}J>\eta (1.6)

for some η>0\eta>0, and

Ju(x,t)=NJ(xy)u(y,t)𝑑y,J*u(x,t)=\int_{\mathbb{R}^{N}}J(x-y)u(y,t)dy,

and F(x,t)=μu2(1kJu)γuF(x,t)=\mu u^{2}(1-kJ*u)-\gamma u is represented as sexual reproduction in the population dynamics system, where γu-\gamma u is population mortality. By [5], assume {Φj}j1\left\{\Phi_{j}\right\}_{j\geq 1} is an orthonormal eigenbasis in L2(Ω),ΩNL^{2}(\Omega),\Omega\subset\mathbb{R}^{N}, associated with the eigenvalues {λj}j1\left\{\lambda_{j}\right\}_{j\geq 1} such that 0<λ1λ2λj,limjλj=.0<\lambda_{1}\leq\lambda_{2}\leq\cdots\leq\lambda_{j}\leq\cdots,\lim_{j\rightarrow\infty}\lambda_{j}=\infty. For all 0<s<10<s<1, we denote by D((Δ)s)D((-\Delta)^{s}) the space defined by

D((Δ)s):={uL2(Ω):j=1λj2s|(u,Φj)|2<},D((-\Delta)^{s}):=\left\{u\in L^{2}(\Omega):\sum_{j=1}^{\infty}\lambda_{j}^{2s}\left|(u,\Phi_{j})\right|^{2}<\infty\right\},

then if uD((Δ)s)u\in D((-\Delta)^{s}), we define the operator (Δ)s(-\Delta)^{s} by

(Δ)su:=j=1λjs(u,Φj)Φj,(-\Delta)^{s}u:=\sum_{j=1}^{\infty}\lambda_{j}^{s}(u,\Phi_{j})\Phi_{j},

which (Δ)s:D((Δ)s)L2(Ω)(-\Delta)^{s}:D((-\Delta)^{s})\rightarrow L^{2}(\Omega), with the following equivalence

uD((Δ)s)=(Δ)suL2(Ω).\left\|u\right\|_{D((-\Delta)^{s})}=\left\|(-\Delta)^{s}u\right\|_{L^{2}(\Omega)}.

Suppose {λjs,Φj}\left\{\lambda_{j}^{s},\Phi_{j}\right\} be the eigenvalues and corresponding eigenvectors of the Laplacian operator Δ-\Delta in ΩN\Omega\in\mathbb{R}^{N} with Dirichlet boundary condition on Ω\partial\Omega:

{ΔΦj=λjsΦj,inΩ,Φj=0,onΩ.\displaystyle\left\{\begin{matrix}-\Delta\Phi_{j}=\lambda_{j}^{s}\Phi_{j},&\text{in}\quad\Omega,\\ \Phi_{j}=0,&\text{on}\quad\partial\Omega.\end{matrix}\right.

Therefore, we will study the solution for equation (1.1)-(1.2) is global boundedness and asymptotic behavior under the above Dirichlet boundary condition in one dimensional space and two dimensional space.

Fractional calculus has gained considerable importance due to its application in various disciplines such as physics, mechanics, chemistry, engineering, etc[6, 7]. Therefore, fractional-order ordinary and partial differential equations have been widely studied by many authors,see [8, 9]. In the process of practical application, researchers found that the solution of fractional differential equation has a lot of properties, and when the solution of the fractional differential equation is proof and analyzed, Laplace transform [10], Fourier transform [11] and Green function method [12]. At present, the research of fractional diffusion equations has become a new field of active research. More than a dozen universities and research institutes at home and abroad have engaged in the research of time fractional, space fractional, time-space fractional diffusion equations and special functions related to it such to Wright, Mittag-Leffler function and so on.

Let us first recall some previous results on fractional diffusion equation. Since there is a large amount of papers for these equations, we mention the ones related to our results.

When α=1\alpha=1 and s=1s=1, then problem (1.1)-(1.2) reduces to the following non-local reaction-diffusion equation in [4]

ut=Δu+μu2(1kJu)γu,\frac{\partial u}{\partial t}=\Delta u+\mu u^{2}(1-kJ*u)-\gamma u,

it denotes by u(x,t)u(x,t) the density of individuals having phenotype xx at time tt and formulate the dynamics of the population density. However, with the further deepening of scientific research, compared with the traditional reaction-diffusion equation, the non-local reaction-diffusion equation has new mathematical characteristics and richer nonlinear dynamic properties [13]. From [14], Kolmogorov and Fish research the following reaction-diffusion equations

ut=Δu+u(1u),x,\frac{\partial u}{\partial t}=\Delta u+u(1-u),\quad x\in\mathbb{R},

in order to describe the row phenomenon of foreign invasion species and animals the transmission process of excellent genes in an infinite habitat. Through the widespread research of the diffusion equations, they found that there are many important applications in many fields such as spatial ecology, evolution of species, and disease dissemination of the non-local reaction-diffusion equation [15, 16, 17].

When s=1s=1 and 0<α<10<\alpha<1, the fractional operator (Δ)s-(-\Delta)^{s} become the standard Laplacian Δ\Delta, which is a time fractional non-local reaction-diffusion equation. Many researchers have studied the corresponding property for solution of reaction-diffusion model with Caputo fractional derivative. In order to study the global boundary of the solution of fractional linear reaction-diffusion equation, [18] uses the maximum regularity to verify the existence of such equations. In general space, we often use the convolutional definition of the Caputo fractional derivative. However, in [19], from the perspective of micro -division operator theory, studied the following time fractional reaction-diffusion equation in fractional Sobolev spaces

tαu(x,t)=Lu(x,t)+F(x,t),xΩn,0<tT\partial_{t}^{\alpha}u(x,t)=-Lu(x,t)+F(x,t),\quad x\in\Omega\subset\mathbb{R}^{n},0<t\leq T

where LL is a differential operator of elliptic type and tα\partial_{t}^{\alpha} denotes the left Caputo fractional derivative that is usually defined by (1.5). In [3], the time fractional reaction-diffusion model with non-local boundary conditions is used to predict the invasion of tumors and its growth. In addition, the Faedo-Galerkin method has also been used to verify that the model has the only weak solution. From [20], Ahmad considers the following time fractional reaction-diffusion equation with boundary conditions

Dαcu=Δuu(1u),xΩ,t>0{}^{c}\textrm{D}^{\alpha}u=\Delta u-u(1-u),\quad x\in\Omega,t>0

with the initial condition u(x,0)=u0(x),xΩu(x,0)=u_{0}(x),x\in\Omega. the initial data u0(x)u_{0}(x) is a given positive and bounded function. Moreover, [20, 21, 22] research the existence, giobal boundary and blow-up of time fractional reaction-diffusion equation with boundary conditions. The study of time fractional diffusion equations has recently attracted a lot of attention. It is worth mentioning that we often use time fractional Duhamel principle [23] and Laplace transform to obtain the solution of time fractional PDE.

When α=1\alpha=1 and 0<s<10<s<1, then problem (1.1)-(1.2) reduces to the non-local reaction-diffusion equation with fractional Laplacian operator, which is called fractional reaction-diffusion equation. Over the past few decades, due to the existence and non-existential research significance of fractional Laplacian equation, which has attracted many scholars to invest in it. For example, [24] study Liouville theorem of fractional Laplacian equation to verify existence/non-existential of the solution. In particular, scholars conducted a lot of research on fractional reaction-diffusion equation and obtained many important results. [25] establish the global existence of weak solutions of non-local energy-weighted fractional reaction–diffusion equations for any bounded smooth domain. In [26], considering the asymptoticity of the following nonlinear non-autonomous fractional reaction–diffusion equations

ut(x,t)=(Δ)α/2h(t)F(u),xN,t>0,\displaystyle u_{t}(x,t)=-(-\Delta)^{\alpha/2}-h(t)F(u),\quad x\in\mathbb{R}^{N},t>0,
u(x,0)=u0(x)0,xN.\displaystyle u(x,0)=u_{0}(x)\geq 0,\quad x\in\mathbb{R}^{N}.

Among them, the initial function u0L1(N)L(N),h:[0,)[0,)u_{0}\in L^{1}(\mathbb{R}^{N})\cap L^{\infty}(\mathbb{R}^{N}),h:[0,\infty)\rightarrow[0,\infty) is a continuous function. [27] study the existence and blow-up of solution of semilinear reaction-diffusion system with the fractional Laplacian. In addition, [28] studied the blow-up and asymptoticity of solution of the following nitial value problem for the reaction–diffusion equation with the anomalous diffusion

tu=(Δ)α+λup,xN,t>0,\displaystyle\partial_{t}u=-(-\Delta)^{\alpha}+\lambda u^{p},\quad x\in\mathbb{R}^{N},t>0,
u(x,0)=u0(x),\displaystyle u(x,0)=u_{0}(x),

where 0<α2,λ{1,1}0<\alpha\leq 2,\lambda\in\left\{-1,1\right\} and p>1p>1. The fractional powers of the classical Laplace operator, namely (Δ)s-(\Delta)^{s} are particul cases of the infinitesimal generators of Le´\acute{e}vy stable diffusion processes and appear in anomalous diffusions in plasmas, flames propagation and chemical reactions in liquids, population dynamics, geophysical fluid dynamics [29, 30].

The tractional diffusion equation tu=Δu\partial_{t}u=\Delta u describes a cloud of spreading particles at the macroscopic level. The space-time fractional diffusion equation tβu=(Δu)α/2\partial_{t}^{\beta}u=-(-\Delta u)^{\alpha/2} with 0<β<10<\beta<1 and 0<α<20<\alpha<2 is used to model anomalous diffusion [31]. The fractional derivative in time is used to describe particle sticking and trapping phenomena and the fractional space derivative is used to model long particle jumps [31]. These two effects combined together produces a concentration profile with a sharper peak, and heavier tails [32]. In [33], basing on a new unique continuation principle for the eigenvalues problem associated with the fractional Laplace operator subject to the zero exterior boundary condition, it study the controllability of the space-time fractional diffusion equation. [34] concerned with boundary stabilization and boundary feed-back stabilization for time-space fractional diffusion equation. In these applications, it is often important to consider boundary value problems. Hence it is useful to develop solutions for space–time fractional diffusion equations on bounded domains with Dirichlet boundary conditions. This paper will consider the following Dirichlet boundary condition [35, 36].

In the classic diffusion equation, the time derivation of the Integer into a time derivation is turned into a time fractional diffusion equation, which is usually used to describe the ultra -diffusion and secondary diffusion phenomenon. However, in some practical situations, part of boundary data, or initial data, or diffusion coefficient, or source term may not be given and we want to find them by additional measurement data which will yield some fractional diffusion inverse problems. [35] anylyzed the inverse source problem for the following space-time fractional diffusion equation by setting up an operator equation

βtβu(t,x)=rβ(Δ)α/2u(t,x)+f(x)h(t,x),(t,x)ΩT,\displaystyle\frac{\partial^{\beta}}{\partial t^{\beta}}u(t,x)=-r^{\beta}(-\Delta)^{\alpha/2}u(t,x)+f(x)h(t,x),\quad(t,x)\in\Omega_{T},
u(t,1)=u(t,1)=0,0<t<T,\displaystyle u(t,-1)=u(t,1)=0,\quad 0<t<T,
u(0,x)=0,xΩ,\displaystyle u(0,x)=0,\quad x\in\Omega,

where (t,x)ΩT:=(0,T)×Ω(t,x)\in\Omega_{T}:=(0,T)\times\Omega and Ω=(1,1)\Omega=(-1,1). For the time fractional diffusion equations cases, the uniqueness of inverse source problems have been widely studied. Basing on the eigenfunction expansion, [37] established the unique existence of the weak solution and the asymptotic behavior as the time tt goes to \infty for fractional diffusion-wave equation. Li and Wei, in [38] proved the existence and uniqueness of a weak solution for the following time-space fractional diffusion equation

0+αu(x,t)=(Δ)β2u(x,t)+f(x)p(t),(x,t)ΩT,\displaystyle\partial_{0+}^{\alpha}u(x,t)=-(-\Delta)^{\frac{\beta}{2}}u(x,t)+f(x)p(t),\quad(x,t)\in\Omega_{T},
u(x,0)=ϕ(x),xΩ¯,\displaystyle u(x,0)=\phi(x),\quad x\in\bar{\Omega},
u(x,t)=0xΩ,t(0,T],\displaystyle u(x,t)=0\quad x\in\partial\Omega,t\in(0,T],

where ΩT:=(0,T)×Ω,Ωd\Omega_{T}:=(0,T)\times\Omega,\Omega\subset\mathbb{R}^{d} and α(0,1),β(1,2)\alpha\in(0,1),\beta\in(1,2). This paper uses the method of proof Theorem 3.2 in [38] to proof global existence and uniqueness of a weak solution. In [39], Wei and Zhang solved an inverse space-dependent source problem by a modified quasi-boundary value method. Wei et al. in [40] identified a time-dependent source term in a multidimensional time-fractional diffusion equation from the boundary Cauchy data. Compared with the problem of classical inverse initial value, the recovery of the fractional inverse initial value is easier and more stable. Through the above two paragraphs, this paper will use the method of verifying the initial value of the inverse initial value under the Dirichlet boundary conditions to verify the existence and uniqueness of time and space fractional non-local reaction diffusion equations.

To the best of our knowledge, until recently there has been still very little works on deal with the existence, decay estimates and blow-up of solutions for time-space fractional diffusion equations. [41] studied the global and local existence, blow-up of solutions of the following time-space fractional diffusion problem by applying the Galerkin method

tβu+(Δ)αu+(Δ)βtβu=λf(x,u)+g(x,t),inΩ×+,\displaystyle\partial_{t}^{\beta}u+(-\Delta)^{\alpha}u+(-\Delta)^{\beta}\partial_{t}^{\beta}u=\lambda f(x,u)+g(x,t),\quad in\,\Omega\times\mathbb{R}^{+},
u(x,t)=0,in(NΩ)×+,\displaystyle u(x,t)=0,\quad in\,(\mathbb{R}^{N}\setminus\Omega)\times\mathbb{R}^{+},
u(x,0)=u0(x),inΩ,\displaystyle u(x,0)=u_{0}(x),\quad in\,\Omega,

where ΩN,0<α<1,0<β<1\Omega\subset\mathbb{R}^{N},0<\alpha<1,0<\beta<1 is a bounded domain with Lipschitz boundary. [42] studys the blow-up, and global existence of solutions to the time-space fractional diffusion problem and give an upper bound estimate of the life span of blowing-up solutions. In [5], By transforming the time-space fractional diffusion equations into an operator equation it investigates the existence, the uniqueness and the instability for the problem. In addition, [40] study the existence and uniqueness of a weak solution of following time-space fractional diffusion equation with homogeneous Dirichlet boundary

0+αu(x,t)=(Δ)β2u(x,t)+f(x)p(x),(x,t)ΩT,\partial_{0+}^{\alpha}u(x,t)=-(-\Delta)^{\frac{\beta}{2}}u(x,t)+f(x)p(x),\quad(x,t)\in\Omega_{T},

where ΩT:=Ω×(0,T],Ωd\Omega_{T}:=\Omega\times(0,T],\Omega\subset\mathbb{R}^{d} and α(0,1),β(1,2)\alpha\in(0,1),\beta\in(1,2) are fractional orders of the time and space derivatives, respectively, T>0T>0 is a fixed final time. Moreover, [43] consider the following time-space non-local fractional reaction-diffusion equation

tαu(x,t)+(Δ)Ωsu(x,t)=u(1u),xΩ,t>0,\displaystyle\partial_{t}^{\alpha}u(x,t)+(-\Delta)_{\Omega}^{s}u(x,t)=-u(1-u),\quad x\in\Omega,t>0, (1.7)
u=0,xnΩ,t>0\displaystyle u=0,\quad x\in\mathbb{R}^{n}\setminus\Omega,t>0
u(x,0)=u0(x),xΩ.\displaystyle u(x,0)=u_{0}(x),\quad x\in\Omega.

And for realistic initial conditions, studying global existence, blow-up in a finite time, asymptotic behavior of bounded solutions of equation (1.7). In the process of blow-up of TSFNRDE, this paper uses the method of the proof of [blow-up, [43]]. However, in addition to the existence and blow-up of time-space fractional diffusion equation mentioned above, we also further study the global boundedness and asymptotic behavior of solution for TSFNRDE.

The outline of this paper is as follows.The global boundedness for the solution of TSFNRDE is analyzed in section 2: First, we introduce the existence and uniqueness of solutions of TSFNRDE with homogeneous Dirichlet boundary condition by Appendix A.2. Secondly, The blow-up for the solution of TSFNRDE is analyzed in Lemma 2.16: for the initial data, blow-up in a finite time TmaxT_{max} satisfies bi-lateral estimate and when tTmaxt\rightarrow T_{max}, then u(,t)\left\|u(\cdot,t)\right\|\rightarrow\infty. Finally, we use the analytical formula for the solution of TSFNRDE and use the fractional Sobolev inequality, Sobolev embedding inequality conversion to the solution of equation to obtain the global boundedness of equation(1.1)-(1.2). In section 3, we introduce the asymptotic behavior of solutions of TSFNRDE in Hilbert space. In section 4, under the condition of J=k=μ=γ=1J=k=\mu=\gamma=1, we first verify existence unique weak solution for problem (1.8)-(1.9). Next, we mainly use Gagliardo-Nirenberg inequality to prove that nonlinear TSFNRDE in LrL^{r} estimates and LL^{\infty} estimates, which is global boundedness of solution in fractional Sobolev space with the nonlinear fractional diffusion terms (Δ)sum-(-\Delta)^{s}u^{m}.

Theorem 1.1.

Suppose (1.6) holds, 0<u0D((Δ)s)L(N).0<u_{0}\in D((-\Delta)^{s})\cap L^{\infty}(\mathbb{R}^{N}). Denote k=0k^{*}=0 for N=1N=1 and k=(μCGN2+1)η1k^{*}=(\mu C_{GN}^{2}+1)\eta^{-1} for N=2N=2, CGNC_{GN} is the constant appears in Gagliardo-Nirenberg inequality in Lemma 2.12, then for any k>k,T>0k>k^{*},T>0, the nonnegative weak solution of (1.1)-(1.2) exists and is globally bounded in time, that is, there exist

K={K(u0L(N),μ,η,k,CGN,T),N=1,K(u0L(N),μ,CGN,T),N=2,K=\begin{cases}K(\Arrowvert u_{0}\Arrowvert_{L^{\infty}(\mathbb{R}^{N})},\mu,\eta,k,C_{GN},T),&N=1,\\ K(\Arrowvert u_{0}\Arrowvert_{L^{\infty}(\mathbb{R}^{N})},\mu,C_{GN},T),&N=2,\end{cases}

such that

0u(x,t)K,(x,t)N×(0,).0\leq u(x,t)\leq K,\qquad\forall(x,t)\in\mathbb{R}^{N}\times(0,\infty).

Theorem 1.2.

Denote u(x,t)u(x,t) the globally bounded solution of (1.1)-(1.2).

  1. 1.

    For any γ>1\gamma>1, there exist μ>0\mu^{*}>0 and m>0m^{*}>0 such that for μ(0,μ)\mu\in(0,\mu^{*}) and u0L(N)<m\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}<m^{*}, we have

    u(x,t)L(N×[0,+))<τμ,\left\|u(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N}\times[0,+\infty))}<\frac{\tau}{\mu},

    and thus

    u(x,t)L(N)u0L(N)e((λ1s+σ))1αt,\left\|u(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N})}\leq\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}e^{(-(\lambda_{1}^{s}+\sigma))^{\frac{1}{\alpha}t}},

    for all t>0,0<α,s<1t>0,0<\alpha,s<1 with σ:=γμu(x,t)L(N×[0,+))>0.\sigma:=\gamma-\mu\left\|u(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N}\times[0,+\infty))}>0.

  2. 2.

    If 1<γ<μ4k1<\gamma<\frac{\mu}{4k}, there exist μ>0\mu^{**}>0 and m>0m^{**}>0 such that for μ(0,μ)\mu\in(0,\mu^{**}) and u0L(N)<m\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}<m^{**}, we have

    u(x,t)L(N×[0,+))<a,\left\|u(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N}\times[0,+\infty))}<a,

    and thus

    limtu(x,t)=0\lim_{t\rightarrow\infty}u(x,t)=0

    locally uniformly in N\mathbb{R}^{N}.

Theorem 1.3.
αutα\displaystyle\frac{\partial^{\alpha}u}{\partial t^{\alpha}} =\displaystyle= (Δ)sum+u2(1Nu𝑑x)u(x,t)\displaystyle-(-\Delta)^{s}u^{m}+u^{2}(1-\int_{\mathbb{R}^{N}}udx)-u(x,t) (1.8)
u(x,0)\displaystyle u(x,0) =\displaystyle= u0(x)xN\displaystyle u_{0}(x)\qquad x\in\mathbb{R}^{N} (1.9)

with (x,t)N×(0,T](x,t)\in\mathbb{R}^{N}\times(0,T]. Then the problem (1.8)-(1.9) has a unique weak solution

uL2((0,T];H01(N))C((0,T];L1(N)).u\in L^{2}((0,T];H_{0}^{1}(\mathbb{R}^{N}))\cap C((0,T];L^{1}(\mathbb{R}^{N})).

Moreover, If 22N<m<1,0<α,s<12-\frac{2}{N}<m<1,0<\alpha,s<1 and initial value 0u0L1(N)L(N)0\leq u_{0}\in L^{1}(\mathbb{R}^{N})\cap L^{\infty}(\mathbb{R}^{N}), the solution uu is globally bounded. There exist M>0M>0 such that

0u(x,t)M,(x,t)N×(0,T].0\leq u(x,t)\leq M,\qquad(x,t)\in\mathbb{R}^{N}\times(0,T].

2 Global boundedness of solutions for TSFNRDE

Definition 2.1.

[44] The Mittag-Leffler function in two parameters is defined as

Eα,β(z)=k=0zkΓ(αk+β),zE_{\alpha,\beta}(z)=\sum_{k=0}^{\infty}\frac{z^{k}}{\Gamma(\alpha k+\beta)},\qquad z\in\mathbb{C}

where α>0,β>0,\alpha>0,\beta>0,\mathbb{C} denote the complex plane.

Lemma 2.1.

[45] If 0<α<1,η>00<\alpha<1,\eta>0, then there is 0Eα,α(η)1Γ(α)0\leq E_{\alpha,\alpha}(-\eta)\leq\frac{1}{\Gamma(\alpha)}. In addition, for η>0\eta>0, Eα,α(η)E_{\alpha,\alpha}(-\eta) is a monotonically decreasing function.

Lemma 2.2.

[46] If 0<α<1,t>0,w>00<\alpha<1,t>0,w>0, for Mittag-Leffler function Eα,1(wtα)E_{\alpha,1}(wt^{\alpha}) , then there is a constant CC such that

Eα,1(wtα)Cew1αt.E_{\alpha,1}(wt^{\alpha})\leq Ce^{w^{\frac{1}{\alpha}t}}. (2.1)

Lemma 2.3.

[44] Assume 0<α<20<\alpha<2, for any β\beta\in\mathbb{R}, there is a constant μ\mu such that πα2<μ<min{π,πα}\frac{\pi\alpha}{2}<\mu<min\left\{\pi,\pi\alpha\right\}, then there is a constant c=c(α,β,μ)>0c=c(\alpha,\beta,\mu)>0 , such that

|Eα,β(z)|c1+|z|,μ|arg(z)|π.\left|E_{\alpha,\beta}(z)\right|\leq\frac{c}{1+\left|z\right|},\qquad\mu\leq\left|arg(z)\right|\leq\pi.

Lemma 2.4.

[45] If 0<α<1,t>00<\alpha<1,t>0, then there is 0<Eα,1(t)<10<E_{\alpha,1}(-t)<1. In addition, Eα,1(t)E_{\alpha,1}(-t) is completely monotonous that is

(1)ndndtnEα,1(t)0,n(-1)^{n}\frac{d^{n}}{dt^{n}}E_{\alpha,1}(-t)\geq 0,\quad\forall n\in\mathbb{N}

Lemma 2.5.

[47] Let 0<α<10<\alpha<1 and uC([0,T],N),uL1(0,T;N)u\in C([0,T],\mathbb{R}^{N}),{u}^{\prime}\in L^{1}(0,T;\mathbb{R}^{N}) and uu be monotone. Then

v(t)tαv(t)12tαv2(t),t(0,T].v(t)\partial_{t}^{\alpha}v(t)\geq\frac{1}{2}\partial_{t}^{\alpha}v^{2}(t),\quad t\in(0,T]. (2.2)

Lemma 2.6.

[47] Let 0s1,xN0\leq s\leq 1,x\in\mathbb{R}^{N} and uC02(N)u\in C_{0}^{2}\left(\mathbb{R}^{N}\right). Then the following inequality holds

2u(Δ)su(x)(Δ)su2(x).2u(-\Delta)^{s}u(x)\geq(-\Delta)^{s}u^{2}(x). (2.3)

Lemma 2.7.

[48] Suppose u:[0,)×Nu:[0,\infty)\times\mathbb{R}^{N}\rightarrow\mathbb{R},the left Caputo fractional derivative with respect to time tt of uu is defined by (1.5). Then there is

Ntαudy=tαNu𝑑y.\int_{\mathbb{R}^{N}}\partial_{t}^{\alpha}udy=\partial_{t}^{\alpha}\int_{\mathbb{R}^{N}}udy. (2.4)

Definition 2.2.

[49] Assume that XX is a Banach space and let u:[0,T]Xu:\left[0,T\right]\rightarrow X. The Caputo fractional derivative operators of uu is defined by

Dtα0Cu(t){}_{0}^{C}\textrm{D}_{t}^{\alpha}u(t) =\displaystyle= 1Γ(1α)0t(ts)αddsu(s)𝑑s.\displaystyle\frac{1}{\Gamma(1-\alpha)}\int_{0}^{t}(t-s)^{-\alpha}\frac{d}{ds}u(s)ds. (2.5)

where Γ(1α)\Gamma(1-\alpha) is the Gamma function. The above integrals are called the left-sided and the right-sided the Caputo fractional derivatives.

Lemma 2.8.

[47] Let 0<α<10<\alpha<1 and uC([0,T],N),uL1(0,T;N)u\in C([0,T],\mathbb{R}^{N}),{u}^{\prime}\in L^{1}(0,T;\mathbb{R}^{N}) and uu be monotone. And when n2n\geq 2,the Caputo fractional derivative with respect to time tt of uu is defined by (2.5). Then there is

un1(0CDtαu)1n(0CDtαun).u^{n-1}(_{0}^{C}\textrm{D}_{t}^{\alpha}u)\geq\frac{1}{n}(_{0}^{C}\textrm{D}_{t}^{\alpha}u^{n}).

Lemma 2.9.

[48] Suppose u:[0,)×Nu:[0,\infty)\times\mathbb{R}^{N}\rightarrow\mathbb{R},the Caputo fractional derivative with respect to time tt of uu is defined by (2.5). Then there is

N(0CDtαu)dy=0CDtαNudy.\int_{\mathbb{R}^{N}}(_{0}^{C}\textrm{D}_{t}^{\alpha}u)dy=_{0}^{C}\textrm{D}_{t}^{\alpha}\int_{\mathbb{R}^{N}}udy.

Lemma 2.10.

[48] Suppose that a nonnegative function y(t)0y(t)\geq 0 satisfies

Dtα0Cy(t)+c1y(t)b{}_{0}^{C}\textrm{D}_{t}^{\alpha}y(t)+c_{1}y(t)\leq b

for almost all t[0,T]t\in[0,T], where b,c1>0b,c_{1}>0 are all contants. then

y(t)y(0)+bTααΓ(α).y(t)\leq y(0)+\frac{bT^{\alpha}}{\alpha\Gamma(\alpha)}.

Definition 2.3.

[50] Let Hs(N)H^{s}(\mathbb{R}^{N}) denote the completion of C0(N)C_{0}^{\infty}(\mathbb{R}^{N}) with respect to the Gagliardo norm

[u]Hs=(NN|u(x)u(y)|2|xy|N+2s𝑑x𝑑y)1/2.[u]_{H^{s}}=(\int_{\mathbb{R}^{N}}\int_{\mathbb{R}^{N}}\frac{\left|u(x)-u(y)\right|^{2}}{\left|x-y\right|^{N+2s}}dxdy)^{1/2}.

Remark 2.1.

The embedding Hs(N)LrH^{s}(\mathbb{R}^{N})\hookrightarrow L^{r} is continuous, that is

uLr(N)C[u]Hs(N)\left\|u\right\|_{L^{r}(\mathbb{R}^{N})}\leqslant C_{*}\left[u\right]_{H^{s}(\mathbb{R}^{N})} (2.6)

for all uHs(N)u\in H^{s}(\mathbb{R}^{N}), and C=c(N)s(1α)(N2α)C_{*}=c(N)\frac{s(1-\alpha)}{(N-2\alpha)} by Theorem 1 of [51].

Lemma 2.11.

[52] Assume 1p<N1\leq p<N, then fCc(N)\forall f\in C_{c}^{\infty}(\mathbb{R}^{N}), there is

fLq(N)C1fLp(N).\left\|f\right\|_{L^{q}(\mathbb{R}^{N})}\leq C_{1}\left\|\bigtriangledown f\right\|_{L^{p}(\mathbb{R}^{N})}. (2.7)

Which is q=NpNpq=\frac{Np}{N-p}, and C1C_{1} only dependent on p,Np,N.

Lemma 2.12.

[53] Let Ω\Omega be an open subset of N\mathbb{R}^{N}, assume that 1p,q1\leq p,q\leq\infty with (Nq)p<Nq(N-q)p<Nq and r(0,p)r\in(0,p). Then there exists constant CGN>0C_{GN}>0 only depending on q,rq,r and Ω\Omega such that for any uW1.q(Ω)Lp(Ω)u\in W^{1.q}(\Omega)\cap L^{p}(\Omega)

Ωup𝑑xCGN(uLq(Ω)(λp)uLp(Ω)(1λ)p+uLp(Ω)p)\int_{\Omega}u^{p}dx\leq C_{GN}(\Arrowvert\bigtriangledown u\Arrowvert^{(\lambda^{*}p)}_{L^{q}(\Omega)}\Arrowvert u\Arrowvert_{L^{p}(\Omega)}^{(1-\lambda^{*})p}+\Arrowvert u\Arrowvert_{L^{p}(\Omega)}^{p})

holds with

λ=NrNp1Nq+Nr(0,1).\lambda^{*}=\frac{\frac{N}{r}-\frac{N}{p}}{1-\frac{N}{q}+\frac{N}{r}}\in(0,1).

Definition 2.4.

[44] The Mittag-Leffler function in two parameters is defined as

Eα,β(z)=k=0zkΓ(αk+β),zE_{\alpha,\beta}(z)=\sum_{k=0}^{\infty}\frac{z^{k}}{\Gamma(\alpha k+\beta)},\qquad z\in\mathbb{C}

where α>0,β>0,\alpha>0,\beta>0,\mathbb{C} denote the complex plane.

Lemma 2.13.

[54] a,b0\forall a,b\geq 0 and ε>0\varepsilon>0, for 1<p,q<,1p+1q=11<p,q<\infty,\frac{1}{p}+\frac{1}{q}=1, then there is

abεapp+εqpbqq.a\cdot b\leq\varepsilon\frac{a^{p}}{p}+\varepsilon^{-\frac{q}{p}}\frac{b^{q}}{q}.

Definition 2.5.

[5] A function uC([0,T];L2(Ω))L2(0,T;D((Δ)s))u\in C([0,T];L^{2}(\Omega))\cap L^{2}(0,T;D((-\Delta)^{s})) is said to be a weak solution to equation (1.1)-(1.2) if the folloing conditions hold

  • 1.

    tαu(t)+(Δ)su(t)=F(t)\partial_{t}^{\alpha}u(t)+(-\Delta)^{s}u(t)=F(t) holds in L2(Ω)L^{2}(\Omega) for t(0,T]t\in(0,T];

  • 2.

    tαu(t)C((0,T];L2(Ω))L2(0,T;L2(Ω))\partial_{t}^{\alpha}u(t)\in C((0,T];L^{2}(\Omega))\cap L^{2}(0,T;L^{2}(\Omega));

  • 3.

    limt0+u(t)u0=0.\lim_{t\rightarrow 0^{+}}\left\|u(t)-u_{0}\right\|=0.

Lemma 2.14.

Let 0<α<1,u0D((Δ)s),F(t)=μu2(1kJu)γuL(0,T;D((Δ)s)).0<\alpha<1,u_{0}\in D((-\Delta)^{s}),F(t)=\mu u^{2}(1-kJ*u)-\gamma u\in L^{\infty}(0,T;D((-\Delta)^{s})). Then there exists a unique weak solution to the problem (1.1)-(1.2) and the solution is given by

u(t)=j=1Eα,1(λjstα)u0,jΦj+j=10t(tτ)α1Eα,α(λjs(tτ)α)Fj𝑑τΦj.u(t)=\sum_{j=1}^{\infty}E_{\alpha,1}(-\lambda_{j}^{s}t^{\alpha})u_{0,j}\Phi_{j}+\sum_{j=1}^{\infty}\int_{0}^{t}(t-\tau)^{\alpha-1}E_{\alpha,\alpha}(-\lambda_{j}^{s}(t-\tau)^{\alpha})F_{j}d\tau\Phi_{j}. (2.8)

Here u0,j=(u0,Φj)u_{0,j}=(u_{0},\Phi_{j}) and Fj=(F(τ),Φj)F_{j}=(F(\tau),\Phi_{j}).

Remark 2.2.

The conclusion of the Lemma 2.14 can be concluded in reference [5]. But in [5], “For brevity, we leave the detail to the reader.”[p255,proof of Theorem 1,[5]]. We have done the above related proof (See Appendix A.2). Therefore, we will use the properties of the eigensystem for the operator (Δ)s(-\Delta)^{s} and use the method as in the proof of [[37],Theorems 2.1-2.2] or [[38], Theorem 3.2] to obtain the existence and uniqueness of the solution (2.9) to the problem.

Remark 2.3.

The solution of the above equation (1.1)-(1.2) is also expressed in [55]. And we set

U(t)u0=j=1Eα,1(λjstα)u0,jΦj,U(t)u_{0}=\sum_{j=1}^{\infty}E_{\alpha,1}(-\lambda_{j}^{s}t^{\alpha})u_{0,j}\Phi_{j},

and

V(t)F=j=1tα1Eα,α(λjstα)FjΦj,V(t)F=\sum_{j=1}^{\infty}t^{\alpha-1}E_{\alpha,\alpha}(-\lambda_{j}^{s}t^{\alpha})F_{j}\Phi_{j},

then equation (2.8) can be written in the following form

u(t)=U(t)u0+0tV(tτ)F(τ)𝑑τ.u(t)=U(t)u_{0}+\int_{0}^{t}V(t-\tau)F(\tau)d\tau. (2.9)

Definition 2.6.

[56] Suppose eigenfunction e1>0e_{1}>0 associated to the first eigenvalue λ1>0\lambda_{1}>0 that satisfies the fractional eigenvalue problem

(Δ)se1(x)=λ1e1(x),xΩ,\displaystyle(-\Delta)^{s}e_{1}(x)=\lambda_{1}e_{1}(x),x\in\Omega, (2.10)
e1(x)=0,xNΩ,\displaystyle e_{1}(x)=0,\quad x\in\mathbb{R}^{N}\setminus\Omega,

normamlized such that Ωe1(x)𝑑x=1\int_{\Omega}e_{1}(x)dx=1.

Lemma 2.15.

[44] The Caputo fractional derivative of an absolutely continous function w(t)w(t) of order 0<α<10<\alpha<1 is defined by (2.5). The function ww satisfies

{Dtα0Cw(t)=w(t)(1+w(t)),w(0)=w0.\displaystyle\left\{\begin{matrix}&{}_{0}^{C}\textrm{D}_{t}^{\alpha}w(t)=w(t)(1+w(t)),\\ &w(0)=w_{0}.\end{matrix}\right. (2.11)

Then, we get the following formula for the solution of problem (2.11)

w(t)=Eα(tα)w0+0t(ts)α1Eα,α((ts)α)w2(s)𝑑s.w(t)=E_{\alpha}(t^{\alpha})w_{0}+\int_{0}^{t}(t-s)^{\alpha-1}E_{\alpha,\alpha}((t-s)^{\alpha})w^{2}(s)ds. (2.12)

Lemma 2.16.

Assume the initial data 0<u0D((Δ)s)<10<u_{0}\in D((-\Delta)^{s})<1. Then there are a maximal existence time

Tmax(0,],uC([0,Tmax];Hs(N))L2(0,Tmax;D((Δ)s))T_{max}\in(0,\infty],u\in C([0,T_{max}];H^{s}(\mathbb{R}^{N}))\cap L^{2}(0,T_{max};D((-\Delta)^{s}))

, such that uu is the unique nonnegative weak solution of (1.1)-(1.2). Furthermore, if 1+λ1Ωu0(x)e1(x)𝑑x=H0,1+\lambda_{1}\leq\int_{\Omega}u_{0}(x)e_{1}(x)dx=H_{0}, and λ1,e1(x)\lambda_{1},e_{1}(x) is the value in Definition 2.6, then the solution of problem (1.1)-(1.2) blow-up in a finite time TmaxT_{max} that satisfies the bi-lateral estimate

(Γ(α+1)4(H0+1/2))1αTmax(Γ(α+1)H0)1α.\left(\frac{\Gamma(\alpha+1)}{4(H_{0}+1/2)}\right)^{\frac{1}{\alpha}}\leq T_{max}\leq\left(\frac{\Gamma(\alpha+1)}{H_{0}}\right)^{\frac{1}{\alpha}}.

Furthermore, if Tmax<T_{max}<\infty, then

limtTmaxu(,t)L(N)=.{\lim_{t\to T_{max}}}\Arrowvert u(\cdot,t)\Arrowvert_{L^{\infty}(\mathbb{R}^{N})}=\infty.

Proof.

Multiplying equations (1.1) by e1(x)e_{1}(x) and integrating over Ω\Omega, we obtain

tαΩu(x,t)e1(x)𝑑x\displaystyle\partial_{t}^{\alpha}\int_{\Omega}u(x,t)e_{1}(x)dx +Ω(Δ)su(x,t)e1(x)𝑑x\displaystyle+\int_{\Omega}(-\Delta)^{s}u(x,t)e_{1}(x)dx
=Ω[μ(1kJu)u(x,t)γ]u(x,t)e1(x)𝑑x.\displaystyle=\int_{\Omega}[\mu(1-kJ*u)u(x,t)-\gamma]u(x,t)e_{1}(x)dx. (2.13)

By (2.10), we have

Ω(Δ)su(x,t)e1(x)𝑑x=Ωu(x,t)(Δ)se1(x)𝑑x=λ1Ωu(x,t)e1(x)𝑑x,\int_{\Omega}(-\Delta)^{s}u(x,t)e_{1}(x)dx=\int_{\Omega}u(x,t)(-\Delta)^{s}e_{1}(x)dx=\lambda_{1}\int_{\Omega}u(x,t)e_{1}(x)dx,

as u=0,e1(x)=0,xNΩu=0,e_{1}(x)=0,x\in\mathbb{R}^{N}\setminus\Omega and

(Ωu(x,t)e1(x)𝑑x)2Ωu2(x,t)e1(x)𝑑x,\left(\int_{\Omega}u(x,t)e_{1}(x)dx\right)^{2}\leq\int_{\Omega}u^{2}(x,t)e_{1}(x)dx,

let function H(t)=Ωu(x,t)e1(x)𝑑xH(t)=\int_{\Omega}u(x,t)e_{1}(x)dx, then satisfies

tαH(t)+(γ+λ1)H(t)μH2(t).\partial_{t}^{\alpha}H(t)+(\gamma+\lambda_{1})H(t)\geq\mu H^{2}(t). (2.14)

Let H~(t)=H(t)(γ+λ1)\tilde{H}(t)=H(t)-(\gamma+\lambda_{1}) and μ(1kJu)>1,λ1>0,γ>1\mu(1-kJ*u)>1,\lambda_{1}>0,\gamma>1, then from (2.14), we get

tαH~(t)H~(t)(H~(t)+γ+λ1)H~(t)(H~(t)+1)\partial_{t}^{\alpha}\tilde{H}(t)\geq\tilde{H}(t)(\tilde{H}(t)+\gamma+\lambda_{1})\geq\tilde{H}(t)(\tilde{H}(t)+1) (2.15)

It is seen that if u(t)u(t)\rightarrow\infty as tTmaxt\rightarrow T_{max}, then H(t)H(t)\rightarrow\infty as tTmaxt\rightarrow T_{max} and vice versa. That is HH and uu will have the same blow-up time. As 0H~0=H~(0)0\leq\tilde{H}_{0}=\tilde{H}(0), then from the results in [57], the solution of inequality (2.15) blows-up in a finite time. The proof of the Lemma is complete.

Remark 2.4.

We will use the method as in the proof of [[43],Theorem 2.1] and [[57], Theorem 3.2]. In addition, in the process of proof, we also need to use comparative principle to proof of blow-up in a finite time.

Proof of Theorem 1.1 1.

We first know that the equation (1.1)-(1.2) has the only weak solution by Lemma 2.14. Next, we will prove the global boundary of this weak solution in one dimensional and two dimensional space. For any xNx\in\mathbb{R}^{N},multiply (1.1) by 2uφε2u\varphi_{\varepsilon}, where φε()C0(N)\varphi_{\varepsilon}(\cdot)\in C_{0}^{\infty}(\mathbb{R}^{N}), and φε()1\varphi_{\varepsilon}(\cdot)\to 1 locally uniformly in N\mathbb{R}^{N} as ε0\varepsilon\to 0. Integrating by parts over N\mathbb{R}^{N}, we obtain

N2uφεαutα𝑑y\displaystyle\int_{\mathbb{R}^{N}}2u\varphi_{\varepsilon}\frac{\partial^{\alpha}u}{\partial t^{\alpha}}dy =N2uφε(Δ)su𝑑y+N2u3μφε(1kJu)𝑑y\displaystyle=-\int_{\mathbb{R}^{N}}2u\varphi_{\varepsilon}(-\Delta)^{s}udy+\int_{\mathbb{R}^{N}}2u^{3}\mu\varphi_{\varepsilon}(1-kJ*u)dy
2γNu2φε𝑑y,\displaystyle\qquad-2\gamma\int_{\mathbb{R}^{N}}u^{2}\varphi_{\varepsilon}dy,

by Lemma 2.5 and Lemma 2.7, then we can get

N2uφεαutα𝑑yαtαNu2φε𝑑y,\int_{\mathbb{R}^{N}}2u\varphi_{\varepsilon}\frac{\partial^{\alpha}u}{\partial t^{\alpha}}dy\geq\frac{\partial^{\alpha}}{\partial t^{\alpha}}\int_{\mathbb{R}^{N}}u^{2}\varphi_{\varepsilon}dy,

and by fractional Laplacian (1.4) and (2.3), then we can get

N2uφε(Δ)su𝑑yNφε(Δ)su2𝑑y,\int_{\mathbb{R}^{N}}2u\varphi_{\varepsilon}(-\Delta)^{s}udy\geq\int_{\mathbb{R}^{N}}\varphi_{\varepsilon}(-\Delta)^{s}u^{2}dy, (2.16)

and a symmetrical estimate of the nuclear function can be obtained (2.16) export

Nφε(Δ)su2𝑑y\displaystyle\int_{\mathbb{R}^{N}}\varphi_{\varepsilon}(-\Delta)^{s}u^{2}dy
=CN,sP.V.NNφεu2(x,t)u2(y,t)|xy|N+2s𝑑x𝑑y\displaystyle=C_{N,s}P.V.\int_{\mathbb{R}^{N}}\int_{\mathbb{R}^{N}}\varphi_{\varepsilon}\frac{u^{2}(x,t)-u^{2}(y,t)}{\left|x-y\right|^{N+2s}}dxdy
CN,sP.V.NNφε(u(x,t)u(y,t))2|xy|N+2s𝑑x𝑑y=CN,sP.V.[u]Hs2.\displaystyle\geq C_{N,s}P.V.\int_{\mathbb{R}^{N}}\int_{\mathbb{R}^{N}}\varphi_{\varepsilon}\frac{(u(x,t)-u(y,t))^{2}}{\left|x-y\right|^{N+2s}}dxdy=C_{N,s}P.V.[u]_{H^{s}}^{2}.

By definition 2.3 and Taking ε0\varepsilon\to 0, we obtain

αtαNu2𝑑y+CN,sP.V.[u]Hs22μNu3(1kJu)𝑑y2γNu2𝑑y\frac{\partial^{\alpha}}{\partial t^{\alpha}}\int_{\mathbb{R}^{N}}u^{2}dy+C_{N,s}P.V.[u]_{H^{s}}^{2}\leq 2\mu\int_{\mathbb{R}^{N}}u^{3}(1-kJ*u)dy-2\gamma\int_{\mathbb{R}^{N}}u^{2}dy

Knowing from (1.6), we have used the fact that y,zN\forall y,z\in\mathbb{R}^{N}, then J(zy)ηJ(z-y)\geq\eta, and then

Ju(y,t)=NJ(yz)u(z,t)𝑑zηNu(y,t)𝑑yJ*u(y,t)=\int_{\mathbb{R}^{N}}J(y-z)u(z,t)dz\geq\eta\int_{\mathbb{R}^{N}}u(y,t)dy

Therefore

αtαNu2𝑑y+CN,sP.V.[u]Hs2\displaystyle\frac{\partial^{\alpha}}{\partial t^{\alpha}}\int_{\mathbb{R}^{N}}u^{2}dy+C_{N,s}P.V.[u]_{H^{s}}^{2}
2μNu3𝑑y2μηkNu3𝑑yNu𝑑y2γNu2𝑑y.\displaystyle\leq 2\mu\int_{\mathbb{R}^{N}}u^{3}dy-2\mu\eta k\int_{\mathbb{R}^{N}}u^{3}dy\int_{\mathbb{R}^{N}}udy-2\gamma\int_{\mathbb{R}^{N}}u^{2}dy. (2.17)

Now we proceed to estimate the term Nu3𝑑y\int_{\mathbb{R}^{N}}u^{3}dy.

Firstly using Gagliardo-Nirenberg inequality in Lemma 2.12, let p=3,q=r=2,λ=N6p=3,q=r=2,\lambda^{*}=\frac{N}{6}, there exists constant CGN>0C_{GN}>0, such that

Nu3𝑑yCGN(N)(uL2(N)N2uL2(N)3N2+uL2(N)3).\int_{\mathbb{R}^{N}}u^{3}dy\leq C_{GN}(N)(\|\bigtriangledown u\|_{L^{2}(\mathbb{R}^{N})}^{\frac{N}{2}}\|u\|_{L^{2}(\mathbb{R}^{N})}^{3-\frac{N}{2}}+\|u\|_{L^{2}(\mathbb{R}^{N})}^{3}). (2.18)

On the one hand, by Lemma 2.13, we can make a=uL2(N)N2,b=CGN(N)uL2(N)3N2,ε=1μ,p=1N,q=44Na=\|\triangledown u\|_{L^{2}(\mathbb{R}^{N})}^{\frac{N}{2}},b=C_{GN}(N)\|u\|_{L^{2}(\mathbb{R}^{N})}^{3-\frac{N}{2}},\varepsilon=\frac{1}{\mu},p=\frac{1}{N},q=\frac{4}{4-N}, then obtain

CGN(N)uL2(N)N2uL2(N)3N21μuL2(N)2+μN4NCGN44N(N)uL2(N)2(6N)4N.C_{GN}(N)\|\triangledown u\|_{L^{2}(\mathbb{R}^{N})}^{\frac{N}{2}}\|u\|_{L^{2}(\mathbb{R}^{N})}^{3-\frac{N}{2}}\leq\frac{1}{\mu}\|\triangledown u\|_{L^{2}(\mathbb{R}^{N})}^{2}+\mu^{\frac{N}{4-N}}C_{GN}^{\frac{4}{4-N}}(N)\|u\|_{L^{2}(\mathbb{R}^{N})}^{\frac{2(6-N)}{4-N}}. (2.19)

Here by Young’s inequality, let a=CGN(N),b=uL2(N)3,p=2(6N)3(4N)a=C_{GN}(N),b=\left\|u\right\|_{L^{2}(\mathbb{R}^{N})}^{3},p=\frac{2(6-N)}{3(4-N)}, q=2(6N)Nq=\frac{2(6-N)}{N}, then

CGN(N)uL2(N)3uL2(N)2(6N)4N+CGN2(6N)N(N).C_{GN}(N)\left\|u\right\|_{L^{2}(\mathbb{R}^{N})}^{3}\leq\left\|u\right\|_{L^{2}(\mathbb{R}^{N})}^{\frac{2(6-N)}{4-N}}+C_{GN}^{\frac{2(6-N)}{N}}(N). (2.20)

By interpolation inequality, we obtain

uL2(N)2(6N)4N\displaystyle\left\|u\right\|_{L^{2}(\mathbb{R}^{N})}^{\frac{2(6-N)}{4-N}} (uL1(N)14uL3(N)34)2(6N)4N\displaystyle\leq(\left\|u\right\|_{L^{1}(\mathbb{R}^{N})}^{\frac{1}{4}}\left\|u\right\|_{L^{3}(\mathbb{R}^{N})}^{\frac{3}{4}})^{\frac{2(6-N)}{4-N}} (2.21)
=(Nu3𝑑yNu𝑑y)6N2(4N).\displaystyle=(\int_{\mathbb{R}^{N}}u^{3}dy\int_{\mathbb{R}^{N}}udy)^{\frac{6-N}{2(4-N)}}.

Combing (2.18)-(2.21), we obtain

2μNu3𝑑y\displaystyle 2\mu\int_{\mathbb{R}^{N}}u^{3}dy (2.22)
2N|u|2𝑑y+2μ(1+μN4NCGN44N(N))(Nu3𝑑yNu𝑑y)6N2(4N)\displaystyle\leq 2\int_{\mathbb{R}^{N}}\left|\bigtriangledown u\right|^{2}dy+2\mu(1+\mu^{\frac{N}{4-N}}C_{GN}^{\frac{4}{4-N}}(N))(\int_{\mathbb{R}^{N}}u^{3}dy\int_{\mathbb{R}^{N}}udy)^{\frac{6-N}{2(4-N)}}
+2μCGN2(6N)N(N).\displaystyle\qquad+2\mu C_{GN}^{\frac{2(6-N)}{N}}(N).

Next we consider the cases N=1N=1 and N=2N=2 respectively.

Case 1. N=1N=1. From (2.22), by Lemma 2.13, let ε=ηk,p=65,q=6\varepsilon=\eta k,p=\frac{6}{5},q=6, we get

2μNu3𝑑y2N|u|2𝑑y+2μηkNu3𝑑yNu𝑑y\displaystyle 2\mu\int_{\mathbb{R}^{N}}u^{3}dy\leq 2\int_{\mathbb{R}^{N}}\left|\bigtriangledown u\right|^{2}dy+2\mu\eta k\int_{\mathbb{R}^{N}}u^{3}dy\int_{\mathbb{R}^{N}}udy (2.23)
+2μ(μ13CGN43(1)+1)6(ηk)5+2ηCGN10(1),\displaystyle\qquad+2\mu(\mu^{\frac{1}{3}}C_{GN}^{\frac{4}{3}}(1)+1)^{6}(\eta k)^{-5}+2\eta C_{GN}^{10}(1),

inserting (2.23) into (2.17), we have

αtαNu2𝑑y+CN,sP.V.[u]Hs2+2γNu2𝑑y\displaystyle\frac{\partial^{\alpha}}{\partial t^{\alpha}}\int_{\mathbb{R}^{N}}u^{2}dy+C_{N,s}P.V.[u]_{H^{s}}^{2}+2\gamma\int_{\mathbb{R}^{N}}u^{2}dy (2.24)
2N|u|2𝑑y+2μ[(μ13CGN43(1))6(ηk)5+CGN10(1)].\displaystyle\leq 2\int_{\mathbb{R}^{N}}\left|\triangledown u\right|^{2}dy+2\mu[(\mu^{\frac{1}{3}}C_{GN}^{\frac{4}{3}}(1))^{6}(\eta k)^{-5}+C_{GN}^{10}(1)].

From Sobolev embedding inequality (Lemma 2.11) and proof of Theorem 1 in [58], we know there exists an embedding constant C2>0C_{2}>0 suth that

uLp(N)2C1uLs2(N)2\left\|\triangledown u\right\|_{L^{p}(\mathbb{R}^{N})}^{2}\geq C_{1}\left\|u\right\|_{L^{s_{2}}(\mathbb{R}^{N})}^{2}

where s2ps_{2}\geq p will be determined later. Here we set s2=p=2s_{2}=p=2,then there is

uL2(N)2C2uL2(N)2.\left\|\triangledown u\right\|_{L^{2}(\mathbb{R}^{N})}^{2}\geq C_{2}\left\|u\right\|_{L^{2}(\mathbb{R}^{N})}^{2}.

By fractional Sobolev inequality (2.6), we can obtain

CN,sP.V.[u]Hs2CN,sP.V.CuL2(N)2C_{N,s}P.V.[u]_{H^{s}}^{2}\geq\frac{C_{N,s}P.V.}{C_{*}}\left\|u\right\|_{L^{2}(\mathbb{R}^{N})}^{2}

for all uHs(N)u\in H^{s}(\mathbb{R}^{N}). Let

C2=CN,sP.V.C,Q1=2μ((μ13CGN43(1)+1)6(ηk)5+CGN10(1)),C_{*}^{-2}=\frac{C_{N,s}P.V.}{C_{*}},\quad Q_{1}=2\mu\left((\mu^{\frac{1}{3}}C_{GN}^{\frac{4}{3}}(1)+1)^{6}(\eta k)^{-5}+C_{GN}^{10}(1)\right),

then, the equation (2.24) is configured above, you can get

αtαNu2𝑑y+C2Nu2𝑑y+2γNu2𝑑yC2Nu2𝑑y+Q1.\displaystyle\frac{\partial^{\alpha}}{\partial t^{\alpha}}\int_{\mathbb{R}^{N}}u^{2}dy+C_{*}^{-2}\int_{\mathbb{R}^{N}}u^{2}dy+2\gamma\int_{\mathbb{R}^{N}}u^{2}dy\leq C_{2}\int_{\mathbb{R}^{N}}u^{2}dy+Q_{1}.

Denote w(t)=Nu2𝑑yw(t)=\int_{\mathbb{R}^{N}}u^{2}dy, the solution of the following fractional differential equation

{Dtα0Cw(t)+(2γ+C2C2)w(t)=Q1w(0)=2δu0L(RN)2\begin{cases}{}_{0}^{C}\textrm{D}_{t}^{\alpha}w(t)+(2\gamma+C_{*}^{-2}-C_{2})w(t)=Q_{1}\\ w(0)=2\delta\left\|u_{0}\right\|_{L^{\infty}(R^{N})}^{2}\end{cases}

for (x,t)×(0,Tmax)\forall(x,t)\in\mathbb{R}\times(0,T_{max}) and from Lemma 2.10, let

C4=2γ+C2C2C_{4}=2\gamma+C_{*}^{-2}-C_{2}

and w(0)=η=2δu0L(RN)2w(0)=\eta=2\delta\left\|u_{0}\right\|_{L^{\infty}(R^{N})}^{2}, we obtain

Nu2𝑑y=uL2(N)2w(t)\displaystyle\int_{\mathbb{R}^{N}}u^{2}dy=\left\|u\right\|_{L^{2}(\mathbb{R}^{N})}^{2}\leq\left\|w(t)\right\|
w(0)+(K+|Q1|)TααΓ(α)Ce[(C4)1α+σ]tσT\displaystyle\leq w(0)+\frac{(K+\left|Q_{1}\right|)T^{\alpha}}{\alpha\Gamma(\alpha)}Ce^{[(-C_{4})^{\frac{1}{\alpha}}+\sigma]t-\sigma T}
2δu0L(RN)2+(K+|2μ((μ13CGN43(1)+1)6(ηk)5+CGN10(1))|)TααΓ(α)C\displaystyle\leq 2\delta\left\|u_{0}\right\|_{L^{\infty}(R^{N})}^{2}+\frac{(K+\left|2\mu\left((\mu^{\frac{1}{3}}C_{GN}^{\frac{4}{3}}(1)+1)^{6}(\eta k)^{-5}+C_{GN}^{10}(1)\right)\right|)T^{\alpha}}{\alpha\Gamma(\alpha)}C
:=M1.\displaystyle:=M_{1}.

Case 2.N=2N=2. From (2.22), we have

2μNu3𝑑y2N|u|2𝑑y+2μ(1+μCGN2(2))\displaystyle 2\mu\int_{\mathbb{R}^{N}}u^{3}dy\leq 2\int_{\mathbb{R}^{N}}\left|\bigtriangledown u\right|^{2}dy+2\mu(1+\mu C_{GN}^{2}(2)) (2.25)
(Nu3𝑑yNu𝑑y)+2μCGN4(2).\displaystyle\qquad(\int_{\mathbb{R}^{N}}u^{3}dy\int_{\mathbb{R}^{N}}udy)+2\mu C_{GN}^{4}(2).

Inserting (2.25) into (2.17), for kk=μCGN2(2)+1ηk\geq k^{*}=\frac{\mu C_{GN}^{2}(2)+1}{\eta}, we obtain

αtαNu2𝑑y+C2Nu2𝑑y+2γNu2𝑑yC2Nu2𝑑y+2μCGN4(2).\displaystyle\frac{\partial^{\alpha}}{\partial t^{\alpha}}\int_{\mathbb{R}^{N}}u^{2}dy+C_{*}^{-2}\int_{\mathbb{R}^{N}}u^{2}dy+2\gamma\int_{\mathbb{R}^{N}}u^{2}dy\leq C_{2}\int_{\mathbb{R}^{N}}u^{2}dy+2\mu C_{GN}^{4}(2).

Denote w(t)=Nu2𝑑yw(t)=\int_{\mathbb{R}^{N}}u^{2}dy the solution of the following ordinary differential equation

{Dtα0Cw(t)+(2γ+C2C2)w(t)=2μCGN4(2),w(0)=(2δ)2u0L(N)2,\begin{cases}{}_{0}^{C}\textrm{D}_{t}^{\alpha}w(t)+(2\gamma+C_{*}^{-2}-C_{2})w(t)=2\mu C_{GN}^{4}(2),\\ w(0)=(2\delta)^{2}\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}^{2},\end{cases} (2.26)

for (x,t)2×(0,Tmax)\forall(x,t)\in\mathbb{R}^{2}\times(0,T_{max}) and let C5=2γ+C2C2C_{5}=2\gamma+C_{*}^{-2}-C_{2}, we obtain

Nu2𝑑y=uL2(N)2w(t)\displaystyle\int_{\mathbb{R}^{N}}u^{2}dy=\left\|u\right\|_{L^{2}(\mathbb{R}^{N})}^{2}\leq\left\|w(t)\right\|
w(0)+(K+|2μCGN4(2)|)TααΓ(α)Ce[(C5)1α+σ]tσT\displaystyle\leq w(0)+\frac{(K+\left|2\mu C_{GN}^{4}(2)\right|)T^{\alpha}}{\alpha\Gamma(\alpha)}Ce^{[(-C_{5})^{\frac{1}{\alpha}}+\sigma]t-\sigma T}
(2δ)2u0L(RN)2+(K+|2μCGN4(2)|)TααΓ(α)C\displaystyle\leq(2\delta)^{2}\left\|u_{0}\right\|_{L^{\infty}(R^{N})}^{2}+\frac{(K+\left|2\mu C_{GN}^{4}(2)\right|)T^{\alpha}}{\alpha\Gamma(\alpha)}C
:=M2.\displaystyle:=M_{2}.

In conclusion, for any (x,t)N×(0,Tmax)(x,t)\in\mathbb{R}^{N}\times(0,T_{max}), we have

uL2(N)M={M1N=1M2N=2\left\|u\right\|_{L^{2}(\mathbb{R}^{N})}\leq M=\begin{cases}\sqrt{M_{1}}\qquad N=1\\ \sqrt{M_{2}}\qquad N=2\end{cases} (2.27)

and then

uL1(N)M.\left\|u\right\|_{L^{1}(\mathbb{R}^{N})}\leq M. (2.28)

Now we proceed to improve the L2L^{2} boundedness of uu to LL^{\infty}, which is based on the fact that for all (x,t)N×(0,Tmax)(x,t)\in\mathbb{R}^{N}\times(0,T_{max}) and equation (2.9), let F(t)=μu2(1kJu)γuF(t)=\mu u^{2}(1-kJ*u)-\gamma u, then we have the solution of equation(1.1)-(1.2)

u(x,t)=U(t)u0+0tV(tr)F(r)𝑑r.u(x,t)=U(t)u_{0}+\int_{0}^{t}V(t-r)F(r)dr.

By [55] and Lemma 2.1-2.4, we can obtain

U(t)u0=j=1Eα,1(λjstα)u0,jΦjj=1C6(u0,Φj)C6u0.\left\|U(t)u_{0}\right\|=\left\|\sum_{j=1}^{\infty}E_{\alpha,1}(-\lambda_{j}^{s}t^{\alpha})u_{0,j}\Phi_{j}\right\|\leq\sum_{j=1}^{\infty}C_{6}(u_{0},\Phi_{j})\leq C_{6}\left\|u_{0}\right\|.

From using Lemma 2.3 and Parseval identity, we have for some function F(t)L2(N)F(t)\in L^{2}(\mathbb{R}^{N})

V(tr)F(r)=j=1(tr)α1Eα,α(λjs(tr)α)FjΦj\displaystyle\left\|V(t-r)F(r)\right\|=\left\|\sum_{j=1}^{\infty}(t-r)^{\alpha-1}E_{\alpha,\alpha}(-\lambda_{j}^{s}(t-r)^{\alpha})F_{j}\Phi_{j}\right\|
C7(j=1(tr)2α2(11+λjs(tr)α)2(Fj,Φj)2)12\displaystyle\leq C_{7}\left(\sum_{j=1}^{\infty}(t-r)^{2\alpha-2}(\frac{1}{1+\lambda_{j}^{s}(t-r)^{\alpha}})^{2}(F_{j},\Phi_{j})^{2}\right)^{\frac{1}{2}}
C7((tr)2α2(tr)12α)12(j=0(Fj,Φj)2((tr)14α1+λjs(tr)α)2)12\displaystyle\leq C_{7}\left((t-r)^{2\alpha-2}(t-r)^{-\frac{1}{2}\alpha}\right)^{\frac{1}{2}}\left(\sum_{j=0}^{\infty}(F_{j},\Phi_{j})^{2}(\frac{(t-r)^{\frac{1}{4}\alpha}}{1+\lambda_{j}^{s}(t-r)^{\alpha}})^{2}\right)^{\frac{1}{2}}
C7(tr)34α1F(r).\displaystyle\leq C_{7}(t-r)^{\frac{3}{4}\alpha-1}\left\|F(r)\right\|.

So there is

0u(x,t)U(t)u0+0tV(ts)F(s)𝑑s\displaystyle 0\leq u(x,t)\leq\left\|U(t)u_{0}\right\|+\int_{0}^{t}\left\|V(t-s)F(s)\right\|ds
C6u0L(N)+C70t(ts)34α1F(s)L(N)𝑑s\displaystyle\leq C_{6}\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}+C_{7}\int_{0}^{t}(t-s)^{\frac{3}{4}\alpha-1}\left\|F(s)\right\|_{L^{\infty}(\mathbb{R}^{N})}ds
C6u0L(N)+μC70t(ts)34α1u2(s)L(N)𝑑s\displaystyle\leq C_{6}\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}+\mu C_{7}\int_{0}^{t}(t-s)^{\frac{3}{4}\alpha-1}\left\|u^{2}(s)\right\|_{L^{\infty}(\mathbb{R}^{N})}ds
C6u0L(N)+μC7M20t(ts)34α1𝑑s\displaystyle\leq C_{6}\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}+\mu C_{7}M^{2}\int_{0}^{t}(t-s)^{\frac{3}{4}\alpha-1}ds
C6u0L(N)+μC8M2T34α134α.\displaystyle\leq C_{6}\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}+\mu C_{8}M^{2}T^{\frac{3}{4}\alpha}\frac{1}{\frac{3}{4}\alpha}.
0u(x,t)u0L(N)+μM2C2(N,T,γ),0\leq u(x,t)\leq\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}+\mu M^{2}C_{2}(N,T,\gamma), (2.29)

with

M={2δu0L(RN)2+(K+|2μ((μ13CGN43(1)+1)6(ηk)5+CGN10(1))|)TααΓ(α)C,N=1(2δ)2u0L(RN)2+(K+|2μCGN4(2)|)TααΓ(α)C,N=2M=\begin{cases}\sqrt{2\delta\left\|u_{0}\right\|_{L^{\infty}(R^{N})}^{2}+\frac{(K+\left|2\mu\left((\mu^{\frac{1}{3}}C_{GN}^{\frac{4}{3}}(1)+1)^{6}(\eta k)^{-5}+C_{GN}^{10}(1)\right)\right|)T^{\alpha}}{\alpha\Gamma(\alpha)}C},&N=1\\ \sqrt{(2\delta)^{2}\left\|u_{0}\right\|_{L^{\infty}(R^{N})}^{2}+\frac{(K+\left|2\mu C_{GN}^{4}(2)\right|)T^{\alpha}}{\alpha\Gamma(\alpha)}C},&N=2\end{cases}

defined in (2.27). As a summary, we have proved that the solution uu is globally bounded in time, the blow-up criterion in Lemma 2.16 shows that uu is the unique weak solution of (1.1)-(1.2) on (x,t)N×(0,+)(x,t)\in\mathbb{R}^{N}\times(0,+\infty). Theorem 1.1 is thus proved.

3 Long time behavior of solutions

Now, we consider the long time behavior of the weak solution of (1.1)-(1.2). To study the long time behavior of solutions for (1.1)-(1.2), by [4], we denote

F(u):=μu2(1kJu)γu.F(u):=\mu u^{2}(1-kJ*u)-\gamma u.

For 1<γ<μ4k1<\gamma<\frac{\mu}{4k}, there are three constant solutions for F(u)=0F(u)=0: 0,a,A0,a,A, where

a=114kγμ2k,A=1+14kγμ2k,a=\frac{1-\sqrt{1-4k\frac{\gamma}{\mu}}}{2k},\qquad A=\frac{1+\sqrt{1-4k\frac{\gamma}{\mu}}}{2k}, (3.1)

and satisfy 1<γμ<a<A1<\frac{\gamma}{\mu}<a<A.

Lemma 3.1.

[47] Let uC02(N)u\in C_{0}^{2}\left(\mathbb{R}^{N}\right) and Φ\Phi be a convex function of one variable. Then

Φ(u)(Δ)su(x)(Δ)sΦ(u(x)).{\Phi}^{\prime}(u)(-\Delta)^{s}u(x)\geq(-\Delta)^{s}\Phi(u(x)).

Proposition 3.1.

Under the assumptions of Theorem 1.1, there is
u(x,t)L(N×(0,+))<a\left\|u(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N}\times(0,+\infty))}<a, the function

H(x,t)=B(x,δ)h(u(y,t))𝑑yH(x,t)=\int_{B(x,\delta)}h(u(y,t))dy

with

h(u)=Aln(1uA)aln(1ua)h(u)=Aln\left(1-\frac{u}{A}\right)-aln\left(1-\frac{u}{a}\right)

is nonnegative and satisfies

αH(x,t)tα(Δ)sH(x,t)+B(x,δ)|u(y,t)|2𝑑yD(x,t)\frac{\partial^{\alpha}H(x,t)}{\partial t^{\alpha}}\leq-(-\Delta)^{s}H(x,t)+\int_{B(x,\delta)}\left|\bigtriangledown u(y,t)\right|^{2}dy-D(x,t) (3.2)

with

D(x,t)=12(Aa)μkB(x,δ)u2(y,t)𝑑y.D(x,t)=\frac{1}{2}(A-a)\mu k\int_{B(x,\delta)}u^{2}(y,t)dy.

Proof.

Fix x0(x10,,xN0)Nx_{0}\triangleq(x_{1}^{0},\cdots,x_{N}^{0})\in\mathbb{R}^{N}, choose 0<δ<12δ0<\delta<\frac{1}{2}\delta, and denote

B(x,δ):={x(x1,,xN)N||xixi0δ|,1iN}.B(x,\delta):=\left\{x\triangleq(x_{1},\cdots,x_{N})\in\mathbb{R}^{N}|\left|x_{i}-x_{i}^{0}\leq\delta\right|,1\leq i\leq N\right\}.

Let k=u(x,t)L(N×[0,))k=\left\|u(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N}\times[0,\infty))}, then 0<k<a0<k<a. From the definition of h()h(\cdot), it is easy to verify that

h(u)=aauAAu=(Aa)u(Au)(au),{h}^{\prime}(u)=\frac{a}{a-u}-\frac{A}{A-u}=\frac{(A-a)u}{(A-u)(a-u)}, (3.3)

and

h′′(u)=a(au)2A(Aa)2=(Aau2)(Aa)(Au)2(au)2.{h}^{\prime\prime}(u)=\frac{a}{(a-u)^{2}}-\frac{A}{(A-a)^{2}}=\frac{(Aa-u^{2})(A-a)}{(A-u)^{2}(a-u)^{2}}. (3.4)

Test (1.1) by h(u)φε{h}^{\prime}(u)\varphi_{\varepsilon} with φε()C0(B(x,δ)),φε()1\varphi_{\varepsilon}(\cdot)\in C_{0}^{\infty}(B(x,\delta)),\varphi_{\varepsilon}(\cdot)\rightarrow 1 in B(x,δ)B(x,\delta) as ε0\varepsilon\rightarrow 0. Integrating by parts over B(x,δ)B(x,\delta). From Definition 2.3 and Lemma 3.1, we obtain

h(u)(Δ)su(x)(Δ)sh(u(x)).{h}^{\prime}(u)(-\Delta)^{s}u(x)\geq(-\Delta)^{s}h(u(x)).

By using Lagrange mean value theorem

(Δ)sB(x,δ)h(u)𝑑y\displaystyle(-\Delta)^{s}\int_{B(x,\delta)}h(u)dy =CN,sP.V.B(x,δ)|B(x,δ)h(u(x))𝑑xB(x,δ)h(u(y))𝑑x||u(x)u(y)|N+2s𝑑y\displaystyle=C_{N,s}P.V.\int_{B(x,\delta)}\frac{\left|\int_{B(x,\delta)}h(u(x))dx-\int_{B(x,\delta)}h(u(y))dx\right|}{\left|u(x)-u(y)\right|^{N+2s}}dy
=CN,sP.V.B(x,δ)B(x,δ)|h(u(x))h(u(y))||u(x)u(y)|N+2s𝑑x𝑑y\displaystyle=C_{N,s}P.V.\int_{B(x,\delta)}\int_{B(x,\delta)}\frac{\left|h(u(x))-h(u(y))\right|}{\left|u(x)-u(y)\right|^{N+2s}}dxdy
B(x,δ)(Δ)sh(u)𝑑y.\displaystyle\leq\int_{B(x,\delta)}(-\Delta)^{s}h(u)dy.

From Lemma 2.7 and Equation (1.4), we get

B(x,δ)h(u)tαudyB(x,δ)tαh(u)dy=tαB(x,δ)h(u)𝑑y.\int_{B(x,\delta)}{h}^{\prime}(u)\partial_{t}^{\alpha}udy\geq\int_{B(x,\delta)}\partial_{t}^{\alpha}h(u)dy=\partial_{t}^{\alpha}\int_{B(x,\delta)}h(u)dy.

Then

αtαB(x,δ)h(u)φε𝑑y\displaystyle\frac{\partial^{\alpha}}{\partial t^{\alpha}}\int_{B(x,\delta)}h(u)\varphi_{\varepsilon}dy
B(x,δ)h(u)(Δ)suφε𝑑y+B(x,δ)[μu2(1kJu)γu]h(u)φε𝑑y\displaystyle\leq-\int_{B(x,\delta)}{h}^{\prime}(u)(-\Delta)^{s}u\varphi_{\varepsilon}dy+\int_{B(x,\delta)}[\mu u^{2}(1-kJ*u)-\gamma u]{h}^{\prime}(u)\varphi_{\varepsilon}dy
(Δ)sB(x,δ)φεh(u)𝑑y+B(x,δ)[μu2(1kJu)γu]h(u)φε𝑑y.\displaystyle\leq-(-\Delta)^{s}\int_{B(x,\delta)}\varphi_{\varepsilon}h(u)dy+\int_{B(x,\delta)}[\mu u^{2}(1-kJ*u)-\gamma u]{h}^{\prime}(u)\varphi_{\varepsilon}dy.

Taking ε0\varepsilon\rightarrow 0, we obtain

αtαB(x,δ)h(u)𝑑y\displaystyle\frac{\partial^{\alpha}}{\partial t^{\alpha}}\int_{B(x,\delta)}h(u)dy
B(x,δ)(Δ)sh(u)𝑑y+B(x,δ)[μu2(1kJu)γu]h(u)𝑑y\displaystyle\leq-\int_{B(x,\delta)}(-\Delta)^{s}h(u)dy+\int_{B(x,\delta)}[\mu u^{2}(1-kJ*u)-\gamma u]{h}^{\prime}(u)dy
(Δ)sB(x,δ)h(u)𝑑y+B(x,δ)[μu2(1kJu)γu]h(u)𝑑y.\displaystyle\leq-(-\Delta)^{s}\int_{B(x,\delta)}h(u)dy+\int_{B(x,\delta)}[\mu u^{2}(1-kJ*u)-\gamma u]{h}^{\prime}(u)dy.

which is

αtαH(x,t)(Δ)sH(x,t)+B(x,δ)[μu2(1kJu)γu]h(u)𝑑y.\displaystyle\frac{\partial^{\alpha}}{\partial t^{\alpha}}H(x,t)\leq-(-\Delta)^{s}H(x,t)+\int_{B(x,\delta)}[\mu u^{2}(1-kJ*u)-\gamma u]{h}^{\prime}(u)dy. (3.5)

By (3.1), we can get

μu2(1ku)γu=kμu(Au)(ua),\mu u^{2}(1-ku)-\gamma u=k\mu u(A-u)(u-a),

and

B(x,δ)h(u)[μu2(1kJu)γu]𝑑y\displaystyle\int_{B(x,\delta)}{h}^{\prime}(u)[\mu u^{2}(1-kJ*u)-\gamma u]dy (3.6)
=B(x,δ)h(u)[μu2(1ku)γu]𝑑y+μkB(x,δ)h(u)u2(uJu)𝑑y\displaystyle=\int_{B(x,\delta)}{h}^{\prime}(u)[\mu u^{2}(1-ku)-\gamma u]dy+\mu k\int_{B(x,\delta)}{h}^{\prime}(u)u^{2}(u-J*u)dy
=(Aa)μkB(x,δ)u2(y,t)𝑑y+μkB(x,δ)h(u)u2(uJu)𝑑y.\displaystyle=-(A-a)\mu k\int_{B(x,\delta)}u^{2}(y,t)dy+\mu k\int_{B(x,\delta)}{h}^{\prime}(u)u^{2}(u-J*u)dy.

Noticing that when 0uk0\leq u\leq k, there is

0h(u)u(Aa)k2(Ak)(ak).0\leq{h}^{\prime}(u)u\leq\frac{(A-a)k^{2}}{(A-k)(a-k)}.

From Young’s inequality and the median value theorem, we can get

μkB(x,δ)h(u)u2(uJu)𝑑y\displaystyle\mu k\int_{B(x,\delta)}{h}^{\prime}(u)u^{2}(u-J*u)dy (3.7)
μkB(x,δ)B(x,δ)h(u)u2(y,t)(u(y,t)u(z,t))J(yz)𝑑z𝑑y\displaystyle\leq\mu k\int_{B(x,\delta)}\int_{B(x,\delta)}{h}^{\prime}(u)u^{2}(y,t)(u(y,t)-u(z,t))J(y-z)dzdy
(Aa)K2(AK)(aK)μkB(x,δ)B(x,δ)u(y,t)|(u(y,t)u(z,t))|J(y,z)𝑑z𝑑y\displaystyle\leq\frac{(A-a)K^{2}}{(A-K)(a-K)}\mu k\int_{B(x,\delta)}\int_{B(x,\delta)}u(y,t)\left|(u(y,t)-u(z,t))\right|J(y,z)dzdy
(Aa)K4μk2(AK)2(aK)2B(x,δ)B(x,δ)(u(z,t)u(y,t))2J(zy)𝑑z𝑑y\displaystyle\leq\frac{(A-a)K^{4}\mu k}{2(A-K)^{2}(a-K)^{2}}\int_{B(x,\delta)}\int_{B(x,\delta)}(u(z,t)-u(y,t))^{2}J(z-y)dzdy
+12(Aa)μkB(x,δ)u2(y,t)𝑑y\displaystyle\qquad+\frac{1}{2}(A-a)\mu k\int_{B(x,\delta)}u^{2}(y,t)dy
(Aa)K4μk2(AK)2(aK)2B(x,δ)B(x,δ)01|u(y+θ(zy),t)|2|zy|2\displaystyle\leq\frac{(A-a)K^{4}\mu k}{2(A-K)^{2}(a-K)^{2}}\int_{B(x,\delta)}\int_{B(x,\delta)}\int_{0}^{1}\left|\bigtriangledown u(y+\theta(z-y),t)\right|^{2}\left|z-y\right|^{2}
J(zy)dθdzdy+12(Aa)μkB(x,δ)u2(y,t)𝑑y,\displaystyle\qquad J(z-y)d\theta dzdy+\frac{1}{2}(A-a)\mu k\int_{B(x,\delta)}u^{2}(y,t)dy,

changing the variables y=y+θ(zy),z=zy{y}^{\prime}=y+\theta(z-y),{z}^{\prime}=z-y, then

|yyyzzyzz|=|1θθ11|=1θ+θ=1.\begin{vmatrix}&\frac{\partial{y}^{\prime}}{\partial y}\qquad\frac{\partial{y}^{\prime}}{\partial z}\\ &\frac{\partial{z}^{\prime}}{\partial y}\qquad\frac{\partial{z}^{\prime}}{\partial z}\\ \end{vmatrix}=\begin{vmatrix}1-\theta\qquad\theta&\\ -1\qquad 1&\end{vmatrix}=1-\theta+\theta=1.

For any θ[0,1],y,zB(x,δ)\theta\in[0,1],y,z\in B(x,\delta), we have yB((1θ)x+θz,(1θ)δ),zB(xy,δ){y}^{\prime}\in B((1-\theta)x+\theta z,(1-\theta)\delta),{z}^{\prime}\in B(x-y,\delta). Noticing B((1θ)x+θz,(1θ)δ)B(x,δ)B((1-\theta)x+\theta z,(1-\theta)\delta)\subseteq B(x,\delta) and B(xy,δ)B(0,2δ)B(x-y,\delta)\subseteq B(0,2\delta), we obtain

(Aa)K4μk2(AK)2(aK)2B(x,δ)B(x,δ)01|u(y+θ(zy),t)|2|zy|2\displaystyle\frac{(A-a)K^{4}\mu k}{2(A-K)^{2}(a-K)^{2}}\int_{B(x,\delta)}\int_{B(x,\delta)}\int_{0}^{1}\left|\bigtriangledown u(y+\theta(z-y),t)\right|^{2}\left|z-y\right|^{2} (3.8)
J(zy)dθdzdy\displaystyle\qquad J(z-y)d\theta dzdy
(Aa)K4μk2(AK)2(aK)201𝑑θB(x,δ)B(x,δ)|u(y,t)|2|z|2J(z)𝑑z𝑑y\displaystyle\leq\frac{(A-a)K^{4}\mu k}{2(A-K)^{2}(a-K)^{2}}\int_{0}^{1}d\theta\int_{B(x,\delta)}\int_{B(x,\delta)}\left|\bigtriangledown u({y}^{\prime},t)\right|^{2}\left|{z}^{\prime}\right|^{2}J({z}^{\prime})d{z}^{\prime}d{y}^{\prime}
(Aa)K4μk(2δ)22(AK)2(aK)2B(x,δ)|u(y,t)|2𝑑y.\displaystyle\leq\frac{(A-a)K^{4}\mu k(2\delta)^{2}}{2(A-K)^{2}(a-K)^{2}}\int_{B(x,\delta)}\left|\bigtriangledown u(y,t)\right|^{2}dy.

Combining (3.6)-(3.8), we obtain

B(x,δ)h(u)[μu2(1kJu)γu]𝑑y12(Aa)μkB(x,δ)u2(y,t)𝑑y\displaystyle\int_{B(x,\delta)}{h}^{\prime}(u)[\mu u^{2}(1-kJ*u)-\gamma u]dy\leq-\frac{1}{2}(A-a)\mu k\int_{B(x,\delta)}u^{2}(y,t)dy (3.9)
+(Aa)k4μk(2δ)22(Aa)2(ak)2B(x,δ)|u(y,t)|2𝑑y.\displaystyle\qquad+\frac{(A-a)k^{4}\mu k(2\delta)^{2}}{2(A-a)^{2}(a-k)^{2}}\int_{B(x,\delta)}\left|\bigtriangledown u(y,t)\right|^{2}dy.

From (3.4), noticing 0u<a0\leq u<a, we obtain

h′′(u)>(Aa)2A2a,{h}^{\prime\prime}(u)>\frac{(A-a)^{2}}{A^{2}a},

inserting (3.9) into (3.5), we obtain

αtαH(x,t)\displaystyle\frac{\partial^{\alpha}}{\partial t^{\alpha}}H(x,t) (Δ)sH(x,t)+(Aa)k4μk(2δ)22(Ak)2(ak)2B(x,δ)|u(y,t)|2𝑑y\displaystyle\leq-(-\Delta)^{s}H(x,t)+\frac{(A-a)k^{4}\mu k(2\delta)^{2}}{2(A-k)^{2}(a-k)^{2}}\int_{B(x,\delta)}\left|\bigtriangledown u(y,t)\right|^{2}dy (3.10)
12(Aa)μkB(x,δ)u2(y,t)𝑑y.\displaystyle\qquad-\frac{1}{2}(A-a)\mu k\int_{B(x,\delta)}u^{2}(y,t)dy.

By choosing δ\delta sufficiently small such that

(Aa)k4μk(2δ)22(Ak)2(ak)21,\frac{(A-a)k^{4}\mu k(2\delta)^{2}}{2(A-k)^{2}(a-k)^{2}}\leq 1,

then

αtαH(x,t)(Δ)sH(x,t)+B(x,δ)|u(y,t)|2𝑑yD(x,t),\frac{\partial^{\alpha}}{\partial t^{\alpha}}H(x,t)\leq-(-\Delta)^{s}H(x,t)+\int_{B(x,\delta)}\left|\bigtriangledown u(y,t)\right|^{2}dy-D(x,t),

making

D(x,t)=12(Aa)μkB(x,δ)u2(y,t)𝑑y.D(x,t)=\frac{1}{2}(A-a)\mu k\int_{B(x,\delta)}u^{2}(y,t)dy.

Lemma 3.2.

[55][59] Assume T>0T>0 is a final time, 0<α,s<10<\alpha,s<1 and ff is a given function. Let u(x,t)u(x,t) is the solution of the following one-dimensional space-time fractional diffusion problem

{αtαu(x,t)=(Δ)su(x,t),1<x<1,0<t<T,u(1,t)=u(1,t)=0,0<t<T,u(x,0)=f(x),1<x<1.\displaystyle\left\{\begin{matrix}\frac{\partial^{\alpha}}{\partial t^{\alpha}}u(x,t)=-(-\Delta)^{s}u(x,t),&-1<x<1,0<t<T,\\ u(-1,t)=u(1,t)=0,&0<t<T,\\ u(x,0)=f(x),&-1<x<1.\end{matrix}\right. (3.11)

Then, we get the following useful formula for the weak solution of the direct problem (3.11)

u(x,t)=n=1f,ψnEα(λntα)ψn(x).u(x,t)=\sum_{n=1}^{\infty}\left\langle f,\psi_{n}\right\rangle E_{\alpha}(-\lambda_{n}t^{\alpha})\psi_{n}(x). (3.12)

the series is convergent in C((0,T];H2s(1,1))C((0,T];H^{2s}(-1,1)) where λn=(λ¯n)s,λ¯n\lambda_{n}=(\bar{\lambda}_{n})^{s},\bar{\lambda}_{n} and {ψn}n1\left\{\psi_{n}\right\}_{n\geq 1} are eigenvalues and eigenvectors of the classical Laplace operator Δ\Delta. ,\left\langle\cdot,\cdot\right\rangle denotes the standard inner product on L2(1,1)L^{2}(-1,1).

Lemma 3.3.

[60] Let ΩN\Omega\subset\mathbb{R}^{N} be a bounded domain with the C2C^{2} boundary Ω\partial\Omega, and define an unbounded linear operator AΩA_{\Omega} on L2(Ω)L^{2}(\Omega) as follows:

{D(AΩ)=H2(Ω)H01(Ω),AΩφ=Δφ,φD(AΩ).\displaystyle\left\{\begin{matrix}D(A_{\Omega})=H^{2}(\Omega)\cap H_{0}^{1}(\Omega),&\\ A_{\Omega}\varphi=-\Delta\varphi,\forall\varphi\in D(A_{\Omega}).&\end{matrix}\right.

We have that

{Dtα0CyAΩsy=λyin(0,+)y(0)=y0.\displaystyle\left\{\begin{matrix}{}_{0}^{C}\textrm{D}_{t}^{\alpha}y-A_{\Omega}^{s}y=-\lambda y&\text{in}\,(0,+\infty)\\ y(0)=y_{0}.&\end{matrix}\right. (3.13)

The solution to (3.13) is

y(x,t)=j=0Eα,1((λjs+λ)tα)y0,jej(x).y(x,t)=\sum_{j=0}^{\infty}E_{\alpha,1}(-(\lambda_{j}^{s}+\lambda)t^{\alpha})y_{0,j}e_{j}(x). (3.14)

where 0<α,s<1,y0,j=y0,ejL2(Ω)0<\alpha,s<1,y_{0,j}=\left\langle y_{0},e_{j}\right\rangle_{L^{2}(\Omega)} and denote by {λj}j=1\left\{\lambda_{j}\right\}_{j=1}^{\infty} with 0<λ1<λ2λ30<\lambda_{1}<\lambda_{2}\leq\lambda_{3}\leq\cdots the eigenvalues of AΩA_{\Omega} and {ej}j=1\left\{e_{j}\right\}_{j=1}^{\infty} with |ej|L2(Ω)=1\left|e_{j}\right|_{L^{2}(\Omega)}=1 the corresponding eigenvectors.

Definition 3.1.

[61] Let X1X_{1} be a Banach space,z0z_{0} belong to X1X_{1}, and fL1(0,T;X1)f\in L^{1}(0,T;X_{1}). The function z(x,t)C([0,T];X1)z(x,t)\in C([0,T];X_{1}) given by

z(t)=etetΔz0+0te(ts)e(ts)Δf(s)𝑑s,0tT,z(t)=e^{-t}e^{t\Delta}z_{0}+\int_{0}^{t}e^{-(t-s)}\cdot e^{(t-s)\Delta}f(s)ds,\qquad 0\leq t\leq T, (3.15)

is the mild solution of (3.15) on [0,T][0,T], where (etΔf)(x,t)=NG(xy,t)f(y)𝑑y(e^{t\Delta}f)(x,t)=\int_{\mathbb{R}^{N}}G(x-y,t)f(y)dy and G(x,t)G(x,t) is the heat kernel by G(x,t)=1(4πt)N/2exp(|x|24t)G(x,t)=\frac{1}{(4\pi t)^{N/2}}exp(-\frac{\left|x\right|^{2}}{4t}).

Lemma 3.4.

[61] Let 0qp,1q1p<1N0\leq q\leq p\leq\infty,\frac{1}{q}-\frac{1}{p}<\frac{1}{N} and suppose that zz is the function given by (3.15) and z0W1,p(N)z_{0}\in W^{1,p}(\mathbb{R}^{N}). If fL(0,;Lq(N))f\in L^{\infty}(0,\infty;L^{q}(\mathbb{R}^{N})), then

z(t)Lp(N)\displaystyle\left\|z(t)\right\|_{L^{p}(\mathbb{R}^{N})} \displaystyle\leq z0Lp(N)+CΓ(γ)sup0<s<tf(s)Lq(N),\displaystyle\left\|z_{0}\right\|_{L^{p}(\mathbb{R}^{N})}+C\cdot\Gamma(\gamma)\underset{0<s<t}{sup}\left\|f(s)\right\|_{L^{q}(\mathbb{R}^{N})},
z(t)Lp(N)\displaystyle\left\|\bigtriangledown z(t)\right\|_{L^{p}(\mathbb{R}^{N})} \displaystyle\leq z0Lp(N)+CΓ(γ~)sup0<s<tf(s)Lq(N),\displaystyle\left\|\bigtriangledown z_{0}\right\|_{L^{p}(\mathbb{R}^{N})}+C\cdot\Gamma(\tilde{\gamma})\underset{0<s<t}{sup}\left\|f(s)\right\|_{L^{q}(\mathbb{R}^{N})},

for t[0,)t\in[0,\infty), where CC is a positive constant independent of p,Γ()p,\Gamma(\cdot) is the gamma function, and γ=1(1q1p)N2,γ~=12(1q1p)N2\gamma=1-(\frac{1}{q}-\frac{1}{p})\cdot\frac{N}{2},\tilde{\gamma}=\frac{1}{2}-(\frac{1}{q}-\frac{1}{p})\cdot\frac{N}{2}.

Proof of Theorem 1.2 1.

From the proof of Theorem 1.1, for any K>0\forall{K}^{\prime}>0 , from (2.29) and the definition of MM, there exist μ(K)>0\mu^{*}({K}^{\prime})>0 and m0(K)>0m_{0}({K}^{\prime})>0 such that for μ(0,μ(K))\mu\in(0,\mu^{*}({K}^{\prime})) and u0L(N)<m0(K)\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}<m_{0}({K}^{\prime}), then for any (x,t)N×[0,+),M\left(x,t\right)\in\mathbb{R}^{N}\times[0,+\infty),M is sufficiently small such that

u(x,t)L(N×[0,+))<K.\left\|u(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N}\times[0,+\infty))}<{K}^{\prime}.
  1. 1.

    The case: u(x,t)L(N×[0,+))<K=γμ\left\|u(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N}\times[0,+\infty))}<{K}^{\prime}=\frac{\gamma}{\mu}.

Noticing σ=γμu(x,t)L(N×[0,))>0\sigma=\gamma-\mu\left\|u(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N}\times[0,\infty))}>0, we have

αutα=(Δ)su+μu2(1kJu)γu\displaystyle\frac{\partial^{\alpha}u}{\partial t^{\alpha}}=-(-\Delta)^{s}u+\mu u^{2}(1-kJ*u)-\gamma u (3.16)
=(Δ)su+u[μu(1kJu)γ]\displaystyle=-(-\Delta)^{s}u+u[\mu u(1-kJ*u)-\gamma]
=(Δ)suu[γμu(1kJu)]\displaystyle=-(-\Delta)^{s}u-u[\gamma-\mu u(1-kJ*u)]
(Δ)suu[γμu]\displaystyle\leq-(-\Delta)^{s}u-u[\gamma-\mu u]
(Δ)suuσ.\displaystyle\leq-(-\Delta)^{s}u-u\sigma.

From Lemma 3.2 and Lemma 3.3, Consider the following equation

{Dtα0Cw+(Δ)sw=σw,w(0)=u0L(N).\begin{cases}{}_{0}^{C}\textrm{D}_{t}^{\alpha}w+(-\Delta)^{s}w=-\sigma w,\\ w(0)=\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}.\end{cases} (3.17)

then, we obtain

w(x,t)=j=0Eα,1((λjs+σ)tα)y0,ejL2(Ω)ej(x).w(x,t)=\sum_{j=0}^{\infty}E_{\alpha,1}(-(\lambda_{j}^{s}+\sigma)t^{\alpha})\left\langle y_{0},e_{j}\right\rangle_{L^{2}(\Omega)}e_{j}(x). (3.18)

Therefore, we have

u(,t)L(N)\displaystyle\left\|u(\cdot,t)\right\|_{L^{\infty}(\mathbb{R}^{N})} w(x,t)L(N)\displaystyle\leq\left\|w(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N})}
j=0Eα,1((λjs+σ)tα)y0,ejL2(Ω)ej(x)L(N)\displaystyle\leq\left\|\sum_{j=0}^{\infty}E_{\alpha,1}(-(\lambda_{j}^{s}+\sigma)t^{\alpha})\left\langle y_{0},e_{j}\right\rangle_{L^{2}(\Omega)}e_{j}(x)\right\|_{L^{\infty}(\mathbb{R}^{N})}
u0L(N)e((λ1s+σ))1αt.\displaystyle\leq\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}e^{(-(\lambda_{1}^{s}+\sigma))^{\frac{1}{\alpha}t}}.
  1. 1.

    The case:u(x,t)L(N×[0,+))<K=a\left\|u(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N}\times[0,+\infty))}<{K}^{\prime}=a.

Denoting H0(y)=H(y,0)H_{0}(y)=H(y,0), from (3.2) in Proposition 3.1, for any (x,t)N×[0,+)(x,t)\in\mathbb{R}^{N}\times[0,+\infty), we have

H(x,t)H0L(N)+0tV(tr)u22𝑑r0tV(tr)D(r)𝑑r,H(x,t)\leq\left\|H_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}+\int_{0}^{t}V(t-r)\left\|\triangledown u\right\|_{2}^{2}dr-\int_{0}^{t}V(t-r)D(r)dr,

from which we obtain

0tV(tr)D(r)𝑑rH0L(N)+0tV(tr)u22𝑑r.\displaystyle\int_{0}^{t}V(t-r)D(r)dr\leq\left\|H_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})}+\int_{0}^{t}V(t-r)\left\|\triangledown u\right\|_{2}^{2}dr.

Due to the fact that uu is a classical solution, we have that

0tV(tr)D(r)𝑑rC2,1(N×[0,)),\int_{0}^{t}V(t-r)D(r)dr\in C^{2,1}(\mathbb{R}^{N}\times[0,\infty)),

which implies that for all xNx\in\mathbb{R}^{N}, the following limit holds:

limtlimrtV(tr)D(r)=0,\lim_{t\rightarrow\infty}\lim_{r\rightarrow t}V(t-r)D(r)=0,

or equivalently

limtlimrtV(tr)Nu2(z,r)𝑑z𝑑y=0,\lim_{t\rightarrow\infty}\lim_{r\rightarrow t}V(t-r)\int_{\mathbb{R}^{N}}u^{2}(z,r)dzdy=0,

which together with the fact that the heat kernel converges to delta function as rtr\rightarrow t, we have that for any xNx\in\mathbb{R}^{N},

limtNu2(y,t)𝑑y=0.\lim_{t\rightarrow\infty}\int_{\mathbb{R}^{N}}u^{2}(y,t)dy=0.

Furthermore, with the uniform boundedness of uu on N×[0,)\mathbb{R}^{N}\times[0,\infty), following Lemma 3.4, we can obtain the global boundedness of u(,t)L(N)\left\|\bigtriangledown u(\cdot,t)\right\|_{L^{\infty}(\mathbb{R}^{N})}, from which and Lemma 2.12 with Ω=N,p=q=,r=2\Omega=\mathbb{R}^{N},p=q=\infty,r=2, the convergence of u(,t)L(N)\left\|u(\cdot,t)\right\|_{L^{\infty}(\mathbb{R}^{N})} follows from the convergences of u(,t)L2(N)\left\|u(\cdot,t)\right\|_{L^{2}(\mathbb{R}^{N})} immediately. This is, we obtain

u(,t)L(N)0\left\|u(\cdot,t)\right\|_{L^{\infty}(\mathbb{R}^{N})}\rightarrow 0

as t0t\rightarrow 0. Therefore, for any compact set in N\mathbb{R}^{N}, by finite covering, we obtain that uu converges to 0 uniformly in that compact set, which means that uu converges locally uniformly to 0 in N\mathbb{R}^{N} as \rightarrow\infty. The proof is complete.

4 Global boundness of solutions for a nonlinear TSFNRDE

In the section, we will make J=k=μ=γ=1J=k=\mu=\gamma=1 and replace fractional diffusion (Δ)su(-\Delta)^{s}u as nonlinear fractional diffusion (Δ)sum(-\Delta)^{s}u^{m} in (1.1)-(1.2) to get equation (1.8)-(1.9). And we make f(x,t)=u2(1Nu𝑑x)u,f(x,t)=u^{2}(1-\int_{\mathbb{R}^{N}}udx)-u, then equation (1.8)-(1.9) can write the following form

{αutα+(Δ)s(um)=f(x,t),xN×(0,T]u(x,0)=u0(x),xN.\displaystyle\left\{\begin{matrix}\frac{\partial^{\alpha}u}{\partial t^{\alpha}}+\left(-\Delta\right)^{s}(u^{m})=f(x,t),&x\in\mathbb{R}^{N}\times(0,T]\\ u(x,0)=u_{0}(x),&x\in\mathbb{R}^{N}.\end{matrix}\right. (4.1)
Definition 4.1.

A function uu is a weak solution to the problem (4.1) if:

  • 1.

    uL2((0,T];H01(N))C((0,T];L1(N))u\in L^{2}((0,T];H_{0}^{1}(\mathbb{R}^{N}))\cap C((0,T];L^{1}(\mathbb{R}^{N})) and

    |u|m1uLloc2((0,T];Hs(N));\left|u\right|^{m-1}u\in L_{loc}^{2}((0,T];H^{s}(\mathbb{R}^{N}));
  • 2.

    identity

    0TNuαφtα𝑑x𝑑t+0TN(Δ)s/2(um)(Δ)s/2φ𝑑x𝑑t\displaystyle\int_{0}^{T}\int_{\mathbb{R}^{N}}u\frac{\partial^{\alpha}\varphi}{\partial t^{\alpha}}dxdt+\int_{0}^{T}\int_{\mathbb{R}^{N}}(-\Delta)^{s/2}(u^{m})(-\Delta)^{s/2}\varphi dxdt
    =0TNfφ𝑑x𝑑t\displaystyle=\int_{0}^{T}\int_{\mathbb{R}^{N}}f\varphi dxdt

    holds for every φC01(N×(0,T])\varphi\in C_{0}^{1}(\mathbb{R}^{N}\times(0,T]);

  • 3.

    u(,0)=u0L1(N)u(\cdot,0)=u_{0}\in L^{1}(\mathbb{R}^{N}) almost everywhere.

Definition 4.2.

We say that a weak solution uu to the problem (4.1) is a strong solution if moreover tαL((τ,);L1(N)),τ>0.\partial_{t}^{\alpha}\in L^{\infty}((\tau,\infty);L^{1}(\mathbb{R}^{N})),\tau>0.

Remark 4.1.

The above two definitions are mainly combined with references [1, 62]. On the one hand, [62] gives the definition of weak existence for time-fractional nonlinear diffusion equations. On the other hand, [1] gives the definition of existence of weak solution/strong solution of fractional nonlinear diffusion equations. For specific content, you can view Appendix A.3.

Lemma 4.1.

[63] Let φC2(N)\varphi\in C^{2}(\mathbb{R}^{N}) and positive real function that is radially symmetric and decreasing in |x|1\left|x\right|\geq 1. Assume also that φ(x)|x|β\varphi(x)\leq\left|x\right|^{-\beta} and that |D2φ(x)|c0|x|β2\left|D^{2}\varphi(x)\right|\leq c_{0}\left|x\right|^{-\beta-2}, for some positive constant β\beta and for |x|\left|x\right| large enough. Then, for all |x||x0|1\left|x\right|\geq\left|x_{0}\right|\gg 1 we have

|(Δ)sφ(x)|{c1|x|β+2s,ifβ<N,c2log|x||x|N+2s,ifβ=N,c3|x|N+2s,ifβ>N,\left|(-\Delta)^{s}\varphi(x)\right|\leq\left\{\begin{matrix}\frac{c_{1}}{\left|x\right|^{\beta+2s}},&\text{if}\quad\beta<N,\\ \frac{c_{2}log\left|x\right|}{\left|x\right|^{N+2s}},&\text{if}\quad\beta=N,\\ \frac{c_{3}}{\left|x\right|^{N+2s}},&\text{if}\quad\beta>N,\end{matrix}\right. (4.2)

with positive constant c1,c2,c3>0c_{1},c_{2},c_{3}>0 that depend only on β,s,N\beta,s,N and φC2(N)\left\|\varphi\right\|_{C^{2}(\mathbb{R}^{N})}. For β>N\beta>N the reverse estimate holds from below if φ0:|(Δ)sφ(x)|c4|x|(N+2s)\varphi\geqslant 0:\left|(-\Delta)^{s}\varphi(x)\right|\geqslant c_{4}\left|x\right|^{-(N+2s)} for all |x||x0|1.\left|x\right|\geqslant\left|x_{0}\right|\gg 1.

Lemma 4.2.

(Weighted L1L^{1} estimates). Let uvu\geq v be two ordered solutions to Eq.(4.1), with 0<m<10<m<1. Let φR=φ(x/R)\varphi_{R}=\varphi(x/R) where R>0R>0 and φ\varphi is as in the previous lemma with 0φ|x|β0\leqslant\varphi\leqslant\left|x\right|^{-\beta} for |x|1\left|x\right|\gg 1 and

N2s1m<β<N+2sm.N-\frac{2s}{1-m}<\beta<N+\frac{2s}{m}.

Then, for all 0τ,t<T0\leqslant\tau,t<T we have

(N(u(x,t)v(x,t))φR(x)𝑑x)1m\displaystyle\left(\int_{\mathbb{R}^{N}}(u(x,t)-v(x,t))\varphi_{R}(x)dx\right)^{1-m}
(N(u(x,τ)v(x,τ))φR(x)𝑑x)1m+C1εTα(1m)(αΓ(α))1mR2sN(1m)\displaystyle\leqslant\left(\int_{\mathbb{R}^{N}}(u(x,\tau)-v(x,\tau))\varphi_{R}(x)dx\right)^{1-m}+\frac{C_{1}\varepsilon T^{\alpha(1-m)}}{(\alpha\Gamma(\alpha))^{1-m}R^{2s-N(1-m)}} (4.3)

with C1,ε>0C_{1},\varepsilon>0 that depends only on β,m,N\beta,m,N

Proof.
  1. 1.

    A fractional differential inequality for the weighted L1L^{1}-norm.

If ψ\psi is a smooth and sufficiently decaying function and Lemma 2.9, we have

|N(0CDtαu(x,t)0CDtαv(x,t))ψ(x)dx|\displaystyle\left|\int_{\mathbb{R}^{N}}(_{0}^{C}\textrm{D}_{t}^{\alpha}u(x,t)-_{0}^{C}\textrm{D}_{t}^{\alpha}v(x,t))\psi(x)dx\right|
=|Dtα0CN(u(x,t)v(x,t))ψ(x)𝑑x|=|N((Δ)sum(Δ)svm)ψ𝑑x|\displaystyle=\left|{}_{0}^{C}\textrm{D}_{t}^{\alpha}\int_{\mathbb{R}^{N}}(u(x,t)-v(x,t))\psi(x)dx\right|=\left|\int_{\mathbb{R}^{N}}((-\Delta)^{s}u^{m}-(-\Delta)^{s}v^{m})\psi dx\right|
=(a)|N(umvm)(Δ)sψ𝑑x|(b)21mN(uv)m|(Δ)sψ|𝑑x\displaystyle=_{(a)}\left|\int_{\mathbb{R}^{N}}(u^{m}-v^{m})(-\Delta)^{s}\psi dx\right|\leqslant_{(b)}2^{1-m}\int_{\mathbb{R}^{N}}(u-v)^{m}\left|(-\Delta)^{s}\psi\right|dx
(c)2(N(uv)ψ𝑑x)m(N|(Δ)sψ|11mψm1m𝑑x)1m.\displaystyle\leqslant_{(c)}2\left(\int_{\mathbb{R}^{N}}(u-v)\psi dx\right)^{m}\left(\int_{\mathbb{R}^{N}}\frac{\left|(-\Delta)^{s}\psi\right|^{\frac{1}{1-m}}}{\psi^{\frac{m}{1-m}}}dx\right)^{1-m}.

Notice that in (a)(a) we used that the fact that (Δ)s(-\Delta)^{s} is a symmetric operator, while in (b)(b) we have used that (umvm)21m(uv)m,(u^{m}-v^{m})\leqslant 2^{1-m}(u-v)^{m}, where um=|u|m1uu^{m}=\left|u\right|^{m-1}u as mentioned. In (c)(c) we have used Hölder inequality with conjugate exponents 1/m>11/m>1 and 1/(1m).1/(1-m). If the last integral factor is bounded, then we get

|Dtα0CN(u(x,t)v(x,t))ψ(x)𝑑x|Cψ1m(N(u(x,t)v(x,t))ψ(x)𝑑x)m\left|{}_{0}^{C}\textrm{D}_{t}^{\alpha}\int_{\mathbb{R}^{N}}(u(x,t)-v(x,t))\psi(x)dx\right|\leqslant C_{\psi}^{1-m}\left(\int_{\mathbb{R}^{N}}(u(x,t)-v(x,t))\psi(x)dx\right)^{m}

and let y(t)=N(u(x,t)v(x,t))ψ(x)𝑑xy(t)=\int_{\mathbb{R}^{N}}(u(x,t)-v(x,t))\psi(x)dx, we can get fractional differential inequality on (τ,t)(\tau,t)

Dtα0Cy(t)Cψ1mym(t){}_{0}^{C}\textrm{D}_{t}^{\alpha}y(t)\leqslant C_{\psi}^{1-m}y^{m}(t)

By Young inequality 2.13, let a=Cψ1m,b=ym(t),q=1m>1,p=11m,a=C_{\psi}^{1-m},b=y^{m}(t),q=\frac{1}{m}>1,p=\frac{1}{1-m}, then

Cψ1mym(t)ε11mCψ+mε1my(t),C_{\psi}^{1-m}y^{m}(t)\leqslant\varepsilon^{\frac{1}{1-m}}C_{\psi}+\frac{m}{\varepsilon^{\frac{1}{m}}}y(t),

So we have

Dtα0Cy(t)ε11mCψ+mε1my(t),{}_{0}^{C}\textrm{D}_{t}^{\alpha}y(t)\leqslant\varepsilon^{\frac{1}{1-m}}C_{\psi}+\frac{m}{\varepsilon^{\frac{1}{m}}}y(t),

and by Lemma 2.10, then we have

(N(u(x,t)v(x,t))φR(x)𝑑x)1m(N(u(x,τ)v(x,τ))φR(x)𝑑x)1m\displaystyle\left(\int_{\mathbb{R}^{N}}(u(x,t)-v(x,t))\varphi_{R}(x)dx\right)^{1-m}-\left(\int_{\mathbb{R}^{N}}(u(x,\tau)-v(x,\tau))\varphi_{R}(x)dx\right)^{1-m}
Cψ1mε(TααΓ(α))1m.\displaystyle\leqslant C_{\psi}^{1-m}\varepsilon(\frac{T^{\alpha}}{\alpha\Gamma(\alpha)})^{1-m}.

Now, we estimate the constant CψC_{\psi}, for a convenient choice of test.

  1. 1.

    Estimating the constant CψC_{\psi}.

Choose ψ(x)=φR(x):=φ(x/R)=φ(y)\psi(x)=\varphi_{R}(x):=\varphi(x/R)=\varphi(y), with φ\varphi as in Lemma 4.1 and y=x/Ry=x/R, so that (Δ)sψ(x)=(Δ)sφR(x)=R2s(Δ)sφ(y)\left(-\Delta\right)^{s}\psi(x)=\left(-\Delta\right)^{s}\varphi_{R}(x)=R^{-2s}\left(-\Delta\right)^{s}\varphi(y), then

Cψ=N|(Δ)sφR(x)|11mφR(x)m1m𝑑x=RN2s1mN|(Δ)sφ(y)|11mφ(y)m1m𝑑y\displaystyle C_{\psi}=\int_{\mathbb{R}^{N}}\frac{\left|(-\Delta)^{s}\varphi_{R}(x)\right|^{\frac{1}{1-m}}}{\varphi_{R}(x)^{\frac{m}{1-m}}}dx=R^{N-\frac{2s}{1-m}}\int_{\mathbb{R}^{N}}\frac{\left|(-\Delta)^{s}\varphi(y)\right|^{\frac{1}{1-m}}}{\varphi(y)^{\frac{m}{1-m}}}dy
=RN2s1m[B2|(Δ)sφ(y)|11mφ(y)m1m𝑑y+B2c|(Δ)sφ(y)|11mφ(y)m1m𝑑y]\displaystyle=R^{N-\frac{2s}{1-m}}\left[\int_{B_{2}}\frac{\left|(-\Delta)^{s}\varphi(y)\right|^{\frac{1}{1-m}}}{\varphi(y)^{\frac{m}{1-m}}}dy+\int_{B_{2}^{c}}\frac{\left|(-\Delta)^{s}\varphi(y)\right|^{\frac{1}{1-m}}}{\varphi(y)^{\frac{m}{1-m}}}dy\right]
=k1RN2s1m\displaystyle=k_{1}R^{N-\frac{2s}{1-m}}

where it is easy to check that first integral is bounded, since φk2>0\varphi\geqslant k_{2}>0 on B2B_{2}, and when |y|>|x0|\left|y\right|>\left|x_{0}\right| with |x0|1\left|x_{0}\right|\gg 1 we know by estimate (4.2) that

|(Δ)sφ(y)|11mφ(y)m1m{k3|y|β+2s1m,ifβ<N,k4log|y||y|N+2s1m,ifβ=N,k5|y|N+2sβm1m,ifβ>N,\frac{\left|(-\Delta)^{s}\varphi(y)\right|^{\frac{1}{1-m}}}{\varphi(y)^{\frac{m}{1-m}}}\leqslant\left\{\begin{matrix}\frac{k_{3}}{\left|y\right|^{\beta+\frac{2s}{1-m}}},&\text{if}\quad\beta<N,\\ \frac{k_{4}log\left|y\right|}{\left|y\right|^{N+\frac{2s}{1-m}}},&\text{if}\quad\beta=N,\\ \frac{k_{5}}{\left|y\right|^{\frac{N+2s-\beta m}{1-m}}},&\text{if}\quad\beta>N,\end{matrix}\right. (4.4)

therefore k1k_{1} is finite whenever N2s1m<β<N+2sm.N-\frac{2s}{1-m}<\beta<N+\frac{2s}{m}. Note that all the constants ki,i=1,2,3,4,5k_{i},i=1,2,3,4,5 depend only on β,m,N.\beta,m,N.

Remark 4.2.

Using the method of proof in [Theorem 2.2, [63]], and when 0<m<mc=(N2s)/N0<m<m_{c}=(N-2s)/N solution corresponding to u0L1(N)Lp(N)u_{0}\in L^{1}(\mathbb{R}^{N})\cap L^{p}(\mathbb{R}^{N}) with pN(1m)/2sp\geqslant N(1-m)/2s. On the other hand, when mc<m<1m_{c}<m<1, the estimate implies the conservation of mass by letting RR\rightarrow\infty. By [1], the above estimates provide a lower bound for the extinction time in such a case, just by letting τ=T,T>0\tau=T,T>0 and t=0t=0 in the above estimates:

1C1RN(1m)2s(Nu0φR𝑑x)1mT.\frac{1}{C_{1}R^{N(1-m)-2s}}\left(\int_{\mathbb{R}^{N}}u_{0}\varphi_{R}dx\right)^{1-m}\leqslant T.

Lemma 4.3.

[8] Suppose that a nonnegative function u(t)0u(t)\geq 0 satisfies

0CDtαu(t)+c1u(t)f(t)_{0}^{C}\textrm{D}_{t}^{\alpha}u(t)+c_{1}u(t)\leq f(t) (4.5)

for almost all t[0,T]t\in[0,T], where c1>0c_{1}>0, and the function f(t)f(t) is nonnegative and integrable for t[0,T]t\in[0,T]. Then

u(t)u(0)+1Γ(α)0t(ts)α1f(s)𝑑s.u(t)\leq u(0)+\frac{1}{\Gamma(\alpha)}\int_{0}^{t}(t-s)^{\alpha-1}f(s)ds. (4.6)

Lemma 4.4.

Assume the function yk(t)y_{k}(t) is nonnegative and exists the Caputo fractional derivative for t[0,T]t\in[0,T] satisfying

0CDtαyk(t)C9(yk(t))k+m1kyk+ak(yk1γ1(t)+yk1γ2(t)),_{0}^{C}\textrm{D}_{t}^{\alpha}y_{k}(t)\leq-C_{9}(y_{k}(t))^{\frac{k+m-1}{k}}-y_{k}+a_{k}(y_{k-1}^{\gamma_{1}}(t)+y_{k-1}^{\gamma_{2}}(t)), (4.7)

where ak=a¯3rk>1a_{k}=\bar{a}3^{rk}>1 with a¯,r\bar{a},r are positive bounded constants and 0<γ2<γ130<\gamma_{2}<\gamma_{1}\leq 3. Assume also that there exists a bounded constant K1K\geq 1 such that yk(0)K3ky_{k}(0)\leq K^{3^{k}}, then

yk(t)(2a¯)3k123r(3k+14k234)max{supt[0,T]y03k(t),K3k}TααΓ(α).y_{k}(t)\leq(2\bar{a})^{\frac{3^{k}-1}{2}}3^{r(\frac{3^{k+1}}{4}-\frac{k}{2}-\frac{3}{4})}max\left\{\underset{t\in[0,T]}{sup}y_{0}^{3^{k}}(t),K^{3^{k}}\right\}\frac{T^{\alpha}}{\alpha\Gamma(\alpha)}. (4.8)

Proof.

From Lemma 4.3 and Eqution (4.6), we make

f(t)=ak(yk1γ1(t)+yk1γ2(t))2akmax{1,supt[0,T]yk13(t)},C9(yk(t))k+m1k0f(t)=a_{k}(y_{k-1}^{\gamma_{1}}(t)+y_{k-1}^{\gamma_{2}}(t))\leq 2a_{k}max\left\{1,\underset{t\in[0,T]}{sup}y_{k-1}^{3}(t)\right\},C_{9}(y_{k}(t))^{\frac{k+m-1}{k}}\geq 0

then

yk(t)yk(0)+1Γ(α)0t(ts)α1f(s)𝑑s\displaystyle y_{k}(t)\leq y_{k}(0)+\frac{1}{\Gamma(\alpha)}\int_{0}^{t}(t-s)^{\alpha-1}f(s)ds (4.9)
K3k+2akΓ(α)0t(ts)α1max{1,supt[0,T]yk13(s)}𝑑s\displaystyle\leq K^{3^{k}}+\frac{2a_{k}}{\Gamma(\alpha)}\int_{0}^{t}(t-s)^{\alpha-1}max\left\{1,\underset{t\in[0,T]}{sup}y_{k-1}^{3}(s)\right\}ds
K3k+2akΓ(α)max{1,supt[0,T]yk13(t)}0t(ts)α1𝑑s\displaystyle\leq K^{3^{k}}+\frac{2a_{k}}{\Gamma(\alpha)}max\left\{1,\underset{t\in[0,T]}{sup}y_{k-1}^{3}(t)\right\}\int_{0}^{t}(t-s)^{\alpha-1}ds
K3k+TααΓ(α)2akmax{1,supt[0,T]yk13(t)}\displaystyle\leq K^{3^{k}}+\frac{T^{\alpha}}{\alpha\Gamma(\alpha)}2a_{k}max\left\{1,\underset{t\in[0,T]}{sup}y_{k-1}^{3}(t)\right\}
TααΓ(α)2akmax{K3k,supt[0,T]yk13(t)}.\displaystyle\leq\frac{T^{\alpha}}{\alpha\Gamma(\alpha)}2a_{k}max\left\{K^{3^{k}},\underset{t\in[0,T]}{sup}y_{k-1}^{3}(t)\right\}.

Then from (4.9) after some iterative steps we have

yk(t)2ak(2ak1)3(2ak2)32(2ak3)33(2a1)3k1max{K3k,supt[0,T]y03k(t)}\displaystyle y_{k}(t)\leq 2a_{k}(2a_{k-1})^{3}(2a_{k-2})^{3^{2}}(2a_{k-3})^{3^{3}}\cdots(2a_{1})^{3^{k-1}}max\left\{K^{3^{k}},\underset{t\in[0,T]}{sup}y_{0}^{3^{k}}(t)\right\}
TααΓ(α)\displaystyle\qquad\frac{T^{\alpha}}{\alpha\Gamma(\alpha)}
=(2a¯)1+3+32+33+3k13r(k+3(k1)+32(k2)++3k1)max{K3k,supt[0,T]y03k(t)}\displaystyle=(2\bar{a})^{1+3+3^{2}+3^{3}+\cdots 3^{k-1}}3^{r(k+3(k-1)+3^{2}(k-2)+\cdots+3^{k-1})}max\left\{K^{3^{k}},\underset{t\in[0,T]}{sup}y_{0}^{3^{k}}(t)\right\}
TααΓ(α)\displaystyle\qquad\frac{T^{\alpha}}{\alpha\Gamma(\alpha)}
=(2a¯)3k123r(3k+14k234)max{K3k,supt[0,T]y03k(t)}TααΓ(α).\displaystyle=(2\bar{a})^{\frac{3^{k}-1}{2}}3^{r(\frac{3^{k+1}}{4}-\frac{k}{2}-\frac{3}{4})}max\left\{K^{3^{k}},\underset{t\in[0,T]}{sup}y_{0}^{3^{k}}(t)\right\}\frac{T^{\alpha}}{\alpha\Gamma(\alpha)}.

Remark 4.3.

The proof of this Lemma is based on the process of Lemma 4 in [64], where the nature of yk(t)0,k=0,1,2,y_{k}(t)\geq 0,k=0,1,2,\cdots is scratched and the desired result is obtained.

Lemma 4.5.

Suppose 0<k<m<10<k<m<1 and b(t)b(t) is continuous and boundary. And let y(t)0y(t)\geq 0 be a solution of the fractional differential inequality

0CDtαy(t)+αyk(t)+βy(t)b(t)ym+c4._{0}^{C}\textrm{D}_{t}^{\alpha}y(t)+\alpha y^{k}(t)+\beta y(t)\leq b(t)y^{m}+c_{4}. (4.10)

For almost all t[0,T]t\in[0,T], then

y(t)\displaystyle y(t) y(0)+[[λky1k(0)+(c4α)(1k)]TααΓ(α)]11k\displaystyle\leq y(0)+\left[\frac{[\lambda_{k}y^{1-k}(0)+(c_{4}-\alpha)(1-k)]T^{\alpha}}{\alpha\Gamma(\alpha)}\right]^{\frac{1}{1-k}}
+(1m)11kε11mTα1kα11kΓ11k(α)b11m(t).\displaystyle\qquad+(1-m)^{\frac{1}{1-k}}\varepsilon^{\frac{1}{1-m}}\frac{T^{\frac{\alpha}{1-k}}}{\alpha^{\frac{1}{1-k}}\Gamma^{\frac{1}{1-k}}(\alpha)}b^{\frac{1}{1-m}}(t).

where λk=mkε1kmkβ(1k)\lambda_{k}=-\frac{m-k}{\varepsilon^{\frac{1-k}{m-k}}}-\beta(1-k) and α,β,c4,ε>0\alpha,\beta,c_{4},\varepsilon>0 are all contants.

Proof.

Multiplying (1k)yk(1-k)y^{-k} to both side of (4.10) yields

Dtα0Cy1k(t)+α(1k)+β(1k)y1k(t)(1k)b(t)ymk+c4(1k)yk(t).{}_{0}^{C}\textrm{D}_{t}^{\alpha}y^{1-k}(t)+\alpha(1-k)+\beta(1-k)y^{1-k}(t)\leq(1-k)b(t)y^{m-k}+c_{4}(1-k)y^{-k}(t).

Let z(t)=y1k(t),0<yk(t)<1z(t)=y^{1-k}(t),0<y^{-k}(t)<1, then we have

Dtα0Cz(t)b(t)(1k)zmk1k(t)β(1k)z(t)+(c4α)(1k).{}_{0}^{C}\textrm{D}_{t}^{\alpha}z(t)\leq b(t)(1-k)z^{\frac{m-k}{1-k}}(t)-\beta(1-k)z(t)+(c_{4}-\alpha)(1-k).

By Young’s inequality, let 1β~=mk1k,a=b(t),b=(z(t))1β~,11β~=q>1,p=1β~1-\tilde{\beta}=\frac{m-k}{1-k},a=b(t),b=(z(t))^{1-\tilde{\beta}},\frac{1}{1-\tilde{\beta}}=q>1,p=\frac{1}{\tilde{\beta}}, then we have

b(t)(1k)zmk1k(t)(1m)εpbp(t)+mkε11β~z(t),b(t)(1-k)z^{\frac{m-k}{1-k}}(t)\leq(1-m)\varepsilon^{p}b^{p}(t)+\frac{m-k}{\varepsilon^{\frac{1}{1-\tilde{\beta}}}}z(t),

then we can get

Dtα0Cz(t)[mkε11β~β(1k)]z(t)+(c4α)(1k)+(1m)εpbp(t).{}_{0}^{C}\textrm{D}_{t}^{\alpha}z(t)\leq[\frac{m-k}{\varepsilon^{\frac{1}{1-\tilde{\beta}}}}-\beta(1-k)]z(t)+(c_{4}-\alpha)(1-k)+(1-m)\varepsilon^{p}b^{p}(t).

Let λk=mkε11β~β(1k)\lambda_{k}=-\frac{m-k}{\varepsilon^{\frac{1}{1-\tilde{\beta}}}}-\beta(1-k) and By Lemma 4.3, then we obtain

z(t)\displaystyle z(t) z(0)+[λkz(0)+(c4α)(1k)]TααΓ(α)\displaystyle\leq z(0)+\frac{[\lambda_{k}z(0)+(c_{4}-\alpha)(1-k)]T^{\alpha}}{\alpha\Gamma(\alpha)}
+(1m)εp1Γ(α)0t(ts)α1bp(s)𝑑s\displaystyle\qquad+(1-m)\varepsilon^{p}\frac{1}{\Gamma(\alpha)}\int_{0}^{t}(t-s)^{\alpha-1}b^{p}(s)ds

then the solution of (4.10) can be estimated as

y(t)\displaystyle y(t) [y1k(0)+[λky1k(0)+(c4α)(1k)]TααΓ(α)\displaystyle\leq[y^{1-k}(0)+\frac{[\lambda_{k}y^{1-k}(0)+(c_{4}-\alpha)(1-k)]T^{\alpha}}{\alpha\Gamma(\alpha)}
+(1m)εp1Γ(α)0t(ts)α1bp(s)ds]11k.\displaystyle\qquad+(1-m)\varepsilon^{p}\frac{1}{\Gamma(\alpha)}\int_{0}^{t}(t-s)^{\alpha-1}b^{p}(s)ds]^{\frac{1}{1-k}}. (4.11)

Using the inequality A+BnAn+Bn,A,B>0,\sqrt[n]{A+B}\leq\sqrt[n]{A}+\sqrt[n]{B},A,B>0, we can transform (Proof) into

y(t)\displaystyle y(t) y(0)+[[λky1k(0)+(c4α)(1k)]TααΓ(α)]11k\displaystyle\leq y(0)+\left[\frac{[\lambda_{k}y^{1-k}(0)+(c_{4}-\alpha)(1-k)]T^{\alpha}}{\alpha\Gamma(\alpha)}\right]^{\frac{1}{1-k}}
+(1m)11kε11mTα1kα11kΓ11k(α)b11m(t).\displaystyle\qquad+(1-m)^{\frac{1}{1-k}}\varepsilon^{\frac{1}{1-m}}\frac{T^{\frac{\alpha}{1-k}}}{\alpha^{\frac{1}{1-k}}\Gamma^{\frac{1}{1-k}}(\alpha)}b^{\frac{1}{1-m}}(t).

Remark 4.4.

By using Young inequality, we first will inequality (4.10) Write the form of fractional differential inequality (4.3). Then using the method of proof [Theorem 1[65]], we can the solution of above mentioned inequality(4.10).

Lemma 4.6.

[66] Assume that (a,b)(+)2,0<α<1(a,b)\in(\mathbb{R}^{+})^{2},0<\alpha<1, then there exist c1,c2,c3>0c_{1},c_{2},c_{3}>0, such that

(a+b)αc1aα+c2bα(a+b)^{\alpha}\leq c_{1}a^{\alpha}+c_{2}b^{\alpha}

and

(ab)(aαbα)c3|aα2bα2|2(a-b)(a^{\alpha}-b^{\alpha})\geq c_{3}\left|a^{\frac{\alpha}{2}}-b^{\frac{\alpha}{2}}\right|^{2} (4.12)

Lemma 4.7.

[66] Assume that 0<s<10<s<1, then there exists a positive constant SS(N,s)S\equiv S(N,s) such tnat for all vC0(N)v\in C_{0}^{\infty}(\mathbb{R}^{N}),

NN|v(x)v(y)|2|xy|N+2s𝑑x𝑑yS(N|v(x)|ps𝑑x)2ps,\int_{\mathbb{R}^{N}}\int_{\mathbb{R}^{N}}\frac{\left|v(x)-v(y)\right|^{2}}{\left|x-y\right|^{N+2s}}dxdy\geq S(\int_{\mathbb{R}^{N}}\left|v(x)\right|^{p_{s}^{*}}dx)^{\frac{2}{p_{s}^{*}}},

where ps=2NN2s.p_{s}^{*}=\frac{2N}{N-2s}.

Lemma 4.8.

[64] Let N3,q>1,m>12/NN\geq 3,q>1,m>1-2/N, assume uL+1(N)u\in L_{+}^{1}(\mathbb{R}^{N}) and um+q12H1(N)u^{\frac{m+q-1}{2}}\in H^{1}(\mathbb{R}^{N}), then

(uqq)1+m1+2/Nq1SN1u(q+m1)/222u11q1(2q/N+m1).(\left\|u\right\|_{q}^{q})^{1+\frac{m-1+2/N}{q-1}}\leq S_{N}^{-1}\left\|\triangledown u^{(q+m-1)/2}\right\|_{2}^{2}\left\|u\right\|_{1}^{\frac{1}{q-1}(2q/N+m-1)}.

Lemma 4.9.

[67][68] When the parameters p,q,rp,q,r meet any of the following conditions:

(i) q>N1q>N\geq 1, r1r\geq 1 and p=p=\infty;

(ii)q>max{1,2NN+2},1r<σq>max\left\{1,\frac{2N}{N+2}\right\},1\leq r<\sigma and r<p<σ+1r<p<\sigma+1 in

σ:={(q1)N+qNq,q<N,,qN.\sigma:=\begin{cases}\frac{(q-1)N+q}{N-q},&q<N,\\ \infty,&q\geq N.\end{cases}

Then the following inequality is established

uLp(N)CGNuLr(N)1λuLq(N)λ\left\|u\right\|_{L^{p}(\mathbb{R}^{N})}\leq C_{GN}\left\|u\right\|_{L^{r}(\mathbb{R}^{N})}^{1-\lambda^{*}}\left\|\bigtriangledown u\right\|_{L^{q}(\mathbb{R}^{N})}^{\lambda^{*}}

among

λ=qN(pr)p[N(qr)+qr].\lambda^{*}=\frac{qN(p-r)}{p[N(q-r)+qr]}.

Lemma 4.10.

[69] Let N1N\geq 1. pp is the exponent from the Sobolev embedding theorem, i.e.i.e.

{p=2NN2,N32<p<,N=2p=,N=1\begin{cases}p=\frac{2N}{N-2},&N\geq 3\\ 2<p<\infty,&N=2\\ p=\infty,&N=1\end{cases} (4.13)

1r<q<p1\leq r<q<p and qr<2r+12p\frac{q}{r}<\frac{2}{r}+1-\frac{2}{p}, then for vH(N)v\in{H}^{\prime}(\mathbb{R}^{N}) and vLr(N)v\in L^{r}(\mathbb{R}^{N}), it holds

vLq(N)q\displaystyle\left\|v\right\|_{L^{q}(\mathbb{R}^{N})}^{q} C(N)c0λq2λqvLp(N)γ+c0uL2(N)2,N>2,\displaystyle\leq C(N)c^{\frac{\lambda q}{2-\lambda q}}_{0}\left\|v\right\|_{L^{p}(\mathbb{R}^{N})}^{\gamma}+c_{0}\left\|\bigtriangledown u\right\|_{L^{2}(\mathbb{R}^{N})}^{2},\quad N>2, (4.14)
vLq(N)q\displaystyle\left\|v\right\|_{L^{q}(\mathbb{R}^{N})}^{q} C(N)(c0λq2λq+c1λq2λq)vLr(N)γ+c0uL2(N)2\displaystyle\leq C(N)(c^{\frac{\lambda q}{2-\lambda q}}_{0}+c^{-\frac{\lambda q}{2-\lambda q}}_{1})\left\|v\right\|_{L^{r}(\mathbb{R}^{N})}^{\gamma}+c_{0}\left\|\bigtriangledown u\right\|_{L^{2}(\mathbb{R}^{N})}^{2}
+c1vL2(N)2,N=1,2.\displaystyle\qquad+c_{1}\left\|v\right\|_{L^{2}(\mathbb{R}^{N})}^{2},\quad N=1,2.

Here C(N)C(N) are constants depending on N,c0,c1N,c_{0},c_{1} are arbitrary positive constants and

λ=1r1q1r1p(0,1),γ=2(1λ)q2λq=2(1qp)2qr2p+1.\lambda=\frac{\frac{1}{r}-\frac{1}{q}}{\frac{1}{r}-\frac{1}{p}}\in(0,1),\gamma=\frac{2(1-\lambda)q}{2-\lambda q}=\frac{2(1-\frac{q}{p})}{\frac{2-q}{r}-\frac{2}{p}+1}.

Proof of Theorem 1.3 1.
  1. 1.

    Existence and unique weak solution for problem (4.1).

(1) We first verify uL2((0,T];H01(N))u\in L^{2}((0,T];H_{0}^{1}(\mathbb{R}^{N})) and |u|m1uLloc2((0,T];Hs(N)).\left|u\right|^{m-1}u\in L_{loc}^{2}((0,T];H^{s}(\mathbb{R}^{N})).

By [62], we denote the space L2(0,T;H01(N))L^{2}(0,T;H_{0}^{1}(\mathbb{R}^{N})) by VV and define Lu:=L^u,φ,L^u:=αutα,L^:D(L^)VVLu:=\left\langle\hat{L}u,\varphi\right\rangle,\hat{L}u:=\frac{\partial^{\alpha}u}{\partial t^{\alpha}},\hat{L}:D(\hat{L})\subset V\rightarrow V^{*} with the domain

D(L^)={uV:αutαV}.D(\hat{L})=\left\{u\in V:\frac{\partial^{\alpha}u}{\partial t^{\alpha}}\in V^{*}\right\}.

We note that the operator LL is linear, densely defined and m-accretive, see [70]. By Proposition 31.5 in [71], the operator LL is a maximal monotone operator. Next, We verify |u|m1uLloc2((0,T];Hs(N)).\left|u\right|^{m-1}u\in L_{loc}^{2}((0,T];H^{s}(\mathbb{R}^{N})). By [1], If ψ\psi and φ\varphi belong to the Schwartz class, definition (1.3) of the fractional Laplacian together with Plancherel’s theorem yields

N(Δ)sψφ𝑑x=N|ξ|2sψ^φ^𝑑x\displaystyle\int_{\mathbb{R}^{N}}(-\Delta)^{s}\psi\varphi dx=\int_{\mathbb{R}^{N}}\left|\xi\right|^{2s}\hat{\psi}\hat{\varphi}dx
=N|ξ|sψ^|ξ|sφ^𝑑x=N(Δ)s/2ψ(Δ)s/2φ𝑑x.\displaystyle=\int_{\mathbb{R}^{N}}\left|\xi\right|^{s}\hat{\psi}\left|\xi\right|^{s}\hat{\varphi}dx=\int_{\mathbb{R}^{N}}(-\Delta)^{s/2}\psi(-\Delta)^{s/2}\varphi dx.

Therefore, if we multiply the equation in (4.1) by a test function φ\varphi and integrate by parts as usual on t[0,T],t\in[0,T], we obtain

0TNuαφtα𝑑x𝑑t+0TN(Δ)s/2(um)(Δ)s/2φ𝑑x𝑑t\displaystyle\int_{0}^{T}\int_{\mathbb{R}^{N}}u\frac{\partial^{\alpha}\varphi}{\partial t^{\alpha}}dxdt+\int_{0}^{T}\int_{\mathbb{R}^{N}}(-\Delta)^{s/2}(u^{m})(-\Delta)^{s/2}\varphi dxdt
=0TNfφ𝑑x𝑑t\displaystyle=\int_{0}^{T}\int_{\mathbb{R}^{N}}f\varphi dxdt (4.15)

This identity will be the basis of our definition of a weak solution. The integrals in (1) make sense if uu and |u|m1u\left|u\right|^{m-1}u belong to suitable spaces. The correct space for |u|m1u\left|u\right|^{m-1}u is the fractional Sobolev space Hs(N)H^{s}(\mathbb{R}^{N}).
(2) We identity (1) hold for every ψC0([0,T]×N).\psi\in C_{0}^{\infty}([0,T]\times\mathbb{R}^{N}).

Let 0u0,nL1(N)L(N)0\leqslant u_{0,n}\in L^{1}(\mathbb{R}^{N})\cap L^{\infty}(\mathbb{R}^{N}) be a non-decreasing sequence of initial data u0,n1u0,nu_{0,n-1}\leqslant u_{0,n}, converging monotonically to u0L1(N,φdx)u_{0}\in L^{1}(\mathbb{R}^{N},\varphi dx),i.e., such that N(u0un,0)φ𝑑x0\int_{\mathbb{R}^{N}}(u_{0}-u_{n,0})\varphi dx\rightarrow 0 as nn\rightarrow\infty, where φ\varphi is as in Lemma 4.2 with decay at infinity |x|β,N2s1m<β<N+2sm.\left|x\right|^{-\beta},N-\frac{2s}{1-m}<\beta<N+\frac{2s}{m}. Consider the unique solution un(x,t)u_{n}(x,t) of Eq.(4.1) with initial data u0,n.u_{0,n}. The weighted estimates (4.2) show that the sequence is bounded in L1(N,φdx)L^{1}(\mathbb{R}^{N},\varphi dx) uniformly in t[0,T]t\in[0,T]. By the monotone convergence theorem in L1(N,φdx)L^{1}(\mathbb{R}^{N},\varphi dx), we know that the solution un(x,t)u_{n}(x,t) converge monotonically as nn\rightarrow\infty to a function u(x,t)L((0,T);L1(N,φdx))u(x,t)\in L^{\infty}((0,T);L^{1}(\mathbb{R}^{N},\varphi dx)). Indeed, the weighted estimates (4.2) show that when u0L1(N,φdx)u_{0}\in L^{1}(\mathbb{R}^{N},\varphi dx) then

(Nu(x,t)φ(x)𝑑x)1m=limn(Nun(x,t)φ(x)𝑑x)1m\displaystyle\left(\int_{\mathbb{R}^{N}}u(x,t)\varphi(x)dx\right)^{1-m}=\lim_{n\rightarrow\infty}\left(\int_{\mathbb{R}^{N}}u_{n}(x,t)\varphi(x)dx\right)^{1-m}
limn(Nun(x,0)φ(x)𝑑x)1m+εC1Tα(1m)(αΓ(α))1mR2sN(1m)\displaystyle\leqslant\lim_{n\rightarrow\infty}\left(\int_{\mathbb{R}^{N}}u_{n}(x,0)\varphi(x)dx\right)^{1-m}+\frac{\varepsilon C_{1}T^{\alpha(1-m)}}{(\alpha\Gamma(\alpha))^{1-m}R^{2s-N(1-m)}}
=(Nu0(x)φ(x)𝑑x)1m+εC1Tα(1m)(αΓ(α))1mR2sN(1m).\displaystyle=\left(\int_{\mathbb{R}^{N}}u_{0}(x)\varphi(x)dx\right)^{1-m}+\frac{\varepsilon C_{1}T^{\alpha(1-m)}}{(\alpha\Gamma(\alpha))^{1-m}R^{2s-N(1-m)}}. (4.16)

At this point we need to show that the function u(x,t)u(x,t) constructed as above is a very weak solution to Eq.(4.1) on [0,T]×N[0,T]\times\mathbb{R}^{N}. By definition 4.2, we make that each unu_{n} is a bounded strong solutions, since the initial data u0L1(N)L(N)u_{0}\in L^{1}(\mathbb{R}^{N})\cap L^{\infty}(\mathbb{R}^{N}), therefore for all ψC0([0,T]×N)\psi\in C_{0}^{\infty}([0,T]\times\mathbb{R}^{N}) we have

0TNunαψtα𝑑x𝑑t+0TN(Δ)s/2(unm)(Δ)s/2ψ𝑑x𝑑t\displaystyle\int_{0}^{T}\int_{\mathbb{R}^{N}}u_{n}\frac{\partial^{\alpha}\psi}{\partial t^{\alpha}}dxdt+\int_{0}^{T}\int_{\mathbb{R}^{N}}(-\Delta)^{s/2}(u_{n}^{m})(-\Delta)^{s/2}\psi dxdt
=0TNfψ𝑑x𝑑t\displaystyle=\int_{0}^{T}\int_{\mathbb{R}^{N}}f\psi dxdt (4.17)

Now, for any ψC0([0,T]×N)\psi\in C_{0}^{\infty}([0,T]\times\mathbb{R}^{N}) we easily have that

limn0TNunαψtα𝑑x𝑑t=0TNuαψtα𝑑x𝑑t\lim_{n\rightarrow\infty}\int_{0}^{T}\int_{\mathbb{R}^{N}}u_{n}\frac{\partial^{\alpha}\psi}{\partial t^{\alpha}}dxdt=\int_{0}^{T}\int_{\mathbb{R}^{N}}u\frac{\partial^{\alpha}\psi}{\partial t^{\alpha}}dxdt

since ψ\psi is compactly supported and we already know that un(x,t)u(x,t)u_{n}(x,t)\rightarrow u(x,t) in Lloc1.L_{loc}^{1}. On the other hand, for any ψC0([0,T]×N)\psi\in C_{0}^{\infty}([0,T]\times\mathbb{R}^{N}) we have that

limn0TN(Δ)s/2(unm)(Δ)s/2ψ𝑑x𝑑t=0TN(Δ)s/2(um)(Δ)s/2ψ𝑑x𝑑t\lim_{n\rightarrow\infty}\int_{0}^{T}\int_{\mathbb{R}^{N}}(-\Delta)^{s/2}(u_{n}^{m})(-\Delta)^{s/2}\psi dxdt=\int_{0}^{T}\int_{\mathbb{R}^{N}}(-\Delta)^{s/2}(u^{m})(-\Delta)^{s/2}\psi dxdt

since unuu_{n}\leq u and

0\displaystyle 0 0TN((Δ)s/2um(x,t)(Δ)s/2unm(x,t))(Δ)s/2ψ𝑑x𝑑t\displaystyle\leqslant\int_{0}^{T}\int_{\mathbb{R}^{N}}((-\Delta)^{s/2}u^{m}(x,t)-(-\Delta)^{s/2}u_{n}^{m}(x,t))(-\Delta)^{s/2}\psi dxdt
0TN(Δ)s/2(um(x,t)unm(x,t))(Δ)s/2ψ𝑑x𝑑t\displaystyle\leqslant\int_{0}^{T}\int_{\mathbb{R}^{N}}(-\Delta)^{s/2}(u^{m}(x,t)-u_{n}^{m}(x,t))(-\Delta)^{s/2}\psi dxdt
0TN(um(x,t)unm(x,t))(Δ)sψ𝑑x𝑑t\displaystyle\leqslant\int_{0}^{T}\int_{\mathbb{R}^{N}}(u^{m}(x,t)-u_{n}^{m}(x,t))(-\Delta)^{s}\psi dxdt
0TN|u(x,t)un(x,t)|mφm(x)|(Δ)sψ|φm(x)𝑑x𝑑t\displaystyle\leqslant\int_{0}^{T}\int_{\mathbb{R}^{N}}\left|u(x,t)-u_{n}(x,t)\right|^{m}\varphi^{m}(x)\frac{\left|(-\Delta)^{s}\psi\right|}{\varphi^{m}(x)}dxdt
0T(N|u(x,t)un(x,t)|φ(x)𝑑x)m(N||(Δ)sψ|φm(x)|11m𝑑x)1m𝑑t\displaystyle\leqslant\int_{0}^{T}\left(\int_{\mathbb{R}^{N}}\left|u(x,t)-u_{n}(x,t)\right|\varphi(x)dx\right)^{m}\left(\int_{\mathbb{R}^{N}}\left|\frac{\left|(-\Delta)^{s}\psi\right|}{\varphi^{m}(x)}\right|^{\frac{1}{1-m}}dx\right)^{1-m}dt
C0TN(u(x,t)un(x,t))φ𝑑x𝑑t0\displaystyle\leqslant C\int_{0}^{T}\int_{\mathbb{R}^{N}}\left(u(x,t)-u_{n}(x,t)\right)\varphi dxdt\rightarrow 0

where we have used Hölder inequality with conjugate exponents 1/m1/m and 1/(1m)1/(1-m), and we notice that

(N||(Δ)sψ|φm(x)|11m𝑑x)1mC\left(\int_{\mathbb{R}^{N}}\left|\frac{\left|(-\Delta)^{s}\psi\right|}{\varphi^{m}(x)}\right|^{\frac{1}{1-m}}dx\right)^{1-m}\leqslant C

since ψ\psi is compactly supported, therefore by Lemma 2.12 we know that |(Δ)sψ(x,t)|c3|x|(N+2s)\left|(-\Delta)^{s}\psi(x,t)\right|\leqslant c_{3}\left|x\right|^{-(N+2s)}, and quotient

||(Δ)sψ|φm(x)|11mc3|x|N+2smβ1m\left|\frac{\left|(-\Delta)^{s}\psi\right|}{\varphi^{m}(x)}\right|^{\frac{1}{1-m}}\leqslant\frac{c_{3}}{\left|x\right|^{\frac{N+2s-m\beta}{1-m}}}

is integrable when N+2smβ1m>N\frac{N+2s-m\beta}{1-m}>N that is when β<N+(2s/m).\beta<N+(2s/m). In the last step we already know that N(u(x,t)un(x,t))φ𝑑x0\int_{\mathbb{R}^{N}}\left(u(x,t)-u_{n}(x,t)\right)\varphi dx\rightarrow 0 when φ\varphi is as above,i.e. as in Lemma 4.2. Therefore we can let nn\rightarrow\infty in (1) and obtain (1).
(3) We verify u(,0)=u0L1(N)u(\cdot,0)=u_{0}\in L^{1}(\mathbb{R}^{N}) almost everywhere.

For the solution constructed above, the weighted estimates (4.2) show that when 0u0L1(N,φdx)0\leqslant u_{0}\in L^{1}(\mathbb{R}^{N},\varphi dx) imply

|Nu(x,t)φR𝑑xNu(x,τ)φR𝑑x|211mC1TααΓ(α)RN2s1m\left|\int_{\mathbb{R}^{N}}u(x,t)\varphi_{R}dx-\int_{\mathbb{R}^{N}}u(x,\tau)\varphi_{R}dx\right|\leqslant 2^{\frac{1}{1-m}}\frac{C_{1}T^{\alpha}}{\alpha\Gamma(\alpha)}R^{N-\frac{2s}{1-m}}

which gives the continuity in L1(N,φdx).L^{1}(\mathbb{R}^{N},\varphi dx). Therefore, the initial trace of this solution is given by u0L1(Nu_{0}\in L^{1}(\mathbb{R}^{N}.

In summary, by definition 4.1 and Theorem 3.1 in [63], we have proved existence of solutions corresponding to initial data u0u_{0} that can grow at infinity as |x|(2s/m)ε\left|x\right|^{(2s/m)-\varepsilon} for any ε>0\varepsilon>0 for problem (4.1). For the uniqueness of the solution, it can be proof through Theorem 3.2 in [63]. Therefore, equation (1.8)-(1.9) existence the unique weak solution.

Next, we will use Gagliardao-Nirenberg inequality, Young inequality and interpolation inequality, and so on. On the other hand, we use N|u|2𝑑x\int_{\mathbb{R}^{N}}\left|\triangledown u\right|^{2}dx and Nuk+1𝑑xNu𝑑x\int_{\mathbb{R}^{N}}u^{k+1}dx\int_{\mathbb{R}^{N}}udx to control nonlinear item Nuk+1𝑑x\int_{\mathbb{R}^{N}}u^{k+1}dx, and get (1.8)-(1.9) LrL^{r} estimate. In the proof process of the section, C(N,k,m)C(N,k,m) and Ci(N,k,m)(i=1,2)C_{i}(N,k,m)(i=1,2) represent the constant that depends on N,k,mN,k,m.

  1. 1.

    The LrL^{r} estimates.

For any xNx\in\mathbb{R}^{N}, multiply (1.8) by uk1,k>1u^{k-1},k>1 and integrating by parts over N\mathbb{R}^{N}, by the proof of Theorem 5.2 in [66], we obtain

Nuk1(Δ)sum𝑑x=CN,sP.V.NNuk1(x)um(x)um(y)|xy|N+2s𝑑x𝑑y\displaystyle\int_{\mathbb{R}^{N}}u^{k-1}(-\Delta)^{s}u^{m}dx=C_{N,s}P.V.\int_{\mathbb{R}^{N}}\int_{\mathbb{R}^{N}}u^{k-1}(x)\frac{u^{m}(x)-u^{m}(y)}{\left|x-y\right|^{N+2s}}dxdy
=12CN,sP.V.NN(uk1(x)uk1(y))um(x)um(y)|xy|N+2s𝑑x𝑑y,\displaystyle=\frac{1}{2}C_{N,s}P.V.\int_{\mathbb{R}^{N}}\int_{\mathbb{R}^{N}}(u^{k-1}(x)-u^{k-1}(y))\frac{u^{m}(x)-u^{m}(y)}{\left|x-y\right|^{N+2s}}dxdy,

from Lemma 4.6 and equation (4.12), let a=um(x),b=um(y),α=k1ma=u^{m}(x),b=u^{m}(y),\alpha=\frac{k-1}{m}, then we have

(um(x)um(y))(uk1(x)uk1(y))c3|uk+m12(x)uk+m12(y)|2.\left(u^{m}(x)-u^{m}(y)\right)\left(u^{k-1}(x)-u^{k-1}(y)\right)\geq c_{3}\left|u^{\frac{k+m-1}{2}}(x)-u^{\frac{k+m-1}{2}}(y)\right|^{2}.

By Sobolev inequality (Lemma 4.7), we reach that

Nuk1(Δ)sum𝑑x\displaystyle\int_{\mathbb{R}^{N}}u^{k-1}(-\Delta)^{s}u^{m}dx c32|uk+m12(x)uk+m12(y)|2\displaystyle\geq\frac{c_{3}}{2}\left|u^{\frac{k+m-1}{2}}(x)-u^{\frac{k+m-1}{2}}(y)\right|^{2}
c3S2(N|uk+m12(x,t)|ps𝑑x)2ps,\displaystyle\geq\frac{c_{3}S}{2}\left(\int_{\mathbb{R}^{N}}\left|u^{\frac{k+m-1}{2}}(x,t)\right|^{p_{s}^{*}}dx\right)^{\frac{2}{p_{s}^{*}}},

where ps=2NN2sp_{s}^{*}=\frac{2N}{N-2s}. Then, we obtain

1k(0CDtαNukdx)+c3S2(N|uk+m12(x,t)|psdx)2ps\displaystyle\frac{1}{k}(_{0}^{C}\textrm{D}_{t}^{\alpha}\int_{\mathbb{R}^{N}}u^{k}dx)+\frac{c_{3}S}{2}\left(\int_{\mathbb{R}^{N}}\left|u^{\frac{k+m-1}{2}}(x,t)\right|^{p_{s}^{*}}dx\right)^{\frac{2}{p_{s}^{*}}} (4.18)
Nuk+1𝑑x(1Nu𝑑x)Nuk𝑑x\displaystyle\leq\int_{\mathbb{R}^{N}}u^{k+1}dx(1-\int_{\mathbb{R}^{N}}udx)-\int_{\mathbb{R}^{N}}u^{k}dx

The following estimate kNuk+1𝑑xk\int_{\mathbb{R}^{N}}u^{k+1}dx. when m3m\leq 3 and

k>max{14(3m)(N2)(m1),(1m2)N1,N(2m)2},k>max\left\{\frac{1}{4}(3-m)(N-2)-(m-1),(1-\frac{m}{2})N-1,N(2-m)-2\right\}, (4.19)

from Lemma 4.9, we obtain

kNuk+1𝑑x=kuk+m12L2(k+1)k+m+1(N)2(k+1)k+m1\displaystyle k\int_{\mathbb{R}^{N}}u^{k+1}dx=k\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2(k+1)}{k+m+1}}(\mathbb{R}^{N})}^{\frac{2(k+1)}{k+m-1}}
kuk+m12L2(k+1)k+m+1(N)2(k+1)k+m1λuk+m12L2(N)2(k+1)(1λ)k+m1.\displaystyle\leq k\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2(k+1)}{k+m+1}}(\mathbb{R}^{N})}^{\frac{2(k+1)}{k+m-1}\lambda^{*}}\left\|\bigtriangledown u^{\frac{k+m-1}{2}}\right\|_{L^{2}(\mathbb{R}^{N})}^{\frac{2(k+1)(1-\lambda^{*})}{k+m-1}}.

Knowing from (4.19)

λ=1+(k+m1)N2(k+1)N21+(k+m1)Nk+2N2ϵ{max{0,2mk+1},1},\lambda^{*}=\frac{1+\frac{(k+m-1)N}{2(k+1)}-\frac{N}{2}}{1+\frac{(k+m-1)N}{k+2}-\frac{N}{2}}\epsilon\left\{max\left\{0,\frac{2-m}{k+1}\right\},1\right\},

using Young’s inequality, there are

kNuk+1𝑑x\displaystyle k\int_{\mathbb{R}^{N}}u^{k+1}dx kuk+m12Lk+2k+m+1(N)2(k+1)k+m1λuk+m12L2(RN)2(k+1)(1λ)k+m1\displaystyle\leq k\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{k+2}{k+m+1}}(\mathbb{R}^{N})}^{\frac{2(k+1)}{k+m-1}\lambda^{*}}\left\|\bigtriangledown u^{\frac{k+m-1}{2}}\right\|_{L^{2}(R^{N})}^{\frac{2(k+1)(1-\lambda^{*})}{k+m-1}} (4.20)
2k(k1)(m+k1)2uk+m12L2(RN)2\displaystyle\leq\frac{2k(k-1)}{(m+k-1)^{2}}\left\|\bigtriangledown u^{\frac{k+m-1}{2}}\right\|_{L^{2}(R^{N})}^{2}
+C1(N,k,m)uk+m12Lk+2k+m1(N)Q2\displaystyle\qquad+C_{1}(N,k,m)\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{k+2}{k+m-1}}(\mathbb{R}^{N})}^{Q_{2}}

where is Q2=2(k+1)λm2+(k+1)λQ_{2}=\frac{2(k+1)\lambda^{*}}{m-2+(k+1)\lambda^{*}}. Next estimate uk+m12Lk+2k+m1(N)Q2\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{k+2}{k+m-1}}(\mathbb{R}^{N})}^{Q_{2}}. We will use the interpolation inequality to get

uk+m12Lk+2k+m1(N)Q2\displaystyle\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{k+2}{k+m-1}}(\mathbb{R}^{N})}^{Q_{2}} uk+m12L2(k+1)k+m1(N)Q2λuk+m12L2k+m1(N)Q2(1λ)\displaystyle\leq\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2(k+1)}{k+m-1}}(\mathbb{R}^{N})}^{Q_{2}\lambda}\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2}{k+m-1}}(\mathbb{R}^{N})}^{Q_{2}(1-\lambda)} (4.21)
(uk+m12L2(k+1)k+m1(N)2(k+1)k+m1uk+m12L2k+m1(N)2k+m1)Q2λ(k+m1)2(k+1)\displaystyle\leq(\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2(k+1)}{k+m-1}}(\mathbb{R}^{N})}^{\frac{2(k+1)}{k+m-1}}\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2}{k+m-1}}(\mathbb{R}^{N})}^{\frac{2}{k+m-1}})^{\frac{Q_{2}\lambda(k+m-1)}{2(k+1)}}
uk+m12L2k+m1(N)Q2(1λλk+1)\displaystyle\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2}{k+m-1}}(\mathbb{R}^{N})}^{Q_{2}(1-\lambda-\frac{\lambda}{k+1})}

in λ=k+1k+2\lambda=\frac{k+1}{k+2}, and

Q2(1λλk+1)=0.Q_{2}(1-\lambda-\frac{\lambda}{k+1})=0.

Then

C1(N,k,m)uk+m12Lk+2k+m1(N)Q2\displaystyle C_{1}(N,k,m)\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{k+2}{k+m-1}}(\mathbb{R}^{N})}^{Q_{2}} (4.22)
C1(N,k,m)(uk+m12L2(k+1)k+m1(N)2(k+1)k+m1uk+m12L2k+m1(RN)2k+m1)Q2λ(k+m1)2(k+1).\displaystyle\leq C_{1}(N,k,m)(\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2(k+1)}{k+m-1}}(\mathbb{R}^{N})}^{\frac{2(k+1)}{k+m-1}}\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2}{k+m-1}}(R^{N})}^{\frac{2}{k+m-1}})^{\frac{Q_{2}\lambda(k+m-1)}{2(k+1)}}.

Noticing that when m>22Nm>2-\frac{2}{N}, it is easy to verify

Q2λ(k+m1)2(k+1)=(k+1)(m2)+λ(k+1)(3m)(k+1)(k+m1)λ<1,\frac{Q_{2}\lambda(k+m-1)}{2(k+1)}=\frac{(k+1)(m-2)+\lambda^{*}(k+1)(3-m)}{(k+1)(k+m-1)\lambda^{*}}<1,

using Young’s inequality, then

C1(N,k,m)(uk+m12L2(k+1)k+m1(N)2(k+1)k+m1uk+m12L2k+m1(N)2k+m1)Q2λ(k+m1)2(k+1)\displaystyle C_{1}(N,k,m)(\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2(k+1)}{k+m-1}}(\mathbb{R}^{N})}^{\frac{2(k+1)}{k+m-1}}\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2}{k+m-1}}(\mathbb{R}^{N})}^{\frac{2}{k+m-1}})^{\frac{Q_{2}\lambda(k+m-1)}{2(k+1)}} (4.23)
kuk+m12L2(k+1)k+m1(N)2(k+1)k+m1uk+m12L2k+m1(N)2k+m1+C2(N,k,m).\displaystyle\leq k\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2(k+1)}{k+m-1}}(\mathbb{R}^{N})}^{\frac{2(k+1)}{k+m-1}}\left\|u^{\frac{k+m-1}{2}}\right\|_{L^{\frac{2}{k+m-1}}(\mathbb{R}^{N})}^{\frac{2}{k+m-1}}+C_{2}(N,k,m).

Substitute (4.20)-(4.23) into (4.18) to get

Dtα0CNuk𝑑x+c3Sk2(N|uk+m12(x,t)|pα𝑑x)2pα+kNuk𝑑x{}_{0}^{C}\textrm{D}_{t}^{\alpha}\int_{\mathbb{R}^{N}}u^{k}dx+\frac{c_{3}Sk}{2}\left(\int_{\mathbb{R}^{N}}\left|u^{\frac{k+m-1}{2}}(x,t)\right|^{p_{\alpha}^{*}}dx\right)^{\frac{2}{p_{\alpha}^{*}}}+k\int_{\mathbb{R}^{N}}u^{k}dx
2k(k1)(m+k1)2uk+m12L2(N)2+C2(N,k,m).\displaystyle\leq\frac{2k(k-1)}{(m+k-1)^{2}}\left\|\triangledown u^{\frac{k+m-1}{2}}\right\|_{L^{2}(\mathbb{R}^{N})}^{2}+C_{2}(N,k,m). (4.24)

Let kk\rightarrow\infty, then 2k(k1)(m+k1)22\frac{2k(k-1)}{(m+k-1)^{2}}\rightarrow 2, so 2k(k1)(m+k1)22.\frac{2k(k-1)}{(m+k-1)^{2}}\leq 2. And from Lemma 4.8 and N3N\geq 3, we obtain

1SN1u11k1(2k/N+m1)(ukk)1+m1+2/Nk1u(k+m1)/222.\frac{1}{S_{N}^{-1}\left\|u\right\|_{1}^{\frac{1}{k-1}(2k/N+m-1)}}(\left\|u\right\|_{k}^{k})^{1+\frac{m-1+2/N}{k-1}}\leq\left\|\triangledown u^{(k+m-1)/2}\right\|_{2}^{2}.

Let k=N(1m)2sk=\frac{N(1-m)}{2s}, then k+m12ps=k,2ps=k+m1k\frac{k+m-1}{2}p_{s}^{*}=k,\frac{2}{p_{s}^{*}}=\frac{k+m-1}{k}. Therefore, we can get

Dtα0CNuk𝑑x+c3Sk2(Nuk𝑑x)k+m1k+kNuk𝑑x{}_{0}^{C}\textrm{D}_{t}^{\alpha}\int_{\mathbb{R}^{N}}u^{k}dx+\frac{c_{3}Sk}{2}\left(\int_{\mathbb{R}^{N}}u^{k}dx\right)^{\frac{k+m-1}{k}}+k\int_{\mathbb{R}^{N}}u^{k}dx
2SN1u11k1(2k/N+m1)(ukk)1+m1+2/Nk1+C2(N,k,m).\displaystyle\leq\frac{2}{S_{N}^{-1}\left\|u\right\|_{1}^{\frac{1}{k-1}(2k/N+m-1)}}(\left\|u\right\|_{k}^{k})^{1+\frac{m-1+2/N}{k-1}}+C_{2}(N,k,m).

Let a=1+m1+2/dk1,f(t)=2SN1u11k1(2k/N+m1),β=k+m1k,y(t)=Nuk𝑑xa=1+\frac{m-1+2/d}{k-1},f(t)=\frac{2}{S_{N}^{-1}\left\|u\right\|_{1}^{\frac{1}{k-1}(2k/N+m-1)}},\beta=\frac{k+m-1}{k},y(t)=\int_{\mathbb{R}^{N}}u^{k}dx. Then, when 0<m<10<m<1, the above-mentioned inequality can be written as

0CDtαy(t)+c3Sk2yβ(t)+ky(t)f(t)ya(t)+C2(N,k,m)_{0}^{C}\textrm{D}_{t}^{\alpha}y(t)+\frac{c_{3}Sk}{2}y^{\beta}(t)+ky(t)\leq f(t)y^{a}(t)+C_{2}(N,k,m) (4.25)

By Lemma 4.5 and 0<β<a,t[0,T]0<\beta<a,t\in[0,T], fractional differential inequality (4.25) has following solution

y(t)\displaystyle y(t) y(0)+[[λky1β(0)+(C2(N,k,m)c3Sk2)(1β)]TααΓ(α)]11β\displaystyle\leq y(0)+\left[\frac{[\lambda_{k}y^{1-\beta}(0)+(C_{2}(N,k,m)-\frac{c_{3}Sk}{2})(1-\beta)]T^{\alpha}}{\alpha\Gamma(\alpha)}\right]^{\frac{1}{1-\beta}}
+(1a)11βε11aTα1βα11βΓ11β(α)f11a(t).\displaystyle\qquad+(1-a)^{\frac{1}{1-\beta}}\varepsilon^{\frac{1}{1-a}}\frac{T^{\frac{\alpha}{1-\beta}}}{\alpha^{\frac{1}{1-\beta}}\Gamma^{\frac{1}{1-\beta}}(\alpha)}f^{\frac{1}{1-a}}(t).

Therefore, we have

y(t)=Nuk𝑑x\displaystyle y(t)=\int_{\mathbb{R}^{N}}u^{k}dx y(0)+[[λky1β(0)+(C2(N,k,m)c3Sk2)(1β)]TααΓ(α)]11β\displaystyle\leq y(0)+\left[\frac{[\lambda_{k}y^{1-\beta}(0)+(C_{2}(N,k,m)-\frac{c_{3}Sk}{2})(1-\beta)]T^{\alpha}}{\alpha\Gamma(\alpha)}\right]^{\frac{1}{1-\beta}}
+(1a)11βε11aTα1βα11βΓ11β(α)f11a(0).\displaystyle\qquad+(1-a)^{\frac{1}{1-\beta}}\varepsilon^{\frac{1}{1-a}}\frac{T^{\frac{\alpha}{1-\beta}}}{\alpha^{\frac{1}{1-\beta}}\Gamma^{\frac{1}{1-\beta}}(\alpha)}f^{\frac{1}{1-a}}(0). (4.26)

where λk=aβε1βaβk(1β),y(0)=u0Lk(N)k\lambda_{k}=-\frac{a-\beta}{\varepsilon^{\frac{1-\beta}{a-\beta}}}-k(1-\beta),y(0)=\left\|u_{0}\right\|_{L^{k}(\mathbb{R}^{N})}^{k} and f(0)=2SN1u011k1(2k/N+m1)f(0)=\frac{2}{S_{N}^{-1}\left\|u_{0}\right\|_{1}^{\frac{1}{k-1}(2k/N+m-1)}}.

  1. 1.

    The LL^{\infty} estimates.

On account of the above arguments, our last task is to give the uniform boundedness of solution for any t>0t>0. Denote qk=2k+2q_{k}=2^{k}+2, by taking k=qkk=q_{k} in (4.18), we have

1qk(0CDtαNuqkdx)+c3S2(N|uqk+m12(x,t)|psdx)2ps\displaystyle\frac{1}{q_{k}}(_{0}^{C}\textrm{D}_{t}^{\alpha}\int_{\mathbb{R}^{N}}u^{q_{k}}dx)+\frac{c_{3}S}{2}\left(\int_{\mathbb{R}^{N}}\left|u^{\frac{q_{k}+m-1}{2}}(x,t)\right|^{p_{s}^{*}}dx\right)^{\frac{2}{p_{s}^{*}}} (4.27)
Nuqk+1𝑑x(1Nu𝑑x)Nuqk𝑑x\displaystyle\leq\int_{\mathbb{R}^{N}}u^{q_{k}+1}dx(1-\int_{\mathbb{R}^{N}}udx)-\int_{\mathbb{R}^{N}}u^{q_{k}}dx

armed with Lemma 4.10, letting

v=m+qk12,q=2(qk+1)m+qk1,r=2qk1m+qk1,c0=c1=12qk,v=\frac{m+q_{k}-1}{2},q=\frac{2(q_{k}+1)}{m+q_{k}-1},r=\frac{2q_{k-1}}{m+q_{k}-1},c_{0}=c_{1}=\frac{1}{2q_{k}},

one has that for N3N\geq 3,

uLqk+1(N)qk+1\displaystyle\left\|u\right\|_{L^{q_{k}+1}(\mathbb{R}^{N})}^{q_{k}+1} C(N)c01δ11(Nuqk1𝑑x)γ1+12qkum+qk12L2(N)2\displaystyle\leq C(N)c_{0}^{\frac{1}{\delta_{1}-1}}(\int_{\mathbb{R}^{N}}u^{q_{k-1}}dx)^{\gamma_{1}}+\frac{1}{2q^{k}}\left\|\bigtriangledown u^{\frac{m+q_{k}-1}{2}}\right\|_{L^{2}(\mathbb{R}^{N})}^{2} (4.28)
+12qkuLm+qk1(N)m+qk1,\displaystyle\qquad+\frac{1}{2q_{k}}\left\|u\right\|_{L^{m+q_{k}-1}(\mathbb{R}^{N})}^{m+q_{k}-1},

where

γ1=1+qk+qk1+1qk1+p(m2)p22,\gamma_{1}=1+\frac{q_{k}+q_{k-1}+1}{q_{k-1}+\frac{p(m-2)}{p-2}}\leq 2,
δ1=(m+qk1)2qk1pqkqk1+1=O(1).\delta_{1}=\frac{(m+q_{k}-1)-2\frac{q_{k-1}}{p^{*}}}{q_{k}-q_{k-1}+1}=O(1).

Substituting (4.28) into (4.27) and with notice that 4qk(qk1)(m+qk1)22\frac{4q_{k}(q_{k}-1)}{(m+q_{k}-1)^{2}}\geq 2. It follows

Dtα0CNuqk𝑑x+c3Sqk2(N|uqk+m12(x,t)|ps𝑑x)2ps+qkNuqk𝑑x\displaystyle{}_{0}^{C}\textrm{D}_{t}^{\alpha}\int_{\mathbb{R}^{N}}u^{q_{k}}dx+\frac{c_{3}Sq_{k}}{2}\left(\int_{\mathbb{R}^{N}}\left|u^{\frac{q_{k}+m-1}{2}}(x,t)\right|^{p_{s}^{*}}dx\right)^{\frac{2}{p_{s}^{*}}}+q_{k}\int_{\mathbb{R}^{N}}u^{q_{k}}dx (4.29)
C(N)qkδ1δ11(Nuqk1𝑑x)γ1+12um+qk1222\displaystyle\leq C(N)q_{k}^{\frac{\delta_{1}}{\delta_{1}-1}}(\int_{\mathbb{R}^{N}}u^{q_{k-1}}dx)^{\gamma_{1}}+\frac{1}{2}\left\|\triangledown u^{\frac{m+q_{k}-1}{2}}\right\|_{2}^{2}
+12uLm+qk1(N)m+qk1Nu𝑑xNuqk+1𝑑x\displaystyle\qquad+\frac{1}{2}\left\|u\right\|_{L^{m+q_{k}-1}(\mathbb{R}^{N})}^{m+q_{k}-1}-\int_{\mathbb{R}^{N}}udx\int_{\mathbb{R}^{N}}u^{q_{k}+1}dx

Applying Lemma 4.10 with

v=um+qk12,q=2,r=2qk1m+qk1,c0=c1=12v=u^{\frac{m+q_{k}-1}{2}},q=2,r=\frac{2q_{k-1}}{m+q_{k}-1},c_{0}=c_{1}=\frac{1}{2}

noticing qk1=(qk+1)+12q_{k-1}=\frac{(q_{k}+1)+1}{2}, and using Young’s inequality, we obtain

12uLm+qk1(N)m+qk1=12Num+qk1𝑑x\displaystyle\frac{1}{2}\left\|u\right\|_{L^{m+q_{k}-1}(\mathbb{R}^{N})}^{m+q_{k}-1}=\frac{1}{2}\int_{\mathbb{R}^{N}}u^{m+q_{k}-1}dx (4.30)
c2(N)(Nuqk1𝑑x)γ2+12um+qk12L2(N)2\displaystyle\leq c_{2}(N)(\int_{\mathbb{R}^{N}}u^{q_{k-1}}dx)^{\gamma_{2}}+\frac{1}{2}\left\|\bigtriangledown u^{\frac{m+q_{k}-1}{2}}\right\|_{L^{2}(\mathbb{R}^{N})}^{2}
Nu𝑑xNuqk+1𝑑x+c3(N)+12um+qk12L2(N)2,\displaystyle\leq\int_{\mathbb{R}^{N}}udx\int_{\mathbb{R}^{N}}u^{q_{k}+1}dx+c_{3}(N)+\frac{1}{2}\left\|\bigtriangledown u^{\frac{m+q_{k}-1}{2}}\right\|_{L^{2}(\mathbb{R}^{N})}^{2},

where

γ2=1+m+qkqk11qk1<2.\gamma_{2}=1+\frac{m+q_{k}-q_{k-1}-1}{q_{k-1}}<2.

By summing up (4.29) and (4.30), with the fact that γ12\gamma_{1}\leq 2 and γ2<2\gamma_{2}<2, we have

Dtα0CNuqk𝑑x+c3Sqk2(N|uqk+m12(x,t)|pα𝑑x)2pα+qkNuqk𝑑x\displaystyle{}_{0}^{C}\textrm{D}_{t}^{\alpha}\int_{\mathbb{R}^{N}}u^{q_{k}}dx+\frac{c_{3}Sq_{k}}{2}\left(\int_{\mathbb{R}^{N}}\left|u^{\frac{q_{k}+m-1}{2}}(x,t)\right|^{p_{\alpha}^{*}}dx\right)^{\frac{2}{p_{\alpha}^{*}}}+q_{k}\int_{\mathbb{R}^{N}}u^{q_{k}}dx
C(N)qkδ1δ11(Nuqk1𝑑x)γ1+um+qk12L2(N)2+C3(N)\displaystyle\leq C(N)q_{k}^{\frac{\delta_{1}}{\delta_{1}-1}}(\int_{\mathbb{R}^{N}}u^{q_{k-1}}dx)^{\gamma_{1}}+\left\|\triangledown u^{\frac{m+q_{k}-1}{2}}\right\|_{L^{2}(\mathbb{R}^{N})}^{2}+C_{3}(N)
max{C(N),C3(N)}qkδ1δ11[(Nuqk1𝑑x)γ1+1+um+qk12L2(N)2]\displaystyle\leq max\left\{C(N),C_{3}(N)\right\}q_{k}^{\frac{\delta_{1}}{\delta_{1}-1}}\left[(\int_{\mathbb{R}^{N}}u^{q_{k-1}}dx)^{\gamma 1}+1+\left\|\triangledown u^{\frac{m+q_{k}-1}{2}}\right\|_{L^{2}(\mathbb{R}^{N})}^{2}\right]
2max{C(N),C3(N)}qkδ1δ11max{(Nuqk1𝑑x)2,1,um+qk12L2(N)2}.\displaystyle\leq 2max\left\{C(N),C_{3}(N)\right\}q_{k}^{\frac{\delta_{1}}{\delta_{1}-1}}max\left\{(\int_{\mathbb{R}^{N}}u^{q_{k-1}}dx)^{2},1,\left\|\triangledown u^{\frac{m+q_{k}-1}{2}}\right\|_{L^{2}(\mathbb{R}^{N})}^{2}\right\}.

Let qk=N(1m)2αq_{k}=\frac{N(1-m)}{2\alpha}, then qk+m12ps=qk,2ps=qk+m1qk\frac{q_{k}+m-1}{2}p_{s}^{*}=q_{k},\frac{2}{p_{s}^{*}}=\frac{q_{k}+m-1}{q_{k}}. Therefore, we can get

Dtα0CNuqk𝑑x+c3Sqk2(Nuqk𝑑x)qk+m1qk+qkNuqk𝑑x{}_{0}^{C}\textrm{D}_{t}^{\alpha}\int_{\mathbb{R}^{N}}u^{q_{k}}dx+\frac{c_{3}Sq_{k}}{2}\left(\int_{\mathbb{R}^{N}}u^{q_{k}}dx\right)^{\frac{q_{k}+m-1}{q_{k}}}+q_{k}\int_{\mathbb{R}^{N}}u^{q_{k}}dx
2max{C(N),C3(N)}qkδ1δ11max{(Nuqk1𝑑x)2,1,um+qk12L2(N)2}.\displaystyle\leq 2max\left\{C(N),C_{3}(N)\right\}q_{k}^{\frac{\delta_{1}}{\delta_{1}-1}}max\left\{(\int_{\mathbb{R}^{N}}u^{q_{k-1}}dx)^{2},1,\left\|\triangledown u^{\frac{m+q_{k}-1}{2}}\right\|_{L^{2}(\mathbb{R}^{N})}^{2}\right\}.

Let

K0=max{1,u0L1(N),u0L(N),u0m+qk12L2(N)2},K_{0}=max\left\{1,\left\|u_{0}\right\|_{L^{1}(\mathbb{R}^{N})},\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})},\left\|\triangledown u_{0}^{\frac{m+q_{k}-1}{2}}\right\|_{L^{2}(\mathbb{R}^{N})}^{2}\right\},

we have the following inequality for initial data

Nu0qk𝑑x(max{u0L1(N),u0L(N),u0m+qk12L2(N)2})qkK0qk.\int_{\mathbb{R}^{N}}u_{0}^{q_{k}}dx\leq\left(max\left\{\left\|u_{0}\right\|_{L^{1}(\mathbb{R}^{N})},\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})},\left\|\triangledown u_{0}^{\frac{m+q_{k}-1}{2}}\right\|_{L^{2}(\mathbb{R}^{N})}^{2}\right\}\right)^{q_{k}}\leq K_{0}^{q_{k}}.

Let d0=δ1δ11d_{0}=\frac{\delta_{1}}{\delta_{1}-1}, it is easy to that qkd0=(2k+2)d0(2k+2k+1)d0q_{k}^{d_{0}}=(2^{k}+2)^{d_{0}}\leq(2^{k}+2^{k+1})^{d_{0}}. By taking a¯=max{C(N),C3(N)}3d0\bar{a}=max\left\{C(N),C_{3}(N)\right\}3^{d_{0}} in the Lemma 4.4, we obtain

uqk𝑑x(2a¯)2k12d0(2k+1k2)max{supt0(Nuq𝑑x)2k,k0qk}TααΓ(α).\int u^{q_{k}}dx\leq(2\bar{a})^{2^{k}-1}2^{d_{0}(2^{k+1}-k-2)}max\left\{\underset{t\geq 0}{sup}(\int_{\mathbb{R}^{N}}u^{q}dx)^{2^{k}},k_{0}^{q_{k}}\right\}\frac{T^{\alpha}}{\alpha\Gamma(\alpha)}. (4.31)

Since qk=2k+2q_{k}=2^{k}+2 and taking the power 1qk\frac{1}{q_{k}} to both sides of (4.31), then the boundedness of the solution u(x,t)u(x,t) is obtained by passing to the limit kk\rightarrow\infty

u(x,t)L(N)2a¯22d0max{supt0Nuq0𝑑x,K0}TααΓ(α).\left\|u(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N})}\leq 2\bar{a}2^{2d_{0}}max\left\{sup_{t\geq 0}\int_{\mathbb{R}^{N}}u^{q_{0}}dx,K_{0}\right\}\frac{T^{\alpha}}{\alpha\Gamma(\alpha)}. (4.32)

On the other hand, by (1) with q0>2,t[0,T]q_{0}>2,t\in[0,T], we know

Nuq0𝑑xNu3𝑑x\displaystyle\int_{\mathbb{R}^{N}}u^{q_{0}}dx\leq\int_{\mathbb{R}^{N}}u^{3}dx [[λky1β(0)+(C2(N,k,m)3c3S2)(1β)]TααΓ(α)]11β\displaystyle\leq\left[\frac{[\lambda_{k}y^{1-\beta}(0)+(C_{2}(N,k,m)-\frac{3c_{3}S}{2})(1-\beta)]T^{\alpha}}{\alpha\Gamma(\alpha)}\right]^{\frac{1}{1-\beta}}
+y(0)+(1a)11βε11aTα1βα11βΓ11β(α)f11a(0).\displaystyle\qquad+y(0)+(1-a)^{\frac{1}{1-\beta}}\varepsilon^{\frac{1}{1-a}}\frac{T^{\frac{\alpha}{1-\beta}}}{\alpha^{\frac{1}{1-\beta}}\Gamma^{\frac{1}{1-\beta}}(\alpha)}f^{\frac{1}{1-a}}(0).

where λk=aβε1βaβ3(1β),y(0)=u0L3(N)3\lambda_{k}=-\frac{a-\beta}{\varepsilon^{\frac{1-\beta}{a-\beta}}}-3(1-\beta),y(0)=\left\|u_{0}\right\|_{L^{3}(\mathbb{R}^{N})}^{3} and f(0)=2SN1u0112(6/N+m1)f(0)=\frac{2}{S_{N}^{-1}\left\|u_{0}\right\|_{1}^{\frac{1}{2}(6/N+m-1)}}. Therefore we finally have

u(x,t)L(N)C(N,u0L1(N),u0L(N),u0m+22L2(N)2,Tα)=M.\left\|u(x,t)\right\|_{L^{\infty}(\mathbb{R}^{N})}\leq C(N,\left\|u_{0}\right\|_{L^{1}(\mathbb{R}^{N})},\left\|u_{0}\right\|_{L^{\infty}(\mathbb{R}^{N})},\left\|\triangledown u_{0}^{\frac{m+2}{2}}\right\|_{L^{2}(\mathbb{R}^{N})}^{2},T^{\alpha})=M.

Remark 4.5.

The solution for problem (1.8)-(1.9) constructed above only need to be integrable with respect to the weight φ\varphi, which has a tail of order less than N+2s/m.N+2s/m. And the method to proof main reference [63]. On the other hand, The method to prove the LrL^{r} estimate to the LL^{\infty} estimate of the equation is also mentioned in [64]. This section mainly uses fractional differential inequalities, Gagliardao-Nirenberg inequalities and so on. Therefore, the LrL^{r} estimate is obtained, and if k=qkk=q_{k} is estimated on LL^{\infty}, the global boundedness of the solution for the nonlinear TSFNRDE is proved in N,N3\mathbb{R}^{N},N\geq 3.

5 Acknowledgements

This work is supported by the State Key Program of National Natural Science of China under Grant No.91324201. This work is also supported by the Fundamental Research Funds for the Central Universities of China under Grant 2018IB017, Equipment Pre-Research Ministry of Education Joint Fund Grant 6141A02033703 and the Natural Science Foundation of Hubei Province of China under Grant 2014CFB865.

6 Appendix A. Definitions, Related Lemma, and complements

A.1 Related Lemma to proof existence weak solution for (1.1)-(1.2)

Lemma 6.1.

[40] Let 0<α<10<\alpha<1 and λ>0\lambda>0, then we have

ddtEα,1(λtα)=λtα1Eα,α(λtα),t>0.\frac{d}{dt}E_{\alpha,1}(-\lambda t^{\alpha})=-\lambda t^{\alpha-1}E_{\alpha,\alpha}(-\lambda t^{\alpha}),\quad t>0.

Lemma 6.2.

[40] Let 0<α<10<\alpha<1 and λ>0\lambda>0, then we have

tαEα,1(λtα)=λEα,α(λtα),t>0.\partial_{t}^{\alpha}E_{\alpha,1}(-\lambda t^{\alpha})=-\lambda E_{\alpha,\alpha}(-\lambda t^{\alpha}),\quad t>0.

Lemma 6.3.

[72] For 0<α<1,λ>00<\alpha<1,\lambda>0 and Let AC[0,T]AC[0,T] be the space of functions ff which are absolutely continuous on [0,T][0,T], if q(t)AC[0,T]q(t)\in AC[0,T] then we have

tα0t(tτ)α1Eα,α(λ(tτ)α)𝑑τ\displaystyle\partial_{t}^{\alpha}\int_{0}^{t}(t-\tau)^{\alpha-1}E_{\alpha,\alpha}(-\lambda(t-\tau)^{\alpha})d\tau
=q(t)λ0tq(τ)(tτ)α1Eα,α(λ(tτ)α)𝑑τ,t(0,T].\displaystyle=q(t)-\lambda\int_{0}^{t}q(\tau)(t-\tau)^{\alpha-1}E_{\alpha,\alpha}(-\lambda(t-\tau)^{\alpha})d\tau,\quad t\in(0,T].

Lemma 6.4.

[40] Suppose p(t)L(0,T),0<α<1,λ0p(t)\in L^{\infty}(0,T),0<\alpha<1,\lambda\geq 0,denote

g(t)=0tp(τ)(tτ)α1Eα,α(λ(tτ)α)𝑑τ,t(0,T],g(t)=\int_{0}^{t}p(\tau)(t-\tau)^{\alpha-1}E_{\alpha,\alpha}(-\lambda(t-\tau)^{\alpha})d\tau,\quad t\in(0,T],

and defines g(0)=0g(0)=0, then g(t)C[0,T]g(t)\in C[0,T]

A.2 The proof of existence weak solution for (1.1)-(1.2)

Remark 6.1.

By consulting the relevant reference, we can get the Lemma 2.14 by [5], but “For brevity, we leave the detail to the reader.”[p255,Theorem 1,[5]]. Therefore, we will give the proof process of Lemma 2.14 as shown below.

Proof of Lemma 2.14 1.

we will show that (2.8) certainly gives a weak solution to (1.1)-(1.2). In the following proof, we denote CC as a generic positive constant and make q(t)=p(t)=1q(t)=p(t)=1 for Lemma 6.3 and Lemma 6.4. Denote

gj(t)=0t(tτ)α1Eα,α(λjs(tτ)α)𝑑τ,g_{j}(t)=\int_{0}^{t}(t-\tau)^{\alpha-1}E_{\alpha,\alpha}(-\lambda_{j}^{s}(t-\tau)^{\alpha})d\tau, (6.1)

then we know

|gj(t)|1Γ(α+1)tα1Γ(α+1)Tα,t[0,T],\displaystyle\left|g_{j}(t)\right|\leq\frac{1}{\Gamma(\alpha+1)t^{\alpha}}\leq\frac{1}{\Gamma(\alpha+1)T^{\alpha}},\quad t\in[0,T], (6.2)
|gj(t)|(1Eα,1(λnTα))λjs1λjs,t[0,T].\displaystyle\left|g_{j}(t)\right|\leq\frac{(1-E_{\alpha,1}(-\lambda_{n}T^{\alpha}))}{\lambda_{j}^{s}}\leq\frac{1}{\lambda_{j}^{s}},\quad t\in[0,T]. (6.3)

The proof is divided into several steps.

(1) We first verify uC([0,T];L2(Ω))u\in C([0,T];L^{2}(\Omega)) and limt0+u(t)u0=0\lim_{t\rightarrow 0^{+}}\left\|u(t)-u_{0}\right\|=0. Define

u1:=j=1Eα,1(λjstα)(u0,Φj)Φj,\displaystyle u_{1}:=\sum_{j=1}^{\infty}E_{\alpha,1}(-\lambda_{j}^{s}t^{\alpha})(u_{0},\Phi_{j})\Phi_{j},
u2:=j=1gj(t)(F(τ),Φj)Φj.\displaystyle u_{2}:=\sum_{j=1}^{\infty}g_{j}(t)(F(\tau),\Phi_{j})\Phi_{j}.

Then we have u(,t)=u1(,t)+u2(,t)u(\cdot,t)=u_{1}(\cdot,t)+u_{2}(\cdot,t). We estimate each term separely. For fixed t[0,T],t\in[0,T], by Lemma 2.4 and (6.2), we have

u1(,t)L2(Ω)2=j=1Eα,12(λjstα)(u0,Φj)2u0L2(Ω)2,\left\|u_{1}(\cdot,t)\right\|_{L^{2}(\Omega)}^{2}=\sum_{j=1}^{\infty}E_{\alpha,1}^{2}(-\lambda_{j}^{s}t^{\alpha})(u_{0},\Phi_{j})^{2}\leq\left\|u_{0}\right\|_{L^{2}(\Omega)}^{2}, (6.4)

and

u2(,t)L2(Ω)2=j=1gj2(t)(F(τ),Φj)2j=1(F(τ),Φj)2Γ2(α+1)t2α.\left\|u_{2}(\cdot,t)\right\|_{L^{2}(\Omega)}^{2}=\sum_{j=1}^{\infty}g_{j}^{2}(t)(F(\tau),\Phi_{j})^{2}\leq\sum_{j=1}^{\infty}\frac{(F(\tau),\Phi_{j})^{2}}{\Gamma^{2}(\alpha+1)}t^{2\alpha}. (6.5)

By (6.5), we know

limt0+u2(,t)L2(Ω)=0.\lim_{t\rightarrow 0^{+}}\left\|u_{2}(\cdot,t)\right\|_{L^{2}(\Omega)}=0. (6.6)

Thus define u2(x,0)=0u_{2}(x,0)=0. From (6.4)-(6.5), we obtain

u2(,t)L2(Ω)C1(u0L2(Ω)+FL2(Ω),t[0,T],\left\|u_{2}(\cdot,t)\right\|_{L^{2}(\Omega)}\leq C_{1}(\left\|u_{0}\right\|_{L^{2}(\Omega)}+\left\|F\right\|_{L^{2}(\Omega)},\quad t\in[0,T],

where C1=max{1,TαΓ(α+1)}C_{1}=max\left\{1,\frac{T^{\alpha}}{\Gamma(\alpha+1)}\right\}. For t,t+h[0,T]t,t+h\in[0,T], we have

u(x,t+h)u(x,t)\displaystyle u(x,t+h)-u(x,t) =j=1(Eα,1(λjs(t+h)α)Eα,1(λjstα))(u0,Φj)Φj\displaystyle=\sum_{j=1}^{\infty}(E_{\alpha,1}(-\lambda_{j}^{s}(t+h)^{\alpha})-E_{\alpha,1}(-\lambda_{j}^{s}t^{\alpha}))(u_{0},\Phi_{j})\Phi_{j}
+j=1(gj(t+h)gj(t))(F(τ),Φj)Φj.\displaystyle+\sum_{j=1}^{\infty}(g_{j}(t+h)-g_{j}(t))(F(\tau),\Phi_{j})\Phi_{j}.
=:I1(x,t;h)+I2(x,t;h).\displaystyle=:I_{1}(x,t;h)+I_{2}(x,t;h).

We estimate each term separately. In fact, by Lemma 2.4, we have

I1(,t;h)L2(Ω)2\displaystyle\left\|I_{1}(\cdot,t;h)\right\|_{L^{2}(\Omega)}^{2} =j=1|Eα,1(λjs(t+h)α)Eα,1(λjstα)|2(u0,Φj)2\displaystyle=\sum_{j=1}^{\infty}\left|E_{\alpha,1}(-\lambda_{j}^{s}(t+h)^{\alpha})-E_{\alpha,1}(-\lambda_{j}^{s}t^{\alpha})\right|^{2}(u_{0},\Phi_{j})^{2}
4u0L2(Ω)2,\displaystyle\leq 4\left\|u_{0}\right\|_{L^{2}(\Omega)}^{2},

since limh0|Eα,1(λjs(t+h)α)Eα,1(λjstα)|=0\lim_{h\rightarrow 0}\left|E_{\alpha,1}(-\lambda_{j}^{s}(t+h)^{\alpha})-E_{\alpha,1}(-\lambda_{j}^{s}t^{\alpha})\right|=0, by using the Lebesgue theorem, we have

limh0I1(,t;h)L2(Ω)2=0.\lim_{h\rightarrow 0}\left\|I_{1}(\cdot,t;h)\right\|_{L^{2}(\Omega)}^{2}=0.

By (6.2), we have

I2(,t;h)L2(Ω)2=j=1(gj(t+h)gj(t))2(F(τ),Φj)2CFL2(Ω)2.\left\|I_{2}(\cdot,t;h)\right\|_{L^{2}(\Omega)}^{2}=\sum_{j=1}^{\infty}(g_{j}(t+h)-g_{j}(t))^{2}(F(\tau),\Phi_{j})^{2}\leq C\left\|F\right\|_{L^{2}(\Omega)}^{2}.

Similarly, by using the Lebesgue theorem and Lemma 6.4, we can prove

limh0I2(,t;h)L2(Ω)2=0.\lim_{h\rightarrow 0}\left\|I_{2}(\cdot,t;h)\right\|_{L^{2}(\Omega)}^{2}=0.

Therefore, uC([0,T];L2(Ω))u\in C([0,T];L^{2}(\Omega)). By Lemma 2.4, we know

u(,t)u0()L2(Ω)\displaystyle\left\|u(\cdot,t)-u_{0}(\cdot)\right\|_{L^{2}(\Omega)} (j=1(u0,Φj)2(Eα,1(λjstα)1)2)12+u2(,t)L2(Ω)\displaystyle\leq\left(\sum_{j=1}^{\infty}(u_{0},\Phi_{j})^{2}(E_{\alpha,1}(-\lambda_{j}^{s}t^{\alpha})-1)^{2}\right)^{\frac{1}{2}}+\left\|u_{2}(\cdot,t)\right\|_{L^{2}(\Omega)}
u0L2(Ω)+u2(,t)L2(Ω).\displaystyle\leq\left\|u_{0}\right\|_{L^{2}(\Omega)}+\left\|u_{2}(\cdot,t)\right\|_{L^{2}(\Omega)}.

Since limt0(Eα,1(λjstα)1)=0\lim_{t\rightarrow 0}(E_{\alpha,1}(-\lambda_{j}^{s}t^{\alpha})-1)=0 and (6.6), we have

limt0+u(t)u0=0.\lim_{t\rightarrow 0^{+}}\left\|u(t)-u_{0}\right\|=0.

(2) We verify uL2(0,T;D((Δ)s)).u\in L^{2}(0,T;D((-\Delta)^{s})). By (2.8), we know

(Δ)su(x,t)\displaystyle(-\Delta)^{s}u(x,t) =j=1λjsEα,1(λjstα)(u0,Φj)Φj+j=1λjsgj(t)(F(τ),Φj)Φj\displaystyle=\sum_{j=1}^{\infty}\lambda_{j}^{s}E_{\alpha,1}(-\lambda_{j}^{s}t^{\alpha})(u_{0},\Phi_{j})\Phi_{j}+\sum_{j=1}^{\infty}\lambda_{j}^{s}g_{j}(t)(F(\tau),\Phi_{j})\Phi_{j}
:=v1(x,t)+v2(x,t),\displaystyle:=v_{1}(x,t)+v_{2}(x,t),

where gj(t)g_{j}(t) is defined in (6.1). For 0<tT0<t\leq T, by Definition 2.4, we obtain

v1(,t)L2(Ω)2\displaystyle\left\|v_{1}(\cdot,t)\right\|_{L^{2}(\Omega)}^{2} =j=1(λjsEα,1(λjstα)(u0,Φj))2\displaystyle=\sum_{j=1}^{\infty}(\lambda_{j}^{s}E_{\alpha,1}(-\lambda_{j}^{s}t^{\alpha})(u_{0},\Phi_{j}))^{2}
j=1(λjs(u0,Φj)2(cλjs1+λjstα)2)Cu0D((Δ)s)2tα.\displaystyle\leq\sum_{j=1}^{\infty}\left(\lambda_{j}^{s}(u_{0},\Phi_{j})^{2}\left(\frac{c\sqrt{\lambda_{j}^{s}}}{1+\lambda_{j}^{s}t^{\alpha}}\right)^{2}\right)\leq C\frac{\left\|u_{0}\right\|_{D((-\Delta)^{s})}^{2}}{t^{\alpha}}. (6.7)

where CC is a constant depending on α\alpha only. For the second term v2v_{2}, by (6.3) we can deduce that

v2(,t)L2(Ω)2=j=1λj2sgj2(t)(F(τ),Φj)2j=1(F(τ),Φj)2FL2(Ω)2.\left\|v_{2}(\cdot,t)\right\|_{L^{2}(\Omega)}^{2}=\sum_{j=1}^{\infty}\lambda_{j}^{2s}g_{j}^{2}(t)(F(\tau),\Phi_{j})^{2}\leq\sum_{j=1}^{\infty}(F(\tau),\Phi_{j})^{2}\leq\left\|F\right\|_{L^{2}(\Omega)}^{2}. (6.8)

By estomates (1)-(6.8), we know v1,v2L2(0,T;L2(Ω)),v_{1},v_{2}\in L^{2}(0,T;L^{2}(\Omega)), hence (Δ)suL2(0,T;L2(Ω))\left(-\Delta\right)^{s}u\in L^{2}(0,T;L^{2}(\Omega)). Moreover, we can obtain the following estimate from (1)-(6.8)

uL2(0,T;D((Δ)s))=(Δ)suL2(0,T;L2(Ω))\displaystyle\left\|u\right\|_{L^{2}(0,T;D((-\Delta)^{s}))}=\left\|\left(-\Delta\right)^{s}u\right\|_{L^{2}(0,T;L^{2}(\Omega))}
C2(α,T,Ω)(u0D((Δ)s)+FL2(Ω)2.\displaystyle\leq C_{2}(\alpha,T,\Omega)(\left\|u_{0}\right\|_{D((-\Delta)^{s})}+\left\|F\right\|_{L^{2}(\Omega)}^{2}.

where C2C_{2} is a positive constant. Therefore, uL2(0,T;D((Δ)s)).u\in L^{2}(0,T;D((-\Delta)^{s})).

(3) We prove that tαu(x,t)C((0,T];L2(Ω))L2(0,T;L2(Ω))\partial_{t}^{\alpha}u(x,t)\in C((0,T];L^{2}(\Omega))\cap L^{2}(0,T;L^{2}(\Omega)) and (1.1)-(1.2) holds in L2(Ω)L^{2}(\Omega) for t(0,T].t\in(0,T]. By Lemma 6.3 and Lemma 6.2, we have

tαu(x,t)\displaystyle\partial_{t}^{\alpha}u(x,t) =j=1λjsEα,1(λjstα)(u0,Φj)Φj\displaystyle=-\sum_{j=1}^{\infty}\lambda_{j}^{s}E_{\alpha,1}(-\lambda_{j}^{s}t^{\alpha})(u_{0},\Phi_{j})\Phi_{j}
+j=1(F(τ),Φj)[1λjs0t(tτ)α1Eα,α(λjs(tτ)α)𝑑τ]Φj\displaystyle+\sum_{j=1}^{\infty}(F(\tau),\Phi_{j})[1-\lambda_{j}^{s}\int_{0}^{t}(t-\tau)^{\alpha-1}E_{\alpha,\alpha}(-\lambda_{j}^{s}(t-\tau)^{\alpha})d\tau]\Phi_{j}
=F(t)(Δ)su(x,t).\displaystyle=F(t)-(-\Delta)^{s}u(x,t).

Hence tαu(x,t)C((0,T];L2(Ω))L2(0,T;L2(Ω))\partial_{t}^{\alpha}u(x,t)\in C((0,T];L^{2}(\Omega))\cap L^{2}(0,T;L^{2}(\Omega)) and (1.1)-(1.2) holds in L2(Ω)L^{2}(\Omega) for t(0,T].t\in(0,T].

(4) We prove the uniqueness of the weak solution to (1.1)-(1.2). Under the condition F(t)=μu2(1kJu)γu=0,u0=0F(t)=\mu u^{2}(1-kJ*u)-\gamma u=0,u_{0}=0, we need to prove that systems (1.1)-(1.2) has only a trivial solution. We take the inner product of (1.1) with Φj(x)\Phi_{j}(x). Using the Green formula and Φj(x)|Ω=0\Phi_{j}(x)|\partial\Omega=0 and setting uj(t):=(u(,t),Φj(x)),u_{j}(t):=(u(\cdot,t),\Phi_{j}(x)), we obtain

{tαuj(t)=λjsuj(t),t(0,T],uj(0)=0.\displaystyle\left\{\begin{matrix}\partial_{t}^{\alpha}u_{j}(t)=-\lambda_{j}^{s}u_{j}(t),&t\in(0,T],\\ u_{j}(0)=0.&\end{matrix}\right.

Due to the existence and uniqueness of the ordinary fractional differential equation in [44], we obtain that uj(t)=0,j=1,2,.u_{j}(t)=0,j=1,2,\cdots. Since {Φj}j1\left\{\Phi_{j}\right\}_{j\geq 1} is an orthonormal basis in L2(Ω),L^{2}(\Omega), we have u=0u=0 in ω×(0,T]\omega\times(0,T], Thus the proof is complete.

Remark 6.2.

The method of this Lemma is used by [37]. However, in reference [37], there is non-local term p(t)f(x)p(t)f(x). In the paper, There is non-local F(x,t)=μu2(1kJu)γuF(x,t)=\mu u^{2}(1-kJ*u)-\gamma u which is different from [37]. From [5], we obtain the conclusion, but “For brevity, we leave the detail to the reader.”[p255,Theorem 1,[5]]. Therefore, We have done the above related proof.

A.3 Definition of weak and strong solution for nonlinear fractional diffusion equation

We call here the definition of weak and strong solution taken from [1]. Considering the following Cauchy problem

{ut+(Δ)σ/2(|u|m1u)=0,xN,t>0u(x,0)=f(x),xN\displaystyle\left\{\begin{matrix}\frac{\partial u}{\partial t}+(-\Delta)^{\sigma/2}(\left|u\right|^{m-1}u)=0,&x\in\mathbb{R}^{N},t>0\\ u(x,0)=f(x),&x\in\mathbb{R}^{N}\end{matrix}\right. (6.9)
Definition 6.1.

[1] A function uu is a weak solution to the problem (6.9) if:

  • 1.

    uC((0,);L1(d))u\in C((0,\infty);L^{1}(\mathbb{R}^{d})) and |u|m1uLloc2((0,);H˙σ/2(d));\left|u\right|^{m-1}u\in L_{loc}^{2}((0,\infty);\dot{H}^{\sigma/2}(\mathbb{R}^{d}));

  • 2.

    The identity

    0duφt𝑑x𝑑t+0d(Δ)σ/4(|u|m1u)(Δ)σ/4φ𝑑x𝑑t=0\displaystyle\int_{0}^{\infty}\int_{\mathbb{R}^{d}}u\frac{\partial\varphi}{\partial t}dxdt+\int_{0}^{\infty}\int_{\mathbb{R}^{d}}(-\Delta)^{\sigma/4}(\left|u\right|^{m-1}u)(-\Delta)^{\sigma/4}\varphi dxdt=0

    holds for every φC01(d×(0,))\varphi\in C_{0}^{1}(\mathbb{R}^{d}\times(0,\infty));

  • 3.

    u(,0)=u0L1(N)u(\cdot,0)=u_{0}\in L^{1}(\mathbb{R}^{N}) almost everywhere.

Note that the fractional Sobolev space H˙σ/2(d)\dot{H}^{\sigma/2}(\mathbb{R}^{d}) is defined as the completion of C0(d)C_{0}^{\infty}(\mathbb{R}^{d}) with the norm

ψH˙σ/2=(d|ξ|σ|ψ|^2𝑑ξ)1/2=(Δ)σ/4ψ2\left\|\psi\right\|_{\dot{H}^{\sigma/2}}=\left(\int_{\mathbb{R}^{d}}\left|\xi\right|^{\sigma}\hat{\left|\psi\right|}^{2}d\xi\right)^{1/2}=\left\|(-\Delta)^{\sigma/4}\psi\right\|_{2}
Definition 6.2.

[1] We say that a weak solution uu to the problem (6.9) is a strong solution if tuL((τ,);L1(N)),τ>0.\partial_{t}u\in L^{\infty}((\tau,\infty);L^{1}(\mathbb{R}^{N})),\tau>0.

On the other hand, we recall the definition of weak solution taken from [62]. Considering the following direct problem:

{βutβ=(a(u)u)+f(x,t),(x,t)ΩT,u(x,0)=0,xΩ¯,u(x,t)=0,(x,t)Γ1×[0,T],Γ1Ωa(u)un=φ(x,t),(x,t)Γ2×[0,T],Γ2Ω,\displaystyle\left\{\begin{matrix}\frac{\partial^{\beta}u}{\partial t^{\beta}}=\triangledown\cdot(a(u)\triangledown u)+f(x,t),&(x,t)\in\Omega_{T},\\ u(x,0)=0,&x\in\bar{\Omega},\\ u(x,t)=0,&(x,t)\in\Gamma_{1}\times[0,T],\Gamma_{1}\subset\partial\Omega\\ a(u)\frac{\partial u}{\partial n}=\varphi(x,t),&(x,t)\in\Gamma_{2}\times[0,T],\Gamma_{2}\subset\partial\Omega,\end{matrix}\right. (6.10)

where ΩT:=Ω×(0,T),\Omega_{T}:=\Omega\times(0,T), the domain Ωn(n1)\Omega\subset\mathbb{R}^{n}(n\geq 1) is assumed to be bounded simple connected with a piecewise smooth boundary Γ\Gamma and Γ1Γ2=,Γ1¯Γ2¯=Γ\Gamma_{1}\cap\Gamma_{2}=\varnothing,\bar{\Gamma_{1}}\cup\bar{\Gamma_{2}}=\Gamma, mean (Γi)0,i=1,2.\left(\Gamma_{i}\right)\neq 0,i=1,2.

Definition 6.3.

[62] A weak solution of problem (6.10) is a function

uL2(0,T;H01(Ω))W2β(0,T;L2(Ω))u\in L^{2}(0,T;H_{0}^{1}(\Omega))\cap W_{2}^{\beta}(0,T;L^{2}(\Omega))

such that the following integral identity holds for a.e t[0,T]:t\in[0,T]:

Ωβutβv𝑑x+Ωa(u)uv𝑑x=Ωfv𝑑x+Γ2φv𝑑x,\int_{\Omega}\frac{\partial^{\beta}u}{\partial t^{\beta}}vdx+\int_{\Omega}a(u)\triangledown u\cdot\triangledown vdx=\int_{\Omega}fvdx+\int_{\Gamma_{2}}\varphi vdx,

for each vL2(0,T;H01(Ω))W2β(0,T;L2(Ω))v\in L^{2}(0,T;H_{0}^{1}(\Omega))\cap W_{2}^{\beta}(0,T;L^{2}(\Omega)), where

W2β(0,T):={uL2[0,T]:βutβL2[0,T]andu(0)=0}W_{2}^{\beta}(0,T):=\left\{u\in L^{2}\in[0,T]:\frac{\partial^{\beta}u}{\partial t^{\beta}}\in L^{2}[0,T]\,\text{and}\,u(0)=0\right\}

is the fractional Sobolev space of order β.\beta.

References

  • [1] A. de Pablo, F. Quirós, A. Rodríguez, J. L. Vázquez, A general fractional porous medium equation, Commun. Pur. App. Math. 65 (2012) 1242–1284.
  • [2] G. Acosta, J. P. Borthagaray, A fractional Laplace equation: regularity of solutions and finite element approximations, SIAM. J. Numer. Anal. 55 (2017) 472–495.
  • [3] Y. Zhou, L. Shangerganesh, J. Manimaran, A. Debbouche, A class of time fractional reaction-diffusion equation with nonlocal boundary condition, Math. Method. Appl. Sci. 41 (2018) 2987–2999.
  • [4] C. Cheng, L. Chen, J. Li, Global boundedness and the Allee effect in a nonlocal bistable reaction-diffusion equation in population dynamics, Nonlinear. Anal-Real. 60 (2021) 103–309.
  • [5] D. T. Dang, D. N. D. Hai, Backward problem for time-space fractional diffusion equations in Hilbert scales, Comput. Math. Appl. 93 (2021) 253–264.
  • [6] K. S. Miller, B. Ross, An Introduction to The Fractional Calculus and Fractional Differential Equations, Wiley, New Jersey, 1993.
  • [7] C. Li, Y. Chen, J. Kurths, Fractional calculus and its applications, Philos. T. R. Soc. A. 371 (2013) 37–45.
  • [8] H. Dong, D. Kim, LpL_{p}-estimates for time fractional parabolic equations with coefficients measurable in time, Adv. Math. 345 (2019) 289–345.
  • [9] T. Sandev, R. Metzler, Ž. Tomovski, Fractional diffusion equation with a generalized Riemann-Liouville time fractional derivative, J. Phys. A-Math Theor. 44 (2011) 203–255.
  • [10] F. Jarad, T. Abdeljawad, Generalized fractional derivatives and Laplace transform, Discrete. Cont. Dyn-S. 13 (2020) 709.
  • [11] Z. Gan, Y. He, L. Meng, Large time behavior and convergence for the Camassa–Holm equations with fractional Laplacian viscosity, Calc. Var. Partial. Dif. 57 (2018) 1–48.
  • [12] F. Mainardi, Y. Luchko, G. Pagnini, The fundamental solution of the space-time fractional diffusion equation, Fract. Calc. Appl. Anal. 4 (2007) 153–192.
  • [13] M. Alfaro, N. Apreutesei, F. Davidson, V. Volpert, Preface to the issue nonlocal reaction-diffusion equations, Math. Model. Natural Phenomena 10 (2015) 1–5.
  • [14] M. Alfaro, J. Coville, G. Raoul, Travelling waves in a nonlocal reaction-diffusion equation as a model for a population structured by a space variable and a phenotypic trait, Commun. Part. Diff. Eq. 38 (2013) 2126–2154.
  • [15] J. Li, L. Chen, C. Surulescu, Global boundedness, hair trigger effect, and pattern formation driven by the parametrization of a nonlocal Fisher-KPP problem, J. Differ. Equations. 269 (2020) 9090–9122.
  • [16] J. Li, E. Latos, L. Chen, Wavefronts for a nonlinear nonlocal bistable reaction–diffusion equation in population dynamics, J. Differ. Equations. 263 (2017) 6427–6455.
  • [17] W. Sheng, Z. Wang, Entire solutions of monotone bistable reaction–diffusion systems in N\mathbb{R}^{N}, Calc. Var. Partial. Dif. 57 (2018) 1–36.
  • [18] A. Alsaedi, M. Kirane, R. Lassoued, Global existence and asymptotic behavior for a time fractional reaction-diffusion system, Comput. Math. Appl. 73 (2017) 951–958.
  • [19] R. Gorenflo, Y. Luchko, M. Yamamoto, Time-fractional diffusion equation in the fractional Sobolev spaces, Fract. Calc. Appl. Anal. 18 (2015) 799–820.
  • [20] B. Ahmad, M. S. Alhothuali, H. H. Alsulami, M. Kirane, S. Timoshin, On a time fractional reaction diffusion equation, Appl. Math. Comput. 257 (2015) 199–204.
  • [21] Q. Zhang, H. Sun, Y. Li, Global existence and blow-up of solutions of the Cauchy problem for a time fractional diffusion system, Comput. math. Appl. 78 (2019) 1357–1366.
  • [22] R. Tapdigoglu, B. T. Torebek, Global existence and blow-up of solutions of the time-fractional space-involution reaction-diffusion equation, Turk. J. Math. 44 (2020) 960–969.
  • [23] S. Umarov, On fractional Duhamel’s principle and its applications, J. Differ. Equations. 252 (2012) 5217–5234.
  • [24] W. Chen, L. D ’Ambrosio, Y. Li, Some Liouville theorems for the fractional Laplacian, Nonlinear. Anal-Theor. 121 (2015) 370–381.
  • [25] M. Chang, H. Wu, Existence of weak solutions for the nonlocal energy-weighted fractional reaction-diffusion equations, J. Evol. Equ. 19 (2019) 883–914.
  • [26] H. Zhang, Asymptotic behavior of non-negative solutions to nonlinear fractional reaction-diffusion equations, Master’s thesis, Huazhong University of Science and Technology (2015).
  • [27] T. Kakehi, Y. Oshita, Blow up and global existence of a solution to a semilinear reaction-diffusion system with the fractional Laplacian, Mathematical Journal of Okayama University 59 (2017) 175–218.
  • [28] F. Ahmad, G. Karch, Decay of mass for nonlinear equation with fractional Laplacian, Monatsh. Math. 160 (2010) 375–384.
  • [29] K. H. Karlsen, ulusoy Suleyman, Stability of entropy solutions for Lévy mixed hyperbolic-parabolic equations, J. Differ. Equations. 538 (2011) 966–967.
  • [30] E. D. Nezza, G. Palatucci, E. Valdinoci, Hitchhiker’s guide to the fractional Sobolev spaces, B. Sci. Math. 136 (5) (2012) 521–573.
  • [31] M. M. Meerschaert, D. A. Benson, H. P. Scheffler, B. Baeumer, Stochastic solution of space-time fractional diffusion equations, Physical. Review. E. 65 (2002) 41–103.
  • [32] Z. Chen, M. M. Meerschaert, E. Nane, Space-time fractional diffusion on bounded domains, J. Math. Anal. Appl. 393 (2011) 479–488.
  • [33] M. Warma, Approximate controllabilty from the exterior of space-time fractional diffusion equations with the fractional Laplacian (2018). arXiv:1802.08028.
  • [34] J. Hua, C. Zhou, Boundary stabilization for time-space fractional diffusion equation, Eur. J. Control. 65 (2022) 639–643.
  • [35] S. Tatar, S. Ulusoy, An inverse source problem for a one-dimensional space–time fractional diffusion equation, Appl. Anal. 67 (2015) 2233–2244.
  • [36] S. Tatar, R. Tinaztepe, S. Ulusoy, Determination of an unknown source term in a space-time fractional diffusion equation, J. Fract. Calc. Appl. 6 (2015) 83–90.
  • [37] K. Sakamoto, M. Yamamoto, Initial value/boundary value problems for fractional diffusion-wave equations and applications to some inverse problems, J. Math. Anal. Appl. 382 (2011) 426–447.
  • [38] Y. S. Li, T. Wei, An inverse time-dependent source problem for a time–space fractional diffusion equation, Appl. Math. Comput. 336 (2018) 257–271.
  • [39] T. Wei, J. Wang, A modified quasi-boundary value method for an inverse source problem of the time-fractional diffusion equation, Appl. Numer. Math. 78 (2014) 95–111.
  • [40] T. Wei, X. L. Li, Y. S. Li, An inverse time-dependent source problem for a time-fractional diffusion equation, Inverse. Probl. 32 (2016) 24–30.
  • [41] R. Shen, M. Xiang, V. D. Râdulescu, Time-space fractional diffusion problems: existence, decay estimates and blow-Up of solutions, Milan. J. Math. (2022) 1–27.
  • [42] A. Nabti, Life span of blowing-up solutions to the Cauchy problem for a time–space fractional diffusion equation, Comput. Math. Appl. 78 (2018) 1302–1316.
  • [43] A. Alsaedi, M. Kirane, B. T. Torebek, Global existence and blow-up for a space and time nonlocal reaction-diffusion equation, Quaest. Math. 44 (2020) 1–7.
  • [44] A. A. Kilbas, H. M. Srivastava, J. J. Trujillo, Theory and applications of fractional differential equations, Vol. 204, Elsevier, New York, 2006.
  • [45] H. Pollard, The completely monotonic character of the Mittag-Leffler function Eα(X)E_{\alpha}(-X), B. Am. Math. Soc. 54 (1948) 1115–1116.
  • [46] E. G. B. jlekova, Fractional Evolution Equations in Banach Spaces, Eindhoven University of Technology Ndhoven 14 (2001) 737–745.
  • [47] A. Alsaedi, B. Ahmad, M. Kirane, A survey of useful inequalities in fractional calculus, Fract. Calc. Appl. Anal. 20 (2017) 574–594.
  • [48] H. Zhan, F. Gao, L. Guo, Global boundedness and Allee effect for a nonlocal time fractional reaction-diffusion equation (2021). arXiv:2112.11143.
  • [49] M. Allen, L. Caffarelli, A. Vasseur, A parabolic problem with a fractional time derivative, Arch. Ration. Mech. An. 221 (2016) 603–630.
  • [50] G. Autuori, P. Pucci, Elliptic problems involving the fractional Laplacian in N\mathbb{R}^{N}, J. Differ. Equations. 255 (2013) 2340–2362.
  • [51] V. Maz’Ya, T. Shaposhnikova, On the bourgain, brezis, and mironescu theorem concerning limiting embeddings of fractional sobolev spaces, J. Funct. Anal. 195 (2002) 230–238.
  • [52] A. Wenhui, The equivalence of Gagliardo-Nirenberg-Sobolev inequality,Nash inequality and Isoperimetric inequality, Central China Normal University (2017) 39.
  • [53] L. Nirenberg, An extended interpolation inequality, Ann. Scuola. Norm-Sci. 20 (1966) 733–737.
  • [54] J. Cheng, Inequalities in partial differential equations and their proofs, Journal of Zaozhuang University 33 (2016) 43–46.
  • [55] S. Tatar, R. Tınaztepe, S. Ulusoy, Simultaneous inversion for the exponents of the fractional time and space derivatives in the space-time fractional diffusion equation, Appl. Anal. 95 (2014) 1–23.
  • [56] L. Brasco, E. Parini, Enea, The second eigenvalue of the fractional pp-plaplacian, Adv. Calc. Var. 9 (2019) 323–355.
  • [57] H. Dorsaf, K. Ferdaous, L. Rafika, Blowing-up solutions and global solutions to a fractional differential equation, Fract. Differ. Calc. 4 (2014) 45–53.
  • [58] B. Zhong, h. Xiang, Extinction behavior of solutions for the pp-Laplacian equations with nonlocal sources, Nonlinear. Anal-Real. 13 (2012) 1780–1789.
  • [59] S. Tatar, S. Ulusoy, A uniqueness result for an inverse problem in a space-time fractional diffusion equation, Electron. J. Differ. Equ. 257 (2013) 1–9.
  • [60] Q. Lü, E. Zuazua, On the lack of controllability of fractional in time ODE and PDE, Math. Control. Signals. 28 (2016) 1–21.
  • [61] Y. Sugiyama, H. Kunii, Global existence and decay properties for a degenerate Keller-Segel model with a power factor in drift term, J. Differ. Equations. 227 (2006) 333–364.
  • [62] S. Tatar, S. Ulusoy, An inverse problem for a nonlinear diffusion equation with time-fractional derivative, J. Inverse. Ill-pose. P. 25 (2017) 185–193.
  • [63] M. Bonforte, J. L. Vázquez, Quantitative local and global a priori estimates for fractional nonlinear diffusion equations, Adv. Math. 250 (2014) 242–284.
  • [64] S. Bian, J. Liu, C. Zou, Ultra-contractivity for Keller-Segel model with diffusion exponent m2/dm\>-2/d, Kinet. Relat. Mod. 7 (2014) 9–28.
  • [65] A. Alsaedi, B. Ahmad, M. Kirane, Maximum principle for certain generalized time and space fractional diffusion equations, Q. Appl. Math. 73 (2015) 163–175.
  • [66] B. Abdellaoui, A. Attar, R. Bentifour, I. Peral, On fractional pp-laplacian parabolic problem with general data, Ann. Mat. Pur. Appl. 197 (2018) 329–356.
  • [67] L. Nirenberg, On elliptic partial differential equations, Ann. Scuola. Norm-Sci. 13 (1959) 115–162.
  • [68] E. Gagliardo, Proprietà di alcune classi di funzioni in più variabili, Matematika 5 (1961) 87–116.
  • [69] S. Bian, L. Chen, A nonlocal reaction diffusion equation and its relation with Fujita exponent, J. Math. Anal. Appl. 444 (2016) 1479–1489.
  • [70] E. Bazhlekova, Existence and uniqueness results for a fractional evolution equation in Hilbert space, Fract. Calc. Appl. Anal. 15 (2012) 232–243.
  • [71] E. Zeidler, Nonlinear Functional Analysis and Its Applications: II/B: Nonlinear Monotone Operators, Springer Science & Business Media, New York, 2013.
  • [72] T. Wei, Z. Q. Zhang, Robin coefficient identification for a time-fractional diffusion equation, Appl. Numer. Math. 111 (2017) 160–180.