Research article Special Issues

Finite element approximation of fractional hyperbolic integro-differential equation

  • In this article, we propose a Galerkin finite element method for numerically solving a type of fractional hyperbolic integro-differential equation, which can be considered as the generalization of the classical hyperbolic Volterra integro-differential equation. Along with Galerkin finite element method in spatial direction, we apply a second order symmetric difference method in time. Next we discuss the regularity analysis of the weak solution and convergence analysis of the semi-discrete scheme. Then we further study the stability analysis and the error estimation of the fully discrete problems, according to the properties of fractional Ritz-Volterra projection, Ritz projection and so on. Numerical examples with comparisons among the proposed schemes verify our theoretical analyses.

    Citation: Zhengang Zhao, Yunying Zheng, Xianglin Zeng. Finite element approximation of fractional hyperbolic integro-differential equation[J]. AIMS Mathematics, 2022, 7(8): 15348-15369. doi: 10.3934/math.2022841

    Related Papers:

    [1] Ailing Zhu, Yixin Wang, Qiang Xu . A weak Galerkin finite element approximation of two-dimensional sub-diffusion equation with time-fractional derivative. AIMS Mathematics, 2020, 5(5): 4297-4310. doi: 10.3934/math.2020274
    [2] Mohammed A. Almalahi, Satish K. Panchal, Fahd Jarad, Mohammed S. Abdo, Kamal Shah, Thabet Abdeljawad . Qualitative analysis of a fuzzy Volterra-Fredholm integrodifferential equation with an Atangana-Baleanu fractional derivative. AIMS Mathematics, 2022, 7(9): 15994-16016. doi: 10.3934/math.2022876
    [3] Sumbal Ahsan, Rashid Nawaz, Muhammad Akbar, Saleem Abdullah, Kottakkaran Sooppy Nisar, Velusamy Vijayakumar . Numerical solution of system of fuzzy fractional order Volterra integro-differential equation using optimal homotopy asymptotic method. AIMS Mathematics, 2022, 7(7): 13169-13191. doi: 10.3934/math.2022726
    [4] Khalid K. Ali, K. R. Raslan, Amira Abd-Elall Ibrahim, Mohamed S. Mohamed . On study the fractional Caputo-Fabrizio integro differential equation including the fractional q-integral of the Riemann-Liouville type. AIMS Mathematics, 2023, 8(8): 18206-18222. doi: 10.3934/math.2023925
    [5] Ahmed Morsy, Kottakkaran Sooppy Nisar, Chokkalingam Ravichandran, Chandran Anusha . Sequential fractional order Neutral functional Integro differential equations on time scales with Caputo fractional operator over Banach spaces. AIMS Mathematics, 2023, 8(3): 5934-5949. doi: 10.3934/math.2023299
    [6] Qun Dai, Shidong Liu . Stability of the mixed Caputo fractional integro-differential equation by means of weighted space method. AIMS Mathematics, 2022, 7(2): 2498-2511. doi: 10.3934/math.2022140
    [7] Lakhlifa Sadek, Tania A Lazǎr, Ishak Hashim . Conformable finite element method for conformable fractional partial differential equations. AIMS Mathematics, 2023, 8(12): 28858-28877. doi: 10.3934/math.20231479
    [8] Yanhua Gu . High-order numerical method for the fractional Korteweg-de Vries equation using the discontinuous Galerkin method. AIMS Mathematics, 2025, 10(1): 1367-1383. doi: 10.3934/math.2025063
    [9] Obaid Algahtani, M. A. Abdelkawy, António M. Lopes . A pseudo-spectral scheme for variable order fractional stochastic Volterra integro-differential equations. AIMS Mathematics, 2022, 7(8): 15453-15470. doi: 10.3934/math.2022846
    [10] Zahra Pirouzeh, Mohammad Hadi Noori Skandari, Kamele Nassiri Pirbazari, Stanford Shateyi . A pseudo-spectral approach for optimal control problems of variable-order fractional integro-differential equations. AIMS Mathematics, 2024, 9(9): 23692-23710. doi: 10.3934/math.20241151
  • In this article, we propose a Galerkin finite element method for numerically solving a type of fractional hyperbolic integro-differential equation, which can be considered as the generalization of the classical hyperbolic Volterra integro-differential equation. Along with Galerkin finite element method in spatial direction, we apply a second order symmetric difference method in time. Next we discuss the regularity analysis of the weak solution and convergence analysis of the semi-discrete scheme. Then we further study the stability analysis and the error estimation of the fully discrete problems, according to the properties of fractional Ritz-Volterra projection, Ritz projection and so on. Numerical examples with comparisons among the proposed schemes verify our theoretical analyses.



    The field of fractional calculus, which deals with mathematical analysis and physical applications of derivatives and integrals of arbitrary order, has become a hot topic nowadays. Fractional derivatives and integral can be considered as the generalization of integer order derivative. However, there are quite different between the fractional derivative and the integer order derivative. One of the reasons is that fractional operators, such as Riemann-Liouville derivative or Caputo derivative, has the nonlocal characteristics and weakly singular properties. But just because of above characteristics, the fractional calculus performs more perfectly than the classical calculus, especially in the field of anomalous diffusion problems. Many published papers reveal that fractional models, such as the fractional differential equations and the fractional integro-differential equations show more realistic dynamic behavior than the classical differential equations and the classical integro-differential ones.

    Fractional integro-differential equations often describe the anomalous diffusion phenomena which come from the dynamic behaviors of viscoelastic materials, heat conduction with memory, and so on [1,2,3,4]. They can be divided into parabolic integro-differential equations and hyperbolic integro-differential equations along the time axis. The fractional hyperbolic integro-differential equations can be modelled by the generalized constitutive relations between stress σ and strain ϵ of the linear viscoelasticity [5]. If the corresponding generalized constitutive relation satisfies

    σ(t)=E0ϵ(t)t0a(ts)ϵ(s)ds, (1.1)

    where a(t) is the stress relaxation modulus and E0 is the Young's modulus. Then substituting (1.1) into its motion equation ρutt=divσ+f, we can obtain a type of generalized hyperbolic integro-differential equations

    utta2uxx=t0K(ts)uxxds+f(x,t), (1.2)

    where u is the displacement, ρ is the density, f is the external force. If the kernel K belongs to the form of power law (e.g. the form tβΓ(1+β), where Γ is the Gamma function), (1.2) is retained to the temporal fractional integro-differential equations, in which the power law widely exists in complex systems [6]. Meanwhile, if the displacement u of continuous media satisfies the 2α(1/2<α<1) order Lévy stable distribution in spatial direction, by applying the power law approximation form of its Fourier transform and inverse Fourier transform, Eq (1.2) turns to be the following fractional hyperbolic integro-differential equation

    utta22αu|x|2α=1Γ(1+β)t0(ts)β2αu|x|2αds+f(x,t), (1.3)

    where 2α|x|2α is the 2α order Riesz fractional derivative, often describes the 2α order Lévy flights [7]. The existence and uniqueness of the analytic solution of the above fractional hyperbolic integro-differential equation can be proved by using the Fourier transform and the Laplace transform. It is omitted here, because the proof is very similar to the corresponding parabolic problem. One can refer to [8,9].

    There are several methods to study the fractional hyperbolic equations. Dassios and Font [10] studied the analytical solution of the time-fractional hyperbolic heat equation, in which the fractional derivatives contain three kinds of definitions. Kumar and Rai [11] presented a fractional hyperbolic bioheat transfer model and used a hybrid numerical scheme based on fractional Legendre wavelet method and finite difference scheme to study the numerical solution. Ashyralyev, Dal and Pinar [12] studied an initial boundary value problem for the fractional hyperbolic equation by difference scheme and discussed the stability details. Qiu et al. [13] constructed a formally second-order BDF finite difference scheme for a integro-differential equations with the multi-term kernels. Qiu et al. [14] presented a formally second-order backward differentiation formula for the Volterra integro-differential equation with a weakly singular kernel. As we all known, finite element method, finite difference method and spectral method are the classical numerical methods. They have been widely applied not only in the classical hyperbolic equations, but also in the fractional parabolic differential equations, e.g. [9,15,16,17,18,19,20,21,22,23,24,25,26,27]. However, for the fractional hyperbolic integro-differental equation, there are relatively few. Moreover, it is quite difficult to get the high accuracy and the high convergence order methods, because the fractional integral and the fractional derivative are mixed into one term.

    In this paper, we consider a Galerkin finite element method to solve the initial-boundary value problem of the fractional hyperbolic integro-differential equation

    uttκ2αu|x|2α=J1+β0,t2αu|x|2α+f(x,t), (1.4)

    together with the homogeneous Dirichlet boundary conditions and the initial conditions

    u(x,0)=u0(x),ut(x,0)=u1(x),xin[a,b], (1.5)

    where J1+β0,t(0<β<1) represents the temporal Riemann-Liouville integral, κ is assumed to be a nonnegative constant, which represents the diffusion rate of particles. Here we use the time stepping method based on the symmetric difference approach (Un1/2+Un+1/2)/2 to approximate u(tn) of the Riesz derivative part in (1.4), and center difference approach (Un12Un+Un+1)/2 to approximate utt(tn), combining with the high order quadrature schemes based on the product trapezoidal formula to treat the Riemann-Liouville integral term. Meanwhile, we use the Galerkin finite element method with r1 order piecewise polynomial as the shape function in space. The expected goal of our convergence order is O(hr+k2). Theoretical analyses and numerical experiments of the designed algorithm will be presented in the following paper.

    We organize the following sections. In Section 2, we introduce the preliminary definitions and properties of fractional integral and fractional derivatives. In Section 3 and Section 4, we construct a Galerkin finite element scheme for the fractional hyperbolic integro-differential equation, and then present the regularity analysis of the weak solution and convergence analysis of the semi-discrete scheme separately. In Section 5, we derive the fully discrete scheme based on the symmetric difference method in time direction. Then we further discuss the stability analysis and error estimate of the fully discrete scheme. In Section 6, we present some numerical experiments to illustrate the efficiency of the theoretical analyses.

    We first introduce some definitions and notations of fractional calculus[28,29], and some properties of fractional derivative space as well.

    Definition 2.1 The Riemann-Liouville integral of order α is defined as

    Jα0,tu(t)=1Γ(α)t0(ts)α1u(s)ds,

    where α>0.

    Definition 2.2 The Riesz derivatives of order α is defined as

    αu|x|α=Cα(Dαa,x+Dαx,b)u(x),

    where the left and right Riemann-Liouville derivatives are defined as

    Dαa,xu(x)=1Γ(nα)dndxnxa(xτ)nα1u(τ)dτ,
    Dαx,bu(x)=(1)nΓ(nα)dndxnbx(τx)nα1u(τ)dτ,

    where Cα=12cos(πα/2),n1<α<n.

    The fractional Sobolev space needs to be denoted. On the fractional Sobolev space Hα(Ω), we denote the semi-norm ||α by

    |u|α=(Dαa,xu,Dαx,bu),

    or

    |u|α=Dαa,xu(x),

    and or

    |u|α=Dαx,bu(x).

    One can prove that they are equivalent to each other [15,17,18,20,29]. Then we define the norm α by

    uα=(u2+|u|2α)1/2.

    Lemma 2.1. ([29]) The Riemann-Liouville integral operator has the following properties

    Jp0,tJq0,tu(t)=Jp+q0,tu(t),
    ddtJ1+p0,tu(t)=Jp0,tu(t),

    where p,q>0.

    Lemma 2.2. ([15]) There exist constants C1,C2,C3,C4>0 such that for any uH2α0(Ω), and vHα0(Ω),

    (D2αa,xu,u)=(Dαa,xu,Dαx,bu)C1u2α,
    (D2αx,bu,u)=(Dαx,bu,Dαa,xu)C2u2α,
    (D2αa,xu,v)=(Dαa,xu,Dαx,bv)C3uαvα,
    (D2αx,bu,v)=(Dαx,bu,Dαa,xv)C4uαvα.

    According to the above Lemma 2.2, the original problem turns to be the following weak form

    (utt,v)κCα[(Dαa,xu,Dαx,bv)+(Dαx,bu,Dαa,xv)]=Cα[(J1+β0,tDαa,xu,Dαx,bv)+(J1+β0,tDαx,bu,Dαa,xv)]+(f,v) (3.1)

    with the homogeneous Dirichlet boundary conditions and the initial conditions

    u(x,0)=u0,ut(x,0)=u1. (3.2)

    Grönwall's inequality is important to our regularity analysis and error estimate.

    Lemma 3.1. (Grönwall's inequality) ([5]) Assume that the nonnegative function y(t) satisfies the integral inequality

    y(t)t0x(s)y(s)ds+f(t), (3.3)

    where x(t)0 and f(t) are absolutely integrable, then

    y(t)f(t)+t0x(τ)f(τ)et0x(s)dsdτ. (3.4)

    Now we consider the regularity of the weak solution. We always denote C as a generic positive constant which may be changed at different situations from now.

    Theorem 3.1 The weak solution u(x,t) in (3.1) satisfies the following energy estimate

    ut+uαC(u1+u0α+t0fdt).

    Proof. Setting v=ut in (3.1), it becomes

    (utt,ut)κCα[(Dαa,xu,Dαx,but)+(Dαx,bu,Dαa,xut)]=Cα[(J1+β0,tDαa,xu,Dαx,but)+(J1+β0,tDαx,bu,Dαa,xut)]+(f,ut),

    i.e.,

    Dtut22κCαDt(Dαa,xu,Dαx,bu)=2Cα[(J1+β0,tDαa,xu,Dαx,but)+(J1+β0,tDαx,bu,Dαa,xut)]+2(f,ut).

    By integration in t and using Lemma 2.2, we get

    ut2+u2αC(u12+u02α+t0(J1+β0,tDαa,xu,Dαx,but)dt+t0(J1+β0,tDαx,bu,Dαa,xut)dt+t0futdt), (3.5)

    where the initial values u0 and u1 are defined in (1.5). Then integrating by parts obtains

    t0(J1+β0,tDαa,xu,Dαx,but)dt=(J1+β0,tDαa,xu,Dαx,bu)t0(Dαx,bu,Jβ0,tDαa,xu)dt, (3.6)
    t0(J1+β0,tDαx,bu,Dαa,xut)dt=(J1+β0,tDαx,bu,Dαa,xu)t0(Dαa,xu,Jβ0,tDαx,bu)dt. (3.7)

    For the first term of the right hand side of (3.6), we have

    (J1+β0,tDαa,xu,Dαx,bu)ϵDαx,bu2+C(ϵ)J1+β0,tDαa,xu2)ϵu2α+C(ϵ)J1+β0,tu2α. (3.8)

    And for the second one, we get

    t0(Dαx,bu,Jβ0,tDαa,xu)dtDαx,bu(¯t)t0Jβ0,tDαa,xudt=Dαx,bu(¯t)J1+β0,tDαa,xuϵu2α+C(ϵ)J1+β0,tu2α. (3.9)

    For the discussion of the right hand side of (3.7), we can prove it similar to (3.8) and (3.9), which is omitted here. For the last term of the right hand side of (3.5), we get

    t0futdtϵut(¯t)2+C(t0fdt)2,

    where

    Dαx,bu(¯t)=sup0stDαx,bu(s),ut(¯t)=sup0stut(s),ut(¯t)α=sup0stut(s)α.

    Hence, we obtain

    ut+uαC(u1+u0α+J1+β0,tu(t)α+t0fdt).

    Then employing Grönwall's inequality (Lemma 3.1) gets that

    ut+uαC(u1+u0α+t0fdt).

    Thus we finish the proof of Theorem 3.1.

    Remark 1. The existence and uniqueness of the weak solution for (3.1) can be derived from Theorem 3.1. The higher order regularities of the weak solution can be guaranteed only by higher differentiabilities of the data and compatibilities, which are omitted here.

    In this section, we will give the error estimate for the following semi-discretization

    (uh,tt,v)κCα[(Dαa,xuh,Dαx,bv)+(Dαx,buh,Dαa,xv)]=Cα[(J1+β0,tDαa,xuh,Dαx,bv)+(J1+β0,tDαx,buh,Dαa,xv)]+(f,v) (4.1)

    for vShr, and Shr={vHα0(Ω)C(Ω),v|rPr1(K)} is the finite element subspace, in which h is the stepsize in space variable x, and Pr1(K) is the r1 degree polynomial space on the subinterval K of Ω.

    In order to carry out the work of error estimation, we first define the projection PhuShr satisfying

    (Phuu,v)=0,vShr. (4.2)

    Obviously, we have that

    PhuuCu.

    Then we define Rhu as the Ritz projection of function uHα0(Ω)Hr(Ω) satisfying

    (Dαa,x(Rhuu),Dαx,bv)+(Dαx,b(Rhuu),Dαa,xv)=0,vShr. (4.3)

    The optimal error estimates about the Phu and Rhu are very useful for our later discussion.

    Lemma 4.1. ([19]) Assume uHα0(Ω)Hr(Ω), then we have that

    uPhu+hαuPhuαChrur, (4.4)
    uRhu+hαuRhuαChrur. (4.5)

    Remark 2. The initial value u0hShr is an approximation of u0, which can take Phu0 or Rhu0, but here it satisfies

    u0hu0αChrαu0r. (4.6)

    Then we define the fractional Ritz-Volterra projection VhuShr satisfying

    κ[(Dαa,x(Vhuu),Dαx,bv)+(Dαx,b(Vhuu),Dαa,xv)]=(J1+β0,tDαa,x(Vhuu),Dαx,bv)+(J1+β0,tDαx,b(Vhuu),Dαa,xv),vShr. (4.7)

    Here we will use the fractional Ritz-Volterra projection Vh and give some useful lemmas to study the error estimate.

    Lemma 4.2. ([9]) Denote by ρ=Vhuu, then we have

    ρ+hαραChr(u0r+J1+β0,tur), (4.8)
    t0(ρt+hαρtα)dsChr(u0r+t0utrds). (4.9)

    The following energy estimation of ρtt is novel and crucial to our convergence analysis.

    Lemma 4.3. For the operator ρtt, we have

    t0(ρtt+hαρttα)dsChr(tβΓ(1+β)u0r+tβ+1Γ(2+β)u1r+t0uttrds).

    Proof. By using the convolution property of Riemann-Liouville integral, we can rewrite (4.7) as follows

    κ[(Dαa,xρ,Dαx,bv)+(Dαx,bρ,Dαa,xv)]=1Γ(1+β)(t0sβDαa,xρ(ts)ds,Dαx,bv)+1Γ(1+β)(t0sβDαx,bρ(ts)ds,Dαa,xv).

    By differentiating with respect to t of both sides of above equation, it becomes

    κ[(Dαa,xρt,Dαx,bv)+(Dαx,bρt,Dαa,xv)]=tβΓ(1+β)[(Dαa,xρ(0),Dαx,bv)+(Dαx,bρ(0),Dαa,xv)]+1Γ(1+β)(t0sβDαa,xρt(ts)ds,Dαx,bv)+1Γ(1+β)(t0sβDαx,bρt(ts)ds,Dαa,xv)=tβΓ(1+β)[(Dαa,xρ(0),Dαx,bv)+(Dαx,bρ(0),Dαa,xv)]+(J1+β0,tDαa,xρt,Dαx,bv)+(J1+β0,tDαx,bρt,Dαa,xv).

    Then by differentiating with respect to t again,

    κ[(Dαa,xρtt,Dαx,bv)+(Dαx,bρtt,Dαa,xv)]=tβ1Γ(β)[(Dαa,xρ(0),Dαx,bv)+(Dαx,bρ(0),Dαa,xv)]+2tβΓ(1+β)[(Dαa,xρt(0),Dαx,bv)+(Dαx,bρt(0),Dαa,xv)]+(J1+β0,tDαa,xρtt,Dαx,bv)+(J1+β0,tDαx,bρtt,Dαa,xv). (4.10)

    Taking v=VhuttRhutt and using the definition of Ritz projection (4.3) obtain that

    CVhuttRhutt2ακ[(Dαa,x(VhuttRhutt),Dαx,b(VhuttRhutt))+(Dαx,b(VhuttRhutt),Dαa,x(VhuttRhutt))]=κ[(Dαa,xρtt,Dαx,b(VhuttRhutt))+(Dαx,bρtt,Dαa,x(VhuttRhutt))]=tβ1Γ(β)[(Dαa,xρ(0),Dαx,b(VhuttRhutt))+(Dαx,bρ(0),Dαa,x(VhuttRhutt))]+2tβΓ(1+β)[(Dαa,xρt(0),Dαx,b(VhuttRhutt))+(Dαx,bρt(0),Dαa,x(VhuttRhutt))]+(J1+β0,tDαa,xρtt,Dαx,b(VhuttRhutt))+(J1+β0,tDαx,bρtt,Dαa,x(VhuttRhutt)).

    Hence,

    VhuttRhuttαC(tβ1Γ(β)ρ(0)α+tβΓ(1+β)ρt(0)α+J1+β0,tρtt(t)α).

    Therefore

    ρttαVhuttRhuttα+RhuttuttαC(tβ1Γ(β)ρ(0)α+tβΓ(1+β)ρt(0)α+J1+β0,tρtt(t)α)+RhuttuttαChrα(tβ1Γ(β)u0r+tβΓ(1+β)u1r+utt(t)r)+CJ1+β0,tρtt(t)α.

    Then by using the Grönwall's inequality, one has

    ρttαChrα(tβ1Γ(β)u0r+tβΓ(1+β)u1r+uttr),

    and finally

    t0ρtt(s)αdsChrα(tβΓ(1+β)u0r+tβ+1Γ(2+β)u1r+t0uttrds).

    Consider now the L2estimate. According to the characteristics of Riesz derivatives, about the left and right Riemann-Liouville derivatives, and also using Eq (4.10), we get that

    (ρtt,y)=κCα[(Dαa,xρtt,Dαx,bω)+(Dαx,bρtt,Dαa,xω)]=κCα[(Dαa,xρtt,Dαx,b(ωv))+(Dαx,bρtt,Dαa,x(ωv))]κCα[(Dαa,xρtt,Dαx,bv)+(Dαx,bρtt,Dαa,xv)]=κCα[(Dαa,xρtt,Dαx,b(ωv))+(Dαx,bρtt,Dαa,x(ωv))]+tβ1Γ(β)[(Dαa,xρ(0),Dαx,bv)+(Dαx,bρ(0),Dαa,xv)]+2tβΓ(1+β)[(Dαa,xρt(0),Dαx,bv)+(Dαx,bρt(0),Dαa,xv)]+(J1+β0,tDαa,xρtt,Dαx,bv)+(J1+β0,tDαx,bρtt,Dαa,xv)=κCα[(Dαa,xρtt,Dαx,b(ωv))+(Dαx,bρtt,Dαa,x(ωv))]+tβ1Γ(β)[(Dαa,xρ(0),Dαx,bv)+(Dαx,bρ(0),Dαa,xv)]+2tβΓ(1+β)[(Dαa,xρt(0),Dαx,bv)+(Dαx,bρt(0),Dαa,xv)]+(J1+β0,tρtt,D2αx,bv)+(J1+β0,tρtt,D2αa,xv),

    where ω satisfies 2α|x|2αω=y in Hα0(Ω), and ω2αC. Choosing v=Rhω, it becomes

    ρtt(t)Chα(ρtt(t)α+J1+β0,tρtt(t)α)+Chr(tβ1Γ(β)u0r+tβΓ(1+β)u1r)+CJ1+β0,tρtt(t),

    in which Rhωωαhαω2α is used. By using the Grönwall's inequality again (Lemma 3.1), one has

    ρtt(t)Chαρtt(t)α+Chr(tβ1Γ(β)u0r+tβΓ(1+β)u1r).

    After integration, we get

    t0ρtt(s)dsChαt0ρtt(s)αds+Chr(tβΓ(1+β)u0r+tβ+1Γ(2+β)u1r)Chr(tβΓ(1+β)u0r+tβ+1Γ(2+β)u1r+t0utt(s)rds).

    Therefore,

    t0(ρtt+hαρttα)dsChr(tβΓ(1+β)u0r+tβ+1Γ(2+β)u1r+t0uttrds).

    Thus the proof is completed.

    Lemma 4.4. ([5]) For each ϵ there is a constant Cϵ=Cϵ(t) such that

    |T0f(t)J1+β0,tf(t)dt|ϵT0f2(t)dt+CϵJ1+β0,TT0f2(σ)dσ.

    Theorem 4.1. Assume that the initial values u0h=Rhu0 and u1hu1Chru1r. Let u(t)Hα0(Ω)Hr(Ω) solve (3.1), and uh(t)Shr solve (4.1), then there exists a constant C satisfies

    uh(t)u(t)Chr(tβΓ(β+1)u0r+tβ+1Γ(β+2)u1r+t0utt(s)rds). (4.11)

    Proof. By using Vhu(t)Shr as an intermediate function, the error can be defined as ε=uhu=(uhVhu)+(Vhuu)=θ+ρ. Therefore, the error equation can be rewritten as

    (θtt,v)κCα[(Dαa,xθ,Dαx,bv)+(Dαx,bθ,Dαa,xv)]=Cα[(J1+β0,tDαa,xθ,Dαx,bv)+(J1+β0,tDαx,bθ,Dαa,xv)](ρtt,v),vShr,

    where θ(0)=0, θt(0)=(Vhu)t(0)u1h.

    Taking v=θt, it becomes

    (θtt,θt)κCα[(Dαa,xθ,Dαx,bθt)+(Dαx,bθ,Dαa,xθt)]=Cα[(J1+β0,tDαa,xθ,Dαx,bθt)+(J1+β0,tDαx,bθ,Dαa,xθt)](ρtt,θt),

    i.e.

    Dtθt2κCαDt(Dαa,xθ,Dαx,bθ)=CαDt[(J1+β0,tDαa,xθ,Dαx,bθ)+(J1+β0,tDαx,bθ,Dαa,xθ)]Cα[(Jβ0,tDαa,xθ,Dαx,bθ)+(Jβ0,tDαx,bθ,Dαa,xθ)](ρtt,θt),

    where Lemma 2.1 is used for the computation of the Riemann-Liouville integral. By integration in t and using Lemma 4.4, we get

    θt2+θ2αC(θt(0)2+θαJ1+β0,tθ(t)α+t0θ(t)αJβ0,tθ(t)αds+t0ρttθtds)ϵ(θ(¯t)2α+θt(¯t)2)+C(ϵ)[θt(0)2+(J1+β0,tθ(t)α)2+(t0ρttds)2],

    where

    t0ρttθtdsϵθt(¯t)2+C(ϵ)(t0ρttds)2,
    t0θ(t)αJβ0,tθ(t)αdsϵθ(¯t)2α+C(ϵ)(t0Jβ0,tθ(t)αds)2=ϵθ(¯t)2α+C(ϵ)(J1+β0,tθ(t)α)2,
    θt(¯t)=sup0stθt(s),θt(¯t)α=sup0stθt(s)α.

    Hence, we obtain

    θt+θαC(θt(0)+J1+β0,tθ(t)α+t0ρttds).

    Then we get

    θt+θαC(θt(0)+t0ρttdt),

    in which the Grönwall's inequality is used. By using Lemma 4.3, we have

    θt+θαChr(tβΓ(β+1)u0r+tβ+1Γ(β+2)u1r+t0utt(s)rds),

    where

    θt(0)=(Vhu)t(0)u1+(u1u1h)Chr(tβΓ(β+1)u0r+tβ+1Γ(β+2)u1r)

    is used. This finishes the error estimates of θ. Then combining Lemma 4.1 for error estimates of ρ, the proof of Theorem 4.1 is completed.

    We now turn to discuss the fully discrete scheme based on a symmetric difference approximation. Let fn=f(tn), tn=nk,n=1,2,,N, where k=T/N is the steplength in time variable t. Denote by Un the approximation solution and by Un=(Un+1Un)/k,¯Un=(UnUn1)/k the forward and backward difference quotient of Un respectively, then

    ¯Un=(UnUn1)/k=(Un+12Un+Un1)/k2

    is the center difference quotient of second order to approximate the second time derivative term utt. We also denote the average

    ˆUj=(Uj+12+Uj12)/2=(Uj+1+2Uj+Uj1)/4attj=jk.

    There are several ways to approximate the fractional integral. Here we select the product trapezoidal technique to approximate the fractional integral J1+β0,tng(t),1nN, under the condition g(t)C2([0,T]). The following is the truncate error estimation.

    Lemma 5.1. ([30,31,32]) Suppose u(t)C2([0,T]), then we have

    J1+β0,tng(t)=k1+βΓ(3+β)nj=0bjg(tnj)+O(k2),

    where

    b0=1,bj=(j+1)2+β2j2+β+(j1)2+β,j=1,2,,n1,bn=n1+β(2+βn)+(n1)2+β.

    The discrete Grönwall's inequality is very important to our stability analysis and the convergence analysis.

    Lemma 5.2. (Discrete Grönwall's inequality) ([33,34,35,36]) Assume that ωn0,fn0 and that for n=0,1,,yn0 satisfies

    ynfn+n1j=0ωjyj,

    then we have

    yNexp(N1i=n+1ωi)max0nNfn.

    Next we denote by Ihu the piecewise polynomial interpolation operator of u in Shr satisfying

    Ihu(tn)=u(tn),n=0,1,,N.

    Lemma 5.3. ([19]) Assume uHα0(Ω)Hr(Ω), then we have that

    RhuIhu+hαRhuIhuαChrur.

    Now we give the fully discrete scheme

    (¯Un,v)κCα[(Dαa,xˆU,Dαx,bv)+(Dαx,bˆU,Dαa,xv)]=σnCα[(Dαa,xˆU,Dαx,bv)+(Dαx,bˆU,Dαa,xv)]+(fn,v),vShr, (5.1)

    with given initial values U0 and U1Shr, and σn(g)=k1+βΓ(3+β)nj=0bnjg(tj). In fact, the above scheme is equivalent to the following form

    (Un+12Un+Un12,v)κCα[(Dαa,xUn+1/2+Un1/22,Dαx,bv)+(Dαx,bUn+1/2+Un1/22,Dαa,xv)]=Cαk1+βΓ(3+β)nj=0bnj[(Dαa,xUj+1/2+Uj1/22,Dαx,bv)+(Dαx,bUj+1/2+Uj1/22,Dαa,xv)]+(fn,v), (5.2)

    where U1/2 can be approximated by U0 and U1, and the second-order accuracy should be guaranteed. Furthermore, we can move the terms Un+1 in the right hand side of (5.2) to the left hand side for explicit processing.

    Then we discuss the stability analysis of the fully discrete scheme (5.1) in the following form.

    Theorem 5.1. The solution of (5.1) satisfies the following stability conclusion

    Un+Un+1/2αCU0+CU1/2α+Cknm=1fm, (5.3)

    for n1,tn+1T.

    Proof. Choosing v=¯Un+1/2 in (5.1) to obtain

    (¯Un,¯Un+1/2)κCα[(Dαa,xˆU,Dαx,b¯Un+1/2)+(Dαx,bˆU,Dαa,x¯Un+1/2)]=Cασn[(Dαa,xˆU,Dαx,b¯Un+1/2)+(Dαx,bˆU,Dαa,x¯Un+1/2)]+(fn,¯Un+1/2):=In1+In2. (5.4)

    Note that

    (¯Un,¯Un+1/2)=(UnUn1,Un+Un1)/2k=12¯Un2,
    (Dαa,xˆU,Dαx,b¯Un+1/2)+(Dαx,bˆU,Dαa,x¯Un+1/2)=12¯[(Dαa,xUn+1/2,Dαx,bUn+1/2)+(Dαx,bUn+1/2,Dαa,xUn+1/2)].

    Multiplying both sides of (5.4) by k and summing from n=1 to N obtain

    UN2+cUN+1/22αU02+CU1/22α+Ck|Nn=1(In1+In2)|.

    For the term k|Nn=1In1|, we have

    k|Nn=1In1|=Cαk|Nn=1σn[(Dαa,xˆUn,Dαx,b¯Un+1/2)+(Dαx,bˆUn,Dαa,x¯Un+1/2)]|=Cαk1+β2Γ(3+β){|Nn=1n1j=0bnj[(Dαa,x(Uj+1/2+Uj1/2),Dαx,b(Un+1/2Un1/2))+(Dαx,b(Uj+1/2+Uj1/2),Dαa,x(Un+1/2Un1/2))]|+|Nn=1[(Dαa,x(Un+1/2+Un1/2),Dαx,b(Un+1/2Un1/2))+(Dαx,b(Un+1/2+Un1/2),Dαa,x(Un+1/2Un1/2))]|}Cαk1+β2Γ(3+β){N1j=0bNjNn=j+1[|(Dαa,x(Uj+1/2+Uj1/2),Dαx,b(Un+1/2Un1/2))|+|(Dαx,b(Uj+1/2+Uj1/2),Dαa,x(Un+1/2Un1/2))|]+Nmax1nNUn+1/22α}C(k1+βN1j=0bNjUj+1/2α+kβmax1nNUn+1/2α)max1nNUn+1/2α, (5.5)

    in which bnj=(nj+1)2+β2(nj)2+β+(nj1)2+β is monotonically increasing for n=1,,N. And for the term k|Nn=1In2|, we have

    k|Nn=1In2|CkNn=1fnmax1nN¯Un.

    After some adjustments and applying the discrete Grönwall's inequality (Lemma 5.2) obtain

    UN+UN+1/2αCU0+CU1/2α+CkNn=1fn.

    Then we finish the proof of the theorem.

    Next we discuss the error estimate of the fully discrete scheme. Denote by the error Unun=(UnVhun)+(Vhunun)=θn+ρn, we have

    (¯θn,v)κCα[(Dαa,xˆθn,Dαx,bv)+(Dαx,bˆθn,Dαa,xv)]=Cασn[(Dαa,xˆθn,Dαx,bv)+(Dαx,bˆθn,Dαa,xv)](en,v),vShr, (5.6)

    where θ0=u0U0,θ1=u1U1, and

    en=en1+en2+en3+en4,en1=¯ρn,en2=¯u(tn)utt(tn),en3=2α|x|2α(Vh^unVhun),en4=σn(2α|x|2αVhˆu)J1+β0,tn2α|x|2αVhu(t).

    To construct the discrete initial values U0 and U1, let u2=utt(0)=f(0)2α|x|2αu0 and define U1=V0+kV1+V2k2/2, where U0=V0,Vj=Phuj,j=0,1,2. According to Lemma 4.1, we get the following conclusions

    V0u0+kV1u1αChrα,V2C,
    V1u1+kV2u2Chrα.

    Then for θn=UnVhun, there is

    θ0+θ1/2αC(u)(hrα+k2). (5.7)

    In the following, we give the error estimation of the fully discrete scheme (5.1).

    Theorem 5.2. The solution U of (5.1) and the solution u of (1.4) at tn+1/2 satisfy the following conclusion

    Un+1/2u(tn+1/2)αC(u)(hrα+k2). (5.8)

    Proof. The proof is similar to Theorem 5.1. Taking v=¯θn+1/2 in (5.6), we have

    (¯θn,¯θn+1/2)κCα[(Dαa,xˆθn,Dαx,b¯θn+1/2)+(Dαx,bˆθn,Dαa,x¯θn+1/2)]=Cασn[(Dαa,xˆθn,Dαx,b¯θn+1/2)+(Dαx,bˆθn,Dαa,x¯θn+1/2)](en,¯θn+1/2), (5.9)

    where

    (¯θn,¯θn+1/2)=(θnθn1,θn+θn1)/2k=12¯θn2,
    (Dαa,xˆθn,Dαx,b¯θn+1/2)+(Dαx,bˆθn,Dαa,x¯θn+1/2)=12¯[(Dαa,xθn+1/2,Dαx,bθn+1/2)+(Dαx,bθn+1/2,Dαa,xθn+1/2)].

    For the first term of the right sides of (5.9), which is similar to (5.5), we obtain

    Cαk|Nn=1σn[(Dαa,xˆθn,Dαx,b¯θn+1/2)+(Dαx,bˆθn,Dαa,x¯θn+1/2)]|C(k1+βN1j=0bNjθj+1/2α+kβmax1nNθn+1/2α)max1nNθn+1/2α.

    Summing (5.9) from n=1 to N obtains that

    θN2+cθN+1/22αθ02+Cθ1/22α+Ck1+βN1j=0bNjθj+1/2αmax1nNθn+1/2α+CkNj=1ejmax1nN¯θn+1/2.

    Then applying Lemma 5.2, we get

    θN+θN+1/2αCθ0+Cθ1/2α+CkNn=1en. (5.10)

    By the Taylor expansion, we know that

    un+1un=u(tn)k+tn+1tnu(s)(tn+1s)ds=u(tn)k+u(tn)k2/2+u(tn)k3/6+1/6tn+1tnu(4)(s)(tn+1s)3ds.

    Therefore

    ken1=k¯ρntn+1tn1ρttdsChrtn+1tn1uttrds,
    ken2=k¯u(tn)utt(tn)Ck2tn+1tn1u(4)(s)ds,
    ken3=k2α|x|2α(Vh^unVhun)k2tn+1tn12α|x|2αVhussdsCk2tn+1tn1uss2αds,

    and

    ken4k2tn02α|x|2αVhussdsCk2tn0utt2αds.

    Thus

    kNn=1enhrtn0uttrds+Ck2tN+10(u(4)+utt2α)ds.

    Combined with (5.7) and (5.10), we finish the proof of Theorem 5.2

    Remark 3. The above stability analysis and the convergence analysis of the fully discrete schemes can be extended to high-dimensional cases without difficulty, which are omitted here.

    In order to test the effectiveness of the designed numerical algorithm, we present the following numerical experiments in this section.

    In the Galerkin finite element approximation, we select the hat function as the shape function, followed by the symmetric difference scheme and fractional trapezoidal formula for the time stepping. Through the theoretical analysis of the previous sections, the expected goal of the convergence order with L2 norm should be O(k2+h2).

    Example 6.1. In this example, we study the following fractional hyperbolic equation

    utt2αu|x|2α=J1+β0,t2αu|x|2α+f(x,t), (6.1)

    with homogeneous boundary conditions. And the corresponding parameters satisfy Ω=[1,1],T=1,α=0.9,β=0.1.

    Case Ⅰ. We choose the source term f=2(1x)(1+x)(t2+2t3+βΓ(4+β))2α|x|2α(1x)(1+x), then the exact solution is u(x,t)=t2(1x)(1+x), which determines that the initial values u0=u1=0.

    Table 1 shows the numerical results and convergence rates of case Ⅰ, which supports the predicted rates of the convergence. Figure 1 shows the exact solution and the numerical solution of case Ⅰ with α=0.9,β=0.1, h=1/64,k=0.01h at t=0.5,x[1,1]. Figure 2 shows the error between the exact solution and the numerical solution of case Ⅰ with α=0.9,β=0.1, h=1/64,k=0.01h, at x=0,t[0,1]. And Figure 3 shows the numerical solution of case Ⅰ under the same conditions.

    Table 1.  Experimental error results of case Ⅰ in Example 6.1.
    h uNuNh cv.rate
    14 1.4686102 -
    18 3.4777103 2.0783
    116 8.1008104 2.1020
    132 1.8438104 2.1354
    164 4.8037105 1.9405

     | Show Table
    DownLoad: CSV
    Figure 1.  The exact solution and the numerical solution of case Ⅰ in Example 6.1 with α=0.9,β=0.1, h=1/64, t=0.5,x[1,1].
    Figure 2.  Error between the exact solution and the numerical solution of case Ⅰ in Example 6.1 with α=0.9,β=0.1, h=1/64 for x[1,1],t[0,1].
    Figure 3.  The numerical solution of case Ⅰ in Example 6.1 with α=0.9,β=0.1, h=1/64 for x[1,1],t[0,1].

    Case Ⅱ. We choose the exact solution u(x,t)=sin(πt)(1x)(1+x). And the source term f is obtained numerically by using the fractional trapezoidal formula. Then the initial values satisfy u0=0,u1=π(1x)(1+x).

    Table 2 shows the numerical results and convergence rates of case Ⅱ, which support the predicted rates of the convergence. Figure 4 shows the exact solution and the numerical solution of case Ⅱ with α=0.9,β=0.1, h=1/64,k=0.01h, at t=0.5,x[1,1]. Figures 5 and 6 show the error between the exact solution and the numerical solution, the numerical solution of case Ⅱ with α=0.9,β=0.1, h=1/64,k=0.01h for x[1,1],t[0,1] separately.

    Table 2.  Experimental error results of case Ⅱ in Example 6.1.
    h uNuNh cv.rate
    14 5.13728102 -
    18 1.3758102 1.9007
    116 4.0785103 1.7542
    132 1.3590103 1.5855
    164 5.1245104 1.4071

     | Show Table
    DownLoad: CSV
    Figure 4.  The exact solution and the numerical solution of case Ⅱ in Example 6.1 with α=0.9,β=0.1, h=1/64, at t=0.5,x[1,1].
    Figure 5.  The error between the exact solution and the numerical solution of case Ⅱ in Example 6.1 with α=0.9,β=0.1, h=1/64 for x[1,1],t[0,1].
    Figure 6.  The numerical solution of case Ⅱ in Example 6.1 with α=0.9,β=0.1, h=1/64 for x[1,1],t[0,1].

    Example 6.2. In this example, we consider the following fractional hyperbolic equation

    uttD2αa,xu=J1+β0,tD2αa,xu+f(x,t), (6.2)

    with homogeneous Dirichlet boundary conditions in Ω=[0,1],T=1.

    We choose the source term f=2(1x)xα(t2+2t3+βΓ(4+β))D2αa,x(1x)xα, then the exact solution is u(x,t)=t2(1x)xα, which has a weak singularity at the boundary point x=0 if 0.5<α<1.

    Table 3 shows the errors and convergence rates with parameters α=0.6,β=0.1, h=1/64,k=0.01h for x[0,1],t[0,1]. Figures 7 and 8 show the numerical solution and the absolute error between the exact solution and the numerical solution of Example 6.2 with α=0.9,β=0.1, h=1/64 for x[0,1],t[0,1] separately. And Figure 9 shows the numerical solution with different values of α at time t=0.5. From Figures 8 and 9, we can see that the numerical solution is basically coincided with the exact solution. Note that the selected exact solution has a weak singularity at the boundary point x=0, therefore the scheme does not work very well near zero.

    Table 3.  Experimental error results of Example 6.2 with α=0.6,β=0.1.
    h uNuNh cv.rate
    14 2.6034E002 -
    18 1.7128E002 0.6041
    116 9.3203E003 0.8779
    132 4.6566E003 1.0011
    164 2.2366E003 1.0580

     | Show Table
    DownLoad: CSV
    Figure 7.  The numerical solution of Example 6.2 with α=0.9,β=0.1, h=1/64 for x[0,1],t[0,1].
    Figure 8.  The absolute error between the exact solution and the numerical solution of Example 6.2 with α=0.9,β=0.1, h=1/64 for x[0,1],t[0,1].
    Figure 9.  The numerical solution and the exact solution of Example 6.2 with α=0.6,0.7,0.8,0.9 separately and β=0.1, h=1/64, t=0.5,x[0,1].

    In this paper, we use the Galerkin finite element method and the symmetric difference method to solve the fractional hyperbolic integro-differential equation, where the space fractional derivative is in Riesz sense and the integro-differential term is compounded of the Riesz space fractional derivative and the Riemann-Liouville time fractional integral. We apply the fractional trapezoidal formula to treat the fractional integral and employ enough points to ensure the convergence order. Numerical examples are presented to test the effectiveness of the convergence analysis. From the numerical results, we can see that the designed numerical algorithm performs well and the convergence orders conform to the convergence analysis.

    As is known to all, fractional calculus has weak singularity and nonlocality from its origin [37]. Not only the fractional differential equation, but also the fractional integro-differential equation, their solutions both behave the weak singularities. In this paper, we design a solution with a weak singularity at the boundary point x=0, which is verified by numerical experiments. Meanwhile, because of its nonlocality, although the above theoretical analyses can be extended to the high-dimensional cases without difficulty, the capacities of computation and memory will become large. So how to reduce the computationally expensive and the storage requirement comes into being the main problem. Maybe the fast algorithm is a good choice. In future, we will continue to study these problems.

    This work was sponsored by Natural Science Foundation of Shanghai under grant No. 19ZR1422000, Natural Science Foundation of Anhui Province under grant No. 2008085MA11, and Natural Science Foundation of Education Department of Anhui Province under grant No. KJ2018A0385.

    The authors wish to thank the anonymous reviewers for their comments and proposals, which greatly improve the quality of the paper. The authors also wish to thank Prof. Changpin Li for his helpful discussions and suggestions.

    All authors declare no conflicts of interest in this paper.



    [1] R. M. Christensen, Theory of viscoelasticity, J. Appl. Mech., 38 (1971), 720. http://dx.doi.org/10.1115/1.3408900 doi: 10.1115/1.3408900
    [2] M. E. Gurtin, A. C. Pipkin, A general theory of heat conduction with finite wave speed, Arch. Rational Mech. Anal., 31 (1968), 113–126. http://dx.doi.org/10.1007/BF00281373 doi: 10.1007/BF00281373
    [3] R. K. Miller, An integro-differential equation for grid heat conductions with memory, J. Math. Anal. Appl., 66 (1978), 313–332. http://dx.doi.org/10.1016/0022-247x(78)90234-2 doi: 10.1016/0022-247x(78)90234-2
    [4] M. Renardy, Mathematical analysis of viscoelastic flows, Ann. Rev. Fluid Mech., 21 (1989), 21–36. http://dx.doi.org/10.1146/annurev.fl.21.010189.000321 doi: 10.1146/annurev.fl.21.010189.000321
    [5] C. M. Chen, S. Tsimin, Finite element methods for integrodifferential equations, Word Scientific, Singapore, 1998. http://dx.doi.org/10.1142/3594
    [6] M. M. Meerschaert, F. Sabzikar, Tempered fractional Brownian motion, Stat. Probabil. Lett., 83 (2013), 2269–2275. http://dx.doi.org/10.1016/j.spl.2013.06.016 doi: 10.1016/j.spl.2013.06.016
    [7] E. W. Montroll, G. H. Weiss, Random walks on lattices. Ⅱ, J. Math. Phys., 6 (1965), 167–181. http://dx.doi.org/10.1063/1.1704269 doi: 10.1063/1.1704269
    [8] I. Podlubny, Fractional differential equations, Academic Press, San Diego, CA, 1998.
    [9] Z. G. Zhao, Y. Y. Zheng, P. Guo, A Galerkin finite element method for a class of time-space fractional diffusion quation with nonsmooth data, J. Sci. Comput., 70 (2017), 386–406. http://dx.doi.org/10.1007/s10915-015-0107-3 doi: 10.1007/s10915-015-0107-3
    [10] I. Dassios, F. Font, Solution method for the time-fractional hyperbolic heat equation, Math. Meth. Appl. Sci., 44 (2021), 11844–11855. http://dx.doi.org/10.1002/mma.6506 doi: 10.1002/mma.6506
    [11] P. Kumar, K. N. Rai, Fractional modeling of hyperbolic bioheat transfer equation during thermal therapy, J. Mech. Medi. Biol., 17 (2017), 1–19. http://dx.doi.org/10.1142/S0219519417500580 doi: 10.1142/S0219519417500580
    [12] A. Ashyralyev, F. Dal, Z. Pinar, A note on the fractional hyperbolic differential and difference equations, Appl. Math. Comput., 217 (2011), 4654–4664. http://dx.doi.org/10.1016/j.amc.2010.11.017 doi: 10.1016/j.amc.2010.11.017
    [13] W. Qiu, D. Xu, H. B. Chen, A formally second-order BDF finite difference scheme for the integro-differential equations with the multi-term kernels, Int. J. Comput. Math., 97 (2020), 2055–2073. https://doi.org/10.1080/00207160.2019.1677896 doi: 10.1080/00207160.2019.1677896
    [14] W. Qiu, D. Xu, J. Guo, A formally second-order backward differentiation formula Sinc-collocation method for the Volterra integro-differential equation with a weakly singular kernel based on the double exponential transformation, Numer. Methods Partial Differ. Equ., 38 (2022), 830–847. http://dx.doi.org/10.1002/num.22703 doi: 10.1002/num.22703
    [15] V. J. Ervin, J. P. Roop, Variational formulation for the stationary fractional advection dispersion equation, Numer. Methods Partial Differ. Equ., 22 (2006), 558–576. http://dx.doi.org/10.1002/num.20112 doi: 10.1002/num.20112
    [16] Z. Sun, X. Wu, A fully discrete difference scheme for a diffusion-wave system, Appl. Numer. Math., 56 (2006), 193–209. http://dx.doi.org/10.1016/j.apnum.2005.03.003 doi: 10.1016/j.apnum.2005.03.003
    [17] V. J. Ervin, N. Heuer, J. P. Roop, Numerical approximation of a time dependent nonlinear, space-fractional diffusion equation, SIAM J. Numer. Anal., 45 (2007), 572–591. http://dx.doi.org/10.1137/050642757 doi: 10.1137/050642757
    [18] W. H. Deng, Finite element method for the space and time fractional Fokker-Planck equation, SIAM J. Numer. Anal., 47 (2008), 204–226. http://dx.doi.org/10.1137/080714130 doi: 10.1137/080714130
    [19] Y. Y. Zheng, C. P. Li, Z. G. Zhao, A note on the finite element method for the space-fractional advection diffusion equation, Comput. Math. Appl., 59 (2001), 1718–1726. http://dx.doi.org/10.1016/j.camwa.2009.08.071 doi: 10.1016/j.camwa.2009.08.071
    [20] C. P. Li, Z. G. Zhao, Y. Q. Chen, Numerical approximation of nonlinear fractional differential equations with subdiffusion and superdiffusion, Comput. Math. Appl., 62 (2011), 855–875. http://dx.doi.org/10.1016/j.camwa.2011.02.045 doi: 10.1016/j.camwa.2011.02.045
    [21] F. H. Zeng, F. W. Liu, C. P. Li, K. Burrage, I. Turner, V. Anh, A crank–nicolson ADI spectral method for a two-dimensional Riesz space fractional nonlinear reaction-diffusion equation, SIAM J. Numer. Anal., 52 (2014), 2599–2622. http://dx.doi.org/10.1137/130934192 doi: 10.1137/130934192
    [22] W. R. Cao, F. H. Zeng, Z. Q. Zhang, G. E. Karniadakis, Implicit-explicit difference schemes for nonlinear fractional differential equations with nonsmooth solutions, SIAM J. Sci. Comput., 38 (2016), A3070–A3093. http://dx.doi.org/10.1137/16M1070323 doi: 10.1137/16M1070323
    [23] Z. G. Zhao, Y. Y. Zheng, P. Guo, A Galerkin finite element scheme for time-space fractional diffusion equation, Int. J. Comput. Math., 93 (2016), 1212–1225. http://dx.doi.org/10.1080/00207160.2015.1044986 doi: 10.1080/00207160.2015.1044986
    [24] Y. M. Liu, Y. B. Yan, M. Khan, Discontinuous Galerkin time stepping method for solving linear space fractional partial differential equations, Appl. Numer. Math., 115 (2017), 200–213. http://dx.doi.org/10.1016/j.apnum.2017.01.009 doi: 10.1016/j.apnum.2017.01.009
    [25] G. A. Zou, A. Atangana, Y. Zhou, Error estimates of a semidiscrete finite element method for fractional stochastic diffusion-wave equations, Numer. Methods Partial Differ. Equ., 34 (2018), 1834–1848. http://dx.doi.org/10.1002/num.22252 doi: 10.1002/num.22252
    [26] Z. J. Zhang, W. H. Deng, G. E. Karniadakis, A Riesz basis Galerkin method for the tempered fractional Laplacian, SIAM J. Numer. Anal., 56 (2018), 3010–3039. http://dx.doi.org/10.1137/17M1151791 doi: 10.1137/17M1151791
    [27] D. Y. Shi, H. J. Yang, Superconvergence analysis of finite element method for time-fractional Thermistor problem, Appl. Math. Comput., 323 (2018), 31–42. http://dx.doi.org/10.1016/j.amc.2017.11.027 doi: 10.1016/j.amc.2017.11.027
    [28] S. G. Samko, A. A. Kilbas, O. I. Maxitchev, Integrals and derivatives of the fractional order and some of their applications, (in Russian), Nauka i Tekhnika, Minsk, 1987.
    [29] C. P. Li, F. H. Zeng, Numerical methods for fractional calculus, Chapman and Hall/CRC, 2015. http://dx.doi.org/10.1201/b18503
    [30] K. Diethelm, N. J. Ford, A. D. Freed, Detailed error analysis for a fractional Adams method, Numer. Algorithms, 36 (2004), 31–52. http://dx.doi.org/10.1023/b:numa.0000027736.85078.be doi: 10.1023/b:numa.0000027736.85078.be
    [31] P. Zhuang, F. Liu, V. Anh, I. Turner, New solution and analytical techniques of the implicit numerical method for the anomalous subdiffusion equation, SIAM J. Numer. Anal., 46 (2008), 1079–1095. http://dx.doi.org/10.1137/060673114 doi: 10.1137/060673114
    [32] C. P. Li, F. H. Zeng, The finite difference mehtods for fractional ordinary differential equations, Numer. Func. Anal. Optim., 34 (2013), 149–179. http://dx.doi.org/10.1080/01630563.2012.706673 doi: 10.1080/01630563.2012.706673
    [33] S. Larsson, V. Thomé, L. B. Wahlbin, Numerical solution of parabolic integro-differential equations by the discontinuous Galerkin method, Math. Comput., 67 (1998), 45–71. http://dx.doi.org/10.1090/S0025-5718-98-00883-7 doi: 10.1090/S0025-5718-98-00883-7
    [34] J. T. Ma, Finite element method for partial Volterra integro-diffeential equations on two-dimensions unbounded spatial domains, Appl. Math. Comput., 186 (2007), 598–609. http://dx.doi.org/10.1016/j.amc.2006.08.004 doi: 10.1016/j.amc.2006.08.004
    [35] F. H. Zeng, J. X. Cao, C. P. Li, Grönwall inequalities, In: Recent advances in applied nonlinear dynamics with numerical analysis, World Scientific, Singapore, 2013. http://dx.doi.org/10.1142/9789814436465_0001
    [36] W. L. Qiu, D. Xu, H. F. Chen, J. Guo, An alternating direction implicit Galerkin finite element method for the distributed-order time-fractional mobile-immobile equation in two dimensions, Comput. Math. Appl., 80 (2020), 3156–3172. http://dx.doi.org/10.1016/j.camwa.2020.11.003 doi: 10.1016/j.camwa.2020.11.003
    [37] C. P. Li, M. Cai, Theory and numerical approximations of fractional integrals and derivatives, SIAM, 2019. http://dx.doi.org/10.1137/1.9781611975888
  • Reader Comments
  • © 2022 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(1951) PDF downloads(116) Cited by(0)

Figures and Tables

Figures(9)  /  Tables(3)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog