Two inverse source problems for a space-time fractional differential equation involving bi-fractional Laplacian operators in the spatial variable and Caputo time-fractional derivatives of different orders between 1 and 2 are studied. In the first inverse source problem, the space-dependent term along with the diffusion concentration is recovered, while in the second inverse source problem, the time-dependent term along with the diffusion concentration is identified. Both inverse source problems are ill-posed in the sense of Hadamard. The existence and uniqueness of solutions for both inverse source problems are investigated. Finally, several examples are presented to illustrate the obtained results for the inverse source problems.
Citation: M. J. Huntul. Inverse source problems for multi-parameter space-time fractional differential equations with bi-fractional Laplacian operators[J]. AIMS Mathematics, 2024, 9(11): 32734-32756. doi: 10.3934/math.20241566
[1] | Bauyrzhan Derbissaly, Makhmud Sadybekov . Inverse source problem for multi-term time-fractional diffusion equation with nonlocal boundary conditions. AIMS Mathematics, 2024, 9(4): 9969-9988. doi: 10.3934/math.2024488 |
[2] | Wei Fan, Kangqun Zhang . Local well-posedness results for the nonlinear fractional diffusion equation involving a Erdélyi-Kober operator. AIMS Mathematics, 2024, 9(9): 25494-25512. doi: 10.3934/math.20241245 |
[3] | Shuang-Shuang Zhou, Saima Rashid, Asia Rauf, Khadija Tul Kubra, Abdullah M. Alsharif . Initial boundary value problems for a multi-term time fractional diffusion equation with generalized fractional derivatives in time. AIMS Mathematics, 2021, 6(11): 12114-12132. doi: 10.3934/math.2021703 |
[4] | Anumanthappa Ganesh, Swaminathan Deepa, Dumitru Baleanu, Shyam Sundar Santra, Osama Moaaz, Vediyappan Govindan, Rifaqat Ali . Hyers-Ulam-Mittag-Leffler stability of fractional differential equations with two caputo derivative using fractional fourier transform. AIMS Mathematics, 2022, 7(2): 1791-1810. doi: 10.3934/math.2022103 |
[5] | Rana Safdar Ali, Saba Batool, Shahid Mubeen, Asad Ali, Gauhar Rahman, Muhammad Samraiz, Kottakkaran Sooppy Nisar, Roshan Noor Mohamed . On generalized fractional integral operator associated with generalized Bessel-Maitland function. AIMS Mathematics, 2022, 7(2): 3027-3046. doi: 10.3934/math.2022167 |
[6] | Shanhao Yuan, Yanqin Liu, Yibin Xu, Qiuping Li, Chao Guo, Yanfeng Shen . Gradient-enhanced fractional physics-informed neural networks for solving forward and inverse problems of the multiterm time-fractional Burger-type equation. AIMS Mathematics, 2024, 9(10): 27418-27437. doi: 10.3934/math.20241332 |
[7] | Mounir Zili, Eya Zougar, Mohamed Rhaima . Fractional stochastic heat equation with mixed operator and driven by fractional-type noise. AIMS Mathematics, 2024, 9(10): 28970-29000. doi: 10.3934/math.20241406 |
[8] | Muhammad Farman, Ali Akgül, J. Alberto Conejero, Aamir Shehzad, Kottakkaran Sooppy Nisar, Dumitru Baleanu . Analytical study of a Hepatitis B epidemic model using a discrete generalized nonsingular kernel. AIMS Mathematics, 2024, 9(7): 16966-16997. doi: 10.3934/math.2024824 |
[9] | Xuemin Xue, Xiangtuan Xiong, Yuanxiang Zhang . Two fractional regularization methods for identifying the radiogenic source of the Helium production-diffusion equation. AIMS Mathematics, 2021, 6(10): 11425-11448. doi: 10.3934/math.2021662 |
[10] | Veliappan Vijayaraj, Chokkalingam Ravichandran, Thongchai Botmart, Kottakkaran Sooppy Nisar, Kasthurisamy Jothimani . Existence and data dependence results for neutral fractional order integro-differential equations. AIMS Mathematics, 2023, 8(1): 1055-1071. doi: 10.3934/math.2023052 |
Two inverse source problems for a space-time fractional differential equation involving bi-fractional Laplacian operators in the spatial variable and Caputo time-fractional derivatives of different orders between 1 and 2 are studied. In the first inverse source problem, the space-dependent term along with the diffusion concentration is recovered, while in the second inverse source problem, the time-dependent term along with the diffusion concentration is identified. Both inverse source problems are ill-posed in the sense of Hadamard. The existence and uniqueness of solutions for both inverse source problems are investigated. Finally, several examples are presented to illustrate the obtained results for the inverse source problems.
Fractional calculus broadens the scope of traditional calculus by extending the concepts of differentiation and integration to non-integer, arbitrary orders. This mathematical framework is particularly valuable for describing systems that exhibit memory and hereditary properties, transcending the limitations of classical differential and integral calculus. Fractional derivatives, which form the cornerstone of fractional calculus, offer a nuanced approach to capturing the dynamics of processes where the past influences the present, a feature not adequately addressed by integer-order derivatives [1,2,3]. These derivatives are instrumental in formulating fractional differential equations (FDEs), which effectively model complex phenomena across various domains, including engineering, physics, biology, and finance [4,5]. The adoption of FDEs allows for a more accurate representation of behaviors such as viscoelastic material response, anomalous diffusion, and memory effects, underscoring the relevance of fractional calculus in complex system analysis [6,7,8,9,10,11].
This paper addresses two inverse source problems (ISPs) defined for the following space-time PDE:
CDξ00+,tu(x,t)+m−1∑j=1ajCDξj0+,tu(x,t)=−(−Δ)η1/2u(x,t)+−(−Δ)η2/2u(x,t)+F(x,t),(x,t)∈ΩT, | (1.1) |
subject to boundary conditions
u(−1,t)=0=u(1,t),t∈(0,T), | (1.2) |
with non-homogeneous initial conditions
u(x,0)=ρ(x),x∈(−1,1), | (1.3) |
ut(x,0)=ν(x),x∈(−1,1), | (1.4) |
where CDξj0+,t stands for the Caputo fractional derivatives in time variable of order ξj,1<ξm−1<...<ξ1<ξ0<2, and is defined by
CDξj0+,th(t):=J2−ξj0+,td2dt2h(t)=1Γ(2−ξj)t∫0d2dτ2h(τ)(t−τ)ξj−1dτ,t>0, |
where ΩT:=Ω×(0,T), Ω∈(−1,1), and aj,j=1,2,...,m−1,m∈N are positive real constants. The Riemann-Liouville fractional integral of order ξj>0 in time is considered here and is defined as:
Jξj0+,th(t):=1Γ(ξj)∫t0(t−τ)ξj−1h(τ)dτξj>0. |
Here, (−Δ)η1/2 and (−Δ)η2/2 denote the fractional Laplacian operators of orders 1<η1≤η2<2 in the spatial domain. These operators are defined through the spectral decomposition of the Laplacian. Consider {¯λn,ψn} as the eigenvalues and eigenfunctions, respectively, associated with the Helmholtz equation for the Laplacian operator in domain Ω, subject to Dirichlet boundary conditions on ∂Ω:
{−ΔXn=ˉλnψn,inΩ,ψn=0,on∂Ω. | (1.5) |
A straightforward calculation shows that ˉλn=(nπ2)2. Consequently, it follows that
ψn(x)=sin(nπ2(x+1)),∀n≥0. |
Define the operator
Qη1,η2Ω:=−(−Δ)η1/2−(−Δ)η2/2,1<η1≤η2<2, |
on Ω for
h∈Dom(Qη1,η2Ω)={h=∞∑n=1cnψn∈L2(Ω):∞∑n=1c2nλ2n<∞}:=˙Hη1,η2, |
and
Qη1,η2Dh(x)=−∞∑n=1cnλnψn(x)withλn=¯λnη1/2+¯λnη2/2∀n=1,2,.... | (1.6) |
Note that the set ψn(x)n=1∞ constitutes an orthonormal basis for L2(Ω). It is evident that ˙Hη1,η2, a Hilbert space, is a subset of L2(Ω). This Hilbert space is equipped with the inner product ⟨⋅,⋅⟩, which denotes the conventional inner product in L2(Ω):
⟨u,v⟩˙Hη1,η2=⟨Qη1,η2Du,Qη1,η2Dv⟩, |
and induced norms
‖v‖˙Hη1,η2=‖Qη1,η2Dv‖L2(Ω)=(∞∑n=1λ2n⟨v,ψn(x)⟩2)1/2. |
For instance, ˙H0,0=L2(Ω), ˙H1,1=H10(D), and ˙H2,2=H2(D)∩H10(D), all of which have equivalent norms. The dual space of ˙Hη1,η2 for η1,η2>0 is denoted as ˙H−η1,−η2, which corresponds to the dual space (˙Hη1,η2)∗. The notation ⟨h,ψ⟩ represents the action of h on a bounded linear functional ψ within ˙Hη1,η2. It is found that ˙H−η1,−η2 is also a Hilbert space, characterized by the norm
‖ψ‖˙H−η1,−η2=(∞∑n=1λ−2n|⟨h,ψn(x)⟩∗|2)1/2. |
Additionally, if h∈L2(Ω) and ψ∈˙Hη1,η2, then ⟨h,ψ⟩∗=⟨h,ψ⟩, as illustrated in ([12], Chap. V). Now, we are going to discuss two ISPs for the given system (1.1)–(1.4).
In this ISP-Ⅰ, we focus on a source term defined as F(x,t)=f(x). To thoroughly analyze the space-dependent source term f(x) alongside u(x,t), we require additional information commonly referred to as an over-specified condition, presented as
u(x,T)=Φ(x),x∈Ω. | (1.7) |
A classical solution to the ISP-Ⅰ, namely a pair of functions {u(x,t),f(x)}, satisfies the conditions that Tξ0+ξj−1f(x)∈C(ˉΩ), ˉΩT:=ˉΩ×[0,T], ˉΩ∈[−1,1], tξ0+ξj−1u(x,t)∈C(ˉΩT), Qη1,η2Ωu(⋅,t)∈C(ˉΩ), t2ξ0+ξj−1CDξ00+,tu(x,⋅)∈C([0,T]), and t2ξ0+ξj−1CDξj0+,tu(x,⋅)∈C([0,T]),j=1,2,...,m−1m∈N. Our investigation will cover the existence and uniqueness results for the solution of the ISP-Ⅰ under specific assumptions about the given data.
In this ISP-Ⅱ, we examine a source term defined as F(x,t)=q(t)f(x,t). To fully reconstruct the pair of functions u(x,t),q(t), additional information, commonly referred to as an over-specified condition, is required and is provided by
∫1−1u(x,t)dx=E(t),t∈[0,T]. | (1.8) |
We define a classical solution for the ISP-Ⅱ as the set {u(x,t),q(t)}, where
q(t)∈C[0,T],tξ0+ξj−1u(x,t)∈C(ˉΩT),Qη1,η2Ωu(⋅,t)∈C(ˉΩ),t2ξ0+ξj−1CDξ00+,tu(x,⋅)∈C([0,T]) |
and
t2ξ0+ξj−1CDξj0+,tu(x,⋅)∈C([0,T]),j=1,2,...,m−1,m∈N. |
We aim to demonstrate that, under specific conditions applied to the given data, a unique classical solution for the ISP-Ⅱ exists.
The ISPs involving FDEs are a significant area of research in applied mathematics and physics. These problems aim to determine unknown parameters or inputs in a system governed by fractional derivatives, which generalize classical integer-order derivatives to non-integer orders. FDEs are particularly useful in modeling processes with memory effects and anomalous diffusion. Solving ISPs in this context involves techniques to reconstruct hidden information from observable data, often leading to applications in fields such as engineering, finance, and biology. Huntul et al. [13,14] considered the IP of recovering the time-dependent source term for time fractional pseudoparabolic equation. The two ISPs for the time FDEs are studied in [15,16]. Direct and ISPs involving the estimation of specific parameters using numerical techniques for a multi-term time FDE are examined in [17]. Li et al. [18] investigated the well-posedness and long-term asymptotic behavior of initial-boundary value problems for multi-term time FDEs. Lin et al. [19] studied the three dimensional meshfree analysis for time-Caputo and space-Laplacian fractional diffusion equation. The governing equation under consideration involves a linear combination of Caputo derivatives in time with decreasing orders in the interval (0,1) and includes positive constant coefficients. The discussion focuses on ISPs involving the determination of a time-dependent source term for higher-order multi-term FDEs that incorporate the Caputo-Fabrizio derivative [20]. The direct and ISPs for integro-differential equations involving generalized fractional derivatives, along with appropriate over-specified conditions, are discussed in [21,22,23]. The ISP for a class of multi-term time FDEs with non-local boundary conditions is examined in [24]. Ilyas et al. [25] focused on examining two ISPs related to a multi-term time-fractional evolution equation that includes an involution term, bridging the characteristics of both the heat and wave equations. Suhaib et al. [26] examined an ISP to identify a time-dependent source term in a multi-term FDE, incorporating a non-local dynamic boundary condition and an integral-type overdetermination condition. The ISP of determining both diffusion and source terms simultaneously in a multi-term FDE is studied in [27]. Ilyas et al. [28] considered an ISP for a diffusion equation involving a fractional Laplacian operator in space and Hilfer fractional derivatives in time with Dirichlet zero boundary conditions.
The rest of the paper is organized as follows: in this section, we provide the definition of the multinomial Prabhakar and Mittag-Leffler functions and describe several of their properties. In Section 3, we formulate the solution of ISP-Ⅰ, investigate the existence and uniqueness results for ISP-Ⅰ, and present the ill-posedness of ISP-Ⅰ. In Section ref{ISPII}, we present the solution of ISP-Ⅱ, discuss the existence and uniqueness results for ISP-Ⅱ, and also discuss the ill-posedness of ISP-Ⅱ. We also provide some examples related to ISP-Ⅰ and ISP-Ⅱ. In the last section, we present the conclusion.
In this section, we will define the multinomial Prabhakar and Mittag-Leffler functions and discuss some of their important properties.
Definition 1. [29] For γ>0, ηi>0, zi∈C,i=1,2,...,m, m∈N, the multinomial Prabhakar function is defined as
Eδ(η1,η2,...,ηm),γ(z1,z2,...,zm):=∞∑k=0∑l1+l2+...+lm=kl1≥0,...,lm≥0(δ)kl1!...lm!m∏i=1zliiΓ(γ+m∑i=1ηili), |
where (δ)k denotes the Pochhammer symbol
(δ)k=δ(δ+1)...(δ+k−1),k∈N,δ0=1. |
Theorem 1. [30] Let 1≥η1>η2>...>ηm>0, 0<η1δ≤γ≤1, and zi>0,i=1,2,...,m. Then
Eδ(η1,η2,...,ηm),γ(z1,z2,...,zm)∈CMF, |
where CMF represents the complete monotone function.
In the special case δ=1, the Pochhammer symbol yields (1)k=k! and Definition 1 is the multinomial Mittag-Leffler function which is defined as follows:
Definition 2. [31] For γ>0, ηi>0, zi∈C,i=1,2,...,m, m∈N, the multinomial Mittag-Leffler function is defined as:
E(η1,η2,...,ηm),γ(z1,z2,...,zm):=∞∑k=0∑l1+l2+...+lm=kl1≥0,...,lm≥0(k;l1,...,lm)m∏i=1zliiΓ(γ+m∑i=1ηili), |
where (k;l1,...,lm)=k!l1!×...×lm!.
Moreover, note that
E(ξ1,ξ2,...,ξm),γ(z1,z2,...,zm)=E(ξm,...,ξ2,ξ1),γ(zm,...,z2,z1). | (2.1) |
Remark 1. For z1≠0 and z2=Z3=...=zm=0, m∈N, the multinomial Mittag-Leffler function reduces to
E(η1,η2,...,ηm),γ(z1,0,...,0)=∞∑k=0zk1Γ(γ+η1k):=Eη1,γ(z1). | (2.2) |
Lemma 1. [18] Let 0<γ<2 and 0<ηm<...<η1<1 be given. Assume that η1π/2<μ<η1π, μ≤|arg(m2τη1)|≤π, and there exist K>0 such that −K≤−m1τη1−η2<0 and −K≤−m2τη1<0. Then there exists a constant C0>0 depending only on μ,K,ηi,i=1,2,...,m, and γ such that
|E(η1−ηm,...,η1−η2,η1),γ(zm,...,z2,z1)|≤C01+|zm|. |
For convenience, we use the following notation:
E(η1,η2,...,ηm),γ(τ;μ1,μ2,...,μm):=τγ−1E(η1,η2,...,ηm),γ(−μ1τη1,−μ2τη2,...,−μmτηm), | (2.3) |
where μi>0,i=1,2,...,m,m∈N.
Lemma 2. [32] For ηi,γ,τ,μi>0, i=1,2,...,m, m∈N the Laplace transform of the multinomial Mittag-Leffler function is given by
£{E(η1,η2,...,ηm),γ(τ;μ1,...,μm)}=s−γ1+m∑i=1μis−ηi,if|m∑i=1μis−ηi|<1. |
Lemma 3. [25] For ηi,β,γ,τ,μi>0,i=1,2,...,m, the Mittag-Leffler-type functions have the following properties:
● cDγ0+,τ(E(η1,η2,...,ηm),1(τ;μ1,μ2,...,μm))=τ−γE(η1,η2,...,ηm),1−γ(τ;μ1,μ2,...,μm),
● cDγ0+,τ(tγ−βE(η1,η2,...,ηm),γ−β+1(τ;μ1,μ2,...,μm))=τ−βE(η1,η2,...,ηm),1−β(τ;μ1,μ2,...,μm),
● RLDγ0+,τ(τγ−1E(η1,η2,...,ηm),γ(τ;μ1,μ2,...,μm))=τ−βE(η1,η2,...,ηm),1−β(τ;μ1,μ2,...,μm),
● J1−γ0+,τ(τγ−1E(η1,η2,...,ηm),γ(τ;μ1,μ2,...,μm))=E(η1,η2,...,ηm),1(τ;μ1,μ2,...,μm).
Lemma 4. [32] For g∈C1([a,b]) and ηi,μi>0, for i=1,2,...,m, we have
|g(τ)∗E(η1,η2,...,ηm),η1(τ;μ1,μ2,...,μm)|≤Cμ1∥g∥C1([0,T]), |
where ∗ represents the Laplace convolution and ‖g‖C1[0,T]=supt∈[0,T]|g(t)|+supt∈[0,T]|g′(t)|.
Lemma 5. [32] For ηi,γ,β,τ,μi>0,i=1,2,...,m, m∈N, we have the following relation:
τγ∗E(η1,η2,...,ηm),β(τ;μ1,μ2,...,μm)=Γ(γ+1)E(η1,η2,...,ηm),β+γ+1(τ;μ1,μ2,...,μm). |
Proposition 1. [32] The following identities hold for Mittag-Leffler functions:
● Eη1,3(τ;μ1)=τ2Γ(3)−μ1Eη1,3+η1(τ;μ1),
● E(η1,η1−η2),3−η2(τ;μ1,μ2)+μ2E(η1,η1−η2),3+η1−2η2(τ;μ1,μ2)=τ2−η2Γ(3−η2)−μ1E(η1,η1−η2),3+η1−η2(τ;μ1,μ2),
The solution of the ISP-Ⅰ (1.1)–(1.4) and (1.7) can be written by using Fourier's method:
u(t,x)=∞∑n=1Tn(t)ψn(x),f(x)=∞∑n=1fnψn(x), |
where Tn(t) and fn are the unknowns and satisfy the following fractional differential equation:
CDξ10+,tTn(t)+m−1∑j=1ajCDξj0+,tTn(t)=−λnTn(t)+⟨f(x),ψn(x)⟩, | (3.1) |
where λn=(ˉλn)η1/2+(ˉλn)η2/2. Applying the Laplace transform and incorporating the initial conditions (1.3) and (1.4), we obtain
L{Tn(t);s}=s(ξ0−1)⟨ρ(x),ψn(x)⟩sξ0+m−1∑j=1ajsξj−λn+m−1∑j=1ajs(ξj−1)⟨ρ(x),ψn(x)⟩sξ0+m−1∑j=1ajsξj−λn+s(ξ0−1)⟨ν(x),ψn(x)⟩sξ0+m−1∑j=1μjsξj−λn+m−1∑j=1ajs(ξj−1)⟨ν(x),ψn(x)⟩sξ0+m−1∑j=1μjsξj−λn+fns(sξ0+m−1∑j=1μjsξj−λn). |
Due to Lemma 2, we obtain
Tn(t)= Eξ,1(..r..)⟨ρ(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+1(..r..)⟨ρ(x),ψn(x)⟩+Eξ,2(..r..)⟨ν(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+2(..r..)⟨ν(x),ψn(x)⟩+Eξ,ξ0+1(..r..)⟨f(x),ψn(x)⟩, | (3.2) |
where ξ and (..r..) are defined as
ξ=(ξ0,ξ0−ξ2,...,ξ0−ξm−1)and(..r..)=(t;λn,a1,a2,...,am−1), |
where fn=⟨f(x),ψn(x)⟩. To calculate the space-dependent source term fn, we will employ the over-specified condition (1.7), which leads to the following expression:
fn=1Eξ,ξ0+1(..r..)|t=T{⟨Φ(x),ψn(x)⟩−(Eξ,1(..r..)|t=T⟨ρ(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+1(..r..)|t=T⟨ρ(x),ψn(x)⟩+Eξ,2(..r..)|t=T⟨ν(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+2(..r..)|t=T⟨ν(x),ψn(x)⟩)}. | (3.3) |
Consequently, the solution to the ISP-Ⅰ, specifically {u(x,t),f(x)}, is provided by
u(x,t)=∞∑n=1(Eξ,1(..r..)⟨ρ(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+1(..r..)⟨ρ(x),ψn(x)⟩+Eξ,2(..r..)⟨ν(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+2(..r..)⟨ν(x),ψn(x)⟩+Eξ,ξ0+1(..r..)⟨f(x),ψn(x)⟩)ψn(x), | (3.4) |
and
f(x)=∞∑n=11Eξ,ξ0+1(..r..)|t=T{⟨Φ(x),ψn(x)⟩−(Eξ,1(..r..)|t=T⟨ρ(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+1(..r..)|t=T⟨ρ(x),ψn(x)⟩+Eξ,2(..r..)|t=T⟨ν(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+2(..r..)|t=T⟨ν(x),ψn(x)⟩)}ψn(x). | (3.5) |
Lemma 6. For g(.,t)∈C2([−1,1) satisfying g(−1,t)=0=g(1,t), we have
|gn(t)|≤ D1|λn|2‖g(ii)(x,t)‖, |
where
gn(t)= ⟨g(x,t),Xn(x)⟩. | (3.6) |
Proof. From the expression of gn(t) given by (3.6) and integration by parts, we obtain
gn=1|λn|2⟨g(ii)(x,t),Xn(x)⟩. |
Using the Cauchy-Schwarz inequality, we have
|gn|≤1|λn|2‖g(ii)(x,t)‖‖Xn(x)‖, |
which implies
|gn|≤D1|λn|2‖g(ii)(x,t)‖, |
where ‖Xn(x)‖≤D1.
In this subsection, we will describe the classical solution of the ISP-Ⅰ under the given data. Before proceeding further, we present the following lemma, which will be used to determine the continuity of the series obtained by taking the Caputo fractional derivative of the series solution.
Lemma 7. (Lemma 15.2 [33], page 278) Let the fractional derivative CDη0+,xgn(x) exist for all n∈N and for every ϵ>0, the series ∞∑n=1gn(x) and ∞∑n=1CDη0+,xgn(x) are uniformly convergent on the subinterval [ϵ,b]. Then
CDη0+,x(∞∑n=1gn(x))=∞∑n=1CDη0+,xgn(x),η>0,0<x<1. |
Theorem 2. Let ρ(x),ν(x), and Φ(x) satisfy the following conditions:
(1) ρ∈C2(Ω) such that ρ(−1)=0=ρ(1).
(2) ν∈C2(Ω) such that ν(−1)=0=ν(1).
(3) Φ∈C2(Ω) such that Φ(−1)=0=Φ(1).
Then, there exists a classical solution of the ISP-I.
Proof. To establish the existence of a solution for the ISP-Ⅰ, it is necessary to demonstrate the uniform convergence of the infinite series related to the functions f(x), u(x,t), CDξ00+,tu(x,t), and CDξj0+,tu(x,t),j=1,2,...,m−1,m∈N. Initially, we demonstrate that Tξ0+ξj−1|f(x)| denotes a continuous function. By utilizing Lemma 1 and Eq (3.5), we derive
|f(x)|≤∞∑n=1|λn|C0{|Φn|−C0|λn|(|ρn|T−ξ0+m−1∑j=1aj|ρn|T−ξj+|νn|T1−ξ0+m−1∑j=1aj|νn|T1−ξj)}. |
By Lemma 6, we obtain
Tξ0+ξj−1|f(x)|≤∞∑n=1D0|λn|{|λn|C0‖Φ″‖Tξ0+ξi−1−(‖ρ″‖Tξj−1+m−1∑j=1aj‖ρ″‖Tξ0−1+‖ν″‖Tξj+m−1∑j=1aj‖ν″‖Tξ0)}. | (3.7) |
Since, λn=(nπ2)η1+(nπ2)η2 and 1<η1≤η2<2, by (3.7), we can conclude that the series Tξ0+ξj−1|f(x)| converges uniformly for x∈ΩT. Consequently, by the Weierstrass M-test, the series Tξ0+ξj−1|f(x)| represents a continuous function. Subsequently, we demonstrate that tξ0+ξj−1|u(x,t)| denotes a continuous function. Utilizing Lemma 1 and Eq (3.4), we derive the ensuing inequality:
|u(x,t)|≤∞∑n=1C0|λn|(|ρn|t−ξ0+m−1∑j=1aj|ρn|t−ξj+|νn|t1−ξ0+m−1∑j=1aj|νn|t1−ξj+|fn|). |
Due to Lemma 6, we obtain
tξ0+ξj−1|u(x,t)|≤∞∑n=1C0|λn|3(‖ρ″‖tξj−1+m−1∑j=1aj‖ρ″‖tξ0−1+‖ν″‖tξj+m−1∑j=1aj‖ν″‖tξ0+‖f″‖tξ0+ξj−1). | (3.8) |
Based on (3.8), the uniform convergence of the tξ0+ξj−1|u(x,t)| is guaranteed due to the Weierstrass M-test. Hence, we can say that the series tξ0+ξj−1|u(x,t)| represents a continuous function.
Next, we will discuss the convergence of Qη1,η2Ωu(x,t). Due to (1.6), we have
Qη1,η2Ωu(x,t)=−∞∑n=1¯λnρ/2Tn(t)Xn(x),λn=λn+¯λnη2/2. | (3.9) |
From (3.8) and under the assumption of Theorem 4, we can concluded that the uniform convergence of Qη1,η2Ωu(x,t) is ensured. Next, we will investigate the uniform convergence of the corresponding infinite series t2ξ0+ξj−1|CDξ00+,tu(x,t)| and t2ξ0+ξj−1|CDξj0+,tu(x,t)|. Due to Lemma 7, we have
CDξ00+,tu(x,t)=∞∑n=1CDξ00+,tTn(t)ψn(x). |
In order to prove the uniform convergence of t2ξ0+ξj−1|CDξ00+,tu(x,t)|, we need to show that t2ξ0+ξj−1|CDξ00+,tTn(t)| is uniformly convergent. By Lemma 3 and Eq (3.2), we obtain the following expression:
CDξ00+,tTn(t)= Eξ,1−ξ0(..r..)⟨ρ(x),ψn(x)⟩+m−1∑j=1ajEξ,1−ξj(..r..)⟨ρ(x),ψn(x)⟩+ Eξ,2−ξ0(..r..)⟨ν(x),ψn(x)⟩+m−1∑j=1ajEξ,2−ξj(..r..)⟨ν(x),ψn(x)⟩+ Eξ,1(..r..)⟨f(x),ψn(x)⟩. |
By using Lemmas 1 and 6, we obtain
t2ξ0+ξj−1|CDξ00+,tTn(t)|≤∞∑n=1C0|λn|3(‖ρ″(x)‖tξj−1+m−1∑j=1aj‖ρ″(x)‖tξ0−1+‖ν″(x)‖tξj+m−1∑j=1aj‖ν″(x)‖tξ0+‖f″(x)‖t2ξ0+ξj−1). |
The series t2ξ0+ξj−1|CDξ00+,tTn(t)| is bounded above. Hence, t2ξ0+ξj−1|CDξ00+,tu(x,t)| is uniformly continuous by the Weierstrass M-test. In a similar way, we can find that the series corresponding to t2ξ0+ξj−1|CDξj0+,tu(x,t)|,j=1,2,3,...,m−1, represents a continuous function.
In this subsection, the uniqueness of the solution of the ISP-Ⅰ is discussed.
Theorem 3. Let {u(x,t),f(x)} and {˜u(x,t),˜f(x)} be two regular solution sets of the ISP-I. If u(x0,t)=˜u(x0,t) for some x0∈(−π,π), then
u(x,t)=˜u(x,t)⇒f(x)=˜f(x),x∈(−π,π)and(x,t)∈ΩT. |
Proof. Consider the following functions:
Tn(t)=∫1−1u(x,t)ψn(x)dx,and˜Tn(t)=∫1−1˜u(x,t)ψn(x)dx. | (3.10) |
Applying CDξj0+,t,j=0,1,2,...,m−1, to both sides of the second equation in (3.10), we obtain
CDξi0+,t˜Tn(t)=∫1−1CDξi0+,t˜u(x,t)ψn(x)dx. |
From (1.1), we obtain the following fractional differential equations:
CDξi0+,t˜Tn(t)=−λn˜Tn(t)+˜fn. | (3.11) |
Using the Laplace transform and initial conditions (1.3) and (1.4), we have
˜Tn(t)= ˜Tn(0)(Eξ,1(..r..)+m−1∑j=1ajEξ,ξ0−ξj+1(..r..))+˜T′n(0)(Eξ,2(..r..)+m−1∑j=1ajEξ,ξ0−ξj+2(..r..))+˜fnEξ,ξ0+1(..r..). |
Similarly, the following expressions of Tn(t) from the first equations in (3.10) is obtained:
Tn(t)= Tn(0)(Eξ,1(..r..)+m−1∑j=1ajEξ,ξ0−ξj+1(..r..))+T′n(0)(Eξ,2(..r..)+m−1∑j=1ajEξ,ξ0−ξj+2(..r..))+fnEξ,ξ0+1(..r..). |
By using the assumption u(x,t)=˜u(x,t), we have Tn(t)=˜Tn(t) and hence
fnEξ,ξ0+1(..r..)= ˜fnEξ,ξ0+1(..r..),⇒(fn−˜fn)Eξ,ξ0+1(..r..)=0. |
Taking the Laplace transform, we get
(fn−˜fn)sξ0+m−1∑j=1μisαi−λn=0,Re s>0,⇒fn−˜fnω+λn=0, | (3.12) |
where sα0+m∑i=1μisαi=ω. By taking a suitable disk D1 which includes only λ1,1 and using the Cauchy integral theorem, integrating (3.12) along the disk, we have
fk,1=˜fk,1,fork=0. |
In a similar way, by taking different disks, we can find that
fk,1=˜fk,1,for allk∈N. |
Similarly, we can find that
fk,2=˜fk,2,for allk∈N. |
Hence, we have f(x)=˜f(x).
In this subsection, we present an example to demonstrate the ill-posedness of ISP-Ⅰ. Before delving into the example, it is important to summarize a few pertinent observations.
Lemma 8. For λn>0, the following result holds:
ddtEξ,1(..r..)=−λntξ0−1Eξ,ξ0(..r..). | (3.13) |
Moreover, for T>0 the following estimate holds true:
∫T0tξ0−1Eξ,ξ0(..r..)|r=τdτ≤C1λn, | (3.14) |
where C1 is a positive constant.
Proof. From Lemma 2, Eq (3.13) can be proved. To obtain the estimate of (3.14), we consider
∫T0Eξ,ξ0(..r..)|r=τdτ=−1λn∫T0ddτEξ,1(..r..)|r=τdτ=1λn(1−Eξ,1(..r..)|r=T). |
This implies
∫T0Eξ,ξ0(..r..)|r=τdτ=1λn(1−λntξ0Eξ,ξ0+1(..r..)|r=T). |
Due to Lemma 1, we obtain
∫T0Eξ,ξ0(..r..)|r=τdτ≤C1λn. |
The above inequality can be written as
Eξ,ξ0+1(..r..)|t=T≤C1λn. |
The following example addresses the result concerning the ill-posedness of ISP-Ⅰ. Considering the initial conditions to be ˜ρ(x)=0 and ˜ν(x)=0, and the final condition as
˜Φ(x)=1√λksin(kπ2(x+1)), |
where k∈N, the following expression for ˜f(x) is obtained:
˜f(x)=1√λkEξ,ξ0+1(..r..)|t=Tsin(kπ2(x+1)). |
By considering another final data Φ(x)=0 and fixing the initial conditions as ˜ρ(x)=0 and ˜ν(x)=0, we obtain f(x)=0. The two input final data have the following error in the L2-norm:
‖˜Φ−Φ‖L2((−1,1))=‖1√λksin(kπ2(x+1))‖L2((−1,1))=1√λk. |
Hence,
limk→+∞‖˜Φ−Φ‖L2((−1,1))=limk→+∞1√λk=0. | (3.15) |
Additionally, the difference between corresponding source terms in the L2-norm is
‖˜f−f‖L2((−1,1))=‖1√λkEξ,ξ0+1(..r..)|t=Tsin(kπ2(x+1))‖L2((−1,1))=1√λkEξ,ξ0+1(..r..)|t=T. |
Using estimate (3.14), we obtain
‖˜f−f‖L2((−1,1))≥√λkC1, |
which leads us to
limk→+∞‖˜f−f‖L2((−1,1))>limk→+∞√λkC1=+∞. | (3.16) |
Hence, based on Eqs (3.15) and (3.16), we conclude that ISP-Ⅰ is ill-posed.
The solution of the inverse problems (1.1)–(1.4) and (1.8) can be written by using Fourier's method:
u(t,x)=∞∑n=1Tn(t)ψn(x), |
where Tn(t) are the unknowns and satisfy the following fractional differential equation:
CDξ10+,tTn(t)+m−1∑j=1ajCDξj0+,tTn(t)=−λnTn(t)+⟨q(t)f(x,t),ψn(x)⟩. | (4.1) |
By using the Laplace transform and the initial conditions (1.3) and (1.4), we get
L{Tn(t);s}= s(ξ0−1)⟨ρ(x),ψn(x)⟩sξ0+m−1∑j=1ajsξj+λn+m−1∑j=1ajs(ξj−1)⟨ρ(x),ψn(x)⟩sξ0+m−1∑j=1ajsξj+λn+s(ξ0−1)⟨ν(x),ψn(x)⟩sξ0+m−1∑j=1μjsξj+λn+m−1∑j=1ajs(ξj−1)⟨ν(x),ψn(x)⟩sξ0+m−1∑j=1μjsξj+λn+L{⟨q(t)f(x,t),ψn(x)⟩;s}sξ0+m−1∑j=1μjsξj+λn. |
Due to Lemma 2, we obtain
Tn(t)= Eξ,1(..r..)⟨ρ(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+1(..r..)⟨ρ(x),ψn(x)⟩+Eξ,2(..r..)⟨ν(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+2(..r..)⟨ν(x),ψn(x)⟩+Eξ,ξ0(..r..)∗⟨q(t)f(x,t),ψn(x)⟩. | (4.2) |
Hence, the solution u(x,t) is given by
u(x,t)=∞∑n=1(Eξ,1(..r..)⟨ρ(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+1(..r..)⟨ρ(x),ψn(x)⟩+ Eξ,2(..r..)⟨ν(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+2(..r..)⟨ν(x),ψn(x)⟩+ Eξ,ξ0(..r..)∗⟨q(t)f(x,t),ψn(x)⟩)ψn(x), | (4.3) |
where q(t) is still to be determined.
In this section, we will discuss the main results for the solution of the ISP-Ⅱ.
In this subsection, we will present the existence of the solution of the ISP-Ⅱ under certain assumptions of the following theorem.
Theorem 4. Let the following conditions hold:
(1) ρ∈C2)(Ω) such that ρ(−1)=0=ρ(1).
(2) ν∈C2(Ω) such that ν(−1)=0=ν(1).
(3) f(⋅,t)∈C2(Ω) such that f(−1,t)=0=f(1,t). Furthermore ∫1−1f(x,t)dx≠0, and
(∫1−1f(x,t)dx)−1≤M1, |
for some positive constant M1,
(4) E∈AC([0,T]) and E(t) satisfies the following consistency condition:
∫1−1ρ(x)dx=E(t). |
Then, there exists a unique regular solution of the ISP-II.
Proof. To prove the unique existence of the time-dependent source term q(t), we will use the over-specified condition (1.8), and then we have the following relation:
∫1−1(CDξ10+,tu(x,t)+m−1∑j=1ajCDξj0+,tu(x,t))dx=(CDξ10+,t+m−1∑j=1ajCDξj0+,t)E(t). |
From (1.1), we have
∫1−1(−(−Δ)η1/2u(x,t)+−(−Δ)η2/2u(x,t)+q(t)f(x,t))dx=CDξ10+,tE(t)+m−1∑j=1ajCDξj0+,tE(t), |
which yields the following expression:
q(t)=[∫1−1f(x,t)dx]−1[CDξ10+,tE(t)+m−1∑j=1ajCDξj0+,tE(t)+∞∑n=1nπ{Eξ,1(..r..)⟨ρ(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+1(..r..)⟨ρ(x),ψn(x)⟩+Eξ,2(..r..)⟨ν(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+2(..r..)⟨ν(x),ψn(x)⟩+∫t0(t−τ)ξ1−1Eξ,ξ1(..r..)|t=t−τ⟨q(τ)f(x,τ),ψn(x)⟩dτ}(cos(nπ)−1)]. | (5.1) |
We let
T(t)=∞∑n=1nπ{Eξ,1(..r..)⟨ρ(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+1(..r..)⟨ρ(x),ψn(x)⟩+ Eξ,2(..r..)⟨ν(x),ψn(x)⟩+m−1∑j=1ajEξ,ξ0−ξj+2(..r..)⟨ν(x),ψn(x)⟩}(cos(nπ)−1), | (5.2) |
and
K(t,τ)=∞∑n=1nπEξ,ξ1(..r..)|t=t−τ⟨f(x,τ),ψn(x)⟩(cos(nπ)−1). | (5.3) |
Hence, (5.1) becomes
q(t)=[∫1−1f(x,t)dx]−1[CDξ00+,tE(t)+m−1∑j=1ajCDξj0+,tE(t)+T(t)+∫t0q(τ)K(t,τ)dτ]. | (5.4) |
Define the mapping S:C([0,T])→C([0,T]) by
S(q(t)):=q(t), | (5.5) |
where q(t) is given by (5.1). First, we will show that for q(t)∈C([0,T]), the mapping S(q(t)) is well-defined and then, second, we will show that the mapping is a contraction.
Due to Lemmas 1 and 6, and Eqs (5.2) and (5.3), we have
tξ0+ξj−1|T(t)|≤∞∑n=1C0|λn|3(‖ρ″(x)‖tξj−1+m−1∑j=1aj‖ρ″(x)‖tξ0−1+‖ν″(x)‖tξj+m−1∑j=1aj‖ν″(x)‖tξ0), | (5.6) |
(t−τ)|K(t,τ)|≤∞∑n=1C0|λn|3‖f″(τ)‖. | (5.7) |
From Eqs (5.6) and (5.7), we conclude that the series tξ0+ξj−1|T(t)| and (t−τ)|K(t,τ)| are bounded above. Hence, by virtue of the Weierstrass M-test, tξ0+ξj−1|T(t)| and (t−τ)K(t,τ) represent continuous functions. Furthermore, we can have M2>0 such that
|K(t,τ)|≤M2. |
Hence, the mapping defined by (5.5) is well-defined.
Next, we will show that the mapping S(q(t)) is a contraction. By virtue of Eq (5.4), we have
|q1(τ)−q2(τ)|=[∫1−1(x,t)dx]−1{∫t0K(t,τ)|q1(τ)−q2(τ)|dτ}. | (5.8) |
By the assumptions of Theorem 4, we obtain
max0≤t≤T|S(q1(t))−S(q2(t))|≤M1M2Tmax0≤t≤T|q1(τ)−q2(τ)|. |
For T<1M1M2, where M1 and M2 are positive constant independent of n,
max0≤t≤T‖S(q1(t))−S(q2(t))‖C([0,T])≤M1M2Tmax0≤t≤T‖q1−q2‖C([0,T]), |
which implies that the mapping S(.) is a contraction. Hence, the unique existence is guarenteed by using the Banach fixed point theorem.
Next, we will prove that the regular solution u(x,t) given by (4.3), Qη1,η2Ωu(x,t), t2ξ0+ξj−1|CDξ00+,tu(x,t)|, and t2ξ0+ξj−1|CDξj0+,tu(x,t)|,j=1,2,...,m−1,m∈N, represent continuous functions. From Eq (4.3), we have
|u(x,t)|≤∞∑n=1(|Eξ,1(..r..)⟨ρ(x),ψn(x)⟩|+|m−1∑j=1ajEξ,ξ0−ξj+1(..r..)⟨ρ(x),ψn(x)⟩|+ |Eξ,2(..r..)⟨ν(x),ψn(x)⟩|+|m−1∑j=1ajEξ,ξ0−ξj+2(..r..)⟨ν(x),ψn(x)⟩|+ |Eξ,ξ0(..r..)∗⟨q(t)f(x,t),ψn(x)⟩|)|ψn(x)|. |
By Lemmas 1 and 4, we have
|u(x,t)|≤∞∑n=1C0|λn|(‖ρ″(x)‖t−ξ0+m−1∑j=1aj‖ρ″(x)‖t−ξj+‖ν″(x)‖t1−ξ0+m−1∑j=1aj‖ν″(x)‖t1−ξj+‖qf‖), |
which yields to
tξ0+ξj−1|u(x,t)|≤∞∑n=1C0|λn|(‖ρ″(x)‖tξj−1+m−1∑j=1aj‖ρ″(x)‖tξ0−1+‖ν″(x)‖tξj+m−1∑j=1aj‖ρ″(x)‖tξ0+M3‖f‖tξ0+ξj−1). | (5.9) |
The uniform convergence of the series involved in (5.9) is ensured by using the Weierstrass M-test. Hence, we deduce that the series tξ0+ξj−1|u(x,t) represents a continuous function.
Next, due to (1.6), (3.8), and under the assumption of Theorem 4, we can show the uniform convergence of Qη1,η2Ωu(x,t).
Similarly, we will show that the convergence of t2ξ0+ξj−1|CDξ00+,tu(x,t)| and t2ξ0+ξj−1|CDξj0+,tu(x,t)|,j=1,2,...,m−1,m∈N, represent continuous functions.
In this subsection, we will discuss the uniqueness of the solution of the ISP-Ⅱ (1.1)–(1.4) and (1.8).
Theorem 5. The regular solution of the ISP-II is unique by satisfying the assumptions of Theorem 4.
Proof. We have already proved the uniqueness of the time-dependent source term q(t) by using the Banach fixed point theorem. It remains to prove the uniqueness of u(x,t). Let ˉu(x,t)=u1(x,t)−u2(x,t), where u1(x,t) and u2(x,t) are two solution sets of the ISP-Ⅱ (1.1)–(1.4) and (1.8). Then, we have the following relation:
CDξ00+,t˜u(x,t)+m−1∑j=1ajCDξj0+,t˜u(x,t)=−(−Δ)η1/2˜u(x,t)+−(−Δ)η2/2˜u(x,t),x∈ΩT, |
with boundary conditions (1.2) and initial conditions:
˜u(x,0)=0,˜ut(x,0)=0,x∈(−1,1). | (5.10) |
Consider the following function:
˜Tn(t)=∫1−1˜u(x,t)ψn(x)dx. | (5.11) |
Taking the Caputo fractional derivatives CDξ0+,t(.), we obtain the following fractional differential equation:
CDξ00+,t˜Tn(t)+m−1∑j=1ajCDξj0+,t˜Tn(t)=−λn˜Tn(t)+ˉq(t)fn(t). |
By using the Laplace transform technique, we obtain
˜Tn(t)=˜Tn(0)(Eξ,1(..r..)+m−1∑j=1ajEξ,ξ0−ξj+1(..r..))+~T′n(0)(Eξ,2(..r..)+m−1∑j=1ajEξ,ξ0−ξj+2(..r..))+Eξ,ξ0(..r..)∗⟨˜q(t)f(x,t),ψn(x)⟩. |
By using the uniqueness of q(t) and the initial conditions (5.10), we get
\begin{align*} \tilde{T}_{n}(t) & = 0, \quad t \in [0,T]. \end{align*} |
Hence, we have
u_{1}(x,t) = u_{2}(x,t). |
In this subsection, we will demonstrate the ill-posedness of the ISP-Ⅱ. We present the following example to illustrate the ill-posedness of ISP-Ⅱ. In Equation (1.1), we consider two fractional derivatives, that is, a_{i} = 0 , i = 2, 3, ..., m-1, and
\begin{equation*} \tilde{f}(x,t) = \lambda_{k}\Bigg(\frac{\Gamma(3-\xi_{1})}{\Gamma(3-\xi_{0}-\xi_{1})}+a_{1}\frac{\Gamma(3-\xi_{1})}{\Gamma(3-2\xi_{1})}t^{\xi_{0}-\xi_{1}}+\big(\big({\pi}/{2}\big)^{\eta_1}+\big({\pi}/{2}\big)^{\eta_2}\big) t^{\xi_{0}}\Bigg)\sin\Big(\frac{\pi}{2}(x+1)\Big). \end{equation*} |
The initial conditions (1.3) and (1.4) are taken to be zero and the over-specified condition is given by
\begin{equation*} \int_{-1}^{1}\tilde{u}(x,t)dx = \frac{4\lambda_{k}t^{2-\xi_{1}}}{\pi}. \end{equation*} |
Using Lemma 5 and Proposition 1, we obtain
\begin{equation*} \tilde{u}(x,t) = \lambda_{k}t^{2-\xi_{1}} \sin\Big(\frac{\pi}{2}(x+1)\Big). \end{equation*} |
The implicit expression for q(t) has the following form:
\begin{align*} q(t) = & \left[ \int_{-1}^{1}f(x,t)dx\right]^{-1}\biggl[{}^{C}D_{0_+,t}^{\xi_0}E(t)+a_{1} {}^{C}D_{0_+,t}^{\xi_1}E(t) +\mathcal{T}(t)+ \int_{0}^{t} q(\tau)K(t,\tau)d\tau\bigg], \end{align*} |
where
\begin{align*} \int_{0}^{1}f(x,t)dx = \ &\frac{4\lambda_{k}}{\pi}\Bigg(\frac{\Gamma(3-\xi_{1})}{\Gamma(3-\xi_{0}-\xi_{1})}+a_{1}\frac{\Gamma(3-\xi_{1})}{\Gamma(3-2\xi_{1})}t^{\xi_{0}-\xi_{1}}+\big(\big({\pi}/{2}\big)^{\eta_1}+\big({\pi}/{2}\big)^{\eta_2}\big) t^{\xi_{0}}\Bigg),\\ {}^{C}D_{0_+,t}^{\xi_0}E(t) = \ &\frac{4\lambda_{k}\Gamma(3-\xi_{1})}{\Gamma(3-\xi_{0}-\xi_{1})}t^{2-\xi_{0}-\xi_{1}},\qquad {}^{C}D_{0_+,t}^{\xi_1}E(t) = \frac{4\lambda_{k}\Gamma(3-\xi_{1})}{\pi \Gamma(3-2\xi_{1})}t^{2-2\xi_{1}}, \qquad \mathcal{T}(t) = 0, \end{align*} |
K(t,\tau) = \frac{4\lambda_{k}}{\pi}\Bigg(\frac{\Gamma(3-\xi_{1})}{\Gamma(3-\xi_{0}-\xi_{1})}+a_{1}\frac{\Gamma(3-\xi_{1})}{\Gamma(3-2\xi_{1})}t^{\xi_{0}-\xi_{1}}+\big(\big({\pi}/{2}\big)^{\eta_1}+\big({\pi}/{2}\big)^{\eta_2}\big) t^{\xi_{0}}\Bigg)\mathcal{E}_{\boldsymbol{\xi},\xi_{1}}(..r..)|_{t = t-\tau}. |
Consequently, we get
q(t) = t^{3-\xi_{0}-\xi_{1}}. |
Now, assume that we have another source term, i.e., f(x, t) . Hence, the solution related to f(x, t) is u(x, t).
An error in L^{2}({\Omega}_{T}) between two corresponding solutions is:
\begin{align*} \|\tilde{u}-u\|_{L^{2}({\Omega}_{T})}& = \int_{0}^{T}\|\tilde{u}-u\|_{L^{2}(-1,1)}dt\;\; \leq \; \frac{4\lambda_{k}}{\pi} \int_{0}^{T}t^{2-\xi_{1}} \;dt. \end{align*} |
This yields
\begin{align*} \|\tilde{u}-u\|_{L^{2}({\Omega}_{T})}& = \frac{4\lambda_{k}{T}^{3-\xi_{1}}}{\pi(3-\xi_{1})}. \end{align*} |
Taking the limit as k\rightarrow \infty , one gets the ill-posedness of the ISP-Ⅱ:
\begin{equation*} \label{uillposed} \lim\limits_{k \to \infty}\|\tilde{u}-u\|_{L^{2}({\Omega}_{T})} = \frac{4{T}^{3-\xi_{1}}}{\pi(3-\xi_{1})} \lim\limits_{k \to \infty} \lambda_{k} = +\infty. \end{equation*} |
In this section, we will present some examples for the ISPs.
Example 1. As a specific example of ISP-I, we consider
\rho(x) = \cos \big(\frac{x}{2}\big),\quad \nu(x) = \frac{t^{\xi_{0}}}{\Gamma(5+\xi_{0})}\cos \big(\frac{\pi x}{2}\big),\quad {and}\quad \varPhi(x) = T^{\xi_{0}}\sin\Big(\frac{ \pi}{2}(x+1)\Big). |
The coefficients of the series expansion of \rho(x) , \nu(x) , and \varPhi(x) for n = 1 are given by
\rho_{1} = \frac{4\pi\cos\big(\frac{x}{2}\big)}{\pi^{2}-1}, \quad \nu_{1} = \frac{t^{\xi_{0}}}{\Gamma(5+\xi_{0})}, \quad {and} \quad \varPhi_{1} = T^{\xi_{0}}. |
By plugging a_i = 0 for i = 1, 2, 3, ..., m-1 and using (3.3), we obtain
\begin{equation} f_{1} = \frac{1}{ \mathcal{E}_{\xi_{0}, \xi_{0}+1}(T;\lambda_{1})}\Bigg(T^{\xi_{0}}-\frac{4\pi\cos\big(\frac{x}{2}\big)}{\pi^{2}-1}\mathcal{E}_{\xi_{0}, 1}(T;\lambda_{1})-\frac{t^{\xi_{0}}}{\Gamma(5+\xi_{0})}\mathcal{E}_{\xi_{0}, 2}(T;\lambda_{1})\Bigg). \end{equation} | (6.1) |
Equation (3.2) yields the following expressions:
\begin{align} T_{1}(t) = &\frac{4\pi\cos\big(\frac{x}{2}\big)}{\pi^{2}-1}\mathcal{E}_{\xi_{0}, 1}(T;\lambda_{1})+\frac{t^{\xi_{0}}}{\Gamma(5+\xi_{0})}\mathcal{E}_{\xi_{0}, 2}(T;\lambda_{1})+f_{1} \mathcal{E}_{\xi_{0}, \xi_{0}+1}(T;\lambda_{1}), \end{align} | (6.2) |
where f_{1} is given in Eq (6.1). By substituting the previously derived expressions for f_{1} and T_{1}(t) , we obtain the solution to the inverse source problem, namely f(x) and u(x, t) , as follows:
\begin{eqnarray*} f(x)& = &\frac{\sin\Big(\frac{ \pi}{2}(x+1)\Big)}{ \mathcal{E}_{\xi_{0}, \xi_{0}+1}(T;\lambda_{1})}\Bigg(T^{\xi_{0}}-\frac{4\pi\cos\big(\frac{x}{2}\big)}{\pi^{2}-1}\mathcal{E}_{\xi_{0}, 1}(T;\lambda_{1})-\frac{t^{\xi_{0}}}{\Gamma(5+\xi_{0})}\mathcal{E}_{\xi_{0}, 2}(T;\lambda_{1})\Bigg), \end{eqnarray*} |
\begin{align*} u(x,t) = \ &\Bigg(\frac{4\pi\cos\big(\frac{x}{2}\big)}{\pi^{2}-1}\mathcal{E}_{\xi_{0}, 1}(T;\lambda_{1})+\frac{t^{\xi_{0}}}{\Gamma(5+\xi_{0})}\mathcal{E}_{\xi_{0}, 2}(T;\lambda_{1})+f_{1} \mathcal{E}_{\xi_{0}, \xi_{0}+1}(T;\lambda_{1})\Bigg)\sin\Big(\frac{ \pi}{2}(x+1)\Big). \end{align*} |
Example 2. In this second example, the solution set u(x, t) and f(x) of the ISP-I is obtained by setting a_1 = 1 and a_i = 0 , i = 2, 3, ..., m-1 :
\begin{align*} f(x) = &\frac{\sin\Big(\frac{ \pi}{2}(x+1)\Big)}{ \mathcal{E}_{(\xi_{0},\xi_{0}-\xi_{1}), \xi_{0}+1}(T;\lambda_{1})}\Bigg(T^{\xi_{0}}-\frac{4\pi\cos\big(\frac{x}{2}\big)}{\pi^{2}-1}\mathcal{E}_{(\xi_{0},\xi_{0}-\xi_{1}), 1}(T;\lambda_{1})-\frac{t^{\xi_{0}}}{\Gamma(5+\xi_{0})}\mathcal{E}_{(\xi_{0},\xi_{0}-\xi_{1}), 2}(T;\lambda_{1})\Bigg), \end{align*} |
\begin{align*} u(x,t) = \ &\Bigg(\frac{4\pi\cos\big(\frac{x}{2}\big)}{\pi^{2}-1}\mathcal{E}_{(\xi_{0},\xi_{0}-\xi_{1}), 1}(T;\lambda_{1})+\frac{t^{\xi_{0}}}{\Gamma(5+\xi_{0})}\mathcal{E}_{(\xi_{0},\xi_{0}-\xi_{1}), 2}(T;\lambda_{1})\nonumber\\+&f_{1} \mathcal{E}_{(\xi_{0},\xi_{0}-\xi_{1}), \xi_{0}+1}(T;\lambda_{1})\Bigg)\sin\Big(\frac{ \pi}{2}(x+1)\Big). \end{align*} |
Example 3. For the specific case of ISP-II, we consider only one fractional derivative in Eq (1.1), where a_{i} = 0 for i = 1, 2, ..., m-1. The function f(x, t) is given by
f(x,t) = \Big(\frac{\Gamma(3)}{\Gamma(3-\xi_{0})}+\big(\big({\pi}/{2}\big)^{\eta_1}+\big({\pi}/{2}\big)^{\eta_2}\big) t^{\xi_{0}}\Big)\sin\Big(\frac{ \pi}{2}(x+1)\Big),\qquad n = 1, |
and the initial conditions in (1.3) and (1.4) are set to zero. The over-specified condition is \int_{-1}^{1}u(x, t)dx = \frac{4t^{2}}{\pi}. Using Lemma 5 and Proposition 1, we obtain
\begin{equation*} u(x,t) = t^{2} \sin\Big(\frac{ \pi}{2}(x+1)\Big). \end{equation*} |
The expression for q(t) given by (5.1) takes the form
\begin{align*} q(t) = & \left[ \int_{-1}^{1}f(x,t)dx\right]^{-1}\biggl[{}^{C}D_{0_+,t}^{\xi_0}E(t) +\mathcal{T}(t)+ \int_{0}^{t} q(\tau)K(t,\tau)d\tau\bigg], \end{align*} |
where
\begin{align*} \int_{-1}^{1}f(x,t)dx = \ &\frac{4}{\pi}\Big(\frac{\Gamma(3)}{\Gamma(3-\xi_{0})}+\big(\big({\pi}/{2}\big)^{\eta_1}+\big({\pi}/{2}\big)^{\eta_2}\big) t^{\xi_{0}}\Big),\\ {}^{C}D_{0_+,t}^{\xi_0}E(t) = \ &\frac{4\Gamma(3)}{\pi\Gamma(3-\xi_{0})}t^{2-\xi_{0}}, \qquad\quad \mathcal{T}(t) = 0,\\ K(t,\tau) = \ &\Big(\frac{\Gamma(3)}{\Gamma(3-\xi_{0})}+\big(\big({\pi}/{2}\big)^{\eta_1}+\big({\pi}/{2}\big)^{\eta_2}\big) \tau^{\xi_{0}}\Big)\mathcal{E}_{{\xi_{0}},\xi_{0}}(t-\tau;\lambda_{1}). \end{align*} |
In this case, we can find an explicit expression for q(t) given by
q(t) = t^{2-\xi_0}. |
Example 4. In this example of ISP-II, we take two fractional derivatives in Eq (1.1), where a_{i} = 0 , i = 2, 3, ..., m-1. Consider
\begin{equation*} {f}(x,t) = \Bigg(\frac{\Gamma(3-\xi_{1})}{\Gamma(3-\xi_{0}-\xi_{1})}+a_{1}\frac{\Gamma(3-\xi_{1})}{\Gamma(3-2\xi_{1})}t^{\xi_{0}-\xi_{1}}+\big(\big({\pi}/{2}\big)^{\eta_1}+\big({\pi}/{2}\big)^{\eta_2}\big) t^{\xi_{0}}\Bigg)\sin\Big(\frac{\pi}{2}(x+1)\Big). \end{equation*} |
The initial conditions (1.3) and (1.4) are taken to be zero and the over-specified condition is \int_{-1}^{1}u(x, t)dx = \frac{4t^{2-\xi_{1}}}{\pi}. Due to Lemma 5 and Proposition 1, we obtain
\begin{equation*} {u}(x,t) = t^{2-\xi_{1}} \sin\Big(\frac{\pi}{2}(x+1)\Big). \end{equation*} |
The implicit expression for q(t) has the following form:
\begin{align*} q(t) = & \left[ \int_{-1}^{1}f(x,t)dx\right]^{-1}\biggl[{}^{C}D_{0_+,t}^{\xi_0}E(t)+a_{1} {}^{C}D_{0_+,t}^{\xi_1}E(t) +\mathcal{T}(t)+ \int_{0}^{t} q(\tau)K(t,\tau)d\tau\bigg], \end{align*} |
where
\begin{align*} \int_{0}^{1}f(x,t)dx = \ &\frac{4}{\pi}\Bigg(\frac{\Gamma(3-\xi_{1})}{\Gamma(3-\xi_{0}-\xi_{1})}+a_{1}\frac{\Gamma(3-\xi_{1})}{\Gamma(3-2\xi_{1})}t^{\xi_{0}-\xi_{1}}+\big(\big({\pi}/{2}\big)^{\eta_1}+\big({\pi}/{2}\big)^{\eta_2}\big) t^{\xi_{0}}\Bigg),\\ {}^{C}D_{0_+,t}^{\xi_0}E(t) = \ &\frac{4\Gamma(3-\xi_{1})}{\Gamma(3-\xi_{0}-\xi_{1})}t^{2-\xi_{0}-\xi_{1}},\quad {}^{C}D_{0_+,t}^{\xi_1}E(t) = \frac{4\Gamma(3-\xi_{1})}{\pi \Gamma(3-2\xi_{1})}t^{2-2\xi_{1}}, \quad \mathcal{T}(t) = 0, \end{align*} |
K(t,\tau) = \frac{4}{\pi}\Bigg(\frac{\Gamma(3-\xi_{1})}{\Gamma(3-\xi_{0}-\xi_{1})}+a_{1}\frac{\Gamma(3-\xi_{1})}{\Gamma(3-2\xi_{1})}t^{\xi_{0}-\xi_{1}}+\big(\big({\pi}/{2}\big)^{\eta_1}+\big({\pi}/{2}\big)^{\eta_2}\big) t^{\xi_{0}}\Bigg)\mathcal{E}_{\boldsymbol{\xi},\xi_{1}}(..r..)|_{t = t-\tau}. |
Consequently, we get
q(t) = t^{3-\xi_{0}-\xi_{1}}. |
In this article, two ISPs for a multi-term space-time fractional differential equation (FDE) incorporating Caputo fractional derivatives with respect to time and a bi-fractional Laplacian operator with respect to space are considered. The series solution for the ISPs is constructed by using the eigenfunction expansion method. First, ISP-Ⅰ is addressed, which determines a space-varying source term from the over-specified condition, i.e., the given data at a specific time T , along with the solution. The series solutions, obtained through the eigenfunction expansion method using an orthonormal set of eigenfunctions of the associated spectral problem, involve multinomial Mittag-Leffler functions. The regularity of the solution is ensured under certain assumptions about the given data and by utilizing results related to multinomial Mittag-Leffler functions. Additionally, the uniqueness and stability of the solution concerning the given data are guaranteed in a similar manner. These ISPs are shown to be ill-posed in the sense of Hadamard. The ISP-Ⅱ is focused on recovering a time-dependent source term from the over-determined condition of integral type, along with the solution. The unique existence of a continuous source term in ISP-Ⅱ is established using the Banach fixed point theorem. Furthermore, the uniqueness of the solution is demonstrated using the completeness of the eigenfunctions.
The authors gratefully acknowledge the funding of the Deanship of Graduate Studies and Scientific Research, Jazan University, Saudi Arabia, through Project Number: GSSRD-24.
The author declares that there is no conflict of interest.
[1] | C. H. Yu, Fractional derivatives of some fractional functions and their applications, Asian J. Appl. Sci. Technol., 4 (2020), 147–158. |
[2] |
H. M. Srivastava, Fractional-order derivatives and integrals: introductory overview and recent developments, Kyungpook Math. J., 60 (2020), 73–116. https://doi.org/10.5666/KMJ.2020.60.1.73 doi: 10.5666/KMJ.2020.60.1.73
![]() |
[3] |
Y. Luchko, Operational calculus for the general fractional derivative and its applications, Fract. Calc. Appl. Anal., 24 (2021), 338–375. https://doi.org/10.1515/fca-2021-0016 doi: 10.1515/fca-2021-0016
![]() |
[4] | A. A. Kilbas, Partial fractional differential equations and some of their applications, (2010). https://doi.org/10.1524/anly.2010.0934 |
[5] | A. A. Kilbas, H. M. Srivastava, J. J. Trujillo, Theory and applications of fractional differential equations, Amsterdam: Elsevier, 2006. |
[6] | G. H. Weiss, R. J. Rubin, Random walks: theory and selected applications, 52 (1982). https://doi.org/10.1002/9780470142769.ch5 |
[7] |
R. B. Bird, D. J. Klingenberg, Multicomponent diffusion-a brief review, Adv. Water Resour., 62 (2013), 238–242. https://doi.org/10.1016/j.advwatres.2013.05.010 doi: 10.1016/j.advwatres.2013.05.010
![]() |
[8] | R. Klages, G. Radons, I. M. Sokolov, Anomalous transport: foundations and applications, Hoboken: Wiley, 2008. https://onlinelibrary.wiley.com/doi/book/10.1002/9783527622979 |
[9] |
Y. Itto, Hetergeneous anomalous diffusion in view of superstatistics, Phys. Lett. A, 378 (2014), 3037–3040. https://doi.org/10.1016/j.physleta.2014.08.022 doi: 10.1016/j.physleta.2014.08.022
![]() |
[10] | R. Hilfer, On fractional diffusion and its relation with continuous time random walks, In: Proceedings of the XIth Max Born Symposium held at Ladek Zdroj, Poland: Springer, 1998. https://doi.org/10.1007/BFb0106834 |
[11] |
R. Matzler, J. Klafter, The random walk's guide to anomalous diffusion: a fractional dynamics approach, Phys. Rep., 339 (2000), 1–77. https://doi.org/10.1016/S0370-1573(00)00070-3 doi: 10.1016/S0370-1573(00)00070-3
![]() |
[12] | H. Brezis, Analyse fonctionnelle, Paris: Masson, 1983. |
[13] |
M. J. Huntul, I. Tekin, M. K. Iqbal, M. Abbas, An inverse problem of recovering the heat source coefficient in a fourth-order time-fractional pseudo-parabolic equation, J. Comput. Appl. Math., 442 (2024), 115712. https://doi.org/10.1016/j.cam.2023.115712 doi: 10.1016/j.cam.2023.115712
![]() |
[14] |
K. Khompysh, M. J. Huntul, M. K. Shazyndayeva, M. Iqbal, An inverse problem for pseudoparabolic equation: existence, uniqueness, stability, and numerical analysis, Quaest. Math., 47 (2024), 1979–2001. https://doi.org/10.2989/16073606.2024.2347432 doi: 10.2989/16073606.2024.2347432
![]() |
[15] |
A. Ilyas, R. A. Khalid, S. A. Malik, Identifying temperature distribution and source term for generalized diffusion equation with arbitrary memory kernel, Math. Meth. Appl. Sci., 47 (2024), 5894–5915. https://doi.org/10.1002/mma.9896 doi: 10.1002/mma.9896
![]() |
[16] |
K. Suhaib, A. Ilyas, S. A. Malik, On the inverse problems for a family of integro-differential equations, Math. Model. Anal., 28 (2023), 255–270. https://doi.org/10.3846/mma.2023.16139 doi: 10.3846/mma.2023.16139
![]() |
[17] |
W. Fan, F. Liu, X. Jiang, I. Turner, Some noval numerical techniques for an inverse problem of the multi-term time fractional partial differential equuation, J. Comput. Appl. Math., 25 (2017), 1618–1638. https://doi.org/10.1016/j.cam.2017.12.034 doi: 10.1016/j.cam.2017.12.034
![]() |
[18] |
Z. Li, Y. Liu, M. Yamamoto, Initial-boundary value problems for multi-term time-fractional diffusion equations with positive constant coefficients, Appl. Math. Comput., 257, (2015), 381–397. https://doi.org/10.1016/j.amc.2014.11.073 doi: 10.1016/j.amc.2014.11.073
![]() |
[19] |
Z. Lin, F. Liu, J. Wu, D. Wang, Y. Gu, Three dimensional meshfree analysis for time-Caputo and space-Laplacian fractional diffusion equation, Eng. Anal. Bound. Elem., 157 (2023), 553–564. https://doi.org/10.1016/j.enganabound.2023.10.005 doi: 10.1016/j.enganabound.2023.10.005
![]() |
[20] | S. Pirnafasov, E. Karimov, On a higher order multi-term time-fractional partial differential equation involving Caputo-Fabrizio derivative, preprint paper, 2017. https://doi.org/10.48550/arXiv.1708.05502 |
[21] |
A. Ilyas, S. A. Malik, Direct and some inverse problems for a generalized diffusion equation with variable coefficients, Comp. Appl. Math., 43 (2024), 364. https://doi.org/10.1007/s40314-024-02869-2 doi: 10.1007/s40314-024-02869-2
![]() |
[22] |
A. Ilyas, Z. Iqbal, S. A. Malik, On some direct and inverse problems for an integro-differential equation, Z. Angew. Math. Phys., 75 (2024), 39. https://doi.org/10.1007/s00033-024-02186-y doi: 10.1007/s00033-024-02186-y
![]() |
[23] |
A. Ilyas, S. A. Malik, S. Saif, On the solvability of direct and inverse problems for a generalized diffusion equation, Phy. Scr., 98 (2023), 125221. https://doi.org/10.1088/1402-4896/ad03c5 doi: 10.1088/1402-4896/ad03c5
![]() |
[24] |
M. Ali, S. Aziz, S. A. Malik, Inverse problem for a multi-term fractional differential equation, Fract. Calc. Appl. Anal., 23 (2020), 799–821. https://doi.org/10.1515/fca-2020-0040 doi: 10.1515/fca-2020-0040
![]() |
[25] |
A. Ilyas, S. A. Malik, S. Saif, Inverse problems for a multi-term time fractional evolution equation with an involution, Inverse Probl. Sci. Eng., 29 (2021), 3377–3405. https://doi.org/10.1080/17415977.2021.2000606 doi: 10.1080/17415977.2021.2000606
![]() |
[26] |
K. Suhaib, S. A. Malik, A. Ilyas, Existence and uniqueness results for a multi-parameters nonlocal diffusion equation, Rep. Math. Phys., 90 (2022), 203–219. https://doi.org/10.1016/S0034-4877(22)00066-0 doi: 10.1016/S0034-4877(22)00066-0
![]() |
[27] |
H. Sun, G. Li, X. Jia, Simultaneous inversion for the diffusion and source coefficients in the multi-term TFDE, Inverse Probl. Sci. Eng., 336 (2018), 114–126. https://doi.org/10.1080/17415977.2016.1275612 doi: 10.1080/17415977.2016.1275612
![]() |
[28] |
A. Ilyas, S. A. Malik, K. Suhaib, Identifying diffusion concentration and source term for anomalous diffusion equation, Rep. Math. Phys., 93, (2024) 145–163. https://doi.org/10.1016/S0034-4877(24)00023-5 doi: 10.1016/S0034-4877(24)00023-5
![]() |
[29] |
E. Bazhlekova, I. Bazhlekov, Identification of a space-dependent source term in a nonlocal problem for the general time-fractional diffusion equation, J. Comput. Appl. Math., 386 (2021), 113213. https://doi.org/10.1016/j.cam.2020.113213 doi: 10.1016/j.cam.2020.113213
![]() |
[30] |
E. Bazhlekova, Completely monotone multinomial Mittag-Leffler type functions and diffusion equations with multiple time-derivatives, Fract. Calc. Appl. Anal., 24 (2021), 88–111. https://doi.org/10.1515/fca-2021-0005 doi: 10.1515/fca-2021-0005
![]() |
[31] |
Y. Luchko, R. Gorenflo, An operational method for solving fractional differential equations with the Caputo derivatives, Acta Math. Vietn., 24 (1999), 207–233. https://doi.org/10.1016/j.amc.2014.05.112 doi: 10.1016/j.amc.2014.05.112
![]() |
[32] |
S. A. Malik, A. Ilyas, A. Samreen, Simultaneous determination of a source term and diffusion concentration for a multi-term space-time fractional diffusion equation, Math. Model. Anal., 26 (2021), 411–431. https://doi.org/10.3846/mma.2021.11911 doi: 10.3846/mma.2021.11911
![]() |
[33] | G. S. Samko, A. A. Kilbas, D. I. Marichev, Fractional integrals and derivatives: theory and applications, Sweden: Gordon and Breach Science Publishers, 1993. |