Skip to main content
Log in

Uniqueness and Numerical Method for Determining a Spatial Source Term in a Time-Fractional Diffusion Wave Equation

  • Published:
Journal of Scientific Computing Aims and scope Submit manuscript

Abstract

This paper is devoted to recovering a spatial source term in a time-fractional diffusion-wave equation by using an additional final time observation. Based on the regularity of solution for the direct problem, the existence and uniqueness of the spatial source term are obtained. For this ill-posed problem, the generalized quasi-boundary value regularization method is proposed, and the convergence rate of the regularized solution under an a priori choice rule of the regularization parameter is achieved. Specially, for the modified quasi-boundary value regularized problem, we propose a numerical method in terms of combining the backward Euler convolution quadrature scheme for approximating the time-fractional derivative and the piecewise linear finite element scheme for dealing with the space variables. Based on a complicated numerical analysis, we give an error estimate between the numerical source solution in a fully discrete regularized problem and the exact source function under certain reasonable smoothness assumptions to the given data including the initial functions and a time-dependent source term. Some numerical experiments are provided for one- and two-dimensional cases to verify the theoretical result and illustrate the efficient of the proposed numerical algorithm.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10

Similar content being viewed by others

Data Availability

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.

References

  1. Mainardi, F.: Fractional relaxation-oscillation and fractional diffusion-wave phenomena. Chaos, Solitons Fractals 7(9), 1461–1477 (1996)

    Article  MathSciNet  Google Scholar 

  2. Mainardi, F.: Fractional Calculus and Waves in Linear Viscoelasticity: An Introduction to Mathematical Models. World Scientific, Singapore (2010)

    Book  Google Scholar 

  3. Agrawal, O.P.: Solution for a fractional diffusion-wave equation defined in a bounded domain. Nonlinear Dyn. 29(1), 145–155 (2002)

    Article  MathSciNet  Google Scholar 

  4. Sakamoto, K., Yamamoto, M.: Initial value/boundary value problems for fractional diffusion-wave equations and applications to some inverse problems. J. Math. Anal. Appl. 382(1), 426–447 (2011)

    Article  MathSciNet  Google Scholar 

  5. Kian, Y., Yamamoto, M.: On existence and uniqueness of solutions for semilinear fractional wave equations. Fract. Calc. Appl. Anal. 20(1), 117–138 (2017)

    Article  MathSciNet  Google Scholar 

  6. Sun, Z.Z., Wu, X.N.: A fully discrete difference scheme for a diffusion-wave system. Appl. Numer. Math. 56(2), 193–209 (2006)

    Article  MathSciNet  Google Scholar 

  7. Chen, A., Li, C.P.: Numerical solution of fractional diffusion-wave equation. Numer. Funct. Anal. Optim. 37(1), 19–39 (2016)

    Article  MathSciNet  Google Scholar 

  8. Huang, J.F., Tang, Y.F., Vázquez, L., Yang, J.Y.: Two finite difference schemes for time fractional diffusion-wave equation. Numer. Algorithms 64(4), 707–720 (2013)

    Article  MathSciNet  Google Scholar 

  9. Du, R., Cao, W.R., Sun, Z.Z.: A compact difference scheme for the fractional diffusion-wave equation. Appl. Math. Model. 34(10), 2998–3007 (2010)

    Article  MathSciNet  Google Scholar 

  10. Jin, B.T., Lazarov, R., Zhou, Z.: Two fully discrete schemes for fractional diffusion and diffusion-wave equations with nonsmooth data. SIAM J. Sci. Comput. 38(1), A146–A170 (2016)

    Article  MathSciNet  Google Scholar 

  11. Sisková, K., Slodicka, M.: Recognition of a time-dependent source in a time-fractional wave equation. Appl. Numer. Math. 121, 1–17 (2017)

    Article  MathSciNet  Google Scholar 

  12. Wei, T., Zhang, Y.: The backward problem for a time-fractional diffusion-wave equation in a bounded domain. Comput. Math. Appl. 75(10), 3632–3648 (2018)

    Article  MathSciNet  Google Scholar 

  13. Floridia, G., Yamamoto, M.: Backward problems in time for fractional diffusion-wave equation. Inverse Probl. 36(12) (2020)

  14. Kian, Y., Li, Z.Y., Liu, Y.K., Yamamoto, M.: The uniqueness of inverse problems for a fractional equation with a single measurement. Math. Ann. 380(3–4), 1465–1495 (2021)

    Article  MathSciNet  Google Scholar 

  15. Zhang, Z.Q., Zhou, Z.: Backward diffusion-wave problem: stability, regularization, and approximation. SIAM J. Sci. Comput. 44(5), A3183–A3216 (2022)

    Article  MathSciNet  Google Scholar 

  16. Wei, T., Xian, J.: Determining a time-dependent coefficient in a time-fractional diffusion-wave equation with the Caputo derivative by an additional integral condition. J. Comput. Appl. Math. 404, 113910 (2022)

    Article  MathSciNet  Google Scholar 

  17. Zhang, Y., Wei, T., Yan, X.B.: Recovery of advection coefficient and fractional order in a time-fractional reaction-advection-diffusion-wave equation. J. Comput. Appl. Math. 411, 114254 (2022)

    Article  MathSciNet  Google Scholar 

  18. Liao, K.F., Wei, T.: Identifying a fractional order and a space source term in a time-fractional diffusion-wave equation simultaneously. Inverse Prob. 35(11), 115002 (2019)

    Article  MathSciNet  Google Scholar 

  19. Cheng, X., Li, Z.Y.: Uniqueness and stability for inverse source problem for fractional diffusion-wave equations. J. Inverse Ill-posed Probl. (2023)

  20. Wei, T., Luo, Y.H.: A generalized quasi-boundary value method for recovering a source in a fractional diffusion-wave equation. Inverse Probl. 38(4), 045001 (2022)

    Article  MathSciNet  Google Scholar 

  21. Kilbas, A.A., Srivastava, H.M., Trujillo, J.J.: Theory and Applications of Fractional Differential Equations, vol. 204. Elsevier, Amsterdam (2006)

    Book  Google Scholar 

  22. Jin, B.T.: Fractional Differential Equations: An Approach via Fractional Derivatives. Springer, Berlin (2021)

    Book  Google Scholar 

  23. Thomée, Vidar: Galerkin Finite Element Methods for Parabolic Problems, vol. 25. Springer, Berlin (2007)

    Google Scholar 

  24. Liao, K.F., Zhang, L., Wei, T.: Identifying a fractional order and a time-dependent coefficient in a time-fractional diffusion wave equation. J. Inverse and Ill-Posed Probl. 31(5), 631–652 (2023)

    Article  MathSciNet  Google Scholar 

  25. Jin, B.T., Lazarov, R., Zhou, Z.: Error estimates for a semidiscrete finite element method for fractional order parabolic equations. SIAM J. Numer. Anal. 51(1), 445–466 (2013)

    Article  MathSciNet  Google Scholar 

  26. Cuesta, E., Lubich, C., Palencia, C.: Convolution quadrature time discretization of fractional diffusion-wave equations. Math. Comput. 75(254), 673–696 (2006)

    Article  MathSciNet  Google Scholar 

  27. Lubich, C.: Discretized fractional calculus. SIAM J. Math. Anal. 17(3), 704–719 (1986)

    Article  MathSciNet  Google Scholar 

  28. Lubich, C.: Convolution quadrature and discretized operational calculus. I. Numerische Mathematik 52(2), 129–145 (1988)

    Article  MathSciNet  Google Scholar 

  29. Lubich, C.: Convolution quadrature and discretized operational calculus. II. Numerische Mathematik 52(4), 413–425 (1988)

    Article  MathSciNet  Google Scholar 

  30. Sousa, E.: How to approximate the fractional derivative of order 1\(<\)\(\alpha \)\(\le \)¡2. Int. J. Bifurc. Chaos 22(04), 1250075 (2012)

    Article  MathSciNet  Google Scholar 

  31. Zhang, Zhengqi, Zhou, Zhi: Numerical analysis of backward subdiffusion problems. Inverse Probl. 36(10), 105006 (2020)

    Article  MathSciNet  Google Scholar 

  32. Jin, B.T., Li, B.Y., Zhou, Z.: Correction of high-order BDF convolution quadrature for fractional evolution equations. SIAM J. Sci. Comput. 39(6), A3129–A3152 (2017)

    Article  MathSciNet  Google Scholar 

  33. Jin, B.T., Zhou, Z.: Numerical Treatment and Analysis of Time-Fractional Evolution Equations. Springer, Berlin (2023)

    Book  Google Scholar 

Download references

Funding

The authors have not disclosed any funding.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Ting Wei.

Ethics declarations

Conflict of interest

The authors have no Conflict of interest to declare that are relevant to the content of this article.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

This paper is supported by the National Natural Science Foundation of China (12171215) and Natural Science Foundation of Gansu Province (22JR5RA391).

Appendix

Appendix

In this appendix, we present the deduction of fully discrete solution in (6.3) and the proof of Lemma 6.2. The following two Lemmas will be utilized.

Lemma 8.1

[10] Let \(\Sigma _\theta =\{z\in {\mathbb {C}}:|z|\ne 0, |\arg {z}|\le \theta \}\) for \(\pi /2<\theta <\pi /\alpha \), there exists positive constants \(c=c(\theta , \alpha )\) s.t.

$$\begin{aligned} \Vert (z^\alpha -L)^{-1}\Vert \le c|z|^{-\alpha },\quad \Vert (z^\alpha -L_h)^{-1}\Vert \le c|z|^{-\alpha },\quad z\in \Sigma _{\theta }. \end{aligned}$$

Lemma 8.2

[15] For any \(\theta \in (\pi /2, \pi )\), there exists \(\theta ^{\prime }\in (\pi /2, \pi )\) and positive constants \(c, c_1, c_2\) independent of \(\tau \) such that for all \(z\in \Gamma _{\theta ,\sigma }^{\tau }\)

$$\begin{aligned}&c_1|z|\le |\delta _{\tau }(e^{-z\tau })|\le c_2|z|,\quad \delta _{\tau }(e^{-z\tau })\in \Sigma _{\theta ^{\prime }},\\&|\delta _{\tau }(e^{-z\tau })-z|\le c\tau |z|^2,\quad |\delta _{\tau }(e^{-z\tau })^{\alpha }-z^{\alpha }|\le c\tau z^{\alpha +1}. \end{aligned}$$

Deduction of fully discrete solution (6.4) Denote \(w_n={\tilde{U}}_n-\phi _h-t_n\psi _h\), then \(w_0=0\). Multiplying \(\xi ^n\) on both sides of the equation in the fully discrete problem (6.3) and summing over n from 1 to \(\infty \), there is

$$\begin{aligned} \sum _{n=0}^{\infty }\overline{\partial }_\tau ^{\alpha }w_n\xi ^n-\sum _{n=0}^{\infty }L_hw_n\xi ^n&=\sum _{n=1}^{\infty }L_h(\phi _h+t_n\psi _h)\xi ^n\nonumber \\&\quad +{\tilde{f}}_{h,\tau }\sum _{n=1}^{\infty }(p(0)+t_np^{\prime }(0)+\overline{\partial }_\tau ^{(-2)} p^{\prime \prime }(t_n))\xi ^n. \end{aligned}$$
(8.1)

In terms of the convolution quadrature (6.1), we have

$$\begin{aligned} \sum _{n=0}^{\infty }\overline{\partial }_\tau ^{\alpha }w_n\xi ^n=\sum _{n=0}^{\infty }\sum _{j=0}^{n}\omega _j^{(\alpha )}w_{n-j}\xi ^n=\sum _{j=0}^{\infty }\omega _j^{(\alpha )}\xi ^{j}\sum _{n=0}^{\infty }w_n\xi ^n=\delta _{\tau }(\xi )^\alpha \sum _{n=0}^{\infty }w_n\xi ^n, \end{aligned}$$

and similarly by the definition of \(\overline{\partial }_\tau ^{(-2)}\), we have

$$\begin{aligned} \sum _{n=1}^{\infty }\overline{\partial }_\tau ^{(-2)} p^{\prime \prime }(t_n)\xi ^n=\sum _{n=0}^{\infty }\overline{\partial }_\tau ^{(-2)} p^{\prime \prime }(t_n)\xi ^n-{\tilde{\omega }}_0 p^{\prime \prime }(0)=\delta _{\tau }(\xi )^{-2}\sum _{n=0}^{\infty }p^{\prime \prime }(t_n)\xi ^n-\tilde{\omega }_0 p^{\prime \prime }(0). \end{aligned}$$

Notice that \(\sum _{n=1}^{\infty }\xi ^n=\frac{\xi }{1-\xi }\) and \(\sum _{n=1}^{\infty }t_n \xi ^n=\frac{\tau \xi }{(1-\xi )^2}\), by (8.1), it implies

$$\begin{aligned} (\delta _{\tau }(\xi )^\alpha -L_h)\sum _{n=0}^{\infty }w_n\xi ^n&=\frac{\xi }{1-\xi }L_h\phi _h+\frac{\tau \xi }{(1-\xi )^2}L_h\psi _h +[p(0)\frac{\xi }{1-\xi }+p^{\prime }(0)\frac{\tau \xi }{(1-\xi )^2} \nonumber \\&\quad +\delta _{\tau }(\xi )^{-2}\sum _{n=0}^{\infty }p^{\prime \prime }(t_n)\xi ^n-\tilde{\omega }_0 p^{\prime \prime }(0)]{\tilde{f}}_{h,\tau }. \end{aligned}$$
(8.2)

In view of the relationship of \(w_n\) and \({{\tilde{U}}}_n\), then \({\tilde{U}}_n\) satisfies

$$\begin{aligned}&(\delta _{\tau }(\xi )^\alpha -L_h)\sum _{n=0}^{\infty }{\tilde{U}}_n\xi ^n\nonumber \\&\quad =\delta _{\tau }(\xi )^\alpha \frac{\xi }{1-\xi }\phi _h+\delta _{\tau }(\xi )^\alpha \frac{\tau \xi }{(1-\xi )^2}\psi _h+(\delta _{\tau }(\xi )^\alpha -L_h)\phi _h\nonumber \\&\qquad +\left( \frac{\xi }{1-\xi }p(0)+ \frac{\tau \xi }{(1-\xi )^2}p^{\prime }(0)+\delta _{\tau }(\xi )^{-2}\sum _{n=0}^{\infty }p^{\prime \prime }(t_n)\xi ^n-\tilde{\omega }_0 p^{\prime \prime }(0)\right) {\tilde{f}}_{h,\tau } :={\tilde{W}}(\xi ). \end{aligned}$$
(8.3)

By the Cauchy integral formula, we have

$$\begin{aligned} {\tilde{U}}_n=\frac{1}{2\pi i}\int _{|\xi |=\rho }(\delta _{\tau }(\xi )^\alpha -L_h)^{-1}{\tilde{W}}(\xi )\xi ^{-n-1}d\xi , \end{aligned}$$
(8.4)

where \(1>\rho >0\) is a small constant.

According to the Cauchy integral theorem, we know \(\frac{1}{2\pi i}\int _{|\xi |=\rho } k(\xi )d\xi =0\) for analytic function \(k(\xi )\) over \(|\xi |\le \rho \), then we have

$$\begin{aligned}&\frac{1}{2\pi i}\int _{|\xi |=\rho }(\delta _{\tau }(\xi )^\alpha -L_h)^{-1}\delta _{\tau }(\xi )^{-2}\sum _{j=0}^{\infty }p^{\prime \prime }(t_j)\xi ^j\xi ^{-n-1}d\xi \\&\quad = \frac{1}{2\pi i}\sum _{j=0}^{n}\int _{|\xi |=\rho }(\delta _{\tau }(\xi )^\alpha -L_h)^{-1}\delta _{\tau }(\xi )^{-2}p^{\prime \prime }(t_j)\xi ^{j-n-1}d\xi , \end{aligned}$$

since for small enough \(\rho \) and \(j\ge n+1\), \((\delta _{\tau }(\xi )^\alpha -L_h)^{-1}\delta _{\tau }(\xi )^{-2}\xi ^{j-n-1}\) is analytic over the disk \(|\xi |\le \rho \), refer to [32]. It is well-known that the following holds

$$\begin{aligned} \frac{1}{2\pi i}\int _{|\xi |=\rho }\xi ^{-n-1}d\xi \phi _h=0,\quad n\ge 1. \end{aligned}$$

Then we can denote

$$\begin{aligned} W(\xi )={\tilde{W}}(\xi )-(\delta _{\tau }(\xi )^\alpha -L_h)\phi _h. \end{aligned}$$
(8.5)

By the transform \(\xi =e^{-z\tau }\) and the Cauchy integral theorem, we have

$$\begin{aligned} {\tilde{U}}_n=\frac{1}{2\pi i}\int _{\Gamma _{\theta ,\delta }^{\tau }}e^{zt_n}(\delta _{\tau }(e^{-z\tau })^\alpha -L_h)^{-1}W(e^{-z\tau })\tau dz. \end{aligned}$$
(8.6)

Therefore, the solution of the fully discrete problem can be obtained easily as (6.4).

Proof of Lemma 6.2

Proof

In terms of (6.5), one knows

$$\begin{aligned} F_j^{h,\tau }(T)&=\frac{1}{2\pi i}\int _{\Gamma ^{\tau }_{\theta ,\sigma }} e^{zT}\delta _{\tau }(e^{-z\tau })^{-1}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _n)^{-1}e^{-z\tau }dzp(0)\nonumber \\&\quad +\frac{1}{2\pi i}\int _{\Gamma ^{\tau }_{\theta ,\sigma }} e^{zT}\delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _n)^{-1}e^{-z\tau }dzp^{\prime }(0)\nonumber \\&\quad +\frac{\tau }{2\pi i}\sum _{k=1}^{N}\int _{\Gamma ^{\tau }_{\theta ,\sigma }} e^{zt_{N-k}}\delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _n)^{-1}dzp^{\prime \prime }(t_k)\nonumber \\&\quad +\frac{\tau }{2\pi i}\int _{\Gamma ^{\tau }_{\theta ,\sigma }} e^{zT}\delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _n)^{-1}dzp^{\prime \prime }(0)\nonumber \\&\quad -\frac{\tau \tilde{\omega }_0}{2\pi i}\int _{\Gamma ^{\tau }_{\theta ,\sigma }} e^{zT}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _n)^{-1}dzp^{\prime \prime }(0)\nonumber \\&=:I_1^{h,\tau }+I_2^{h,\tau }+I_3^{h,\tau }+I_4^{h,\tau }+I_5^{h,\tau }. \end{aligned}$$
(8.7)

Due to (6.9), it follows

$$\begin{aligned} F_j^h(T)&=\frac{1}{2\pi i}\int _{\Gamma _{\theta ,\sigma }} e^{zT}z^{-1}(z^\alpha +\lambda _j^h)^{-1}dzp(0)+\frac{1}{2\pi i}\int _{\Gamma _{\theta ,\sigma }} e^{zT}z^{-2}(z^\alpha +\lambda _j^h)^{-1}dzp^{\prime }(0)\nonumber \\&\quad +\frac{1}{2\pi i}\int _0^{T} \int _{\Gamma _{\theta ,\sigma }} e^{z t_{N-s}}z^{-2}(z^\alpha +\lambda _j^h)^{-1}p^{\prime \prime }(s)ds\nonumber \\&=:I_1^h+I_2^h+I_3^h. \end{aligned}$$
(8.8)

In the following, we try to prove

$$\begin{aligned} \lambda _j^h|F_j^{h,\tau }(t_n)-F_j^h(t_n)|&\le \lambda _j^h|I_1^{h,\tau }-I_1^h|+\lambda _j^h|I_2^{h,\tau }-I_2^h|+\lambda _j^h|I_3^{h,\tau }-I_3^h|\\&\quad +\lambda _j^h|I_4^{h,\tau }|+\lambda _j^h|I_5^{h,\tau }|\le C\tau . \end{aligned}$$

Firstly, it is easy to deduce

$$\begin{aligned}&\lambda _j^h(I_1^h-I_1^{h,\tau })\\&\quad =\frac{1}{2\pi i}\int _{\Gamma _{\theta ,\sigma }\backslash \Gamma ^{\tau }_{\theta ,\sigma }} e^{zT}\lambda _j^hz^{-1}(z^\alpha +\lambda _j^h)^{-1}dzp(0)\\&\qquad +\frac{1}{2\pi i}\int _{\Gamma ^{\tau }_{\theta ,\sigma }} e^{zT}\lambda _j^h[z^{-1}(z^\alpha +\lambda _j^h)^{-1}- \delta _{\tau }(e^{-z\tau })^{-1}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}]dzp(0)\\&\qquad +\frac{1}{2\pi i}\int _{\Gamma ^{\tau }_{\theta ,\sigma }} e^{zT}\lambda _j^h\delta _{\tau }(e^{-z\tau })^{-1}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}(1-e^{-z\tau })dzp(0)\\&\quad =:e_{11}+e_{12}+e_{13}. \end{aligned}$$

According to Lemma 8.1, it can be proved that

$$\begin{aligned} | \lambda _j^h(z^\alpha +\lambda _j^h)^{-1}|=|(\lambda _j^h+z^\alpha -z^\alpha )(z^\alpha +\lambda _j^h)^{-1}|=|1-z^\alpha (z^\alpha +\lambda _j^h)^{-1}|\le 2. \end{aligned}$$

Hence, by simple calculations, we can obtain

$$\begin{aligned} |e_{11}|&\le \frac{|p(0)|}{\pi }\int _{\Gamma _{\theta ,\sigma }\backslash \Gamma ^{\tau }_{\theta ,\sigma }}|e^{zT}z^{-1}||dz| \le C\int _{\frac{\pi }{\tau \sin {\theta }}}^{\infty }e^{\rho \cos {\theta }T}\rho ^{-1} d\rho \le C\tau \int _{0}^{\infty }e^{\rho \cos {\theta }T}d\rho \\&\le C\tau , \end{aligned}$$

where \(C>0\) depends on \(\Vert p\Vert _{C[0,T]}\).

By Lemma 8.2, it is easy to prove \(|z^{-1}-\delta _{\tau }(e^{-z\tau })^{-1}|\le C\tau \) and

$$\begin{aligned} |z^{\alpha -1}(z^\alpha +\lambda _j^h)^{-1}-\delta _{\tau }(e^{-z\tau })^{\alpha -1}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}|\le C\tau . \end{aligned}$$

Then by taking \(\sigma =T^{-1}\), noting that \(\cos \theta <0\), one can derive at

$$\begin{aligned} |e_{12}|&\le \frac{|p(0)|}{2\pi }\int _{\Gamma _{\theta ,\sigma }^{\tau }}|e^{zT}(z^{-1}-z^{\alpha -1}(z^\alpha +\lambda _j^h)^{-1} -\delta _{\tau }(e^{-z\tau })^{-1}\\&\quad +\delta _{\tau }(e^{-z\tau })^{\alpha -1}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}||dz|\\&\le C\tau \int _{\Gamma ^{\tau }_{\theta ,\sigma }} |e^{zT}||dz| \le C\tau \left( \int _{\sigma }^{\frac{\pi }{ \tau \sin {\theta }}}e^{\rho {\textit{cos}}{\theta }T}d\rho + \int _{-\theta }^{\theta }e^{\sigma {\textit{cos}}{\varphi }T}\sigma d\varphi \right) \\&\le C\tau \left( T^{-1}\int _{1}^{\infty }e^{\cos {\theta }y}dy+ T^{-1}\int _{-\theta }^{\theta }e^{ \cos {\varphi }}d\varphi \right) \le C\tau . \end{aligned}$$

Further, we have

$$\begin{aligned} |e_{13}|&\le \frac{|p(0)|}{2\pi }\int _{\Gamma ^{\tau }_{\theta ,\sigma }} |e^{zT}\lambda _j^h(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}\tau ||dz| \le C\tau \int _{\Gamma ^{\tau }_{\theta ,\sigma }} |e^{zT}||dz|\\&\le C\tau . \end{aligned}$$

Therefore, we have \(\lambda _j^h|I_1^h-I_1^{h,\tau }|\le C\tau \) where \(C>0\) depending on \(\Vert p\Vert _{C[0,T]}\).

Secondly, we estimate \(\lambda _j^h(I_2^h-I_2^{h,\tau })\), it can be divided into

$$\begin{aligned}&\lambda _j^h(I_2^h-I_2^{h,\tau }) \\&\quad =\frac{1}{2\pi i}\int _{\Gamma _{\theta ,\sigma }\backslash \Gamma ^{\tau }_{\theta ,\sigma }} e^{zT}\lambda _j^hz^{-2}(z^\alpha +\lambda _j^h)^{-1}dz p^{\prime }(0)\\&\qquad +\frac{1}{2\pi i}\int _{\Gamma ^{\tau }_{\theta ,\sigma }} e^{zT}\lambda _j^h[z^{-2}(z^\alpha +\lambda _j^h)^{-1}dz- \delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}]dz p^{\prime }(0)\\&\qquad +\frac{1}{2\pi i}\int _{\Gamma ^{\tau }_{\theta ,\sigma }} e^{zT}\lambda _j^h\delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}(1-e^{-z\tau })dzp^{\prime }(0)\\&\quad =:e_{21}+e_{22}+e_{23}. \end{aligned}$$

Each term is estimated respectively as follows

$$\begin{aligned} |e_{21}|&\le \frac{|p^{\prime }(0)|}{2\pi }\int _{\Gamma _{\theta ,\sigma }\backslash \Gamma ^{\tau }_{\theta ,\sigma }} |e^{zT}z^{-2}||dz| \le C\int _{\frac{\pi }{\tau \sin {\theta }}}^{\infty }e^{\rho \cos {\theta }T}\rho ^{-2} d\rho \le C\tau ^2\int _0^{\infty }e^{\rho \cos {\theta }T}d\rho \\&\le C\tau ^2. \end{aligned}$$

And by Lemma 8.2, it is easy to prove \(|z^{-2}-\delta _{\tau }(e^{-z\tau })^{-2}|\le C\tau |z|^{-1}\) and

$$\begin{aligned} |z^{\alpha -2}(z^\alpha +\lambda _j^h)^{-1}-\delta _{\tau }(e^{-z\tau })^{\alpha -2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}|\le C\tau |z|^{-1}, \end{aligned}$$

from which we have

$$\begin{aligned} |e_{22}|&=\frac{|p^{\prime }(0)|}{2\pi }\int _{\Gamma ^{\tau }_{\theta ,\sigma }} |e^{zT}(z^{-2}-z^{\alpha -2}(z^\alpha +\lambda _j^h)^{-1} -\delta _{\tau }(e^{-z\tau })^{-2}\\&\quad +\delta _{\tau }(e^{-z\tau })^{\alpha -2} (\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1})||dz|\\&\le C\tau \int _{\Gamma ^{\tau }_{\theta ,\sigma }} |e^{zT}z^{-1}||dz| \le C\tau \left( \int _{\sigma }^{\frac{\pi }{\tau \sin {\theta }}}e^{\rho {\textit{cos}}{\theta }T}\rho ^{-1}d\rho + \int _{-\theta }^{\theta }e^{\sigma {\textit{cos}}{\varphi }T} d\varphi \right) \\ {}&\le C\tau \left( \int _{1}^{\infty }e^{{\textit{cos}}{\theta }y}y^{-1}dy+\int _{-\theta }^{\theta }e^{ {\textit{cos}}{\varphi }}d\varphi \right) \le C\tau , \end{aligned}$$

and note that \(1-e^{-\tau z}=\delta _{\tau }(e^{-z\tau })\tau \), we have

$$\begin{aligned} |e_{23}|&\le |\frac{|p^{\prime }(0)|}{2\pi }\int _{\Gamma ^{\tau }_{\theta ,\sigma }} |e^{zT}\lambda _j^h\delta _{\tau }(e^{-z\tau })^{-1}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}\tau ||dz| \le C\tau \int _{\Gamma ^{\tau }_{\theta ,\sigma }} |e^{zT}z^{-1}||dz|\\&\le C\tau . \end{aligned}$$

Therefore, we have \(\lambda _j^h|I_2^h-I_2^{h,\tau }|\le C\tau \) where \(C>0\) depending on \(\Vert p'\Vert _{C[0,T]}\).

Thirdly, we estimate \(\lambda _j^h(I_3^h-I_3^{h,\tau })\) as follow.

$$\begin{aligned}&\lambda _j^h(I_3^h-I_3^{h,\tau })\\&\quad =\frac{1}{2\pi i}\int _0^T\int _{\Gamma _{\theta ,\sigma }} e^{z(T-s)}\lambda _j^hz^{-2}(z^\alpha +\lambda _j^h)^{-1}dz p^{\prime \prime }(s)ds\\&\qquad -\frac{1}{2\pi i}\sum _{k=1}^{N}\int _{\Gamma _{\theta ,\sigma }^{\tau }} \tau e^{z t_{N-k}}\lambda _j^h\delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}dz p^{\prime \prime }(t_k)\\&\quad =\frac{1}{2\pi i}\int _0^T\int _{\Gamma _{\theta ,\sigma }\backslash \Gamma _{\theta ,\sigma }^{\tau }} e^{z(T-s)}\lambda _j^hz^{-2}(z^\alpha +\lambda _j^h)^{-1}dz p^{\prime \prime }(s)ds\\&\qquad +\frac{1}{2\pi i}\int _0^T\int _{\Gamma _{\theta ,\sigma }^{\tau }} e^{z(T-s)}\lambda _j^h[z^{-2}(z^\alpha +\lambda _j^h)^{-1}{-}\delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha {+}\lambda _j^h)^{-1}]dz p^{\prime \prime }(s)ds\\&\qquad +\frac{1}{2\pi i}\int _0^T\int _{\Gamma _{\theta ,\sigma }^{\tau }} e^{z(T-s)}\lambda _j^h\delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}dz p^{\prime \prime }(s)ds \\&\qquad -\frac{1}{2\pi i}\sum _{k=1}^{N}\int _{\Gamma _{\theta ,\sigma }^{\tau }} \tau e^{zt_{N-k}}\lambda _j^h\delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}dz p^{\prime \prime }(t_k)\\&\quad =:e_{31}+e_{32}+e_{33}, \end{aligned}$$

where \(e_{33}\) contains the last two terms. Then we analyse each term as follows.

$$\begin{aligned} |e_{31}|&\le \frac{1}{2\pi }\int _0^T\int _{\frac{\pi }{\tau \sin {\theta }}}^{\infty }e^{\rho \cos {\theta }(T-s)}\rho ^{-2}d\rho |p^{\prime \prime }(s)|ds \le C\int _0^T\int _{\frac{\pi }{\tau \sin {\theta }}}^{\infty }\rho ^{-2}d\rho |p^{\prime \prime }(s)|ds\\&\le C\tau \Vert p^{\prime \prime }(t)\Vert _{L^1(0,T)}, \end{aligned}$$

and for estimating \(e_{32}\), we take \(\sigma =(T-s)^{-1}\), then we have

$$\begin{aligned} |e_{32}|&\le \frac{1}{2\pi }\int _0^T\int _{\Gamma _{\theta ,\sigma }^{\tau }} |e^{z(T-s)}[z^{-2}-z^{\alpha -2}(z^\alpha +\lambda _j^h)^{-1}\\&\quad -\delta _{\tau }(e^{-z\tau })^{-2} +\delta _{\tau }(e^{-z\tau })^{\alpha -2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}]||dz||p^{\prime \prime }(s)|ds\\&\le C\tau \int _0^T\int _{\Gamma _{\theta ,\sigma }^{\tau }} |e^{z(T-s)}z^{-1}||dz||p^{\prime \prime }(s)|ds \le C\tau \int _0^T\bigg (\int _\sigma ^{\frac{\pi }{\tau \sin {\theta }}}e^{\rho \cos {\theta }(T-s)}\rho ^{-1}d\rho \\&\quad +\int _{-\theta }^\theta e^{\sigma \cos {\varphi }(T-s)}d\varphi \bigg )|p^{\prime \prime }(s)|ds\\&\le C\tau \int _0^T\left( \int _1^{\infty }e^{\cos {\theta }y}y^{-1}dy+\int _{-\theta }^\theta e^{\cos {\varphi }}d\varphi \right) |p^{\prime \prime }(s)|ds \le C\tau \Vert p^{\prime \prime }(t)\Vert _{L^1(0,T)}. \end{aligned}$$

As for \(e_{33}\), we have

$$\begin{aligned} e_{33}&=\frac{1}{2\pi i}\sum _{k=1}^{N}\int _{t_{k-1}}^{t_k}\int _{\Gamma _{\theta ,\sigma }^{\tau }} e^{z(T-s)}\lambda _j^h\delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}dz p^{\prime \prime }(s)ds\\&\quad -\frac{1}{2\pi i}\sum _{k=1}^{N}\int _{\Gamma _{\theta ,\sigma }^{\tau }} \tau e^{zt_{N-k}}\lambda _j^h\delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}dz p^{\prime \prime }(t_k), \end{aligned}$$

noting that \(e^{-zs}p^{\prime \prime }(s)=e^{-zt_k}p^{\prime \prime }(t_k)-\int _{s}^{t_k}e^{-z\xi }(-zp^{\prime \prime }(\xi )+p^{\prime \prime \prime }(\xi ))d\xi \) for \(s\in [t_{k-1},t_k]\), and substituting it into \(e_{33}\), we have

$$\begin{aligned} |e_{33}|&\le \frac{1}{2\pi }\sum _{k=1}^{N}|\int _{t_{k-1}}^{t_k}\int _{\Gamma _{\theta ,\sigma }^{\tau }} \int _{s}^{t_k} e^{z(T-\xi )} \lambda _j^h\delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j^h)^{-1}\\&\quad \times (-zp^{\prime \prime }(\xi )+p^{\prime \prime \prime }(\xi )) d\xi dzds|\\&\le C\sum _{k=1}^{N}\int _{t_{k-1}}^{t_k}\int _{s}^{t_k} \int _{\Gamma _{\theta ,\sigma }^{\tau }} |e^{z(T-\xi )}|(|z|^{-1}|p^{\prime \prime }(\xi )|+|z|^{-2}|p^{\prime \prime \prime }(\xi )|)|dz| d\xi ds\\&\le C\sum _{k=1}^{N}\int _{t_{k-1}}^{t_k}\int _{s}^{t_k} \bigg [\left( \int _{\sigma }^{\frac{\pi }{\tau \sin {\theta }}} e^{\rho \cos {\theta }(T-\xi )}\rho ^{-1}d\rho +\int _{-\theta }^\theta e^{\sigma \cos {\varphi }(T-\xi )}d\varphi \right) |p^{\prime \prime }(\xi )|\\&\quad +\left( \int _{\sigma }^{\frac{\pi }{\tau \sin {\theta }}} e^{\rho \cos {\theta }(T-\xi )}\rho ^{-2}d\rho +\int _{-\theta }^\theta e^{\sigma \cos {\varphi }(T-\xi )}\sigma ^{-1}d\varphi \right) |p^{\prime \prime \prime }(\xi )|\bigg ]d\xi ds, \end{aligned}$$

then by taking \(\sigma =(T-\xi )^{-1}\), \(y=\rho (T-\xi )\), it follows

$$\begin{aligned} |e_{33}|&\le C\sum _{k=1}^{N}\int _{t_{k-1}}^{t_k}\int _{s}^{t_k} \bigg [\left( \int _1^{\infty } e^{\cos {\theta }y}y^{-1}dy+\int _{-\theta }^\theta e^{\cos {\varphi }}d\varphi \right) |p^{\prime \prime }(\xi )|\\&\quad +(T-\xi )\left( \int _1^{\infty } e^{\cos {\theta }y}y^{-2}dy+\int _{-\theta }^\theta e^{\cos {\varphi }}d\varphi \right) |p^{\prime \prime \prime }(\xi )|)\bigg ]d\xi ds\\&\le C\sum _{k=1}^{N}\int _{t_{k-1}}^{t_k}\int _{s}^{t_k}(|p^{\prime \prime }(\xi )|+(T-\xi )|p^{\prime \prime \prime }(\xi )|)d\xi ds\\&=C\sum _{k=1}^{N}\int _{t_{k-1}}^{t_k}\int _{t_{k-1}}^{\xi }(|p^{\prime \prime }(\xi )|+(T-\xi )|p^{\prime \prime \prime }(\xi )|)ds d\xi \\&\le C\sum _{k=1}^{N}\int _{t_{k-1}}^{t_k}(\xi -t_{k-1})(|p^{\prime \prime }(\xi )|+(T-\xi )|p^{\prime \prime \prime }(\xi )|)d\xi \\&\le C\tau (\Vert p^{\prime \prime }(t)\Vert _{L^1(0,T)}+\Vert p^{\prime \prime \prime }(t)\Vert _{L^1(0,T)}). \end{aligned}$$

Therefore, we have \(\lambda _j^h|I_3^h-I_3^{h,\tau }|\le C\tau \) where \(C>0\) depending on \(\Vert p^{\prime \prime }\Vert _{L^1(0,T)}+\Vert p^{\prime \prime \prime }\Vert _{L^1(0,T)}\).

Finally, we deduce the estimates of the last two terms.

$$\begin{aligned} |\lambda _j^hI_4^{h,\tau }|&\le \frac{|p^{\prime \prime }(0)|\tau }{2\pi }\int _{\Gamma ^{\tau }_{\theta ,\sigma }} |e^{zT}\lambda _j^h\delta _{\tau }(e^{-z\tau })^{-2}(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j)^{-1}||dz|\\&\le \frac{|p^{\prime \prime }(0)|\tau }{2\pi }\int _{\Gamma ^{\tau }_{\theta ,\sigma }} |e^{zT}z^{-2}||dz|\\ {}&\le C\tau \left( \int _\sigma ^\infty e^{\rho \cos {\theta }T}\rho ^{-2}d\rho +\int _{-\theta }^\theta e^{\sigma \cos {\varphi }T}\sigma ^{-1}d\varphi \right) \\ {}&\le C\tau \left( T\int _1^\infty e^{\cos {\theta }y}y^{-2}dy +T\int _{-\theta }^\theta e^{\cos {\varphi }}d\varphi \right) \le C\tau , \end{aligned}$$

where \(C>0\) depends on \(\Vert p^{\prime \prime }\Vert _{C[0,T]}\).

It is known the convolution weight \({\tilde{\omega }}_0=\tau ^2\), then we have

$$\begin{aligned} |\lambda _j^hI_5^{h,\tau }|&\le \frac{\tau \tilde{\omega }_0|p^{\prime \prime }(0)|}{2\pi }\int _{\Gamma ^{\tau }_{\theta ,\sigma }} |e^{zT}\lambda _j^h(\delta _{\tau }(e^{-z\tau })^\alpha +\lambda _j)^{-1}||dz| \le C\tau ^3\int _{\Gamma ^{\tau }_{\theta ,\sigma }}\! |e^{zT}||dz| \le C\tau ^3. \end{aligned}$$

Combining the estimates above for each term, one can obtain the following estimate

$$\begin{aligned} |F_j^h(T)-F_j^{h,\tau }(T)|\le \frac{C\tau }{\lambda _j^h}. \end{aligned}$$

By \(F_j^h(T)\ge \frac{C}{\lambda _j^h}\), for a small enough \(\tau \), there exists a positive constant C such that

$$\begin{aligned} F_j^{h,\tau }(T)=F_j^h(T)-F_j^h(T)+F_j^{h,\tau }(T)\ge F_j^h(T)-|F_j^h(T)-F_j^{h,\tau }(T)|\ge \frac{C}{\lambda _j^h}>0. \end{aligned}$$

\(\square \)

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Luo, Y., Wei, T. Uniqueness and Numerical Method for Determining a Spatial Source Term in a Time-Fractional Diffusion Wave Equation. J Sci Comput 99, 51 (2024). https://doi.org/10.1007/s10915-024-02523-3

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1007/s10915-024-02523-3

Keywords

Navigation