Abstract
We consider a pencil of the first-order integro-differential operators with the convolution kernel dependent on the spectral parameter. The inverse problem is studied, which consists in recovering the kernel from the spectrum. We develop a constructive procedure for solution and obtain necessary and sufficient conditions for the solvability of the inverse problem.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction and Main Results
The paper concerns the inverse spectral theory for integro-differential operators. Inverse spectral problems consist in reconstruction of operators, by using their spectral characteristics. The most complete theory of such problems was developed for differential operators (see the monographs [1,2,3,4]). However, integro-differential operators are often more adequate for description of various processes in physics, biology, economics and engineering [5]. Inverse spectral theory for intergo-differential operators has not been sufficiently developed yet. It consists of fragmentary results, not forming a general picture (see [6,7,8,9,10,11,12,13] and references therein).
In this paper, we solve an inverse problem for a pencil of integro-differential operators with a kernel, depending on the spectral parameter. We note that investigation of inverse problems for differential pencils causes principal difficulties, comparing with usual differential operators (see e.g. [14]). Inverse problems for integro-differential equations with coefficients, depending on the spectral parameter, as far as we know, have not been studied before.
We investigate the boundary value problem L for the first-order integro-differential equation
with the boundary condition \(y(0) = y(\pi )\), where the convolution kernel depends linearly on the spectral parameter: \(M(x, \lambda ) = M_0(x) + \lambda M_1(x)\).
Define the class of functions
We assume that the functions \(M_0\) and \(M_1\) are real-valued, the function \(M_1\) is absolutely continuous on \([0, \pi )\), \(M_0\) and \(M_1'\) belong to \(L_{2, \pi }\), and \(M_1(0) = 0\).
The main results of the paper are Theorems 1 and 2, providing necessary and sufficient conditions on the spectrum of the problem L.
Theorem 1
The boundary value problem L has a countable set of complex eigenvalues, which can be numbered as \(\{ \lambda _n \}_{n \in \mathbb Z}\), counting with their multiplicities, so that the following asymptotic relation holds
Theorem 2
For arbitrary complex numbers \(\{ \lambda _n \}_{n \in {\mathbb {Z}}}\), satisfying the asymptotic relation (2), there exists a unique boundary value problem L in the form described above, such that \(\{ \lambda _n \}_{n \in {\mathbb {Z}}}\) is the spectrum of L.
Moreover, we develop a constructive procedure for solving the following inverse problem.
Inverse Problem 1
Given the spectrum \(\{ \lambda _n \}_{n \in {\mathbb {Z}}}\) of L, construct the functions \(M_0\) and \(M_1\).
Note that the case, when \(M_1 = 0\) and \(M_0\) is complex-valued, has been studied in [15]. In that case, the spectrum \(\{ \lambda _n \}_{n \in {\mathbb {Z}}}\) is also sufficient for recovering L, and the theorem similar to Theorem 2 is valid.
In order to solve Inverse Problem 1 and to prove Theorem 2, we develop an approach of [7, 12]. Our method is based on the reduction of the inverse problem to the system of nonlinear integral Eq. (15), called the main equations. In Sect. 2, we derive the system (15) and prove its unique solvability (Lemma 2). In Sect. 3, we prove the main results and obtain Algorithm 1 for solution of the inverse problem.
2 Main Equations of the Inverse Problem
Denote by \(e(x, \lambda )\) the solution of Eq. (1), satisfying the initial condition \(e(0, \lambda ) = 1\). Obviously, the eigenvalues of L coincide with the zeros of the entire characteristic function \(\Delta (\lambda ) := e(\pi , \lambda ) - e(0, \lambda )\). Introduce the following notations for convolutions
The solution \(e(x, \lambda )\) admits the representation
where
The relations (3) and (4) can be obtained similarly to [7], where Eq. (1) with \(M(x, \lambda ) \equiv M(x)\) has been considered.
Formal calculations show that
where
Below we use the symbol C for various constants, independent of t, \(\lambda \), etc. Denote the functions
Lemma 1
For \(n \ge 0\), the function \(F_n\) belongs to \(W_2^n[0, \pi ]\), and the following estimate holds
Moreover, \(F_n^{(k)}(0) = F_n^{(k)}(\pi ) = 0\), \(0 \le k < n\).
Proof
Rewrite the relation (7) in the form
Note that
Using the latter formulas, we obtain
Since the functions \(M_0\) and \({{\tilde{M}}}_1\) belong to \(L_{2,\pi }\), one can easily show (see e.g. [12]), that for \(n + s \ge 2\), \(\nu = \overline{0, n + s}\), the function \(g_{\nu }(\pi - t) (M_0^{*s} * {{\tilde{M}}}^{*n} * g_{n - 1 - \nu })(t)\) is absolutely continuous on \([0, \pi ]\), and
where [x] is an integer part of x. Consequently, the relation (9) for \(0 \le k < n\) and \(n + s \ge 2\) yields that the functions \(F_{ns}^{(k)}\) are absolutely continuous on \([0, \pi ]\), and \(F_{ns}^{(k)}(0) = F_{ns}^{(k)}(\pi ) = 0\). For \(n + s \ge 2\) the functions \(F_{ns}^{(n)}\) are also absolutely continuous on \([0, \pi ]\) and satisfy the estimate
It remains to consider the two cases, when \(n + s = 1\):
Thus, we have for all \(n, s \ge 0\), \(n + s \ge 1\), that \(F_{ns}^{(n)} \in L_2(0, \pi )\) and
Note that
where \(C_1\) is a constant. Consequently, the series
converge in \(L_2(0, \pi )\). Taking (8) into account, we arrive at the assertion of the Lemma. \(\square \)
In view of Lemma 1, integration by parts yields
The relations (5), (6) and (10) imply
where
Lemma 1 implies \(w \in L_2(0, \pi )\).
Define the following functions for \(n \ge 1\), \(j = \overline{0, n}\):
Substituting (7) into (12), we derive the relation
Considering the real part and the imaginary part of (14) separately, we arrive at the system of two nonlinear integral equations with respect to real functions \(M_0\) and \({{\tilde{M}}}_1\):
where
The following Lemma claims the unique solvability of (15), and plays a crucial role in investigation of Inverse Problem 1.
Lemma 2
The system of main Eq. (15) with the coefficients, defined by (16), has the unique solution \((M_0, \tilde{M}_1)\), \(M_0 \in L_{2, \pi }\), \({{\tilde{M}}}_1 \in L_{2, \pi }\), for any \(w \in L_2(0, \pi )\).
The proof of Lemma 2 is based on Proposition 1, which is a special case of [16, Theorem 3].
Proposition 1
Let \(\psi _{\nu n j}(x)\) and \(\Psi _{\nu n j}(x, t)\), \(\nu = 0, 1\), \(n \in {\mathbb {N}}\), \(j = \overline{0, n}\), be arbitrary functions, square integrable on (0, b) and \({\mathcal {S}} := \{ (x, t) :0< t< x < b \}\), respectively, and satisfying the estimates
for some fixed \(A > 0\) independent of \(\nu \), n and j. Assume that \(\psi _{\nu 1 j}(x) = \delta _{\nu j}\), \(\nu , j = 0, 1\), where \(\delta _{\nu j}\) is the Kronecker delta. Then for every functions \(f_{\nu } \in L_2(0, b)\), \(\nu = 0, 1\), the system (15) has the unique solution \((M_0, {{\tilde{M}}}_1)\), \(M_0 \in L_2(0, b)\), \({{\tilde{M}}}_1 \in L_2(0, b)\).
Proof of Lemma 2
Let w be an arbitrary function from \(L_2(0, \pi )\). Obviously, the functions \(\psi _{\nu n j}\), \(\Psi _{\nu n j}\) and \(f_{\nu }\), \(\nu = 0, 1\), \(n \in {\mathbb {N}}\), \(j = \overline{0, n}\), defined by (16), satisfy the conditions of Proposition 1 for every \(b \in (0, \pi )\). Hence (15) has the unique solution \((M_0, {{\tilde{M}}}_1)\), \(M_0 \in L_2(0, b)\), \({{\tilde{M}}}_1 \in L_2(0, b)\) for every \(b \in (0, \pi )\). It remains to prove that \(M_0 \in L_{2, \pi }\) and \({{\tilde{M}}}_1 \in L_{2, \pi }\).
We represent the functions in the form \(M_0(x) = M_{01}(x) + M_{02}(x)\), \({{\tilde{M}}}_1(x) = M_{11}(x) + M_{12}(x)\), so that \(M_{\nu 1}(x) = 0\) for \(x \in (\pi /2, \pi )\) and \(M_{\nu 2}(x) = 0\) for \(x \in (0, \pi /2)\), \(\nu = 0, 1\). Note that \(M_{\nu _1 2} * M_{\nu _2 2} \equiv 0\) on \((0, \pi )\), \(\nu _k = 0, 1\), \(k = 1, 2\). Consequently, we have
where
Denote
Substituting (18) into (15), we obtain the following relation for \(x > \pi /2\):
Multiplying this relation by \((\pi - x)\), we arrive at the system of linear Volterra integral equations
where
Note that \(f_{\nu } \in L_2(0, \pi )\), \(\Psi _{\nu 1 \xi } \in L_2({\mathcal {T}})\), \(\nu , \xi = 0, 1\), \({\mathcal {T}} := \{ (x, t) :\pi /2< t< x < \pi \}\). Using the estimates
together with (17), we conclude that \(\mu _{\nu } \in L_2(\pi /2, \pi )\), \(K_{\nu \xi } \in L_2({\mathcal {T}})\), \(\nu , \xi = 0, 1\). Consequently, the Volterra integral Eq. (19) has the unique solution \((z_0, z_1)\), \(z_{\nu } \in L_2(0, \pi )\), \(\nu = 0, 1\), so we arrive at the assertion of the Lemma. \(\square \)
3 Proofs of the Main Results
In this section, we prove Theorems 1 and 2, and also provide an algorithm for solving Inverse Problem 1.
Proof of Theorem 1
Using (11), we obtain the following relation for the characteristic function:
Applying to (20) the standard technique (see [4, Theorem 1.1.3]), based on Rouché‘s theorem, we derive the asymptotic relations (2) for the zeros of \(\Delta (\lambda )\). \(\square \)
Relying on (2) and (20), one can prove the following Proposition similarly to [11, Lemmas 1 and 2].
Proposition 2
The characteristic function is uniquely determined by its zeros by the formula
For arbitrary complex numbers \(\{\lambda _n\}_{n \in {\mathbb {Z}}}\) of the form (2), the function \(\Delta (\lambda )\), determined by (21), has the form (20) with a certain function \(w \in L_2(0, \pi )\).
Proof of Theorem 2
Consider an arbitrary sequence of complex numbers \(\{ \lambda _n \}_{n \in {\mathbb {Z}}}\), satisfying the asymptotic relations (2). Let \(\Delta (\lambda )\) be the functions, constructed by (21). By Proposition 2, \(\Delta (\lambda )\) admits the representation (20) with some function \(w \in L_2(0, \pi )\). Define the functions \(f_{\nu }\), \(\psi _{\nu n j}\) and \(\Psi _{\nu n j}\) for \(\nu = 0, 1\), \(n \in {\mathbb {N}}\), \(j = \overline{0, n}\), by (16). Then, by Lemma 2, the main Eq. (15) have the unique solution \((M_0, {{\tilde{M}}}_1)\), \(M_0 \in L_{2, \pi }\), \(\tilde{M}_1 \in L_{2, \pi }\). Define \(M_1(x) := \int _0^x {{\tilde{M}}}_1(t) \, dt\), and conisder the boundary value problem L with the kernel \(M(x, \lambda ) = M_0(x) + \lambda M_1(x)\), constructed by the found functions. By necessity, the characteristic function of L has the form (20) with the function w, satisfying the relation (14), equivalent to the system of the main Eq. (15). Thus, the characteristic function of L coincides with the function \(\Delta (\lambda )\), constructed by the given numbers \(\{ \lambda _n \}_{n \in {\mathbb {Z}}}\). Hence the spectrum of L coincides with \(\{ \lambda _n \}_{n \in {\mathbb {Z}}}\). \(\square \)
The proof of Theorem 2 leads to the following algorithm for solving Inverse Problem 1.
Algorithm 1
Let the complex numbers \(\{ \lambda _n \}_{n \in {\mathbb {Z}}}\) be given.
- 1.
Construct the function \(\Delta (\lambda )\) as an infinite product by (21).
- 2.
Find the function w(t), inverting the Fourier transform (20) by the formula
$$\begin{aligned} w(t) = \frac{1}{\pi } \sum _{n = -\infty }^{\infty } \Delta (2 n) \exp (- 2 i n t). \end{aligned}$$ - 3.
Construct the functions \(\varphi _{nj}(x)\), \(\Phi _{nj}(x, t)\), \(n \in {\mathbb {N}}\), \(j = \overline{0, n}\), using (13), and then \(f_{\nu }(x)\), \(\psi _{\nu n j}(x)\), \(\Psi _{\nu n j}(x, t)\), \(\nu = 0, 1\), \(n \in {\mathbb {N}}\), \(j = \overline{0, n}\), using (16).
- 4.
Find the functions \(M_0(x)\) and \({{\tilde{M}}}_1(x)\) as the solution of the main Eq. (15), put \(M_1(x) := \int _0^x {{\tilde{M}}}_1(t) \, dt\).
References
Marchenko, V.A.: Sturm–Liouville Operators and Their Applications. Naukova Dumka, Kiev (1977) (Russian); English transl., Birkhauser (1986)
Levitan, B.M.: Inverse Sturm–Liouville Problems. Nauka, Moscow (1984) (Russian); English transl. VNU Science Press, Utrecht (1987)
Pöschel, J., Trubowitz, E.: Inverse Spectral Theory. Academic Press, New York (1987)
Freiling, G., Yurko, V.: Inverse Sturm–Liouville Problems and Their Applications. Nova Science Publishers, Huntington, NY (2001)
Lakshmikantham, V., Rama Mohana Rao, M.: Theory of Integro-Differential Equations. Stability and Control: Theory, Methods and Applications, vol. 1. Gordon and Breach Science Publishers, Singapore (1995)
Yurko, V.A.: An inverse problem for integro-differential operators. Math. Notes 50(5–6), 1188–1197 (1991)
Buterin, S.A.: On an inverse spectral problem for a convolution integro-differential operator. Results Math. 50(3–4), 73–181 (2007)
Kuryshova, Y.: An inverse spectral problem for differential operators with integral delay. Tamkang J. Math. 42(3), 295–303 (2011)
Wang, Y., Wei, G.: The uniqueness for Sturm–Liouville problems with after effect. Acta Math. Sci. 32A(6), 1171–1178 (2012)
Buterin, S.A., Choque Rivero, A.E.: On inverse problem for a convolution integrodifferential operator with Robin boundary conditions. Appl. Math. Lett. 48, 150–155 (2015)
Bondarenko, N., Buterin, S.: On recovering the Dirac operator with an integral delay from the spectrum. Results Math. 71(3–4), 1521–1529 (2017)
Bondarenko, N.P., Buterin, S.A.: An inverse spectral problem for integro-differential Dirac operators with general convolution kernels. Applica. Anal. (2018). https://doi.org/10.1080/00036811.2018.1508653
Ignatyev, M.: On an inverse spectral problem for the convolution integro-differential operator of fractional order. Results Math. 73, 34 (2018). https://doi.org/10.1007/s00025-018-0800-2
Yurko, V.A.: An inverse problem for pencils of differential operators. Matem. Sbornik 191, 137–160 (2000)
Buterin, S.A.: On an inverse spectral problem for first-order integro-differential operators with discontinuities. Appl. Math. Lett. 78, 65–71 (2018)
Buterin, S., Malyugina, M.: On global solvability and uniform stability of one nonlinear integral equation. Results Math. 73, 117 (2018)
Acknowledgements
This work was supported by Grant 17-11-01193 of the Russian Science Foundation.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Bondarenko, N.P. An Inverse Problem for an Integro-Differential Equation with a Convolution Kernel Dependent on the Spectral Parameter. Results Math 74, 148 (2019). https://doi.org/10.1007/s00025-019-1073-0
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00025-019-1073-0
Keywords
- Pencil of integro-differential operators
- inverse spectral problems
- nonlinear integral equations
- necessary and sufficient conditions