![MathJax Logo](/templates/jsp/_style2/_tandf/pb2/images/math-jax.gif)
ABSTRACT
The Dirac system with an integral delay in the form of convolution is considered. We suppose that the convolution kernel, having four independent components, is known a priori on a part of the interval. A partial inverse problem is studied, which consists in recovering the kernel on the remaining part of the interval from some fractional parts of two spectra. Uniqueness theorems are proved, and a constructive algorithm is developed for solving this inverse problem.
1. Introduction and main results
The paper concerns the inverse spectral theory for integro-differential operators. Operators of this kind have many applications in science and engineering, in particular, in biological population models, in the theory of nuclear reactors, in the elasticity theory (see [Citation1,Citation2]). Integro-differential operators are nonlocal, and therefore they are often more adequate for description of physical processes, than differential operators. However, the spectral theory for integro-differential operators has not been completely constructed yet, and contains only a few fragmentary results.
Inverse problems of spectral analysis consist in recovering operators from their spectral characteristics. For differential operators, there exist powerful methods, that reduce nonlinear inverse spectral problems to some linear equations (e.g. Gelfand-Levitan method, the method of spectral mappings, see [Citation3–5]). For integro-differential operators, those methods do not work because of nonlocality. In recent years, the new method has been developed for solving inverse problems for integral operators [Citation6] and, later on, for integro-differential operators [Citation7–15]. This method reduces inverse spectral problems to uniquely solvable nonlinear integral equations, and allows to prove uniqueness theorems, obtain algorithms for solution together with necessary and sufficient conditions for solvability of inverse problems. In particular, this method yielded the results for the Sturm-Liouville (second order) integro-differential operators [Citation7,Citation8], Dirac integro-differential systems [Citation9–11], first-order operators with discontinuities [Citation12], integro-differential operators of arbitrary order [Citation13] and fractional order [Citation14,Citation15]. Some other issues of inverse problem theory for integro-differential operators were studied in [Citation16–21]. However, this topic still contains many open questions.
This paper is devoted to the Dirac system with the integral delay in the form of convolution:
(1)
(1) where
the functions
,
, belong to
and are complex-valued. For j=1, 2, denote by
the boundary value problems for the system (Equation1
(1)
(1) ) with the boundary conditions
(2)
(2) The spectra of these boundary value problems are described by the following lemma.
Lemma 1.1
[Citation9,Citation10]
The spectra of
are countable sets of complex eigenvalues, which can be numbered as
counting with their multiplicities
so that
In [Citation10] the inverse problem has been studied, that consists in recovering the kernel from two spectra
, j=1, 2. In the earlier paper [Citation9], the special symmetric case was considered, when
,
. In that case, the only spectrum
is sufficient for reconstruction of
.
In the present paper, we study the partial inverse problem, formulated below. Let a be a real number, , and let
, j=1, 2, be some subsets of
.
Inverse Problem 1
Suppose that the matrix-function for
is known a priori. Given the subspectra
,
, construct
on
.
Inverse Problem 1 is an analogue of the famous Hochstadt-Lieberman problem, which consists in recovering the potential of the differential Sturm-Liouville operator from the spectrum, while the potential is known a priori on a half of the interval. The Hochstadt-Lieberman problem and its generalizations for differential operators were investigated in [Citation22–29,Citation30]. In the paper [Citation8], the first results has been obtained in this direction for integro-differential operators (namely, for the integro-differential Sturm-Liouville operator). The goals of the present paper are to prove uniqueness theorems and to develop a constructive algorithm for solving Inverse Problem 1 under some assumptions on the sets , j=1, 2.
We proceed to the formulation of the main results. Along with ,
, we consider the boundary value problems
, j=1, 2, of the same form (Equation1
(1)
(1) )–(Equation2
(2)
(2) ), but with a different convolution kernel
. We agree that if a certain symbol γ denotes an object related to
, then
with tilde denotes an analogous object related to
. Define
. Denote by
the complex Hilbert space of vector functions
with the scalar product
Define the sequences of vector functions
(3)
(3) The uniqueness theorem for Inverse Problem 1 is formulated as follows.
Theorem 1.1
Suppose that a.e. on
,
,
, j=1, 2, and the sequences
, j=1, 2, are complete in
. Then
a.e. on
.
Remark 1.1
The numeration of eigenvalues in the sequences and
, j=1, 2, is not uniquely fixed by Lemma 1.1. Therefore the formulation of Theorem 1.1 should be understood in the following sense: suppose that there exists such numeration of the eigenvalues, that the conditions of the theorem hold.
Remark 1.2
The completeness of the sequences implies, that the eigenvalues in the subspectra
are distinct for each fixed j=1, 2. However, our results can be generalized for the case of multiple eigenvalues, by using the approach of [Citation11].
For the case when the sequences , j=1, 2, are Riesz bases in
, we develop a constructive procedure for solving Inverse Problem 1 (see Algorithm 3.1). One can find more information about Riesz bases in [Citation5, Section 1.8.5].
We also prove the following theorem for the special case.
Theorem 1.2
Let ,
,
, and
,
,
. Suppose that
a.e. on
,
,
, j=1, 2, and for j=1, 2, the subspectrum
consists of distinct eigenvalues. Then
a.e. on
. Thus, in this case the solution of Inverse Problem 1 is unique and can be found by using Algorithm 3.1.
Roughly speaking, if the convolution kernel is unknown on the interval of length
, then we need a
th part of each spectrum to solve the inverse problem.
Our approach is based on the so-called main equation, derived in [Citation10]. It is a nonlinear integral equation with respect to the unknown convolution kernels ,
. We show that, if the matrix-function
is known a priori on the interval
,
, the main equation is linear on the remaining interval
. Under the certain conditions on the subspectra
, j=1, 2, this equation can be used for solving the partial Inverse Problem 1.
The paper is organized as follows. In Section 2, we provide the nonlinear main equation and other preliminaries. In Section 3, we prove the uniqueness Theorems 1.1, 1.2 and develop an algorithm for solving the partial inverse problem. In Section 4, we illustrate our method by a numerical example.
2. Preliminaries
In this section, we formulate some results of [Citation10], that will be important for investigation of the partial inverse problem.
Let be the solution of the system (Equation1
(1)
(1) ), satisfying the initial conditions
,
. Here and below T is the transposition sign. Note that
are entire functions of λ for each fixed
, j=1, 2. For j=1, 2, the eigenvalues of the boundary value problem
coincide with the zeros of the characteristic function
, counting with their multiplicities. The characteristic functions admit the representation
(4)
(4) where
,
, and
, l, m=1, 2, are functions from
.
Introduce the notations
Denote by
,
, all possible convolution monomials of the form
(5)
(5)
In [Citation10] the following nonlinear integral main equation with respect to the convolution kernels ,
, has been derived:
(6)
(6) Here
are certain constant coefficients, independent of M and satisfying the estimate
where C is an absolute constant, and
(7)
(7)
It has been shown in [Citation10], that, if ,
, then the functions
and
are absolutely continuous on
and
,
. Furthermore, one can calculate the functions
, l, m=1, 2, by the following formulae:
(8)
(8) where
the coefficients
are certain constants and
3. Proofs of the main results
Suppose that the functions ,
, are known on the subinterval
,
. First we modify the main equation (Equation6
(6)
(6) ) for this case. Represent the functions
in the following form:
(9)
(9) where
on
and
on
,
. Note that
(10)
(10) Substituting (Equation9
(9)
(9) ) into (Equation5
(5)
(5) ) and using (Equation10
(10)
(10) ), we get the following representation for
,
,
:
where
are certain integer coefficients from the interval
and
Then the main equations (Equation6
(6)
(6) ) for
can be rewritten in the form
Thus, we obtain the Volterra integral equation of the second kind
(11)
(11) where
and
are column vector functions and
is the
matrix function with the elements defined as follows:
(12)
(12)
(13)
(13)
One can easily show that
and
, where we suppose that
for
,
. Consequently, Equation (Equation11
(11)
(11) ) has the unique solution z, such that
.
In order to solve Equation (Equation11(11)
(11) ), we have to construct the functions
,
, for
. Using the functions
,
, on
, we can find the functions
, l, m=1, 2, on
,
, by the formulae (Equation8
(8)
(8) ). Note that
, l, m=1, 2. Using (Equation4
(4)
(4) ), we get the relation
Define
(14)
(14) Then
(15)
(15) Therefore one can determine the numbers
,
,
, by (Equation14
(14)
(14) ), and then use them to find the functions
, l, m=1, 2, on
from (Equation15
(15)
(15) ). Now we are ready to prove the uniqueness theorem for Inverse Problem 1.
Proof of Theorem 1.1.
Let boundary value problems L and satisfy the conditions of Theorem 1.1. In particular,
a.e. on
,
. Then, in view of (Equation8
(8)
(8) ),
a.e. on
for l, m=1, 2. Since
,
, j=1, 2, by virtue of (Equation14
(14)
(14) ), we have
,
, j=1, 2. Then it follows from (Equation15
(15)
(15) ), that
Recall that the sequences
, j=1, 2, defined by (Equation3
(3)
(3) ), are complete in
. Hence
a.e. on
, l, m=1, 2. Consequently, the relations (Equation7
(7)
(7) ) imply
a.e. on
for
. Thus, Equation (Equation11
(11)
(11) ) for L coincides with the similar equation for
. By virtue of the unique solvability of (Equation11
(11)
(11) ), we conclude that
a.e. on
,
.
Suppose that , j=1, 2, are Riesz bases in
. Summarizing the above arguments, we arrive at the following algorithm for solving Inverse Problem 1.
Algorithm 3.1
Let the functions ,
,
, and the eigenvalues
, j=1, 2, be given.
Construct the functions
,
,
, by formulae (Equation8
(8)
(8) ).
Find the numbers
, using (Equation14
(14)
(14) ).
Construct the functions
on
by the formula
where
are the biorthonormal bases to
, j=1, 2.
Find the functions
,
,
, using (Equation7
(7)
(7) ).
Find
and
, using (Equation12
(12)
(12) ) and (Equation13
(13)
(13) ).
Solve the Volterra integral equation of the second kind (Equation11
(11)
(11) ), and find
,
,
.
Further in this section we assume ,
,
, and
,
, j=1, 2. The proof of Theorem 1.2 is based on the following lemma.
Lemma 3.1
Under the conditions of Theorem 1.2, the sequences
are Riesz bases in
.
The assertion of Lemma 3.1 easily follows from Lemmas 3.2 and 3.3.
Lemma 3.2
Let be a sequence of distinct complex numbers. Then the sequence of vector functions
is a Riesz basis in
if and only if
is a Riesz basis in
.
Proof.
The sequence is transformed into
by the linear bounded transform
, acting as follows:
The inverse transform is also bounded:
so it preserves the Riesz basis property.
Lemma 3.3
Let
. Then the system of exponents
is a Riesz basis in
.
Proof.
Note that the shift s does not influence the Riesz basis property, since the linear transform is bounded and has a bounded inverse. Thus, it is sufficient to study the system
. This system is
-close to the orthonormal system
. Using the Theorem by Levin and Ljubarskiĭ [Citation31], one can easily show that the system
in complete in
. Consequently, it is a Riesz basis.
Lemma 3.1 implies that the sequences ,
, are complete in
, so the uniqueness assertion of Theorem 1.2 immediately follows from Theorem 1.1. Since
, j=1, 2, are Riesz bases, one can apply Algorithm 3.1 to solve Inverse Problem 1 for the considered case.
4. Example
In this section, we apply Algorithm 3.1 to the following example. Suppose that
(16)
(16)
where
is a given number.
Using formulae (Equation8(8)
(8) ), we get
for
, l, m=1, 2. Therefore the relations (Equation14
(14)
(14) ) and (Equation15
(15)
(15) ) together with (Equation16
(16)
(16) ) imply
(17)
(17)
(18)
(18)
(19)
(19)
In view of (Equation18(18)
(18) ), the Fourier coefficients of the vector functions
with respect to the orthogonal system
in
equal zeros for all
, so we have
Substituting the latter representation into (Equation17
(17)
(17) ) and solving the linear system, we get
,
so
,
,
. Similarly we obtain
,
, j=1, 2, from (Equation19
(19)
(19) ). Then one can easily calculate
,
,
, using (Equation7
(7)
(7) ). Since
, the main equation (Equation11
(11)
(11) ) takes the form
,
,
. Hence we get
(20)
(20) for
. Note that the functions
and
are continuous at
.
Now let us check for a particular value , that the boundary value problems
, j=1, 2, with
,
, and
, given by (Equation20
(20)
(20) ) for
, have eigenvalues in the form (Equation16
(16)
(16) ). For
, we have
,
. For
, the Dirac system (Equation1
(1)
(1) ) for
takes the form
Solving the latter system by the Runge-Kutta method, using Simpson's formula for integration, we find the following real eigenvalues in the interval :
Obviously, the spectra include subsequences in the form (Equation16(16)
(16) ).
Disclosure statement
No potential conflict of interest was reported by the author.
Additional information
Funding
References
- Lakshmikantham V, Rama Mohana Rao M. Theory of integro-differential equations, stability and control: theory, methods and applications. Vol. 1. Singapore: Gordon and Breach Science Publishers; 1995.
- Bažant ZP, Jirásek M. Nonlocal integral formulation of plasticity and damage: survey of progress. ASCE J Eng Mech. 2002;1119–1149. doi: 10.1061/(ASCE)0733-9399(2002)128:11(1119)
- Marchenko VA. Sturm-Liouville operators and their applications. Kiev: Naukova Dumka; 1977. (Russian) English transl., Birkhauser (1986).
- Levitan BM. Inverse Sturm-Liouville problems. Moscow: Nauka; 1984. (Russian); English transl., VNU Sci. Press, Utrecht (1987).
- Freiling G, Yurko V. Inverse Sturm-Liouville problems and their applications. Huntington, NY: Nova Science Publishers; 2001.
- Buterin SA. The invese problem of recovering the Volterra convolution operator from the incomplete spectrum of its rank-one perturbation. Inverse Probl. 2006;22:2223–2236. doi: 10.1088/0266-5611/22/6/019
- Buterin SA. On an inverse spectral problem for a convolution integro-differential operator. Results Math. 2007;50(3–4):73–181. doi: 10.1007/s00025-007-0244-6
- Buterin S, Sat M. On the half inverse spectral problem for an integro-differential operator. Inverse Probl Sci Eng. 2017;25(10):1508–1518. doi: 10.1080/17415977.2016.1267171
- Bondarenko N, Buterin S. On recovering the Dirac operator with an integral delay from the spectrum. Results Math. 2017;71(3–4):1521–1529. doi: 10.1007/s00025-016-0568-1
- Bondarenko NP, Buterin SA. An inverse spectral problem for integro-differential Dirac operators with general convolution kernels. Appl Anal. 2018. Published online. doi:10.1080/00036811.2018.1508653.
- Bondarenko NP. An inverse problem for the integro-differential Dirac system with partial information given on the convolution kernel. J Inverse Ill-Posed Probl. 2018; Published online: 2018-06-19. doi:10.1515/jiip-2017-0058.
- Buterin SA. On an inverse spectral problem for first-order integro-differential operators with discontinuities. Appl Math Lett. 2018;78:65–71. doi: 10.1016/j.aml.2017.11.005
- Buterin SA, Vasiliev SV. On uniqueness of recovering the convolution integro-differential operator from the spectrum of its non-smooth one-dimensional perturbation. Bound Value Probl. 2018;2018:15. doi:10.1186/s13661-018-0974-2.
- Ignatiev M. On an inverse spectral problem for one integro-differential operator of fractional order. J Inverse Ill-Posed Probl. 2018. Published online. doi:10.1515/jiip-2017-0121.
- Ignatyev M. On an inverse spectral problem for the convolution integro-differential operator of fractional order. Results Math. 2018;73:34. doi:10.1007/s00025-018-0800-2
- Eremin MS. An inverse problem for a second-order integro-differential equation with a singularity. Differ Uravn. 1988;24(2):350–351.
- Kuryshova JV. Inverse spectral problem for integro-differential operators. Math Notes. 2007;81(6):767–777. doi: 10.1134/S0001434607050240
- Kuryshova Y. An inverse spectral problem for differential operators with integral delay. Tamkang J Math. 2011;42(3):295–303. doi: 10.5556/j.tkjm.42.2011.295-303
- Wang Y, Wei G. The uniqueness for Sturm-Liouville problems with aftereffect. Acta Math Sci. 2012;32A(6):1171–1178.
- Yurko VA. Inverse problems for second order integro-differential operators. Appl Math Lett. 2017;74:1–6. doi: 10.1016/j.aml.2017.04.013
- Yurko V. Inverse problems for arbitrary order integral and integro-differential operators. Results Math. 2018;73:72. doi:10.1007/s00025-018-0835-4.
- Hochstadt H, Lieberman B. An inverse Sturm-Liouville problem with mixed given data. SIAM J Appl Math. 1978;34:676–680. doi: 10.1137/0134054
- Gesztesy F, Simon B. Inverse spectral analysis with partial information on the potential. II. The case of discrete spectrum. Trans AMS. 2000;352(6):2765–2787. doi: 10.1090/S0002-9947-99-02544-1
- Sakhnovich L. Half-inverse problems on the finite interval. Inverse Probl. 2001;17:527–532. doi: 10.1088/0266-5611/17/3/311
- Martinyuk O, Pivovarchik V. On the Hochstadt-Lieberman theorem. Inverse Probl. 2010;26:035011. (6pp). doi: 10.1088/0266-5611/26/3/035011
- Hryniv RO, Mykytyuk YV. Half-inverse spectral problems for Sturm-Liouville operators with singular potentials. Inverse Probl. 2004;20:1423–1444. doi: 10.1088/0266-5611/20/5/006
- Yang C-F, Huang ZY. A half-inverse problem with eigenparameter dependent boundary conditions. Numer Funct Anal Optim. 2010;31(6):754–762. doi: 10.1080/01630563.2010.490934
- Shieh C-T, Buterin SA, Ignatiev M. On Hochstadt-Liberman theorem for Sturm-Liouville operators. Far East J Appl Math. 2011;52(2):131–146.
- Wei G, Xu H-K. Inverse spectral problem with partial information given on the potential and norming constants. Trans AMS. 2012;364(6):3265–3288. doi: 10.1090/S0002-9947-2011-05545-5
- Wang YP. Inverse problems for discontinuous Sturm-Liouville operators with mixed spectral data. Inverse Probl Sci Eng. 2015;23(7):1180–1198. doi: 10.1080/17415977.2014.981748
- Levin BJ, Ljubarskiĭ JI. Interpolation by means of special classes of entire functions and related expansions in series of exponentials. Math USSR-Izv. 1975;9(3):621–662. doi: 10.1070/IM1975v009n03ABEH001493