Abstract
The inverse problem for the discrete analogue of the transmission eigenvalue problem for absorbing media with a spherically symmetric index of refraction is considered. Some uniqueness results are provided which imply that can be recovered uniquely if only the all transmission eigenvalues (counting with their multiples) are given together with partial information on the entries of .
AMS Subject Classifications:
1. Introduction
Transmission eigenvalue problems have received widespread attention and have become an important area of research in inverse scattering theory, especially in the case of acoustic waves and the electromagnetic waves (see [Citation1–Citation8] and the references therein). The main interest is motivated by the fact that transmission eigenvalues carry information about the material properties of the scattering object and that these eigenvalues can in principle be determined from the scattering data (cf. [Citation3,Citation9]).
The interior transmission problem is a non-selfadjoint boundary-value problem for a pair of fields and in a bounded and simply connected domain of with the sufficiently smooth boundary Let be a ball of radius b in and be a spherically symmetric function corresponding to the square of the refractive index of the medium and be another spherically symmetric function i.e. the medium with support and index of refraction is embedded in a background with index of refraction at location in the electromagnetic case of the reciprocal or the sound speed in the acoustic case [Citation7].
It is straightforward to see that the corresponding boundary value problem is [Citation10]:(1.1) (1.1)
Here and for . When and absorption is not present in either the background or inhomogeneity.
This paper is concerned with the inverse discrete transmission eigenvalue problem for absorbing media, which is defined as follows. Let be a complex-valued function of the discrete variable n and be the discrete version of the operator , defined by(1.2) (1.2)
where The discrete transmission eigenvalue problem for absorbing media can be formulated as(1.3) (1.3)
with the transmission conditions(1.4) (1.4)
Here both the natural numbers M, N are given, denote the wave-function , for all , and , for all denote the potential at n, are the discrete indexes of refractions [Citation11]. The in (Equation1.3(1.3) (1.3) ) is the spectral parameter as usual. Those -values for which there exists a nontrivial solution pair to (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ) are called the transmissions eigenvalues. It will be shown in Section 2 below that the number of the transmission eigenvalues (repeated eigenvalues are allowed) is at most . The eigenvalue problem (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ) can be viewed as the discrete version of the transmission eigenvalue problem (Equation1.1(1.1) (1.1) ) with N difference nodes for the first equation and M difference nodes for the second equation.
Let us mention that, for problem (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ), if M and N become large enough, then the difference operators and defined by (Equation1.2(1.2) (1.2) ) converges to the differential operator Together with the accurate description of Borcea et al. (for details, see [Citation12,Citation13]) for the connection between the discrete problems and the continuous boundary problems, one knows that the transmission eigenvalues of the discrete problem (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ) converge to the eigenvalues of (Equation1.1(1.1) (1.1) ).
In recent years, all of the research on transmission eigenvalue and its inverse problem concentrates mainly on considering the case that and i.e. without absorbing media(see [Citation1,Citation4,Citation8,Citation11,Citation14,Citation17] and the references therein). For (Equation1.1(1.1) (1.1) ), Aktosun et al. [Citation1] and Wei and Xu [Citation17] gave the conditions to determine uniquely from the transmission eigenvalues; On the other hand, for the inverse discrete transmission eigenvalue problem (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ), Papanicolaou and Doumas [Citation11] and the author [Citation16] provided the uniqueness theorem in the situation that and all .
The transmission eigenvalue problem for absorbing media was studied by [Citation5,Citation18–Citation20], Particularly, Cakoni et al. [Citation5] showed that the transmission eigenvalues of (Equation1.1(1.1) (1.1) ) exist and form at most a discrete set. Though nowadays there are only a rather limited number results on the inverse transmission eigenvalue problem for absorbing media in our knowledge (see [Citation10]).
The main purpose of this paper is, for the cases and in Problem (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ), to study the problem whether () can be recovered by all of the transmission eigenvalues provided that () are known a priori. More precisely, we proved that, when , and are uniquely determined by all transmission eigenvalues; When , and are uniquely determined by all transmission eigenvalues as well as one of , or constant known a priori; And when , all transmission eigenvalues together with some partial information on the entries of and can uniquely determine all of and
The technique we use to prove our uniqueness theorems is the extended Euclidean algorithm for polynomial. We should like to express our thanks to the anonymous referee for his/her suggestion to use this technique, which have improved the manuscript better than those originally submitted. In fact, the process to prove our results provides a reconstruction procedure of functions and .
Our paper proceeds as follows. In the next section, we provide some preparatory material, which contains some polynomials and their properties associated with our inverse problem. In Sections 3 and 4, we establish our main results.
2. Preliminaries
In this section, we provide some polynomials and their properties associated with Problem (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ), which will be used later to prove our main results.
We begin by defining polynomials of a complex variable , say which is the unique solution of equation(2.1) (2.1)
with initial conditions(2.2) (2.2)
By straightforward induction, is of degree and formulated as(2.3) (2.3)
for . Here, and the coefficients can be obtained inductively by using the following recursive formula(2.4) (2.4)
Thus has the form(2.5) (2.5)
It should be noted that the coefficients of the first and second terms of are(2.6) (2.6)
and(2.7) (2.7)
respectively. Similarly, for the second equation of (Equation1.3(1.3) (1.3) ), we will refer is the solution of(2.8) (2.8)
with the initial conditions(2.9) (2.9)
Thus(2.10) (2.10)
for .
Lemma 2.1:
The polynomials (in ) and ( and ) are relatively prime.
Proof If we suppose and are not relatively prime, that is, there exists satisfying , in virtue of (Equation2.4(2.4) (2.4) ), then , it follows that for all n, which contradicts in (Equation2.2(2.2) (2.2) ). Similarly, we can prove the polynomials and are relatively prime.
Let us consider the transmission eigenvalues of Problem (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ). If is a nontrivial solution, then (Equation1.3(1.3) (1.3) ) implies that
where are nonzero constants. Thus (Equation1.4(1.4) (1.4) ) can be written as
which is equivalent to(2.11) (2.11)
If we set(2.12) (2.12)
then the transmission eigenvalues of (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ) and the zeros of the polynomial all coincide. It should be noted that the multiplicity of some zeros of may be greater than 1. We refer to the multiplicity of a zero of also as the multiplicity of the transmission eigenvalue and call the characteristic polynomial of (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ).
Substituting (Equation2.3(2.3) (2.3) ) and (Equation2.10(2.10) (2.10) ) into (Equation2.12(2.12) (2.12) ), we get(2.13) (2.13)
It is clear that the degree of is at most , and if , then the leading coefficient of vanishes. When in addition , together with (Equation2.4(2.4) (2.4) ) and (Equation2.8(2.8) (2.8) ), we infer that . Moreover, if , (Equation2.13(2.13) (2.13) ) implies that , which yields that 0 is always a transmission eigenvalue in the case .
Assume and let be the zeros of (repeated roots are allowed), we have(2.14) (2.14)
where(2.15) (2.15)
In particular, when we also have from (Equation2.13(2.13) (2.13) ) and (Equation2.14(2.14) (2.14) )(2.16) (2.16)
3. The case
In this section, we will consider the following inverse problem for : Given all transmission eigenvalues, reconstruct and .
Note that if () are known a priori, then and are known. The first result in this section is presented as the following.
Theorem 3.1:
Let . If () and all transmission eigenvalues of the problem defined by (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ) are known a priori, as well as or is known, then and are uniquely determined.
Proof From (Equation2.12(2.12) (2.12) ) and (Equation2.14(2.14) (2.14) ), we have(3.1) (3.1)
Firstly we prove and are known and are determined uniquely by using the Euclidean algorithm for polynomial.
Note that and are relatively prime, there exist polynomials such that(3.2) (3.2)
and can be computed by the extended Euclidean algorithm for polynomial. Then doing the Euclidean division of polynomial by we have(3.3) (3.3)
where Multiply (Equation3.2(3.2) (3.2) ) by and multiply (Equation3.3(3.3) (3.3) ) by respectively, and add the two results up, we obtain that(3.4) (3.4)
Here Since and one knows that we then have Which together with (Equation2.12(2.12) (2.12) ), yields(3.5) (3.5)
Since and are relatively prime, it implies that and In addition and if one has
The values at 0 of the polynomials give
Which imply that are known and are determined uniquely by(3.6) (3.6)
and(3.7) (3.7)
respectively, where polynomials(3.8) (3.8)
and(3.9) (3.9)
are known a priori since are computed by the above process. Note that , and Now, both sides of (Equation3.6(3.6) (3.6) ) are polynomials of degree . Equating the coefficients of the powers and , we obtain(3.10) (3.10)
and(3.11) (3.11)
Similarly, equating the coefficients of the powers and in (Equation3.7(3.7) (3.7) ), we have(3.12) (3.12)
and(3.13) (3.13)
For notational convenience, we set(3.14) (3.14)
and(3.15) (3.15)
Furthermore, we denote
Thus the unknown can be expressed as(3.16) (3.16)
Using the new notation, (Equation3.10(3.10) (3.10) )–(Equation3.13(3.13) (3.13) ) can be rewritten as(3.17) (3.17) (3.18) (3.18) (3.19) (3.19) (3.20) (3.20)
Multiplying (Equation3.17(3.17) (3.17) ) by and adding up (Equation3.19(3.19) (3.19) ), we obtain(3.21) (3.21)
The fact that (Equation3.21(3.21) (3.21) ) is always true implies that equations (Equation3.17(3.17) (3.17) ) and (Equation3.19(3.19) (3.19) ) are not independent. In the following we discard (Equation3.19(3.19) (3.19) ) and work with (Equation3.17(3.17) (3.17) ).
From (Equation3.17(3.17) (3.17) ), we obtain(3.22) (3.22)
Next we multiply (Equation3.17(3.17) (3.17) ) by , (Equation3.18(3.18) (3.18) ) by . Adding up the two results gives(3.23) (3.23)
Adding up (Equation3.23(3.23) (3.23) ) and (Equation3.20(3.20) (3.20) ) yields(3.24) (3.24)
From (Equation3.22(3.22) (3.22) ) and (Equation3.24(3.24) (3.24) ) we obtain(3.25) (3.25)
If or is known a priori, then we can use (Equation3.25(3.25) (3.25) ) to compute the other. It follows that x is known from (Equation3.22(3.22) (3.22) ), and is also known from (Equation3.16(3.16) (3.16) ). Now (Equation3.6(3.6) (3.6) ) and (Equation3.7(3.7) (3.7) ) imply and are completely known. We continue by observing that (Equation2.4(2.4) (2.4) ) can be written as(3.26) (3.26)
From the above discussion it follows that is completely known. We have from (Equation2.6(2.6) (2.6) ) and (Equation2.7(2.7) (2.7) ),(3.27) (3.27)
are also known. So is completely known, continuing in the same way, are known, and so on. Thus, eventually, all are determined.
Moreover, if and the transmission eigenvalues are known, in virtue of (Equation3.6(3.6) (3.6) ) and (Equation3.7(3.7) (3.7) ), are also known. By a similar argument as Theorem 3.1, we have:
Theorem 3.2:
Let . If () and all transmission eigenvalues of the problem defined by (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ) are known a priori, as well as is known, then and are uniquely determined.
In the following, we consider the case . It should be noted that, in this case, and . In virtue of (Equation3.5(3.5) (3.5) ), one has(3.28) (3.28)
Which together with (Equation3.5(3.5) (3.5) ) yields Since and are relatively prime, we get
and
Thus and are completely known, then we have
Corollary 3.3:
Let . If () and all transmission eigenvalues of the problem defined by (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ) are known a priori, then and are uniquely determined.
4. The case
In this section, we consider the case . On the condition that we set we begin by defining polynomials for as(4.1) (4.1) (4.2) (4.2)
for It is not hard to see that is of degree .
Lemma 4.1:
For , the polynomials and are relatively prime.
Proof This can be justified as follows. If the polynomials and are not relatively prime, then there exists such that , it follows from (Equation4.1(4.1) (4.1) ) and (Equation4.2(4.2) (4.2) ) that , which contradicts Lemma 2.1.
By proceeding inductively, in terms of (Equation2.12(2.12) (2.12) ), (Equation4.1(4.1) (4.1) ) and (Equation4.2(4.2) (4.2) ), it is easy to see that
Lemma 4.2:
If , then(4.3) (4.3)
Theorem 4.3:
Let . If () and all transmission eigenvalues of the problem defined by (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ) are known, as well as and are known, where for is odd and for is even, then and are uniquely determined.
Proof From Lemma 4.2, we obtain(4.4) (4.4)
Here the polynomials and are known if and are known a priori.
Note that and are relatively prime, there exist polynomials such that(4.5) (4.5)
where can be computed by the extended Euclidean algorithm for polynomial. Notice that is known from (Equation2.16(2.16) (2.16) ), that is is known. Then doing the Euclidean division of polynomial by we have(4.6) (4.6)
where Multiply (Equation4.5(4.5) (4.5) ) by and multiply (Equation4.6(4.6) (4.6) ) by respectively, and add the two results up, we have(4.7) (4.7)
Here and . In virtue of (Equation4.3(4.3) (4.3) ), thus we have(4.8) (4.8)
Since and are relatively prime, together with and it implies that and Set(4.9) (4.9)
Which together with (Equation4.8(4.8) (4.8) ) yields Since and are relatively prime, we get
and
Thus and are completely known, and can be written as(4.10) (4.10)
and(4.11) (4.11)
where the coefficients and are known. Denote(4.12) (4.12)
and(4.13) (4.13)
Together with (Equation4.10(4.10) (4.10) ), equating the coefficients of the powers and of , we obtain(4.14) (4.14)
and(4.15) (4.15)
Similarly, equating the coefficients of the powers and of and by using (Equation4.11(4.11) (4.11) ), one has(4.16) (4.16)
and(4.17) (4.17)
From (Equation4.14(4.14) (4.14) ) and (Equation4.16(4.16) (4.16) ), we have(4.18) (4.18)
Furthermore, (Equation4.14(4.14) (4.14) ) and (Equation4.15(4.15) (4.15) ) yield(4.19) (4.19)
In virtue of (Equation4.16(4.16) (4.16) ) and (Equation4.17(4.17) (4.17) ), one has
which implies
By using (Equation3.28(3.28) (3.28) ), is completely known. Continuing in the same way as in the proof of Theorem 3.1, all can be determined.
We proceed to denote the transmission eigenvalues of the problem (Equation1.3(1.3) (1.3) ) with the transmission conditions(4.20) (4.20)
by (repeated roots are allowed, of course). which coincide with the zeros of the polynomial(4.21) (4.21)
In virtue of (Equation2.12(2.12) (2.12) ), we have(4.22) (4.22)
If (Equation4.22(4.22) (4.22) ) yields that
Furthermore, if the multiplicity of the root is , that is for , then
Proceeding by induction by differentiating (Equation4.22(4.22) (4.22) ) for k times, we obtain(4.23) (4.23)
for , and is known. We give the following result:
Theorem 4.4:
Let . Suppose for . If () and all transmission eigenvalues of the problem defined by (Equation1.3(1.3) (1.3) )–(Equation1.4(1.4) (1.4) ) are known, as well as m different zeros of , denoted by and their corresponding multiplicities , ,satisfying , are known a priori, then and are uniquely determined.
Proof Let(4.24) (4.24)
It should be noted that for Repeating the proof of Theorem 3.1, we can prove that (Equation3.2(3.2) (3.2) )–(Equation3.5(3.5) (3.5) ) can be obtained again. We set(4.25) (4.25)
and(4.26) (4.26)
where and Thus, together with (Equation3.5(3.5) (3.5) ), we have(4.27) (4.27)
following one has(4.28) (4.28)
Furthermore, proceeding by induction by differentiating (Equation4.27(4.27) (4.27) ) for k times and by using (Equation4.23(4.23) (4.23) ), we obtain(4.29) (4.29)
for and is known. Thus, Hermite interpolation yields the polynomial is known.
Moreover, (Equation3.5(3.5) (3.5) ) together with (Equation4.25(4.25) (4.25) ) and (Equation4.26(4.26) (4.26) ) yields(4.30) (4.30)
then we have(4.31) (4.31)
Proceeding by induction by differentiating (Equation4.30(4.30) (4.30) ) for k times, one yields(4.32) (4.32)
for and is known. Thus, Hermite interpolation yields the polynomial is known. We proceed by using (Equation4.30(4.30) (4.30) ), obtain(4.33) (4.33)
where and is known. Then doing the Euclidean division of polynomial by we have(4.34) (4.34)
Which together with (Equation3.2(3.2) (3.2) ), yields(4.35) (4.35)
where
Since and are coprime, similarly to the proof of Theorem 3.2, (Equation4.35(4.35) (4.35) ) implies that
that is(4.36) (4.36)
and(4.37) (4.37)
Here and are known. Thus the polynomial and are completely known.
Similarly to the proof of Theorem 3.1, by equating the coefficients of the first and second terms of and , we obtain
It follows that is completely known, continuing in the same way, are known, and so on. Thus, eventually, all are determined.
5. One example
The mathematical treatment of the scattering of a time harmonic electromagnetic plane wave by an inhomogeneous infinite cylinder with cross section D such that the electric field E is polarized parallel to the axis of the cylinder leads to an interior transmission eigenvalue problem, that is a boundary value problem for electromagnetic waves (see [Citation6] for details). In this section, we consider the discrete version of the corresponding problem for the case .
If we take and set the transmission eigenvalues are known a priori, which satisfy(5.1) (5.1)
From the relations (Equation2.8(2.8) (2.8) ) and (Equation2.9(2.9) (2.9) ), we have(5.2) (5.2)
and(5.3) (5.3)
Since and are coprime, we have(5.4) (5.4)
by using the extended Euclidean algorithm for polynomial, we obtain(5.5) (5.5)
Then doing the Euclidean division of polynomial by we have(5.6) (5.6)
where
and
Thus(5.7) (5.7)
following we have,(5.8) (5.8)
and(5.9) (5.9)
Note that we know that If is known also, then from the relation (Equation3.25(3.25) (3.25) ), one has
which yields
In virtue of (Equation3.22(3.22) (3.22) ), one obtains
thus
Then one has(5.10) (5.10)
and(5.11) (5.11)
By using (Equation3.26(3.26) (3.26) ), we obtain(5.12) (5.12)
That is In virtue of (Equation3.27(3.27) (3.27) ),then we have .
Acknowledgements
The authors would like to thank the referees for their helpful comments and suggestions. Special thanks to the referee who advices us using the extended Euclidean algorithm which led to a much improved manuscript.
Additional information
Funding
Notes
No potential conflict of interest was reported by the authors.
References
- Aktosun T, Gintides D, Papanicolaou VG. The uniqueness in the inverse problem for transmission eigenvalues for the spherically symmetric variable-speed wave equation. Inverse Prob. 2011;27:115004 (17pp).
- Aktosun T, Gintides D, Papanicolaou VG. Reconstruction of the wave speed from transmission eigenvalues for the spherically symmetric variable-speed wave equation. Inverse Prob. 2013;29:065007 (19pp).
- Cakoni F, Colton D, Haddar H. The interior transmission problem for regions with cavities. SIAM J Math Anal. 2010;42:145–162.
- Cakoni F, Gintides D, Haddar H. The existence of an infnite discrete set of transmission eigenvalues. SIAM J Math Anal. 2010;42:237–255.
- Cakoni F, Colton D, Haddar H. The interior transmission problem for absorbing media. Inverse Prob. 2012;28:045005 (15pp).
- Cakoni F, Colton D, Monk P. On the use of transmission eigenvalues to estimate the index of refraction from far field data. Inverse Prob. 2007;23:507–522.
- Colton D, Kress R. Inverse acoustic and electromagnetic scattering theory. Vol. 93, Applied Mathematical Sciences. 2nd ed. New York (NY): Springer; 1998.
- Kirschl A. On the existence of transmission eigenvalues. Inverse Prob Imaging. 2009;3(2):155–172.
- Cakoni F, Colton D. Qualitative methods in inverse scattering theory. Berlin: Springer; 2006.
- Chen L. A uniqueness theorem on the eigenvalues of spherically symmetric interior transmission problem in absorbing medium. Complex Var Elliptic Equ. 2015;60:145–167.
- Papanicolaou VG, Doumas AV. On the discrete one-dimensional inverse transmission eigenvalue problem. Inverse Prob. 2011;27:015004 (14pp).
- Borcea L, Druskin V, Knizhnerman L. On the continuum limit of a discrete inverse spectral problem on optimal finite difference grids. Commun Pure Appl Math. 2005;58:1231–1279.
- Borcea L, Druskin V. Optimal finite difference grids for direct and inverse Sturm--Liouville problems. Inverse Prob. 2002;18:979–1001.
- Hitrik M, Krupchyk K, Ola P, et al. Transmission eigenvalues for operators with constant coefficients. SIAM J Math Anal. 2010;42:2965–2986.
- Päivärinta L, Sylvester J. Transmission eigenvalues. SIAM J Math Anal. 2008;40:738–753.
- Wei GS. The uniqueness for inverse discrete transmission eigenvalue problems. Linear Algebra Appl. 2013;439:3699–3712.
- Wei GS, Xu HK. Inverse spectral analysis for the transmission eigenvalue problem. Inverse Prob. 2013;29:115012 (24pp).
- Delbary F. Transmission eigenvalues for Maxwell’s equations in isotropic absorbing media with frequency-dependent electrical parameters. Inverse Prob. 2013;29: ap.104005, 20.
- Delbary F, Brignone M, Bozza G, et al. A visualization method for breast cancer detection using microwaves. SIAM J Appl Math. 2010;70:2509–2533.
- Rynne BP, Sleeman BD. The interior transmission problem and inverse scattering from inhomogeneous media. SIAM J Math Anal. 1991;22:1755–1762.