Abstract
The aim of this article is to investigate the uniqueness of solution of an inverse source problem for ultrahyperbolic equations. We first reduce the inverse problem to a Cauchy problem for an integro-differential equation and then by using a pointwise Carleman type inequality we prove the uniqueness.
Export citation and abstract BibTeX RIS
1. Introduction and the main result
In this article, we consider an inverse problem for an ultrahyperbolic equation. One of our motivations to deal with this equation is its interesting structure from the point of view of the theory of partial differential equations. For instance, depending on the specific form of initial conditions, solutions possess both hyperbolic and elliptic properties (see [15]). Another motivation is recent discussions on the possibility of physics in multiple time dimensions, (e.g. [4, 20, 21]). Namely, in some superstring theories which attempt to unify the general theory of relativity and the quantum mechanics, extra dimensions are required for the consistency of theory. When the presence of more than one temporal dimension is considered, the mathematical model occurs as an ultrahyperbolic equation (e.g. [9]). More precisely, the paper [9] asserts that the equation in a form of
is of central physical importance, which describes the dynamical evolution of many physical quantities of classical and quantum field theories including the components of the electromagnetic fields in the case of a single time dimension, while the equation in a form of
is fundamental where and are, respectively, space-like and time-like variables.
Throughout the paper, we set
and use the following notation
Moreover for short, we write
Henceforth for , let be a bounded doman and where I is an open bounded interval and is a bounded domain. We set
and we assume that is supported by the plane x1 = 0, that is, for any there exists a cone which is contained in and includes the point , and whose base lies on the plane x1 = 0.
Similarly to [9, 20, 21], we here consider an ultrahyperbolic equation in , which is associated with general geometry in the time variables y :
in the domain .
Throughout we assume that we can choose constant M > 0 such that
and , and . We note that in this article, we do not assume the ellipticity for
The purpose of this article is to investigate the uniqueness of solution of the following problem:
Inverse source problem.
For given and , find a pair of functions in satisfying equation (1), Cauchy data
and the additional information
This is an inverse problem of determining a factor g which is independent of one component y m of the time-like variable of the right-hand side of (1). When we admit the multiple time dimensions related to a superstring theory (e.g. [9, 20, 21]), as a governing equation, we introduce an ultrahyperbolic equation (1) and then we should discuss an original cause initiating dynamical evolution of physical quantities. In our case, the right-hand side in (1) is assumed to be such a cause and our inverse problem is the determination of a factor g.
Our main result is stated in theorem 1 and asserts that our data can determine the factor g uniquely.
Theorem 1. We assume that and there exists a constant such that
Then the inverse source problem has at most one solution such that .
For the uniqueness in the inverse problem, we need to assume (5) and in general we cannot expect the uniqueness for aij satisfying ellipticity condition and (2) even for the case m = n = 1. We can interpret condition (5) in the case of m = 1 that wave propagation speed increases at a neighborhood of the plane x1 = 0, which implies the absence of waveguides in a medium with x1 > 0 and as a related work, see Amirov and Yamamoto [3].
Inverse problems for ultrahyperbolic equations were studied in [1, 2, 6, 11], where the key method is based on weighted a priori estimates and was firstly developed by Bukhgeim and Klibanov [6]. A uniqueness theorem for ultrahyperbolic equations is given by [6] for a bounded domain with Dirichlet and Neumann type condition on a part of the boundary. In [1] and [2], uniqueness is investigated in an unbounded domain with an additional information for the solution of direct problem at y = 0. In [11], Hölder stability estimates were obtained in a bounded domain by some lateral boundary data. A major difference of our work from the existing results is that, in the inverse problem, additional information is given only at y m = 0.
As for the direct problem (1) and (3) with given fg, it is known that the problem of determination of the function u from relations (1) and (3) is ill-posed in the Hadamard sense (see [16], chapter IV). By using the mean-value theorem of Asgeirsson (see e.g. [13]), it was shown by [8] that the existence of solutions fails if the initial conditions are not properly prescribed. We refer to [7, 10, 17–19], as for the uniqueness results for various Cauchy, Dirichlet and Neumann problems for ultrahyperbolic equations. Finally, in [9] it is proved that under a nonlocal constraint, the initial value problem is well-posed for initial data given on a codimension-one hypersurface.
This paper consists of four sections and one appendix. In section 2, we state proposition 1 which is the key Carleman estimate for the proof of theorem 1 and the proof of the proposition is given in section 4. In section 3, we prove theorem 1 on the basis of proposition 1. In appendix, we prove two lemmata which are used for the proofs of proposition 1 and theorem 1.
2. Key Carleman estimate
In order to prove theorem 1, the key tool is an Carleman type inequality which will be presented in proposition 1 below.
We now introduce a Carleman weight function in the form of where
where and are large positive parameters.
We define the domain as a level set by . More precisely, we set
Then we see
First, we reduce equation (1) to a more suitable form by introducing a new variable , that is, , where . Without loss of generality, we assume that i.e. , and so we have . Then, for the new function , by using the relations
we have
where , , are functions in , and or y , and given by ; for example, , and and
For the sake of simplicity, let us denote , by aij, x1, ak, bj , g respectively, where Then we can write
We set
Henceforth C > 0 denotes generic constants which are independent of and
Then we can have the following, which is the key for the proof of theorem 1.
Proposition 1. (Carleman estimate for an ultrahyperbolic equation) Let condition (5) be satisfied and be so small that
where . Then there exists a constant such that for any and there exists a positive number such that for any , the following estimate holds
for all . Here and the vector-valued function U satisfies the following inequality:
The Carleman estimate proposition 1 is an estimate which is uniform in large parameter , that is, the constant C > 0 is independent of provided that is sufficiently large. This uniformity is essential for the application to the inverse problem.
As for Carleman estimates for ultrahyperbolic equation (1) with constant principal parts aii = 1 and aij = 0 for , see section 2 of chapter 4 of [1, 2] and section 4 of chapter IV of [16]. As for a Carleman estimate for an ultrahyperbolic equation with variable aij, see also Romanov [19] which assumes the ellipticity for aij unlike our paper.
We note that reflecting a similar property to the hyperbolicity, the ultrahyperbolic equations require some condition (5) to the principal coefficients aij for proving a Carleman estimate.
As general but very limited references on Carleman estimates, we refer, for example, to Bellassoued and Yamamoto [5], Hörmander [12], and Klibanov and Timonov [14]. Our Carleman estimates are different from more common Carleman estimates in e.g. [5, 12] in the following senses:
- Our Carleman estimate asserts a pointwise inequality, while the conventional Carleman estimates are proved in terms of weighted L2 -integrals of solutions u and boundary data.
- In conventional Carleman estimates, the weight function has a form with suitable , and the estimate holds uniformly in all large .
One can prove a conventional type of Carleman estimate for (1), and we can apply both types similarly to inverse problems and establish the uniqueness for the inverse problem. We note that our weight function is the same as in [1, 2], while [16] applies a weight in the form of .
3. The proof of theorem 1
Once that a relevant Carleman estimate proposition 1 is established, we can apply it for the proof of the uniqueness of the inverse problem by the method originating from Bukhgeim and Klibanov [6]. As for such a method, see also [3, 5, 14]. In this section, postponing the proof of proposition 1 to section 4, we first complete the proof of theorem 1.
Let be a solution to (1), (3) and (4) with in . Since and , there exists a number such that also in We assume that , which was introduced before, satisfies this condition. We define a new unknown function in Then dividing equation (6) by and taking into account relations (3)–(4), we obtain
where , , , depend on the coefficients of equation (1), the function and its derivatives.
Differentiating equation (9) with respect to y m, setting and using (11), we obtain the integro-differential equation
with the Cauchy data
where
We now prove that, if satisfies (12) and (13), then in .
From (12), we obtain
where the constant M0 > 0 depends on the coefficients aij, ak, bj , and . Here we shall use the following lemma, whose proof is given in appendix.
Lemma 1. The following relations hold:
where
By lemma 1, we can write inequality (14) in the following form
On the other hand, by proposition 1, we can write
for . From (15) and (16), it follows that
Integrating inequality (17) in and using the fact that , we have
for and . Taking into account condition (3) and inequality (8), from (18) we can write
Now let us choose small Then we see that and
in Hence inequality (19) implies that
Since and
passing to the limit as in (20), we conclude that
which means that in . Since is an arbitrary number, in
Varying the point of the plane x1 = 0, we establish that z = 0 on that is, on . Then from equation (9) and by condition (11) we conclude that on , where is the projection of in . Repeating the same argument leads to z = 0 in and on . Thus, continuing in this way we complete the proof.
4. The proof of proposition 1
In the proof of proposition 1, we shall use two lemmata. The first of them is lemma 2, the proof of which is technical and lengthy, and we postpone it to appendix.
Lemma 2. Under the hypothesis of proposition 1, there exists a constant such that for any and , there exists a positive number such that for all and for all the following inequality is valid
where is a polynomial in and of degree 6, and the term is given in the proof explicitly.
In (21), the sign of the term is minus and therefore we need to perform another estimation:
Lemma 3. The following equality holds for any function :
where is a polynomial in and of degree 4. Moreover, the term is given as follows
Lemma 3 can be proved easily by direct calculations and we omit the proof here.
We now proceed to the completion of the proof of proposition 1. We multiply equality (22) by and add to inequality (21) to have
for , where
Setting and using the inequality
we can estimate the coefficient of in (23):
for
As for the coefficient of u2, we can write in the form
where the expressions depend on , aij and . Since the functions aij and are bounded in the space , it can be seen that the function is bounded uniformly with respect to and , M1 > 0. Then, noting that we have
for . It is obvious that the functions and are also bounded on for fixed , , that is, there exist constants M2, M3 > 0 such that Hence, by using (25), we obtain
for which yields
Consequently, inequalities (23), (24) and (26) imply that
for and , where . It is easy to see that the vector function U satisfies relation (8). Thus the proof of proposition 1 is completed.
Acknowledgments
The second author was supported by Grant-in-Aid for Scientific Research (S) 15H05740 of Japan Society for the Promotion of Science and by The National Natural Science Foundation of China (no. 11771270, 91730303). This work was prepared with the support of the "RUDN University Program 5-100".
Appendix A.: Proof of lemma 1
Let K be the part of the domain produced by the plane y m = 0 and let be the distance between the boundary point of and the point
Since is monotone with respect to y m, we have
for . Here we used the relations
and
Similarly we can prove the rest part.
Appendix B.: Proof of lemma 2
The proof is based on the method described in section 4 of chapter IV in [16].
We introduce a new function
Then using the relations
we obtain
where we set
and
Here we set and if . Noting that , , and , in we estimate the terms Ti, as follows:
where ;
where
Next
where and we set
We have
where
where
where
where
where
where
where
where
Similarly to (B.10) we obtain
where
Finally we obtain
where
Then by relations (B.2)–(B.15), we can write
where
and
Now we shall evaluate the expressions in (B.16), respectively.
Since , and by (7), we obtain
which implies
Next, by choosing and setting we see that
for where is the constant in (5) and
It is clear that
where Since the functions aij and are bounded in the space the function is bounded uniformly with respect to and that is there exists a number M4 > 0 such that Then
for On the other hand, is also bounded on for fixed and , say, , M5 > 0, then by (B.20) we obtain
for
Thus, inequalities (B.17)–(B.21) lead to
for , ,
Finally, returning to the original function in (B.22), we obtain (21), where
and Thus the proof of lemma 2 is completed.