![MathJax Logo](/templates/jsp/_style2/_tandf/pb2/images/math-jax.gif)
ABSTRACT
The one-dimensional Saint-Venant equation describes unsteady water flow in channels and is derived from the one-dimensional Euler equation by imposing several physical assumptions. In this paper, we consider the linearized and simplified equation in the one-dimensional case featuring a mixed derivative term and prove the global Lipschitz stability of the inverse source problem via the global Carleman estimate.
1. Introduction and main result
We consider
(1)
(1) where a>0,
and T>0 are positive constants and H is the source term. The differential operator A has the form of a one-dimensional wave operator plus the mixed derivative term
. This term appears when we linearize the one-dimensional Saint-Venant equation, which is the equation introduced by Saint-Venant in [Citation1] to describe unsteady water flow in channels. The one-dimensional Saint-Venant equation comprises continuity and momentum equations. Their formulations and physical meanings are written in Cunge et al. [Citation2]. Even though one-dimensional flow does not exist in nature, mathematically speaking, it is important to consider the simplified equation and observe its properties.
We consider the uniqueness and stability for the inverse source problem to determine H on from the boundary observation data of the solution to (Equation1
(1)
(1) ). The argument is based on the Carleman estimate and the Bukhgeim–Klibanov method in [Citation3]. The Carleman estimate was first introduced in Carleman [Citation4] to prove the unique continuation property for the elliptic operator whose coefficients are not necessarily analytic. Using the Carleman estimate, Bukhugeim and Klibanov proved global uniqueness results for multidimensional coefficient inverse problems. This methodology is widely applicable to various partial differential equations provided that we can prove the Carleman estimate for the considered equations. For a hyperbolic equation, Imanuvilov and Yamamoto [Citation5] considered the global Lipschitz stability for wave equations through interior observations. Baudouin et al. [Citation6] proved the global Carleman estimate for the wave equation and considered its applications to controllability, inverse problems, and reconstructions. Bellassoued and Yamamoto [Citation7] considered the inverse source and coefficient problems for the wave equation on a compact Riemannian manifold with a boundary.
In proving the Carleman estimate for the operator A, the main difficulties lie in the existence of the mixed derivative term . There are also difficulties when we apply the Carleman estimate to the extended solution to (Equation1
(1)
(1) ). In the usual case of the wave equation, an evenly extended solution with respect to time t satisfies the wave equation as well. However, considering (Equation1
(1)
(1) ), the evenly extended solution no longer satisfies the equation. We therefore need to consider a different extension.
To prove the global Lipschitz stability for inverse source problems of the hyperbolic partial differential equation, the observation time should be given for the distant wave to reach the boundary owing to the finite propagation speed. We define constants to describe this situation mathematically. Let be a given point and
where
is the square of the wave speed.
Theorem 1.1
Assume where
and
Let
. We assume there exists a unique solution u to (Equation1
(1)
(1) ) in the class
There then exists a constant C>0 that is independent of u and f such that
2. Proof of Theorem 1.1
2.1. Preliminary
To prove Theorem 1.1 using the Bukhugeim–Klibanov method, we need to prove the Carleman estimate for the operator A in extended domain . We consider
where a>0 is a constant. The next proposition is the global Carleman estimate, whose proof is postponed to the Appendix section.
Proposition 2.1
Choose
and β such that
and set
where
and s>0 is some parameters. There then exists a constant
such that for all
, the following holds.
There exist constants and
such that
for all
and
such that
on
and
on
.
Proving our main theorem requires several energy estimates as follows.
Lemma 2.2
Assume
and
. Let
be a solution to
There then exists a constant C such that
holds.
Proof.
Set . Multiplying the first equation by
and integrating over
yield
and so
Furthermore, integration over
yields
Lemma 2.3
Assume . Let
be a solution to
and
There then exists a constant C>0 such that
holds.
Proof.
Multiplying the first equation by and integrating over
, we have
Hence, we get
The proof for
is similar.
2.2. Proof of the main theorem
Proof
Proof of Theorem 1.1
We use the weight function for
. We assume
, i.e.
, and there thus exists
such that
There then exists
such that
Thus, for all
and
, we have
Let u be the solution in the class
and take the extension of u to
:
We also extend R:
We can then prove
and
. Indeed, we assume
and we thus have
and
Furthermore, we assume the symmetry of the source term f and
and thus get
Considering R, we have
owing to the symmetry of
. This extended u satisfies
From Lemma 2.2, we have
(2)
(2) Let
. v satisfies
Also from Lemma 2.2, we have
(3)
(3) We define a cut-off function η satisfying
:
We set
. w satisfies
Moreover,
holds. We can therefore apply Proposition 2.1 to w to obtain
for all
, where
We here consider sufficiently large
as a fixed constant. Considering the second, third, and fourth terms on the right-hand side and using (Equation2
(2)
(2) ) and (Equation3
(3)
(3) ), we get
because
, and for all
and
, we have
. We therefore have
(4)
(4) for all
.
We next set and z satisfies
where
From Lemma 2.3, we have
(5)
(5) and we also find
and thus obtain
Furthermore, we have
by the Lebesgue's dominated convergence theorem. We apply this estimate and (Equation4
(4)
(4) ) to (Equation5
(5)
(5) ) to get
There exists
such that for all
,
holds. Therefore, taking sufficiently large
, we finally have
for a constant C>0 independent of f.
Acknowledgments
The author would like to thank Professor Masahiro Yamamoto (The University of Tokyo) for many valuable discussions and comments. The author also thank anonymous referees for invaluable comments and Glenn Pennycook, NSc, from Edanz Group (www.edanzediting.com/ac) for editing a draft for this manuscript.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Additional information
Funding
References
- Saint-Venant. Théorie du mouvement non permanent des eaux, avec application aux crues des rivières et a l'introduction de marées dans leurs lits. C R Acad Sci. 1871;73:147–154 and 237–240.
- Cunge JA, Holly FM, Verwey A. Practical aspects of computational river hydraulics. Boston: Pitman Advanced Publishing Program; 1980.
- Bukhgeim AL, Klibanov MV. Global uniqueness of class of multidimensional inverse problems. Soviet Math Dokl. 1981;24:244–247.
- Carleman T. Sur un problème d'unicité pur les systèmes d'équations aux dérivées partielles à deux variables indépendantes. Ark Mat Astr Fys. 1939;2B:1–9.
- Imanuvilov OY, Yamamoto M. Global Lipschitz stability in an inverse hyperbolic problem by interior observations. Inverse Probl. 2001;17(4):717–728. doi: https://doi.org/10.1088/0266-5611/17/4/310
- Baudouin L, de Buhan M, Ervedoza S. Global carleman estimates for waves and applications. Comm Partial Differ Equ. 2013;38(5):823–859. doi: https://doi.org/10.1080/03605302.2013.771659
- Bellassoued M, Yamamoto. M. Carleman estimates and applications to inverse problems for hyperbolic systems. Tokyo: Springer Japan; 2017.
Appendix
Proof
Proof of Proposition 2.1
For simplicity, we write
Simple calculation yields
We introduce
Then,
We decompose Pw into
and
as follows.
and
Note that
. We wish to make the lower bound of
, and we therefore try to estimate
.
Through integration by parts, we get
First, we consider a part of the sum
. Set
and
(A1)
(A1) To estimate the terms
(first-order terms), we use the next energy inequality.
Integration by parts yields
Hence, we have
(A2)
(A2) Let
be a constant to be determined later. Multiplying (EquationA2
(A2)
(A2) ) by
, and adding it to (EquationA1
(A1)
(A1) ) yields
for all
to be determined later. We wish to choose
such that both coefficients are positive, i.e.
(A3)
(A3) holds. This claim follows only if μ satisfies
. Therefore, if we choose
, by our assumption of
, this inequality holds and (EquationA3
(A3)
(A3) ) also holds. Hence, there exists a positive constant C>0 such that
(A4)
(A4) Finally, we estimate zeroth-order terms. We have
Hence, we have
Adding this equality to (EquationA4
(A4)
(A4) ) yields
(A5)
(A5) We consider the coefficient of
. Noting
, we have
for all
to be determined later. We wish to choose
such that
(A6)
(A6) Such δ exists only if ν satisfies
. Therefore, if we take
, both the above inequality and (EquationA6
(A6)
(A6) ) hold. Hence, there exists constant a C>0 such that
We apply this estimate to (EquationA5
(A5)
(A5) ) to obtain
for sufficiently large
.
The Cauchy–Schwarz inequality then yields
We can choose s>0 large enough to ensure
We define
and thus obtain
and
. Again by choosing s>0 large, we can rewrite the inequality in terms of
:
Here, we get