Abstract
We consider a 3d multi-structure composed of two joined perpendicular thin films: a vertical one with small thickness \(h^a_n\) and a horizontal one with small thickness \(h^b_n\). We study the asymptotic behavior, as \(h^a_n\) and \(h^b_n\) tend to zero, of an eigenvalue problem for the Laplacian defined on this multi-structure. We shall prove that the limit problem depends on the value \(q=\displaystyle {\lim _n\dfrac{h^b_n}{h^a_n}.}\) Precisely, we pinpoint three different limit regimes according to q belonging to \(]0,+\infty [\), q equal to \(+\infty \), or q equal to 0. We identify the limit problems and we also obtain \(H^1\)-strong convergence results.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let \(\Omega _n\), n in \( \mathbb N\), be a 3d multi-structure composed of two joined perpendicular thin films (see Fig. 1): a vertical one \(\Omega ^a_n\) with small thickness \(h^a_n\) and a horizontal one \(\Omega ^b_n\) with small thickness \(h^b_n\) (from now on, the exponent ’a’ stands for above, while ’b’ for below).
In \(\Omega _n\) consider the following eigenvalue problem with mixed boundary conditions
where \(\Gamma _n\) denotes the part of the boundary of \(\Omega _n\) having small thickness (see dotted area in Fig. 1) and \(\nu \) denotes the exterior unit normal to \(\Omega _n\) (see Sect. 2 for the rigorous definition of \(\Omega _n\) and \(\Gamma _n\), and for the weak formulation of problem (1.1)).
For any n in \( \mathbb N\), problem (1.1) has a discrete positive spectrum \(\{\lambda _{n,k}\}_{k\in \mathbb {N}}\) with corresponding eigenfunctions \(\{U_{n,k}\}_{k\in \mathbb {N}}\) forming an orthonormal basis in \(L^2(\Omega _n)\) (see Sect. 2), equipped with the inner product
This means that the following normalization
is considered, but it does not restrict the generality of our results.
Problem (1.1) arises, for instance, from the Fourier analysis in the study of the heat problem or the propagation of sound waves (cf. [23], see also [14] in connection with elastic waves).
For reasons of simplicity and economy, especially from a numerical point of view, one tries to remodel the 3d problem with a problem defined on a multi-structure composed of 2d components. In this paper, it will be obtained by an asymptotic process based on the so-called “dimensional reduction”, i.e., by the study of the asymptotic behavior of problem (1.1) as \(h^a_n\) and \(h^b_n\) tend to zero.
We shall prove that the limit problem depends on a nonegative parameter q defined by
Precisely, we pinpoint three different limit regimes according to q belonging to \(]0,+\infty [\), q equal to \(+\infty \), or q equal to 0.
\(\bullet \) When q belongs to \(]0,+\infty [\), i.e., when the thicknesses of the two thin films vanish with the same rate, we obtain the following limit eigenvalue problem,
where \(\omega ^a\) is the cross-section of the vertical film, \(\omega ^b_+\) and \(\omega ^b_-\) are the two parts into which \(\omega ^b\), the cross-section of the horizontal film, is divided by the intersection with \( \partial \omega ^a\) (see Fig. 2),
Problem (1.3) is a \(2d-2d-2d\) eigenvalue problem with coupled conditions on \(\gamma \) (see the last two lines of (1.3)).
The weak formulation of (1.3) is given by (3.4) (see also (3.1), (3.2), and (3.3)). This problem has a discrete positive spectrum \(\{\lambda _k\}_{k\in \mathbb {N}}\) with the corresponding eigenfunctions \(\{(u^a_k, u^b_{k+}, u^b_{k-})\}_{k\in \mathbb {N}}\) forming a basis in \(L^2(\omega ^a)\times L^2(\omega ^b_+)\times L^2(\omega ^b_-)\) subjected to the orthonormal condition
where \(\delta _{h,k}\) denotes the Kronecker delta.
In Theorem 3.1 we prove the convergence of the eigenvalues of problem (1.1), as \(n\rightarrow +\infty \), to the eigenvalues of problem (1.3) with conservation of the multiplicity. We prove also a strong \(H^1\)-convergence result for the corresponding eigenfunctions (see (3.5), (3.6), (3.7), and Corollary 3.2).
\(\bullet \) When \(q=+\infty \), i.e., when the thickness of the vertical thin film vanishes faster than the thickness of the horizontal thin film, the limit spectrum is the union of the spectra of the following two uncoupled 2d eigenvalue problems with homogeneous Dirichlet boundary condition
Precisely, one has to collect together the eigenvalues of these two problems and order the obtained set in an increasing sequence \(\{\lambda _k\}_{k\in \mathbb {N}}\) with the convention of repeated eigenvalues. The corresponding eigenfunctions form an orthonormal basis in \(L^2(\omega ^a)\times L^2(\omega ^b)\).
In Theorem 3.3 we prove the convergence of the eigenvalues of problem (1.1), as \(n\rightarrow +\infty \), to the sequence \(\{\lambda _k\}_{k\in \mathbb {N}}\) with conservation of the multiplicity. Moreover, by means of renormalization in \(\Omega ^b_n\), we prove a strong \(H^1\)-convergence result for the corresponding eigenfunctions (see (3.10), (3.11), (3.12), and Corollary 3.5).
\(\bullet \) When \(q=0\), i.e., when the thickness of the horizontal thin film vanishes faster than the thickness of the vertical thin film, we choose the sequence \(\{U_{n,k}\}_{k\in \mathbb {N}}\) of eigenfunctions associated to the discrete positive spectrum \(\{\lambda _{n,k}\}_{k\in \mathbb {N}}\) of problem (1.1) such that it forms an orthonormal basis in \(L^2(\Omega _n)\) equipped with the inner product
i.e., the following normalization
is considered.
In this case, the limit spectrum is the union of the spectra of the following three uncoupled 2d eigenvalue problems, the first one with mixed boundary condition, while the other two with homogeneous Dirichlet boundary condition
As above, one has to collect together the eigenvalues of these three problems and order the obtained set in an increasing sequence \(\{\lambda _k\}_{k\in \mathbb {N}}\) with the convention of repeated eigenvalues. The corresponding eigenfunctions form an orthonormal basis in \(L^2(\omega ^a)\times L^2(\omega ^b_+)\times L^2(\omega ^b_-)\).
In Theorem 3.6 we prove the convergence of the eigenvalues of problem (1.1), as \(n\rightarrow +\infty \), to the sequence \(\{\lambda _k\}_{k\in \mathbb {N}}\) with conservation of the multiplicity. Also in this case we prove a strong \(H^1\)-convergence result for the corresponding eigenfunctions, but by means of a renormalization in \(\Omega ^a_n\) (see (3.16), (3.17), (3.18), and Corollary 3.8).
Notice that, when q belongs to \(]0,+\infty [\), choosing (1.2) or (1.4) as normalization leads to the same limit result. Instead, to obtain a meaningful result, normalization (1.2) must be used when q is \(+\infty \) and normalization (1.4) when q is 0.
In Sect. 2, following [3], problem (1.1) is rescaled on a fixed domain. Section 4 is devoted to obtaining a priori estimates of the eigenvalues \(\lambda _{n,k}\) of problem (1.1): below by a positive constant independent of n and k, and above by an explicit constant independent of n but dependent on k (see also Remark 4.2). The upper bound of \(\lambda _{n,k}\) implies \(H^1\)-a priori estimates of the eigenfunctions. In Sect. 3, the main results are stated. Section 5 contains some results that are crucial for proving the main results, i.e., Theorems 3.1, 3.3, and 3.6. Precisely, in Proposition 5.1 we give a trace convergence result, written in a very general way, which will allow us to identify junction and boundary conditions in the limit problems. In Proposition 5.2, we prove a density result for approximating the elements of the space of setting of the limit problem, when q belongs to \(]0,+\infty [\), with regular functions. Although this result was used in other works, to our knowledge, there are no previous proofs of it. Our proof is rather technical and it works also for domains which are not "symmetric". Proposition 5.3 is devoted to building a recovery sequence which will be used in the proof of all three main results. Sections 6, 7, and 8 are devoted to proving the main results in the case where q belongs to \(]0,+\infty [\), q equal to \(+\infty \), or q equal to 0, respectively. The three proofs follow the same pattern. In them, we highlight the novelties and refer to [9] and [22] for the classical parts.
In this paper we consider the Laplace operator in order to investigate the effect of the junction condition on the limit problem. It is of course possible to replace the Laplacian by an elliptic operator with a symmetric and positive definite thermal conductivity matrix. Taking into account our analysis and arguing as in [10] easily lead to the limit problem. Moreover, we just considered two perpendicular thin films. Of course, the whole analysis works with the appropriate modifications if the two films form an angle other than \(\frac{\pi }{2}\). We leave the study of these cases to an interested reader.
The asymptotic behavior of a spectral problem for an homogeneous isotropic elastic body consisting of two folded and perpendicular plates with the same thickness h but with the requirement of large elastic coefficients, of order \(O(h^{-2})\), was studied in [12] (see also [13]). This assumption technically avoids a rescaling of the eigenvalues and gives very different asymptotic behaviors from our problem. Also, we refer to [4, 11, 16], and [17] for different eigenvalue problems in plate theory.
The modelling of spectral problems for the Laplace operator in joined \(1d-1d\) and \(1d-2d\) multi-structures were obtained in [7, 9, 10, 15], and [20]. The modelling of the spectrum for the linear water-wave system in a joined \(1d-2d\) multi-structure was obtained in [1].
For other problems in joined thin films, we refer to [2, 6], and [8].
Eventually, we refer to [5, 18, 19], and references therein, for problems on thin structures.
2 Position of the problem and rescalings
Let \(l_1^+\), \(l_1^-\), \(l_2\), and \(l_3\) be four positive real numbers such that
Set (see Fig. 2)
Let \(\left\{ h^a_n\right\} _{n \in \mathbb N},\) \( \left\{ h^b_n\right\} _{n \in \mathbb N}\) be two sequences in ]0, 1[ such that
For every n in \(\mathbb N\) set (see Fig. 1)
For every n in \( \mathbb N\), consider the space \(L^2(\Omega _n)\) equipped with the inner product
and the space
equipped with the inner product
The classical spectral theory (for instance, see [21]) ensures the existence of an increasing diverging sequence of positive numbers \(\{\lambda _{n,k}\}_{k \in \mathbb N}\) forming the set of all the eigenvalues of Problem (1.1), i.e.,
Moreover, there exists a \(L^2(\Omega _n)\)-Hilbert orthonormal basis \(\{U_{n,k}\}_{k \in \mathbb N}\) such that, for every k in \(\mathbb N\), \(U_{n,k}\) belongs to \(\mathcal{V}_n\) and it is an eigenvector of (2.5) with eigenvalue \(\lambda _{n,k}\); hence, \(\left\{ \lambda _{n,k}^{-\frac{1}{2}}U_{n,k}\right\} _{k \in \mathbb N}\) is a \(\mathcal{V}_n\)-Hilbert orthonormal basis.
Set now
From now on,
As it is usual (see [3]), problem (2.5) will be reformulated on the fixed domain \(\Omega ^a\cup \Omega ^b\cup \left( ]-\frac{1}{2},\frac{1}{2}[\times ]0,l_2[ \right) \) through the following maps
To this aim, for every n in \(\mathbb N\), let \(H_n\) be the space \(L^2(\Omega ^a)\times L^2(\Omega ^b) \) equipped with the inner product
and let \(V_n\) be the space defined by
equipped with the inner product
Moreover, for every n and k in \(\mathbb {N}\), set
Then, for every n in \(\mathbb N\), \(\{\lambda _{n,k}\}_{k \in \mathbb N}\) is an increasing diverging sequence of positive numbers forming the set of all the eigenvalues of the following problem
\(\{u_{n,k}\}_{k \in \mathbb N}\) is a \(H_n\)-Hilbert orthonormal basis such that, for every k in \(\mathbb N\), \(u_{n,k}\) belongs to \(V_n\) and it is an eigenvector of (2.10) with eigenvalue \(\lambda _{n,k}\). Moreover, \(\left\{ \lambda _{n,k}^{-\frac{1}{2}}u_{n,k}\right\} _{k \in \mathbb N}\) is a \(V_n\)-Hilbert orthonormal basis. In particular, one has
Furthermore, for every k in \(\mathbb N\), \(\lambda _{n,k}\) is characterized by the following min-max Principle
where \(\mathcal{F}_k\) is the set of the subspaces \(\mathcal{E}_k\) of \(V_n\) with dimension k (for instance, see [21]).
Problem (2.10) is obtained from (2.5) by means of rescaling of variables, once multiplied by \(\dfrac{1}{h^a_n}\).
3 The main results
This section is devoted to stating the main results of this paper.
The limit problem will depend on q defined by (2.1) which acts as a weight on \(\omega ^b\) in the scalar product. Precisely, three different limit regimes will appear according to q belonging to \(]0,+\infty [\), q equal to \(+\infty \), or q equal to 0.
3.1 The case q in \(]0,+\infty [\)
Fix q in \(]0,+\infty [\).
Consider \(L^2(\omega ^a)\times L^2(\omega ^b)\) equipped with the inner product
Moreover, let
be equipped with the inner product
Both are Hilbert spaces. Moreover, the norm induced on V by the inner product \({\alpha }_{q}(\cdot ,\cdot )\) is equivalent to the usual \(\left( H^1(\omega ^a)\times H^1(\omega ^b)\right) \)-norm, and the norm induced on \(L^2(\omega ^a)\times L^2(\omega ^b)\) by the inner product \([\cdot ,\cdot ]_q\) is equivalent to the usual \(\left( L^2(\omega ^a)\times L^2(\omega ^b)\right) \)-norm. Consequently, V is continuously and compactly embedded into \(L^2(\omega ^a)\times L^2(\omega ^b)\). Furhtermore, V is dense in \(L^2(\omega ^a)\times L^2(\omega ^b)\) since \(C_0^\infty (\omega ^a)\times \{v\in C_0^\infty (\omega ^b):v=0 \hbox { on }\{0\}\times ]0,l_2[\}\) is included in V. Then, all classic results hold true for the eigenvalue problem (see [21])
and the following result will be proved.
Theorem 3.1
For every n in \(\mathbb N\), let \(H_n\) be the space \(L^2(\Omega ^a)\times L^2(\Omega ^b) \) equipped with the inner product \((\cdot ,\cdot )_n\) defined by (2.6) and \(V_n\) be the space defined by (2.7) equipped with the inner product \(a_n(\cdot ,\cdot )\) defined by (2.8).
For every n in \(\mathbb N\), let \(\{\lambda _{n,k}\}_{k \in \mathbb N}\) be the increasing diverging sequence of all the eigenvalues of Problem (2.10) and let \(\{u_{n,k}\}_{k \in \mathbb N}\) be a \(H_n\)-Hilbert orthonormal basis such that \(\left\{ \lambda _{n,k}^{-\frac{1}{2}}u_{n,k}\right\} _{k \in \mathbb N}\) is a \(V_n\)-Hilbert orthonormal basis and, for every \(k\in \mathbb N\), \(u_{n,k}=(u_{n,k}^a,u_{n,k}^b)\) is an eigenvector of Problem (2.10) with eigenvalue \(\lambda _{n,k}\).
Assume that (2.1) holds true with q in \(]0,+\infty [\).
Let \(L^2(\omega ^a)\times L^2(\omega ^b) \) be equipped with the inner product \([\cdot ,\cdot ]_{q}\) defined by (3.1) and V be the space defined by (3.2) equipped with the inner product \({\alpha }_{q}(\cdot ,\cdot )\) defined by (3.3).
Then, there exists an increasing diverging sequence of positive numbers \(\{\lambda _{k}\}_{k \in \mathbb N}\), depending on q, such that
and \(\{\lambda _{k}\}_{k \in \mathbb N}\) is the set of all the eigenvalues of Problem (3.4). Moreover, there exist an increasing sequence of positive integer numbers \(\{n_i\}_{i \in \mathbb N}\) and a \((L^2(\omega ^a)\times L^2(\omega ^b),[\cdot ,\cdot ]_q)\)-Hilbert orthonormal basis \(\{u_{k}\}_{k \in \mathbb N}\) (depending possibly on the selected subsequence \(\{n_i\}_{i \in \mathbb N}\) and q) such that, for every k in \(\mathbb N\), \(u_{k}\) belongs to V and it is an eigenvector of Problem (3.4) with eigenvalue \(\lambda _{k}\), and
as i diverges,
as n diverges. Furthermore, \(\{\lambda _k^{-\frac{1}{2}}u_{k}\}_{k\in \mathbb {N}}\) is a \((V,\alpha _q)\)-Hilbert orthonormal basis.
As far as the original problem (1.1) is concerned, one has the following result which is an immediate corollary of Theorem 3.1, by change of variable.
Corollary 3.2
For every n in \(\mathbb N\), let \(L^2(\Omega _n)\) be equipped with the inner product defined by (2.2) and let \(\mathcal{V}_n\) be the space defined by (2.3) equipped with the inner product defined by (2.4).
For every n in \(\mathbb N\), let \(\{\lambda _{n,k}\}_{k \in \mathbb N}\) be the increasing diverging sequence of all the eigenvalues of Problem (2.5) and let \(\{U_{n,k}\}_{k \in \mathbb N}\) be a \(L^2(\Omega _n)\)-Hilbert orthonormal basis such that \(\left\{ \lambda _{n,k}^{-\frac{1}{2}}U_{n,k}\right\} _{k \in \mathbb N}\) is a \(\mathcal{V}_n\)-Hilbert orthonormal basis and, for every \(k\in \mathbb N\), \(U_{n,k}\) is an eigenvector of Problem (2.5) with eigenvalue \(\lambda _{n,k}\).
Assume that (2.1) holds true with q in \(]0,+\infty [\).
Let \(L^2(\omega ^a)\times L^2(\omega ^b) \) be equipped with the inner product \([\cdot ,\cdot ]_{q}\) defined by (3.1) and V be the space defined by (3.2) equipped with the inner product \({\alpha }_{q}(\cdot ,\cdot )\) defined by (3.3).
Then, there exists an increasing diverging sequence of positive numbers \(\{\lambda _{k}\}_{k \in \mathbb N}\), depending on q, such that
and \(\{\lambda _{k}\}_{k \in \mathbb N}\) is the set of all the eigenvalues of Problem (3.4). Moreover, there exist an increasing sequence of positive integer numbers \(\{n_i\}_{i \in \mathbb N}\) and a \((L^2(\omega ^a)\times L^2(\omega ^b),[\cdot ,\cdot ]_q)\)-Hilbert orthonormal basis \(\{u_{k}\}_{k \in \mathbb N}\) (depending possibly on the selected subsequence \(\{n_i\}_{i \in \mathbb N}\) and q) such that, for every k in \(\mathbb N\), \(u_{k}=(u_{k}^a,u_{k}^b)\) belongs to V and it is an eigenvector of Problem (3.4) with eigenvalue \(\lambda _{k}\), and
where, from now on, \(\displaystyle {\int \!\!\!\!\!\!-}_{\Omega ^a_{n_i}}\) means \(\displaystyle \frac{1}{\vert \Omega ^a_{n_i}\vert }\int _{\Omega ^a_{n_i}}\) and \(\displaystyle {\int \!\!\!\!\!\!-}_{\Omega ^b_{n_i}}\) means \(\displaystyle \frac{1}{\vert \Omega ^b_{n_i}\vert }\int _{\Omega ^a_{n_i}}\).
Furthermore, \(\{\lambda _k^{-\frac{1}{2}}u_{k}\}_{k\in \mathbb {N}}\) is a \((V,\alpha _q)\)-Hilbert orthonormal basis.
3.2 The case \(q=+\infty \)
Let \([\cdot ,\cdot ]_{1}\) be the inner product on \(L^2(\omega ^a)\times L^2(\omega ^b)\) defined by (3.1) with \(q=1\). Moreover, still denote by \({\alpha }_{1}\) the inner product on \(H_0^1(\omega ^a)\times H_0^1(\omega ^b)\) defined by (3.3) with \(q=1\), i.e.,
Then, both are Hilbert spaces and all classic results hold true for the eigenvalue problem
(see [21]) and the following result will be proved when q is equal to \(+\infty \).
Theorem 3.3
For every n in \(\mathbb N\), let \(H_n\) be the space \(L^2(\Omega ^a)\times L^2(\Omega ^b) \) equipped with the inner product \((\cdot ,\cdot )_n\) defined by (2.6) and \(V_n\) be the space defined by (2.7) equipped with the inner product \(a_n(\cdot ,\cdot )\) defined by (2.8).
For every n in \(\mathbb N\), let \(\{\lambda _{n,k}\}_{k \in \mathbb N}\) be the increasing diverging sequence of all the eigenvalues of Problem (2.10) and let \(\{u_{n,k}\}_{k \in \mathbb N}\) be a \(H_n\)-Hilbert orthonormal basis such that \(\left\{ \lambda _{n,k}^{-\frac{1}{2}}u_{n,k}\right\} _{k \in \mathbb N}\) is a \(V_n\)-Hilbert orthonormal basis and, for every \(k\in \mathbb N\), \(u_{n,k}=(u_{n,k}^a,u_{n,k}^b)\) is an eigenvector of Problem (2.10) with eigenvalue \(\lambda _{n,k}\).
Assume that (2.1) holds true with \(q=+\infty \).
Let \(L^2(\omega ^a)\times L^2(\omega ^b) \) be equipped with the inner product \([\cdot ,\cdot ]_{1}\) defined by (3.1) and \(H_0^1(\omega ^a)\times H_0^1(\omega ^b)\) be equipped with the inner product \({\alpha }_{1}(\cdot ,\cdot )\) defined by (3.8).
Then, there exists an increasing diverging sequence of positive numbers \(\{\lambda _{k}\}_{k \in \mathbb N}\) such that
and \(\{\lambda _{k}\}_{k \in \mathbb N}\) is the set of all the eigenvalues of Problem (3.9). Moreover, there exist an increasing sequence of positive integer numbers \(\{n_i\}_{i \in \mathbb N}\) and a \((L^2(\omega ^a)\times L^2(\omega ^b),[\cdot ,\cdot ]_1)\)-Hilbert orthonormal basis \(\{u_{k}\}_{k \in \mathbb N}\) (depending possibly on the selected subsequence \(\{n_i\}_{i \in \mathbb N}\)) such that, for every k in \(\mathbb N\), \(u_{k}\) belongs to \(H_0^1(\omega ^a)\times H_0^1(\omega ^b)\) and it is an eigenvector of Problem (3.9) with eigenvalue \(\lambda _{k}\), and
as i diverges,
as n diverges. Furthermore, \(\{\lambda _k^{-\frac{1}{2}}u_{k}\}_{k\in \mathbb {N}}\) is a \((H_0^1(\omega ^a)\times H_0^1(\omega ^b),\alpha _1)\)-Hilbert orthonormal basis.
Remark 3.4
Notice that (3.10) and (3.12) imply that
As far as the original problem (1.1) is concerned, one has the following result which is an immediate corollary of Theorem 3.3, by change of variable.
Corollary 3.5
For every n in \(\mathbb N\), let \(L^2(\Omega _n)\) be equipped with the inner product defined by (2.2) and let \(\mathcal{V}_n\) be the space defined by (2.3) equipped with the inner product defined by (2.4).
For every n in \(\mathbb N\), let \(\{\lambda _{n,k}\}_{k \in \mathbb N}\) be the increasing diverging sequence of all the eigenvalues of Problem (2.5) and let \(\{U_{n,k}\}_{k \in \mathbb N}\) be a \(L^2(\Omega _n)\)-Hilbert orthonormal basis such that \(\left\{ \lambda _{n,k}^{-\frac{1}{2}}U_{n,k}\right\} _{k \in \mathbb N}\) is a \(\mathcal{V}_n\)-Hilbert orthonormal basis and, for every \(k\in \mathbb N\), \(U_{n,k}\) is an eigenvector of Problem (2.5) with eigenvalue \(\lambda _{n,k}\).
Assume that (2.1) holds true with \(q=+\infty \).
Let \(L^2(\omega ^a)\times L^2(\omega ^b) \) be equipped with the inner product \([\cdot ,\cdot ]_{1}\) defined by (3.1) and \(H_0^1(\omega ^a)\times H_0^1(\omega ^b)\) be equipped with the inner product \({\alpha }_{1}(\cdot ,\cdot )\) defined by (3.8).
Then, there exists an increasing diverging sequence of positive numbers \(\{\lambda _{k}\}_{k \in \mathbb N}\) such that
and \(\{\lambda _{k}\}_{k \in \mathbb N}\) is the set of all the eigenvalues of Problem (3.9). Moreover, there exist an increasing sequence of positive integer numbers \(\{n_i\}_{i \in \mathbb N}\) and a \((L^2(\omega ^a)\times L^2(\omega ^b),[\cdot ,\cdot ]_1)\)-Hilbert orthonormal basis \(\{u_{k}\}_{k \in \mathbb N}\) (depending possibly on the selected subsequence \(\{n_i\}_{i \in \mathbb N}\)) such that, for every k in \(\mathbb N\), \(u_{k}=(u^a_k,u^b_k)\) belongs to \(H_0^1(\omega ^a)\times H_0^1(\omega ^b)\) and it is an eigenvector of Problem (3.9) with eigenvalue \(\lambda _{k}\), and
Furthermore, \(\{\lambda _k^{-\frac{1}{2}}u_{k}\}_{k\in \mathbb {N}}\) is a \((H_0^1(\omega ^a)\times H_0^1(\omega ^b),\alpha _1)\)-Hilbert orthonormal basis.
3.3 The case \(q=0\)
Let \([\cdot ,\cdot ]_{1}\) be the inner product on \(L^2(\omega ^a)\times L^2(\omega ^b)\) defined by (3.1) with \(q=1\).
Set
and still denote by \({\alpha }_{1}\) the inner product on \(H_{\gamma ^a}^1(\omega ^a)\times W_0\) defined by (3.3) with \(q=1\), i.e.,
Then, both are Hilbert spaces and all classic results hold true for the following eigenvalue problem
(see [21]) and the following result will be proved when \(q=0\).
Theorem 3.6
With an abuse of notation, for every n in \(\mathbb N\), let \(L^2(\Omega ^a)\times L^2(\Omega ^b) \) be equipped with the inner product \(\dfrac{h^a_n}{h^b_n}(\cdot ,\cdot )_n\), where \((\cdot ,\cdot )_n\) is defined by (2.6), still denoted by \(H_n\) and be the space defined by (2.7) equipped with the inner product \(\dfrac{h^a_n}{h^b_n}a_n(\cdot ,\cdot )\), where \(a_n(\cdot ,\cdot )\) is defined by (2.8), still denoted by \(V_n\).
For every n in \(\mathbb N\), let \(\{\lambda _{n,k}\}_{k \in \mathbb N}\) be the increasing diverging sequence of all the eigenvalues of Problem (2.10) and let \(\{u_{n,k}\}_{k \in \mathbb N}\) be a \(H_n\)-Hilbert orthonormal basis such that \(\left\{ \lambda _{n,k}^{-\frac{1}{2}}u_{n,k}\right\} _{k \in \mathbb N}\) is a \(V_n\)-Hilbert orthonormal basis and, for every \(k\in \mathbb N\), \(u_{n,k}=(u_{n,k}^a,u_{n,k}^b)\) is an eigenvector of Problem (2.10) with eigenvalue \(\lambda _{n,k}\).
Assume that (2.1) holds true with \(q=0\).
Let \(L^2(\omega ^a)\times L^2(\omega ^b)\) be the space equipped with the inner product \([\cdot ,\cdot ]_{1}\) defined by (3.1), \(W_0\) be defined by (3.13), and \(H_{\gamma ^a}^1(\omega ^a)\times W_0\) be the space equipped with the inner product \({\alpha }_{1}(\cdot ,\cdot )\) defined by (3.14).
Then, there exists an increasing diverging sequence of positive numbers \(\{\lambda _{k}\}_{k \in \mathbb N}\) such that
and \(\{\lambda _{k}\}_{k \in \mathbb N}\) is the set of all the eigenvalues of Problem (3.15). Moreover, there exist an increasing sequence of positive integer numbers \(\{n_i\}_{i \in \mathbb N}\) and a \((L^2(\omega ^a)\times L^2(\omega ^b),[\cdot ,\cdot ]_1)\)-Hilbert orthonormal basis \(\{u_{k}\}_{k \in \mathbb N}\) (depending possibly on the selected subsequence \(\{n_i\}_{i \in \mathbb N}\)) such that, for every k in \(\mathbb N\), \(u_{k}\) belongs to \(H_{\gamma ^a}^1(\omega ^a)\times W_0\) and it is an eigenvector of Problem (3.15) with eigenvalue \(\lambda _{k}\), and
as i diverges,
as n diverges. Furthermore, \(\{\lambda _k^{-\frac{1}{2}}u_{k}\}_{k\in \mathbb {N}}\) is a \((H_{\gamma ^a}^1(\omega ^a)\times W_0,\alpha _1)\)-Hilbert orthonormal basis.
Remark 3.7
Notice that (3.16) and (3.17) imply that
As far as the original problem (1.1) is concerned, one has the following result which is an immediate corollary of Theorem 3.6, by change of variable.
Corollary 3.8
For every n in \(\mathbb N\), let \(L^2(\Omega _n)\) be equipped with the inner product defined
and let \(\mathcal{V}_n\) be the space defined by (2.3) equipped with the inner product defined by
For every n in \(\mathbb N\), let \(\{\lambda _{n,k}\}_{k \in \mathbb N}\) be the increasing diverging sequence of all the eigenvalues of Problem (2.5) and let \(\{U_{n,k}\}_{k \in \mathbb N}\) be a \(L^2(\Omega _n)\)-Hilbert orthonormal basis such that \(\left\{ \lambda _{n,k}^{-\frac{1}{2}}U_{n,k}\right\} _{k \in \mathbb N}\) is a \(\mathcal{V}_n\)-Hilbert orthonormal basis and, for every \(k\in \mathbb N\), \(U_{n,k}\) is an eigenvector of Problem (2.5) with eigenvalue \(\lambda _{n,k}\).
Assume that (2.1) holds true with \(q=0\).
Let \(L^2(\omega ^a)\times L^2(\omega ^b)\) be the space equipped with the inner product \([\cdot ,\cdot ]_{1}\) defined by (3.1), \(W_0\) be defined by (3.13), and \(H_{\gamma ^a}^1(\omega ^a)\times W_0\) be the space equipped with the inner product \({\alpha }_{1}(\cdot ,\cdot )\) defined by (3.14).
Then, there exists an increasing diverging sequence of positive numbers \(\{\lambda _{k}\}_{k \in \mathbb N}\) such that
and \(\{\lambda _{k}\}_{k \in \mathbb N}\) is the set of all the eigenvalues of Problem (3.15). Moreover, there exist an increasing sequence of positive integer numbers \(\{n_i\}_{i \in \mathbb N}\) and a \((L^2(\omega ^a)\times L^2(\omega ^b),[\cdot ,\cdot ]_1)\)-Hilbert orthonormal basis \(\{u_{k}\}_{k \in \mathbb N}\) (depending possibly on the selected subsequence \(\{n_i\}_{i \in \mathbb N}\)) such that, for every \(k\in \mathbb N\), \(u_{k}=(u^a_k,u^b_k)\) belongs to \(H_{\gamma ^a}^1(\omega ^a)\times W_0\) and it is an eigenvector of Problem (3.15) with eigenvalue \(\lambda _{k}\), and
4 A priori estimates on the eigenvalues
This section is devoted to proving lower and upper bounds for the eigenvalues of Problem (2.10)
Proposition 4.1
For every n in \(\mathbb N\), let \(H_n\) be the space \(L^2(\Omega ^a)\times L^2(\Omega ^b) \) equipped with the inner product \((\cdot ,\cdot )_n\) defined by (2.6), \(V_n\) be the space defined by (2.7) equipped with the inner product \(a_n(\cdot ,\cdot )\) defined by (2.8), and \(\{\lambda _{n,k}\}_{k \in \mathbb N}\) be the increasing diverging sequence of all the eigenvalues of Problem (2.10). Then,
where \(l_2\) is the positive real number involved in the definition of \(\omega ^a\) and \(\omega ^b\) (see Sect. 2).
Proof
As far as the proof of (4.1) is concerned, at first note that the boundary conditions on \(u^a_{n,k}\) and \(u^b_{n,k}\) provide that
where \(l_2\) is the positive real number involved in the definition of \(\omega ^a\) and \(\omega ^b\).
Combining now (2.13), (4.3), and (2.12) gives
i.e., (4.1) holds true.
As far as the proof of (4.2) is concerned, let \(\{\lambda _j\}_{j \in \mathbb N}\) be the increasing diverging sequence of all the eigenvalues of the following problem
Then, for every \(j\in \mathbb N\) there exists an eigenvector \(y_j\) in \(H_0^1(\omega ^a)\) of (4.4) with eigenvalue \(\lambda _j\) such that \(\{y_j\}_{j \in \mathbb N}\) is a \(L^2(\omega ^a)\)-Hilbert orthonormal basis and \(\left\{ \lambda _j^{-\frac{1}{2}}y_j\right\} _{j\in \mathbb N}\) is a \(H_0^1(\omega ^a)\)-Hilbert orthonormal basis.
For every j in \(\mathbb N\), set
Fix k in \(\mathbb N\) and set
Then, for every n in \(\mathbb N\), \(Z_k\) is a subspace of \(V_n\) with dimension k. Consequently, the min-max Principle (2.14) provides that
i.e. (4.2) holds true with \(c_k=\lambda _k\). \(\square \)
Remark 4.2
It is possible to give an estimate of the constant \(c_k\) in Proposition 4.1. Indeed, it is well known that the set of all the eigenvalues of problem (4.4) is given by
Then,
Recall that \(A_{im}\sin \left( \dfrac{i\pi }{l_2}x_2\right) \sin \left( \dfrac{m\pi }{l_3}x_3\right) ,\) with \(A_{im}\) in \(\mathbb {R}\), is an eigenfunction of (4.4) with eigenvalue \(\left( \dfrac{i^2}{l_2^2}+\dfrac{m^2}{l_3^2}\right) \pi ^2\).
Remark 4.3
Proposition 4.1 is independent of the asymptotic behavior of \(\left\{ h^a_n\right\} _{n \in \mathbb N}\) and \( \left\{ h^b_n\right\} _{n \in \mathbb N}\).
Choosing \(k=h\) in (2.13) and taking into account Proposition 4.1 provide the following result.
Corollary 4.4
For every n in \(\mathbb N\), let \(H_n\) be the space \(L^2(\Omega ^a)\times L^2(\Omega ^b) \) equipped with the inner product \((\cdot ,\cdot )_n\) defined by (2.6) and \(V_n\) be the space defined by (2.7) equipped with the inner product \(a_n(\cdot ,\cdot )\) defined by (2.8).
For every n in \(\mathbb N\), let \(\{\lambda _{n,k}\}_{k \in \mathbb N}\) be the increasing diverging sequence of all the eigenvalues of Problem (2.10) and let \(\{u_{n,k}\}_{k \in \mathbb N}\) be a \(H_n\)-Hilbert orthonormal basis such that \(\left\{ \lambda _{n,k}^{-\frac{1}{2}}u_{n,k}\right\} _{k \in \mathbb N}\) is a \(V_n\)-Hilbert orthonormal basis and, for every \(k\in \mathbb N\), \(u_{n,k}\) is an eigenvector of Problem (2.10) with eigenvalue \(\lambda _{n,k}\). Then,
5 Some preliminary results
This section contains some results that are crucial for proving Theorems 3.1, 3.3, and 3.6. Precisely, Proposition 5.1 will give a trace convergence result, written in a very general way, which will allow us to identify junction and boundary conditions in the limit problems. Proposition 5.2 will give a density result for approximating the elements of V defined in (3.2) by regular functions. Although this result was used in other works, to our knowledge, there are no previous proofs of it. Our proof is rather technical and it works also for domains which are not “symmetric”. Proposition 5.3 is devoted to building a recovery sequence which will be used in the proof of all three main results.
Proposition 5.1
Let \(\{h_i\}_{i\in \mathbb { N}}\) be a sequence in \(]0,+\infty [\) such that
Let \(\{w_i\}_{i\in \mathbb { N}}\) be sequence in \(H^1(\Omega ^b)\) such that
and
Then,
Notice that assumption (5.2) ensures that the function w given by (5.3) is independent of \(x_3\), i.e.,
Then, it makes sense to write \(w(0,x_2) \) in (5.4).
Proof
At first, one proves the existence of \(\overline{x}_3\) in \(]-1,0[\) and of an increasing sequence of positive integer numbers \(\{i_j\}_{j\in \mathbb N}\) such that
as j diverges.
Indeed, set
Then, Fatou’s Lemma combined with assumption (5.3) provides that
Consequently, there exist two constants c in \(]0,+\infty [\) and \(\overline{x}_3\) in \(]-1,0[\), and an increasing sequence of positive integer numbers \(\{i_j\}_{j\in \mathbb N}\) such that
which provides (5.6), thanks to (5.3) and (5.5).
Now, for proving (5.4), fix \(\varphi \) in \( C_0^\infty \left( ]0,l_2[\right) \) and split the first integral in (5.4), written with index \(i_j\), as
One will pass to the limit, as j diverges, in each term of this decomposition.
As far as the first integral on the right-hand side of (5.7) is concerned, assumption (5.2) implies that
As far as the second integral on the right-hand side of (5.7) is concerned, assumption (5.1) and (5.6) imply
As far as the last integral on the right-hand side of (5.7) is concerned, (5.6) implies
Eventually, passing to the limit in (5.7), as j diverges, and taking into account (5.8), (5.9), and (5.10) give (5.4) for the subsequence \(\{n_{i_j}\}_{j\in \mathbb {N}}\). Notice that (5.4) holds true for the whole subsequence \(\{n_i\}_{i\in \mathbb {N}}\) too, since the limit \(\varphi w\) does not depend on \(\{n_{i_j}\}_{j\in \mathbb {N}}\). \(\square \)
The following proposition is devoted to approximating the elements of the space V defined by (3.2) by more regular functions belonging to the space \(V_{reg}\) defined by
Proposition 5.2
Let V and \( V_{reg} \) be defined by (3.2) and (5.11), respectively. Then, \( V_{reg} \) is dense in V.
Proof
Fix \(\left( v^{a},v^{b}\right) \) in V. The goal is to find a sequence \( \left\{ \left( v_{n}^{a},v_{n}^{b}\right) \right\} _{n \in \mathbb {N}}\) in \( V_{reg} \) such that
The proof of (5.12) will be split into two steps.
Step 1. The first step is devoted to proving (5.12) when
Split \( v^{b}\) in the even part and in the odd part with respect to \(x_{1}\), i.e.,
where
As far as the approximation of \(v^o\) is concerned, since it belongs to \(H_0^1(\omega ^b)\) and
one has that \(v^o_{|\omega ^b_-}\) belongs to \(H_0^1(\omega ^b_-)\) and \(v^o_{|\omega ^b_+}\) belongs to \(H_0^1(\omega ^b_+)\) (see Sect. 2 for the definition of \(\omega ^b_+\) and \(\omega ^b_-\)). Consequently, there exist two sequences \( \left\{ v_{n}^{o-}\right\} _{n \in \mathbb {N} }\) in \(C_0^{\infty }\left( \omega ^b_-\right) \) and \( \left\{ v_{n}^{o+}\right\} _{n \in \mathbb {N} }\) in \(C_0^{\infty }\left( \omega ^b_+\right) \) such that
Then, setting for every n in \(\mathbb {N}\)
one has that
and
As far as the approximation of \(v^a\) and \(v^e\) is concerned, set
Since
it is easy to see that \(\widehat{v}\) belongs to \( H_0^1( \omega ^b_+\cup \omega ^a_R)\). Consequently, there exists a sequence \(\{ \widehat{v}_n\}_{n\in \mathbb {N}}\) in \(C_0^\infty (\omega ^b_+\cup \omega ^a_R)\), such that
which implies, thanks to definition (5.18), that
and
Set now, for every n in \(\mathbb {N}\),
Then, the sequence \( \left\{ v_{n}^{a}\right\} _{n \in \mathbb {N}}\) is included in \( C^{\infty }_0\left( ]0,l_2[\times [0,l_3[ \right) \) and, thanks to (5.21), (5.20), and (5.17), it converges strongly in \(H^1( \omega ^a)\) to the function given by
i.e.
Moreover, setting for every n in \(\mathbb {N}\),
one has
and by virtue of (5.19)
Now, setting for every n in \(\mathbb {N}\),
(5.13), (5.14), (5.15), (5.16), (5.23), (5.24), and (5.25) imply that
Eventually, the sequence \( \left\{ \left( v_{n}^{a},v_{n}^{b}\right) \right\} _{n \in \mathbb {N}}\), so built, is in \( V_{reg} \) and satisfies (5.12).
Step 2. The second step is devoted to proving (5.12) when
For instance, assume
Let \(\widetilde{v}_b\) be the function defined on \(]-l_1^-, l_1^-[\times ]0,l_2[\) by
By virtue of the previous step, there exists a sequence \( \left\{ \left( v_{n}^{a},\widetilde{v}_{n}^{b}\right) \right\} _{n \in \mathbb {N}}\subset C_0^{\infty }( ]0,l_2[\times [0,l_3[ \times C_0\left( ]-l_1^-, l_1^-[\times ]0,l_2[\right) \) such that
for every \(n\in \mathbb {N}\), and
Now, for every \(n\in \mathbb {N}\), let \(v^b_n\) be the function defined on \(\omega ^b=]-l_1^-, l_1^+[\times ]0,l_2[\) by
Then, the sequence \( \left\{ \left( v_{n}^{a},v_{n}^{b}\right) \right\} _{n \in \mathbb {N}}\) belongs to \( V_{reg} \) and satisfies (5.12).
The proof of (5.12) is similar if \(l_1^-<l_1^+.\) \(\square \)
This section concludes with the building of a recovery sequence for functions in \( V_{reg} \) with functions in \(V_n\) defined by (2.7).
Proposition 5.3
Let \( V_{reg} \) be defined by (5.11). Let \(v=(v^a,v^b)\) be in \(V_{reg}\). Then, there exists a sequence \(\{g_n\}_{n\in \mathbb {N}}\subset H_{\Gamma ^a}^1(\Omega ^a)\) such that
Proof
For every \(n\in \mathbb N\) set
Obviously, \(\{g_n\}_{n\in \mathbb {N}}\) is included in \( H_{\Gamma ^a}^1(\Omega ^a)\) and the last line of (5.26) is satisfied. Moreover, by the definition of \(V_{reg}\), it is easy to see that
and
as n diverges, which imply the convergences in (5.26). \(\square \)
Eventually, introduce the space
which will be used in the following sections.
6 Proof of Theorem 3.1
The proof will be split into several steps.
Step 1. The first step is devoted to proving the existence of an increasing sequence of positive integer numbers \(\{n_i\}_{i \in \mathbb N}\), an increasing sequence of positive numbers \(\{\lambda _{k}\}_{k \in \mathbb N}\), a sequence \(\left\{ u_{k}=(u^a_{k},u^b_{k})\right\} _{k \in \mathbb N}\) in \(\widetilde{V}\), where \(\widetilde{V}\) is the space defined by (5.27), and a sequence \(\left\{ (\xi ^a_{k},\xi ^b_k)\right\} _{k \in \mathbb N}\) in \(L^2(\Omega ^a)\times L^2(\Omega ^b)\) (depending possibly on the selected subsequence \(\{n_i\}_{i \in \mathbb N}\)) such that, for every k in \(\mathbb N\),
as i diverges, and
Estimates in (4.1) and in (4.5), assumption (2.1) with q in \(]0,+\infty [\), and a diagonal argument ensure that (6.1), (6.2), and (6.3) hold true for a suitable increasing sequence of positive integer numbers \(\{n_i\}_{i \in \mathbb N}\) and for suitable sequences \(\{\lambda _{k}\}_{k \in \mathbb N}\) in \([\frac{1}{l_2^2},+\infty [ \), \(\left\{ u_{k}=(u^a_{k},u^b_{k})\right\} _{k \in \mathbb N}\) in \(\widetilde{V}\) and \(\left\{ (\xi ^a_{k},\xi ^b_k)\right\} _{k \in \mathbb N}\) in \(L^2(\Omega ^a)\times L^2(\Omega ^b)\).
Eventually, (6.4) follows by passing to the limit in
as i diverges, thanks to assumption (2.1) with q in \(]0,+\infty [\) and the strong \(L^2\)-convergence in (6.2).
For asserting that \(u_k=(u^a_k,u^b_k)\) belongs to V, it remains to prove the following result.
Step 2.
Fix k in \(\mathbb {N}\).
The transmission condition in (2.7) gives
As far as the first integral in (6.6) is concerned, the weak \(H^1\)-convergence in (6.2) and the fact that \(u^a_k\) is independent of \(x_1\) imply
As far as the last integral in (6.6) is concerned, note that estimate in (4.5) provides that
Then, combining (6.8) with the weak \(H^1\)-convergence in (6.2) and using Proposition 5.1 yield
Eventually, the junction condition in (6.5) follows from (6.6), (6.7), and (6.9).
Step 3. This step is devoted to proving that
Fix k in \(\mathbb {N}\). To prove (6.10), by following the classic idea of \(\Gamma -\)convergence, a recovery sequence will be constructed for regular function \(v=(v^a,v^b)\) in \(V_{reg}\), where \(V_{reg}\) is defined by (5.11). Precisely, for \(v=(v^a,v^b)\) in \(V_{reg}\), let \(\{g_n\}_{n\in \mathbb {N}}\) be a sequence in \( H_{\Gamma ^a}^1(\Omega ^a)\) satisfying (5.26) in Proposition 5.3. Choosing \((g_{n_i},v^b)\) as test function in (2.11) written with index \(n_i\) yields
Passing to the limit, as i diverges, in (6.13) and using (2.1) with q in \(]0,+\infty [\), (6.1), (6.2), (6.3), and (5.26) provide that
which implies (6.10), thanks to the density of \(V_{reg}\) in V proved in Proposition 5.2.
Relations in (6.11) follow from (6.10), (6.4), and from the fact that \(\lambda _{k}\) are all positive.
As far as (6.12) is concerned, either (6.12) holds true, or \(\{\lambda _k\}_{k\in N}\) is a finite set. In the second case, by virtue of (6.4), Problem (6.10) would admit an eigenvalue of infinite multiplicity. But this is not possible, due to the Fredholm’s alternative Theorem.
Step 4. This step is devoted to proving (3.5), (3.6), and (3.7).
Fix k in \(\mathbb {N}\).
Combining (2.13), (6.1), and (6.11) gives the convergence of the energies
which implies (3.5), (3.6), and (3.7), thanks to (2.1) with q in \(]0,+\infty [\), (6.2), and (6.3).
Step 5. Conclusion.
It is proved that \(\{\lambda _{k}\}_{k\in \mathbb {N}}\) is included in \([\frac{1}{l_2^2},+\infty [ \) and it is an increasing and diverging sequence of eigenvalues of Problem (6.10), \(\{u_{k}\}_{k\in \mathbb {N}}\) is an orthonormal sequence in \((L^2(\omega ^a)\times L^2(\omega ^b),[\cdot ,\cdot ]_q)\), \(\{\lambda _k^{-\frac{1}{2}}u_{k}\}_{k\in \mathbb {N}}\) is an orthonormal sequence in \((V,\alpha _q)\), for every \(k\in \mathbb {N} \) \(u_{k} \) is an eigenvector for Problem (6.10), with eigenvalue \(\lambda _{k}\), and convergences (3.5), (3.6), and (3.7) hold true.
Moreover, arguing as in [9] (see step 2 in the proof of Theorem 2.5) or as in [22] (see Theorem 9.2), one can prove that there does not exist \((\overline{u}, \overline{\lambda })\in V\times \mathbb {R}\) satisfying the following problem
As in [9] (see step 3 in the proof of Theorem 2.5), this implies that the sequence \(\{\lambda _{k}\}_{k\in \mathbb {N}}\) forms the whole set of the eigenvalues of Problem (3.4), that \(\{\lambda _k^{-\frac{1}{2}}u_{k}\}_{k\in \mathbb {N}}\) is a \((V,\alpha _q)\)-Hilbert orthonormal basis, and that \(\{u_{k}\}_{k \in \mathbb N}\) is a \((L^2(\omega ^a)\times L^2(\omega ^b),[\cdot ,\cdot ]_q)\)-Hilbert orthonormal basis.
In conclusion, since the sequence \(\{\lambda _k\}_{k\in \mathbb {N}}\) can be characterized by the min-max Principle, for every \(k\in \mathbb {N}\) convergence (6.1) holds true for the whole sequence \(\{\lambda _{n,k}\}_{n\in \mathbb {N}}\).
7 Proof of Theorem 3.3
The proof will be split into several steps.
Step 1. The first step is devoted to proving the existence of an increasing sequence of positive integer numbers \(\{n_i\}_{i \in \mathbb N}\), an increasing sequence of positive numbers \(\{\lambda _{k}\}_{k \in \mathbb N}\), a sequence \(\left\{ u_{k}=(u^a_{k},u^b_{k})\right\} _{k \in \mathbb N}\) in \(\widetilde{V}\), where \(\widetilde{V}\) is the space defined by (5.27), and a sequence \(\left\{ (\xi ^a_{k},\xi ^b_k)\right\} _{k \in \mathbb N}\) in \(L^2(\Omega ^a)\times L^2(\Omega ^b)\) (depending possibly on the selected subsequence \(\{n_i\}_{i \in \mathbb N}\)) such that, for every k in \(\mathbb N\),
as i diverges, and
Estimates in (4.1) and in (4.5), and a diagonal argument ensure that (7.1), (7.2), and (7.3) hold true for a suitable increasing sequence of positive integer numbers \(\{n_i\}_{i \in \mathbb N}\), for suitable sequences \(\{\lambda _{k}\}_{k \in \mathbb N}\) in \([\frac{1}{l_2^2},+\infty [\) and for suitable sequences \(\left\{ u_{k}=(u^a_{k},u^b_{k})\right\} _{k \in \mathbb N}\) in \(\widetilde{V}\) and \(\left\{ (\xi ^a_{k},\xi ^b_k)\right\} _{k \in \mathbb N}\) in \(L^2(\Omega ^a)\times L^2(\Omega ^b)\).
Eventually, (7.4) follows by passing to the limit in
as i diverges, thanks to the strong \(L^2\)-convergence in (7.2).
For asserting that \(u^a_k\) belongs to \(H_0^1(\omega ^a)\), it remains to prove the following result.
Step 2.
Fix k in \(\mathbb {N}\).
The transmission condition in (2.7) ensures that
As far as the first integral in (7.6) is concerned, the weak \(H^1\)-convergence in (7.2) and the fact that \(u^a_k\) is independent of \(x_1\) imply
As far as the last integral in (7.6) is concerned, note that estimate in (4.5) provides that
moreover, weak \(H^1\)-convergences in (7.2) and assumption (2.1) with \(q=+\infty \) provide
as i diverges. Then, combining (7.8) with (7.9) and using Proposition 5.1 yield
Eventually, boundary condition (7.5) follows from (7.6), (7.7), and (7.10).
Step 3. This step is devoted to proving that
Fix k in \(\mathbb {N}\).
To obtain (7.11), it is enough to prove that
Equation (7.12) follows immediately by passing to the limit, as i diverges, in (2.11) written with index \(n_i\) and with a test function \(v=(v^a,0)\), \(v^a\) in \(H_0^1(\omega ^a)\), and using (7.1) and (7.2).
As far as the proof of (7.13) is concerned, for \(v^b\) in \(C_0^\infty (\omega ^b)\), it is easy to construct \(v^a\) in \(C_0^\infty (]0,l_2[\times [0,l_3[)\) such that
Then, \(v=(v^a,v^b)\) belongs to \(V_{reg}\), where \(V_{reg}\) is defined by (5.11). Let \(\{g_n\}_{n\in \mathbb {N}}\) be a sequence in \(H_{\Gamma ^a}^1 (\Omega ^a)\) satisfying (5.26) in Proposition 5.3. Choosing \(\left( \sqrt{\frac{h^a_{n_i}}{h^b_{n_i}}}\,g_{n_i},\sqrt{\frac{h^a_{n_i}}{h^b_{n_i}}}\,v^b\right) \) as test function in (2.11) written with index \(n_i\) yields
Passing to the limit, as i diverges, in (7.14) and using (2.1) with \(q=+\infty \), (5.26), (7.1), (7.2), and (7.3) provide (7.13) with \(v^b\) in \(C_0^\infty (\omega ^b)\). Then, (7.13) holds true for any \(v^b\) in \(H_0^1(\omega ^b)\), by a density argument.
Step 4. Conclusion.
By arguing as in the proof of Theorem 3.1, one proves that
and that (3.10), (3.11), and (3.12) hold true.
Moreover in a classical way (for instance, see [9] or [22]) one can prove that the sequence \(\{\lambda _{k}\}_{k\in \mathbb {N}}\) forms the whole set of the eigenvalues of Problem (3.9), that \(\{\lambda _k^{-\frac{1}{2}}u_{k}\}_{k\in \mathbb {N}}\) is a \((H_0^1(\omega ^a)\times H_0^1(\omega ^b),\alpha _1)\)-Hilbert orthonormal basis, and that \(\{u_{k}\}_{k \in \mathbb N}\) is a \((L^2(\omega ^a)\times L^2(\omega ^b),[\cdot ,\cdot ]_1)\)-Hilbert orthonormal basis.
In conclusion, since the sequence \(\{\lambda _k\}_{k\in \mathbb {N}}\) can be characterized by the min-max Principle, for every \(k\in \mathbb {N}\) convergence (7.1) holds true for the whole sequence \(\{\lambda _{n,k}\}_{n\in \mathbb {N}}\).
8 Proof of Theorem 3.6
The proof will be split into several steps.
Step 1. The first step is devoted to proving the existence of an increasing sequence of positive integer numbers \(\{n_i\}_{i \in \mathbb N}\), an increasing sequence of positive numbers \(\{\lambda _{k}\}_{k \in \mathbb N}\), a sequence \(\left\{ u_{k}=(u^a_{k},u^b_{k})\right\} _{k \in \mathbb N}\) in \(\widetilde{V}\), where \(\widetilde{V}\) is the space defined by (5.27), and a sequence \(\left\{ (\xi ^a_{k},\xi ^b_k)\right\} _{k \in \mathbb N}\) in \(L^2(\Omega ^a)\times L^2(\Omega ^b)\) (depending possibly on the selected subsequence \(\{n_i\}_{i \in \mathbb N}\)) such that, for every k in \(\mathbb N\),
as i diverges, and
Thanks to Proposition 4.1,
Then, a diagonal argument ensures that (8.1), (8.2), and (8.3) hold true for a suitable increasing sequence of positive integer numbers \(\{n_i\}_{i \in \mathbb N}\), for suitable sequences \(\{\lambda _{k}\}_{k \in \mathbb N}\) in \([\frac{1}{l_2^2},+\infty [\) and for suitable sequences \(\left\{ u_{k}\right\} _{k \in \mathbb N}\) in \(\widetilde{V}\) and \(\left\{ (\xi ^a_{k},\xi ^b_k)\right\} _{k \in \mathbb N}\) in \(L^2(\Omega ^a)\times L^2(\Omega ^b)\).
Eventually, (8.4) follows by passing to the limit in
as i diverges, and using the strong \(L^2\)-convergence in (8.2).
For asserting that \(u^b_k\) belongs to \(W_0\), it remains to prove the following result.
Step 2.
Fix k in \(\mathbb {N}\).
The transmission conditions in (2.7) gives
As far as the first integral in (8.7) is concerned, the weak \(H^1\)-convergence in (8.2), the fact that \(u^a_k\) is independent of \(x_1\) and that assumption (2.1) holds true with \(q=0\) imply
As fas as the last integral in (8.7) is concerned, note that estimate (8.5) and assumption (2.1) with \(q=0\) provide that
Then, combining the weak \(H^1\)-convergence in (8.2) with (8.9), and using Proposition 5.1 yield
Eventually, boundary condition (8.6) follows from (8.7), (8.8), and (8.10).
Step 3. This step is devoted to proving that
Fix k in \(\mathbb {N}\).
To obtain (8.11), it is enough to prove that
As far as the proof of (8.12) is concerned, for \(v^a\) in \(C_0^\infty (]0,l_2[\times [0,l_3[)\), it is easy to construct \(v^b\) in \(C_0^\infty (\omega ^b)\) such that
Then, \(v=(v^a,v^b)\) belongs to \(V_{reg}\), where \(V_{reg}\) is defined by (5.11). Let \(\{g_n\}_{n\in \mathbb {N}}\) be a sequence in \(H_{\Gamma ^a}^1 (\Omega ^a)\) satisfying (5.26) in Proposition 5.3. Choosing \(\left( \sqrt{\frac{h^a_{n_i}}{h^b_{n_i}}}\,g_{n_i},\sqrt{\frac{h^a_{n_i}}{h^b_{n_i}}}\,v^b\right) \) as test function in (2.11) written with index \(n_i\) yields
Passing to the limit, as i diverges, in (8.14) and using (2.1) with \(q=0\), (5.26), (8.1), (8.2), and (8.3) provide (8.12) with \(v^a\) in \(C_0^\infty (]0,l_2[\times [0,l_3[)\). Then, (8.12) holds true for any \(v^a\) in \(H_{\gamma _a}^1(\omega ^a)\), by a density argument.
As far as the proof of (8.13) is concerned, set
(see Sect. 2 for the definition of \(\omega ^b_+\) and \(\omega ^b_-\)). Obviously, \(\widetilde{W}_0\) is dense in \(W_0\).
Passing to the limit, as i diverges, in (2.11) written with index \(n_i\) and with a test function \(\dfrac{h^a_{n_i}}{h^b_{n_i}}(0,v^b)\), \(v^b\) in \(\widetilde{W}_0\) (note that \((0,v^b)\) belong to \(V_{n_i}\), for i large enough), and using (8.1) and (8.2) provide (8.13) with \(v^b\) in \(\widetilde{W}_0\). Then, (8.13) holds true for any \(v^b\) in \(W_0\), by a density argument.
Step 4. Conclusion.
By arguing as in the proof of Theorem 3.1, one proves that
and that (3.16), (3.17), and (3.18) hold true.
Moreover in a classical way (for instance, see [9] or [22]) one can prove that the sequence \(\{\lambda _{k}\}_{k\in \mathbb {N}}\) forms the whole set of the eigenvalues of Problem (3.15), that \(\{\lambda _k^{-\frac{1}{2}}u_{k}\}_{k\in \mathbb {N}}\) is a \((H_{\gamma ^a}^1(\omega ^a)\times W_0,\alpha _1)\)-Hilbert orthonormal basis, and that \(\{u_{k}\}_{k \in \mathbb N}\) is a \((L^2(\omega ^a)\times L^2(\omega ^b),[\cdot ,\cdot ]_1)\)-Hilbert orthonormal basis.
In conclusion, since the sequence \(\{\lambda _k\}_{k\in \mathbb {N}}\) can be characterized by the min-max Principle, for every \(k\in \mathbb {N}\) convergence (8.1) holds true for the whole sequence \(\{\lambda _{n,k}\}_{n\in \mathbb {N}}\).
Data Availability
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.
References
Cancedda, A., Chiadó Piat, V., Nazarov, S.A., Taskinen, J.: Spectral gaps for the linear water-wave problem in a channel with thin structures. Math. Nachr. 295(4), 657–682 (2022)
Carbone, L., Chacouche, K., Gaudiello, A.: Fin junction of ferroelectric thin films. Adv. Calc. Var. 11(4), 341–371 (2018)
Ciarlet, P.G., Destuynder, P.: A justification of the two-dimensional linear plate model. J. Mécanique 18(2), 315–344 (1979)
Ciarlet, P.G., Kesavan, S.: Two-dimensional approximation of three-dimensional eigenvalue problems in plate theory. Comput. Methods Appl. Mech. Engrg. 26, 149–172 (1981)
Cioranescu, D., Damlamian, A., Griso, G.: The Periodic Unfolding Method Theory and Applications to Partial Differential Problems. Springer, Singapore (2018)
De Maio, U., Faella, L., Soueid, S.: Junction of quasi-stationary ferromagnetic thin films. Asympt. Anal. 94(3–4), 211–240 (2015)
Gaudiello, A., Gómez, D., Pérez, M.E.: Asymptotic analysis of the high frequencies for the Laplace operator in a thin T-like shaped structure. J. Math. Pures Appl. 134, 299–327 (2020)
Gaudiello, A., Hadiji, R.: Junction of ferromagnetic thin films. Calc. Var. Partial Differential Equations 39(3), 593–619 (2010)
Gaudiello, A., Sili, A.: Asymptotic analysis of the eigenvalues of a Laplacian problem in a thin multidomain. Indiana Univ. Math. J. 56(4), 1675–1710 (2007)
Gaudiello, A., Sili, A.: Asymptotic analysis of the eigenvalues of an elliptic problem in an anisotropic thin multidomain. Proc. Roy. Soc. Edinburgh Sect. A 141(4), 739–754 (2011)
Gómez, D., Lobo, M., Pérez, M.E.: On the vibrations of a plate with a concentrated mass and very small thickness. Math. Methods Appl. Sci. 26(1), 27–65 (2003)
Le Dret, H.: Vibrations of a folded plate. RAIRO Modél. Math. Anal. Numér. 24(4), 501–521 (1990)
Le Dret, H.: Problèmes variationnels dans le multi-domaines: modélisation des jonctions et applications. Research in Applied Mathematics, 19 . Masson, Paris (1991)
Landau, L., Lifchitz, E.: Physique théorique. Théorie de l’élasticité, 7 Edition MIR Moscu (1991)
Mel’nyk, T.: Asymptotic approximations for eigenvalues and eigenfunctions of a spectral problem in a thin graph-like junction with a concentrated mass in the node. Anal. Appl. (Singap.) 19(5), 875–939 (2021)
Nazarov, S.A.: On the asymptotics of the spectrum of a problem in elasticity theory for a thin plate. Siberian Math. J. 41(4), 744–759 (2000)
Nazarov, S.A., Pérez, M.E., Taskinen, J.: Localization effect for Dirichlet eigenfunctions in thin non-smooth domains. Trans. Amer. Math. Soc. 368, 4787–4829 (2016)
Panasenko, G.: Multi-Scale Modelling for Structures and Composites. Springer, New York (2005)
Panasenko, G.: Introduction to Multiscale Mathematical Modeling. World Scientific (2022)
Panasenko, G., Pérez, M.E.: Asymptotic partial decomposition of domain for spectral problems in rod structures. J. Math. Pures Appl. 87(1), 1–36 (2007)
Sanchez-Hubert, J., Sanchez-Palencia, E.: Vibration and Coupling of Continuous Systems. Asymptotic Methods. Springer-Verlag, Heidelberg (1989)
Vanninathan, M.: Homogenization of eigenvalue problems in perforated domains. Indian Acad. Sci. (Math. Sci.) 90(3), 239–271 (1981)
Weinberger, H.F.: A First Course in Partial Differential Equations with Complex Variables and Transform Methods. Dover Publications Inc, New York (1995)
Acknowledgements
The work has partially been supported by Grant PGC2018-098178-B-I00 funded by MCIN and by “ERDF A way of making Europe” and by Grant Ref. 20.VP66.64662 funded by Gobierno de Cantabria-UC.
Funding
Open access funding provided by Universitá degli Studi della Campania Luigi Vanvitelli within the CRUI-CARE Agreement.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Susanna Terracini.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Gaudiello, A., Gómez, D. & Pérez-Martínez, ME. A spectral problem for the Laplacian in joined thin films. Calc. Var. 62, 129 (2023). https://doi.org/10.1007/s00526-023-02464-z
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00526-023-02464-z