Finite-time stabilization of a network of strings

We investigate the finite-time stabilization of a tree-shaped network of strings. Transparent boundary conditions are applied at all the external nodes. At any internal node, in addition to the usual continuity conditions, a modified Kirchhoff law incorporating a damping term $\alpha u_t$ with a coefficient $\alpha$ that may depend on the node is considered. We show that for a convenient choice of the sequence of coefficients $\alpha$, any solution of the wave equation on the network becomes constant after a finite time. The condition on the coefficients proves to be sharp at least for a star-shaped tree. Similar results are derived when we replace the transparent boundary condition by the Dirichlet (resp. Neumann) boundary condition at one external node.


Introduction
Solutions of certain ODE ẋ = f (x) may reach the equilibrium state in finite time.This phenomenon, when combined with the stability, was termed finite-time stability in [4,11].
A finite-time stabilizer is a feedback control for which the closed-loop system is finite-time stable around some equilibrium state.In some sense, it satisfies a controllability objective with a control in feedback form.On the other hand, a finite-time stabilizer may be seen as an exponential stabilizer yielding an arbitrarily large decay rate for the solutions to the closed-loop system.Indeed, any solution of the closed-loop system can be estimated as where h(δ) → 0 as δ → 0, and λ > 0 is arbitrarily large.This explains why some efforts were made in the last decade to construct finite-time stabilizers for controllable systems, including the linear ones.See [15] for some recent developments and up-to-date references, and [3] for some connections with Lyapunov theory.
To the best knowledge of the authors, the analysis of the finite-time stabilization of PDE is not developed yet.However, since [14], it is well-known that solutions of the wave equation on certain bounded domains may disappear when using transparent boundary conditions.For instance, the solution of the 1-D wave equation u tt − c 2 u xx = 0, in (0, T ) × (0, L), (1.1) cu x (L, t) = −u t (L, t), in (0, T ), (1.2) cu x (0, t) = u t (0, t), in (0, T ), ( (u(0), u t (0)) = (u 0 , u 1 ), in (0, L), (1.4) is finite-time stable in the space {(u, v) ∈ H 1 (0, L) × L 2 (0, L); c u(0) + u(L) + L 0 v(x)dx = 0}, with T = L/c as extinction time (see e.g.[12,Theorem 0.5] for the details.)The condition (1.2) is "transparent" in the sense that a wave u(x, t) = f (x − ct) traveling to the right satisfies 1 (1.2) and leaves the domain at x = L without generating any reflected wave.Note that the solution issued from any state (u 0 , u 1 ) ∈ H 1 (0, L) × L 2 (0, L) is not necessarily vanishing, but constant, for t ≥ L/c.Note also that if we replace (1.3) by the boundary condition u(0, t) = 0 (or u x (0, t) = 0), then a finite-time extinction still occurs (despite the fact that waves bounce at x = 0) with an extinction time T = 2L/c.We refer to [5] for the analysis of the finite-time extinction property for a nonhomogeneous string with a viscous damping at one extremity, to [8] for the finite-time stabilization of a string with a moving boundary, to [16] (resp.[17]) for the finite-time stabilization of a system of conservation laws on an interval (resp.on a tree-shaped network).
The stabilization of networks of strings has been considered in e.g.[1,2,7,9,10,18,20].In [10], the authors considered a star of vibrating strings, and derived the finite time stability (resp.the exponential stability) when transparent boundary conditions are applied at all external nodes (resp.at all external nodes but one, which is changing as times proceeds).For a more general network, we guess that the finite time stability cannot hold without the introduction of additional feedback controls at the internal nodes.Indeed, it is proved here that for a bone-shaped tree, if the feedback controls are applied only at the external nodes, then the finite time stability fails.
The aim of this paper is to investigate the finite-time stabilization of a tree-shaped network of strings.At each internal node n connecting k edges, we assume that the usual continuity condition hold u i (n, t) = u j (n, t), ∀i = j, (1.5) while the usual Kirchhoff law is modified by incorporating a damping term inside: (1.6) In (1.6), the sum is over the indices i of the edges having n as one end, α(n) ∈ R is a coefficient depending on the node n, and we have set u(n, t) := u i (n, t) (for any i) and taken n as the origin of each edge to define the derivative along the space variable.The case α = 0 corresponds to the usual (conservative) Kirchhoff law.Note that we can assume without loss of generality that the length of each edge is one, by scaling the variable x and the coefficient c i along each edge.
Even if the finite-time stabilization of 2 × 2 hyperbolic systems on tree-shaped networks was already considered in [17] (and applied to the regulation of water flows in networks of canals, with k − 1 controls at any node connecting k canals), the novelty (and difficulty) here comes from the fact that only one control is applied at each internal node.The present work can be seen as a first step in the understanding of the finite-time stabilization of systems of conservation laws with a few controls.
A natural guess is that the finite-time stability cannot hold if one can find in the tree a pair of adjacent nodes that are free of any control, because of the (partial but standing) bounces of waves at these nodes.This conjecture will be demonstrated here for a star-shaped tree and a bone-shaped tree.
Actually, we shall prove that the finite-time stabilization can be achieved for a very particular choice of the coefficient α at each internal node.One of the main results proved in this paper is the following Theorem 1.Consider any tree-shaped network of strings, with transparent boundary conditions at the external nodes, continuity conditions and the modified Kirchhoff law at the internal nodes.If at each internal node n connecting k edges we have α(n) = k − 2, then each solution of the wave equation on the network becomes constant after some finite time.
Similar results will be obtained when replacing at one given external node the transparent boundary condition by the homogeneous Dirichlet (resp.Neumann) boundary condition.We shall also see that the condition about α is sharp for a star-shaped tree by explicit computation of the discrete spectrum.The same approach gives for a bone-shaped tree a necessary and sufficient condition for the finite time stability, which differs slightly from those stated in Theorem 1.
The paper is outlined as follows.In Section 2, we provide a sharp condition on the coefficients α(n) for the system to be well-posed.It is obtained by expressing the conditions (1.5)-(1.6)at the internal nodes in terms of the Riemann invariants.In Section 3, we prove the finite-time stability results when the coefficients α are chosen as in Theorem 1.We discuss in Section 4 the necessity of that condition by considering tree-shaped networks and bone-shaped networks.

Well-posedness
We introduce some notations inspired by [6].Let T be a tree, whose vertices (or nodes) are numbered by the index n ∈ N = {0, ..., N }, and whose edges are numbered by the index i ∈ I = {1, ..., N }.We choose a simple vertex (i.e. an external node), called the root of T and denoted by R, and which corresponds to the index n = 0.The edge containing R has i = 1 as index, and its other endpoint has for index n = 1.We choose an orientation of the edges in the tree such that R is the "first" encountered vertex.The depth d of the tree is the number of generations (d = 1 for a tree reduced to a single edge, d = 2 for a star-shaped tree, etc.) Once the orientation of the tree is chosen, each point of the i-th edge (of length 1) is identified with a real number x ∈ [0, 1].The points x = 0 and x = 1 are termed the initial point and the final point of the i-th edge, respectively.Renumbering the edges if needed, we can assume that the edge of index i has as final point the vertex with the (same) index n = i for all i ∈ I. (See Figure 1.)The set of indices of simple and multiple nodes are denoted by N S and N M , respectively.
For n ∈ N M we denote by I n the set of indices of those edges having the vertex of index n as initial point.As we consider a network of strings whose constants c i may vary from one edge to another one, the case #(I n ) = 1 (one child) is possible.The number of edges having the vertex of index n as one as their extremities is We consider the following system with the following boundary conditions ) where the sequence (α n ) n∈N M is still to be defined.For the boundary condition at the root R, we shall consider one of the following conditions Replacing u i,t by v i and dropping the variable t, conditions (2.3) -(2.8) may be rewritten respectively as (2.14) ) and (2.12) we obtain (2.16) Introduce the operator A D , A N and A T defined as The main result in this section is concerned with the well-posedness of system (2.1)-(2.5)and (2.6) (or (2.7), or (2.8)).
Theorem 2. Let T be a tree and let (α n ) n∈N M be a given family of real numbers.Then A T generates a strongly continuous semigroup of operators on H if, and only if, (2.17) The same conclusion holds for A N on H (resp. for A D on H 0 ).
Proof.We sketch the proof only for A T .We need a preliminary result about the Riemann invariants around an internal node.Consider any internal node connecting edges whose indices range over {1, ..., k} (to simplify the notations).Consider any solution of (2.1) satisfying Introduce the Riemann invariants for all i ∈ I. Then the following result holds.
Letting t = 0 yields (2.23).Replacing s i and d i by their expressions in terms of u i and v i and using (2.10), we obtain (2.24).
Let us proceed to the proof of Theorem 2. If (2.17) is not satisfied, picking some initial data (u 0 i , v 0 i ) i∈I ∈ D(A T ) that does not satisfies (2.24) around an internal node for which (2.17) fails, we infer from Lemma 1 that system (2.1)-(2.5)and (2.8) does not admit any solution (u i , v i ) i∈I ∈ C(R + ; D(A T )).This shows A T is not the generator of a continuous semigroup on H. Conversely, assume that (2.17) is satisfied.We aim to construct by a fixed-point procedure a solution to (2.1)-(2.5)and (2.8).Pick any Pick a number ρ ∈ (0, 1).We introduce the Hilbert space E = L 2 ρ t dt (0, T ) N endowed with the norm X(t) stands for the vector (..., d n (1, t), s n+1 (0, t), ..., s n+kn−1 (0, t), ...) where n ranges over N M .Let We define a map P : X = (x 1 , ..., x N ) ∈ E 0 → X = (x 1 , ..., xN ) ∈ E 0 as follows.Pick any n ∈ N M .By Lemma 1, there exists a matrix A n ∈ R kn×kn such that the Riemann invariants associated with the solution of (2.1)-(2.5)and (2.8) satisfy . . .
Then it can be seen that P is a map from E 0 into itself.Let us check that it is a contraction for ρ small enough.Let X 1 = (x 1 1 , ..., x 1 N ) and X 2 = (x 2 1 , ..., x 2 N ) be given in E 0 .In what follows, c denotes a constant that may vary from line to line.Then we have x ∈ [0, 1].We can therefore define for all i ∈ I and all T > 0 a function u i ∈ H 1 ((0, 1) × (0, T )) by the constant of integration being chosen so that )ds for a.e.(x, t) ∈ (0, 1) × (0, T ).
Then it can be seen that S(t) t≥0 is a strongly continuous semigroup in H whose generator is A T .The proof of Theorem 2 is complete.

Finite-time extinction
Pick any tree of depth d ≥ 1, and define the sequence (t i ) i∈I as follows Then it is easily seen that T (R) is the maximum of the quantities and the final point of the edge of index i q is an external node (different from R). Define T (T) as the largest of the T (R)'s when the root R ranges over N S ; that is, we take as root of the tree any external node, change the numbering of the edges and nodes, and define the corresponding sequences (I i ) i∈I and (t i ) i∈I .Obviously, Example 1.Consider again the tree drawn in Figure 1, and assume for simplicity that c i = 1 for all i ∈ [1,11].Then T (R) = 5 and T (T) = 7.Indeed, if we take the node of index n = 12 as (new) root, we obtain T (R n=12 ) = 7.Similarly, we see that Theorem 3. Let T be a tree of root R, and let T (R) and T (T) be as above.Assume that the sequence (α n ) n∈N M satisfies the condition Proof of Lemma 2. We argue by induction on the depth d of the tree.If d = 1, then there is only one edge (I = {1}) and s 1 solves ) Then it is easily seen that Assume now Lemma 1 established for any tree of depth at most d − 1, where d ≥ 2. Pick a tree T of depth d, and a sequence (α n ) n∈N M satisfying (P).Denote by R the node of index n = 1, and by T i , for i = 2, ..., k 1 , the subtree of T of root R and of first edge the edge of T of index i.Since T i is of depth at most d − 1, we infer from the induction hypothesis that for i > 1 It remains to prove (3.7) for i = 1.Since the condition (3.3) is satisfied for n = 1, we infer from (2.22) that It follows then from (3.12) that Finally, using (3.8), we infer that The proof of Lemma 2 is complete.
Let us go back to the proof of Theorem 3.
(i) Assume first that U 0 ∈ H 0 , and let (u i ) i∈I denote the solution of (2.1)-(2.5)and (2.6).From Lemma 2, we have that for all i ∈ I From (2.6), we infer that d 1 (0, t) + s 1 (0, t) = 0 for all t ≥ 0, and hence Using (2.25), we infer that (3.13), this yields Using the second definition of T (R) and proceeding inductively, we arrive to Gathering together (3.13) and (3.14), we infer the existence of some constant C ∈ R such that Using (2.6), we see that C = 0.This proves that solutions of (2.1)-(2.5)and (2.6) are null for t ≥ T (R).Combined with the strong continuity of the semigroup (e tA D ) t≥0 in H 0 , this yields the finite time stability.
(ii) Assume now that u 0 ∈ H and let (u i ) i∈I denote the solution of (2.1)-(2.5)and (2.7).From (2.6), we infer that d 1 (0, t) − s 1 (0, t) = 0 for all t ≥ 0. The same proof as in (i) then yields Thus there exists a constant C ∈ R such that (iii) Pick a solution (u i ) i∈I of (2.1)-(2.5)and (2.8).Then it follows from Lemma 2 that for all i ∈ I s For any given i ∈ I, we pick a sequence i 1 < i 2 < • • • < i p such that i 1 = 1, i = i q for some q ∈ [1, p], and the final point of the edge of index i p is an external point, that we call R. If we exchange R and R, we notice that d i is linked to the sj 's (associated with the new root R) by: We infer that (3.16) Therefore, there exists a constant C ∈ R such that The proof of Theorem 3 is complete.
4. Sharpness of the condition (3. 3) The condition (3.3), which is sufficient to yield the finite-time stability, is expected to be also necessary.A way to prove it is to search for an eigenvalue of the underlying operator.Indeed, if we can find an eigenvalue, then the corresponding exponential solution will not steer 0 in finite time.This program can be achieved when the geometry is sufficiently simple, namely when d = 2, 3. Actually, we will consider any value of the sequence of coefficients (α n ) n∈N M , and exhibit an eigenvalue of the underlying operator when (2.17) holds and (3.3) fails.We shall consider (1) a star-shaped tree, with the homogeneous Dirichlet boundary condition at one external node and the transparent boundary conditions at the other external nodes; (2) a tree with two internal nodes, for which a transparent boundary condition is applied at each external node.We assume that α 1 = N , so that the system (2.1)-(2.5)and (2.6) is well-posed in H 0 according to Theorem 2. According to Theorem 3, there is finite-time stabilization when α 1 = N − 2.
We shall show that this condition is sharp, i.e. that a finite-time stabilization cannot hold if Let α 1 ∈ R be given.The operator A D reads where = d/dx, = d 2 /dx 2 , etc. Setting U := (u , v i ) i∈I , we see that (2.1)-(2.5)and (2.6) may be written as If A D U 0 = λU 0 with U 0 = 0, then the solution U of (4.1)-(4.2) reads U (t) = e λt U 0 (exponential solution), and hence ||U (t)|| H = e (Re λ)t ||U 0 || H > 0 for all t ≥ 0. Thus if the operator A D has at least one eigenvalue, then the finite-time stabilization cannot hold.
Proposition 4.1.Let T denote a star-shaped tree with N edges, and assume that α 1 = N .Then the operator A D has at least one eigenvalue if, and only if, Furthermore, if (4.3) holds, then the discrete spectrum of and log − π 2 denotes the usual determination of the logarithm in C \ iR − .In particular, if (4.3) holds, then the finite-time stabilization of (2.1)-(2.5)and (2.6) in H 0 fails.
2. If we replace the Dirichlet boundary condition u 1 (0, t) = 0 by the transparent boundary condition u 1,t (0, t) = c 1 u 1,x (0, t) and take any value Note that the conditions v 1 (0) = 0 and v i (0) = v 1 (1) for 2 ≤ i ≤ N are satisfied whenever (4.5)-(4.6)and (4.9) hold.(4.5) is easily solved as where a i , b i ∈ C are constants to be determined.Substituting the above expression of u i (x) in (4.6)-(4.9)yields the system If λ = 0, we infer from (4.10)-(4.11)and (4.14) that U = 0, which is excluded.Assume from now on that λ = 0. Then the system (4.11)-(4.14) is found to be equivalent to the system −(N − 1)a 1 (e λ/c 1 − e −λ/c 1 ) − a 1 (e λ/c 1 + e −λ/c 1 ) = −α 1 a 1 (e λ/c 1 − e −λ/c 1 ), (4.17) The existence of a nontrivial solution (a 1 = 0) holds if, and only if, the coefficient above a 1 in (4.17For α 1 = N , (4. 19) is equivalent to ) has a solution λ ∈ C if and only if α 1 = N − 2, and in that case the solutions of (4.1) read Remark 3.For k ∈ Z and λ k as in (4.20), we introduce the sequence of eigenfunctions U k = ((u i,k , v i,k )) 1≤i≤N,k∈Z where Then the family (a k U k ) k∈Z may fail to be a Riesz basis in H 0 for any choice of the sequence of numbers (a k ) k∈Z .Consider e.g.N = 2 and c 2 = c 1 /2.Then, for N − 2 < α 1 < N , k∈Z is a Riesz basis in H 0 , then U can be expended in terms of the U k 's in H 0 as for some sequence we have, by harmonicity, that and hence in L 2 (0, 1).Therefore, u 1 is uniquely determined by the c k 's, and hence by u 2 , which is a property much stronger than the conditions u 1 (0) = 0 and u 1 (1) = u 2 (0) present in the definition of H 0 .This shows that the family (a k U k ) k∈Z is not total in H 0 .
It is natural to conjecture a decay of all the trajectories like ) Without a Riesz basis of eigenvectors in the full space H 0 , the validity of (4.21) seems hard to check.

4.2.
The tree with two internal nodes.We assume now that T is a tree with N + 1 nodes, two of which being multiple ), and we consider the boundary conditions (2.3)-(2.5)and (2.8).(See Figure 3.) We will let α 1 and α 2 range over R, assuming only that (2.17) holds.In particular, when α 1 = α 2 = 0, there is no damping at the internal nodes n = 1, 2. We shall show that the finite-time stabilization cannot hold in that case, because of the (partial but continuous) bounces of waves at the internal nodes.Note that for this geometry, condition (3.3) reads Here, we shall show that there is an eigenvalue (so that the finite-time stability fails) if, and only if, both α . We shall prove that, when then the finite-time stability (to constant functions) occurs.We conclude that, when d = 3 and transparent boundary conditions are imposed at all the external nodes, a necessary and sufficient condition for the finite-stability (to constant functions) is (4.23).The interpretation is that the nodes satisfying (3.3) and for which all the adjacent nodes but one are external, are "transparent" and can be "removed" from the tree.Let (α 1 , α 2 ) ∈ R 2 be given.The operator A T reads then Setting U = (u i , v i ) i∈I , we see that (2.1)-(2.5)and (2.8) may be written as   Combined with the continuity condition u 1 (1, t) = u 2 (0, t), this yields c −1 i .