ESAIM: Control, Optimisation and Calculus of Variations URL: http://www.emath.fr/cocv/,, BOUNDARY CONTROLLABILITY IN PROBLEMS OF TRANSMISSION FOR A CLASS OF SECOND ORDER HYPERBOLIC SYSTEMS JOHN E. LAGNESE Abstract. We consider transmission problems for general second order linear hyperbolic systems having piecewise constant coecients in a bounded, open connected set with smooth boundary and controlled through the Dirichlet boundary condition. It is proved that such a system is exactly controllable in an appropriate function space provided the interfaces where the coecients have a jump discontinuity are all star-shaped with respect to one and the same point and the coecients satisfy a certain monotonicity condition.. Introduction and main results Let, be bounded, open, connected sets in R n with smooth boundaries ; and ;, respectively, such that, and set = n, whose boundary is ; =;[ ;. Let m and n be positive integers. For i k = ::: m,forj ` = ::: n and for =, let be real numbers having the symmetry = a k`ij = i k = ::: m j ` = ::: n: (.) When m = n we further assume that The = a jik` = a ij`k = i j k ` = ::: n: (.) are also assumed to satisfy the following ellipticity condition: if m 6= n then ij k` c ij ij c > 8( ij ) R mn (.3) if m = n then ij k` c ij ij c > 8( ij ) S n (.4) where S n denotes the set of all real symmetric n n matrices. In (.3), (.4) and in what follows, a repeated Roman index indicates summation with respect to that index over an appropriate range (either to m or to n, as the context suggests), unless otherwise noted. Fix T>and let W = B w. w m C A W = B w. w m C A Department of Mathematics, Georgetown University, Washington, DC 57 USA. Received by the journal April 5, 997. Accepted for publication October 3, 997. c Societe de Mathematiques Appliquees et Industrielles. Typeset by L A TEX.
344 JOHN E. LAGNESE be vector functions dened on the cylinders = ( T)and = ( T), respectively. Weset=; ( T) and =; ( T) and consider the following problem of transmission: ( w i ; w k `j = in w i ; (.5) w k `j = in i = ::: m where _ = =t and where a subscript following a comma indicates dierentiation with respect to the corresponding spatial variable w i = w i W = U on (.6) w k ` j = w k ` j on i = ::: m (.7) where = ( ::: n ) is the unit normal to ; pointing into (and towards the exterior of ), W j t= = _ W j t= = in, = : (.8) The function U in (.6) is viewed as a control function which ischosen to eect the evolution of the dynamics. More precisely, we consider below the reachability problem, which is to determine the set R T = f(w (T ) W (T ) _W (T ) _W (T )) : U Ug where U is an appropriate space of controls. The question of boundary controllability in problems of transmission has been considered by several authors. In particular, J.-L. Lions considered the system (.5) - (.8) in the special case of two wave equations, that is, he considered the problem ( w ; aw = in (.9) w ; a w = in where a > and is the ordinary Laplacian in R n, w = u on (.) w w = w a = a w on (.) w j t= = _w j t= = in, = : (.) For this problem the following result is proved in [5, Chapter VI]. Theorem.. Assume that ; is star-shaped with respect to some point x and set ;(x )=fx ;j (x ; x ) >g, (x )=;(x ) ( T), where is the unit outer normal to ;. Let R T = f(w (T ) w (T ) _w (T ) _w (T )) : u L () u=on n(x )g: If a >a and T>T(x ):=R(x )= p a,where R(x ) = max x jx ; x j, then R T = H V where H = L ( ) L ( ) V = f(' ' ) H ( ) H ( ): ' j ; = ' j ; = ' j ; g:
BOUNDARY CONTROLLABILITY IN PROBLEMS OF TRANSMISSION 345 The main purpose of this paper is prove the following generalization of Theorem. to the system (.5) - (.8). Let us dene ja j = 8 >< >: kl ij inf m 6= n 6=( ij )R mn ij ij kl ij inf m = n: 6=( ij )S n ij ij (.3) Theorem.. Assume that ; is star-shaped with respect to some point x and let ;(x ), (x ) and R(x ) be asintheorem.. Let R T =f(w (T ) W (T ) _W (T ) _W (T )):U [L ()] m U= on n(x )g: If and if 8 < : ij k` ij k` 8( ij ) R mn m6= n then R T = H V where ij k` ij k` 8( ij ) S n m= n T>T(x )= 8 >< >: R(x ) p ja j m 6= n p R(x ) p m = n: ja j H =[L ( )] m [L ( )] m V = f( ) [H ( )] m [H ( )] m : j ; = j ; = j ; g: (.4) (.5) Remark.3. Whether or not the monotonicity condition (.4) is necessary for exact controllability in dimension n is unknown, even in the case of ordinary wave equations considered in Theorem.. It is known that no such condition is necessary for exact boundary controllability of transmission problems in dimension one or, more generally, for exact controllability of - dimensional networks in R 3 [3, Chapter IV]. Remark.4. Set H ;( )=f' H ( ): 'j ; =g: The space V is a closed subspace of [H ( )] m [H ; ( )] m and V may be identied with the restriction to V of elements in the product space [(H ( )) ] m [(H ; ( )) ] m (cf. [5, Remark 5., p. 376]. Remark.5. With only minor modications in the proof, Theorem. can be extended to the situation involving and p nested open sets! :::! p with Set! i! i+ i = ::: p;! p =: =! i =! i n! i; i = ::: p: Setting ; i :=! i, one should assume that ; i is star-shaped with respect to a point x, i = ::: p;, and that the parameters, = ::: p,
346 JOHN E. LAGNESE associated with satisfy appropriate monotonicity conditions with respect to. Example.6. (Wave equations) Suppose that m =. In this case the system (.5) may be written ( w ; a j` w j` in w ; a j` w j` in where a j` = a`j.ifwe set A =(a j` ), the condition (.4) with m =<n becomes the requirement that the matrix A ; A be positive semidenite. The control time T (x ) is given by R(x )= p ja j,where A ja j = inf 6=R n jj : In particular, if A = a I,whereI is the identity matrix in R n,werecover Theorem.. Example.7. (Linear isotropic elastodynamics) Here n = m = 3 and the system of equations is (we do not use summation convention in this example) 8 >< >: w i ; w i ; ( + ) w i ; w i ; ( + ) 3X 3X j= j= w j ji = in w j ji = in i = 3 where is the (constant) mass density per unit of reference volume and, are the Lame parameters for the elastic material occupying the region.we assume that > and : (.6) By a change of notation we may assume that = =. Then the nonzero coecients are given by a ijij a iiii = + i = 3 = a ijji = a iijj = i 6= j = 3 and therefore for ( ij ) S 3 wehave X 3X a ijk` ij k` =( + ) ii + = 3X i= i= ii! + 3X i j= i6=j 3X i j= ii jj + ij 3X i j= 3X i j= i6=j ij: ( ij + ji ) In the same way, the monotonicity condition (.4) (with strict inequality) will be satised if and : The control time T (x )isgiven by p R(x )= p ja j =R(x )= p, which is what is expected since the two wave speeds associated with the region are p and p +.
BOUNDARY CONTROLLABILITY IN PROBLEMS OF TRANSMISSION 347 Let us briey recount other works on boundary controllability in problems of transmission or, more generally, in interface problems. These works are for the most part disjoint from the present paper. For one-dimensional interface problems (for example, string or beam equations on one-dimensional graphs in R 3 ), one may refer to [3] for a comprehensive analysis of the reachability problem utilizing controls supported at the nodes of the graph. These systems, however, do not generally fall within the framework of the problem (.5) - (.8). For higher dimensional systems, one may cite[], where boundary controllability in transmissions problems for the Reissner-Mindlin system of thin plate theory is considered. This system consists of two coupled second order equations with constant coecients but with lower order terms as well, so they too fall outside of the scope of the present work. There are also several works dealing with boundary controllability in problems of transmission or in interface problems for the wave equation in the singular case. If, for example, ; \ ; 6= in the problem (.9) - (.) above, the solution will generally suer a loss of regularity in a neighborhood of points belonging to this set. In this case the proof technique for Theorem. given in [5] fails that is the reason for the assumption. In the singular case, we can refer to two recent papers of S. Nicaise [6], [7], where the question of exact boundary controllability for the wave equation on two-dimensional networks in R n is treated. In these papers it is shown, among other things, that a result analogous to Theorem. may be obtained provided that, in addition to boundary controls, controls are also imposed near the singular vertices. The reader is referred to the cited papers for the precise statements of Nicaise's results. References to other works related to the topic of the present paper may be found in the bibliographies of the books and papers cited above.. A priori estimates Let H and V be as in Theorem.. For k set H k () = [H k ( )] m [H k ( )] m : Let := ( ), :=( ) be elements of V and consider the bilinear form ( ) = X = ' k ` i j dx V: This form is continuous on V V in the topology induced by H () and is symmetric by virtue of (.). If m 6= n, is also coercive onv because of the ellipticity assumption (.3) and j ; =, that is, for some k>, ( ) kkk H () 8 V: On the other hand, if m = n, because of the symmetry assumptions (.) we may write ( ) = X = " k`( )" ij ( ) dx
348 JOHN E. LAGNESE where " k`() = (' k ` + '` k ): If V then the functions ( ' i (x) x ' i (x)= ' i (x) x i = ::: m are in H () and therefore the following Korn's inequality holds [8, Theorem.]: X = " ij ( )" ij ( ) dx It follows from (.) and (.4) that ( ) c o X = X = " ij ( )" ij ( ) dx kkk H () ' i j ' i j dx: (.) 8 V Therefore in all cases [( )] = denes a norm on V equivalent to the one induced by H (). We retopologize V with this norm and let V denote the dual space of V. With H identied with its dual space we have the dense and continuous embeddings V H V. Let us consider the homogeneous boundary value problem ( 'i ; ' k `j = in ' i ; (.) ' k `j = in i = ::: m = on (.3) ' i = ' i ' k ` j = ' k ` j on i = ::: m (.4) j t= = _ j t= = in, = : (.5) Assume that := ( ) V, := ( ) H. According to standard variational theory (.) - (.5) has a unique solution in the following sense: there is a unique function := ( ) C([ T] V ) \ C ([ T] H) \ C ([ T] V ) such that h i V + ( ) = 8 V <t T and which satises (.5), where h i V denotes the pairing between elements V and V in the V ; V duality. Furthermore, the mapping S(t) : ( ) 7! ((t) (t)) _ is a denes a unitary group on V H, that is k(t)k V + k _ (t)k H = k k V + k k H t : (.6) Let A denote the canonical isomorphism of V onto V, so that ( ) = ha i V, 8 V, and set D(A) =f V : A Hg k k D(A) = ka k H : If ( )D(A)V, then C([ T] D(A))\C ([ T] V )\C ([ T] H) and S(t) is a unitary group on D(A) V. Because of our assumptions on the regions and, an element D(A) has the regularity
BOUNDARY CONTROLLABILITY IN PROBLEMS OF TRANSMISSION 349 H () \ V and A =;( ' k j` ' k j`). Moreover, satises both of the transmission conditions in (.4) in the trace sense and the norm kk D(A) is equivalent to that induced on D(A) by H (). These results are standard and may beproved by the usual methods. Lemma.. Suppose that ( ) D(A) V, and let h [W ()] n. Then the solution of (.) - (.5) satises the following identity: X = _' i (h r' i ) dx + = T (div h)[ _' i _' i ; + + (h ) + (h ) 'i ' k ` ' i p h p j dxdt 'k 'i ' k ` ' i j ] dxdt j ` d h ( ; )' i j ' k ` ; ' i 'k Proof. The proof utilizes classical multipliers. Form = and integrate by parts. One has ' i (h r' i ) dxdt = + ; ' k j ` d: (.7) (' i ; ' k `j )h r' i dxdt (.8) _' i (h r' i ) dx (div h) _' i _' i dxdt ; T (h ) _' i _' i d (.9) where = ( T) and is the unit normal vector pointing into the exterior of. In addition, ( ' k `j )h r' i dxdt = ; ' k `' i p h p j dxdt ; From the symmetry property (.)we obtain ' k `' i p h p j d ' k `' i pj h p dxdt: a ijk` ' k `' i pj = (aijk` ' k `' i j ) p
35 JOHN E. LAGNESE and therefore ( ' k `j )h r' i dxdt = ' k `' i p h p j d ; ; ' k `' i p h p j dxdt + We then obtain from (.8) - (.) the identity = _' i (h r' i ) dx + T + (h )( _' i _' i ; (div h)[ _' i _' i ; (h ) ' k `' i j d (div h) ' k ` ' i j dxdt: (.) ' k `' i p h p j dxdt ' k `' i j ) d+ We sum (.) over =, noting the relations ' i j = ' i j ' i j j = ' i j ' k ` ' i j ] dxdt ' k `' i p h p j d: i = ::: m j = ::: n (.) and observing that along the boundary region ; that is common to and the unit normal vector pointing out of points into.we obtain X ( = _' i (h r' i ) dx + + = T + (div h)( _' i _' i ; (h ) 'k ' k `' i p h p j dxdt ' k ` ' i j ) dxdt j ` d 'i (h )( ' k `' i j ; ' k `' i j ) d + ( ' k `' i p ; ' k `' i p )h p j d: (.) To complete the proof we need to show that the last two integrals in (.) agree with the last integral in (.7). Consider the last integral in (.). By choosing suitable extensions of and wemay assume that ' i H () for = and i = ::: n. We have ( ' k `' i p ; ' k `' i p )h p j = ' k ` j h p (' i p ; ' i p )=(h ) ' k ` j 'i ; ' i (.3)
on, since BOUNDARY CONTROLLABILITY IN PROBLEMS OF TRANSMISSION 35 (' i p ; ' i p )j ; = 'i It also follows from (.4) that on and ; ' i p i = ::: m p = ::: n: (.4) ' k `' i j = ' k ` ' i j + 'i ' k `' i j = 'i ' k ` ' i j ; ; ' i j ; ' i j : If the last expression is subtracted from the preceding one we obtain ' k `' i j ; ' k `' i j =( ; )' k `' i j + ' k ` j 'i ; ' i on (.5) where we have used the symmetry property (.) and the second transmission condition in (.4). Itfollows from (.3) and (.5) that (h )( ' k `' i j ; ' k `' i j ) d + ( ' k `' i p h p j ; ' k `' i p h p j ) d = We use (.4) once again to obtain on ( ; )' k `' i j (h )( ; )' k `' i j d: (.6) =( ; )' k `' i j +( ; 'i )' k ` j =( ; )' k `' i j + ' k ` j 'i ; ' k ` j 'i ; ' i ; ' i =( ; )' k `' i j + 'i (' k ` ; ' k `) j =( ; )' k `' i j + 'k ; ' k 'i Now substitute (.7) into (.6). ; ' i ; ' i ; ' i j `: (.7) Theorem.. Let ( ) V H. Then the solution of (.) - (.5) satises 'k for some constant C. 'i j ` d C(T + )(k k V + k k H ) (.8)
35 JOHN E. LAGNESE Proof. By density it suces to prove (.8) for ( ) D(A) V.Since ; and ; are smooth, there is a vector eld h [C ()] n such that h = on ; and h = on ; (see Komornik [, Lemma.]). We utilize this vector eld in (.7) and observe that the left side of (.7) is bounded above in absolute value by C o (k (T _ )k H +k(t )k V +k k H +k k V )+C T sup(k(t)k V +k (t)k _ H) [ T ] in view of (.6). C(T +)(k k H + k k V ) Remark.3. If m 6= n is follows from (.8) and (.3) that ' i L () i = ::: m: (.9) If m = n, then (.8) and (.4) imply that ' i j + ' j i L () i j = ::: n so, in particular, i ' i = L () for each i = ::: n. Since nx 'i j + ' nx j 'i nx ' i i = + i j i j= i= i j= ' j and the second sum is in L (), it follows that (.9) holds true also in this case. Theorem.4. Assume that ; is star-shaped with respect to some point x and suppose that the coecients satisfy (.4). Let ( ) V H. Then the solution of (.) - (.5) satises (T ; T (x ))(k k H + k k V ) (h ) 'k (x ) where T (x ) is given by (.5) and h(x) =x ; x. 'i j ` d (.) Proof. Because of (.8) it suces to prove (.) for ( ) D(A) V. We apply (.7) with h(x) =x ; x. Then h on;. On the other hand, assumption (.4) and the standing assumption (.) (or (.)) imply that ( ; )' i j ' k ` + Therefore X = _' i (h r' i ) dx T 'i + n +(; n ) ' k ` ' j dxdt + (x ) (h ) ; ' i 'k _' i _' i dxdt 'k ; ' k j ` : ' k ` ' i p h p j dxdt 'i j ` d: (.)
Now form BOUNDARY CONTROLLABILITY IN PROBLEMS OF TRANSMISSION 353 = X = and integrate by parts to obtain X T _' i ' i ( = ; = (' i ; (_' i _' i ; ' k `j )' i dxdt ' k `' i j ) dxdt ( ' k `' i j ; ' k `' i j ) d: (.) The right side vanishes because of the transmission conditions (.4). Multiply (.) by (n ; )= and add the result from (.) to obtain X = _' i (h r' i + n ; ' i ) dx + (_' i _' i + (x ) (h ) T ' k ` ' j ) dxdt 'k 'i ) j ` d: (.3) One has X = (_' i _' i + ' k ` ' j ) dxdt = T (k k V + k k H) (.4) and (dropping for the moment summation convention over repeated Roman indices) X mx = i= X mx _' i (h r' i + n ; ' i) dx K = i= _' i dx + 4K (h r' i +(n ; )' i ) dx for any K>. From Komornik [, Lemma 3.] we have (hr' i +(n;)' i ) dx (n;) Hence X mx = i= _' i (h r' i + n ; ' i) dx X mx K = i= ; (h)' i d+4 _' i dx + K (hr' i ) dx: (h r' i ) dx
354 JOHN E. LAGNESE where we have utilized (.3) and (.4). Nowchoose K = K(x )tobeaconstant such that (returning to summation convention over repeated Roman indices) X = (h r' i )(h r' i ) dx K We then obtain X = X K _' i (h r' i + n ; ' i) dx = X = ' k l ' i j dx 8 V: (_' i _' i + ' k `' i j ) dx = K (k k V + k k H): (.5) Use of (.4) and (.5) in (.3) yields the inequality (T ; K)(k k V + k k H) (h ) 'k (x ) 'i j ` d: One may choose K = 8 >< >: R(x ) p ja j m 6= n p R(x ) p m = n: ja j In fact, if m 6= n one has X = (R(x )) ja j (h r' i )(h r' i ) dx (R(x )) X = X = a ijkl ' k `' i j dx (R(x )) ja j while if m = n we have from (.) X = (h r' i )(h r' i ) dx (R(x )) (R(x )) ja j (R(x )) ja j = (R(x )) ja j X = ' i j ' i j dx X = X = X = X = a ijkl ' k `' i j dx " ij ( )" ij ( ) dx " kl ( )" ij ( ) dx " kl ( )" ij ( ) dx ' k `' i j dx:
Set BOUNDARY CONTROLLABILITY IN PROBLEMS OF TRANSMISSION 355 3. Proof of Theorem. U = fu [L ()] m : U = onn(x )g: We rst make precise the sense in which (.5) - (.8) is solvable when U U. Proceeding formally for the moment, let D(A) and form = X = (w i ; w k `j )' i dx: If we integrate by parts and use (.) and properties of we obtain = X = (w i ' i dx ; = ; The last equation is the same as ' k `j w i dx ;(x ) ( W ) H +(W A) H = ; Let H and set = A ;. One has ;(x ) ' k ` j u i d; ' k ` j u i d; U = ;(x ) and therefore there is a B L(U H)such that ; ;(x ) Then (3.) may be written which is the same as B u. u m C A : ' k ` j u i d;: (3.) CkUk U kk D(A) = CkUk U k k H ' k ` j u k d;=(bu ) H 8 H: ( W A ; ) H +(W ) H =(BU ) H 8 H W + AW = ABU in (D(A)) (3.) where (D(A)) is the dual of D(A) with respect to H and A is an isomorphism of H onto (D(A)) dened by ha i D(A) =( A) H 8 H D(A): Since ABU L ( T (D(A)) )itfollows that given initial data W() = W V _W() = W (D(A)) (3.3) the problem (3.), (3.3) has a unique solution with W C([ T] V ), _W C([ T] (D(A)) ). Further, because of Theorem. and utilizing a \lifting theorem" of Lasiecka and Triggiani [4] we have that (W W ) H V =) W C([ T] H) \ C ([ T] V ) and, moreover, k(w _W)k C([ T ] HV ) C T [k(w W )k HV + kuk U ]:
356 JOHN E. LAGNESE In particular, the problem (.5) - (.8) has a unique solution in the sense of (3.) and W() = _W() = : (3.4) To complete the proof we apply the Hilbert uniqueness method [5]. Let =( ) be the solution of the problem ( 'i ; ' k `j = in ' i ; ' k `j = in i = ::: m = on ' i = ' i ' k ` j = ' k ` j on i = ::: m j t=t = V _j t=t = H: According to Theorems. and., for T>T(x ) one may dene on V H an equivalent norm kk F by " k( 'k = 'i )k F = j ` d# : (x ) Now letw =(W W ) be the solution of (.5) - (.8) in which wechoose From = 8 < U = : ; on (x ) on n(x ): X = we obtain by integration by parts (w i ; =h _ W(T ) i V ; (W(T ) ) H ; which may be rewritten h( W(T _ ) ;W(T )) ( )i V H = 'k (x ) w k `j )' i dx (x ) 'i ' k ` j d 'i j ` d =k( )k F : Therefore for T > T(x ) the map : ( ) 7! ( _ W(T ) ;W(T )) is the canonical isomorphism of F onto F = V H. Hence, given any (W ;W ) V H one may choose ( ) V H such that that is, ( )=(W ;W ) W(T )=W _W(T )=W : This proves that R T = H V if T>T(x ).
BOUNDARY CONTROLLABILITY IN PROBLEMS OF TRANSMISSION 357 References [] V. Komornik: Exact Controllability and Stabilization: The Multiplier Method,Masson, Paris, 994. [] J. E. Lagnese: Boundary controllability in transmission problems for thin plates, in Dierential Equations, Dynamical Systems, and Control Science, K.D.Elworthy, W. N. Everitt and E. B. Lee, Eds., Lecture Notes in Pure and Applied Mathematics, vol. 5, Marcel Dekker, New York, 994, 64{658. [3] J. Lagnese, G. Leugering, G. Schmidt: Modeling, Analysis and Control of Dynamic Elastic Multi-Link Structures, Birkhauser, Boston, 994. [4] I. Lasiecka, R. Triggiani: A lifting theorem for the time regularity of solutions of abstract equations with unbounded operators and applications to hyperbolic equations, Proc. AMS 4 (988), 745{755. [5] J.-L. Lions: Contr^olabilite Exacte, Perturbations et Stabilisation de Systemes Distribues: Tome, Contr^olabilite Exacte, Coll. RMA, vol. 8, Masson, Paris, 988. [6] S. Nicaise: Boundary exact controllability of interface problems with singularities I: addition of the coecients of singularities, SIAM J. Control & Opt. 34 (996), 5{ 53. [7] S. Nicaise: Boundary exact controllability of interface problems with singularities II: addition of internal controls, SIAM J. Control & Opt. 35 (997), 585{63. [8] O. A. Oleinik, A. S. Shamaev, G. A. Yosian: Mathematical Problems in Elasticity and Homogenization, Studies in Mathematics and Its Applications, vol. 6, North-Holland, Amsterdam, 99.