I nternat. J. Mah. Mh. Si. Vol. 3 No. 3 (1980) 505-520 5O5 A NON-LINEAR HYPERBOLIC EQUATION ELIANA HENRIQUES de BRITO Instituto de Matemtica UFRJ Caixa Postal 1835, ZC-00 Rio de Jane iro RJ BRAS IL (Received January i0, 1979) ABSTRACT. In this paper the following Cauchy problem, in a Hilbert space H, is considered: (I + XA)u" + u(0) u u’(0) A2u + [ + M(IA1/2ul 2)]Au f O u i M and f are given functions, A an operator in H, satisfying convenient hypothesis, X >_ 0 and For when Uo is a real number. in the domain of A and u I 1/2, in the domain of A if > 0, and u I in H, 0, a theorem of existence and uniqueness of weak solution is proved. KEY WORDS AND PHRASES. Non,near Wave Equan, Cauchy Problem, Existence and Uniquess 1980 MATHEMATICS SUBJECT CLASSIFICATION CODES. 34G20, 35L05. 506 I. E.H. de BRITO INTRODUCTION. The physical origin of the problem here considered lies in the theory of vibrations of an extensible beam of length , whose ends are held a fixed distance apart, hinged or clamped, and is either stretched or compressed by an axial force, taking into account the fact that, during vibration, the elements of a beam perform not only a translatory motion, but also rotate; see Timoshenko [9]. A mathematical model for this problem is an initial-boundary value problem for the non-linear hyperbolic equation 2u t 2 4u t22 4u +------ 4 [ + I[ u (s t) ]2 ds3 o 2u 2 0 (I.I) where u(,t) is the deflection of point l at time t, a is a real constant, proportional to the axial force acting on the beam when it is constrained to lle along the axis, and is a nonnegative constant (y%= 0 means neglecting the The non-linearity of the rotatory inertia, while % > 0 means considering it). equation is due to considering the extensibility of the beam. This model, when % 0, was treated by Dickey [2], Ball [I] and, in a Hilbert space formulation, by Medelros [5]. Lions For related problems, see Pohozaev [?], [4], Menzala [6] and Rivera [B]. In this paper, a theorem of existence and uniqueness of weak solution for a Cauchy problem in a Hilbert space H, is proved for the equation (I + kA)u" + A2u + [ + M(IA1/2uI2)] (1.2) Au-- f, with suitable conditions on the operator A and the given functions M and f. This paper is divided in three parts. existence of a weak solution is proved. In Part i, the theorem is taed and In Part 2, its uniqueness is established. Finally, an application is given, in Part 3, when H is n, set with regular boundary in R L2(), and A is the Laplace operator a bounded open A. A NON-LINEAR HYPERBOLIC EQUATION 2. 507 EXISTENCE OF WEAK SOLUTION. Let H be a real Hilbert space, with inner product ( Let A be a linear operator in H, with domain D(A) graph norm of A, denoted II II, and norm V dense in H. With the i.e. for v V, V is a real Hilbert space and its injection in H is continuous. We assume this injection compact. Suppose A self-adjoint and positive, i.e., there is a constant k > 0 such that klvl 2 (Av,v) > for v in V. (2.1) Let V’ be the dual of V and <,> denote the pairing between V’ and V. fylng H and H’, it follows that V H c c V’. Injections being continuous and dense, it is known that, for h in H and v in V, <h,v> Define A < It follows that A 2 A2u, 2: (h,v). V’ by V + (Au,Av), for u, v in V. v > (2.2) V’. is a bounded linear operator from V into 1/2) Let a(u,v) denote the bilinear form in D(A (Au, A1/2v), a(u,v) Identl- associated to A, i.e., ) for u, v in D(A a(u) means a(u,u). Given X > 0, consider in W IX, D((IA) 1/2) the graph norm of (IA) 1/2, denoted i-e., for w in W Note that W Let H, if I D(A), 0, and W be a real number, M a real C I > 0; hence V is dense in W. if function, with M’(s)>0, for s > 0. Assume the existence of positive constants mo I such that M(s) and m >mo + mls’ 508 E.H. de BRITO for s > 0. by ( + Notice that, should M be the identity function, replacement of mo + (m + s), with arbitrary m s > 0, ensures the fulfilment of the above condition on M. The theorem can now be stated. THEOREM. u u(t), 0 Given f in L2(0,T;H) u in V, u [a + O in W I there is a unique function < t < T, such that: a) u L (0,T;V) b) u’. L(0,T;W) c) u is a weak solution of (I + kA)u" + A2u + i.e., for every v in V, u satisfies in d [ (u’ (t),v) dt M([A1/2u 2 )U Au (2.3a) f, D’ (0,T): + %a(u’ (t) v) + (Au(t),Av) + + [a + M(a(u(t)))] a(u(t),v) (f(t),v) (2.3b) d) The following initial conditions hold: u(0) Uo u’(0) u (2.4ab) I Before proving the theorem, some remarks are pertinent. Equation (2.3a) makes sense, because (a) and (b) above imply that u, Au, u’, (kA) u’ belong to L Au, (0,T;H). Initial condition (2.4a) makes sense, because it is known,(see Lions that if u and u’ are in L (O,T;H), then u belongs to C (O,T;H), (2.5) Now, initial condition (2.4b) must be understood. Remember u’ L(O,T;W) implies that (l+kA)u’ <(I + A)u’,v> L(O,T;V’), (u’,v) + %a(u’ v) because for v in V A NON-LINEAR HYPERBOLIC EQUATION From (2.3a), it follows that (I L2(O,T;V’). L2(O,T;V ’) ensures + kA)u" (I + A)u’ and (I + A)u" belong to + A)u’ (0) (I + %A)ul, in V’. is defined. Given u It follows that u’ (0) (I + A)w I in W, set (I 0, for w in W, implies w lw-v l 2 + kA)u’ (0) uI, because, it will be proved below, lw-v l + Xa(w- (2.7) 0. (vj)j , W, i.e., as j + that (2.6) Indeed, V being dense in W, there is a sequence to w in The fact that both C(O,T;V ’) (I + A)u’ Therefore (I 509 vj) N in V that converges --+ 0 and <(I + 0 kA)w,vj> (w,vj) + %a(w,vj) tends to lw’12 (w,w) + %a (w) Hence w 0. Proof of Existence: It will follow Galerkin method. Let, then, wI, ,Wm be a sequence in V such that, for each m, the set is linearly independent and the finite linear combinations of w I, w2, are dense in V. (i) (wj) jN Let Vm denote Approximate Solutions m Search for u (t) m j =i ((I + Suppose, for simplicity, v separable. %A)u(t),v) + the finite subspace of V, spanned by gjm(t)wj (AUm(t),Av) u (0)= u om m in V m + such that Wl,... ,Wm. for all v in V m’ [a+M(a(Um(t))](Aum(t),v)=(f(t),v) u’(O)= m Ulm’ (2.8) (2 9) 510 E.H. de BRITO where Uom converges to Uo in V and Ulm to u I in W. This system of ordinary differential equations with initial conditions has a solution u (t), defined for 0 < t < t m < T. m that the matrix ((l+lA)wj,wl), It is convenient to emphasize ,m, is invertible, for otherwise the i,j--i homogeneous system of linear equations , m [ ((+ j=l 0 A)wj,w+/-)x. would have a non-trivial solution al,...,sm, i-- hence m m m ((I + %A) X j--I m, 7. mjwj, j--I I miwi i 0, a contradiction to the linear independence of w.,...,w m (ii) A Priori Estimates For v 2u(t), (2.8) becomes: + dt + [(o) Set M(a(Um(t))) t a(Um(t)) a d a(Um(t)) + 2(f(t)’um’(t)) (2.10) M(s)ds o We integrate (2.10) from 0 to t < t Ju(t) 2 + m Xa(u(t)) and obtain: + JAum(t) <Km + JaJa(Um(t)) where K m Ulm j2 + la(Ulm) By choice, (remember that + 2 + o (a(um(t))) Ju(s) 12 I JAUom 12 + M(a(Uom)) + Uom and Ulm converge respectively Ulm j2 + Xa(Ulm)) Juzml2 + to T o Uo (2. d, Jf(s) J2ds" in V and to u I in W 511 A NON-LINEAR HYPERBOLIC EQUATION Therefore, there is a constant C o > 0, independent of m and greater than Km such that (2.11) still holds, with K replaced by C o m mo + s _> Now, M(s) a(Um(t)) For M() implies _> mo + (2.12) lels from (2.11), (2.12) and __Isl < 2m + I m I o2 2 one obtains 2 lu’(t) m where C C + %a(u(t)) + 2m 1 o 2 IAUm(t) + + moa(Um(t)) < C + o lu(s) 12d s (2.13) a constant independent of m. In particular, lu’(t) 12 m < C + o lug(s) 12ds. Hence, applying Gronwall inequality lu’(t) 12 m < C e T (2.14) It follows from (2.13) and (2.14) that lU’m(t) 12 + la(u(t)) where K TeT), C(I + In particular, as bounded hence + IAUm(t) 12 + moa(Um(t)) for all t in [0,t 2 _< klUm(t) m (2.15) K, and all m. a(Um(t)), it can be extended to < [0,T-I. it follows that urn(t) remains Therefore, (2.15) holds, in fact, for all m and t in [0,T]. (iii) Passase to the Limit It follows that there is a sub-sequence of which, as m / ,, (urn), still dentoed (urn), for the following is true, in the weak star convergence of L (0,T;H): u m a(u m --+ u, (2.16) a(u), (2.17) 512 E.H. de BRITO Au --+ -- m U m %a(ul) -+ Au, (2.18) U (2.19) la(u’), (2.20) (2.21) M(a (um) )AUm --’+ It must still be proved that, in fact (2.22) M(a(u))Au (2.22) will be shown to follow from the Lemma below, whose proof, here reproduced, was given by J.L. Lions [ 3] and [4]. LEMMA. The mapping v --+ PROOF. The function M(a(v))Av from V into H is monotonic. M(s)ds is non-decreasing (because M’ () () O M(O) > 0) and convex (because "() M’ () > 0). Take (a(v)), (v) It is easy to see that ’ for v in V has a Gateau derivative, (v) 2M(a(v))Av, for v in V, and that # is convex, i.e., for 0 < p < i, #(pv + (l-o)w) < O#(v) + (l-o)#(w), for v, w in V. This inequality can be written in the form. (w + ,(v-w)) (w) -< (v) (w) Passing to the limit, as p + 0 it follows that (’ (w). v-w) -< (v) (w) (w) (v) and, interchanging the roles of v and w, (’ (v). w-v) < Adding the two inequalities above, one obtains: (’(w) ’(v). w-v) >_ 0. A NON-LINEAR HYPERBOLIC EQUATION 513 This proves the Lemma. It can now be shon that (2.22) holds. Indeed, because of the Lemma, for all v in ITcMcaCum))Au m) Because (u is bounded in Jection of V in H strongly in MCaCv))Av), um m o L=(0,T;V) m) is compact, (u L2(O,T;H). L2(0,T;V), and (u) v)dt 0 L=(0,T;H) in and the in- can, further, be supposed to converge to u Hence, as m / (R): M(a(v))Av, u- v)dt T( it is true that > 0 o Set u v pw, p > O, divide the inequality by p and let p T(@_ M(a(u))Au, w)dt / 0, to obtain: > 0 o This holds for all w in L (0,T;V), hence M(a(u))Au. In the following, let k be fixed, k < m; take v in V and let m k (2.19) and (2.20) imply that, in D’(0,T) dt (u(t) V) -+ d-(u’ (t),v) / . (2.23) (2.24) Passing to the limit in (2.8), then (2.23) and (2.24), with (2.17), (2.18), (2.21) and (2.22) ensure that _d [(u,(t) v) + la(u’(t) v)] + (Au(t) Av) + dt + holds in D’ (0,T), all v in V. [a + M(a(u(t)))] a(u(t),v) for all v-in V k. (2.25) (f(t),v), By density, (2.25) holds in ’ (0,T), for 514 E.H. de BRITO Therefore, u is, indeed, a weak solution of (2.3a). It must still be shown, in order to complete the proof of existence, that u satisfies (iv) (2.4ab). Initial Conditions (2.19) means that, for v in V and 0 in C’ (0,T) 0(T) such that 0(0) i and 0, as m o v)O(t)dt (u’(t) m (2.26) (u’(t),v)(t)dt -+ o Because of (2.5) and (2.16), integrating (2.26) by parts, it follows that But u om (Uom, V) - -+ (u(0),v), for v (2.27) in V. u in V; hence (2.27) yields o (u(0),v), for v in V, i.e., u satisfies (2.4a). (Uo,V) To show that u satisfies (2.4b), consider equations (2.3b) and (2.8) for as It follows, using (2.17)-(2.18), (2.21) and (2.22), that, w., j --1,2 v 3 m/o __d dt converges to t (u(t) w.) + (u’ (t),wj) + Xa(u’ 3 (t),wj) (2.28) means that for v in V, 0 in - IT t o o (2.28) Xa(u(t),wj)] weak star in CI(0,T) L=(0,T). such that 0(0)=i, O(T) [(u(t)’wj) + Xa(u’(t) wj )] 0(t)dt m [(u’(t),wj) + Xa(u’(t),Wo)] 0(t)dt. 0, (2.29) 3 Because of (2.6), (2.19) and (2.20), integrating (2.29) by parts, it follows that A NON-LINEAR HYPERBOLIC EQUATION + (Ulm,Wj) But to (u u Ulm--+ I in %a(Ulm,Wj) --+ 515 (u’ (0),wj) + %a(u’ (0),wj), (2.30) W, hence the left-hand side of (2.30) converges also l,wj + a (ul,wj). Therefore + (u’(O),wj) + (Ul,Wj) %a(u’(O),wj) (2.31) %a(ul,wj) 1,2,..., it follows that, in fact, for all v in V: As (2.31) holds for j (u’ (0),v) + la(u’ (0),v) (Ul,V) + la(ul,v). In other words (I + lA)u’(0) [6]) u’(0) But this implies,[see 3. (I + lA)u Ul, I in V’ i e. u satisfies (2 45) UNIQUENESS Let u and 5 be two solutions of (2.3a) with the same initial conditions (2.4ab). Thus w (I + kA)w" + satisfies u A2w + Aw + M(a(u))w +[M(a(u)) 0, w(0) w’(0) M(a(5))] A5 (3.z) 0, (3.2ab) 0. The standard energy method cannot be used to prove uniqueness, because, and not to L(0,T;V). L2(0,T;V’), (3.1) belongs to while the left-hand side of u’ belongs to L(0,T;W) A modification has to be made; this procedure can be f.ound in Lions [3]. Consider: s -ftw()d for t < s 0 for t > s (3.3) z(t) and wl(t) w()d, (3.4) 516 E.H. de BRITO Then o z(s) z’(t) and As w L(R)(0,T;V), LI(0,T;V). (3.6) w(s) z(O) (3.7) (3.8) w(t) [see it is clear (3.8).] (3.3) and that z and z’ are in Hence, it follows from (3.1) that < (I + lA)w"(t), z(t)-> dt (Aw(t), Az(t))dt + + O O + I a s (Aw(t),z(t))dt + o I s M(a(u(t)))(Aw(t),z(t))dt + o s + [M(a(u(t))) M(a((t)))](AS(t),z(t))dt (3.9) 0. o Bu, [ (3.8)] < (l+IA)w"(t) ,z (t) > t((l+IA)w’(t),z(t))- ((I+IA)w’(t),z’(t)) ((Z+A)w’ (t),z()) Therefore, using (3 2ab) and (3 7) + s + )w"(t),z() o No, [, ( 3. ) ] Thus, [see (Aw(t) Az(t)) (3.6) and (3.7)] (Az’(t) Az(=))d= o ((Z+XA)w(t),w(=)) it follows that (remember ka (w)) < ( y > d - Az(t)) 2 i l2 __d 1/2lAw(s)[2 dt ’[w[ 2 (3.0) ]Az(t) (3.n) A NON-LINEAR HYPERBOLIC EQUATION As lwll >-lwl, I() 12 + + 517 (3.9), (3.10) and (3.11)yield IA (s)12 -< 21=I l(w(t), Az())Idt O M(a(u(t)))l(w(t) Az(t))Idt 2 O IM(a(u(t))) + 2 M(a((t))) l((t), Az(t))Idt (3.12) O I As u, uEL (0,T;V) and, for s > 0, M > 0 is a C function, with there is a constant C > 0 such that M(a(u(t)))](w(t) Az(t))[dt 2 < 2C O O [w(t) IAz(t)[dt (3.13) O And IM(a(u(t)) 2 l(5(t), M(a((t))) Az(t)) O s ]a(u(t)) -< 2C O < 2C 2 o < 2C 3 o [see I I s ](A(u(t) + (t)), w(t))] ]Az(t) ldt O s [w(t) [Az (t) (3.14) O (3.5) ] 2[w(t) ]Az(t)[ < Notice that, a((t))] [(t)[ IAz(t)[dt O 2[Iw(t)]2 + iAwl(t [2 ] + IAwl(s) 12 (3.15) Hence, it follows from (3.15) that 2 O Hence (3.13) and (3.15) give O + iAWl(t)12 ]dr + a(3.16) E.H. de BRITO 518 s M(a(u(t)})l(w(t), 2 Az(t)} o <_ E 2co IAw(t)12]d + Co IAw ()12 lw(t)12 + o (3.17) Now (3.14) and (3.15) give s IMCaCuCt))) 2 MCaCCt)))l ICe,t), AzCt)) o co -< [ o lw(t)12 + IAWl(t)i 2 ]dt + C3o It now follows from (3.12), with (3.16) lw(s) 12 + where C Take s I=l o ( IAw (s) 12 s (3.17) that E lw(t)12 + cs)lAw()12 -< 2C (3.8) o IAwl(t) (3.19) + C o + Co3. such that for 0 -< s So’ 1 <i- CS < I. Hence (3.19) yields for0Sss: o lw() 12 + I IAw()l 2 I < 2O s IAWl(t) 123dt. E lw(t) 2 + 2 + lw(t) 123dt. A fortiori, for 0 lw()l 2 + IAl(,)l 2 _< [lwCt) 4c o Applying Gronwall inequality, it then follows that w(s) 0, for 0 < s < s o 0, for s o < s 0, for 0 < s < T. Similarly, it is proved that w(s) then follows that, in fact, w(s) The proof of uniqueness is complete. < s o + , with 0. It A NON-LINEAR HYPERBOLIC EQUATION 4. 519 APPLICATION n, For a bounded open set in R H L with regular boundary, consider 2() V H I() H o 2(D) n Let A be the Laplace and V the gradient operators in R A * A, hence A Hypothesis on A are satisfied. V. case, the condition (Av,v) > klvl 2, respectively. Take Notice that, in this for v in V, is the Friedrichs Poincar inequality; the compactness of the injection of V in H is the Rellich theorem. It is clear that Now (,) and are W L2(), if I W HI(), if I > 0 respectively 0 the inner product and the norm in L2(). Given u u f H o I() 2 H () o I L L2(), if I 2(0,T; L ()), O; u I t HI(), if I > 0, 2 the theorem proved above ensures existence and uniqueness of weak solution for the non-linear hyperbolic equation (I satisfying u(0) ACKNOW__EDGBMENT. 2 l&)u + A u- [e Uo, u’ (0) u M(IVu 2) Au f, I. This research was supported by FIMEP and CEPG- UFRJ. Special acknowledgement is due to Professor L.A. Medeiros, who drew my attention to this problem, for his generous help and encouragement. REFERENCES 13 Ball, J.M. Initial Boundary Value Problems for an Extensible Beam, Journal of Functional Analysis and Applications 42, (1973) 61-90. 520 2] E.H. de BRITO Dickey, R.W. Free Vibrations and Dynamic Buckling of the Extensible Beam, J. Math. Anal. and Appl. 29, (1970), 443-454. Quelques Mthodes de Rsolution des Problmes aux Limites Non-Linaires, Dunod, Gauthier-Villars, Paris, 1969. Lions, J.L. 4] Lions, J.L. On Some Questions in Boundary Value Problems of Mathematical Physics, Lecture Notes, Instituto de Matemtica, Universidade Federal do Rio de Janeiro, 1977. IS] Medeiros, L.A. On a New Class of Non-Linear Wave Equations, to appear in in Journal of Mathematical Analysis and Applications. [6] Menzala, G.P. Une Solution d’une Equation Non-Linaire d’Evolution, C.R. Acad. Sc. Paris t. _286, (1978), 273-275. [? ] Pohozaev, S.I. On a Class of Quasi Linear Hyperbolic Equations, Mat. USSR Sbornik, 25, (1975) i, 145-158. 8] [9] Rivera, P.H. On a Non-Linear Hyperbolic Equation in Hilbert Spaces, Anais Acad. Bras. Ciencias, Vol. 50, No. 2, 1978. Timoshenko, S. D.H. Young, and W. Weavem John Wiley, 1974. Vibration Problems in Engineering,