UDC 517.944+531/534
MATHEMATICS
Submitted 1967-01-01 | RussiaRxiv: ru-196701.70603 | Translated from Russian

Abstract

Full Text

UDC 517.944+531/534

MATHEMATICS

Yu. A. DUBINSKII

ON AN OPERATOR SCHEME AND THE SOLVABILITY OF A NUMBER OF QUASILINEAR EQUATIONS OF MECHANICS

(Presented by Academician A. Yu. Ishlinskii, 2 XII 1966)

The paper proves theorems on the solvability of a certain class of nonlinear equations in Banach spaces, making it possible to consider from a general point of view questions of the existence of solutions of quasilinear equations and systems occurring in mechanics.

§ 1. A general scheme for stationary (elliptic) equations. Let \(V, X, Y\) be a triple of Banach spaces; \(A, B\) nonlinear operators. The interrelation of these spaces and operators is expressed by the following diagram

\[ \begin{array}{ccccc} & \xrightarrow{\ i\ } & X & \xrightarrow{\ A\ } & X^* \\ & & & & \\ V & & & & \xrightarrow{\ i\ } V^*,\\ & & & & \\ & \xrightarrow{\ i\ } & X & \xrightarrow{\ B\ } & Y \end{array} \]

where \(i\) is the embedding operator; \(V^*, X^*\) are the conjugate spaces.

Assume that \(X\) is separable and that the embedding \(V \subset X\) is dense everywhere. Consider the equation

\[ A(u)+B(u)=h, \tag{1} \]

where \(u \in X\) is the unknown solution, and \(h \in X^*\) is given.

Definition. A solution of equation (1) is an element \(u \in X\) such that for every \(v \in V\) the identity

\[ \langle A(u),v\rangle+\langle B(u),v\rangle=\langle h,v\rangle \tag{2} \]

holds.

(Here \(\langle w,v\rangle\) is the value of \(w \in V^*\) on \(v \in V\).)

Assumptions. I. Ellipticity of \(A(u)\). For every \(u \in V\)

\[ \lim_{\|u\|_X\to\infty}\frac{\langle A(u),u\rangle}{\|u\|_X}=+\infty . \]

II. Orthogonality of \(B(u)\). For every \(u \in V\), \(\langle B(u),u\rangle=0\).

III. The operators \(A\) and \(B\) are weakly continuous as operators from \(X\) into \(V^*\), i.e., if \(u_n \to u\) weakly in \(X\), then \(\langle A(u_n),v\rangle \to \langle A(u),v\rangle\) and \(\langle B(u_n),v\rangle \to \langle B(u),v\rangle\) for every \(v \in V\).

Theorem 1. If conditions I–III are satisfied, then for every \(h \in X^*\) equation (1) has at least one solution in the sense of (2).

The proof of the theorem is carried out by B. G. Galerkin’s method; here the solvability of the moment equations and the a priori estimate follow from conditions I, II, while the possibility of passage to the limit follows from condition III.

§ 2. Applications to stationary problems.

  1. The Navier–Stokes system of equations. Let \(G \subset \mathbb{R}^n\), \(u=(u_1,\ldots,u_n)\),

\[ -\Delta u+\sum_{k=1}^{n}u_k\frac{\partial u}{\partial x_k}+\operatorname{grad} p=h(x),\qquad \operatorname{div}u=0,\qquad u\big|_{\partial G}=0 . \tag{3} \]

We apply Theorem 1 to prove solvability of system (3); accordingly, set

\[ X=\{u(x)\mid u(x)\in \dot W_2^{(1)},\ \operatorname{div}u=0\}, \]

\[ V=\{u(x)\mid u(x)\in \dot W_p^{(1)},\ p>n,\ \operatorname{div}u=0\},\quad Y=L_1,\quad A(u)=-\Delta u, \]

\[ B(u)=\sum_{k=1}^n u_k\frac{\partial u}{\partial x_k}. \]

Theorem 2. If \(h(x)\in X^*\), then system (3) is solvable.

This result is known; see, for example, \((^1)\).

Remark. It is not difficult to show that \(X^*=W_2^{(-1)}/\Pi\), where \(\Pi\) is the subspace of potentials of the form \(u(x)=\operatorname{grad}p(x)\), \(p(x)\in L_2\) (cf. \((^1)\)).

2. The system of large deflection of an A. Föppl plate \((^2)\)

\[ \Delta^2\xi+b_1(\chi,\xi) = \Delta^2\xi+ \frac{\partial^2\chi}{\partial y^2}\frac{\partial^2\xi}{\partial x^2} - 2\frac{\partial^2\chi}{\partial x\partial y}\frac{\partial^2\xi}{\partial x\partial y} + \frac{\partial^2\chi}{\partial x^2}\frac{\partial^2\xi}{\partial y^2} = P(x,y), \]

\[ 2\Delta^2\chi+b_2(\xi) = 2\Delta^2\chi - 2\frac{\partial^2\xi}{\partial x^2}\frac{\partial^2\xi}{\partial y^2} + 2\left(\frac{\partial^2\xi}{\partial x\partial y}\right)^2 = 0. \]

In the domain \(G\subset \mathbf R^2\) it is required to find a solution of the first boundary-value problem. Set

\[ X=\{u=(\xi,\chi)\mid u\in \dot W_2^{(2)}\},\quad Y=L_1,\quad V=\dot W_2^{(2)},\quad A(u)=(\Delta^2\xi,2\Delta^2\chi),\quad B(u)=(b_1(\xi,\chi),b_2(\xi)). \]

Theorem 3 (see also \((^3,^4)\)). If \(P(x,y)\in W_2^{(-2)}\), then the system of A. Föppl equations is solvable in the space \(\dot W_2^{(2)}\).

Remark. From the smoothness theorem for the biharmonic equation it follows that if \(P(x,y)\) is smooth, then the solution is also smooth.

§ 3. General scheme for nonstationary equations

Notation:

\[ L_p(X)=\left\{u(t):[0,T]\to X\mid \|u\|^p=\int_0^T\|u\|_X^p\,dt<\infty\right\},\quad 1\leq p<\infty. \]

Consider the Cauchy problem for the equation

\[ u'+A(u)+B(u)=h(t),\quad u(0)=0, \tag{4} \]

where the operators \(A\) and \(B\) act in the following scheme:

\[ \begin{array}{ccccc} & \xrightarrow{\ i\ } & L_p(X) & \xrightarrow{\ A\ } & L_{p'}(X^*) \\ L_\infty(V) & & & & \downarrow i \\ & & & & L_1(V^*),\quad p'=\dfrac{p}{p-1}.\\ & \xrightarrow{\ i\ } & L_p(X) & \xrightarrow{\ B\ } & Y \\ & & & & \uparrow i \end{array} \]

Definitions:

\[ F_1=\{u(t)\mid u\in L_\infty(V),\ u'\in L_1(V^*),\ u(0)=0\}, \]

\[ F_2=\{u(t)\mid u\in L_p(X),\ u'\in L_1(V^*),\ u(0)=0\}; \quad [w,v]=\int_0^T \langle w,v\rangle\,dt, \]

\[ w\in L_1(V^*),\quad v\in L_\infty(V). \]

Assumptions. I. Parabolicity. \(u'+A(u)\). For any function \(u(t)\in F_1\),

\[ \lim_{\|u\|\to\infty}\frac{[u'+A(u),u]}{\|u\|}=+\infty. \]

II. Orthogonality of \(B(u)\). If \(u(t)\in F_1\), then

\[ [B(u),u]=0. \]

III. The operators \(A\) and \(B\) are weakly continuous from \(F_2\) into \(L_1(V^*)\).

Theorem 4. If conditions I–III are fulfilled, then for every function \(h(t)\in L_{p'}(X^*)\) there exists at least one solution \(u(t)\in F_2\) of equation (4), in the sense of the identity

\[ [u'+A(u)+B(u),v]=[h,v],\quad \forall v(t)\in L_\infty(V). \]

The proof is carried out by the Galerkin method with the introduction of a vanishing “viscosity” (see, for example, \((^5)\) or \((^6)\)).

§ 4. Applications to Nonstationary Problems

  1. The Euler equations of motion of a rigid body in the principal axes of inertia have the form

\[ I_m \frac{d\Omega_m}{dt}+(I_{m+2}-I_{m+1})\Omega_{m+1}\Omega_{m+2}=K_m(t),\qquad \Omega_m(0)=0, \tag{5} \]

where \(m\) runs through the cyclic permutation of the indices \(1,2,3\).

Put \(G=\cdot;\ X=V=\mathbf{R}^3;\ A\equiv 0;\ B(\vec{\Omega})=\{(I_{m+2}-I_{m+1})\Omega_{m+1}\Omega_{m+2}\},\ m=1,2,3;\ p=2\). Then \(L_2(X)=L_2(0,T)\), \(L_1(V^*)=L_1(0,T)\); further, \(F_1=F_2=\{\vec{\Omega}(t)\mid \vec{\Omega}'(t)\in L_1(0,T),\ \vec{\Omega}(0)=0\}\).

Theorem 5. If \(K_m(t)\) are summable functions, then system (5) has at least one solution \(\vec{\Omega}(t)\in F_1\).

Remark. Since \(F_1\subset C(0,T)\), for continuous \(K_m(t)\) it follows from equations (5) that \(\vec{\Omega}'(t)\) is also continuous.

  1. The nonstationary Navier—Stokes system. Let \(Q=G\times[0,T]\), \(G\subset \mathbf{R}^n\), \(S=\partial G\times[0,T]\);

\[ u'-\Delta u+\sum_{k=1}^{n}u_k\frac{\partial u}{\partial x_k}+\operatorname{grad} p=h(x,t),\qquad \operatorname{div} u=0, \tag{6} \]

\[ u(x,0)=0,\qquad u|_S=0. \tag{7} \]

Put \(X=\{u(x)\mid u\in \mathring{W}_2^{(2)}(G),\ \operatorname{div}u=0\}\), \(V=\{u(x)\mid u\in \mathring{W}_q^{(1)}, q>n,\ \operatorname{div}u=0\}\); \(p=2;\ Y=L_1(Q)\), \(A(u)=-\Delta u\), \(B(u)=\sum_{k=1}^{n}u_k\frac{\partial u}{\partial x_k}\). Then

\[ F_2=\{u(x,t)\mid u\in L_2(\mathring{W}_2^{(1)}),\ u'\in L_1(W_q^{(-1)}),\ u(x,0)=0\}. \]

Theorem 6. If \(h(x,t)\in L_2(X^*)\), then system (6), (7) has at least one solution in \(F_2\).

This theorem refines the theorem of E. Hopf (9) (see also (8)); in particular, the finiteness of \(u'\) in \(L_1(W_q^{(-1)})\) ensures the continuity of \(u(x,t)\) in \(t\) in the sense of the space \(F_2\).

  1. Analogy with example 2 of § 2. Consider in the cylinder \(Q=G\times[0,T]\), \(G\in \mathbf{R}^{2n}\), the first boundary-value problem for the system

\[ \xi' + \Delta^2\xi-\sum_{i=1}^{n}b_1^{i}(\xi,\chi) =P_1(x_1,y_1,\ldots,x_n,y_n,t), \]

\[ \chi' + \Delta^2\chi-\sum_{i=1}^{n}b_2^{i}(\xi) =P_2(x_1,y_1,\ldots,x_n,y_n,t), \tag{8} \]

where \(b_1^{i}(\xi,\chi)\), \(b_2^{i}(\xi)\) are the same as in § 2, with \((x,y)\) replaced by \((x_i,y_i)\).

Putting \(X=\mathring{W}_2^{(2)}(G)\), \(V=\mathring{W}_p^{(2)}\), \(p>n/2\), \(Y=L_1(Q)\), we obtain:

Theorem 7. If \(P_i(\ldots)\in L_2(W_2^{(-2)})\), \(i=1,2\), then system (8) is solvable in the space \(F_2=\{u=(\xi,\chi)\mid u\in L_2(\mathring{W}_2^{(2)}),\ u'\in L_1(W_p^{(-2)}),\ u(0)=0\}\).

Moscow Power Engineering Institute

Received
25 XI 1966

CITED LITERATURE

  1. O. A. Ladyzhenskaya, Mathematical Problems of the Dynamics of a Viscous Incompressible Fluid, Moscow, 1961.
  2. J. L. Lions and E. Magenes, Non-Homogeneous Boundary Value Problems, Moscow, 1965.
  3. I. I. Vorovich, Izv. AN SSSR, Ser. Mat., 19, No. 4, 173 (1955).
  4. N. F. Morozov, DAN, 114, No. 5, 968 (1957).
  5. J. Lions, Bull. Soc. Math. France, 93, 155 (1965).
  6. Yu. A. Dubinskii, Mat. Sb., 67 (109), 4, 609 (1965).
  7. L. Schwartz, Ann. Inst. Fourier, 7, 1 (1957); 8, 1 (1958).
  8. J. Lions, Preprint Singular Perturbations and Some Non-linear Boundary Value Problems, Univ. Wisconsin, Math. Res. Center, 1963.
  9. E. Hopf, Math. Nachr., 4, 213 (1950—1951).

Submission history

UDC 517.944+531/534