On the stability of nonsymmetric equilibrium figures of a rotating viscous incompressible liquid (Q1769403): Difference between revisions

The author considers the classical problem of the stability of a rotating fluid. A finite fluid mass rotates about the \(x_3\)-axis with constant angular velocity \(\omega_0\) and is subject to capillari forces at the boundary and to the forces of self gravitation. The fluid is in equilibrium if it satisfies the equation \[ \sigma H(x)+ \textstyle{{1\over 2}}\cdot \omega^2_0(x^2_1+ x^2_2)+ \kappa U(x)+ p_0= 0,\quad x\in G=\partial F,\tag{1} \] where \(F\) is the volume occupied by the fluid , \(\sigma\), \(\kappa\) are the coefficients of fluid tension and the gravitational constant resp. Moreover, \(p_0\) is a constant, \(H(x)\) twice the mean curvature at \(x\in G\), \(U(x)= \int_F|x- y|^{-1}dx\) the Newtonian potential. It turns out that (1) is the Euler equation of the functional \[ R= \sigma|\Gamma|+ \beta^2/2 \int_\Omega(x^2_1+ x^2_2)\,dx- {1\over 2} \kappa \int_\Omega\int_\Omega|x- y|^{-1} \,dx\,dy- p_0,\tag{2} \] where \(\Gamma= \partial\Omega\), with \(\Omega\subseteq \mathbb R^3\) a domain close to \(F\), having the same volume as \(F\), i.e. \(|\Omega|= |F|\), and with \(|\Gamma|\) the surface measure of \(\Gamma\); moreover, \(\beta= \omega_0 \int_F(x^2_1+ x^2_2)\,dx\). It is also assumed that \(\Omega\) and \(F\) have the same barycentric position. In order to state that \(\Omega\) is close to \(F\) one uses the representation \[ x= y+ n(y)\rho(y),\quad y\in G=\partial F,\quad x\in\partial\Omega,\tag{3} \] where \(n(y)\) is the outward unit normal at \(y\in G\), while \(\rho\in C^1(G)\) is supposed to be small. The various side conditions on \(\Omega\) may then be expressed in terms of \(\rho\). In paricular one can express the functional \(R\) in (1) in terms of \(\rho\) in order to obtain a functional \(R[\rho]\). It turns out that the first variation of \(R\) is \(=0\), \[ \partial_0 R= \partial_\lambda R[\lambda\rho]|_{\lambda=0 ]}= 0.\tag{4} \] It was conjectured by Poincaré and A. M. Lyapunov that sufficient condition of the stability of the equilibrium figure \(F\) underying (2) is the positivity of the second variation of \(R\), i.e. \[ \partial^2_0 R[\rho]= \partial^2_\lambda R[\lambda\rho]|_{\lambda=0}\geq 0.\tag{5} \] This criterion, while never rigorously proved, is now generally accepted. In this paper, the author gives a rigorous proof of a variant of this conjecture . He considers the Navier-Stokes equations \[ \partial_t v= \nu\Delta v-\nabla p-(v\nabla)v,\;\text{div}(v)= 0\quad\text{on }\Omega_t,\tag{6} \] supplied by dynamic and kinematic boundary conditions which involve \(\partial\Omega_t\), the pressure \(p\) and the vector field \(v\). He assumes that he equilibrium figure \(F\) under scrutiny gives rise to a functional \(R\) subject to an inequality \[ c_1\cdot\|\rho\|^2_X\leq \partial^2_0 R[\rho]\leq c_2\cdot\| \rho\|^2_X,\quad X= W^1_2(G),\text{ some }c_i.\tag{7} \] Under some additional smallness conditions on the initial data, the principal result, Thm. 2.1 , then asserts that the equilibrium solution \(F\) is exponentially stable under small perturbations in suitable Hölder spaces. The proof is very difficult and extends over many pages.