Leading order down-stream asymptotics of non-symmetric stationary Navier-Stokes flows in two dimensions (Q813305): Difference between revisions

The authors investigate the 2D Navier-Stokes equation \[ -(u \nabla)u+\Delta u-\nabla p=0,\quad \text{div}(u)=0,\quad u=(u_1,u_2),\tag{1} \] in the half space \(\Omega= \{(x,y)\mid x\geq 1\}\). They consider the situation where the fluid enters the surface \(\Sigma=\{(x,y)\mid x=1\}\), and where the flow at infinity is parallel to the \(x\)-axis at nonzero constant speed \(u_\infty=(1,0)\). This leads to the boundary conditions \[ \lim u(x,y)=u_\infty\quad\text{as } x^2+y^2\to\infty,\;x\geq 1,\tag{2} \] \[ u|_\Sigma= u_\infty+ u_*, \quad u_*(y)= (u^*(y),v^*(y)),\quad \lim u_*(y)=0 \text{ as }|y|\to\infty. \] Here \(u_*\) belongs to a Banach space of vector fields \((u(y),v(y))\) which is defined in the course of the arguments. The main result, Theorem 1, states that for \(u_*=(u^*,v^*)\) properly choosen, there are \(U= u_\infty+(u,v)\) and \(p\) such that (1), (2) hold, and exhibiting the asymptotic behaviour \[ \begin{aligned}\lim_{x\to\infty}&x^{\frac 12}\left(\sup_{y\in R}\bigl|(u-u_a) (x,y)\bigr|\right)=0,\\ \lim_{x\to\infty}&x \biggl(\sup_{y\in R}\bigl |(v-v_a)(x,y)\bigr| \biggr)=0\end{aligned}\tag{3} \] where \[ \begin{aligned} u_a&= cx^{-\frac 12}e^{-y^2/4x}+ (x^2+y^2)^{-1}(dx+by),\\ v_a&= (c/2)yx^{-3/2} e^{-y^2/4x}+ (x^2+y^2)^{-1}(dy-bx), \end{aligned} \] with \(b,c,d\) given explicitely in terms of \(u_*\). Preparatory to the proof, the authors transform problem (1), (2) into an equivalent one, having the form of a dynamical system, with \(x\) in the role of time. That is, they consider the system \[ \partial_x\omega=\eta,\;\partial_x\eta= \eta-\partial^2_y\omega+q,\quad \partial_xu=-\partial_yv,\;\partial_xv= \partial_yu+\omega,\tag{4} \] with \(q= \partial_x(u\omega)+\partial_y (v\omega)\). After a series of further transformations involving Fourier transform the authors replace (3) by an equivalent system (3*) of equations, which lives in Fourier space, and involves four unknowns \(\omega_+,\omega_-,u_+,u_-\), related linearly to \(\omega,\eta,u, v\). The system (3*) is then transformed into a complicated system of nonlinear integral equations which forms the basis for the further arguments of the proof. After the introduction of suitable function spaces, the authors use a contraction mapping argument in order to show that the integral equation in question has a fixed point, which, after backward computation, gives rise to a solution of the original system (1), (2). The difficulties encountered in the verification of the estimates required by the fixed point argument are quite formidable and are relegated to a lengthy appendix.