The Liouville theorem and the \(L^{2}\) decay for the FENE dumbbell model of polymeric flows (Q522186)

The authors consider the finite extensible nonlinear elastic dumbbell model \( u_{t}+(u\cdot \nabla u)u-\nu \Delta u+\nabla P=\operatorname{div} \tau \), \(\operatorname{div} u=0\), \(\psi _{t}+(u\cdot \nabla u)\psi =\operatorname{div}(-\sigma (u)\cdot R\psi +\nabla _{R}\psi +\nabla _{R} \mathcal{U} \psi )\), posed in \(B\times (0,\infty )\) where \( B\) is the unit ball of \(\mathbb{R}^{d}\), \(d\geq 2\). Here \(u\) is the velocity of the polymeric liquid, \(\psi \) is the distribution function for the internal configuration, \(P\) is the pressure, \(\sigma (u)\) is the drag term (equal to \(\nabla u\), in the general case or to \(\nabla u-(\nabla u)^{T}\) , in the co-rotation case) and \( \mathcal{U} (R)=-k\log (1-|R|^{2})\) is the potential for some positive constant \(k\). Initial conditions \(u_{0}\) and \(\psi _{0}\) are respectively imposed to \(u\) and \(\psi \). The boundary condition \((\nabla _{R}\psi +\nabla _{R} \mathcal{U} \psi )\cdot n=0\) is imposed on the boundary of the ball \(B(0,1)\). The authors observe that \((0,\psi _{\infty })\) is a solution to this problem with \(\psi _{\infty }=\exp (-\mathcal{U} (R))/\int_{B}\exp (-\mathcal{U} (R))dR\). The first main result of the paper proves that if \((u,\psi )\) is a bounded suitable stationary weak solution to the problem such that \( \int_{B(0,1)}\psi dR=1\) and there exist constants \(C_{1}\), \(C_{2}\) such that \(0<C_{1}\leq \psi /\psi _{\infty }\leq C_{2}\), and if \(u\in (L^{\frac{3d}{d-1 }}(\mathbb{R}^{d}))^{d}\) then \(u=0\) and \(\psi =\psi _{\infty }\). The second main result proves that if \((u,\psi )\) is a bounded suitable stationary weak solution to the problem such that \(\int_{B(0,1)}\psi dR=1\), there exist constants \(C_{1}\), \(C_{2}\) such that \(0<C_{1}\leq \psi /\psi _{\infty }\leq C_{2}\) and if there exist constants \(1\leq p_{i}\), \(q_{i}\), \(r_{i}<\infty \) ( \(i=1,2,3\)) such that \(u_{i}\in L_{x_{1}}^{p_{1}}L_{x_{2}}^{p_{2}}L_{x_{3}}^{p_{3}}\) and \( 1/p_{i}+1/q_{i}+1/r_{i}=2/3\) (\(i=1,2,3\)) then \(u=0\) and \(\psi =\psi _{\infty }\). The authors first define the notion of suitable weak stationary solution using a variational formulation of the problem, they introduce a smooth cut-off function \(\eta _{K}\) equal to 1 if \(\left| x\right| \leq K\) and to 0 if \(\left| x\right| \geq 2K\) and they take \(v=u\eta _{K}\) as a test function in this variational formulation. They then estimate the different terms of this variational formulation using classical arguments and direct computations. In the last part of their paper, the authors prove \( L^{2}\) decay estimates, first in the co-rotation case and finally in the general case, assuming appropriate hypotheses on the initial data of the problem. In the co-rotation case, the authors assume that \(u_{0}\in L^{2}\cap L^{1}\), that \(\psi _{0}-\psi _{\infty }\in L_{x}^{2}(\mathcal{L} ^{2})\) with \(\int_{B}\psi _{0}=1\) for almost every \(x\). In the co-rotation case, the authors prove that \(\int_{\mathbb{R}^{d}\times B}\frac{\left| \psi -\psi _{\infty }\right| ^{2}}{\psi _{\infty }}dxdR\leq Ce^{-Ct}\), \( \left\| u\right\| _{L^{2}}\leq C(1+t)^{-d/4}\) if \(d\geq 3\) and \( \left\| u\right\| _{L^{2}}\leq C_{l}\ln ^{-l}(e+t)\) if \(d=2\), for every integer \(l\), \(C_{l}\) depending on \(l\). In the general case, the authors first recall the existence of a unique global strong solution proved by \textit{N. Masmoudi} [Invent. Math. 191, No. 2, 427--500 (2013; Zbl 1258.35160)], assuming \(H^{s}\), \(s>1+d/2\), regularity properties of the initial data and that \(\left\| u_{0}\right\| _{H^{s}}^{2}+\left\| \psi _{0}-\psi _{\infty }\right\| _{H^{s}(\mathcal{ L}^{2})}^{2}\) is less than some constant \(\varepsilon _{0}\). They here prove \(\int_{\mathbb{R}^{d}}\left| u\right| ^{2}dx\leq +\int_{ \mathbb{R}^{d}\times B}\frac{\left| \psi -\psi _{\infty }\right| ^{2} }{\psi _{\infty }}dxdR\) is less than \(C(1+t)^{1-d/2}\) if \(d\geq 3\) or than \( C\ln ^{-1}(1+t)\) if \(d=2\). For the proof, the authors start from the variational formulation in which they take \(v=u\) as test function and they proceed through direct computations. In the co-rotation case, they link their results to \textit{M. E. Schonbek}'s conjecture in [SIAM J. Math. Anal. 41, No. 2, 564--587 (2009; Zbl 1189.35263)].