Exponential speed of uniform convergence of the cell density toward equilibrium for subcritical mass in a Patlak-Keller-Segel model (Q473058)

This paper is concerned with the following initial-boundary-value problem in one spatial dimension: \[ \begin{alignedat}{2} &u_t=x^{2-2/N}u_{xx}+uu_x^{2/N}&\quad &\text{for }t>0,x\in(0,1],\\ & u(t,0)=0,\,u(t,1)=m &\quad &\text{for }t\geq 0,\\ &u_x(t,x)\geq 0 &\quad &\text{for }t>0,x\in[0,1],\end{alignedat} \] where \(N\in\mathbb N\) with \(N\geq 2\), \(m\geq 0\), and an initial function \(u(0)\in C([0,1])\) compatible with the given boundary conditions. This system is related to the Patlak-Keller-Segel model posed on the unit ball in \(\mathbb R^N\) by a change of variables: Given a solution \(u\) to the system above, \(\rho(t,r)=N^{N}u_x(N^2t,r^N)\) defines a radially symmetric solution to \[ \rho_t=\Delta \rho-\operatorname{div}(\rho^{2/N}\nabla c),\quad -\Delta c=\rho, \] with no-flux and Dirichlet boundary conditions, respectively. The parameter \(m\) in the model above corresponds (up to a multiplicative factor) to the total mass of \(\rho\), which is a conserved quantity. It was shown in [the author, Nonlinearity 26, No. 9, 2669--2701 (2013; Zbl 1273.35157)] and [Discrete Contin. Dyn. Syst., Ser. B 19, No. 1, 231--256 (2014; Zbl 1286.35005)] that the system at hand exhibits a mass threshold phenomenon: There exists a critical mass \(M>0\) such that if \(m\in [0,M)\), the (unique classical) solution \(u\) exists globally in time and converges to the unique stationary state \(U_a\) w.r.t. \(C^1([0,1])\) as \(t\to\infty\). On the other hand, if \(m>M\), the solution blows up in finite time. For example, \(N=2\) yields the critical mass \(M=2\), corresponding to a critical mass of \(8\pi\) for radial solutions of the Patlak-Keller-Segel system in two dimensions. The main result of this work is a quantitative one for the subcritical regime \(m<M\): The solution \(u\) converges exponentially fast to the stationary state \(U_a\) as \(t\to\infty\) in \(C^1([0,1])\).\newline The proof is based on the following modified Hardy inequality on the space \newline \(H=\{h\in L^2((0,1),\frac{dx}{x^{2-2/N}}):\,\dot h\in L^2(0,1),\,h(0)=0=h(1)\}\): \[ \int_0^1 \frac{\dot h^2}{\dot U_a^{2/N}} dx\geq \lambda_1 \int_0^1 \frac{h^2}{x^{2-2/N}}dx \] for some \(\lambda_1>1\), which is established in Section 3 of the paper. Heuristically speaking (for \(N\geq 3\)), this Hardy-type inequality is equivalent to uniform convexity of the functional \[ \mathcal{F}[u]=\int_0^1\left(\frac{\dot u^{2-2/N}}{(2-\frac2{N})(1-\frac2{N})}-\frac{u^2}{2x^{2-2/N}}\right)dx \] on the ``manifold'' \(\mathcal{M}=\{u\in C^1([0,1]):\,\dot u>0,\,u(0)=0,\,u(1)=m\}\), with respect to the ``Riemannian metric'' \[ g_u(h,k)=\int_0^1 \frac{hk}{x^{2-2/N}\dot u^q}dx \] for tangent vectors \(h,k\) at \(u\in\mathcal{M}\). Formally, the problem above possesses a gradient flow structure w.r.t. these objects. However, this structural observation mostly serves as a motivation for the subsequent proof. The author first proves exponential convergence (at a rate \(\lambda\in(0,\lambda_1-1)\)) in \(L^2(0,1)\) (cf. Section 4). In a nutshell, Gronwall's lemma is applied to the map \(g_{U_a}(u(t)-U_a,u(t)-U_a)\). Transforming to a problem in \(\mathbb R^N\) and using heat semigroup estimates after a regularization process, exponential convergence to equilibrium is shown also in \(C^1([0,1])\), see Section 5.