Asynchronous exponential growth of the growth-fragmentation equation with unbounded fragmentation rate (Q2188070): Difference between revisions

The rate of convergence to equilibrium is studied for solutions to the growth-fragmentation equation \[ \partial_t f(t,x) + \partial_x (\tau(x) f(t,x)) = \int_0^1 B\left( \frac{x}{z} \right) f\left( t , \frac{x}{z} \right) \frac{\mathfrak{p}(dz)}{z} - B(x) f(t,x) \] for \((t,x)\in (0,\infty)^2\), supplemented with the boundary condition \(f(t,0)=0\) for \(t>0\) and an initial condition \(f(0,x)=f^{in}(x)\) for \(x>0\). The growth rate \(\tau\in C^1(0,\infty)\) is positive with \(1/\tau\in L^1(0,1)\) and there are \(\nu_0\le 1\) and \(\tau_1\ge \tau_0> 0\) such that \[ \tau_0 \mathbf{1}_{(1,\infty)}(x) x^{\nu_0} \le \tau(x) \le \tau_1 \max\{1,x\}, \quad x>0, \] and the total fragmentation rate \(B\in C(0,\infty)\) is non-negative with connected support and there are \(0<\gamma_0\le \gamma_1\), \(0<B_0\le B_1\), and \(x_0>0\) such that \[ B_0 \mathbf{1}_{(x_0,\infty)}(x) x^{\gamma_0} \le B(x) \le B_1 \max\{ 1 , x^{\gamma_1}\}, \quad x>0. \] Finally, the fragmentation kernel \(\mathfrak{p}\) is a finite positive measure on \((0,1)\) satisfying \[ \int_0^1 z \mathfrak{p}(dz)=1\ \text{ and }\ \underline{\alpha} := \inf\left\{ \alpha\in \mathbb{R}: \int_0^1 z^\alpha \mathfrak{p}(dz) < \infty \right\} \in [-\infty,0]. \] It is by now well-known that these assumptions guarantee that there are \(\lambda>0\) and a non-negative function \(G\in L^1(0,\infty)\) such that \[ G\in \bigcap_{\alpha\ge 0} L^1((0,\infty),(1+x)^\alpha), \quad \|G\|_{L^1}=1, \] and \((t,x) \mapsto e^{\lambda t} G(x)\) is a particular solution to the growth-fragmentation equation. In addition, this solution is an attractor for the dynamics, in the sense that \(f(t) e^{-\lambda t}\) converges to \(C_* G\) as \(t\to\infty\) in a suitable topology with a well-identified constant \(C_*\) depending only on \(\tau\), \(B\), \(\mathfrak{p}\), and \(f^{in}\). It is the question of temporal decay estimates for the difference \(f(t)e^{-\lambda t} - C_* G\) which is investigated in the paper under review. Specifically, it is shown that, given \(\alpha>\max\{1 , \underline{\alpha} + 2(\gamma_1-\gamma_0)\}\), there are \(M>0\) and \(\sigma>0\) such that \[ \| f(t)e^{-\lambda t} - C_* G \|_{L^1((0,\infty),(1+x)^\alpha dx)} \le M e^{-\sigma t} \|f^{in}\|_{L^1((0,\infty),(1+x)^\alpha dx)} \] for all \(f^{in}\in L^1((0,\infty),(1+x)^\alpha dx)\) when, either \(\mathfrak{p}\) is absolutely continuous with respect to the Lebesgue measure on \((0,1)\), or \(\mathfrak{p}\) is supported in \((\epsilon,1-\epsilon)\) for some \(\epsilon\in (0,1)\) and \(\tau\) is constant. Besides, the optimality of the integrability of \(1/\tau\) on \((0,1)\) is discussed in the last section. Owing to the linearity of the growth-fragmentation equation, the proof mainly relies on semigroup arguments. An intermediate result is that some moments, which are initially infinite, become instantaneously finite for positive times under the action of the evolution equation.