Degenerate parabolic equation with critical exponent derived from the kinetic theory. III: \(\varepsilon \)-regularity. (Q454513)

Consider \(n\geq 3\), \(m=2(n-1)/n\), and a nonnegative integrable and bounded initial condition \(u_0\), and denote the Poisson potential in \(\mathbb {R}^n\) by \(\Gamma \). The solution \(u\) to the generalised Smoluchowski-Poisson equation \[ \partial _t u = \text{div}\left ( \frac {m-1}{m} \nabla u^m - u\;\nabla (\Gamma \ast u) \right ) \quad \text{ in }\quad \mathbb {R}^n\times (0,\infty ) \] with \(u(0)=u_0\) has a different behaviour according to the value of \(| u_0| _1\). Indeed, there is a threshold value \(\lambda _{*}>0\) such that \(u\) exists for all times if \(| u_0| _1\leq \lambda _{*}\) while \(u\) might blow up in finite time if \(| u_0| _1>\lambda _{*}\). For solutions blowing up at a finite time \(T_{\max }\), it is of interest to describe more precisely the blow-up mechanism and it is shown here that it is induced by concentration as for the classical Smoluchowski-Poisson equation in \(\mathbb {R}^2\). The main result actually states that there is a positive number \(\varepsilon _0\) depending only on \(n\) such that, if \[ \limsup _{t\to T_{\max }} \int _{B(x_0,R)} u(x,t)\;dx < \varepsilon _0 \] for some \(x_0\in \mathbb {R}^n\) and \(R>0\), then \(\limsup _{t\to T_{\max }} \sup _{x\in B(x_0,R/2)} u(x,t)\) is finite. In particular, \(x_0\) is not a blow-up point. A similar result is proved in [\textit{Y. Sugiyama}, in: Noncommutativity and singularities. Proceedings of French-Japanese symposia. Tokyo: Mathematical Society of Japan. Adv. Stud. Pure Math. 55, 137--160 (2009; Zbl 1194.35475)] for the same equation but with the Bessel (or Yukawa) potential instead of the Poisson potential. The main difference is that the latter has a slow decay at infinity.