Partial regularity at the first singular time for hypersurfaces evolving by mean curvature (Q1952279)

The author considers a family of smooth properly immersed hypersurfaces \(\mathcal{M}=(M_t)_{t\in (0,t_0)}\) in \(\mathbb{R}^{n+1 }\) evolving by mean curvature. A point \(x_0\) in the set \(\mathrm{sing}_{t_0}\mathcal{M}\) of singular points of \(\mathcal{M}\) at \(t=t_0\), is a point such that for some sequences \(t_j\nearrow t_0\) and \(x_j\in M_{t_j}\) , \(x_j\to x_0\), \((M_t)_{t\in (0,t_0)}\) has no smooth extension for some \(t\geq t_0\) in any neighbourhood of \(x_0\). A smooth properly immersed solution of the mean curvature flow \((M_t)_{t\in (0,t_0)}\) inside an open set \(U\subset \mathbb{R}^{n+1}\) satisfies the area continuity and unity density hypothesis at time \(t_0\), if in the sense of Radon measures, and inside \(U\), the hypersurfaces \(M_t\) converge for \(t\nearrow t_0\) to an \(\mathcal{H}^n\)-measurable, countably \(n\)-rectifiable set \(M_{t_0}\) of locally finite Hausdorff \(\mathcal{H}^n\)-measure. Under this condition, Brakke's regularity theorem states that at \(t_0\), \(\mathcal{H}^n(\mathrm{sing}_{t_0}\mathcal{M}\cap U)=0\). In particular, the evolving hypersurfaces do not form multiple sheets in a set of positive measure, as for example Cantor sets. The author derives in main Theorem 1.1 some Brakke-type theorems for closed, immersed, mean-convex solutions without the area continuity and unity density assumption, generalizing results of \textit{H.-W.\ Xu} et al. [Asian J.\ Math.\ 15, No. 4, 549--556 (2011; Zbl 1242.53085)] and of \textit{N. Q.\ Le} and \textit{N.\ Sesum} [Math.\ Z.\ 267, No. 3--4, 583--604 (2011; Zbl 1216.53060)]: If \(\mathcal{M}=(M_t)_{t\in(0, t_0)}\) is a smooth properly immersed solution of the mean curvature flow inside an open ball \(B_R(y)\) of \(\mathbb{R}^{n+1}\) with finite total \(L^p\)-norm \[ \int_0^{t_0}\int_{M_t\cap B_{R}(y)}|A|^p<\infty \] of the second fundamental form for some \(p\geq 2\), then \(\mathcal{H}^{n+2-p}(\mathrm{sing}_{t_0}\mathcal{M}\cap B_R(y))=0\) for \(p\in [2,n+2]\), and \(\mathrm{sing}_{t_0}\mathcal{M}\cap B_R(y)=\emptyset\) for \(p\geq n+2\). This result also implies Brakke's result for closed immersed mean convex hypersurfaces, not necessarily embedded, and with no need of area continuity and unit density hypothesis. Furthermore, the author recalls that even in the embedded case, we do not know yet if the integrability assumption holds with \(p=3-\alpha\) for all small \(\alpha>0\). It holds for known smooth examples, and does not hold for several other examples, namely the case of the two-dimensional shrinking cylinder with axis containing the origin, that has a singular set of dimension one. Some considerations are given concerning the dimension of the singular set, as well on how valid or how optimal is the integrability condition. Several corollaries are derived, as for example, among others of other type, for solution hypersurfaces that are closed and \(k\)-convex the integrability condition holds with \(p=n+3-k-\alpha\), for small positive \(\alpha\). Examples and counterexamples are given illustrating the conditions of the corollaries. The proof of Theorem 1.1 is a consequence of a local \(\epsilon\)-regularity result with the \(L^p\)-norm of the second fundamental form \(A\) given in Theorem 1.8., improving a previous result of the author in [Calc. Var. Partial Differ. Equ. 3, No. 1, 107--126 (1995; Zbl 0814.53004)], using a similar scaling argument.