Horizons non-differentiable on a dense set (Q1270145)

The authors prove the sharpness of a result in [\textit{S. W. Hawking} and \textit{G. F. R. Ellis}, `The large scale structure of space-time' (Camb. Mon. Math. Phys. 1, Camb. Univ. Pr., London) (1973; Zbl 0265.53054) (Prop. 6.3.1) and \textit{R. Penrose}, `Techniques of differential topology in relativity' (Regional Conference Series in Applied Mathematics 7, Society for Industrial and Applied Mathematics, Philadelphia) (1972; Zbl 0321.53001) (Lemma 3.17)] on the Lipschitz regularity of the boundary of a future set. They construct a connected set \(K\subset{\mathbb{R}}^2\) such that the union of future cones in \(2+1\) Minkowski space with vertex in \(K\) has a non-differentiable boundary. This boundary \(H^+(K)\) is interpreted as a Cauchy horizon in a spacetime obtained by cutting off suitable sets from Minkowski spacetime. Such examples are causal and asymptotically flat, but have an incomplete null infinity. The proof is achieved by constructing a decreasing sequence \(K_i\) of sets having more and more cusps on their boundary, and converging to the desired set \(K\). To prove that the creases in the \(K_i\) do lead to points of non-differentiability in \(H^+(K)\), one needs to control convergence properties of semi-tangents to \(H^+(K_i)\), which requires most of the work. It seems that some of the technicalities could be simplified by a systematic use of the contingent and paratingent cones.