Non-quadratic Euclidean complete affine maximal type hypersurfaces for $\theta\in(0,(N-1)/N]$

Abstract

Bernstein problem for affine maximal type equation \begin{equation}\label{e0.1} u^{ij}D_{ij}w=0, \ \ w\equiv[\det D^2u]^{-\theta},\ \ \forall x\in\Omega\subset{\mathbb{R}}^N \end{equation} has been a core problem in affine geometry. A conjecture proposed firstly by Chern (Proc. Japan-United States Sem., Tokyo, 1977, 17-30) for entire graph and then extended by Trudinger-Wang (Invent. Math., {\bf140}, 2000, 399-422) to its full generality asserts that any Euclidean complete, affine maximal type, locally uniformly convex $C^4$-hypersurface in ${\mathbb{R}}^{N+1}$ must be an elliptic paraboloid. At the same time, this conjecture was solved completely by Trudinger-Wang for dimension $N=2$ and $\theta=3/4$, and later extended by Jia-Li (Results Math., {\bf56} 2009, 109-139) to $N=2, \theta\in(3/4,1]$ (see also Zhou (Calc. Var. PDEs., {\bf43} 2012, 25-44) for a different proof). On the past twenty years, much efforts were done toward higher dimensional issues but not really successful yet, even for the case of dimension $N=3$. Recently, counter examples were found in \cite{Du2} (J. Differential Equations, {\bf269} (2020), 7429-7469) for $N\geq3$ and $\theta\in(1/2,(N-1)/N)$ using a much more complicated argument. In this paper, we will construct explicitly various new Euclidean complete affine maximal type hypersurfaces which are not elliptic paraboloid for the improved range $$N\geq2, \ \ \theta\in(0,(N-1)/N].$$