The $2+1$ convex hull of a finite set.

2020 
We study $\mathbb{R}^2\oplus\mathbb{R}$-separately convex hulls of finite sets of points in $\mathbb{R}^3$, as introduced in \cite{KirchheimMullerSverak2003}. When $\mathbb{R}^3$ is considered as a certain subset of $3\times 2 $ matrices, this notion of convexity corresponds to rank-one convex convexity $K^{rc}$. If $\mathbb{R}^3$ is identified instead with a subset of $2\times 3$ matrices, it actually agrees with the quasiconvex hull, due to a recent result \cite{HarrisKirchheimLin18}. We introduce "$2+1$ complexes", which generalize $T_n$ constructions. For a finite set $K$, a "$2+1$ $K$-complex" is a $2+1$ complex whose extremal points belong to $K$. The "$2+1$-complex convex hull of $K$", $K^{cc}$, is the union of all $2+1$ $K$-complexes. We prove that $K^{cc}$ is contained in the $2+1$ convex hull $K^{rc}$. We also consider outer approximations to $2+1$ convexity based in the locality theorem \cite[4.7]{Kirchheim2003}. Starting with a crude outer approximation we iteratively chop off "$D$-prisms". For the examples in \cite{KirchheimMullerSverak2003}, and many others, this procedure reaches a "$2+1$ $K$-complex" in a finite number of steps, and thus computes the $2+1$ convex hull. We show examples of finite sets for which this procedure does not reach the $2+1$ convex hull in finite time, but we show that a sequence of outer approximations built with $D$-prisms converges to a $2+1$ $K$-complex. We conclude that $K^{rc}$ is always a "$2+1$ $K$-complex", which has interesting consequences.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    33
    References
    0
    Citations
    NaN
    KQI
    []