Non-compact lamination convex hulls. (Q1810730)

The lamination convex hull of a set \(K\) of \(m\times n\) real matrices is the smallest set \(L(K)\) that contains \(K\) and has the property that \((1-\lambda)A+\lambda B\in L(K)\) whenever \(A, B\in K\), \(\text{rank}(A-B) = 1\) and \(\lambda \in (0,1).\) The author proves by means of an example that a compact set \(K\) of symmetric \(2\times 2\) matrices need not have a compact lamination convex hull. A similar result is proved for separately lamination convex hulls in \(\mathbb R^3\) (the separately lamination convex hull of \(K\subset \mathbb R^n\) is the smallest set containing \(K\) and every segment with endpoints in \(K\) which is parallel to one of the coordinate axis) and for bi-convex hulls in \(\mathbb R^2\times \mathbb R\) (the bi-convex hull of \(A\subset \mathbb R^k\times \mathbb R^l\) is the smallest set with convex sections that contains \(A\)). It is also shown that the mapping assigning to \(K\) the closure of \(L(K)\) is not upper semicontinuous on the space of diagonal \(3\times 3\) matrices with respect to the Hausdorff metric.