On nondegenerate triangulation with condensation at the boundary of a domain (Q548728)

Let \(\Omega\) be a connected domain in the plane with smooth boundary \(\partial\Omega\). The authors construct an infinite triangulation of the neighborhood \(\Pi_{(\epsilon)}\) of the boundary consisting of those points in the interior of \(\Omega\) that lie within a distance \(\epsilon\) from \(\partial\Omega\), where \(\epsilon>0\) is suitably small. The construction starts by partitioning \(\Pi_{(\epsilon)}\) into strips \(\Pi_i=\Pi_{(\epsilon/2^i)}\setminus\Pi_{(\epsilon/2^{i+1})}\) of width \(\epsilon/2^{i+1}\) (for \(i\geq 0\)), and proceeds by associating to each strip a triangulated region \({\mathcal T}_i\), whose boundary consists of two polygonal paths that approximate the two boundary curves of \(\Pi_i\), and whose vertices lie only on this boundary. The region \({\mathcal T}_{i+1}\) is constructed to contain twice as many triangles as \({\mathcal T}_i\), and the angles of the triangles in \({\mathcal T}_{i+1}\) naturally correspond to angles of the triangles in \({\mathcal T}_i\). It is shown that the angles of the triangles of \({\mathcal T}_{i+1}\) are sufficiently close to the corresponding angles of \({\mathcal T}_i\) to guarantee that the resulting triangulation of \(\Pi_{(\epsilon)}\) is nondegenerate, provided that the triangles in the initial strip \({\mathcal T}_0\) are chosen appropriately. The article is intended for a general audience and uses essentially only elementary arguments, although the estimates carried out are intricate. There a few typographical errors and the exposition is rough in spots, although this is possibly a byproduct of the translation process.