Triangulation and general position of PL diagrams (Q909294)

Consider the diagram (henceforth referred to as the Diagram) consisting of locally compact polyhedra P, Q, R and PL maps f: \(P\to Q\), g: \(P\to R\). It is well known that this Diagram cannot, in general, be triangulated, i.e. there do not exist triangulations of P, Q, and R with respect to which both f and g are simplicial. In the present paper the author gives a sufficient condition for the triangulabity of the Diagram; there should not exist arbitrarily long finite sequences \((q_ n)\) of points in Q such that \(gf^{-1}(q_ n)\cap gf^{-1}(q_{n+1})\neq \emptyset\) for each n. If, moreover, g is surjective, then the Diagram admits a PL (and simplicial) pushout (the surjectivity condition is stated correctly in the abstract but is unfortunately missing in the statement of the theorem). The second major result of the paper is some sort of general position theorem: If R in the Diagram is a PL m-manifold, f is onto, and dim P\(+\dim f^{-1}(q)<m\) for each \(q\in Q\), then g can be approximated by maps which satisfy the triangulability condition stated above. This theorem is then used to prove \textit{R. T. Miller}'s theorem [Ann. Math., II. Ser. 95, 406-416 (1972; Zbl 0208.518)] that every codimension 3 topological embedding of one PL manifold into another can be approximated by PL embeddings. This theorem was first announced by \textit{T. Homma} in his mimeographed notes ``Piecewise linear approximations of embeddings of manifolds'' [Math. Dept., Florida State Univ. (1965)], but later it was realized that this proof required a more restrictive dimension condition than just codimension 3 [cf. \textit{R. T. Miller}, loc. cit.]. The proof in the present paper follows Homma's original idea.