14.02.2013 Views

Gruber P. Convex and Discrete Geometry

Gruber P. Convex and Discrete Geometry

Gruber P. Convex and Discrete Geometry

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

14 Preliminaries <strong>and</strong> the Face Lattice 257<br />

the contrary, there is a hyperplane HFG which contains all vertices of F <strong>and</strong> G <strong>and</strong><br />

thus F <strong>and</strong> G. LetHG be a support hyperplane of P with G = HG ∩ P. Clearly<br />

HG �= HFG. Rotate HG keeping the (d − 2)-dimensional plane HG ∩ HFG fixed to<br />

the first position, say H, where it contains a further vertex v of P, not in HFG. Then<br />

H is a support hyperplane of P <strong>and</strong> H ∩ P is a face E of P with E � G <strong>and</strong> such<br />

that E ∩ HFG = G. Hence E ∩ F =∅. This contradicts the maximality of G. ⊓⊔<br />

14.4 <strong>Convex</strong> Polytopes <strong>and</strong> Simplicial Complexes<br />

A (finite) simplicial complex C in E d is a family of finitely many simplices in E d such<br />

that for any S ∈ C each face of S is also in C <strong>and</strong> for any S, T ∈ C, the intersection<br />

S ∩ T is a face of both S <strong>and</strong> T . In topology a (convex or non-convex) polytope is<br />

defined to be the union of all simplices of a simplicial complex in E d , see Alex<strong>and</strong>roff<br />

<strong>and</strong> Hopf [9] or Maunder [698]. The problem arises, whether a convex polytope P in<br />

the sense of convex geometry can be obtained in this way. In some cases even more is<br />

dem<strong>and</strong>ed, the vertices of the simplices all should be vertices of P. If the latter holds,<br />

we speak of a simplification of P. That simplifications always exist seems to be well<br />

known. Anyhow, this result is used by several authors, including Macdonald [675],<br />

Ehrhart [292, 293] <strong>and</strong> Betke <strong>and</strong> Kneser [108] without further comment. Thus it is<br />

important in our context, see Sects. 19.1, 19.2 <strong>and</strong> 19.4. For convex polytopes a proof<br />

was communicated to the author by Peter Mani-Levitska [685] <strong>and</strong> only then we<br />

found the proof of Edmonds [286] in the literature. Edmonds’s idea is also described<br />

by Lee [635, 636]. Since these proofs are essentially different, <strong>and</strong> each contains an<br />

interesting idea, both are presented.<br />

<strong>Convex</strong> Polytopes have Simplifications<br />

Our aim is to prove the following result.<br />

Theorem 14.9. Let P ∈ P. Then P has a simplification.<br />

The k-skeleton k-skel P of a convex polytope P in E d is the union of its faces of<br />

dimension ≤ k. Byasimplification of k-skel P we mean simplifications of the faces<br />

of P of dimension ≤ k which fit together at common sub-faces. Let f0(P) denote<br />

the number of vertices of P.<br />

Proof (by Edmonds). We may assume that P is proper. By induction, the following<br />

will be shown.<br />

(6) Let k ∈{0, 1,...,d}. Then each face of P of dimension ≤ k has a simplification.<br />

These simplifications together form a simplification of k-skel P.<br />

Consider a linear ordering of the set of vertices of P. Clearly, (6) holds for k = 0.<br />

Assume now, that k > 0 <strong>and</strong> that (6) holds for k − 1. We construct a simplification<br />

of k-skel P as follows. Let F be a face of P with dim F = k. By the induction<br />

assumption, all proper faces of F have a simplification. Let p be the first vertex of

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!