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.

98 <strong>Convex</strong> Bodies<br />

Thus<br />

Second, Minkowski’s theorem 6.5 on mixed volumes shows that, in particular,<br />

V (εC + P) = V (P,...,P) + dV(C, P,...,P)ε + ...<br />

(5) V (C, P,...,P) = 1<br />

d lim<br />

ε→+0<br />

+ dV(C,...,C, P)ε d−1 + V (C,...,C)ε d<br />

= V (P) + dV(C, P,...,P)ε +···+V (C)ε d .<br />

V (εC + P) − V (P)<br />

.<br />

ε<br />

Third, in the following we distinguish two cases. In the first case let o ∈ C.<br />

Choose σ>0 such that C ⊆ σ Bd . The convex body εC + P then may be dissected<br />

into P, into cylinders, possibly slanting, with the facets as bases <strong>and</strong> into the remaining<br />

part of εC + P. The cylinders can be described as follows: for a facet F of P with<br />

exterior unit normal vector u F choose a point p ∈ C ∩ HC(u F). The corresponding<br />

cylinder then is F + ε[o, p]. The remaining part of εC + P is contained in the union<br />

of sets of the form εC + G ⊆ εσ Bd + G, where G ranges over the faces of P with<br />

dim G ≤ d − 2 <strong>and</strong> thus has volume O(ε2 ) by (4). Hence<br />

(6) V (εC + P) = V (P) + ε �<br />

hC(u F)v(F) + O(ε 2 ) as ε → 0.<br />

F facet of P<br />

Finally, (5) <strong>and</strong> (6) together yield (3) in case where o ∈ C.<br />

In the second case let o �∈ C. Choose t ∈ E d such that o ∈ C + t. Then<br />

V (C, P,...,P) = V (C + t, P,...,P)<br />

= 1 �<br />

hC+t(u F)v(F)<br />

d<br />

F facet of P<br />

= 1 �<br />

hC(u F)v(F) +<br />

d<br />

1<br />

d<br />

F facet of P<br />

by Proposition 6.5 <strong>and</strong> the first case. Now note that<br />

�<br />

F facet of P<br />

t · u F v(F) =�t� �<br />

F facet of P<br />

�<br />

F facet of P<br />

t · u F v(F)<br />

t<br />

�t� · u F v(F) = 0<br />

since the latter sum equals the sum of the area of the projection of bd P onto the<br />

hyperplane through o orthogonal to t, taken once with a + sign <strong>and</strong> once with a −<br />

sign. This concludes the proof of (3) in case where o �∈ C. ⊓⊔<br />

For later reference we state the following consequence of (5), where for the<br />

Minkowski surface area, see Sect. 6.4.<br />

Corollary 6.1. Let P ∈ P. Then d V (B d , P,...,P) equals the (Minkowski) surface<br />

area of P.<br />

The next result is a refinement of Lemma 6.4.

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

Saved successfully!

Ooh no, something went wrong!