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

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

31 Covering with <strong>Convex</strong> Bodies 461<br />

by (1). Since the mean value extended over all choices of m points in F of the<br />

proportion in question is (1 − 1/s d ) m , there is at least one choice X ={x1,...,xm}<br />

of m points in F for which this proportion is at most (1 − 1/s d ) m , concluding the<br />

proof of (4).<br />

Second, we prove the following proposition where the set X is as in (4).<br />

(5) Let 0 < t ≤ 1/d <strong>and</strong> Y ={y1,...,yn} ⊆F with n maximum such that<br />

the family � − tC + l + y : l ∈ L, y ∈ Y � is a packing contained in<br />

E d \(C + L + X). Then<br />

n ≤ sd<br />

td �<br />

1 − 1<br />

sd �m .<br />

This is easy. First, the proportion of space covered by the packing � −tC+l + y : l ∈<br />

L, y ∈ Y � , that is its density, is nt d V (C)/V (F) = nt d /s d by (1) <strong>and</strong> the choice of<br />

L = s Z d . Second, this packing is contained in E d \(C + L + X) <strong>and</strong> the proportion<br />

of space covered by the latter set is at most (1 − 1/s d ) m by (4). The proof of (5) is<br />

complete.<br />

Third, note (4) <strong>and</strong> (5). In order to prove that<br />

(6)<br />

� (1 + t)C + l + z : l ∈ L, z ∈ X ∪ Y � is a covering,<br />

let w ∈ Ed . We have to show that w is contained in at least one of the bodies in (6).<br />

By (2), (3) <strong>and</strong> since by (5) 0 < t ≤ 1 d , the bodies −tC+m+w, m ∈ L, are pairwise<br />

disjoint. We distinguish two cases. First, (−tC + L + w) ∩ (C + L + X) �= ∅. Then<br />

there are p, q ∈ C, l, m ∈ L, xi ∈ X such that −tp + l + w = q + m + xi, or<br />

w = tp + q − l + m + xi ∈ (1 + t)C + L + X. Second, if the first alternative does<br />

not hold, then −tC + L + w ⊆ Ed � � \(C + L + X). By the maximality of the family<br />

− tC + l + y : l ∈ L, y ∈ Y we then have (−tC + L + w) ∩ (−tC + L + Y ) �= ∅.<br />

Thus there are p, q ∈ C, l, m ∈ L, y j ∈ Y such that −tp+l + w =−td+ m + y j ,<br />

or w = tp + tq − l + m + y j ∈ (1 + t)C + L + Y by (2) <strong>and</strong> since 0 < t ≤ 1/d<br />

by (5). This concludes the proof of (6).<br />

Fourth, noting (1), the definition of L = s Zd , Propositions (4) <strong>and</strong> (5) show that<br />

(7) the density of the covering in (6) is<br />

(1 + t) d (m + n)<br />

s d<br />

≤ (1 + t) d<br />

�<br />

m 1<br />

+<br />

sd td �<br />

1 − 1<br />

sd �m� .<br />

By choosing s, m, t suitably this finally yields the desired upper bound for a covering<br />

of E d by translates of (1 + t)C, <strong>and</strong> thus for a corresponding covering of E d by<br />

translates of C:Let<br />

m 1<br />

= d log<br />

sd t<br />

, t =<br />

1<br />

d log d .

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

Saved successfully!

Ooh no, something went wrong!