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.

458 <strong>Geometry</strong> of Numbers<br />

Proof. Call x, y ∈ Ed congruent modulo L if x − y ∈ L. The points of the form<br />

1<br />

2 l with l ∈ L then fall into 2d congruence classes modulo L. (To see this, let<br />

{b1,...,bd} be a basis of L. Then each point of the form 1 2 l with l ∈ L is congruent<br />

to one of the 2d points<br />

u1<br />

2 b1 +···+ ud<br />

2 bd where ui ∈{0, 1}<br />

<strong>and</strong> these points are pairwise incongruent.) Let o, 1 2 l1,..., 1 2 ln, with li ∈ L <strong>and</strong><br />

n = 2 d − 1 be representatives of these congruence classes. Then, as the translates of<br />

C by the vectors of L cover E d , there are m1,...,mn ∈ L such that 1 2 li ∈ C + mi,<br />

or<br />

1<br />

2 li − mi ∈ C for i = 1,...,n.<br />

These points lie in different congruence classes modulo L <strong>and</strong> are not congruent<br />

to o. Since −( 1 2 li − mi) ≡ 1 2 li − mi modulo L, we see that the points −( 1 2 li − mi)<br />

<strong>and</strong> 1 2 li − mi lie in the same congruence class but clearly are different. Since C is<br />

o-symmetric it follows that<br />

the 2n = 2 d+1 − 2 points ±( 1 2 li − mi), i = 1,...,n,<br />

are pairwise different, different from o <strong>and</strong> all are contained in C.<br />

Thus C meets each of the 2n + 1 = 2 d+1 − 1 distinct bodies C, C ± (li − 2mi),<br />

i = 1,...,n, of the covering. ⊓⊔<br />

31.3 Rogers’s Upper Bound for ϑT (C)<br />

Numerous early attempts to construct or to establish the existence of dense lattice<br />

<strong>and</strong> non-lattice coverings of E d with convex bodies were relatively unsuccessful.<br />

The upper bounds which were obtained all were of the form c d with suitable constants<br />

c > 1. The breakthrough finally was achieved by Rogers who introduced<br />

averaging methods to this problem which led to surprisingly small upper estimates<br />

for the minimum density of a covering of E d by translates of a convex body <strong>and</strong> for<br />

lattice coverings.<br />

In this section we prove Rogers’ upper estimate for the minimum density of a<br />

covering of E d by translates of a convex body. His proof is based on periodic sets<br />

<strong>and</strong> a mean value argument.<br />

For an exposition of the work of Rogers on coverings, the reader may wish to<br />

consult Rogers’ classical Cambridge tract [851].<br />

An Inequality Between δT (C) <strong>and</strong> ϑT (C)<br />

Before embarking on Rogers’ result, we present a result which relates the packing<br />

<strong>and</strong> the covering case <strong>and</strong>, as a consequence, yields a (large) upper bound for ϑT (C).<br />

We point out the nice idea of proof, which might have been a starting point for<br />

Rogers’ ingenious proof, see later.

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

Saved successfully!

Ooh no, something went wrong!