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.

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

Hlawka [514] described the idea of his proof of the Minkowski–Hlawka theorem<br />

as follows:<br />

This theorem requires one to pick out from the infinite set of lattices, lattices which<br />

have a particular property. Since it is difficult to find such lattices by chance, it is<br />

plausible to make the following comparison: consider the problem of catching fish<br />

of given length from a pond. Making one haul, one may catch such a fish only by<br />

chance. For this reason it makes sense to catch many fish, hoping that a fish of the<br />

desired length is among them. In probability theory this is called a r<strong>and</strong>om sample.<br />

For this sample one considers the mean value, in our case the mean value of the<br />

length, to get information on the sample. This was the idea which I applied in 1942 ...<br />

The two proofs given earlier – as all other known proofs, including Hlawka’s<br />

original proof – are based on mean value arguments dealing with huge sets of lattices.<br />

This fact prevents the effective construction of lattices as specified in the theorem. In<br />

particular, this explains why, so far, there is no effective algorithm available to construct<br />

lattice packings of o-symmetric convex bodies of density at least 2 −d , although<br />

such packings exist by the Minkowski–Hlawka theorem, compare Theorem 30.4. For<br />

balls, the situation is slightly better but by no means satisfactory. There are constructions<br />

of rather dense lattice packings of balls using codes. In many cases the codes<br />

can be given effectively. Unfortunately the codes used by Rush [862] to reach the<br />

Minkowski–Hlawka bound 2 −d+o(d) are not of this type. See Sect. 29.3.<br />

Refinements<br />

There are several refinements of the Minkowski–Hlawka theorem, see <strong>Gruber</strong> <strong>and</strong><br />

Lekkerkerker [447]. The best known estimate is due to Schmidt [893]: there is an<br />

absolute constant α > 0 such that, for each Borel set B in E d with V (B) ≤<br />

d log √ 2 − α, there is a lattice in E d of determinant 1 which contains no point of<br />

B except, possibly, o. At present, it is the belief of many people working in the<br />

geometry of numbers that, in essence, the theorem of Minkowski–Hlawka cannot be<br />

refined. In the past, this was not always so, but Edmund Hlawka [517] told the author<br />

that he was always convinced that no essential refinement was possible.<br />

24.2 Siegel’s Mean Value Theorem <strong>and</strong> the Variance Theorem<br />

of Rogers–Schmidt<br />

A natural question to ask in the geometry of numbers is the following: Given a property<br />

which a lattice may or may not have, is the set of lattices which have this property,<br />

large or small? Tools which sometimes help to give an answer are measure<br />

<strong>and</strong> Baire categories. Measure has turned out to be a versatile tool which applies<br />

to many such questions, ever since Siegel [936] defined <strong>and</strong> put to use a natural<br />

measure on the space of all lattices of determinant 1. Later contributions are due to<br />

Rogers, Macbeath <strong>and</strong> Schmidt. Because of great technical difficulties, in particular<br />

in the work of Schmidt, the development seems to have reached a deadlock. Despite<br />

this, we believe that many important measure results in the geometry of numbers

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

Saved successfully!

Ooh no, something went wrong!