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.

11 Approximation of <strong>Convex</strong> Bodies <strong>and</strong> Its Applications 203<br />

Hausdorff metric δ H <strong>and</strong> then by <strong>Gruber</strong> [427] for the symmetric difference metric<br />

δ V . These results gave rise to a long series of further investigations.<br />

Early contributions to r<strong>and</strong>om approximation are due to Crofton, Czuber <strong>and</strong><br />

Blaschke, the first asymptotic results were proved by Rényi <strong>and</strong> Sulanke [831] in<br />

case d = 2. Modern results for general d are due, amongst others, to Bárány <strong>and</strong><br />

Buchta [69] <strong>and</strong> Reitzner [829].<br />

In this section we first consider John’s characterization of the ellipsoid of maximum<br />

volume contained in a convex body <strong>and</strong> use it to prove Ball’s reverse<br />

isoperimetric inequality. Then the author’s asymptotic formula for the best volume<br />

approximation of a convex body by circumscribed convex polytopes is given<br />

as the number of facets tends to infinity. This is then applied to the isoperimetric<br />

problem for convex polytopes.<br />

For references concerning John’s theorem <strong>and</strong> asymptotic best approximation,<br />

see the following sections. Recent surveys on r<strong>and</strong>om approximation are due to the<br />

author [435] <strong>and</strong> Schneider [909].<br />

11.1 John’s Ellipsoid Theorem <strong>and</strong> Ball’s Reverse Isoperimetric Inequality<br />

To a given convex body one may assign several ellipsoids in a canonical way.<br />

One example is the ellipsoid of inertia, for other examples, see the articles of<br />

Milman [725] <strong>and</strong> Lutwak, Yang <strong>and</strong> Zhang [670] <strong>and</strong> the book of Pisier [802].<br />

Among the ellipsoids which are inscribed, respectively, circumscribed to a proper<br />

convex body in E d , there is precisely one of maximum, respectively, minimum volume.<br />

Simple proofs for this result are due to Löwner (unpublished), Behrend [90]<br />

(d = 2) <strong>and</strong> Danzer, Laugwitz <strong>and</strong> Lenz [242] (general d). John [549] characterized<br />

inscribed ellipsoids of maximum volume, a complement being due to Ball [51]. Both<br />

results are of interest in convex geometry <strong>and</strong> in the geometry of normed spaces.<br />

John’s theorem implies, in particular, that for any origin symmetric convex body<br />

there is an ellipsoid which approximates it up to a factor √ d.<br />

Below these results are proved for convex bodies which are symmetric in o. The<br />

proof of the characterization result is taken from <strong>Gruber</strong> <strong>and</strong> Schuster [452]. It is<br />

based on the idea of Voronoĭ to identify symmetric, positive definite d × d matrices,<br />

respectively, positive definite quadratic forms on E d with points in E 1 2 d(d+1) ,<br />

compare Sect. 29.4. This idea was applied earlier in the same context by the author<br />

[421]. We give two classical applications, one to the group of affinities which map<br />

a convex body onto itself <strong>and</strong> one to the Banach–Mazur distance between norms on<br />

E d . A third application is the reverse isoperimetric inequality. For convex bodies C,<br />

the isoperimetric quotient<br />

S(C) d<br />

V (C) d−1<br />

is bounded below by the isoperimetric quotient of the unit ball B d , but it is clearly not<br />

bounded above. Behrend [89], for d = 2, <strong>and</strong> Ball [50], for general d, asked whether<br />

for any given convex body C there is an affine image, the isoperimetric quotient of<br />

which is bounded above in terms of d, <strong>and</strong> what is the worst case.

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

Saved successfully!

Ooh no, something went wrong!