Annales Academiæ Scientiarum Fennicæ
Mathematica
Volumen 44, 2019, 141-165
University of St. Andrews,
Department of Mathematics
St. Andrews, Fife KY16 9SS, Scotland;
lo 'at' st-and.ac.uk
Abstract. Let X be a metric space and write K(X) for the family of non-empty compact subsets of X equipped with the Hausdorff metric. The lower and upper box dimensions, denoted by \underline dimB(E) and \overline dimB(E), of a subset E of X are defined by
\underline dimB(E) = liminfr&rarr0 log Nr(E)/–log r, \overline dimB(E) = limsupr→0 log Nr(E)/–log r,
where Nr(E) is the smallest number of closed balls with centres in E and radii equal to r that are needed to cover E. In the 1980's, Gruber proved that the box counting function
(*) log Nr(C)/–log r
of a typical compact set C ∈ K(X) diverges in the worst possible way as r → 0. For example, Gruber proved that \underline dimB(C) = 0 and \overline dimB(E) = N for a typical C ∈ K(RN).
In this paper we prove that the box counting function (*) of a typical compact set C ∈ K(X) is spectacularly more irregular than suggested by Gruber's result. In particular, we show the following surprising result: not only is the box counting function (*) of a typical compact set C ∈ K(X) divergent as r → 0, but it is so irregular that it remains spectacularly divergent as r → 0 even after being "averaged" or "smoothened out" using powerful averaging methods including, for example, all higher order Hölder and Cesaro averages. As an application of our results we obtain strengthened versions of Gruber's result.
2010 Mathematics Subject Classification: Primary 28A78, 28A80.
Key words: Box dimension, compact set, Hölder mean, Cesaro mean, Baire category.
Reference to this article: L. Olsen: Average box dimensions of typical compact sets. Ann. Acad. Sci. Fenn. Math. 44 (2019), 141-165.
https://doi.org/10.5186/aasfm.2019.4406
Copyright © 2019 by Academia Scientiarum Fennica