Journal of Statistical Physics, Vol. W, Nos. 3/4, IWX
A Short Existence Proof for Correlation Dimension
Anthony Manning1 and Karoly Simon2
Received October 2K, 7997
The Grassberger Hentschel Procaccia correlation dimension has been put on a
rigorous basis by Pesin and Tempelman. We simplify their proof that this
dimension is given in terms of the measure of neighborhoods of the diagonal.
KEY WORDS: Correlation dimension; ergodic measure; neighborhood of
diagonal.
Let (X, p) be a separable metric space. Suppose that n is an ergodic prob-
ability measure for the continuous map/: X-> X. The r neighbourhood of
the diagonal in XxX is denoted by Sr. That is Sr := {(.v, v)e Xx X:
p(x,y)^r}. The function q>(r) = \'(Sr) is monotone increasing where v is
the product measure /*x/;. For xeX and neN we let C(.v, n, r) denote
l/n2${(i, j): (/''(x), fJ(x))eSr,0^i, j<n}, the proportion of pairs of
points in part of the orbit that are closer than r. Roughly speaking, if, for
/i almost every x, for large n and small r, we have C(,v, n, r) ~ ra then a is
called the correlation dimension' 3' of the measure //. To give a precise
definition of the correlation dimension it is fundamental to prove the
following theorem, as was done for invertible / by Pesin' u and, in a more
general context, by Pesin and Tempelman. <2>
Theorem 1. There is a set YcX of full ^-measure such that for
each xe Y
provided (p is continuous at r.
1 Mathematics Institute, University of Warwick, Coventry, CV4 7AL, U.K.
2 Institute of Mathematics, University of Miskolc, Miskolc-Egyetemvaros H-3515, Hungary.
1047
0022-4715/98/0200-1047115.00/0 © 1998 Plenum Publishing Corporation