A note on the Gaussian correlation conjecture (Q4524741)

The probably most interesting and most challenging open problem for Gaussian measures is the so-called correlation conjecture: Given two symmetric convex sets \(A\) and \(B\) in \(\mathbb R^n\) and a centered Gaussian measure \(\mu\), does it follow that \(\mu(A\cap B)\geq \mu(A) \mu(B)\)? A weaker (also open) form of this question is whether or not \(\mu(A)=\mu(B)=\rho\) always implies \(\mu(A\cap B)\geq\rho^2\). It is well-known that the correlation conjecture is of great importance for the investigation of small ball problems for Gaussian random processes. In particular, in several cases the validity of this conjecture implies the existence of so-called small ball constants. The aim of this paper is to show that already the above mentioned weaker form of the conjecture leads to similar results. As an example, the fractional Brownian motion on \([0,1]\) is treated. From our point of view, the basic result of this paper is Lemma 2 asserting \[ \mu(A\cap B)\geq \mu(\lambda A)\cdot\mu\big((1-\lambda^2)^{1/2} B\big) \] for any \(0<\lambda<1\), which already has turned out to be very useful in many circumstances. For the entire collection see [Zbl 0948.00040].