The correlation length for the high-density phase of Bernoulli perlocation (Q910839)

Two standard types of connectivity functions \(\tau^ f(x,y)\), \({\tilde \tau}\)(x,y) in the high-density phase of nearest-neighbor Bernoulli (bond) percolation are examined. Let p be the bond density of the percolation model on \(Z^ d\). \(\tau^ f(x,y)\) for a given p and for the given \(x,y\in Z^ d\) is the probability that the points x and y belong to the same finite occupied cluster. \(\tau\) (x,y) is the probability that the points x and y are connected by an occupied path on the lattice \(Z^ d\) and \({\tilde \tau}\)(x,y)\(=\tau (x,y)-P^ 2_{\infty}(p)\). Here \(P_{\infty}(p)\) is the percolation probability for \(p>p_ c\) \((p_ c\) is the percolation threshold). The authors proved that the limits \((1)\quad \xi^ f(p)=-\log \tau^ f(0,x)/| x| \quad and(2)\quad {\tilde \xi}(p)=-\log {\tilde \tau}(0,x)/| x|\)exist, when x moves out to \(\infty\) in a fixed direction. Thus two connectivity functions decay exponentially. Besides that the authors proved that the limits (1),(2) are equal and depend on the direction vector. The quantity \(\xi^ f(p)\) is the proper choice for the correlation length in the percolation model. In addition the exponential upper bounds for the connectivity functions for \(p>p_ c\) are obtained and it is established that the inverse of the correlation length is convex, as a function of the direction vector, and upper semicontinuous as a function of the bond density. It is explicitly demonstrated that [\({\tilde \tau}\)(0,x)- \(\tau^ f(0,x)]\geq 0.\) For the bond percolation in two dimensions it is shown that the correlation length at bond density \(p_ 1>p_ c=1/2\) is exactly half the correlation length at the subcritical bond density \(p_ 2=1-p_ 1<p_ c\).