Estimates on path delocalization for copolymers at selective interfaces (Q2575675)

The authors consider a model for a random heterogeneous polymer in the proximity of an interface separating two selective solvents. More precisely, they look at the transformed path measure that weights the probability of a simple random walk \((S_ n)_{n\in\mathbb N_0}\) on \(\mathbb Z\) with a factor \(\exp\{\lambda\sum_{n=1}^ N(\omega_n+h)\text{sign}(S_n)\}\), where \(\lambda,h\in[0,\infty)\) are coupling resp.~drift parameters, and \((\omega_ n)_{n\in\mathbb N_0}\) is an i.i.d.~random sequence, independent of the walk. In earlier work by \textit{E. Bolthausen} and \textit{F. den Hollander} [Ann. Probab. 25, No.~3, 1334--1366 (1997; Zbl 0885.60022)] it was shown that this model undergoes a (de)localization transition: there exists a continuous increasing function \(\lambda\mapsto h_{\text{c}}(\lambda)\) such that the model is localized for \(h<h_{\text{c}}(\lambda)\), while it is delocalized for \(h\geq h_{\text{c}}(\lambda)\). This distinction was made only in terms of properties of the free energy. Later \textit{M. Biskup} and \textit{F. den Hollander} [Ann. Appl. Probab. 9, No.~3, 668--687 (1999; Zbl 0971.60098)] sharpened the picture by showing that the (de)localization also occurs in terms of path properties. The paper presents a new method, based on concentration bounds for suitably restricted partition functions, that yields much more information about the delocalized region. In particular, the upper bound for the number of visits to the lower half plane is strengthened from \(o(N)\) to \(O(\log N)\). Furthermore, the limit of \(h_{\text{c}}(\lambda)/\lambda\) as \(\lambda\downarrow 0\) is shown to exist and to be independent on the law of \(\omega_ 1\), subject to some mild technical restrictions.