Annealed upper tails for the energy of a charged polymer (Q629800)

Consider the path \((S(0),\ldots,S(n-1))\) of a simple random walk on the \(d\)-dimensional (\(d\geq 3\)) integer lattice as the model of a random polymer. Denote by \[ \ell_n(z)=\sum_{i=0}^{n-1}\mathbf{1}_{\{S(i)=z\}} \] its local time at \(z\). Each monomer \(i=0,\dots,n-1\) is assigned a random charge \(\eta(i)\), where \(\eta(0),\eta(1),\dots\) is an i.i.d.~family of centered random variables with variance 1. The Hamilton function is defined by \[ H_n=\sum_{0\leq i<j\leq n-1}\eta(i)\eta(j)\mathbf{1}_{\{S(i)=S(j)\}}. \] In the paper under review, the asymptotics as \(n\to\infty\) of the upper tails of the annealed distribution of \(H_n\) is studied. More precisely, the energy is represented as \(H_n=\check X_n+Y_n\), where \[ \check X_n=\sum_{x\in\mathbb{Z}^d}\left[\left(\sum_{k=0}^{n-1}\eta(k)\mathbf{1}_{\{S(k)=x\}}\right)^2-\ell_n(x)\right] \] and \[ Y_n=\sum_{i=0}^{n-1}\big(1-\eta^2(i)\big). \] Since the large deviations for the i.i.d.~sum \(Y_n\) is well understood, only the upper tails of \(\check X_n\) are studied. Assume that \(n^{2/3}\ll\xi_n\ll n^2\) and that, for some \(\alpha>1\), the power \(|\eta|^\alpha\) has an exponential moment. The main result states that there exists a constant \(\mathcal{Q}_2\) such that \[ \lim_{n\to\infty}\xi_n^{-1/2}\log\operatorname{P}[\check X_n\geq \xi_n]=-\mathcal{Q}_2. \] Furthermore, the main contributions to the probability comes from configurations where \(\ell_n(z)\) is of order \(\sqrt{n}\) for a finite set of \(z\). If the distribution of \(\eta_i\) fulfils additional conditions, then the constant \(\mathcal{Q}_2\) can be computed explicitly. In fact, assume the log-Laplace transform \(\Gamma(x)=\log \operatorname{E}[\exp(x\eta_0)]\) is twice differentiable and that \(x\mapsto\Gamma(\sqrt{x})\) is convex. Denote by \(\Gamma^{-1}\) the inverse function. Let \(r_d\) be the return probability of a \(d\)-dimensional symmetric simple random walk and define \(\chi_d=-\log(r_d)\). Then the author shows that \(\mathcal{Q}_2=\Gamma^{-1}(\chi_d)\).