Concentration under scaling limits for weakly pinned Gaussian random walks (Q1017884)

The authors study scaling limits for inhomogeneous Markov chains \(\phi\) called the weakly pinned Gaussian random walks. The state space is \(\mathbb{R}^d_{(+)}\), i.e. either \(\mathbb{R}^d\) or \(\mathbb{R}^d_+\) = \(\mathbb{R}^{d-1} \times \mathbb{R}_+\), which corresponds to absence or presence of a wall at the boundary \(\partial \mathbb{R}^d_+\). The ``microscopic'' time values are \(\{0,1,\dots,N\}\), \(N\in \mathbb N\), the starting point being \(Na, a\in \mathbb{R}^d_{(+)}\). The two types of perturbed Gibbs kind distributions \(\mu_N\) on \(\mathbb{R}^{Nd}_{(+)}\) are considered, the perturbation (of magnitude \(\varepsilon\)) accounts for the force attracting to a subspace \(M\subset\partial \mathbb{R}^d_+\) with \(0\leq \mathrm {dim}M \leq d-1\). Namely, such \(\mu_N\) are attributed to the Dirichlet or free cases (\(\phi_N=Nb\) or \(\phi_N\) unrestricted respectively). The scaled ``macroscopic'' trajectories \(h^N(t), t\in [0,1],\) arise by polygonal interpolation of the vertices \(h^N(i/N)=N^{-1}\phi_i, 0\leq i\leq N\). The large deviation principle with certain rate functional holds (as \(N\to \infty\)) for \(h^N\), distributed under (each of) indicated \(\mu_N\), on the due space \(C([0,1],\mathbb{R}^d_{(+)})\). Among the other results established note the concentration properties, phase transitions, a version of CLT, and so called coexistence in the limit with specified probabilities found in the critical situation of two minimizers admitted by the appropriate rate functional. The motivation comes from random polymer and stochastic interface models.