The paper studies a class of equations from mathematical physics called Hamilton-Jacobi equations, which describe how certain quantities evolve over space and time. The specific setting involves a situation where the underlying environment is periodic, like a crystal lattice, but with a localized disturbance added in one region. The question being asked is: as a particular regularization parameter called the "discount factor" shrinks to zero, do the solutions to these modified equations converge to a unique, well-defined limiting solution? This is called the vanishing discount selection problem, and identifying which solution gets picked out is the central challenge.
The authors consider two cases depending on whether the localized disturbance acts as an attractive well or a repulsive bump. When the disturbance is attractive (negative potential), they prove that solutions do converge locally to a specific limiting solution, and they give a precise description of what that solution looks like. The key technical insight involves decomposing a certain probability measure that captures long-run behavior into two parts: one part that stays near the disturbance and one part that drifts off toward the periodic background far away. This decomposition lets them carefully track how mass is distributed between the local perturbation and the periodic environment at infinity.
When the disturbance is repulsive (positive potential), convergence is harder to guarantee, and the authors need an extra condition on the unperturbed periodic problem to make their arguments work. They also highlight a subtle and somewhat surprising finding: even when convergence does occur, the most natural formulas one might write down to guess the limiting solution can give the wrong answer. This serves as a warning that the selection mechanism in these problems is genuinely delicate and cannot always be captured by simple heuristics. Overall, the paper advances the mathematical understanding of how local features of an environment influence global long-run behavior in Hamiltonian systems.