The study examines a mathematical object called a Gaussian free field, which you can think of as a randomly fluctuating surface defined on a large grid. Here, instead of a single such surface, the researchers work with several independent copies stacked together as a vector. The key twist is a constraint: the vector formed by the field values at every interior point must stay outside a fixed ball, meaning the combined magnitude of all the field components cannot get too small in a certain central region. This is a rare event that forces the fields to collectively avoid being near zero.
The main finding is a striking "freezing" phenomenon. When the grid size grows large and the field values are rescaled by dividing by the logarithm of the grid size, the random fluctuations essentially disappear at the macroscopic level. What remains is not random in shape, only in direction. The rescaled field converges to a fixed spatial profile, called the equilibrium potential of the forbidden region, multiplied by a random unit vector that points uniformly in any direction. In other words, the conditioning event forces the field into an almost deterministic shape, with randomness surviving only as a single random orientation. The paper also precisely quantifies how unlikely the conditioning event is, in terms of a quantity called capacity.
The proof brings together several sophisticated tools. One is a recent result about how the norm of a vector field behaves under repulsion constraints. Another is the theory of large deviations for Gaussian processes, which describes the probabilities of rare behaviors. A third is a classical variational problem from potential theory, where the equilibrium potential naturally appears as the minimizer of an energy functional subject to the avoidance constraint. The combination of these tools allows the authors to pin down not just the limiting shape but also the precise probability cost of the constraint, and to show that the field components align with each other on essentially all grid points.