CLT for quadratic additive functionals of three-dimensional and four-dimensional voter models

In this paper, we prove central limit theorems for the quadratic additive functionals of the three-dimensional and four-dimensional voter models. Our main results show that, under proper time-space scalings, the centered quadratic additive functional of the four-dimensional voter model converges weakly to a Brownian motion and that in the three-dimensional case converges weakly to a Gaussian process without independent increment. The proof of the main result in the four-dimensional case extends the martingale decomposition method. In the three-dimensional case, we introduce a quadratic fluctuation limit theorem which shows that the quadratic fluctuation density field of the voter model converges weakly to a generalized Ornstein-Uhlenbeck process. Then our main result in the three-dimensional case follows from an argument where the quadratic additive functional is approximated by an integral of a test function with respect to the quadratic fluctuation density field. The duality relationship between the voter model and the coalescing random walk is repeatedly utilized in proofs of the main results. The proof strategies introduced in this paper are applicable in deducing CLTs for general additive functionals of three-dimensional and four dimensional voter models. We only deal with those with the quadratic forms to make our exposition easy to catch.