Percolation of strongly correlated Gaussian fields I. Decay of subcritical connection probabilities

We study the decay of connectivity of the subcritical excursion sets of a class of strongly correlated Gaussian fields. Our main result shows that, for smooth isotropic Gaussian fields whose covariance kernel $K(x)$ is regularly varying at infinity with index $\alpha \in [0, 1)$, the probability that $\{f \le \ell\}$, $\ell < \ell_c$, connects the origin to distance $R$ decays sub-exponentially in $R$ at log-asymptotic rate $c_\alpha (\ell_c-\ell)^2 / K(R)$ for an explicit $c_\alpha > 0$. If $\alpha = 1$ and $\int_0^\infty K(x) dx = \infty$ then the log-asymptotic rate is $c_1 (\ell_c-\ell)^2 R (\int_0^R K(x) dx)^{-1}$, and if $\alpha > 1$ the decay is exponential. Our findings extend recent results on the Gaussian free field (GFF) on $\mathbb{Z}^d$, $d \ge 3$, and can be interpreted as showing that the subcritical behaviour of the GFF is universal among fields with covariance $K(x) \sim c|x|^{d-2}$. Our result is also evidence in support of physicists' predictions that the correlation length exponent is $\nu = 2/\alpha$ if $\alpha \le 1$, and in $d=2$ we establish rigorously that $\nu \ge 2/\alpha$. More generally, our approach opens the door to the large deviation analysis of a wide variety of percolation events for smooth Gaussian fields. This is the first in a series of two papers studying subcritical level-set percolation of strongly correlated Gaussian fields, which can be read independently.