Linear response for intermittent maps

Abstract

We consider the one parameter family $\alpha \mapsto T_\alpha$ ($\alpha \in [0,1)$) of Pomeau-Manneville type interval maps $T_\alpha(x)=x(1+2^\alpha x^\alpha)$ for $x \in [0,1/2)$ and $T_\alpha(x)=2x-1$ for $x \in [1/2, 1]$, with the associated absolutely continuous invariant probability measure $\mu_\alpha$. For $\alpha \in (0,1)$, Sarig and Gou\"ezel proved that the system mixes only polynomially with rate $n^{1-1/\alpha}$ (in particular, there is no spectral gap). We show that for any $\psi\in L^q$, the map $\alpha \to \int_0^1 \psi\, d\mu_\alpha$ is differentiable on $[0,1-1/q)$, and we give a (linear response) formula for the value of the derivative. This is the first time that a linear response formula for the SRB measure is obtained in the setting of slowly mixing dynamics. Our argument shows how cone techniques can be used in this context. For $\alpha \ge 1/2$ we need the $n^{-1/\alpha}$ decorrelation obtained by Gou\"ezel under additional conditions.