Effective decay of multiple correlations in semidirect product actions (Q316983): Difference between revisions

Following [\textit{Z. J. Wang}, J. Funct. Anal. 267, No. 1, 15--79 (2014; Zbl 1330.22011)], the author first considers a connected semisimple almost algebraic \(\mathbb{R}\)-group \(G\), an irreducible faithful representation \(\varrho : G\to \mathrm{GL}(V)\) and a measure-preserving action of the semi-direct product group \(G\propto_\varrho V\) on a probability space \((X,\mu)\). He assumes that any almost simple factor of \(G\) fixes 0 only and that the representation of \(1\propto V\) on \(L_0^2(X)\) has no invariant vector. Restricting to \(G\propto_\varrho \{0\}\) yields an action of \(G\). Let \(G= KD^+FK\) be a Cartan decomposition, and \(a_0,\ldots, a_k\) be distinct elements in \(D^+\). Assuming \(\mathrm{rank}(G)=1\) allows to assume \(0< a_0<\cdots< a_k\in \mathbb{R}\). Let \[ R(a) := \min\left\{1, \sum_{n=0}^{k-1}{a_n\over a_k} \times \sum_{\ell=1}^{k}{a_0\over a_\ell}\right\}. \] Then the main result corresponding to this rank-one setting ensures the existence of an \(L_0^2(X)\)-dense subspace \(\mathcal{D}\) of bounded, zero-mean functions, such that for any \(f_0,\ldots, f_k \in\mathcal{D}\), the following exponential control holds: \[ \left|\int_X (a_0^{-1}\cdot x)\, f_0(x)\, f_1(a_1^{-1}\cdot x)\cdots f_k(a_k^{-1}\cdot x)\,\mu(dx)\right| \leq C(f_0,\ldots, f_k)\times R(a)^q, \] with an exponent \(q\) depending only on the action. The above hypotheses can be weakened somewhat. Then the author turns to the higher-rank case, first with the basic example \(G_0:= \mathrm{SL}(3,\mathbb{R})\propto \mathbb{R}^3\). He establishes a result of the same type as above, with a modified rate \(R(a)\), which can be explicited in terms of the highest and lowest weights of \(\varrho\) (and must be multiplied by \(k^2\)). He exhibits examples showing that often \(R(a)=1\). In this case his method is useless. Finally the author extends the results to some class of split semi-simple algebraic groups of higher rank. The method followed by the author is spectral and based on approximations by nicely behaved functions.