Uniform estimates for spherical functions on complex semisimple Lie groups (Q5956634)

Let \(G\) be a connected complex semisimple Lie group, \({\mathfrak g}\) its Lie algebra (considered over \(\mathbb{R}\)), \(K\) a maximal compact subgroup. Let \({\mathfrak a}\) be a maximal Abelian subalgebra orthogonal to the Lie algebra of \(K\). Let \(\varphi_\lambda\), \(\lambda\in{\mathfrak a}^*_{\mathbb{C}}\), be spherical functions on \(G\) with respect to \(K\). Here, the authors' notation differs from Harish-Chandra's: they write \(\lambda\) where he writes \(i\lambda\). In the paper under review the authors give some new estimates for \(\varphi_\lambda\) and their derivatives which are uniform in \(\lambda\) and apply these estimates to the study of singular sphere averages. Let us show some estimates. Take a regular element \(H\in{\mathfrak a}\) with norm 1, then for \(\lambda\in{\mathfrak a}^*_{\mathbb{C}}\), \(t\in\mathbb{R}^+\), and \(k\in\mathbb{N}\), we have \[ \Biggl|{d^k\over dt^k} \varphi_\lambda(\exp(tH))\Biggr|\leq C\cdot B(H)\cdot {(1+\|\lambda\|)^k\over (1+\|\lambda\|)^\gamma}\cdot \Omega(tH)\cdot {S_{\text{Re }\lambda}(tH)\over S_\rho(tH)}.\tag{1} \] Here \(C\) and \(\gamma\) depend only on \({\mathfrak g}\), \(B(H)= \prod |\alpha(H)|^{- 1}\), \(\Omega(H)= \prod(1+ |\alpha(H)|)\), \(S_\lambda(H)= \sum\exp\sigma\lambda(H)\), \(\rho= (1/2) \sum m^\alpha \alpha\), the products and the sums are taken over positive roots \(\alpha\) of the pair \(({\mathfrak g},{\mathfrak a})\) and/or elements \(\sigma\) of the Weyl group \(W= W({\mathfrak g},{\mathfrak a})\). For a complex group, the multiplicities \(m^\alpha\) are equal to 2. Estimate (1) improves the Harish-Chandra estimate, it replaces polynomial growth in \(\lambda\) by polynomial decay (at least when \(k< \gamma\)). In particular, for positive definite spherical functions \(\varphi_\lambda\) we have the following: let \(G\) be of rank \(>1\), then there exists \(\kappa\in (0,1)\) such that \(\text{Re }\lambda\) lies in the closed convex hull of \(W(\kappa\rho)\), \(\varphi_\lambda\) belongs to \(L^p(G)\) for \(p> 2/(1-\kappa)\) and one can replace \(\text{Re }\lambda\) in (1) by \(\kappa\rho\). Now let \(Y\) be a \(G\)-space, i.e. a Borel space with a Borel measurable \(G\)-action preserving a Borel measure, for example, \(Y= K\setminus G/\Gamma\) with \(\Gamma\) discrete. For \(H\in {\mathfrak a}\), \(t\geq 0\), define the measure \(\sigma^H_t\) on \(G\) with support \(K\exp(tH)K\) (a \(K\)-sphere) by \(\sigma^H_t= m_K* \delta_{\exp(tH)}* m_K\) where \(m_K\) is the normalized Haar measure on \(K\). For a function \(f\) on \(Y\), the convolution \((\sigma^H_t* f)(x)\) is an average of \(f\) over a \(K\)-sphere with center \(x\) and radius \(t\). Define the maximal operator \({\mathcal M}_{\sigma^H}\) on \(L^p(Y)\) by \({\mathcal M}_{\sigma^H}f= \sup|(\sigma^H_t* f)|\), \(t\geq 0\). Let \(H\) be as before (1) and \(p> 1+ (2\gamma)^{-1}\), then for all \(f\in L^p(Y)\) we have the following statements (maximal inequalities, singular differentiation, and ergodicity): \(\|{\mathcal M}_{\sigma^H} f\|_p\leq C_p\cdot B(H)\cdot\|f\|_p\), limits of \(\sigma^H_t* f\) as \(t\to 0\) and \(t\to\infty\) are equal almost everywhere to \(\sigma^H_0* f\) and \(Ef\), respectively. Similar results are true for Riemannian spheres (with some conditions on \(Y\)).