Hardy spaces on Lie groups of polynomial growth (Q2379243): Difference between revisions

Let \(G\) be a connected Lie group of polynomial growth with Lie algebra \(\mathfrak{g}\) and \(a_1,\dots,a_{d'}\) an algebraic basis of \(\mathfrak{g}\). For all \(i\in\{1,2,\dots,d'\}\), let \(A_i=dL(a_i)\) denote the generator of left translations in the direction \(a_i\). Consider the second order subelliptic operator \[ H=-\sum^{d'}_{i,j=1}c_{i,j}A_iA_j, \] where \(C=(c_{i,j})\) is a \(d'\times d'\) matrix of complex coefficients satisfying \(2^{-1}(C+C^*)\geq \mu I\) for some \(\mu>0\). It is known that the closure of the subelliptic operator \(H\) generates a holomorphic contraction semigroup \(S_t\) which has a smooth kernel \(K_t\) satisfying the Gaussian bounds. In terms of the semigroup \(S_t\), the maximal Hardy space \(H^1_{\max,H}(G)\) can be defined as follows: \[ H^1_{\max,H}(G)=\{\varphi: \varphi^*_{H}\in L_1(G)\} \] with norm \(\|\varphi\|_{H^1_{\max,H}}= \|\varphi^*_H\|_{L_1}\), where the maximal function \(\varphi^*_H\) is given as \[ \varphi^*_H(g)= \sup_{\{(h,t)\in G\times (0,\infty): |gh^{-1}|<\sqrt{t}\}}|(S_t\varphi)(h)|. \] Similarly, the Hardy space \(H^1_{\max,H^\gamma}(G)\) is defined via the semigroup \(S^\gamma_t\) generated by \(H^\gamma\) which is the fractional power of \(H\). The result can be stated as follows: Let \(\varphi\in L_1(G)\). The following statements are equivalent. (1) There exists \(\gamma\in [2^{-1},1]\) such that \(\varphi\in H^1_{\max,H^\gamma}(G)\). (2) \(\varphi\in H^1_{\max,H^\gamma}(G)\) for all \(\gamma\in [2^{-1},1]\). (3) \(\sup_{t>0}|S_t\varphi|\in L_1(G)\). (4) \(\sup_{t>0}|S^\gamma_t\varphi|\in L_1(G)\) for all \(\gamma\in [2^{-1},1]\). (5) \(\varphi\in H^1_{\text{atom}}(G)\). Moreover, if the coefficients \(c_{i,j}\) of \(H\) are real, then the conditions are equivalent to the following. (3') There exists a \(\gamma\in [2^{-1},1]\) such that \(\sup_{t>0}|S^{\gamma}_t\varphi|\in L_1(G)\). The main result of this paper is a nontrivial generalization of standard results for the Laplacian on \(\mathbb{R}^d\).