Boundedness of generalized Riesz transforms on Orlicz-Hardy spaces associated to operators (Q2376114)

In recent years, the theory of Hardy spaces associated with operators has been extensively developed. The idea is to replace square functions (or maximal functions) defined in terms of the functional calculus of the Laplacian (or, more explicitly, by convolutions) by analogues defined in terms of the functional calculus of a given operator. One of the seminal papers on the subject [\textit{P. Auscher, A. McIntosh} and \textit{E. Russ}, J. Geom. Anal. 18, No. 1, 192--248 (2008; Zbl 1217.42043)], for instance, introduces Hardy spaces associated with the Hodge Dirac operator on a Riemannian manifold (where the measure is doubling). Spectral multipliers, such as Riesz transforms, are bounded on these spaces, while they can be unbounded on the classical Hardy space over a doubling metric measure space introduced by \textit{R. R. Coifman} and \textit{G. Weiss} [Bull. Am. Math. Soc. 83, 569--645 (1977; Zbl 0358.30023)]. After several interesting special cases had been investigated, the theory was developed in a natural general context in the memoir [\textit{S. Hofmann} et al., Mem. Am. Math. Soc. 1007 (2011; Zbl 1232.42018)]. That memoir considered a self-adjoint operator \(L\) satisfying Gaffney-Davies estimates on \(L^{2}(X)\), where \(X\) is a doubling metric measure space, and studied the associated \(H^{p}_{L}\) (\(1\leq p< \infty\)) and \(\text{BMO}_{L}\) spaces. While this context is natural, it is not the most general in which the methods can be applied. Gaffney-Davies estimates -- an integrated version of Gaussian kernel bounds valid for many second order differential operators -- can be generalised to handle higher order differential operators. Two of the authors have considered such a generalisation in [\textit{J. Cao} and \textit{D.-C. Yang}, Sci. China, Math. 55, No. 7, 1403--1440 (2012; Zbl 1266.42057)]. An alternative is to use Poisson bounds, as the third and fourth authors have done in [\textit{Y.-Y. Liang}, \textit{D.-C. Yang} and \textit{S.-B. Yang}, Sci. China, Math. 54, No. 11, 2395--2426 (2011; Zbl 1245.42019)]. Hardy spaces can also be replaced by Orlicz-Hardy spaces (taking an Orlicz norm of square functions rather than an \(L^p\) norm). This generalisation was considered in [\textit{R.-J. Jiang} and \textit{D.-C. Yang}, J. Funct. Anal. 258, No. 4, 1167--1224 (2010; Zbl 1205.46014)]. In the paper under review, the authors put all these generalisations together, and prove boundedness of the relevant Riesz transforms (including end point estimates). Conditions on the functions \(\phi\) used to define the Orlicz norms are given in terms of two critical parameters \(p_{-}(L),p_{+}(L)\) introduced in [\textit{P. Auscher}, Mem. Am. Math. Soc. 871 (2007; Zbl 1221.42022)]. These parameters are the optimal values for which \(-L\) generates a bounded semigroup on \(L^p\) for \(p \in (p_{-}(L),p_{+}(L))\). An important part of the paper under review consists of proving, using methods similar to those employed in [\textit{P. Auscher}, loc.\,cit.], that the non-tangential square function associated with \(L\) is \(L^p\) bounded in that interval.