Transference for Banach space representations of nilpotent Lie groups. II. Pedersen multipliers (Q2665615)

Let \(G\) be a connected simply connected nilpotent Lie group and \(\pi:G\to\mathcal{B}(\mathcal{H})\) be an irreducible unitary representation of \(G\) associated with a coadjoint orbit \(\mathcal{O}=\mathrm{Ad}_G^*\, \xi_0 \subset\mathfrak{g}^*\). Let \(\mathfrak{g}_e\subset\mathfrak{g}\) denote the predual of \(\mathcal{O}\). The Pedersen transform \(T^\pi:\mathcal{S}(\mathfrak{g}_e)\to\mathcal{B}(\mathcal{H})_\infty\), on the set of smooth operators, is defined for \(b\in\mathcal{S}(\mathfrak{g}_e)\) by \[ T^\pi(b)=\int_{\mathfrak{g}_e}b(X)\pi(\exp_GX)\,dX. \] We can define uniquely the twisted convolution product \(a*_eb\) in \(\mathcal{S}(\mathfrak{g}_e)\) such that \(T^\pi(a*_eb)=T^\pi(a)T^\pi(b)\). Then \((\mathcal{S}(\mathfrak{g}_e),*_e)\) is an involutive associated algebra such that \(T^\pi\) is an involutive algebra isomorphism. In this paper the authors introduce a Pedersen \(L^p\)-multiplier \(M\in\mathcal{B}(\mathcal{H})\), \(1\leq p<\infty\), meaning there exists a bounded linear operator \(C_M:L^p(\mathfrak{g}_e)\to L^p(\mathfrak{g}_e)\) such that \(T^\pi(C_M\phi)=MT^\pi(\phi)\) for all \(\phi\in\mathcal{S}(\mathfrak{g}_e)\). Since \(T^\pi\) is a \(*\)-isomorphism, \(C_M\) satisfies \(C_M(\phi*_e\psi)=(C_M\phi)*_e\psi\) for all \(\phi,\psi\in L^1(\mathfrak{g}_e)\cap L^p(\mathfrak{g}_e)\). Characterizing the twisted convolution, they show that, if a bounded linear operator \(C:L^p(\mathfrak{g}_e)\to L^p(\mathfrak{g}_e)\) satisfies \(C(\phi*_e\psi)=(C\phi)*_e\psi\) for all \(\phi,\psi\in\mathcal{S}(\mathfrak{g}_e)\), then \(C=C_M\) with \(M=T^\pi(u)\), where \(u\in\mathcal{S}'(\mathfrak{g}_e)\). We assume additionally \((\pi, L^2(\mathcal{V}))\) is square-integrable modulo \(Z(G)\). Since the twisted convolution has a more explicit form, the above result can be strengthened. Let \(1\leq p,q<\infty\) and a bounded linear operator \(C:L^p(\mathfrak{g}_e)\to L^q(\mathfrak{g}_e)\) satisfy \(C(\phi*_e\psi)=(C\phi)*_e\psi\) for all \(\phi,\psi\in L^1(\mathfrak{g}_e)\cap L^p(\mathfrak{g}_e)\). If \(p>q\), then \(C\equiv 0\) and if \(p\leq q\), then \(C=C_M\) with \(M=T^\pi(u)\), where \(u\in \mathcal{S}'(\mathfrak{g}_e)\). Moreover, the corresponding multiplier \(M\in \mathcal{B}(L^2(\mathcal{V}))\) has an \(L^p\) extension, that is, the restriction \(M|_{L^2(\mathcal{V})\cap L^p(\mathcal{ V})}\) extends to an operator in \(\mathcal{B}(L^p(\mathcal{V}))\). For part~I see [\textit{I.~Beltiţă} et al., Proc. Am. Math. Soc. 146, No.~12, 5065--5075 (2018; Zbl 1439.22017)].