Vectorial Hankel operators, Carleson embeddings, and notions of BMOA

Let $\mathrm{BMOA}_{\mathcal{NP}}$ denote the space of operator-valued analytic functions $\phi$ for which the Hankel operator $\Gamma_\phi$ is $H^2(\mathcal{H})$-bounded. Obtaining concrete characterizations of $\mathrm{BMOA}_{\mathcal{NP}}$ has proven to be notoriously hard. Let $D^\alpha$ denote differentiation of fractional order $\alpha$. Motivated originally by control theory, we characterize $H^2(\mathcal{H})$-boundedness of $D^\alpha\Gamma_\phi$, where $\alpha>0$, in terms of a natural anti-analytic Carleson embedding condition. We obtain three notable corollaries: The first is that  $\mathrm{BMOA}_{\mathcal{NP}}$ is not characterized by said embedding condition. The second is that when we add an adjoint embedding condition, we obtain a sufficient but not necessary condition for boundedness of $\Gamma_\phi$ . The third is that there exists a bounded analytic function for which the associated anti-analytic Carleson embedding is unbounded. As a consequence, boundedness of an analytic Carleson embedding does not imply that the anti-analytic ditto is bounded. This answers a question by Nazarov, Pisier, Treil, and Volberg.