The natural rearrangement invariant structure on tensor products (Q2427243)

Let \((\Omega_i,\Sigma_i,\lambda_i)\), \(i=1,2\), be measure spaces, and \(X,Y\) and \(Z\) be rearrangement invariant (r.i.) spaces on \(\Omega_1,\Omega_2\) and \(\Omega_1 \times \Omega_2\), respectively. A classical question is the following: In which cases is the bilinear operator \(B: X(\Omega_1) \times Y(\Omega_2) \rightarrow Z(\Omega_1\times\Omega_2)\) defined by \(B(x,y)(s,t):=x(s)y(t)\) bounded? Or equivalently, when do we have a continuous embedding of the projective tensor product \(X(\Omega_1) \widehat{\otimes}_\pi Y(\Omega_2)\) into \(Z(\Omega_1 \times \Omega_2)\)? In the article under review, the authors state the related question of which r.i. spaces \(X(\Omega_1)\) and \(Y(\Omega_2)\) and which crossnorms \(\alpha\) satisfy that the tensor product \(X(\Omega_1) \widehat{\otimes}_\alpha Y(\Omega_2)\) is again an r.i. space, i.e., there exists an r.i. space \(Z(\Omega_1 \times \Omega_2)\) such that the extension operator \(\widehat{B}: X(\Omega_1) \widehat{\otimes}_\alpha Y(\Omega_2) \rightarrow Z(\Omega_1 \times \Omega_2)\) is defined and a topological isomorphism. They prove that for a separable measure space \(\Omega=\Omega_1=\Omega_2\) this only happens when \(X=Y=Z=L_p\) and \(\alpha=\Delta_p\) for some \(p \in [1,\infty)\). The proof is divided into three parts: The cases \(\Omega=[0,1]\) and \(\Omega=[0,\infty)\) (each with the Lebesgue measure) and the case where \(\Omega\) is the set of natural numbers with the counting measure. In the latter case, the problem on separable spaces can be translated into bases. The structure of the proof in each of the cases is always as follows: Using classical characterizations of \(L_p\)-spaces, it is shown in a first step that necessarily \(X=Y=L_p\) for some \(p \in [1,\infty]\). In a second step, the case \(p=\infty\) is ruled out, and finally, by a density argument, the crossnorm \(\alpha\) has to be equivalent to the norm \(\Delta_p\).