Distributionally concave symmetric spaces and uniqueness of symmetric structure (Q1929204): Difference between revisions

The authors extend the well-known result that every \(L_p[0,1]\)-space has a unique symmetric structure. The results of the paper are based on the following concept introduced by \textit{S. Montgomery-Smith} and \textit{E. Semenov} [``Random rearangements and operators'', in: P. Kuchment (ed.) et al., Voronezh winter mathematical schools. Transl., Ser.~2, Am. Math. Soc. 184\,(37), 157--183 (1998; Zbl 0924.46025)]. Given functions \(x_1,x_2,\dots,x_n\) on \([0,1]\), their dilated disjoint sum is the function on \([0,1]\) given by \[ C(x_1,\dots,x_n)(t)=x_k(nt-k+1)\text{ for }(k-1)/n\leq t<k/n, \;k=1,\dots,n. \] A symmetric space \(E\) on the segment \([0,1]\) is called distributionally concave if there is a constant \(c_E>0\) such that \[ \|C(x_1,\dots,x_n)\|_E\geq c_E\min_{1\leq k\leq n}\|x_k\|_E. \] Let us present two typical results of the paper. Theorem 1.3. If \(E\) is a distributionally concave separable symmetric function space on \([0,1]\), such that \(E\) satisfies an upper \(p\)-estimate for some \(p>2\), then the following conditions are equivalent. (1) The spaces \(L_2[0,1]\) and \(E\) are the only symmetric spaces on \([0,1]\) isomorphic to a subspace of \(E\). (2) The tensor product operator \(x\otimes y\) is bounded from the Cartesian square \(E\times E\) into \(E\).\smallskip Corollary 1.5. Let \(M\) be a \(p\)-convex (for some \(p>2\)) Orlicz function satisfying the condition \(M(uv)\leq CM(u)M(v)\) for some \(C>0\) and any \(u,v\geq 1\), and let \(\psi\) be a concave increasing function on \([0,1]\) such that \(\psi(0)=0\) and \(\psi(st)\leq C'\psi(s)\psi(t)\) for some \(C'>0\) and any \(0<s,t\leq 1\). Then \(L_2[0,1]\) and the Orlicz-Lorentz space \(\Lambda(M,\psi)\) are the only symmetric spaces on \([0,1]\) isomorphic to a subspace of \(\Lambda(M,\psi)\). In particular, \(\Lambda(M,\psi)\) has a unique symmetric structure.