Ultrapowers of \(L_1(\mu)\) and the subsequence splitting principle (Q5935808): Difference between revisions

Let \({\mathcal U}\) be any ultrafilter on a set \(I\) and let \(I_1(\mu)= L_1(\Omega,\Sigma,\mu)\) be any \(L_1\)-space. It is well known that the ultrapower of \(L_1(\mu)\) admits a repreentation \[ L_1(\mu)_{\mathcal U}= L_1(\mu_{\mathcal U})\oplus L_1(\nu_{\mathcal U}), \] where \(\mu_{\mathcal U}\) is the ultrapower of \(\mu\) on \(\Omega^{\mathcal U}\) and \(\nu_{\mathcal U}\) is some other measure. In the present paper a very useful characterization of the both components is given in the following way: A function \(f\in L_1(\mu)_{\mathcal U}\) belongs to \(L_1(\mu_{\mathcal U})\) iff \(f\) admits a representation \(f= [f_i]\) with a relative \(w\)-compact family \(\{f_i: i\in I\}\subseteq L_1(\mu)\), and \(f\) belongs to \(L_1(\nu_{\mathcal U})\) if there is a representation \(f= [f_i]\) with \(\mu_{\mathcal U}(\{f_i\neq 0\}^{\mathcal U}= \lim_{\mathcal U}\mu(\{f_i\neq 0\})= 0\). If \({\mathcal U}\) is a selective ultrafilter (= rare p-point) then even more is true: \(f\in L_1(\mu_{\mathcal U})\) iff there is a representation \(f= [f_i]\) with \((f_i)\) weakly convergent following \({\mathcal U}\), and \(f\in L_1(\nu_{\mathcal U})\) iff there is a representation \(f= [f_i]\) with pairwise disjoint functions. In particularly, for each bounded sequence \()f_i)\) in \(L_1(\mu)\) there is a splitting \([f_i]= [g_i]+ [h_i]\) in the ultrapower with \((h_i)\) pairwise disjoint and \(\omega\)-\(\lim_{\mathcal U}g_i\) existing. This last statement could be called the ultrapower version of Rosenthal's subsequence splitting principle for bounded sequences in \(L_1(\mu)\). The methods are also used to give a new proof of the fact that reflexive subspaces of \(L_1\)-spaces are superreflexive. Counterexamples for the case of non-selective ultrafilters are given.