The Banach space T and the fast growing hierarchy from logic (Q1056866)

Tsirelson's Banach space T is by now a famous counterexample. The space T has an unconditional basis \(\{t_ n\}\), but T contains no subsymmetric basic sequence and hence no \(c_ 0\) or \(\ell_ p\) subspaces. In particular, there is a subsequence \(\{\) i(n)\(\}\) so that \(\{t_ n\}\) and \(\{t_{i(n)}\}\) are not equivalent. However, until now, there was only an existence proof of such a subsequence \(\{\) i(n)\(\}\). A construction of such an \(\{\) i(n)\(\}\) is given. The growth rate of \(\{\) i(n)\(\}\) turns out to be rather larger than expected from an analysist point of view. We use ''the fast growing hierarchy'' from logic which are inductively defined by \[ g_ 0(n)=n+1\quad and\quad g_{i+1}(n)=g^ n_ i(n) \] where \(f^ n(k)\) is the n-fold composite of f, that is \(f^{n+1}(k)=f(f^ n(k))\) and \(f^ 1(k)=k.\) The main theorem is that \(t_ n\sim t_{i(n)}\) if and only if \(i(n)\leq g_ m(n)\) for some m and large n. In particular, if \(\{\) i(n)\(\}\) is a primitive recursive subsequence, then \(\{t_{i(n)}\}\) is equivalent to \(\{t_ n\}\). And if \(i(n)=g_{\omega}(n)=g_ n(n),\) then \(\{t_{i(n)}\}\) is not equivalent to \(\{t_ n\}\).