On the Banach-Stone theorem (Q5916313): Difference between revisions

\(C(T)\) denotes the complex Banach space of supremum-normed continuous complex-valued functions on the compact Hausdorff space \(T\), \(B(T)\) is its open unit ball, and \(\Gamma (T)\) is the set of unimodular functions in \(C(T)\) (or, as Lemma 1 says, the set of all (complex = real) extreme points of \(\overline {B(T)}\)). The paper under review is concerned with linear isometries from \(C(S)\) to \(C(T)\), and with holomorphic mappings from \(B(X)\) to \(B(T)\) that are isometries in the infinitesimal sense given by the Carathéodory-Kobayashi differential metric. The first major result, Theorem 1, is an extension of the Banach-Stone theorem that, really, is familiar to many, but may not have been written down before: If \(A\) is a norm 1 linear mapping from \(C(S)\) to \(C(T)\) that carries \(\Gamma (S)\) into \(\Gamma (T)\), then \(A\) has the form \((Ax) (t)= \alpha (t) (x\circ \tau) (t)\) for some \(\alpha\in \Gamma (T)\) and some continuous mapping \(\tau\) from \(T\) into \(S\); furthermore, \(A\) is an isometry (into) iff \(A\) is injective iff \(\tau\) is surjective. The second major result, Theorem 2, extends standard results from complex function theory with a sort of ``Schwarz lemma'': If \(F\) is a holomorphic function from \(B(S)\) to \(B(T)\) for which there is an \(x\in B(S)\) such that \((1-|F(x) |^2 )^{-1} dF(x) ((1-|x|^2) v)\in \Gamma (T)\) for all \(v\in \Gamma (S)\), and if \(dF (x)\) is injective, then \(F\) is an (infinitesimal) isometry. A third theorem gives a representation of \(F\) from Theorem 2 as a sort of Möbius transformation, extending the standard theorem on holomorphic automorphisms of the unit disc. The paper concludes with examples exhibiting, for instance, linear isometries on \(C(T)\) (with \(T= [0, 1])\) whose restrictions to \(B(T)\) are not holomorphic isometries. \{Reviewer's remarks: Lemma 2, used to prove Theorem 1, is easily proven, but the proof here fails: the two equations preceding equation (6) are entirely equivalent, and equation (6) does not follow from them. It also appears that the proof of Theorem 3 invokes Theorem 1 even though \(F\) need not satisfy the linearity hypothesis of the latter. Finally, there are a fair number of minor misprints and errors\}.