A proof of uniqueness of the Gurariĭ space (Q375860): Difference between revisions

One has wondered for some years whether there is an elementary proof of Lusky's theorem that the separable Gurariĭ space \(\mathbb{G}\) is unique (up to a linear, bijective isometry), and also of Gevorkjan's observation that it is universal among separable Banach spaces. The paper under review shows that there is! Not an easy one, but no theorem is used that Banach did not know. Recall the definition of Gurariĭ spaces. A Gurariĭ space is a Banach space \(Z\) with the following property: Given finite-dimensional Banach spaces \(X\subset Y\), \(\varepsilon>0\) and a linear isometric embedding \(f:X\to Z\), then \(f\) extends to a linear \(g:Y\to Z\) with \(\|g\|\cdot\|g^{-1}\|<1+\varepsilon\) (an \(\varepsilon\)-isometry into). These spaces are named after \textit{V. I. Gurariĭ} who constructed, in [{Sib. Mat. Zh. 7, 1002--1013 (1966; Zbl 0166.39303)], even a separable such Banach space. Gurariĭ himself observed that all separable spaces of the type he constructed are almost isometric. That they are indeed isometric was proved by \textit{W. Lusky} [Arch. Math. 27, 627--635 (1976; Zbl 0338.46023)] using deep techniques of \(L_1\)-preduals due to Lazar and Lindenstrauss. \textit{J. L. Gevorkjan}'s observation from [Funct. Anal. Appl. 8, 157 (1974); translation from Funkts. Anal. Prilozh. 8, No.~2, 72 (1974; Zbl 0296.46019)] is a much more direct application of the same techniques. A little about the proof: Both Lusky's and Gevorkjan's results follow by a clever use of the following principle for finite-dimensional Banach spaces \(X\) and \(Y\): When \(f:X\to Y\) is an \(\varepsilon\)-isometry, there exist a finite-dimensional Banach space \(Z\) and isometric embeddings \(i:X\to Z\) and \(j:Y\to Z\) such that \(\|j\circ f-i\|\leq\varepsilon\).}