A localization of the Artstein-Avidan-Milman theorem (Q2134442)

The Artstein-Avidan-Milman (``A-A-M'') theorem [\textit{S. Artstein-Avidan} and \textit{V. Milman}, Ann. Math. (2) 169, No. 2, 661--674 (2009; Zbl 1173.26008)] states that every fully order preserving (resp. reversing) mapping on \(\mathrm{Conv}(\mathbb{R}^n)\) is essentially the identity (resp. Legendre transform). This result was the starting point of several interesting developments, and a variety of representation theorems concerning fully order preserving (resp. reversing) mappings for specific classes of convex functions have been established. \textit{S. Artstein-Avidan} et al. [Lect. Notes Math. 2050, 61--122 (2012; Zbl 1266.26021)] showed that if \(K\), \(L\) are two windows (i.e., convex sets with nonempty interiors), then for every fully order preserving mapping \(\varphi\) between \(K\) and \(L\) , there exists a bijective fractional linear mapping \(F\) between \(K\times\mathbb{R}\) and \(L\times\mathbb{R}\) such that \(\mathrm{epi}(\varphi)f= F(\mathrm{epi} f)\), where \(\mathrm{epi}(f)\) is the epigraph of \(f\) and \(f\in\mathrm{Conv}(K)\). This was the first theorem concerning localization of the classical ``A-A-M'' theorems. However, unlike the classical ``A-A-M'' first theorem, this is just an implicit characterization of \((\varphi)f\). In this paper, the authors investigate the localization of the first ``A-A-M'' theorem with a total different approach, and they obtain an explicit characterization. Later, they also show a similar nontrivial localization for the second ``A-A-M'' theorem. The results are, however too complicated to be stated here.