A variational approach to norm attainment of some operators and polynomials (Q627555)

A classical result due to Bishop and Phelps asserts that the set of norm attaining functionals is dense in the dual space. This result inspired some variational principles such as the famous Ekeland Variational Principle. As the title suggests, in the paper under review the authors present a variational approach to norm attaining operators and polynomials. We recall some terminology: For a bounded and closed subset \(A\) of a Banach space \(X\), a bounded above function \(g:A\rightarrow\mathbb{R}\) strongly exposes \(A\) if \(g\) attains its maximum at a point \(x_{0}\in A\) and \(x_{n}\rightarrow x\) for every sequence \((x_{n})\) in \(A\) such that \(g(x_{n})\rightarrow\max g(A)\). If \(X\) is a complex Banach space, let \(\mathcal{A}_{u}(B_{X})\) denote the Banach space of bounded and uniformly continuous functions on the closed unit ball of \(X\) which are holomorphic on the open unit ball, endowed with the supremum norm. A Banach space is called Asplund if every real-valued continuous convex function defined on a nonempty open convex subset \(O\) of \(X\) is Fréchet differentiable at each point of some dense \(G_{\delta}\)-subset of \(O\). The authors prove that if \(X\) is an Asplund space and \(E\subset X^{\ast}\) is a norm closed and bounded set, and \(\Phi:E\rightarrow\mathbb{R}\) is a bounded above and upper semicontinuous function, then the set \[ \left\{ x\in X:\Phi+\text{Re}x\text{ strongly exposes }E\right\} \] is \(G_{\delta}\)-dense in \(X\). This result has several interesting applications. For example, a combination of this result and a theorem by \textit{M. D. Acosta, J. Alaminos, D. García} and \textit{M. Maestre} [J. Math. Anal. Appl. 297, No.~2, 625--644 (2004; Zbl 1086.46034)] provides the following Bishop-Phelps type result: Let \(X\) be an Asplund space. Then the set of norm attaining elements of \(\mathcal{A}_{u}(B_{X^{\ast}})\) is dense in \(\mathcal{A}_{u}(B_{X^{\ast}}).\) In the final section, the authors use some other optimization results in order to obtain results for polynomials whose transposes attain the norm. For example, it is shown that if \(P:X\rightarrow Y\) is a continuous \(n\)-homogeneous polynomial and \(\varepsilon>0\), then there is a continuous \(n\)-homogeneous polynomial \(Q:X\rightarrow Y\) so that \(\left\| Q\right\| <\varepsilon\) and \(\left( P+Q\right) ^{t}\) attains its norm.