Frame potential for finite-dimensional Banach spaces (Q2321346): Difference between revisions

\textit{J. J. Benedetto} and \textit{M. Fickus} proved in [Adv. Comput. Math 18, 357--385 (2003; Zbl 1028.42022)] that for every \(N\geq n,\) every \(n\)-dimensional Hilbert space \(H\) has a tight frame consisting of \(N\) unit vectors. The main tool in their proof is the study of the frame potential of a finite set \(\left(x_j\right)_{j=1}^N \subset H,\) defined as the value \(\sum_{i,j=1}^{N}\left|\langle x_j,x_i\rangle\right|^2.\) Let \(X\) be a finite dimensional Banach space. In this paper, the authors define the frame potential of a finite sequence of pairs \(\left(x_j, x_j^\ast\right)_{j=1}^N \subset X \times X^\ast\) as the square of the \(2\)-summing norm of the associated frame operator \(S(x) = \sum_{j=1}^N x_j^\ast(x)x_j.\) The previous sequence of pairs is said to be a finite unit norm tight frame (FUNTF) for \(X\) if \(\|x_j\| = \|x_j^\ast\| = x_j^\ast(x_j) = 1\) for all \(1\leq j \leq N\) and its frame operator is a scalar multiple of the identity. The authors prove that, for every complex \(n\)-dimensional Banach space with a \(1\)-unconditional basis and each \(N\geq n,\) there exists a FUNTF of length \(N\) for \(X.\) Moreover, a sequence \(\left(x_j, x_j^\ast\right)_{j=1}^N \subset X \times X^\ast\) satisfying \(\|x_j\| = \|x_j^\ast\| = x_j^\ast(x_j) = 1\) for all \(1\leq j \leq N\) is a FUNTF if and only if it minimizes the frame potential. These are the main results of Section~4. The characterization of FUNTFs in terms of frame potential depends on the fact that, for a finite dimensional Banach space \(X,\) the identity map is smooth in the space \(\Pi_2(X,X)\) of \(2\)-summing operators on \(X\) (Theorem 3.2). Section 5 contains several results related to smoothness and strict convexity of \(\Pi_2(X,X)\) and Section 6 discusses a result concerning maximal error due to the loss of one coordinate. Section 7 presents several partial results on the existence of FUNTFs of given length \(N\geq n\) for real \(\ell_1^n\). The final section collects some related open problems.