Multilinear absolutely summing operators associated to a tensor norm (Q6158444)
From MaRDI portal
scientific article; zbMATH DE number 7690510
Language | Label | Description | Also known as |
---|---|---|---|
English | Multilinear absolutely summing operators associated to a tensor norm |
scientific article; zbMATH DE number 7690510 |
Statements
Multilinear absolutely summing operators associated to a tensor norm (English)
0 references
31 May 2023
0 references
Let \(1\le p\le \infty\) and \(\beta_n\) be a tensor norm on the $n$-fold full tensor product. The author introduces the norm \(d_{p^*}(\beta_n)\) on the \((n+1)\)-fold tensor product, \(d_{p^*}(\beta_n)\colon X_1\otimes\cdots X_n\otimes X_{n+1}\to [0,\infty)\), by \[ d_{p^*}(\beta_n)(u)=\inf\left\{l_{p^*}(x_i^{n+1}),\sup_{\|\alpha\|_{p^*}\le 1} \beta_n\left(\sum_{i=1}^m \alpha_i x_i^1\otimes\cdots\otimes x_i^n\right)\right \} \] where the infimum is taken over all representations \(\sum_{i=1}^mx_i^1 \otimes\cdots\otimes x_i^n\otimes x_i^{n+1}\) of \(u\). It is shown that \(d_{p^*}(\beta_n)\) is an \((n+1)\)-tensor norm on the full tensor product. Let \(1\le p\le \infty\) and \(\beta_n\) be a tensor norm on the full tensor product. An \(n\)-linear operator \(T\colon X_1\times \cdots\times X_n\to Y\) is said to be \((\beta_n,p)\)-summing if there is \(C>0\) so that, for \((x_i^j)_{1\le i\le m}\) in \(X_j\) and for \(1\le j\le n\), \[ \left(\sum_{i=1}^m |T(x_i^1,\ldots,x_i^n)|^p\right)^{1/p}\le C \sup_{\|\alpha\|_{p^*}\le 1} \beta_n\left(\sum_{i=1}^m \alpha_ix_i^1\otimes \cdots\otimes x_i^n\right) \] and one defines \(\pi_p^{\beta_n}(T)\) to be the infimum of all \(C\) which satisfy this inequality. Using \(\Pi_p^{\beta_n}(X_1,\ldots,X_n:Y)\) to denote the space of all \((\beta_n,p)\)-summing operators from \(X_1\times \cdots\times X_n\) to \(Y\), the author shows that the spaces \(\left(\Pi_p ^{\beta_n}(X_1,\ldots,X_n:Y^*),\pi_p^{\beta_n}\right)\) and \(\left(X_1\otimes \cdots\otimes X_n\otimes Y, d_{p^*}(\beta_n)\right)^*\) are isometrically isomorphic. An \((n+1)\)-linear operator \(T\colon X_1\times \cdots\times X_{n+1}\to Y\) is said to be Saphar \((\beta_n,p)\)-summing if there is \(C>0\) so that for \((x_i^j)_{1\le i\le m}\) in \(X_j\) for \(1\le j\le n+1\) \[ \left(\sum_{i=1}^m |T(x_i^1,\ldots,x_i^{n+1})|^p\right)^{1/p}\le C\sup_{ A\in \Pi_p^{\beta_n}(X_1,\ldots,X_n:X_{n+1}^*),\pi_p^{\beta_n}(A)\le1}\left(\sum _{i=1}^m \left|\langle x_i^{n+1},A(x_i^1,\ldots, x_i^n)\rangle\right|^p \right)^{1/p} \] and one defines \(\pi_p^{\mathrm{Saphar} \beta_n}(T)\) to be the infimum of all \(C\) which satisfy this inequality. It is shown that an \((n+1)\)-linear mapping is Saphar \((\beta_n,p)\)-summing if and only if it is \((d_{p^*}(\beta_n),p) \)-summing and that the \(\pi_p^{\mathrm{Saphar}\beta_n}\) and \(\pi_p^{d_{p^*}( \beta_n)}\) norms coincide. A bilinear operator \(T\colon X_1\times X_{2}\to Y\) is said to be 2-Saphar \(p\)-summing if there is \(C>0\) so that, for \((x_i^j)_{1\le i\le m}\) in \(X_j\) and for \(j=1,2\), \[ \left(\sum_{i=1}^m |T(x_i^1,x_i^{2})|^p\right)^{1/p}\le C\sup_{ A\in \Pi_p^{p}(X_1,:X_{2}^*),\pi_p(A)\le 1}\left(\sum _{i=1}^m \left|\langle x_i^{2},A(x_i^1)\rangle\right|^p \right)^{1/p}, \] and one defines \(\pi_p^{\mathrm{Saphar}_2}(T)\) to be the infimum of all \(C\) which satisfy this equation. The space of all such bilinear mappings is denoted by \(\Pi_p^{\mathrm{Saphar}_2}(X_1,X_2,Y)\). Supposing that the space \(\Pi_p^{\mathrm{Saphar}_n}\) of \(n\)-Saphar \(p\)-summing mappings is defined, the author says that an \((n+1)\)-linear operator \(T\colon X_1\times \cdots\times X_{n+1}\to Y\) is Saphar \((\beta_n,p)\)-summing if there is \(C>0\) so that, for \((x_i^j)_{1\le i\le m}\) in \(X_j\) and for \(1\le j\le n+1\), \begin{align*} & \left(\sum_{i=1}^m |T(x_i^1,\ldots,x_i^{n+1})|^p\right)^{1/p} \\ & \le C\sup_{ A\in \Pi_p^{\mathrm{Saphar}_n}(X_1,\ldots,X_n:X_{n+1}^*),\pi_p^{\mathrm{Saphar}_n}(A)\le1} \left(\sum_{i=1}^m \left|\langle x_i^{n+1},A(x_i^1,\ldots, x_i^n)\rangle \right|^p\right)^{1/p}. \end{align*} Starting with \(d_{p^*}^1=\|\cdot\|\), the author inductively defines \(d_{p^*}^{n+1}\) by \(d_p^{n+1}:=d_{p^*}(d_{p^*}^n)\) and shows that an \((n+1)\)-linear mapping is \((n+1)\)-Saphar \(p\)-summing if and only if it is \((d_{p^*}^{n+1},p)\)-summing. In an analogous way, starting with the ideals of strongly \(p\)-summing and semi \(p\)-summing mappings, the author defines the ideals of Saphar strongly \(p\)-summing and Saphar semi \(p\)-summing \((n+1)\)-linear mappings and shows that the ideal of Saphar strongly \(p\)-summing \((n+1)\)-linear mappings is isomorphic to the space of \((d_{p^*}^{n+1}(\pi_2),p)\)-summing \((n+1)\)-linear mappings, while the ideal of Saphar semi \(p\)-summing \((n+1)\)-linear mappings is isomorphic to the space of \((d_{p^*}^{n+1}(\epsilon_2),p)\)-summing \((n+1)\)-linear mappings. The paper concludes with some inclusion results for summing multilinear mappings.
0 references
\(p\)-summing operators
0 references
tensor norm
0 references
Saphar tensor norm associated to a tensor norm
0 references
absolutely summing operators associated to a tensor norm
0 references
Saphar \(p\)-summing
0 references
Saphar strongly \(p\)-summing
0 references
Saphar semi \(p\)-summing operators
0 references
dominated
0 references
multiple summing operators
0 references
0 references
0 references
0 references
0 references