\((p,q)\)-dominated multilinear operators and Lapresté tensor norms (Q1630615)

From MaRDI portal
scientific article
Language Label Description Also known as
English
\((p,q)\)-dominated multilinear operators and Lapresté tensor norms
scientific article

    Statements

    \((p,q)\)-dominated multilinear operators and Lapresté tensor norms (English)
    0 references
    10 December 2018
    0 references
    The authors consider Banach spaces \(X_1,\ldots, X_n\) and \(Y\), \(1\le p,q\le\infty\) such that \(\frac{1}{p}+\frac{1}{q}\le 1\) and a reasonable tensor norm \(\beta\) on \(X_1\otimes\cdots\otimes X_n\). Setting \(r\) to satisfy \(\frac{1}{r}+\frac{1}{p^*}+\frac{1}{q^*}=1\) and using the weak \(q^*\)- and weak \(p\)-norms \[ \|(p_i-q_i)_{i=1}^m \|_p^{w,\beta}=\sup_{ \varphi\in B_{(X_1\otimes\cdots\otimes X_n,\beta)^*}}\left(\sum_{i=1}^m| f_\varphi(p_i)-f_\varphi(q_i)|^p\right)^{1/p} \] where \(f_\varphi(x_1\otimes\cdots \otimes x_n)=\varphi(x_1)\ldots \varphi(x_n)\), and similarly for \(\|\cdot\|_{q^*}^{w,\beta}\) they define the Lapresté norm \(\alpha_{p,q}^\beta\) on \(X_1\otimes\cdots\otimes X_n\otimes Y\) by \[ \alpha_{p,q}^\beta(u)=\inf\left\{\|(\lambda_i)_{i=1}^m\|_r\|(p_i-q_i)_{i=1}^m \|_{q^*}^{w,\beta}\|(y_i)_{i=1}^m\|_{p^*}^w :u=\sum_{i=1}^m\lambda_i(p_i-q_i) \otimes y_i\right\} \] and say that a multilinear mapping \(T: X_1\times\cdots\times X_n\to Y\) is \(\beta\)-\((p,q)\)-summing if there is \(C>0\) such that \[ \|\langle y_i^*, T(x_i^1,\dots, x_i^n)-T(z_i^1,\dots, z_i^n)\rangle)_{i=1}^m \|_{r^*}\le C\|(x_i^1\otimes\cdots\otimes x_i^n-z_i^1\otimes\cdots\otimes z_i^n)_{i=1}^m\|_p^{w,\beta} \|(y_i^*)_{i=1}^m\|_q^w \] for all finite sequence \((x_i^1,\ldots x_i^n)_{i=1}^m\), \((z_i^1,\ldots z_i^n )_{i=1}^m\) in \(X_1\times\cdots\times X_n\) and \((y_i^*)_{i=1}^m\) in \(Y^*\). The space of all \(\beta\)-\((p,q)\)-summing multilinear mapping from \(X_1\times\cdots \times X_n\) to \(Y\) is denoted \(\mathcal{D}_{p,q}^\beta(X_1,\ldots, X_n;Y)\). When \(\beta=\pi\), the projective tensor norm, the authors show that for a multilinear mapping \(T: X_1\times\cdots\times X_n\to Y\) the following are equivalent: \begin{itemize} \item[(i)] \(T\) is \(\pi\)-\((p,q)\)-dominated; \item[(ii)] For any weakly compact norming subsets \(K\) in \(B_{\mathcal{L}(X_1,\ldots,X_n)}\) and \(L\) in \(B_{Y^{**}}\) there are \(C>0\) and regular probability measures \(\mu\) on \(K\) and \(\nu\) on \(L\) so that \[ |\langle y^*, T(x)-T(z)\rangle|\le C\left(\int_K|\psi(x)-\psi(x)|^p\,d\mu(\psi) \right)^{\/p}\left(\int_L|y^{**}(y^*))|^q\,d\mu(y^{**})\right)^{\/q}; \] \item[(iii)] \(T\) admits a factorization through a Banach space \(Z\) as the composition of a \(p\)-summing multilinear operator and a linear operator whose adjoint is \(q\)-summing. \end{itemize} For general reasonable crossnorms \(\beta\), the authors prove that \(\alpha_{p,q}^ \beta\) is a norm on \(X_1\otimes\cdots\otimes X_n\otimes Y\) and that \((X_1\otimes\cdots\otimes X_n\otimes Y, \alpha_{p,q}^\beta)^*\) is isometrically isomorphic to \(\mathcal{D}_{p,q}^\beta(X_1,\ldots,X_n;Y^*)\). Finally, the authors show how the Lapresté tensor norm can be used to deduce results concerning the Chevet-Saphar tensor norms.
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    dominated operators
    0 references
    multilinear and polynomial mappings
    0 references
    tensor products
    0 references
    ideals of multilinear mappings
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references