Proximinality in operator spaces

Abstract approximation theory (approximation in normed linear spaces and other abstract spaces) (41A65) Classical Banach spaces in the general theory (46B25) Spaces of operators; tensor products; approximation properties (46B28)

Review text

The paper is concerned with proximinality in spaces of operators. More exactly, one studies the proximinality of \(L(X,G)\) in \(L(X,Y)\), where \(X,Y\) are Banach spaces and \(G\) is a subspace of \(T\). If \(L(X,G)\) is proximinal in \(L(X,Y)\), then \(G\) is proximinal in \(Y\) (Theorem 1). The converse of this result is more difficult to prove and it is known to hold only under some supplementary hypotheses, e.g., if \(G\) is a tensorial subspace of \(Y\), a notion introduced in this paper. One studies also the proximinality in spaces of nuclear operators: if \(G\) is a complemented subspace of \(X\), then \(N_1(G,Y)\) is proximinal in \(N_1(X,Y)\). If \(X,Y\) are Banach spaces, then a nuclear operator \(T:X\to Y\) is of the form \(Tx=\sum_{n=1}^\infty x^*_n(x)y_n\), \(x\in X\), for two sequences, \((x^*_n)\) in \(X^*\) and \((y_n)\) in \(Y\), with the norm \(\|T\|=\inf\sum_{n=1}^\infty \|x^*_n\|\|y_n\|,\) the infimum being taken with respect to all representations of \(T\) in the given form. If \(G\) is a complemented subspace of \(X\) with continuous projection \(P:X\to G,\) then every \(y^*\in G^*\) defines through \(x^*:=y^*\circ P\) a continuous linear functional \(x^*\in X^*\). The best approximation problem is understood in this sense: approximating nuclear operators \(T\in N_1(X,Y)\) by operators \(S\) of the form \(Sx= \sum_{n=1}^\infty (y^*_n\circ P)(x) z_n\), \(x\in X\), where \((y^*_n)\) is a sequence in \(G^*\) and \((z_n)\) a sequence in \(Y\). A similar result holds for the spaces \(N_p(G,\ell^p)\) and \(N_p(X,\ell^p)\) of \(p\)-nuclear operators, \(1<p<\infty\).

Recommendations

Cited in

(11)

This page was built for publication: Proximinality in operator spaces

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2839855)