Absolute widths of some embeddings (Q1890589)

In this review capital letters will stand for arbitrary Banach spaces and \(B_Z\) for the unit ball of the Banach space \(Z\). The Kolmogorov \(m\)-width of an operator \(v:E \to F\) is defined by \(d_m(v) =\inf_L |Q_Lv |\), where \(L\) is an arbitrary subspace of \(F\), with \(\dim L\leq m\), and \(Q_L: F\to F/L\) denotes the quotient mapping. \textit{R. S. Ismagilov} [Russ. Math. Surveys 29, No. 3, 169-186 (1974; Zbl 0303.41039)], defined the absolute width of an operator \(u:X\to Y\) by \(d^a_m (u)= \inf_j |ju |\), where \(j:Y \to\overline Y\) is an arbitrary isometric embedding of \(Y\) in a Banach space \(\overline Y\). R. S. Ismagilov proved that \(d^a_m (u)= d_m (iu)\), where \(i\) denotes the natural embedding of \(Y\) into \(M(B_{Y^*})\) -- the Banach space of all bounded functions on \(B^{Y^*}\). The aim of the present paper is to prove the formula \(d^a_m (I^n_{pq}) =(n-m)^{1/p- 1/q}\), where \(I^n_{pq}\) stands for the natural identity operator from \(\ell^n_p\) to \(\ell^n_q\), \(1\leq q\leq p\leq \infty\). In the case of Kolmogorov width the above formula is well known (see e.g. \textit{A. Pietsch}, Eigenvalues and \(s\)-numbers, (1987; Zbl 0615.47019)].