3.5 Hilbert numbers \(h_n\)
For the Hilbert numbers we work over \(\mathbb {R}\) and use \(L_2 := \ell _2(\mathbb {N}; \mathbb {R})\).
For \(S \in \mathcal{L}(X, Y)\) and \(n \in \mathbb {N}_0\),
Let \(X, Y\) be Banach spaces. Then the Hilbert numbers \(h = (h_n)_n\) form an s-number sequence: they satisfy the axioms (S1)–(S5). The sequence \(h\) is not strict: (S5\('\)) fails in general.
Each axiom is obtained by pushing the corresponding property of the approximation numbers \(a_n\) through the supremum defining \(h_n\): nonnegativity, antitonicity, subadditivity (S2), the ideal property (S3) and the rank axiom (S4) transfer directly. For the normalisations, \(h_0(S) = \| S\| \) combines the upper bound \(h_n(S) \le \| S\| \) with a rank-one factorisation of arbitrary unit vectors through \(L_2\), and \(h_n(\mathrm{id}_{\ell _2^{n+1}}) = 1\) (S5) tests the supremum with the canonical embedding and projection between \(\ell _2^{n+1}\) and \(L_2\).