The subspace problem for weighted inductive limits revisited (Q5928660): Difference between revisions

Let \(G\) be an open subset of \(\mathbb{C}^N\), \(N\geq 1\), and let \({\mathcal V}= (v_n)_{n\in\mathbb{N}}\) be a decreasing sequence of strictly positive continuous functions on \(G\). For each \(n\) let \(Hv_n(G)\) denote the Banach space of all holomorphic functions \(f\) on \(G\) for which \(\|f\|_n= \sup_G v_n|f|\) is finite. The corresponding weighted inductive limit is denoted by \({\mathcal V}H(G)= \text{ind}_nHv_n(G)\). In an effort to describe the locally convex inductive limit topology of this space by a system of weighted sup-seminorms, \textit{K. D. Bierstedt}, \textit{R. Meise} and \textit{W. H. Summers} [Trans. Am. Math. Soc. 272, 107-160 (1982; Zbl 0599.46026)] introduced the associated system \(\overline V=\overline V({\mathcal V})\), consisting of all strictly positive continuous functions \(\overline v\) on \(G\) for which \(\overline v/v_n\) is bounded on \(G\) for each \(n\in\mathbb{N}\), and the locally convex space \(H\overline V(G)\) of all holomorphic functions \(f\) on \(G\) with \(p_{\overline v}(f)= \sup_G\overline v|f|< \infty\) for each \(\overline v\in\overline V\). In fact, \({\mathcal V}H(G)=H\overline V(G)\) always holds algebraically, and the projective description problem asks when the two spaces have the same topology. Considering spaces of continuous functions instead of spaces of holomorphic functions, one defines \(Cv_n(G)\), \({\mathcal V}C(G)\) and \(C\overline V(G)\) in an analogous way. Under condition (D), a mild restriction on \({\mathcal V}\), the projective description problem is equivalent to asking when \({\mathcal V}H(G)\) is a topological subspace of \({\mathcal V}C(G)\). In the article quoted above, it was proved that projective description holds whenever \({\mathcal V}H(G)\) is a (semi-) Montel space. For weighted inductive limits of spaces of continuous functions, projective description holds whenever the above-mentioned condition (D) is satisfied and, in particular, if \({\mathcal V}\) is regularly decreasing; i.e., for each \(n\) there exists \(m\geq n\) such that for each subset \(Y\) of \(G\), \(\inf_Y v_m/v_n> 0\) implies \(\inf_Y v_k/v_n> 0\) for \(k= m+1,m+2,\dots\)\ . The authors of the present paper, in [Mich. Math. J. 42, No. 2, 259-268 (1995; Zbl 0841.46014)], constructed the first counterexamples to projective description for spaces of holomorphic functions. In particular, they proved that it may happen that \(H\overline V(G)\) is semi-Montel, but \({\mathcal V}H(G)\) is not. In an appendix to the present paper, the authors now show that this cannot happen if \({\mathcal V}\) is, in addition, regularly decreasing. It was asked in \textit{K. D. Bierstedt} and \textit{R. Meise} [Doğa, Turk. J. Math. J. 10, No. 1, Spec. Iss. 54-82 (1986; Zbl 0970.46541)] if, also in the case of spaces of holomorphic functions, projective description holds whenever \({\mathcal V}\) is regularly decreasing. The authors try to solve this question in the article under review and come up with a complicated example of a very disconnected open subset \(\Omega\) of \(\mathbb{C}^2\) and a regularly decreasing sequence \({\mathcal V}= (v_n)_n\) of strictly positive functions on \(\Omega\) such that \(H\overline V(\Omega)\) is not bornological, whence projective description does not hold, and \({\mathcal V}H(\Omega)\) is not a topological subspace of \({\mathcal V}C(\Omega)\). This is done by showing that \(H\overline V(\Omega)\) contains a complemented subspace isomorphic to a non-bornological sequence space \(X^\perp\). And the main points in the proof are the construction of this sequence space and the embedding of an isomorphic copy of \(X^\perp\) in \(H\overline V(\Omega)\). For the last point, the second complex dimension is essential, as is an application of the classical Liouville theorem. Unfortunately, the authors point out that it is not possible to modify their example in such a way that the weights \(v_n\) become at least upper semicontinuous, and hence the original question remains open. For a recent survey on weighted inductive limits of spaces of holomorphic functions and on open problems in this area, see \textit{K. D. Bierstedt} [Bull. Soc. Roy. Sc. Liège 70, 167-182 (2001)].