Lim colim versus colim lim. II: Derived limits over a pospace
Abstract
\v{C}ech cohomology of a separable metrizable space is defined in terms of cohomology of its nerves (or ANR neighborhoods) whereas Steenrod-Sitnikov homology is defined in terms of homology of compact subsets . We show that one can also go vice versa: in a sense, can be reconstructed from , and if is finite dimensional, can be reconstructed from . The reconstruction is via a Bousfield-Kan/Araki-Yoshimura type spectral sequence, except that the derived limits have to be "corrected" so as to take into account a natural topology on the indexing set. The corrected derived limits coincide with the usual ones when the topology is discrete, and in general are applied not to an inverse system but to a "partially ordered sheaf". The "correction" of the derived limit functors in turn involves constructing a "correct" (metrizable) topology on the order complex of a partially ordered metrizable space (such as the hyperspace of nonempty compact subsets of with the Hausdorff metric). It turns out that three natural approaches (by using the space of measurable functions, the space of probability measures, or the usual embedding ) all lead to the same topology on .
Cite
@article{arxiv.1809.00022,
title = {Lim colim versus colim lim. II: Derived limits over a pospace},
author = {Sergey A. Melikhov},
journal= {arXiv preprint arXiv:1809.00022},
year = {2022}
}
Comments
29 pages. v2: Minor changes (Proposition 3.5 from v1 has migrated to arXiv:1106.3249, where it is now called Proposition 26.14)