Enveloppes of Semicontinuous functions on Hereditarily Lindelรถf spaces #
In this file, we show that, if X is a hereditarily Lindelรถf space and ๐ is any family
of upper semicontinuous functions on X, then there is a countable subfamily ๐'
with the same infimum / lower enveloppe. Most importantly, this applies whenever X has a
SecondCountableTopology.
The codomain E of the functions is assumed to be linearly ordered, and to admit a countable
order-dense subset. In particular we do not assume that arbitrary infima exist in E, so our
result is of the form "if IsGLB ๐ s, then there is a countable ๐' โ ๐ such that IsGLB ๐' s".
Of course we also provide the dual statements for lower semicontinuous functions.
Implementation Notes #
There is currently no way to state the hypothesis "E admits a countable order-dense subset"
which would be inferrable by typeclass inference. Instead, we assume
[TopologicalSpace E] [OrderClosedTopology E] [DenselyOrdered E] [SeparableSpace E], and
use Dense.exists_between to show that a chosen countable dense subset is order-dense.
This is not completely satisfying because the hypotheses on E should be purely order-theoretic,
but in practice E is either Real, NNReal, ENNReal or EReal, all of which are already
equipped with the order topology.
References #
- [N. Bourbaki, Topologie Gรฉnรฉrale, Chapitre IX, Appendice I][bourbaki1974] (this appendix does not seem to exist in the English translation)
If a function s : X โ E can be written as the infimum of a family ๐ of upper semicontinuous
functions then, assuming that X is hereditarily Lindelรถf (for example, second countable),
s can in fact be written as the infimum of some countable subfamily ๐'.
This is implication a) โ b) in [N. Bourbaki, Topologie Gรฉnรฉrale, Chapitre IX, Appendice I, Proposition 3][bourbaki1974]
See the module docstring for a discussion of the assumptions on E.
If a function s : X โ E can be written as the supremum of a family ๐ of lower semicontinuous
functions then, assuming that X is hereditarily Lindelรถf (for example, second countable),
s can in fact be written as the supremum of some countable subfamily ๐'.
This is implication a) โ b) in [N. Bourbaki, Topologie Gรฉnรฉrale, Chapitre IX, Appendice I, Proposition 3][bourbaki1974]
See the module docstring for a discussion of the assumptions on E.