4
$\begingroup$

Can the Brouwer fixed point theorem be formulated for non-Hausdorff spaces ?

More particularly, is there a formulation of the Brouwer fixed point theorem which covers both the standard case of metrizable spaces, and the case of finite topological spaces ?

For example, is the following true:

Each endomorphism of a quasi-compact $T_0$ space which is an absolute extensor for normal spaces, necessarily has a fixed point.

As pointed out in comments by Michael Greinecker, Kinoshita, 1953 gives a counterexample to the following which is a contractible compact subset of $\Bbb R^3$, see also Bing,The elusive fixed point property and references in The generalization of Brouwer's fixed point theorem?.

Each endomorphism of a contractible quasi-compact $T_0$ space necessarily has a fixed point. (false!)

hence we replace "contractible" by "absolute extensor"

Note that being $T_0$ is necessary: the indiscrete space with two points is quasi-compact and contractible, yet the permutation of the two points has no fixed point. Perhaps one can relax the notion of a fixed point, e.g. by requiring that $x$ and $f(x)$ are topologically indistinguishable.

By the Brouwer theorem for finite topological spaces I mean what is implied by the Brouwer fixed point theorem for the geometric realisation of a finite topological space: namely, an endomorphism $f:K\to K$ of a finite topological space $K$ has a fixed point provided the endomorphism $\lvert f\rvert\colon\lvert K\rvert\to\lvert K\rvert$ of the geometric realisation of $K$ has a fixed point.

$\endgroup$
18
  • 2
    $\begingroup$ There is a famous example by Kinoshita of a contractible compact subset of $\mathbb{R}^3$ that lacks the topological fixed-point property. $\endgroup$ Commented Dec 11, 2022 at 19:34
  • $\begingroup$ Thanks! I replaced "contractible" with "absolute extensor for normal spaces", which is different. Would that help ? Is this counterexample an absolute extensor for normal spaces ? $\endgroup$
    – user494312
    Commented Dec 11, 2022 at 21:35
  • 1
    $\begingroup$ I'm not that knowledgable when it comes to algebraic topology, but "acyclic absolute neighborhood retract" is enough by a fixed-point theorem of Eilenberg and Montgomery. $\endgroup$ Commented Dec 11, 2022 at 21:55
  • $\begingroup$ What is the geometric realization in the last paragraph? $\endgroup$
    – Z. M
    Commented Dec 12, 2022 at 10:09
  • 1
    $\begingroup$ @Z.M this is what is defined in the Clader's thesis. ( I seem to recall people writing there was an inaccuracy/error there but do not recall at the moment). Maybe also see this How much of homotopy theory can be done using only finite topological spaces? $\endgroup$
    – user494312
    Commented Dec 12, 2022 at 13:05

0

You must log in to answer this question.

Browse other questions tagged .