Suppose $X$ is a compact connected metric connected space and for every $\epsilon >0$ , there exists a continuous surjective function $f : X \rightarrow [0,1]$, that for all $y \in [0,1]$, the diameter of the set $f^{-1} (y)$ is less than $\epsilon$.

Prove that $X$ has the fixed point property; that is, for all continuous $g : X \rightarrow X$, there exists $x_0 \in X$ that $g(x_0) = x_0$.

If we know that for a continuous surjective function $f : X \rightarrow [0,1]$ the diameter of set $f^{-1} (y)$ is zero, then we know that $X$ and $[0,1]$ are homeomorphic, so the statement is proven. But I have no idea for that.

Hints are appreciated.

  • $\begingroup$ You may be able to prove that $X$ is simply connected. Choose two paths on $X$ with the same start and end-point, they will also be paths on $[0,1]$ which is simply connected, and so are homotopic. By Brouwers fixed point theorem any continuous function will then have a fixed point. $\endgroup$ – Damien Feb 15 '18 at 20:06
  • $\begingroup$ @DanielBeale: Simply connected and compact does not imply the fixed point property. $\endgroup$ – Moishe Kohan Feb 16 '18 at 0:41
  • 1
    $\begingroup$ It's a very neat problem, but you should not be trying to prove the existence of a homeomorphism $X\to [0,1]$. Instead, show that there is a sequence of points $x_n\in X$ such that $\lim_{n\to\infty} d(g(x_n), x_n)=0$. This even works if you replace $[0,1]$ with a closed $N$-disk. $\endgroup$ – Moishe Kohan Feb 16 '18 at 3:03
  • $\begingroup$ @MoisheCohen I was thinking of compact sets in $\mathbb{R}^N$. Is it not true that the Lefschetz fixed point theorem works for simply connected compact spaces? One can take a convex subset of $Y \subset X$ and then create homotopies to the continuous functions in $Y$. $\endgroup$ – Damien Feb 16 '18 at 15:13
  • 1
    $\begingroup$ @DanielBeale: Think of the antipodal map of the 2-sphere to itself. How many fixed points does it have? $\endgroup$ – Moishe Kohan Feb 16 '18 at 15:15

Here is one solution which works even when you replace the closed unit interval with the closed unit $N$-ball $B$. I will describe steps of the proof and leave you to fill in the details. (Since the problem smells like a homework.)

(0). Prove the following lemma akin to the uniform continuity of continuous functions on compact metric spaces (and proven in a similar fashion):

Lemma. Let $f: (X,d_X)\to (Y,d_Y)$ be a continuous map between two compact metric spaces such that for every $y\in Y$, $diam(f^{-1}(y))\le \epsilon$. Then there exists $\sigma>0$ (depending on $f$ and on $\epsilon$) such that $$ \forall x_1, x_2\in X, ~~~d_Y(f(x_1), f(x_2))<\sigma \Rightarrow d_X(x_1, x_2)< 2\epsilon. $$

Proof. Suppose the claim fails. Then for every natural number $n$ there exists a pair of points $x_n, x'_n\in X$ such that $d_X(x_n, x'_n)\ge 2\epsilon$ while $$ d_Y(f(x_n), f(x'_n))<\frac{1}{n}.$$ In view of compactness of $X$, after passing to a subsequence, we can assume that $$ \lim_{n\to\infty} x_n=x, \lim_{n\to\infty} x'_n=x', $$ with $d_X(x,x')\ge 2\epsilon$, while $$ \lim_{n\to\infty} f(x_n)= \lim_{n\to\infty} f(x'_n)=y\in Y, $$ and $f(x)=f(x')=y$. Thus, the preimage of $y$ has diameter $\ge 2\epsilon$. A contradiction. qed

(1). Given $f$ and $g$ as in your problem ($f$ is surjective and $diam(f^{-1}(y))<\epsilon$ for all $y\in B$), construct a piecewise-linear continuous map $h: B\to B$ such that the following diagram is "almost commutative": $$ \require{AMScd} \begin{CD} X @>{g}>> X\\ @VVfV @VVfV \\ B @>{h}>> B \end{CD} $$ meaning that $d(f\circ g, h\circ f)<\delta(\epsilon)$, where $$ \lim_{\epsilon\to 0}\delta(\epsilon)=0. $$

I will explain how to do this in the case $B=[0,1]$, the extension to the case when $B$ is higher-dimensional is quite straightforward. In order to construct such $h$ first pick a finite subset $0=y_0< y_1 < y_2 < ... <y_n\in [0,1]$ (with $|y_i- y_{i-1}|$ sufficiently small for all $i$) such that $$ \forall i\in \{1,...,n\}, ~~diam(f^{-1}([y_{i-1}, y_i])) < 2\epsilon $$ Then define $h$ on the finite subset $\{y_0,...,y_n\}$ so that $$ \forall y_i, \exists x_i\in f^{-1}(y_i), h(y_i)=fg(x_i). $$ Then extend $h$ to the rest of $[0,1]$ linearly on each interval $[y_{i-1},y_i]$.

(2). Use the fact that $B$ has the fixed point property to show that $g$ "almost" has a fixed point, i.e.: For every $\epsilon>0$ there is a point $x\in X$ such that $d(g(x), x)\le \eta(\epsilon)$, where $$ \lim_{\epsilon\to 0}\eta(\epsilon)=0. $$

To prove this, take $y\in B$ such that $h(y)=y$ and think of its preimage $f^{-1}(y)$.

(3). Conclude that there is a sequence $(x_n)$ in $X$ such that $$ \lim_{n\to\infty} d(g(x_n), x_n)=0. $$

(4). Use compactness of $X$ to show that $g$ has a fixed point in $X$.

| cite | improve this answer | |
  • $\begingroup$ thank you very much.I'm sorry to tell I don't know the proof of the Lemma,although I know the proof of the statement u mentioned. $\endgroup$ – user115608 Feb 17 '18 at 19:19
  • $\begingroup$ @user115608: Well, try to prove it (by analogy with the uniform continuity proof) since you do not know a proof of the lemma. Math is not about knowing some proofs but about coming up with new proofs. I also did not know how to prove this lemma before I proved it. $\endgroup$ – Moishe Kohan Feb 17 '18 at 19:55
  • $\begingroup$ I did,I tried.But I don't know if the diameter of $f^{-1}(y)$ is small, how can I use it to prove that the diameter of the preimage of a small interval is small. $\endgroup$ – user115608 Feb 18 '18 at 3:48
  • $\begingroup$ @user115608: How would you prove uniform continuity of a continuous function on a compact metric space (say, on an interval)? $\endgroup$ – Moishe Kohan Feb 18 '18 at 13:44
  • $\begingroup$ @Mouse Cohen would you please see this:math.stackexchange.com/questions/2653248/… $\endgroup$ – user115608 Feb 19 '18 at 6:15

consider $h: X \rightarrow X$ continuous. I want to prove there's $x \in X$ that $h(x)=x$. for $\epsilon > 0$ consider $foh -f : X \rightarrow [0,1]$.because $f$ is surjective, there is $x \in X$ that $(foh -f) (x) \leq 0$ and there is $y \in X$ that $(foh -f) (y)\geq 0$. $foh -f $ is continuous and $X$ is connected, so there exists $t \in X$ that $(foh -f)(t) =0$. then $t , h(t) \in f^{-1}(f(t))$. so $d(t , h(t)) < \epsilon$.so we can have a sequence $t_n$ that $d(t_n , h(t_n)) < 1/n$. because $X$ is compact we can find $s \in X$ that $s = h(s)$.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.