Could you please verify my proof?

Let $f:X \to X$ be a shrinking map on a compact metric space $(X,d)$.

In other words: $d(f(x),f(y)) < d(x,y)$ for all $x,y \in X$.

Prove: $f$ has a unique fixed point.


Define $f^n=f^{n-1} \circ f.$ Let $A_n = f^n(X)$.

(1) By induction I proved that $\bigcap_{i=1}^n A_i$ is non empty

(2) Each of the $A_i$'s is a continous mapping of a compact set hence it is compact. A compact subset of a Hausdorff space (metric spaces are Hausdorff) is closed and thus $A_i$ is closed for all $i$.

(3) $\{A_n\}_{n \in \mathbb{N}}$ is a collection of closed sets with the finite intersection property by (1) thus their intersection is non-empty. Let $A = \bigcap_{n \in N} A_n$ denote that intersection.

(4) $x \in A \implies x \in f^n(X)$ for all $n$. It follows that for every $n$ there exist $x_n \in X$ such that $f^{n}(x_n)=x$.

(5) By sequential compactness (which is equivalent to compactness for metric spaces) $x_n$ has a convergent subsequence $x_{p_n} \to a$. By continuity $f(x_{p_n}) \to f(a)$. (there is a a gap here!) But $f^{p_n}(x_{p_n}) = x$ by construction and we have $f(a)=x.$

(6) By (4) and (5) $A \subseteq f(A)$.

(7) It simple to prove that if $x \in f(A)$ then $x \in A$ thus $f(A) \subseteq A$.

(8) by (6) and (7) $A=f(A)$. The metric is a continuous function in both variables. $f$ is a continuous map from a compact space $A \times A$ to $\mathbb{R}$ (which is ordered) thus the metric obtains a maximum value. let $(x,y) \in A \times A$ be the point such that $diam(A)=d(x,y)$.

(9) Let $(a,b) \in A$ be such that $f(a) = x, f(b) =y$. Assume $x \neq y$

$$d(x,y) = d(f(a),f(b)) < d(a,b) \le diam(A) = d(x,y)$$

And we have a contradiction.

I this proof valid?

  • 2
    $\begingroup$ math.stackexchange.com/q/118536/4280 has another proof of the exact same fact, simpler. $\endgroup$ – Henno Brandsma Dec 18 '13 at 20:30
  • 3
    $\begingroup$ In (5), how do you argue that $a\in A$? Or how do you obtain $A \subset f(A)$ in (6)? $\endgroup$ – Daniel Fischer Dec 18 '13 at 20:32
  • $\begingroup$ @DanielFischer See my edit? $\endgroup$ – Saal Hardali Dec 18 '13 at 20:37
  • $\begingroup$ In (5) you say 'by continuity'. But of what? Notice that on the LHS you have $f^n$ which depends on $n$, not a single function. $\endgroup$ – Marek Dec 18 '13 at 20:40
  • $\begingroup$ oh, it's a typo. nevertheless there is a gap there i didn't notice. $\endgroup$ – Saal Hardali Dec 18 '13 at 20:42

Here is a complete proof:

Consider the function $d(x,f(x))$.

It is a continuous function from a compact space into the reals. Therefore it obtains its minimum.

If its minimum is $0$, we are done.

Otherwise suppose $d(z,f(z))=m>0$. Then $d(f(z),ff(z))<d(z,f(z))=m$ a contradiction.

And hence the minimum must indeed be $0$.

| 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.