10

If two continuous mappings $f$ and $g$ of an interval into ifself conumute,that is $$f(g(x))=g(f(x))$$,then they have a common fixed point?

This problem is from Mathemmatical Analysis (Zorich) PP169,Thank you everyone.

math110
  • 93,304
  • 2
    http://mathoverflow.net/questions/3332/two-commuting-mappings-in-the-disk – Will Jagy Jul 24 '13 at 03:37
  • 1
    As @WillJagy posted, it looks like this doesn't hold unless you apply more conditions to at least one of the functions. For example, if one of the functions is invertible, it's fairly easy to prove. – qaphla Jul 24 '13 at 04:04
  • 1
    In one dimension, there is a point with $f(x) = g(x).$ However, this is weaker than talking about fixed points. – Will Jagy Jul 24 '13 at 04:12
  • Hello,can you example why doesn't hold, – math110 Jul 24 '13 at 09:15

3 Answers3

5

In the textbook by Zorich published in 2002 this exercise states the following:

If two continuous maps $f$ and $g$ of a line segment into itself commutate, i.e. $f\circ g = g \circ f$, then they have a common fixed point.

In the latest version of the textbook by Zorich (2015) this exercise is modified to almost the opposite statement:

If two continuous maps $f$ and $g$ of a line segment into itself commutate, i.e. $f\circ g = g \circ f$, then they do not always have a common fixed point, though for linear maps and polynomials in general we always have such a point.

Here is the proof for linear functions.

Let $f(x), g(x)$ are commuting linear functions. Suppose that none of them is constant or identical (otherwise our task is trivial). Then they intersect the line $e(x) = x$ in a single point. This means that both of them has a single fixed point.

Let $f(x_0) = x_0$. Suppose, that $g(x_0) = x_1 \neq x_0$. Then $f(x_1) = f(g(x_0)) = g(f(x_0)) = g(x_0) = x_1$ and $f$ has two fixed points, which is impossible for non-identical linear functions.

For polynomials I can only give some hints, because I do not have a complete proof.

Let $L = [a, b] \subset \mathbb{R},\ f = f|_L:L \to L, g = g|_L:L \to L$ are polynomial functions, $F$ and $G$ are sets of fixed points of $f$ and $g$ respectively.

1) Suppose, we have $f(x_i) = x_i \in F$ and $g(x_i) = x_j$. Then we apply the same logic, as above: $f(x_j) = f(g(x_i)) = g(f(x_i)) = g(x_i) = x_j$. This means, that $g(x_i) \in F$ for any $x_i \in F$, i.e. $g(F) \subset F$. Similarly, $\ f(G) \subset G$.

2) So if $|F| = 1$ or $|G| = 1$, then we are done. Let $|F| > 1$ and $|G| > 1$.

3) $f$ is a polynomial $\Rightarrow$ it has a finite degree $\Rightarrow$ it has a finite amount of extremums $\Rightarrow$ if $f \not \equiv x$, then $|F|$ is finite. So as $|G|$. We will not consider the case when $f \equiv x$ or $g \equiv x$ because it is trivial.

4) As $g(F) \subset F$ and $|F|$ is finite, then $g$ has a periodic point in it. This means, that $g^n(x_i) = x_i$. So $f$ and $g^n$ have a common fixed point. So as $g$ and $f^m$.

  • I don't see how you conclude in the very beginning that they have a fixed point. For example $x+1$ is linear and has no fixed points; also $2x, x \ge 1$ has no fixed points but it's linear – Ant Jan 16 '17 at 19:49
  • Condition states, that it is a mapping of a line segment into itself. Any continuous mapping of a line segment into itself has at least one fixed point: http://math.stackexchange.com/questions/453908/fixed-point-of-a-continuous-map – Wunsch Punsch Jan 16 '17 at 19:53
  • Ah, I had missed that. Thanks :) – Ant Jan 16 '17 at 20:46
0

I can't comment, so I will post it as answer. Additional condition is that interval must have endpoints, i.e. it must be closed interval. For now I can't prove this exercise.

I suppose that map $x - \frac{1}{x}$ doesn't have fixed point in $(0,1)$ interval, and it commute with identity map, $x$, so it really doesn't work for open intervals.

ckorzhik
  • 101
  • 2
0

This is only a hint, because the question is seriously lacking any sort of clarity!!!

I would have liked to have used the "usual trick" from measure theory. In three steps: first I would have proven it for constant functions, then I would have shown it for simple functions, finally for general functions.

Step 1

Suppose $g(x) = c$ where $c\in \mathbb{R}$ is a constant. And $f$ is a continuous function. From the assumption, that they commute, it follows that:

$$f(c)=f(g(x))=g(f(x))=c$$ $$g(c)=c$$

So they share a fixed point.

Step 2

Clean up your question!

Your statement isn't true in general!

Squirtle
  • 6,698
  • "Your statement isn't true in general!" While I like your Step 1, it would be much more helpful in Step 2 if you would explicitly provide a counterexample. – Caleb Stanford Jan 05 '14 at 07:48