Let $p\in\mathbb{N}$. Let $f:I\to\mathbb{R}$ differentiable in the closed interval $I$ (bounded or not), with $f(I) \subset I$, and let $g = f\circ f\circ \cdots \circ f = f^p$, where $\circ$ means composition. If $\lvert g'(x) \rvert \leq c < 1, \forall x\in I$ where $c$ is a constant, show that there is only one $a\in I$ such that $f(a) = a$.

Some thoughts

I proved it for $p=1$, i.e. $g = f$ and $\lvert g'(x) \rvert = \lvert f'(x) \rvert \leq c < 1$. Notice that $I$ can be unbounded so the existence of the fixed point is not trivial.

For the task, I use two results:

  1. Let $f:I\to\mathbb{R}$ differentiable in the interval $I$. The function $f$ satisfies $\lvert f(x) - f(y) \rvert \leq c\lvert x - y \rvert$ for all $x,y\in I$ if and only if $\lvert f'(x) \rvert \leq c$ for all $x\in I$.

  2. Let $0\leq c < 1$. If the sequence $(x_n)$ is such that $\lvert x_{n+2} - x_{n+1}\rvert \leq c\lvert x_{n+1} - x_n\rvert $ for all $n\in \mathbb{N}$, then the sequence $(x_n)$ converges.

Now take any point $x_0\in I$ and build the following sequence: $x_1 = g(x_0), x_2 = g(x_1), \cdots, x_n = g(x_{n-1}), \cdots$. Since $\lvert g'(x) \rvert \leq c$, from result 1 we have $\lvert g(x_{n+1}) - g(x_{n}) \rvert \leq c\lvert x_{n+1} - x_{n} \rvert$, then $\lvert x_{n+2} - x_{n+1} \rvert \leq c\lvert x_{n+1} - x_{n} \rvert$. Since $0\leq c < 1$, from result 2 we have $(x_n)\to a$. Since $I$ is closed, $a\in I$. Finally, since $x_n = g(x_{n-1})$, this means $g(a) = a$. The uniqueness comes from supposing another $b$ and applying IVT to get contradiction.

I am having trouble trying to prove for $p > 1$.

Update: I marked Matthew's suggestion because it contained the key idea for the solution. The remaining bits for the solution are in the comments to his suggestion.

  • $\begingroup$ You asked the same question 13 hours ago, and did a substantial edit one hour ago. You should've flagged that one for reopening instead of making a new one that's basically a copy. Actually, as of now it's officially reopened, but it's deleted instead. $\endgroup$ – Arthur Jul 20 '16 at 13:29
  • $\begingroup$ I'm sorry, I didn't know that. I deleted it instead. $\endgroup$ – Michael Jul 20 '16 at 13:30
  • $\begingroup$ Fair enough, you seem fairly new here. Just keep it in mind for next time. $\endgroup$ – Arthur Jul 20 '16 at 13:31

Assuming your proof of the $p=1$ case is valid, you can apply it to $g=f^p$. You know $g$ has a unique fixed point $a$. You can show that $f(a)$ is a fixed point of $g$, too. But this means $a=f(a)$.

  • $\begingroup$ Thanks for the idea. This was the key. $\endgroup$ – Michael Jul 22 '16 at 3:04
  • 2
    $\begingroup$ Extending from the original post, let now $g = f\circ f\circ \cdots \circ f = f^p$. We know there is a unique $a$ such that $g(a)=a$. We want to show that such $a$ is also the fixed point of $f$. Now, $g(a) = f^{p}(a) = a$, then $g(f(a)) = f^{p+1}(a) = f(a)$, therefore $f(a)$ is also fixed point of $g$, hence $a=f(a)$ by uniqueness. $\endgroup$ – Michael Jul 22 '16 at 3:10

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.