6

Suppose $f:(a,b)\to(a,b)$ satisfies $|f(x)-f(y)|<|x-y|,\forall\ x\neq y\in(a,b)$, $x_1\in(a,b),x_{n+1}=f(x_n)$, prove that $\{x_n\}$ is convergent. Here $-\infty<a<b<\infty$.

Any comments and other methods will welcome.Two goals for this post: (1) check my proof and give comments (2) better alternative solutions.
This question comes from the 2025 Graduate Entrance Examination for the Department of Mathematics of Nankai University.

The following is my proof:

(1) It is easy to see that $f$ is uniformly continuous on $(a,b)$, so the both the right limit $f(a+0)=\lim\limits_{x\to a^+}f(x)$ and the left limit $f(b-0)=\lim\limits_{x\to b^-}f(x)$ exist and $$f(a+0)\geq a,\quad f(b-0)\leq b.$$ Let $f(a)=f(a+0),f(b)=f(b-0)$, then $f:[a,b]\to[a,b]$ is continuous on $[a,b]$.

(2) For $x<y\in(a,b)$, we know that $$f(y)-f(x)\leq|f(y)-f(x)|<y-x\implies f(y)-y<f(x)-x;$$ $$f(x)-f(y)\leq|f(y)-f(x)|<y-x\implies f(x)+x<f(y)+y,$$ we get that the function $f(x)-x$ is strictly decreasing on $(a,b)$ and the function $f(x)+x$ is strictly increasing on $(a,b)$. Due to the continuity of $f(x)-x$ and $f(x)+x$ on $[a,b]$, we know that the function $f(x)-x$ is strictly decreasing on $[a,b]$ and the function $f(x)+x$ is strictly increasing on $[a,b]$.

(3) Claim: $$|f(x)-f(y)|<|x-y|,\quad\forall\ x\neq y\in[a,b].$$ If $|f(a)-f(x)|=|a-x|$, then $f(a)-f(x)=a-x$ or $f(a)-f(x)=x-a$. In both cases we can get $x=a$ by the facts of $(2)$. Similarly, if $|f(b)-f(x)|=|b-x|$, we can get $x=b$.

(4) By $(1),(3)$, $f$ has a unique fixed point $x_0\in[a,b]$(for proof, we can refer unique fixed point or note that: $f(x)-x$ satisfies $f(a)-a\geq0,f(b)-b\leq0$ and is is strictly decreasing on $[a,b]$, those imply that $f(x)-x$ has a unique zero, so $f$ has a unique fixed point. ). we will prove that the sequence $x_1\in(a,b),x_{n+1}=f(x_n)$ converges to $x_0$. The following idea is from @copper.hat's Proof. It is clear that $|x_{n+1}-x_0|\leq|x_n-x_0|$ for all $n$. Let $\delta=\inf\limits_n|x_n-x_0|$. If $\delta=0$, we are finished.

Suppose $\delta>0$. Let $$\lambda=\max_{\substack{|x-x_0|\ge\delta \\x\in[a,b]}}\frac{|f(x)-x_0|}{|x-x_0|}.$$ (the $\max$ exists because it is the $\sup$ of a continuous function over a compact domain), and note that $\lambda<1$ due to the claim of $(3)$. Since $|x_n-x_0|\ge\delta$ for all $n$, we have $|x_{n+1}-x_0|\le\lambda|x_n-x_0|$ which implies $\delta=0$, this contradicts $\delta>0$. Hence $\delta=0$ and $\lim\limits_{n\to\infty}x_n=x_0$.

Riemann
  • 11,801
  • I think you need the interval $(a,b)$ be closed and your constant of contraction be least than 1 – Ataulfo Dec 25 '24 at 14:13
  • @Piquito There is no doubt that constant of contraction be least than $1$ makes the my result right. In my condition , I gave a proof. If my proof has mistake, I hope you can point out, thanks ! – Riemann Dec 25 '24 at 14:17
  • This is true without a constant of contraction in a compact metric space (see Exercise 56 in Rouvière's "Petit guide de calcul différentiel à l'usage de la licence et de l'agrégation"), I'll write the corresponding answer – reded Dec 25 '24 at 14:21
  • @ reded That will be nice! Thank you! Can you check my proof? – Riemann Dec 25 '24 at 14:24
  • Yes I will too. – reded Dec 25 '24 at 14:25
  • @Riemann: I remember a counterexample when $k=1$ but that was longtime ago. Maybe it was for the non-unicity of fixed point. I am sorry if your proof is correct and I hope you are right. In this moment i am make something other thing and i do not read your proof. Regards. Merry Christmas (sorry for bad English) – Ataulfo Dec 25 '24 at 14:26
  • I have no knowledge of an English version sorry. I'm writing an answer anyway – reded Dec 25 '24 at 14:33
  • Are you asking for an alternative solution or a solution verification? – Sine of the Time Dec 25 '24 at 15:02
  • @Riemann I see. Note that "For posts looking for feedback or verification of a proposed solution. "Is this proof correct?" is too broad or missing context. [...]" – Sine of the Time Dec 25 '24 at 15:10
  • @ Sine of the Time Thank you for your comment, I will try my best to post a question! – Riemann Dec 25 '24 at 15:12

1 Answers1

1

The wording is not precise, and I am worried that you may not understand the issues. In particular, in the case $a=-\infty$ and $b= \infty,$ there is no (finite) fixpoint for the example

$$ f(x) = \frac{x + \sqrt{1 + x^2}}{2} $$

Note that we always have $f(x) > x$ as well as $0 < f'(x) < 1.$

enter image description here

Will Jagy
  • 146,052
  • In my post, I listed $-\infty<a<b<\infty$. Thank you for your concern! – Riemann Dec 26 '24 at 00:12
  • 1
    @Riemann The application of this to a finite interval is simply this: let $f(x) = \frac{x+b}{2}.$ Then there is a fixpoint on the entire real line, but not in your interval, as the fixpoint is simply $b.$ The sequence does not converge in the given interval. – Will Jagy Dec 26 '24 at 00:41
  • We only need to prove that the sequence ${x_n}$ is convergent, not need the limit point in $(a,b)$. – Riemann Dec 26 '24 at 00:53