1

Question:
Let $f:[0;1] \rightarrow [0;1]$ a strictly increasing fct. Prove that there exists at least one $x \in [0;1]$ s.t. $f(x)=x$.


The prove is false read selected answer for further explanation

My answer:
First by absurd let's suppose that it exist a function $f(x) \neq x \; \forall x \in [0;1]$ which is strictly increasing and define as given in the question.
That means especially that $ 0 \leq f(x) \leq 1 $.

(i) - First let prove, wlog, that such function $f(x)$ can not satisfy $f(x)<x \; \forall x \in [0;1]$. Indeed in such a case in the point $x=0$ the function has to satisfy $f(0)<0$ and that impossible because the question explicitly states that: $f(x) \geq 0 \; \forall \; x \in [0;1]$
The proof is very similar to prove that such function $f(x)$ can not satisfy $f(x)>x \; \forall x \in [0;1]$.

(ii) - Now let write the two set $A_>=\left \{ x \in [0;1] \; | f(x)>x \right \}$ and $A_<=\left \{ x \in [0;1] \; | f(x)<x \right \}$. Obviously $x=0 \in A_>$ and $x=1 \in A_<$.
Because in the other case we will get wlog: $f(1)>1$ while $f(x) \leq 1$ by assumption.

(iii) - Now let choose $x_1 \in A_>$ and $x_2 \in A_<$ s.t. $x_1<x_2$ (exists according to (ii) ) and let define the two following sequence $a_n$ and $b_n$ as follow:

  • $a_1=x_1$ , $a_2=f(x_1)=f(a_1)$ , ... , $a_{n} = f(a_{n-1}) \; \Rightarrow a_{n-1}< a_n = f(a_{n-1})$
    We should remark that $a_n \in A_>$ too as it is defined from $a_{n-1} \in A_>$
    Indeed i.e. $f(a_2)=f(f(x_1)) > f(x_1) = a_2$ as $f(x_1)>x_1$ and the fct is strictly increasing, hence $\forall \; b>a \Rightarrow f(b)>f(a)$ so in particular if $b=f(x_1)>a=x_1$ (which is true as $x_1 \in A_>$.
  • $b_1=x_2$ , $b_2=f(x_2)=f(b_1)$ , ... , $b_{n} = f(b_{n-1}) \; \Rightarrow b_{n-1}< b_n = f(b_{n-1})$
    Similar remark for $b_n$ to those for $a_n$.

Conclusion:
Hence: $a_{n} = f(a_{n-1})$ and $b_{n} = f(b_{n-1}) \Rightarrow a_{n-1}< a_n = f(a_{n-1}) < b_{n-1}< b_n = f(b_{n-1})$ That means that $a_n$ and $b_n$ are two strictly increasing or decreasing bounded sequences. So they have a limit.
So concerning $a_n$ we get on one side: $\lim_{n \to \infty }a_n = l \in (0;1) $ and on the other side: $\lim_{n \to \infty }a_n = \lim_{n \to \infty }f(a_{n-1}) = f(l)$ so by the uniqueness of the limit we get: $f(l)=l$ and that there exists a point $l \in [0;1]$ that is the limit of $a_n$ and $f(a_{n-1})$ and that satisfies (by the uniqueness of the limit of a sequence) $f(l)=l$

Q.E.D.

Is it correct? If anyone has a better solution feel free to post it. Thank for your help.

  • 2
    Hint: $g(x)=f(x)-x$ – oliverjones Dec 10 '22 at 15:46
  • @oliverjones i thought about it but it didn't really help me. Did my solution correct? – X0-user-0X Dec 10 '22 at 15:47
  • 2
    As an aside, you get the same conclusion $f(x)=x$ for at least one $x$ if $f$ is just continuous, with no increasing assumption. You can think of this as Brouwer’s fixed point theorem for $n=1$. – FShrike Dec 10 '22 at 16:04
  • @FShrike Thk for your answer but i don't know the "Brouwer’s fixed point theorem" + is my prove correct? – X0-user-0X Dec 10 '22 at 16:05
  • 3
    Your proof idea is good, though I confess it is hard to read (e.g. your proof that $a_n\in A_{>}$ for all $n$ is quite unclear). There is one problem; in: $$\lim_{n\to\infty}f(a_{n-1})=f(\ell)$$You need to be careful because $f$ is not necessarily continuous! Checking the last limit (if it is even true) requires more careful analysis. – FShrike Dec 10 '22 at 16:11
  • 3
    I think $f$ is not assumed to be continuous, just strictly increasing. Then Tarski's fixed point theorem applies. The shortest proof without that is showing that $x:=\sup {y \in [0,1]:y \le f(y)}$ is a fixed point. – Gerd Dec 10 '22 at 16:11
  • 2
    @oliverjones Their $f$ isn’t necessarily continuous, so perhaps your intended proof involving $g$ doesn’t work – FShrike Dec 10 '22 at 16:14
  • @FShrike no $f$ is not forced to be continuous. but i don't use it int the prove. It's just by the definition of $a_n$ and by the fact that $0 \leq f(x) \leq 1$ that it cames that $a_n$ converges to a point $l$ in (0;1). – X0-user-0X Dec 10 '22 at 16:15
  • @FShrike $f(x)$ is not continuous BUT $a_n=f(a_{n-1})$ is too a point in [0;1] and too an element of the sequence. In fact it's like this that the other elements of the sequence are define. So if all $a_n$ cvge to $l$ is true for all his element, even for the element which can be written: $f(a_{n-1})$ which converges too to $l$. – X0-user-0X Dec 10 '22 at 16:16
  • 1
    See https://math.stackexchange.com/q/2700152/42969 – Martin R Dec 10 '22 at 16:30
  • @MartinR Thk i ll read it. But is my proove correst? I ve spent time to find it and i ll really would like to have a feed back. – X0-user-0X Dec 10 '22 at 16:38
  • @FShrike i ve made the prove more clearer it it can help you – X0-user-0X Dec 10 '22 at 16:41
  • up. Can some one please tell me if my proove is correct? I ve spent time find and writte it, i would really like a feedback – X0-user-0X Dec 10 '22 at 18:44
  • 1

1 Answers1

3

No, your proof is incorrect.

A counterexample

Here is a summary of your proof. Suppose $f(a_1)>a_1$. Let $a_n=f(a_{n-1})$. Then $a_n$ converges to some number $l$ when $n\to\infty$. Then $f(l)=l$.

Here is a counterexample to your proof.

Consider function $f(x)=\begin{cases}\frac14+\frac x2&\text{if }x\in[0;\frac12)\\\frac34+\frac{x-\frac12}2&\text{if }x\in[\frac12;1]\end{cases}$

Since $f(0)=\frac14>0$, let $a_1=0$ in your proof. The sequence of $a_i$'s is $0,\frac14, \frac38, \frac7{16}, \frac{15}{32},\cdots$. Check that $a_n=\frac12-\frac{1}{2^n}$ for all $n$. $l=\lim_{n\to\infty}a_n=\frac12$.

All are fine so far. However, $f(l)=\frac34\not=l$.

The mistake in the proof is, as pointed by FShrike, the following equality does not hold necessarily,

$$ \lim_{n \to \infty }f(a_{n-1}) = f(l)$$ since $f$ may not be continuous at $l$.

A correct proof

I will assume $f$ is weakly increasing only. A strictly increasing function is also weakly increasing.

Let $A_\ge=\{x\in[0;1]\mid f(x)\ge x\}\ni 0$. Since $A_\ge$ is a non-empty set bounded from above, its least upper bound is well-defined. . Let it be $\ell$.

For any $a\in A_\ge$, $\ell\ge a$. So $f(\ell)\ge f(a)\ge a$, i.e., $f(\ell)$ is also an upper bound for $A_\ge$. Hence $f(\ell)\ge\ell$.

Applying $f$ to both sides, we get $f(f(\ell))\ge f(\ell)$, i.e., $f(\ell)\in A_\ge$. Since $\ell$ is an upper bound for $A_\ge$, $\ell\ge f(\ell)$.

So $f(\ell)=\ell$.

Apass.Jack
  • 13,396
  • 1
  • 20
  • 33