Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

A friend of mine gave me the following problem:

Find all functions $f:[0,1]\to[0,1]$, which are one-to-one and onto and satisfy the functional relation $$ f\big(2x-f(x)\big)=x, \tag{1} $$ for all $x\in [0,1]$.

Clearly, the identity function $f(x)=x$ is one such function.

Also, as $f$ is a bijection $f^{-1}$ exists, and by $(1)$ we have $$ f^{-1}(x)=2x-f(x), \tag{2} $$ but I have no idea that how should I continue. It will be great if someone can give me some hints.

Thanks in advance.

share|improve this question
    
The only one as I think is $y=x$ –  Semsem Jan 28 at 18:03
    
@Semsem, so you suggest to prove that $f(x)=x?$ –  Xabat Jan 28 at 18:07
    
let us do it together I have to prove or you have to give one different example –  Semsem Jan 28 at 18:11
    
How do I interpret (1) when $2x-f(x)\notin[0,1]$, or are we assuming that $f$ has the property that $2x-f(x)\in [0,1]$ for all $x$? –  Jason DeVito Jan 28 at 19:03
    
@JasonDeVito, The assumption says $f$ is a function on $[0,1]$, so "Yes", $2x-f(x)\in [0,1]$ for all $x\in [0,1]$. –  Xabat Jan 28 at 19:09

4 Answers 4

up vote 14 down vote accepted

In the formulation of the question the function $f$ is one-to-one and onto. Below two proofs are presented. The first one requires the one-to-one and assumption, while the second one does not.


First answer requiring that $f$ is one-to-one.

Clearly, $f$ is also onto, as an arbitrary $x\in[0,1]$, is the image of $2x-f(x)$, and since $f$ is one-to-one and onto, $f$ possesses an inverse $f^{-1}:[0,1]\to [0,1]$.

The functional relation $f\big(2x-f(x)\big)=x$, implies $$ f(x)-x=x-f^{-1}(x). \tag{1} $$ We shall show that $f(x)=x$. Assume not. Then $$ f(x_0)-x_0\ne 0, $$ for some $x_0\in(0,1)$. Let's assume that $f(x_0)-x_0=a>0$. Then $(1)$ implies that $$ f(x_0)-x_0=x_0-f^{-1}(x_0)=f^{-1}(x_0)-f^{-2}(x_0)=\cdots=f^{-k}(x_0)-f^{-(k+1)}(x_0), $$ for every $k\in\mathbb N$, where $f^{-k}$ is $f^{-1}\circ\cdots\circ f^{-1}$ $k$ times. But this means that $$ f^{-1}(x_0)=x_0-a,\,\,f^{-2}(x_0)=x_0-2a,\ldots,f^{-k}(x_0)=x_0-ka, $$ which means that $\lim_{k\to\infty}f^{-k}(x_0)=-\infty$. A contradiction, $f^{-k}(x_0)\in [0,1]$.

We would reach to a contradiction even if we had assumed that $a<0$.


Second answer not requiring that $f$ is one-to-one.

As $f(2x - f(x)) = x,\,$ for all $x \in \left[ {0,1} \right]$, then $$ 0 \le 2x - f(x) \le 1 \quad \Longrightarrow\quad 2x - 1 \le f\left( x \right) \le 2x,\,\,\, \text{for all $x \in \left[ {0,1} \right]$.} $$ Replacing in the above $x$ with $2x - f(x)$ we obtain \begin{align} 2\left( {2x - f(x)} \right) - 1 \le f\left( {2x - f(x)} \right) \le 2\left( {2x - f(x)} \right) \end{align} or \begin{align} 2\left( {2x - f(x)} \right) - 1 \le x \le 2\left( {2x - f(x)} \right) \end{align} which implies that $$ \frac{3x-1}{2}\le f(x) \le \frac{3x}{2} $$ Repetition of this process produces the following inequalities $$ \frac{(n+1)x-1}{n}\le f(x)\le \frac{(n+1)x}{n},\,\,\, \text{for all $x \in \left[ {0,1} \right]\,\,$ and $\,\,n\in\mathbb N$,} $$ and therefore $$ f(x)=x ,\,\,\, \text{for all $x \in \left[ {0,1} \right]$.} $$

share|improve this answer
    
Can you explain a little more that why $f$ is increasing? –  Xabat Jan 28 at 21:22
3  
@Xabat: If a function $f:[a,b]\to[a,b]$ is continuous and 1-1, then it is necessarily strictly increasing or strictly decreasing. –  Yiorgos S. Smyrlis Jan 28 at 21:25
    
@Xabat: Actually, I am not using this in the proof, after all! –  Yiorgos S. Smyrlis Jan 28 at 21:30
    
@Xabat: See update. –  Yiorgos S. Smyrlis Jan 28 at 21:31
    
I see, that's nice and perfect. Thanks and (+1). What do you think about @woso's claim, that 1-1 condition is not necessary? –  Xabat Jan 28 at 21:41

OK guys, thanks to @woso's suggestion I found an answer. Let me share it with you.


First Answer

By $(2)$ we have $$f(x)-x=x-f^{-1}(x)\tag{3}$$ Now let $x_0\in[0,1]$ be arbitrary. We define the sequence $(x_n)_{n=1}^{\infty}$ as follows $$x_n=f(x_{n-1})\tag{4}$$ for all $n\ge1$. From $(3)$ and $(4)$ we have $$\begin{align*}x_n-x_{n-1}&=x_{n-1}-x_{n-2}\\ &= x_{n-2}-x_{n-3}\\ &= x_{n-3}-x_{n-4}\\&\vdots\\ &= x_{1}-x_{0}\tag{5}\end{align*}$$ Now from $(5)$ we have $$x_n-x_0=\sum_{k=1}^{n}{(x_k-x_{k-1})}=\sum_{k=1}^{n}{(x_1-x_0)}=n(x_1-x_0)\tag{6}$$ Now since $x_n,x_0\in[0,1]$ so by $(6)$ we have $$n|x_1-x_0|=|x_n-x_0|\le 1\tag{7}$$ so by $(7)$ for alll $n\in\Bbb{N}$ we have $|x_1-x_0|\le \frac{1}{n}$, but it's true only if $|x_1-x_0|=0$, so finaly $$f(x_0)=x_1=x_0$$ since $x_0\in[0,1]$ was arbitrary so $f(x)=x$ for all $x\in[0,1]$ and it's the only solution.


Second Answer

We can show that $f(x)=x$ is the only solution even without considering $1$-$1$ and onto conditions.

For all $x\in[0,1]$ consider $g(x)=x-f(x)$. Let $x_0\in[0,1]$ be arbitrary, then $g(x_0)=x_0-f(x_0)$ and by assumption $x_1=x_0+g(x_0)=2x_0-f(x_0)\in[0,1]$, also we can see that $g(x_1)=g(x_0)$, the same process shows that $x_2=x_1+g(x_1)\in[0,1]$ and $g(x_2)=g(x_1)=g(x_0)$,so the sequence $(x_n)$ defined by $x_n=x_{n-1}+g(x_{n-1})$ has these properties:

1. for all $n\ge0$, $x_n\in[0,1]$;

2. for all $n\ge1$, $g(x_n)=g(x_0)$.

so $x_n-x_{n-1}=g(x_{n-1})=g(x_0)$, and $$x_n-x_0=\sum_{k=1}^{n}{(x_k-x_{k-1})}=\sum_{k=1}^{n}{g(x_0)}=ng(x_0) \tag{*}$$

Now since $x_n,x_0\in[0,1]$ so by $(*)$ $$n|g(x_0)|=|x_n-x_0|\le 1\tag{**}$$ so by $(**)$ for all $n\in\Bbb{N}$ we have $|g(x_0)|\le\frac{1}{n}$, but it's true only if $g(x_0)=0$, so finally $$f(x_0)=x_0-g(x_0)=x_0$$ since $x_0\in[0,1]$ was arbitrary so $f(x)=x$ for all $x\in[0,1]$and it's the only solution.


If you find something wrong with it, please add a comment.

share|improve this answer
    
(+1) for me too - We need to make sure that the 1-1 condition is necessary! –  Yiorgos S. Smyrlis Jan 28 at 23:08

I think the answers so far are great, but I thought I'd record a geometrically motivated idea I had and see if people could make it work. The given equality is equivalent to (2) which is equivalent to $$ \frac{f(x) + f^{-1}(x)}2 = x. $$ In other words, the graph of $y=f(x)$, when reflected across $y=x$, has the property that vertical line segments from $(t,f(t))$ to $(t,f^{-1}(t))$ have their midpoints at $(t,t)$ for all $t\in[0,1]$. But that shouldn't be compatible with the fact that all northwest-southeast line segments $(t,f(t))$ and $(f(t),t)$ are also bisected by $y=x$.

If $f(t)>t$ ever, maybe we should choose $t$ such that $f(t)-t$ is maximized, in fact letting $t$ be the smallest such $t$ that maximizes $f(t)-t$. Then we should be able to show that $t-f^{-1}(t)$ is strictly smaller than $f(t)-t$. And since the problem is symmetric in $f$ and $f^{-1}$, the case $f(t)>t$ suffices. (My intuition here has $f$ continuous, but it might be possible to make it work without assuming continuity a priori, by replacing max/min by sup/inf.)

share|improve this answer

Suppose that, for some $x_0$, $$ f(x_0)-x_0=\Delta\tag{1} $$ where $\Delta\ne0$. By the hypothesis, $$ f^{-1}(x_0)=2x_0-f(x_0)=x_0-\Delta\tag{2} $$ Thus, subtracting $x_0$ from $(2)$ gives $$ f^{-1}(x_0)-x_0=-\Delta\tag{3} $$ Considering $(1)$ and $(3)$, we may assume $\Delta\gt0$ by switching the roles of $f$ and $f^{-1}$.


Applying $f$ to $(2)$ yields $$ f(x_0-\Delta)=x_0\tag{4} $$ which gives $$ f(x_0-\Delta)-(x_0-\Delta)=\Delta\tag{5} $$ which is $(1)$ shifted by $\Delta$. Therefore, inductively, we get $$ f(x_0-k\Delta)-(x_0-k\Delta)=\Delta\tag{6} $$ and $$ f^{-1}(x_0-k\Delta)-(x_0-k\Delta)=-\Delta\tag{7} $$ $\hspace{3cm}$enter image description here


Since $\Delta\gt0$, we can find a $k$ so that $x_0-k\Delta\ge0$ and $x_0-(k+1)\Delta\lt0$. For this $k$, $(7)$ says $$ f^{-1}(x_0-k\Delta)=x_0-(k+1)\Delta\lt0\tag{8} $$ but $f^{-1}:[0,1]\mapsto[0,1]$.

Thus, $\Delta=0$, and therefore, $f(x)=x$.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.