17
$\begingroup$

I would like to find all functions $f:\mathbb{R}\backslash\{0,1\}\rightarrow\mathbb{R}$ such that

$$f(x)+f\left( \frac{1}{1-x}\right)=x.$$

I do not know how to solve the problem. Can someone explain how to solve it?

In one of my attempts I did the following, which is confusing to me: By the substitution $y=1-\frac{1}{x}$ one gets

$f(y)+f\left( \frac{1}{1-y}\right)=\frac{1}{1-y}$. So with $x=y$ it follows that $0=x-\frac{1}{1-x}$. So it would follow that there is no solution. Is that possible or is there a mistake?

Best regards

$\endgroup$
6
  • $\begingroup$ You need to substitute for all x not only one x $\endgroup$ May 4, 2017 at 17:28
  • $\begingroup$ Didn't I do that? $\endgroup$ May 4, 2017 at 17:29
  • $\begingroup$ Your second bracket seems incorrect $\endgroup$ May 4, 2017 at 17:34
  • $\begingroup$ By substituiting $y=1-\frac 1x$ you don't get $$f(y)+f\left( \frac{1}{1-y}\right)=\frac{1}{1-y}$$ $\endgroup$ May 4, 2017 at 17:37
  • 3
    $\begingroup$ Possible duplicate of Find $f$ if $ f(x)+f\left(\frac{1}{1-x}\right)=x $ $\endgroup$
    – Hanno
    May 28, 2017 at 7:32

3 Answers 3

37
$\begingroup$

I would like to shed some light on this issue by taking a more abstract point of view.

In my answer to this recent question : (How to solve an equation of the form $f(x)=f(a)$ for a fixed real a.), I used the following group of functions (with the algebraic meaning of the word "group")

$$\begin{cases}\phi_1(x)=x, & \ \ \ \ \phi_2(x)=1-x, & \ \ \ \ \ \phi_3(x)=\tfrac{1}{x},\\ \phi_4(x)=1-\tfrac{1}{x}, & \ \ \ \ \phi_5(x)=\tfrac{1}{1-x}, & \ \ \ \ \ \phi_6(x)=\tfrac{x}{x-1}.\end{cases}$$

Here also, the presence of this group is natural because it provides all the potentially fruitful changes of variables leading ultimately to the solution.

Let us take the following notation:

$$\psi_k(x):=f(\phi_k(x))$$

Thus, the given functional equation can be written:

$$\tag{1} f(x)+f(\phi_5(x))=x \ \ \ \iff \ \ \ \color{red}{f(x)+\psi_5(x)=x},$$

Substitution $x \to \phi_4(x)$ in (1) gives:

$$\tag{2}f(\phi_4(x))+f(\underbrace{\phi_5(\phi_4(x))}_{\phi_1(x)=x})=\phi_4(x) \ \iff \ \color{red}{\psi_4(x)+f(x)=1-\tfrac{1}{x}},$$

Substitution $x \to \phi_5(x)$ in (1) gives:

$$\tag{3}f(\phi_5(x))+f(\underbrace{\phi_5(\phi_5(x))}_{\phi_4(x)})=\phi_5(x) \ \iff \ \color{red}{\psi_5(x)+\psi_4(x)=\tfrac{1}{1-x}}.$$

It suffices now to make the following combination of equations (1)+(2)-(3) (the parts in red) to obtain:

$$f(x)=\frac12\left(x+1-\frac{1}{x}-\frac{1}{1-x}\right)$$

enter image description here

Remark: the group of functions $\phi_k$ has been recognized by Kummer in the mid-nineteenth century in connection with hypergeometric differential equations. See p. 306 of (http://www.springer.com/la/book/9781461457244), a fascinating book about the rise of complex function theory.

This group has also an interest in projective geometry; for this reason, it is sometimes called the "cross-ratio group". For a modern presentation of the projective invariant called the cross-ratio, take a look for example at (http://www.maths.gla.ac.uk/wws/cabripages/klein/pinvariant.html).

$\endgroup$
5
  • $\begingroup$ (+1), very instructive solution. I didn't know about "cross-ratio group". $\endgroup$
    – Arnaldo
    May 4, 2017 at 21:51
  • $\begingroup$ @Jean Marie How to check obtained solution is only solution satisfying given condition? $\endgroup$
    – mathophile
    Feb 2 at 16:56
  • $\begingroup$ @mathophile I don't understand your question: we have seen that necessarily we are "driven" to this function: it is the only one. See the same process in the solution by Arnaldo. $\endgroup$
    – Jean Marie
    Feb 2 at 17:56
  • $\begingroup$ @JeanMarie Like in this question math.stackexchange.com/questions/4371595/… . It seemed to be there is only one function at start. But later Reinhard pointed out that there are many such function. $\endgroup$
    – mathophile
    Feb 3 at 6:35
  • $\begingroup$ @mathophile I understand your concern. In the reference you give, there is an issue whether the function is assumed continuous or not. Here , my solution (and that of Arnaldo) do not use continuity arguments. $\endgroup$
    – Jean Marie
    Feb 3 at 6:40
27
+50
$\begingroup$

make $x:= \frac{1}{1-x}$ then

$$f\left( \frac{1}{1-x}\right)+f\left( \frac{1}{1-\frac{1}{1-x}}\right)=\frac{1}{1-x}\to f\left( \frac{1}{1-x}\right)+f\left(1- \frac{1}{x}\right)=\frac{1}{1-x}\quad (1)$$

do it again in the last equation:

$$f\left( \frac{1}{1-\frac{1}{1-x}}\right)+f(x)=\frac{1}{1-\frac{1}{1-x}}\to f\left(1- \frac{1}{x}\right)+f(x)=1- \frac{1}{x}\quad (2)$$

now make $(1)-(2)$ and get:

$$f\left( \frac{1}{1-x}\right)-f(x)=\frac{1}{1-x}-1+\frac{1}{x}$$

Subtract the equation is the statement and this last one.

$$2f(x)=x-\frac{1}{1-x}+1-\frac{1}{x}\to f(x)=\frac12\left(x-\frac{1}{1-x}+1-\frac{1}{x}\right)$$

$\endgroup$
4
  • 1
    $\begingroup$ [+1] I am amazed at the way you have found your way with a machete toward the hidden unique solution. My answer, which in fact parallels yours, is guided by a group attached to the functional equation. $\endgroup$
    – Jean Marie
    May 4, 2017 at 21:23
  • $\begingroup$ Thanks for the comments, @JeanMarie . $\endgroup$
    – Arnaldo
    May 4, 2017 at 21:29
  • 1
    $\begingroup$ Wow... what a beautiful and unique solution. This will definitely help me in the future when I am working with functional equations. I think I'm going to give you a +50 bounty for this one. $\endgroup$ May 20, 2017 at 21:40
  • $\begingroup$ @Frpzzd: Thank you for the comment. I am glad to help. $\endgroup$
    – Arnaldo
    May 20, 2017 at 21:44
4
$\begingroup$

By replacing $x$ with $\frac{1}{1-x}$ and $\frac{x-1}{x}$ sequentially, you obtain a system of 3 equations. Then, you can get the solution.

$\endgroup$

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.