Find all continuous functions $f$ over real numbers such that $f(x)+x = 2f(2x)$.

We have $f(0) = 0$ and $f(x) = 2f(2x) - x$, but I am not sure how to convert this functional equation into something that is easier to solve. Maybe using induction may work, but I don't see an easy way to induct since we only have one variable.

  • $\begingroup$ $f(x)=\frac{x}{3}$ works. $\endgroup$ – vadim123 Apr 15 '16 at 20:30

Set $g(x)=f(x)-ax$; then $f(x)=g(x)+ax$ and $$ g(x)+ax+x=2(g(2x)+2ax)=2g(2x)+4ax $$ and we can choose $a=1/3$, so the equation becomes $$ g(x)=2g(2x) $$ For $x/2$, we get $$ g(x)=\frac{1}{2}g(x/2)=\frac{1}{4}g(x/4)=\dots=2^{-n}g(2^{-n}x) $$ Since $g$ is continuous, we have $$ g(x)=\lim_{n\to\infty}2^{-n}g(2^{-n}x)=0 $$

  • $\begingroup$ How is it true that $g(x)=\lim_{n\to\infty}2^{-n}g(2^{-n}x)=0$? $\endgroup$ – user19405892 Apr 22 '16 at 19:16
  • $\begingroup$ @user19405892 $\lim_{n\to\infty}2^{-n}=0$ and $g(0)$ is continuous at $0$, so you get $0\cdot g(0)$. $\endgroup$ – egreg Apr 22 '16 at 19:32

We prove that there is only one function $f:\mathbb R\to\mathbb R$ that satisfies $2f(2x)=f(x)+x$ and \begin{align*} \lim_{t\to0}tf(xt)=0 \end{align*} for all $x\in\mathbb R$, namely $f(x)=x/3$. Note that if $f$ is continuous, it satisfies this limit condition (why?)!

By iterating $f(x)=\frac{1}{2}f(x/2)+x/4$ one finds the formula \begin{align*} f(x)=\frac{1}{2^n}f\left(\frac{x}{2^n}\right)+\frac{x}{4^n}\sum_{k=0}^{n-1}4^k,\quad n\geq 1,x\in\mathbb R.\qquad (\star) \end{align*} This can also be shown by induction: For $n=1$, this is just the definition of $f$. Assume that $(\star)$ has been proven for some $n\geq 1$. Then, again by definition, \begin{align*} f\left(\frac{x}{2^n}\right)=\frac{1}{2}f\left(\frac{x}{2^{n+1}}\right)+\frac{x}{2^{n+2}}, \end{align*} and if we put this into our formula (induction!), we obtain \begin{align*} f(x)= \frac{1}{2^n}\left(\frac{1}{2}f\left(\frac{x}{2^{n+1}}\right)+\frac{x}{2^{n+2}}\right)+\frac{x}{4^n}\sum_{k=0}^{n-1}4^k= \frac{1}{2^{n+1}}f\left(\frac{x}{2^{n+1}}\right)+\frac{x}{4^{n+1}}+\frac{x}{4^n}\sum_{k=0}^{n-1}4^k= \frac{1}{2^{n+1}}f\left(\frac{x}{2^{n+1}}\right)+\frac{x}{4^{n+1}}\sum_{k=0}^{n}4^k, \end{align*} which proves $(\star)$.

Now, if we fix $x\in\mathbb R$ and take the limit $n\to\infty$ in $(\star)$ we obtain due to the limit condition that $f(x)=x/3$.


Try with a polynomial of degree $n$ $$\sum_{k=0}^{k=n}a_kx^k$$ We have $$\sum_{k=0}^{k=n}a_kx^k+x=\sum_{k=0}^{k=n}2^{k+1}a_kx^k$$ It follows $$\begin {cases}a_k=2^{k+1}a_k;\space k=2,3,.....,n\\a_1x+x=4a_1x\\a_0=2a_0\end{cases}$$

Hence the only polynomial satisfying the property is $$f(x)=\frac x3$$

Similarly with an homography $$\frac{ax+b}{cx+d}$$ one gets at once $c=0$ and $d=3a$ which comes to $f(x)=\frac x3$

It is doubtful that the property is worth for rational or transcendental function. I stop here.


Exists an elegant and general proof to show that $f(x)=\frac{x}{3}$ is the only one solution.

Elegant proof

Using complex variable, rewrite the functional equation as $2f(2z)-f(z)=z$. Since $2z$ and $z$ are complex-differentiable at $z\in\mathbb{C}$ it follows that $f(z)$ is necessarily complex-differentiable for all $z\in\mathbb{C}$ and hence necessarily continuous. Moreover, since $2$, $2z$, $-1$ and $z$ are finite-grade polynomials then $f(z)$ too and

$$ \deg(f) = \max( \frac{\deg(z) - \deg(2)}{\deg(2z)} , \frac{\deg(z) - \deg(-1)}{\deg(z)} ) = \max(1,1) = 1.$$

As a consequence, the general solution is $f(z) = c_0+zc_1$ where $c_0 = 0$ and $c_1 = \frac{1}{3}$ by replacing in the functional equation. Therefore, $f(z) = \frac{z}{3}$ is the only one solution.

On a general proof

Of course, $f(z)$ is necessarily continuous because it is necessarily complex-differentiable. Note that the fact that being a complex-differentiable function is an even stronger property than being a continuous function. Hence if you have a functional equation

$$A(z) f(U(z)) + B(z) f(V(z)) = P(z),$$

where $A(z)$, $B(z)$, $U(z)$, $V(z)$ and $P(z)$ are complex-differentiable in an open $\Omega\subset\mathbb{C}$, then $f(z)$ is also complex-differentiable for all $z\in\Omega$ (HINT: you can use the Cauchy-Riemann equations to prove this). Therefore, in order to get a general solution you can always compare the Taylor's coefficients of both members of the functional equation.

In particular, if you have that $A(z)$, $B(z)$, $U(z)$, $V(z)$ and $P(z)$ are finite order polynomials, as in your case, then $f(z)$ is complex-differentiable for all $z\in\mathbb{C}$ and its Taylor series exists, is unique and converges for all $z\in\mathbb{C}$. Moreover, for these cases it can easily be shown that $f(z)$ will always be a finite polynomial. In fact,

$$ \deg f = \max( \frac{\deg P - \deg A}{\deg U} , \frac{\deg P - \deg B}{\deg V} ),$$

that you can use to avoid comparing infinite coefficients.


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.