The following is the final problem from this page:

Find all the functions $f:\mathbb{R}\rightarrow \mathbb{R}$ such that $$f(x+y)=f(x)+f(y) \; \; \; \forall \,x,y\in \mathbb{R}$$ and also (this is the important part) $$f(x^{2019})=f(x)^{2019}\tag{$*$}$$

My idea is to prove that $f(x)=x \; \; \forall x \in \mathbb{R}$, $f(x)=-x \; \; \forall x \in \mathbb{R}$ or $f\equiv 0$.

If we change $2019$ for an even number this is easy because it implies that the image of a positive number is positive and from there $f$ is linear and hence the identity or zero.

If we change $2019$ by $3$ then this is related (although I don't know how to deal with the case $f(1)=0$ or $f(1)=-1$)

But in this case I don't know how to prove any type of regularity from $(*)$ to conclude that $f$ must be linear.

  • 2
    $\begingroup$ You have tagged this as "contest-math". Can you please provide a link to the contest that this problem comes from? $\endgroup$ – Xander Henderson Aug 16 '19 at 19:14
  • $\begingroup$ @XanderHenderson Oh sorry, but I don't know if it comes from a contest. My source is that it is problem 10 of the proposed problems of this blog entry how-did-i-get-here.com/61 $\endgroup$ – A123 Aug 16 '19 at 19:51
  • $\begingroup$ I have edited your question to include the link (in the future, you should provide such references yourself---these provide valuable context). Given that this problem is not from a contest, is there any particular reason that you have chosen to tag it as a contest problem? $\endgroup$ – Xander Henderson Aug 16 '19 at 23:29
  • $\begingroup$ @XanderHenderson Similar problems (e.g. replacing 2019) have appeared as training exercises before. I can't link one off the top of my head, but I think the tag is perfectly fine here, even if it does not come from an explicit contest. $\endgroup$ – user574848 Aug 17 '19 at 0:35

Let $f$ satisfy the premises. Then $f(ax)=af(x)$ for any $x\in\mathbb{R}$ and $a\in\mathbb{Q}$. Now $$f\big((a+x)^{2019}\big)=f(a+x)^{2019}$$ (with both sides expanded using the binomial formula and the above), being a polynomial identity in $a\in\mathbb{Q}$, implies $$f(x^k)=f(1)^{2019-k}f(x)^k\qquad(0\leqslant k\leqslant 2019).$$ Taking $k=2$, we get $f(x^2)=f(1)f(x)^2$. This reduces to the case you have worked out (after replacing $f$ by $-f$ if needed).

| cite | improve this answer | |
  • 1
    $\begingroup$ That's really cool, thanks!! $\endgroup$ – A123 Aug 16 '19 at 21:33
  • $\begingroup$ @metamorphy I still seem to be missing something. Why cannot $f$ satisfy something like e.g., $f(e)=e^2$? [where $e$ is the base of the natural logarithm]. $\endgroup$ – Mike Aug 17 '19 at 1:53
  • 1
    $\begingroup$ @Mike If $f(x+y)=f(x)+f(y)$ and $x\geqslant 0\implies f(x)\geqslant 0$, then $f$ is (necessarily) linear. The Wikipedia article suggests this as an easy consequence of the fact that nonlinear solutions must have graphs (everywhere) dense in $\mathbb{R}^2$. $\endgroup$ – metamorphy Aug 17 '19 at 9:07
  • $\begingroup$ I think I see it now: the condition $f(x^2)=f(1)(f(x)^2$ implies that for all positive $y$, the sign of $f(y)$ has to be the same as the sign of $f(1)$. Which implies that there is at least one quadrant of $\mathbb{R}^2$ not covered by the graph of $f$, which implies (via the reasoning in the link you posted) that $f$ has to be linear. Thank you for explaining @metamorphy ! $\endgroup$ – Mike Aug 17 '19 at 16:01

The first equation is Cauchy's functional equation, and hence the existence of nonlinear solutions depends on the axiom of choice.

If we assume that $f$ is linear, then it must be of the form $f(x)=ax$ for some $a \in \mathbb{R}$. The second equation then says that $\forall x \in \mathbb{R} (ax)^{2019} = ax^{2019}$. Setting $x=1$, it follows that $a$ must be its own 2019th power. The only real numbers that are their own 2019th (or nth for any odd $n>1$) powers are $0$, $1$, and $-1$. Hence, the three linear solutions are $f(x)=0$, $f(x)=x$, and $f(x)=-x$.

If instead of considering solutions over $\mathbb{R}$, we had considered solutions over $\mathbb{C}$, then there would be 2016 more linear solutions, with one corresponding to each non-real 2018th root of unity.

For nonlinear solutions (assuming AC), we don't know.

| cite | improve this answer | |

We can prove that $f$ is continuous at $\mathbb{R}$. Furthermore $f$ has the first derivative on $(0, 0)$. Using these properties leads to show that $f(x) = x$ for every $x \in \mathbb{R}$ or $f \equiv 0$.

| cite | improve this answer | |
  • 3
    $\begingroup$ Can you please show the continuity? $\endgroup$ – Virtuoz Aug 16 '19 at 18:49
  • $\begingroup$ I agree that there is more work to be done. If $f$ were a function on $\mathbb{Q}$ instead of $\mathbb{R}$ then this would be much easier. But e.g., just because one has $f(e)$ does not mean that one has $f(e^2)$. $\endgroup$ – Mike Aug 16 '19 at 19:48
  • $\begingroup$ For example, let $X$ be the subset of $\mathbb{R}$ of the form $\{a +be+ce^2\}$; $a,b,c \in \mathbb{Q}$. Consider $f$ defined as follows: $f(1)=1$; $f(e)=0$; $f(e^2)=100000$. Then this could be extended to an additive real-valued function on $X$. Now there may very well be "something else" happening for the entire real line $\mathbb{R}$ but if so what is it. $\endgroup$ – Mike Aug 16 '19 at 19:51

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.