Does there exist a function $f:\Bbb R \to \Bbb R$ such that $f$ is discontinuous at a point and satisfies $f(x+y)=f(x)+f(y)$ for all $x,y$ in $\Bbb R$?


Yes. The reason is that we have a lot of $\mathbb{Q}$-linear independence over $\mathbb{R}$, i.e. $\mathbb{R}$ is an infinite dimensional vector space over $\mathbb{Q}$.

First note we must have $f(0) = 0$. Then we can define $f(1)$ to be anything. Once we have $f(1)$ determined, it's easy to see we must have $f(k) = kf(1)$ for $k \in \mathbb{Z}$ and then $f(q) = qf(1)$ for $q \in \mathbb{Q}$. However, there are no other restrictions this puts on $f$! Note if $f$ were continuous everywhere, then we would know what $f$ is completely, since we know what it is on rationals.

Therefore, we can define $f(\sqrt{2})$ to be whatever we want, and then once again, we can deduce what $f(x)$ is for $x \in \mathbb{Q}[\sqrt{2}]$. And then we can define $f(\alpha)$ to be whatever we want, for some $\alpha \not \in \mathbb{Q}[\sqrt{2}]$. And we can keep doing this on and on, infinitely many times since $\mathbb{R}$ has infinite dimension as a vector space over $\mathbb{Q}$.

It should be clear that the resulting $f$ need not be continuous. To see this directly, note that $\sqrt{2}$ is a limit of rational numbers so once $f$ is defined on $\mathbb{Q}$, if it were continuous, it would be defined at $\sqrt{2}$ already. But we chose $f(\sqrt{2})$ to be whatever we wanted.

The moral of the story is that continuity is more of a metric property of the reals whereas the equation $f(x+y) = f(x)+f(y)$ only uses the algebraic properties of the real line and allows freedom metrically.

  • $\begingroup$ And the graph would be dense as well as far as I know $\endgroup$ – CoffeeCCD Apr 15 '17 at 5:21
  • $\begingroup$ @RonitDebnath It can be dense. It doesn't have to be. Can keep the image of the function inside $(0,1)$ if you want. $\endgroup$ – Stella Biderman Apr 15 '17 at 5:22
  • $\begingroup$ @RonitDebnath, $f$ can be identically $0$ so no. This is the only counter example I believe, so you're basically correct. $\endgroup$ – mathworker21 Apr 15 '17 at 5:23
  • $\begingroup$ @StellaBiderman, I don't know if that's true... I think if some $x$ satisfies $f(x) \not = 0$, then $f(qx) = qf(x)$ which is dense in $\mathbb{R}$ since $f(x) \not = 0$. $\endgroup$ – mathworker21 Apr 15 '17 at 5:23
  • $\begingroup$ Yeah sorry you are right.Maybe upto its range it is graph is dense in R2.I am talking about the horizontal strip. $\endgroup$ – CoffeeCCD Apr 15 '17 at 5:24

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.