# Find the function $f(x)$ if $f(x+2f(y))=f(x)+y+f(y)$

Let $f:\mathbb{R}\to \mathbb{R}$ such $f(x)$ at $x=0$ continuous, and for any $x,y\in \mathbb{R}$ such $$f(x+2f(y))=f(x)+y+f(y)$$ Find $f(x)$.

Let $x=0,y=0$ then we have $$2f(2f(0))=f(0)++f(0)$$ Let $y=2f(0)$ then we have $$f(x+2f(2f(0))=f(x)+2f(0)+f(2f(0))$$ so we have $$f(x+4f(0))=f(x)+4f(0)$$ on the other hand we have $$f(x+4f(0))=f(x+2f(0))+0+f(0)=2f(0)+f(x)$$ so we have $f(0)=0$ Then I can't deal with this problem

• Maybe I am off base but what happens if you plug in $y=x$? Sep 23, 2016 at 5:29
• @CameronWilliams you get that $x+2f(x)$ is a fixed point. Sep 23, 2016 at 5:30
• Am I missing something or does $f(x)=x$ work? Sep 23, 2016 at 6:21
• @Mastrem It does, but presumably the question is to find all such functions. Sep 23, 2016 at 6:22
• @Mastrem it works, but no one says this equation has only one solution. You have to prove then that it's the only one. Sep 23, 2016 at 6:23

Substitute $x\mapsto y$ to get $$f(y+2f(y))=y+2f(y).$$ Now substitute $y\mapsto y+2f(y)$: $$f(x+2f(y+2f(y)))=f(x)+y+2f(y)+f(y+2f(y)).$$ Thus $$f(x+2y+4f(y))=f(x)+2y+4f(y).$$ Substituting $x\mapsto x+2y+2f(y)$, $$f(x+2y+4f(y))=f(x+2y+2f(y))+y+f(y).$$ Substituting $x\mapsto x+2y$, $$f(x+2y+2f(y))=f(x+2y)+y+f(y).$$ Combining the last 3 equations, $$f(x+2y)=f(x)+2f(y).$$ Thus $$f(x)=f(0+2x/2)=f(0)+2f(x/2)=2f(x/2),$$ so $$f(x+y)=f(x+2y/2)=f(x)+2f(y/2)=f(x)+f(y).$$ Since you are given that $f$ is continuous at $0$, it must be continuous everywhere. Now there is a standard argument to show that $f(x)=f(1)x$ (prove it on rationals and extend by continuity). Finally using the original equation, we see $f(1)=1$ or $-\frac12$, so the solutions are $f(x)=x$ or $f(x)=-x/2$.

Setting $x=y$ gives $f(x+2f(x))=x+2f(x)$, and so $f(y)=y$ whenever $y$ is of the form $x+2f(x)$ for some $x$. Note also that if $f(y)=y$, then for any $x$, $f(x+2y)=f(x)+2y$.

Now consider the function $g(x)=f(x)-x$. From the previous paragraph, we see that if $f(y)=y$ then $g(x+2y)=g(x)$. In particular, every number of the form $2x+4f(x)$ is a period of $g$. Let $P\subseteq\mathbb{R}$ be the group of periods of $g$, i.e. the set of $p$ such that $g(x)=g(x+p)$ for all $x$. There are now two cases.

The first case is that $P$ is not cyclic. Then $P$ is dense in $\mathbb{R}$, and so each coset of $P$ is dense in $\mathbb{R}$, and in particular $0$ is in the closure of each coset. But by definition of $P$, $g$ is constant on each coset of $P$, and so by continuity of $g$ at $0$ this constant value must be equal to $g(0)$. Thus $g(x)=g(0)$ for all $x$, and $g$ is constant everywhere. Thus $f(x)$ has the form $f(x)=x+c$ for some constant $c$. Plugging this into the functional equation gives $$x+2(y+c)+c=x+c+y+y+c,$$ so $c=0$. Thus $f(x)=x$.

The second case is that $P$ is cyclic, generated by some $p\in\mathbb{R}$ (possibly $0$). Then every number of the form $2x+4f(x)$ is an integer multiple of $p$. But $2x+4f(x)$ is continuous at $0$, so this means $2x+4f(x)$ is constant in a neighborhood of $0$. As you've shown, $f(0)=0$, so $2x+4f(x)=0$ for all $x$ in some neighborhood of $0$, so $f(x)=-x/2$ for all $x$ in some neighborhood of $0$.

Now suppose $\epsilon>0$ is such that $f(x)=-x/2$ whenever $|x|\leq\epsilon$. If $0\leq a\leq \epsilon$, the functional equation with $x=a$ and $y=-\epsilon$ then gives $$f(a+\epsilon)=-\frac{a+\epsilon}{2}.$$ That is, $f(x)=-x/2$ is also valid if $\epsilon\leq x\leq 2\epsilon$. Similarly, using $y=\epsilon$ gives that $f(x)=-x/2$ is also valid if $-2\epsilon\leq x\leq -\epsilon$.

That is, if $f(x)=-x/2$ whenever $|x|\leq\epsilon$, $f(x)=-x/2$ whenever $|x|\leq 2\epsilon$ as well. Repeating this argument over and over, we get that $f(x)=-x/2$ for all $x$.

Thus the only possibilities for $f$ are $f(x)=x$ and $f(x)=-x/2$, and you can easily check that both of these work.

• Nice try! Even if using machinery which .. outguns the problem! I downvoted by accident.(and I thought I had already cancelled it. Seems I hadn't!) I will make a trivial edit so that I'll be able to cancel it. P.S.: the part where you are arguing about the cosets of the group of periods is not very clear. Could you expand it a little? (what is the equivalence relation here? why does the continuity of $g$ implies that every coset accumulates at $0$ ?). Sep 23, 2016 at 18:35
• I've clarified that part a bit. Sep 23, 2016 at 19:12
• Thanks for your response. As I've already said: very nice ! +1 Sep 25, 2016 at 19:49