13

I need to find all functions $f:\mathbb R \rightarrow \mathbb R$ such that $f(x+y)=f(x)+f(y)$. I know that there are other questions that are asking the same thing, but I'm trying to figure this out by myself as best as possible. Here is how I started out:

Try out some cases:

$x=0:$ $$f(0+y)=f(0)+f(y) \iff f(y)=f(0)+f(y) \iff 0=f(0) $$ The same result is for when $y=0$

$x=-y:$ $$f(-y+y)=f(-y)+f(y) \iff f(0)=f(-y)+f(y) \iff 0=f(-y)+f(y)\iff \quad f(-y)=-f(y)$$ I want to extend the result of setting $x=-y$ to numbers other that $-1$, perhaps all real numbers or all rational numbers. I got a little help from reading other solutions on the next part:

Let $q=1+1+1+...+1$. Then $$f(qx)=f((1+1+...+1)x)=f(x+x+...+x)=f(x)+f(x)+...+f(x)=qf(x)$$ I understood this part, but I don't understand why this helps me find all the functions that satisfy the requirement that $f(x+y)=f(x)+f(y)$, but here is how I went on:

Thus $$f(qx)=qf(x)$$ and it should follow that $$f \bigg (\frac {1}{q} x\bigg)= \frac{1}{q}f(x)$$ where $q\not =0$, then it further follows that $$f \bigg (\frac {p}{q} x\bigg)= \frac{p}{q}f(x)$$ where $\frac{p}{q}$ is rational, and lastly it further follows that $$f (ax)= af(x)$$ where $a$ is real. Thus functions of the form $f(ax)$ where $a$ is real satisfies the requirement of $f(x+y)=f(x)+f(y)$.

I don't know how much of what I did is correct\incorrect, and any help would be greatly appreciated. Also is there any way that I can say that functions of the form $f(ax)$ where $a$ is real are the only functions that satisfy the requirement of $f(x+y)=f(x)+f(y)$? Or do other solutions exist?

Again, thanks a lot for any help! (Hints would be appreciated, I'll really try to understand the hints!)

Asaf Karagila
  • 393,674
user66807
  • 1,313
  • 2
    Think about $f(x)=f(1 x)=f(\frac{q}{q} x)$. In order to extend the result to $a$ real you need to use the continuity of $f$. – Quimey Apr 10 '13 at 02:43
  • @Quimey What do you mean that I need an extra hypothesis like continuity? I think I mentioned that I am looking for the functions f from $\mathbb R \rightarrow \mathbb R$ that are continuous, is that what you are talking about? – user66807 Apr 10 '13 at 02:46
  • Yes. I had misread the question. I've edited the comment. – Quimey Apr 10 '13 at 02:51
  • 3
    @user66807, Instead of "functions of the form $f(ax)$," I think you mean "functions of the form $f(x) = ax$". – Paul Gustafson Apr 10 '13 at 02:52
  • @user66345 You're right, that makes a lot more sense than what I wrote – user66807 Apr 10 '13 at 02:54
  • @Quimey so then should I say "lastly because of the continuity of $f$ it further follows that $f(ax)=af(x)$ with $a$ real. Thus the functions of the form $f(x)=ax$ where $a$ is real satisfies the requirement of $f(x+y)=f(x)+f(y)$? – user66807 Apr 10 '13 at 02:58
  • 6
    Once you have proven that $f(rx)=rf(x)$ for every rational number $r$, you have $f(r)=rf(1)$ for every rational number $r$, and by continuity (and the density of the rationals in the reals) for every real number $r$. Ie any solution $f$ is completely determined by its value at $1$. Moreover, every constant $a$ gives rise to a solution of the form $f(x)=ax$ (trivial to check). Thus we have found the full solution set. – goonfiend Apr 10 '13 at 02:59
  • @SeanGomes I understood the beginning of what you said until you wrote "Moreover, every constant $a$ gives rise to a solution of the form $f(x)=ax$ (trivial to check). Thus we have found the full solution set." Could you explain that a little more please? – user66807 Apr 10 '13 at 03:05
  • So, you will have shown that any continuous function $f : \mathbb{R}\to\mathbb{R}$ satisfying $f(x+y) = f(x)+f(y)$ is of the form $f(x) = ax$ for some $a \in \mathbb{R}$; on the other hand, if $f(x) = ax$ for some $a \in \mathbb{R}$, then it is easy to check that $f$ is continuous and satisfies $f(x+y) = f(x)+f(y)$. – Branimir Ćaćić Apr 10 '13 at 09:03
  • @SeanGomes I think I found where my confusion is. When user66807 said that I mean the solutions are the functions of the form $f(x)=ax$ I know that that is the answer, but I'm not exactly sure how. You wrote that "every constant a gives rise to a solution of the form $f(x)=ax$ (trivial to check)." In order to check this, this is what I came up with: $f(ax)=af(x)=axf(1)$ and I think this is close to proving\checking the solution, but what happens to the $f(1)$? Does it just equal $1$, and if so, how\why? Thanks a lot! – user66807 Apr 10 '13 at 14:28
  • What I mean is that there are two halves to this problem as with most functional equations. Find a form that every solution must have, and verify that all functions of this form are in fact solutions.

    By letting $f$ be an arbitrary solution and deducing that $f(x)=xf(1)$ for all real x, we have deduced that any possible solution must just be a constant ($f(1)$) multiplied by $x$. Ie every solution must be of the form $g(x)=cx$ for some constant $c$. The last (and in this case rather trivial) step is to show that any $c$ will do.

    $g(x+y)=c(x+y)=cx+cy=g(x)+g(y)$.

     – goonfiend Apr 11 '13 at 02:26
  • @SeanGomes Thanks so much for your help, I understand it now. Wish you would have put an actual answer up so that I could select it as accepted. – user66807 Apr 11 '13 at 02:50
  • 1
    Wikipedia: Cauchy functional equation. BTW you mention continuity of $f$ in the title, but I do not see it in the body of your question. – Martin Sleziak May 14 '13 at 13:08
  • @Goonfiend I am still unclear about waht is happeneing to the f(1). My book has this exercise and it is asking me to show that f(ax) = ax. Your last comment only shows that f(x) = cx. – ngc1300 May 20 '22 at 20:51

3 Answers3

7

For $f(\frac1q x)$:

$$f(x) = f(q\cdot\frac1q x) = f(\frac1q x+\ldots+\frac1q x) = f(\frac1q x)+\ldots+f(\frac1q x) = qf(\frac1q x)$$

For $f(\frac pqx)$: Set $y=\frac xq$ to get $$f(\frac pqx) = f(p\frac xq) = f(py) = p\,f(y) = p\,f(\frac1q x) = p\cdot\frac1qf(x)$$

So now you know $f(\alpha x)=\alpha f(x)$ for all $x\in \mathbb Q$. Let $(\alpha_n)$ be a sequence of rational numbers converging to the real number $r$. Since $f$ is continuous, we have $\lim_{n\to\infty}f(\alpha_n x)=f(r x)$. On the other hand, since all $\alpha_n$ are rational, we have $\lim_{n\to\infty}f(\alpha_n x) = \lim_{n\to\infty} \alpha_n f(x) = r\,f(x)$. Since for every $r\in R$ there exists a sequence of rational numbers converging to it, we therefore have $f(rx)=r\,f(x)$ for all $r\in \mathbb R$.

Finally, we can get an explicit form by observing that $f(x) = f(x\cdot 1) = x\,f(1)$. Therefore with $f(1)=c$ arbitrary, we get $$f(x) = cx$$

celtschk
  • 43,384
2

If you do not assume continuity of $f$, then it depends on the Axiom of Choice, stated as Zorn's Lemma. Consider the family $\cal L$ of linearly independent over $\mathbf Z$ subsets of $\mathbf{R}$. Given a chain of such sets,

$$L_1 \subset L_2 \subset \dotsm \subset L_n \subset \dotsm$$

we can prove $L = \bigcup_n L_n \in \cal L$. Suppose $L$ were not linearly independent over $\mathbf Z$. Then, there would be distinct $x_i \in L$ and nonzero integers $a_i$ such that

$$ a_1 x_1 + \dotsm + a_m x_m = 0 $$

But each $x_i$ must be in $L_n$ for some $n$. Let $N$ be the maximum of all such $n$. Then, each $x_i \in L_N$, and so $L_N$ is a $\mathbf Z$-dependent set, contrary to the assumption.

Since every chain has an upper bound, Zorn's Lemma states that there is a maximal member, $M$, of $\cal L$. Every real number $r$ must be a $\mathbf Z$=linear combination of a finite set of elements of $M$; otherwise, $M \cup \{r\}$ would be $\mathbf Z$-independant, contrary to $M$ being maximal. $M$ is a Hamel basis of $\mathbf R$.

And now, any permutation of $M$ extends by linearity to a discontinuous $f$ such that $f(x+y) = f(x) + f(y)$.

Eric Jablow
  • 1,747
1

$$f(x+y)=f(x)+f(x)$$ $$f(0)=0$$ $$f(-x)=-f(x)$$ All these are enough to state a linear function with no constant. Also,$$f'(x)=\dfrac{f(x+h)-f(x)}h$$ if $$f'(x)=\dfrac{f(h)}{h}=c$$

So, The only solution set is $f(x)=cx$

A confirmation too from wolfram. .

ABC
  • 6,034
  • 3
    It is not given that $f$ has a derivative, only that $f$ is continuous. – Aryabhata Apr 10 '13 at 19:05
  • And any solution must make use of continuity in an essential way. Without that assumption, highly irregular solutions exist (assuming choice). – goonfiend Apr 11 '13 at 02:32
  • I hope $f'$ was evaluated because function is continuous.$f(0)=0$ so we can say that $f(h)=0^+$ for $h\to 0^+$ and the $f'$ comes out to be $c$, if it had not been the case then function would have been non-differentiable. – ABC Apr 11 '13 at 02:36