1

I am having a hard time solving this rather interesting question below:

Prove that any function defined on the entire real axis can be written as the sum of two functions whose graphs admit centers of symmetry.

I know we have to set the two functions that are mentioned in the question as an odd function (because it is a function with a graph that is symmetrical with respect to the origin) but I am having a hard time trying to come up with the proof as there are some cases that we need to consider but I don't know what are the appropriate cases we should imply. Thanks!

matcha_
  • 193

1 Answers1

5

A method that almost works is letting $$ \begin{aligned} f_e&= \dfrac{f(x) + f(-x)}{2} \\ f_0&= \dfrac{f(x) - f(-x)}{2} \end{aligned} $$ Then $f(x)= f_e(x) + f_o(x)$ and $f_e,f_o$ are even and odd, respectively. Then $f_o$ has a point of symmetry $(0,0)$ because it is odd. But $f_e$ has an axis of symmetry rather than necessarily a point of symmetry. So we will need something else. I will outline the method suggested by the book, and leave you to fill in the gaps and be sure you understand it.

$1$. Let $g$ be an odd function on $[-1,1]$ with $g(1)=f(1)$. [Be sure you can give at least one concrete example for any give point $(1,f(1))$. It shouldn't be too hard. For example, try with line segments! But it may be beneficial to try to create some 'fancier' examples.]

$2$. Now $g$ is odd (thus far) so it will have $(0,0)$ as a point of symmetry. We need to define another function, which along with $g$, will sum to $f$. An easy way is to define $h(x)= f(x) - g(x)$ on $[-1,1]$. Check that $f= g + h$.

$3$. Now we need to extend $g,h$ so that they are defined everywhere. For $n \geq 1$, define $h(x):= -h(2-x)$ and then extend $g$ via $g(x):= f(x) - h(x)$ for $x \in (2n-1,2n+1]$. Now we need to extend the functions 'on the other side.' Define $g(x):= -g(-x)$ and $h(x):= f(x) - g(x)$ for $x \in [-2n-1,-2n+1)$.

Try this for $n=1,2$ with an explicit function, something simple like $f(x)=x^2$ to understand the process. Then go to the 'generic' $f(x)$. For $n=1,2$, check that $g$ is odd and $h(x)$ has symmetry point $(1,0)$, then check that $f(x)= g(x)+h(x)$. Once you have this done. You should be able to finish the induction argument to show that $f= g+h$ and each have a point of symmetry. Remember, an explicit example always helps!

  • Hi! Thanks for clearing up with clear explanation. However I don't particularly understand why for the domain for function h mentioned in #3 is $2-x$ and for function g we consider the x interval $(2n-1,2n+1]$, could you elaborate more on the choice of interval? Sorry :-/ – matcha_ May 31 '20 at 06:11
  • 2
    @matcha_ I have not worked through everything thoroughly myself. I would strongly suggest going through 1-3 using $f(x)= x^2$ or $f(x)$ linear to get the idea. What they are doing is defining things one side (the 'right' side) and then forcing $g,h$ to have the properties by then mirroring them properly on the 'left' side. The choice of interval seems a bit 'arbitrary.' You could have started with something larger/smaller, say $[-2,2]$, $[-\frac{4}{3},\frac{4}{3}]$, etc. But eventually you are then defining things on an interval of the same 'radius' on one side and then mirroring on the – mathematics2x2life May 31 '20 at 06:15
  • other side and continuing things inductively. The intervals are not so important as that you have some start and then build up the functions inductively from there. The only thing you wanted was the original interval to by symmetric about $0$. The choice was probably because $[-1,1]$ was 'nice' and the other choices because it was simple to write down the process for continuing to build the functions. – mathematics2x2life May 31 '20 at 06:17
  • 1
    Note that once the work here is done, $g$ and $h$ have much simpler definitions on $\mathbb R$: $$g(x) = \begin{cases}-g(-x) & x \le -1 \ x f(1) & -1 \le x \le 1 \text{ (for example)} \ f(x) - h(x) & x \ge 1\end{cases}$$ and $$h(x) = \begin{cases}f(x) - g(x) & x \le 1 \ -h(2-x) & x \ge 1\end{cases}$$ We can check that $g(x) = -g(-x)$, that $h(x) = -h(2-x)$, and that $f(x) = g(x) + h(x)$ for all $x$ very easily from these definitions; the other thing to check is that for any $x$, these definitions give a value for $g(x)$ and $h(x)$ after finitely many "recursive calls". – Misha Lavrov May 31 '20 at 14:08