MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

$a \in \mathbb{R}$

$f:\mathbb{R} \rightarrow \mathbb{R}$

$g:\mathbb{R} \rightarrow \mathbb{R}$

For generic functions $f$ and $g$, how isolate $f(x)$ in the equation below?

$f(x+a)=f(x)+a\times g(x)$

I tried to use Fourier Transform and Inverse Fourier Transform but looks like this don't work very well.

$f(x - a)=$ $e^{-2\pi i a \xi} \hat{f}(\xi)$

$\hat{f}(\xi)=$ $\int_{-\infty}^{\infty}f(x) e^{-2\pi i x\xi}\, dx \quad$ (Fourier Transform)

I tried ZTransform too, but again, didn't worked very well.

share|cite|improve this question
Does your equation have to hold for every $x\in\mathbb{R}$? Most importantly, what do you mean by "isolate"? – Qfwfq Oct 31 '11 at 22:45
You are actually trying to calculate the indefinite sum of terms defined by $f$ (just look at the case $a=1$), so you really should narrow down the class of functions to have summation algorithms. – user11235 Oct 31 '11 at 23:01
@Qfwfq, isolate is find f or $f(x)=\cdots$, where $\cdots$ is something using $g$ and $a$. – GarouDan Oct 31 '11 at 23:02
This should not have the special-functions tag – Yemon Choi Nov 1 '11 at 19:18
I think use DiracDelta is special functions no? And using Fourier Transforms, for example, DiracDelta appears frequently. – GarouDan Nov 1 '11 at 21:10

The answer is not unique - you can add any function of period $a$ to f. This is what the singularities are trying to tell you. In a distributional sense, the Fourier transform of a function of period a is supported exactly on the zeros of $e^{2 \pi i a \xi}-1$.

share|cite|improve this answer
@shrdlu Yes I agree, but if $f$ is unknown, how to find, at least, one solution to $f$? – GarouDan Nov 1 '11 at 14:09

If you're not concerned with continuity, integrability, or any of the niceties of real analysis (and there's nothing in the problem that says you are), then, given $a\ne 0$ and $g$, you can let $f$ be any function on the interval $[0,|a|)$ and simply extend it to all real numbers by repeated applications of the given functional equation (written in the form $f(x) = f(x+a) - ag(x)$ to extend it in the direction opposite to the sign of $a$). If $a=0$, it's clear $f$ can be anything.

Addition: Let me be somewhat more explicit. If $a=1$, we can let $f$ be identically zero on $[0,1)$ and equal to $\sum_{k=1}^{[x]} g(x-k)$ for $x\ge1$, with a similar formula for $x<0$.

share|cite|improve this answer
Didn't understand very well. Can you give an example? For example, using $a=1$ ang $g(x)=\frac{-1}{x(x-1)}$ who mathematica don't solves. – GarouDan Nov 2 '11 at 0:04
@GarouDan: Barry Cipra's point is that not every function is of the nice form that Mathematica would give you - not every function is given by a formula. If you are only interested in functions that are "physically realistic" in some sense then you should probably make this clear in the question – Yemon Choi Nov 2 '11 at 2:38
Great answer! :) – B R Nov 2 '11 at 14:15

Just as you said:$e^{2\pi i a \xi} \hat{f}(\xi)= \hat{f}(\xi) +a \hat{g}(\xi)$. In this way you get $\hat{f}(\xi)$ and you can use it to find $f(x)$.

share|cite|improve this answer
In Mathematica, if we try $a=1$ and $g(x)=\frac{-1}{x(x+1)}$ $f(x)=a (\mathcal{F}_{\xi }^{-1}[\frac{\mathcal{F}_x[G(x)](\xi )}{-1+e^{2 i \pi a \xi }}])(x)$ We have no answer, but, is easy to know the answer. $f(x)=\frac{1}{x}$ because, $f(x+1)=f(x)+1\times g(x) \iff \frac{1}{x+1}-\frac{1}{x}=g(x) \iff \frac{-1}{x(x+1)}=g(x)$ – GarouDan Oct 31 '11 at 23:17
@GarouDan: Shall I write more details? – user16974 Oct 31 '11 at 23:18
I don't know what's wrong with mathematica, my answer is clear. – user16974 Oct 31 '11 at 23:21
I was going to say "unfortunately I don't know a good source for this, as I learnt it from Folland's excellent Lectures on Partial Differential Equations, which is very hard to find nowadays", but I see that the Tata Institute has made it available online (and re-typeset)! – B R Nov 1 '11 at 2:49
I mean, what the poles are expressing is that you only know $f$ up to a periodic function with period $a$. It's up to you to coax a reasonable closed form out of it, and Fourier shouldn't work in general. In fact, nothing should. – Will Sawin Nov 1 '11 at 5:00

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.