Fourier Basis functions question

[mnb96]

Hi,
the continuous Fourier Transform is often defined on a finite interval, usually [-\pi,\pi]:

\hat{f_k} = \frac{1}{2\pi}\int_{-\pi}^{\pi}f(x)e^{-ikx}dx

If I understood correctly, this allows the basis functions to be defined so that they have norm=1, and they form an orthonormal basis for L^2([-\pi,\pi]).
Now, I get confused when one tries to compute the FT of a function f in the whole \mathcal{R} because:

1) The 2-norm of the basis functions goes to +\infty
2) They are not square integrable
3) Should I conclude that the basis-functions are orthogonal but NOT orthonormal?
4) If (1,2,3) are correct, then what is the space spanned by the basis-functions?
5) Is it possible to define an orthonormal basis for L^2(\mathcal{R}) with Fourier basis ?

 

[bartek2009]

I'm not familiar with a 2-norm but the basis is normalised "in my book" like so:

\int_{-\pi}^{\pi}\phi_{k}^{*}\phi_{k}dx = \frac{1}{2\pi}\int_{-\pi}^{\pi}e^{+ikx}e^{-ikx}dx = \frac{1}{2\pi}\int_{-\pi}^{\pi}dx = 1

---

 

[mnb96]

Maybe I used an incorrect term, but indeed, you proved that the squared norm of the basis-functions are always 1.
The important points of my question (1,2,3,4,5) arise when the bounds of the integral are not anymore [-\pi,\pi] but they become (-\infty,+\infty)

Basically, I wanted you to consider:
\int_{-\infty}^{+\infty}\phi_{k}^{*}\phi_{k}dx
and then, answer those questions.

 

[bartek2009]

I see what you mean now. In this case, I think you cannot define the basis.
But maybe there is a way to expand functions in an unnormalised basis.

---