1
$\begingroup$

Suppose you have a function that can be described as $$f(s) = \sum_{n=0}^{\infty} a_n e^{f_n s}$$ where each $f_n$ is a complex number. I am looking for a transform $T$ to act on $f$ which produces a function $T(f): \mathbb{C} \rightarrow \mathbb{C}$ such that if $f_n$ was one of the frequencies in our original $f(s)$ then the value of $T(f)$ at the point $f_n$ in the complex plane will be equal to $a_n$

This is basically just a fourier transform that also can detect exponential growth and decay.

The goal is ultimately given a function defined on the real line such as $$f(s) = e^{-s} \sin(s) + 3\pi e^{2s} \cos(s)$$ to be able to produce the list of (frequency: weight) such as for this example $$(f = -1+i : w = \frac{1}{2i}),\\ (f = -1-i : w= -\frac{1}{2i}),\\ (f = 2+i : w= \frac{3\pi}{2}),\\ (f= 2-i : w= \frac{3\pi}{2})$$

This feels like something that probably has been invented and has a name, but i'm not sure where to begin. I found this paper which tries to create a generalized fourier transform that combines the fourier transform and laplace transform into one entity but I wanted to ask this question before I dig through and read all of that.

My concerns:

If such a thing doesn't exist it might because the representation of a function with such a series is not UNIQUE. In which case thats fine too, I'd just be interested in ANY such representation

$\endgroup$
2
  • 4
    $\begingroup$ Not sure, why Laplace does not work? You can express sin/cos functions by combination if complex exponents $\endgroup$ Jun 27 at 21:36
  • 1
    $\begingroup$ Ah wait let me think about that, I suppose you could do a laplace transform. And then for each pole consider the “residue” at the pole to generate the desired result $\endgroup$ Jun 27 at 21:43

1 Answer 1

1
$\begingroup$

So Gideon Genadi Kogan provided the following solution:

We can consider a laplace transform of $f(t)$ that is $\int_{0}^{\infty} e^{-st} f(t) dt $

For each term in the expansion this would be something like

$$ \int_{0}^{\infty} e^{-st} a_n e^{f_n t} dt = a_n \int_{0}^{\infty} e^{(f_n - s)t} = a_n \frac{e^{(f_n - s)t}}{f_n - s} |_{0}^{\infty} $$

For this to converge numerically we need to assume the real part of $f_n - s$ is negative so assuming that or just treating this "formally" we end up with

$$ \frac{a_n}{s - f_n} $$

So we have

$$ \mathcal{L} \left\lbrace \sum_{n=0}^{\infty} a_n e^{f_n x} \right \rbrace = \sum_{n=0}^{\infty} \left[ \frac{a_n}{s - f_n} \right] $$

From here to finally wrap up our problem we use the residue theorem. We can define the operator $O[f]$ at a point p as the following

$$ O[f](p) = \lim_{r \rightarrow 0} \frac{1}{2\pi i} \oint_{\text{circle of radius} \ r \ \text{centered at p} } \left[ f(t) \right] dt $$

It then follows that

$$ O \circ \mathcal{L} $$

Is the transform we are seeking.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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