Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I have the following system of equations: $$\alpha f_1(x) = \int_\mathbb{R} g(k) h_1(k) e^{\mathrm{i}kx} dk$$ $$\beta f_2(x) = \int_\mathbb{R} g(k) h_2(k) e^{\mathrm{i}kx} dk$$

with known functions $f_i(x)$ and $h_i(x)$ and unknown $\alpha$, $\beta$ and $g(k)$.
Before outlining my ansatz, here directly the question:

Under which conditions on $f_1(x)$ and $f_2(x)$ can we calculate the ratio $\frac\beta\alpha$ uniquely?

My approach

The first step was to integrate one of the equations, $$\alpha \int_\mathbb{R} f_1(x) e^{-\mathrm{i}\lambda x}dx = \int_{\mathbb{R}^2} g(k) h_1(k) e^{\mathrm{i}(k-\lambda)x} dxdk$$ resulting in $$\alpha F_1(k) = 2\pi g(k) h_1(k)$$ because of $$\delta\left(k-\lambda\right) = \frac{1}{2\pi}\int_{-\infty}^{\infty}e^{\mathrm{i}\left(k-\lambda\right)x}dx$$

hence $g(k)$ can be seen as calculated. Plugging this into the second equation, one gets

$$2\pi\frac\beta\alpha f_2(x) = \int_\mathbb{R} F_1(k) \frac{h_2(k)}{h_1(k)} e^{\mathrm{i}kx} dk\, .$$

Now I think the answer to the question relates to a further integration of this equation via some kernel $K(x)$. I would think that this is somehow the dual of the function $f_2(x)$ like in a Hilbert space. But I don't know if this really gives me a well defined unique result, and, how to calculate the dual of $f_2(x)$.

I would be thankful for any hint
Sincerely

Robert

A little excuse
As some might have noticed, I am at the moment occupied with some questions arising in, say, applied functional analysis. As a physicist I am often not very sure about the justification of what I am doing so I hope it is ok to ask these questions here even though the answers might be obvious to most of the people in this community.

share|improve this question
    
What do you mean with the dual of a function? –  Jonas Teuwen Jan 20 '11 at 16:48
    
Hmm, if you just take the inverse Fourier transform of both equations and you divide them then you get $\beta/\alpha = G(x)$ for some $G(x)$, but that $G(x)$ is uniquely determined because the Fourier transform is. Do I miss something? In this case it is sufficient for $h_i g$ to be a tempered distribution. –  Jonas Teuwen Jan 20 '11 at 19:21
    
@JonasT: Thank you for your thoughts! By the dual I mean it in the sense as in the Dirac notation, en.wikipedia.org/wiki/Bra-ket_notation . What you meantion in your second comment is exactly the problem. If $\alpha$ and $\beta$ are just constants (as I might not have pointed out very direct), taking the Fourier transform and dividing, they will be functions... This is one thing I am puzzled with :) Greets –  Robert Filter Jan 21 '11 at 9:12
add comment

1 Answer

up vote 1 down vote accepted

Fold each integral with $\int_{\mathbb{R}} d x\, e^{-i\lambda x}$, $$\begin{eqnarray*} \alpha F_1(k) &=& 2\pi g(k) h_1(k) \\ \beta F_2(k) &=& 2\pi g(k) h_2(k). \end{eqnarray*}$$ Therefore, $$\alpha h_2(k) F_1(k) = \beta h_1(k) F_2(k),$$ or $$\alpha \mathcal{F}\{ \eta_2 * f_1\} = \beta \mathcal{F}\{ \eta_1 * f_2\},$$ where $*$ is the convolution and $\eta_i = \mathcal{F}^{-1}\{h_i\}$. Thus, if $$\eta_2 * f_1 \propto \eta_1 * f_2$$ and $$\eta_1 * f_2 \ne 0$$ the ratio $\beta/\alpha$ can be found uniquely, it is $(\eta_2 * f_1)/(\eta_1 * f_2)$.

As @JonasTeuwen mentioned in the comments, a sufficient condition for this all to work is $f_i$, $h_i$, and $g$ are tempered distributions.

share|improve this answer
    
Thank you oenamen for your answer! –  Robert Filter Apr 28 '12 at 16:51
    
@RobertFilter: Thank you for the question. By the way, how did you come by this problem? –  user26872 Apr 28 '12 at 17:25
    
I wanted to know if a certain approach I was using in some paper (arxiv.org/abs/1107.4934) was unique :) –  Robert Filter Apr 28 '12 at 18:31
add comment

Your Answer

 
discard

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.