I'm trying to solve this integral equation, but I can't deal with the sine! $$\int_0^\infty f(y)\sin(xy)dy=e^{-x} , x>0$$

How do I proceed?

I tried using the fourier transform but then I can't deal with the fourier transform of the sine.

The definition of Fourier transform I'm using is:

$$f(\xi)=\frac{1}{\sqrt{2\pi}}\int_{-\infty}^\infty f(x)e^{-i\xi x}dx$$

EDIT: I think I know what my problem was, I was doing the fourier transform of the sine individualy instead of doing the fourier transform of $f(y)\times sin(xy)$. Doing this and substituing the sine for its exponential formula will give the answer right?

  • 2
    $\begingroup$ It may help from Euler's formula that $$\sin a = \frac{e^{ia} - e^{-ia}}{2i}$$ $\endgroup$
    – gt6989b
    Jan 3 '20 at 17:03

Note that as $x \to 0+$, $\sin(xy) \to 0$ so if $f$ is integrable, you should have $\int_0^\infty f(y) \sin(xy)\; dy \to 0$ by Dominated Convergence. But of course $e^{-x}$ goes to $1$, not $0$. So there is certainly no integrable solution.

I suppose it's possible that there is a non-integrable solution, but that makes convergence of your integral problematic. Do you have any reason to expect a solution to exist?

  • $\begingroup$ Yes, this was a question on a last year exam! I'll check if I saw it right $\endgroup$ Jan 4 '20 at 2:10
  • $\begingroup$ Yes I saw it right...maybe my professor got something wrong? The question says explicitly to calculate f by solving the integral equation.... $\endgroup$ Jan 4 '20 at 2:12
  • $\begingroup$ The question also says to prove that f belongs to L2. I don't know if that helps. $\endgroup$ Jan 4 '20 at 2:15
  • $\begingroup$ Maybe the exam paper is wrong, if the sine changes to sin(x-y) does it have a solution? I'll try this tomorrow $\endgroup$ Jan 4 '20 at 2:31

The solution is $$f(y)=\frac{2}{\pi}\frac{y}{1+y^2}$$ I will add the derivation soon.


Let $$g(y)= \begin{cases} f(y)&&y\ge0 \\ 0&&y<0 \end{cases} $$ Then, with Euler's formula, we can rephrase the problem as $$\int^\infty_{-\infty}g(y)\left(\frac{e^{ixy}-e^{-ixy}}{2i}\right)dy=e^{-x}$$

Equivalently, $$\mathcal F\{g(y)\}(-x)-\mathcal F\{g(y)\}(x)=i\sqrt{\frac2\pi}e^{-x}$$ $$G(-x)=G(x)+i\sqrt{\frac2\pi}e^{-x}\qquad{x>0}$$

This functional equation does not tell much, as the solution is not unique. The best we can do is defining $$G(x)= \begin{cases} \varphi (x) && x>0 \\ \varphi(-x)+i\sqrt{\frac2\pi}e^{x} && x<0 \end{cases} $$ for some $\varphi (x):\mathbb R^+$ with sufficiently nice properties.

Then, $$ \begin{align} g(y)&=\frac{1}{\sqrt{2\pi}}\int^\infty_{-\infty}G(x)e^{ixy}dx \\ \sqrt{2\pi}g(y)&=\int^\infty_{0}G(x)e^{ixy}dx+\int^{\infty}_{0}G(-x)e^{-ixy}dx \\ &=\int^\infty_{0}\varphi(x)e^{ixy}dx+\int^{\infty}_{0}\left(\varphi(x)+i\sqrt{\frac2\pi}e^{-x}\right)e^{-ixy}dx \\ &=2\int^\infty_{0}\varphi(x)\cos(xy)dx+i\sqrt{\frac2\pi}\int^{\infty}_{0}e^{-x}e^{-ixy}dx \\ &=2\int^\infty_{0}\varphi(x)\cos(xy)dx+i\sqrt{\frac2\pi}\frac1{1+iy} \\ g(x)&=\sqrt{\frac2\pi}\int^\infty_{0}\varphi(x)\cos(xy)dx+\frac i\pi\frac1{1+iy} \\ \end{align} $$

Let $a>0$. By assumption, $g(-a)=0$. Therefore, $$\sqrt{\frac2\pi}\int^\infty_{0}\varphi(x)\cos(ay)dx+\frac i\pi\frac1{1-ia}=0$$

Then, $$\begin{align} g(a)&=\sqrt{\frac2\pi}\int^\infty_{0}\varphi(x)\cos(ay)dx+\frac i\pi\frac1{1+ia} \\ &=-\frac i\pi\frac1{1-ia}+\frac i\pi\frac1{1+ia} \\ &=\frac2\pi\frac{a}{1+a^2} \end{align} $$

Hence, $$f(y)=\frac{2}{\pi}\frac{y}{1+y^2}$$


(The calculations below are a bit sloppy as I assumed that the integral and differentiation can be interchanged.)

$$\begin{align} L.H.S. &=\frac2\pi \int^\infty_0\frac{y}{1+y^2}\sin(xy)dy \\ &=-\frac2\pi\frac{\partial}{\partial x}\int^\infty_0\frac{\cos(xy)}{1+y^2}dy \\ &=-\frac1\pi\frac{\partial}{\partial x}\int^\infty_{-\infty}\frac{\cos(xy)}{1+y^2}dy \\ &=-\frac1\pi\frac{\partial}{\partial x}\pi e^{-x} \\ &=e^{-x}\\ &=R.H.S. \end{align} $$

Here we utilized the well-known integral identity $$\int^\infty_{-\infty}\frac{\cos(ax)}{1+x^2}dx=\pi e^{-|a|}$$ For its proof, see the accepted answer here.

  • $\begingroup$ Under the assumption that $x\geq 0$ that is =) $\endgroup$ Jan 4 '20 at 13:21

If $f_o$ is the odd extension of $f$ to $\mathbb{R}$, then $$ e^{-x} = \int_0^{\infty}f(y)\sin(xy)dy \\ = \frac{1}{2}\int_{-\infty}^{\infty}f_o(y)\sin(xy)dy \\ = \frac{1}{2i}\int_{-\infty}^{\infty}f_o(y)e^{ixy}dy, $$ which works out by residues if you close in the upper half plane and you use $$ f_o(y)=\frac{1}{\pi}\left(\frac{1}{y-i}+\frac{1}{y+i}\right) = \frac{2}{\pi}\frac{y}{y^2+1}. $$ As anticipated by Robert Israel, $f(y)$ is not integrable on $(0,\infty)$. The solution is valid when interpreted as an improper integral at $\infty$, which fits nicely with Complex Analysis and with Fourier Analysis. Any real function that is locally integrable on $[0,\infty)$ and is eventually monotonically increasing or decreasing to $0$ has a Fourier transform because the sine and cosine transforms can be interpreted as sums of integrals over a finite interval plus an alternating series with a general term that tends to $0$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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