1

I am looking for the Fourier transform of $x \tanh(x)$ in $\mathbb{R}$. The Fourier transform on $\mathbb{R}\setminus\{0\}$ is $$2\sum_{n=0}^{\infty}(2n+1)\exp\left(-(2n+1)\frac{|t|}{2}\right).$$ But there is singularity at $0$. How can I get the Fourier transform of $x \tanh(x)$ in $\mathbb{R}$?

Gary
  • 31,845
RIM
  • 51
  • 1
    I revised your post for spelling, grammar, LaTeX issues, and removing casual words like "hi." All of these things are standard on this site, and you should familiarize yourself with the rules and expectations surrounding asking a good question: https://math.meta.stackexchange.com/questions/9959/how-to-ask-a-good-question

    There are still some things you can do to improve this question further, such as providing the source of the question and your own efforts.

    – A. Thomas Yerger Feb 13 '24 at 19:37
  • Since $x \tanh(x)$ is an even function of $x$ you can just evaluate your series as $$2 \sum\limits_{n=0}^{\infty } (2 n+1) \exp \left(-(2 n+1), \frac{t}{2}\right)=\coth\left(\frac{t}{2}\right), \text{csch}\left(\frac{t}{2}\right)$$ and $\underset{t\to 0}{\text{lim}}\left(\coth\left(\frac{t}{2}\right), \text{csch}\left(\frac{t}{2}\right)\right)=\infty$. – Steven Clark Feb 13 '24 at 21:42
  • @StevenClark That expression is not the Fourier Transform of the Tempered Distribution $x\tanh(x)$. – Mark Viola Feb 15 '24 at 20:25
  • 1
    @rim Please let me know how I can improve my answer. I really want to give you the best answer I can. – Mark Viola Mar 04 '24 at 18:26

1 Answers1

4

The Fourier Transform of the distribution $t\tanh(t)$ is a tempered distribution. We can calculate its Fourier Transform in terms of the sum of the Fourier tranform of the $L^1$ funcion $f(t)=t\tanh(t)-|t|$ and the Fourier Transform of the tempered distribution $g(t)=|t|$. We now proceed.



FOURIER TRANSFORM OF $\displaystyle |t|$:

In THIS ANSWER, I developed the Fourier Transform for $g_a(t)=|t|^a$ for all real values of $a$. In particular, for $a=1$, we have

$$\mathscr{F}\{ g_1\}(\omega)=-\frac2{\omega^2}$$

Note that the object $\mathscr{F}\{ g_1\}(\omega)=-2/\omega^2$ is not a function; it is a Tempered distribution. It does not assign a value for each $\omega$. Rather, it is defined as how it acts on any Schwartz function $\phi$ in the following way. We have

$$\langle \mathscr{F}\{ g_1\},\phi\rangle =\lim_{\delta\to0^+}\int_{|\omega|\ge\delta|} \left(\frac{-2}{\omega^2}\right)\left(\phi(\omega)-\phi(0)\right)\,d\omega $$



FOURIER TRANSFORM OF $\displaystyle \tanh(t)-|t|$:

Note that the function $f(t)=\tanh(t)-|t|$ is in $L^1$. Therefore, its Fourier Transform is given by

$$\begin{align} \mathscr{F}\{f\}(\omega)&=\int_{-\infty}^\infty \left(t\tanh(t)-|t|\right)\,e^{i\omega t}\,dt\\\\ &=2\int_0^\infty t(\tanh(t) - 1)\cos(\omega t)\,dt\\\\ &=4\int_0^\infty t\left(\frac{e^{-2t}}{1+e^{-2t}}\right)\cos(\omega t)\,dt\\\\ &=4\frac{d}{d\omega}\int_0^\infty \left(\frac{e^{-2t}}{1+e^{-2t}}\right)\sin(\omega t)\,dt\\\\ &=4\frac{d}{d\omega}\Im\int_0^\infty \left(\frac{e^{-(2t-i\omega)t}}{1+e^{-2t}}\right)\,dt\\\\ &=4\frac{d}{d\omega}\Im\int_0^\infty e^{-(2t-i\omega)t} \sum_{n=0}^\infty (-1)^ne^{-2nt}\,dt\\\\ &=4\frac{d}{d\omega}\Im\sum_{n=0}^\infty (-1)^n\int_0^\infty e^{-(2t-i\omega)t} e^{-2nt}\,dt\\\\ &=4\frac{d}{d\omega}\Im\sum_{n=0}^\infty \frac{(-1)^n}{2(n+1)-i\omega}\\\\ &=4\frac{d}{d\omega}\sum_{n=0}^\infty \frac{(-1)^n\omega }{4(n+1)^2+\omega^2} \\\\ &=\frac{d}{d\omega}\sum_{n=1}^\infty \frac{(-1)^{n-1}\omega}{n^2+(\omega/2)^2}\\\\ &=\frac{d}{d\omega}\frac{\pi \omega \,\operatorname{csch}(\pi \omega/2)-2}{\omega}\\\\ &=-\frac{\pi^2}2 \frac{\cosh(\pi \omega /2)}{\sinh^2(\pi \omega/2)}+\frac2{\omega^2} \end{align}$$

where we exploited the partial fraction representation of the hyperbolic cosecant function given by

$$\operatorname{csch}(z)=\frac1z +2z\sum_{n=1}^\infty \frac{(-1)^{n-1}}{(n\pi)^2+z^2}$$

The value of the Fourier Transform of $f$ at $\omega=0$ is equal to $-\pi^2/12$.



Putting it all together, we find that the Fourier Transform of $t\tanh(t)$ is the tempered distribution

$$\mathscr{F}\{|f|\}(\omega)=-\frac{\pi^2}2 \frac{\cosh(\pi \omega /2)}{\sinh^2(\pi \omega/2)}$$

where (again), the object $\mathscr{F}\{|f|\}(\omega)$ is defined as a Tempered Distribution. Therefore, for any Schwartz Function $\phi$, we have

$$\langle \mathscr{F}\{|f|\}, \phi\rangle =\lim_{\delta\to 0^+}\int_{|\omega|\ge\delta} \left(-\frac{\pi^2}2 \frac{\cosh(\pi \omega /2)}{\sinh^2(\pi \omega/2)}\right)(\phi(\omega)-\phi(0))\,d\omega$$

Mark Viola
  • 179,405
  • actually i try to understand why there is problem at 0? – RIM Feb 15 '24 at 19:49
  • 1
    The integral $\int_{-\infty} t\tanh(t) e^{i\omega t},dt$ does not exist for any value of $\omega$. However, the Fourier Transform of $t\tanh(t)$ does exist as a Tempered Disrtibution. We note that the Fourier Transform of the function $t\tanh(t)-|t|$ does exist as the integral $$\int_{-\infty} \left(t\tanh(t)-|t|\right) e^{i\omega t},dt$$In fact, when $\omega =0$, its value is $-\pi^2/12$. However, the Fourier Transform of $|t|$ exists as the Tempered Distribution $-2/\omega^2$, which has no value at $\omega=0$. See the answer I referenced for more details. Does that clarify? – Mark Viola Feb 15 '24 at 20:17
  • @ Mark Viola yes thank you – RIM Feb 15 '24 at 22:58
  • @rim You're welcome. My pleasure – Mark Viola Feb 16 '24 at 03:06
  • @Mark Viola just for curiosity the transform of fourier of f in $\mathbb{R}-{0}is given if i wanted to prove it,how can i do that ? – RIM Feb 16 '24 at 11:32
  • 1
    @rim The Fourier Transform of $|t|$ exists only as a tempered distribution. It does not exist as an $L^1$ integration. It is NOT equal to the ordinary function $-2/\omega^2$. It is equal to the Tempered Distribution $-2/\omega^2$. The same applies to $f$ – Mark Viola Feb 16 '24 at 14:10
  • yes as a tempered distribution how can i show that fourier transform of$f(x)= xtanh(x)$ coincides in $\mathbb{R}\setminus{0}$ with the function g(t)=$2\sum_{n=0}^{\infty}(2n+1)\exp\left(-(2n+1)\frac{|t|}{2}\right)$? – RIM Feb 16 '24 at 16:52
  • @rim What do you think is the closed-form for that series when $t\ne0$? – Mark Viola Feb 16 '24 at 17:45
  • its the derivative of $-4\sum_{k=0,k even}^{\infty}\exp(-\frac{(k-1)|t|}{2})$ ? – RIM Feb 16 '24 at 19:24
  • but i dont think that it will provide an additional information –(even if i take inverse fourier transform i will not have the result) – RIM Feb 16 '24 at 19:47
  • @rim What result do you mean? I've tried to explain that the FT here is not defined as an integral. – Mark Viola Feb 16 '24 at 20:56
  • yes i understood that , to answer your question the closest form of that series is $-4\frac{dt}{dt}\sum_{k=0,even}\exp(-\frac{(k-1)|t|}{2}$but can that give me that the series coincides with fourier transform of f? – RIM Feb 16 '24 at 21:05
  • HI Mark Viola ,its given in the exercice(the series) – RIM Mar 05 '24 at 19:40