I have the following integral as:
$$
\int_{-\infty}^{\infty} d k \frac{\sinh a k}{\sinh (\pi k) \sinh (\frac{\pi k}{2})}e^{ikx}
$$
where $0\leq a \leq \frac{\pi}{2}$.
Can any one help me calculataing this integral? Many thanks!
-
1$\begingroup$ Are you familiar with contour integration? $\endgroup$– Mark ViolaApr 8, 2021 at 16:54
-
$\begingroup$ Could $a \in (-3 \pi/2,3 \pi/2)$? $\endgroup$– Ron GordonApr 8, 2021 at 18:55
-
$\begingroup$ Not exactly, but the approach is the same (integration in the complex plane) math.stackexchange.com/questions/395822/… or math.stackexchange.com/questions/61605/… $\endgroup$– SvyatoslavApr 8, 2021 at 20:53
-
$\begingroup$ @RonGordon Yes it could be. $\endgroup$– HJamesApr 8, 2021 at 21:32
-
$\begingroup$ @MarkViola just a little bit, any hint will be appreciated. $\endgroup$– HJamesApr 11, 2021 at 9:49
1 Answer
$$J=\int_{-\infty}^{\infty} \frac{\sinh a k}{\sinh (\pi k) \sinh (\frac{\pi k}{2})}e^{ikx}dk=\int_{-\infty}^{\infty} \frac{\sinh a k}{2\sinh^2 (\frac{\pi k}{2}) \cosh (\frac{\pi k}{2})}e^{ikx}dk=\int_{-\infty}^{\infty} \frac{\sinh 2a k}{\sinh^2 (\pi k) \cosh (\pi k)}e^{2ikx}dk$$
For the convenience we will evaluate the integral
$$J=\int_{-\infty}^{\infty} \frac{\sinh a k}{\sinh^2 (\pi k) \cosh (\pi k)}e^{ikx}dk=\frac{1}{2}\int_{-\infty}^{\infty} \frac{e^{ikx}}{\sinh^2 (\pi k) \cosh (\pi k)}\Bigl(e^{ak}-e^{-ak}\Bigr)dk=\frac{1}{2}\bigr(I(a)-I(-a)\bigl)$$
We will consider $I(a)$ and $I(-a)$ separately. This is not rigorous, but we know that at $a \in (-3 \pi/2,3 \pi/2)$ integral is convergent, and we don't have singularity at $x\to0$: due to the symmetry only $i\sin(kx)$ survive from $e^{ikx}$, what gives additional power of $x$ at $x\to0$ and makes $x=0$ a removable singular point. $$I(a)=\int_{-\infty}^{\infty} \frac{e^{ikx}}{\sinh^2 (\pi k) \cosh (\pi k)}e^{ak}dk$$
We choose the following contour in the complex plane:
To close the contour we also add two small half-circles ($C_1$ and $C_2$ above the point $x=0$ and below $x=i$ (we go counter clockwise in both cases) and vertical lines $1$ and $2$ at $R\to\pm\infty$
Integral along the closed contour looks $$\oint =I(a)+\int_{C_1}+\int_1+\,\,e^{ia-x}I(a)+\int_{C_2}+\int_2=2\pi i \operatorname{Res}_{x=\frac{i}{2}}\biggl(\frac{e^{ikx}}{\sinh^2 (\pi k) \cosh (\pi k)}e^{ak}\biggr)$$
It can be shown that $\int_1$ and $\int_2\, \to0$ at $R\to\infty$. We have a single pole of first order inside the contour at $x=\frac{i}{2}$, so $$I(a)(1+e^{ia-x})+\int_{C_1}+\int_{C_2}=-2\exp{(\frac{ia-x}{2})}$$
$$\int_{C_1}=\lim_{r\to0}\int_{-\pi}^0\frac{\exp{(are^{i\phi})}\exp{(ix re^{i\phi})}}{\bigl(\pi re^{i\phi}+\frac{(\pi re^{i\phi})^3}{3!}+...\bigr)^2}ire^{i\phi}d\phi=\lim_{r\to0}\int_{-\pi}^0\frac{1+re^{i\phi}(a+ix)+ O(r^2)}{\bigl(\pi re^{i\phi}+\frac{(\pi re^{i\phi})^3}{3!}+...\bigr)^2}ire^{i\phi}d\phi$$
Integral contains the divergent term, but it does not depend on $a$, so it will be canceled when we evaluate $I(a)-I(-a)$ - as it should be. Keeping only converging terms we get: $$\int_{C_1}=-\frac{i}{\pi} (a+ix)$$
In the same fashion we get $$\int_{C_2}=\frac{i}{\pi} (a+ix)e^{ia-x}$$
Taking all together $$I(a)=\frac{i}{\pi}\,(a+ix)\frac{1-e^{ia-x}}{1+e^{ia-x}}-2\frac{e^{\frac{ia-x}{2}}}{1+e^{ia-x}}$$ $$I(a)-I(-a)=\frac{2ia}{\pi}\frac{\sinh x}{\cosh x+\cos a}+\frac{2 ix}{\pi}\frac{\sin a}{\cosh x+\cos a}-4i\frac{\sin\frac{a}{2}\sinh \frac{x}{2}}{\cosh x+\cos a}$$
Taking one half and switching to the initial $a$ and $x$ ($a\to 2a$ and $x\to 2x$) $$J=\frac{2i}{\pi}\,\frac{a\sinh 2x+x\sin 2a-\pi\sin a\sinh x}{\cosh 2x+\cos 2a}$$
Please check the calculations.
-
$\begingroup$ Just corrected several typos (lost $\frac{1}{\pi^2}$ in the first term - $\sinh^2\pi x=(\pi x)^2+...$ at $x\to0\,$) $\endgroup$ Apr 11, 2021 at 17:10
-
$\begingroup$ First of all thank you very much @Svyatoslav for your detailed answer. I guess a minus sign is lost on the way $2\pi i Res_{x=\frac{i}{2}}=2\pi i (\frac{i}{\pi}e^{\frac{1}{2}(i a-x)})=-2e^{\frac{1}{2}(i a-x)}$. I think also that there is an extra pre-factor of 1/2 in the second definition of I(a) (just before the the contour plot ). By divergent term you mean : $\frac{i}{\pi^{2}r e^{i\phi}}$ ? Finally, I am having slightly different pre-factors in the final result. Your answer was very helpful for me ... Thanks a lot! $\endgroup$– HJamesApr 12, 2021 at 14:02
-
$\begingroup$ Thank you. This is good that you checked the result - I did only quick check ($I=0$ at $k=0$ and $a=0$). You are right: the factor $\frac{1}{2}$ just before the contour plot is extra (but, luckily, it is not used in the further calculations :) - I will remove it from the text. Divergent term - also correct. You are also right regarding the residual at $x=\frac{i}{2}$: $\,\sinh^2(\pi i/2)=-\sin^2(\pi/2)$ - I lost this sign. Sorry for these mistakes. Such calculations require attention and thoroughness... $\endgroup$ Apr 12, 2021 at 14:26
-
$\begingroup$ Dear @Svyatoslav this is what I got as a final result : $\frac{i}{\pi} \frac{(2x-\pi)\sin(2a)+2a\sinh(2x)}{\cos(2a)+\cosh(2x)}$ and I hope I am not mistaken. $\endgroup$– HJamesApr 12, 2021 at 15:57
-
$\begingroup$ I'm afraid $I(x\to0)\to0$ - because of the symmetry (in fact we integrate $\sin kx$ from $e^{ikx}=\cos k x+i\sin kx$ in the nominator - integral with $\cos k x$ - wanish). $\endgroup$ Apr 12, 2021 at 16:49