7

I would like to evaluate analytically the improper integral $$ f(z,r,\rho,\nu) = \int_0^\infty \sin(k z)\, K_{i\nu}(k \rho)\, K_{i\nu}(k r)\, dk \, , $$ where $K_{i\nu}$ is the modified Bessel function of the second kind with imaginary order. Here, $\nu \ge 0$. This integral arises in solving a boundary value problem using the Kontorovich–Lebedev transform.

When cosine replaces sine, the integral is known and can be found in the classical reference by Gradshteyn and Ryzhik, Eq. ET I 50(51) (6.672 page 719). It is given by $$ \int_0^\infty \cos(k z)\, K_{i\nu}(k \rho)\, K_{i\nu}(k r)\, dk = \frac{\pi^2}{4} \frac{\operatorname{sech}(\pi\nu)}{\sqrt{\rho r}} \, P_{i\nu-\frac{1}{2}} (\xi) \, , \tag{A} $$ with $$ \xi = \frac{\rho^2+r^2+z^2}{2\rho r} \, . $$

Here are the steps I followed. I used the integral representation of the modified Bessel function for real argument $x>0$, namely $$ K_{i\nu}(x) = \int_0^\infty e^{-x\cosh u}\cos(\nu u)\,du \, , $$ and substituted this representation for both Bessel functions into $I(z,r,\rho,\nu)$. By noting that $$ \int_0^\infty \sin(kz)\, e^{-a k}\, dk = \frac{z}{a^2+z^2}, \qquad a>0 \, , $$ the resulting double integral reads $$ f(z,r,\rho,\nu) = z \int_0^\infty \int_0^\infty \frac{\cos(\nu u)\cos(\nu t)}{(\rho\cosh u + r\cosh t)^2 + z^2}\,du\,dt. $$

Any suggestions on how to proceed further? I’d appreciate your help, thanks.

Jack D'Aurizio
  • 372,937
  • 42
  • 421
  • 888
Eulerian
  • 274
  • 1
    (+1) Nice question. Have you already extended the last integral to $\mathbb{R}^2$ then set $x=\frac{u+t}{\sqrt{2}}$ and $y=\frac{u-t}{\sqrt{2}}$ in order to simplify the numerator? – Jack D'Aurizio Aug 28 '25 at 13:43
  • @JackD'Aurizio Thanks for the comment. I will do that and update the question. – Eulerian Aug 28 '25 at 13:45
  • @JackD'Aurizio What one gets is some more complex integral $\int_0^\infty dx \int_{-x}^x dy$. Not sure how this simplifies the numerator – Eulerian Aug 28 '25 at 13:53
  • 2
    $$\frac{\pi}{2\sqrt{\rho r}}P_{-1/2+i\nu}\Bigl(\frac{\rho^2+r^2+z^2}{2\rho r}\Bigr)$$ – Martin.s Aug 28 '25 at 14:22
  • 1
    @Martin.s No, it is not. It looks like the result with cosine not sine, for which we have $\pi^2 \text{sech}(\pi\nu)/4$ instead of $\pi/2$ in your expression. I have made some numerical checks. – Eulerian Aug 28 '25 at 15:17
  • 1
    I guess it is enough to differentiate both sides of $(A)$ with respect to $z$, am I wrong? – Jack D'Aurizio Aug 28 '25 at 16:22
  • @JackD'Aurizio No, you get $-k\sin(kz)$ while i need $\sin(kz)$. – Eulerian Aug 28 '25 at 16:25
  • 1
    As a reference: https://apps.dtic.mil/sti/tr/pdf/AD0267210.pdf , but the table only contains $k\sin(kz)$. Maybe there is some parity obstruction. – Jack D'Aurizio Aug 28 '25 at 18:40
  • 2
    Related: https://math.stackexchange.com/questions/4936195/evaluting-int-0-infty-k-nuax-k-nubx-sincx-dx-missing-from-erdelyi – Jack D'Aurizio Aug 28 '25 at 18:55
  • @JackD'Aurizio Thus the integral is unlikely to be evaluated analytically. – Eulerian Aug 29 '25 at 07:28
  • 1
    $$\int_0^{\infty } \sin (z x) K_{i v}(a x) K_{i v}(b x) , dx=\int_0^{\infty } \frac{1}{4} \sqrt{\pi } G_{0,2}^{1,0}\left(\frac{x z}{2},\frac{1}{2}| \begin{array}{c} \frac{1}{2},0 \ \end{array} \right) G_{0,2}^{2,0}\left(\frac{a x}{2},\frac{1}{2}| \begin{array}{c} \frac{i v}{2},-\frac{1}{2} (i v) \ \end{array} \right) G_{0,2}^{2,0}\left(\frac{b x}{2},\frac{1}{2}| \begin{array}{c} \frac{i v}{2},-\frac{1}{2} (i v) \ \end{array} \right) , dx=\text{2 variable FoxH function}$$ – Mariusz Iwaniuk Aug 29 '25 at 09:37

0 Answers0