3

$$\int_{-\pi/2}^{\pi/2} \tan x \cos (A \cos x +B \sin x) \, dx$$

Is it possible to calculate this? Both A and B are non-zero and assumed to be real numbers.

I tried Integrate[Tan[x]*Cos[A*Cos[x]+B*Sin[x]],{x,-Pi/2,Pi/2},PrincipalValue->True], but it didn't work.

I would be very grateful if you could share some of the good integration skills, ideas, or any advice.

p.s. I think the integration result should be expressed as a combination of Bessel functions.

1 Answers1

5

One can find a representation in terms of infinite series

Integrate[Tan[x]Series[Cos[b Sin[x]+a Cos[x]],{a,0,5}]//Normal,{x,-Pi/2,Pi/2},PrincipalValue->True]

It yields $$\frac1{\pi}I= -a J_1(b)+\frac{a^3}{6b} J_2(b)-\frac{a^5}{40b^2}J_3(b)+\ldots$$

We can continue to find the following representation

$$ \frac1{\pi}I=\sum_{i=1}^{\infty} \frac{(-1)^{i}a^{2i-1}}{2^{i-1} (i-1)! (2i-1)b^{i-1}} J_{i}(b). $$

So, you were right about the Bessel functions.

Note

I am cautiously optimistic that a closed form may be found due to the existence of recursive relations that can be used to reduce $J_i$ to a sum of just few Bessel functions of lower order. However, Mathematica seem not to recognize these relations. Maybe someone here can help.

yarchik
  • 1,181
  • 1
    Bravo! I was hacking away at this myself and I didn't think to take the series in $a$. – flinty Jun 17 '20 at 21:17
  • 1
    @flinty Just wondering if a closed-form expression exists. Replacing all $J_i$ in the sum with a constant leads to the erf-function. But this does not bring me further. – yarchik Jun 17 '20 at 21:22