The integral solution lost its symmetry

87 Views Asked by At

I solved this integral using Mathematica:

$H(z,t)=\int_{-\infty}^{\infty}\int_{-\infty}^{\infty}dxdy(x\cdot y)^aE_{-t}(-ix)E_{-t}(-iy)E_{-z}(ix)E_{-z}(iy) $

Where $E_{n}(x)$ is the Exponential Integral function (ExpIntegralE[n,x] in Mathematica)

And the solution was:

$\frac{2\pi^2 i}{(1-a+z+t)^2sin(\pi a)}(\frac{\Gamma(a-z)^2}{\Gamma(-z)^2}+\frac{\Gamma(a-t)^2}{\Gamma(-t)^2}-\frac{2e^{i\pi a}\Gamma(a-z)\Gamma(a-t)}{\Gamma(-z)\Gamma(-t)})$

As you can see, this integral satisfy this symmetry: $H(z,t)=H(t,z)^*$ When I assumed that a,t,z are real (this assumption was included in Mathematica). But weirdly enough the solution does not satisfy this symmetry.

What happen?

Few notes:

the only condition I got was $a>-1$

I know a little bit more about $a$. It has the value of $a=C+z+t$ where $C$ is just a real constant that I can freely fix.

the solution above was not written like this in Mathematica. I used the identity $\Gamma(-x)\Gamma(1+x)sin(\pi x)=-\pi$ to write it in this form.

Moreover,this integral cannot be solved separately, you get different answers (for more info about it you can ask).

1

There are 1 best solutions below

3
On BEST ANSWER

The answer is, in fact, correct. You only need to evaluate the integral $$I(z, t) = \int_0^\infty x^a E_{-z}(i x) E_{-t}(-i x) dx = \\ \int_0^{i \infty} x^a E_{-z}(i x) E_{-t}(-i x) dx = i^{a + 1} \int_0^\infty x^a E_{-z}(-x) E_{-t}(x) dx,$$ which can be done by using the general formula for the integral of a product of two Meijer G-functions: $$I(z, t) = i^{a + 1} \int_0^\infty x^a G_{1, 2}^{2, 0} \left(x \middle| {-t \atop -t - 1, 0} \right) G_{1, 2}^{2, 0} \left(-x \middle| {-z \atop -z - 1, 0} \right) dx = \\ i^{a + 1} G_{3, 3}^{2, 2} \left(-1 \middle| {-a, t + 1 - a, -z \atop 0, -z - 1, t - a} \right),$$ which reduces to a sum of hypergeometric functions with the argument equal to $1$ and then, by Gauss's theorem, to gamma functions.

Then, integrating over each of the four quadrants separately, we obtain $$H(z, t) = I(z, t)^2 + 2 (-1)^a I(z, t) I(t, z) + I(t, z)^2,$$ which simplifies to the solution that you provided. $I(z, t) = I(t, z)^*$, but the relation doesn't hold for $H(z, t)$.

The convergence conditions, determined by the behavior of the exponential integral at zero and at infinity, are $$ -1 < a < 1 \land \max(t, -1) + \max(z, -1) < a - 1.$$