I am looking for a closed form of this integral for reals $a,b>0$ and integers $n,m>0$
$I(n,m)=\int_0^{+\infty} e^{-ax^n-\frac{b}{x^m}}$
I read about $I(2,2)$ in the book Irresistible integrals.
However, the neat proof included in the book does not lend itself to generality, or even other particular cases that I know of...
The proof for $I(2,2)$ is essentially subsituting $u=\sqrt{a}x$, then $t=\frac{\sqrt{ab}}{u}$ to get :
$I(2,2)=\int_0^{+\infty}e^{-t^2-\frac{ab}{t^2}}\times\frac{1}{\sqrt{a}}\times\frac{1}{\sqrt{ab}}\times\frac{ab}{t^2}dt=\sqrt{b}\times\int_0^{+\infty}e^{-t^2-\frac{ab}{t^2}}\times \frac{1}{t^2}dt$
So adding this new form with the previous one (after the first subsitution) yields :
$2I(2,2)=\frac{1}{\sqrt{a}}\int_0^{+\infty}e^{-t^2-\frac{ab}{t^2}}\times(\frac{\sqrt{ab}}{t^2}+1)dt=\frac{1}{\sqrt{a}}\int_0^{+\infty}e^{-t^2-\frac{ab}{t^2}}\times d(t-\frac{\sqrt{ab}}{t})$
After completing the square in the exponential, and using $\int_{-\infty}^{+\infty}e^{-x^2}=\sqrt{\pi}$, one eventually gets :
$I(2,2)=\frac{1}{2} \sqrt{\frac{\pi}{a}}e^{-2\sqrt{ab}}$
Trying the same method and expecting to be able to "complete the cube" for the $3,3$ case is already not working for me (I can't get the integral to simplify to something I know already)
I think the probability for a closed form to exist for all $n,m$ is pretty low... but is there a closed form for some particular cases ? Like $n=3, m=3$ ?