I am running MMA to find the solution, but it is taking a long time. Before I wait for long time, Do we have solution for this integral?

$$\int_{-\infty }^{\infty } \frac{ (v-{\alpha_1}) (v-{\alpha_2})}{(v-{\beta_1}) (v-{\beta_2}) (v-{\beta_3})} e^{-v^2/u^2}\, dv$$

where all the roots are complex number.

  • $\begingroup$ Do any of the roots lie on the real line? Also, you can get rid of the $/ u^2$ by rescaling $v$ and all the roots. $\endgroup$ – avs Sep 14 '16 at 0:54

I am not surprised that a CAS has problems.

Thinking about the antiderivative and using partial fraction decomposition we have $$\frac{ (v-{\alpha_1}) (v-{\alpha_2})}{(v-{\beta_1}) (v-{\beta_2}) (v-{\beta_3})}=\sum_{i=1}^3\frac{A_i}{v-{\beta_i}} $$ and so we face the problem of $$\int \frac{e^{-\frac{v^2}{u^2}}}{v-\beta} \, dv$$ for which I suspect that no closed form exists.

However, using a CAS, what it produced is $$\int_{-\infty}^\infty \frac{e^{-\frac{v^2}{u^2}}}{v-\beta} \, dv=e^{-\frac{\beta^2}{u^2}} \left(-\pi \, \text{erfi}\left(\beta \sqrt{\frac{1}{u^2}}\right)+\log \left(-\frac{1}{\beta}\right)+\log (\beta)\right)$$ provided $\Re\left(u^2\right)>0\land \Im(\beta)\neq 0$.

So, it seems that you could obtain your result.


I was too lazy to compute the $A_i$'s which look quite nice $$A_1=\frac{(\alpha_1-\beta_1) (\alpha_2-\beta_1)}{(\beta_1-\beta_2) (\beta_1-\beta_3)} \qquad A_2=\frac{(\alpha_1-\beta_2) (\beta_2-\alpha_2)}{(\beta_1-\beta_2) (\beta_2-\beta_3)} \qquad A_3=\frac{(\alpha_1-\beta_3) (\beta_3-\alpha_2)}{(\beta_1-\beta_3) (\beta_3-\beta_2)} $$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.