So I'm taking a complex analysis course on EdX, and this is the first problem on the first problem set. It's far overdue, so I'm not looking to submit it. I just want to know what are the inner workings of how one gets from the question to the answer.

The integral is: $$K(\lambda) = \int_{-\infty}^{\infty}e^{-\lambda t^2}\frac{dt}{t-ib}$$

The equation that it apparently satisfies is (and I don't know this to start with, this is just from clicking on Show Answer; there are blanks in place of the $'$, $b^2$, and $\displaystyle \frac {b}{\sqrt{\lambda}}$):

$$K' - b^2K=-i\sqrt{\pi}\frac{b}{\sqrt{\lambda}}$$

This is a differentiation w.r.t. parameter problem, so $\displaystyle K'\equiv\frac{dK}{d\lambda}$.

I've tried expanding the $\displaystyle \frac{1}{t-ib}$ term, but that doesn't seem to lead anywhere. I've tried differentiating the $e^{-\lambda t^2}$ term w.r.t. parameter, but that doesn't make the expression look like anything nice either.

I don't know how to go about this, and this is quite literally the first problem of a course in which I can follow the lectures.


I'll assume the instructor has shown why $\int_{-\infty}^\infty e^{-\lambda t^2}dt=\sqrt{\pi/\lambda}$ (if not, look up "Gaussian integral"). We have

$$ K'(\lambda)-b^2K(\lambda)=-\int_{-\infty}^\infty e^{-\lambda t^2}\left(\frac{t^2+b^2}{t-ib}\right)dt=-\int_{-\infty}^\infty e^{-\lambda t^2}(t+ib)dt. $$

The $t$ term integrates to zero since $t\mapsto t\,e^{-\lambda t^2}$ is an odd function, and for any integrable $f$, we have $\int_{-\infty}^\infty f(t)dt=\int_{-\infty}^\infty f(-t)dt$ after changing variables.

  • $\begingroup$ I should probably edit this, but I don't know the end-point differential equation before-hand. I'm supposed to reach that end point. So I don't a priori know that -b^2 coefficient. Thank you, though. This gives me a place to attempt again. $\endgroup$ – Chronum Jun 15 '17 at 14:38
  • $\begingroup$ Fair enough. I wish you luck! $\endgroup$ – user254433 Jun 15 '17 at 14:43
  • $\begingroup$ On second thoughts, this appears to be more of an intuition problem (I should be able to intuitively see that the t^2+b^2 has something to do with the denominator), and your answer appears to be the way to go. $\endgroup$ – Chronum Jun 15 '17 at 14:45
  • $\begingroup$ Here's some extra intuition in case it helps. We know how to evaluate $\int e^{-t^2}dt$ and $\int t e^{-t^2}dt$, but not $\int \frac{1}{t+i}e^{-t^2}dt$, so it's natural to try to "increase the polynomial order" of the integrand. Differentiating with respect to $\lambda$ gives $\frac{t^2}{t-ib}$, so we're nearly there, but we want the denominator to cancel with the numerator. To make it cancel, add and subtract: $\frac{(t^2-ibt)+ibt}{t-ib}=t+\frac{ibt}{t-ib}$. Add and subtract again: $\frac{ibt-(ib)^2+(ib)^2}{t-ib}=ib-\frac{b^2}{t-ib}$. But the second term corresponds to $-b^2K(\lambda)$. $\endgroup$ – user254433 Jun 15 '17 at 14:58
  • $\begingroup$ Thank you so much for clarifying all of this. I'm still getting used to the way of thinking about all of this. It's been a long time since I've attempted to learn something this far out of my league. Can't say it feels bad, but then again, being stuck doesn't feel good. $\endgroup$ – Chronum Jun 15 '17 at 15:04

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.