So, I need to prove the identity

$$\int_{-\infty}^\infty \cos t^2 dt = \int_{-\infty}^\infty \sin t^2 dt = \sqrt{\frac{\pi}{2}}$$

and as a hint I have the Gaussian integral

$$\int_{-\infty}^\infty e^{-xt^2} dt = \sqrt{\frac{\pi}{x}} \;\;\;\forall x>0.$$

I suspect I have to take the real/imaginary part of $e^{-t^2}$ at some point, but I can't quite figure how. I.e., $\int e^z dz = e^z$ gives me nothing. So, how do I do it?

  • $\begingroup$ Do you know Euler's identity? $\endgroup$ – Matthew Leingang Jan 10 '15 at 21:11
  • 2
    $\begingroup$ check en.wikipedia.org/wiki/Fresnel_integral $\endgroup$ – janmarqz Jan 10 '15 at 21:11
  • 1
    $\begingroup$ see here $\endgroup$ – r9m Jan 10 '15 at 21:11
  • $\begingroup$ @MatthewLeingang Err, yes. $\endgroup$ – Minethlos Jan 10 '15 at 21:11
  • 1
    $\begingroup$ @r9m Thank you, I understand robjohn's derivation fully. $\endgroup$ – Minethlos Jan 10 '15 at 21:24

Substitute $x=i$ in your hint and simplify the resulting complex number and then you need to consider real and imaginary parts of the integral. See here for a justification which it depends on your background.

  • 3
    $\begingroup$ I think it also should be explained why one could let $x=i$ when the equality really says $x>0$. $\endgroup$ – mickep Jan 10 '15 at 21:27
  • $\begingroup$ @mickep: I was about to provide him with a link that explains that but it is better to stick to the hint he was given. $\endgroup$ – Mhenni Benghorbal Jan 10 '15 at 21:32
  • $\begingroup$ The hint is meant to hold only for real $x$, yes. $\endgroup$ – Minethlos Jan 10 '15 at 21:36
  • $\begingroup$ @Minethlos: They want you to use the hint and I showed you how to do it. $\endgroup$ – Mhenni Benghorbal Jan 10 '15 at 21:44
  • $\begingroup$ After some research I found that they probably wanted me to use the Identity theorem after which your advice does make sense :) $\endgroup$ – Minethlos Jan 10 '15 at 22:54

For a nice real analytic derivation, we can notice that: $$I^2=\left(\int_{0}^{+\infty}\cos(t^2)\,dt\right)^2 = \frac{1}{2}\iint_{[0,+\infty)^2}\cos(s^2-t^2)\,ds\,dt = \frac{1}{4}\int_{0}^{+\infty}\int_{-v}^{v}\cos(uv)\,du\,dv$$ so: $$ I^2 = \frac{1}{2}\int_{0}^{+\infty}\frac{\sin v^2}{v}\,dv =\frac{1}{4}\int_{0}^{+\infty}\frac{\sin z}{z}\,dz=\frac{\pi}{8}.$$ Some care is needed in proving that $\iint_{[0,+\infty)^2}\cos(s^2+t^2)\,ds\,dt$ vanishes and the manipulations are legit, since $\frac{\sin z}{z}$ is a Riemann integrable function but it does not belong to $L^1(\mathbb{R})$. Beside that, this proof just mimics the proof that $\int_{-\infty}^{+\infty}e^{-x^2}\,dx = \sqrt{\pi}$ through the Fubini-Tonelli's theorem.

  • $\begingroup$ As I just commented to the question, I have a real-variable approach. Do you have a nice real-variable method to evaluate $\int_0^\infty\frac{\sin(x)}{x}\mathrm{d}x$? $\endgroup$ – robjohn Jan 10 '15 at 22:27
  • $\begingroup$ @robjohn: I usually prove it with the (inverse) Laplace transform or through the identity $$\sum_{n=-\infty}^{+\infty}\frac{(-1)^n}{x+n\pi}=\frac{1}{\sin x}.$$ $\endgroup$ – Jack D'Aurizio Jan 10 '15 at 22:32
  • $\begingroup$ @robjohn: in fact, we can deal with convergence issues by introducing an exponential smoothing factor $e^{-\alpha x}$ and letting $\alpha\to 0$. This gives that our approaches are essentially the same. $\endgroup$ – Jack D'Aurizio Jan 10 '15 at 22:46
  • 1
    $\begingroup$ @JackD'Aurizio: that is most likely so since my approach also mimics the classic proof of $\int_{-\infty}^\infty e^{-\pi x^2}\,\mathrm{d}x=1$ $\endgroup$ – robjohn Jan 10 '15 at 22:55


$\Im e^{iz} = \sin(z) \implies \sin(z^2) = \Im e^{iz^2}$

  • $\begingroup$ Yes, I know Euler's identity. But I don't see any straightforward way to use it here. Do you mean there is simpler solution than here? $\endgroup$ – Minethlos Jan 10 '15 at 21:27

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.