2
$\begingroup$

I want to prove that: \begin{equation} \lim_{t\to 0} \int_{|x|>\epsilon} \frac{e^{-\frac{x^2}{2t}}}{\sqrt{2 \pi t}} dx=0, \end{equation} for any $\epsilon >0$

I've shown using polar coordinates that \begin{equation} \int \limits_{-\infty}^{+\infty} \frac{e^{-\frac{x^2}{2t}}}{\sqrt{2 \pi t}}=1. \end{equation}

I tried the same approach for the limit I need to prove. But I obtained a result that depended on $\epsilon$ only, and not on $t$, and therefore it does not go to $0$. Where is my mistake?

Other ways to approach it would be greatly appreciated too (preferably, with real analysis methods only).

$\endgroup$
  • $\begingroup$ it is very difficult to answer what goes wrong in your computation that gave you "a result that depended on $\epsilon$ only" unless you actually show that computation. $\endgroup$ – hmakholm left over Monica May 5 '15 at 23:38
  • $\begingroup$ The computation I did is similar with the one here: mathworld.wolfram.com/GaussianIntegral.html $\endgroup$ – odnerpmocon May 5 '15 at 23:41
  • $\begingroup$ The only important difference is that $r$ is from $2 \epsilon ^2$ to $\infty$ instead of from 0 to $\infty$ $\endgroup$ – odnerpmocon May 5 '15 at 23:42
1
$\begingroup$

We can use the same polar coordinates trick here and in fact obtain a good bound on the integral. Let $$I(\epsilon,t) = \int_{\vert x \vert > \epsilon} \dfrac{e^{-\frac{x^2}{2t}}}{\sqrt{2\pi t}}dx$$ We now have $$I^2(\epsilon,t) = \int_{\vert x \vert > \epsilon} \dfrac{e^{-\frac{x^2}{2t}}}{\sqrt{2\pi t}}dx \cdot \int_{\vert y \vert > \epsilon} \dfrac{e^{-\frac{y^2}{2t}}}{\sqrt{2\pi t}}dy = \int_{\vert x \vert > \epsilon} \int_{\vert y \vert > \epsilon} \dfrac{e^{-\frac{x^2+y^2}{2t}}}{2\pi t}dxdy \leq \int_{x^2+y^2 > \epsilon^2} \dfrac{e^{-\frac{x^2+y^2}{2t}}}{2\pi t}dxdy$$ Now moving to polar coordinates, we have $$I^2(\epsilon,t) \leq \int_{\theta=0}^{2\pi} \int_{r=\epsilon}^{\infty} \dfrac{e^{-\frac{r^2}{2t}}}{2\pi t}rdrd\theta = \dfrac1t \int_{r=\epsilon}^{\infty}re^{-\frac{r^2}{2t}}dr = \dfrac1t \cdot te^{-\frac{\epsilon^2}{2t}} = e^{-\frac{\epsilon^2}{2t}}$$ This gives us that $$\boxed{\color{blue}{I(\epsilon,t) \leq e^{-\frac{\epsilon^2}{4t}}}}$$ Take $\lim t \to 0$ to obtain what you want.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks. My approach was pretty much similar, other than a change of variable in the beginning which I shouldn't have done. $\endgroup$ – odnerpmocon May 5 '15 at 23:49

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.