2

Let $t \in \mathbb{R}_{>0}$ and $k \in \mathbb{R}$. I want to find $$\int_{-\infty}^\infty e^{-\frac{x^2}{2t}} e^{-ikx} \, \mathrm dx.$$ A hint told me to first determine $\int_{-\infty}^\infty e^{-\frac{x^2}{2}} \, \mathrm dx$ which I found to be equal to $\sqrt{2 \pi}$. Now I am told to compute the integral using Cauchy's formula for a convenient Cauchy contour. As I have not yet practically applied Cauchy's formula and I have no idea how it would be helpful in this case, I ask for a little help, a hint would be enough really. Thanks in advance.

studeth
  • 347

3 Answers3

5

Here is a method which circumvents complex analysis. As DonAntonio pointed out, we have

$$ \exp\left\{ -\frac{x^2}{2t} - ikx \right\} = \exp\left\{ -\frac{1}{2t}\left( x + ikt \right)^2 - \frac{k^2t}{2} \right\} $$

Since

$$ \frac{d}{du} \exp\left\{ -\frac{1}{2t}\left( x + iut \right)^2 \right\} = -i(x+iut) \exp\left\{ -\frac{1}{2t}\left( x + iut \right)^2 \right\}, $$

we have

$$ \begin{align*} & \int_{-\infty}^{\infty} \exp\left\{ -\frac{1}{2t}\left( x + ikt \right)^2 \right\} \, dx - \int_{-\infty}^{\infty} \exp\left\{ -\frac{x^2}{2t} \right\} \, dx \\ &= \int_{-\infty}^{\infty} \left[ \frac{d}{du} \exp\left\{ -\frac{1}{2t}\left( x + iut \right)^2 \right\} \right]_{u=0}^{u=k} \, dx \\ &= -i \int_{-\infty}^{\infty} \int_{0}^{k} (x+iut) \exp\left\{ -\frac{1}{2t}\left( x + iut \right)^2 \right\} \, dudx. \end{align*}$$

Since the integrand is Lebesgue integrable on $(x,u) \in \Bbb{R} \times [0, k]$, we can apply Fubini's theorem and we have

$$ \begin{align*} & \int_{-\infty}^{\infty} \exp\left\{ -\frac{1}{2t}\left( x + ikt \right)^2 \right\} \, dx - \int_{-\infty}^{\infty} \exp\left\{ -\frac{x^2}{2t} \right\} \, dx \\ &= -i \int_{0}^{k} \int_{-\infty}^{\infty} (x+iut) \exp\left\{ -\frac{1}{2t}\left( x + iut \right)^2 \right\} \, dxdu. \\ &= \int_{0}^{k} \left[ it \exp\left\{ -\frac{1}{2t}\left( x + iut \right)^2 \right\} \right]_{x=-\infty}^{x=\infty} \, du = 0. \end{align*}$$

Therefore

$$ \begin{align*} \int_{-\infty}^{\infty} \exp\left\{ -\frac{x^2}{2t} - ikx \right\} \, dx &= \int_{-\infty}^{\infty} \exp\left\{ -\frac{1}{2t}\left( x + ikt \right)^2 - \frac{k^2t}{2} \right\} \, dx \\ &= \exp\left\{ - \frac{k^2t}{2} \right\} \int_{-\infty}^{\infty} \exp\left\{ -\frac{x^2}{2t} \right\} \, dx \\ &= \sqrt{2\pi t} \exp\left\{ - \frac{k^2t}{2} \right\}. \end{align*}$$

Sangchul Lee
  • 167,468
2

After completing the square of the exponential argument, use the following contour to justify the change of variables, enter image description here

Here is what $c$ is $$\displaystyle \int_{-\infty}^\infty e^{-p(t+c)^2}dt = \sqrt{\frac{\pi}{p}}, \quad p,c\in {\bf C},\;\mathrm{Re}\displaystyle \left\{p\right\}>0$$

0

Complete the square:

$$\frac{x^2}{2t}+ikx=\frac{1}{2t}(x^2+2tikx)=\frac{1}{2t}(x+tik)^2+\frac{t^2k^2}{2t}=\frac{1}{2t}\left[(x+tik)^2+t^2k^2\right]\Longrightarrow$$

$$\Longrightarrow e^{-\frac{x^2}{2t}-ikx}=e^{-\frac{1}{2t}(x+tik)^2}\,e^{-\frac{1}{2}tk^2}$$

Now, substituting

$$u:=\frac{1}{\sqrt{2t}}(x+tik)\Longrightarrow du=\frac{1}{\sqrt{2t}}dx\Longrightarrow$$

$$\Longrightarrow \int_{-\infty}^\infty e^{-\frac{1}{2t}(x+tik)^2}dx=\sqrt{2t}\int_{-\infty}^\infty e^{-u^2}du=\sqrt {2t\pi}$$

End now the exercise.

DonAntonio
  • 211,718
  • 17
  • 136
  • 287
  • 4
    Your answer may be right, but the last step isn't since $u$ is complex. How did you get the domain of integration in the last integral? – HorizonsMaths Oct 27 '12 at 11:53
  • Well, I assume the OP knows what to do with integration limits of the form $,\pm\infty +ia,,,,a+i\infty,$ and etc., but I can see your point. I'll try to add some explanation later. Thanx. – DonAntonio Oct 27 '12 at 12:05
  • @Mercy: Deformation of the infinite coutour of integration is justified by easy estimates on the function. – Alexander Shamov Oct 27 '12 at 12:08
  • Perhaps so, @AlexanderShamov, yet I think Mercy meant that in my answer there must be a little more info that it shows and, perhaps, the OP won't be able to apply this as he write he's hardly used Cauchy formula and etc. I've no time right now but I'll try to come up with some idea later to help the OP out, if in the meantime someone doesn't fill up this hole. – DonAntonio Oct 27 '12 at 12:12
  • @AlexanderShamov I don't get your point. What do you mean exactly? – HorizonsMaths Oct 27 '12 at 12:12
  • @Mercy: I mean that $\int_{-\infty+tik}^{+\infty+tik} e^{-x^2/2t} dx = \int_{-\infty}^{+\infty} e^{-x^2/2t} dx$, since the integrals over the left-hand and right-hand side of the rectangle drawn in Mhenni Benghorbal's answer vanish as $T$ goes to $\infty$. – Alexander Shamov Oct 27 '12 at 20:16
  • @AlexanderShamov And because the function is holomorphic and with no pole in the rectangle. – Did Oct 28 '12 at 07:52