Integral from $e^{-itx}$ over $\mathbb R$

1.8k Views Asked by At

As a part of my task considering characteristic functions I have to compute $\int_{\mathbb R} e^{-itx}dx$

The result i get is $\frac{1}{-it}e^{-itx}|^{\infty}_{-\infty}$, but I don't really know where to go from here. Could anyone help, please?

(Wolfram says the integral does not converge - my starting problem was given characteristic function $\phi (t)=\frac{1}{4}+\frac{1}{4}e^{-it}+\frac{1}{2}\frac{3}{4-e^{2it}}$ find the the density of random variable X - so how to find it given the brute force method (integration) does not work?)

1

There are 1 best solutions below

0
On BEST ANSWER

In fact the integral does not converge in the classical sense. However, we have that the characteristic function of a random variable with probability $p>0$ of taking on the value $x$ will have a term of $p e^{i t x}$. In particular the characteristic function of a random variable which has probability $p$ of taking on the value $0$ will have a term of $p$. A consequence of that is that in the sense of distribution theory, the integral you have written is $\delta_0$, that is, the Dirac delta centered at zero.

In your case, let us expand the last term with the geometric series:

$$\frac{1}{2} \frac{3}{4-e^{2 i t}} = \frac{3}{8} \frac{1}{1-e^{2 i t}/4} = \frac{3}{8} \sum_{n=0}^\infty \left ( \frac{e^{2 i t}}{4} \right )^n.$$

This means that your variable has probability $1/4$ of being $-1$, probability $1/4+3/8=5/8$ of being $0$, and probability $\frac{3 \cdot 4^{-n}}{8}$ of being $2n$ for each natural number $n$.

The fact that this variable is actually discrete should explain why the integral defining its density fails to converge: it is because its density doesn't exist as a function, but rather only as a distribution.