3
$\begingroup$

Here is the question.

Suppose we have a sequence of probability measures $\{\mu_n\}$ defined \begin{align*} \mu_n(\{x\}) = \frac{1}{n} \end{align*} for $x = 0,\frac{1}{n}, \frac{2}{n},...,\frac{n-1}{n}$. Show that $\mu_n \Rightarrow \lambda$ where $\lambda$ is the Lebesgue measure on $[0,1]$.

By the continuity theorem we have the equivalence between weak convergence and pointwise convergence of characteristic functions. So it suffices to show that the characteristic functions $\{\phi_n\}$ of the sequence of measures $\{\mu_n\}$ converges pointwise to the characteristic function of $\lambda$. That is, I want to show that for any $t \in [0,1]$, \begin{align*} \phi_n(t) = \int e^{itx} \mu_n(dx) \to \int e^{itx} \lambda(dx) = \phi(t) \end{align*} First I computed the characteristic function of $\mu_n$, \begin{align*} \phi_n(t) = \int e^{itx} \mu_n (dx) = \sum_{k=0}^{n-1} \frac{1}{n} e^{it\frac{k}{n}} \end{align*} And the characteristic function of $\lambda$, \begin{align*} \phi(t) = \int e^{itx} \lambda(dx) = \int_0^1 e^{itx} dx = \frac{e^{it} - 1}{it} \end{align*} I am stuck here. I can't show that \begin{align*} \lim_n \sum_{k=0}^{n-1} \frac{1}{n} e^{it\frac{k}{n}} = \frac{e^{it} - 1}{it} \end{align*} but I computed some values in Matlab and I'm fairly certain its true. How can I evaluate this limit / show it is equal to the expression on the right?

$\endgroup$
1
$\begingroup$

Partial sum of the geometric series \begin{align}\sum_{k=0}^{n-1}\frac1n e^{it\frac{k}n}&=\frac1n\sum_{k=0}^{n-1}\left(e^{it\frac1n}\right)^k=\frac{1}{n}\cdot\frac{1-(e^{it}\frac1n)^n}{1-e^{it\frac1n}}=\frac1n\cdot\frac{1-e^{it}}{1-e^{it\frac1n}}=(1-e^{it})\cdot\frac{\frac1n}{1-e^{it\frac1n}}\end{align} and L'Hopital's rule \begin{align}\lim_{n\to\infty}\frac{\frac1n}{1-e^{it\frac1n}}=\lim_{x\to 0}\frac{x}{1-e^{itx}}\overset{\text{L'Hopital}}=\lim_{x\to 0}\frac{1}{-ite^{itx}}=\frac{1}{-it}\end{align} give $$\lim_{n\to \infty}\sum_{k=0}^{n-1}\frac1n e^{it\frac{k}n}=(1-e^{it})\cdot\frac{1}{-it}=\frac{e^{it}-1}{it}$$

$\endgroup$

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.