A question about part of the proof of the Central Limit Theorem

85 Views Asked by At

I've been reading the proof the Central limit theorem from Rick Durret's book. In Theorem 3.3.8 he proves the following

If $E[X^2]<\infty $ then $$\tag{1}\phi(t)=1+itE[X]-\frac{t^2}{2}E[X^2]+o(t^2)$$ where $\phi$ denotes the characteristic function of $X$ and $\lim_{t\to 0}o(t^2)/t^2=0$.

For the proof of this theorem he uses the estimate $$\tag{2}\biggl|e^{ix}-\sum_{k=0}^{n}\frac{(ix)^k}{k!}\biggr|\leq \min\biggl\{\frac{2|x|^n}{n!},\frac{|x|^{n+1}}{(n+1)!}\biggr\}$$

Although, just before the proof of this estimate he makes the observation that if $X$ has finite $n-$th moment $E[|X|^n]<\infty$ then $\phi$ is $n$ times differentiable at $0$ and $\phi^{(n)}(0)=E[(iX)^n]$. So, it follows from the Taylor expansion of $\phi$ at $0$ that $$\tag{3}\phi(t)=\sum_{k=0}^{n}\frac{E(iX)^k}{k!}t^k+o(t^n)$$

My question is on the proof of Theorem 3.3.8 why do we need the estimate of the difference of $e^{ix}$ and it's partial sums since before we have proved $(3)$ which gives us the result for $n=2$. After the proof of Theorem 3.3.8 he makes a comment that roughly says "because we dont have to assume that $E[|X|^3]<\infty$", but i cannot see why. Anyone knows why do we have to prove $(2)$ and use it to prove $(1)$ rather than use $(3)$?

Thanks in advance!