A local limit theorem for attraction to the standard normal law: The case of infinite variance (Q799295)

Let \(\{X_ n\}\) be a sequence of i.i.d. random variables with an absolutely continuous distribution function F and a characteristic function f. If F belongs to the domain of attraction of the standard normal law and \(EX^ 2_ 1<\infty\) then it is known that for the probability density function \(f_ n\) and for every \(\beta \leq 2\), \[ \sup_{-\infty <x<\infty}(1+| x|)^{\beta}| f_ n(x)-\phi (x)| =o(1)\quad as\quad n\to\infty \] if and only if f is absolutely integrable in the r-th power for some \(r\geq 1\). The author shows that the same conclusion is true for every \(\beta <2\) if \(EX^ 2_ 1=\infty\).