On Mukhin's necessary and sufficient condition for the validity of the local limit theorem (Q6177642)

Let \((S_n)_{n\geq1}\) be a sequence of integer-valued random variables such that there exist sequences of real numbers \((a_n)_{n\geq1}\) and \((b_n)_{n\geq1}\) such that \(b_n\to\infty\) as \(n\to\infty\) and \(b_n^{-1}(S_n-a_n)\) converges weakly to an absolutely continuous distribution \(G\) with density \(g\). The corresponding local limit theorem is said to hold if \[ \mathbb{P}(S_n=m)=b_n^{-1}g\left(\frac{m-a_n}{b_n}\right)+o(b_n^{-1}), \] uniformly in \(m\). Necessary and sufficient conditions for the local limit theorem to hold are stated by \textit{A. B. Mukhin} [Dokl. Akad. Nauk USSR 8, 7--8 (1984; Zbl 0553.60034)], but from the proof presented there the author of the present paper is able to establish only a weaker result: letting \[ \varepsilon_n = \sup_{x\in\mathbb{R}}\left|\mathbb{P}\left(\frac{S_n-a_n}{b_n}<x\right)-G(x)\right|\to 0, \] a necessary and sufficient condition for the local limit theorem to hold is that \[ \sup_{|m-k|\leq\max\{1,[\sqrt{\varepsilon_n}b_n]\}}\left|\mathbb{P}(S_n=m)-\mathbb{P}(S_n=k)\right|=o(b_n^{-1}). \]