Compactness of families of convolution operators with respect to convergence almost everywhere (Q2501032)

Let \(\{\mu_n: n= 1,2,\dots\}\) be a sequence of probability measures on the unit circle \(\mathbb{T}:= [0,2\pi)\). The authors consider the convolution operators \[ f\to (f^*\mu_n)(x):= \int f(x- t)\,d\mu_n(t). \] If each \(\mu_n\) is an absolutely continuous measure, then the following Theorem 1 is proved. If the sequence \(\{K_n\}\) of kernels is such that \[ K_n(t)\geq 0,\quad \int K_n(t)\,dt= 1, \] and for every \(r> 0\), \[ \int_{|t|> r} K_n(t)\,dt\to 0\qquad\text{as }n\to\infty, \] then there exists an increasing sequence \(\{n(j)\}\) of integers such that for every \(f\in L^1\), \[ (f* K_{n(j)})(x)\to f(x)\quad \text{as }j\to\infty\qquad\text{a.e.}. \] It is also proved that an analogous theorem is not true in the case of discrete measures or singular nonatomic measures.