Truncation and smoothing for relative \((C,1)\) limits (Q1175727)

The author introduces for any sequence \(\xi=(\xi(j))_{j\in\mathbb{N}}\) and any subset \(E\) of \(\mathbb{N}\) the doubly truncated sequence \(\xi_ b^ a\), where \(\xi^ a_ b(j)=\min\{\xi(j),a\}\), if this is \(\geq b\) and 0 otherwise \((a,b>0)\). It is shown that \(\xi^ a_ b\) has a relative \((C,1)\) limit \(C(\xi^ a_ b/E)\) relative to \(E\) for all but countably many positive values of \(b\) (``truncation sequence relative to \(E\)''). Truncation sequences relative to \(E\) are precisely the (not-negative) sequences for which the relative asymptotic density \(\delta_ E(\xi,t)\) of \(\xi^{-1}[t,\infty)\), relative to \(E\), exists for all but countably many values of \(t>0\). Furthermore, the integral \(\int_ 0^ \infty \delta_ E(\xi,t)dt\) is precisely the limit of \(C(\xi^ a_ b/E)\) as \(a\to\infty\) and \(b\to 0\). The proof makes use of the author's theory of ``improper Riemann integration''. In particular, it is shown that there exists some finitely additive measure \(\nu\) on a suitable field of subsets of \(\mathbb{N}\) (dependent on \(\xi\) and \(E\)) for which \(\int_ 0^ \infty \delta_ E(\xi,t)dt\) is just the ``abstract'' integral \(I(\xi/E)\), where \(\xi\) is called ``smooth relative to \(E\)'' in the case, where the relative \((C,1)\) limit \(C(\xi/E)\) coincides with \(I(\xi/E)\). It is proved that \(\xi\) can be ``smoothed'' relative to \(E\) by replacing its values on a thin (``negligible'') set of integers by 0. Throughout this article numerous examples involving uniformly distributed sequences are included to illustrate the concepts and results and their limitations. The author points out that a measure theoretical investigation of the behavior of other summation methods or generalized limits should be of interest.