Good Ramanujan Expansions: A suitably enhanced decay of coefficients has important consequences

Abstract

In this self-contained short note, we introduce the new definition of Good Ramanujan Expansion, say G.R.E., for a fixed arithmetic function $F$, building upon a good decay of its coefficients $G$; this, gains $\log-$powers w.r.t. the trivial bound for $G$ and precisely $\log^{1+η}$, where the present parameter $η>0$ is real. This property alone has important consequences for all the $F$ having a G.R.E. : mainly, 1) the Eratosthenes Transform $F'$ of our $F$ is infinitesimal (see in Theorem 1); 2) when $η>1$ (an enhanced decay) we have uniqueness of $G$ (actually, these are the classic Wintner-Carmichael coefficients, see Th.2); 3) we get a bound for $F$ (in Th.3); 4) an important new class of arithmetic functions $F$ can't have a G.R.E. (see Th.4). These are a generalization of Correlations; which in this way, if are, say, a kind of "far from constants", may not have a G.R.E., whence, a fortiori, can't have the R.E.E.F. This is the Ramanujan Exact Explicit Formula, that we introduced with Prof. Ram Murty. On the Hardy-Ramanujan Journal, I proved that: any "fair" Correlation may be "well-approximated" by a (BH) Correlation. For a (BH) Correlation, we prove here: it has a G.R.E. IF AND ONLY IF it has the (R.E.E.F.). On the other hand, next Counterexample 1 in section 3 is a (BH) Correlation without the (R.E.E.F.); also, it provides here a (BH) Correlation behaving like in Theorem 4.