Number theoretic correlation inequalities for Dirichlet densities (Q677622): Difference between revisions

For a set \(A\) of integers, \(\underline{\mathbf D} A=\liminf_{s\to 1}(s-1)\sum _{a\in A}a^{-s}\) denotes its lower Dirichlet density. The principal result of the authors is the inequality \(\underline{\mathbf D}(A,B)\underline{\mathbf D}[A,B]\geq\underline{\mathbf D}A\underline{\mathbf D}B\), where \(A,B\) are arbitrary sets of positive integers and \((A,B)=\{(a,b):a\in A,\;b\in B\}\), \([A,B]=\{[a,b]:a\in A,\;b\in B\}\) are the sets of greatest common divisors and least common multiples, respectively. This inequality was previously established by the authors for sets of multiples [J. Number Theory 55, 170-180 (1995; Zbl 0849.11013)]. The present inequality is the first of its kind, valid for arbitary sets without structural restrictions. The proof is a rather straightforward deduction from an inequality of \textit{R. Ahlswede} and \textit{D. E. Daykin} [Z. Wahrsch. Verw. Geb. 43, 183-185 (1978; Zbl 0369.04004)] (and could be even more straightforward if the authors stated this for general lattices rather than subsets). As the authors say, they ``discovered a method to produce density inequalities from combinatorial correlation inequalities'', and they demonstrate this by giving several other examples. The applied concept of lower Dirichlet density is not the most natural one (the authors claim erroneously that it equals the lower logarithmic density), and it would be interesting to decide whether other densities admit analogous inequalities.