Published online by Cambridge University Press: 20 November 2018
How few three-term arithmetic progressions can a subset $S\,\subseteq \,{{\mathbb{Z}}_{N}}\,:=\,\mathbb{Z}\,/\,N\mathbb{Z}$ have if
$|S|\,\ge \,\upsilon N$ (that is,
$S$ has density at least
$\upsilon$ )? Varnavides showed that this number of arithmetic progressions is at least
$c(v)\,{{N}^{2}}$ for sufficiently large integers
$N$ . It is well known that determining good lower bounds for
$c\left( \upsilon\right)\,>\,0$ is at the same level of depth as Erdös's famous conjecture about whether a subset
$T$ of the naturals where
$\sum{_{n\in T}\,1/n}$ diverges, has a
$k$ -term arithmetic progression for
$k\,=\,3$ (that is, a three-term arithmetic progression).
We answer a question posed by B. Green about how this minimial number of progressions oscillates for a fixed density $\upsilon$ as
$N$ runs through the primes, and as
$N$ runs through the odd positive integers.