Integer colorings with no rainbow 3-term arithmetic progression (Q2138586)

For an additive set~\(A\), that is, a subset of some abelian group, let~\(g_r(A)\) be the number of \(r\)-colourings of~\(A\) with \(r\)~colours that contain no rainbow \(3\)-term arithmetic progression (\(3\)AP). That is, let~\(g_r(A)\) be the number of functions \(\chi\colon A\to[r]=\{1,\dots,r\}\) such that for every non-trivial \(3\)AP \(\{a,b,c\}\subset A\), at least two of~\(\chi(a)\), \(\chi(b)\) and~\(\chi(c)\) coincide. The authors prove an upper bound for~\(g_r(A)\) for the case in which~\(A\subset[n]\subset\mathbb{Z}\) that allows them to conclude that \(g_r([n])=(\binom{r}{2}+o(1))2^n\), and hence that almost all rainbow \(3\)AP-free \(r\)-colourings of~\([n]\) are actually \(2\)-colourings. Also, they are able to show that among all \(A\subset[n]\), the quantity~\(g_r(A)\) is maximized when \(A=[n]\) as long as~\(n\) is large enough. The authors also consider the case in which \(A\subset\mathbb{Z}/n\mathbb{Z}\), obtaining again upper and lower bounds that allow them to conclude results similar to the ones for \(A\subset[n]\). In the case that \(A=\mathbb{Z}/p\mathbb{Z}\) where~\(p\) is a prime, \(g_r(A)\)~is determined exactly. The proofs rely on the container method of Balogh, Morris and Samotij and Saxton and Thomason. For similar results for \(k\)-term arithmetic progressions, see [\textit{H.~Lin} et al., Eur. J. Comb. 104, Article ID 103547, 12 p. (2022; Zbl 1490.05268)].