Roth's theorem in many variables (Q2017118)

The authors study solubility of translation invariant equations of type \[ a_1 x_1 + \dots + a_k x_k = 0 \] with \(a_1 + \dots + a_k = 0\) in subsets of integers. Their main theorem says that if \(N\) and \(k \geq 6\) are integers and a set \(A \subset \{1, \dots, N\}\) does not have solutions to the equation above, then \(A = O(N \exp(- (\log N)^{1/7}))\). Behrend's well-known construction shows that there are solution-free sets \(A\) with at least \(N \exp(- (\log N)^{1/2})\) elements, so the obtained upper bound is quite close to this. An important tool in the proof is \textit{T. Sanders}' [Anal. PDE 5, No. 3, 627--655 (2012; Zbl 1320.11009)] work on polynomial Freiman-Ruzsa conjecture which allows an efficient density increment argument. The authors first prove a simpler finite field case of the theorem, so that the reader can get grasp of the main ideas before going into technicalities about Bohr sets etc.