In this post I want to discuss a combinatorial problem that is very appealing in its own right, but also important as a potential first step towards solving a central open problem in Ramsey theory. It is the second in a series of three posts (I may add to this number later, but three is what I have written or semi-written so far) that give further details about possible Polymath projects. This one was number 2 in the post in which I first discussed these projects. In that post, I said nothing beyond the fact that the project had close connections with the density Hales-Jewett theorem. Unlike the origin-of-life suggestion, this project is a straightforward mathematical one.
Let me very briefly indicate what the central open problem in Ramsey theory is. The density Hales-Jewett theorem, which has been discussed at great length on this blog, is the assertion that every dense subset of contains a combinatorial line (provided that is large enough in terms of and the density). This implies Szemerédi’s theorem.
Now there is an amazing generalization of Szemerédi’s theorem, due to Bergelson and Leibman, known as the polynomial Szemerédi’s theorem. This is the following assertion. For any and any choice of polynomials with integer coefficients and constant terms equal to zero, there exists such that every subset of size at least contains a subset of the form with . To see that this generalizes Szemerédi’s theorem, just let (unless you feel that is more natural).
Another special case of this theorem is when and . Then one is trying to find a subset of the form , or in other words a pair of elements of that differ by a perfect square. This was proved independently by Sarkozy and Furstenberg, though my favourite proof is a more recent argument due to Ben Green that follows more closely the general structure of Roth’s proof of Roth’s theorem.
An obvious question now arises: is there a generalization of DHJ that implies both DHJ and the polynomial Szemerédi’s theorem? So far, the answer is no, and that is the central problem I was referring to earlier. However, we do at least know the colouring version: that is, there is a colouring statement that simultaneously generalizes the Hales-Jewett theorem and van der Waerden’s theorem. This is a result of Bergelson and McCutcheon — alternative proofs have been given by Mark Walters (who has the shortest and simplest argument) and Saharon Shelah (who has produced a primitive recursive bound). (more…)