Get Complete Project Material File(s) Now! »

## Strategically Ramsey sets and the pigeonhole principle

The aim of this section is to prove a version of Rosendal’s theorem I.13 in the general setting of Gowers spaces. We also introduce the notion of the pigeonhole principle for a Gowers space and see that the last result can be strengthened in the case where this principle holds. This will enable us to see the fundamental dierence between the Mathias{Silver space and the Rosendal space over a eld with at least three elements. We start by introducing Gowers’ game and the asymptotic game in the setting of Gowers spaces, and the notion of a strategically Ramsey set. In this whole section, we x a Gowers space G pP;X;¤;¤; q.

1. Gowers’ game below p, denoted by Gp, is dened in the following way: where the xi’s are elements of X, and the pi’s are elements of P. The rules are the following:

for I: for all i P !, pi ¤ p.

for II: for all i P !, px0; : : : ; xiq pi.

### The strength of the adversarial Ramsey principle

In section II.1, we proved the adversarial Ramsey property for Borel sets using Borel determinacy, and we saw on the trivial example of the space with only one subspace that, given a suitable class of subsets of Polish spaces, the adversarial Ramsey property for -sets implied the determinacy for -games on integers. This had two consequences: on one hand, the use of a suciently large fragment of ZFC is necessary to prove the adversarial Ramsey property for Borel sets, and on the other hand, it is not possible to prove it for analytic or coanalytic sets in ZFC. However, the space we used to make this remark is quite articial. Of course, we made the same remark in the introduction of this thesis using the Rosendal space, however we did it by making players play according to the norms f the vectors, which is quite articial too (we would not do that, for example, in the applications to Banach-space geometry, where we usually restrict our attention to normalized vectors). Therefore, is it natural to ask in which cases using a large fragment of ZFC is necessary to prove the adversarial Ramsey property for Borel sets, or in which cases this property could be provable in ZFC for analytic and coanalytic sets; the aim of this section is to give an answer to this question. We will see, in particular, that Gowers spaces where the pigeonhole principle holds, and those where it does not hold, behave very dierently.

In this section and the next one, we x a suitable class of subsets of Polish spaces. Given a Gowers space G pP;X;¤;¤; q, we denote by AdvGpq the statement \every -subset of X! is adversarially Ramsey », and by StratGpq the statement \every -subset of X! is strategically Ramsey ». We let Advpq be the statement \for every analytic Gowers space G, AdvGpq holds », and Stratpq be the statement \for every analytic Gowers space G, StratGpq holds ». We proved in the two previous sections the following implications.

#### Closure properties and limitations for strategically Ramsey sets

In this section, we show the same kind of dierence of behavior between spaces with and without a pigeonhole principle as in the previous section, but this kind for strategically Ramsey sets. We x, in the whole section, a Gowers space G pP;X;¤;¤; q and a suitable class of subsets of Polish spaces. The rst thing to remark is that if G satises the pigeonhole principle, then by corollary II.21, the class of strategically Ramsey sets is closed under taking complements: X X! is strategically Ramsey if and only if Xc is so. In particular, in ZFC, every 11 subset of X! is strategically Ramsey. In spaces where the pigeonhole principle does not hold, the situation is very dierent; we rstly state the two main results of this section and present their consequences, before proving them.

The rst result is a generalisation of a theorem proved by Lopez-Abad [37] in the context of strategically Ramsey sets in Banach spaces. This result only holds for forgetful Gowers spaces, and to prove it, we need the negation of a slight weakening of the pigeonhole principle. We will say that the forgetful space G satises the weak pigeonhole principle if for every A X, there exists p P P such that either p A, or p Ac (where p A abusively denotes the fact that for every x P X, if x p, then x P A). Of course, in most of the concrete spaces we consider, P has a maximum 1 that is isomorphic to every subspace (meaning, here, that for every p0 P P, there are bijections : P ÝÑ tp P P | p ¤ p0u and ‘ : X Ñ tx P X | x p0u that preserve the relations ¤, ¤ and ); this is, for example, the case of the Mathias{Silver space or of the Rosendal space. In these spaces, the weak pigeonhole principle is equivalent to the pigeonhole principle. Our result is the following: Proposition II.26. Suppose that G is forgetful and does not satisfy the weak pigeonhole principle. Then StratGpq ñ StratGpDq.

**Table of contents :**

Remerciements

Courte introduction en francais

Notations and conventions

**I Introduction and history **

I.1 Determinacy

I.2 Innite-dimensional Ramsey theory

I.3 Gowers’ Ramsey-type theorem in Banach spaces and adversarial Gowers’ games

I.4 Banach-spaces dichotomies and complexity of the isomorphism

I.5 Organisation of the results

**II Ramsey theory with and without pigeonhole principle **

II.1 Gowers spaces and the aversarial Ramsey property

II.2 Strategically Ramsey sets and the pigeonhole principle

II.3 The strength of the adversarial Ramsey principle

II.4 Closure properties and limitations for strategically Ramsey sets

II.5 The adversarial Ramsey property under large cardinal assumptions

**IIIRamsey theory in uncountable spaces **

III.1 A counterexample

III.2 Approximate Gowers spaces

III.3 Eliminating the asymptotic game

**IV Hilbert-avoiding dichotomies and ergodicity **

IV.1 Preliminaries

IV.2 The rst dichotomy

IV.3 The second dichotomy

IV.4 Links with ergodicity and Johnson’s problem

IV.5 A simple proof of Gowers{Maurey’s theorem

**Bibliography**