FI-sets with relations (Q2004046)

This article studies the structure of finitely generated \textit{FI}-sets, that is, functors from the category \(FI\) of finite sets with injections to the category of sets, and applications. Let us say that two \(FI\)-sets \(X\) and \(Y\) \textit{eventually coincide} if there is some integer \(N\) such that the restrictions of the functors \(X\) and \(Y\) to the full subcategory of \(FI\) consisting of finite sets with cardinality \(\ge N\) are isomorphic. If \(n\) is a non-negative integer and \(X\) a \(\mathfrak{S}_n\)-set, where \(\mathfrak{S}_n\) is the group of permutations of the set \([n]:=\{1,\dots,n\}\), the \textit{induced \(FI\)-set} associated to \(X\) can be defined as \(FI([n],-)\underset{\mathfrak{S}_n}{\times}X\). The main general result (Theorem 1.1) asserts that each finitely generated \(FI\)-set eventually coincides with a finite disjoint union of induced \(FI\)-sets. (Warning to the reader: the statement given in Theorem 1.1 is weaker but, as Remark 1.2 following just if mentions, the proof of the article gives this result.) This result is similar to -- but also has interesting differences with -- the following one, due to \textit{R. Nagpal} [``FI-modules and the cohomology of modular representations of symmetric groups'', Preprint, \url{arXiv:1505.04294}]. Let us fix a commutative Noetherian ring \(k\) and consider \(FI\)-modules over \(k\), that is, functors from \(FI\) to \(k\)-modules. If \(n\) is a non-negative integer and \(V\) a representation of \(\mathfrak{S}_n\) over \(k\), the induced \(FI\)-module associated to \(V\) is \(k[FI([n],-)]\underset{k[\mathfrak{S}_n]}{\otimes}V\) (where \(k[-]\) denotes the linearisation over \(k\)). So, the linearisation of an induced \(FI\)-set is a typical induced \(FI\)-module. One of Nagpal's main theorems states that, for each finitely generated \(FI\)-module \(F\), there is some non-negative integer \(n\) such that the shift \(\tau_n(F):=F\circ ([n]\sqcup -)\) of \(F\) has a finite filtration whose subquotients are induced \(FI\)-modules. So, for linearisation of \(FI\)-sets, Theorem 1.1 of the article under review implies Nagpal's theorem, but it gives also a strictly stronger statement, because: \begin{itemize} \item[--] an \(FI\)-module having a finite filtration with induced subquotients need \textit{not} to be itself induced; \item[--] if two \(FI\)-modules \(F\) and \(G\) eventually coincide, some shifts \(\tau_n(F)\) and \(\tau_n(G)\) are isomorphic, but the conserve is false. Indeed, Nagpal's Theorem can \textit{not} be strengthened by replacing the shift by an eventual isomorphism. \end{itemize} So, even if some easy lemmas are proven by using linearisation and known facts about \(FI\)-modules, most of the arguments to establish Theorem 1.1 (and other results of the article) are specific to \(FI\)-sets, with a more combinatorial flavour. In a second part, the authors study relations between \(FI\)-sets. A \textit{relation} between two \(FI\)-sets \(X\) and \(Y\) is by definition a sub-\(FI\)-set of their cartesian product \(X\times Y\). Let \(k\) be some ground field of characteristic \(0\), \(X\) and \(Y\) be two \(FI\)-sets and \(R\subset X\times Y\) be a relation. For each non-negative integer \(n\), let us define a linear map \(r_n: k[X_n]\to k[Y_n]\) (where \(X_n:=X([n])\)) by sending \([x]\) on \(\sum_{(x,y)\in R_n} [y]\). It shall be noticed that the collection of arrows \((r_n)\) does \textit{not} define a morphism between the \(FI\)-modules \(k[X]\) and \(k[Y]\). But \(r_n\) is \(\mathfrak{S}_n\)-equivariant, so one can look at its isotypic components associated to partitions of \([n]\). The second main result of the article (Theorem 1.5) gives a polynomial dependence in \(n\) of these components (all but the biggest part of the partition being fixed). The article gives also applications, and a lot of examples, in particular related to graphs. It proves, in Theorem 4.14, a result announced in [\textit{E. Ramos} and \textit{G. White}, Sel. Math., New Ser. 25, No. 5, Paper No. 70, 42 p. (2019; Zbl 1428.05321)]. This paper is generally written in a clear way, but we warn the unexpert reader of a misprint in Definition 3.2, where the inequality should be reversed (one should read: finitely generated in degree \(\le d\)). This work is a nice contribution to a combinatorial and representation-theoretic study of functors to sets and their interplay with functors to modules. In a similar spirit (but with different methods and motivations), one can mention the recent article [\textit{G. Powell}, Ann. Inst. Fourier 69, No. 5, 2169--2204 (2019; Zbl 1444.55010)]. This aspect of representation stability seems to have received very few attention but could give other fruitful results.