# The ring generated by all functions from a set to itself

Let $S$ be a finite set. Now $\mathop{End}(S)$ is a monoid, and we may build a ring $R$ by allowing formal sums of functions.

Preliminary questions, since $R$ is surely well-known: What is it called? In a general category, what is the name of the construction that builds a ring out of the endomorphisms of an object?

My real question is about $A = R \otimes \mathbb{C}$. Restricting to the subalgebra generated by the automorphisms of $S$, it is clear that understanding the representation theory of symmetric groups is a prerequisite for understanding $A$. Are the symmetric groups on sets of size smaller than $| S |$ responsible for all nontrivial properties of $A$? If so, what suitably strong formulation of inclusion/exclusion is at work?

-

The monoid of all maps on $n$ letters is denoted $T_n$ and called the full transformation monoid. Your intuition is both right and wrong. The irreducible representations of $T_n$ are in bijection with irreducible representations of all symmetric groups of degree at most $n$. The character table is block upper triangular with diagonal blocks character tables of symmetric groups of degree at most $n$. Thus inverting the table, which is how you decompose characters into irreducibles, is a sort of generalized inclusion-exclusion. Putcha computed the character table, although Hewitt and Zuckermann stated a portion of the result without a complete proof in 1957. But this is only the semisimple part of the story.

The algebra of $T_n$ is not semisimple. For example, nobody knows what the projective indecomposables look like in general. The algebra is quasihereditary and so has finite global dimension. Putcha computed a lot of information about the quiver. Ponizovsky showed $T_1,T_2,T_3$ has finite representation type, Ringel showed $T_4$ has finite representation type and Putcha showed the representation type is infinite in all other cases. I think this part of the representation theory is not completely controlled by symmetric groups.

Update 1/15. In case you are still interested I just put up a paper http://arxiv.org/abs/1502.00959 computing the global dimension of this algebra. From the paper you can get a very precise idea of how the representations of the symmetric groups of different degrees interact to control the Ext spaces between simple modules in a non trivial way. Also the paper gives a more accurate survey of the representation theory of this algebra than my answer.

-
Here is a link to Putcha's paper. I think it will answer to a certain extent your question. plms.oxfordjournals.org/content/s3-73/3/623.abstract – Benjamin Steinberg Jun 27 '11 at 3:22
This is great! I have often wondered about these homological questions since the answers should have influence over certain representation stability phenomena. – John Wiltshire-Gordon Feb 4 '15 at 16:56
I agree. Things are tougher than in the case of the category of injective mappings because the category of all mappings is not EI. I do think this algebra should be studied in greater detail. – Benjamin Steinberg Feb 4 '15 at 17:12