4
$\begingroup$

We shall first define a functor $$ F:\mathsf{Mon}^{\text{op}}\to\mathsf{CRing}, $$ where $\mathsf{Mon}^{\text{op}}$ is the category opposite to the category of monoids and $\mathsf{CRing}$ is the category of commutative rings with one.

Let $\mathsf{CSRing}$ be the category of commutative semirings with one, and $L:\mathsf{CSRing}\to\mathsf{CRing}$ the left adjoint to the forgetful functor (in particular $L(\mathbb N)=\mathbb Z$).

Our functor $F$ will be the composite $L\circ F'$, where $F':\mathsf{Mon}^{\text{op}}\to\mathsf{CSRing}$ is defined as follows:

Let $M$ be a monoid, $M$-$\mathsf{Set}_{\text{fin}}$ the category of finite $M$-sets, and $$ S\subset M\text{-}\mathsf{Set}_{\text{fin}} $$ a skeleton. Set $0:=\varnothing\in S$, let $1\in S$ be the terminal object, and for $X,Y\in S$ let $X+Y\in S$ be the coproduct of $X$ and $Y$, and $XY\in S$ the product of $X$ and $Y$.

It is straightforward to check that the formula $F'(M):=S$ defines a functor $F':\mathsf{Mon}^{\text{op}}\to\mathsf{CSRing}$, and we can set $F:=L\circ F'$.

As $F$ sends the trivial monoid to $\mathbb Z$, we have a natural morphism $F(M)\to\mathbb Z$. In other words, it would be better to view $F$ as a (contravariant) functor from monoids to commutative rings over $\mathbb Z$. In particular $F(M)$ contains $\mathbb Z$.

Question 1. Is the Krull dimension of $F(M)$ always equal to one?

Question 2. If the monoid $M$ is a group $G$, is $F(G)$ always integral over $\mathbb Z$?

The answer is Yes if

$\bullet$ the finite index subgroups of $G$ are normal,

or if

$\bullet\ G$ is finite.

(See below.)

Clearly, if $M$ is a group $G$, and $f:G\to\hat G$ is the morphism to the profinite completion, then $Ff:F(\hat G)\to F(G)$ is an isomorphism.


The ring $F(M)$ can be described as follows:

Say that an $M$-set is indecomposable if it cannot be written as a disjoint union of sub-$M$-sets in a nontrivial way, and let $I$ be the set of indecomposable objects of cardinality at least two in our skeleton $S$. For all $X,Y\in I$ we have $$ XY=\sum_{Z\in I}\ c_{XY}^Z\ Z, $$ where each $(c_{XY}^Z)_{Z\in I}$ is a finitely supported family of nonnegative integers. We get $$ F(M)\simeq\mathbb Z\left[(T_X)_{X\in I}\right]/\mathfrak a, $$ where the $T_X$ are indeterminates and $\mathfrak a$ is the ideal generated by the $$ T_XT_Y-\sum_{Z\in I}\ c_{XY}^Z\ T_Z. $$ In particular the element $1\in F(M)$ and the images $t_X$ of the $T_X$ form a $\mathbb Z$-basis of $F(M)$, and the natural morphism $F(M)\to\mathbb Z$ sends $t_X$ to zero.

Also note that if $M$ is a group $G$, and $N$ a normal subgroup of index $i<\infty$, then we have $(t_{G/N})^2=it_{G/N}$, and $t_{G/N}$ is integral over $\mathbb Z\subset F(G)$. This justifies the first claim after Question 2.

To prove the second claim after Question 2, recall that $I$ is the set of indecomposable objects of cardinality at least two in the skeleton $S$, and that $F(G)$ is generated by a family $(t_X)_{X\in I}$.

Assume first that the monoid $M$ is a (possibly infinite) group $G$. Order $I$ by setting $X\le Y$ if there is a surjective morphism $Y\to X$. We claim

(a) for all $X\in I$ the ring $F(G)$ is integral over the subring generated by the $t_Y$ with $Y>X$.

More precisely:

(b) for all $X\in I$ we have $$ X^2=nX+\sum_{Y>X}n_YY $$ with $n,n_Y\in\mathbb N$.

To prove (b) note that, for all $x_1,x_2\in X$, the stabilizer $H$ of $(x_1,x_2)\in X^2$ is the intersection of the stabilizers $H_1$ and $H_2$ of $x_1$ and $x_2$. Thus we have either $H=H_1=H_2$ and $G(x_1,x_2)\simeq X$ , or $H<H_i$ for $i=1,2$, and $G(x_1,x_2)>X$ (more correctly $Y>X$ if $Y$ is the unique element of $I$ isomorphic to $G(x_1,x_2)$). This proves (b), and thus (a).

Clearly, if $G$ is finite, (a) implies that $F(G)$ is integral over $\mathbb Z$, which is the second claim after Question 2.


Here are some examples:

As already indicated, if $M$ is the trivial monoid, then $F(M)\simeq\mathbb Z$. We also have $F(\mathbb Q)\simeq\mathbb Z$.

The ring $F(\mathbb Z)$ admit a $\mathbb Z$-basis $\{1,t_2,t_3,\dots\}$ with $$ t_it_j=(i\land j)\ t_{i\lor j}, $$ where $i\land j$ and $i\lor j$ denote the gcd and the lcm of $i$ and $j$.

If $M$ is the monoid $\{0,1\}$ with the obvious multiplication, then the ring $F(M)$ admit a $\mathbb Z$-basis $\{1,t_1,t_2,\dots\}$ with $$ t_it_j=t_{(i+1)(j+1)-1}. $$ If $S_3$ denotes the symmetric group on three letters, then the ring $F(S_3)$ admit a $\mathbb Z$-basis $\{1,t_2,t_3,t_6\}$ with $$ t_2^2=2t_2,\quad t_3^2=t_3+t_6,\quad t_it_6=it_6,\quad t_2t_3=t_6. $$ If $G$ is the Klein four-group (that is, the non-cyclic group of order $4$), then the ring $F(G)$ admit a $\mathbb Z$-basis $\{1,t_1,t_2,t_3,u\}$ with $$ t_i^2=2t_i,\quad u^2=4u,\quad t_iu=2u,\quad t_it_j=u\text{ for }i\ne j. $$

$\endgroup$
5
$\begingroup$

First, a remark: your construction is known as the Burnside ring (usually considered only in the case of a group) which I imagine will aid you in finding more information about it.

Question 1: No. Indeed, the monoid $M=\{0,1\}$ you mentioned is a counterexample. Let me write $x_n$ for what you have written $t_{n-1}$, so $x_n$ is the indecomposable $M$-set with $n$ elements (the $n$-element set with only one point in the image of $0$). Then these elements $x_n$ satisfy $x_nx_m=x_{nm}$. This makes it clear that actually $F(M)$ is just the polynomial ring $\mathbb{Z}[x_2,x_3,x_5,\dots]$ on the elements $x_p$ where $p$ is prime. This ring has infinite Krull dimension.

Question 2: Yes. To prove this, let $X$ be any finite $G$-set and let $K$ be the kernel of the action of $G$ on $X$. Note that $K$ is a finite index normal subgroup of $G$, and acts trivially on $X^n$ for all $n$. It follows that actually the subring of $F(G)$ generated by $X$ is isomorphic to the subring of $F(G/K)$ generated by $X$. Since $G/K$ is finite, this shows $X$ is integral over $\mathbb{Z}$ by the work you have done.

Incidentally, there is a quicker way to see $F(G)$ is integral over $\mathbb{Z}$ when $G$ is a finite group. Just note that $F(G)$ is generated as an abelian group by the $G$-sets $G/H$ where $H$ ranges over all subgroups of $G$. In particular, $F(G)$ is a finitely generated $\mathbb{Z}$-module and thus is integral over $\mathbb{Z}$.

$\endgroup$
  • $\begingroup$ My impression was that the terminology "Burnside ring" is used only in connection with groups, not other monoids. I agree that extending it to monoids is reasonable, but is that commonly done? $\endgroup$ – Andreas Blass Mar 19 at 19:39
  • $\begingroup$ That's a fair question. I don't think I've ever seen it defined explicitly for monoids, but I also haven't ever seen the concept explored for monoids. $\endgroup$ – Eric Wofsey Mar 19 at 20:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.