I am following the (german) construction of the Grothendieck group at https://de.wikipedia.org/wiki/Grothendieck-Gruppe. It is very precise and easy to follow until to the point of the universal property.

I understand that for a commutative semi group $(H,+)$ we define $$(x,y) \sim (x',y') : \iff \exists t \in H : x+y'+t=x'+y+t $$ which gives an equivalence relation on $H^2$. The equivalence class of said relation will be denoted as $G(H)$ the Grothendieck group of $H$ and with component wise addition it forms an abelian group.

In the article they give $\varphi_H : H \to G(H)$ as $\varphi_H(x)=[x+x,x]$ and then they claim that for any Group $G$ and semi group morphism $\Psi: H \to G$ there exists exactly one groupmorphism $\lambda: G(H) \to G$ such that $\lambda \circ \varphi_H = \Psi$ (Universal Property)

I can't deduce from that how $\lambda$ must be defined, of course it would be a group morphism, because $\Psi$ is, and furthermore it would be uniquely determined by $\Psi$, but how can I show that $\lambda$ is well defined? For this I suppose that I need more details about how $\lambda$ actually looks i.e. $$\lambda([x,y])=? $$ I thought about taking $\lambda([x,y])=\Psi(y)$ as a projection, but then realized that with such a definition I cannot deduce that $$G(\mathbb{N},+) \cong (\mathbb{Z},+) $$

  • 1
    $\begingroup$ $[a,b]$ should be thought of as $a-b$. Does that help? (I assume $G$ is an abelian group too? Because I'm using additive notation here instead of multiplicative.) $\endgroup$ – whacka Dec 1 '15 at 23:25
  • $\begingroup$ In the mentioned wikipedia article (not that I take it as granted) $G$ is an arbitrary group, but I personally am already happy if I understand it for abelian groups. I have also read before that $[a,b]$ can be thought of as $a-b$, do you recommend that $$\lambda([x,y])=\Psi(x-y)? $$ $\endgroup$ – Spaced Dec 1 '15 at 23:29
  • $\begingroup$ $\Psi(x-y)$ doesn't make sense because $x-y$ doesn't make sense because inverses don't necessarily exist in a semigroup. What can you write instead of $\Psi(x-y)$ that does make sense? $\endgroup$ – whacka Dec 1 '15 at 23:30
  • $\begingroup$ Maybe I can write $\Psi (x) - \Psi(y)$ because inverses exists in $G$ but I suppose that is equivalent because of the semi groupmorphism property of $ \Psi$ other than that, I can always go for $\Psi(x+y)$ which makes sense, but then I would have a hard time showing that it is well defined. $\endgroup$ – Spaced Dec 1 '15 at 23:36
  • $\begingroup$ Yes. Since $[x,y]$ is supposed to be $x-y$, we want $\Psi([x,y])$ to be $\Psi(x)-\Psi(y)$. $\endgroup$ – whacka Dec 1 '15 at 23:45

The idea behind constructing $\mathbb{Z}$ formally out of $\mathbb{N}$ is to create a set of "formal differences" $x-y$ encoded as ordered pairs $(x,y)$ modulo the relation $(x,y)\sim(x+z,y+z)$. Similarly we construct $\mathbb{Q}$ formally out of $\mathbb{Z}$ by making a set of "formal quotients $x/y$ encoded as ordered pairs $(x,y)$ with $y\ne 0$ modulo the relation $(xz,yz)$ for all $z\ne 0$. (Then we define addition.)

The same holds in general for Grothendieck constructions. In particular, given a semigroup $H$ (it is most natural to consider cancellative semigroups) the Groethendieck construction consists of formal differences once again encoded as ordered pairs $(x,y)$. Perhaps $(x,y)\sim (x',y')$ should mean that "$x-y=x'-y'$" (which doesn't make sense in a semigroup without all inverses) which we would write down as $x+y'=x'+y$, and I assume we add $+t$ to both sides for extra padding in case $H$ is not cancellative.

For the universal property, since $(x,y)$ is supposed to be $x-y$, we want $\Psi(x,y)$ to be $\Psi(x)-\Psi(y)$, so we can just define it that way and verify the universal property holds.


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.