Strategies for $\text{Hom}$ Computations
Huh
Right now it feels like I kind of suck at computing $\text{Hom}$ between two sets. We’ve had a few problems like this on our problem sets for algebra, and my go to methods have always been what I think is naive and not super fun: just looking at what elements could map where and figuring out what restrictions there might be on the maps. In this sketch I want to explore the usage of exact sequences, tensor products, and category theory for discovering $\text{Hom}$ sets (or modules or etc. etc. in enriched structures).
Date Started: February 29, 2024
Date Finished: On Hiatus
My Ideas
$\mathsf{Grp}$
I’ll start with the category of groups. I’ll start with some simple examples, which I have even previously computed in previous sketches. Suppose we would like to compute (\Hom(\mathbb{Z}, \Z)). Many people would find this one to be quite simple: any homomorphism out of $\Z$ is completely determined by its value at one, since $\Z$ is generated by 1. In other words, if $\varphi \in \Hom(\Z,\Z)$, then $\varphi(k) = k \varphi(1)$ (here I’m implicitely utilizing the $\Z$-module structure on $\Z$ since it is an abelian group). Notice that in general, if we have fixed the value of $\varphi$ mapping out of a group $G$ on a subset $S \subseteq G$ in a consistent way, the values of $\varphi$ are determined on the smallest subgroup of $G$ containing $S$. This comes down to the classic idea on how I approach finding hom sets:
Claim. If $G$ is generated by ${g_{\alpha} \mid \alpha \in S}$ for some indexing set $S$, then any morphism out of $G$ is determined by its values on the generating set.
After we know this, we can look at the image. In the case of $\Hom(\Z,\Z)$, we know that we simply have to determine the value of any homomorphism on $1$, and we can see that $1$ can take any value in $\Z$. Thus, $\Hom(\Z,\Z) \cong \Z$. Similarly, we can see that $\Hom(\Z,\Q) \cong \Q$ and $\Hom(\Z,\R) \cong \R$. So far so good. In fact, we can see that for any group $G$, $\boxed{\Hom(\Z, G) \cong G}$ since for every element $g \in G$, we have a homomorphism mapping $1$ to $G$, so that $k$ maps to $g^k$. This is quite nice.
But things aren’t as nice when we consider mapping into $\Z$. Consider $\varphi \in \Hom(\Q,\Z)$. Again, notice that the values of $\varphi$ are determined on $\Z \subseteq \Q$ simply by choosing $\varphi(1)$, namely $\varphi(a) = a\varphi(1)$. However, this becomes a problem because we need to map $\varphi(\frac{1}{n})$ to a value such that $n\varphi(\frac{1}{n}) = \varphi(1) \in \Z$. On the other hand, if we choose $N \in \Z$ such that $N > \varphi(1)$, there clearly does not exist an integer such that $N\varphi(\frac{1}{N}) = \varphi(1)$, that is, unless $\varphi(1) = 0$, in which case we have that $\varphi(\frac{1}{n}) = 0$ for all $n$, so $\varphi$ is the zero map. As such, there is only one possible map, so $\boxed{\Hom(\Q, \Z) = 0}$. Notice that in this case there existed a larger subgroup of $\Q$ (the one generated by $\frac{1}{N}$) which was more restrictive than the one we initially picked ($\Z$). From this there’s one thing in particular that I would like to notice: $\Hom(\Z,\Z)$ is a superset of $\Hom(\Q,\Z)$ since $\Z$ is a subset of $\Q$; any homomorhism from $\Q$ to $\Z$ must also be a homomorphism when restricted to $\Z$.
Claim. There is no unique maximal element in the lattice of subgroups of $\Q$.
Proof Sketch.
Suppose for contradiction that there was a unique maximal element, $H \le \Q$. If $\frac{1}{p^k} \in H$ for all $p$ prime and $k \in \Z$, then $H=\Q$. Thus there must be some prime $p$ and $j \in \Z$ such that $\frac{1}{p^j}$ is not in $\Q$, so simply take the smallest subgroup contianing $H$ and $\frac{1}{p^j}$. This subgroup still doesn’t contain $\frac{1}{p^{j+1}}$ so it is not all of $\Q$, but it is larger than $H$, a contradiction. $\lendshow$
To go back to another idea, maybe the maybe this idea that $\Hom$ reverses inclusions in one component will end up motivating categorical limits or functoriality, I guess we’ll have to see.
Notice that the same issue happens with $\Z[\frac{1}{p}]$ for any $p$, we have that $\boxed{\Hom(\Z[\frac{1}{p}] ,\Z) = 0}$. We might instead consider $\Hom(\Z[\frac{1}{n}], \Z[\frac{1}{m}])$ for $n$ and $m$ integers. This becomes a much more tedious problem. We can see that $\phi(a) = a\phi(1)$ again on the integers, and $n^k\phi(\frac{1}{n^k}) = \phi(1)$, so we must have that $\phi(1)$ is always “divisible” by $n^k$ for all $k$, or that $n$ is invertible in $\Z[\frac{1}{m}]$. $n$ being invertible in $\Z[\frac{1}{m}]$ would imply that $n^k$ divides some power of $m$, so the primes dividing $n$ also divide $m$. It is interesting to note that this condition is precisely that $V(n) \subseteq V(m)$ in $\text{Spec}(\Z)$. In this case, regardless of where we map $\phi(1)$, things work, so that $\boxed{\Hom(\Z[\frac{1}{n}], \Z[\frac{1}{m}]) = \Z[\frac{1}{m}]}$.