7

I am reading the first chapter of Finite Groups by Serre, in which he invokes Goursat's Lemma for subgroups of a direct product $G\times H$ of groups. Using this link and this post by Arturo Magidin I came to an understanding of this lemma, and of why although it seems to be a classification of only subdirect products of $G\times H$, it actually is a classification of all subgroups of $G\times H$.

First of all, I am now trying to test my knowledge on the following elementary practice problem, which is to determine all subgroups of the direct product $C_5\times S_4$.

The only subgroups of $C_5$ are $1$ and $C_5$ itself. An isomorphism from the trivial group (quotient of $C_5$ by itself) needs to go to the trivial group (quotient of $S_4$ by itself), which gives rise by Goursat's Lemma to $C_5\times S_4$ itself. On the other hand, since $\#S_4=2^3\cdot 3$, there don't exist $H,K$ such that $H\lhd K<S_4$ and $|K/H|=5$, so by Goursat's lemma there doesn't exist a subgroup induced by an isomorphism $C_5\stackrel{\sim}{\to}H/K$. What am I missing here? In particular, what is the required isomorphism from the lemma from which I get the trivial subgroup?

The second question is about the application of Goursat's Lemma in Galois theory. I read about it in Serre, but it would be very useful to have a concrete example of it being applied.

enter image description here

Any help is much appreciated.

rae306
  • 9,742
  • 1
    If you can get your hands on the following, have a look at that too: D.D. Anderson and V. Camillo, Subgroups of direct products of groups, ideals and subrings of direct products of rings, and Goursat’s lemma, Rings, modules and representations. International conference on rings and things in honor of Carl Faith and Barbara Osofsky, Zanes- ville, OH, USA, June 15–17, 2007, American Mathematical Society, 2009, pp. 1–12. – the_fox Jan 20 '20 at 18:57

1 Answers1

7

Added. A couple of words: you ran into trouble because you didn’t recognize that you had two ways of getting the trivial subgroup as a quotient.

In principle, the exhaustive/exhausting way of using Goursat’s Lemma to list all possible subgroups of $A\times B$ would be the following:

  1. Find all subgroups of $A$.
  2. For each subgroup $H$ of $A$, find its normal subgroups $N$.
  3. Make a list of the quotients $H/N$.
  4. Repeat with $B$.
  5. Identify pairs, one from each list, of isomorphic subgroups.
  6. List all isomorphisms between such pais.
  7. Each isomorphism listed yields a subgroup.

So here you would start by taking all subgroups of $C_5$, and then list all its quotients. You get: (i) trivial and all of $C_5$ for the subgroup $C_5$; and (ii) trivial for the subgroup $\{e\}$. Then do the same for $S_4$, though the fact that you are only aiming for $C_5$ and the trivial group simplify matters, as done below.


So, as you know, Goursat’s Lemma tells you all subgroups of $C_5\times S_4$ arise from isomophisms of quotients of subgroups of $C_5$ and $S_4$.

So a subgroup of $C_5\times S_4$ corresponds to five pieces of information:

  1. A subgroup $H$ of $C_5$;
  2. A subgroup $K$ of $S_4$;
  3. A normal subgroup $N$ of $H$;
  4. A normal subgroup $M$ of $K$;
  5. An isomorphism $\phi\colon H/M\to K/N$.

The subgroup is then the “graph of $\phi$”, given by $$\{ (x,y)\in C_5\times S_4\mid x\in H, y\in K, \phi(xM)=yN\}.$$

As you note, the only quotients of subgroups of $C_5$ are $C_5$ and $\{1\}$. But there are two ways of “getting” $\{1\}$. One is to take the trivial subgroup and quotient out by itself; the other is to take $C_5$ and quotient out by itself.

Now, every quotient of a subgroups of $S_4$ has order prime to $5$, so your isomorphism will never involve $C_5/\{e\}$ on “the left side”. Since you will always be taking the trivial subgroup on the left, that amounts to looking at any subgroup of $K$ of $S_4$, moding out by itself (that is, $M=K$), and identifying it with the trivial subgroup on the right in either of the two ways of getting it. You will have either the trivial isomorphism $\phi\colon C_5/C_5\to K/K$, or the trivial isomorphism $\phi\colon \{e\}/\{e\} \to K/K$.

So you end up with two types of subgroups:

  1. Those that are obtained by taking $H=C_5$, $N=H$, $K$ a subgroup of $S_4$, and $M=K$. The corresponding subgroup is $$\{ (x,y)\in C_5\times S_4\mid y\in K\} = C_5\times K.$$
  2. Those that are obtained by taking $H=\{e\}$, $N=\{e\}$, $K$ a subgroup of $S_4$, and $M=K$. The corresponding subgroup is $$\{ (x,y)\in C_5\times S_4\mid x=e, y\in K\} = \{e\}\times K.$$

The trivial subgroup is obtained in Type 2, when you take $K=\{e\}=M$.

Here’s two trivial examples in Galois Theory.

  1. Consider the extension $L=\mathbb{Q}(\sqrt{2},\sqrt{3})$ over $\mathbb{Q}$. You have the intermediate extensions $L_1=\mathbb{Q}(\sqrt{2})$, with Galois group $C_2$ over $\mathbb{Q}$, and $L_2=\mathbb{Q}(\sqrt{3})$ with Galois group $C_3$. Thus, the Galois group of $L$ over$\mathbb{Q}$ embedds into $C_2\times C_2$; because $L_1\cap L_2=\mathbb{Q}$, so we get $\mathrm{Gal}(L/\mathbb{Q}) = C_2\times C_2$.

  2. Now consider $L$, the splitting field of $(x^4-2)(x^4-3)$ over $\mathbb{Q}$, with $L_1$ the splitting field of $x^4-2$ and $L_2$ the splitting field of $x^4-3$. Each of them is obtained by first adding $i$ and then adding $\sqrt[4]{r}$, with $r=2$ and $3$, giving you a dihedral group of order $8$. Thus, the Galois group of $L/\mathbb{Q}$ is a subdirect product of $D_4\times D_4$ (with $D_n$ the dihedral group of degree $n$ and order $2n$). In this case, $M=L_1\cap L_2=\mathbb{Q}(i)$, so you do not get the whole direct product. Instead, note that $\mathrm{Gal}(L_i/M)$ is cyclic of order $4$. So $\mathrm{Gal}(L/M) \cong C_4\times C_4$ with $\mathrm{Gal}(M/\mathbb{Q}) \cong C_2$. The group $\mathrm{Gal}(L/\mathbb{Q})$ is a subdirect product of $D_4\times D_4$, given by taking the cyclic group of order $4$ in each copy, and taking the graph of the identity isomorphism of $(D_4/C_4)\times(D_4/C_4)$.

Arturo Magidin
  • 398,050
  • Thank you very much for your answer, it it all clear for me now. I just added a screen of the part in Serre's book about the application to Galois theory. I hope you can look at it later :) – rae306 Jan 20 '20 at 18:44
  • @rae306: Added two examples that come to mind, where the extensions are disjoint and non-disjoint. Don’t know if they will help, though... – Arturo Magidin Jan 20 '20 at 19:30
  • Thank you for the additional comments! I accepted the answer, thanks for your time :) – rae306 Jan 20 '20 at 20:03