Last time we considered how you can sometimes define an operation out of the set of cosets. The group is called a factor group or quotient group. We will write this as \(G/H\) and say β\(G\) mod \(H\text{,}\)β where \(G\) is the larger group and \(H\) is the subgroup. When this worked, the operation was defined as \(aHbH = abH\text{.}\) But it didnβt always work. Today, we will explore the properties of subgroups \(H\) that make this work.
Those subgroups \(H\) that work will be called normal subgroups. But thatβs not how we define normal subgroup. We define it as a property of the subgroup, and then notice that this is exactly the property that allows us to properly well-define the coset operation.
Our definition is: a subgroup \(N\) of a group \(G\) is normal provided \(aN = Na\) for all \(a \in G\text{.}\) That is, left cosets give the same partition as right cosets.
What if \(G\) is abelian? Well, then left cosets will be exactly the same as right cosets (even element by element). For non-abelian groups, we can sometimes have left-cosets be right-cosets, as in \(H = \{(1), (123), (132)\}\text{.}\)
The previous example suggests that any time \(H\) has index 2 in \(G\) (i.e., there are exactly two cosets), we will have \(H\) normal in \(G\text{.}\) You are asked to prove this as a bonus on your homework.
Deciding whether a subgroup is normal is important, and we will want to prove that subgroups are normal. What can we do if \(G\) is not abelian and \([G : H] \ne 2\text{?}\) It would be nice to have an easier test than writing out all the left and right cosets.
Here is a clue: what does it mean to say \(gN = Ng\text{?}\) It does not mean that \(gn = ng\) for all \(g \in G\text{,}\)\(n \in N\text{.}\) Rather, we have \(gn = n'g\text{.}\) But this means that \(gng\inv = n'\text{.}\)
So to show that \(N\) is a normal subgroup of \(G\text{,}\) we just need to show that every conjugate of every element of \(N\) is still in \(n\text{.}\) In other words, that \(N\) is closed under conjugates.
Important: the conjugates of \(n\) are the result of conjugating by all elements of \(G\text{.}\) It doesnβt help at all to check conjugates by elements just from \(N\) (these will always be in \(N\text{,}\) right?).
To prove that \(Z(G)\) is normal, we will show that it is closed under conjugates. So let \(n \in Z(G)\) and \(g \in G\text{.}\) What can we say about \(gng\inv\text{?}\)
Well, since \(g \in G\) and \(n \in Z(G)\text{,}\) we have \(gn = ng\text{.}\) So \(gng\inv = ngg\inv = n\text{.}\) So indeed, \(gng\inv \in Z(G)\text{.}\)
What is really going on? We need every element of \(aHbH\) to be an element of \(abH\text{.}\) The elements of \(aH bH\) have the form \(ah_1bh_2\text{.}\) Now we can use associativity to think of this as \(a(h_1b)h_2\text{.}\)
If there is time, consider how the factor group of rotations in \(D_4\) shows us the coarse structure of the group \(D_4\text{.}\) Or how \(\Z/2\Z\) shows us the structure of evens and odds.