Homework #7, due March 18th
1. If A,B are normal in G and have intersection {1}, show they commute.

Ans. We want to know that for all a in A and b in B, that ab=ba. All we're given is a way of guaranteeing that some group element is 1. So let's massage our equation into something being 1:
a b a^-1 b^-1 = 1, that's what we want.
If we group as (a b a^-1) b^-1, we see that this is an element of B.
If we group as a (b a^-1 b^-1), we see that this is an element of A.
Therefore it's 1.


2. If H is a subgroup of G, what is the kernel of the action of G on G/H?

By definition, it's those {g in G : gC = C for all C in G/H}.
Unwinding the definition of G/H, it's {g in G : gkH = kH for all k in G}.
We can rewrite this as [k^-1 g k H = H]. But this is the same as k^-1 g k in H.
So now it looks like {g in G : k^-1 g k in H for all k in G},
or {g in G : g in k H k^-1, for all k in G}.
That's the intersection over all k in G, of k H k^-1.

So what? Why should this be preferable, as an answer, to our first line?

  • This doesn't have as many quantifiers in it -- one, instead of two or three.
  • If H is normal, then this formula makes it obvious that the kernel is H. (If H isn't normal, then the kernel isn't H -- kernels are always normal!) That's not nearly so transparent from the first formula.

  • 4.3 #5. The size of the conjugacy class through b is the number of cosets of G by the centralizer of b. (This is just our usual |orbit| = |G/stabilizer|, applied to the conjugation action.) Since the centralizer contains the center (follow the definitions), the partition of G by cosets of the center -- into n parts -- is a refinement of the partition of G by cosets of the centralizer. So the conjugacy class has at most n elements.
    4.3 #8. Let pi be a nonidentity permutation (a b ...) (...) ... (...), i.e. the first cycle listed has length > 1. Let c be different from a and b = pi(a), which uses n at least 3. We now claim that pi does not commute with (b c).

    Proof: pi (b c) a = pi a = b, whereas (b c) pi a = (b c) b = c.

    Note that we need c different from b to learn that these two answers are different, and c different from a to learn that (b c) a = a. Anyway, the upshot is that we have a witness (b c) to the fac that pi doesn't commute with everyone. So it's not in the center. Only the identity is.


    4.3 #10. sigma = (12345), sigma^2 = (13524), sigma^-1 = (15432), sigma^-2 = (14253).
    Find taus conjugating sigma to these powers. All you have to do is match up elements.
  • e.g. (12345) -> (13524), use (1->1, 2->3, 3->5, 4->2, 5->4), so tau = (2354)
  • tau = (25)(34)
  • tau = (2453)

  • 4.3 #13. If G has two conjugacy classes, then one of them is the identity, and the other is of size n-1 (where G is size n). But then n-1 | n. That forces n=2. So G is necessarily of order 2, and every group of order 2 has two conjugacy classes.

    If you're disappointed by this answer, you should try three conjugacy classes, and show the only groups are of size 3 and 6. I don't know the answer for 4. I also don't know if the original question has more infinite examples.


    4.3 #35. For a permutation to be of order k, the lcm of the cycle lengths should be k. If k is a prime p, this means that all the cycles are of length 1 or p. So our permutation is either a p-cycle and all 1s, or two p-cycles, or three, or ... up to floor(n/p), the maximum number of p-cycles we could cram in n elements. That's the greatest integer less than or equal to n/p.