D-MATH Algebra I HS 14

Prof. Emmanuel Kowalski

Solutions of exercise sheet 5

The content of the marked exercises (*) should be known for the exam.

1. Let G be a group, and consider the set of maps C(G) = {f : G → C}.

1. Prove that defining (g · f )(x) = f (xg), for every g, x ∈ G and f ∈ C(G) we obtain
an action of G on C(G). Is it faithful?
2. If |G| =
6 1, find a non-trivial invariant subset of C(G).

Solution:

1. First, notice that the association (g, f ) 7→ g · f defines a map G × C(G) → C(G).
Moreover, we can easily check compatibility with the group structure:

((gh) · f )(x) = f (xgh) = (h · f )(xg) = (g · (h · f ))(x), ∀g, h, x ∈ G, f ∈ C(G),

so that (gh) · f = g · (h · f ) for every g, h ∈ G and f ∈ C(G). Finally, the identity

1G ∈ G acts trivially on C(G):

(1G · f )(x) = f (x1G ) = f (x), ∀x ∈ X, ∀f ∈ C(G),

so that 1G · f = f for every f ∈ C(G). This proves that we have a group

action, which can also be seen as a group map % : G → Sym(C(G)) sending
C
g 7→ g · −. Consider the characteristic function χ : G → sending 1G 7→ 1C and
1G 6= g 7→ 0. Then if g ∈ StabG (χ), one has that g · χ(1G ) = χ(1G ) = 1C , where
(g · χ)(1G ) = χ(g), which is one if and only g = 1G , so that StabG (χ) = {1G }.
Hence
\
ker(%) = StabG (f ) ⊆ StabG (χ) = {1G }
f ∈C(G)

Hence the action is faithful, the kernel being trivial.

2. We just need to find a subset Y ⊆ C(G) such that G · Y ⊆ Y . Of course this is
true for Y = C(G)G , the set of invariant elements. This is nothing but the set
of constant functions on G. It is clear by definition that a constant function is
invariant by G. On the other hand, if f is invariant by G, then for every x ∈ G we
have f (x) = (x−1 · f )(x) = f (1), so that f is constant. Since constant functions
C
are parametrized by , we obtained C(G)G is a non-trivial invariant subset of
C(G).

Please turn over!

 Notice that there are many more invariant subsets of C(G). We have that Y ⊆
C(G) if and only if for every f ∈ Y and g ∈ G, one has g · f ∈ Y , that is, if Y
contains only whole orbits of the action of G on C(G). Since functions in the same
orbit have the same image (easy to check), every orbit is completely contained in
the set of maps having image inside the image of the functions in that orbit. Hence
C
if we take T ⊆ and Y = {f ∈ C(G) : Im(f ) ⊆ T }, then we obtain an invariant
subset of C(G).

2. Let G be a group and suppose there is an action of G on a set X. For H ⊆ G, define

X H = {x ∈ X|∀h ∈ H, h · x = x}. Prove: if H E G, then the action of G on X induces
an action of G/H on X H .

Solution:

We consider the following map:

l : G/H × X H → X H
(gH, x) 7→ g · x,

where the dot (·) is the given action of G on X. First, we have to prove that l is a map,
that is, the image of any gH lies in X H and does not depend on the representative g:

• For every g ∈ G, x ∈ X H and h ∈ H, we have that h·(g ·x) = g ·((g −1 hg)·x) = g ·x

as H E G. Hence g · x ∈ X H .
• If gH = g 0 H, then g = g 0 h for h ∈ H. Then g · x = (g 0 h) · x = g 0 · (h · x) = g 0 · g
for every x ∈ X H , so that the image of gH does not depend on the choice of g.

Then the map l is an action of G/H on X H , since l(1G · H, x) = 1G · x = x for every

x ∈ X H , and compatibility is inherited from compatibility of the given action.

3. Let G act transitively on a finite set X, with |X| ≥ 2. Show that there exists at least
one element of g ∈ G such that g has no fixed point.

Solution:
S
We have that g ∈ G has a fixed point if and only if g ∈ y∈X StabG (y). Applying
our solution of Exercise 4 (independent from other exercises) and using the fact that
the
S action is transitive gives that the set of elements in G fixing some point in X is
−1
S
y∈X StabG (y) = g∈G gStabG (x)g , where x ∈ X is fixed. This set cannot be equal
to the whole group G, as seen in class, being StabG (x) a subgroup of G of finite index
|X| by the orbit-stabiliser theorem.

4. Let G be a group acting on X. Show that the stabilizers of two elements in the same
orbit are conjugate. What happens if for x ∈ X one has StabG (x) E G?

See next page!

 Solution:

Let x, y ∈ X lie in the same orbit, that is, y = g · x for some g ∈ G. Then

StabG (y) = {u|u ∈ G : u · y = y} = {u|u ∈ G : ug · x = g · x} =

u0 =g −1 ug
= {u|u ∈ G : g −1 ug ∈ StabG (x)} = {gu0 g −1 |u0 ∈ StabG (x)} =
= gStabG (y)g −1 ,

so that StabG (x) and StabG (y) are conjugate in G (precisely by g), as desired.

If StabG (x) E G, then by definition it coincides with its conjugates, so that all the
elements in the orbit of x have the same stabilizer.

R
5. Consider the group G = GLn ( ), where n is a positive integer, and let H be the
subgroup consisting of diagonal matrices.

1. Suppose that g ∈ H has distinct eigenvalues. Compute CG (g). Try to generalize

this for g ∈ G a (non-necessarily diagonal) diagonalizable matrix with distinct
eigenvalues.
2. Now suppose that n = 2. Compute NG (H) and show that NG (NG (H)) = NG (H).

Solution:

Notation: for u1 , . . . , un ∈ R×, we denote

 
u1 0 ... 0
 .. .. 
 0 u2 . . 
 ∈ H.
diag(u1 , . . . , un ) = 
 .. .. .. 
 . . . 0 
0 ... 0 un

Recall that the centralizer in G of g ∈ G is the set of elements in G commuting with

g. It is easily seen that CG (x) = StabG (x) for every x ∈ G, the action of G on itself
being done by conjugation. We will denote this action with g ∗ h = ghg −1 in order not
to confuse it with the product in G.

1. If g is a diagonal matrix g = diag(λ1 , . . . , λn ) with dinstict eigenvalues λi , we claim

that CG (g) is the subgroup H consisting of diagonal matrices. It is clear that all
diagonal matrices commute with each other, so we only need to prove that for a a
matrix commuting with g we have a ∈ H. To do so, write a = (aij )1≤i,j≤n , where
i is the row index and j is the column index. Then denoting a · g = (bij )1≤i,j≤n
and g · a = (cij )1≤i,j≤n we get bij = λj aij and cij = λi aij and imposing equality
for each index we get (λi − λj )aij = 0. For every i 6= j, we have λi 6= λj by
hypothesis, giving aij = 0. Hence a ∈ H, and we proved CG (g) = H.

Please turn over!

 Now suppose g is a diagonalizable matrix with dinstinct eigenvalues λ1 , . . . , λn .
This means that g = p · d · p−1 , for some p ∈ G and d = diag(λ1 , . . . , λn ) ∈ H.
Then applying Exercise 4 and what we got for the diagonal case, we get

CG (g) = StabG (g) = StabG (p ∗ d) = pStabG (d)p−1 = pHp−1

Hence CG (g) = pHp−1 where p is any matrix diagonalizing g.

2. Suppose that s ∈ NG (H). Then for every d ∈ H we have sds−1 ∈ H, and
since sds−1 has the same eigenvalues as d there are only two possibilities: either
sds−1 = d (i.e., s ∈ CG (d)), or sds−1 is equal to the matrix obtained by 
switching

−1 −1 0 1
the elements in the diagonal of d, that is, sds = ede , where e =
1 0
[indeed, conjugating with the matrix e corresponds to change the basis of 2 byR
switching the basis elements]. But observing that e = e−1 , this second case is
just equivalent to es ∈ CG (d), that is, s ∈ eCG (d). Those two conditions on s are
verified by any element in G when d has equal eigenvalues, and they are satisfied
precisely by the elements of H ∪ eH when d has distinct eigenvalues. This proves
that NG (H) ⊆ H ∪ eH. We claim that we have in fact equality, since for any
h ∈ H one has hH = Hh (as diagonal matrices commute with each others) and
ehH = Heh (which is equivalent to ehHh−1 e = H, i.e., eHe = H, true since
interchanging the two elements in the diagonal gives a bijection of H with itself).
In conclusion, NG (H) = H ∪ eH, that is,

R
    
× 0 a a 0
NG (H) = s ∈ G|∃a, b ∈ :s= or s = .
b 0 0 b

Now we want to check that NG (NG (H)) = NG (H). The inclusion “⊇” is trivial
since NG (H) is a subgroup of G. For the other inclusion, let a ∈ NG (NG(H)).
1 0
Then we can choose h ∈ H with distinct eigenvalues, e.g. h = and
0 −1
the condition a ∈ NG (H ∪ eH) implies aha−1 ∈ H ∪ eH. If aha−1 ∈ H, by the
previous argument we obtain a ∈ NG (H). Else aha−1 ∈ eH, meaning that this
matrix has zero values in the principal diagonal, but same eigenvalues as h, so
that the only two possibilities are aha−1 = eh and aha−1 = he. But those give
both a contradiction, by considering the determinants, as det(aha−1 ) = det(h) 6=
− det(h) = det(eh) = det(he). In conclusion, we have NG (NG (H)) = NG (H) as
desired.

6. Let G be a finite group. Prove that any subgroup of index equal to the smallest prime
dividing |G| is normal. [Hint: Consider an action of G on the coset space with respect
to the subgroup, and find its kernel.]

Solution:

Let G be a finite group, p the minimal prime dividing |G| and H ≤ G subgroup of
index [G : H] = p. It is easily seen that G acts on G/H by left multiplication, so that

See next page!

 we have a group homomorphism

% : G → Sym(G/H).

The action has kernel

ker(%) = {g ∈ G|∀x ∈ G, gxH = xH} = {g ∈ G|∀x ∈ G, x−1 gx ∈ H} =

\
= {g ∈ G|∀x ∈ G, g ∈ xHx−1 } = xHx−1 ⊆ H 6= G.
x∈G

so that the action is not trivial. By the First Isomorphism Theorem, we have |G|/|ker(%)| =
| Im(%)|, so that the image of the action has order dividing |G|. By Lagrange’s Theo-
rem, the image’s order also divides the order of its supergroup Sym(G/H), which is p!.
Hence | Im(%)| divides both |G| (whose factorization contains only primes ≥ q) and p!
(whose factorization contains only primes < p and the prime p once), so that the image
can only contain 1 or p elements. The only possibility is that | Im(%)| = p, being the
action non-trivial. Then [G : ker(%)] = p, and the containment ker(%) ⊆ H (together
with the multiplicativity of the index) gives H = ker(%). In particular, H is a normal
subgroup of G.

7. (*) We want to give a proof of Sylow theorems. Given a prime number p and a finite
group G, we call p-subgroup of G any subgroup of order equal to a power of p. We
call p-Sylow subgroup of a finite group G any subgroup of order equal to the maximal
power of p dividing |G|. (For instance, if G = S4 , then a 2-Sylow subgroup of G is a
subgroup of order 8, and the only 5-Sylow subgroup is {1G }).

1. Let G be a finite group, and write G = pn h, with p a prime number, and n, h

positive integers such that p does not divide h. Consider the set P = {I ⊆ G :
|I| = pn }:
a) Prove that the following defines an action of G on P:

∀g ∈ G, ∀I ∈ P, g · I := gI = {gi|i ∈ I};

b) Prove that p does not divide |P|, and deduce that there exists an orbit O ⊆ P
of the action above whose cardinality is not divisible by p. Deduce that |O|
divides h;
S
c) Prove that S∈O S = G, and deduce from this that |O| ≥ h. Find the cardi-
nality of H = StabG (S0 ), for S0 ∈ O.
Conclude: any finite group G has a p-Sylow subgroup (First Sylow Theorem).
2. Second Sylow Theorem. Let P be a p-Sylow subgroup of G and Q a p-subgroup
of G.
d) Prove that the following defines an action of Q on G/P :

∀q ∈ Q, ∀g ∈ G, q · gP := (qg)P ;

Please turn over!

 e) Prove that the cardinality of any orbit is 1 or is divisible by p. Deduce that
there is a fixed point gP ∈ G/P , and that P contains a conjugate of Q.
Conclude: p-Sylow subgroups of G are conjugate in G (Second Sylow Theorem).
3. Let np be the number of p-Sylow subgroup of G, and P a p-Sylow subgroup of G.
f) Prove that P acts on X := {Q p-Sylow in G} by conjugation;
g) Prove that the action above has precisely one fixed point, and that p divides
the size of the other orbits.
Conclude: p divides np − 1, that is, np ≡ 1 (mod p) (Third Sylow Theorem).

Solution (sketch):

1. a) It is clear that for each g ∈ G the left-translation map g · − : G → G is a

bijection, so that for I ⊆ G we have that |g · I| = |I|, and sets with pn elements
are sent to sets with pn elements. This implies that it makes sense to define
G × P → P via (g, I) 7→ gI. Of course 1G I = I, and (gh)I = g(hI), so that
we have indeed a group action.
b) We have
 n  n −1
pY
p h pn h · · · · (pn h − pn + 1) pn h − i
|P| = = =
pn pn pn − i
i=0

and it is easily seen that each pn h − i is divisible by p as many time as pn − i

is (we say that a number is divisible l times by p if it is divisible by pl but not
by pl+1 ). This implies that p does not divide |P|, and since the orbits form
a partition of P, we have an orbit O whose cardinality is not divisible by p.
Since the cardinality of O divides the cardinality of the group G (why?), |O|
divides h.
S S
c) Fixing S0 ∈ O, one can easily see that S∈O S = g∈G gS0 , and that this
union is the whole G. But the first union consisted of |O| sets of pn elements,
and in order for it to be equal to the whole G we need that |O| ≥ h. This,
together with previous point, gives |O| = h. Then, |StabG (S0 )| = pn (why?),
and this stabilizer is a p-Sylow of G, proving the First Sylow Theorem.
2. d) This is just the restriction of the action G → Sym(G/P ) (see previous Exercise
for H = P ).
e) The cardinality of any orbit divides |Q|, which is a power of p. Hence orbits
with more than 1 point have number of points equal to a multiple (positive
power) of p. As the orbits form a partition of G/P which has h elements,
some orbit must have just one point, meaning that there exists a fixed point.
This means that for some g ∈ G one has QgP = gP , i.e. P ⊇ g −1 Qg. If now
we suppose that Q is also a p-Sylow, then the inclusion above needs to be an
equality (why?) and P and Q are conjugate.
3. f) Since conjugation leaves the cardinality unchanged it makes sense to let G act
on X by conjugation. Then we can restrict the action to the subgroup P

See next page!

g) We have that each orbit has a cardinality dividing |P | = pn , so that X \ X P
is the union of all the orbits with a number of element divisible by p, meaning
that p divides nP − |X P |. Hence it is enough to prove that there is precisely
one fixed point and we can conclude. Clearly, P ∈ X P . Conversely, if Q ∈ X P
then gQg −1 = Q for every g ∈ P , so that P ≤ NG (Q). As Q E NG (Q), and
P and Q are conjugate in NG (Q), we obtain P = Q (why?). Hence the only
fixed point is P .
Notice that for a fixed Sylow subgroup P , we have np = |{gP g −1 }| = [G : NG (P )].
Since P ≤ NG (P ), we have that h = [G : P ] divides np . Moreover, we have that
np = 1 if and only if P coincides with all its conjugates, that is nP E G.