Math 412 Adventure sheet on cosets

D EFINITION : Fix a group G and a subgroup K. A right K-coset of K is any subset of G of the form
K ◦ b = {k ◦ b | k ∈ K}
where b ∈ G. Similarly, a left K-coset of K is any set of the form b ◦ K = {b ◦ k | k ∈ K}.
P ROPOSITION : Fix a group G and a subgroup K. The total number of right K-cosets is equal to the total
number of left K-cosets.
D EFINITION : Fix a group G and a subgroup K. The index of K in G is the total number of distinct right
K-cosets of K in G. We write this index [G : K].
L AG RANGE ’ S T HEOREM : Fix a group G and a subgroup K. Then |G| = |K|[G : K].
D EFINITION : Let a, b ∈ G. We say a is congruent to b modulo K if ab−1 ∈ K.

A. E XAMPLE IN THE G ROUP OF I NTEGERS . Let G = (Z, +) and let K be the subgroup generated by 7.
(1) Verify that K = 7Z = {7k|k ∈ Z}.
(2) Describe the right K-coset K + 0.
(3) Explain why the left/right K-coset containing a is the same as the set [a]7 ⊆ Z.
(4) Find the index [G : K]. Verify LaGrange’s theorem.

Solution.
(1) The elements in this subgroup are the integers that can be obtained by adding or subtracting 7
any number of times, so the multiples of 7.
(2) K + 0 = {7k|k ∈ Z}.
(3) K + a = {7k + a|k ∈ Z} = [a]7 .
(4) [G : K] = 7, and |G| = |H| = ∞, so even though we have some orders that are infinite,
Lagrange’s Theorem still holds!

B. E XAMPLE IN S3 . Consider the subgroup K of S3 generated by (1 2).

(1) List out all the elements of K. What does Lagrange’s Theorem predict about the number of right
cosets of K?
(2) Find the right K-coset Ke. Show that it is the same as the right coset K(1 2).
(3) Find the right coset K(2 3). Show that it is the same as the right coset K(1 2 3).
(4) Find the right coset K(1 3). Show that it is the same as the right coset K(1 3 2).
(5) Write out all the elements of S3 explicitly, grouping them together if they are in the same right
K-coset.
(6) Express S3 as a disjoint union of right K-cosets. How many right K-cosets are there in total?
(7) Verify Lagrange’s Theorem for K ⊆ S3 .

Solution.
(1) Ke = {e, (1 2)} = K (1 2).
(2) K (2 3) = {(2 3), (1 2)(2 3)} = {(2 3), (1 2 3)} = {(1 2 3), (1 2)(1 2 3)} = K (1 2 3).
(3) K (1 3) = {(1 3), (1 2)(1 3)} = {(1 3), (1 3 2)} = {(1 3 2), (1 2)(1 3 1)} = K (1 3 2).
(4) Ke = {e, (1 2)}, K (2 3) = {(2 3), (1 2 3)}, K (1 3) = {(1 3), (1 3 2)}
 (5) S3 = Ke ∪ K(1 2) ∪ K(1 3).
(6) |S3 | = 6 = 3 × 2 = [S3 : K] |K|.

C. R IGHT K- COSETS AND CONGRUENCE MODULO K. Fix a group G and a subgroup K.

(1) Prove that a is congruent to b modulo K if and only if a ∈ Kb. So the set of all elements congruent
to b mod K is precisely the right coset Kb.
(2) Prove that congruence modulo K is an equivalence relation.
(3) Discuss: the concept of right K-coset is the group analog of the concept of congruence class modulo
an ideal for rings.
(4) Show that if b ∈ Ka, then Ka = Kb. Show also that if b ∈ / Ka, then Ka ∩ Kb = ∅. That is, two
cosets are either exactly the same subset of G or they do not overlap at all.

Solution.
(1) If a is congruent to b modulo K, then ab−1 ∈ K, and a = ab−1 b ∈ Kb. On the other hand, if
a ∈ Kb, then a = kb for some k ∈ K. Then ab−1 = k ∈ K.
(2) Reflexive: for any a ∈ G, aa−1 = e ∈ K, so a is congruent to a modulo K.
Symmetric: for any a, b ∈ G, if ab−1 = e ∈ K, then ba−1 = (ab−1 )−1 ∈ K. So if a is
congruent to b modulo K, then b is congruent to a modulo K.
Transitive: suppose that a is congruent to b modulo K and b is congruent to c modulo K. Then
ab , bc−1 ∈ K. Since K is closed for products, ac−1 = (ab−1 )(bc−1 ) ∈ K, so a is congruent to
−1

c modulo K.
(4) Suppose that b ∈ Ka, which we have shown is equivalent to a being congruent to b modulo K.
Given any element g ∈ G, g ∈ K if and only if gab−1 ∈ K (why?). Then
Kb = {kb | k ∈ K} = {(kab−1 )b | k ∈ K} = {ka | k ∈ K} = Ka.
On the other hand, if b ∈ / Ka, then by (1) we know ab−1 ∈ / K, and so for every k1 , k2 ∈ K,
k1 a 6= k2 b, or else we could write ab−1 = k1−1 k2 ∈ K. Therefore, Ka ∩ Kb = ∅.

D. T HE PROOF OF L AGRANGE ’ S T HEOREM. Fix a group G and a subgroup K. Let a, b ∈ G.

(1) Prove that there is a bijection
Ka → Kb
−1
given by right multiplication by a b.
(2) Prove that G is the disjoint union of its distinct right K-cosets, all of which have cardinality |K|.
(3) Prove that if G is finite, then |G| = [G : K]|K|.
(4) Conclude that the order of any subgroup K must divide the order of G.
(5) Conclude that the order of any element in G must divide the order of G.

Solution.
(1) The map Ka → Kb given by right multiplication by a−1 b has inverse Kb → Ka given by right
multiplication by b−1 a. This is easy to check: na 7→ (na)(a−1 b) 7→ (na)(ab−1 )(b−1 a) = na and
nb 7→ (nb)(b−1 a) 7→ (nb)(b−1 a)(a−1 b) = nb so these maps are mutually inverse.
(2) We already know that every element of G is in one coset, so G is the disjoint union of its cosets.
By (1), each coset has the same cardinality as K.
(3) Each coset has |K| elements. so |G| = |K|[G : K].
(4) Lagrange’s Theorem says that |K| divides |G|.
(5) The order of an element g is the same as the order of the cyclic subgroupof G generated by g.
E. L EFT VS RIGHT COSETS. Let G be a group and K be a subgroup of G.
(1) With the notation we used in A, is K + 0 = 0 + K? How about K + a and a + K for some a ∈ Z?
(2) With the notation we used in B, is K(1 2 3) = (1 2 3)K?
(3) T RUE OR FALSE : In an arbitrary group G, for any subgroup K, Kg = gK for all g ∈ K.
(4) T RUE OR FALSE : In an arbitrary abelian group G, for any subgroup K, Kg = gK for all g ∈ K.
(5) T RUE OR FALSE : In an arbitrary group G, every right K-coset is a subgroup of G.

Solution.
(1) Yes! In particular, because this group is abelian.
(2) K (1 2 3) = {(2 3), (1 2 3)} and (1 2 3) K = {(1 2 3), (1 3)}.
(3) False. For a counterexample, consider the subgroup generated by (1 2) in S3 .
(4) True, because g commutes with all the elements in K.
(5) False. In particular, only one of the cosets contains the identity.

F. Fix a subgroup K of a group (G, ◦).

(1) Show that Ke = K = eK.
(2) Show that for any a ∈ G, there is a bijection K −→ Ka.
(3) Prove that |K ◦ a| = |a ◦ K|, even if in general K ◦ a 6= ◦K.
(4) Prove that if G is finite, the number of left K-cosets is the same as the number of right K-cosets.

Solution.
(1) Ke = {ke|k ∈ K} = {ek|k ∈ K} = eK.
(2) The map k 7→ ka is a bijection, with inverse b 7→ ba−1 .
(3) The bijection k 7→ ka shows that |K ◦ a| = |K|. Similarly, there is a bijection between K and
aK.
(4) We have shown that the right K-cosets partition G into subsets of the size |K|; that means there
|G|
must be |K| right K-cosets. Similarly, the left K-cosets partition G into subsets all of size |K|,
|G|
so there must be |K|
left K-cosets.

G. A CAUTIONARY EXAMPLE . Let G be a group and let K be a subgroup. Consider the set G/K of all right
K-cosets. It is tempting to try to define a quotient group as we defined quotient rings. That is, we can try to
define a binary operation ? on G/K by (K ◦ g) ? (K ◦ h) := K(g ◦ h).
(1) Show that in the example of 7Z in Z from A, ? is a well-defined binary operation.
(2) Show that in the example of K = h(1 2)i in S3 as in B, ? is not a well-defined binary operation. In
fact, there is no natural way to induce a quotient group structure on the set of cosets G/K.
(3) For R4 in D4 in A, is ? a well-defined binary operation on the set of right cosets D4 /R4 ? Is
(D4 /R4 , ?) a group?

Solution.
(1) The operation ? is the operation + we have previously defined on Z7 , and we have shown that is
well-defined.
 (2) (1 2 3)(1 2 3) = (1 3 2), so if ? is well-defined we should have K(1 2 3) ? K(1 2 3) = K(1 3 2) 6=
Ke. However, (2 3) ∈ K(1 3 2) as well, and (2 3)(2 3) = e, which should mean that K(1 2 3) ?
K(1 2 3) = Ke.
(3) Yes! We will come up with a better justification for this soon; for now, the best we can do is
check all possible products.

 
1 17
H. A MATRIX EXAMPLE. Consider G = GL2 (R), the subgroup K = SL2 (R), and A = .
0 π
(1) Prove that the right K-coset KA in GL2 (R) is {B ∈ GL2 (R) | det B = π}.
(2) Prove that the left K-coset AK = KA.
(3) Prove that the right K-cosets KC and KD are the same in this case if and only if det C = det D.
(4) What is the index [GL2 (R) : SL2 (R)]?

Solution.
(1) A matrix B is in KA if and only if B is congruent to A modulo K, which means that AB −1 ∈ K.
Equivalently,
1 = det(BA−1 ) = det(B)π −1 ,
which is equivalent to det(B) = π.
(2) A matrix B is in AK if and only if A−1 B ∈ K. Equivalently,
1 = det(A−1 B) = π −1 det(B),
which is equivalent to det(B) = π.
(3) We have shown that KC = KD if and only if C is congruent to D modulo K. So KC = KD
if and only if det(CD−1 ) = 1, or equivalently, by 217, det(C) det(D)−1 = 1.
(4) It’s infinite: there is one coset for each real number.