Why partition a group?
In a group , multiplication need not commute. The word same in the next definition matters: we do not invent a new operation on a subset.
Definition 7.3.1
(Subgroup and left coset)
A subgroup is a nonempty subset that is itself a group under the same multiplication of . Equivalently, it contains the identity and is closed under products and inverses.
If , its left coset is .
A left coset need not be a subgroup and it need not equal the right coset . The useful question is whether cosets overlap partially or partition the whole group. We will prove the second statement, then count the parts.
The coset test and the partition
Proposition 7.3.2
(Coset test and partition)
Let be a subgroup of . For , the following are equivalent:
Consequently the distinct left cosets of partition .
Proof
(Overlap forces equality)
The first condition implies the second because a coset is nonempty. If for , multiplying by on the left and by on the right gives . Conversely, if , then and
Here : multiplication by maps into , while every can be written . The order of factors was preserved throughout; no commutativity was used.
Every belongs to , since the identity lies in . The cosets cover ; the test shows that overlapping cosets are equal. Thus the distinct left cosets partition .
The test also settles when a coset is a subgroup. Since contains the identity exactly when , it is a subgroup only for the coset itself.
Common mistake
(Cosets are not subgroups)
In with , the coset does not contain the identity , so it is not a subgroup. Only the coset of an element of is itself; every other coset is merely a translate of .
The next claim is a plausible one that the coset test does not support.
Counterexample mode
Left and right cosets can differ
Claim: for every subgroup and every element . In the symmetric group , with permutations composed from right to left, take and . Then sends , , , so it is and . On the other side sends , , , so it is and . Since , the sets differ. The correct statement is the coset test for left cosets only; the right cosets form their own partition, and the two partitions agree only for special subgroups.
Lagrange's theorem and what it actually says
Definition 7.3.3
(Index and order of an element)
When the left cosets of in are finitely many, their number is the index . For , the order of is the least positive integer with , when one exists.
Theorem 7.3.4
(Lagrange's theorem)
If is finite and is a subgroup, then divides ; in fact .
Proof
(Counting equal-sized cosets)
Each left coset has exactly elements. Indeed is a bijection : surjectivity is the definition of the coset and injectivity follows by multiplying on the left by . If there are distinct left cosets, their disjoint union is , so
The number is the index . Finite counting is used here; the expression is not justified by the same arithmetic when is infinite.
Proof X-ray
Where the counting proof gets its power
The proof has two independent ingredients. The partition comes from the coset test and uses that is a subgroup: closure and inverses are what make . The equal sizes come from the bijection , whose injectivity uses cancellation, that is, the inverse in . Neither step uses commutativity. The key move is that the same number counts every piece, so the count collapses to a product . Finiteness is used last: it makes and ordinary integers, so divisibility has a meaning. For infinite groups the same two ingredients still give a partition into pieces of equal size, but they cannot be turned into divisibility.
Common mistake
(Counting cosets in infinite groups)
The subgroup of has exactly two cosets, the even and the odd integers, so its index is . Yet and are both infinite, and no quotient gives . The identity is a statement about finite counting.
Corollary 7.3.5
(The order of an element divides the group order)
If is finite and , then has an order , the subgroup generated by has exactly elements, and . In particular .
Proof
(The cyclic subgroup generated by an element)
For any in a finite group, some positive power equals the identity: among the powers two agree, and multiplying by an inverse gives for a positive . Let be the least such exponent. The cyclic subgroup generated by consists exactly of : division of an integer exponent by reduces it to one of these powers, and equality of two listed powers would contradict minimality. Therefore this subgroup has elements, and Lagrange gives . Writing gives .
The converse about element orders is false: even when divides , there need not be an element of order . For example, in every nonzero element has order two, so there is no element of order four although the group has four elements. This example does not claim the converse about subgroups in general.
Quick check
(Let . Is ?)
Use the coset test with the additive inverse.
Solution · Answer
In additive notation the condition reads . Here , so the cosets are equal. Directly, .
Quick check
(A group has elements. Can it have a subgroup with elements?)
Apply Lagrange's theorem.
Solution · Answer
No. A subgroup of order would force , but . The possible subgroup orders are among the divisors .
Quick check
(Let be a group of prime order and . Why does generate ?)
The cyclic subgroup generated by has order dividing .
Solution · Answer
The cyclic subgroup has order dividing , so . Since , the subgroup has at least two elements, so and .
Worked examples
Worked example 7.3.6
(Cosets in a cyclic group)
Question. In the additive group , let . Prove is a subgroup, list its cosets and find the index.
Solution. consists of the multiples of ; it contains , sums of its elements remain multiples of , and the negative of is while the negative of is . Thus is a subgroup. Its distinct additive cosets are
These are disjoint and cover all twelve residue classes, so and . In additive notation the identity is and the coset operation is addition, not multiplication.
Worked example 7.3.7
(A divisor need not be an element order)
Question. In under coordinatewise addition, find the order of each element and explain why there is no element of order four.
Solution. The identity has order one. For any nonzero , doubling gives modulo two, while the element itself is not the identity. Its least positive order is therefore two. The four elements have orders ; although , no element has order four. Lagrange's theorem still holds: each of the orders one and two divides four. An element order and a subgroup order are distinct questions.
Worked example 7.3.8
(The smallest and largest subgroups)
Question. In , find the cosets and index of and of .
Solution. For the coset of is , so there are six singleton cosets, and . For every coset equals , since for all . There is a single coset, and . These edge cases are consistent with the formula . As a final check, the element has order three: , and .
Exercises
Exercise 7.3.9
(Let be a finite group with elements and . Prove that , and conclude that .)
CoreProof
Hint(1)
Let be the order of and use Lagrange's theorem for the cyclic subgroup generated by .
Solution(3)
- 1
Let be the order of and count the cyclic subgroup
Key step
The subgroup has elements and is a subgroup of , so by Lagrange's theorem.
Because: The subgroup generated by has exactly elements.
- 2
Write and raise to a power
Choose the integer with . Then
- 3
Multiply by
∎Therefore .