Motivation
A group can contain many elements that differ in ways irrelevant to a particular question. An integer may matter only through its remainder modulo six; a permutation may matter only through its parity. A quotient makes this loss of information precise. Its elements are whole equivalence classes, and its multiplication must give the same answer however those classes are represented. That requirement explains why normal subgroups appear.
These notes assume elementary set maps, equivalence relations, and integer divisibility. Read products of permutations from right to left. All groups here have an identity, denoted , and multiplication need not commute.
Definitions
Definition 1.1.1
(Groups and subgroups)
A group is a nonempty set with an associative operation, an identity , and an inverse for every . A subset is a subgroup, written , if it is a group under the inherited operation. Equivalently, is nonempty and whenever .
For the criterion, taking gives ; taking gives inverses; replacing by its inverse gives products. Associativity is inherited. Cancellation follows by multiplying an equality by an inverse, so it does not require commutativity.
The cyclic subgroup generated by is . Its size is the order of . If the least positive with exists, Euclidean division of an exponent by shows that are all its elements and are distinct. Otherwise all integer powers are distinct. In particular, implies when is finite.
A left coset is ; a right coset is . The relation when is an equivalence relation, whose class at is . Reflexivity, symmetry, and transitivity follow from the subgroup axioms. This observation proves that left cosets partition ; it does not yet give them a group operation.
A homomorphism satisfies . Its kernel is , and its image is . The homomorphism law forces and . A bijective homomorphism is an isomorphism. Its inverse is automatically a homomorphism: apply the original map to the proposed product identity and use injectivity.
Definition 1.1.2
(Normal subgroup)
A subgroup is normal, written , when for every .
Theorems
Theorem 1.1.3
(Lagrange's theorem and normality)
If is finite and , then , where is the number of left cosets. Thus divides . A subgroup is normal exactly when for every . Every subgroup of index two is normal.
Theorem 1.1.4
(Quotient and first isomorphism theorem)
For , the rule defines a group . The projection is surjective with kernel . For every homomorphism , its kernel is normal and
Proof sketch or proof idea
Despite the heading, the following arguments prove the displayed assertions in full.
Proof
(Counting cosets and comparing the two partitions)
Multiplication by is a bijection , with inverse multiplication by . Counting the disjoint cosets proves Lagrange. Applying it to proves that every element order divides ; the converse, that every divisor occurs as an element order, does not follow.
Multiplying on the right by gives , and reversing that step proves equivalence. For index two, inversion gives a bijection from left cosets to right cosets, sending to , so there are two of each. If , both and equal . If , both are the complement : each is the sole coset other than in its respective partition. Hence for every .
Proof
(Changing representatives inside a normal subgroup)
For quotient multiplication, choose and with . Then
Normality ensures the conjugate lies in , so changing representatives changes no product coset. Associativity descends from ; the identity is and the inverse of is . Conversely, if this multiplication makes the projection a homomorphism, its kernel is , so normality is necessary as well.
Indeed, if , then . This proves ; replacing by proves equality. For the induced isomorphism, equality is equivalent to , hence to . This simultaneously proves well-definedness and injectivity. The map preserves products and is onto the image by definition, completing the proof.
Proof X-ray
The exact role of normality
The quotient proof has two separate tasks. Cosets already partition the set without normality. Normality enters when a changed first representative must pass a second factor: the conjugate must remain in the proposed kernel. After that check, associativity is inherited rather than reproved by enumerating cosets. The same conjugation computation, applied to a kernel, shows that kernels are normal; and the single equivalence carries both well-definedness (read from left to right) and injectivity (read from right to left) of the induced map. The proof identifies a precise obstruction, so it explains both the construction and why an arbitrary subgroup is insufficient.
The theorem explains how to identify an unfamiliar quotient: construct a homomorphism whose kernel is the desired subgroup, then determine its image. Guessing a familiar group with the same size is insufficient; nonisomorphic groups can have the same order.
The correspondence and third isomorphism theorems
Theorem 1.1.5
(Correspondence and third isomorphism theorems)
Let and let be the projection. Then is an inclusion-preserving bijection from the subgroups of containing onto the subgroups of , with inverse . Under it, exactly when , and in that case
Proof
(Pulling subgroups back along the projection)
If , subgroups of correspond bijectively to subgroups of containing . The maps are and . Preimages of subgroups are subgroups. If , then : an element with the same coset as some lies in . Surjectivity gives . These identities prove that the maps are inverse and preserve inclusion.
Normality corresponds too. Conjugating by gives ; thus is normal exactly when every lies in . The reverse inclusion follows by conjugating with . If , the homomorphism , , is well defined, onto, and has kernel . The first isomorphism theorem therefore gives . This is a precise way to say that forgetting information in two stages agrees with forgetting it at once.
Products of subgroups and a second isomorphism
Theorem 1.1.6
(Second isomorphism theorem)
Let and . Then is a subgroup of , , , and
Proof
(Restricting the projection to a subgroup)
There is another useful comparison between a subgroup and a quotient. Let and . The set is a subgroup. Indeed, normality rewrites a product as , which is again in ; an inverse can likewise be rewritten as an element of followed by one of . The subgroup is normal in , being normal in the larger group.
Restrict the projection onto to . Its image is , because , and its kernel is . The first isomorphism theorem proves
This is the second isomorphism theorem. Notice that need not be normal in , and that is normal in because it is the kernel of this restricted map. A theorem's hypotheses should be attached to the groups in which they are used, rather than strengthened unnecessarily.
For a finite group, the same construction counts . One direct justification is to consider the onto map , . Two pairs give the same product exactly when belongs to . Choosing that common element determines each pair in a fiber uniquely, so every fiber has elements. This argument concerns set sizes; it does not claim that the multiplication map from the direct product is a homomorphism.
The largest abelian quotient
Definition 1.1.7
(Commutator subgroup)
Define a commutator by , and let be the subgroup generated by all commutators. To say generated means all finite products of those elements and their inverses, including the empty product.
Conjugating a commutator by gives , so conjugation preserves the generating set. It preserves the generated subgroup as well; using gives equality. Hence is normal.
For any normal subgroup , the cosets and commute exactly when . Thus is abelian exactly when . In particular is abelian, and every homomorphism from to an abelian group kills . Such a homomorphism factors uniquely through the projection onto : define the factor on by its original value at ; well-definedness follows from containment in the kernel, and uniqueness follows from the projection being onto.
This factorization is a universal property: it characterizes what information the quotient retains, beyond merely listing its elements. It is often more useful than a multiplication table. For example, in every commutator lies in since is abelian. The commutator of and is a nonidentity three-cycle, which generates . Therefore , and its largest abelian quotient has order two. This example also shows that proving two subgroups equal requires both containments; finding a single commutator does not complete the argument by itself.
A practical way to start a quotient problem is to write down the original group, the proposed kernel, and the intended target before doing any coset arithmetic. Check that the original operation is the one preserved by the proposed map: an additive group requires a sum identity, whereas a permutation group requires composition. Next prove both containments in the kernel calculation. Showing that every proposed kernel element maps to the identity leaves open the possibility of additional elements in the actual kernel. Finally, calculate the image explicitly; a map into a familiar target may reach only a proper subgroup. The first isomorphism theorem identifies the quotient with that image, not automatically with the whole target. Each step settles a separate question, and none may be skipped merely because a group order looks right.
Concept lensGeometric
Cosets as the level sets of a map
For a homomorphism , the coset is exactly the fiber ; this is the equivalence from the proof. The quotient is therefore the set of level sets of , and the product of two level sets is the level set of the product of their values. For on , treated below, the level sets are the lattice points on the parallel lines , one line for each integer ; adding two cosets adds their levels.
Worked examples
Worked example 1.1.8
(A quotient of a cyclic group)
In the additive group , the subgroup is normal because the group is abelian. Reduction modulo six is onto and has that kernel. Inside the quotient, the element has order three: and neither nor is zero. More generally, in with , has order , since is equivalent to .
Worked example 1.1.9
(A normal subgroup and a subgroup that is not normal)
In , let . It is a cyclic subgroup of order three and index two, so it is normal. Its two-element quotient is cyclic because any group of prime order is generated by each nonidentity element, by Lagrange. By contrast is not normal: conjugating by gives , outside . Consequently the tempting rule is not a quotient-group operation here.
Worked example 1.1.10
(Collapsing two integer coordinates)
Let be , with addition on both groups. It is onto since . Solving gives , so its kernel is . Therefore . The quotient is infinite although one direction has been identified with zero. A coset is determined by the integer , rather than by either coordinate separately.
Common mistakes
Common mistake
(A coset is not a subgroup)
A coset is a set, not usually a subgroup. It contains the identity exactly when it is itself. Nor does mean : it means the two representatives differ by an element of .
Common mistake
(An abelian quotient does not make the group abelian)
The quotient has order two, so it is abelian, yet while . A quotient forgets exactly the information in the kernel, including the failure of commutativity recorded by commutators lying in . The extra hypothesis in the fourth exercise, that the kernel is central, is what restores commutativity when the quotient is cyclic.
Common mistake
(Lagrange's theorem has no general converse)
Lagrange gives a necessary condition on subgroup orders and element orders, not an existence theorem for each divisor. The group has order four, but each of its elements satisfies , so none has order four. At the level of subgroups, the alternating group of order twelve has no subgroup of order six.
Do not exchange and during a quotient calculation. The valid replacement above used conjugation and normality; the next computation shows what happens without normality.
Counterexample mode
Coset multiplication by representatives needs normality
Claim: the rule defines an operation on the left cosets of any subgroup. In , with right-to-left composition, take , so that and both represent the coset . Multiply by . With the representative , the rule gives . With the representative , it gives . These two cosets are disjoint, so the proposed product depends on the representative. The failure is exactly the conjugate in the proof above: with and , we get . The repaired statement requires to be normal.
Summary
Cosets partition a group and make counting possible. Normality is the exact condition that allows those classes to multiply by representatives. Homomorphisms explain quotients through their kernels, while correspondence identifies the subgroup structure that survives a quotient. Each step separates a set-theoretic assertion, an operation-preservation assertion, and a bijectivity assertion; keeping these checks distinct prevents most quotient errors.
Exercises
Quick check
(Prove that every subgroup of a cyclic group is cyclic, covering the infinite case.)
Solution
Write . If it is generated by . Otherwise there is a positive integer with : replace a negative exponent by its opposite if needed; in the finite case positive representatives are available. Choose the least such . For any , division with gives . Minimality forces . Thus . This proof uses least-positive-exponent reasoning and works for infinite cyclic groups too.
Quick check
(Find the kernel, image, and quotient associated with , . First check whether the proposed map is well defined.)
Solution · Answer
The proposed map is well defined because changing by twelve changes by twenty-four, a multiple of eight. Its image is , cyclic of order four. Its kernel consists of modulo twelve, because means . The quotient by this kernel is therefore cyclic of order four, not of order eight.
Quick check
(Let and in . Describe and .)
Solution · Answer
, cyclic of order three. The third isomorphism theorem identifies the final quotient with . Its two elements record whether a class modulo six is even or odd.
Quick check
(If is cyclic and lies in the center , prove that is abelian.)
Solution · Answer
Choose generating the quotient. Write any two elements as and for integers and . Centrality allows all to commute with each other and with , so both product orders give . Thus arbitrary elements commute. Without centrality, shows the assertion fails.