September 22nd, 2026
Notes (PDF)Recall that subgroups can be pushed forward and pulled back by a homomorphism. Indeed, the image and preimage of a subgroup are subgroups. Moreover, the preimage of a normal subgroup is also a normal subgroup, and the image of a normal subgroup is normal as well, if the image is surjective.
Suppose is a surjective homomorphism. Let .
There is a bijective correspondence
given by and .
Suppose , a subgroup of containing , corresponds to under this mapping.
is a normal subgroup of if and only if is a normal subgroup of .
The restriction is a surjective homomorphism with kernel .
.
We first verify the mappings are well-defined and mutual inverses. We know the image and preimage of a subgroup under a homomorphism are always subgroups. Furthermore, for any subgroup , note , meaning . Thus the domains and codomains of the bijection are correct.
To show they are inverses, note that follows immediately from the surjectivity of . Conversely, we show for any subgroup containing . The inclusion is trivial. Let . Then for some . Rewriting yields , meaning . Because , we have , and since , closure forces . Thus , meaning , establishing the bijection.
Suppose and .
We previously proved that the preimage of a normal subgroup is always normal, and the image of a normal subgroup is normal provided the homomorphism is surjective. The result comes immediately.
The restriction is inherently a homomorphism. It is surjective because . The kernel of this restricted map is precisely . Because , this collapses to , and so .
By applying the standard counting formula to the restricted map , we immediately yield .
This proof is mostly straightforward, but some remarks are in order for . In general, the identity is false. If you push a set forward and then pull it back, the set can get bigger because you accidentally grab extra elements that map to the same place. By our previous theorem, however, the set of all elements mapping to is exactly the coset . This is precisely why the assumption is so crucial. By the closure of , multiplying any by elements of keeps you strictly inside . Thus, the entire pulled-back coset lies completely within . The “extra” elements we grabbed were already inside to begin with, ensuring the set does not actually grow.
Let be a group together with a subgroup . We already know that , as a set, is the set of all left cosets of . We ask if can be endowed with some natural operation that can make it into a group. The answer is yes, but we require to be a normal subgroup, for reasons we will see soon.
Suppose is a group together with a normal subgroup . For any two cosets , define coset multiplication by
Then forms a group under this operation.
Note that is not a subgroup of , which is simultaneously obvious and surprising. It is not immediately clear that this forms a well-defined group operation. Indeed, since if , we need to show that the choice of representative for the coset does not alter the result of multiplication.
Let be a subgroup of . The operation is well-defined on the set of left cosets if and only if is a normal subgroup of .
We show that if is normal, the operation is strictly independent of the chosen representatives. Suppose and . We wish to show that .
By the equivalence class properties of cosets, we can write and for some . Then note
Because is normal, its left and right cosets coincide, meaning . Thus, for our element , there exists some such that . Substituting this yields
Since is a subgroup and is closed under multiplication, . Thus can be written as multiplied by an element of , which immediately implies . Consequently, .
If were not normal, the element could permanently escape , shifting the product into a completely different coset and causing the entire operation to be ill-defined.
Thus the subgroup for which quotienting by yields a group is precisely a normal subgroup. Having established that the operation is completely independent of the chosen representatives when is normal, we may proceed to quickly verify the standard group axioms.
Let be a normal subgroup of . The set of left cosets forms a group under coset multiplication.
Closure. Since , their product by closure of the parent group, and so is indeed a valid coset in .
Associativity. For any , note that
Because the operation in is inherently associative, . Thus
satisfying associativity.
Identity. The coset of the identity, , serves as the identity element. For any , we have
Inverse. The inverse of is precisely . Indeed,
and symmetrically .
Earlier, we proved that the kernel of any homomorphism is inherently a normal subgroup. We now resolve the natural converse: is every normal subgroup the kernel of some homomorphism? Having constructed the quotient group , the answer is yes.
Let be a group and a normal subgroup of . We define the canonical projection map by
The canonical projection map is a surjective homomorphism with .
First, we show is a homomorphism. For any , the definition of coset multiplication yields
and so preserves the group structure. Surjectivity is trivial, as any coset is immediately mapped to by . Finally, we determine the kernel. The identity element in is the coset . Thus, if and only if , meaning . This occurs if and only if . Therefore, .
Thus, a subgroup is normal if and only if it is the kernel of a homomorphism.
Having classically constructed the quotient group by defining operations on cosets, we now take a step back. The quotient construction is not something exclusive to groups; it is a fundamental, structural mechanism that works with several constructions (sets, rings, topological spaces, and more). To truly understand what a quotient is, we must examine it by working with abstract objects, not simply groups. We start with sets.
Let and be arbitrary sets, and let be an equivalence relation on . Recall that the equivalence classes of partition the set , and we denote the equivalence class of an element by . We can define the projection map by .
Suppose we have a function . We ask if we can push this function down through the quotient to define a new map on . For this to make sense, the original function must be blind to the differences between equivalent elements.
Let be a function and be an equivalence relation on . We say that respects the equivalence relation if is constant on equivalence classes. That is, for all :
If respects the equivalence relation, it guarantees that acts on entire equivalence classes uniformly, allowing for a powerful factorization of the action of .
Let be an equivalence relation on a set , and let be the canonical projection map. For any set and any function that respects the equivalence relation , there exists a unique function such that .
Equivalently, the following diagram commutes:
We must establish both the existence and the uniqueness of .
Existence. We define directly on the equivalence classes in by setting . Because elements of are sets (equivalence classes) rather than individual elements of , we must verify this function is well-defined. Suppose . This implies . Because we assumed respects the equivalence relation, , and therefore . The definition is independent of the chosen representative, so is a well-defined function. Furthermore, for any , we have , ensuring the diagram commutes.
Uniqueness. Suppose there is another function such that . We wish to show . Let be an arbitrary element of . Because is inherently surjective, for some . Then . But we already know . Thus for all classes in the quotient, forcing .
The universal property of quotients gives us a powerful tool to mechanically “repair” arbitrary functions. Let be any generic set function. Generally, fails to be a bijection (an isomorphism in the category of sets) for two reasons. First, it may fail surjectivity (missing elements in ). Second, it may fail injectivity (collapsing distinct elements of to the same target).
We can resolve both of these flaws. First, we fix surjectivity by restricting our codomain to the image, . The map from to is trivially surjective, and we can seamlessly include back into via an injective inclusion map .
Second, we fix injectivity by defining a natural equivalence relation on induced by the function itself. Define if and only if . By definition, trivially respects the equivalence relation . Quotienting by “glues” all elements that map to the same target into a single equivalence class. Immediately, we see that the universal property of quotients yields a unique map .
Because we restricted the codomain to , is surjective. Because we quotiented out the exact elements that caused collisions, is strictly injective. Thus, is a forced bijection.
Any function can be uniquely factored as the composition of a surjection, a bijection, and an injection. This factorization, , is captured by the following commutative diagram:
where is the surjective projection map, is a natural bijection, and is the injective inclusion map.
Any function is secretly just a bijection wrapped in layers of redundancy (collapsing elements) and deficiency (missing the target). In our canonical decomposition diagram, the first part of the sequence (the projection ) acts as the quotient fix to resolve injectivity issues, while the last part (the inclusion ) acts as the embedding fix to resolve surjectivity issues. By quotienting the domain and restricting the codomain, we strip away these flaws to isolate , a faithful representation capturing the core mechanism as to how operates.
We now transport this idea into the category of groups. Previously, we classically constructed the quotient group by defining coset multiplication, proving that it forms a group if and only if is a normal subgroup of . Instead of building the quotient from the ground up and verifying it works, we can equivalently define the quotient entirely by demanding it satisfies a universal property analogous to the one in , and then locate the object that fits the description.
Let be a group and a normal subgroup of . The quotient of by is a group together with a surjective homomorphism with , satisfying the following universal property.
For any group and any homomorphism where , there exists a unique homomorphism such that .
One might wonder why we still constrain to be a normal subgroup in a supposedly pure, structural definition. Because we explicitly demand to be a homomorphism with , and the kernel of any homomorphism is inherently normal, the rigid structure of groups strictly prohibits non-normal subgroups from ever acting as kernels. Thus, in the categorical lens, a normal subgroup is not an arbitrary set-theoretic restriction, but simply the name we give to a subgroup capable of being a kernel for some homomorphism.
Furthermore, the condition is exactly the group-theoretic translation of a set function respecting an equivalence relation. If elements in get crushed to the identity by , then is completely blind to differences between elements that differ by an element of .
Before we confirm our classical coset space fits this definition, we must show that this universal property is actually rigid. By a standard diagram chase, any two objects satisfying this property are uniquely isomorphic, meaning the quotient is structurally unique.
Let and be two quotients of by satisfying the universal property. Then there exists a unique isomorphism such that .
Since satisfies the universal property and is a homomorphism with (trivially), there exists a unique homomorphism such that .
Symmetrically, since satisfies the universal property and is a homomorphism with , there exists a unique homomorphism such that .
Substituting the first identity into the second yields , which we reassociate as .
Now, consider the universal property of applied to the homomorphism . We seek a unique map such that . The identity map trivially satisfies this. However, we just showed that also satisfies this. By the strict uniqueness guaranteed by the universal property, we are forced to conclude .
By an identical symmetric argument, . Thus is a bijection, and since it is inherently a homomorphism, it is an isomorphism. This establishes that .
We now confirm that our classically constructed set of cosets is indeed the exact object requested by this universal property.
The classically constructed coset space equipped with coset multiplication , paired with the canonical projection , satisfies the universal property of the quotient group.
Let be a homomorphism such that . We must show there exists a unique homomorphism making the diagram commute.
Well-definedness. We propose . Because we are defining this on cosets (equivalence classes), we must check it is strictly independent of the representative. Suppose . This means . Because , this implies . Since is a homomorphism, , meaning . Thus , so is well-defined.
Homomorphism. We verify that respects the group operation. Indeed, note
Thus is a valid group homomorphism. By construction, , ensuring the diagram commutes.
Uniqueness. Suppose is another homomorphism satisfying . For any coset , we have . But . Since and agree on all elements, .
We conclude by applying our canonical decomposition setup from to group homomorphisms. Let be an arbitrary homomorphism. Like any general function, it may fail to be an isomorphism due to a lack of injectivity or a lack of surjectivity.
We resolve surjectivity by mapping onto the image , which is a valid subgroup of . We resolve injectivity by defining the equivalence relation . In the language of groups, implies . Thus, the equivalence classes are exactly the left cosets .
Because the kernel of any homomorphism is inherently a normal subgroup of , the quotient space naturally forms a group. By its very definition, respects the equivalence relation generated by its own kernel. The universal property immediately gives us a unique, well-defined homomorphism .
Because we quotiented by the exact subgroup causing injectivity collisions, is injective. Because we restricted the codomain, is surjective. Therefore, is an isomorphism.
The First Isomorphism Theorem is not a standalone group-theoretic trick. It is the realization of the universal canonical decomposition of maps in an arbitrary category, simply applied to . Injectivity is factored out by quotienting by the kernel, surjectivity is factored out by restricting to the image, and we are left with , the underlying isomorphism that faithfully represents .
Let be a group homomorphism. Then induces a canonical isomorphism
given by the map .