Group Homomorphisms


Besides studying groups by decomposing them into substructures, it's also possible to study group by building a special map between two groups that preserves some group structure.

Consider the homogeneous matrix group under multiplication and the non-zero real number under multiplication . It's possible to draw a map defined by , and there's when , which means the elements' multiplication is preserved under such map. And such map is common in mathematical world.

In this text, you will soon find out the possibility of drawing a map between two groups reveals some non-trivial structural similarities between them, which can be utilized as a powerful tool to create solutions and proofs to group theory problems.

Group homomorphisms

In group theory, among possible maps between elements of two groups and , we're especially interested in maps such that holds. Such maps are called group homomorphisms from to .

Since a group homomorphism is a map in the first place, when working with a group homomorphism, you'll have to keep its map instincts in mind, and consider whether it has injectivity and surjectivity.

The most ideal case is that group homomorphism is injective and surjective, in which case the elements are just renamed correspondingly between two groups. So you will see there's a one-to-one correspondence between that , so the group structure is kept identical. Such group homomorphism is called a group isomorphism. We also denote such relation between two isomorphic groups as .

However, in order to get anywhere further, we'll have to do some algebraic reasoning on the homomorphism.

Element constrains

Although a map in the first place, a group homomorphism is constrained by the group structure of and , and the requirement of . So a group homomorphism is not on the loose and should have some part of it restrained.

Let's substitute the identity of in, we get . So is the element that fulfils in group and is the identity of , which means a group homomorphism always map identity to identity.

Let's select , we get . So , which means a group homomorphism should always map the inverse element to a image that is inverse of the element's image.

Let's select , and we will get , , which means the binary operation of image of 's element is still in the image of . Noted that it won't be hard to show and , so by the element constrain of a subgroup .

This is also applicable to subgroup of , by trimming the domain and codomain of to , and repeat the reasoning of .

So when the group homomorphism is injective, noted that there's a bijection between and . And since is a subgroup of , we will get .

Other group homomorphism might seem wild since there's not yet been applicable properties to restrain their behaviour. But we will see how these "wild" group homomorphisms are constrained generally under the first isomorphism theorem.

First isomorphism theorem

Will it sound a little bit crazy when the structure of the homomorphically mapped group completely depends on the kernel of the homomorphism on the group, rather than concrete map definition of the homomorphism? Or more precisely, the relation between the kernel and the image is defined by:

The statement leads to possible homomorphisms between and completely depends on the structure of and . And if or is finite, the amount of possible homomorphism is also finite.

This sounds even more crazy, but once you've got the point of its proof, you will get a step closer to the beauty of abstraction and symmetry.

First, let's show that . For the identity, obviously so . For the inverse element, noted that and , so and . Finally given two element , , so and so does . So there's in the first place.

Second, let's show that . , , so , and there's .

Finally, since , there exists a quotient group , and we are about to show , which means we need to find a group homomorphism which is an isomorphism. It won't be hard to find that , so let's begin with .

To show it's group homomorphism, noted that , consider , so is indeed a group homomorphism.

To show it's injective, let's choose some coset such that while . Multiplying by yields , so and , which is a contradiction to .

To show it's surjective, let's assume there's some such that . Noted that so it is a contradiction to such coset's non-existence.

By now, we've shown there exists a group isomorphism defined by , which concludes our proof of .

The first isomorphism theorem of group shows some extra symmetries of group homomorphism constrained by the group structure: the group homomorphism uses the kernel on the group as the mould, putting together every elements (based on the coset formed by , of course) and forges them into a pinpoint in .

So if a group homomorphism wants to be injective, it must only put together exactly one element in each coset, so , in which case the only possibility is .

More specifically, consider the normal subgroup , the map defined by is group homomorphism since , the kernel of which is since . Such group homomorphism is also called the natural homomorphism or canonical homomorphism of . Under the first isomorphism theorem, all group homomorphisms such that is equivalent to the natural homomorphism of .

One must take extra attention to a pitfall that merely having does not necessarily mean is bijective. You might have seen examples of infinite sets mapping to itself: given non-zero complex number , which forms a group of multiplication. Consider defined by , its codomain is still since the range of remains unchanged, while , which means is not bijective.

Direct products

Given two group and , their external direct product is defined as group such that . Let's prove is a group:

  1. (Closed) Noted that and , so .
  2. (Identity) Consider , obviously , so is the identity.
  3. (Associative) , , so binary operation is associative.
  4. (Invertible) , , so binary operation is invertible.

On the other hand, consider another group constructed in similar way as above. Obviously there exists an group isomorphism defined by . In this way there's . So when we decompose a group into external direct product of a set of groups, their order does not matter.

Given two subgroup of , define , when there're and , their internal direct product exists, is group , and :

  1. Since , by definition .
  2. , there exists which is the inverse element for .
  3. , .

When is an external direct product and is an internal direct product, consider a map defined by . Since , it is indeed a group homomorphism. Consider the kernel , assume there exists some , due to there's , and since is a group, there's , similarily there's . Which means and is a contradiction to . So , and according to the first isomorphism theorem, is a group isomorphism, and the connection between external and internal direct product can be restated as:

Conversely, how do we find the internal direct product when we know ? Denote the associated group isomorphism as , obviously there're and . Let's extract some properties from them:

  1. Since , .
  2. Since , and . Similarily we have , and , so .
  3. Obviously , .

So obviously by utilizing and we can conveniently find the internal direct product corresponding to an external direct product. This shows us the connection and unity between the two types of direct products.

Someone might be lazily denoting the existence of the internal direct product as . However here comes another common pitfall that being does not necessarily mean is an internal direct product of , given that it is not necessarily the case 1 . So we will always claim that " is the internal direct produt of " instead of flawed description "". Again, it is time to ring an alarm that clarifying which group isomorphism the current context is talking about is at uttermost priority.

Factorizing a group into direct product is an even stronger decomposition of group structure than normal subgroup and quotient group decomposition: Elements from different internal direct product component are commutative, and operations on them do not affect each other, which enables us to study the behaviours of its every direct product components first and them combine them together to describe the group's behaviour.

Second isomorphism theorem

As you might have noticed, when we slack the constrain for to be direct product of that , to , though no longer a direct product, remains to be a group. This is done by showing :

  1. Since , by definition .
  2. , consider its inverse element , by definition , so .
  3. , , .

So and is a group, moreover, it is obvious that and exists. However, defined by is not even a homomorphism, and is not a direct product.

From the example above, it won't be hard to find out that a group has direct product decomposition is just an extremely ideal case of a group has normal subgroup decomposition . Is it possible to generalize both cases in one formula?

To do so, consider their intersection , let's show :

  1. Since , by definition .
  2. , since is group, so that , similarily , so .
  3. , since is group, , similarily , so .

Since and is a group, we've proved that and is also a group. What's more, consider any , since , by the closed property , and since , by the normal property of , and . Now we have another associated quotient group .

Consider a map defined by , since , , so is a group homomorphism. Noted that iff , and by group 's closed property there will also be , so . By first isomorphism theorem:

The second isomorphism theorem, which is the statement we've deduced above, show the relationship between the product with a normal subgroup , and the normalizing intersection with the normal subgroup. As you might have noticed, the theorem is in the form of .

Consider the normal subgroup of , where . Given that , apply the second isomorphism theorem and we get . However, is not a direct product of the subgroups since , which means the binary operation between elements from the two subgroups are not commutative.

Consider a special case that . Apply the second isomorphism theorem and we get both and . Assume there's , right multiply by and we get , by the condition that there must be . And since , , there's . So and is the internal direct product of .

Conversely, when is the internal direct product of , there's , and thus implies and .

It won't be hard to show when at least one of is finite 2, combined with are equivalent to is internal direct product of . Let's assume is finite, is only possible when , and thus falls back to the special case we've discussed above.

Third isomorphism theorem

Given a normal subgroup , consider some (not necessarily normal) subgroup , it is intuitive that . Conversely, consider some subgroup , it is also intuitive that . However, intuition without rigid proof is the root of all mistakes, so let's begin with drawing some proof.

For any and , let's show that :

  1. (Identity) .
  2. (Inverse) .
  3. (Closed) .

For any subgroup , let's show that :

  1. (Identity) .
  2. (Inverse) , such that .
  3. (Closed) .

Finally, we would like to show the map (not a homomorphism) is bijective. Since by that , the map should be surjective, all remaining to be done is to show is injective.

Assume there're such that . For each , , so and , conversely we can get and . Which concludes our proof of 's injectivity.

So we've proved the map is bijective, when there's a normal subgroup partitioning the group into quotient group , each super group uniquely corresponds to a subgroup . This is also called the correspondence theorem.

More specifically, consider the case that and , consider the map ( must be on the right since might have more elements and might be smaller) defined by , since , is a group homomorphism. And obviously , is the kernel of such homomorphism. By applying the first isomorphism theorem we get:

The third isomorphism theorem, which we've proved above, alongside with the correspondence theorem we've proved earlier, describe the symmetry of quotient group hierarchy over the same normal subgroup.

Conclusion

In this text, we introduce the concept of group homomorphism, isomorphism and direct product, alongside with the three isomorphism theorems.

The first isomorphism theorem tells us that it is only possible to build homomorphisms from group to when there's some quotient group isomorphic to some subgroup . And all group homomorphism is fundamentally equivalent to their kernel's natural homomorphism, where their kernel is some normal subgroup of the group.

The second isomorphism theorem is a generalization of group decomposition into quotient group with normal subgroup or direct products. Specifically, it states will become a direct product when and normalize each other and their intersection .

The third isomorphism theorem, alongside with the correspondence theorem, describe the symmetry of quotient group hierarchy over the same normal subgroup.

The group homomorphism, isomorphism and direct product arm us with another powerful tool for studying group theory, and their intrinsic connection with normal subgroup and quotient group is a show-off of how symmetry of group will constrain and affect other components in mathematical world.


  1. For such kind of non-corresponding counterexample, see also https://math.stackexchange.com/questions/4644058. [return]
  2. It is false when both are infinite, see also https://math.stackexchange.com/questions/4644058. [return]
March 4, 2023