Group Actions


A group action is a binary operation involving a group and a set, resulting in another element in that set, where consecutive group actions are identical to associating the group elements first, and then letting it act on the set element.

Consider how one can compute the a term in Fibonacci sequence through , which can be rewritten as . The complexity of computing in Fibonacci sequence in this way is if one repeat such multiplication from up to . But just as you know, if we factorize in binary and compute , , the complexity falls to drastically. While this is due to the associativity of matrix multiplication, it can also be seen as a group action of acting on , and that we are utilizing associativity of instead of letting act on step by step.

Another example of group action we've seen is permutation group's acting on polynomials, where we utilize the normality of the permutation group to construct polynomials that fit into solution of algebraic equations, analyzing the solvability of equation of certain order.

While these examples of group action are somewhat dedicated to scenarios, in this text, we are going to see what's the general view of group action in group theory, and how we can turn group action into a powerful tool for studying group theory, deriving some most fundamental theorems in group theory: Cauchy's theorem and the Sylow theorems.

Group actions

For a group and any set , a (left) group action on is a binary operator such that 1. Just like all maps will have properties of group homomorphism once it fulfils , all binary operators will have properties of group actions once they fulfil the criteria we showed above.

Orbits and stabilizer subgroups

The orbit of is the set of the elements reachable from 's group actions on . Each element in the orbit is called a point in the orbit.

's group actions partition into disjoint orbits: consider equivalence relation defined by :

  1. (Reflexive) .
  2. (Symmetric) .
  3. (Transitive) .

And obviously s are the elements of quotient set .

An element is said to be fixed by if . What's more, all elements fixing forms a subgroup :

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

Such a subgroup is called a stabilizer subgroup of .

For each point , it won't be hard to see that , since .

Orbit-stabilizer theorem

We know that contains elements fixing , so outside lies element that moves . Consider a left coset , obviously there's must be some point distinct from , and is one-to-one correspondence between left cosets of and points in orbit . It won't be hard to see .

Such correspondence relationship, alongside with its derived numberic relationship, is called the orbit-stabilizer theorem.

When a point is fixed by the entire , there's and the orbit for is trivial. Let's put all elements with trivial orbit together into group , for all non-trivial orbits , we get the equation of group action:

The class equation

While can be any set that group can act on, there's nothing forbidding us to put in the place of . In fact, doing so turns group action into one of the most powerful tool for studying group theory.

Center and centralizer subgroups

Consider a binary operation , it's a group action since and is called the conjugation. Two elements of are said to be conjugate when they lies in the same orbit. Each distinct non-trivial orbit is also called a conjugacy class.

The center of a group, which is denoted as , is the set of fixed points under conjugation. Specially, there's :

  1. .
  2. .
  3. .

Noted that means , is also the set of elements which commute with any other element in the group. And since , there's .

While is a normal subgroup rather than some arbitrary set, we can tie constrains of normal groups to it.

Similarily, for each , we the stablizer subgroup for under conjugation as , and call it the centralizer subgroup. It's obvious that binary operations of elements in and elements in conjugacy class corresponding to commute.

For distinct centralizer subgroups corresponding to non-trivial conjugacy classes, which are proper subgroups of , we obtain the class equation that:

The class equation imposes numberic constrains on the size of the center of a group (which is a normal subgroup) and the number of elements in a group under conjugation.

Groups of prime-power order have non-trivial center

For a group with , is prime, we would like to show its center is not trivial.

First, for each centralizer subgroup , if exists, since it's a proper subgroup, . If there's no centralizer subgroup, . Either case we turn the class equation into a congruence relation that .

Then, by , is divisible by , it's impossible for and is not trivial. Up to this meaning, the case of groups of order is special and such a group is sometimes called a -group.

Groups of prime-squared order are abelian

For a group with , is prime, there's either or .

When , the entire group is abelian and we are done.

Assume , since has prime order, there's . And for any , there's , which means any binary operation in commutes rather than just those with , and is a contradiction.

So when , we have and the group is abelian.

Cauchy's theorem

While in the text of cyclic group, we've proved a special case of Cauchy's theorem applied to finite abelian groups, and in this text, we will prove the complete version of it using the class equation.

The Cauchy's theorem states that given a prime and a group , if is divisible by , then there must be such that .

First, if there exists some centralizer subgroup whose order is divisible by , by , we continue the search in .

Second, none of the centralizer subgroup has order divisible by , for each centralizer subgroup , , and we can turn the class equation into a congrence relation that , so does the case of no centralizer subgroup. Since divides and is a finite abelian group, by applying Cauchy's theorem of finite abelian group we conclude the proof.

A group is called a -group if any element in the group has order divisible by prime . And by the Cauchy's theorem, if a group's order is divisible by prime , there must be a subgroup of order which is cyclic and there's element of order . So a group is a -group iff its order is .

The Sylow Theorems

Following the proof of the Cauchy's theorem and based on the class equation, Sylow made some development and came up with some subgroup existence theorems which are very useful for studying group theory.

First Sylow theorem

The first Sylow theorem moves a step further from the Cauchy's theorem that if is divisible by , then there must be such that .

The proof is done by mathematical induction: while proving the first Sylow theorem parameterized by , we assume the theorem is true when parameterized by . The case of is true by the Cauchy's theorem.

First, if there's some centralizer subgroup with order divisible by , by , we continue the search in .

Second, if none of the centralizer subgroup has order divisible by , then for every centralizer subgroup , there's , and thus , so does the case of no centralizer subgroup. By Cauchy's theorem and 's being prime, there's some subgroup . Since 's binary operations with commute and so do 's, there's . Consider the quotient group , since is divisible by , is divisible by . By applying the first Sylow theorem parameterized by and correspondence theorem, we get some subgroup of quotient group of order , where is of order and is the subgroup we are seeking for.

So for a group such that , we are able to find a subgroup such that , which is maximal in and called the Sylow -subgroup.

Does it mean we can construct subgroup of any size using the first Sylow theorem? You see, starting from a specific size dividing where are primes, by taking and , since , does it mean either or are the subgroup of size we are seeking for? The answer is no, since is not necessarily a group, there's nothing constraining to be still in . Neither does . They can be a group iff 2, however first Sylow theorem has no guarantee on such property for the -subgroup and -subgroup retrieved.

Normalizer

In the group , consider binary operation , It's a group action on 's subgroups since and the resulting is also a subgroup of :

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

Such group action is analogous to the conjugaction of elements, in fact we extend the concept of conjugation of elements to subgroups: two subgroups are said to be conjugate when they lies in the same orbit, that is, the conjugacy class of subgroups.

Obviously both and are fixed points by , and subgroups of different size must lie in differnt conjugacy class. For each subgroup of , we define its stabilizer subgroup as and call it the normalizer of in .

For each subgroup , there's . First since , is a subgroup of . Then since , is normal in . And since contains all elements such that its left coset equals right coset, is maximal subgroup in that is normal in .

For a subgroup , sometimes we distinguish where we're looking for normalizer of , so we denote as the normalizer of in . And given a subgroup , there's . Please notice it is not necessarily the case , which it requires to hold, however we can still have . Once you notice is a subgroup of both deriving properties of both of them, things become obvious.

Consider the group such that , and is prime, let be its Sylow -subgroup, then for any of order , if is fixed by under conjugation, then .

To prove, when it is trivially true, and is feasible by the first Sylow theorem, so we just need to consider the case of . Since is fixed by under conjugation, there're and . By the second isomorphism theorem there's . Assume , there must be , otherwise there're and . So there's , which is a contradiction to 's being the Sylow -subgroup, and is infeasible in this case.

Second Sylow theorem

The second Sylow theorem states that all Sylow -subgroups are conjugate, that is, for two Sylow -subgroups of , , or equivalently, under conjugation .

Let the be a group such that , and it's obvious that for 's being Sylow -subgroups.

First, by orbit-stabilizer theorem, there's while . This is because so is divisible by . And if , is divisible by then, which is a contradiction to .

Second, let act on by conjugation of subgroups and see what will happen. Let be the set of fixed points under group action of , and we have the equation of 's group action of conjugation (the equation following orbit-stabilizer theorem, not the class equation):

Finally, since , we can turn the equation into , which means might never be empty. Let's draw some from the set, all non-trivial element has order , and fixes by conjugation, so , and combined with their sharing identity element we have . Finally by , we have , which concludes our proof.

Third Sylow theorem

By the second Sylow theorem, all Sylow -subgroups are conjugate and must lie in the same conjugacy class . For a group such that , the third Sylow theorem states there's and dividing .

Following the proof of the second Sylow theorem, any element taken from is identical to , so actually there're , and .

According to the orbit-stabilizer theorem, we have , and since divides , there can only be 's dividing .

Following the third Sylow theorem, we would like to show that a group such that , where are distinct primes and , the Sylow -subgroup is normal in . What's more, when , there's , which means is cyclic.

Consider the Sylow -group in , by 's dividing there can only be or . And by , , the only candidate is , rendering and .

Consider the Sylow -group in , similarily we have candidates and , and it's possible for the latter case, e.g. and . By setting we eliminate the latter case so that and . Since normalizes each other, is a subgroup. And by , according to the second isomorphism theorem, there's , finally by the fundamental theorem of finite abelian groups, is cyclic.

For the example of group of order but and is not cyclic, consider the group , there're but , and is not cyclic.

From the proof above, we can clearly see how group actions and the Sylow theorems augment our capability of studying much complexer groups drastically, in a very different way.

Finally, let's linger on the question of taking subgroup of order following first Sylow theorem: consider taking a subgroup of order from , whose order is . Any group of order is cyclic, which means if taking a subgroup of order is feasible, there must be some element of order . However after partitioning elements in into disjoint cycles, the maximal order of elements in is , formed by a transposition plus a disjoint -cycle. So taking a subgroup of order is impossible in , rendering taking subgroup of following first Sylow theorem is infeasible in general.

Conclusion

In this text, we introduce the concept of group action such that .

Just like the consequence of group homomorphism, involving a group exhibits some symmetric behaviours: this time, a group action partitions the set on into disjoint orbits s, each point in the orbit corresponds to a left coset of the stabilizer subgroup of , so , which is called the orbit-stabilizer theorem.

By interleaving the points fixed by the entire into from those by a proper subgroup of , we derive the equation of group action that:

Then, we study an group action of a group acting on itself by conjugation : points that are fixed by the whole group forms a subgroup called the center of the group, and those fixed by stabilizer subgroup , which is called the centralizer subgroup of , forms a conjugacy class of .

By putting together the center of group and non-trivial conjugacy classes, we derive the class equation of that:

The class equation ties the group action of conjugacy, the center and centralizer subgroups, and the order of the group together. With it we can inspect how the order of group affects the center and centralizer subgroups, and therefore constrains the structure of a group.

A classical application of the class equation is to prove the Cauchy's theorem of any finite group: for finding a subgroup of order dividing the order of the group , either we can find in of order dividing , which is a proper subgroup of and the descent cannot go on indefinitely, or divides and by is a finite abelian group we can find in .

Based on the idea of Cauchy's theorem, Sylow tweaked its process of finding subgroup and extended its capability to finding dividing the order of . First it's possible to find subgroup of order dividing the order of a group. And if findinding subgroup of order of a group is possible, the subgroup of order found in is a cyclic group, and is normal in . In the quotient group of , by the induction we are able to find subgroup of order , and by correspondence theorem, is the subgroup of order which we are seeking for. This is also called the first Sylow theorem.

The second and third Sylow theorems are about Sylow -subgroups. A Sylow -subgroup is the maximal subgroup of order such that does not divide the order of the group. The second Sylow theorem states that all Sylow -subgroups are conjugate, which means for two Sylow -subgroup of , there's some such that , which can be proved to be 's acting on set of subgroups of . And the third Sylow theorem states that given a group such that , for the number the Sylow -subgroups there's and dividing .

The proofs of both second and third Sylow theorem are done by first inspecting the orbit of a Sylow -subgroups under 's group action of conjugation on subgroups, and then choose another Sylow -subgroup and let it act on the orbit again by conjugation. The consequence is that the other Sylow -subgroup lies in the orbit of the previous one, and is the only one fixed under the group action of itself in the second step. So we derive that any two Sylow -subgroups are conjugate, and their numberic relations.

So group action is indeed a powerful tool for utilizing symmetries of a group, and with appropriate group action chosen (like conjugation), we even utilize a group's own symmetries into constraining itself, which is a quite useful way for studying group theory.


  1. We can define right group actions in analogous way, in which case discussions on them are just mirrored and trival. So every group actions discussed in this text are left group actions. [return]
  2. When , there's and , so is a group. Conversely, when is a group, . By letting we get so . And since is invertible by in , , . Finally we get . [return]
April 15, 2023