← IA Groups – Full Text
These are Zixuan’s notes for Part IA – Groups at the University of Cambridge in 2025. The notes are not endorsed by the lecturers or the University, and all errors are my own.
The latest version of this document is available at academic.micfong.space. Please direct any comments to my CRSid email or use the contact details listed on the site.
This document is typeset using Typst. All figures are created using Inkscape.
1 Introduction on Groups
One can think about groups in two ways:
-
on the one hand, they are related to algebra,
-
on the other hand, they are related to symmetry.
1.1 Motivation
An equilateral triangle has rotational symmetry, reflective symmetry, and the identity symmetry.
Let us list these symmetries:
So an equilateral triangle has exactly 6 symmetries.
Things get more interesting when we start to compose (or multiply) symmetries. Note that, after composition of symmetries, our result must also be a symmetry. To find out exactly which one is the end result, we can label the vertices on the triangle.
Here are some important features of symmetries:
- symmtries can be composed,
- there is an identity,
- every symmetry has an inverse,
- comoposition of symmetries is associative,
We can extend this method to algebra to handle more complex cases, e.g. for a 17-gon.
1.2 Introduction on Groups
A group is a triple where
- is a set
- is a binary operation on ,
that satisfies the following four axioms:
-
Closure. For all , . [This can be deduced by definition of binary operations.]
-
Associativity. For all , .
-
Identity. For all , .
-
(Right) Inverse. For all , there exists such that .
We can also think about the definitions as encompassing algebra with one operation, as shown in the following example.
The definition has some important consequences.
Let be a group, and .
- If , then . [right inverses are left inverses.]
- [left identities are right identities.]
- If , then . [inverses are unique.]
- If , then . [the identity is unique.]
Proof.
-
Using Definition 1.2, we have
By the inverse axiom, there is a such that . Multiplying both sides on the right by , we get
-
Using Definition 1.2, and the previous part, there exists such that .
Now
-
We have
-
Using the inverse axiom from Definition 1.2, and part (1), there is a such that .
Multiplying by on both sides of gives
Since , we have .
Notation. By Proposition 1.4, inverses are unique. Therefore, if , we may write
Part (1) of Proposition 1.4 tells us that
which tells us the following.
For in a group , we have
Notation. It makes sense to extend this notation. For any ,
- for any
- for any
Exercise. Show that, if is a group, then for and ,
Recall that it is not necessarily true that in a group . Hence, it is also not necessarily true that
Let be a group and . Then
Proof. We have
Since inverses are unique, the result follows.
1.3 Familar Examples from Arithmetic
-
are all abelian groups. The inverse of is in each case.
-
is not a group due to a lack of inverses.
-
is an abelian group. Note that is not a group since has no inverse.
Similarly, and are abelian groups.
1.4 Finite Groups
Most of the groups above are infinite.
The order of a group is the number of elements of , denoted by .
If , then is finite.
-
For a specific , let
then is an abelian group.
-
Let . For , let . Then is an abelian group.
1.5 Symmetric Groups
We need to introduce some definitions before we introduce the notion of a symmetric group.
Let be sets. A bijection is a map that has an inverse such that
i.e. for all , and for all .
is defined to be the set of permutations of a set .
We will prove that this is a group in Proposition 1.16.
Recall that .
Consider the following maps of sets:
Then .
Proof. For any ,
This makes it easy to see that is a group.
Proof.
- Closure is automatic, since the composition of two permutations is still a permutaion.
- Associativity follows from Lemma 1.15.
- The identity map is the identity element.
- Inverses exist by definition of a bijection.
Therefore the result follows.
-
is the group of ways to rearrange three flower pots on a windowsill.
-
is the group of ways to shuffle a deck of cards.
Notation. Writing is cumbersome. Henceforth, we will just write .
If we want to emphasize that is the identity element of , we will write for . Likewise, we can also write for the operation on .
1.6 Subgroups
Sometimes, we want to restrict our attention to smaller groups. For instance, inside , the rotations of a triangle instead of all symmetries.
Let be a group and . If we have
- ,
- for all ,
- for all .
Then we say that is a subgroup of . We write .
-
Every group is a subgroup of itself.
-
For any group , is the trivial subgroup of .
-
.
-
For , let .
Now we have
- Let . Then
- Let . Then
Therefore, . In fact, these are all of the subgroups .
Proof. For the trivial case , we can construct it by .
Otherwise, if , we may choose to be the smallest positive . [Note that, if and , then and . Therefore, unless , contains a positive element.]
By induction, we see that for all . By the closure of inverses, we conclude that for all . Hence .
It remains to prove that . We shall prove this by contradiction. Suppose that , so there exists some such that . Dividing by and taking remainders, we get
for some and .
But now , so . However, we have which contradicts with our construction of .
If , then
Similarly, for any family of subgroups , we have
Let be a group, and be a subset of . Then
which is the intersection of all the subgroups of that contain , and is called the subgroup generated by .
If , we say generates , or is a generating set for .
Intuitively, is the smallest subgroup containing . If generates , it means that every can be written as
for some , where all . [Note that and can be repetitive elements from .]
1.7 Geometric Examples of Subgroups
Let be the plane, equipped with the usual notion of distance.
For any an isometry of is a bijection
that preserves distance:
Proof. Let us check Definition 1.20. Let and .
-
Clearly, .
-
Since and are isometries,
So .
-
Let . Then
because is an isometry. So
and as required.
Let be the -gon with vertices for .
Define the th dihedral group to be
We can fast-forward to take a look at Theorem 1.32:
For a dihedral group, we have for .
The proof will also give us a good desciption of all the elements. But first, we need some geometric lemmas.
Let . If
then is perpendicular to .
Proof. By symmetry,
But is a straight line, so .
Proof. The proof is by contradiction. Suppose that for some . Then
for .
Now apply Lemma 1.30, with , , then we get
Similarly,
Hence, are collinear, contradicting the hypothesis.
Now, to prove the theorem we stated above, we define two elements of :
For a dihedral group, we have for , and we have
In particular, generates .
Proof. This will be a long proof.
Outline.
- We show that .
-
We show that by showing
- that , and
- that .
- We show that there are no duplicate elements in .
Let the polygon be . First, we show that .
-
For :
Indeed, For any , then
so is indeed an isometry. Also,
so sends vertices of to vertices, and hence preserves .
-
For :
Similarly for any , we have
so is indeed an isometry. Also,
so sends vertices of to vertices, and hence preserves .
We have shown that . Therefore, by induction,
To see that this is all the elements, let . We aim to prove that . We shall apply Lemma 1.31 to complete this proof.
Consider , and .
-
For :
Since , is a vertex of . So
for some . We will try to undo using and .
Therefore,
-
For and :
Now, , so
and hence .
For the same reasons, . Therefore, there are two cases:
- and ,
- and .
-
Case 1. forms .
Since are not collinear,
by Lemma 1.31. Hence, .
-
Case 2. We have
Therefore,
Simiarly to Case 1, by the Lemma 1.31, we get
Hence, as required.
This proves that
Now, we need to check that this list does not contain duplicate elements, so that
First, if such that
then
so .
Now, if then
so and .
But then , which is a contradiction. Therefore, for any .
Now, if , then , contradicting the previous case.
Finally, if there exists such that , then multiplying by inverses to the right by gives
as above.
To understand the group operations on , we need to understand the different ways to multiply and .
For where represents the rotation and represents the reflection, we have
Proof. By Lemma 1.31, it suffices to check that the two expressions do the same thing to . Indeed, with , and ,
2 Homomorphism and Isomorphism
2.1 Introduction on Homomorphisms and Isomorphisms
Some groups are not set-theoretically equal, but nonetheless have the same structure. e.g. and . We wish to encompass this underlying strucutre.
A map between groups
is called a homomorphism if
for all .
-
For any two groups and , the map
for all is a homomorphism, called the trivial homomorphism.
-
If , then the map
is the inclusion homomorphism.
-
Recall that .
Exercise. Show that if , then
is a homomorphism.
-
Since , the determinant function
is a homomorphism.
If is a homomorphism, then
-
.
-
for all .
Proof.
-
We have
Note that , being multiplied by a group element and yielding itself, must be the identity element . This is by Proposition 1.4 (4).
-
We have
Thus, by Proposition 1.4 (3), we have .
To discuss the notion of two groups being “structurally the same”, we can do this using isomorphisms.
If a homomorphism
is a bijection, then is an isomorphism. In this case, we write
From the perspective of group theory, isomorphic groups are considered the same.
-
Recall that with operation , and with operation .
Let
Clearly, is bijective. Further more, for any ,
for some , so
Therefore, is indeed an homomorphism, and hence an isomorphism. That is,
for all .
-
The exponential map
is a homomorphism, because
Since , is also a bijection. Hence is an isomorphism.
The following lemma justifies the claim that we may think of isomorphic groups as “the same”.
- If is a isomorphism, so is .
- If are homomorphisms, so is .
- is an equivalence relation.
We have seen that every subgroup leads to an inclusion homomorphism. This is conversely true, that homomorphism leads to subgroups.
Let be a homomorphism.
-
The image of is
-
The kernel of is
If is a homomorphism, then
- , and
- .
Proof.
-
-
In Lemma 2.3, we showed that .
-
For , we have
-
For ,
Therefore .
-
-
-
-
For ,
So .
-
For , we have
So .
Therefore .
-
Let be a homomorphism.
-
is surjective if and only if .
-
is injective if and only if .
Proof.
-
This is immediate from the definition of surjectivity and image.
-
Indeed, if is injective, then
This has at most one element, and since , we have .
Conversely, suppose . If , then
This calculation shows that , so by assumption, and hence . Thus is injective.
2.2 Cyclic Groups
The groups that we have seen are examples of cyclic groups.
A group is cyclic if there exists an element such that
Such an element is called a generator of .
-
is cyclic, with generator .
-
is cyclic, with generator .
-
is cyclic, since .
If is cyclic, then either
- for some , or
- .
Proof. Let be a cyclic group with generator . Let
and let
Case 1. If , Define
We need to show that is an isomorphism. Since
is certainly a homomorphism.
By the definition of cyclic groups, is surjective.
To prove that is injective, for the purpose of contradiction, suppose that . Since , we may replace by if necessary, and assume . Then , which is a contradiction because .
Therefore, so is injective. Hence, is an isomorphism and .
Case 2. If , define
Since for some , we have
Thus, is a homomorphism.
To prove that is surjective, since is cyclic, every element can be written as for some . By the division algorithm, we can write for some and . Therefore,
This proves that is surjective.
To prove injectivity, suppose that for some [this is equivalent to saying ]. Then or . Since is minimal in , and that and , it follows that , because . Therefore, and is injective.
Therefore,
Because of this theorem, we will write for any cyclic group of order , and .
For any group , and element , let
Note that is cyclic, so for some . This number is called the order of , denoted by .
2.3 Dihedral Groups, Revisited
Remark. Whenever satisfy the dihedral relation
Then, for any , we have
by induction on .
If , we also have
In summary, we have shown that
for all .
Let be a group, that for some , the following relations hold:
- for some
Then and defines a homomorphism that sends the generators of to respectively.
Proof. There are 4 cases to check:
Suppose has generating set
satisfying:
- for some
Then
Proof. By Lemma 2.15, there is a homomorphism
sending and . Since and generate , is surjective. Also, since , is bijective. Therefore, is an isomorphism, and
3 Lagrange’s Theorem
This very important theorem helps us to think about the orders of a groups and subgroups.
The idea of this theorem is to somehow partition the group into cosets of .
3.1 Cosets
Let and . The corresponding left coset is
The set of all left cosets of in is denoted by
Remark. We may similarly define right cosets
The set of right cosets of in is denoted by
If , the left cosets partition . That is,
-
-
If , then for any
Proof.
-
For any , we have since . Thus, . Hence, . The reverse inclusion is obvious. Hence the equality holds.
-
Suppose , so there is a in the intersection. Then,
for some . Thus,
Since is a group, . Thus
Furthermore, for any ,
Hence . By symmetry, we have the reverse inclusion. Thus, the equality holds.
A schematic picture of the lemma above is shown below.
However, we can say more about the sizes of the cosets.
Let , then there is a bijection
for any . In particular, .
So the schematic picture above can be redrawn, where each coset has the same size as .
Let . The index of in is defined as
If and , then
Proof. Since left cosets partition ,
3.2 Consequences of Lagrange’s Theorem
If and , then
Proof. Corollary 3.7 says that for some , so
Proof. Choose any . Then
So, since is prime, either or . Since , we must have . Therefore,
So generates , and hence is cyclic.
3.3 Applications of Lagrange’s Theorem
Lagrange’s theorem 3.6 implies an important result in number theory.
The Euler totient function
Let denote multiplication modulo on , which is associative with identity .
Recall that by the division algorithm, has a multiplicative inverse modulo , iff there are such that
Hence,
is a group with operation .
Let . If , then
4 Group Actions
Groups become groups of symmetries when they act.
4.1 Introduction
An action of a group on a set is a function
with
such that
- for all ,
- for all and .
-
For any group , there is a trivial action of on any set defined by for all and .
-
by .
-
If and then by restriction.
In particular, since .
-
Similarly, dihedral groups acts on (the regular -gon). It also acts on the set of vertices of the regular -gons.
-
Every group acts on itself by left multiplication: for all .
This is called the left regular action of .
Proof. We will show the two directions separately.
Suppose . Consider defined by for all . Then
So . Similarly, . Thus is invertible, so . Therefore, we can define a homomorphism by .
We shall now prove that is a homomorphism. For any and , we have
Thus, , so . Hence, is a homomorphism.
Conversely, given a homomorphism
we may define an action by
Let us check the two properties of an action.
- For any ,
- For any and ,
Thus, the two properties hold, so we have an action of on .
Every group is isomorphic to a subgroup of some .
Furthermore, if , we may choose with .
Proof. Let . Consider the left regular action of on itself. By the previous theorem, this corresponds to a homomorphism
Let . We may think of as a surjective homomorphism from to .
Proof. Indeed, if for all . In particular,
Hence, is bijective, and thus an isomorphism. So so required.
Automatically, if , then .
A lot of important results in group theory come from studying group actions.
Let , and let .
- The orbit of is the set
- The stabiliser of is the set
If , we say that the action is transitive.
If every element (except ) has such that , then is faithful.
4.2 Orbit-Stabiliser Theorem
Suppose . Then
- For any , .
- The orbits form a partition of .
Proof.
-
We need to check satisfies the definition of a subgroup.
-
If , then
so .
-
since .
-
If , then
so .
So all subgroup criteria are satisfied.
-
-
Similarly to the proof of Lemma 3.3,
-
so orbits cover .
-
If , then for some . Hence,
Moreover, any can now be written as
Hence . By symmetry, we have the reverse inclusion. Thus, the equality holds.
-
Consider , where is the set of the regular -gon.
For any ,
Therefore,
The above calculation also shows that . Hence,
Suppose acts on a set . Then for any , the formula
defines a well-defined bijection
If and , then
Proof. Theorem 4.9 gives
Consider again as in the previous example. We saw that, if is a vertex, then
This gives us an easy (but circular for now) proof that
Proof. [For Theorem 4.9]
For notational convenience, let , and define
We have to check several things about .
-
well-defined. That is, for any such that , we have , i.e. .
Now, means that there is such that . Then,
Thus, is well-defined.
-
is surjective. For any , we have
Thus, is surjective.
-
is injective. Suppose for some . By definition, this means that
We need to show that . Let . Now,
Thus, , so
Since cosets paritition, it follows that . Hence, is injective.
Let be the group of isometries of a cube, and let be the centre of a face. Then,
since it is just the number of faces of the cube.
Now, the face is essentially a square, so since the stabiliser must permute the four edges of the face. Thus,
Therefore, by the orbit-stabiliser theorem,
The next theorem is a different kind of application of the orbit-stabiliser theorem.
Proof. Consider the set of -tuples (distinct entries not required)
Define an action of on as follows.
If , let , which is just a cyclic rotation of the -tuple. We need check that this is a well-defined action.
It is easy to see that
We also need to check that
Suppose, therefore, that
For convenience, let and . Then, we know that . So, and therefore . Hence,
Thus, the action is well-defined.
Now compute . For any choices of , there is a unique choice of such that
as inverses are unique. Thus, there are choices for , so
The action on partitions into orbits. So let
By orbit-stabiliser, for each ,
so either or . Let be the number of orbits with size . After renumbering, we may assume that
Since the orbits partition , we have
Since, by our assumption, divides , it follows that divides .
Now, note that has an orbit of size if and only if . In particular, has an orbit of size . Thus, . Since divides and , we must have
so there is at least one more orbit with .
The definition of implies that , so , whence as required.
4.3 Conjugation
Intuition. One way to think about conjugation is that has the same shape as .
Another way is to think about it is that corresponds to changing the coordinates of .
The conjugacy class of an element is the set
Exercise. acts on itself by conjugation:
defines . This is very different from the left regular action. Then is just the orbit of under this action.
In Example Sheet 2 Q6, it is proven that has 1 conjugacy class of reflections if is odd, and 2 conjugacy classes if is even.
Note that the red reflections cannot be obtained by conjugating the blue reflections, and vice versa.
The centraliser of is defined to be
which is just the stabiliser of under the conjugation action.
Remark. Note that
so is the set of elements that commutes with .
The centre of is defined to be
This is exactly the set of elements that commutes with every element of .
5 The Möbius Group
This chapter is about an interesting group. It is almost a group of bijections of , but we need to add a formal point at .
5.1 Riemann Sphere and Möbius Transformations
The Riemann sphere is the set
This set is called a sphere because we can visualize it as follows.
We can define a map called the stereographic projection. It identifies with all the points on the sphere except the north pole at .
Let be complex numbers with . If , then the corresponding Möbius transformation is the map defined by, if ,
If , then is defined by
We usually just write and interpreting the cases of and appropriately.
Then
together with composition of functions forms a group, called the Möbius group.
If
then
Remark. Compare this with matrix multiplication:
5.2 The Möbius Group
Proof.
-
Composition of functions is associative.
-
The identity map is a Möbius transformation with and .
-
For , let .
Then is also a Möbius transformation and and . Hence is the inverse of .
One important way to study Möbius transformations is via their fixed points.
Proof. Let . A fixed point satisfies
Case 1. If there is a fixed point , WLOG let , then . So and satisfy
This is a linear equation with at least 2 roots, so we cannot have two distinct fixed points and unless and . Then .
Case 2. If none of the fixed points is , then we have three distinct complex numbers satisfying
This is a quadratic equation with at least 3 roots, so we cannot have three distinct fixed points unless , , and . Then .
-
Consider . Then .
-
Consider . Then .
Proof. Instead of constructing directly, we construct two auxiliary Möbius transformations and such that and . Then we can let .
Let be defined by
[Modify appropriately if any of is .]
Then , , and .
Similarly, let be defined by
[Modify appropriately if any of is .]
Then , , and .
Note that and are both in . Then let
Then for .
Let be distinct points. Because sharply triply transitively, there exists a unique such that
The cross-ratio is defined as
Note that we saw that can be computed by
by the proof of Lemma 5.8.
Just like for dihedral groups, we can use the 3-point lemma to find a generating set for .
is generated by the set of elements of the following 3 forms:
- , where
- , where
Proof. Let be arbitrary, and let , , .
Step 1. Construct such that .
Either and or
where . Then
Let and .
Step 2. Let , and let , i.e.
Note that and . By construction,
Let
Step 3. Let and
By construction,
By the three-point lemma, , so
5.3 Circles
A circle in is either
-
a Euclidean circle in , or
-
where is a Euclidean straight line in .
Euclidean circles are described by the equation
for some and ,
while lines are described by
Proof. Recall that is generated by
So it suffices to show that each of these generators maps circles to circles.
It is clear that and map circles to circles. So we just need to show that maps circles to circles.
If is a Euclidean circle in , then has equation
If , then we have
This is the equation of a line. Otherwise, , and the equation becomes
which is the equation of a circle.
We are left with the case where is a line, which follows a very similar calculation to the above and is left as an exercise.
Proof. Let such that , , , and so
[] If lie on a circle , then by the previous theorem, is also a circle containing . The only such circle is the real line , so .
[] If , then lie on the circle . So by the previous theorem, their preimages also lie on a circle.
6 Finite Groups
We have already seen some nice theorems about finite groups, including Lagrange’s, orbit-stabilizer, Cayley’s, and Cauchy’s theorems.
We will now develop some small examples of finite groups. We shall proceed naively, trying to list examples by order.
Now, if , we know that is always an option. But is there another?
If are groups, the direct product is
with operation
Note that is a identity, and .
If and
-
,
-
,
-
, i.e. for every , there exist such that ,
then .
Proof. Define with . We need to show that is an isomorphism.
For any , , by definition,
This is immediate from (3).
Recall that we need to show that . Suppose that . Then , so . But also, so by (1), we must have and . Thus , as required.
Proof. By Lagrange’s theorem, every non-trivial element of has order 2 or 4. If there is a such that , then .
Otherwise, every non-trivial element has order 2. Let be distinct elements such that . Let and . It is immediate that [this can be seen by writing out the elements explicitly], which gives us (1). Following the remark, (3) holds.
Finally, since , we have , so , giving us (2) [this was mentioned in Example Sheet 1, Q11]. Thus by the direct product theorem, , as required.
Another application of the direct product theorem 6.3 is to find out when a product of two cyclic groups is cyclic.
Proof. Let . Set
We will check against DPT 6.3.
-
Note that
Similarly,
Therefore, iff divides . Since , this happens iff . Thus, as required.
-
Since is abelian, this is immediate.
-
This follows from (1) and the remark after DPT.
We shall now move on to groups of order 5 and above.
Proof. By Cauchy’s theorem 4.13, there exist such that and . Since , , and . Therefore,
since the other coset must be . So for some .
-
If , then , so , a contradiction.
-
If , then . Then and satisfy the conditions of DPT 6.3, so
-
If , then . In this case, with relations , , which is satisfies the dihedral relation 1.33. Thus, .
Continuing on,
Consider .
We have and and as abelian groups of order 8.
We also have as a non-abelian group of order 8.
Let
It is easy to check that these form a group. We usually use Hamilton’s notation:
So the elements of are , with relations
- ,
- , etc.
- , , ,
- commutes with everything.
We call this group the quaternion group.
Since , is non-abelian and so
By considering the orders of elements, we can also see that . has 5 elements of order 2 (one 180° rotation and the 4 reflections), whereas has only one (the element ).
It can be shown that these are all the groups of order 8.
Proof. By Lagrange’s theorem, the possible orders of elements in are 1, 2, 4, or 8. We have the following cases:
-
If there is an element of order 8, then .
-
If every non-trivial element has order 2, then by a similar argument as in Lemma 6.4. We will start by choosing non-trivial elements with and . We can see that
by using Theorem 6.3 twice. [Check Example Sheet 1, Q11 for related details.]
-
If there is an element of order 4, but no element of order 8, let with . Let . By Lagrange’s theorem,
so
In particular, for some .
- If , then , so .
- If , then , so for all , then is abelian. [Case A.]
- If , then , so But by Example Sheet 2, Q1. However, .
- If , then , which looks similar to the dihedral relation. [Case B.]
Next, note that if , then and so .
Thus . We have several more cases:
- , which we will handle later. [Case I.]
- If , then , so .
- , which we will handle later. [Case II.]
- If , then , so .
There are now four subcases to consider:
-
[Case A, I.] If is abelian and , then , so by Theorem 6.3,
-
[Case A, II.] If is abelian, , so
Again, by Theorem 6.3.
-
[Case B, I.] We have and with relation . These are exactly the relations for (using Proposition 2.16), so .
-
[Case B, II.] We have and with .
Let , , , .
Then, the elements of are which are in Hamilton’s notation.
It is easy to check that the relations in match those in . This defines an isomorphism between and .
Therefore , as required.
In summary, the groups of order up to 8 are as follows:
- ,
- ,
- , , , ,
7 Quotient Groups
7.1 Normal Subgroups
Let be a group homomorphism.
is always a special kind of subgroup.
-
, for any group .
-
If is abelian and , then .
-
since
by the dihedral relation [we don’t have to check since it is in ]. But is not normal since
-
Suppose is a homomorphism. If and then
so . Thus .
Suppose . Then iff
for all .
Proof.
[] Let and . Since , we have . Therefore
Therefore . Similarly, for any , we have so
Thus , and so .
[] Suppose for all . Let . Then
so there exists such that .
Thus .
7.2 Quotient Groups
If , the set of (left) cosets is a group with operation
Proof. We need to check that the operation is well-defined and satisfies the group axioms.
Suppose and since .
That is, there are such that
Therefore,
by Lemma 7.3, for some . Thus,
so
since cosets partition.
-
Associativity. Immediate from associativity in .
-
Identity. The identity is .
-
Inverses. The inverse of is since
-
Closure. Immediate from the definition of the operation.
Therefore is a group.
-
, .
-
Since is abelian, for any . Thus, for any , we have the quotient group
with generator of order .
-
Let be a group, , and suppose . Then, for any ,
and . So by Lemma 7.3, . Furthermore, since its order is 2.
-
An example of (3) is
and hence .
Important. It is a common error that one might think we can multiply groups using direct products and get back to the original group.
-
Note that
But since is cyclic but is not. This shows that quotient groups do not undo direct products. In particular, for groups ,
7.3 The Isomorphism Theorem
If is a homomorphism, then
Proof. Since , the quotient is a group. Let us define
We first check that is a well-defined isomorphism.
Suppose . We need to show that .
We have
for some since . Then,
For ,
Recall that a homomorphism is injective iff its kernel is trivial.
If , then
so , and hence .
That is, , so is injective.
A typical element of is for some . But
so is surjective.
-
Because defined by is a homomorphism with image and kernel , by isomorphism theorem 7.7, we have
-
Similarly, defined by is a homomorphism with
so by isomorphism theorem 7.7, we have
An important question in group theory is to find and understand examples of non-abelian simple groups.
8 Permutations
8.1 Permutations and Cycle Notation
Recall from Definition 1.13 that a permutation of a set is a bijection , and from Definition 1.14 that is the set of all permutations of .
If , then . So examples of permutations include
We can compute by representing permutations as lists.
This is nonetheless a bit cumbersome. A more compact notation is to write permutations in cycle notation.
Any list of distinct elements
defines a -cycle:
which sends
and leaves all other elements fixed.
For Example 8.1, we have
The important rule about cycle multiplication is that the rightmost cycle acts first.
Every can be written as a product of disjoint cycles. This expression is unique up to
-
shifting the elements within each cycle, and
-
reordering the cycles.
Proof. The action of on partitions into orbits. Let
Let . We see that
which proves existence of the decomposition.
The choices we have made are on the representatives of each orbit, and the order of the orbits, which proves uniqueness up to the stated conditions.
Consider
If
then is called a -cycle.
The (multi)set of numbers is called the cycle type of . We often omit singletons from the cycle type.
8.2 Transpositions and the Sign Homomorphism
Proof. We shall prove by induction on .
Base case. Consider . Then is generated by the transposition .
Inductive step. Assume that is generated by transpositions. Consider . Let .
-
If , then , so by the inductive hypothesis is generated by transpositions.
-
Otherwise, let . Then
So , so by the inductive hypothesis is generated by transpositions. Since
we see that is also generated by transpositions.
Proof. Assume . Then the proof is by induction on .
Base case. If , then is already an adjacent transposition.
Inductive step. Assume that we can write as a product of an odd number of adjacent transpositions. Then
and since can be written as a product of an odd number of adjacent transpositions by the inductive hypothesis, so can .
In particular, is generated by adjacent transpositions.
This discussion leads to a notion of parity for permutations.
If are all transpositions and
then is even.
Proof. By Lemma 8.13, we may assume that all are adjacent transpositions.
We say that a pair is called an inversion of a permutation if but .
Proof. We prove this by induction on .
Base case. If , then has inversions, which is even.
Inductive step. Let
Since are all adjacent transpositions, we have for some . Consider which pairs would be inversions of but not , or vice versa.
The only such pair is such that and . This is because only swaps and , so any other pair would remain an inversion or non-inversion in both and .
For this pair, we have
but
Therefore, if and is an inversion for but not for , while if and is an inversion for but not for .
In either case,
as required.
By the claim, since has inversions, we see that must be even.
This enables us to define the sign homomorphism.
The map
is a well-defined homomorphism.
Proof. To see that this is well-defined, suppose are all transpositions such that
Then
so by the previous lemma, is even. Therefore, , so .
To see that this is a homomorphism, note that
The subgroup
is called the alternating group on elements. i.e. it is the set of all even permutations in .
In , the permutations are
We have and , so the even permutations are
Remark. The cycle type makes it easy to determine the sign of a permutation.
Indeed,
so is even iff is odd.
More generally, a -cycle is even iff is even.
- is even
- is odd
8.3 Conjugacy in and
We shall apply what we have obtained so far to study conjugacy in in . Recall that since is a normal subgroup of , the conjugacy class of any element in is contained in .
Proof.
[] Suppose
is a product of disjoint cycles. We have
Since the cycle types are the same, can be written as
Now
defines a permutation of . We can compute
Therefore, .
[] Suppose . The above argument shows that, if
then we can define so that
for all and . Therefore, and have the same cycle type.
This makes it easy to count conjugacy classes in .
Consider
Then we have
Consider without knowing its exact elements. Then we have
Recall that Conjugation Classes 4.16 are essentially orbits under the conjugation action of on itself, and the Centraliser 4.18 of an element is its stabiliser under this action.
Then, Orbit-Stabiliser Theorem 4.9 implies that
Therefore, it is also easy to count the sizes of centralisers.
In , we have
Indeed, we can make a list:
We can list all the conjugacy classes in .
We can write out a table:
| Typical element | ||
We should verify that the sizes of the conjugacy classes add up to .
Indeed, as expected.
Now, counting conjugacy classes in is slightly more subtle. Recall that is the set of elements of that commute with .
Let .
-
If some odd element of commutes with , then
-
Otherwise, if every element of that commutes with is even, then splits into two:
where is any transposition (or any odd permutation).
Proof. Orbit-Stabiliser Theorem 4.9 gives
Since , this gives
is the even permutations that commute with , and is all permutations that commute with . [So is the even bits of .] Therefore, we can write as the kernel of the sign homomorphism restricted to :
The image of has size or [since it is a subgroup of ], so by the Isomorphism Theorem 7.7, we have
-
If there is an odd element of that commutes with [, then this element is in but not in ], so
Using Lagrange’s theorem 3.6, [we have , and] Equation * then becomes
Since , as required.
-
The hypothesis means that
so Equation * becomes
So is half as big as .
Now consider a transposition . Note that .
For the sake of contradiction, suppose . Then there exists such that
But then, by rearranging, we have
which means that commutes with . So . But is odd[, so some odd permutation now commutes with ].
Therefore, , as required.
This makes it possible to determine the conjugacy classes in .
Consider . The even elements of are , -cycles and -cycles. Note that commutes with every element, so its conjugacy class in is the same as in .
Since there is an odd number of -cycles in , the conjugacy class of remains intact in [we cannot split it into two equal parts].
Finally, consider a -cycle, say . We have
and since we know that the cyclic group generated by definitely commutes with , we have
and the conjugacy class of splits into two in . In summary,
| Typical element | |
Finally, let us look at conjugacy in and .
We can write out the conjugacy classes in as follows:
| Even | Typical element | ||
The sizes of the conjugacy classes add up to as expected.
Consider . The even elements of are , -cycles, -cycles and -cycles.
Note that
-
commutes with every element, so its conjugacy class in is the same as in .
-
, so the conjugacy class of remains intact in .
-
Since is odd, the conjugacy class of remains intact in .
-
, so . Therefore, the conjugacy class of splits into two in .
| Typical element | |
The sizes of the conjugacy classes add up to as expected.
Proof. Suppose . By Example Sheet 3 Q5, is a union of conjugacy classes in . At this point, we can list the possible union sizes of conjugacy classes in and see if they divide (by Lagrange’s Theorem 3.6).
The possible union sizes are [note that must be included]:
Therefore, the only possible sizes for are and , so or . Hence, is simple.
9 Matrix Groups
Let be the set of all matrices with real entries.
From IA Vectors and Matrices, we know that matrix multiplication is associative and has an identity element, the identity matrix , though not all matrices have inverses under multiplication.
Here is another result from IA Vectors and Matrices.
This implies that is a homomorphism
By the Isomorphism Theorem 7.7,
For any ,
Hence and so
Remark. We can replace with in the above and get similar results. Therefore we have
9.1 Change of Basis
This is a familiar concept from IA Vectors and Matrices. There is a natural action by conjugation:
Let be an -dimensional vector space over , and a linear map. If that represents in some basis, then the orbit
consists of all matrices that represent in any basis.
Proof. A basis for defines an isomorphism of vector spaces
The claim that represents in this basis means that
and so .
Likewise, another basis for corresponds to another isomorphism
and a matrix represents in these coordinates if
Therefore,
where [because its inverse exists, namely the matrix representing ] represents the isomorphism in the standard basis. Thus, the set of all matrices representing in any basis is contained in the orbit .
Conversely, if
for some , then setting
we get a basis
for . In this basis, represents .
9.2 Möbius Transformations, Revisited
Recall that multiplication in looked similar to multiplication of matrices.
Identify
then
and
Proof. We can prove both statements by constructing a surjective homomorphism from onto with kernel , by the Isomorphism Theorem 7.7. Consider the map
By our previous computation of multiplication in , we see that is a homomorphism. Also, is surjective since for any Möbius transformation with , the matrix is in .
A matrix iff its image fixes , and by the Three Point Lemma for 5.6. Hence
Thus, which we have identified with .
Therefore, by the Isomorphism Theorem 7.7
9.3 Orthogonal Groups
Let us write for the normal notion of length on , i.e.
The -dimensional orthogonal group is the subgroup of that preserves distance in :
In fact, the dot product
is often more convenient to work with.
For any ,
Proof.
It follows that we can characterise using the dot product.
Proof. If for all , then for any ,
Therefore .
Conversely, if , then ,
Hence for all as required.
This quickly leads to a nice characterisations of matrices in .
Let . The following are equivalent:
-
.
-
The columns of form an orthonormal basis of .
-
.
Proof. Let .
[(1) (2).] Let be the standard basis for . The th column of is . since
The columns of form an orthonormal basis.
[(2) (3).] As explained above, (2) means that
Since , this means that
But is the th entry of the matrix , so this shows that the th entry of is for all . Therefore .
[(3) (1).]
Suppose . then
Hence as required.
Recall that . Therefore,
So for any .
The special orthogonal group is the subgroup
Note that
Thus,
Examples of elements of are provided by reflections.
Any defines an orthogonal plane .
The reflection in is defined to be
Remark.
-
We will sometimes write for the reflection in the plane .
-
We may replace by and assume that . then
Proof. We may assume that . From the definition, is linear in . So we can think of as a matrix . Now,
So,
So indeed . In particular, is invertible with inverse , So
Finally, for any ,
Hence as required.
Remark. Let , and pick an orthonormal basis for . In the basis for , has matrix
so and hence .
Proof. We will prove this by induction on .
Base case. When , . The matrix is the reflection in the origin, so the result holds.
Inductive step. Let be the standard basis for . Let .
Then , [and since is an orthogonal transformation, by Lemma 9.9, dot products are preserved, and hence vectors that are orthogonal to are sent to some vector that is still orthogonal to ,] so preserves .
By induction, there are such that
Since both sides also fix , they also agree on . Therefore,
Orthogonal transformations are especially easy to analyse in low dimensions.
Let .
- If then is a reflection.
- If then is a rotation about .
Proof. Recall that , so . By Theorem 9.14, we may take .
-
If , then is odd and hence . So is a reflection.
-
If , then is even, so unless , we can write for some that are not parallel.
We claim that only fixes the origin. [Here, we define a rotation to be an orthogonal transformation that only fixes the origin.] Indeed, for , suppose
But is parallel to and is parallel to , so this implies that is parallel to .
Hence only fixes the origin, and is therefore a rotation about .
Remark. Let . We have seen that the columns form an orthonormal basis of , so we may write
with . Thus, there exists such that and , so
Proof. By Theorem 9.14, is a product of at most reflections. Since , either or for some that are not parallel. Since ,
where is a line through the origin. Since fixes pointwise and also fixes pointwise, their composition also fixes pointwise. [We shall define a rotation in to be an orthogonal transformation that fixes a line pointwise.]
Also , similar to Lemma 9.15, either
- , in which case is fixed by , or
- is parallel to .
Thus, only fixes the line pointwise, and is therefore a rotation.
10 Platonic Solids
While there are infinitely many regular 2-dimensional polygons, in three dimensions there are only five regular solids, known as the Platonic solids.
A convex polyhedron is a Platonic solid if
-
every face of is a regular -gon for some ,
-
acts transitively on the faces.
-
if is the midpoint of a face, then
There are, up to similarity, exactly five Platonic solids:
- the tetrahedron, with 4 triangular faces and 4 vertices,
- the cube, with 6 square faces and 8 vertices,
- the octahedron, with 8 triangular faces and 6 vertices,
- the dodecahedron, with 12 pentagonal faces and 20 vertices,
- the icosahedron, with 20 triangular faces and 12 vertices.
Two solids are dual if can be constructed from by putting vertices in the centers of each face, and then joining vertices in adjacent faces by edges.
The cube and the octahedron are dual.
In particular, if and are dual, then , so we only have three distinct isometry groups of Platonic solids to consider.
Let . By definition, acts transitively on the four faces, and the so by the orbit-stabilizer theorem,
where is the center of a face.
Furthermore, the action of on the four vertices defines a homomorphism
We shall prove that is injective. Suppose . Then fixes each vertex of the tetrahedron, which are not coplanar. So by the 4-point lemma. Therefore is injective, and we may identify with a subgroup of .
But and , so in fact .
Let us also identify the group of rotational symmetries [which are the ones we can actually realize by rotating the solid in space]:
where the tetrahedron is centered at the origin.
Proof. Because , , and . We therefore have surjective homomorphism with .
Since transpositions generate , there is a transposition with . Because all transpositions are conjugate [they have the same cycle type], so for any other transposition and some . Thus
Therefore , so
Therefore, since , we have .
Let . By definition, acts transitively on the six faces, and the so by the orbit-stabilizer theorem,
where is the center of a face.
In particular, the index-two rotational subgroup has order 24.
In Example Sheet 2 Q7, we saw that acts on the set of the four long diagonals of the cube, giving a homomorphism
Since both and have order 24, to show that is injective it suffices to show that it is surjective.
Proof. Since Transpositions Generate 8.11, it suffices to show that all transpositions are contained in . Indeed,
rotation about the axis through the midpoints of an edge maps to a transposition under .
We get different transpositions this way, which is all of them [because it happens that ]. Therefore .
Hence comparing orders, we have .
Now, since commutes with everything in , by Direct Product Theorem 6.3 we have
Remark. We have and , so
Since cubes have reflectional symmetries, we have in that case. The same applies to the other Platonic solids as well.
Now, for the final two Platonic solids.
Let and the rotational subgourp of index two.
By definition, acts transitively on the twelve faces, and the so by the orbit-stabilizer theorem,
where is the center of a face. And so .
By drawing diagonoals on faces, we may inscribe 5 cubes into the dodecahedron.
Since the 5 cubes are built symmetrically from the geometry of the dodecahedron, acts on the set of these 5 cubes, giving a homomorphism
Rotation around the axes through an opposite pair of vertices leads to a 3-cycle. There are 10 diagonals between opposite pairs of vertices, so we get 10 inverse pairs of 3-cycles. So, we get all 10 3-cycles in .
Claim. Let be the set of 3-cycles. Then
Proof. By Example Sheet 4 Q2,
Since is simple, either or . Since , we must have .
In summary, if is the set of 3-cycles, we have seen that
Therefore we have
so is surjective and hence an isomorphism. Therefore .
Finally, since commutes with everything in , by Direct Product Theorem 6.3 we have
