Lecture Notes
Lecture Notes
Amit Kulshrestha
[email protected]
3
4
Lecture 1
Course logistics January 06, 2025
Course Hours
Course TAs
Grading Policy
4. Attendance: 10 points
Every student starts with 10 points. Four unexcused absences are permissible
without a loss of point. From then onward every unexcused absence will attract a
penalty of 1 point per absence.
5
Syllabus
• Symmetries of regular planar shapes and their types, composition tables of sym-
metries.
Suggested readings
Some of the keywords that catch our visual attention while thinking of symmetry
are–reflection about a mirror, rotation about an axis, invariance under a transformation,
repetition, similarity. We will assign a precise meaning to symmetry in the next lecture.
In addition to a mathematical definition of symmetry, it is essential for us to learn how
to identify symmetry, how to keep track of multiple symmetries, and most importantly–
how to make calculations using symmetry. We will see that our existing understanding
of linear algebra and groups would be playing a pivotal role in accomplishing all this.
In the mean time, here is a food for thought. Should we say that the following picture
(Barnsley’s fern) possesses symmetry?
6
Barnsley’s Fern
7
8
Lecture 2
Symmetry in R2 and R3 January 08, 2025
Let us fix the following notation: R2 is the plane whose points are characterized by an
arbitrary choice of origin and x and y coordinates with respect to two non-parallel (and
non-antiparallel) directions. Similarly, R3 is the space whose points are characterized by
an arbitrary choice of origin and x, y and z coordinates with respect to three non-planar
directions. We denote the points of R2 and R3 by column vectors. Further, the directions
x, y, and z are be taken to be mutually perpendicular. For v, w ∈ R2 or R3 , the distance
between v and w will be denoted by d(v, w). We will assign more precise meaning to
distance in the next lecture.
A subset S of R2 will be called a planar shape and a subset of R3 will be called a spatial
shape.
While we make our definition for planar shapes, these hold for spatial shapes as well.
Let S ⊆ R2 be a planar shape. A map σ : R2 → R2 is called a symmetry of S if the
following hold for σ and S.
(i). σ is distance preserving. That is, for every v, w ∈ R2 we have d(σv, σw) = d(v, w).
(ii). σ(S) = S.
9
A nontrivial translation Tu is not a symmetry for circle S 1
4. For n ∈ Z, define
n
Sn1 := +v :v ∈S 1
0
n
Note that Sn1 is simply a translate of the unit circle S by u =1
∈ R2 .
0
[
1
Let S := S2n . Can you recognize this shape? It is infinite strip of circles
n∈Z
touching each other.
2 2 2
Note that in this case, Tu : R → R , where u = , is a symmetry of S. Are
0
there other translations that are symmetries of S?
5. With the definition above, what are all symmetries of Barnsley’s fern? The only
symmetry is the identity map id : R2 → R2 .
10
2.2. Symmetries of a rectangle
Now, we are going to observe symmetries in a rectangle and tabulate them. In the
rectangular figure below, we have marked X and Y axes, and the two diagonals d1 and
d2 .
(ii). The flipping fX about X-axis. Observe that this flipping map fX : R2 → R2 is
defined by
a a
fX =
b −b
This is actually the rotation map that rotates a point by the angle π (in anticlock-
wise direction). We denote it by rπ . What about fY fX ? We note that in this case
fY fX = fX fY = rπ . That the two symmetries fX and fY commute in this case is
incidental. In general the compositions στ and τ σ of two symmetries σ and τ
need not be equal.
(v). Let us try to observe more symmetries? What about flipping along a diagonal?
11
Note that the flipping along diagonal d2 takes our (green) rectangle to another (red)
rectangle. Thus fd2 is not a symmetry, unless our rectangle is a square. Similarly,
flipping along the other diagonal is also not a symmetry of the rectangle.
Hence fX rπ = fY , and the composition did not yield a new symmetry. Instead of
tracking coordinates to recognize symmetries we can just track the movement of
four vertices and conclude that fX rπ = fY .
A rectangle has only four symmetries: id, fX , fY , rπ . After a few lectures we will be able
to convince ourselves that there are no more symmetries of a rectangle . For notational
12
convenience, we denote id by 1 and consider the set of symmetries of the rectangle.
sym(R) := {1, fX , fY , rπ }
1 fX fY rπ
1 1 fX fY rπ
fX fX 1 rπ fY
fY fY rπ 1 fX
rπ rπ fY fX 1
The entries highlighted blue indicate that fX rπ = fY . Thus, in this composition table,
the header column signifies the symmetry operation to be performed first and the
header row signifies the symmetry operation to be performed next in order to get the
symmetry at corresponding entry in the composition table. Incidentally, in this table
we have ab = ba for every a, b ∈ sym(R). Hence, this composition table is commutative
or abelian.
(ii). r2π/3 : rotation by the angle 2π/3 in anti-clockwise direction about the centre of the
triangle.
(iii). r4π/3 : rotation by 4π/3 in anti-clockwise direction about the centre of the triangle.
(iv). f1 , f2 , f3 : where fi denotes the flipping about median of the triangle passing
through vertex i, where i is as per the labeling in the initial configuration.
13
Thus, the set of symmetries of an equilateral triangle ∆ is
1 r2π/3 r4π/3 f1 f2 f3
1 1 r2π/3 r4π/3 f1 f2 f3
r2π/3 r2π/3 r4π/3 1 f2 f3 f1
r4π/3 r4π/3 1 r2π/3 f3 f1 f2
f1 f1 f3 f2 1 r4π/3 r2π/3
f2 f2 f1 f3 r2π/3 1 r4π/3
f3 f3 f2 f1 r4π/3 r2π/3 1
Here, the entries highlighted blue signify the composition f2 f1 = r4π/3 . To see this, we
may go through the following sequence of transformations. In the figure below, the
initial configuration is labeled blue and the movement of vertices is tracked by the
labels colored red.
The six symmetric movements of an equilateral triangle and their relations can be read
through the following picture. It is a good task to read this picture together with the
composition table of sym(∆).
14
2.4. What is common between the two composition tables?
Recall that if a, b ∈ sym(R) then ab = ba. However, in contrast, there are a, b ∈ sym(∆)
such that ab ̸= ba. For example, f2 f1 = r2π/3 , but f1 f2 = r4π/3 . Despite this, the
composition tables for sym(R) and sym(∆) have the following common features.
(ii). For every a ∈ sym(·), there is some a′ ∈ sym(·) such that aa′ = a′ a = 1.
You should verify these. In the next lecture, we will axiomatize these features to say
that symmetries of a shape form a group.
15
16
Lecture 3
Groups: the storekeepers of symmetries January 10, 2025
In the last lecture, we observed that the composition on the set of symmetries is
associative and that every symmetry comes equipped with an opposite symmetry.
These features will be axiomatized to the concept of groups. At this point, you are
encouraged to revise Lecture 18 (October 26, 2023) of your MTH101 (Basic Linear
Algebra) course.
3.1. Groups
A group consists of two mathematical objects — first one - a set X, and second one - a
binary operation ∗ : X × X → X that satisfies the following properties.
I. Associativity.
(a ∗ b) ∗ c = a ∗ (b ∗ c)
for every triplet a, b, c in X. Here, the notation a ∗ b stands for the image of
(a, b) ∈ X × X under the binary operation ∗.
II. Existence of a neutral element. There exists an element 1 ∈ X such that for every
a ∈ X,
a ∗ 1 = a,
and, 1∗a=a
a ∗ a′ = 1,
and, a′ ∗ a = 1
17
Here the first equality holds because 1′ is a neutral element, and the second equality
holds because 1 is a neutral element. Thus, a group has unique neutral element.
The same question for inverses — Can an element a ∈ X have two distinct inverses a′ and
a′′ ? No, again, as the following computation shows.
We attribute
Thus, uniqueness of neutral element and inverses is not a part of group axioms, but
rather a consequence of these. When the binary operation is clear from the context,
then we quietly drop ∗ from the notation (X, ∗) and say that X is a group.
Number of elements in X is called the order of the group X. Groups that have infinitely
many elements are called infinite groups.
1. First and the foremost, symmetry groups of shapes are examples of groups. Thus,
sym(R) is an abelian group of order four. Interestingly, in this group every element
is inverse of itself! That is, a = a′ for every a ∈ sym(R).
The group sym(∆) is a nonabelian group of order six. The three flippings f1 , f2
and f3 , and the neutral element 1 are their own inverses, and two rotations are
inverses of each other.
The group sym(R2 ) is an infinite nonabelian group. This group is nonabelian
because rotation and translation maps do not commute, in general. The inverse
of translation Tu is the translation T−u .
2. (Z, ∗). integers under the binary operation of multiplication do not form a group.
While the integer 1 is a neutral element for multiplication, there are elements in
Z, such as 0 and 2, that do not have inverse. In fact, no integer except ±1 has
multiplicative inverse within integers.
3. (Q, +). rational numbers under the binary operation of addition form a group.
4. Consider the set of clock hours: X := {12, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11}, and the binary
operation of addition of clock hours. Thus, 10 + 3 = 1, and 12 + 7 = 7 under this
binary operation. It forms a group, and the neural element is 12. The inverse of 8
is 4, and the inverse of 1 is 11.
18
5. Consider the set X := {1, 3, 7, 9}. For a, b ∈ X, define a ∗ b to be the rightmost digit
in the usual multiplication of a and b. It is interesting to observe that it forms a
group! In this group, 1 is the neutral element, 3 and 7 are inverses of each other,
and 9 is its own inverse.
6. Consider a set X := {a, b, c} and put a binary operation on it through the following
table.
* a b c
a b a c
b c b a
c a c b
This b ∗ c = c and c ∗ a = c. Does this binary operation define a group? No! The
operation is neither associative, nor does it have a neutral element. Existence of
inverse is, of course, out of question!
7. Consider the set X that consists of 2 × 2 matrices whose entries are integers
and determinant is nonzero. On X, consider the binary operation of matrix
multiplication. Does
it form
a group? Well, while there is indeed
a neutral
element,
1 0 1 0
namely the matrix ∈ X, there are elements such as ∈ X, which
0 1 0 2
do not have inverses. Thus, X is not a group under matrix multiplication.
8. Orthogonal matrices. Consider the set On (R) consisting of n × n invertible
matrices with entries in R such that At = A−1 . Here, At denotes the transpose of
A, and A−1 is the multiplicative inverse of A. Such matrices are called orthogonal
matrices. On this set, consider the binary operation of matrix multiplication? Do
we have a group?
First of all, if A, B ∈ On (R), then AB is invertible (being the product of two
invertible matrices). Further, (AB)t = B t At = B −1 A−1 = (AB)−1 . Thus, AB
indeed belongs to On (R). Also, if A ∈ On (R), then
(A−1 )t = (At )t = A = (A−1 )−1
and thus, A ∈ On (R). The identity matrix In also belongs to On (R) and A ∈ On (R)
inherits associativity from the associativity of multiplication of matrices. Thus,
the set On (R) of orthogonal matrices forms a group under matrix multiplication.
9. This one is a curious exercise. Consider the set
a a ×
X := :a∈Q
a a
On this set, consider the binary operation of matrix multiplication. Show that X
is a group
under this
binary operation. The neutral element of this group is the
1/2 1/2
matrix ∈ X.
1/2 1/2
In the following lectures, our task will be to understand the group sym(R2 ) and utilize
it to understand symmetry of various shapes.
19
20
Lecture 4
Pinpointing symmetries in R2 January 13, 2025
We have seen that translations, rotations and reflections are symmetries. We know that
their compositions are also symmetries. In this lecture, we aim to classify all distance
preserving maps σ : R2 → R2 . The central role in this classification is played by the dot
product that helps us measure distances and angles.
Before we talk about preserving distances, we need to equip ourselves with the tools
that help us measure distances. Let us systematically recall and develop these.
21
• Triangle inequality. d(v1 , v2 ) + d(v2 , v3 ) ≥ d(v1 , v3 ) for every v1 , v2 , v3 ∈ R2 .
• Symmetry. d(v, w) = d(w, v) for every v, w ∈ R2 .
⟨v, w⟩
θ(v, w) :=
ℓ(v)ℓ(w)
(v). Dot product in terms of length. For v, w ∈ R2 , we perform the following calcula-
tion.
ℓ(v − w)2 = ⟨v − w, v − w⟩
= ⟨v, v⟩ − ⟨v, w⟩ − ⟨w, v⟩ + ⟨w, w⟩
= ℓ(v)2 − 2⟨v, w⟩ + ℓ(w)2
Therefore,
1
ℓ(v)2 + ℓ(w)2 − ℓ(v − w)2
⟨v, w⟩ =
2
We replace ⟨v, w⟩ by θ(v, w)ℓ(v)ℓ(w), using the definition of angle cosine, and
obtain
1
ℓ(v)2 + ℓ(w)2 − ℓ(v − w)2
θ(v, w)ℓ(v)ℓ(w) =
2
Thus,
ℓ(v)2 + ℓ(w)2 − ℓ(v − w)2
θ(v, w) =
2ℓ(v)ℓ(w)
This is nothing but the following cosine formula that we saw in our school.
22
4.2. Tu and O2 (R) is all that it takes
Theorem 4.1. Let σ : R2 → R2 be a distance preserving map. Then there exists u ∈ R2 and an
orthogonal matrix A ∈ O2 (R) such that σ = Tu ◦ A. Therefore, every distance preserving map
in R2 is obtained by composing multiplication by an orthogonal matrix and translation.
Further, since θ has expression in terms of dot product and ℓ, which are preserved
by σ ′ , it follows that angles are preserved by σ ′ .
(iv). σ ′ is linear.
This is most surprising claim. We are required to show that for every v, w ∈ R2
and λ ∈ R, we have
σ ′ (v + λw) = σ ′ v + λσ ′ w
For this, is is enough to show that the length ℓ(σ ′ (v + λw) − σ ′ v − λσ ′ w) is 0. The
following calculation achieves this.
ℓ(σ ′ (v + λw) − σ ′ v − λσ ′ w)2 = ℓ(σ ′ (v + λw) − σ ′ v)2 + ℓ(λσ ′ w)2 − 2⟨σ ′ (v + λw) − σ ′ v, λσ ′ w⟩
= d(σ ′ (v + λw), σ ′ v)2 + λ2 ℓ(σ ′ w)2 − 2⟨σ ′ (v + λw) − σ ′ v, λσ ′ w⟩
23
Observe that 2⟨σ ′ (v + λw) − σ ′ v, λσ ′ w⟩ = 2λ⟨σ ′ (v + λw), σ ′ w⟩ − 2λ⟨σ ′ v, σ ′ w⟩. Since
dot product is invariant under σ ′ , this is equal to 2λ⟨v + λw, w⟩ − 2λ⟨v, w⟩. This,
due to the bilinearity of dot product, is equal to 2λ⟨v + λw − v, w⟩ = 2λ⟨λw, w⟩ =
2λ2 ℓ(w)2 . Therefore, we have
ℓ(σ ′ (v + λw) − σ ′ v − λσ ′ w)2 = d(σ ′ (v + λw), σ ′ v)2 + λ2 ℓ(σ ′ w)2 − 2λ2 ℓ(w)2
= d(v + λw, v)2 + λ2 ℓ(w)2 − 2λ2 ℓ(w)2
(since σ preserves d and ℓ)
= ℓ(v + λw − v)2 + λ2 ℓ(w)2 − 2λ2 ℓ(w)2
= ℓ(λw)2 + λ2 ℓ(w)2 − 2λ2 ℓ(w)2
= λ2 ℓ(w)2 + λ2 ℓ(w)2 − 2λ2 ℓ(w)2 = 0
The situation is now reduced to show that the linear map σ ′ : R2 → R2 comes from an
orthogonal matrix. We will continue with this in the next lecture.
24
Lecture 5
Orthogonal matrices uphold bones of Rn January 15, 2025
Orthopaedic surgeons try to preserve relative distances between our bone cells. Orthog-
onal matrices also do the same! In fact, orthogonal matrices do much more by rotating
and reflecting our bones!
Recall that an invertible matrix A is called an orthogonal matrix if At = A−1 . The set
On (R) of n × n orthogonal matrices forms a group under matrix multiplication. Note
that if A ∈ On (R), then det(A) ∈ {−1, 1}.
Though our bones are not planar, let us focus on matrices in R2 . Arguments are the
same for Rn , when n ≥ 3. Let τ : R2 → R2 be a distance preserving linear map. Since
it is linear, τ (v) = Av for some 2 × 2 matrix A; and as we have seen in the last lecture,
demanding τ to preserve distance can be rephrased as
Therefore, we have
25
In Theorem 4.1, the map σ ′ := T−u ◦ σ was distance preserving. By the above discussion,
σ ′ comes from an orthogonal matrix. Thus, the above discussion completes the proof of
that theorem, and establishes that every symmetry of R2 is of the form Tu ◦ A.
a11 a12
Why the name orthogonal is justified for such matrices? Let A = . We
a21 a22
expand the equation At A = I2 to
a11 a21 a11 a12 1 0
=
a12 a22 a21 a22 0 1
a11 a12
Let v1 = and v2 = .
a21 a22
Then the above equation translates to
⟨v1 , v1 ⟩ ⟨v1 , v2 ⟩ 1 0
=
⟨v2 , v1 ⟩ ⟨v2 , v2 ⟩ 0 1
and we conclude that the columns of orthogonal matrices form an orthonormal set of
R2 , that is, both v1 and v2 have length 1 and the two vectors are orthogonal to each
other. In fact, by considering the equation AAt = I2 we arrive at the conclusion that
rows of orthogonal matrices also form an orthonormal set of R2 .
a11 a12
How does a 2 × 2 orthogonal matrix look like? A = , then the condition
a21 a22
At A = I2 poses the follow restrictions
a211 + a221 = 1
a212 + a222 = 1
a11 a12 + a21 a22 = 0
that is, sin(α + β) = 0. In other words, α + β ∈ {0, π}, and we have two cases.
Case 1. β = −α. In this case we have sin β = − sin α and cos β = cos α. Thus, in this
case
a11 a12 cos α sin β cos α − sin α
A= = =
a21 a22 sin α sin α sin α cos α
26
Case 2. β = π − α. In this case we have sin β = sin α and cos β = − cos α. In this case
a11 a12 cos α sin β cos α sin α
A= = =
a21 a22 sin α sin α sin α − cos α
cos α − sin α cos 2α sin 2α
We denote Rα := and Fα := .
sin α cos α sin 2α − cos 2α
Recall that the transformation given by v 7→ Rα v is the rotation by angle α in anticlock-
wise direction, and the one given by v 7→ Fα v is the reflection about a mirror kept along
the line y = tan αx. Thus, Rα is called a rotation matrix and Fα is called a reflection matrix.
What we have shown here is that if A ∈ O2 (R) then A is either a rotation matrix or a
reflection matrix.
Observe that F0 is the reflection about x-axis. Note that Fα/2 = Rα F0 . We summarize
this discussion to the following.
Proposition 5.1. Let A ∈ O2 (R). Then there exists a unique α ∈ [0, 2π) such that either
A = Rα or A = Rα F0 .
In what follows, we state various composition rules. These are left as an exercise for
the reader.
Exercise 5.2 By a direct calculation, obtain the following composition rules for sym-
metries.
(i). Tu Tv = Tu+v .
(ii). Rα Rβ = Rα+β .
(iii). F02 = 1.
(vi). F0 Rα = R−α F0 .
In light of Proposition 5.1 and Exercise 5.2, we revisit Theorem 4.1 to state a much
precise version of it.
27
Proof. It follows from Theorem 4.1 and Proposition 5.1 that σ = Tu Rα F0i , for some
u ∈ R2 , α ∈ [0, 2π) and i ∈ {0, 1}. To establish uniqueness of this decomposition,
assume that σ = Tu Rα F0i = Tv Rβ F0j be two such decompositions. Then, we use Exercise
5.2 (i) to get
Tu−v Rα F0i = T−v Tu Rα F0i = T−v Tv Rβ F0j = Rβ F0j
Where does it take 0 ∈ R2 ? To u − v, if we look at the left hand side, and to 0, if we
look at the right hand side. Hence u = v. This establishes uniqueness of the translation
vector u. Further, it yields
Rα F0i = Rβ F0j
Now, we compare determinant of these matrices to conclude that i = j. Having done
that, Rα = Rβ is immediate. Since, we restrict the range of α within the interval [0, 2π),
the uniqueness of α is also evident.
28
Lecture 6
Symmetries of regular polygons: Dn January 20, 2025
We have already seen symmetries of equilaterral triangles and their composition table.
The following depicts the symmetries of a square.
What is common between the composition tables of an equilateral triangle and a square?
29
6.1. Symmetries of regular polygons
In this lecture, we study symmetries of regular polygons. We first look at the following
proposition.
Proposition 6.1. Let R be a finite group of rotations in O2 (R). Then, there exists a rotation
r ∈ R and m ∈ N such that R = {1, r, r2 , · · · , rm }.
Here 0 < α1 < α2 < · · · < αm < 2π, and α0 = 0. Therefore, rα0 corresponds to the trivial
rotation. The smallest denomination of the angle of a nontrivial rotation is α1 . Since R
is a group, each rαi 1 ∈ R. We note that rαi 1 = riα1 , and that αi ≤ iα1 .
We claim that αi = iα1 for every i ∈ {1, 2, · · · , m}. To justify our claim, imagine the
smallest i where αi ̸= iα1 . Denote such i by k. Then,
Thus, αk − αk−1 < α1 . Observe that rαk −αk−1 = rαk rα−1k−1 ∈ R. Reading it along with
αk − αk−1 < α1 suggests that R contains a rotation of a denomination smaller than α1 .
That is a contradiction, from which we achieve our claim αi = iα1 . We follow a similar
argument to obtain,
αm + α1 = mα1 + α1 = 2π.
Thus, α1 = 2π/(m + 1). To sum it up, we have R = {1, r, r2 , · · · , rm } where r =
r2π/(m+1) .
where r := r2π/n denotes the rotation by angle 2π/n in anti-clockwise direction and f := f0
denotes the reflection about x-axis. The composition table of sym(Pn ) follows the following
laws:
rn = 1, f 2 = 1, f r = r−1 f
30
Thus, sym(Pn ) ⊆ sym(R2 ). Since Pn is a bounded shape, its group of symmetries cannot
contain a nontrivial translation. Hence an element of sym(Pn ) is either a reflection or a
rotation. In other words, sym(Pn ) ⊆ O2 (R).
Let us focus on rotations first, and denote the set of rotations in sym(Pn ) by rot(Pn ).
Note that the set rot(Pn ) is itself a group. Since a rotation must map a vertex of Pn
to a vertex of Pn , there are finitely many elements in rot(Pn ). By Proposition 6.1,
there is a rotation r ∈ rot(Pn ) and m ∈ N such that rot(Pn ) = {1, r, r2 , · · · , rm } where
r = r2π/(m+1) .
What is m in terms of n? Fix a vertex v0 of Pn . Since r ∈ rot(Pn ) is solely determined
by where it takes v0 , the points v0 , r(v0 ), r2 (v0 ), · · · , rm (v0 ) are n distinct vertices of Pn .
Thus, m+1 = n and we have rot(Pn ) = {1, r, r2 , · · · , rn−1 }. This completes our counting
of rotations.
Now, about reflections: the polygon Pn is placed in such a way that f , the reflection
about x-axis is a symmetry of Pn . That is, f ∈ sym(Pn ). Since sym(Pn ) is a group,
we have ri f ∈ sym(Pn ) for every i ∈ {0, 1, . . . , n − 1}. These are n distinct reflections
in sym(Pn ). Have we missed accounting for some other reflection? For a moment,
assume that we have indeed missed a reflection, say f ′ . Then, since composition of two
reflections is a rotation, we have that f ′ f is a rotation. Since our counting of rotations
was complete, we have f ′ f = ri for some i ∈ {0, 1, . . . , n − 1. Thus, f ′ = f ′ f f = ri f ,
that we have already counted. This suggests that every reflection of Pn is of the form is
ri f , and that
sym(Pn ) = {1, r, r2 , . . . , rn−1 , f, rf, r2 f, . . . , rn−1 f }
Theorem 6.2 captures the composition tables of both, an equilateral triangle and a
square. The group sym(Pn ) is called a dihendral group of size 2n.
31
32
Thanks
The following readers are acknowledged for careful reading and pointing out errors.
33