Isometries

1. Isometries versus moves

Isometries
In this course, the notion of move is initial and undefinable. The
notion of congruent figures was introduced in terms of moves: two figures
are called congruent if there exists a move mapping one of them to the
other one. For some classes of figures there are easy congruence tests.
For example, it is easy to check if two line segments are congruent.
One may ask whether a map of the plane to itself is a move if it
maps figures of some type to congruent figures. A positive answer may
give an additional insight on the notion of move.
In the context of this course, an isometry is a mapping of the plane
to itself such that for any two points A, B the segment AB connecting
them is congruent to the segment connecting their images.
The notion of isometry is a general notion commonly accepted in
mathematics. The word isometry means “preserving distances”. The word
metric is a synonym to the word distance.
A move maps any figure to a congruent figure. In the definition of
isometry this is required only for pairs of points, but for other figures
this is not required. Therefore each move is an isometry.
Does the converse hold true? For some simple figures it is easy to
prove that each isometry maps them to congruent figures.

1.A Theorem. Any isometry f maps a circle c to a circle centered at the

image of the center of c with radius congruent to the radius of c.

1
2 Isometries

Proof. Recall that a circle is the set of points X such that the segment
OX is congruent to a fixed segment. So, c = {X ∣ XO = AB}. An isometry
f maps it to {f (X) ∣ XO = AB} that is to {f (X) ∣ f (X)f (O) = AB},
because f (X)f (O) = XO. ◻

1.B Theorem. Any isometry maps a segment to a congruent segment;

Proof. Let us show that an isometry f maps a segment AB to segment

f (A)f (B). It is clear (from the definition of isometry) that the segment
f (A)f (B) congruent to AB. However, it is not clear that the image
of the segment AB is a straight line segment. Points belonging to the
segment AB are characterized as those points X for which the triangle
inequality AB ≤ AX + XB turns to identity. In other words, X belongs to
the segment AB if and only if AX + XB = AB. Since f is an isometry,
f (A)f (B) = AB, f (A)f (X) = AX and f (X)f (B) = XB for any point X.
Thus f (A)f (B) = f (A)f (X) + f (X)f (B) if and only if AX + XB = AB.
Therefore f (X) belongs to the segment f (A)f (B) if and only if X
belongs to the segment AB. ◻

1 Prove that any isometry maps

(1) a line to a line;
(2) a ray to a ray;
(3) a triangle to a congruent triangle;
(4) an angle to a congruent angle.

In fact, each isometry of the plane is a move. We will prove this

below by classifying all the isometries.

Recovering an isometry from images of three points

1.C Theorem. An isometry of the plane can be recovered from its restriction
to any triple of non-collinear points.

Proof. Given images f (A), f (B) and f (C) of non-collinear points A, B,

C under an isometry f , let us find the image f (X) of an arbitrary point
X.
Draw circles cA and cB centered at A and B and passing through X.
By theorem 1.A, the images f (cA ) and f (cB ) of these circles are circles
of the same radii centered at f (A) and f (B). The circles cA and cB
intersect in X. Therefore the circles f (cA ) and f (cB ) intersect in f (X).
1. Isometries versus moves 3

There may be at most one more

intersection point. If f (cA ) and
f (cB ) intersect in one point only,
this point is f (X), and we are done.
If not, we have narrowed our search cA
for f (X) to two points, the two in-
tersection points of f (cA ) and f (cB ), A
the images of intersection points X C
and Y of cA and cB .
Point C is not equidistant from X Y
X and Y , because the locus of
equidistant points is the perpendic-
ular to the segment XY bisecting it, B
points A, B are equidistant and be-
long to the locus, while C is not
cB
collinear with A and B. In order to
choose the image of X, choose the in-
tersection point of f (cA ) and f (cB )
such that the segment connecting it
to f (C) is congruent to XC.
◻
In fact, as we will see later, there are exactly two isometries with the
same restriction to a pair of distinct points.
Isometries as compositions of reflections
1.D Theorem. Any isometry of the plane is a composition of at most three
reflections.
Proof. Choose three non-collinear points A, B, C. By theorem 1.C, it
would suffice to find a composition of at most three reflections which
maps A, B and C to their images under a given isometry f .
First, find a reflection R1 which maps A to f (A). The axis of such

f (C)

B
f f (A)
A
f (B)

Figure 1.
4 Isometries

a reflection is a perpendicular bisector of the segment Af (A). It is

uniquely defined, unless f (A) = A. If f (A) = A, one can take either a
reflection about any line passing through A, or take, instead of reflection,
an identity map for R1 .

C R1 R1 (C)

f (C)

B R1 (B)

A f (A) = R1 (A)
f (B)

Figure 2.

Second, find a reflection R2 which maps R1 (B) to f (B). The axis of

such a reflection is the perpendicular bisector of the segment R1 (B)f (B).
The segments connecting f (A) with f (B) and R1 (B) are congruent to
the segment AB. Therefore f (A) lies on the axis of reflection R2 and is
not moved by R2 .

Now either R2 (R1 (C)) = C and then S = R2 ○ R1 , or the triangles

R2 ○ R1 (△ABC) and f (△ABC) are symmetric in their common side
f (AB). In the former case f = R2 ○ R1 , in the latter case denote by R3

C R1 (C)

f (C)

B R1 (B)
f (A)
A R2
f (B)

R2 ○ R1 (C)

Figure 3.
2. Translations 5

the reflection in f (AB) and observe that f = R3 ○ R2 ○ R1 .

C

f (C) R3

A f (A)
f (B)

R2 ○ R1 (C)
◻

1.E Corollary. Any isometry is a move. In particular, any isometry maps a

figure to a congruent figure.

Proof. By theorem 1.D, any isometry is a composition of reflections.

Reflections are moves. A composition of moves is a move. ◻

Theorem 1.D gives an opportunity to investigate isometries by studying

compositions of reflections. We will do this in the next sections.

2. Translations
B
A map of the plane to itself is called a
translation if, for some fixed points A and A

B, it maps a point X to a point Y = T (X) Y

such that XY BA is a parallelogram.
X
Here we have to be careful with the notion of parallelogram, because
a parallelogram may degenerate to a figure in a line. Not any degener-
ate quadrilateral fitting in a line deserves to be called a parallelogram,
although any two sides of such a degenerate quadrilateral are parallel. In
a (non-degenerate) parallelogram opposite sides are congruent and con-
tinuous degeneration cannot make them non-congruent. This motivates
the following definition. By a parallelogram we mean a sequence of four
segments KL, LM , M N and M K such that KL is congruent and parallel
6 Isometries

to M N and LM is congruent and parallel to M K.

This definition describes both usual parallelograms, N
K
for which congruence of opposite sides can be L
M
deduced from parallelness and vice versa, and the
K L N M
degenerate parallelograms. KL = M N and LM = N K
In fact, in this definition congruence of the sides M
K N
from one of the pairs of opposite sides does not L
follow from congruence of the sides from the other
K L MN
pair. KL = M N , but LM ≠ N K

2.A Theorem. For any points A and B there exists a translation mapping
A to B. Any translation is an isometry.
Proof. Any three points A, B and X can be complemented in a unique
way to a parallelogram ABX ′ X. Define T (X) = X ′ . Obviously, T satisfies
the definition of translation and T (A) = B.
B
For any points X, Y , the quadrilat-
eral XY T (Y )T (X) is a parallelogram, since A
XT (X) ∣∣ AB ∣∣ Y T (Y ) and XT (X) = AB = T (X) T (Y )
Y T (Y ). Therefore, XY = T (X)T (Y ), so T
is an isometry. X Y
◻
The translation moving a point A to a point B will be denoted below
by TÐ→
AB
. The reflection in line l will be denoted by Rl .

Vectors
A pair of points A, B determines a segment AB. An ordered pair
(A, B) of points determine an oriented segment. Orientation of a segment
is nothing but an order of its end points. Usually in a picture an oriented
segment is presented by an arrow and called an arrow, its first end point
is called arrow tail, its second end point is called arrowhead. In formulas
Ð→
an arrow with tail A and head B is denoted by AB.
Ð→ ÐÐ→
Arrows AB and CD are called equivalent if ABDC is a parallelogram.
2 Prove that this equivalence of vectors is an equivalence relation (i.e., it is
reflective, symmetric and transitive). Compare to the proof of Theorem 2.A.

An equivalence class of arrows is called a vector. A vector is de-

termined by any of arrows belonging to it (i.e., by any element of the
equivalence class).
Ð→
An arrow AB defines the translation TÐ→
AB
which maps A to B. Two
arrows are equivalent if and only if they define the same translation.
2. Translations 7

Historical remarks. Vectors were introduced much later than other

objects that we consider. They have been devised by William Rowan
Hamilton He discovered quaternions (in 1843) as a generalization of com-
plex numbers. In 1846 Hamilton divided his quaternions into the sum of
real and imaginary parts that he respectively called ”scalar” and ”vector”.
The word vector comes from Latin and means carrier. Together with
the term vector, Hamilton introduced a few other terms which did not
survived. For example, he called the initial point of the arrow by the
word ’vehend’ (the one who is carried) and the final point (our arrowhead)
by ’vectum’ (the one who has been brought). Nowadays the term vector
used in a huge number of meanings in many fields. Check with wikipedia.

Reflections in parallel lines

2.B Theorem. The composition of two reflections in parallel lines is a transla-
tion in the direction perpendicular to the lines, by twice the distance between
the lines.
More precisely, if lines l and m are parallel, and the line A ∈ l and B is
the image of A under the reflection in m, then

Rm ○ Rl = TÐ→
AB
.

Proof. Obviously, Rl (A) = A, as A ∈ l. Therefore Rm ○Rl (A) = Rm (A) = B.

Pick a point X near l, but outside
the strip between l and m. The seg-
ment connecting X and Rl (X) is bi-
sected by l, the segment connecting A B = Rm (A) =
Rl (X) and Rm ○Rl (X) is bisected by Rm ○ Rl (A)
m. Therefore the segment connecting
X Rl (X) Rm ○ Rl (X)
X and Rm ○ Rl (X) splits into four
segments, the first two of them are l m
congruent and the last two are con-
gruent.
Thus the whole segment is congruent to the sum of its part enclosed
between l and m and two segments outside the strip between l and m,
and the whole segment is twice its part enclosed between l and m. Hence
it is congruent to the segment connecting A and its image under Rm ○ Rl .
Both the segments are perpendicular to l and m. Thus they are opposite
sides of a parallelogram. We can take three such points close to each
other and non-collinear. Restriction of the composition to them coincide
with the translation TÐ→
AB
. Hence, by Theorem 1.C, these two isometries
coincide. ◻
8 Isometries

2.1. Non-uniqueness of presentations for a translation

Now we have two ways of presenting a translation: the original de-
scription via an arrow, and a presentation as a composition of reflections
in two parallel lines. Both are not unique.
Non-uniqueness of the first one is well understood. Two arrows define
the same translation if and only if they are equivalent, i.e., they represent
Ð→ ÐÐ→
the same vector. (Recall that arrows AB and CD are equivalent if
ABDC is a parallelogram.)
A presentation of a translation as a composition of reflections is quite
similar to presentations via arrows. An arrow is an ordered pair of points.
Reflections forming presentation of the second type are defined by their
axes, which are parallel lines, and this pair of lines is ordered, because
one of the reflection is performed first followed by the other one.
Which ordered pairs of parallel lines determine the same translation?
The answer to this question has been already obtained above, in The-
orem 2.B. The lines should be perpendicular to the arrows that determine
the translation, the distance between the lines is one half of the length of
the arrows. These are the only restrictions on the pair of parallel lines.

2.C Corollary of Theorem 2.B. If l, m a pair of parallel lines and l′ , m′

is another pair of parallel lines, then Rm ○ Rl = Rm′ ○ Rl′ if and only if lines
l′ , m′ are parallel to l and m, the distance between l′ and m′ equals the
distance between l and m and the intersection of the half-plane bounded by
l′ and containing m′ with the half-plane bounded by l and containing m is
a half-plane.

l m l′ m′

The following problem gives an elegant reformulation of 2.C .

3 Let l, m, l′ , m′ be lines, l∣∣m. Prove that Rm ○ Rl = Rm′ ○ Rl′ if and only if

there exists a translation T such that l′ = T (l) and m′ = T (m).
3. Rotations 9

3. Rotations

3.A Theorem. The composition of two reflections in non-parallel lines is

a rotation about the intersection point of the lines by the angle equal to
doubled angle between the lines.
Rm ○ Rl (A)
m

β Rl (A)
β
α
α l
A

Proof. Pick some points whose images under reflections are easy to track.
From symmetries/congruent triangles in the picture, it is clear that effect
of two refections is that of a rotation. Since we know that an isometry
is determined by the image of 3 non-collinear points, there is no need to
consider all possible positions of the points.
◻

Non-uniqueness of presentation for a rotation

As for translations, a rotation can be presented as a composition of
two reflections in many ways. However, the intersection point of the axes
should be the center of the rotation. Therefore it should be the same for
any two lines composition of reflections in which is the rotation. Similarly,
the angle between the lines is the half of the angle of the rotation.
Conversely, any couple of lines passing through the center of a rotation,
forming angle which equals the half of the rotation angle, provides a
presentation of the rotation as a composition of two reflections. The
reflections in the lines should be taken in the appropriate order, which
depends on the direction of rotation.

4 Let l, m, l′ , m′ be lines, l ∩ m = O. Prove that rotation Rm ○ Rl coincides with

Rm′ ○ Rl′ if and only there exists a rotation R centered at O such that R(l) = l′
and R(m) = m′ .
10 Isometries

4. Glide reflections

A reflection about a line l followed by a translation along l is called a

glide reflection. In this definition, the order of reflection and translation
does not matter, because they commute: Rl ○ TÐ→ AB
= TÐ→
AB
○ Rl if l ∥ AB. If
A = B, then TÐ→ AB
is identity and the glide reflection is a usual reflection.
Due to theorem ??, TÐ→ AB
can be presented as a composition of reflec-
tion in two parallel lines m and n perpendicular to l.

Compositions of three reflections

4.A Theorem. A composition of three reflections is either a reflection, or a
glide reflection.

Proof. If all three axes of the reflections are parallel, then the firs two
can be translated without changing of their composition (the composition
of reflections about two parallel lines depends only on the direction of
lines and the distance between them). By translating the first two lines,
make the second of them coinciding with the third line. Then in the
total composition they cancel, and the composition is just the reflection
in the first line.

l1 l2 l1′ l3 l1′ l3 = l2′

Rl3 ○ Rl2 ○ Rl1 = Rl3 ○ Rl‘2 ○ Rl1′ = Rl1′

If not all three lines are parallel, then the second is not parallel to one
of the rest. The composition of reflections about these two non-parallel
lines is a rotation, and the lines can be rotated simultaneously about their
intersection point by the same angle without changing of the composition.

m′

l m n l m n′ n

By an appropriate rotation, make the middle line perpendicular to the

line which was not rotated. Then by rotating of these two perpendicular
lines about their intersection point, make the middle one parallel to the
other line. Now the configuration of lines consists of two parallel lines
5. Classification of plane isometries 11

and a line perpendicular to them. The composition of reflections about

them (the order does not matter any more, because they commute) is a
glide reflection.
l
′
l

m′ m′

l n′ m′′ n′
◻

5. Classification of plane isometries

5.A Theorem. Any isometry of the plane is either the identity, or a reflection
about a line, or rotation, or translation, or gliding reflection.
Proof. By Theorem 1.D, any isometry of the plane is a composition of
at most one reflections about lines. If the number of reflections in the
composition is one, then the composition is a reflection. By Theorems
3.A and 2.B, a composition of two reflections is either rotation about a
point, or translation. The identity isometry appears here, as the two lines
may coincide (it appears also as the composition of zero reflections). By
Theorem 4.A, a composition of three reflections is either a reflection or a
glide reflection. ◻