Chapter 3
Classification of Isometries
From our work in Chapter 1 we know that reflections, translations, glide reflec-
tions and rotations are isometries. Furthermore, the Fundamental Theorem tells
us that every isometry is a product of three or fewer reflections. In this chapter
we consider the following question: Is there a way to compose three reflections
and obtain an isometry other than a reflection, translation, glide reflection or
rotation? As we shall soon see, the surprising answer is no! We begin with
a special case of the Angle Addition Theorem, which will be discussed in full
generality later in this chapter.
3.1 The Angle Addition Theorem, part I
Rotations and translations are closely related. We shall observe that the product
of two rotations with arbitrary centers is either a rotation or a translation. We
begin by identifying the isometries that are involutions. Recall that the identity
can be thought of as a rotation about an arbitrary point F through directed
angle 5 0 or as a translation from a point S to itself, i.e., PP .
Theorem 97 Every involutory isometry is either a reflection or a halfturn.
Proof. Reflections are involutory isometries, so consider an involutory isom-
etry that is not a reflection. We must show that is a halfturn. Since 6= >
there exist distinct points S and S 0 such that (S ) = S 0 = Apply to both sides
and obtain
2 (S ) = (S 0 )=
Since is an involution, 2 (S ) = S= Hence (S 0 ) = S and we see that
interchanges the points S and S 0 . Let P be the midpoint of S and S 0 ; then
S P = P S 0=
I claim that P is a fixed point. Let P 0 = (P )= Since is an isometry,
P S 0 = S P 0 = P 0S 0
59
60 CHAPTER 3. CLASSIFICATION OF ISOMETRIES
in which case P 0 is equidistant from S and S 0 and lies on the perpendicular
bisector of S S 0 = But
S S 0 = S P + P S 0 = S P 0 + P 0S 0
so that S> P 0 > and S 0 are collinear by the triangle inequality. Thus S S 0 = 2S P 0
and P 0 is the midpoint of S and S 0 = It follows that
(P ) = P 0 = P=
Since P is a fixed point, Theorem 91 tells us that is either a non-identity
rotation about the point P or a reflection in some line containing P= By as-
sumption, is not a reflection so it must be a non-identity rotation about the
point P> i.e., = P> 6= > where 5
@ 0 = By Proposition 71,
2 = 2P> = P>2 =
Since 2 = we have
P>2 = >
in which case 2 5 0 = 360 and 5 180 = Therefore the rotation is a halfturn
as claimed.
We know from Proposition 55 that a halfturn *F fixes a line c if and only if
F is on line c= But which lines are fixed by a general rotation?
Theorem 98 Non-identity rotations that fix a line are halfturns.
Proof. Let F> be a non-identity rotation that fixes line c> i.e., c = F> (c)=
Let p be the line through F perpendicular to c= By Corollary 85, there is a line
q through F such that
F> = q p
(see Figure 3.1).
Figure 3.1.
3.1. THE ANGLE ADDITION THEOREM, PART I 61
Since c and p are perpendicular, p (c) = c= Therefore
c = F> (c) = ( q p ) (c) = q ( p (c)) = q (c)
and q fixes line c= Either q = c or q B c= But lines p and q intersect at point
F and p B c= So we must have q = c= Therefore F> = c p > which is a
halfturn by Corollary 87.
A technique similar to the one used in Section 2.3 to transform a product of
halfturns into a product of reflections in parallel lines can be applied to a pair
of general rotations and gives part I of the Angle Addition Theorem:
Theorem 99 (The Angle Addition Theorem, part I) Let and be real
@ 0 = Then there is a unique point F such that
numbers such that + 5
E> D> = F>+ =
Proof. If D = E> then F = D and the result was proved earlier in Propo-
#$
sition 71. So assume that D 6= E and let c = DE= By Corollary 85, there exist
unique lines p and q passing through D and E> respectively, such that
E> = q c and D> = c p
(see Figure 3.2).
Figure 3.2.
The measure of the directed angle from p to c is 12 and the measure of the
directed angle from c to q is 12 = Since 12 ( + ) ? 180 > lines p and q intersect
at some unique point F= The Exterior Angle Theorem applied to 4DEF implies
that the measure of the angle from p to q is 12 ( + )= Hence
E> D> = (q c ) ( c p ) = q p = F>+ =
62 CHAPTER 3. CLASSIFICATION OF ISOMETRIES
Exercises
£¤ £¤
1. Let R = 00 and F = 20 =
a. Find equations of lines c> p and q such that F>90 = p q and
R>90 = c p =
b. Find {|-coordinates of the point G such that *G = R>90 F>90 =
c. Find {|-coordinates for the point H such that *H = F>90 R>90 =
£¤ £¤
2. Let R = 00 and F = 01 =
a. Find equations of the lines c> p and q such that *R = p c and
F>120 = q p =
b. Find {| coordinates for the point G and the angle of rotation such
that G> = F>120 *R =
c. Find {|-coordinates for the point H and the angle of rotation such
that H> = F>120 R>60 =
£¤ £¤
3. Let D = 40 and E = 04 =
a. Find equations of lines c> p and q such that D>90 = p c and
E>120 = q p =
b. Find {| coordinates for the point F and the angle of rotation such
that F> = E>120 D>90 =
c. Find {|-coordinates for the point G and the angle of rotation such
that G> = D>90 E>120 =
4. If c> p and q are the angle bisectors of a triangle, prove that the line s
such that s = q p c is perpendicular to a side of the triangle.
3.2 Parity
An isometry has even parity if it factors as an even number of reflections; other-
wise it has odd parity. In this section we shall observe that every even isometry
is the identity, a non-identity translation, or a non-identity rotation and every
odd involutory isometry is a reflection. The case of non-involutory isometries
with odd parity will be addressed in a later section.
Before we can make the notion of parity precise, we must be sure that a
product of reflections with an even number of factors can never be simplified to
3.2. PARITY 63
an odd number of factors and visa-versa, which is the point of our next lemma
and theorems that follow.
Lemma 100 Given a point S and lines d and e> there exist lines f and g with
f passing through S such that
e d = g f =
Proof. If dke> let f be the line through S parallel to d and e= By Corollary
81, there exists a (unique) line g parallel to d, e, and f such that g = e d f =
Multiplying both sides of this equation on the right by f gives the result in
this case. On the other hand, suppose that d and e intersect at a point F= Let
#$
f = FS ; then lines d, e, and f are concurrent at point F= By Corollary 86, there
exists a (unique) line g passing through F such that g = e d f (see Figure
3.3). Multiplying both sides of this equation on the right by f gives the result
in this case as well.
D
C
E
2
B
4
4
Figure 3.3.
Theorem 101 A product of four reflections can always be simplified to a prod-
uct of two reflections, i.e., given lines s> t> u> and v> there exist lines c and p
such that
v u t s = p c=
Proof. Consider the product v u t s = Let S be a point on line s=
Apply Lemma 100 to the point S and given lines t and u to obtain lines t 0 and
u0 with t 0 passing through S such that
u t = u0 t0 =
Next apply Lemma 100 to the point S and lines u0 and v to obtain lines u00 and
p with u00 passing through S such that
v u0 = p u00
64 CHAPTER 3. CLASSIFICATION OF ISOMETRIES
(see Figure 3.4).
N
S''
Q
RR'
4 4
l
2 )
T
SS'
Figure 3.4.
Since s> t 0 > and u00 are concurrent at point S> Corollary 86 tells us that there
exists a (unique) line c passing through S such that
c = u00 t0 s =
Therefore,
v u t s = v u0 t0 s = p u00 t0 s = p c =
By repeatedly applying Theorem 101 to an even length string of reflections,
we can reduce the string to two reflections. On the other hand, repeated appli-
cations of Theorem 101 to an odd length string of reflections reduces it to either
one or a product of three reflections. Therefore we need:
Theorem 102 A product of two reflections never equals one or a product of
three reflections.
Proof. The product of two reflections is the identity, a rotation, or a trans-
lation, which fix every point, exactly one point, or no points, respectively. On
the other hand, a reflection fixes every point on its reflecting line and no other
point. So the product of two reflections can never equal one reflection. Now
arguing indirectly, suppose there exist lines s> t> u> v> and w such that
u t s = v w =
Multiplying both sides of this equation on the left by v gives
v u t s = w =
3.2. PARITY 65
But by Theorem 101, there exist lines c and p such that
p c = v u t s = w >
which contradicts our earlier conclusion that two reflections can never equal one
reflection. Therefore a product of three reflections can never equal a product of
two reflections.
Corollary 103 Every product of reflections with an even number of factors is
a product of two reflections. Every product of reflections with an odd number of
factors is either a single reflection or a product of three reflections.
Definition 104 An isometry is even if and only if is a product of an even
number of reflections; otherwise is odd.
In light of Corollary 103, Theorem 78, and Theorem 83, we see that every
even isometry is the identity, a non-identity translation, or a non-identity rota-
tion.
Theorem 105 Every even involutory isometry is a halfturn; every odd involu-
tory isometry is a reflection.
Proof. By Theorem 97, every involutory isometry is a halfturn or a re-
flection. As noted above, every even isometry is a translation or a rotation.
Therefore every even involutory isometry is a halfturn. Furthermore, Corollary
103 tells us that every odd isometry is either a single reflection or a product of
three reflections. Therefore every odd involutory isometry is a reflection.
The notion of parity is closely related to “orientation”, which we now define.
Choose an ordered basis {v1 > v2 } for the plane R2 as a vector space over R and
position v1 and v2 with their initial points at the origin R. Let D and E be
their respective terminal points and let = p_DRE= (see Figure 3.5).
Figure 3.5.
66 CHAPTER 3. CLASSIFICATION OF ISOMETRIES
Definition 106 An orientation of the plane is a choice of ordered basis {u> v}.
Given an orientation {u> v} > let be the measure of the (undirected) angle
with initial side u and terminal side v= The orientation is negative if 180 ?
? 0 and positive if 0 ? ? 180 .
n £ ¤ £ ¤o
Theorem 107 The sign of an orientation u = xx12 > v = yy12 is the sign of
the determinant
det [u|v] = x1 y2 x2 y1 =
£kukProof. Let £ and ¤ be respective direction angles for u and v; then u =
cos ¤
= kvk cos
kuk sin
> v kvk sin
and the (undirected) angle from u to v is ( ) =
Hence
d1 e2 e1 d2 = (kuk cos ) (kvk sin ) (kvk cos ) (kuk sin )
= kuk kvk (sin cos cos sin )
= kuk kvk sin ( ) >
which is negative if and only if 180 ? ( ) ? 0 and positive if and only
if 0 ? ( ) ? 180 =
Definition 108 A transformation is orientation preserving if the orienta-
tions {u> v} and { (u) > (v)} have the same sign; otherwise is orientation
reversing.
The following fact is intuitively obvious. We leave the proof to the reader
as a series of exercises in this and the next sections (see Exercises 10, 11, 9 and
10).
Proposition 109 Even isometries preserve orientation; odd isometries reverse
orientation.
Here is a chart that summarizes the results obtained so far:
3.2. PARITY 67
Exercises
In problems 1-5, the respective equations of lines s> t> u and v are given. In each
case, find lines c and p such that p c = v u t s and identify the
isometry p c as a translation, a rotation or the identity.
1. \ = [> \ = [ + 1> \ = [ + 2 and \ = [ + 3=
2. \ = [> \ = [ + 1> \ = [ + 2 and \ = [=
3. [ = 0> \ = 0> \ = [ and \ = [=
4. [ = 0> \ = 0> \ = 2> and [ = 2=
5. [ = 0> \ = 0> [ = 1 and \ = [ + 2=
6. [ = 0> \ = 0> \ = [ + 1 and \ = [ + 2=
7. In the diagram below, use a MIRA to construct lines c and p such that
p c = v u t s .
68 CHAPTER 3. CLASSIFICATION OF ISOMETRIES
S
T
Q R
8. Consider the following diagram:
B C
o
2
o
o
E
D
a. Use a MIRA to construct the point T such that T> = e *S d
and find the rotation angle =
b. Use a MIRA to construct the point U such that U> = g f e d
and find the rotation angle =
9. Prove that an even isometry fixing two distinct points is the identity.
10. Prove that a translation preserves orientation.
11. Prove that the reflection in a line through the origin reverses orientation.
3.3 The Geometry of Conjugation
In this section we give geometrical meaning to algebraic conjugation, which plays
a key role in the classification of isometries. You’ve seen conjugation before.
s For
example, to rationalize the denominator of 3+1s2 we conjugate by 3 2> i.e.,
s
we multiply by 3 2 and its multiplicative inverse:
³ μ ¶³
1 s ´ 1 s ´1
s = 3 2 s 3 2 (3.1)
3+ 2 3+ 2
³ μ ¶μ ¶ s
s ´ 1 1 3 2
= 3 2 s s = =
3+ 2 3 2 7
3.3. THE GEOMETRY OF CONJUGATION 69
s
However, you are probably used tos seeing conjugation by 3 2 expressed
as multiplication by the fraction 3 s2 rather than the seemingly awkward ex-
3 2
pression given above. We are free to use the familiar fraction form precisely
because multiplication of real numbers
s is commutative; thus we can interchange
the positions of real numbers 3 2 and 3+1s2 above and multiply. But when
multiplication non-commutative, as in the study of isometries in which multi-
plication is function composition, conjugation takes the form in (3.1) above.
Definition 110 Let and be isometries. The conjugate of by is the
isometry
1 =
In general, = 1 if and only if and commute i.e., = .
We saw an important example of conjugation in Chapter 1 when we derived
the equations for a general rotation about a point F other than the origin R by
conjugating a rotation about R by the translation from R to F (see the proof
of Theorem 69). Precisely, F> is obtained by
1. translating from F to R followed by
2. rotating about R through angle followed by
3. translating from R to F=
Thus
F> = OC R> 1
OC > (3.2)
i.e., F> is the conjugate of R> by OC = Indeed, since rotations and trans-
lation do not commute, the equations for rotations about F have a form quite
dierent from those for rotations about R=
Here are some algebraic properties of conjugation.
Theorem 111
a. The square of a conjugate is the conjugate of the square.
b. The conjugate of an involution is an involution.
Proof. Let and be isometries.
¡ ¢2
(a). 1 = 1 1 = 2 1 =
(b). If is an involution then 1 6= (otherwise 1 = implies
¡ ¢2
= 1 = ). By part (a), 1 = 2 1 = 1 =
1 = and 1 is an involution.
Here are some geometrical consequences.
70 CHAPTER 3. CLASSIFICATION OF ISOMETRIES
Proposition 112 Conjugation preserves parity, i.e., if and are isometries,
then and 1 have the same parity; both either preserve orientation or
reverse orientation.
Proof. By Theorem 92 we know that factors as a product of reflections.
Since the inverse of a product is the product of the inverses in reverse order,
and 1 have the same parity and together contribute an even number of factors
to every factorization of 1 as a product of reflections. Therefore the
parity of an isometry is the same as the parity of 1 . The fact that
and 1 both either preserve orientation or reverse orientation follows
immediately from Proposition 109.
Theorem 113 Let be an isometry.
a. Every conjugate of a halfturn is a halfturn. Furthermore, if S is any
point, then
*S 1 = *(S ) =
b. Every conjugate of a reflection is a reflection. Furthermore, if c is any
line, then
c 1 = (c) =
Proof. (a) Since *S is even by Corollary 87, so is *S 1 by Proposition
112. Furthermore, *S 1 is an involution by Theorem 111. By Theorem 97,
involutory isometries are either reflections (which are odd) or halfturns (which
are even), so *S 1 is a halfturn. To locate its center, observe that
¡ ¢ ¡ ¢
*S 1 ((S )) = *S 1 (S ) = ( *S ) (S ) = (S )
so (S ) is a fixed point for the halfturn *S 1 and is consequently its
center, i.e.,
*S 1 = *(S ) =
MW(2)(3) = W 3 '' 3 '' = M2 (3 ')
2
W(2)
3
3 ' = W (3)
Figure 3.6: Conjugation of *S by =
3.3. THE GEOMETRY OF CONJUGATION 71
(b) Similarly, since c is odd, so is c 1 has odd parity by Proposition
112. Furthermore c 1 is an involution by Theorem 111. By Theorem
97 involutory isometries are either halfturns or reflections, so c 1 is a
reflection. To determine its reflecting line, observe that for any point S on c
¡ ¢ ¡ ¢
c 1 ((S )) = c 1 (S ) = ( c ) (S ) = (S )
so the line (c) is fixed pointwise by the reflection c 1 = Consequently,
(c) is the reflecting line and
c 1 = (c) =
W (3 '') = VW (l) (3) W (l)
3 '' = Vl (3 ')
3
3 ' = W-1 (3)
Figure 3.7: Conjugation of c by =
Theorem 114 Every conjugate of a translation is a translation. If is an
isometry, D and E are points, D0 = (D)> and E 0 = (E), then
AB 1 = A0 B0 =
Proof. Let P be the midpoint of D and E; then
AB = *P *D
by Theorem 56= Since is an isometry, P 0 = (P ) is the midpoint of D0 = (D)
and E 0 = (E)= So again by Theorem 56,
A0 B0 = *P 0 *D0 =
72 CHAPTER 3. CLASSIFICATION OF ISOMETRIES
Therefore, by Theorem 113 (part a) we have
AB 1 = (*P *D ) 1
= ( *P 1 ) ( *D 1 )
= *P 0 *D0 = A0 B0 =
2 ' = Vl (2) 2 '' = W#$ (2 ')
l
# $
Vl (2 '') = W#'$ ' (2)
$ ' = Vl ($)
2
# ' = Vl (#)
Figure 3.8: Conjugation of AB by c =
Our next theorem generalizes the remarks related to (3.2) above:
Theorem 115 The conjugate of a rotation is a rotation. If is an isometry,
F is a point and 5 R> then
½
1 (F)> if is even
F> = =
(F)> if is odd
Proof. Since is the a product of three or fewer reflections, we consider
each of the three possible factorizations of separately.
Case 1: Let u be a line and let = u = Then = 1 is odd and we must show
that u F> u = u (F)> = Let p be the line through F perpendicular to
u= By Corollary 85, there is a (unique) line q passing through F such that
F> = q p
(see Figure 3.10).
3.3. THE GEOMETRY OF CONJUGATION 73
Figure 3.9.
Hence
u F> u = u q p u = ( u q u ) (u p u )
and Theorem 113 we have
( u q u ) ( u p u ) = u (q) u (p) = (3.3)
Since pBu we know that
u (p) = p=
Furthermore, F = p _ q so
u (F) = p _ u (q)=
Since the measure of the directed angle from p to q is 12 > the measure of the
directed angle from p to u (q) is 12 = So by Theorem 82 the right-hand side
in (3.3) becomes
u (q) p = u (F)>
and we conclude that
u F> u = u (F)> = (3.4)
Case 2: Let u and v be lines. Since = v u is even, we must show that
F> 1 = (F)> = But two successive applications of (3.4) give the desired
result:
F> 1 = ( v u ) F> (v u )1
= v ( u F> u ) v
= v u (F)> v = (F)> =
74 CHAPTER 3. CLASSIFICATION OF ISOMETRIES
Case 3: Let u> v> and w be lines. Since = w v u is odd, we must show
that F> 1 = (F)> = This time, three successive applications of (3.4)
in the manner of Case 2 give the result, as the reader can easily check.
Example 116 Look again at the discussion on the equations for general rota-
tions above. Equation (3.2) indicates that OC R> 1
OC = F> = Since OC
is even and OC (R) = F we have
OC R> 1
OC = OC (R)> >
which confirms the conclusion of Theorem 115.
Let’s apply these techniques to draw some interesting geometrical conclu-
sions.
Theorem 117 Two non-identity commuting rotations have the same center of
rotation.
Proof. Let and F> be non-identity commuting rotations. Since is
even, apply Theorem 115 to conjugate F> by and obtain
F> 1 = (F)> =
Since and F> commute by assumption,
(F)> = F> 1 = F> 1 = F> =
Therefore (F) = F= Since is a non-identity rotation, it has exactly one fixed
point, namely F> which is its center of rotation. Therefore and F> have the
same center.
Theorem 118 Let m and n be lines. Then the reflections p and q commute
if and only if p = q or pBq=
Proof. If p and q commute, then
p q = q p
and equivalently
q p q = p = (3.5)
By Theorem 113,
q p q = q (p) = (3.6)
By combining (3.5) and (3.6) we obtain
p = q (p) =
3.3. THE GEOMETRY OF CONJUGATION 75
Therefore p = q (p) in which case q = p or qBp= Conversely, if q = p or
qBp> then q (p) = p in which case p = q (p) = By Theorem 113 we have
q p q = q (p) =
Substituting gives
q p q = p >
which is equivalent to
p q = q p =
Exercises
1. Given a line d and a point E o d> construct the line e such that *E d
*E = e =
2. Given a line e and a point D o e> construct the point E such that e
*D e = *E =
3. Given a line e and a point D such that e *D e = *D > prove that D
lies on e=
4. Let D and E be distinct points and let f be a line. Prove that D>E f =
f D>E if and only if D>E (f) = f=
@ 0 > E> D> =
5. Let D and E be distinct points. Prove that if + 5
F>+ and D> E> = G>+ > then G = # $ (F) =
DE
6. Let d be a line and let E be a point o d. Prove that *E d *E d
*E d *E is a reflection in some line parallel to d=
7. Let D be a point and let be a non-identity translation . Prove that
*D 6= *D =
8. Let D be a point and let f be a line. Prove that f *D 6= *D f > for
every glide reflection f with axis f=
9. Using Exercises 10 and 11, prove that a reflection reverses orientation.
10. Using Exercise 9, prove Proposition 109: Even isometries preserve orien-
tation; odd isometries reverse orientation.
11. Complete the proof of Theorem 115 by proving Case 3.
76 CHAPTER 3. CLASSIFICATION OF ISOMETRIES
3.4 The Angle Addition Theorem
In this section we answer the following questions:
1. What is the result of composing two rotations with dierent centers the
sum of whose rotation angles is a multiple of 360?
2. What is the result of composing a translation and a non-identity rotation
(in either order)?
These results will give us the second and third parts of the Angle Addition
Theorem.
Let’s summarize the facts about even isometries we’ve discovered so far. By
Theorems 78, 83 and repeated applications of Theorem 101, we know that every
even isometry is either a translation or a rotation (we think of the identity as
a trivial rotation or translation). Furthermore, the product of two translations
is a translation (Proposition 33), the product of two halfturns is a translation
(Theorem 51), the product of two rotations with the same center is a rotation
(Proposition 71), and the product of two rotations with dierent centers is a
rotation as long as the sum of the rotation angles is not a multiple of 360
(Theorem 99).
Consider a product of two rotations G> F> about distinct centers F 6= G
#$
and suppose that + 5 0 = Let c = FG= By Corollary 85, there exist unique
lines p and q passing through F and G> respectively, such that
F> = c p and G> = q c >
where the measure of the directed angle from p to c is 12 and the measure of the
directed angle from c to q is 12 = But + = 360 > so that 12 + 12 = 180 .
Therefore the measure of the directed angle from c to q and from c to p are
both elements of 12 (see Figure 3.10).
N O
4/2 )/2 )/2
l
% &
Figure 3.10: G> F> is a translation.
Thus pkq since c is transverse to p and q with equal corresponding angles and
G> F> = ( q c ) ( c p ) = q p
3.4. THE ANGLE ADDITION THEOREM 77
is a translation by Theorem 78. We have proved:
Theorem 119 (The Angle Addition Theorem, part II): If F and G are
distinct points and + = 0 then G> F> is a translation.
Finally, consider the composition (in either order) of a translation and a
non-identity rotation F> .
Theorem 120 (The Angle Addition Theorem, part III): The composi-
tion of a non-identity rotation of and a translation (in either order) is a
rotation of .
Proof. Let c be the line through F perpendicular to the direction of transla-
tion; let p and q be the unique lines such that F> = c p with F = c_p and
= q c with qkc and p a transversal Thus F> = q c c p = q p
is the rotation about the point E = p _ q through angle since the angles
from p to c and from p to q are corresponding. A similar argument for the
composition F> is left to the reader.
We conclude by gathering together the various parts of the Angle Addition
Theorem:
Theorem 121 (The Angle Addition Theorem)
a. A rotation of followed by a rotation of is a rotation of +
unless + = 0 , in which case the composition is a translation.
b. A translation followed by a non-identity rotation of is a rotation of
=
c. A non-identity rotation of followed by a translation is a rotation of
=
d. A translation followed by a translation is a translation.
Exercises
1. Given distinct points D and E> use a MIRA to construct a point S such
that D>60 = S>E E>60 =
2. Given distinct non-collinear points D> E and F> use a MIRA to construct
the point G such that G>60 = D>E F>60 =
78 CHAPTER 3. CLASSIFICATION OF ISOMETRIES
3. Let F be a point and let be a translation. Prove that there is a point U
such that *F = *U =
4. Complete the proof of Theorem 120: Given a translation and a non-
identity rotation F> > prove there is a point E such that F> = E> =
3.5 The Classification Theorem
The Fundamental Theorem proved in chapter 2, tells us that a transformation
is an isometry if and only if it is a product of three or fewer reflections. Thus the
classification of isometries reduces to analyzing the isometries that arise from
the various configurations of three or fewer lines in the plane. Two lines are
either parallel or intersecting; reflecting in two lines is either a translation or
a rotation. Three lines can be either parallel, concurrent, or neither parallel
nor concurrent. Since reflecting in three parallel or concurrent lines is a single
reflection, the only issue that remains is to identify those isometries that arise
as products of reflections in three non-parallel non-concurrent lines.
The main result of this section is the fact that an isometry is a glide reflec-
tion if and only if it factors as a product of reflections in three non-parallel non-
concurrent lines. This key fact leads us to the conclusion that every isometry
is either a reflection, a translation, a glide reflection or a rotation, and com-
pletely settles the classification problem. This result is profoundly significant–
mathematics par excellence! Indeed, the goal of all mathematical inquiry is to
classify the objects studied, and having done so is cause for great celebration.
Recall that an isometry is defined to be a glide reflection if and only if
there exist if there exists a line f and a non-identity translation fixing f such
that = f = In light of Theorem 78, is a glide reflection if and only if
there exist distinct parallel lines d and e with common perpendicular f such
that = f e d = In fact, a glide reflection can be thought of in many
dierent ways:
Theorem 122 For a non-identity isometry , the following are equivalent:
a. is a glide reflection.
b. is a non-identity translation fixing some line f followed by a reflection
f> or vise versa.
c. is a reflection in some line d followed by a halfturn about some point E
o d= The axis of is the line through E perpendicular to d=
d. is a halfturn about some point D followed by a reflection in some line e
o D= The axis of is the line through D perpendicular to e=
Proof. Statement (a) and the first statement in (b) are equivalent by defin-
ition. To establish the equivalence of the two statements in (b), choose distinct
3.5. THE CLASSIFICATION THEOREM 79
parallels d and e and a common perpendicular f such that = f e d . Set
D = d _ f and E = e _ f; then = f 2AB > where AB is a non-zero vector in
the direction of f. But
= f e d = e f d = e d f = 2AB f =
To show that statement (a) implies statements (c) and (d), note that
*D = d f = f d and *E = e f = f e =
Then for d> e> f> D> E as above, we have
= f e d = *E d >
i.e., is a reflection in some line d perpendicular to f followed by a halfturn
about some point E o d (see Figure 3.11), and furthermore,
= f e d = e f d = e f d = e *D >
i.e., is a halfturn about some point D followed by a reflection in some line e
perpendicular to f o D=
D
2
2 ' = V (2)
B
$ M (2 ') = J2)
$
Figure 3.11: = *E d =
Conversely, if = *E d with point E o d> let f be the line through E
perpendicular to d> let e be the line perpendicular to f at E and let D = d _ f=
Then d and e are distinct since E is on e and o d> and
*E d = f e d = f 2AB
is a glide reflection by definition. Furthermore, if = e *D with line e o D>
let f be the line through D perpendicular to e> let d be the line perpendicular
to f at D and let E = e _ f= Then
e *D = e d f = 2AB f
is a glide reflection since the second statement in (b) implies (a).
80 CHAPTER 3. CLASSIFICATION OF ISOMETRIES
Theorem 123 Let be a glide reflection with axis f and glide vector v=
a. 1 is a glide reflection with axis f and glide vector v=
b. If is any translation fixing f> then = =
c. 2 is a non-identity translation.
Proof. The proof of statement (a) is left to the reader.
(b) If is the identity it fixes f and commutes with . So assume that is a
non-identity translation fixing f. Let D be a point on f; then E = (D) 6= D is
a point on f and = AB = Thus f is a glide reflection, and by Theorem 122
(part a),
f = f = (3.7)
On the other hand, by definition there exists a non-identity translation 0 fixing
f such that = f 0 = By Proposition 33, any two translations commute. This
fact together with equation (3.7) gives
= f 0 = f 0 = f 0 = =
(c) Write = f 0 , where 0 is some non-identity translation that fixes f= By
Theorem 122 (part a) we have
2 2
2 = (f 0 ) = f 0 f 0 = f f 0 0 = ( 0 ) >
which is the non-identity translation whose vector is twice the vector of 0 .
Now if = u t s is a glide reflection, then is not a reflection and lines
s> t> and u are neither concurrent nor parallel (see Corollary 81 and Corollary
86). The converse is also true:
Theorem 124 An isometry = u t s is a glide reflection if and only if
lines p, q and r are neither concurrent nor parallel.
Proof. Implication (,) was settled in the preceding remark. To prove the
converse, assume that lines s> t> and u are neither concurrent nor parallel and
show that = u t s is a glide reflection. We consider two cases:
Case 1: Suppose that lines s and t intersect at point T= Since s> t> and u are
not concurrent, T lies o u= Let S be the foot of the perpendicular from T to
u> and let p be the line through S and T. By Corollary 86, there is a (unique)
line c passing through T such that
t s = p c =
Since s 6= t> c 6= p and S (which is distinct from T and lies on p) lies o c=
Thus
= u t s = u p c = *S c
3.5. THE CLASSIFICATION THEOREM 81
with S o c is a glide reflection by Theorem 122 (part b).
Case 2: Suppose that lines s and t are parallel. Then u is not parallel to either
s or t and must intersect t at some point, call it T. Consider the composition
s t u = By Case 1 above, there is some point S o some line c such that
s t u = *S c =
Hence,
= u t s = ( s t u )1 = (*S c )1 = c *S
with point S o c is once again a glide reflection by Theorem 122 (part c).
We can now obtain the long awaited classification theorem:
Theorem 125 (Classification of Plane Isometries) Every non-identity
isometry is exactly one of the following: a translation, a rotation, a reflection,
or a glide reflection.
Proof. By Corollary 103, every isometry is either even or odd. By Theorem
88, every even isometry is either a translation or a rotation. By Corollary 81,
Corollary 86, and Theorem 124, every odd isometry is either a reflection or a
glide reflection.
We can now determine the conjugates of a glide reflection.
Theorem 126 The conjugate of a glide reflection is a glide reflection. If is
an isometry and is a glide reflection with axis f> then 1 is a glide
reflection with axis (f)=
Proof. Suppose that is a glide reflection with axis f and let be any
isometry. Consider the line (f); then
¡ ¢ ¡ ¢
1 ((f)) = 1 (f) = ( ) (f) = ((f)) = (f)
so that 1 fixes line (f)= Since is an odd isometry, so is 1 =
By Theorem 125, every odd isometry is either a reflection or a glide reflection.
If 1 were a reflection it would be an involution. But by Theorem 123
(part b), 2 is a non-identity translation = Therefore
¡ ¢2
1 = 1 1 = 2 1 = 1
is a non-identity translation by Theorem 114. Hence 1 is not an
involution and consequently is not be a reflection. We conclude that 1
is a glide reflection with axis (f)=
82 CHAPTER 3. CLASSIFICATION OF ISOMETRIES
Exercises
1. Identify the isometries that are dilatations and explain.
2. Given a translation with vector v, find the glide vector and axis of a
glide-reflection such that 2 = .
3. Let d, e and f be three non-parallel non-concurrent lines intersecting at
points D> E and F as shown in the diagram below. Problems (b)-(d) use
the result in part (a).
C < B
# 4 ) $
D
a. Prove that the axis of the glide reflection = d e f contains the
feet of the perpendiculars to d and f from D and F> respectively.
b. Use a MIRA and the result in part (a) to construct the axis of and
its glide vector v=
c. Use a MIRA and the result in part (a) to construct the axis of 0 =
f d e and its glide vector v0 =
d. Use a MIRA and the result in part (a) to construct the axis of
00 = e f d and its glide vector v00 =
e. Show that F>2 E>2 D>2 = 2 but D>2 E>2 F>2 = =
4. Prove Theorem 123 (a): If is a glide reflection with axis f and glide
vector v> then 1 is a glide reflection with axis f and glide vector v=
5. Let D and E be distinct points and let f be a glide reflection with axis
f. Prove that D>E f = f D>E if and only if D>E (f) = f=