DIHEDRAL GROUPS

KEITH CONRAD

1. Introduction
For n ≥ 3, the dihedral group Dn is defined as the rigid motions1 taking a regular n-gon
back to itself, with the operation being composition. These polygons for n = 3, 4, 5, and
6 are pictured below. The dotted lines are lines of reflection: reflecting the polygon across
each line brings the polygon back to itself, so these reflections are in D3 , D4 , D5 , and D6 .

In addition to reflections, a rotation by a multiple of 2π/n radians around the center carries
the polygon back to itself, so Dn contains some rotations.
We will look at elementary aspects of dihedral groups: listing its elements, relations
between rotations and reflections, the center, and conjugacy classes. Throughout, n ≥ 3.

2. Finding the elements of Dn

Points in the plane at a specified distance from a given point form a circle, so points with
specified distances from two given points are the intersection of two circles, which is two
points (non-tangent circles) or one point (tangent circles). For instance, the blue points in
the figure below have the same distances to each of the two black points.

Lemma 2.1. Every point on a regular polygon is determined, among all points on the
polygon, by its distances from two adjacent vertices of the polygon.
Proof. In the picture above, let the blue dots be adjacent vertices of a regular polygon. The
line segment connecting them is an edge of the polygon and the polygon is entirely on one
side of the line through the blue dots. So the two black dots can’t both be on the polygon,
1A rigid motion is a distance-preserving transformation, such as a rotation, a reflection, and a translation,
and is also called an isometry.
1
2 KEITH CONRAD

which means each point on the polygon is distinguished from all other points on the polygon
(not from all other points in the plane!) by its distances from two adjacent vertices. 
Theorem 2.2. The size of Dn is 2n.
Proof. Our argument has two parts: an upper bound and then a construction of enough
rigid motions to achieve the upper bound.
Step 1: |Dn | ≤ 2n.
Pick two adjacent vertices of a regular n-gon, and call them A and B as in the figure
below. An element g of Dn is a rigid motion taking the n-gon back to itself, and it must
carry vertices to vertices (how are vertices unlike other points in terms of their distance
relationships with all points on the polygon?) and g must preserve adjacency of vertices,
so g(A) and g(B) are adjacent vertices of the polygon.

A
g(B)

P
g(A)
For each point P on the polygon, the location of g(P ) is determined by g(A) and g(B),
because the distances of g(P ) from the adjacent vertices g(A) and g(B) equal the distances
of P from A and B, and therefore g(P ) is determined on the polygon by Lemma 2.1. To
count |Dn | it thus suffices to find the number of possibilities for g(A) and g(B).
Since g(A) and g(B) are a pair of adjacent vertices, g(A) has at most n possibilities (there
are n vertices), and for each choice of that g(B) has at most 2 possibilities (one of the two
vertices adjacent to g(A)). That gives us at most n · 2 = 2n possibilities, so |Dn | ≤ 2n.
Step 2: |Dn | = 2n.
We will describe n rotations and n reflections of a regular n-gon.
A regular n-gon can be rotated around its center in n different ways to come back to
itself (including rotation by 0 degrees). Specifically, we can rotate around the center by
2kπ/n radians where k = 0, 1, . . . , n − 1. This is n rotations.
To describe reflections taking a regular n-gon back to itself, look at the pictures on the
first page: if n is 3 or 5 there are lines of reflection connecting each vertex to the midpoint
of the opposite side, and if n is 4 or 6 there are lines of reflection connecting opposite
vertices and lines of reflection connecting midpoints of opposite sides. These descriptions
of reflections work in general, depending on whether n is even or odd:
• For odd n, there is a reflection across the line connecting each vertex to the midpoint
of the opposite side. This is a total of n reflections (one per vertex). They are
different because each one fixes a different vertex.
• For even n, there is a reflection across the line connecting each pair of opposite
vertices (n/2 reflections) and across the line connecting midpoints of opposite sides
(another n/2 reflections). The number of these reflections is n/2+n/2 = n. They are
different because they have different types of fixed points on the polygon: different
pairs of opposite vertices or different pairs of midpoints of opposite sides.
The rotations and reflections are different in Dn since a non-identity rotation fixes no point
on the polygon, the identity rotation fixes all points, and a reflection fixes two points. 
 DIHEDRAL GROUPS 3

In Dn it is standard to write r for the counterclockwise rotation by 2π/n radians. This

rotation depends on n, so the r in D3 means something different from the r in D4 . However,
as long as we are dealing with one value of n, there shouldn’t be confusion.
Theorem 2.3. The n rotations in Dn are 1, r, r2 , . . . , rn−1 .
Here and below, we designate the identity rigid motion as 1.
Proof. The rotations 1, r, r2 , . . . , rn−1 are different since r has order n. 
Let s be a reflection across a line through a vertex. See examples in the polygons below.2
A reflection has order 2, so s2 = 1 and s−1 = s.

s s s s

Theorem 2.4. The n reflections in Dn are s, rs, r2 s, . . . , rn−1 s.

Proof. The rigid motions s, rs, r2 s, . . . , rn−1 s are different since 1, r, r2 , . . . , rn−1 are different
and we just multiply them all on the right by s. No rk s is a rotation because if rk s = r`
then s = r`−k , but s is not a rotation.
Since Dn has n rotations and n reflections, and no rk s is a rotation, they’re all reflections.

Since each element of Dn is a rotation or reflection, there is no “mixed rotation-reflection”:
the product of a rotation ri and a reflection rj s (in either order) is a reflection.
The geometric interpretation of the reflections s, rs, r2 s, and so on is this: drawing all
lines of reflection for a regular n-gon and moving clockwise around the polygon starting
from a vertex fixed by s, we meet successively the lines fixed by rs, r2 s, . . . , rn−1 s. See the
polygons below. Convince yourself, for instance, that if s is the reflection across the line
through the rightmost vertex then rs is the next line of reflection counterclockwise.

r2 s r2 s
r2 s r3 s r4 s r3 s r2 s
r4 s
rs r3 s rs rs r5 s rs

s s s s

Let’s summarize what we have now found.

2The convention that s fixes a line through a vertex matters only for even n, where there are some reflections
across a line that doesn’t pass through a vertex, namely a line connecting midpoints of opposite sides. When
n is odd, all reflections fix a line through a vertex, so any of them could be chosen as s.
4 KEITH CONRAD

Theorem 2.5. The group Dn has 2n elements. As a list,

(2.1) Dn = {1, r, r2 , . . . , rn−1 , s, rs, . . . , rn−1 s},
In particular, all elements of Dn with order greater than 2 are powers of r.
Watch out: although each element of Dn with order greater than 2 has to be a power
of r, because each element that isn’t a power of r is a reflection, it is false in general that
the only elements of order 2 are reflections. When n is even, rn/2 is a 180-degree rotation,
which has order 2. Clearly a 180-degree rotation is the only rotation with order 2, and it
lies in Dn only when n is even.

3. Relations between rotations and reflections

The rigid motions r and s do not commute. Their commutation relation is a fundamental
formula for computations in Dn , and goes as follows.
Theorem 3.1. In Dn ,
(3.1) srs−1 = r−1 .
Proof. A short proof comes from rs being a reflection: (rs)2 = 1 ⇒ rsrs = 1 ⇒ srs = r−1 ,
and s = s−1 since s has order 2.
We now want to prove (3.1) in a longer way using a geometric interpretation of both
sides. Since every rigid motion of a regular n-gon is determined by its effect on two adjacent
vertices, to prove srs−1 = r−1 in Dn it suffices to check srs−1 and r−1 have the same values
at a pair of adjacent vertices.
Recall s is a reflection fixing a vertex of the polygon. Let A be a vertex fixed by s and
write its adjacent vertices as B and B 0 , with B appearing counterclockwise from A and B 0
appearing clockwise from A. This is illustrated in the figure below, where the dashed line
through A is fixed by s. We have r(A) = B, r−1 (A) = B 0 , s(A) = A, and s(B) = B 0 .

B0
The values of srs−1 and r−1 at A are
(srs−1 )(A) = (srs)(A) = sr(s(A)) = sr(A) = s(B) = B 0 and r−1 (A) = B 0 ,
while their values at B are
(srs−1 )(B) = (srs)(B) = sr(s(B)) = sr(B 0 ) = s(A) = A and r−1 (B) = A.
Since srs−1 and r−1 agree at A and at B, they agree on the polygon, so srs−1 = r−1 . 

Equivalent ways of writing srs−1 = r−1 are (since s−1 = s)

(3.2) sr = r−1 s, rs = sr−1 .
 DIHEDRAL GROUPS 5

What these mean is that when calculating in Dn we can move r to the other side of s by
inverting it. By induction (or by raising both sides of (3.1) to an integral power) check
(3.3) srk = r−k s, rk s = sr−k
for every integer k. In other words, every power of r can be moved to the other side of s
by inversion. This also follows from rk s being a reflection:
1 = (rk s)2 = rk srk s ⇒ srk = r−k s−1 = r−k s.
Example 3.2. In D7 , using (3.3)
r2 sr6 sr3 = r2 (sr6 )sr3 = r2 (r−6 s)sr3 = r2 r−6 ssr3 = r−4 r3 = r−1 = r6
and
sr4 sr3 sr2 = s(r4 s)r3 (sr2 ) = s(sr−4 )r3 (r−2 s) = ssr−4 r3 r−2 s = r−3 s = r4 s.
The relation (3.2) involves a particular rotation and a particular reflection in Dn . In
(3.3), we extended (3.2) to any rotation and a particular reflection in Dn . We can extend
(3.3) to any rotation and any reflection in Dn : a general reflection in Dn is ri s, so by (3.3)
(ri s)rj = ri r−j s
= r−j ri s
= r−j (ri s).
In the other order,
rj (ri s) = ri rj s
= ri sr−j
= (ri s)r−j .
This has a nice geometric meaning: when multiplying in Dn , every rotation can be moved
to the other side of every reflection by inverting the rotation. This geometric description
makes such algebraic formulas easier to remember.
Knowing how rotations and reflections interact under multiplication lets us compute the
center of Dn . The answer depends on whether n is even or odd.
Theorem 3.3. When n ≥ 3 is odd, the center of Dn is trivial. When n ≥ 3 is even, the
center of Dn is {1, rn/2 }.
Proof. No reflections are in the center of Dn since reflections do not commute with r:
(ri s)r = ri (sr) = ri r−1 s = ri−1 s, r(ri s) = ri+1 s
so if ri s commutes with r then ri−1 = ri+1 , which implies r2 = 1, but r has order n ≥ 3.
Which rotations rj could be in the center of Dn ? Without loss of generality 0 ≤ j ≤ n−1.
We would need rj to commute with s, so rj s = srj , which is equivalent to rj s = r−j s, which
implies r2j = 1. Since r has order n, r2j = 1 only if n | 2j. For odd n this implies n | j, so
j is a multiple of n and thus rj = 1. Hence for odd n the only rotation that could be in the
center of Dn is 1. Certainly 1 is in the center, so for odd n the center of Dn is {1}.
For even n, the condition n | 2j is equivalent to n/2 | j, and for 0 ≤ j ≤ n − 1 the
only choices for j are j = 0 and j = n/2. Thus rj = r0 = 1 or rj = rn/2 . Certainly 1 is
in the center, and to show rn/2 is in the center we check it commutes with every rotation
and reflection in Dn . That rn/2 commutes with rotations is obvious since all rotations are
powers of r and thus they all commute with each other. To check rn/2 commutes with every
6 KEITH CONRAD

reflection in Dn , the key point is that rn/2 = r−n/2 , which follows from rn = 1. (This also
makes sense geometrically since rn/2 is a 180◦ rotation, and rotating by 180◦ or −180◦ has
the same effect.) Now we check rn/2 commutes with each reflection ri s:
rn/2 (ri s) = rn/2+i s, (ri s)rn/2 = ri r−n/2 s = ri rn/2 s = ri+n/2 s = rn/2+i s.

Example 3.4. The group D3 has trivial center. The group D4 has center {1, r2 }.
Geometrically, rn/2 for even n is a 180-degree rotation, so Theorem 3.3 is saying in words
that the only nontrivial rigid motion of a regular polygon that commutes with all other
rigid motions of the polygon is a 180-degree rotation (when n is even).

4. Conjugacy
In Dn the geometric description of reflections depends on the parity of n: for odd n,
the lines of reflection look the same – each line connects a vertex and the midpoint on the
opposite side – but for even n the lines of reflection fall into two types – lines through pairs
of opposite vertices and lines through midpoints of opposite sides. See Figures 1 and 2.

Figure 1. Lines of Reflection for n = 3 and n = 5.

Figure 2. Lines of Reflection for n = 4 and n = 6.

These different geometric descriptions of reflections in Dn for even and odd n manifest
themselves in the algebraic structure of the group Dn when we describe its conjugacy classes.
Theorem 4.1. The conjugacy classes in Dn are as follows.
(1) If n is odd,
• the identity element: {1},
• (n − 1)/2 conjugacy classes of size 2: {r±1 }, {r±2 }, . . . , {r±(n−1)/2 },
• all the reflections: {ri s : 0 ≤ i ≤ n − 1}.
 DIHEDRAL GROUPS 7

(2) If n is even,
n
• two conjugacy classes of size 1: {1}, {r 2 },
n
• n/2 − 1 conjugacy classes of size 2: {r±1 }, {r±2 }, . . . , {r±( 2 −1) },
n
• the reflections fall into two conjugacy classes: {r2i s : 0 ≤ i ≤ 2 − 1} and
{r2i+1 s : 0 ≤ i ≤ n2 − 1}.
In words, the theorem says each rotation is conjugate only to its inverse (which is another
rotation) except for the identity and (if n is even) except for the 180-degree rotation rn/2 .
Also the reflections are all conjugate for odd n but break up into two conjugacy classes
for even n. The two conjugacy classes of reflections for even n are the two types we see in
Figure 2: those whose fixed line connects opposite vertices (reven s) and those whose fixed
line connects midpoints of opposite sides (rodd s).

Proof. Every element of Dn is ri or ri s for some integer i. Therefore to find the conjugacy
class of an element g we will compute ri gr−i and (ri s)g(ri s)−1 .
The formulas
ri rj r−i = rj , (ri s)rj (ri s)−1 = r−j
as i varies show the only conjugates of rj in Dn are rj and r−j . Explicitly, the basic formula
srj s−1 = r−j shows us rj and r−j are conjugate; we need the more general calculation to
be sure there is nothing further that rj is conjugate to.
To find the conjugacy class of s, we compute
ri sr−i = r2i s, (ri s)s(ri s)−1 = r2i s.
As i varies, r2i s runs through the reflections in which r occurs with an exponent divisible
by 2. If n is odd then every integer modulo n is a multiple of 2 (since 2 is invertible mod n
we can solve k ≡ 2i mod n for i given k). Therefore when n is odd
{r2i s : i ∈ Z} = {rk s : k ∈ Z},
so every reflection in Dn is conjugate to s. When n is even, however, we only get half the
reflections as conjugates of s. The other half are conjugate to rs:
ri (rs)r−i = r2i+1 s, (ri s)(rs)(ri s)−1 = r2i−1 s.
As i varies, this gives us {rs, r3 s, . . . , rn−1 s}. 

Since elements in the center of a group are those whose conjugacy class has size 1, the
calculation of the conjugacy classes in Dn gives another proof that the center of Dn is trivial
for odd n and {1, rn/2 } for even n: we see in Theorem 4.1 that for odd n the only conjugacy
class of size 1 is {1}, while for even n the only conjugacy classes of size 1 are {1} and {rn/2 }.

Appendix A. Commutators in Dn
In a group, a commutator is a product of the form ghg −1 h−1 , which is denoted [g, h].
The set of commutators is not necessarily closed under multiplication (it is closed under
inversion since [g, h]−1 = hgh−1 g −1 = [h, g]), so this set may not be a subgroup. We have
[g, h] = e if and only if gh = hg, so the commutator is related to commuting. What are the
commutators in dihedral groups?
Theorem A.1. The commutators in Dn form the subgroup hr2 i.
8 KEITH CONRAD

Proof. The commutator [r, s] is rsr−1 s−1 = rrss−1 = r2 , so r2 is a commutator. More

generally, [ri , s] = ri sr−i s−1 = ri ri ss−1 = r2i , so every element of hr2 i is a commutator.
To show every commutator is in hr2 i, we will compute [g, h] = ghg −1 h−1 when g and h
are rotations or reflections and check the answer is always a power of r2 .
Case 1: g and h are rotations.
Writing g = ri and h = rj , these commute so ghg −1 h−1 is trivial.
Case 2: g is a rotation and h is a reflection.
Write g = ri and h = rj s. Then h−1 = h, so
ghg −1 h−1 = ghg −1 h = ri rj sr−i rj s = ri+j r−(j−i) ss = r2i .
Case 3: g is a reflection and h is a rotation.
Since (ghg −1 h−1 )−1 = hgh−1 g −1 , by Case 2 the commutator hgh−1 g −1 is a power of r2 ,
so passing to its inverse tells us that ghg −1 h−1 is a power of r2 .
Case 4: g and h are reflections.
Write g = ri s and h = rj s. Then g −1 = g and h−1 = h, so
ghg −1 h−1 = ghgh = (gh)2 = (ri srj s)2 = (ri−j ss)2 = r2(i−j) .

Remark A.2. If n is odd, hr2 i = hri. If n is even, hr2 i is a proper subgroup of hri.