A Gallery of Groups

Study Sheet

A Gallery of Groups

Integers mod n, units mod n, symmetries of polygons, and matrix groups

Integers Mod n Under Addition

Concept
The group (Zn,+)(\mathbb Z_n, +)

Zn={0,1,2,…,n−1}\mathbb Z_n = \{0, 1, 2, \ldots, n-1\} under addition mod nn is a group of order nn: the identity is 00, and the inverse of aa is n−an - a (taking a=0a = 0 to its own inverse 00).

A clock face is Z12\mathbb Z_{12} in disguise: the hour hand wraps from 1111 straight back to 00 because the hours are added mod 1212, not as ordinary integers. Closure and associativity cost nothing new, since addition mod nn inherits them from ordinary addition; the only axiom worth a sentence is inverses, because −a-a itself is rarely one of 0,1,…,n−10, 1, \ldots, n-1, but its remainder n−an - a always is, and a+(n−a)=n≡0(modn)a + (n - a) = n \equiv 0 \pmod n. Addition is commutative no matter what, so (Zn,+)(\mathbb Z_n, +) is abelian for every single nn — a uniformity the next section loses the moment addition is swapped for multiplication.

Example
The Cayley table of (Z6,+)(\mathbb Z_6, +)
·012345001234511234502234501334501244501235501234
Addition mod 6 is symmetric about the diagonal: Z6 is abelian.

Build the addition table for Z6={0,1,2,3,4,5}\mathbb Z_6 = \{0,1,2,3,4,5\}: row aa, column bb holds a+b mod 6a + b \bmod 6. For instance row 44, column 55 holds 4+5=9≡3(mod6)4 + 5 = 9 \equiv 3 \pmod 6. The table is symmetric across the main diagonal, the Cayley-table signature of an abelian group; every row is also the row above it shifted one step, since adding 11 slides every entry forward and wraps the overflow back to the front.

Example
Solving a linear equation

Solve 3+x≡1(mod7)3 + x \equiv 1 \pmod 7 for x∈Z7x \in \mathbb Z_7. Add the inverse of 33 to both sides: the inverse of 33 in (Z7,+)(\mathbb Z_7, +) is 7−3=47 - 3 = 4, so x≡1+4=5(mod7)x \equiv 1 + 4 = 5 \pmod 7. Check: 3+5=8≡1(mod7)3 + 5 = 8 \equiv 1 \pmod 7. Unlike solving 3x=13x = 1 (which needs 33 to be a unit, the subject of the next section), an additive equation like this one is always solvable, because every element of (Zn,+)(\mathbb Z_n, +) has an additive inverse.

Example
A subgroup inside Z12\mathbb Z_{12}

List every multiple of 44 inside Z12\mathbb Z_{12}: 4⋅0=04\cdot0 = 0, 4⋅1=44\cdot1 = 4, 4⋅2=84\cdot2 = 8, 4⋅3=12≡04\cdot3 = 12 \equiv 0 — the list repeats after 33 steps, giving the set {0,4,8}\{0, 4, 8\}. This set is closed under addition (4+4=84 + 4 = 8, 4+8=12≡04 + 8 = 12 \equiv 0, 8+8=16≡48 + 8 = 16 \equiv 4) and contains 00 and every inverse, so it is a subgroup of Z12\mathbb Z_{12} — the cyclic subgroup generated by 44, of order 33 (more on subgroups in Topic 3).

Side note
Where it goes wrong

A natural but wrong guess is that the identity of (Zn,+)(\mathbb Z_n, +) is 11, carried over from thinking of 11 as "the" identity number. Test it: in Z6\mathbb Z_6, 1+2=3≠21 + 2 = 3 \ne 2, so 11 fails to be an identity the moment it is added to anything but itself. The identity of an additive group is always 00: 0+a=a+0=a0 + a = a + 0 = a for every aa, with no exception.

Try it
Try it: an additive inverse

Find the additive inverse of 99 in Z13\mathbb Z_{13}.

Show the answer

The inverse of aa in (Zn,+)(\mathbb Z_n, +) is n−an - a, so the inverse of 99 in Z13\mathbb Z_{13} is 13−9=413 - 9 = \mathbf4. Check: 9+4=13≡0(mod13)9 + 4 = 13 \equiv 0 \pmod{13}.

Units Mod n: the Group U(n)

Concept
The group U(n)U(n)

U(n)={a∈{1,…,n−1}:gcd⁡(a,n)=1}U(n) = \{a \in \{1, \ldots, n-1\} : \gcd(a, n) = 1\} under multiplication mod nn is a group of order φ(n)\varphi(n) (Euler's totient), the group of units mod nn.

Multiplication mod nn is always closed and associative, and 11 is always the identity, so building U(n)U(n) really comes down to one question: which elements have inverses? An element aa has a multiplicative inverse mod nn exactly when gcd⁡(a,n)=1\gcd(a, n) = 1, because Bezout's identity then gives integers x,yx, y with ax+ny=1ax + ny = 1, so ax≡1(modn)ax \equiv 1 \pmod n and x mod nx \bmod n is the inverse; when gcd⁡(a,n)=d>1\gcd(a, n) = d > 1, every multiple of aa mod nn is also a multiple of dd, so a⋅(anything)a\cdot(\text{anything}) can never hit 11. This is a sharp contrast with (Zn,+)(\mathbb Z_n, +): addition mod nn is a group for every nn, but multiplication mod nn only becomes one after the non-invertible elements are thrown out.

Example
The Cayley table of U(8)U(8)
×135711357331755571377531
Every diagonal entry is 1: every non-identity element of U(8) is its own inverse.

U(8)={1,3,5,7}U(8) = \{1, 3, 5, 7\}, the four numbers below 88 coprime to it (2,4,62, 4, 6 all share the factor 22). The multiplication table mod 88 shows every element squares to 11: 32=9≡13^2 = 9 \equiv 1, 52=25≡15^2 = 25 \equiv 1, 72=49≡1(mod8)7^2 = 49 \equiv 1 \pmod 8. A group in which every non-identity element has order 22 is the Klein four-group, U(8)≅Z2×Z2U(8) \cong \mathbb Z_2\times\mathbb Z_2, not the cyclic group of order 44.

Example
Is U(15)U(15) cyclic?

The units mod 1515 are the 88 numbers below 1515 coprime to both 33 and 55: U(15)={1,2,4,7,8,11,13,14}U(15) = \{1, 2, 4, 7, 8, 11, 13, 14\}. Computing the order of each by its powers (e.g. 21=2,22=4,23=8,24=16≡12^1=2, 2^2=4, 2^3=8, 2^4=16\equiv1, so ord⁡(2)=4\operatorname{ord}(2) = 4) shows every element has order 11, 22 or 44: only 11 has order 11; 4,11,144, 11, 14 have order 22; and 2,7,8,132, 7, 8, 13 have order 44. Since ∣U(15)∣=8|U(15)| = 8 but no element reaches order 88, U(15)U(15) is not cyclic.

Example
A cyclic example: U(7)U(7)

In U(7)={1,2,3,4,5,6}U(7) = \{1,2,3,4,5,6\}, the powers of 33 are 3,32≡2,33≡6,34≡4,35≡5,36≡1(mod7)3, 3^2\equiv2, 3^3\equiv6, 3^4\equiv4, 3^5\equiv5, 3^6\equiv1 \pmod7: all six nonzero residues appear before returning to 11, so ord⁡(3)=6=∣U(7)∣\operatorname{ord}(3) = 6 = |U(7)|, and U(7)=⟨3⟩U(7) = \langle3\rangle is cyclic. In fact U(n)U(n) is cyclic exactly when nn is 11, 22, 44, pkp^k, or 2pk2p^k for an odd prime pp — which is why U(7)U(7) succeeds and U(15)=U(3⋅5)U(15) = U(3\cdot5) above does not.

Side note
Where it goes wrong

It is tempting to think U(n)U(n) is just "everything but 00" mod nn. It is not, whenever nn is composite: 2∉U(8)2 \notin U(8), because gcd⁡(2,8)=2≠1\gcd(2, 8) = 2 \ne 1, and indeed 22 has no inverse mod 88 — checking all eight values, 2x mod 82x \bmod 8 for x=0,…,7x = 0,\ldots,7 gives only 0,2,4,60, 2, 4, 6, and 11 never appears.

Try it
Try it: the size of U(20)U(20)

List U(20)U(20) and give ∣U(20)∣|U(20)|.

Show the answer

The numbers below 2020 coprime to it are {1,3,7,9,11,13,17,19}\{1,3,7,9,11,13,17,19\}, so ∣U(20)∣=φ(20)=8|U(20)| = \varphi(20) = \mathbf8.

Symmetries of a Polygon: the Dihedral Groups

Concept
The dihedral group DnD_n

DnD_n is the group of symmetries of a regular nn-gon — rotations and reflections, composed like functions — and it has order 2n2n: it is generated by a rotation rr of order nn and a reflection ss of order 22, subject to the relation srs−1=r−1srs^{-1} = r^{-1}. For n≥3n \ge 3, DnD_n is non-abelian.

Every symmetry of a regular polygon either fixes the center and rotates it, or flips it over through an axis, and composing two symmetries gives a third, so these 2n2n motions close up into a group under composition — the first group in this gallery not built from ordinary arithmetic. The relation srs−1=r−1srs^{-1} = r^{-1} says a reflection reverses the direction of a rotation: rotating then reflecting lands in the same place as reflecting first and rotating the other way. That one relation already forces DnD_n to be non-abelian whenever r≠r−1r \ne r^{-1}, i.e. whenever n≥3n \ge 3, since srsr and rsrs cannot be equal unless r=r−1r = r^{-1}. Every element is uniquely rkr^k or rksr^ks for 0≤k<n0 \le k < n: the nn rotations and the nn reflections, 2n2n elements in all.

Example
The Cayley table of D3D_3
·err2srsr2seerr2srsr2srrr2ersr2ssr2r2err2ssrsssr2srser2rrsrssr2srer2r2sr2srssr2re
D3 is not symmetric about its diagonal, the signature of a non-abelian group.

The multiplication table of D3={e,r,r2,s,rs,r2s}D_3 = \{e, r, r^2, s, rs, r^2s\} (symmetries of an equilateral triangle) is NOT symmetric across the diagonal, unlike every table drawn so far: the entry in row ss, column rr is sr=r2ssr = r^2s, while the entry in row rr, column ss is rsrs — a different element. This asymmetry is the Cayley-table signature of a non-abelian group.

Example
Composing a rotation and a reflection in either order
012
The mirror axis through vertex 1 belongs to the reflection r2s, not rs.

Label the triangle's vertices 0,1,20,1,2 counterclockwise; let rr rotate 0→1→2→00\to1\to2\to0 and let ss fix vertex 00 while swapping 1,21,2. Apply rr then ss (that is, srsr): 0→r1→s20\xrightarrow r1\xrightarrow s2, 1→r2→s11\xrightarrow r2\xrightarrow s1, 2→r0→s02\xrightarrow r0\xrightarrow s0, so srsr sends 0→2,1→1,2→00\to2, 1\to1, 2\to0 — the reflection r2sr^2s, fixing vertex 11. Apply ss then rr (that is, rsrs): 0→s0→r10\xrightarrow s0\xrightarrow r1, 1→s2→r01\xrightarrow s2\xrightarrow r0, 2→s1→r22\xrightarrow s1\xrightarrow r2, so rsrs sends 0→1,1→0,2→20\to1, 1\to0, 2\to2 — the reflection rsrs, fixing vertex 22 instead. Since sr=r2ssr = r^2s fixes vertex 11 while rsrs fixes vertex 22, they are different reflections: sr≠rssr \ne rs.

Example
Find all elements of order 2 in D6D_6

D6D_6, the symmetries of a regular hexagon, has order 12=2⋅612 = 2\cdot6: six rotations e,r,r2,r3,r4,r5e, r, r^2, r^3, r^4, r^5 and six reflections s,rs,r2s,…,r5ss, rs, r^2s, \ldots, r^5s. Every reflection has order 22, since reflecting twice returns every vertex to its start — that is already 66 elements. Among the rotations, rkr^k has order 22 only when r2k=er^{2k} = e, i.e. 6∣2k6 \mid 2k, i.e. 3∣k3\mid k: only k=3k = 3 works among 0,…,50,\ldots,5, giving the single rotation r3r^3 (the 180°180° rotation). So D6D_6 has exactly 6+1=76 + 1 = 7 elements of order 22: the 66 reflections together with r3r^3.

Side note
Where it goes wrong

It is tempting to think a reflection commutes with rotations, since geometrically "flipping" feels independent of "turning." It does not: the defining relation is srs−1=r−1srs^{-1} = r^{-1}, not srs−1=rsrs^{-1} = r. Concretely in D4D_4 (the square, rr a 90°90° rotation), conjugating rr by ss gives srs−1=r3=r−1srs^{-1} = r^3 = r^{-1}, the rotation by 270°270° — the opposite rotation, never rr itself (unless r=r−1r = r^{-1}, which only happens for the order-22 rotation r2r^2).

Try it
Try it: the order of a rotation

In D12D_{12} (symmetries of a regular 12-gon), rr has order 1212. What is the order of r4r^4?

Show the answer

By the order formula for powers (proved in Why It Works below), ord⁡(r4)=ord⁡(r)gcd⁡(ord⁡(r),4)=12gcd⁡(12,4)=124=3\operatorname{ord}(r^4) = \dfrac{\operatorname{ord}(r)}{\gcd(\operatorname{ord}(r), 4)} = \dfrac{12}{\gcd(12,4)} = \dfrac{12}4 = \mathbf3. Check directly: r4r^4 applied three times is r12=er^{12} = e, while r4,r8≠er^4, r^8 \ne e, so 33 really is smallest.

Matrix Groups

Concept
The general linear group GL(n,R)GL(n,\mathbb R)

The n×nn\times n invertible real matrices form a group under matrix multiplication, GL(n,R)GL(n,\mathbb R): the identity is II, and AA has an inverse exactly when det⁡A≠0\det A \ne 0. For n≥2n \ge 2, GL(n,R)GL(n,\mathbb R) is non-abelian.

Closure needs one check: if det⁡A≠0\det A \ne 0 and det⁡B≠0\det B \ne 0 then det⁡(AB)=det⁡Adet⁡B≠0\det(AB) = \det A\det B \ne 0, so the product of two invertible matrices is again invertible — singular matrices are thrown out of the set for exactly the reason non-units were thrown out of Zn\mathbb Z_n to build U(n)U(n). Matrix multiplication is associative for any matrices, invertible or not, so that axiom is inherited free of charge. Unlike every group met so far in this gallery, GL(n,R)GL(n,\mathbb R) is infinite, and for n≥2n \ge 2 it is also non-abelian: invertible matrices do not generally commute, even with each other.

Example
Two invertible matrices that do not commute

Let A=(1201)A = \begin{pmatrix}1&2\\0&1\end{pmatrix} and B=(1031)B = \begin{pmatrix}1&0\\3&1\end{pmatrix}; both have determinant 11, so both lie in GL(2,R)GL(2,\mathbb R). Multiplying in one order, AB=(7231)AB = \begin{pmatrix}7&2\\3&1\end{pmatrix}; in the other order, BA=(1237)BA = \begin{pmatrix}1&2\\3&7\end{pmatrix}. Since AB≠BAAB \ne BA, GL(2,R)GL(2,\mathbb R) is non-abelian — confirmed by nothing more exotic than 2×22\times2 matrix multiplication.

Example
The inverse of a specific matrix

Find the inverse of M=(2312)M = \begin{pmatrix}2&3\\1&2\end{pmatrix}. First det⁡M=2⋅2−3⋅1=1≠0\det M = 2\cdot2 - 3\cdot1 = 1 \ne 0, so MM is invertible. The 2×22\times2 inverse formula swaps the diagonal, negates the off-diagonal, and divides by the determinant: M−1=1det⁡M(2−3−12)=(2−3−12)M^{-1} = \dfrac1{\det M}\begin{pmatrix}2&-3\\-1&2\end{pmatrix} = \begin{pmatrix}2&-3\\-1&2\end{pmatrix} (since det⁡M=1\det M = 1). Check: MM−1=(2312)(2−3−12)=(1001)=IMM^{-1} = \begin{pmatrix}2&3\\1&2\end{pmatrix}\begin{pmatrix}2&-3\\-1&2\end{pmatrix} = \begin{pmatrix}1&0\\0&1\end{pmatrix} = I.

Example
How many invertible 2×22\times2 matrices are there over Z2\mathbb Z_2?

Over the field Z2={0,1}\mathbb Z_2 = \{0,1\} there are 24=162^4 = 16 matrices (abcd)\begin{pmatrix}a&b\\c&d\end{pmatrix} in all; a matrix is invertible exactly when det⁡=ad−bc≢0(mod2)\det = ad - bc \not\equiv 0 \pmod2. Checking all 1616 leaves exactly 66 invertible matrices, so ∣GL(2,Z2)∣=6|GL(2,\mathbb Z_2)| = 6. Multiplying them out shows the group is non-abelian with one identity, three elements of order 22, and two of order 33 — exactly the order pattern of D3D_3 from the previous section: GL(2,Z2)≅D3GL(2,\mathbb Z_2) \cong D_3, the smallest non-abelian group, realized here as matrices instead of triangle symmetries.

Side note
Where it goes wrong

The inverse of a product reverses the order: (AB)−1=B−1A−1(AB)^{-1} = B^{-1}A^{-1}, not A−1B−1A^{-1}B^{-1}. With C=(2111)C = \begin{pmatrix}2&1\\1&1\end{pmatrix} and D=(1101)D = \begin{pmatrix}1&1\\0&1\end{pmatrix} (both determinant 11), (CD)−1=(2−3−12)(CD)^{-1} = \begin{pmatrix}2&-3\\-1&2\end{pmatrix} while C−1D−1=(1−2−13)C^{-1}D^{-1} = \begin{pmatrix}1&-2\\-1&3\end{pmatrix} — a different matrix — and it is D−1C−1=(2−3−12)D^{-1}C^{-1} = \begin{pmatrix}2&-3\\-1&2\end{pmatrix} that matches (CD)−1(CD)^{-1} exactly. Undoing two steps means undoing the second one first.

Try it
Try it: inverting a matrix

Find the inverse of N=(1213)N = \begin{pmatrix}1&2\\1&3\end{pmatrix} and check your answer.

Show the answer

Here det⁡N=1⋅3−2⋅1=1\det N = 1\cdot3 - 2\cdot1 = 1, so N−1=(3−2−11)N^{-1} = \begin{pmatrix}3&-2\\-1&1\end{pmatrix}. Check: NN−1=(1213)(3−2−11)=(1001)NN^{-1} = \begin{pmatrix}1&2\\1&3\end{pmatrix}\begin{pmatrix}3&-2\\-1&1\end{pmatrix} = \begin{pmatrix}1&0\\0&1\end{pmatrix}, so N−1=(3−2−11)N^{-1} = \begin{pmatrix}\mathbf3&\mathbf{-2}\\\mathbf{-1}&\mathbf1\end{pmatrix}.

The Order of an Element

Concept
Order of an element

The order of g∈Gg \in G, written ∣g∣|g| or ord⁡(g)\operatorname{ord}(g), is the smallest positive integer kk with gk=eg^k = e (additively, the smallest kk with kg=0kg = 0); if no such kk exists, gg has infinite order. Every element of a finite group has finite order, since the powers g,g2,g3,…g, g^2, g^3, \ldots cannot all be distinct, and a repeat gi=gjg^i = g^j (i<ji<j) forces gj−i=eg^{j-i} = e.

Order measures how long it takes an element, repeated, to cycle back to doing nothing — a 120°120° rotation returns after 33 steps, a reflection after 22. Every worked example so far secretly computed an order: the smallest kk with 4k≡0(mod10)4k \equiv 0 \pmod{10}, or the smallest kk with 3k≡1(mod7)3^k \equiv 1 \pmod7. The subgroup {e,g,g2,…,gk−1}\{e, g, g^2, \ldots, g^{k-1}\} generated by an element of order kk has exactly kk elements, since it cannot repeat before gkg^k by the minimality of kk, and it must repeat by then — so order and subgroup size are really the same number in disguise, the start of the story that Lagrange's theorem (Topic 5) finishes.

Example
Order in an additive group

Find ord⁡(8)\operatorname{ord}(8) in Z20\mathbb Z_{20}. Add 88 to itself repeatedly: 8,16,24≡4,12,20≡08, 16, 24\equiv4, 12, 20\equiv0 — reached 00 after 55 steps, with no earlier partial sum equal to 00. So ord⁡(8)=5\operatorname{ord}(8) = 5 in Z20\mathbb Z_{20}, matching the general formula ord⁡(a)=n/gcd⁡(n,a)\operatorname{ord}(a) = n/\gcd(n,a) for (Zn,+)(\mathbb Z_n,+): here 20/gcd⁡(20,8)=20/4=520/\gcd(20,8) = 20/4 = 5.

Example
Order in a multiplicative group

Find ord⁡(2)\operatorname{ord}(2) in U(11)U(11). The powers of 22 mod 1111 are 2,4,8,16≡5,10,20≡9,18≡7,14≡3,6,12≡12, 4, 8, 16\equiv5, 10, 20\equiv9, 18\equiv7, 14\equiv3, 6, 12\equiv1 — all ten nonzero residues appear before returning to 11, so ord⁡(2)=10=∣U(11)∣\operatorname{ord}(2) = 10 = |U(11)|: 22 is a generator of U(11)U(11).

Example
If ord⁡(g)=12\operatorname{ord}(g) = 12, what is ord⁡(g9)\operatorname{ord}(g^9)?

By the formula ord⁡(gk)=ord⁡(g)gcd⁡(ord⁡(g),k)\operatorname{ord}(g^k) = \dfrac{\operatorname{ord}(g)}{\gcd(\operatorname{ord}(g),k)} (proved below), ord⁡(g9)=12gcd⁡(12,9)=123=4\operatorname{ord}(g^9) = \dfrac{12}{\gcd(12,9)} = \dfrac{12}3 = 4. Test it concretely with g=rg = r, the rotation of order 1212 in D12D_{12}: r9r^9 applied four times is r36=r36 mod 12=r0=er^{36} = r^{36\bmod12} = r^0 = e, and r9,r18=r6,r27=r3r^9, r^{18} = r^6, r^{27} = r^3 are all different from ee, so 44 really is smallest.

Side note
Where it goes wrong

A common guess is ord⁡(ab)=lcm⁡(ord⁡(a),ord⁡(b))\operatorname{ord}(ab) = \operatorname{lcm}(\operatorname{ord}(a), \operatorname{ord}(b)). This is only valid when aa and bb commute and generate subgroups meeting only at ee — exactly the situation in a direct product (Topic 8). It fails badly otherwise: in D3D_3, ord⁡(r)=3\operatorname{ord}(r) = 3 and ord⁡(s)=2\operatorname{ord}(s) = 2, yet ord⁡(rs)=2\operatorname{ord}(rs) = 2, not lcm⁡(3,2)=6\operatorname{lcm}(3,2) = 6 — indeed D3D_3 has only 66 elements total and is not cyclic, so no element of D3D_3 could have order 66.

Try it
Try it: order in U(18)U(18)

Find ord⁡(7)\operatorname{ord}(7) in U(18)U(18).

Show the answer

Powers of 77 mod 1818: 7,72=49≡13,73=7⋅13=91≡17, 7^2 = 49\equiv13, 7^3 = 7\cdot13 = 91\equiv1 — back to 11 after three steps. So ord⁡(7)=3\operatorname{ord}(7) = \mathbf3 in U(18)U(18) (which has order φ(18)=6\varphi(18) = 6, so 77 generates a subgroup of size 33, not the whole group).

Why It Works

Let gg have finite order mm. Claim: gk=eg^k = e if and only if m∣km \mid k. (⇐\Leftarrow) If k=mqk = mq then gk=(gm)q=eq=eg^k = (g^m)^q = e^q = e. (⇒\Rightarrow) Suppose gk=eg^k = e; write k=mq+rk = mq + r with 0≤r<m0 \le r < m (the division algorithm). Then e=gk=gmq+r=(gm)qgr=gre = g^k = g^{mq+r} = (g^m)^qg^r = g^r, so gr=eg^r = e with 0≤r<m0 \le r < m; since mm is the smallest positive exponent killing gg, this forces r=0r = 0, i.e. m∣km \mid k. Corollary (order of a power): ord⁡(gk)=mgcd⁡(m,k)\operatorname{ord}(g^k) = \dfrac m{\gcd(m,k)}. Let d=gcd⁡(m,k)d = \gcd(m,k). Then (gk)m/d=gkm/d=(gm)k/d=e(g^k)^{m/d} = g^{km/d} = (g^m)^{k/d} = e, since m∣km/dm \mid km/d (as d∣kd \mid k makes km/d=(k/d)mkm/d = (k/d)m a whole multiple of mm); so ord⁡(gk)\operatorname{ord}(g^k) divides m/dm/d. Conversely, if (gk)j=e(g^k)^j = e then gkj=eg^{kj} = e, so by the claim m∣kjm \mid kj; dividing through by dd gives md∣kd j\dfrac md \mid \dfrac kd\,j, and since gcd⁡(m/d,k/d)=1\gcd(m/d, k/d) = 1, this forces md∣j\dfrac md \mid j. Hence ord⁡(gk)=md\operatorname{ord}(g^k) = \dfrac md.

Represent a symmetry of the regular nn-gon with vertices labeled 0,1,…,n−10, 1, \ldots, n-1 around a circle by an affine map x↦ax+b(modn)x \mapsto ax+b \pmod n with a∈{1,−1}a \in \{1,-1\}: a=1a=1 preserves the cyclic order of the vertices (a rotation), a=−1a=-1 reverses it (a reflection). Composing two such maps, x↦a1(a2x+b2)+b1=a1a2x+(a1b2+b1)x \mapsto a_1(a_2x+b_2)+b_1 = a_1a_2x + (a_1b_2+b_1), is again of this form with new coefficient a1a2∈{1,−1}a_1a_2 \in \{1,-1\}, so rotations and reflections close under composition; every symmetry of the nn-gon is one of these 2n2n maps (one for each a∈{1,−1}a \in \{1,-1\} and b∈{0,…,n−1}b \in \{0,\ldots,n-1\}), so ∣Dn∣=2n|D_n| = 2n. Writing r ⁣:x↦x+1r\colon x \mapsto x+1 and s ⁣:x↦−xs\colon x \mapsto -x: since s(s(x))=−(−x)=xs(s(x)) = -(-x) = x, s−1=ss^{-1}=s, so (srs−1)(x)=s(r(s(x)))=s(−x+1)=−(−x+1)=x−1=r−1(x)(srs^{-1})(x) = s(r(s(x))) = s(-x+1) = -(-x+1) = x - 1 = r^{-1}(x). Hence srs−1=r−1srs^{-1} = r^{-1}, for every nn.