Structure and Classification of Cyclic Groups

What does every cyclic group actually look like, up to relabeling? This is a concrete question. It is also surprisingly total. A cyclic group is generated by a single element. So its entire structure is dictated by how powers of that element repeat. It turns out there are only two possibilities in the whole universe of groups. These are the integers under addition, or the integers modulo nn. Every cyclic group is a perfect copy of exactly one of these. This matters because cyclic groups are the atomic building blocks of abelian group theory. The Fundamental Theorem of Finite Abelian Groups decomposes every finite abelian group into a product of cyclic pieces. So understanding cyclic groups completely is a prerequisite for understanding all finite abelian groups. The big picture is one of the cleanest classification theorems in algebra. A single integer invariant (the order, possibly infinite) completely determines a cyclic group up to isomorphism. Beneath this classification lies an equally clean secondary structure. The subgroups of a cyclic group correspond exactly to the divisors of its order. They form a lattice that mirrors the divisor lattice of an integer. Generators, too, are pinned down precisely. An element aˉ\bar a generates Z/nZ\mathbb{Z}/n\mathbb{Z} exactly when gcd⁡(a,n)=1\gcd(a,n)=1. This gives φ(n)\varphi(n) generators. This article gives the complete classification of cyclic groups via Group Isomorphisms. It proves the correspondence between subgroups and divisors. It also works out generator-counting in full rigor. We assume familiarity with Cyclic Groups and Generators — the basic definitions and first examples. We build from there to the structural theorems. These theorems make cyclic groups so tractable and so useful as building blocks elsewhere in algebra.

Key takeaways
  • Every cyclic group is isomorphic to exactly one of Z\mathbb{Z} (infinite order) or Z/nZ\mathbb{Z}/n\mathbb{Z} (order nn).
  • The isomorphism type of a cyclic group is completely determined by a single number: its order.
  • Subgroups of Z/nZ\mathbb{Z}/n\mathbb{Z} correspond bijectively to positive divisors dd of nn, one subgroup of each order dd.
  • Subgroups of Z\mathbb{Z} are exactly nZn\mathbb{Z} for n≥0n \geq 0, all infinite cyclic except the trivial group.
  • The element aˉ∈Z/nZ\bar a \in \mathbb{Z}/n\mathbb{Z} generates the group iff gcd⁡(a,n)=1\gcd(a,n) = 1; there are φ(n)\varphi(n) generators.
  • Key idea: cyclic groups are classified entirely by their order, and their internal subgroup lattice is literally the divisor lattice of that order.

Where this lives today

Cyclic groups are the computational backbone of modern cryptography: the Diffie–Hellman key exchange and ElGamal encryption operate inside the cyclic group (Z/pZ)×(\mathbb{Z}/p\mathbb{Z})^\times or cyclic subgroups of elliptic curve groups, where the classification tells you exactly how many generators exist and the subgroup lattice tells you which smaller subgroups an attacker's element might accidentally fall into (a real concern exploited in small-subgroup attacks). In number theory, the group (Z/nZ)×(\mathbb{Z}/n\mathbb{Z})^\times being cyclic exactly when n∈{1,2,4,pk,2pk}n \in \{1,2,4,p^k,2p^k\} for odd prime pp underlies the theory of primitive roots and index calculus algorithms. In abstract algebra itself, cyclic groups are the indecomposable pieces in the Fundamental Theorem of Finite Abelian Groups, so every finite abelian group is literally built by taking products of the cyclic groups classified here. In coding theory, cyclic codes (a major class of error-correcting codes, including Reed–Solomon and BCH codes) are defined via ideals in F[x]/(xn−1)\mathbb{F}[x]/(x^n-1), whose additive structure is a cyclic group of order nn. In computer science, hash table probing sequences and pseudo-random number generators (linear congruential generators) rely on the divisor-lattice structure of Z/nZ\mathbb{Z}/n\mathbb{Z} to guarantee full-period cycles.

Ask the tutor: How does the classification of cyclic groups extend to prove the Fundamental Theorem of Finite Abelian Groups?

DefinitionCyclic Group

A group GG is cyclic if there exists an element g∈Gg \in G, called a generator, such that

G=⟨g⟩:={ gk:k∈Z }.G = \langle g \rangle := \{\, g^k : k \in \mathbb{Z} \,\}.

Equivalently, every element of GG is an integer power of gg (written additively, an integer multiple of gg). The order of GG, denoted ∣G∣|G|, is the number of elements of GG if finite, or ∞\infty otherwise. The order of an element g∈Gg \in G is the smallest positive integer nn with gn=eg^n = e, if such nn exists; otherwise gg has infinite order. Two groups GG and HH are isomorphic, written G≅HG \cong H, if there is a bijection ϕ:G→H\phi: G \to H satisfying ϕ(xy)=ϕ(x)ϕ(y)\phi(xy) = \phi(x)\phi(y) for all x,y∈Gx,y \in G.

Intuition

Picture a cyclic group as a single arrow being repeated: start at the identity, apply the generator gg over and over, and record where you land. If the arrow never returns to its starting point, you get an infinite ray of distinct positions — indexed exactly by the integers, since gi=gjg^i = g^j forces i=ji = j. If the arrow does eventually return after nn steps, the positions wrap around into a closed loop of exactly nn points, and the arithmetic of "where you land after kk steps" is precisely addition modulo nn. This single picture — an infinite line, or a closed loop with nn beads — is the entire classification. There is no room for a cyclic group to look like anything else, because the whole group is generated by iterating one element, and iterating one element can only ever produce a line or a loop. The subgroup structure follows the same logic: a subgroup of a loop of nn beads is a sub-loop that evenly divides the big loop, so subgroups correspond to divisors of nn.

Figure
The two shapes of a cyclic group: infinite line (Z\mathbb{Z}) and 6-bead loop (Z/6Z\mathbb{Z}/6\mathbb{Z})
Infinite cyclic group ℤ -2 -1 0 1 2 3 … … Finite cyclic group ℤ/6ℤ (loop of 6 beads) 0 1 2 3 4 5 g² = 2, g⁶ = 0 (wraps around)
TheoremClassification of Cyclic Groups

Let G=⟨g⟩G = \langle g \rangle be a cyclic group. 1. If gg has infinite order, then G≅ZG \cong \mathbb{Z} (under addition). 2. If gg has finite order nn, then G≅Z/nZG \cong \mathbb{Z}/n\mathbb{Z} (under addition), and ∣G∣=n|G| = n. Consequently, every cyclic group is isomorphic to exactly one of Z\mathbb{Z} or Z/nZ\mathbb{Z}/n\mathbb{Z} for some positive integer nn, and two cyclic groups are isomorphic if and only if they have the same order (allowing the value ∞\infty).

Proof

We use the First Isomorphism Theorem framework by explicitly constructing and analyzing the natural surjective homomorphism from Z\mathbb{Z} onto GG; this is a direct construction, not a use of any theorem built on top of the present classification, so there is no circularity.

Step 1: A surjective homomorphism ϕ:Z→G\phi: \mathbb{Z} \to G. Define ϕ:Z→G\phi: \mathbb{Z} \to G by ϕ(k)=gk\phi(k) = g^k. This is well-defined for every integer kk since GG is a group (inverses and all powers exist). It is a homomorphism because

ϕ(k+l)=gk+l=gkgl=ϕ(k)ϕ(l)\phi(k + l) = g^{k+l} = g^k g^l = \phi(k)\phi(l)

for all k,l∈Zk, l \in \mathbb{Z}, using only the group axioms (associativity and the laws of exponents, which follow from associativity by induction). It is surjective because G=⟨g⟩G = \langle g \rangle means every element of GG is exactly gkg^k for some k∈Zk \in \mathbb{Z}, i.e. every element of GG lies in the image of ϕ\phi.

Step 2: The kernel of ϕ\phi is a subgroup of Z\mathbb{Z}, hence of the form nZn\mathbb{Z}. Let K=ker⁡ϕ={k∈Z:gk=e}K = \ker \phi = \{ k \in \mathbb{Z} : g^k = e \}. Since ϕ\phi is a homomorphism, KK is a subgroup of Z\mathbb{Z} (kernels of homomorphisms are always subgroups: ϕ(0)=e\phi(0)=e so 0∈K0 \in K; if k,l∈Kk,l \in K then ϕ(k−l)=ϕ(k)ϕ(l)−1=e\phi(k-l) = \phi(k)\phi(l)^{-1} = e so k−l∈Kk - l \in K). We claim every subgroup of Z\mathbb{Z} has the form nZn\mathbb{Z} for some integer n≥0n \geq 0. If K={0}K = \{0\}, take n=0n = 0. Otherwise KK contains a nonzero integer, and since KK is closed under negation it contains a positive integer; let nn be the least positive element of KK (this exists by the well-ordering principle). We show K=nZK = n\mathbb{Z}.

  • nZ⊆Kn\mathbb{Z} \subseteq K: since n∈Kn \in K and KK is closed under addition and negation, all integer multiples of nn lie in KK.
  • K⊆nZK \subseteq n\mathbb{Z}: let k∈Kk \in K. By the division algorithm, k=qn+rk = qn + r with 0≤r<n0 \leq r < n. Then r=k−qn∈Kr = k - qn \in K since k∈Kk \in K, n∈Kn \in K, and KK is a subgroup (closed under subtraction and integer combinations). Since nn was the least positive element of KK and 0≤r<n0 \le r < n, we must have r=0r = 0. Hence k=qn∈nZk = qn \in n\mathbb{Z}.

So K=nZK = n\mathbb{Z} for a unique n≥0n \geq 0.

Step 3: Interpreting nn. Case n=0n = 0: Then K={0}K = \{0\}, so ϕ\phi is injective (its kernel is trivial: if ϕ(k)=ϕ(l)\phi(k) = \phi(l) then ϕ(k−l)=e\phi(k - l) = e so k−l∈K={0}k - l \in K = \{0\}, i.e. k=lk = l). An injective surjective homomorphism is an isomorphism, so ϕ:Z→G\phi: \mathbb{Z} \to G is an isomorphism and G≅ZG \cong \mathbb{Z}. Note K={0}K = \{0\} means no positive integer kk satisfies gk=eg^k = e, i.e. gg has infinite order — matching case (1) of the theorem. Case n>0n > 0: Then gn=eg^n = e and nn is the least positive integer with this property (by minimality of nn as the least positive element of KK), so gg has order exactly nn — matching case (2). We now build an explicit isomorphism ϕ‾:Z/nZ→G\overline{\phi}: \mathbb{Z}/n\mathbb{Z} \to G. Define ϕ‾(kˉ)=gk\overline{\phi}(\bar k) = g^k for kˉ∈Z/nZ\bar k \in \mathbb{Z}/n\mathbb{Z}.

  • Well-defined: if kˉ=lˉ\bar k = \bar l in Z/nZ\mathbb{Z}/n\mathbb{Z}, then k−l∈nZ=K=ker⁡ϕk - l \in n\mathbb{Z} = K = \ker\phi, so ϕ(k)=ϕ(l)\phi(k) = \phi(l), i.e. gk=glg^k = g^l. So the value ϕ‾(kˉ)\overline\phi(\bar k) does not depend on the choice of representative kk.
  • Homomorphism: ϕ‾(kˉ+lˉ)=ϕ‾(k+l‾)=gk+l=gkgl=ϕ‾(kˉ)ϕ‾(lˉ)\overline\phi(\bar k + \bar l) = \overline\phi(\overline{k+l}) = g^{k+l} = g^k g^l = \overline\phi(\bar k)\overline\phi(\bar l).
  • Surjective: for any x∈Gx \in G, since G=⟨g⟩G = \langle g\rangle, x=gk=ϕ‾(kˉ)x = g^k = \overline\phi(\bar k) for some kk.
  • Injective: if ϕ‾(kˉ)=ϕ‾(lˉ)\overline\phi(\bar k) = \overline\phi(\bar l) then gk−l=eg^{k - l} = e, so k−l∈K=nZk - l \in K = n\mathbb{Z}, so kˉ=lˉ\bar k = \bar l in Z/nZ\mathbb{Z}/n\mathbb{Z}.

So ϕ‾\overline\phi is an isomorphism, giving G≅Z/nZG \cong \mathbb{Z}/n\mathbb{Z}. Since Z/nZ\mathbb{Z}/n\mathbb{Z} has exactly the nn elements 0ˉ,1ˉ,…,n−1‾\bar 0, \bar 1, \dots, \overline{n-1}, we get ∣G∣=n|G| = n.

Step 4: Same order ⇒\Rightarrow isomorphic; different order ⇒\Rightarrow not isomorphic. By Steps 1–3, every cyclic group of order nn (finite or ∞\infty) is isomorphic to the same fixed model (Z\mathbb{Z} or Z/nZ\mathbb{Z}/n\mathbb{Z}), so any two cyclic groups of the same order are isomorphic to each other (isomorphism is transitive and symmetric, since composing/inverting bijective homomorphisms gives bijective homomorphisms). Conversely, isomorphic groups have the same cardinality, so groups of different order cannot be isomorphic. This completes the classification. ■\blacksquare

∎
LemmaSubgroups of Z\mathbb{Z}

Every subgroup of Z\mathbb{Z} equals nZn\mathbb{Z} for a unique integer n≥0n \geq 0. Every such subgroup is cyclic, generated by nn (or trivial if n=0n=0).

Proof

This was proved as Step 2 of the theorem above, via the well-ordering principle and the division algorithm; uniqueness of n≥0n \ge 0 follows since nZ=mZn\mathbb Z = m\mathbb Z with n,m≥0n,m \ge 0 forces n∣mn \mid m and m∣nm \mid n, hence n=mn = m. Each nZn\mathbb{Z} is generated by nn by definition, so is cyclic. ■\blacksquare

∎
TheoremDivisor Lattice of Subgroups of Z/nZ\mathbb{Z}/n\mathbb{Z}

Let n≥1n \geq 1 and let G=Z/nZG = \mathbb{Z}/n\mathbb{Z}. Then: 1. For every positive divisor dd of nn, there is a unique subgroup of GG of order dd, namely

Hd=⟨n/d‾⟩={ 0‾,n/d‾,2n/d‾,… }.H_d = \langle \overline{n/d} \rangle = \{\, \overline{0}, \overline{n/d}, \overline{2n/d}, \dots \,\}.

2. Every subgroup of GG is cyclic and arises this way; hence subgroups of GG correspond bijectively to positive divisors of nn. 3. For divisors d1∣d2d_1 \mid d_2 of nn, the corresponding subgroups satisfy Hd1⊆Hd2H_{d_1} \subseteq H_{d_2}; the subgroup lattice of GG is isomorphic (as a poset under inclusion) to the divisor lattice of nn under divisibility.

Proof

We use the correspondence with subgroups of Z\mathbb Z established in the preceding Lemma, transported through the quotient map — a standard and non-circular technique since it only uses facts about Z\mathbb{Z} already proved.

Setup. Let π:Z→Z/nZ\pi: \mathbb{Z} \to \mathbb{Z}/n\mathbb{Z} be the canonical surjective homomorphism π(k)=kˉ\pi(k) = \bar k, with kernel nZn\mathbb{Z}.

Every subgroup of GG is the image of a subgroup of Z\mathbb{Z} containing nZn\mathbb{Z}. Let H≤GH \leq G be any subgroup. Let H~=π−1(H)={k∈Z:kˉ∈H}\widetilde H = \pi^{-1}(H) = \{k \in \mathbb{Z} : \bar k \in H\}. This is a subgroup of Z\mathbb{Z}: it contains 00 (since 0ˉ=e∈H\bar 0 = e \in H), and if k,l∈H~k, l \in \widetilde H then kˉ,lˉ∈H\bar k, \bar l \in H so k−l‾=kˉ−lˉ∈H\overline{k - l} = \bar k - \bar l \in H (as HH is a subgroup), so k−l∈H~k - l \in \widetilde H. Also nZ⊆H~n\mathbb{Z} \subseteq \widetilde{H} since π(nZ)={0ˉ}⊆H\pi(n\mathbb Z) = \{\bar 0\} \subseteq H. By the Lemma, H~=mZ\widetilde H = m\mathbb{Z} for a unique m≥0m \geq 0; since n∈nZ⊆H~=mZn \in n\mathbb Z \subseteq \widetilde H = m\mathbb{Z} and n>0n > 0, we must have m>0m > 0 and m∣nm \mid n (because n∈mZn \in m\mathbb{Z} means mm divides nn). Then

H=π(H~)=π(mZ)={0‾,m‾,2m‾,… }=⟨mˉ⟩.H = \pi(\widetilde H) = \pi(m\mathbb{Z}) = \{\overline{0}, \overline{m}, \overline{2m}, \dots\} = \langle \bar m \rangle.

So every subgroup HH of GG is cyclic, generated by mˉ\bar m for some positive divisor mm of nn.

Order of H=⟨mˉ⟩H = \langle \bar m\rangle. The order of mˉ\bar m in GG is the least positive tt with tm≡0(modn)t m \equiv 0 \pmod n, i.e. the least positive tt with n∣tmn \mid tm, i.e. t=n/gcd⁡(m,n)t = n/\gcd(m,n). Since m∣nm \mid n, gcd⁡(m,n)=m\gcd(m,n) = m, so the order of mˉ\bar m, and hence of H=⟨mˉ⟩H = \langle \bar m \rangle, is exactly n/mn/m. Writing d=n/md = n/m (a positive divisor of nn, since m∣nm \mid n), we get ∣H∣=d|H| = d and H=⟨n/d‾⟩=HdH = \langle \overline{n/d}\rangle = H_d as defined in the statement.

Existence for every divisor d∣nd \mid n. Given any positive divisor dd of nn, set m=n/dm = n/d (also a positive divisor of nn); then Hd:=⟨mˉ⟩H_d := \langle \bar m \rangle is a subgroup of order dd by the computation above. This proves existence.

Uniqueness for each order dd. Suppose H,H′H, H' are both subgroups of GG of order d∣nd \mid n. By the argument above, H=⟨n/d‾⟩H = \langle \overline{n/d}\rangle and H′=⟨n/d′‾⟩H' = \langle \overline{n/d'}\rangle where d,d′d, d' are both forced (by the order computation) to equal the actual order of the respective subgroup; since both have order dd, both equal ⟨n/d‾⟩=Hd\langle \overline{n/d} \rangle = H_d. Hence H=H′H = H': there is only one subgroup of each order d∣nd \mid n.

Bijectivity of d↦Hdd \mapsto H_d. We have shown: every subgroup of GG equals HdH_d for a unique divisor d∣nd \mid n (its order), and every divisor d∣nd \mid n gives rise to exactly one subgroup HdH_d. So d↦Hdd \mapsto H_d is a bijection from positive divisors of nn to subgroups of GG.

Inclusion reflects divisibility. Suppose d1∣d2d_1 \mid d_2, both dividing nn. Write d2=d1ed_2 = d_1 e. Then

Hd1=⟨n/d1‾⟩,Hd2=⟨n/d2‾⟩=⟨n/(d1e)‾⟩.H_{d_1} = \langle \overline{n/d_1} \rangle, \qquad H_{d_2} = \langle \overline{n/d_2}\rangle = \langle \overline{n/(d_1 e)} \rangle.

Since n/d1=e⋅(n/d2)n/d_1 = e \cdot (n/d_2), the generator of Hd1H_{d_1} is an integer multiple of the generator of Hd2H_{d_2} modulo nn — more precisely n/d1‾=e⋅n/d2‾∈Hd2\overline{n/d_1} = e\cdot \overline{n/d_2} \in H_{d_2} — so Hd1⊆Hd2H_{d_1} \subseteq H_{d_2} (as Hd2H_{d_2} is closed under taking multiples of its generator, being a subgroup generated by it). Conversely if Hd1⊆Hd2H_{d_1} \subseteq H_{d_2} then since ∣Hd1∣=d1|H_{d_1}| = d_1 divides ∣Hd2∣=d2|H_{d_2}| = d_2 by Lagrange's Theorem (a subgroup's order divides the group's order), we get d1∣d2d_1 \mid d_2. So Hd1⊆Hd2  ⟺  d1∣d2H_{d_1} \subseteq H_{d_2} \iff d_1 \mid d_2, exhibiting the bijection d↦Hdd \mapsto H_d as an order isomorphism between the divisor poset of nn and the subgroup poset of GG under inclusion. ■\blacksquare

∎

Worked examples

Example 1: The full subgroup lattice of Z/24Z\mathbb{Z}/24\mathbb{Z}. Let n=24=23⋅3n = 24 = 2^3 \cdot 3. Its positive divisors are 1,2,3,4,6,8,12,241, 2, 3, 4, 6, 8, 12, 24 — eight of them. By the theorem, Z/24Z\mathbb{Z}/24\mathbb{Z} has exactly eight subgroups, one of each of these orders, and the subgroup of order dd is Hd=⟨24/d‾⟩H_d = \langle \overline{24/d}\rangle. Let us compute all eight explicitly:

  • d=1d=1: H1=⟨24‾⟩=⟨0ˉ⟩={0ˉ}H_1 = \langle \overline{24}\rangle = \langle \bar 0 \rangle = \{\bar 0\}.
  • d=2d=2: H2=⟨12‾⟩={0ˉ,12‾}H_2 = \langle \overline{12}\rangle = \{\bar 0, \overline{12}\}.
  • d=3d=3: H3=⟨8‾⟩={0ˉ,8‾,16‾}H_3 = \langle \overline{8}\rangle = \{\bar 0, \overline{8}, \overline{16}\}.
  • d=4d=4: H4=⟨6‾⟩={0ˉ,6‾,12‾,18‾}H_4 = \langle \overline{6}\rangle = \{\bar 0, \overline 6, \overline{12}, \overline{18}\}.
  • d=6d=6: H6=⟨4‾⟩={0ˉ,4‾,8‾,12‾,16‾,20‾}H_6 = \langle \overline 4\rangle = \{\bar0, \overline4,\overline8,\overline{12},\overline{16},\overline{20}\}.
  • d=8d=8: H8=⟨3‾⟩={0ˉ,3‾,6‾,9‾,12‾,15‾,18‾,21‾}H_8 = \langle \overline 3 \rangle = \{\bar0,\overline3,\overline6,\overline9,\overline{12},\overline{15},\overline{18},\overline{21}\}.
  • d=12d=12: H12=⟨2‾⟩H_{12} = \langle \overline 2 \rangle, all even residues, 12 elements.
  • d=24d=24: H24=⟨1‾⟩=Z/24ZH_{24} = \langle \overline 1 \rangle = \mathbb{Z}/24\mathbb{Z} itself.

Check divisibility gives inclusion, e.g. 4∣124 \mid 12 so H4⊆H12H_4 \subseteq H_{12}: indeed {0ˉ,6ˉ,12‾,18‾}⊂{0ˉ,2ˉ,4ˉ,…,22‾}\{\bar0,\bar6,\overline{12},\overline{18}\} \subset \{\bar0,\bar2,\bar4,\dots,\overline{22}\} (all even residues), which is true since 6,12,186,12,18 are even. Also 6∤86 \nmid 8 (since 8/68/6 is not an integer), and correspondingly H6⊈H8H_6 \not\subseteq H_8: indeed 4‾∈H6\overline 4 \in H_6 but 4‾∉H8={0ˉ,3ˉ,6ˉ,… }\overline 4 \notin H_8 = \{\bar0,\bar3,\bar6,\dots\} (multiples of 33 mod 2424), confirming the lattice faithfully mirrors divisibility rather than mere size comparison.

Example 2: Counting and identifying generators of Z/18Z\mathbb{Z}/18\mathbb{Z}. We determine exactly which elements generate G=Z/18ZG=\mathbb{Z}/18\mathbb{Z}, using the fact (proved just below as a corollary) that aˉ\bar a generates Z/nZ\mathbb{Z}/n\mathbb{Z} iff gcd⁡(a,n)=1\gcd(a,n)=1. For n=18=2⋅32n = 18 = 2\cdot 3^2, we scan a=0,…,17a = 0, \dots, 17 and compute gcd⁡(a,18)\gcd(a, 18):

a01234567891011121314151617gcd⁡(a,18)1812321612921612321\begin{array}{c|cccccccccccccccccc} a & 0&1&2&3&4&5&6&7&8&9&10&11&12&13&14&15&16&17\\\hline \gcd(a,18)&18&1&2&3&2&1&6&1&2&9&2&1&6&1&2&3&2&1 \end{array}

The values with gcd⁡(a,18)=1\gcd(a,18) = 1 are a=1,5,7,11,13,17a = 1,5,7,11,13,17 — exactly 66 values. Euler's totient confirms this: φ(18)=φ(2)φ(9)=1⋅6=6\varphi(18) = \varphi(2)\varphi(9) = 1 \cdot 6 = 6 (using φ(32)=32−3=6\varphi(3^2) = 3^2 - 3 = 6 and multiplicativity of φ\varphi over coprime factors 22 and 99). So Z/18Z\mathbb{Z}/18\mathbb{Z} has exactly 66 generators: 1‾,5‾,7‾,11‾,13‾,17‾\overline 1, \overline 5, \overline 7, \overline{11}, \overline{13}, \overline{17}. As a sanity check, 5‾\overline 5 should generate the whole group: its multiples mod 18 are 5,10,15,2,7,12,17,4,9,14,1,6,11,16,3,8,13,05,10,15,2,7,12,17,4,9,14,1,6,11,16,3,8,13,0 — eighteen distinct values before repeating, confirming 5‾\overline5 has order 1818 and generates GG. Contrast with 6‾\overline 6: gcd⁡(6,18)=6≠1\gcd(6,18)=6\ne 1, and indeed 6,12,0,6,…6,12,0,6,\dots cycles after only 33 steps, generating the small subgroup {0,6,12}\{0,6,12\} of order 33, not the whole group — consistent with the classification, since 6‾\overline 6 generates H3H_3 (the unique subgroup of order 33) rather than all of Z/18Z\mathbb Z/18\mathbb Z.

CorollaryGenerators of Z/nZ\mathbb{Z}/n\mathbb{Z}

For n≥1n \geq 1, the element aˉ∈Z/nZ\bar a \in \mathbb{Z}/n\mathbb{Z} generates Z/nZ\mathbb{Z}/n\mathbb{Z} if and only if gcd⁡(a,n)=1\gcd(a,n) = 1. Consequently Z/nZ\mathbb{Z}/n\mathbb{Z} has exactly φ(n)\varphi(n) generators, where φ\varphi is Euler's totient function.

Proof

By the subgroup-order computation in the proof of the Divisor Lattice theorem (applied with mm replaced by aa, not assumed to divide nn; the same computation generalizes): the order of aˉ\bar a in Z/nZ\mathbb{Z}/n\mathbb{Z} is n/gcd⁡(a,n)n/\gcd(a,n). Indeed, the order is the least positive tt with n∣tan \mid ta; writing g=gcd⁡(a,n)g=\gcd(a,n), a=ga′a = g a', n=gn′n = g n' with gcd⁡(a′,n′)=1\gcd(a',n')=1, the condition n∣tan \mid ta becomes gn′∣tga′gn' \mid tga', i.e. n′∣ta′n' \mid ta', and since gcd⁡(a′,n′)=1\gcd(a',n')=1 this forces n′∣tn' \mid t; the least such positive tt is t=n′=n/gt=n'=n/g. Now aˉ\bar a generates Z/nZ\mathbb{Z}/n\mathbb{Z} iff ⟨aˉ⟩\langle \bar a\rangle has order nn (equal to the whole group), i.e. iff n/gcd⁡(a,n)=nn/\gcd(a,n) = n, i.e. iff gcd⁡(a,n)=1\gcd(a,n)=1. The number of generators is therefore the number of a∈{0,1,…,n−1}a \in \{0,1,\dots,n-1\} (a complete set of representatives) with gcd⁡(a,n)=1\gcd(a,n)=1, which is exactly the definition of φ(n)\varphi(n). ■\blacksquare

∎
CounterexampleWhy the "single generator" hypothesis is needed

The entire classification collapses without cyclicity. Consider the Klein four-group V=Z/2Z×Z/2Z={(0,0),(1,0),(0,1),(1,1)}V = \mathbb{Z}/2\mathbb{Z} \times \mathbb{Z}/2\mathbb{Z} = \{(0,0),(1,0),(0,1),(1,1)\} with componentwise addition mod 2. It has order 44, the same order as Z/4Z\mathbb{Z}/4\mathbb{Z}, yet

V≇Z/4Z.V \not\cong \mathbb{Z}/4\mathbb{Z}.

Indeed, in Z/4Z\mathbb{Z}/4\mathbb{Z} the element 1ˉ\bar 1 has order 44 (it generates the whole group), so Z/4Z\mathbb{Z}/4\mathbb{Z} is cyclic. But in VV, every non-identity element has order 22: for instance (1,0)+(1,0)=(0,0)(1,0)+(1,0) = (0,0), and likewise for (0,1)(0,1) and (1,1)(1,1). So VV has no element of order 4, hence no single element generates VV — VV is not cyclic at all. This shows that "same order" alone never implies isomorphism among all groups; the classification theorem's power comes specifically from the hypothesis that GG is cyclic (generated by one element). Without that hypothesis, groups of the same order can have entirely different structure — here, different subgroup lattices (VV has three subgroups of order 22, while Z/4Z\mathbb{Z}/4\mathbb{Z} has exactly one, matching the unique-subgroup-per-divisor conclusion of the Divisor Lattice theorem, which also requires cyclicity and fails for VV).

Scope, edge cases & common traps

SituationWhat holdsWhat can go wrong if misapplied
GG cyclic, infinite orderG≅ZG \cong \mathbb{Z}Assuming infinite cyclic groups can differ "in size" — they cannot; all are isomorphic to Z\mathbb Z
GG cyclic, order nnG≅Z/nZG \cong \mathbb{Z}/n\mathbb{Z}, subgroups ↔\leftrightarrow divisors of nnForgetting subgroup order must divide nn (Lagrange) before hunting for a subgroup of a given order
GG non-cyclic, order nnNo such classification; may have many isomorphism typesAssuming "same order ⇒\Rightarrow isomorphic" for general groups (false, see Klein four counterexample)
Trivial group {e}\{e\}Cyclic, order 11, corresponds to d=1d=1 divisorSometimes mistakenly excluded from "cyclic" — it is cyclic (generated by ee)
n=0n = 0 "Z/0Z\mathbb Z/0\mathbb Z"By convention Z/0Z≅Z\mathbb{Z}/0\mathbb{Z} \cong \mathbb{Z}Confusing this degenerate notation with a genuine finite group

The mathematician behind it

Origins · 1 minCarl Friedrich Gauss and how this was discovered
CF
Carl Friedrich Gauss
1777–1855 · Modular arithmetic and the founding of number theory as a rigorous science

Gauss did not phrase his work in the language of "cyclic groups" — that abstraction came later — but his Disquisitiones Arithmeticae (1801) is the true origin of the entire subject. He systematically studied arithmetic modulo nn, introduced the congruence notation still used today, proved what amounts to the structure of (Z/nZ)×(\mathbb{Z}/n\mathbb{Z})^\times and the existence of primitive roots for prime moduli, and effectively computed with generators and orders of elements decades before "group" was a formal word. His clock-arithmetic viewpoint is exactly the loop-of-beads picture underlying finite cyclic groups.

History
- 1801: Gauss publishes Disquisitiones Arithmeticae, developing modular arithmetic and orders of elements mod nn — the arithmetic core of finite cyclic groups.
- 1830s–1840s: Galois and Cauchy develop permutation group ideas, setting the stage for abstract group axioms.
- 1854: Cayley gives an early abstract definition of a group, detached from permutations or numbers.
- 1870: Kronecker states and essentially proves structure results for finite abelian groups, closely tied to cyclic decomposition.
- Late 1800s–early 1900s: The modern axiomatic treatment (Weber, Dedekind, and later Noether's school) crystallizes the cyclic group classification as a clean isomorphism theorem, exactly as presented here.
Is every subgroup of a cyclic group also cyclic?

Yes. If G=⟨g⟩G = \langle g \rangle is cyclic and H≤GH \le G is any subgroup, then HH is cyclic. For G≅ZG \cong \mathbb{Z}, this is the Lemma on subgroups of Z\mathbb{Z}: every subgroup is nZn\mathbb{Z}, generated by nn. For G≅Z/nZG \cong \mathbb{Z}/n\mathbb{Z}, the Divisor Lattice theorem shows every subgroup equals Hd=⟨n/d‾⟩H_d = \langle \overline{n/d}\rangle for some divisor d∣nd \mid n, hence is cyclic, generated by a single element. This is a genuinely useful shortcut: to show a subgroup of a cyclic group is cyclic, you never need to search for a generator by hand — its existence is guaranteed automatically by the classification.

Why can't two cyclic groups of different finite order be isomorphic?

Isomorphisms are bijections, so isomorphic groups must have exactly the same number of elements. A cyclic group of order mm has mm elements and one of order n≠mn \ne m has nn elements; if m≠nm \ne n no bijection between the underlying sets can exist, let alone a bijective homomorphism. This is why "order" is a complete isomorphism invariant for cyclic groups specifically — for non-cyclic groups, order alone is not enough, as the Klein four-group versus Z/4Z\mathbb{Z}/4\mathbb{Z} counterexample shows.

Does the classification theorem also tell us about infinite subgroups of Z/nZ\mathbb{Z}/n\mathbb{Z}?

There are none to worry about: Z/nZ\mathbb{Z}/n\mathbb{Z} is a finite group, so every subgroup is automatically finite, with order dividing nn by Lagrange's Theorem. The Divisor Lattice theorem completely lists all of them: exactly one subgroup for each positive divisor of nn. Infinite subgroups only become relevant when G≅ZG \cong \mathbb{Z}, where the Lemma on subgroups of Z\mathbb{Z} shows every nonzero subgroup nZn\mathbb{Z} (for n≥1n \ge 1) is itself infinite (isomorphic to Z\mathbb{Z} again), while only the trivial subgroup {0}\{0\} is finite.

How does this classification foreshadow the Fundamental Theorem of Finite Abelian Groups?

Every finite abelian group decomposes as a direct product of cyclic groups of prime-power order, uniquely up to reordering — this is the Fundamental Theorem of Finite Abelian Groups. That theorem is only meaningful because we first know, as proved here, exactly what a cyclic group is up to isomorphism: a single number (its order) suffices to pin it down completely. The finite abelian classification is thus a direct generalization, replacing "one number" with "one multiset of prime-power numbers," but it rests entirely on the cyclic case being airtight first.

If gcd⁡(a,n)=d>1\gcd(a,n) = d > 1, what does aˉ\bar a generate in Z/nZ\mathbb{Z}/n\mathbb{Z}?

By the order formula proved in the Generators corollary, aˉ\bar a has order n/gcd⁡(a,n)=n/dn/\gcd(a,n) = n/d. By the uniqueness part of the Divisor Lattice theorem, there is exactly one subgroup of that order, namely Hn/d=⟨d‾⟩H_{n/d} = \langle \overline{d} \rangle (since gcd⁡(a,n)=d\gcd(a,n)=d implies aa and dd generate the same cyclic subgroup — both have order n/dn/d, and by uniqueness of the subgroup of that order they must coincide). So ⟨aˉ⟩=⟨dˉ⟩\langle \bar a\rangle = \langle \bar d\rangle, a strictly smaller subgroup than the whole group whenever d>1d > 1.

Can a cyclic group have more than one generator, and if so how many for Z\mathbb{Z} itself?

Finite cyclic groups typically have several generators — Z/nZ\mathbb{Z}/n\mathbb{Z} has exactly φ(n)\varphi(n) of them. The infinite cyclic group Z\mathbb{Z} is a boundary case: it has exactly two generators, 11 and −1-1, since ⟨k⟩=kZ\langle k \rangle = k\mathbb{Z} equals all of Z\mathbb{Z} only when k=±1k = \pm 1 (any ∣k∣≥2|k|\ge 2 gives kZ⊊Zk\mathbb Z \subsetneq \mathbb Z, e.g. 2∉3Z2 \notin 3\mathbb Z). This matches the general pattern that a generator must have order equal to the whole group's order, and in Z\mathbb{Z} only ±1\pm 1 generate every integer by repeated addition.

Is Z/nZ\mathbb{Z}/n\mathbb{Z} under multiplication also relevant to this classification?

No — Z/nZ\mathbb{Z}/n\mathbb{Z} under multiplication is not even a group (0 has no inverse, and other elements may lack inverses too), so it falls outside this theorem entirely. The relevant multiplicative object is the unit group (Z/nZ)×(\mathbb{Z}/n\mathbb{Z})^\times, consisting of residues coprime to nn under multiplication; it has order φ(n)\varphi(n) but is cyclic only for special nn (namely n=1,2,4,pk,2pkn = 1,2,4,p^k,2p^k for odd prime pp). This is a genuinely separate and harder theorem than the one proved here, since it concerns a different group built from Z/nZ\mathbb{Z}/n\mathbb{Z}, not Z/nZ\mathbb{Z}/n\mathbb{Z} itself under addition.

Before this

Where this leads