Group Homomorphisms and Isomorphisms

Maps between groups that respect the operation. Examples from the sign of a permutation, the rotation-or-reflection sign of a dihedral symmetry, reduction modulo n, powers of a fixed element and the vertex permutations of the octagon, and maps that fail; first properties, proved in full; the kernel, a subgroup that decides when a homomorphism is one-to-one; homomorphisms out of a cyclic group and how to count them; isomorphisms and the properties they preserve, used to sort five groups of order 16 and to identify U(32) with Z_2 x Z_8; the classification of cyclic groups as Z and Z_n; and conjugation as an automorphism.

The group of units $U(9)$ and the additive group $\mathbb{Z}_6$ have the same Cayley table after a renaming. The first article of this unit listed $U(9)$ as the powers $1, 2, 4, 8, 7, 5$ of 2. Its multiplication table then became the addition table of $\mathbb{Z}_6$, because multiplying powers of 2 adds their exponents. The renaming is

$k$ in $\mathbb{Z}_6$ 0 1 2 3 4 5
$2^k$ in $U(9)$ 1 2 4 8 7 5

and the rule $2^j \cdot 2^k = 2^{j + k}$ says that it turns sums in $\mathbb{Z}_6$ into products in $U(9)$. Not every renaming works. Sending each unit $a$ to its remainder modulo 6 sends $2 \cdot 4 = 8$ to 2, but $2 + 4 = 6 = 0$ in $\mathbb{Z}_6$, and the script confirms that this map fails. What exactly must a map between two groups do for the tables to match, and what can a map tell us when it merges elements instead of renaming them?

A renaming that matches the tables is an isomorphism. The broader notion, a map that respects the operation while possibly merging elements, is a homomorphism. The ideas repeat ring homomorphisms with one operation instead of two, and the proofs are often the ring proofs with addition removed. But groups have more homomorphisms than rings, and more varied ones. The sign of a permutation is one. So are the passage from a symmetry of the octagon to the permutation of its corners, the powers of a single element, and conjugation by a fixed element.

First we collect examples, and maps that fail. Then we prove the first properties, meet the kernel and count the homomorphisms out of a cyclic group. Finally we ask what an isomorphism preserves, sort five groups of order 16, classify the cyclic groups and meet conjugation.

This is the seventh article of Groups. It covers §7.4 of the course text (Hungerford, 2014), except Cayley's theorem, which has its own article, with Judson (2026, §9.1) as a free companion. It uses Basic Properties of Groups and Subgroups and Cyclic Groups. It is the home of group homomorphisms, kernels and isomorphisms, on which the next unit builds.

Definitions

Definition. Let $G$ and $H$ be groups. A function $f: G \to H$ is a homomorphism if

$$f(ab) = f(a)\,f(b) \qquad \text{for all } a, b \in G.$$

A homomorphism that is a bijection is an isomorphism, and $G$ and $H$ are isomorphic, written $G \cong H$, if there is an isomorphism from $G$ to $H$. An isomorphism from $G$ to itself is an automorphism of $G$.

On the left the product is taken in $G$, on the right in $H$, and the two operations may be written differently. For a homomorphism from the additive group $\mathbb{Z}_6$ to the multiplicative group $U(9)$, the condition reads $f(j + k) = f(j)f(k)$. From a multiplicative group to an additive one it reads $f(ab) = f(a) + f(b)$, as for logarithms. Unlike the ring case, there is only one condition to check. There is no separate question about identities either: a group homomorphism always sends the identity to the identity, as shown below.

What do homomorphisms look like in this unit's groups?

Examples

The sign of a permutation. $\operatorname{sgn}: S_n \to \{1, -1\}$ is a homomorphism to the multiplicative group $\{1, -1\}$, because $\operatorname{sgn}(\alpha\beta) = \operatorname{sgn}(\alpha)\operatorname{sgn}(\beta)$ (Symmetric Groups and Permutations). The script checks all $120^2$ products in $S_5$.

Rotation or reflection. The map $D_n \to \{1, -1\}$ sending rotations to $1$ and reflections to $-1$ is a homomorphism; Dihedral Groups read this off its product rule, in which the exponent of $s$ adds modulo 2. In the notation $x \mapsto \varepsilon x + k$ for a symmetry acting on vertex labels, the map is $\varepsilon$.

Reduction modulo $n$. The map $\mathbb{Z} \to \mathbb{Z}_n$, $a \mapsto [a]$, is a homomorphism of additive groups, since $[a + b] = [a] + [b]$; it is also a ring homomorphism (Ring Homomorphisms and Isomorphisms). Changing the modulus works when the new modulus divides the old: $\mathbb{Z}_{36} \to \mathbb{Z}_9$, $[a]_{36} \mapsto [a]_9$, is a well-defined homomorphism onto $\mathbb{Z}_9$.

Powers of an element. For any element $a$ of a group $G$, the map

$$\exp_a: \mathbb{Z} \to G, \qquad k \mapsto a^k,$$

is a homomorphism from the additive group $\mathbb{Z}$, because $a^{j + k} = a^j a^k$ is the first law of exponents. Its image is the cyclic subgroup $\langle a \rangle$. This map turns out to be the whole story of cyclic groups.

Symmetries as permutations. Sending each symmetry of the octagon to the permutation it induces on the eight vertex labels is a one-to-one homomorphism $D_8 \to S_8$, with an image of 16 elements; Dihedral Groups describes the cycle types of these permutations. In the same way each move of the Rubik's cube permutes the eight corner positions, giving a homomorphism from the cube group to $S_8$.

Trivial maps and inclusions. The map sending every element of $G$ to $e_H$ is a homomorphism. If $K$ is a subgroup of $G$, the inclusion $K \to G$, $x \mapsto x$, is a one-to-one homomorphism, and the identity map of $G$ is an automorphism.

A renamed operation. The operation $x * y = (x + y)/(1 + xy)$ on the interval $(-1, 1)$, from Groups: Definition and First Examples, is the addition rule of the hyperbolic tangent: $\tanh(a + b) = \tanh a * \tanh b$. So $\tanh$ is a homomorphism from the additive group $\mathbb{R}$ to $((-1, 1), *)$. Since $\tanh$ is a bijection from $\mathbb{R}$ onto $(-1, 1)$, it is an isomorphism.

Maps that are not homomorphisms

A map can look natural and still fail; one bad product rules it out.

  • Remainders modulo 6 on $U(9)$, as in the opening: $2 \cdot 4 = 8 \mapsto 2$, but $2 + 4 \mapsto 0$.
  • The exponent of $r$ on $D_8$. Sending $r^a$ and $r^a s$ to $a$ in $\mathbb{Z}_8$ looks natural, but $r^3 s \cdot r^2 = r^{3 - 2}s = rs$, which goes to 1, while $3 + 2 = 5$. The map forgets that moving $r^2$ past $s$ inverts it.
  • Moving the identity. On $\mathbb{Z}$, $h(a) = a + 1$ fails at once: $h(0 + 0) = 1$ but $h(0) + h(0) = 2$. Every homomorphism sends the identity to the identity, so a map that moves it is ruled out.

The last item used a fact not yet proved. Which properties does the single condition $f(ab) = f(a)f(b)$ force?

First properties

Theorem. Let $f: G \to H$ be a homomorphism. Then:

  1. $f(e_G) = e_H$;
  2. $f(a^{-1}) = f(a)^{-1}$ for every $a \in G$;
  3. $f(a^k) = f(a)^k$ for every $a \in G$ and every integer $k$;
  4. the image $f(G) = \{f(a) : a \in G\}$ is a subgroup of $H$, abelian if $G$ is abelian;
  5. if $a$ has finite order $n$, then $f(a)$ has finite order, and $\lvert f(a) \rvert$ divides $n$.

Here is the idea in one sentence: the identity is the only element that equals its own square, and a homomorphism carries that equation from $G$ to $H$; everything else follows by multiplying.

Proof. (1) $f(e_G)\,f(e_G) = f(e_G e_G) = f(e_G) = f(e_G)\,e_H$, and left cancellation in $H$ gives $f(e_G) = e_H$.

(2) $f(a)\,f(a^{-1}) = f(aa^{-1}) = f(e_G) = e_H$ by (1). One product equal to the identity identifies the inverse (Basic Properties of Groups), so $f(a^{-1}) = f(a)^{-1}$.

(3) For $k \ge 0$ by induction: $f(a^0) = e_H = f(a)^0$ by (1), and $f(a^{k+1}) = f(a^k a) = f(a)^k f(a) = f(a)^{k+1}$. For $k = -p$ with $p \ge 1$, $f(a^{-p}) = f((a^p)^{-1}) = f(a^p)^{-1} = (f(a)^p)^{-1} = f(a)^{-p}$, by (2).

(4) The image contains $e_H$ by (1). For $f(a)$ and $f(b)$ in the image, $f(a)f(b)^{-1} = f(a)f(b^{-1}) = f(ab^{-1})$ is in the image, so by the one-step test the image is a subgroup. If $G$ is abelian, $f(a)f(b) = f(ab) = f(ba) = f(b)f(a)$.

(5) By (3) and (1), $f(a)^n = f(a^n) = f(e_G) = e_H$, so $f(a)$ has finite order, and by the theorem on orders that order divides $n$. $\blacksquare$

Parts (1), (2) and (4) are the course text's Theorem 7.20. Part (5) is an obstacle to the existence of homomorphisms: an element of order 9 cannot go to an element of order 2, since 2 does not divide 9. Under the reduction $\mathbb{Z}_{36} \to \mathbb{Z}_9$, the element 1 of order 36 goes to 1 of order 9, and 30 of order 6 goes to 3 of order 3. Each image order divides the original.

Check your understanding. The element 8 of $\mathbb{Z}_{36}$ has order $36/\gcd(36, 8) = 9$. What is the order of its image 8 in $\mathbb{Z}_9$?

Also 9, as the script confirms: 8 is coprime to 9, so it generates $\mathbb{Z}_9$. Part (5) allows 1, 3 or 9. A map that merges elements need not shrink every order.

Which elements does a homomorphism merge, and how can we tell?

The kernel

Definition. The kernel of a homomorphism $f: G \to H$ is

$$\ker f = \{\, a \in G : f(a) = e_H \,\}.$$

Theorem. Let $f: G \to H$ be a homomorphism with kernel $K$.

  1. $K$ is a subgroup of $G$.
  2. For $a, b \in G$, $f(a) = f(b)$ if and only if $a^{-1}b \in K$.
  3. $f$ is one-to-one if and only if $K = \{e_G\}$.

The idea: $f(a) = f(b)$ says that $f(a^{-1}b)$ is the identity, so every coincidence of images is recorded in the kernel.

Proof. (1) $e_G \in K$ by part (1) of the first theorem. If $a, b \in K$, then $f(ab^{-1}) = f(a)f(b)^{-1} = e_H e_H^{-1} = e_H$, so $ab^{-1} \in K$, and the one-step test applies.

(2) $f(a^{-1}b) = f(a)^{-1}f(b)$, which is $e_H$ exactly when $f(a) = f(b)$.

(3) If $f$ is one-to-one, the only element sent to $e_H = f(e_G)$ is $e_G$. Conversely, if $K = \{e_G\}$ and $f(a) = f(b)$, then $a^{-1}b \in K$ by (2), so $a^{-1}b = e_G$ and $a = b$. $\blacksquare$

The kernel plays the part it played for ring homomorphisms: two elements have the same image exactly when they differ by an element of the kernel, here in the multiplicative form $a^{-1}b \in K$. The kernels of the examples above:

Homomorphism Kernel
$\operatorname{sgn}: S_n \to \{1, -1\}$ the alternating group $A_n$ (60 elements for $n = 5$)
rotation or reflection, $D_8 \to \{1, -1\}$ the eight rotations $\langle r \rangle$
$\mathbb{Z} \to \mathbb{Z}_n$ $n\mathbb{Z}$
$\mathbb{Z}_{36} \to \mathbb{Z}_9$ $\{0, 9, 18, 27\} = \langle 9 \rangle$
$\exp_a: \mathbb{Z} \to G$ $n\mathbb{Z}$ if $\lvert a \rvert = n$; $\{0\}$ if $a$ has infinite order
vertices, $D_8 \to S_8$ $\{1\}$
corner positions, cube group $\to S_8$ the moves that return every corner to its own position

The kernel of $\exp_a$ follows from the theorem on orders: $a^k = e$ exactly when $\lvert a \rvert$ divides $k$. The last kernel is large. It contains the superflip, which flips all twelve edges and moves no corner. It also contains the move $(R U R' U')^2$, which the script confirms leaves all eight corners in their positions (twisting four of them) while moving three edges.

Check your understanding. Under the reduction $\mathbb{Z}_{36} \to \mathbb{Z}_9$, which elements of $\mathbb{Z}_{36}$ go to 5?

Four of them: 5, 14, 23 and 32. By part (2) of the theorem, these are the elements that differ from 5 by an element of the kernel $\{0, 9, 18, 27\}$, written additively. The script lists the same four.

Kernels of group homomorphisms have one more property, which arbitrary subgroups lack. If $k \in K$ and $g \in G$, then $f(gkg^{-1}) = f(g)e_H f(g)^{-1} = e_H$, so $gkg^{-1} \in K$. Subgroups with this property are the normal subgroups, the group counterpart of ideals. They are the subject of Normal Subgroups and Quotient Groups, and the first isomorphism theorem for groups identifies the image of $f$ with the quotient of $G$ by its kernel.

Homomorphisms out of a cyclic group

How much must be known to pin down a homomorphism? Its values on a generating set suffice, because by part (3) of the first theorem $f(a_1^{k_1} \cdots a_m^{k_m}) = f(a_1)^{k_1} \cdots f(a_m)^{k_m}$. For a cyclic group one value decides everything, and the possible values can be described exactly.

Proposition. Let $H$ be a group and $n$ a positive integer. For each $h \in H$ with $h^n = e_H$, the rule $f([k]) = h^k$ is a well-defined homomorphism $f: \mathbb{Z}_n \to H$ with $f([1]) = h$, and every homomorphism from $\mathbb{Z}_n$ to $H$ arises in this way from exactly one such $h$. Similarly, every homomorphism from $\mathbb{Z}$ to $H$ is $\exp_h$ for exactly one $h \in H$.

The idea: the image $h$ of $[1]$ decides every other value, and the only constraint is that $n$ steps around $\mathbb{Z}_n$ return to $[0]$, so $h^n$ must be $e_H$.

Proof. Well defined: if $[k] = [k']$, then $n \mid k - k'$, and since $h^n = e_H$ the order of $h$ divides $n$, hence divides $k - k'$, so $h^k = h^{k'}$ by the theorem on orders. Homomorphism: $f([j] + [k]) = f([j + k]) = h^{j + k} = h^j h^k$. Conversely, if $g: \mathbb{Z}_n \to H$ is any homomorphism, put $h = g([1])$. Then $h^n = g(n[1]) = g([0]) = e_H$ by part (3) of the first theorem written additively, and $g([k]) = g(k[1]) = h^k = f([k])$. Different $h$ give different homomorphisms, since $h = f([1])$. The statement for $\mathbb{Z}$ is the same argument without the condition $h^n = e$, since $\exp_h$ is a homomorphism for every $h$. $\blacksquare$

So the homomorphisms from $\mathbb{Z}_n$ to $H$ correspond one to one with the elements $h$ of $H$ whose order divides $n$. The script counts them by brute force, trying every possible image of $[1]$, and gets agreement in every case it tries:

  • $\mathbb{Z}_{36} \to U(9)$: 6 homomorphisms, since every element of $U(9)$ has order dividing 6, hence dividing 36.
  • $\mathbb{Z}_6 \to D_8$: 10 homomorphisms, from the identity and the nine elements of order 2.
  • $\mathbb{Z}_9 \to D_8$: only the trivial homomorphism, since no element of $D_8$ other than 1 has order dividing 9.

Check your understanding. How many homomorphisms are there from $\mathbb{Z}_{12}$ to the stop sign's group $D_8$?

Twelve, one for each element whose order divides 12. Only the four rotations of order 8 are ruled out, leaving the identity, nine elements of order 2, and $r^2$ and $r^6$. The script's brute-force count is also 12.

Now to the bijective maps. Is "isomorphic" a sensible way to say "the same"?

Composites, inverses and the isomorphism relation

The idea: a composite of maps that respect the operation respects it, and undoing a renaming is a renaming.

Proposition. Let $f: G \to H$ and $g: H \to K$ be homomorphisms.

  1. $g \circ f: G \to K$ is a homomorphism, and an isomorphism if $f$ and $g$ are.
  2. If $f$ is an isomorphism, its inverse function $f^{-1}: H \to G$ is an isomorphism.

Proof. (1) $g(f(ab)) = g(f(a)f(b)) = g(f(a))\,g(f(b))$, and a composite of bijections is a bijection. (2) $f^{-1}$ is a bijection. For $x, y \in H$ let $a = f^{-1}(x)$ and $b = f^{-1}(y)$. Then $f(ab) = f(a)f(b) = xy$, so $f^{-1}(xy) = ab = f^{-1}(x)f^{-1}(y)$. $\blacksquare$

With the identity maps, this makes "is isomorphic to" an equivalence relation on groups, as for rings (Judson, 2026, Theorem 9.1.10). Group theory studies groups up to isomorphism. Two isomorphic groups are the same group with different names for the elements, and anything stated in terms of the operation is true of one exactly when it is true of the other.

What an isomorphism preserves

Which statements "in terms of the operation" are worth checking first? Here is the idea of the next theorem: an isomorphism matches elements one to one and turns every equation $ab = c$ in $G$ into an equation in $H$, so any property defined by such equations survives.

Theorem. Let $f: G \to H$ be an isomorphism. Then:

  1. $\lvert G \rvert = \lvert H \rvert$;
  2. $G$ is abelian if and only if $H$ is;
  3. $\lvert f(a) \rvert = \lvert a \rvert$ for every $a \in G$, so for each $k$ the two groups have the same number of elements of order $k$;
  4. $G$ is cyclic if and only if $H$ is, and $f$ sends generators to generators;
  5. $f$ matches the subgroups of $G$ one to one with those of $H$, preserving their orders, and $f(Z(G)) = Z(H)$.

Proof. As for rings, each statement about $H$ can be pulled back to $G$ with the isomorphism $f^{-1}$, so it is enough to prove the forward direction. (1) $f$ is a bijection. (2) The image of an abelian group is abelian, and the image is $H$. (3) For $k \ge 1$, $f(a)^k = f(a^k)$, and this equals $e_H = f(e_G)$ exactly when $a^k = e_G$, since $f$ is one-to-one; so the least such $k$ is the same for $a$ and $f(a)$, and if there is none for one there is none for the other. (4) If $G = \langle a \rangle$, every element of $H$ is $f(a^k) = f(a)^k$, so $H = \langle f(a) \rangle$. (5) The image $f(K)$ of a subgroup $K$ is a subgroup of $H$, by part (4) of the first theorem applied to $f$ restricted to $K$, with $\lvert f(K) \rvert = \lvert K \rvert$; and $K \mapsto f(K)$ has the inverse $L \mapsto f^{-1}(L)$. Finally $z$ commutes with every $g$ exactly when $f(z)$ commutes with every $f(g)$, that is, with every element of $H$. $\blacksquare$

These properties prove that groups are not isomorphic, by finding one that differs. Dihedral Groups already used part (3): $D_8$ and $\mathbb{Z}_{16}$ both have 16 elements, but $\mathbb{Z}_{16}$ has an element of order 16 and $D_8$ does not.

Five groups of order 16

Counting the elements of each order is a quick test. For five groups of order 16 the script finds:

Group Abelian Cyclic Order 1 Order 2 Order 4 Order 8 Order 16
$\mathbb{Z}_{16}$ yes yes 1 1 2 4 8
$\mathbb{Z}_2 \times \mathbb{Z}_8$ yes no 1 3 4 8 0
$\mathbb{Z}_4 \times \mathbb{Z}_4$ yes no 1 3 12 0 0
$U(32)$ yes no 1 3 4 8 0
$D_8$ no no 1 9 2 4 0

The profiles of $\mathbb{Z}_{16}$, $\mathbb{Z}_2 \times \mathbb{Z}_8$, $\mathbb{Z}_4 \times \mathbb{Z}_4$ and $D_8$ are all different, so no two of these four groups are isomorphic. The table also shows that a single statistic is often not enough: the column of elements of order 2, for instance, separates only $\mathbb{Z}_{16}$ and $D_8$ from the rest.

The group of units $U(32)$, the 16 odd classes modulo 32, has the same profile as $\mathbb{Z}_2 \times \mathbb{Z}_8$. Are they therefore isomorphic? That is the tempting conclusion, but the preserved properties never prove that two groups are isomorphic, however many agree; for that an isomorphism must be written down. Equal profiles can belong to different groups. The script gives the pairs $(i, j)$ of $\mathbb{Z}_4 \times \mathbb{Z}_4$ the twisted product $(i, j)(k, l) = (i + (-1)^j k, \, j + l)$. The result is a nonabelian group with exactly the profile of $\mathbb{Z}_4 \times \mathbb{Z}_4$.

What the table can do is say which map to look for: an element of order 8, such as 3, and an element of order 2 outside $\langle 3 \rangle$, such as $-1 = 31$. The idea of the proof is that the 16 products $\pm 3^j$ are all different, as remainders modulo 8 show.

Proposition. The map $\psi: \mathbb{Z}_2 \times \mathbb{Z}_8 \to U(32)$, $\psi(i, j) = (-1)^i 3^j \bmod 32$, is an isomorphism.

Proof. Well defined. $(-1)^i$ depends only on $i$ modulo 2, and $3^j$ modulo 32 depends only on $j$ modulo 8, because 3 has order 8 in $U(32)$: its powers are $3^0, \ldots, 3^7 = 1, 3, 9, 27, 17, 19, 25, 11$, and $3^8 = 1$.

Homomorphism. In the additive group $\mathbb{Z}_2 \times \mathbb{Z}_8$, $(i, j) + (i', j') = (i + i', j + j')$, and

$$\psi(i + i', j + j') = (-1)^{i + i'} 3^{j + j'} = (-1)^i 3^j \cdot (-1)^{i'} 3^{j'} = \psi(i, j)\,\psi(i', j'),$$

using commutativity of multiplication modulo 32.

One-to-one. The eight powers $3^j$ are different, since 3 has order 8. They are all congruent to 1 or 3 modulo 8 ($3^2 = 9 \equiv 1$), so the eight numbers $-3^j$ are congruent to 7 or 5 modulo 8, and no $3^j$ equals any $-3^{j'}$ modulo 32. So the 16 values of $\psi$ are different, and $\psi$ is one-to-one. Since both groups have 16 elements, it is also onto. $\blacksquare$

The isomorphism from Z2 x Z8 to U(32), drawn as a grid: the cell in row i and column j holds (-1)^i 3^j modulo 32. Notice that each of the 16 odd classes modulo 32 appears exactly once, which is why the map is a bijection. Moving right multiplies by 3 and moving between rows multiplies by -1 = 31. The top row is the cyclic subgroup generated by 3, whose elements are 1 or 3 modulo 8; the bottom row holds the classes that are 5 or 7 modulo 8.
The isomorphism from Z_2 x Z_8 to U(32), drawn as a grid: the cell in row i and column j holds (-1)^i 3^j modulo 32. Notice that each of the 16 odd classes modulo 32 appears exactly once, which is why the map is a bijection. Moving right multiplies by 3 and moving between rows multiplies by -1 = 31. The top row is the cyclic subgroup generated by 3, whose elements are 1 or 3 modulo 8; the bottom row holds the classes that are 5 or 7 modulo 8.

Cyclic groups are $\mathbb{Z}$ and $\mathbb{Z}_n$

For cyclic groups no search is needed: the homomorphism $\exp_a$ classifies them all. Here is the idea: $\exp_a$ is onto $\langle a \rangle$, and the order of $a$ decides its kernel. If the order is infinite the kernel is trivial; if it is $n$, the same rule on $\mathbb{Z}_n$ has a trivial kernel.

Theorem. Let $G = \langle a \rangle$ be a cyclic group.

  1. If $a$ has infinite order, then $\exp_a: \mathbb{Z} \to G$, $k \mapsto a^k$, is an isomorphism, so $G \cong \mathbb{Z}$.
  2. If $a$ has order $n$, then $[k] \mapsto a^k$ is an isomorphism from $\mathbb{Z}_n$ onto $G$, so $G \cong \mathbb{Z}_n$.

In particular two cyclic groups are isomorphic exactly when they have the same order.

Proof. (1) $\exp_a$ is a homomorphism onto $\langle a \rangle = G$, and its kernel is $\{0\}$ because $a$ has infinite order, so it is one-to-one by the kernel theorem.

(2) By the proposition on homomorphisms out of $\mathbb{Z}_n$, with $h = a$ (which satisfies $a^n = e$), the rule $f([k]) = a^k$ is a well-defined homomorphism. It is onto, since every element of $G$ is a power of $a$. Its kernel is the set of $[k]$ with $a^k = e$, that is, with $n \mid k$, which is $\{[0]\}$; so it is one-to-one. The last statement follows because isomorphism is an equivalence relation and $\mathbb{Z}$, $\mathbb{Z}_n$ and $\mathbb{Z}_m$ for $m \neq n$ have different orders. $\blacksquare$

This is the course text's Theorem 7.19 and Judson's Theorems 9.1.7 and 9.1.8. It confirms several identifications made informally earlier in the unit:

  • $U(9) \cong \mathbb{Z}_6$ by $[k] \mapsto 2^k$, the renaming in the opening. Its inverse sends each unit to its exponent, a discrete logarithm to the base 2: $\log_2 7 = 4$ in $U(9)$.
  • The 8th roots of unity $\mu_8 = \langle \omega \rangle$, $\omega = e^{2\pi i/8}$, form a group isomorphic to $\mathbb{Z}_8$ by $[k] \mapsto \omega^k$, as Groups: Definition and First Examples observed.
  • The rotations of $D_8$ form a copy of $\mathbb{Z}_8$, by $[k] \mapsto r^k$.
  • $\mathbb{Z}_5 \times U(9) \cong \mathbb{Z}_{30}$, since $(3, 2)$ has order 30 (Basic Properties of Groups); the isomorphism is $[k] \mapsto (3, 2)^k$, and for example $(3,2)^7 = (21 \bmod 5, 2^7 \bmod 9) = (1, 2)$.
  • $\mathbb{Z}_4 \times \mathbb{Z}_9 \cong \mathbb{Z}_{36}$ by $[k] \mapsto (k \bmod 4, k \bmod 9)$, the additive half of the Chinese remainder theorem.
  • The cyclic subgroup $\langle RU \rangle$ of the cube group is a copy of $\mathbb{Z}_{105}$, and the subgroup $\langle 1 + i \rangle$ of $\mathbb{C}^\times$ is a copy of $\mathbb{Z}$.

So there is, up to isomorphism, exactly one cyclic group of each order. All the facts about cyclic groups in Subgroups and Cyclic Groups, such as one subgroup for each divisor and $\varphi(n)$ generators, can be read off from $\mathbb{Z}_n$. Isomorphisms carry them to every other cyclic group of order $n$.

Check your understanding. Subgroups and Cyclic Groups found that 5 has order 22 in $U(23)$, so $U(23)$ is cyclic. Without computing any powers modulo 23, how many elements of $U(23)$ have order 11?

Ten. By part (2) of the theorem, $[k] \mapsto 5^k$ is an isomorphism from $\mathbb{Z}_{22}$ onto $U(23)$, and isomorphisms preserve orders, so the count is the same as in $\mathbb{Z}_{22}$. There $k$ has order $22/\gcd(22, k)$, which is 11 exactly when $\gcd(22, k) = 2$, that is, for the ten even $k$ from 2 to 20. The elements of order 11 in $U(23)$ are the ten powers $5^k$ with those exponents. The script checks the isomorphism on all $22^2$ sums and counts ten elements of order 11.

Conjugation

Every group also has isomorphisms onto itself, one for each element. For a fixed element $c$ of a group $G$, conjugation by $c$ is the map

$$\gamma_c: G \to G, \qquad \gamma_c(x) = c x c^{-1}.$$

Proposition. Each $\gamma_c$ is an automorphism of $G$, with inverse $\gamma_{c^{-1}}$. Consequently $cxc^{-1}$ and $x$ have the same order, for all $c$ and $x$.

The idea: in a product of two conjugates the inner $c^{-1}c$ cancels, and conjugating by $c^{-1}$ undoes conjugating by $c$.

Proof. Homomorphism: $\gamma_c(x)\gamma_c(y) = cxc^{-1}cyc^{-1} = cx(c^{-1}c)yc^{-1} = cxyc^{-1} = \gamma_c(xy)$. Bijection: for every $x$,

$$\gamma_{c^{-1}}(\gamma_c(x)) = c^{-1}(cxc^{-1})c = x \qquad \text{and} \qquad \gamma_c(\gamma_{c^{-1}}(x)) = c(c^{-1}xc)c^{-1} = x,$$

so $\gamma_{c^{-1}}$ is an inverse function for $\gamma_c$, and a function with an inverse function is a bijection. The statement about orders is part (3) of the preservation theorem. $\blacksquare$

The automorphisms $\gamma_c$ are called inner automorphisms. In an abelian group every one of them is the identity map, since $cxc^{-1} = xcc^{-1} = x$. Conjugation has already appeared three times in this unit:

  • In $D_8$. Conjugation by $s$ sends each rotation $r^a$ to $r^{-a}$ and each reflection $r^a s$ to $r^{-a}s$, by the rule $sr^a = r^{-a}s$. Flipping, rotating and flipping back rotates the other way. Conjugation by $r$ fixes every rotation and sends $r^a s$ to $r^{a+2}s$, turning each mirror line through 45 degrees. The script confirms that both are automorphisms.
  • In $S_n$. Conjugating a permutation by $\rho$ renames each object $x$ as $\rho(x)$ (Symmetric Groups and Permutations). That explains why conjugate permutations have the same cycle type.
  • On the cube. A conjugate $XMX^{-1}$ does the work of $M$ on other pieces, and has the same order as $M$. The cube article found that $RF = F'(FR)F$ is a conjugate of $FR$ (in the cube's left-to-right reading). Both have order 105, as the proposition requires.

Check your understanding. In the stop sign's group, what is $s r^3 s^{-1}$, and what is its order?

It is $r^{-3} = r^5$, the same turn the other way, and like $r^3$ it has order 8. The script also sends $r^3$ to $r^5$.

Where it leads

What we have learned

A homomorphism is a map with $f(ab) = f(a)f(b)$. That one condition forces it to respect identities, inverses and powers, and to shrink orders only to divisors. Its kernel records which elements it merges, and a trivial kernel means one-to-one. An isomorphism preserves every property stated through the operation. So order counts can prove that groups differ, but only an explicit map, such as $(i, j) \mapsto (-1)^i 3^j$, proves that $U(32) \cong \mathbb{Z}_2 \times \mathbb{Z}_8$. Every cyclic group is $\mathbb{Z}$ or some $\mathbb{Z}_n$.

The opening showed $U(9)$ renamed as $\mathbb{Z}_6$. Can every group, however abstract, be renamed as a group of permutations? The next article, Cayley's Theorem, proves that it can.

References

  • Hungerford, T. W. (2014). Abstract Algebra: An Introduction (3rd ed.), §7.4. Brooks/Cole, Cengage Learning. The course text: isomorphisms and homomorphisms of groups; infinite cyclic groups are isomorphic to $\mathbb{Z}$ and cyclic groups of order $n$ to $\mathbb{Z}_n$ (Theorem 7.19); a homomorphism sends the identity to the identity and inverses to inverses, its image is a subgroup, and a one-to-one homomorphism is an isomorphism onto its image (Theorem 7.20). record · 44 pages
  • Judson, T. W. (2026). Abstract Algebra: Theory and Applications (annual edition, August 2026). Free online at judsonbooks.org. §9.1 (isomorphisms; Theorem 9.1.6, properties preserved by an isomorphism; Theorems 9.1.7 and 9.1.8, every cyclic group is isomorphic to $\mathbb{Z}$ or $\mathbb{Z}_n$; Theorem 9.1.10, isomorphism is an equivalence relation). record · 44 pages