Multiplying every element of $U(9)$ by 2 rearranges them:
$$1, 2, 4, 5, 7, 8 \quad \longmapsto \quad 2, 4, 8, 1, 5, 7.$$
This list is the row of 2 in the Cayley table of $U(9)$. Every row of a Cayley table contains each element exactly once (Basic Properties of Groups), so the row is a permutation of the six elements, and so is the row of every other element. Multiplying first by 4 and then by 2 is the same as multiplying by $2 \cdot 4 = 8$, so the six permutations multiply exactly as the six units do. Up to the names of its elements, $U(9)$ is a group of permutations of a six-element set. The dihedral groups, the Rubik's cube group and the symmetric groups were also met as permutations, of vertices, stickers and labels. Is every group a group of permutations in disguise?
It is, and nothing in the argument above depends on $U(9)$. Every group acts on itself by multiplication, and the permutations it produces form a copy of the group. That is Cayley's theorem. It means that the permutation groups, where group theory began, include every group there is.
First we prove the theorem and see what the copy of $U(9)$ looks like. Then we find that each permutation's cycles all have one length, which describes the copy of the stop sign's group $D_8$ without writing it out. Finally we look at infinite groups, smaller representations, and what the theorem does not tell us.
This is the eighth and last article of Groups. It covers the end of §7.4 of the course text (Hungerford, 2014, Theorem 7.21 and Corollary 7.22), with Judson (2026, §9.1.1) as a free companion. It uses Symmetric Groups and Permutations for cycle notation and Group Homomorphisms and Isomorphisms for one-to-one homomorphisms.
The theorem
For a set $X$, $\mathrm{Sym}(X)$ denotes the group of all permutations (bijections) of $X$ under composition (Groups: Definition and First Examples); products of permutations are composites with the right factor applied first, as in Symmetric Groups and Permutations.
Definition. For an element $g$ of a group $G$, left multiplication by $g$ is the function
$$\lambda_g: G \to G, \qquad \lambda_g(x) = gx.$$
Lemma. Each $\lambda_g$ is a permutation of $G$, and for all $g, h \in G$,
$$\lambda_g \circ \lambda_h = \lambda_{gh}, \qquad \lambda_e = \mathrm{id}_G.$$
The idea: multiplying by $h$ and then by $g$ is multiplying by $gh$, by associativity, and multiplying by $g^{-1}$ undoes multiplying by $g$.
Proof. For every $x$, $(\lambda_g \circ \lambda_h)(x) = g(hx) = (gh)x = \lambda_{gh}(x)$ by associativity, and $\lambda_e(x) = ex = x$. Then $\lambda_g \circ \lambda_{g^{-1}} = \lambda_{gg^{-1}} = \lambda_e$ and likewise $\lambda_{g^{-1}} \circ \lambda_g = \lambda_e$, so $\lambda_{g^{-1}}$ is an inverse function for $\lambda_g$, and $\lambda_g$ is a bijection. $\blacksquare$
The order of composition is what makes left multiplication the right choice. Multiplying on the right, $\rho_g(x) = xg$, also gives permutations, but $(\rho_g \circ \rho_h)(x) = (xh)g = x(hg)$. So $\rho_g \circ \rho_h = \rho_{hg}$: the order of the factors is reversed, and $g \mapsto \rho_g$ respects products only when the group is abelian.
Here is the plan of the proof: the lemma already makes $g \mapsto \lambda_g$ a homomorphism, and its kernel shows that different elements give different permutations.
Theorem (Cayley). Every group $G$ is isomorphic to a group of permutations. Precisely, the map
$$\lambda: G \to \mathrm{Sym}(G), \qquad g \mapsto \lambda_g,$$
is a one-to-one homomorphism, and $G$ is isomorphic to its image $\{\lambda_g : g \in G\}$, a subgroup of $\mathrm{Sym}(G)$.
Proof. By the lemma, $\lambda$ takes values in $\mathrm{Sym}(G)$ and $\lambda(gh) = \lambda_{gh} = \lambda_g \circ \lambda_h = \lambda(g)\lambda(h)$, so $\lambda$ is a homomorphism. Its kernel is $\{e\}$: if $\lambda_g$ is the identity permutation, then in particular $ge = \lambda_g(e) = e$, so $g = e$. By the kernel theorem $\lambda$ is one-to-one. The image of a homomorphism is a subgroup (first properties), and $\lambda$, viewed as a map from $G$ onto its image, is a bijective homomorphism, hence an isomorphism. $\blacksquare$
The homomorphism $\lambda$ is the left regular representation of $G$ (Judson, 2026, §9.1.1). The proof is short because the work was done earlier: the lemma is the Latin square property of Cayley tables in another form, and the rest is the general fact, Theorem 7.20 of the course text, that a one-to-one homomorphism is an isomorphism onto its image.
For a finite group the permutations can be written in the notation of $S_n$.
Corollary. A finite group of order $n$ is isomorphic to a subgroup of $S_n$.
The idea: number the elements $1, \ldots, n$, and a permutation of the group becomes a permutation of the numbers.
Proof. Number the elements of $G$ as $g_1, \ldots, g_n$, and let $\beta: G \to \{1, \ldots, n\}$ be the numbering, $\beta(g_i) = i$. For a permutation $\pi$ of $G$, the function $\beta \circ \pi \circ \beta^{-1}$ is a permutation of $\{1, \ldots, n\}$: it sends $i$ to the number of $\pi(g_i)$. The map $\pi \mapsto \beta\pi\beta^{-1}$ from $\mathrm{Sym}(G)$ to $S_n$ is a homomorphism, since $\beta\pi\beta^{-1} \cdot \beta\pi'\beta^{-1} = \beta\pi\pi'\beta^{-1}$, and it is a bijection, with inverse $\tau \mapsto \beta^{-1}\tau\beta$. Composing it with $\lambda$ gives a one-to-one homomorphism $G \to S_n$, and $G$ is isomorphic to its image. $\blacksquare$
The renumbering changes nothing essential. It is the same idea as conjugation renaming the objects of a permutation (Symmetric Groups and Permutations), carried out between two different sets.
$U(9)$ as permutations of six objects
Back to the opening. What does the theorem's copy of $U(9)$ look like? Following the row of 2 from the opening, multiplication by 2 sends $1 \to 2 \to 4 \to 8 \to 7 \to 5 \to 1$, so $\lambda_2$ is a single cycle through all six elements. The other five permutations need no further work. $U(9) = \langle 2 \rangle$ is cyclic, so every element is a power $2^k$, and the lemma gives $\lambda_{2^k} = \lambda_2^k$. The copy of $U(9)$ is therefore the cyclic group generated by one 6-cycle. Once the six elements are numbered, as in the corollary, it is the cyclic subgroup of $S_6$ generated by a 6-cycle, and the script confirms that $\lambda$ respects all 36 products.
The same argument applies to every finite cyclic group. If $G = \langle g \rangle$ has order $n$, its copy is generated by the single permutation $\lambda_g$, and the proposition in the next section shows that $\lambda_g$ is one cycle through all $n$ elements.
The cycles of a left multiplication
In $U(9)$ the generator gave one long cycle. What happens for an element that does not generate? In $\mathbb{Z}_{12}$, adding 4 sends $0 \to 4 \to 8 \to 0$, $1 \to 5 \to 9 \to 1$, $2 \to 6 \to 10 \to 2$ and $3 \to 7 \to 11 \to 3$. That is four cycles, each of length 3, and 3 is the order of 4. Is every cycle of $\lambda_g$ always as long as the order of $g$? Here is the idea of the proof: $x$ returns to itself after $k$ steps exactly when $g^k x = x$, and cancellation turns that into $g^k = e$.
Proposition. Let $G$ be a finite group and $g \in G$ of order $m$. Then $\lambda_g$ is a product of $\lvert G \rvert / m$ disjoint cycles, each of length $m$. In particular, for $g \neq e$, $\lambda_g$ moves every element of $G$.
Proof. By the theorem in Symmetric Groups and Permutations, the cycles of a permutation $\pi$ are its orbits: the cycle containing $x$ is $x, \pi(x), \pi^2(x), \ldots$, and its length is the least $L \ge 1$ with $\pi^L(x) = x$. For $\pi = \lambda_g$, the lemma gives $\lambda_g^k = \lambda_{g^k}$, so $\lambda_g^k(x) = g^k x$. Now $g^k x = x$ holds exactly when $g^k = e$, by right cancellation, and the least $k \ge 1$ with $g^k = e$ is $m$. So the cycle through every $x$ has length exactly $m$. The cycles partition the $\lvert G \rvert$ elements, so there are $\lvert G \rvert / m$ of them. If $g \neq e$, then $m \ge 2$, so no element lies in a cycle of length 1. $\blacksquare$
A permutation group is called regular if only the identity fixes a point and, as here, any element can be carried to any other (the element $yx^{-1}$ carries $x$ to $y$). That is the source of the name "regular representation". Two consequences of the proposition deserve notice.
The order of an element divides the order of the group. The $\lvert G \rvert$ elements fall into cycles of length $\lvert g \rvert$, so $\lvert g \rvert$ divides $\lvert G \rvert$. In $D_8$, the elements have orders 1, 2, 4 and 8, all dividing 16; in $U(9)$, orders 1, 2, 3 and 6, all dividing 6. The cycles of $\lambda_g$ are the sets $\{x, gx, g^2x, \ldots\}$, which are the cosets $\langle g \rangle x$ of the cyclic subgroup $\langle g \rangle$. The cosets of any subgroup split a group into pieces of equal size, so the order of every subgroup divides the order of the group. That is Lagrange's theorem, the subject of the next unit.
The sign of $\lambda_g$ can be read off. A cycle of length $m$ has sign $(-1)^{m-1}$ (Symmetric Groups and Permutations), so
$$\operatorname{sgn}(\lambda_g) = (-1)^{(m - 1)\,\lvert G \rvert / m}, \qquad m = \lvert g \rvert.$$
In $U(9)$ the elements 2 and 5 have order 6 and 8 has order 2, so for them the exponent is $5 \cdot 1$ or $1 \cdot 3$, both odd, and the sign is $-1$. The elements 4 and 7 have order 3, with exponent $2 \cdot 2$, and the identity has exponent 0, so their sign is $+1$. So the image of $U(9)$ in $S_6$ contains odd permutations, and it is not inside the alternating group $A_6$. In $D_8$, $\lvert G \rvert / m$ is even for every element (16 divided by 1, 2, 4 or 8). So every $\lambda_g$ is even, and the image of $D_8$ lies in $A_{16}$. The script checks the formula for every element of both groups.
Check your understanding. In $\mathbb{Z}_{36}$, the element 9 has order 4. Without writing it out, describe $\lambda_9$, the permutation $x \mapsto 9 + x$ of the 36 elements. Is it even or odd?
It is nine disjoint 4-cycles, such as $0 \to 9 \to 18 \to 27 \to 0$. Each 4-cycle has sign $-1$, and nine of them give $(-1)^9 = -1$, so $\lambda_9$ is odd. The script finds nine cycles of length 4 and sign $-1$.
The octagon's group in $S_{16}$
The same construction works for a nonabelian group. The stop sign's group $D_8$ (Dihedral Groups) has 16 elements, so once they are numbered the theorem places a copy of $D_8$ inside $S_{16}$. What do its sixteen permutations look like? Tracing each one through sixteen objects is not needed: the proposition gives the cycle type of $\lambda_g$ from the order of $g$ alone. By the orders in $D_n$, the rotation $r^k$ has order $8/\gcd(8, k)$ and every reflection has order 2.
Check your understanding. Before reading the table below, predict how many of the sixteen permutations $\lambda_g$ are products of eight disjoint 2-cycles.
Nine. By the proposition these are the $\lambda_g$ with $\lvert g \rvert = 2$. The eight reflections have order 2, and among the rotations only the half turn $r^4$ does, since $8/\gcd(8, k) = 2$ forces $\gcd(8, k) = 4$, that is, $k = 4$.
Sorting all the elements by order gives the shapes of all sixteen permutations:
| Elements $g$ of $D_8$ | $\lvert g \rvert$ | $\lambda_g$ | How many |
|---|---|---|---|
| $1$ | 1 | the identity | 1 |
| $r^4$ and the eight reflections | 2 | eight disjoint 2-cycles | 9 |
| $r^2$, $r^6$ | 4 | four disjoint 4-cycles | 2 |
| $r$, $r^3$, $r^5$, $r^7$ | 8 | two disjoint 8-cycles | 4 |
The script computes all sixteen permutations and confirms each shape. Cycle type does not determine the element, or even how it behaves in the group: $\lambda_{r^4}$ and $\lambda_s$ have the same shape, yet the half turn $r^4$ commutes with every element of $D_8$ and $s$ does not. Since $D_8 = \langle r, s \rangle$, the copy is the subgroup of $S_{16}$ generated by $\lambda_r$ and $\lambda_s$, and the script checks that $\lambda$ respects all 256 products.
The representation is wasteful. $S_{16}$ has $16! = 20{,}922{,}789{,}888{,}000$ elements, of which the image uses 16. The symmetries of the octagon already permute its eight vertices, which embeds $D_8$ in $S_8$, a group of $8! = 40{,}320$ elements (Dihedral Groups). Cayley's theorem uses the one set that every group is guaranteed to act on, the group itself, and pays for that generality in size.
Infinite groups
Does the theorem need finiteness? It makes no such assumption. For the additive group $\mathbb{Z}$, $\lambda_k$ is the translation $x \mapsto k + x$. Cayley's theorem identifies $\mathbb{Z}$ with the group of translations of the integers, a subgroup of the uncountable group $\mathrm{Sym}(\mathbb{Z})$. For the multiplicative group $\mathbb{R}^\times$, $\lambda_c$ is the scaling $x \mapsto cx$ of the nonzero real numbers. In each case the abstract group is matched with a group of transformations of a set, the setting in which the axioms were first abstracted.
Smaller permutation representations
Cayley's theorem places a group of order $n$ inside $S_n$, but often a much smaller symmetric group will do. What is needed is a set on which the group acts faithfully: each element permutes the set, products act as composites, and different elements give different permutations. That is, a one-to-one homomorphism into the permutations of the set.
- $D_8$ acts faithfully on the 8 vertices of the octagon, which gives a copy inside $S_8$ instead of $S_{16}$.
- The Rubik's cube group, of order 43,252,003,274,489,856,000, acts faithfully on the 48 movable stickers, which gives a copy inside $S_{48}$ (The Rubik's Cube Group). The regular representation would need the symmetric group on $4.3 \times 10^{19}$ objects.
- $U(9)$, a cyclic group of order 6, needs only five objects. The permutation $(1\;2)(3\;4\;5)$ in $S_5$ has order $\operatorname{lcm}(2, 3) = 6$ (Symmetric Groups and Permutations), so its cyclic subgroup has six elements. That subgroup is isomorphic to $\mathbb{Z}_6 \cong U(9)$ by the classification of cyclic groups, and the script confirms that $2^k \mapsto ((1\;2)(3\;4\;5))^k$ respects all 36 products. Four objects are too few. The orders of the elements of $S_4$ are 1, 2, 3 and 4, read from its cycle types $(1,1,1,1)$, $(2,1,1)$, $(2,2)$, $(3,1)$ and $(4)$. So $S_4$ has no element of order 6, and by the preservation of orders no subgroup isomorphic to $U(9)$.
In each case the smaller representation comes from a set on which the group acts naturally, such as vertices, stickers, or (for $U(9)$) two separate cycles running at once. The regular representation comes from the group acting on itself.
Check your understanding. The stop sign has four long diagonals, each joining two opposite corners, and every symmetry permutes them. Does this give a copy of $D_8$ inside $S_4$?
No. The half turn $r^4$ swaps the ends of every diagonal, so it leaves all four in place, as the identity does. The script finds the kernel $\{1, r^4\}$ and an image of 8 elements. No action on four objects could be faithful: Lagrange's theorem will show that a subgroup's order divides the group's, and 16 does not divide $4! = 24$. Studying a group through the ways it acts on sets, and through the matrices it maps into, is the beginning of representation theory (Judson, 2026, §9.1.1).
Why the theorem matters
Cayley's theorem settles a question about the definition. The axioms of a group were abstracted from groups of permutations and symmetries (Groups: Definition and First Examples). One might ask whether the abstraction admitted anything new, some group that is not a group of transformations in disguise. The answer is no: every group, given only by its Cayley table, is a group of permutations of its own elements. In that sense the abstract definition and the concrete examples describe the same class of objects, and $S_n$ contains every group of order $n$.
A natural misreading is that the theorem reduces group theory to the study of $S_n$. It does not make the symmetric groups an easy way to study groups. $S_n$ is enormous, and knowing that a group of order 16 sits somewhere among the $16!$ permutations of 16 objects says little about which subgroup it is. Its value is conceptual. It guarantees that facts proved for all permutation groups hold for all groups, and it gives every group a concrete model in which its elements can be computed with cycle notation. Arthur Cayley, born in 1821, gave his name to both the theorem and the tables. He wrote more than 900 papers, many of them while practicing law, before becoming a professor at Cambridge in 1863 (Judson, 2026, §9.1.2).
Where it leads
- The cycles of $\lambda_g$ are cosets of $\langle g \rangle$. Cosets of an arbitrary subgroup, and the theorem that the order of a subgroup divides the order of the group, are in Cosets and Lagrange's Theorem.
- The next unit, Cosets, Normal Subgroups and Quotient Groups, continues with Normal Subgroups and Quotient Groups and the The First Isomorphism Theorem for Groups.
- The homomorphisms used in the proof, and the preservation of orders that rules out $U(9)$ inside $S_4$, are in Group Homomorphisms and Isomorphisms.
What we have learned
Left multiplication by $g$ permutes the elements of any group, and $g \mapsto \lambda_g$ is a one-to-one homomorphism. So every group is a group of permutations, and a group of order $n$ lies in $S_n$. Each $\lambda_g$ splits the group into cycles of length $\lvert g \rvert$, which fixes its sign and shows that $\lvert g \rvert$ divides $\lvert G \rvert$. A natural action, such as $D_8$ on the vertices or $U(9)$ in $S_5$, usually needs far fewer objects.
The cycles of $\lambda_g$ were cosets of the subgroup $\langle g \rangle$, and they all had the same size. Do the cosets of every subgroup cut a group into equal pieces? The first article of the next unit, Cosets and Lagrange's Theorem, proves that they do.
References
- Hungerford, T. W. (2014). Abstract Algebra: An Introduction (3rd ed.), §7.4. Brooks/Cole, Cengage Learning. The course text: a one-to-one homomorphism is an isomorphism onto its image (Theorem 7.20); Cayley's theorem, every group is isomorphic to a group of permutations (Theorem 7.21); a finite group of order $n$ is isomorphic to a subgroup of $S_n$ (Corollary 7.22). record · 44 pages
- Judson, T. W. (2026). Abstract Algebra: Theory and Applications (annual edition, August 2026). Free online at judsonbooks.org. §9.1.1 (Cayley's theorem as a representation theorem; Theorem 9.1.12, Cayley's theorem, with the left regular representation $g \mapsto \lambda_g$); §9.1.2 (historical note on Arthur Cayley). record · 44 pages