To undo a sequence of moves on a Rubik's cube, the last move is undone first: the inverse of $RU$ is $U'R'$, not $R'U'$. To divide by 2 in the units modulo 9, multiply by 5. To know where the move $RU$ leaves the cube after 100 repetitions, it is enough to know that 105 repetitions restore it. The cube, the units modulo 9 and the permutations of Symmetric Groups and Permutations obey these rules for different-looking reasons. Do they hold in every group, using nothing but the four axioms?
They do, and each follows from the axioms of the definition in a few lines. These are the working rules of the subject, used without comment in every later article. Some govern how elements combine, and the rest govern the powers of a single element and its order.
First we prove uniqueness of inverses, cancellation and the unique solution of $ax = b$, which makes every Cayley table a Latin square. Then we reverse products, drop parentheses and prove the laws of exponents. Finally we turn to orders: when a power is the identity, why every element of a finite group has finite order, the order of a power, and orders in a direct product.
This is the fifth article of Groups. It covers §7.2 of the course text (Hungerford, 2014), with Judson (2026, §3.2 and §4.1) as a free companion. It assumes the definitions of Groups: Definition and First Examples, which is the home of the order of an element. The examples use the group of units $U(9)$ from that article, the dihedral group $D_8$ of the octagon from Dihedral Groups, the additive group $\mathbb{Z}_{36}$, and moves of the Rubik's cube. Throughout, $G$ is a group written multiplicatively, with identity $e$.
Identities and inverses
The definition showed that a group has only one identity: if $e$ and $e'$ both are identities, then $e = ee' = e'$. Can an element have two different inverses? The idea of the answer: sandwich one candidate against the other, and associativity forces them to agree.
Proposition. Each element $a$ of $G$ has exactly one inverse.
Proof. Suppose $b$ and $c$ are both inverses of $a$, so that $ab = e = ba$ and $ac = e = ca$. Then
$$b = be = b(ac) = (ba)c = ec = c. \qquad \blacksquare$$
The proof uses only $ba = e$ and $ac = e$: a left inverse and a right inverse of the same element must agree. This justifies the notation $a^{-1}$ for the inverse. Two consequences follow at once.
Corollary. For all $a$ in $G$, $(a^{-1})^{-1} = a$. If $ab = e$, then $b = a^{-1}$ and $a = b^{-1}$.
Proof. The equations $a^{-1}a = e = aa^{-1}$ say that $a$ is an inverse of $a^{-1}$, and by uniqueness it is the inverse. For the second statement, multiply $ab = e$ on the left by $a^{-1}$: $b = (a^{-1}a)b = a^{-1}(ab) = a^{-1}$. So $b = a^{-1}$, and then $b^{-1} = (a^{-1})^{-1} = a$. $\blacksquare$
The second statement is a labor saver: to show that $b$ is the inverse of $a$, it is enough to check one of the two products $ab$ and $ba$. For instance, $2 \cdot 5 = 10 = 1$ in $U(9)$ already shows that $5 = 2^{-1}$. The script lists all the inverses of $U(9)$:
- $1$ and $8$ are their own inverses;
- the pairs $\{2, 5\}$ and $\{4, 7\}$ are inverse to each other.
Uniqueness of inverses is part of the course text's Theorem 7.5 and is Judson's Proposition 3.2.11. The rule $(a^{-1})^{-1} = a$ is the course text's Corollary 7.6 and Judson's Proposition 3.2.13. In a ring, where the multiplication need not have inverses, the same argument showed that a unit has only one inverse.
Cancellation and equations
With inverses unique, can we divide? In the real numbers, $2x = 2y$ implies $x = y$ because 2 can be divided out. In a group every element can be divided out, on the side where it stands. The idea: multiply by $a^{-1}$ on that side, and associativity lets it meet $a$ and cancel.
Theorem (cancellation laws). For all $a$, $b$, $c$ in $G$:
- if $ab = ac$, then $b = c$;
- if $ba = ca$, then $b = c$.
Proof. (1) Multiply both sides of $ab = ac$ on the left by $a^{-1}$ and use associativity: $b = eb = (a^{-1}a)b = a^{-1}(ab) = a^{-1}(ac) = (a^{-1}a)c = c$. (2) Multiply $ba = ca$ on the right by $a^{-1}$ in the same way. $\blacksquare$
The canceled element must stand on the same side of both products. In a nonabelian group $ab = ca$ does not imply $b = c$. The rule $sr = r^{-1}s$ of Dihedral Groups is an example. In $D_8$, $sr = r^7 s$, so the products $s \cdot r$ and $r^7 \cdot s$ are equal, but $r \neq r^7$. Canceling the $s$ that stands on the left of one product and on the right of the other would be an error.
The same multiplication by $a^{-1}$ solves equations.
Theorem. For all $a$ and $b$ in $G$, the equation $ax = b$ has exactly one solution $x$ in $G$, namely $x = a^{-1}b$, and the equation $ya = b$ has exactly one solution, $y = ba^{-1}$.
Proof. The element $x = a^{-1}b$ is a solution, since $a(a^{-1}b) = (aa^{-1})b = eb = b$. If $x'$ is another, then $ax' = b = ax$, and left cancellation gives $x' = x$. The equation $ya = b$ is handled in the same way, with $y = ba^{-1}$ and right cancellation. $\blacksquare$
How are the two solutions related? Since $b = ax$,
$$y = ba^{-1} = (ax)a^{-1} = a x a^{-1}.$$
So $y$ is obtained from $x$ by multiplying by $a$ on the left and by $a^{-1}$ on the right, and $y = x$ exactly when $axa^{-1} = x$, that is, when $a$ commutes with $x$. The script checks the relation for every pair $a$, $b$ in $D_8$ and finds 144 of the 256 pairs for which the two solutions differ. Elements of the form $axa^{-1}$ are called conjugates of $x$, and they reappear in Symmetric Groups and Permutations and Group Homomorphisms and Isomorphisms.
Check your understanding. In $U(9)$, solve $4x = 2$ and $y \cdot 4 = 2$. Do the two solutions differ?
Both are $4^{-1} \cdot 2 = 7 \cdot 2 = 14 = 5$, and indeed $4 \cdot 5 = 20 = 2$. They agree because $U(9)$ is abelian, so $y = 4 x 4^{-1} = x$. The two-sided distinction is a feature of nonabelian groups.
Cayley tables are Latin squares
The theorem has a picture. In the Cayley table of a finite group, the row of $a$ lists the products $ax$ as $x$ runs through $G$. Each element $b$ appears in that row for exactly one $x$, because $ax = b$ has exactly one solution. So every row lists each element of $G$ exactly once, and by the equation $ya = b$ so does every column: the table is a Latin square. The script confirms this for the tables of $U(9)$, $D_8$ and $\mathbb{Z}_{36}$. A row of the table is therefore a rearrangement of the elements of $G$, a permutation, and Cayley's theorem is built on this observation.
The converse fails: a Latin square need not be the table of a group, since nothing in the Latin property forces associativity.
The inverse of a product
Back to the cube of the opening, and to undoing a product. To undo putting on socks and then shoes, take off the shoes first. The same holds in every group. The idea of the proof: by the corollary above, it is enough to check that the reversed product cancels $ab$ on one side.
Theorem. For all $a$ and $b$ in $G$, $(ab)^{-1} = b^{-1}a^{-1}$. More generally, for elements $a_1, a_2, \ldots, a_k$ of $G$,
$$(a_1 a_2 \cdots a_k)^{-1} = a_k^{-1} \cdots a_2^{-1} a_1^{-1}.$$
Proof. By associativity, $(ab)(b^{-1}a^{-1}) = a(bb^{-1})a^{-1} = aea^{-1} = aa^{-1} = e$. By the corollary above, one product equal to $e$ is enough, so $b^{-1}a^{-1} = (ab)^{-1}$. For $k$ factors use induction on $k$: the case $k = 1$ is trivial, and if the formula holds for $k$ factors, then writing $a_1 \cdots a_{k+1} = (a_1 \cdots a_k)\,a_{k+1}$ and applying the case of two factors gives
$$(a_1 \cdots a_{k+1})^{-1} = a_{k+1}^{-1}\,(a_1 \cdots a_k)^{-1} = a_{k+1}^{-1} a_k^{-1} \cdots a_1^{-1}. \qquad \blacksquare$$
A common slip is to write $(ab)^{-1} = a^{-1}b^{-1}$, which in general is false. The cube of the opening shows why. The rule says that $(RU)^{-1} = U^{-1}R^{-1} = U'R'$, and the script confirms that this move undoes $RU$ while $R'U'$ does not. The cube article reads its products left to right, but the reversal rule is the same in either convention. In an abelian group the order of the factors does not matter, and $(ab)^{-1} = a^{-1}b^{-1}$.
Longer products reverse in the same way. In $D_8$, $(r^3 s r^2)^{-1} = r^{-2} s^{-1} r^{-3}$, and the script finds that both sides equal the reflection $rs$.
The rule is the course text's Corollary 7.6 and Judson's Proposition 3.2.12. It was already used for units in a ring, where the inverse of a product of units $ab$ is $b^{-1}a^{-1}$, and for permutations, where $(\sigma\tau)^{-1} = \tau^{-1}\sigma^{-1}$.
Products without parentheses
The associative law is stated for three elements, but it is used for products of any length: $abcd$ is written without parentheses on the understanding that every way of inserting them gives the same result. For four factors there are five ways,
$$((ab)c)d, \quad (a(bc))d, \quad (ab)(cd), \quad a((bc)d), \quad a(b(cd)),$$
and for five factors there are 14. The script evaluates all 5 bracketings of the product $r^3 \cdot s \cdot r^5 \cdot r^2 s$ in $D_8$ and gets $r^4$ every time. It evaluates all 14 bracketings of that product followed by $r$ and gets $r^5$ every time. That this always happens needs a proof. Here is the plan: show that every bracketing equals the one that multiplies from the left, one factor at a time, by moving the last factor out with a single use of the associative law.
Theorem (generalized associativity). For any elements $a_1, \ldots, a_k$ of $G$, every way of inserting parentheses into $a_1 a_2 \cdots a_k$ gives the same element, namely the left-normed product $L_k = (\cdots((a_1 a_2) a_3) \cdots ) a_k$, defined by $L_1 = a_1$ and $L_{j+1} = L_j \, a_{j+1}$.
Proof. By strong induction on $k$. For $k \le 2$ there is only one bracketing. Let $k \ge 3$ and assume the statement for every product of fewer than $k$ factors (with any elements). A bracketing of $a_1 \cdots a_k$ is a product $XY$ of a bracketing $X$ of $a_1 \cdots a_j$ and a bracketing $Y$ of $a_{j+1} \cdots a_k$, for some $j$ with $1 \le j < k$. By the induction hypothesis $X = L_j$, and $Y$ equals the left-normed product of $a_{j+1}, \ldots, a_k$. If $j = k - 1$, then $Y = a_k$ and $XY = L_{k-1}a_k = L_k$. If $j < k - 1$, the left-normed product of $a_{j+1}, \ldots, a_k$ is $Y' a_k$, where $Y'$ is the left-normed product of $a_{j+1}, \ldots, a_{k-1}$. Then, by the associative law for the three elements $X$, $Y'$, $a_k$,
$$XY = X(Y'a_k) = (XY')\,a_k.$$
Now $XY'$ is a bracketing of the $k - 1$ factors $a_1, \ldots, a_{k-1}$, so by the induction hypothesis it equals $L_{k-1}$, and $XY = L_{k-1}a_k = L_k$. $\blacksquare$
Many textbooks take this theorem for granted; it is the reason the group axioms list associativity only for three elements. Commutativity, by contrast, cannot be added for free. In a nonabelian group the order of the factors matters even though the grouping does not.
Powers and the laws of exponents
A product of one element with itself needs no parentheses either, so powers are well defined. Do they obey the familiar rules for exponents? For an element $a$ and an integer $n$, the powers of $a$ are defined by
$$a^0 = e, \qquad a^n = \underbrace{a a \cdots a}_{n \text{ factors}} \ (n \ge 1), \qquad a^{-n} = (a^{-1})^n \ (n \ge 1).$$
By the inverse of a product, $(a^n)^{-1} = (a^{-1})^n = a^{-n}$ for $n \ge 1$, since reversing the order of $n$ equal factors changes nothing. So the two possible meanings of $a^{-n}$, the inverse of $a^n$ and the $n$th power of $a^{-1}$, agree.
For positive exponents the laws are just counting factors. The work is in the negative exponents. Here is the plan: prove the case of one extra factor $a$ or $a^{-1}$ for every integer $m$, then induct upward and downward from $n = 0$.
Theorem (laws of exponents). For every $a$ in $G$ and all integers $m$ and $n$,
$$a^m a^n = a^{m+n} \qquad \text{and} \qquad (a^m)^n = a^{mn}.$$
Proof. The first step is the case $n = 1$: $a^m a = a^{m+1}$ for every integer $m$. For $m \ge 0$ this is the definition. For $m = -p$ with $p \ge 1$, $a^{-p}a = (a^{-1})^p a = (a^{-1})^{p-1}(a^{-1}a) = (a^{-1})^{p-1} = a^{-(p-1)} = a^{m+1}$, where for $p = 1$ the factor $(a^{-1})^0$ is $e$. Similarly $a^m a^{-1} = a^{m-1}$ for every integer $m$: for $m = -p \le 0$ this is the definition $(a^{-1})^p a^{-1} = (a^{-1})^{p+1}$, and for $m \ge 1$, $a^m a^{-1} = (a^{m-1}a)a^{-1} = a^{m-1}(aa^{-1}) = a^{m-1}$.
Now fix $m$ and prove $a^m a^n = a^{m+n}$ by induction on $n \ge 0$: it holds for $n = 0$, and if it holds for $n$, then $a^m a^{n+1} = a^m (a^n a) = (a^m a^n) a = a^{m+n} a = a^{m+n+1}$, using the first step twice. For negative $n$, induct downward in the same way: if $a^m a^n = a^{m+n}$, then $a^m a^{n-1} = a^m(a^n a^{-1}) = (a^m a^n)a^{-1} = a^{m+n}a^{-1} = a^{m+n-1}$.
For the second law, fix $m$. For $n \ge 0$ induct on $n$: $(a^m)^0 = e = a^0$, and $(a^m)^{n+1} = (a^m)^n a^m = a^{mn} a^m = a^{mn + m}$ by the first law. For $n = -p$ with $p \ge 1$, $(a^m)^{-p}$ is the inverse of $(a^m)^p = a^{mp}$, and the inverse of $a^{mp}$ is $a^{-mp}$, because $a^{mp}a^{-mp} = a^0 = e$ by the first law. $\blacksquare$
The course text states these laws as Theorem 7.7 and Judson as Theorem 3.2.16. The script checks both laws for every element of $U(9)$ and of $D_8$ and all exponents from $-30$ to $30$. In particular all powers of one element commute with each other, since $a^m a^n = a^{m+n} = a^n a^m$.
Check your understanding. On the cube, which move undoes $(RU)^2 = RURU$: $(U'R')^2$ or $(R'U')^2$?
It is $(U'R')^2$. By the laws of exponents $((RU)^2)^{-1} = ((RU)^{-1})^2$, and $(RU)^{-1} = U'R'$ by the reversal rule. The script confirms that $(U'R')^2$ undoes $RURU$ and that $(R'U')^2$ does not.
A third law of the real numbers, $(ab)^n = a^n b^n$, needs the factors to commute. The idea: if $a$ and $b$ commute, every $b$ in $abab \cdots ab$ can be slid to the right past the $a$'s. Suppose $ab = ba$. Then $b^n a = a b^n$ for every $n \ge 0$, by induction: $b^{n+1}a = b^n(ba) = b^n(ab) = (b^n a)b = (ab^n)b = ab^{n+1}$. So $(ab)^n = a^n b^n$ for $n \ge 0$, again by induction: $(ab)^{n+1} = (ab)^n(ab) = a^n (b^n a) b = a^n (a b^n) b = a^{n+1}b^{n+1}$. For $n = -p$ with $p \ge 1$, $(ab)^{-p}$ is the inverse of $(ab)^p = a^p b^p = b^p a^p$. (The last step is the same induction with $a$ and $b$ exchanged.) By the reversal rule that inverse is $a^{-p}b^{-p}$.
Without commutativity the law fails. In $D_8$, the script finds that $(ab)^2 \neq a^2 b^2$ for 144 of the 256 pairs $(a, b)$. On the cube, $(RU)^2 = RURU$ and $R^2U^2$ place 18 of the 48 movable stickers differently.
In additive notation the laws of exponents become the familiar rules for multiples, which were proved for rings in Basic Properties of Rings:
| Multiplicative | Additive |
|---|---|
| $a^0 = e$ | $0a = 0$ |
| $a^{-n} = (a^{-1})^n = (a^n)^{-1}$ | $(-n)a = n(-a) = -(na)$ |
| $a^m a^n = a^{m+n}$ | $ma + na = (m + n)a$ |
| $(a^m)^n = a^{mn}$ | $n(ma) = (nm)a$ |
| $(ab)^n = a^n b^n$ if $ab = ba$ | $n(a + b) = na + nb$ (always, since $+$ is commutative) |
The order of an element
When are two powers of an element equal? The order $\lvert a \rvert$ is defined in Groups: Definition and First Examples, where the powers of 2 in $U(9)$ repeat with period 6. For units of $\mathbb{Z}_n$, Fermat's little theorem proved that $a^m = 1$ exactly when the order divides $m$, for $m \ge 0$. That proof works in any group; the version below adds negative exponents and infinite order.
Here is the idea in one sentence. If two different powers agree, dividing one by the other gives a positive power equal to $e$, and the division algorithm then reduces every exponent to its remainder modulo the order.
Theorem. Let $a \in G$.
- If $a$ has infinite order, then the powers $a^k$, $k \in \mathbb{Z}$, are all different.
- If $a$ has finite order $n$, then for all integers $k$, $i$ and $j$: (a) $a^k = e$ if and only if $n \mid k$; (b) $a^i = a^j$ if and only if $i \equiv j \pmod n$; (c) the powers of $a$ are exactly the $n$ different elements $e, a, a^2, \ldots, a^{n-1}$.
Proof. The key step: if $a^i = a^j$ with $i > j$, multiplying both sides by $a^{-j}$ and using the laws of exponents gives $a^{i-j} = a^0 = e$ with $i - j$ positive. So if two different powers of $a$ coincide, then some positive power of $a$ is $e$, and $a$ has finite order. This proves (1).
For (2), suppose $\lvert a \rvert = n$. If $n \mid k$, say $k = nq$, then $a^k = (a^n)^q = e^q = e$. Conversely, suppose $a^k = e$. By the division algorithm, $k = nq + r$ with $0 \le r < n$, and
$$e = a^k = a^{nq} a^r = (a^n)^q a^r = a^r.$$
Since $n$ is the least positive integer with $a^n = e$ and $0 \le r < n$, $r$ cannot be positive, so $r = 0$ and $n \mid k$. This proves (a). For (b), $a^i = a^j$ holds if and only if $a^{i-j} = e$ (multiply by $a^{-j}$ or by $a^j$), which by (a) happens exactly when $n \mid i - j$. For (c), every integer $k$ is congruent modulo $n$ to exactly one of $0, 1, \ldots, n - 1$, namely its remainder; so $a^k$ equals one of $e, a, \ldots, a^{n-1}$, and these $n$ powers are different from each other because no two of $0, \ldots, n - 1$ are congruent modulo $n$. $\blacksquare$
The theorem combines the course text's Theorems 7.8 and 7.9 (Judson proves the same facts for cyclic groups in Proposition 4.1.12 and §4.1). It turns computations with large exponents into arithmetic modulo the order. In $U(9)$, $100 = 6 \cdot 16 + 4$, so $2^{100} = 2^4 = 7$. Negative exponents work the same way: $2^{-5} = 2^{1} = 2$, since $-5 \equiv 1 \pmod 6$.
Check your understanding. What is $2^{-100}$ in $U(9)$?
It is $2^2 = 4$, since $-100 = 6 \cdot (-17) + 2$, so $-100 \equiv 2 \pmod 6$. As a cross-check, $2^{-100}$ is the inverse of $2^{100} = 7$, and $4 \cdot 7 = 28 = 1$ in $U(9)$. The script agrees.
Exponents behave like elements of $\mathbb{Z}_n$, an observation that Group Homomorphisms and Isomorphisms turns into an isomorphism between the powers of $a$ and the additive group $\mathbb{Z}_n$.
Every element of a finite group has finite order
Must an element have finite order at all? In a finite group it must, by the argument that Fermat's little theorem used for units: a finite group has no room for infinitely many different powers.
Proposition. If $G$ is finite, every element $a$ of $G$ has finite order, and $\lvert a \rvert \le \lvert G \rvert$.
Proof. If $a$ had infinite order, part (1) of the theorem would make the powers $a^k$, $k \in \mathbb{Z}$, infinitely many different elements of the finite set $G$, which is impossible. So $a$ has some finite order $n$, and by part (2)(c) the $n$ powers $e, a, \ldots, a^{n-1}$ are different elements of $G$; hence $n \le \lvert G \rvert$. $\blacksquare$
The same finiteness argument was used for permutations in Symmetric Groups and Permutations. It is why every move of the Rubik's cube has finite order: some number of repetitions of any move restores the cube. The bound $\lvert a \rvert \le \lvert G \rvert$ is far from the truth for the cube, whose largest element order is 1260. The next unit improves it: by Lagrange's theorem the order of an element divides the order of the group.
In an infinite group the proposition says nothing, and an element can have infinite order. In the group $\mathbb{C}^\times$ of nonzero complex numbers, $1 + i$ has infinite order, since $\lvert (1 + i)^k \rvert = 2^{k/2} > 1$ for every $k \ge 1$.
The order of a power
The cube's move $RU$ has order 105. What is the order of $(RU)^{15}$, or of any other power? If $a$ has order $n$, the powers of $a^k$ are among the powers of $a$, and their number is the order of $a^k$. Place $a^0, a^1, \ldots, a^{n-1}$ around a circle. The powers of $a^k$ are reached by stepping $k$ places at a time, and the walk closes when it returns to $a^0$.
Here is the idea in one sentence: the walk closes after $m$ steps exactly when $km$ is a multiple of $n$, and dividing out the common factor $\gcd(n, k)$ leaves the answer.
Theorem. Let $a$ have finite order $n$ and let $k$ be an integer. Then
$$\lvert a^k \rvert = \frac{n}{\gcd(n, k)}.$$
If $a$ has infinite order and $k \neq 0$, then $a^k$ has infinite order.
Proof. Let $d = \gcd(n, k)$ (for $k = 0$, $d = n$ and the formula gives $\lvert e \rvert = 1$). Write $n = dn'$ and $k = dk'$. Then $\gcd(n', k') = 1$, since a common divisor greater than 1 would make $d$ not the greatest. For a positive integer $m$, the theorem on orders gives
$$(a^k)^m = a^{km} = e \iff n \mid km \iff dn' \mid dk'm \iff n' \mid k'm \iff n' \mid m,$$
where the last step holds because $n'$ and $k'$ are relatively prime (by the theorem in Divisibility and the Euclidean Algorithm that $a \mid bc$ and $\gcd(a, b) = 1$ imply $a \mid c$). The least positive $m$ with $n' \mid m$ is $n' = n/d$. If $a$ has infinite order and $(a^k)^m = a^{km} = e$ for some $m \ge 1$, then $km = 0$ by part (1) of the theorem on orders, which is impossible for $k \neq 0$. $\blacksquare$
The course text proves the special case $k \mid n$ (Theorem 7.9(3)), and Judson the general formula for cyclic groups (Theorem 4.1.13). The rotation case $\lvert r^k \rvert = n/\gcd(n, k)$ in $D_n$ was proved in Dihedral Groups by the same argument.
Check your understanding. In the figure, stepping by 4 visits the 9 even positions. Which positions does stepping by 14 visit?
The same 9 even positions, in a different order. Since $\gcd(18, 14) = 2 = \gcd(18, 4)$, both walks have $18/2 = 9$ steps, and every position they visit is even. So $a^{14}$ and $a^4$ have the same order and the same powers. The script confirms both. Subgroups and Cyclic Groups proves that in general the powers of $a^k$ depend only on $\gcd(n, k)$.
Some consequences of the formula:
- Which powers have full order. $a^k$ has the same order $n$ as $a$ exactly when $\gcd(n, k) = 1$. In $U(9)$, where $\lvert 2 \rvert = 6$, the exponents 1 and 5 are the ones prime to 6, so $2^1 = 2$ and $2^5 = 5$ are the elements of order 6.
- The cube, read backward. The formula also answers the reverse question: which powers of an element have a given order? The move $RU$ has order 105 (The Rubik's Cube Group), and $105 = 3 \cdot 5 \cdot 7$. A power $(RU)^k$ has order 7 exactly when $105/\gcd(105, k) = 7$, that is, when $\gcd(105, k) = 15$. Among the exponents from 0 to 104 these are $k = 15, 30, 45, 60, 75$ and 90, six of them. At the other extreme, $(RU)^k$ has the full order 105 for the 48 exponents $k$ prime to 105. The script computes the powers on the cube's 48 stickers and finds exactly the six of order 7. One power deserves a name: $(RU)^{35}$ has order $105/35 = 3$, and it is the move described in the cube article that leaves every cubie in place with six corners twisted, so that three applications of it restore the cube.
- Additive notation. In the additive group $\mathbb{Z}_n$, the element $k$ is the $k$th multiple of 1, and 1 has order $n$, so the order of $k$ is $n/\gcd(n, k)$. In $\mathbb{Z}_{36}$, $\lvert 8 \rvert = 36/4 = 9$, $\lvert 30 \rvert = 36/6 = 6$ and $\lvert 27 \rvert = 36/9 = 4$. The same division by $d$ was used to count the solutions of linear congruences.
Orders in a direct product
The last tool combines two groups. In the direct product $G \times H$, powers are computed coordinate by coordinate: $(g, h)^k = (g^k, h^k)$. This holds by induction on $k$ for $k \ge 0$, and by the inverse rule $(g, h)^{-1} = (g^{-1}, h^{-1})$ for negative $k$. The idea of the next result: a pair returns to the identity exactly when both coordinates do at once.
Proposition. If $g \in G$ and $h \in H$ have finite orders, then $\lvert (g, h) \rvert = \operatorname{lcm}(\lvert g \rvert, \lvert h \rvert)$.
Proof. $(g, h)^k = (g^k, h^k)$ is the identity $(e_G, e_H)$ exactly when $g^k = e_G$ and $h^k = e_H$, that is, when $\lvert g \rvert \mid k$ and $\lvert h \rvert \mid k$, by the theorem on orders. The least positive such $k$ is the least common multiple. $\blacksquare$
Judson proves this as Theorem 9.2.5. It is the group version of the rule that the order of a permutation is the least common multiple of the lengths of its disjoint cycles (Symmetric Groups and Permutations). Disjoint cycles act independently, like the coordinates of a pair. The check uses $\mathbb{Z}_5 \times U(9)$, the group of order 30 from Groups: Definition and First Examples, whose first coordinates are added modulo 5 and second coordinates multiplied modulo 9:
| Element | Order of first coordinate | Order of second coordinate | Order |
|---|---|---|---|
| $(3, 2)$ | 5 | 6 | 30 |
| $(2, 7)$ | 5 | 3 | 15 |
| $(1, 8)$ | 5 | 2 | 10 |
| $(4, 1)$ | 5 | 1 | 5 |
| $(0, 4)$ | 1 | 3 | 3 |
The script checks the formula for all 30 elements.
Check your understanding. How many elements of $\mathbb{Z}_5 \times U(9)$ have order 30?
Eight. The order is $\operatorname{lcm}(\lvert u \rvert, \lvert v \rvert) = 30$ exactly when $u$ has order 5 and $v$ has order 6. That allows the 4 nonzero classes of $\mathbb{Z}_5$ and the 2 elements 2 and 5 of $U(9)$, so $4 \cdot 2 = 8$ pairs. The script counts 8.
Since $(3, 2)$ has order 30, its powers are 30 different elements, so they make up the whole group: $\mathbb{Z}_5 \times U(9)$ consists of the powers of a single element. Groups of this kind are the cyclic groups of Subgroups and Cyclic Groups, where the proposition decides exactly when a product $\mathbb{Z}_m \times \mathbb{Z}_n$ is cyclic.
Summary
| Fact | Statement |
|---|---|
| uniqueness | one identity; one inverse for each element; $(a^{-1})^{-1} = a$ |
| one-sided check | $ab = e$ implies $b = a^{-1}$ |
| cancellation | $ab = ac \Rightarrow b = c$ and $ba = ca \Rightarrow b = c$ |
| equations | $ax = b$ and $ya = b$ have the unique solutions $a^{-1}b$ and $ba^{-1}$ |
| inverse of a product | $(a_1 \cdots a_k)^{-1} = a_k^{-1} \cdots a_1^{-1}$ |
| exponents | $a^m a^n = a^{m+n}$, $(a^m)^n = a^{mn}$; $(ab)^n = a^n b^n$ if $ab = ba$ |
| order | if $\lvert a \rvert = n$: $a^k = e \iff n \mid k$, and $a^i = a^j \iff i \equiv j \pmod n$ |
| finite groups | every element has finite order, at most $\lvert G \rvert$ |
| powers | $\lvert a^k \rvert = n/\gcd(n, k)$ |
| products | $\lvert (g, h) \rvert = \operatorname{lcm}(\lvert g \rvert, \lvert h \rvert)$ |
Where it leads
- The powers of a single element form a subgroup with $\lvert a \rvert$ elements, the cyclic subgroup $\langle a \rangle$; subgroups, cyclic groups and their subgroups are in Subgroups and Cyclic Groups.
- The theorem on orders says that the powers of an element of order $n$ multiply like the elements of $\mathbb{Z}_n$; Group Homomorphisms and Isomorphisms makes this an isomorphism and shows that homomorphisms preserve the rules proved here.
- The Latin square property makes each row of a Cayley table a permutation, the starting point of Cayley's theorem.
- That the order of an element divides the order of a finite group is a consequence of Lagrange's theorem.
What we have learned
Four axioms are enough for the working rules of group theory. Inverses are unique, elements cancel on the side where they stand, and every Cayley table is a Latin square. Products reverse under inversion, and the laws of exponents hold for all integers. Exponents reduce modulo the order, and $\lvert a^k \rvert = n/\gcd(n, k)$, as for the powers of $RU$.
The powers of one element form a group of their own inside $G$. Which subsets of a group are groups in their own right? The next article, Subgroups and Cyclic Groups, answers that.
References
- Hungerford, T. W. (2014). Abstract Algebra: An Introduction (3rd ed.), §7.2. Brooks/Cole, Cengage Learning. The course text: uniqueness of the identity and of inverses and the cancellation laws (Theorem 7.5); $(ab)^{-1} = b^{-1}a^{-1}$ and $(a^{-1})^{-1} = a$ (Corollary 7.6); the laws of exponents (Theorem 7.7); powers of an element of infinite order are distinct (Theorem 7.8); for an element of order $n$, $a^k = e$ exactly when $n \mid k$, $a^i = a^j$ exactly when $i \equiv j \pmod n$, and $a^t$ has order $d$ when $n = td$ (Theorem 7.9); multiplicative and additive notation. record · 44 pages
- Judson, T. W. (2026). Abstract Algebra: Theory and Applications (annual edition, August 2026). Free online at judsonbooks.org. §3.2 (Propositions 3.2.11 to 3.2.15: uniqueness of inverses, $(ab)^{-1} = b^{-1}a^{-1}$, $(a^{-1})^{-1} = a$, unique solutions of $ax = b$ and $xa = b$, cancellation; Theorem 3.2.16, the laws of exponents); §3.5, Exercise 2 (tables to be tested against the group axioms); §4.1 (Proposition 4.1.12 and Theorem 4.1.13, the order of $a^k$ in a cyclic group of order $n$); §9.2 (Theorem 9.2.5, the order of an element of a direct product). record · 44 pages