The sixteen symmetries of the stop sign's octagon fall into two sets of eight: the rotations $1, r, \ldots, r^7$ and the reflections $s, rs, \ldots, r^7 s$ (Dihedral Groups). The rotations form a subgroup. The reflections do not, but they are a copy of it moved over: multiplying every rotation on the right by $s$ produces every reflection, once each. In $\mathbb{Z}_{36}$ the multiples of 4 form a subgroup of 9 elements. Adding 1, 2 or 3 to each of them produces three more sets of 9, and the four sets fill $\mathbb{Z}_{36}$ without overlap. Each time, a subgroup and its translates cut the group into pieces of the same size. Does every subgroup cut every group this way, and what does that force on the sizes of subgroups?
The translates are called cosets, and the answer to the first question is yes. The answer to the second is the first general theorem about finite groups: the order of a subgroup divides the order of the group. That is Lagrange's theorem, and it constrains every finite group:
- a group of order 16 has no subgroup of order 6 and no element of order 3;
- a group of prime order has no subgroups other than the two it must have, and is therefore cyclic;
- for every unit $a$ modulo $n$, $a^{\varphi(n)} \equiv 1$. This is Euler's theorem from Fermat's Little Theorem, recovered as a fact about the group $U(n)$.
First we define cosets and see in $D_8$ that left and right cosets can differ. Then we prove that the cosets split a group into equal pieces, which is Lagrange's theorem. Finally we follow its consequences, from the orders of elements to the orbit-stabilizer relation behind the order of the Rubik's cube group.
This is the first article of Cosets, Normal Subgroups and Quotient Groups, and the home of cosets of a subgroup and of Lagrange's theorem. It covers §8.1 of the course text (Hungerford, 2014), with Judson (2026, Chapter 6) as a free companion. The course text works with right cosets $Ka$. This article, like Judson and the class sessions, uses left cosets $gH$, and every statement holds with the sides exchanged. It uses Subgroups and Cyclic Groups and the equivalence relations of an earlier unit.
Cosets
Definition. Let $H$ be a subgroup of a group $G$ and $g \in G$. The left coset of $H$ with representative $g$ is
$$gH = \{\, gh : h \in H \,\},$$
and the right coset is $Hg = \{\, hg : h \in H \,\}$. In an additive group they are written $g + H$ and $H + g$, and they are equal, since addition is commutative.
The subgroup itself is a coset, $H = eH = He$. It is the only coset that is a subgroup, since a coset containing $e$ is $H$ itself, as shown below. Two families of examples have been met already.
Congruence classes. In $\mathbb{Z}$, the cosets of $n\mathbb{Z}$ are the sets $a + n\mathbb{Z}$, which are the congruence classes modulo $n$. In $\mathbb{Z}_{36}$, the subgroup $\langle 4 \rangle$ of multiples of 4 has the four cosets
| Coset | Elements |
|---|---|
| $0 + \langle 4 \rangle$ | 0, 4, 8, 12, 16, 20, 24, 28, 32 |
| $1 + \langle 4 \rangle$ | 1, 5, 9, 13, 17, 21, 25, 29, 33 |
| $2 + \langle 4 \rangle$ | 2, 6, 10, 14, 18, 22, 26, 30, 34 |
| $3 + \langle 4 \rangle$ | 3, 7, 11, 15, 19, 23, 27, 31, 35 |
Each is the set of classes with a given remainder modulo 4. Since 4 divides 36, that remainder is well defined on $\mathbb{Z}_{36}$. The cosets of an ideal were sets of exactly this kind, $a + I$, for the additive group of a ring. Rotations and reflections. In $D_8$, the rotation subgroup $\langle r \rangle$ has two left cosets: $\langle r \rangle$ itself, and $s\langle r \rangle = \{ s r^a \} = \{ r^{-a} s \}$, which is the set of all eight reflections. The right cosets are the same two sets.
Left and right cosets can differ
Is $gH$ always the same set as $Hg$? In a nonabelian group it need not be. Take the subgroup
$$H = \langle r^4, s \rangle = \{1, \; r^4, \; s, \; r^4 s\}$$
of $D_8$, of order 4, made of the half turn, the identity and the reflections in two perpendicular mirror lines. Its left cosets are computed with the rule $r^a \cdot r^b s = r^{a+b}s$. Its right cosets need the rule $s r^a = r^{-a} s$ of Dihedral Groups as well; for example $Hr$ contains $sr = r^{-1}s = r^7 s$. The script finds:
| Representative | Left coset $gH$ | Right coset $Hg$ |
|---|---|---|
| $1$ | $1, r^4, s, r^4 s$ | $1, r^4, s, r^4 s$ |
| $r$ | $r, r^5, rs, r^5 s$ | $r, r^5, r^3 s, r^7 s$ |
| $r^2$ | $r^2, r^6, r^2 s, r^6 s$ | $r^2, r^6, r^2 s, r^6 s$ |
| $r^3$ | $r^3, r^7, r^3 s, r^7 s$ | $r^3, r^7, rs, r^5 s$ |
Both families consist of four sets of four that fill $D_8$ without overlap, but they are different families. The left coset $rH$ and the right coset $Hr$ share the rotations $r, r^5$ and differ in their reflections. For $g = r^2$ the two cosets agree, and the script finds that $gH = Hg$ holds for exactly eight of the sixteen elements, those of $\langle r^2, s \rangle$. Subgroups for which left and right cosets always agree are the normal subgroups of the next article.
On either side, the cosets were pieces of one size that did not overlap. The next two sections prove that this always happens.
Congruence modulo a subgroup
The cosets of $H$ are the classes of an equivalence relation, as the congruence classes modulo $n$ were.
Definition. Let $H \le G$. For $a, b \in G$, say $a$ is congruent to $b$ modulo $H$ if $a^{-1}b \in H$.
In additive notation the condition reads $-a + b \in H$, or $b - a \in H$. For $H = n\mathbb{Z}$ that is ordinary congruence modulo $n$, and for an ideal $I$ it is congruence modulo $I$. In a nonabelian group the order of the factors matters: the condition $ab^{-1} \in H$, which the course text uses, describes the right cosets instead.
Theorem. Congruence modulo $H$ is an equivalence relation on $G$, and the equivalence class of $a$ is the left coset $aH$.
The idea: the three properties of an equivalence relation are the three subgroup properties (identity, inverses, closure) read through the element $a^{-1}b$.
Proof. Reflexive: $a^{-1}a = e \in H$. Symmetric: if $a^{-1}b \in H$, then its inverse $(a^{-1}b)^{-1} = b^{-1}a$ is in $H$, by the reversal rule and closure of $H$ under inverses. Transitive: if $a^{-1}b \in H$ and $b^{-1}c \in H$, then their product $a^{-1}b\,b^{-1}c = a^{-1}c$ is in $H$.
The class of $a$. If $b$ is congruent to $a$, then $h = a^{-1}b$ is in $H$ and $b = ah \in aH$. Conversely, an element $ah$ of $aH$ satisfies $a^{-1}(ah) = h \in H$. So the class of $a$ is exactly $aH$. $\blacksquare$
Because the classes of an equivalence relation partition the set they live in (Equivalence Relations and Partitions), the left cosets of $H$ partition $G$: every element lies in exactly one of them. The theorem also gives a practical test for when two cosets are the same.
Corollary. For $a, b \in G$ the following are equivalent:
- $aH = bH$;
- $b \in aH$;
- $a^{-1}b \in H$;
- $aH$ and $bH$ have an element in common.
The idea: each statement says, in its own way, that $a$ and $b$ lie in the same class.
Proof. Statements 2 and 3 both say that $b$ lies in the class of $a$, by the theorem. If 3 holds, $a$ and $b$ are in the same class, so their classes $aH$ and $bH$ are equal, which is 1. If 1 holds, then $b = be \in bH = aH$, which is 2, and in particular $b$ is a common element, which is 4. If 4 holds, a common element $c$ is congruent to both $a$ and $b$, so $a$ and $b$ are congruent to each other, which is 3. $\blacksquare$
So two cosets are either equal or disjoint, and any element of a coset can serve as its representative: if $b \in aH$, then $bH = aH$. A coset has many names, one for each of its elements, and none of them is the correct one. With $a = e$, a coset $bH$ equals $H$ exactly when $b \in H$. That is why $H$ is the only coset containing $e$, and the only coset that is a subgroup. Back to the table above: $r^3 H$ contains $r^7 s$, as the test confirms: $(r^3)^{-1} r^7 s = r^4 s \in H$. And $r^3 s$ is not in $rH$, since $r^{-1} \cdot r^3 s = r^2 s \notin H$.
The cosets never partly overlap. Do they also all have the same size?
All cosets have the same size
Lemma. For every $g \in G$, the map $h \mapsto gh$ is a bijection from $H$ onto $gH$. In particular, if $H$ is finite, every left coset of $H$ has exactly $\lvert H \rvert$ elements, and so does every right coset.
The idea: multiplying by $g$ can be undone by multiplying by $g^{-1}$, so it never merges two elements.
Proof. The map is onto $gH$ by the definition of $gH$. It is one-to-one by left cancellation: $gh = gh'$ implies $h = h'$ (Basic Properties of Groups). For right cosets use $h \mapsto hg$ and right cancellation. $\blacksquare$
The lemma is where cancellation is used. In a set with an operation but no cancellation, multiplying a set by an element can shrink it.
Check your understanding. The cosets cover $G$, never partly overlap, and all have $\lvert H \rvert$ elements. Which of these facts survive for the translates of a subset that is not a subgroup, such as $S = \{1, r\}$ in $D_8$?
Covering and equal size survive, since $g \in gS$ and cancellation still holds. Disjointness fails: $1S = \{1, r\}$ and $rS = \{r, r^2\}$ share $r$ without being equal. The script finds 16 different translates of 2 elements each, too many to be separate pieces of a 16-element group. Disjointness came from the equivalence relation, which needed the subgroup properties.
Definition. The index of $H$ in $G$, written $[G : H]$, is the number of left cosets of $H$ in $G$.
The index counts cosets, which may be infinitely many: $[\mathbb{Z} : n\mathbb{Z}] = n$ for $n \ge 1$, while $\{0\}$ has infinitely many cosets in $\mathbb{Z}$. The definition chose left cosets. Could a subgroup have more right cosets than left ones? The next proposition shows that it cannot.
Proposition. The rule $gH \mapsto Hg^{-1}$ is a well-defined bijection from the set of left cosets of $H$ onto the set of right cosets. So $H$ has as many right cosets as left cosets.
The idea: taking inverses of all the elements of a set turns the left coset $gH$ into the right coset $Hg^{-1}$, since $(gh)^{-1} = h^{-1}g^{-1}$.
Proof. Right cosets satisfy the mirror image of the corollary: $Hx = Hy$ exactly when $xy^{-1} \in H$. Now $aH = bH$ holds exactly when $a^{-1}b \in H$, and $a^{-1}b = a^{-1}(b^{-1})^{-1}$, so this holds exactly when $Ha^{-1} = Hb^{-1}$. Read from left to right, this says the rule does not depend on the representative, so it is a function. Read from right to left, it says the rule is one-to-one. It is onto, since $Hg$ is the image of $g^{-1}H$. $\blacksquare$
Judson (2026, Theorem 6.1.8) proves the count with the same map.
Lagrange's theorem
Theorem (Lagrange). Let $G$ be a finite group and $H$ a subgroup of $G$. Then
$$\lvert G \rvert = [G : H] \cdot \lvert H \rvert.$$
In particular, the order of every subgroup of $G$ divides the order of $G$, and $[G : H] = \lvert G \rvert / \lvert H \rvert$.
The idea: $G$ is cut into $[G : H]$ separate pieces of $\lvert H \rvert$ elements each.
Proof. Let $g_1 H, \ldots, g_k H$ be the different left cosets of $H$, so $k = [G : H]$. By the theorem on congruence modulo $H$, every element of $G$ lies in exactly one of them, so
$$\lvert G \rvert = \lvert g_1 H \rvert + \lvert g_2 H \rvert + \cdots + \lvert g_k H \rvert.$$
By the lemma, each of the $k$ terms equals $\lvert H \rvert$. $\blacksquare$
This is the course text's Theorem 8.5 and Judson's Theorem 6.2.2. The proof has three ingredients, each proved above: the cosets cover the group, they do not overlap, and they all have the same size. The theorem carries the name of Joseph-Louis Lagrange, whose memoir of 1770 to 1771 on the solution of equations studied permutations of the roots of polynomials (Groups: Definition and First Examples), long before groups had been defined.
Back to the opening examples. In $\mathbb{Z}_{36}$ the subgroup $\langle 4 \rangle$ gave $36 = 4 \cdot 9$, and in $D_8$ the subgroup $\langle r^4, s \rangle$ gave $16 = 4 \cdot 4$. The script checks the theorem on every subgroup of several groups met in this course:
| Group | Order | Subgroups | Orders of subgroups |
|---|---|---|---|
| $\mathbb{Z}_{36}$ | 36 | 9 | 1, 2, 3, 4, 6, 9, 12, 18, 36 |
| $D_8$ | 16 | 19 | 1, 2, 4, 8, 16 |
| $U(9)$ | 6 | 4 | 1, 2, 3, 6 |
| $U(45)$ | 24 | 16 | 1, 2, 3, 4, 6, 8, 12, 24 |
Lagrange's theorem also gives indices without listing a single coset. Take the Rubik's cube group $G$, of order 43,252,003,274,489,856,000. Every subgroup order in that article's table divides $\lvert G \rvert$. The subgroup $\langle U, D, R2, L2, F2, B2 \rangle$ of order 19,508,428,800 has index
$$\frac{43{,}252{,}003{,}274{,}489{,}856{,}000}{19{,}508{,}428{,}800} = 2{,}217{,}093{,}120,$$
These cosets are the pieces into which Rokicki, Kociemba, Davidson and Dethridge (2013, §4) divided the positions of the cube in their proof that every position can be solved in 20 moves.
Consequences
The idea: each element generates a cyclic subgroup, so its order is a subgroup order too.
Corollary. Let $G$ be a finite group and $a \in G$. Then the order of $a$ divides $\lvert G \rvert$, and $a^{\lvert G \rvert} = e$.
Proof. The order of $a$ is the order of the cyclic subgroup $\langle a \rangle$ (Subgroups and Cyclic Groups), which divides $\lvert G \rvert$ by Lagrange's theorem. Write $\lvert G \rvert = \lvert a \rvert \cdot m$; then $a^{\lvert G \rvert} = (a^{\lvert a \rvert})^m = e^m = e$. $\blacksquare$
This is the course text's Corollary 8.6. The script confirms it for every element of $\mathbb{Z}_{36}$, $D_8$, $U(9)$, $U(32)$ and $U(45)$. The elements of $D_8$ have orders 1, 2, 4 and 8. Those of $U(32)$, also a group of order 16, have orders 1, 2, 4 and 8 as well, three of them of order 2. So neither group has an element of order 3 or 16. On the cube, the move $RU$ has order 105, and the move $R\,U2\,D'\,B\,D'$ has order 1260, the largest possible; both divide
$$\lvert G \rvert = 2^{27} \cdot 3^{14} \cdot 5^3 \cdot 7^2 \cdot 11,$$
and since 13 is not a factor, no move of the cube has order 13 or any multiple of 13.
Cayley's theorem already contained a proof of this corollary. There the permutation $x \mapsto gx$ of $G$ was shown to split $G$ into cycles $\{x, gx, g^2 x, \ldots\}$, all of length $\lvert g \rvert$. Those cycles are the right cosets $\langle g \rangle x$, so that argument was Lagrange's theorem for a cyclic subgroup, counted with right cosets.
Corollary. Every group of prime order $p$ is cyclic, and is generated by each of its elements other than $e$. So every group of order $p$ is isomorphic to $\mathbb{Z}_p$.
The idea: the only divisors of $p$ are 1 and $p$, and only the identity has order 1.
Proof. Let $\lvert G \rvert = p$ and $a \ne e$. The order of $a$ divides $p$ and is not 1, so it is $p$, and the $p$ powers of $a$ fill $G$. A cyclic group of order $p$ is isomorphic to $\mathbb{Z}_p$ by the classification of cyclic groups. $\blacksquare$
This is the course text's Theorem 8.7 and Judson's Corollary 6.2.4. For instance, any group with two elements is a copy of $\{1, -1\}$, the target of the sign of a permutation. The rotations of a heptagon, the seventh roots of unity and $\mathbb{Z}_7$ are one group up to isomorphism. A group of prime order has only the trivial subgroup and itself as subgroups, since no other orders are available.
Euler's and Fermat's theorems
The group of units $U(n)$ consists of the classes modulo $n$ that are relatively prime to $n$, and has $\varphi(n)$ elements (Groups: Definition and First Examples).
Corollary (Euler's theorem). If $n \ge 2$ and $\gcd(a, n) = 1$, then $a^{\varphi(n)} \equiv 1 \pmod n$.
Proof. The class of $a$ lies in $U(n)$, a group of order $\varphi(n)$, so its $\varphi(n)$th power is the identity of $U(n)$, the class of 1, by the previous corollary. $\blacksquare$
Corollary (Fermat's little theorem). If $p$ is prime and $p \nmid a$, then $a^{p-1} \equiv 1 \pmod p$.
Proof. Take $n = p$ in Euler's theorem; $\varphi(p) = p - 1$. $\blacksquare$
The article on Fermat's little theorem proved both by multiplying all the units together, an argument that uses commutativity. The proof by Lagrange's theorem needs only that the units form a group. It also says more: the order of each unit divides $\varphi(n)$.
Check your understanding. $U(23)$ has order 22. Before reading on, list the orders its elements can have. Must an element of each of these orders exist?
The possible orders are the divisors of 22: 1, 2, 11 and 22. Lagrange's theorem allows them but does not promise them. Here all four occur: the script finds one element of order 1, one of order 2, and 10 of each of the orders 11 and 22. Among the latter is 5, so $5^{22} \equiv 1 \pmod{23}$ with no smaller positive exponent.
In $U(45)$, of order $\varphi(45) = 24$, an allowed order is missing. The script finds that the largest order of an element is 12, so $a^{12} \equiv 1 \pmod{45}$ for every unit $a$, and Euler's exponent is not the smallest that works. Lagrange's theorem gives a multiple of each order, not the order itself, a point the Fermat article made for $\mathbb{Z}_{21}$.
The converse is false
A common misreading is "6 divides 12, so a group of order 12 has a subgroup of order 6." Lagrange's theorem says which orders a subgroup may have, not that each of them occurs.
For cyclic groups every divisor does occur, with exactly one subgroup of each order (Subgroups and Cyclic Groups). The same happens in $D_8$ and in $U(45)$: in the table above, their subgroups take every order dividing 16 and 24. But not in every group. The alternating group $A_4$ has order 12, and the script finds subgroups of orders 1, 2, 3, 4 and 12 only: none of order 6. Judson (2026, Proposition 6.2.7) gives a proof by hand. The next article proves that a subgroup of index 2 is always normal, which is the key to such arguments.
Orbits and stabilizers
Lagrange's theorem counts more than subgroups. Many groups are groups of permutations of some set: $D_8$ permutes the eight vertices of the octagon, and the cube group permutes the 48 movable stickers. How many places can a symmetry carry a given point to? For such a group $G$ acting on a set $X$ and a point $x \in X$:
- the orbit of $x$ is the set $G x = \{ g(x) : g \in G \}$ of points that $x$ can be carried to;
- the stabilizer of $x$ is $G_x = \{ g \in G : g(x) = x \}$, the elements that leave $x$ where it is.
The stabilizer is a subgroup: it contains the identity, if $g(x) = x$ and $h(x) = x$ then $gh(x) = g(x) = x$, and applying $g^{-1}$ to $g(x) = x$ gives $x = g^{-1}(x)$.
Theorem (orbit-stabilizer). If $G$ is a finite group of permutations of a set $X$ and $x \in X$, then
$$\lvert G \rvert = \lvert G x \rvert \cdot \lvert G_x \rvert.$$
Here is the idea in one sentence: two symmetries send $x$ to the same point exactly when they differ by a symmetry that fixes $x$, that is, when they lie in one coset of the stabilizer.
Proof. For $g, h \in G$, $g(x) = h(x)$ holds exactly when $h^{-1}g(x) = x$, that is, when $h^{-1}g \in G_x$. By the coset test this means $gG_x = hG_x$. So the rule $gG_x \mapsto g(x)$ is well defined and one-to-one, and it is onto the orbit by the definition of the orbit. The left cosets of $G_x$ therefore correspond one to one with the points of the orbit. So $[G : G_x] = \lvert Gx \rvert$, and Lagrange's theorem gives the formula. $\blacksquare$
The proof shows more than the count: the symmetries sending $x$ to a given point $y$ form one coset of the stabilizer. Take $D_8$ acting on the vertex labels $0, \ldots, 7$. The orbit of vertex 0 is all eight vertices, and its stabilizer is $\{1, s\}$, the identity and the reflection in the mirror line through vertex 0. The formula reads $16 = 8 \cdot 2$, and the script confirms that the two symmetries taking vertex 0 to vertex 3 are $r^3$ and $r^3 s$, which form the coset $r^3\{1, s\}$. Judson (2026, Theorem 14.1.11) states the theorem as $\lvert Gx \rvert = [G : G_x]$.
Check your understanding. The midpoint of the edge from vertex 0 to vertex 1 can be carried to all eight edge midpoints. How many symmetries fix it, and which one besides the identity?
Two, since $16 / 8 = 2$. The other is the reflection $rs$, which sends vertex $k$ to $1 - k$ and so exchanges vertices 0 and 1. The script confirms that the stabilizer is $\{1, rs\}$.
The relation explains the computation of the order of the cube group in The Rubik's Cube Group. The Schreier-Sims algorithm fixes a sticker, finds its orbit, and then works inside the stabilizer of that sticker. The first sticker can be carried to 24 positions, the stickers of the corner cubies, so its stabilizer has index 24 and order
$$\frac{43{,}252{,}003{,}274{,}489{,}856{,}000}{24} = 1{,}802{,}166{,}803{,}103{,}744{,}000.$$
Repeating the step inside each stabilizer, with a new sticker each time, until the stabilizer that remains is trivial, writes $\lvert G \rvert$ as the product of the eighteen orbit sizes, from 24 down to 2, that the cube article lists.
Where it leads
- When left and right cosets agree, the cosets can be multiplied, and they form a group: Normal Subgroups and Quotient Groups.
- The cosets of the kernel of a homomorphism are the sets of elements with the same image. Counting them relates the sizes of a group, its kernel and its image: The First Isomorphism Theorem for Groups.
- The RSA cryptosystem rests on Euler's theorem, here a one-line consequence of Lagrange's theorem in the group $U(n)$.
What we have learned
The cosets of a subgroup $H$ split the group into pieces of $\lvert H \rvert$ elements, so $\lvert G \rvert = [G : H] \cdot \lvert H \rvert$. That one count limits the orders of subgroups and elements, makes groups of prime order cyclic, reproves Euler's and Fermat's theorems, and gives the orbit-stabilizer relation. It does not run backward, as $A_4$ shows.
The cosets of the rotations in $D_8$ multiplied like the signs $1$ and $-1$, but those of $\langle r^4, s \rangle$ did not even agree on the two sides. Which subgroups have cosets that can be multiplied? The next article, Normal Subgroups and Quotient Groups, answers that.
References
- Hungerford, T. W. (2014). Abstract Algebra: An Introduction (3rd ed.), §8.1. Brooks/Cole, Cengage Learning. The course text: congruence modulo a subgroup, defined by $ab^{-1} \in K$, as an equivalence relation (Theorem 8.1) whose classes are the right cosets (Theorem 8.2), which are disjoint or identical (Corollary 8.3) and all of the same size (Theorem 8.4); the index; Lagrange's theorem (Theorem 8.5); the order of an element divides the order of the group, and $a^{\lvert G \rvert} = e$ (Corollary 8.6); groups of prime order are cyclic (Theorem 8.7). record · 44 pages
- Judson, T. W. (2026). Abstract Algebra: Theory and Applications (annual edition, August 2026). Free online at judsonbooks.org. §6.1 (left and right cosets; Lemma 6.1.3, conditions for equal cosets; Theorem 6.1.4, cosets partition the group; Theorem 6.1.8, left and right cosets are equally many); §6.2 (Proposition 6.2.1, cosets have the size of the subgroup; Theorem 6.2.2, Lagrange's theorem; Corollaries 6.2.3 and 6.2.4; Proposition 6.2.7, $A_4$ has no subgroup of order 6); §6.3 (Euler's and Fermat's theorems from Lagrange's theorem); §14.1 (Theorem 14.1.11, orbits and stabilizers). record · 44 pages
- Rokicki, T., Kociemba, H., Davidson, M., and Dethridge, J. (2013). The diameter of the Rubik's cube group is twenty. SIAM Journal on Discrete Mathematics, 27(2), 1082 to 1105. §4: the positions of the cube partitioned into cosets of the subgroup $\langle U, D, R2, L2, F2, B2 \rangle$ (as described in The Rubik's Cube Group). record · 3 pages