Normal subgroups
The left cosets of a subgroup can differ from its right cosets, and then a product of two cosets need not be a coset. Which subgroups have cosets that multiply the way elements do?
Where you are 4 in Cosets and quotients
Why this exists
In Cosets, Example 4, a subgroup H of \operatorname{GL}_2(\mathbb{Z}_2) had left cosets that were different pairs from its right cosets. Its cosets also fail to multiply: (XH)(YH) = \{XY, XP, QY, QP\} = \{P, Y, J, I_2\} has four elements, so it is not a coset. In \mathbb{Z} neither thing happens: the cosets of n\mathbb{Z} add as classes, which is the arithmetic of \mathbb{Z}_n in Relations, equivalence classes and partitions. Normal subgroups are exactly the subgroups whose cosets multiply by (aN)(bN) = abN (Theorem 7). They are the subgroups a group can be divided by, giving the Quotient groups, and exactly the kernels of Homomorphisms and kernels.
Definition
For g, x \in G, the element gxg^{-1} is the conjugate of x by g, and the map c_g : G \to G, c_g(x) = gxg^{-1}, is conjugation by g. For a subset S \subseteq G, gSg^{-1} = \{gxg^{-1} : x \in S\}. If g and x commute, then gxg^{-1} = x.
A subgroup N of G is normal if conjugating an element of N by any element of G always gives an element of N:
that is, gNg^{-1} \subseteq N for every g \in G. We write N \trianglelefteq G. Every group has the normal subgroups \{e\} and G.
For subsets A, B \subseteq G, AB = \{ab : a \in A, b \in B\} and aB = \{a\}B. For H \le G, the normalizer of H is N_G(H) = \{g \in G : gHg^{-1} = H\}. For S \subseteq G, the subgroup generated by S, written \langle S \rangle (or \langle a, b \rangle), is the intersection of all subgroups containing S: a subgroup (Subgroups, Theorem 6) inside every subgroup containing S, and for one element the \langle a \rangle of Cyclic groups. S generates G when \langle S \rangle = G.
Intuition
Conjugation is relabeling. In S_n, by Permutation groups and cycle notation, Exercise 14,
the conjugate does what the cycle does, to the renamed points. In the symmetries of a square, conjugating a reflection by a rotation gives the reflection in the rotated axis (Example 1). So gxg^{-1} is x seen after the labels have been moved by g, and a normal subgroup is one that no relabeling can leave. Such a subgroup is cut out by a property that does not mention labels: “even permutation”, “rotation”, “determinant 1”. The permutations in S_n (n \ge 3) fixing the point 1, or the subgroup of D_4 generated by one reflection, name a label, and are not normal.
The second picture is the cosets. By Theorem 2, N is normal exactly when gN = Ng for every g: multiplying by g on the left or on the right moves N to the same block of the partition, and that is what lets blocks be multiplied.
Results
Throughout, G is a group.
Theorem 1 (conjugation). For each g \in G, c_g is a bijection of G with inverse c_{g^{-1}}, and c_g(xy) = c_g(x)c_g(y) for all x, y. If H \le G, then gHg^{-1} \le G and c_g restricts to a bijection H \to gHg^{-1}; so |gHg^{-1}| = |H| when H is finite.
Theorem 2 (four ways to say normal). For N \le G the following are equivalent: (i) N \trianglelefteq G; (ii) gNg^{-1} = N for every g \in G; (iii) gN = Ng for every g \in G; (iv) every left coset of N is also a right coset of N.
Theorem 3 (checking on generators). For every H \le G, N_G(H) \le G. Now let N \le G, let S generate G, and let T generate N. If sts^{-1} \in N and s^{-1}ts \in N for all s \in S and t \in T, then N \trianglelefteq G. If N is finite, the condition s^{-1}ts \in N may be dropped.
Corollary 4 (the only one of its order). If H \le G is finite and is the only subgroup of G of order |H|, then H \trianglelefteq G.
Theorem 5 (sources of normal subgroups). (a) If G is abelian, every subgroup of G is normal. (b) Every subgroup of index 2 is normal. (c) Every subgroup of the center Z(G) is normal in G; in particular Z(G) \trianglelefteq G. (d) The intersection of a nonempty collection of normal subgroups of G is normal in G.
Theorem 6 (products). Let N \trianglelefteq G and H \le G. Then NH = HN and NH \le G. If also H \trianglelefteq G, then NH \trianglelefteq G.
Theorem 7 (cosets multiply). Let N \le G. Then N \trianglelefteq G if and only if (aN)(bN) = abN for all a, b \in G.
Why it is true
Proof of Theorem 1. c_g(xy) = gxyg^{-1} = gx(g^{-1}g)yg^{-1} = c_g(x)c_g(y). Also c_{g^{-1}}(c_g(x)) = g^{-1}gxg^{-1}g = x, and likewise c_g(c_{g^{-1}}(x)) = x, so c_g is a bijection with inverse c_{g^{-1}} (Composing functions between sets, and when an inverse exists). Let H \le G. Then e = c_g(e) \in gHg^{-1}; c_g(h_1)c_g(h_2) = c_g(h_1h_2); and (ghg^{-1})^{-1} = gh^{-1}g^{-1} = c_g(h^{-1}) by the inverse of a product (Binary operations and the group axioms, Theorem 3). So gHg^{-1} is a subgroup (Subgroups, Theorem 2). The restriction of c_g to H is one to one because c_g is, and onto gHg^{-1} by definition. \square
Proof of Theorem 2. (i) \Rightarrow (ii). Let g \in G. By (i) for g, gNg^{-1} \subseteq N. By (i) for g^{-1}, g^{-1}Ng \subseteq N; applying c_g gives N = g(g^{-1}Ng)g^{-1} \subseteq gNg^{-1}. So gNg^{-1} = N.
(ii) \Rightarrow (iii). Each gn equals (gng^{-1})g, so gN = (gNg^{-1})g = Ng.
(iii) \Rightarrow (iv). The left coset gN equals the right coset Ng.
(iv) \Rightarrow (i). Let g \in G and say gN = Nb. Then g = ge \in Nb, and g \in Ng; two right cosets that share an element are equal (Cosets, Theorem 3), so gN = Nb = Ng. For n \in N, gn \in Ng, so gn = n'g for some n' \in N, and gng^{-1} = n' \in N. \square
Proof of Theorem 3. First, N_G(H) \le G: eHe^{-1} = H; if g, h \in N_G(H), then (gh)H(gh)^{-1} = g(hHh^{-1})g^{-1} = gHg^{-1} = H; and applying c_{g^{-1}} to gHg^{-1} = H gives H = g^{-1}Hg, so g^{-1} \in N_G(H).
Now fix s \in S. By Theorem 1, s^{-1}Ns \le G, so K = N \cap s^{-1}Ns is a subgroup of G (Subgroups, Theorem 6) lying inside N, hence a subgroup of N. Each t \in T lies in K, since t = s^{-1}(sts^{-1})s with sts^{-1} \in N. As T generates N, K = N; that is, N \subseteq s^{-1}Ns, and applying c_s, sNs^{-1} \subseteq N. The same argument with s^{-1} in place of s, using s^{-1}ts \in N, gives s^{-1}Ns \subseteq N, and applying c_s, N \subseteq sNs^{-1}. So s \in N_G(N) for every s \in S. Thus N_G(N) is a subgroup containing S, so it is G, and gNg^{-1} = N for all g. If N is finite, then sNs^{-1} \subseteq N and |sNs^{-1}| = |N| (Theorem 1) already give sNs^{-1} = N, without the second condition. \square
Proof of Corollary 4. For g \in G, gHg^{-1} is a subgroup of order |H| (Theorem 1), so it is H. \square
Proof of Theorem 5. (a) gng^{-1} = ngg^{-1} = n. (b) If [G : H] = 2, then gH = Hg for all g (Cosets, Exercise 13); apply Theorem 2. (c) Z(G) \le G (Subgroups, Example 3). If H \le Z(G), z \in H and g \in G, then gzg^{-1} = zgg^{-1} = z \in H. (d) The intersection is a subgroup (Subgroups, Theorem 6), and if x lies in every N_i, so does gxg^{-1}. \square
Proof of Theorem 6. For n \in N and h \in H, nh = h(h^{-1}nh) and hn = (hnh^{-1})h, with h^{-1}nh, hnh^{-1} \in N; so NH \subseteq HN and HN \subseteq NH. Next, e = ee \in NH. For n_1h_1, n_2h_2 \in NH, write h_1n_2 = n'h_1 with n' = h_1n_2h_1^{-1} \in N; then (n_1h_1)(n_2h_2) = (n_1n')(h_1h_2) \in NH. And (nh)^{-1} = h^{-1}n^{-1} \in HN = NH. So NH \le G. If also H \trianglelefteq G, then g(nh)g^{-1} = (gng^{-1})(ghg^{-1}) \in NH. \square
Proof of Theorem 7. The set product is associative, because the operation of G is. If N \trianglelefteq G, then Nb = bN by Theorem 2, and NN = N, since N is closed and n = en. So
Conversely, suppose (aN)(bN) = abN for all a, b, and let a \in G, n \in N. Then ana^{-1} = (an)(a^{-1}e) \in (aN)(a^{-1}N) = aa^{-1}N = N. So aNa^{-1} \subseteq N. \square
Worked examples
The symmetries of a square
D_4 is the symmetry group of a square, with elements e, r, r^2, r^3 and s, rs, r^2s, r^3s, where r is the quarter turn, s a reflection, r^4 = s^2 = e and sr^k = r^{-k}s for all k (Dihedral groups). Decide which of A = \langle r \rangle, B = \{e, r^2\}, C = \{e, s\} and K = \{e, r^2, s, r^2s\} are normal in D_4.
SolutionA has 4 elements, so [D_4 : A] = 2 and A \trianglelefteq D_4 by Theorem 5(b).
K is a subgroup: sr^2 = r^{-2}s = r^2s, so s and r^2 commute, (r^2s)^2 = r^2sr^2s = r^4s^2 = e, and r^2 \cdot s = r^2s; the set is closed, hence a subgroup by the finite test (Subgroups, Theorem 5). It has index 2, so K \trianglelefteq D_4.
For B, use Theorem 3. Every element is r^k or r^ks, so S = \{r, s\} generates D_4, and T = \{r^2\} generates B, which is finite. Now r r^2 r^{-1} = r^2 and s r^2 s^{-1} = sr^2s = r^{-2}s^2 = r^2, both in B. So B \trianglelefteq D_4.
For C:
So C is not normal. Geometrically, rsr^{-1} is the reflection in the axis of s turned a quarter turn, a different mirror line.
Notice that C \le K with [K : C] = 2, so C \trianglelefteq K, and K \trianglelefteq D_4, yet C is not normal in D_4. A normal subgroup of a normal subgroup need not be normal.
Triangular matrices
In G = \operatorname{GL}_2(\mathbb{R}), let B be the subgroup of invertible upper triangular matrices and U = \{U_t : t \in \mathbb{R}\}, where U_t = \begin{pmatrix} 1 & t \\ 0 & 1 \end{pmatrix}. Show that U \trianglelefteq B, that neither U nor B is normal in G, and that \operatorname{SL}_2(\mathbb{R}) \trianglelefteq G.
SolutionU_sU_t = U_{s + t}, U_0 = I_2 and U_t^{-1} = U_{-t}, so U \le B. Let g = \begin{pmatrix} a & b \\ 0 & d \end{pmatrix} \in B, so ad \ne 0. Then
So gUg^{-1} \subseteq U for every g \in B, and U \trianglelefteq B.
Now take J = \begin{pmatrix} 0 & 1 \\ 1 & 0 \end{pmatrix}, with J^{-1} = J. Multiplying by J on the left swaps rows and on the right swaps columns, so
which for t \ne 0 is not upper triangular. So JUJ^{-1} \not\subseteq U and JBJ^{-1} \not\subseteq B: neither is normal in G.
Finally, for A \in \operatorname{SL}_2(\mathbb{R}) and g \in G, the product rule for determinants (Properties of determinants: row operations, products, transposes) gives \det(gAg^{-1}) = \det g \cdot \det A \cdot (\det g)^{-1} = 1. So \operatorname{SL}_2(\mathbb{R}) \trianglelefteq G; the same line works for every n.
The quaternion group
Q_8 = \{\pm 1, \pm i, \pm j, \pm k\}, where -1 commutes with everything, (-1)^2 = 1, i^2 = j^2 = k^2 = -1, ij = k, jk = i, ki = j, ji = -k, kj = -i and ik = -j. (These are the products of the matrices \pm I_2, \pm\begin{pmatrix} i & 0 \\ 0 & -i \end{pmatrix}, \pm\begin{pmatrix} 0 & 1 \\ -1 & 0 \end{pmatrix}, \pm\begin{pmatrix} 0 & i \\ i & 0 \end{pmatrix} in \operatorname{GL}_2(\mathbb{C}), so associativity comes from matrix multiplication.) Find every subgroup of Q_8 and show that each is normal, although Q_8 is not abelian.
SolutionQ_8 is not abelian, as ij = k \ne -k = ji. The element -1 has order 2, and \pm i, \pm j, \pm k have order 4, since for instance i^2 = -1 and i^4 = 1. By Lagrange’s theorem (Lagrange's theorem, Theorem 1) a subgroup has order 1, 2, 4 or 8. Order 2: it is \{1, x\} with x of order 2, so it is \{\pm 1\}. Order 4: if it holds an element of order 4, it is generated by it, giving \langle i \rangle = \{\pm 1, \pm i\}, \langle j \rangle or \langle k \rangle; otherwise its three elements other than 1 would all have order 2, but Q_8 has only one. So the subgroups are \{1\}, \{\pm 1\}, \langle i \rangle, \langle j \rangle, \langle k \rangle and Q_8: six in all.
Each of \langle i \rangle, \langle j \rangle, \langle k \rangle has index 2, so is normal by Theorem 5(b). And \{\pm 1\} \le Z(Q_8), since -1 commutes with everything, so it is normal by Theorem 5(c). Every subgroup is normal, yet Q_8 is not abelian: Theorem 5(a) has no converse.
Conjugating permutations
Show that A_n \trianglelefteq S_n for n \ge 2, that the set V of e, (1\ 2)(3\ 4), (1\ 3)(2\ 4) and (1\ 4)(2\ 3) is a normal subgroup of S_4, and that P = \langle (1\ 2\ 3\ 4) \rangle is not normal.
Solution\displaystyle |A_n| = \frac{n!}{2} (Permutation groups and cycle notation, Theorem 8), so [S_n : A_n] = 2 and A_n \trianglelefteq S_n by Theorem 5(b).
V is a subgroup: each element is its own inverse and the product of any two of the three nonidentity elements is the third, as in (1\ 2)(3\ 4)\,(1\ 3)(2\ 4) = (1\ 4)(2\ 3). A product of two disjoint transpositions in S_4 is fixed by which point is paired with 1, so S_4 has exactly three of them, all in V. For \sigma \in S_4, Theorem 1 and the cycle formula give
again a product of two disjoint transpositions, since \sigma is one to one. So \sigma V \sigma^{-1} \subseteq V and V \trianglelefteq S_4.
P consists of e, (1\ 2\ 3\ 4), (1\ 3)(2\ 4) and (1\ 4\ 3\ 2). Conjugating by \sigma = (1\ 2),
so P is not normal, and (1\ 2)P \ne P(1\ 2) by Theorem 2.
Normal subgroups that meet trivially
Let M, N \trianglelefteq G with M \cap N = \{e\}. Prove that mn = nm for all m \in M and n \in N, that MN \trianglelefteq G, and that each element of MN is mn for exactly one pair (m, n).
SolutionProof. Let x = mnm^{-1}n^{-1}. Grouped as m(nm^{-1}n^{-1}), it lies in M, since nm^{-1}n^{-1} \in M by normality of M. Grouped as (mnm^{-1})n^{-1}, it lies in N, since mnm^{-1} \in N. So x \in M \cap N = \{e\}, which says mn = nm. By Theorem 6, MN \trianglelefteq G. If m_1n_1 = m_2n_2, then m_2^{-1}m_1 = n_2n_1^{-1} lies in M \cap N = \{e\}, so m_1 = m_2 and n_1 = n_2. \square
For instance, in D_6 (the hexagon, r^6 = e), M = \{e, r^3\} lies in the center, as sr^3 = r^{-3}s = r^3s, and N = \{e, r^2, r^4, s, r^2s, r^4s\} has index 2. They meet in \{e\}, and |MN| = 2 \cdot 6 = 12, so every element of D_6 is mn in exactly one way. Direct products and finite abelian groups builds on this.
Every normal subgroup of S_4
Find all normal subgroups of S_4.
SolutionA normal subgroup N contains, with each x, every conjugate gxg^{-1}. Two permutations in S_4 of the same cycle type are conjugate: write the cycles of x (fixed points as cycles of length 1) above those of y, lengths matched, and let \pi send each entry to the one below it; the cycle formula gives \pi x \pi^{-1} = y. Conjugates have the same cycle type by the same formula. So N is a union of whole cycle types. The types and their sizes are
| type | number |
|---|---|
| e | 1 |
| (a\ b) | 6 |
| (a\ b)(c\ d) | 3 |
| (a\ b\ c) | 8 |
| (a\ b\ c\ d) | 6 |
and 1 + 6 + 3 + 8 + 6 = 24. Since |N| divides 24, |N| - 1 is one of 0, 1, 2, 3, 5, 7, 11, 23, and it is a sum of some of the sizes 6, 3, 8, 6. Running through the sixteen subsets, only 0, 3, 11 = 3 + 8 and 23 occur, each in one way. They give \{e\}; V; V with the eight 3-cycles, which is A_4, as these twelve are the even permutations; and S_4. All four are normal (Example 4), so these are all of them.
Boundaries
- Not transitive. C \trianglelefteq K \trianglelefteq D_4 but C is not normal in D_4 (Example 1).
- Every g, not one. For a single g, gNg^{-1} \subseteq N can hold with gNg^{-1} \ne N when N is infinite (Exercise 17); the proof of Theorem 2 used g^{-1} too. For finite N, gNg^{-1} \subseteq N already gives gNg^{-1} = N (Theorem 1).
- Normal is not commuting. gN = Ng is an equality of sets: gn = n'g with n' possibly different from n. In Example 1, A \trianglelefteq D_4, yet sr = r^3s \ne rs.
- Abelian is enough, not needed. Every subgroup of Q_8 is normal (Example 3).
- Simple groups. G \ne \{e\} is simple if its only normal subgroups are \{e\} and G. Groups of prime order are (Lagrange's theorem, Corollary 3); S_4 is not (Example 6).
Where this goes next
Theorem 7 starts Quotient groups: for N \trianglelefteq G the cosets form a group G/N. Homomorphisms and kernels shows that kernels are normal, and the natural map G \to G/N makes every normal subgroup a kernel; the The isomorphism theorems use Theorem 6 and Exercise 13. Example 5 becomes the internal direct product. Conjugation is a group action whose orbits are the conjugacy classes, and Corollary 4 is how the The Sylow theorems show that a Sylow subgroup with no others of its order is normal.
Exercises
Deciding normality
- In D_4 (Example 1), decide whether \{e, rs\} and \{e, r^2, rs, r^3s\} are normal subgroups.
- Show that e, (1\ 2), (3\ 4) and (1\ 2)(3\ 4) form a subgroup H of S_4, and decide whether it is normal.
- In \operatorname{GL}_2(\mathbb{R}), decide whether the invertible diagonal matrices form a normal subgroup, and whether the scalar matrices cI_2 (c \ne 0) do.
- In Q_8 (Example 3), find the set \{gig^{-1} : g \in Q_8\}.
- Show that \langle (1\ 2\ 3) \rangle is not normal in A_4.
- In S_5, compute \sigma\pi\sigma^{-1} for \sigma = (1\ 2\ 3\ 4\ 5) and \pi = (1\ 3\ 5)(2\ 4).
- Find every normal subgroup of D_4.
- Show that V (Example 4) is the only subgroup of order 4 in A_4, and use Corollary 4 to conclude V \trianglelefteq A_4.
Multiplying cosets
- In D_4 with N = \{e, r^2\}, list rN and sN, compute the set (rN)(sN), and compare it with rsN.
- In D_4 with C = \{e, s\}, compute the set (rC)(r^3C) and show that it is not a coset of C.
- With V from Example 4 and H = \{e, (1\ 2)\}, explain why VH \le S_4, find |VH|, and list its elements.
- Kofi, a weaver, can hang a finished square rug in eight ways, related by the elements of D_4. Amara records each rehanging only as “turn” (in \langle r \rangle) or “flip” (in \langle r \rangle s) and claims that the records of two rehangings fix the record of the two done in a row. Kofi tries the same with the two-element cosets of C = \{e, s\}. Who is right?
Proofs
- Let N \trianglelefteq G and H \le G. Prove that N \cap H \trianglelefteq H.
- Prove that a normal subgroup of order 2 lies in Z(G).
- Let N \le G contain every commutator [a, b] = aba^{-1}b^{-1}. Prove that N \trianglelefteq G.
- Let H \le G. Prove that \bigcap_{g \in G} gHg^{-1} is a normal subgroup of G that lies in H and contains every normal subgroup of G lying in H.
- In \operatorname{GL}_2(\mathbb{R}), let H = \{U_n : n \in \mathbb{Z}\} (Example 2) and g = \begin{pmatrix} 2 & 0 \\ 0 & 1 \end{pmatrix}. Show that gHg^{-1} is a proper subset of H, and explain why this does not contradict Theorem 2.
- Let N \trianglelefteq G be finite and cyclic. Prove that every subgroup of N is normal in G.
Answers
- r(rs)r^{-1} = r^2sr^3 = r^2r^{-3}s = r^3s \notin \{e, rs\}: not normal. For K' = \{e, r^2, rs, r^3s\}: (rs)^2 = rr^{-1}s^2 = e, (r^3s)^2 = e, rs \cdot r^3s = rr^{-3} = r^2, r^2 \cdot rs = r^3s, r^2 \cdot r^3s = rs; closed, so a subgroup by the finite test, of index 2, hence normal by Theorem 5(b).
- (1\ 2) and (3\ 4) are disjoint, so they commute (Permutation groups and cycle notation, Theorem 2); each of the three nonidentity elements squares to e and the product of two is the third, so H is closed. Not normal: (2\ 3)(1\ 2)(2\ 3)^{-1} = (1\ 3) \notin H by the cycle formula.
- Diagonal: not normal. With X = \begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix} and A = \begin{pmatrix} 1 & 0 \\ 0 & 2 \end{pmatrix}, XAX^{-1} = \begin{pmatrix} 1 & 1 \\ 0 & 2 \end{pmatrix}, not diagonal. Scalar: normal, since cI_2 commutes with every matrix, so the scalar matrices form a subgroup of the center (Theorem 5(c)).
- \{i, -i\}. Elements of \langle i \rangle commute with i. jij^{-1} = ji(-j) = -(ji)j = kj = -i, and kik^{-1} = -(ki)k = -jk = -i; -j and -k give the same as j and k, as -1 is central.
- \langle (1\ 2\ 3) \rangle = \{e, (1\ 2\ 3), (1\ 3\ 2)\}. With \sigma = (1\ 2)(3\ 4) \in A_4, \sigma(1\ 2\ 3)\sigma^{-1} = (2\ 1\ 4) = (1\ 4\ 2), not in it.
- Rename each entry of \pi by \sigma (the cycle formula): \sigma\pi\sigma^{-1} = (2\ 4\ 1)(3\ 5), that is, (1\ 2\ 4)(3\ 5).
- Orders are 1, 2, 4, 8. Order 2: \{e, r^2\} and the four \{e, r^ks\}. Order 4: \langle r \rangle (the only elements of order 4 are r, r^3), or e with three elements of order 2, two of them reflections r^as, r^bs whose product r^{a - b} must be r^2: K or K'. Normal: \{e\}, \{e, r^2\}, \langle r \rangle, K, K', D_4, six of the ten. The four \{e, r^ks\} are not, as r(r^ks)r^{-1} = r^{k + 2}s.
- A subgroup of order 4 has element orders dividing 4 (Lagrange, Corollary 2), so it has no 3-cycle. The elements of A_4 that are not 3-cycles are exactly the four of V. So V is the only one, and V \trianglelefteq A_4.
- rN = \{r, r^3\}, sN = \{s, r^2s\}. Products: rs, r^3s, r^3s, r^5s = rs, so (rN)(sN) = \{rs, r^3s\} = rsN, since rsr^2 = r^{-1}s = r^3s.
- rC = \{r, rs\}, r^3C = \{r^3, r^3s\}. Products: rr^3 = e, rr^3s = s, rsr^3 = r^{-2}s = r^2s, rsr^3s = r^{-2} = r^2. The set \{e, s, r^2s, r^2\} has four elements, but cosets of C have two.
- V \trianglelefteq S_4, so VH \le S_4 by Theorem 6. V \cap H = \{e\}, so |VH| = 4 \cdot 2 / 1 = 8 (Lagrange's theorem, Exercise 15). VH is V with (1\ 2), (3\ 4), (1\ 4\ 2\ 3) and (1\ 3\ 2\ 4).
- Amara. \langle r \rangle has index 2, so it is normal, and by Theorem 7 the coset of a product depends only on the cosets of the factors: turn then turn is a turn, turn and flip in either order is a flip, flip then flip is a turn. Kofi is wrong: by Exercise 10, r and rs share a coset of C, and so do r^3 and r^3s, but rr^3 = e \in C while rs \cdot r^3 = r^2s \in r^2C.
- Proof. N \cap H \le H (Subgroups, Theorem 6). For h \in H and x \in N \cap H, hxh^{-1} \in N as N is normal, and hxh^{-1} \in H as H is closed. \square
- Proof. Let N = \{e, n\} and g \in G. Then gng^{-1} \in N, and gng^{-1} \ne e, as otherwise n = g^{-1}g = e. So gng^{-1} = n, that is gn = ng, and N \subseteq Z(G). \square
- Proof. For g \in G and n \in N, gng^{-1} = (gng^{-1}n^{-1})n = [g, n]\,n, a product of two elements of N. \square
- Proof. Each gHg^{-1} is a subgroup (Theorem 1), so the intersection C is one (Subgroups, Theorem 6), and C \subseteq eHe^{-1} = H. Let x \in C and a \in G. For each g, x \in (a^{-1}g)H(a^{-1}g)^{-1} = a^{-1}(gHg^{-1})a, so axa^{-1} \in gHg^{-1}; hence axa^{-1} \in C. If N \trianglelefteq G and N \subseteq H, then N = gNg^{-1} \subseteq gHg^{-1} for every g, so N \subseteq C. \square
- U_mU_n = U_{m + n} and U_n^{-1} = U_{-n}, so H is a subgroup. By Example 2 with a = 2, d = 1, gU_ng^{-1} = U_{2n}, so gHg^{-1} = \{U_{2n} : n \in \mathbb{Z}\}, which misses U_1. No contradiction: Theorem 2(i) asks xHx^{-1} \subseteq H for every x, and x = g^{-1} gives g^{-1}U_1g = U_{1/2} \notin H. So H is not normal, and for infinite H one g does not force equality.
- Proof. Let K \le N and g \in G. Then gKg^{-1} \subseteq gNg^{-1} = N, and gKg^{-1} is a subgroup of order |K| (Theorem 1), so a subgroup of N of order |K|. A finite cyclic group has exactly one subgroup of each order dividing its order (Cyclic groups, Theorem 10), so gKg^{-1} = K. \square