Subgroups

The even integers form a group inside the integers, and the matrices of determinant 1 a group inside the invertible matrices. When is a subset of a group a group in its own right, and how can we tell without checking every axiom again?

Where you are 6 in Groups

Why this exists

Groups seldom come alone. Inside (\mathbb{Z}, +) sit the even integers; inside \operatorname{GL}_n(\mathbb{R}) sits \operatorname{SL}_n(\mathbb{R}) (Integers mod n, symmetries and matrices as groups); inside \mathbb{R}^n under addition sits the kernel of every matrix (Kernel). Each is a subset that is a group under the operation of the larger group. Checking every axiom anew is wasted work: associativity is inherited, and the identity and inverses are forced to be those of the larger group. What remains is a short test. Much of the later course (Lagrange’s theorem, quotients, kernels) is about subgroups.

Definition

Let G be a group. A subset H \subseteq G is closed under the operation if ab \in H whenever a, b \in H; then the operation of G, restricted to pairs from H, is a binary operation on H.

A subgroup of G is a subset H that is closed under the operation of G and is a group under that restricted operation. We write H \le G, and H < G when also H \ne G, a proper subgroup. Every group has the subgroups G and \{e\}, the trivial subgroup. If K \le H and H \le G, then K \le G, since all three use the same operation.

In an additive group the operation is +, the identity 0 and the inverse -a; for a positive integer n, n\mathbb{Z} \le \mathbb{Z}.

Intuition

A subgroup is a part of the group that computing cannot leave: combine two of its elements, or invert one, and the result is inside again. Associativity costs nothing; the real questions are closure and inverses. Ordered by inclusion, the subgroups of a finite group form a diagram in which each sits above the ones it contains (Example 5).

Results

Throughout, G is a group with identity e, and H is a subset of G.

Theorem 1 (inherited identity and inverses). If H \le G, then the identity of H is e, and for each h \in H the inverse of h in H is its inverse h^{-1} in G.

Theorem 2 (subgroup test). H \le G if and only if (a) e \in H; (b) ab \in H for all a, b \in H; and (c) a^{-1} \in H for all a \in H.

Theorem 3 (one-step test). H \le G if and only if H \ne \varnothing and ab^{-1} \in H for all a, b \in H. In additive notation: H \ne \varnothing and a - b \in H for all a, b \in H.

Lemma 4 (powers stay inside). If H \le G and a \in H, then a^n \in H for every integer n (in additive notation, na \in H).

Theorem 5 (finite test). If H is finite, nonempty and closed under the operation, then H \le G.

Theorem 6 (intersections). If \{H_i : i \in I\} is a nonempty collection of subgroups of G, then \bigcap_{i \in I} H_i \le G.

Why it is true

Proof of Theorem 1. Let e_H be the identity of H. In H, e_H e_H = e_H, and in G, e_H e = e_H. So e_H e_H = e_H e, and left cancellation in G (Binary operations and the group axioms, Corollary 5) gives e_H = e. If h' \in H is the inverse of h in H, then hh' = h'h = e_H = e, so h' is an inverse of h in G, and inverses in G are unique (Theorem 2 there). \square

Proof of Theorem 2. If H \le G, then (b) is part of the definition, and (a) and (c) hold by Theorem 1, since the group H contains its own identity and inverses. Conversely, assume (a), (b) and (c). By (b) the restricted operation is a binary operation on H. It is associative, because (ab)c = a(bc) holds for all elements of G. By (a), e \in H, and ea = ae = a for a \in H. By (c) each a \in H has an inverse in H. So H is a group. \square

Proof of Theorem 3. If H \le G, then e \in H, so H \ne \varnothing, and for a, b \in H Theorem 2 gives b^{-1} \in H and then ab^{-1} \in H. Conversely, pick a \in H. Then e = aa^{-1} \in H. For b \in H, b^{-1} = eb^{-1} \in H. For a, b \in H, then, b^{-1} \in H, so ab = a(b^{-1})^{-1} \in H, using (b^{-1})^{-1} = b (Binary operations and the group axioms, Theorem 3). Theorem 2 applies. \square

Proof of Lemma 4. a^0 = e \in H by Theorem 1. If a^n \in H for some n \ge 1, then a^{n + 1} = a^n a \in H by closure (brackets may be dropped, Theorem 7 of Binary operations and the group axioms), so a^n \in H for all n \ge 1 by induction. As a^{-1} \in H, the same argument gives a^{-n} = (a^{-1})^n \in H. \square

Proof of Theorem 5. Let a \in H. By closure and induction, every power a^n with n \ge 1 lies in H. As H is finite, these powers repeat: a^i = a^j for some 1 \le i < j. Put k = j - i \ge 1. Then a^i a^k = a^j = a^i = a^i e, and cancellation gives a^k = e. If k = 1, then a = e and a^{-1} = a \in H. If k \ge 2, then a \, a^{k - 1} = a^{k - 1} a = a^k = e, so a^{-1} = a^{k - 1} \in H. Thus H is closed under inverses, and e = aa^{-1} \in H; Theorem 2 applies. \square

Proof of Theorem 6. Let K = \bigcap_{i \in I} H_i. Each H_i contains e, so K \ne \varnothing. If a, b \in K, then a, b \in H_i for every i, so ab^{-1} \in H_i for every i by Theorem 3; that is, ab^{-1} \in K. Theorem 3 applies. \square

Worked examples

Example 1

Four subsets of the nonzero complex numbers

Which of these are subgroups of \mathbb{C}^\times under multiplication? (a) The circle S^1 = \{z : |z| = 1\}; (b) the positive real numbers \mathbb{R}_{>0}; (c) D = \{z : 0 < |z| \le 1\}; (d) C = \{z : |z| = 2\}.

Solution

We use Theorem 2 and the fact that moduli multiply, |zw| = |z||w| (Polar form of a complex number), so that |z^{-1}| = 1/|z|.

(a) |1| = 1, so 1 \in S^1. If |z| = |w| = 1, then |zw| = 1 and |z^{-1}| = 1. So S^1 \le \mathbb{C}^\times.

(b) 1 > 0; a product of positive reals is positive, and so is the reciprocal of one. So \mathbb{R}_{>0} \le \mathbb{C}^\times.

(c) D contains 1 and is closed, since |zw| = |z||w| \le 1. But \displaystyle \frac{1}{2} \in D while its inverse 2 \notin D: condition (c) fails, and D is not a subgroup.

(d) 1 \notin C, and C is not closed: 2 \cdot 2 = 4 has modulus 4. Not a subgroup.

By Theorem 6, S^1 \cap \mathbb{R}_{>0} = \{1\} is a subgroup, the trivial one.

Example 2

Orthogonal matrices

Let O = \{A \in \operatorname{GL}_2(\mathbb{R}) : A^{\mathsf T} A = I_2\}. Prove that O \le \operatorname{GL}_2(\mathbb{R}). Then, for \displaystyle A = \frac{1}{5}\begin{pmatrix} 3 & -4 \\ 4 & 3 \end{pmatrix} and B = \begin{pmatrix} 0 & 1 \\ 1 & 0 \end{pmatrix}, find AB^{-1} and confirm that it lies in O.

Solution

For a square matrix, A^{\mathsf T} A = I_2 makes A invertible with A^{-1} = A^{\mathsf T} (The invertible matrix theorem), so also AA^{\mathsf T} = I_2. Now I_2 \in O, so O \ne \varnothing. Let A, B \in O. Then B^{-1} = B^{\mathsf T}, and with (XY)^{\mathsf T} = Y^{\mathsf T} X^{\mathsf T} and (X^{\mathsf T})^{\mathsf T} = X (The transpose and symmetric matrices),

\begin{aligned} (AB^{-1})^{\mathsf T}(AB^{-1}) &= (AB^{\mathsf T})^{\mathsf T}(AB^{\mathsf T}) \\ &= BA^{\mathsf T}AB^{\mathsf T} = BB^{\mathsf T} = I_2. \end{aligned}

So AB^{-1} \in O, and O \le \operatorname{GL}_2(\mathbb{R}) by Theorem 3.

For the given matrices, \displaystyle A^{\mathsf T}A = \frac{1}{25}\begin{pmatrix} 25 & 0 \\ 0 & 25 \end{pmatrix} = I_2 and B^{\mathsf T}B = I_2, with B^{-1} = B. Multiplying by B on the right swaps the columns, so

AB^{-1} = AB = \frac{1}{5}\begin{pmatrix} -4 & 3 \\ 3 & 4 \end{pmatrix}.

Its columns \displaystyle \frac{1}{5}(-4, 3) and \displaystyle \frac{1}{5}(3, 4) have length 1 and dot product \displaystyle \frac{1}{25}(-12 + 12) = 0, which says exactly that (AB)^{\mathsf T}(AB) = I_2. (A is a rotation, Finding the matrix of a transformation, and \det(AB) = -1.)

Example 3

The center of a group

The center of G is Z(G) = \{z \in G : zg = gz \text{ for all } g \in G\}. Prove that Z(G) \le G, that Z(G) is abelian, and that Z(G) = G exactly when G is abelian. Then find Z(\operatorname{GL}_2(\mathbb{R})).

Solution

We check Theorem 2. (a) eg = g = ge for all g, so e \in Z(G). (b) Let z, w \in Z(G) and g \in G. By associativity, then wg = gw, then zg = gz,

\begin{aligned} (zw)g &= z(wg) = z(gw) \\ &= (zg)w = (gz)w = g(zw), \end{aligned}

so zw \in Z(G). (c) Let z \in Z(G) and g \in G. Multiplying zg = gz by z^{-1} on both the left and the right gives z^{-1}(zg)z^{-1} = z^{-1}(gz)z^{-1}, that is, gz^{-1} = z^{-1}g. So z^{-1} \in Z(G), and Z(G) \le G. Two elements of Z(G) commute, since each commutes with everything, so Z(G) is abelian. And Z(G) = G says that every element commutes with every element, which is the definition of abelian.

For G = \operatorname{GL}_2(\mathbb{R}), let M = \begin{pmatrix} a & b \\ c & d \end{pmatrix} lie in Z(G). It commutes with X = \begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix} and Y = \begin{pmatrix} 1 & 0 \\ 1 & 1 \end{pmatrix}, both in G. Now

MX = \begin{pmatrix} a & a + b \\ c & c + d \end{pmatrix}, \qquad XM = \begin{pmatrix} a + c & b + d \\ c & d \end{pmatrix},

and comparing entries gives c = 0 and a = d. Likewise MY = \begin{pmatrix} a + b & b \\ c + d & d \end{pmatrix} and YM = \begin{pmatrix} a & b \\ a + c & b + d \end{pmatrix} give b = 0. So M = aI_2, with a \ne 0 as M is invertible. Conversely (aI_2)N = aN = N(aI_2) for every N. Hence Z(\operatorname{GL}_2(\mathbb{R})) = \{aI_2 : a \ne 0\}, the nonzero scalar matrices.

Example 4

Stalls of a roundhouse

At a railroad roundhouse, 18 stalls stand around a rotating bridge at equal angles, numbered 0 to 17 counterclockwise, and the bridge starts aligned with stall 0. Hiro’s control panel turns the bridge 12 or 8 stalls at a time, in either direction. Which stalls can Leah’s locomotive, parked on the bridge, be sent into? How does Hiro reach stall 10, and can he reach stall 7?

Solution

Writing k for [k]_{18}, a turn by k stalls counterclockwise is the element k of \mathbb{Z}_{18}, and turns combine by addition. After a net turns of 12 and b of 8 (negative for clockwise) the bridge faces stall 12a + 8b, so the reachable stalls form

R = \{12a + 8b : a, b \in \mathbb{Z}\} \subseteq \mathbb{Z}_{18}.

R contains 0, and the difference (12a + 8b) - (12a' + 8b') is 12(a - a') + 8(b - b') \in R, so R \le \mathbb{Z}_{18} by Theorem 3. It contains 12 - 8 = 4, so by Lemma 4 it contains 5 \cdot 4 = 20 = 2, and then every multiple of 2: all nine even classes 0, 2, \dots, 16. It contains no odd class: 12a + 8b is an even integer, and since 18 is even, every integer in the class of an even integer is even. So R is exactly the set of even classes, and Leah’s locomotive can be sent into the nine even stalls and no others.

For stall 10, one turn of 12 and two of 8 give 12 + 16 = 28 \equiv 10. Stall 7 is odd, so it cannot be reached. The smallest step is 2 = \gcd(12, 8, 18) = 5 \cdot 12 - 5 \cdot 8 - 18, a combination of the kind Bezout's identity and the extended Euclidean algorithm finds.

Example 5

Every subgroup of the units mod 20

Find all subgroups of G = \mathbb{Z}_{20}^\times = \{1, 3, 7, 9, 11, 13, 17, 19\} and order them by inclusion.

Solution

For a \in G let \langle a \rangle = \{a, a^2, a^3, \dots\}, the set of positive powers of a. It is finite, nonempty and closed, since a^m a^n = a^{m + n}, so it is a subgroup by Theorem 5; and every subgroup containing a contains \langle a \rangle by Lemma 4. (For an arbitrary group, Cyclic groups defines \langle a \rangle with all integer powers.) Mod 20, 1^2, 9^2, 11^2 and 19^2 are 1, while 3^2, 7^2, 13^2 and 17^2 are 9. With 3^3 = 27 \equiv 7, 13^3 = 2197 \equiv 17 and 3^4 = 13^4 = 81 \equiv 1, this gives \langle 3 \rangle = \langle 7 \rangle = \{1, 3, 7, 9\}, \langle 13 \rangle = \langle 17 \rangle = \{1, 9, 13, 17\}, \langle 9 \rangle = \{1, 9\}, \langle 11 \rangle = \{1, 11\} and \langle 19 \rangle = \{1, 19\}.

A subgroup H is the union of the \langle a \rangle with a \in H. Suppose first that H contains none of 3, 7, 13, 17, so H \subseteq K = \{1, 9, 11, 19\}. The products 9 \cdot 11 = 99 \equiv 19, 9 \cdot 19 = 171 \equiv 11 and 11 \cdot 19 = 209 \equiv 9 show that any two of 9, 11, 19 produce the third and that K is closed. So H is \{1\}, \langle 9 \rangle, \langle 11 \rangle, \langle 19 \rangle or K, each a subgroup by Theorem 5.

Otherwise H contains \langle 3 \rangle or \langle 13 \rangle. If it contains both, it contains 3 \cdot 13 = 39 \equiv 19 and 3 \cdot 17 = 51 \equiv 11, so H = G. If H \supseteq \langle 3 \rangle also contains 11 or 19, it contains 3 \cdot 11 = 33 \equiv 13 or 3 \cdot 19 = 57 \equiv 17, so both. Likewise 13 \cdot 11 = 143 \equiv 3 and 13 \cdot 19 = 247 \equiv 7. So the remaining subgroups are \langle 3 \rangle, \langle 13 \rangle and G: eight in all.

G
\langle 3 \rangle
\langle 13 \rangle
K
\langle 9 \rangle
\langle 11 \rangle
\langle 19 \rangle
\{1\}
Figure 1The eight subgroups of \mathbb{Z}_{20}^\times (Example 5), each joined to the subgroups just above it that contain it.

The three subgroups of order 4 meet pairwise in \langle 9 \rangle (Theorem 6). Every order here, 1, 2, 4 or 8, divides 8.

Boundaries

  • The operation must be the same. \mathbb{R}_{>0} is a subset of \mathbb{R} and a group under multiplication, but it is not a subgroup of (\mathbb{R}, +): it lacks 0, and 1 - 2 \notin \mathbb{R}_{>0}.
  • Closure alone suffices only for finite sets. The positive integers form a nonempty subset of (\mathbb{Z}, +) closed under addition, yet -1 is missing. The proof of Theorem 5 used finiteness to find a^k = e.
  • Nonempty is needed. The empty set meets “ab^{-1} \in H for all a, b \in H” vacuously, but it is not a group.
  • Unions fail. \langle 11 \rangle \cup \langle 19 \rangle in \mathbb{Z}_{20}^\times is not closed, since 11 \cdot 19 \equiv 9; see Exercise 11.

Where this goes next

Groups that equal a single \langle a \rangle are the Cyclic groups. A subgroup H cuts G into Cosets gH of equal size, which gives Lagrange's theorem: |H| divides |G| in a finite group. The subgroups with gHg^{-1} = H for every g (compare Exercise 14) are the Normal subgroups. Kernels and images of Homomorphisms and kernels are subgroups, and so are the stabilizers of a group action.

Exercises

Testing subsets

  1. Which of these are subgroups of \mathbb{Z}_{24}? (a) \{0, 6, 12, 18\}; (b) \{0, 9, 18\}; (c) \{0, 8, 16\}; (d) \{1, 13\}.
  2. Which of these are subgroups of \mathbb{R}^\times? (a) \{2^n : n \in \mathbb{Z}\}; (b) \{x \in \mathbb{R}^\times : x^2 \in \mathbb{Q}\}; (c) \{x \in \mathbb{R} : |x| \ge 1\}; (d) \{1, -1\}.
  3. Which of these are subgroups of \operatorname{GL}_2(\mathbb{R})? (a) The invertible diagonal matrices; (b) the matrices \begin{pmatrix} 1 & b \\ 0 & 1 \end{pmatrix} with b \in \mathbb{R}; (c) the invertible symmetric matrices; (d) the matrices with \det A = \pm 1.
  4. Decide whether \{1, 7, 11\} and \{1, 7, 8\} are subgroups of \mathbb{Z}_{19}^\times.

Finding subgroups

  1. Find all subgroups of \mathbb{Z}_{14}^\times = \{1, 3, 5, 9, 11, 13\}.
  2. Find the smallest subgroup of \mathbb{Z}_{15}^\times that contains 4 and 11.
  3. In \mathbb{Z}_{24}, let H be the set of multiples of 4 and K the set of multiples of 6. Find H \cap K, and decide whether H \cup K is a subgroup.
  4. Let T be the group of real matrices \begin{pmatrix} a & b \\ 0 & d \end{pmatrix} with ad \ne 0. Find Z(T).
  5. Leah’s second roundhouse has 16 stalls, and its bridge turns 6 or 10 stalls at a time, in either direction, starting at stall 0. Which stalls can it reach, and how does it reach stall 2?
  6. How many subgroups with exactly two elements does \operatorname{GL}_2(\mathbb{Z}_3) have?

Proofs

  1. Let H, K \le G. Prove that H \cup K \le G if and only if H \subseteq K or K \subseteq H.
  2. For x \in G, the centralizer of x is C_G(x) = \{g \in G : gx = xg\}. Prove that C_G(x) \le G and that Z(G) = \bigcap_{x \in G} C_G(x).
  3. Let G be abelian and n \ge 1. Prove that \{g \in G : g^n = e\} \le G. Then show that in \operatorname{GL}_2(\mathbb{R}) the set of g with g^2 = I_2 is not closed, using A = \begin{pmatrix} 1 & 0 \\ 0 & -1 \end{pmatrix} and B = \begin{pmatrix} 0 & 1 \\ 1 & 0 \end{pmatrix}.
  4. Let H \le G and g \in G. Prove that gHg^{-1} = \{ghg^{-1} : h \in H\} is a subgroup of G.
  5. For H, K \le G let HK = \{hk : h \in H, k \in K\}. Prove that HK \le G when G is abelian. With A and B as in Exercise 13, show that H = \{I_2, A\} and K = \{I_2, B\} are subgroups of \operatorname{GL}_2(\mathbb{R}) but HK is not.
  6. Prove that \displaystyle \{a + b\sqrt{2} : a, b \in \mathbb{Q}\} \setminus \{0\} is a subgroup of \mathbb{R}^\times, and find the inverse of \displaystyle 3 + 2\sqrt{2} in it.
  7. Let H be a nonempty subset of G, closed under the operation, such that each h \in H has h^n = e for some n \ge 1. Prove that H \le G, even if H is infinite.
Answers
  1. (a) Yes: as 6 \mid 24, a sum of multiples of 6 reduces to a multiple of 6, so the set is closed and Theorem 5 applies. (b) No: 9 + 18 = 27 \equiv 3. (c) Yes: 8 + 8 = 16, 8 + 16 \equiv 0, 16 + 16 = 32 \equiv 8, and Theorem 5 applies. (d) No: 0 \notin \{1, 13\} (and 1 + 1 = 2 is outside).
  2. By the one-step test. (a) Yes: 2^m(2^n)^{-1} = 2^{m - n}. (b) Yes: it contains 1, and (x/y)^2 = x^2/y^2 \in \mathbb{Q}. (c) No: 2 is in it, but \displaystyle 2^{-1} = \frac{1}{2} is not. (d) Yes: a quotient of two of \pm 1 is \pm 1.
  3. By the one-step test. (a) Yes: \operatorname{diag}(a, d)\operatorname{diag}(a', d')^{-1} = \operatorname{diag}(a/a', d/d'). (b) Yes: \begin{pmatrix} 1 & b \\ 0 & 1 \end{pmatrix}\begin{pmatrix} 1 & b' \\ 0 & 1 \end{pmatrix}^{-1} = \begin{pmatrix} 1 & b - b' \\ 0 & 1 \end{pmatrix}. (c) No: \begin{pmatrix} 1 & 1 \\ 1 & 0 \end{pmatrix}\begin{pmatrix} 1 & 0 \\ 0 & 2 \end{pmatrix} = \begin{pmatrix} 1 & 2 \\ 1 & 0 \end{pmatrix} is not symmetric. (d) Yes: \det(AB^{-1}) = \det A / \det B = \pm 1.
  4. 7 \cdot 7 = 49 \equiv 11, 7 \cdot 11 = 77 \equiv 1 and 11 \cdot 11 = 121 \equiv 7, so \{1, 7, 11\} is closed and is a subgroup by Theorem 5. But 7 \cdot 8 = 56 \equiv 18, so \{1, 7, 8\} is not closed and is not a subgroup.
  5. Four: \{1\}, \{1, 13\}, \{1, 9, 11\} and \mathbb{Z}_{14}^\times. The powers of 3 are 3, 9, 13, 11, 5, 1 and those of 5 are 5, 11, 13, 9, 3, 1, so a subgroup containing 3 or 5 is everything. \langle 9 \rangle = \langle 11 \rangle = \{1, 9, 11\} (as 9^2 = 81 \equiv 11, 9 \cdot 11 = 99 \equiv 1) and \langle 13 \rangle = \{1, 13\}. A subgroup containing 13 and one of 9, 11 contains 9 \cdot 13 = 117 \equiv 5, so is everything.
  6. \{1, 4, 11, 14\}. It must contain 4^2 = 16 \equiv 1, 11^2 = 121 \equiv 1 and 4 \cdot 11 = 44 \equiv 14, and this set is closed (4 \cdot 14 = 56 \equiv 11, 11 \cdot 14 = 154 \equiv 4, 14^2 = 196 \equiv 1), so it is a subgroup by Theorem 5.
  7. H = \{0, 4, 8, 12, 16, 20\} and K = \{0, 6, 12, 18\}, so H \cap K = \{0, 12\}, a subgroup by Theorem 6. H \cup K is not a subgroup: 4 + 6 = 10 is in neither.
  8. Z(T) = \{aI_2 : a \ne 0\}. Let M = \begin{pmatrix} a & b \\ 0 & d \end{pmatrix}. With X = \begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix}, the upper right entries of MX and XM are a + b and b + d, so a = d. With D = \operatorname{diag}(2, 1), those of MD and DM are b and 2b, so b = 0. Scalar matrices commute with every matrix.
  9. The eight even stalls 0, 2, \dots, 14. As in Example 4, the set of 6a + 10b is a subgroup of \mathbb{Z}_{16}; it holds only even classes because 16 is even, and it contains 2 \cdot 6 - 10 = 2, hence every even class by Lemma 4. For stall 2: two turns of 6 counterclockwise and one of 10 clockwise.
  10. 13. A set \{e, g\} with g \ne e is closed only if g^2 = e (g^2 = g would force g = e), and then it is a subgroup by Theorem 5; so we count the A \ne I_2 with A^2 = I_2. Multiplying out shows A^2 - (\operatorname{tr} A)A + (\det A)I_2 = 0 for 2 \times 2 matrices, so A^2 = I_2 gives (\operatorname{tr} A)A = (1 + \det A)I_2. If \operatorname{tr} A \ne 0, A = \lambda I_2 with \lambda^2 = 1, so A = -I_2. Otherwise \operatorname{tr} A = 0 and \det A = -1, which conversely give A^2 = I_2: A = \begin{pmatrix} a & b \\ c & -a \end{pmatrix} with a^2 + bc = 1 in \mathbb{Z}_3. For a = 0, bc = 1 has 2 solutions; for a = 1 and for a = 2, bc = 0 has 5 each. That makes 12, and with -I_2, 13.
  11. Proof. If H \subseteq K, then H \cup K = K \le G, and similarly if K \subseteq H. Conversely, let H \cup K \le G and suppose neither contains the other: pick h \in H \setminus K and k \in K \setminus H. Then hk \in H \cup K. If hk \in H, then k = h^{-1}(hk) \in H; if hk \in K, then h = (hk)k^{-1} \in K. Both contradict the choice of h and k (Proof by contradiction). \square
  12. Proof. ex = x = xe, so e \in C_G(x). If g, h \in C_G(x), then (gh)x = g(hx) = g(xh) and g(xh) = (gx)h = (xg)h = x(gh). If gx = xg, multiplying by g^{-1} on both sides gives xg^{-1} = g^{-1}x. So C_G(x) \le G by Theorem 2. And z \in Z(G) exactly when zx = xz for every x, that is, when z \in C_G(x) for every x \in G; Theorem 6 then reproves Z(G) \le G. \square
  13. Proof. Let H = \{g : g^n = e\}; e^n = e, so H \ne \varnothing. In an abelian group (ab)^n = a^n b^n, by induction: (ab)^{n + 1} = a^n b^n ab = a^n a b^n b = a^{n + 1}b^{n + 1}, as b^n a = ab^n. Also (b^{-1})^n b^n = e by regrouping and canceling pairs, so (b^{-1})^n = (b^n)^{-1}. For a, b \in H, then, (ab^{-1})^n = a^n (b^n)^{-1} = e, and Theorem 3 applies. \square In \operatorname{GL}_2(\mathbb{R}), A^2 = B^2 = I_2, but AB = \begin{pmatrix} 0 & 1 \\ -1 & 0 \end{pmatrix} has (AB)^2 = -I_2 \ne I_2.
  14. Proof. geg^{-1} = e, so the set is nonempty. For h_1, h_2 \in H, by Theorem 3 of Binary operations and the group axioms (gh_2g^{-1})^{-1} = gh_2^{-1}g^{-1}, so (gh_1g^{-1})(gh_2g^{-1})^{-1} equals gh_1g^{-1}gh_2^{-1}g^{-1} = g(h_1h_2^{-1})g^{-1}, and h_1h_2^{-1} \in H. Theorem 3 applies. \square
  15. Proof. e = ee \in HK. For h_1k_1, h_2k_2 \in HK, (h_1k_1)(h_2k_2)^{-1} = h_1k_1k_2^{-1}h_2^{-1}, which by commutativity is (h_1h_2^{-1})(k_1k_2^{-1}) \in HK. Theorem 3 applies. \square As A^2 = B^2 = I_2, H and K are closed, so subgroups by Theorem 5. Then HK = \{I_2, A, B, AB\} with AB = \begin{pmatrix} 0 & 1 \\ -1 & 0 \end{pmatrix}, but BA = \begin{pmatrix} 0 & -1 \\ 1 & 0 \end{pmatrix} is a product of two elements of HK that is not in HK.
  16. Proof. Call the set Q; it contains 1. Since \displaystyle \sqrt{2} is irrational (Proof by contradiction), c^2 - 2d^2 \ne 0 for rationals c, d not both 0: if d = 0 then c \ne 0, and if d \ne 0 then c^2 = 2d^2 would make \displaystyle \sqrt{2} = |c/d| rational. For nonzero \displaystyle x = a + b\sqrt{2} and \displaystyle y = c + d\sqrt{2} in Q, \displaystyle xy^{-1} = \frac{(a + b\sqrt{2})(c - d\sqrt{2})}{c^2 - 2d^2} = \frac{(ac - 2bd) + (bc - ad)\sqrt{2}}{c^2 - 2d^2} has rational coefficients and is nonzero, so lies in Q; Theorem 3 applies. \square As \displaystyle (3 + 2\sqrt{2})(3 - 2\sqrt{2}) = 9 - 8 = 1, the inverse is \displaystyle 3 - 2\sqrt{2}.
  17. Proof. Let h \in H with h^n = e, n \ge 1. If n = 1, then h = e and h^{-1} = h \in H. If n \ge 2, then h^{n - 1} \in H by closure, and hh^{n - 1} = h^{n - 1}h = h^n = e, so h^{-1} = h^{n - 1} \in H. As H is nonempty, e = hh^{-1} \in H for any h \in H, and Theorem 2 applies. \square
math.wiki · updated 2026-11-17 · CC-BY-SA-4.0