陶哲轩Analysis II(第4版)习题的参考解答及思考(第1章)
第1章
章节1.1
练习1.1.1
题目:
Prove Lemma 1.1.1.
Lemma 1.1.1的内容:
Let \( (x_n)_{n = m}^{\infty} \) be a sequence of real numbers, and let \( x \) be another real number. Then \( (x_n)_{n = m}^{\infty} \) converges to \( x \) if and only if \( \lim_{n \to \infty} d(x_n, x) = 0 \).
证明:
必要性:
如果\( (x_n)_{n = m}^{\infty} \)收敛于\( x \),则\( \forall \epsilon > 0, \exists N \geq m, \forall n \geq N, |x - x_n| \leq \epsilon \),特别的,有\( |d(x_n, x) - 0| = d(x_n, x) = |x - x_n| \leq \epsilon \),即\( \lim_{n \to \infty} d(x_n, x) = 0 \)。
充分性:
如果\( \lim_{n \to \infty} d(x_n, x) = 0 \),则\( \forall \epsilon > 0, \exists N \geq m, \forall n \geq N, |d(x_n, x) - 0| = d(x_n, x) \leq \epsilon \),而\( |x - x_n| = d(x_n, x) \leq \epsilon \),即\( (x_n)_{n = m}^{\infty} \)收敛于\( x \)。
证毕。
练习1.1.2
题目:
Show that the real line with the metric \( d(x, y) := |x - y| \) is indeed a metric space. (Hint: you may wish to review your proof of Proposition 4.3.3 from Analysis I.)
附注:
虽然定理I.4.3.3仅证明了有理数的情况,没有证明实数的情况,但是实数也满足定理I.4.3.3,这里简短点,直接使用定理I.4.3.3,而不仿照我们定理I.4.3.3的证明再证明一遍了。
证明:
证明满足条件(a)和条件(b):
条件(a)和(b)等价于\( \forall x, y \in \mathbf{R}, d(x, y) = 0 \)当且仅当\( x = y \),而由定理I.4.3.3(a),可得\( \forall x, y \in \mathbf{R}, |x - y| = 0 \)当且仅当 \( x = y \),进而可得\( d(x - y) = |x - y| = 0 \)当且仅当\( x = y \)。
证明满足条件(c):
\( \forall x \neq y \in \mathbf{R} \),由定理I.4.3.3(a),可得\( d(x, y) = |x - y| > 0 \)(因为\( x \neq y \))。
证明满足条件(d):
直接由定理I.4.3.3的(g)可得。
证毕。
练习1.1.3
题目:
Let \( X \) be a set, and let \( d: X \times X \to [0, \infty) \) be a function.
- Give an example of a pair \( (X, d) \) which obeys axioms (bcd) of Definition 1.1.2, but not (a). (Hint: modify the discrete metric.)
- Give an example of a pair \( (X, d) \) which obeys axioms (acd) of Definition 1.1.2, but not (b).
- Give an example of a pair \( (X, d) \) which obeys axioms (abd) of Definition 1.1.2, but not (c).
- Give an example of a pair \( (X, d) \) which obeys axioms (abc) of Definition 1.1.2, but not (d). (Hint: try examples where X is a finite set.)
例子1:
令\( X := \{ 1, 3, 5 \} \),定义度量函数\( d \),\( \forall x, y \in X \),如果\( x \neq y \),则令\( d(x, y) := |x - y| \),如果\( x = y \),则令\( d(x, y) := 1 \),易证\( (X, d) \)满足(bcd),但不满足(a)。
例子2:
给定任意集合\( X \),定义度量函数\( d \),\( \forall x, y \in X \),令\( d(x, y) := 0 \),易证\( (X, d) \)满足(acd),但不满足(b)。
例子3:
令\( X := \mathbf{R} \),定义度量函数\( d \),\( \forall x, y \in X \),如果\( x \geq y \),令\( d(x, y) := x - y \),如果\( x < y \),则令\( d(x, y) := 1 \),易证\( (X, d) \)满足(abd),但不满足(c)。
例子4:
令\( X := \{ 1, 3, 5 \} \),定义度量函数\( d \),\( \forall x, y \in X \),
- 如果\( x = y \),则令\( d(x, y) := 0 \)。
- 如果\( x = 1, y = 3 \)或者\( x = 3, y = 2 \),则令\( d(x, y) := 1 \)。
- 如果\( x = 3, y = 5 \)或者\( x = 5, y = 3 \),则令\( d(x, y) := 1 \)。
- 如果\( x = 1, y = 5 \)或者\( x = 5, y = 1 \),则令\( d(x, y) := 3 \)。
易证\( (X, d) \)满足(abc),但不满足(d)。
练习1.1.4
题目:
Shwo that the pair \( (Y, d|_{Y \times Y}) \) defined in Example 1.1.5 is indeed a metric space.
Example 1.1.5的内容:
Example 1.1.5 (Induced metric spaces) Let \( (X, d) \) be any metric space, and let \( Y \) be a subset of \( X \). Then we can restrict the metric function \( d : X \times X \to [0, +\infty) \) to the subset \( Y \times Y \) of \( X \times X \) to create a restricted metric function \( d|_{Y \times Y} : Y \times Y \to [0, +\infty) \) of Y ; this is known as the metric on \( Y \) induced by the metric \( d \) on \( X \). The pair \( (Y, d|_{Y \times Y}) \) is a metric space (Exercise 1.1.4) and is known the subspace of \( (X, d) \) induced by \( Y \). Thus for instance the metric on the real line in the previous example induces a metric space structure on any subset of the reals, such as the integers \( \mathbf{Z} \), or an interval \( [a, b] \).
证明:
\( \forall x \in Y \),有\( x \in X \),进而有\( d|_{Y \times Y}(x, x) = d(x, x) = 0 \),满足条件(a)。
\( \forall x, y \in Y \),有\( x \neq y \in X \),进而有\( d|_{Y \times Y}(x, y) = d(x, y) > 0 \),满足条件(b)。
\( \forall x, y \in Y \),有\( x, y \in X \),进而有\( d|_{Y \times Y}(x, y) = d(x, y) = d(y, x) = d|_{Y \times Y}(y, x) \),满足条件(c)。
\( \forall x, y, z \in Y \),有\( x, y, z \in X \),进而有\( d|_{Y \times Y}(x, y) = d(x, y) \leq d(x, y) + d(y, z) = d|_{Y \times Y}(x, y) + d|_{Y \times Y}(y, z) \),满足条件(d)。
证毕。
练习1.1.5
题目:
Let \( n \geq 1 \), and let \( a_1, a_2, \dots, a_n \) and \( b_1, b_2, \dots, b_n \) be real numbers. Verify the identity
\[ (\sum_{i = 1}^{n} a_i b_i)^2 + \dfrac{1}{2} \sum_{i = 1}^{n} \sum_{j = 1}^{n} (a_i b_j - a_j b_i)^2 = (\sum_{i = 1}^{n} a_i^2) (\sum_{j = 1}^{n} b_j^2), \]
and conclude the Cauchy-Schwarz inequality
\[ |\sum_{i = 1}^{n} a_i b_i| \leq (\sum_{i = 1}^{n} a_i^2)^{1/2} (\sum_{j = 1}^{n} b_j^2)^{1/2}. \]
Then use the Cauchy-Schwarz inequality to prove the triangle inequality
\[ (\sum_{i = 1}^{n} (a_i + b_i)^2)^{1/2} \leq (\sum_{i = 1}^{n} a_i^2)^{1/2} + (\sum_{j = 1}^{n} b_j^2)^{1/2}. \]
证明:
证明第1个等式:
使用数学归纳法证明。
当\( n = 1 \)时,等式左边\( = a_1 b_1 + \dfrac{1}{2} (a_1 b_1 - a_1 b_1)^2 = a_1 b_1 \),等式右边\( = a_1 b_1 \),左边\( = \)右边,等式成立。
归纳假设当\( n = k \)时等式成立,当\( n = k + 1 \)时,
\[ \begin{aligned} \\ \text{左边} &= (\sum_{i = 1}^{k + 1} a_i b_i)^2 + \dfrac{1}{2} \sum_{i = 1}^{k + 1} \sum_{j = 1}^{k + 1} (a_i b_j - a_j b_i)^2 \\ &= (\sum_{i = 1}^{k} a_i b_i + a_{k + 1} b_{k + 1})^2 + \dfrac{1}{2} \sum_{i = 1}^{k + 1} \sum_{j = 1}^{k + 1} (a_i b_j - a_j b_i)^2 \\ &= (\sum_{i = 1}^{k} a_i b_i)^2 + 2(\sum_{i = 1}^{k} a_i b_i) a_{k + 1} b_{k + 1} + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \dfrac{1}{2} \sum_{i = 1}^{k + 1} \sum_{j = 1}^{k + 1} (a_i b_j - a_j b_i)^2 \\ &= (\sum_{i = 1}^{k} a_i b_i)^2 + 2(\sum_{i = 1}^{k} a_i b_i) a_{k + 1} b_{k + 1} + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \dfrac{1}{2} \sum_{i = 1}^{k + 1} \sum_{j = 1}^{k} (a_i b_j - a_j b_i)^2 + \dfrac{1}{2} \sum_{i = 1}^{k + 1} (a_i b_{k + 1} - a_{k + 1} b_i)^2 \\ &= (\sum_{i = 1}^{k} a_i b_i)^2 + 2(\sum_{i = 1}^{k} a_i b_i) a_{k + 1} b_{k + 1} + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \dfrac{1}{2} \sum_{i = 1}^{k} \sum_{j = 1}^{k} (a_i b_j - a_j b_i)^2 + \dfrac{1}{2} \sum_{j = 1}^{k} (a_{k + 1} b_j - a_j b_{k + 1})^2 + \\ &\qquad \dfrac{1}{2} \sum_{i = 1}^{k + 1} (a_i b_{k + 1} - a_{k + 1} b_i)^2 \\ &= (\sum_{i = 1}^{k} a_i^2) (\sum_{j = 1}^{k} b_j^2) + 2(\sum_{i = 1}^{k} a_i b_i) a_{k + 1} b_{k + 1} + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \dfrac{1}{2} \sum_{j = 1}^{k} (a_{k + 1} b_j - a_j b_{k + 1})^2 + \dfrac{1}{2} \sum_{i = 1}^{k + 1} (a_i b_{k + 1} - a_{k + 1} b_i)^2 \\ &= (\sum_{i = 1}^{k} a_i^2) (\sum_{j = 1}^{k} b_j^2) + 2(\sum_{i = 1}^{k} a_i b_i) a_{k + 1} b_{k + 1} + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \dfrac{1}{2} \sum_{j = 1}^{k} (a_{k + 1} b_j - a_j b_{k + 1})^2 + \\ &\qquad \dfrac{1}{2} \sum_{i = 1}^{k} (a_i b_{k + 1} - a_{k + 1} b_i)^2 + \dfrac{1}{2} (a_{k + 1} b_{k + 1} - a_{k + 1} b_{k + 1})^2 \\ &= (\sum_{i = 1}^{k} a_i^2) (\sum_{j = 1}^{k} b_j^2) + 2(\sum_{i = 1}^{k} a_i b_i) a_{k + 1} b_{k + 1} + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \dfrac{1}{2} \sum_{j = 1}^{k} (a_{k + 1} b_j - a_j b_{k + 1})^2 + \dfrac{1}{2} \sum_{i = 1}^{k} (a_i b_{k + 1} - a_{k + 1} b_i)^2 \\ &= (\sum_{i = 1}^{k} a_i^2) (\sum_{j = 1}^{k} b_j^2) + 2(\sum_{i = 1}^{k} a_i b_i) a_{k + 1} b_{k + 1} + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \dfrac{1}{2} \sum_{j = 1}^{k} (a_{k + 1} b_j - a_j b_{k + 1})^2 + \dfrac{1}{2} \sum_{i = 1}^{k} (-1(a_{k + 1} b_i - a_i b_{k + 1}))^2 \\ &= (\sum_{i = 1}^{k} a_i^2) (\sum_{j = 1}^{k} b_j^2) + 2(\sum_{i = 1}^{k} a_i b_i) a_{k + 1} b_{k + 1} + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \dfrac{1}{2} \sum_{j = 1}^{k} (a_{k + 1} b_j - a_j b_{k + 1})^2 + \dfrac{1}{2} \sum_{i = 1}^{k} (a_{k + 1} b_i - a_i b_{k + 1})^2 \\ &= (\sum_{i = 1}^{k} a_i^2) (\sum_{j = 1}^{k} b_j^2) + 2(\sum_{i = 1}^{k} a_i b_i) a_{k + 1} b_{k + 1} + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \sum_{j = 1}^{k} (a_{k + 1} b_j - a_j b_{k + 1})^2 \\ &= (\sum_{i = 1}^{k} a_i^2) (\sum_{j = 1}^{k} b_j^2) + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad 2(\sum_{i = 1}^{k} a_i b_i) a_{k + 1} b_{k + 1} + \sum_{j = 1}^{k} (a_{k + 1} b_j - a_j b_{k + 1})^2 \\ &= (\sum_{i = 1}^{k} a_i^2) (\sum_{j = 1}^{k} b_j^2) + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \sum_{i = 1}^{k} 2 a_i b_i a_{k + 1} b_{k + 1} + \sum_{i = 1}^{k} (a_{k + 1} b_i - a_i b_{k + 1})^2 \\ &= (\sum_{i = 1}^{k} a_i^2) (\sum_{j = 1}^{k} b_i^2) + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \sum_{i = 1}^{k} (2 a_i b_i a_{k + 1} b_{k + 1} + (a_{k + 1} b_i - a_i b_{k + 1})^2) \\ &= (\sum_{i = 1}^{k} a_i^2) (\sum_{j = 1}^{k} b_i^2) + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \sum_{i = 1}^{k} (2 a_i b_i a_{k + 1} b_{k + 1} + a_{k + 1}^2 b_i^2 + a_i^2 b_{k + 1}^2 - 2 a_{k + 1} b_i a_i b_{k + 1}) \\ &= (\sum_{i = 1}^{k} a_i^2) (\sum_{j = 1}^{k} b_i^2) + (a_{k + 1} b_{k + 1})^2 + \\ &\qquad \sum_{i = 1}^{k} (a_{k + 1}^2 b_i^2 + a_i^2 b_{k + 1}^2) \end{aligned} \]
我们接着展开右边,这样比较容易看出左边\( = \)右边,
\[ \begin{aligned} \\ \text{右边} &= (\sum_{i = 1}^{k + 1} a_i^2) (\sum_{j = 1}^{k + 1} b_j^2) \\ &= (\sum_{i = 1}^{k} a_i^2 + a_{k + 1}^2) (\sum_{j = 1}^{k} b_j^2 + b_{k + 1}^2) \\ &= (\sum_{i = 1}^{k} a_i^2)(\sum_{j = 1}^{k} b_j^2) + (\sum_{i = 1}^{k} a_i^2) b_{k + 1}^2 + a_{k + 1}^2(\sum_{j = 1}^{k} b_j^2) + a_{k + 1}^2 b_{k + 1}^2 \\ &= (\sum_{i = 1}^{k} a_i^2)(\sum_{j = 1}^{k} b_j^2) + (a_{k + 1} b_{k + 1})^2 + (\sum_{i = 1}^{k} a_i^2) b_{k + 1}^2 + a_{k + 1}^2(\sum_{i = 1}^{k} b_i^2) \\ &= (\sum_{i = 1}^{k} a_i^2)(\sum_{j = 1}^{k} b_j^2) + (a_{k + 1} b_{k + 1})^2 + (\sum_{i = 1}^{k} a_i^2 b_{k + 1}^2) + (\sum_{i = 1}^{k} a_{k + 1}^2 b_i^2) \\ &= (\sum_{i = 1}^{k} a_i^2)(\sum_{j = 1}^{k} b_j^2) + (a_{k + 1} b_{k + 1})^2 + (\sum_{i = 1}^{k} a_{k + 1}^2 b_i^2 + a_i^2 b_{k + 1}^2) \end{aligned} \]
综上,\( 左边 = 右边 \),故\( n = k + 1 \)时等式也成立,归纳完毕。
证明Cauchy-Schwarz不等式:
由第1个等式可得,
\[ (\sum_{i = 1}^{n} a_i b_i)^2 = (\sum_{i = 1}^{n} a_i^2) (\sum_{j = 1}^{n} b_j^2) - \dfrac{1}{2} \sum_{i = 1}^{n} \sum_{j = 1}^{n} (a_i b_j - a_j b_i)^2, \]
这里由于\( \dfrac{1}{2} \sum_{i = 1}^{n} \sum_{j = 1}^{n} (a_i b_j - a_j b_i)^2 \geq 0 \),于是可得,
\[ (\sum_{i = 1}^{n} a_i b_i)^2 \leq (\sum_{i = 1}^{n} a_i^2) (\sum_{j = 1}^{n} b_j^2), \]
进而可得,
\[ |\sum_{i = 1}^{n} a_i b_i| = ((\sum_{i = 1}^{n} a_i b_i)^2)^{1/2} \leq (\sum_{i = 1}^{n} a_i^2)^{1/2} (\sum_{j = 1}^{n} b_j^2)^{1/2}。 \]
证明三角不等式:
由Cauchy-Schwarz不等式可得, \( \sum_{i = 1}^{n} a_i b_i \leq |\sum_{i = 1}^{n} a_i b_i| \leq (\sum_{i = 1}^{n} a_i^2)^{1/2} (\sum_{j = 1}^{n} b_j^2)^{1/2} \),进而可得,
\[ \begin{aligned} \sum_{i = 1}^{n} (a_i + b_i)^2 &= \sum_{i = 1}^{n} (a_i^2 + 2 a_i b_i + b_i^2) \\ &= \sum_{i = 1}^{n} a_i^2 + 2 \sum_{i = 1}^{n} a_i b_i + \sum_{i = 1}^{n} b_i^2 \\ &\leq \sum_{i = 1}^{n} a_i^2 + \sum_{i = 1}^{n} b_i^2 + 2((\sum_{i = 1}^{n} a_i^2)^{1/2} (\sum_{j = 1}^{n} b_j^2)^{1/2}) \\ &= ((\sum_{i = 1}^{n} a_i^2)^{1/2} + (\sum_{i = 1}^{n} b_i^2)^{1/2})^2 \end{aligned} \]
最终可得,
\[ \begin{aligned} (\sum_{i = 1}^{n} (a_i + b_i)^2)^{1/2} \leq (\sum_{i = 1}^{n} a_i^2)^{1/2} + (\sum_{j = 1}^{n} b_j^2)^{1/2}。 \end{aligned} \]
证毕。
练习1.1.6
题目:
Show that \( (\mathbf{R}^n , d_{l^2}) \) in Example 1.1.6 is indeed a metric space. (Hint: use Exercise 1.1.5.)
证明:
证明满足条件(a):
易证,略。
证明满足条件(b):
\( \forall x \neq y \in \mathbf{R}^n \),记\( x = (x_1, \dots, x_n), y = (y_1, \dots, y_n) \),由\( x \neq y \),可得\( \exists 1 \leq k \leq n \),满足\( x_k \neq y_k \),此时有\( (x_k - y_k)^2 > 0 \),而\( \forall 1 \leq i \leq n, i \neq k \),我们有\( (x_i - y_i)^2 \geq 0 \),可得\( (x_1 - y_1)^2 + \dots + (x_n - y_n)^2 > 0 \),进而可得\( d_{l^2}(x, y) = \sqrt{(x_1 - y_1)^2 + \dots + (x_n - y_n)^2} > 0 \)。
证明满足条件(c):
易证,略。
证明满足条件(d):
\( \forall x, y, z \in \mathbf{R}^n \),记\( x = (x_1, \dots, x_n), y = (y_1, \dots, y_n), z = (z_1, \dots, z_n) \), \( \forall 1 \leq i \leq n \),令\( a_i = x_i - y_i, b_i = y_i - z_i \),由练习1.1.5证明的三角不等式,可得,
\[ \begin{aligned} d_{l^2}(x, z) &= ((x_1 - z_1)^2 + \dots + (x_n - z_n)^2)^{1/2} \\ &= (\sum_{i = 1}^{n} (a_i + b_i)^2)^{1/2} \\ &\leq (\sum_{i = 1}^{n} a_i^2)^{1/2} + (\sum_{j = 1}^{n} b_j^2)^{1/2} \\ &= d_{l^2}(x, y) + d_{l^2}(y, z)。 \end{aligned} \]
证毕。
练习1.1.7
题目:
Show that the pair \( (\mathbf{R}^n, d_{l^1}) \) in Example 1.1.7 is indeed a metric space.
证明:
\( \forall x, y, z \in \mathbf{R}^n \),记\( x = (x_1, \dots, x_n), y = (y_1, \dots, y_n), z = (z_1, \dots, z_n) \)。
证明满足条件(a):
易证,略。
证明满足条件(b):
由\( x \neq y \),可得\( \exists 1 \leq k \leq n \),满足\( x_k \neq y_k \),此时有\( |x_k - y_k| > 0 \),而\( \forall 1 \leq i \leq n, i \neq k \),我们有\( |x_i - y_i| \geq 0 \),可得\( |x_1 - y_1| + \dots + |x_n - y_n| > 0 \),这意味着\( d_{l^1}(x, y) > 0 \)。
证明满足条件(c):
易证,略。
证明满足条件(d):
\( \forall 1 \leq i \leq n \),我们有\( |x_i - z_i| = |(x_i - y_i) + (y_i - z_i)| \leq |x_i - y_i| + |y_i - z_i| \),进而可得\( d_{l^1}(x, z) = \sum_{i = i}^{n} |x_i - z_i| \leq \sum_{i = 1}^{n} (|x_i - y_i| + |y_i - z_i|) = \sum_{i = 1}^{n} |x_i - y_i| + \sum_{i = 1}^{n} |y_i - z_i| = d_{l^1}(x, y) + d_{l^1}(y, z) \)。
证毕。
练习1.1.8
题目:
Prove the two inequalities in (1.1). (Hint: For the first inequality, square both sides. For the second inequality, use Exercise (1.1.5).)
(1.1)中的两个不等式:
\( d_{l^2}(x, y) \leq d_{l^1}(x, y) \leq \sqrt{n} d_{l^2}(x, y) \)
证明:
记\( x = (x_1, \dots, x_n), y = (y_1, \dots, y_n) \)。
证明第1个不等式:
首先我们有:
\[ \begin{aligned} (d_{l^1}(x, y))^2 &= (\sum_{i = 1}^{n} |x_i - y_i|)^2 \\ &= \sum_{i = 1}^{n} \sum_{j = 1}^{n} |x_i - y_i||x_j - y_j| \\ &= (\sum_{(i, j) \in \{ 1, \dots ,n \} \times \{ 1, \dots ,n \}, i \neq j} |x_i - y_i||x_j - y_j|) + (\sum_{(i, j) \in \{ 1, \dots ,n \} \times \{ 1, \dots ,n \}, i = j} |x_i - y_i||x_j - y_j|) \\ &= (\sum_{(i, j) \in \{ 1, \dots ,n \} \times \{ 1, \dots ,n \}, i \neq j} |x_i - y_i||x_j - y_j|) + (\sum_{i = 1}^{n} |x_i - y_i||x_i - y_i|) \\ &= (\sum_{(i, j) \in \{ 1, \dots ,n \} \times \{ 1, \dots ,n \}, i \neq j} |x_i - y_i||x_j - y_j|) + (\sum_{i = 1}^{n} (x_i - y_i)^2) \end{aligned} \]
上面\( \sum_{(i, j) \in \{ 1, \dots ,n \} \times \{ 1, \dots ,n \}, i \neq j} |x_i - y_i||x_j - y_j| \geq 0 \),加上\( \sum_{i = 1}^{n} (x_i - y_i)^2 = (d_{l^2}(x, y))^2 \),于是可得\( (d_{l^1}(x, y))^2 \geq (d_{l^2}(x, y))^2 \),进一步可得\( d_{l^1}(x, y) \geq d_{l^2}(x, y) \)。
证明第2个不等式:
使用Cauchy-Schwarz不等式可得,
\[ \begin{aligned} d_{l^1}(x, y) &= \sum_{i = 1}^{n} |x_i - y_i| = |\sum_{i = 1}^{n} (1 \times |x_i - y_i|)| \\ &\leq (\sum_{i = 1}^{n} 1^2)^{1/2} (\sum_{i = 1}^{n} (|x_i - y_i|)^2)^{1/2} \\ &= \sqrt{n} d_{l^2}(x, y) \end{aligned} \]
证毕。
练习1.1.9
题目:
Show that the pair \( (\mathbf{R}^n, d_{l^{\infty}}) \) in Example 1.1.9 is indeed a metric space.
证明:
证明满足条件(a):
易证,略。
证明满足条件(b):
由\( x \neq y \),可得\( \exists 1 \leq k \leq n \),满足\( x_k \neq y_k \),此时有\( |x_k - y_k| > 0 \),而\( \forall 1 \leq i \leq n, i \neq k \),我们有\( |x_i - y_i| \geq 0 \),可得\( \sup \{ |x_i - y_i| : 1 \leq i \leq n \} > 0 \),这意味着\( d_{l^{\infty}}(x, y) > 0 \)。
证明满足条件(c):
易证,略。
证明满足条件(d):
\( \forall 1 \leq i \leq n \),我们有\( |x_i - z_i| = |(x_i - y_i) + (y_i - z_i)| \leq |x_i - y_i| + |y_i - z_i| \),进而可得\( d_{l^1}(x, z) = \sup \{ |x_i - z_i| : 1 \leq i \leq n \} \leq \sup \{ |x_i - y_i| + |y_i - z_i| : 1 \leq i \leq n \} = \sup \{ |x_i - y_i| : 1 \leq i \leq n \} + \sup \{ |y_i - z_i| : 1 \leq i \leq n \} = d_{l^{\infty}}(x, y) + d_{l^{\infty}}(y, z) \)。
注:这里\( \sup \{ |x_i - y_i| + |y_i - z_i| : 1 \leq i \leq n \} = \sup \{ |x_i - y_i| : 1 \leq i \leq n \} + \sup \{ |y_i - z_i| : 1 \leq i \leq n \} \)是比较容易证明的,这里就不证了(反证法:假设\( |x_i - y_i|, |y_j - z_j| \)不都取最大值的话,则\( |x_i - y_i| + |y_i - z_i| \)能达到最大值吗?)。
证毕。
练习1.1.10
题目:
Prove the two inequalities in (1.2).
(1.2)中的两个不等式:
\( \dfrac{1}{\sqrt{n}} d_{l^2}(x, y) \leq d_{l^{\infty}}(x, y) \leq d_{l^2}(x, y) \)
证明:
记\( x = (x_1, \dots, x_n), y = (y_1, \dots, y_n) \)。
证明第1个不等式:
\( \forall 1 \leq i \leq n \),我们有\( (x_i - y_i)^2 \leq (\sup \{ |x_i - y_i| : 1 \leq i \leq n \})^2 \),可得\( (d_{l^2}(x, y))^2 \leq n (d_{l^{\infty}}(x, y))^2 \),进而可得\( \dfrac{1}{\sqrt{n}} d_{l^2}(x, y) \leq d_{l^{\infty}}(x, y) \)。
证明第2个不等式:
由\( \exists 1 \leq k \leq n \)满足\( |x_k - y_k| = \sup \{ |x_i - y_i| : 1 \leq i \leq n \} \),可得\( (d_{l^2}(x, y))^2 = (\sum_{i = 1}^{n} (x_i - y_i)^2) \geq (\sup \{ |x_i - y_i| : 1 \leq i \leq n \})^2 = (d_{l^{\infty}}(x, y))^2 \),进而可得\( d_{l^2}(x, y) \geq d_{l^{\infty}}(x, y) \)。
证毕。
练习1.1.11
题目:
Show that the discrete metric \( (X, d_{\text{disc}}) \) in Example 1.1.11 is indeed a metric space.
证明:
证明满足条件(a):
易证,略。
证明满足条件(b):
易证,略。
证明满足条件(c):
易证,略(分\( x = y, x \neq y \)两种情况讨论)。
证明满足条件(d):
\( \forall x, y, z \in \mathbf{R}^n \),分情况讨论:
- 如果\( x = z \),则\( d(x, z) = 0 \leq d(x, y) + d(y, z) \)。
- 如果\( x \neq z \),则此时\( x = y, y = z \)不能同时成立,于是有\( x \neq y \)或\( y \neq z \),进而\( d(x, y) + d(y, z) \geq 1 = d(x, z) \)。
证毕。
练习1.1.12
题目:
Prove Proposition 1.1.18.
Proposition 1.1.18的内容:
(Equivalence of \( l^1 \), \( l^2 \), \( l^{\infty} \)) Let \( \mathbf{R}^n \) be a Euclidean space, and let \( (x^{(k)})_{k = m}^{\infty} \) be a sequence of points in \( \mathbf{R}^n \). We write \( x^{(k)} = (x_1^{(k)}, x_2^{(k)}, \dots, x_n^{(k)}) \), i.e., for \( j = 1, 2, \dots, n, x_j^{(k)} \in \mathbf{R} \) is the jth co-ordinate of \( x^{(k)} \in \mathbf{R}^n \). Let \( x = (x_1, \dots, x_n) \) be a point in \( \mathbf{R}^n \). Then the following four statements are equivalent:
- \( (x^{(k)})_{k = m}^{\infty} \) converges to \( x \) with respect to the Euclidean metric \( d_{l^2} \).
- \( (x^{(k)})_{k = m}^{\infty} \) converges to \( x \) with respect to the taxicab metric \( d_{l^1} \).
- \( (x^{(k)})_{k = m}^{\infty} \) converges to \( x \) with respect to the sup norm metric \( d_{l^{\infty}} \).
- For every \( 1 \leq j \leq n \), the sequence \( (x_j^{(k)})_{k = m}^{\infty} \) converges to \( x_j \) (Notice that this is a sequence of real numbers, not of points in \( \mathbf{R}^n \).)
证明:
证明\( 1 \rightarrow 2 \):
由\( (x^{(k)})_{k = m}^{\infty} \)在Euclidean度量下收敛到\( x \),可得\( \forall \epsilon > 0, \exists N \geq m, \forall n' \geq N, d_{l^2}(x^{(n')}, x) \leq \epsilon \),进而由不等式(1.1),可得\( d_{l^1}(x^{(n')}, x) \leq \sqrt{n} d_{l^2}(x^{(n')}, x) \leq \sqrt{n} \epsilon \)( 注: 这里\( n \)是维度,是常数),这意味着\( (x^{(k)})_{k = m}^{\infty} \)在taxicab度量下收敛到\( x \)。
证明\( 2 \rightarrow 3 \):
由\( (x^{(k)})_{k = m}^{\infty} \)在taxicab度量下收敛到\( x \),可得\( \forall \epsilon > 0, \exists N \geq m, \forall n' \geq N, d_{l^2}(x^{(n')}, x) \leq \epsilon \),进而由不等式(1.2),可得\( d_{l^{\infty}}(x^{(n')}, x) \leq d_{l^2}(x^{(n')}, x) \leq \epsilon \),这意味着\( (x^{(k)})_{k = m}^{\infty} \)在sup norm度量下收敛到\( x \)。
证明\( 3 \rightarrow 4 \):
由\( (x^{(k)})_{k = m}^{\infty} \)在sup norm度量下收敛到\( x \),可得\( \forall \epsilon > 0, \exists N \geq m, \forall n' \geq N, d_{l^{\infty}}(x^{(n')}, x) \leq \epsilon \),进而由\( \exists 1 \leq j' \leq n \)满足\( |x_{j'}^{(n')} - y_{j'}^{(n')}| = \sup \{ |x_i^{(n')} - y_i^{(n')}| : 1 \leq i \leq n \} \),可得\( \forall 1 \leq j \leq n, d_{l^{\infty}}(x_j^{(n')}, x_j) \leq d_{l^{\infty}}(x_{j'}^{(n')}, x_{j'}) = d_{l^{\infty}}(x^{(n')}, x) \leq \epsilon \),这意味着\( (x_j^{(k)})_{k = m}^{\infty} \)收敛到\( x_j \)。
证明\( 4 \rightarrow 1 \):
\( \forall 1 \leq j \leq n \),由\( (x_j^{(k)})_{k = m}^{\infty} \)收敛到\( x_j \),可得\( \forall \epsilon > 0, \exists N \geq m, \forall n' \geq N, |x_j^{(n')} - x_j| \leq \epsilon \),于是可得\( (d_{l^2}(x^{(n')}, x))^2 = \sum_{j = 1}^{n} (x_j^{(n')} - x_j)^2 \leq n \epsilon^2 \),进而可得\( d_{l^2}(x^{(n')}, x) \leq \sqrt{n} \epsilon \),这意味着\( (x^{(k)})_{k = m}^{\infty} \)在Euclidean度量下收敛到\( x \)。
证毕。
练习1.1.13
题目:
Prove Proposition 1.1.19.
Proposition 1.1.19的内容:
(Convergence in the discrete metric) Let \( X \) be any set, and let \( d_{\text{disc}} \) be the discrete metric on \( X \). Let \( (x^{(n)})_{n = m}^{\infty} \) be a sequence of points in \( X \), and let \( x \) be a point in \( X \). Then \( (x^{(n)})_{n = m}^{\infty} \) converges to \( x \) with respect to the discrete metric \( d_{\text{disc}} \) if and only if there exists an \( N \geq m \) such that \( x^{(n)} = x \) for all \( n \geq N \).
证明:
由\( (x^{(n)})_{n = m}^{\infty} \)收敛于\( x \)可得, \( \exists N \geq m, \forall n \geq N, d_{\text{disc}}(x^{(n)}, x) \leq \dfrac{1}{2} \),而所有不同点之间的度量都是\( 1 > \dfrac{1}{2} \),唯一\( \leq \dfrac{1}{2} \)的度量就是\( 0 \),而度量要为\( 0 \),则要求两个点是同一个点,这意味着\( \forall n \geq N, x = x^{(n)} \)。
证毕。
练习1.1.14
题目:
Prove Proposition 1.1.20. (Hint: modify the proof of Proposition 6.1.7 from Analysis I.)
Proposition 1.1.20的内容:
(Uniqueness of limits) Let \( (X, d) \) be a metric space, and let \( (x^{(n)})_{n = m}^{\infty} \) be a sequence in \( X \). Suppose that there are two points \( x, x' \in X \) such that \( (x^{(n)})_{n = m}^{\infty} \) converges to x with respect to \( d \), and \( (x^{(n)})_{n = m}^{\infty} \) also converges to \( x' \) with respect to \( d \). Then we have \( x = x' \).
证明:
假设\( x \neq x' \),则\( \epsilon_0 = \dfrac{|x - x'|}{3} > 0 \),此时由\( (x^{(n)})_{n = m}^{\infty} \)收敛于\( x \),可得\( \exists N_1 \geq m, \forall n \geq N_1, d(x^{(n)}, x) \leq \epsilon_0 \),再由\( (x^{(n)})_{n = m}^{\infty} \)收敛于\( x' \),可得\( \exists N_2 \geq m, \forall n \geq N_2, d(x^{(n)}, x') \leq \epsilon_0 \),令\( N = \max(N_1, N_2) \),则\( \forall n \geq N \),有\( \epsilon = d(x, x') \leq d(x, x^{(n)}) + d(x^{(n)}, x') \leq \dfrac{2|x - x'|}{3} \),这里\( \epsilon \leq \dfrac{2|x - x'|}{3} \),矛盾,因此假设不成立,有\( x = x' \)。
证毕。
练习1.1.15
题目:
Let
\[ X := \{ (a_n)_{n = 0}^{\infty} : \sum_{n = 0}^{\infty} |a_n| < \infty \} \]
be the space of absolutely convergent sequences. Define the \( l^1 \) and \( l^{\infty} \) metrics on this space by
\[ d_{l^1}((a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) := \sum_{n = 0}^{\infty} |a_n - b_n|; \]
\[ d_{l^{\infty}}((a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) := \sup_{n \in \mathbf{N}} |a_n - b_n|. \]
Show that these are both metrics on \( X \), but show that there exist sequences \( x^{(1)}, x^{(2)}, \dots \) of elements of \( X \) (i.e., sequences of sequences) which are convergent with respect to the \( d_{l^{\infty}} \) metric but not with respect to the \( d_{l^1} \) metric. Conversely, show that any sequence which converges in the \( d_{l^1} \) metric automatically converges in the \( d_{l^{\infty}} \) metric.
解答:
证明\( d_{l^1} \)为\( X \)的度量:
首先得确保\( \forall (a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty} \in X, d_{l^1}((a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) \)都是有定义的,由\( (a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty} \)绝对收敛(即序列对应的绝对值级数收敛),可得\( \sum_{n = 0}^{\infty} (|a_n| + |b_n|) \)也收敛,而\( \forall n \geq 0, |a_n - b_n| \leq |a_n| + |b_n| \),加上\( \sum_{n = 0}^{\infty} (|a_n| + |b_n|) \)收敛,根据推论I.7.3.2,可得\( \sum_{n = 0}^{\infty} |a_n - b_n| \)收敛,这意味着\( d_{l^1}((a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) \)是有定义的。
证明满足条件(a):
\( \forall (a_n)_{n = 0}^{\infty} \in X \),我们有 \( d_{l^1}((a_n)_{n = 0}^{\infty}, (a_n)_{n = 0}^{\infty}) = \sum_{n = 0}^{\infty} |a_n - a_n| = \sum_{n = 0}^{\infty} 0 = 0 \)。
证明满足条件(b):
\( \forall (a_n)_{n = 0}^{\infty} \neq (b_n)_{n = 0}^{\infty} \in X \),可得\( \exists k \geq 0, a_k \neq b_k \),于是有\( |a_k - b_k| > 0 \),进而可得\( d_{l^1}((a_n)_{n = 0}^{\infty}, (a_n)_{n = 0}^{\infty}) = \sum_{n = 0}^{\infty} |a_n - b_n| \geq |a_k - b_k| > 0 \)。
证明满足条件(c):
易证,略。
证明满足条件(d):
\( \forall (a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}, (c_n)_{n = 0}^{\infty} \in X \),我们有\( \forall n \geq 0, |a_n - c_n| \leq |a_n - b_n| + |b_n - c_n| \),进而根据推论I.7.3.2,可得\( d_{l^1}((a_n)_{n = 0}^{\infty}, (c_n)_{n = 0}^{\infty}) = \sum_{n = 0}^{\infty} |a_n - c_n| \leq \sum_{n = 0}^{\infty} (|a_n - b_n| + |b_n - c_n|) = \sum_{n = 0}^{\infty} |a_n - b_n| + \sum_{n = 0}^{\infty} |b_n - c_n| = d_{l^1}((a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) + d_{l^1}((b_n)_{n = 0}^{\infty}, (c_n)_{n = 0}^{\infty}) \)。
综上,\( d_{l^1} \)为\( X \)的度量,证毕。
证明\( d_{l^{\infty}} \)为\( X \)的度量:
首先得确保\( \forall (a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty} \in X, d_{l^1}((a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) \)都是实数,而不会是\( \infty \),假设\( \sup_{n \in \mathbf{N}} |a_n - b_n| = \infty \),则\( (|a_n - b_n|)_{n = 0}^{\infty} \)无上界,进而\( \sum_{n = 0}^{\infty} |a_n - b_n| \)不收敛,这和上面的结论矛盾,因此假设不成立,有\( d_{l^{\infty}}((a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) = \sup_{n \in \mathbf{N}} |a_n - b_n| \)为实数。
证明满足条件(a):
易证,略。
证明满足条件(b):
\( \forall (a_n)_{n = 0}^{\infty} \neq (b_n)_{n = 0}^{\infty} \in X \),可得\( \exists k \geq 0, a_k \neq b_k \),于是有\( |a_k - b_k| > 0 \),进而可得\( d_{l^{\infty}}((a_n)_{n = 0}^{\infty}, (a_n)_{n = 0}^{\infty}) = \sup_{n \in \mathbf{N}} |a_n - b_n| \geq |a_k - b_k| > 0 \)。
证明满足条件(c):
易证,略。
证明满足条件(d):
\( \forall (a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}, (c_n)_{n = 0}^{\infty} \in X \),我们有\( \forall n \geq 0, |a_n - c_n| \leq |a_n - b_n| + |b_n - c_n| \),进而可得\( d_{l^{\infty}}((a_n)_{n = 0}^{\infty}, (c_n)_{n = 0}^{\infty}) = \sup_{n \in \mathbf{N}}^{\infty} |a_n - c_n| \leq \sup_{n \in \mathbf{N}} (|a_n - b_n| + |b_n - c_n|) = \sup_{n \in \mathbf{N}} |a_n - b_n| + \sup_{n \in \mathbf{N}} |b_n - c_n| = d_{l^{\infty}}((a_n)_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) + d_{l^{\infty}}((b_n)_{n = 0}^{\infty}, (c_n)_{n = 0}^{\infty}) \)。
综上,\( d_{l^{\infty}} \)为\( X \)的度量,证毕。
给出一个序列满足在度量\( d_{l^{\infty}} \)下收敛,但在\( d_{l^1} \)下不收敛:
已知级数\( \sum_{n = 1}^{\infty} \dfrac{1}{n^2} \)绝对收敛,我们可以在它基础上改,但是题目要求下标从\( 0 \)开始,而当\( n = 0 \)时,\( \dfrac{1}{n^2} \)是未定义的,故我们当\( n = 0 \)时直接给一个特殊值\( 0 \),构造序列\( (b_n)_{n = 0}^{\infty} \),当\( n = 0 \)时,令\( b_n = 0 \),当\( n > 0 \)时,令\( b_n = \dfrac{1}{n^2} \),易得\( \sum_{n = 0}^{\infty} b_n \)也绝对收敛,进而\( (b_n)_{n = 0}^{\infty} \in X \)。
接着,我们构造一堆\( X \)中的“点”(每个“点”都是一个序列,而不仅仅是一个数),我们希望构造的这些“点”收敛于序列\( (b_n)_{n = 0}^{\infty} \in X \),容易想到,序列的前几项不影响对应级数的收敛,故我们可以在序列\( (b_n)_{n = 0}^{\infty} \)的基础上改前几项,同时,我们希望修改到的这前几项对级数和的影响越来越小,进而级数和最终会和\( \sum_{n = 0}^{\infty} b_n \)越来越接近,而不是越来越远,容易想到,在\( \dfrac{1}{n^2} \)的基础上,加上类似\( \dfrac{1}{k}, \dfrac{1}{k^2}, \dfrac{1}{k^3} \)这种会变得越来越小的项,或许可以达到我们的目的,选择前几项加上\( \dfrac{1}{k} \)试试。
具体的,\( \forall k \geq 1 \),我们构造序列\( (x_n^{(k)})_{n = 0}^{\infty} \),当\( n = 0 \)时,令\( x_n^{(k)} = \dfrac{1}{k} \),当\( 1 \leq n \leq k \)时,令\( x_n^{(k)} = \dfrac{1}{n^2} + \dfrac{1}{k} \),当\( n > k \)时,令\( x_n^{(k)} = b_n \)。
\( \forall k \geq 1 \),令\( x^{(k)} = (x_n^{(k)})_{n = 0}^{\infty} \),令\( b = (b_n)_{n = 0}^{\infty} \),我们现在证明序列\( (x^{(k)})_{k = 1}^{\infty} \)(序列的序列)在度量\( d_{l^{\infty}} \)下收敛于\( b \):
\( \forall k \geq 1, \forall 0 \leq n \leq k \),我们有\( |x_n^{(k)} - b_n| = \dfrac{1}{k} \),而\( \forall n > k \)时,我们则有\( |x_n^{(k)} - b_n| = 0 \),可得\( d_{l^{\infty}}((x_n^{(k)})_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) = \sup_{n \in \mathbf{N}} |x_n^{(k)} - b_n| = \dfrac{1}{k} \),而\( \forall \epsilon > 0 \),存在\( K \geq 1 \),满足\( \dfrac{1}{K} < \epsilon \),于是可得\( \forall k \geq K \),有\( d_{l^{\infty}}((x_n^{(k)})_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) = \sup_{n \in \mathbf{N}} |x_n^{(k)} - b_n| = \dfrac{1}{k} \leq \dfrac{1}{K} < \epsilon \),这说明\( (x^{(k)})_{k = 1}^{\infty} \)在度量\( d_{l^{\infty}} \)下收敛于\( b \)。
我们接着证明\( (x^{(k)})_{k = 1}^{\infty} \)在度量\( d_{l^1} \)下则不收敛,这说明\( d_{l^1} \)和\( d_{l^{\infty}} \)在无限维的情况下是不等价的(这里每个\( x^{(k)} \)都是一个序列,相当于有无限个坐标分量,所以说是无限维):
\( \forall k \geq 1, \forall 0 \leq n \leq k \),我们有\( |x_n^{(k)} - b_n| = \dfrac{1}{k} \),而\( \forall n > k \)时,我们则有\( |x_n^{(k)} - b_n| = 0 \),可得\( d_{l^1}((x_n^{(k)})_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) = \sum_{n = 0}^{\infty} |x_n^{(k)} - b_n| = \dfrac{k + 1}{k} > 1 \),距离没办法任意缩小,明显有\( (x^{(k)})_{k = 1}^{\infty} \)在度量\( d_{l^1} \)下不收敛。
证明任意序列如果在度量\( d_{l^1} \)下收敛的话,则在度量\( d_{l^{\infty}} \)下也收敛:
\( \forall X \)下的序列\( (x^{(k)})_{k = 1}^{\infty} \),如果该序列在度量\( d_{l^1} \)下收敛于某个\( b = (b_n)_{n = 0}^{\infty} \in X \)的话,则\( \forall k \geq 1 \),我们记\( x^{(k)} = (x_n^{(k)})_{n = 0}^{\infty} \),此时由收敛可得,\( \forall \epsilon > 0, \exists K \geq 1 \)满足 \( \forall k \geq K, d_{l^1}((x_n^{(k)})_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) = \sum_{n = 0}^{\infty} |x_n^{(k)} - b_n| \leq \epsilon \),而\( \sup_{n \in \mathbf{N}} |x_n^{(k)} - b_n| \leq \sum_{n = 0}^{\infty} |x_n^{(k)} - b_n| \),可得\( d_{l^{\infty}}((x_n^{(k)})_{n = 0}^{\infty}, (b_n)_{n = 0}^{\infty}) = \sup_{n \in \mathbf{N}} |x_n^{(k)} - b_n| \leq \sum_{n = 0}^{\infty} |x_n^{(k)} - b_n| \leq \epsilon \),综上,有\( (x^{(k)})_{k = 1}^{\infty} \)在度量\( d_{l^{\infty}} \)下也收敛,且收敛于同一个“点”\( b \),证毕。
练习1.1.16
题目:
Let \( (x_n)_{n = 1}^{\infty} \) and \( (y_n)_{n = 1}^{\infty} \) be two sequences in a metric space \( (X, d) \). Suppose that \( (x_n)_{n = 1}^{\infty} \) converges to a point \( x \in X \), and \( (y_n)_{n = 1}^{\infty} \) converges to a point \( y \in X \). Show that \( \lim_{n \to \infty} d(x_n, y_n) = d(x, y) \). (Hint: use the triangle inequality several times.)
证明:
由\( (x_n)_{n = 1}^{\infty} \)收敛于\( x \)可得, \( \forall \epsilon > 0, \exists N_1 \geq 1, \forall n \geq N_1, d(x_n, x) \leq \dfrac{\epsilon}{2} \),而由\( (y_n)_{n = 1}^{\infty} \)收敛于\( y \)可得, \( \forall \epsilon > 0, \exists N_2 \geq 1, \forall n \geq N_2, d(y_n, y) \leq \dfrac{\epsilon}{2} \),取\( N = max(N_1, N_2) \),则\( \forall n \geq N \),有\( d(x_n, y_n) - d(x, y) \leq (d(x_n, x) + d(x, y_n)) - d(x, y) \leq d(x_n, x) + (d(x, y) + d(y, y_n)) - d(x, y) = d(x_n, x) + d(y, y_n) = \epsilon \),还有\( d(x, y) - d(x_n, y_n) \leq (d(x, x_n) + d(x_n, y)) - d(x_n, y_n) \leq d(x, x_n) + (d(y, y_n) + d(y_n, x_n)) - d(x_n, y_n) = d(x, x_n) + d(y, y_n) = \epsilon \),综上,有\( d(x_n, y_n) - d(x, y) \leq \epsilon, d(x, y) - d(x_n, y_n) \leq \epsilon \),而由\( d(x, y) - d(x_n, y_n) \leq \epsilon \),可得\( d(x_n, y_n) - d(x, y) \geq -\epsilon \),再加上\( d(x_n, y_n) - d(x, y) \leq \epsilon \),有\( |d(x_n, y_n) - d(x, y)| \leq \epsilon \),至此,我们有了结论:\( \lim_{n \to \infty} d(x_n, y_n) = d(x, y) \)。
证毕。
章节1.2
练习1.2.1
题目:
Verify the claims in Example 1.2.8.
Example 1.2.8的内容:
When we give a set \( X \) the discrete metric \( d_{\text{disc}} \), and \( E \) is any subset of \( X \), then every element of \( E \) is an interior point of \( E \), every point not contained in \( E \) is an exterior point of \( E \), and there are no boundary points; see Exercise 1.2.1.
证明:
证明\( \forall x \in E \),\( x \)均为\( E \)的内点:
\( \forall x \in E \),令\( r = \dfrac{1}{2} \),则\( B(x, r) = \{ y \in X : d(y, x) < r \} = \{ x \} \),可得\( B(x, r) \cap E = \{ x \} \subseteq E \),这意味着\( x \)为\( E \)的内点。
证明\( \forall x \in X \setminus E \),\( x \)均为\( E \)的外点:
\( \forall x \in X \setminus E \),令\( r = \dfrac{1}{2} \),则\( B(x, r) = \{ y \in X : d(y, x) < r \} = \{ x \} \),可得\( B(x, r) \cap E = \empty \)(因为\( x \notin E \)),这意味着\( x \)为\( E \)的外点。
证明\( X \)中没有边界点:
\( \forall x \in X \),如果\( x \in E \),则\( x \)为内点,不是边界点,如果\( x \notin E \),则\( x \in X \setminus E \),此时\( x \)为外点,不是边界点。综上,\( X \)中没有边界点。
证毕。
练习1.2.2
题目:
Prove Proposition 1.2.10. (Hint: for some of the implications one will need the axiom of choice, as in Lemma 8.4.5 from Analysis I.)
Proposition 1.2.10的内容:
Let \( (X, d) \) be a metric space, let \( E \) be a subset of \( X \), and let \( x_0 \) be a point in \( X \). Then the following statements are logically equivalent.
- \( x_0 \) is an adherent point of \( E \).
- \( x_0 \) is either an interior point or a boundary point of \( E \).
- There exists a sequence \( (x_n)_{n = 1}^{\infty} \) in \( E \) which converges to \( x_0 \) with respect to the metric \( d \).
证明:
证明\( 1 \rightarrow 2 \):
为了证明\( x_0 \)是\( E \)的内点或者边界点,我们仅仅需要证明 \( x_0 \)不是\( E \)的外点即可。
假设\( x_0 \)是\( E \)的外点,则\( \exists r > 0 \)满足 \( B(x_0, r) \cap E = \empty \),然而这和\( x_0 \)为\( E \)的附着点矛盾,因此假设不成立,有\( x_0 \)不是\( E \)的内点。
证明\( 2 \rightarrow 3 \):
由\( x_0 \)是\( E \)的内点或者边界点,可得\( x_0 \)不是\( E \)的外点,进而可得\( \forall r > 0 \),均有\( B(x_0, r) \cap E \neq \empty \),于是\( \forall n \geq 1 \),可得\( B(x_0, \dfrac{1}{n}) = \{ x \in X : d(x, x_0) < \dfrac{1}{n} \} \cap E \neq \empty \),任取\( B(x_0, \dfrac{1}{n}) \)中的任意一个元素作为序列\( (x_n)_{n = 1}^{\infty} \)的第\( n \)个元素\( x_n \)(注:这里我们隐式地用到了选择公理),有\( x_n \in E \)。
我们现在证明\( (x_n)_{n = 1}^{\infty} \)收敛于\( x_0 \), \( \forall \epsilon > 0, \exists N \geq 1 \)满足\( \dfrac{1}{N} < \epsilon \),进而\( \forall n \geq N \),有\( d(x_n, x_0) < \dfrac{1}{n} \leq \dfrac{1}{N} < \epsilon \),这意味着\( (x_n)_{n = 1}^{\infty} \)收敛于\( x_0 \)。
证明\( 3 \rightarrow 1 \):
由\( (x_n)_{n = 1}^{\infty} \)收敛于\( x_0 \),可得\( \forall r > 0 \),\( \exists N \geq 1 \)满足\( \forall n \geq N, d(x_n, x_0) \leq \dfrac{r}{2} < r \),可得\( x_N \in B(x_0, r) \),又\( x_N \in E \),这意味着\( B(x_0, r) \cap E \neq \empty \),综上,有\( x_0 \)是\( E \)的附着点。
证毕。
练习1.2.3
题目:
Prove Proposition 1.2.15. (Hint: you can use earlier parts of the proposition to prove later ones.)
Proposition 1.2.15的内容:
(Basic properties of open and closed sets) Let \( (X, d) \) be a metric space.
- Let \( E \) be a subset of \( X \). Then \( E \) is open if and only if \( E = \text{int}(E) \). In other words, \( E \) is open if and only if for every \( x \in E \), there exists an \( r > 0 \) such that \( B(x, r) \subseteq E \).
- Let \( E \) be a subset of \( X \). Then \( E \) is closed if and only if \( E \) contains all its adherent points. In other words, \( E \) is closed if and only if for every convergent sequence \( (x_n)_{n = m}^{\infty} \) in \( E \), the \( \lim_{n \to \infty} x_n \) of that sequence also lies in \( E \).
- For any \( x_0 \in X \) and \( r > 0 \), the ball \( B(x_0, r) \) is an open set. The set \( \{ x \in X : d(x, x_0) \leq r \} \) is a closed set. (This set is sometimes called the closed ball of radius \( r \) centered at \( x_0 \).)
- Any singleton set \( \{ x_0 \} \), where \( x_0 \in X \), is automatically closed.
- If \( E \) is a subset of \( X \), then \( E \) is open if and only if the complement \( X \setminus E := \{ x \in X : x \notin E \} \) is closed.
- If \( E_1, \dots, E_n \) is a finite collection of open sets in \( X \), then \( E_1 \cap E_2 \cap \dots \cap E_n \) is also open. If \( F_1, \dots, F_n \) is a finite collection of closed sets in \( X \), then \( F_1 \cup F_2 \cup \dots \cup F_n \) is also closed.
- If \( \{ E_{\alpha} \}_{\alpha \in I} \) is a collection of open sets in \( X \) (where the index set \( I \) could be finite, countable, or uncountable), then the union \( \bigcup_{\alpha \in I} E_{\alpha} := \{ x \in X : x \in E_{\alpha} \text{ for some } \alpha \in I \} \) is also open. If \( \{ F_{\alpha} \}_{\alpha \in I} \) is a collection of closed sets in \( X \), then the intersection \( \bigcap_{\alpha \in I} F_{\alpha} := \{ x \in X : x \in F_{\alpha} \text{ for all } \alpha \in I \} \) is also closed.
- If \( E \) is any subset of \( X\), then \( \text{int}(E) \) is the largest open set which is contained in \( E \); in other words, \( \text{int}(E) \) is open, and given any other open set \( V \subseteq E \), we have \( V \subseteq \text{int}(E) \). Similarly \( \overline{E} \) is the smallest closed set which contains \( E \); in other words, \( \overline{E} \) is closed, and given any other closed set \( K \supseteq E \), \( K \supseteq \overline{E} \).
证明:
证明1:
必要性:
由\( E \)是开集,可得\( \forall x \in E \),有\( x \)不是\( E \)的边界点, \( x \)也不可能是\( E \)的外点,因为\( E \)的外点要求\( x \notin E \),剩下的唯一可能性就是\( x \)是\( E \)的内点了,于是有\( x \in \text{int}(E) \)了,至此可得\( E \subseteq \text{int}(E) \);而\( \text{int}(E) \subseteq E \)是显然的(因为所有\( E \)的内点都得是\( E \)的成员);综上,有\( E = \text{int}(E) \)。
充分性:
假设\( E \)不是开集,则\( E \)包含至少一个边界点\( x_0 \in \partial E \),但这意味着\( x_0 \notin \text{int}(E) \),可得\( E \)不是\( \text{int}(E) \)的子集,然而这和\( E = \text{int}(E) \)矛盾,因此假设不成立,有\( E \)是开集。
证明2:
我们先证明2的前半段命题。
必要性(前半段命题):
针对\( E \)的所有附着点\( x \),由定理1.2.10可得,\( x \)为内点或者边界点,如果\( x \)为内点,则直接有\( x \in E \),如果\( x \)为边界点,则由\( E \)是闭集,可得\( x \in E \),综上,有\( E \)包含自身所有的附着点。
充分性(前半段命题):
由定理1.2.10可得,\( E \)的所有边界点均是\( E \)的附着点,而\( E \)包含它所有的边界点,这意味着它也包含它所有的附着点,也就是说\( E \)是闭集。
我们最后证明2的后半段命题。
必要性(后半段命题):
针对所有\( E \)上的收敛序列\( (x_n)_{n = m}^{\infty} \),记它收敛于\( x \),我们得证明\( x \in E \),易得\( x \)是\( E \)的附着点,进而由定理1.2.10可得, \( x \)是内点或者边界点,如果\( x \)是内点,则直接有\( x \in E \),如果\( x \)是边界点,则由\( E \)为闭集,也可得\( x \in E \)。
充分性(后半段的命题):
假设\( E \)不封闭,则存在\( E \)不包含的边界点\( x_0 \),进而由定理1.2.10,可得\( x_0 \)为\( E \)的附着点且存在\( E \)上的序列\( (x_n)_{n = m}^{\infty} \)收敛于\( x_0 \),但这意味着\( x_0 \in E \),矛盾,因此假设不成立,有\( E \)为封闭集。
证明3:
先证明\( B(x_0, r) \)为开集:
直接证明它里面的所有元素都是内点就行了, \( \forall x \in B(x_0, r) \),我们有\( x \in X, d(x, x_0) < r \),令\( r' = r - d(x, x_0) \),则\( \forall y \in B(x, r') \),有\( y \in X, d(y, x_0) \leq d(y, x) + d(x, x_0) \leq r' + d(x, x_0) = r - d(x, x_0) + d(x, x_0) = r \),可得\( B(x, r') \subseteq B(x_0, r) \),这意味着\( x \)是\( B(x_0, r) \)的内点。
最后证明\( \{ x \in X : d(x, x_0) \leq r \} \)为闭集:
令\( E = \{ x \in X : d(x, x_0) \leq r \} \),假设\( E \)不是闭集,则存在\( E \)的边界点\( z \)不属于\( E \),可得\( z \in X, d(z, x_0) > r \),令\( r' = d(z, x_0) - r \),则\( \forall y \in B(z, r') \), \( y \in X, d(y, z) < r' \),此时假设\( y \in E \),则\( d(y, x_0) \leq r \),而\( d(z, x_0) \leq d(z, y) + d(y, x_0) < r' + r = d(z, x_0) \),有\( d(z, x_0) < d(z, x_0) \)矛盾,因此假设不成立,有\( y \notin E \),综上,有\( B(z, r') \cap E = \empty \),也就是说\( z \)是\( E \)的外点,不是\( E \)的边界点,这和一开始说的\( z \)为\( E \)的边界点矛盾,因此一开始的假设不成立,有不存在\( E \)的边界点\( z \)不属于\( E \),换句话来说就是\( E \)包含它所有的边界点,可得\( E \)是闭集。
证明4:
针对所有\( \{ x_0 \} \)上的序列\( (b_n)_{n = m}^{\infty} \),有\( \forall n \geq m, b_n = x_0 \),可得\( (b_n)_{n = m}^{\infty} \)收敛于\( x_0 \in E \),再由2可得,\( \{ x_0 \} \)是闭集。
证明5:
从内点的定义,易得\( x \)是\( E \)的内点当且仅当\( x \)是\( X \setminus E \)的外点,也就是\( \text{int}(E) = \text{ext}(X \setminus E) \),同理,也易得\( x \)是\( E \)的外点当且仅当\( x \)是\( X \setminus E \)的内点,也就是\( \text{ext}(E) = \text{int}(X \setminus E) \) (注:这意味着\( \text{int} \)和\( \text{ext} \)在差集操作下是对偶的),于是可得\( \partial (X \setminus E) = X \setminus \text{int}(X \setminus E) \setminus \text{ext}(X \setminus E) = X \setminus \text{ext}(E) \setminus \text{int}(E) = \partial E \)。
必要性:
由\( E \)是开集以及1,可得\( E = \text{int}(E) \),进而可得\( X \setminus E = X \setminus \text{int}(E) = \partial E \cup \text{ext}(E) = \partial (X \setminus E) \cup \text{int}(X \setminus E) \),特别的,有\( X \setminus E \)包含它所有的边界点,于是有\( X \setminus E \)为开集。
充分性:
由\( X \setminus E \)是闭集,可得\( \partial (X \setminus E) \subseteq X \setminus E \),又\( \partial (X \setminus E) = \partial E \),于是有\( \partial E \subseteq X \setminus E \),可得\( \forall x \in \partial E, x \notin E \),也就是说\( E \)不包含它的任意边界点,这意味着\( E \)是开集。
证明6:
先证明\( E_1 \cap E_2 \cap \dots \cap E_n \)为开集:
\( \forall x \in E_1 \cap E_2 \cap \dots \cap E_n, \forall 1 \leq i \leq n \),由\( E_i \)为开集,可得\( \exists r_i > 0 \)满足\( B(x, r_i) \subseteq E_i \),取\( r = \min \{ r_i : 1 \leq i \leq n \} \),则有\( \forall 1 \leq i \leq n, B(x, r) \subseteq E_i \),可得\( B(x, r) \subseteq E_1 \cap E_2 \cap \dots \cap E_n \),此时再由1,可得\( E_1 \cap E_2 \cap \dots \cap E_n \)为开集。
证明完有限个开集的交集仍然是开集,容易联想到,是否无限个开集的交集也仍然是开集呢?答案是否定的,举个反例即可,\( \bigcap_{n = 1}^{\infty} (-\dfrac{1}{n}, \dfrac{1}{n}) = \{ 0 \} \),在常规的实数集度量下,有\( \forall n \geq 1, (-\dfrac{1}{n}, \dfrac{1}{n}) \)均是开集,但是\( \{ 0 \} \)并不是开集,它包含了自身的边界点\( 0 \)。
接着证明\( F_1 \cup F_2 \cup \dots \cup F_n \)为闭集:
\( \forall 1 \leq i \leq n \),由\( F_i \)为闭集以及5,可得\( X \setminus F_i \)为开集,进而可得\( \bigcap_{i = 1}^{n} X \setminus F_i \)为开集,而\( F_1 \cup F_2 \cup \dots \cup F_n = X \setminus (\bigcap_{i = 1}^{n} X \setminus F_i) \),再由5可得,\( F_1 \cup F_2 \cup \dots \cup F_n \)为闭集。
同上,证明完有限个闭集的并集仍然闭集,容易联想到,是否无限个闭集的并集也仍然是闭集呢?答案同样是否定的,举个反例即可,令\( (0, 1) = \bigcup_{x \in (0, 1)} \{ x \} \),这里由\( 4 \)可得,\( \forall x \in (0, 1), \{ x \} \)均为闭集,但是这些闭集的并集是\( (0, 1) \),在常规的实数集度量下,\( (0, 1) \)并不是闭集,它不包含自身的所有边界点\( \{ 0, 1 \} \)。
证明7:
先证明多个开集的并集仍然是开集:
我们通过1来证明,\( \forall x \in \bigcup_{\alpha \in I} E_{\alpha} \),有\( \exists \alpha \in I, x \in E_{\alpha} \),这里\( E_{\alpha} \)是开集,由1可得,\( \exists r > 0, B(x, r) \subseteq E_{\alpha} \subseteq \bigcup_{\alpha \in I} E_{\alpha} \),进而言之,\( \forall x \in \bigcup_{\alpha \in I} E_{\alpha}, \exists r > 0, B(x, r) \subseteq \bigcup_{\alpha \in I} E_{\alpha} \),再由1,可得\( \bigcup_{\alpha \in I} E_{\alpha} \)为开集。
接着证明多个闭集的交集仍然是闭集:
我们通过2来证明,针对所有\( \bigcap_{\alpha \in I} F_{\alpha} \)上的收敛序列\( (x_n)_{n = m}^{\infty} \),我们有\( \forall \alpha \in I \),该序列为\( F_{\alpha} \)上的序列,进而由2,可得\( \lim_{n \to \infty} x_n \in F_{\alpha} \),至此有\( \lim_{n \to \infty} x_n \in \bigcap_{\alpha \in I} F_{\alpha} \),再由2,可得\( \bigcap_{\alpha \in I} F_{\alpha} \)为闭集。
证明8:
证明\( \text{int}(E) \)是开集:
我们通过1来证明, \( \forall x \in \text{int}(E) \),可得\( \exists r > 0, B(x, r) \subseteq E \), \( \forall y \in B(x, \dfrac{r}{2}) \),易得\( B(y, \dfrac{r}{4}) \subseteq B(x, r) \),进而可得\( B(y, \dfrac{y}{4}) \subseteq E \),有\( y \)为\( E \)的内点,综上有\( B(x, \dfrac{r}{2}) \)中所有点均是\( E \)的内点,即\( B(x, \dfrac{r}{2}) \subseteq \text{int}(E) \),此时由1可得,\( \text{int}(E) \)是开集。
证明\( \text{int}(E) \)是所有被包含在\( E \)中的开集中最大的,即\( \text{int}(E) \)是\( E \)下最大的开集:
\( \forall \)开集\( V \subseteq E \),我们得证明\( V \subseteq \text{int}(E) \),即得证明\( V \)中所有元素均是\( E \)的内点, \( \forall x \in V \),由\( V \)是开集,可得\( \exists r > 0, B(x, r) \subseteq V \),而\( V \subseteq E \),可得\( B(x, r) \subseteq E \),至此有\( x \)为\( E \)的内点,于是有\( x \in \text{int}(E) \)。
证明\( \overline{E} \)是闭集:
我们有\( \text{ext}(E) = \text{int}(X \setminus E) \),这里\( \text{int}(X \setminus E) \)通过刚才的证明可得是开集,于是\( \text{ext}(E) \)也是开集,而\( \overline{E} = X \setminus \text{ext}(E) \) (即\( \overline{E} = \text{int}(E) \cup \partial E \),这点由定理1.2.10(b)可得),此时通过5可得,\( \overline{E} = X \setminus \text{ext}(E) \)是闭集。
证明\( \overline{E} \)是所有包含\( E \)的闭集中最小的:
给定任意闭集\( K \supseteq E \), \( \forall x \in \overline{E} \),我们证明\( x \in K \),这里\( x \)为\( E \)的附着点,可得\( \forall r > 0, B(x, r) \cap E \neq \empty \),而\( K \supseteq E \),于是也有\( B(x, r) \cap K \neq \empty \),综上有\( x \)也为\( K \)的附着点,而\( K \)是闭集,它包含自身所有的附着点,可得\( x \in K \),至此可得\( X \subseteq K \),即\( K \supseteq X \)。
证毕。
练习1.2.4
题目:
Let \( (X, d) \) be a metric space, \( x_0 \) be a point in \( X \), and \( r > 0 \). Let \( B \) be the open ball \( B := B(x_0, r) = \{ x \in X : d(x, x_0) < r \} \), and let \( C \) be the closed ball \( C := \{ x \in X : d(x, x_0) \leq r \} \).
- Show that \( \overline{B} \subseteq C \).
- Give an example of a metric space \( (X, d) \), a point \( x_0 \), and a radius \( r > 0 \) such that \( \overline{B} \) is not equal to \( C \).
解答:
证明1:
\( \forall x \in \overline{B} \),假设\( d(x, x_0) > r \),则令\( r' = d(x, x_0) - r \),则\( \forall y \in B(x, r') \),有\( d(y, x) < r' \),假设\( d(y, x_0) < r \),则\( d(x, x_0) \leq d(x, y) + d(y, x_0) < r' + r = d(x, x_0) \),也就是\( d(x, x_0) < d(x, x_0) \),矛盾,因此假设不成立,有\( d(y, x_0) \geq r \),于是可得\( y \notin B \),综上有\( B(x, r') \cap B = \empty \),这意味着\( x \)不是\( B \)的附着点,有\( x \notin \overline{B} \),矛盾,因此一开始的假设不成立,有\( d(x, x_0) \leq r \),至此可得\( x \in C \),综上,有\( \overline{B} \subseteq C \)。
证毕。
解答2:
令\( X \)为任意有两个或者两个以上元素的集合,令\( d = d_{\text{disc}} \),任取\( x_0 \in X \),令\( r = 1, B = B(x_0, r) = \{ x_0 \}, C = \{ x \in X : d(x, x_0) \leq r \} = X \),由定理1.2.15(d)可得,\( B = \{ x_0 \} \)是闭集,再由\( \overline{B} \)是最小的闭集,可得\( \overline{B} = B \),而\( C = X \)且\( X \)有两个或者以上的元素,但\( \overline{B} \)只有一个元素,有\( \overline{B} \neq C \)。
章节1.3
练习1.3.1
题目:
Prove Proposition 1.3.4(2).
Proposition 1.3.4的内容:
Let \( (X, d) \) be a metric space, let \( Y \) be a subset of \( X \), and let \( E \) be a subset of \( Y \).
- \( E \) is relatively open with respect to \( Y \) if and only if \( E = V \cap Y \) for some set \( V \subseteq X \) which is open in \( X \).
- \( E \) is relatively closed with respect to \( Y \) if and only if \( E = K \cap Y \) for some set \( K \subseteq X \) which is closed in \( X \).
证明:
必要性:
\( E \)相对\( Y \)是相对闭集当且仅当\( Y \setminus E \)相对\( Y \)是相对开集,进而由定理1.3.4(1),可得\( E \)相对\( Y \)是相对闭集当且仅当\( \exists X \)下的开集\( V \subseteq X \)满足\( Y \setminus E = V \cap Y \),令\( K = X \setminus V \),我们证明\( E = K \cap Y \)。
先证明\( K \cap Y \subseteq E \),\( \forall x \in K \cap Y \),可得\( x \in K, x \in Y \),由\( K \)的定义,可得\( x \in X, x \notin V \),于是有\( x \notin V \cap Y = Y \setminus E \),这里\( x \notin Y \setminus E \)加上\( x \in Y \),意味着\( x \in E \),综上,有\( K \cap Y \subseteq E \)。
接着证明\( E \subseteq K \cap Y \),\( \forall x \in E \),由\( E \subseteq Y \subseteq X \),可得\( x \in X \),另外由\( x \in E \),可得\( x \notin Y \setminus E = V \cap Y \),这意味着\( x \notin V \)或者\( x \notin Y \),但是\( E \subseteq Y \),我们知道\( x \in Y \),于是只剩下\( x \notin V \)这个可能性,加上之前得到的\( x \in X \),可得\( x \in X \setminus V = K \),至此加上已有的\( x \in Y \),可得\( x \in K \cap Y \),综上,有\( E \subseteq K \cap Y \)。
至此可得,\( E = K \cap Y \)。
充分性:
由\( \exists X \)下的闭集\( K \subseteq X \)满足\( E = K \cap Y \),可得\( X \setminus K \)为\( X \)下的开集,令\( V = X \setminus K \),由定理1.3.4(1),可得\( V \cap Y \)相对\( Y \)是相对开集,进而可得,\( Y \setminus (V \cap Y) = (Y \setminus V) \cup (Y \setminus Y) = (Y \setminus V) \cup \empty = Y \setminus V \)相对\( Y \)是相对闭集。
我们现在证明\( E = Y \setminus V \),从而说明\( E \)相对\( Y \)是相对闭集:
先证明\( E \subseteq Y \setminus V \): \( \forall x \in E = K \cap Y \),我们有\( x \in K, x \in Y \),这里由\( x \in K \),可得\( x \notin X \setminus K = V \),加上\( x \in Y \),可得\( x \in Y \setminus V \),综上,有\( E \subseteq Y \setminus V \)。
接着证明\( Y \setminus V \subseteq E \): \( \forall x \in Y \setminus V \),有\( x \in Y, x \notin V \),由\( x \notin V = X \setminus K \),可得\( x \notin X \)或者\( x \in K \),又\( Y \subseteq X \),可得\( x \in X \),于是只剩下\( x \in K \)这种可能性,加上\( x \in Y \),可得\( x \in K \cap Y = E \),综上,有\( Y \setminus V \subseteq E \)。
至此可得,\( E = Y \setminus V \),加上\( Y \setminus V \)相对\( Y \)是相对闭集,可得,\( E \)相对\( Y \)是相对闭集。
证毕。
章节1.4
练习1.4.1
题目:
Prove Lemma 1.4.3. (Hint: review your proof of Proposition 6.6.5.)
Lemma 1.4.3的内容:
Let \( (x^{(n)})_{n = m}^{\infty} \) be a sequence in \( (X, d) \) which converges to some limit \( x_0 \). Then every subsequence \( (x^{(n_j)})_{j = 1}^{\infty} \) of that sequence also converges to \( x_0 \).
证明:
\( \forall (x^{(n)})_{n = m}^{\infty} \)的子序列\( (x^{(n_j)})_{j = 1}^{\infty} \), \( \forall \epsilon > 0 \),由\( (x^{(n)})_{n = m}^{\infty} \)收敛于\( x_0 \),可得\( \exists N \geq m, \forall n \geq N \),有\( d(x^{(n)}, x_0) \leq \epsilon \),而\( \forall j \geq N \),有\( n_j \geq n_N \geq N \),可得\( d(x^{(n_j)}, x_0) \leq \epsilon \),综上,有\( (x^{(n_j)})_{j = 1}^{\infty} \)收敛于\( x_0 \)。
证毕。
练习1.4.2
题目:
Prove Proposition 1.4.5. (Hint: review your proof of Proposition 6.6.6.)
Proposition 1.4.5的内容:
Let \( (x^{(n)})_{n = m}^{\infty} \) be a sequence of points in a metric space \( (X, d) \), and let \( L \in X \). Then the following are equivalent:
- \( L \) is a limit point of \( (x^{(n)})_{n = m}^{\infty} \).
- There exists a subsequence \( (x^{(n_j)})_{j = 1}^{\infty} \) of the original sequence \( (x^{(n)})_{n = m}^{\infty} \) which converges to \( L \).
证明:
证明\( 1 \Longrightarrow 2 \):
我们递归定义子序列\( (x^{(n_j)})_{j = 1}^{\infty} \),归纳假设\( \forall 1 \leq j' < j \),我们有\( n_{j'} \)已定义且满足\( d(x^{(n_{j'})}, L) \leq \dfrac{1}{j'}, m \leq n_1 < n_2 < \dots < n_{j'} \),我们定义\( n_j \),如果\( j = 1 \),则令\( N_j = m \),否则的话,有\( j > 1 \),此时\( \{ k : 1 \leq k < j \} \)非空,令\( N_j = \max \{ n_k : 1 \leq k < j \} + 1 \),由\( L \)是\( (x^{(n)})_{n = m}^{\infty} \)的极限点,可得\( \exists n \geq N_j \),使得\( d(x^{(n)}, L) \leq \dfrac{1}{j} \),令\( n_j = n \),有\( d(x^{(n_j)}, L) \leq \dfrac{1}{j} \),以及\( m \leq n_1 < n_2 < \dots < n_j \),递归定义完毕。
现在证明\( (x^{(n_j)})_{j = 1}^{\infty} \)收敛于\( L \), \( \forall \epsilon > 0, \exists j \geq 1 \)满足\( \dfrac{1}{j} \leq \epsilon \),于是\( \forall k \geq j \),有\( d(x^{(n_k)}, L) \leq \dfrac{1}{k} \leq \dfrac{1}{j} \leq \epsilon \),综上,有子序列\( (x^{(n_j)})_{j = 1}^{\infty} \)收敛于\( L \)。
证明\( 2 \Longrightarrow 1 \):
\( \forall \epsilon > 0, \forall N \geq m \),由子序列\( (x^{(n_j)})_{j = 1}^{\infty} \)收敛于\( L \),可得\( \exists K \geq 1 \)满足\( \forall k \geq K, d(x^{(n_k)}, L) \leq \epsilon \),取\( j = \max(K, N) \),则有\( n_j \geq j \geq N \)以及\( d(x^{(n_j)}, L) \leq \epsilon \),综上,有\( L \)为\( (x^{(n)})_{n = m}^{\infty} \)的极限点。
证毕。
练习1.4.3
题目:
Prove Lemma 1.4.7. (Hint: review your proof of Proposition 6.1.12.)
Lemma 1.4.7的内容:
(Convergent sequences are Cauchy sequences) Let \( (x^{(n)})_{n = m}^{\infty} \) be a sequence in \( (X, d) \) which converges to some limit \( x_0 \). Then \( (x^{(n)})_{n = m}^{\infty} \) is also a Cauchy sequence.
证明:
\( \forall \epsilon > 0 \),由\( (x^{(n)})_{n = m}^{\infty} \)收敛于\( x_0 \),可得\( \exists N \geq m, \forall n \geq N, d(x^{(n)}, x_0) \leq \dfrac{\epsilon}{3} < \dfrac{\epsilon}{2} \),进而可得\( \forall j, k \geq N \),有\( d(x^{(j)}, x^{(k)}) \leq d(x^{(j)}, x_0) + d(x_0, x^{(k)}) < \dfrac{\epsilon}{2} + \dfrac{\epsilon}{2} = \epsilon \),综上,有\( (x^{(n)})_{n = m}^{\infty} \)为柯西序列。
证毕。
练习1.4.4
题目:
Prove Lemma 1.4.9.
Lemma 1.4.9的内容:
Let \( (x^{(n)})_{n = m}^{\infty} \) be a Cauchy sequence in \( (X, d) \). Suppose that there is some subsequence \( (x^{(n_j)})_{j = 1}^{\infty} \) of this sequence which converges to a limit \( x_0 \) in \( X \). Then the original sequence \( (x^{(n)})_{n = m}^{\infty} \) also converges to \( x_0 \).
证明:
\( \forall \epsilon > 0 \),由\( (x^{(n_j)})_{j = 1}^{\infty} \)收敛于\( x_0 \),可得\( \exists N_1 \geq 1, \forall j \geq N_1, d(x^{(n_j)}, x_0) \leq \dfrac{\epsilon}{2} \),再由\( (x^{(n)})_{n = m}^{\infty} \)是柯西序列,可得\( \exists N_2 \geq m, \forall j, k \geq N_2, d(x^{(j)}, x^{(k)}) < \dfrac{\epsilon}{2} \),取\( N = \max(N_1, N_2) \),则\( \forall n \geq N \),有\( d(x^{(n)}, x_0) \leq d(x^{(n)}, x^{(n_{N_1})}) + d(x^{(n_{N_1})}, x_0) < \dfrac{\epsilon}{2} + \dfrac{\epsilon}{2} = \epsilon \),综上,有\( (x^{(n)})_{n = m}^{\infty} \)也收敛于\( x_0 \)。
证毕。
练习1.4.5
题目:
Let \( (x^{(n)})_{n = m}^{\infty} \) be a sequence of points in a metric space \( (X, d) \), and let \( L \in X \). Show that if \( L \) is a limit point of the sequence \( (x^{(n)})_{n = m}^{\infty} \), then L is an adherent point of the set \( \{ x^{(n)} : n \geq m \} \). Is the converse true?
解答:
证明如果是上述序列的极限点,则是上述集合附着点:
\( \forall r > 0 \),由\( L \)是\( (x^{(n)})_{n = m}^{\infty} \)的极限点,可得\( \exists n \geq m \),满足\( d(x^{(n)}, L) \leq \dfrac{r}{2} < r \),加上\( x^{(n)} \in X \),可得\( x^{(n)} \in B(L, r) \),又明显有\( x^{(n)} \in \{ x^{(n)} : n \geq m \} \),可得\( x^{(n)} \in B(L, r) \cap \{ x^{(n)} : n \geq m \} \),进而可得,\( B(L, r) \cap \{ x^{(n)} : n \geq m \} \neq \empty \),综上,有\( L \)为\( \{ x^{(n)} : n \geq m \} \)的附着点。
证毕。
反命题为真吗?
不为真,举个例子,令\( x^{(n)} = \dfrac{1}{n} \),则\( \dfrac{1}{1} = 1 \)为序列的附着点,但是\( \dfrac{1}{1} = 1 \)明显不是序列的极限点。
练习1.4.6
题目:
Show that every Cauchy sequence can have at most one limit point.
证明:
给定任意柯西序列\( (x^{(n)})_{n = m}^{\infty} \),假设它有两个或者两个以上的极限点,任取它的两个极限点\( x_1, x_2 \), \( \forall \epsilon > 0 \),由\( (x^{(n)})_{n = m}^{\infty} \)是柯西序列,可得\( \exists N \geq m, \forall j, k \geq N, d(x^{(j)}, x^{(k)}) < \dfrac{\epsilon}{3} \),再由\( x_1 \)是\( (x^{(n)})_{n = m}^{\infty} \)的极限点,可得\( \exists n_1 \geq N, d(x^{(n_1)}, x_1) \leq \dfrac{\epsilon}{3} \),同理由\( x_2 \)是\( (x^{(n)})_{n = m}^{\infty} \)的极限点,可得\( \exists n_2 \geq N, d(x^{(n_2)}, x_2) \leq \dfrac{\epsilon}{3} \),由\( n_1 \geq N, n_2 \geq N \),可得\( d(x^{(n_1)}, x^{(n_2)}) < \epsilon \),于是可得\( d(x_1, x_2) \leq d(x_1, x^{(n_1)}) + d(x^{(n_1)}, x_2) \leq d(x_1, x^{(n_1)}) + d(x^{(n_1)}, x^{(n_2)}) + d(x^{(n_2)}, x_2) < \dfrac{\epsilon}{3} + \dfrac{\epsilon}{3} + \dfrac{\epsilon}{3} = \epsilon \),简而言之,\( \forall \epsilon > 0 \),有\( d(x_1, x_2) < \epsilon \),可得\( x_1 = x_2 \)。
证毕。
练习1.4.7
题目:
Prove Proposition 1.4.12.
Proposition 1.4.12的内容:
- Let \( (X, d) \) be a metric space, and let \( (Y, d|_{Y \times Y}) \) be a subspace of \( (X, d) \). If \( (Y, d|_{Y \times Y}) \) is complete, then \( Y \) must be closed in \( X \).
- Conversely, suppose that \( (X, d) \) is a complete metric space, and \( Y \) is a closed subset of \( X \). Then the subspace \( (Y, d|_{Y \times Y}) \) is also complete.
证明:
证明1:
\( \forall (Y, d|_{Y \times Y}) \)上的收敛序列\( (x^{(n)})_{n = m}^{\infty} \),由引理1.4.7,可得\( (x^{(n)})_{n = m}^{\infty} \)是柯西序列,进而由\( (Y, d|_{Y \times Y}) \)是完备度量空间,可得\( \lim_{n \to \infty} x^{(n)} \in Y \),此时由\( d \)和\( d|_{Y \times Y} \)在\( Y \times Y \)上一致,可得在\( (X, d) \)上,也有\( (x^{(n)})_{n = m}^{\infty} \)收敛且\( \lim_{n \to \infty} x^{(n)} \in Y \),此时根据定理1.2.15的(b),可得\( Y \)是闭合集。
证明2:
\( \forall (Y, d|_{Y \times Y}) \)上的柯西序列\( (x^{(n)})_{n = m}^{\infty} \),由\( d \)和\( d|_{Y \times Y} \)在\( Y \times Y \)上一致,可得\( (x^{(n)})_{n = m}^{\infty} \)也是\( (X, d) \)上的柯西序列,再由\( (X, d) \)是完备度量空间,可得\( (x^{(n)})_{n = m}^{\infty} \)收敛于\( (X, d) \),此时根据\( Y \)是闭合集以及定理1.2.15的(b),可得\( \lim_{n \to \infty} x^{(n)} \in Y \),再次根据\( d \)和\( d|_{Y \times Y} \)在\( Y \times Y \)上一致,可得在\( (Y, d|_{Y \times Y}) \)上,也有\( (x^{(n)})_{n = m}^{\infty} \)收敛且\( \lim_{n \to \infty} x^{(n)} \in Y \),综上,有\( (Y, d|_{Y \times Y}) \)是完备度量空间。
证毕。
练习1.4.8
题目:
The following construction generalizes the construction of the reals from the rationals in Chap.5, allowing one to view any metric space as a subspace of a complete metric space. In what follows we let \( (X, d) \) be a metric space.
- Given any Cauchy sequence \( (x_n)_{n = 1}^{\infty} \) in \( X \), we introduce the formal limit \( \text{LIM}_{n \to \infty} x_n \). We say that two formal limits \( \text{LIM}_{n \to \infty} x_n \) and \( \text{LIM}_{n \to \infty} y_n \) are equal if \( \lim_{n \to \infty} d(x_n, y_n) \) is equal to zero. Show that this equality relation obeys the reflexive, symmetry, and transitive axioms.
- Let \( \overline{X} \) be the space of all formal limits of Cauchy sequences in \( X \), with the above equality relation. Define a metric \( d_{\overline{X}}: \overline{X} \times \overline{X} \to [0, +\infty) \) by setting \( d_{\overline{X}}(\text{LIM}_{n \to \infty} x_n, \text{LIM}_{n \to \infty} y_n) := \lim_{n \to \infty} d(x_n, y_n) \). Show that this function is well-defined (this means not only that the limit \( \lim_{n \to \infty} d(x_n, y_n) \) exists, but also that the axiom of substitution is obeyed; cf. Lemma 5.3.7) and gives \( \overline{X} \) the structure of a metric space.
- Show that the metric space \( (\overline{X}, d_{\overline{X}}) \) is complete.
- Show that the closure of X in \( \overline{X} \) is \( \overline{X} \) (which explains the choice of notation \( \overline{X} \)).
- Show that the formal limit agrees with the actual limit, thus if \( (x_n)_{n = 1}^{\infty} \) is any Cauchy sequence in \( X \), then we have \( \lim_{n \to \infty} x_n = \text{LIM}_{n \to \infty} x_n \) in \( \overline{X} \).
证明:
证明1:
证明满足自反性:
\( \forall X \)上的柯西序列\( (x_n)_{n = 1}^{\infty} \),我们有\( \lim_{n \to \infty} d(x_n, x_n) = \lim_{n \to \infty} 0 = 0 \),于是有\( \text{LIM}_{n \to \infty} x_n = \text{LIM}_{n \to \infty} x_n \)。
证明满足对称性:
\( \forall X \)上的柯西序列\( (x_n)_{n = 1}^{\infty}, (y_n)_{n = 1}^{\infty} \),如果\( \text{LIM}_{n \to \infty} x_n = \text{LIM}_{n \to \infty} y_n \), 则有\( \lim_{n \to \infty} (d(x_n, y_n) = d(y_n, x_n)) = 0 \),于是有\( \text{LIM}_{n \to \infty} y_n = \text{LIM}_{n \to \infty} x_n \)。
证明满足传递性:
\( \forall X \)上的柯西序列\( (x_n)_{n = 1}^{\infty}, (y_n)_{n = 1}^{\infty}, (z_n)_{n = 1}^{\infty} \),如果\( \text{LIM}_{n \to \infty} x_n = \text{LIM}_{n \to \infty} y_n, \text{LIM}_{n \to \infty} y_n = \text{LIM}_{n \to \infty} z_n \), 则可得\( \lim_{n \to \infty} d(x_n, y_n) = 0, \lim_{n \to \infty} d(y_n, z_n) = 0 \),进而可得\( \lim_{n \to \infty} (d(x_n, y_n) + d(y_n, z_n)) = 0 \),而\( 0 \leq d(x_n, z_n) \leq d(x_n, y_n) + d(y_n, z_n) \),由推论I.6.4.14(Squeeze test),可得\( \lim_{n \to \infty} d(x_n, z_n) = 0 \),于是有\( \text{LIM}_{n \to \infty} x_n = \text{LIM}_{n \to \infty} z_n \)。
证明2:
给定\( X \)上的柯西序列\( (x_n)_{n = 1}^{\infty}, (y_n)_{n = 1}^{\infty} \)。
证明\( \lim_{n \to \infty} d(x_n, y_n) \)存在:
直接利用\( \mathbf{R} \)与它的常规度量共同组成完备度量空间这点,我们证明\( (d(x_n, y_n))_{n = 1}^{\infty} \)是柯西序列就行。
\( \forall \epsilon > 0 \),由\( (x_n)_{n = 1}^{\infty} \)是柯西序列,可得\( \exists N_1 \geq 1, \forall j, k \geq N_1, d(x_j, x_k) < \dfrac{\epsilon}{2} \),由\( (y_n)_{n = 1}^{\infty} \)是柯西序列,可得\( \exists N_2 \geq 1, \forall j, k \geq N_2, d(y_j, y_k) < \dfrac{\epsilon}{2} \),取\( N = \max(N_1, N_2) \),则\( \forall j, k \geq N \),有\( d(d(x_j, y_j), d(x_k, y_k)) = |d(x_j, y_j) - d(x_k, y_k)| \leq |d(x_j, y_j)| + |-d(x_k, y_k)| = d(x_j, y_j) + d(x_k, y_k) < \epsilon \),至此可得,\( (d(x_n, y_n))_{n = 1}^{\infty} \)是柯西序列,进而由\( \mathbf{R} \)与它的常规度量共同组成完备度量空间这点,可得\( (d(x_n, y_n))_{n = 1}^{\infty} \)收敛,即\( \lim_{n \to \infty} d(x_n, y_n) \)存在。
证明\( d_{\overline{X}} \)满足代入公理:
给定\( X \)上的柯西序列\( (x_n)_{n = 1}^{\infty}, (y_n)_{n = 1}^{\infty}, (z_n)_{n = 1}^{\infty}, (w_n)_{n = 1}^{\infty} \),如果\( \text{LIM}_{n \to \infty} x_n = \text{LIM}_{n \to \infty} y_n, \text{LIM}_{n \to \infty} z_n = \text{LIM}_{n \to \infty} w_n \),则我们要证明\( d_{\overline{X}}(\text{LIM}_{n \to \infty} x_n, \text{LIM}_{n \to \infty} z_n) = d_{\overline{X}}(\text{LIM}_{n \to \infty} y_n, \text{LIM}_{n \to \infty} w_n) \),也就是我们得证明\( \lim_{n \to \infty} d(x_n, z_n) = \lim_{n \to \infty} d(y_n, w_n) \)。
这里的思路是,由\( \text{LIM}_{n \to \infty} x_n = \text{LIM}_{n \to \infty} y_n \),可得\( d(x_n, z_n) \)最终会和\( d(y_n, z_n) \)值差不多,再由\( \text{LIM}_{n \to \infty} z_n = \text{LIM}_{n \to \infty} w_n \),可得\( d(y_n, z_n) \)最终会和\( d(y_n, w_n) \)值差不多,连起来就是\( d(x_n, z_n) \)最终会和\( d(y_n, w_n) \)值差不多,这是我们想要的,所谓的“值差不多”通过两者相减极限为\( 0 \)来刻画。
\( \forall \epsilon > 0 \),由\( (x_n)_{n = 1}^{\infty} \)收敛于\( \lim_{n \to \infty} d(x_n, z_n) \),可得\( \exists N_1 \geq 1, \forall n \geq N_1, |d(x_n, z_n) - \lim_{n \to \infty} d(x_n, z_n)| \leq \dfrac{\epsilon}{4} \),可得\( -\dfrac{\epsilon}{4} + \lim_{n \to \infty} d(x_n, z_n) \leq d(x_n, z_n) \leq \dfrac{\epsilon}{4} + \lim_{n \to \infty} d(x_n, z_n) \),由\( \text{LIM}_{n \to \infty} x_n = \text{LIM}_{n \to \infty} y_n \),可得\( \lim_{n \to \infty} d(x_n, y_n) = 0 \),进而可得\( \exists N_2 \geq 1, \forall n \geq N_2, d(x_n, y_n) \leq \dfrac{\epsilon}{4} \),取\( N_3 = \max(N_1, N_2) \),则\( \forall n \geq N_3 \),有\( d(y_n, z_n) \leq d(y_n, x_n) + d(x_n, z_n) \leq \dfrac{\epsilon}{4} + \dfrac{\epsilon}{4} + \lim_{n \to \infty} d(x_n, z_n) = \dfrac{\epsilon}{2} + \lim_{n \to \infty} d(x_n, z_n) \),还有\( d(x_n, z_n) \leq d(x_n, y_n) + d(y_n, z_n) \leq \dfrac{\epsilon}{4} + d(y_n, z_n) \),也就是\( d(x_n, z_n) \leq \dfrac{\epsilon}{4} + d(y_n, z_n) \),可得\( -\dfrac{\epsilon}{4} + d(x_n, z_n) \leq d(y_n, z_n) \),又\( -\dfrac{\epsilon}{4} + \lim_{n \to \infty} d(x_n, z_n) \leq d(x_n, z_n) \),可得\(-\dfrac{\epsilon}{4} + (-\dfrac{\epsilon}{4} + \lim_{n \to \infty} d(x_n, z_n)) \leq d(y_n, z_n) \),也就是\( -\dfrac{\epsilon}{2} + \lim_{n \to \infty} d(x_n, z_n) \leq d(y_n, z_n) \),加上前面得到的\( d(y_n, z_n) \leq \dfrac{\epsilon}{2} + \lim_{n \to \infty} d(x_n, z_n) \),可得\( |d(y_n, z_n) - \lim_{n \to \infty} d(x_n, z_n)| \leq \dfrac{\epsilon}{2} \),进而可得\( -\dfrac{\epsilon}{2} + \lim_{n \to \infty} d(x_n, z_n) \leq d(y_n, z_n) \leq \dfrac{\epsilon}{2} + \lim_{n \to \infty} d(x_n, z_n) \),由\( \text{LIM}_{n \to \infty} z_n = \text{LIM}_{n \to \infty} w_n \),可得\( \lim_{n \to \infty} d(z_n, w_n) = 0 \),进而可得\( \exists N_4 \geq 1, \forall n \geq N_4, d(z_n, w_n) \leq \dfrac{\epsilon}{2} \),取\( N_5 = \max(N_3, N_4) \),则\( \forall n \geq N_5 \),有\( d(y_n, w_n) \leq d(y_n, z_n) + d(z_n, w_n) \leq \dfrac{\epsilon}{2} + \lim_{n \to \infty} d(x_n, z_n) + \dfrac{\epsilon}{2} = \epsilon + \lim_{n \to \infty} d(x_n, z_n) \),还有\( d(y_n, z_n) \leq d(y_n, w_n) + d(w_n, z_n) \leq d(y_n, w_n) + \dfrac{\epsilon}{2} \),也就是\( d(y_n, z_n) \leq d(y_n, w_n) + \dfrac{\epsilon}{2} \),可得\( -\dfrac{\epsilon}{2} + d(y_n, z_n) \leq d(y_n, w_n) \),又\( -\dfrac{\epsilon}{2} + \lim_{n \to \infty} d(x_n, z_n) \leq d(y_n, z_n) \),可得\( -\dfrac{\epsilon}{2} + (-\dfrac{\epsilon}{2} + \lim_{n \to \infty} d(x_n, z_n)) \leq d(y_n, w_n) \),也就是\( -\epsilon + \lim_{n \to \infty} d(x_n, z_n) \leq d(y_n, w_n) \),加上前面得到的\( d(y_n, w_n) \leq \epsilon + \lim_{n \to \infty} d(x_n, z_n) \),可得\( |d(y_n, w_n) - \lim_{n \to \infty} d(x_n, z_n)| \leq \epsilon \)。简而言之,有\( \forall \epsilon, \exists N \geq 1, \forall n \geq N, |d(y_n, w_n) - \lim_{n \to \infty} d(x_n, z_n)| \leq \epsilon \),于是可得\( \lim_{n \to \infty} d(y_n, w_n) = \lim_{n \to \infty} d(x_n, z_n) \)。
证明3:
给定任意\( (\overline{X}, d_{\overline{X}}) \)上的柯西序列\( (Y_n)_{n = 1}^{\infty} \),这里\( \forall n \geq 1 \),针对\( Y_n \),均有\( \exists X \)上的柯西序列\( (x_n^{(k)})_{k = 1}^{\infty} \)使得 \( Y_n = \text{LIM}_{k \to \infty} x_n^{(k)} \),这里\( (x_n^{(k)})_{k = 1}^{\infty} \)为柯西序列,可得\( \exists N_{n'} \geq 1, \forall j_1, j_2 \geq N_{n'}, d(x_n^{(j_1)}, x_n^{(j_2)}) < \dfrac{1}{n} \),令\( N_n = \max(N_{n'}, n) \),则还是有\( \forall k \geq N_n, d(x_n^{(k)}, x_n^{(N_n)}) < \dfrac{1}{n} \),且同时有\( N_n \geq n \),令\( y_n = x_n^{(N_n)} \),这里,我们隐式通过选择公理构造出了序列\( (y_n)_{n = 1}^{\infty} \),下面,我们会证明\( (y_n)_{n = 1}^{\infty} \)是\( X \)上的柯西序列,且\( (Y_n)_{n = 1}^{\infty} \)收敛于\( \text{LIM}_{n \to \infty} y_n \)。
先证明\( (y_n)_{n = 1}^{\infty} \)是\( X \)上的柯西序列, \( \forall \epsilon > 0, \) 由\( (Y_n)_{n = 1}^{\infty} \)为\( \overline{X} \)上的柯西序列,可得\( \exists K \geq 1, \forall j_1, j_2 \geq K, d_{\overline{X}}(Y_{j_1}, Y_{j_2}) = \lim_{k \to \infty} d(x_{j_1}^{(k)}, x_{j_2}^{(k)}) < \dfrac{\epsilon}{2} \)。还有\( \exists N \geq K \)满足\( \dfrac{1}{N} < \dfrac{\epsilon}{4} \),此时针对\( \forall i, j \geq N \),则\( \forall k \geq \max(N_i, N_j) \),有\( d(y_i, y_j) = d(x_i^{(N_i)}, x_j^{(N_j)}) \leq d(x_i^{(N_i)}, x_i^{(k)}) + d(x_i^{(k)}, x_j^{(k)}) + d(x_j^{(k)}, x_j^{(N_j)}) \leq \dfrac{1}{N_i} + d(x_i^{(k)}, x_j^{(k)}) + \dfrac{1}{N_j} \leq \dfrac{1}{N} + d(x_i^{(k)}, x_j^{(k)}) + \dfrac{1}{N} < \dfrac{\epsilon}{4} + d(x_i^{(k)}, x_j^{(k)}) + \dfrac{\epsilon}{4} = \dfrac{\epsilon}{2} + d(x_i^{(k)}, x_j^{(k)}) \) (注:这里用到了\( N_i \geq i \geq N \),从而\( \dfrac{1}{N_i} \leq \dfrac{1}{N} \),\( N_j \)也类似),这里由于\( i, j \geq N \geq K \),根据前面,可得\( \lim_{k \to \infty} d(x_{i}^{(k)}, x_{k}^{(k)}) < \dfrac{\epsilon}{2} \),这意味着\( \exists k_1 \geq 1, \forall k \geq k_1, d(x_{i}^{(k)}, x_{k}^{(k)}) < \dfrac{\epsilon}{2} \),特别的,取\( k_2 = \max(k_1, N_i, N_j) \),我们有\( d(x_i^{(N_i)}, x_j^{(N_j)}) \leq \dfrac{\epsilon}{2} + d(x_i^{(k_2)}, x_j^{(k_2)}) < \dfrac{\epsilon}{2} + \dfrac{\epsilon}{2} = \epsilon \)。简而言之,\( \forall \epsilon > 0, \exists N \geq 1, \forall i, j \geq N, d(y_i, y_j) < \epsilon \),可得\( (y_n)_{n = 1}^{\infty} \)是\( X \)上的柯西序列。
接着证明\( (Y_n)_{n = 1}^{\infty} \)收敛于\( \text{LIM}_{n \to \infty} y_n \),令\( L = \text{LIM}_{n \to \infty} y_n \), \( \forall \epsilon > 0 \),我们得证明\( \exists N \geq 1, \forall n \geq N, d_{\overline{X}}(Y_n, L) = \lim_{k \to \infty} d(x_{n}^{(k)}, y_k) < \epsilon \)。 \( \forall \epsilon > 0 \),可得\( \exists N_1 \geq 1, \dfrac{1}{N_1} < \dfrac{\epsilon}{2} \),由\( (y_n)_{n = 1}^{\infty} \)是\( X \)上的柯西序列,可得\( \exists N_2 \geq 1, \forall j_1, j_2 \geq N_2, d(y_j, y_k) < \dfrac{\epsilon}{2} \),取\( N_3 = \max(N_1, N_2) \),则\( \forall n \geq N_3 \),由\( y_n \)的定义,可得\( \forall k \geq \max(N_{n}, N_3) \),有\( d(x_n^{(k)}, y_n) < \dfrac{1}{n} < \dfrac{\epsilon}{2} \),可得\( d(x_n^{(k)}, y_k) \leq d(x_n^{(k)}, y_n) + d(y_n, y_k) < \dfrac{\epsilon}{2} + \dfrac{\epsilon}{2} = \epsilon \) 这意味着\( \lim_{k \to \infty} d(x_{n}^{(k)}, y_k) < \epsilon \)。简而言之,\( \forall \epsilon > 0 \),我们证明了\( \exists N \geq 1, \forall n \geq N, d_{\overline{X}}(Y_n, L) = \lim_{k \to \infty} d(x_{n}^{(k)}, y_k) < \epsilon \),可得\( (Y_n)_{n = 1}^{\infty} \)收敛于\( L \)。
证明4:
令\( Y \)为\( X \)在\( \overline{X} \)上的闭包,我们要证明\( Y = \overline{X} \),显然有\( Y \subseteq \overline{X} \),还得证明\( \overline{X} \subseteq Y \), \( \forall y \in \overline{X} \),可得存在\( X \)上的柯西序列\( (x_n)_{n = 1}^{\infty} \) 使得\( \text{LIM}_{n \to \infty} x_n = y \),为了证明\( y \in Y \),我们得证明\( y \)是\( X \)的附着点(注意,得在度量\( d_{\overline{X}} \)的标准下,而不是\( d \)下), \( \forall r > 0 \),由\( (x_n)_{n = 1}^{\infty} \)为柯西序列,可得\( \exists N \geq 1, \forall n \geq N, d(x_n, x_N) < \dfrac{r}{2} \),进而由推论I.6.4.14(Squeeze test),可得\( \lim_{n \to \infty} d(x_n, x_N) \leq \dfrac{r}{2} \),此时有\( d_{\overline{X}}(y, x_N) = d_{\overline{X}}(y, \text{LIM}_{n \to \infty} x_N) = \lim_{n \to \infty} d(x_n, x_N) \leq \dfrac{r}{2} < r \),可得\( x_N \in B_{(\overline{X}, d_{\overline{X}})}(y, r) \),又\( x_N \in X \),这意味着\( B_{(\overline{X}, d_{\overline{X}})}(y, r) \cap X \neq \empty \),可得\( y \)是\( X \)的附着点,有\( y \in \overline{X} \),至此有\( \overline{X} \subseteq Y \),综上有\( Y = \overline{X} \)。
证明5:
由4以及\( (x_n)_{n = 1}^{\infty} \)为柯西序列,可得\( \lim_{n \to \infty} x_n \)在\( (\overline{X}, d_{\overline{X}}) \)上存在(注:严格来讲,\( (x_n)_{n = 1}^{\infty} \)本身是\( (X, d) \)上的序列,而我们这里指的是\( (x_n)_{n = 1}^{\infty} \)在\( (\overline{X}, d_{\overline{X}}) \)上对应的序列的极限存在,后者序列比前者序列“高”一层级,但两者是同构的,这里不区分这两个层级下对应的元素,实际上在4中我们已经这么干了),有\( \lim_{n \to \infty} x_n \in \overline{X} \),于是存在\( X \)上的柯西序列\( (y_n)_{n = 1}^{\infty} \)使得\( \lim_{n \to \infty} x_n = \text{LIM}_{n \to \infty} y_n \), \( \forall \epsilon > 0 \),由\( \lim_{n \to \infty} x_n = \text{LIM}_{n \to \infty} y_n \),可得\( \exists K_1 \geq 1, \forall k \geq K_1, d_{\overline{X}}(x_k, \text{LIM}_{n \to \infty} y_n) = \lim_{n \to \infty} d(x_k, y_n) \leq \dfrac{\epsilon}{2} \),进一步可得针对这个固定的\( k \),有\( \exists N_1 \geq 1, \forall n \geq N_1, d(x_k, y_n) \leq \dfrac{\epsilon}{2} \),而由\( (x_n)_{n = 1}^{\infty} \)为柯西序列,可得\( \exists N_2 \geq 1, \forall j_1, j_2 \geq N_2, d(x_{j_1}, x_{j_2}) < \dfrac{\epsilon}{2} \),于是\( \forall n \geq \max(K_1, N_1, N_2) \),取\( N_3 = \max(N_2, K_1) \),有\( d(x_n, y_n) \leq d(x_n, x_{N_3}) + d(x_{N_3}, y_n) < \dfrac{\epsilon}{2} + \dfrac{\epsilon}{2} = \epsilon \),至此可得\( d_{\overline{X}}(\text{LIM}_{n \to \infty} x_n, \text{LIM}_{n \to \infty} y_n) = \lim_{n \to \infty} d(x_n, y_n) = 0 \),根据1中相等的定义,可得\( \text{LIM}_{n \to \infty} x_n = \text{LIM}_{n \to \infty} y_n = \lim_{n \to \infty} x_n \)。
证毕。
章节1.5
练习1.5.1
题目:
Show that Definition 9.1.22 from Analysis I and Definition 1.5.3 match when talking about subsets of the real line with the standard metric.
Definition 9.1.22 from Analysis I的内容:
(Bounded sets) A subset \( X \) of the real line is said to be bounded if we have \( X \subseteq [-M, M] \) for some real number \( M > 0 \). A subset \( X \) of the real line is unbounded if it is not bounded.
Definition 1.5.3的内容:
(Bounded sets) Let \( (X, d) \) be a metric space, and let \( Y \) be a subset of \( X \). We say that \( Y \) is bounded iff for every \( x \in X \) there exists a ball \( B(x, r) \) in \( X \) of some finite radius \( r \) which contains \( Y \). We call the metric space \( (X, d) \) bounded if \( X \) is bounded.
证明:
给定\( X \subseteq R \),我们证明\( X \)在定义9.1.22下为有界集当且仅当 \( X \)在定义1.5.3下为有界集。
如果\( X \)在定义9.1.22下为有界集,则\( \exists M > 0, X \subseteq [-M, M] \), \( \forall x \in X \),令\( r = 3M \),则\( B(x, r) = \{ y \in X : d(y, x) < r = 3M \} \),针对\( \forall y \in X \),由\( X \subseteq [-M, M] \),可得\( y \in [-M, M] \),加上\( x \)也\( \in [-M, M] \)(因为\( x \in X \)),可得\( d(y, x) \leq 2M < 3M = r \),至此有\( y \in B(x, r) \),综上有\( X \subseteq B(x, r) \),可得\( X \)在定义1.5.3下为有界集。
反之,如果\( X \)在定义1.5.3下为有界集,先考虑\( X \)为空集的情况,则令\( M = 1 \),有\( X \subseteq [-M, M] \),可得\( X \)在定义9.1.22下为有界集,最后考虑\( X \)不为空集的情况,此时任取\( x_0 \in X \),由\( X \)在定义1.5.3下为有界集,可得\( \exists r > 0 \),使得\( X \subseteq B(x_0, r) = \{ x \in X : d(x, x_0) < r \} \),取\( M = \max \{ x_0 + r, 1, x_0 - r \} \),则易得\( X \subseteq B(x_0, r) = \{ x \in X : d(x, x_0) < r \} \subseteq [-M, M] \),有\( X \)在定义9.1.22下为有界集。
证毕。
练习1.5.2
题目:
Prove Proposition 1.5.5. (Hint: prove the completeness and boundedness separately. For both claims, use proof by contradiction. You will need the axiom of choice, as in Lemma 8.4.5 from Analysis I.)
Proposition 1.5.5的内容:
Let \( (X, d) \) be a compact metric space. Then \( (X, d) \) is both complete and bounded.
证明:
先证明\( (X, d) \)为完备度量空间: \( \forall (X, d) \)上的柯西序列\( (x_n)_{n = 1}^{\infty} \),我们得证明它收敛于\( (X, d) \)中的某个元素,由\( (X, d) \)为紧度量空间,可得存在\( (x_n)_{n = 1}^{\infty} \)的子序列 \( (x_{n_j})_{j = 1}^{\infty} \)收敛于某个元素\( x_0 \in X \),此时根据引理1.4.9,可得\( (x_n)_{n = 1}^{\infty} \)也收敛于\( x_0 \) ,综上,有\( (X, d) \)为完备度量空间。
接着证明\( (X, d) \)为有界集:假设\( (X, d) \)非有界集,则\( \exists x_0 \in X, \forall r > 0 \),有\( X \setminus d(x_0, r) \neq \empty \),进而\( \forall n \geq 1 \),我们任取\( x_n \in X \setminus d(x_0, n) \),有\( x_n \geq n \),这里序列\( (x_n)_{n = 1}^{\infty} \)为\( (X, d) \)上的序列,根据\( (X, d) \)为紧度量空间,可得它存在子序列\( (x_{n_j})_{j = 1}^{\infty} \)收敛于某个数\( L \in X \),我们证明该子序列不收敛,从而产生矛盾,由\( (x_{n_j})_{j = 1}^{\infty} \)收敛于某个数\( L \),可得\( \exists N_1 \geq 1, \forall j \geq N_1, d(x_{n_j}, L) \leq 1 \),存在正整数\( N_2 \geq d(x_0, L) + 2 \),取\( j_0 = \max(N_1, N_2) \),则有\( d(x_0, x_{n_{j_0}}) \leq d(x_0, L) + d(L, x_{n_{j_0}}) \leq d(x_0, L) + 1 \),但根据\( x_{n_{j_0}} \)的定义,有\( d(x_0, x_{n_{j_0}}) \geq n_{j_0} \geq j_0 \geq N_2 \geq d(x_0, L) + 2 > d(x_0, L) + 1 \),矛盾,因此假设不成立,有\( (X, d) \)为有界集。
证毕。
练习1.5.3
题目:
Prove Theorem 1.5.7. (Hint: use Proposition 1.1.18 and Theorem 9.1.24 from Analysis I.)
Theorem 1.5.7的内容:
(Heine–Borel theorem) Let \( (\mathbf{R}^n , d) \) be a Euclidean space with either the Euclidean metric, the taxicab metric, or the sup norm metric. Let \( E \) be a subset of \( \mathbf{R}^n \). Then \( E \) is compact if and only if it is closed and bounded.
Proposition 1.1.18的内容:
见练习1.1.12。
Theorem 9.1.24 from Analysis I的内容:
注:
Theorem 9.1.24和Proposition 1.1.18中的序列起始坐标不一致,但是序列起始坐标并不影响收敛性,故这里无视这点。
证明:
必要性:
由推论1.5.6可得成立。
充分性:
如果\( E \)为空集,则显然有\( E \)为紧集,下面考虑\( E \)不为空集的情况。
\( \forall 1 \leq j \leq n \),我们令\( E_j = \{ x_j : \forall x = (x_1, \dots, x_n) \in E \} \),即\( E_j \)为\( E \)中所有元素的第\( j \)个坐标分量组成的集合。
我们证明所有集合\( E_j \)均封闭且有界。
\( \forall 1 \leq j \leq n \),假设\( E_j \)不封闭,则存在\( \forall E_j \)上收敛的\( (x^{(k)})_{k = 1}^{\infty} \),它的极限\( x \notin E_j \),\( \forall 1 \leq i \leq n, i \neq j \),任取\( x_i \in E_i \),\( \forall k \geq 1 \),令\( y^{(k)} = (x_1, \dots, x_{j - 1}, x^{(k)}, x_{j + 1}, \dots, x_n) \),即\( y^{(k)} \)第\( j \)个分量取\( x^{(k)} \),其他分量都取各个\( E_i \)中某个固定的常量,令\( y = (x_1, \dots, x_{j - 1}, x, x_{j + 1}, \dots, x_n) \),根据定理1.1.18,易得\( (y^{(k)})_{k = 1}^{\infty} \)在度量\( d_{l^2}, d_{l^1}, d_{l^{\infty}} \)下收敛于\( y \),而\( (y^{(k)})_{k = 1}^{\infty} \)为\( E \)上的序列,由\( E \)为封闭集,可得\( y \in E \),进而有\( x \in E_j \),这和\( x \notin E_j \)矛盾,因此假设不成立,有\( E_j \)封闭。
\( \forall 1 \leq j \leq n \),假设\( E_j \)非有界,则\( \exists x_0 \in E, \forall r > 0, E_j \setminus \{ x \in E_j : |x - x_0| < r \} \neq \empty \),易得在度量\( d_{l^2}, d_{l^1}, d_{l^{\infty}} \)下,均有\( E \setminus B(x_0, r) \neq \empty \),这意味着\( E \)非有界,和\( E \)有界矛盾,因此假设不成立,有\( E_j \)有界。
现在我们回去证明\( E \)为紧集。
给定任意\( E \)上的序列\( (x^{(k)})_{k = m}^{\infty} \), \( \forall k \geq m \),令\( x^{k} = (x_1^{(k)}, \dots, x_n^{(k)}) \),我们有\( \forall 1 \leq j \leq n, (x_j^{(k)})_{k = m}^{\infty} \)为\( E_j \)上的序列,由\( E_j \)封闭且有界,根据定理9.1.24,可得\( (x_j^{(k)})_{k = m}^{\infty} \)收敛于\( x_j \in E_j \),进而根据定理1.1.18,可得\( (x^{(k)})_{k = m}^{\infty} \)在度量\( d_{l^2}, d_{l^1}, d_{l^{\infty}} \)下收敛于\( x = (x_1, \dots, x_n) \in E \),综上,有\( E \)为紧集。
证毕。
练习1.5.4
题目:
Let \( (\mathbf{R}, d) \) be the real line with the standard metric. Give an example of a continuous function \( f: \mathbf{R} \to \mathbf{R} \), and an open set \( V \subseteq \mathbf{R} \), such that the image \( f(V) := \{ f(x) : x \in V \} \) of \( V \) is not open.
解答:
随便找个非空开集\( V \),然后构造一个常数函数就行了,此时像为单个点,非开集,具体的,令\( V = (-1, 1) \)为开集,\( \forall x \in V \),令\( f(x) = 1 \),则\( f(V) = \{ 1 \} \)非开集(注:\( 1 \)是\( \{ 1 \} \)的边界点,而开集不能包含边界点,进而\( \{ 1 \} \)非开集)。
练习1.5.5
题目:
Let \( (\mathbf{R}, d) \) be the real line with the standard metric. Give an example of a continuous function \( f: \mathbf{R} \to \mathbf{R} \), and a closed set \( F \subseteq \mathbf{R} \), such that \( f(F) \) is not closed.
解答:
令\( F = [1, \infty) \)为闭集,\( \forall x \in F \),令\( f(x) = \dfrac{1}{x} \),则\( f(F) = (0, 1] \)非闭集(注:\( 0 \)是\( (0, 1] \)的边界点,而闭集得包含所有边界点,进而\( (0, 1] \)非闭集)。
练习1.5.6
题目:
Prove Corollary 1.5.9. (Hint: work in the compact metric space \( K_1, d|_{K_1 \times K_1} \), and consider the sets \( V_n = K_1 \setminus K_n \), which are open on \( K_1 \). Assume for sake of contradiction that \( \bigcap_{n = 1}^{\infty} K_n = \empty \), and then apply Theorem 1.5.8.)
Corollary 1.5.9的内容:
Let \( (X, d) \) be a metric space, and let \( K_1, K_2, K_3, \dots \) be a sequence of non-empty compact subsets of \( X \) such that \( K_1 \supseteq K_2 \supseteq K_3 \supseteq \dots \). Then the intersection \( \bigcap_{n = 1}^{\infty} K_n \) is non-empty.
证明:
根据推论1.5.6,我们知道所有集合\( K_1, K_2, \dots \)都是闭集,实际上,根据定理1.5.5,我们知道它们都是完备的,进而都是本质闭集(intrinsically closed)。
假设\( \bigcap_{n = 1}^{\infty} K_n = \empty \),则\( \forall n \geq 1 \),我们令\( V_n = K_1 \setminus K_n \),此时由\( K_n \)在\( K_1 \)下是闭集,可得\( V_n \)在\( K_1 \)下为开集,进而根据定理1.2.15(g),可得\( K_1 \setminus (\bigcap_{n = 1}^{\infty} K_n) = \bigcup_{n = 1}^{\infty} (K_1 \setminus K_n) = \bigcup_{n = 1}^{\infty} V_n \)为开集,而\( K_1 \setminus (\bigcap_{n = 1}^{\infty} K_n) = K_1 \setminus \empty = K_1 \),这意味着\( K_1 = \bigcup_{n = 1}^{\infty} V_n \),也就是说\( \{ V_1, V_2, \dots \} \)为\( K_1 \)的开覆盖,进而根据定理1.5.8,可得存在有限集合\( A \subseteq (\mathbf{N} \setminus \{ 0 \}) \),使得\( K_1 = \bigcup_{n \in A} V_n \),也就是\( K_1 \)能被有限个\( V_n \)集合覆盖,而这有限个\( V_n \)集合中必然有一个“最大”的集合,即该集合是所有其他集合的超集,假设该集合的下标为\( k \in A \),则\( \forall n \in A, V_n \subseteq V_k \),于是\( K_1 = \bigcup_{n \in A} V_n = V_k = K_1 \setminus K_k \),这意味着\( K_k = \empty \),然而这和\( K_1, K_2, \dots \)中所有集合均非空矛盾,因此假设不成立,有\( \bigcap_{n = 1}^{\infty} K_n \)非空。
证毕。
练习1.5.7
题目:
Prove Theorem 1.5.10. (Hint: for part 3, you may wish to use 2, and first prove that every singleton set is compact.)
Theorem 1.5.10的内容:
Let \( (X, d) \) be a metric space.
- If \( Y \) is a compact subset of \( X \), and \( Z \subseteq Y \), then \( Z \) is compact if and only if \( Z \) is closed.
- If \( Y_1, \dots, Y_n \) are a finite collection of compact subsets of \( X \), then their union \( Y_1 \cup \dots \cup Y_n \) is also compact.
- Every finite subset of \( X \) (including the empty set) is compact.
证明:
证明1:
必要性:
如果\( Z \)为紧集,则根据推论1.5.6,可得\( Z \)为闭集。
充分性:
如果\( Z \)为闭集,则给定任意\( Z \)上的序列\( (x_n)_{n = 1}^{\infty} \),该序列也是\( Y \)上的序列,而\( Y \)为紧集,因此可得该序列存在收敛的子序列\( (x_{n_j})_{j = 1}^{\infty} \)且 \( \lim_{j \to \infty} x_{n_j} \in Y \),这里\( (x_{n_j})_{j = 1}^{\infty} \)也是\( Z \)上收敛的序列,此时由\( Z \)为闭集以及定理1.2.15(b),可得\( \lim_{j \to \infty} x_{n_j} \in Z \),综上,有\( Z \)为紧集。
证明2:
给定任意\( Y_1 \cup \dots \cup Y_n \)上的序列\( (x_n)_{n = 1}^{\infty} \),则其中必然有一个集合\( Y_k, 1 \leq k \leq n \)包含\( (x_n)_{n = 1}^{\infty} \)中无限个下标对应的序列元素,具体的就是,\( \forall N \geq 1, \exists n \geq N, x_n \in Y_k \),否则如果所有集合都仅包含有限个下标对应的序列元素,则必存在某个下标对应的序列元素不在\( Y_1 \cup \dots \cup Y_n \)中,这和\( (x_n)_{n = 1}^{\infty} \)是\( Y_1 \cup \dots \cup Y_n \)上的序列矛盾(注意,这点针对无限个集合的并集\( Y_1 \cup Y_2 \cup \dots \)是不成立的)。回去继续考虑\( Y_k \),根据上述讨论,容易使用选择公理构造出一个\( (x_n)_{n = 1}^{\infty} \)的子序列 \( (x_{n_j})_{j = 1}^{\infty} \)满足\( \forall j \geq 1, x_{n_j} \in Y_k \),也就是说\( (x_{n_j})_{j = 1}^{\infty} \)为\( Y_k \)上的序列,进而根据\( Y_k \)为紧集可得\( (x_{n_j})_{j = 1}^{\infty} \)存在收敛的子序列 \( (x_{n_{f(i)}})_{i = 1}^{\infty} \)且\( \lim_{i \to \infty} (x_{n_{f(i)}}) \in Y_k \),进一步可得\( (x_{n_{f(i)}})_{i = 1}^{\infty} \)是\( (x_n)_{n = 1}^{\infty} \)收敛的子序列且\( \lim_{i \to \infty} (x_{n_{f(i)}}) \in Y_1 \cup \dots \cup Y_n \),综上,有\( Y_1 \cup \dots \cup Y_n \)为紧集。
证明3:
任意\( X \)的有限子集都是有限个单例集的并集,因此我们先证明单例子集也为紧集,进一步使用2即可。
\( \forall x \in X \),我们证明\( \{ x \} \)为紧集, \( \forall \{ x \} \)上的序列\( (y_n)_{n = 1}^{\infty} \),有\( (y_n)_{n = 1}^{\infty} \)为常数序列,进而可得\( (y_n)_{n = 1}^{\infty} \)收敛且收敛于\( \{ x \} \)中,综上,有\( \{ x \} \)为紧集。
\( \forall \)有限集\( Y \subseteq X \),令\( n = \#(Y) \),可得存在\( n \)个元素使得\( Y = \{ y_1, \dots, y_n \} \),这里根据刚才的证明,可得\( \forall 1 \leq i \leq n, \{ y_i \} \)为紧集,进而根据2,可得\( Y = \bigcup_{i = 1}^{n} \{ y_i \} \)为紧集。
证毕。
练习1.5.8
题目:
Let \( X, d_{l^1} \) be the metric space from Exercise 1.1.15. For each natural number \( n \), let \( e^{(n)} = (e_j^{(n)})_{j = 0}^{\infty} \) be the sequence in \( X \) such that \( e_j^{(n)} := 1 \) when \( n = j \) and \( e_j^{(n)} := 0 \) when \( n \neq j \). Show that the set \( \{ e^{(n)} : n \in \mathbf{N} \} \) is closed and bounded subset of \( X \), but is not compact. (This is despite the fact that \( (X, d_{l^1}) \) is even a complete metric space—a fact which we will not prove here. The problem is that not that \( X \) is incomplete, but rather that it is “infinite-dimensional”, in a sense that we will not discuss here.)
注:
这里\( e^{(0)} = (1, 0, 0, \dots) \),这里\( e^{(1)} = (0, 1, 0, \dots) \),以此类推。读者如果有一些线性代数知识的话,则容易看出, \( e^{(0)}, e^{(1)}, \dots \)为\( X \)的基。
本题目想说明的一点就是,在某些度量下,紧与(闭且有界)并不等价。
证明:
令\( E = \{ e^{(n)} : n \in \mathbf{N} \} \)。
先证明\( E \)为闭集:
我们打算使用定理1.2.15(b)来证明,给定任意\( E \)上的收敛序列\( (a_n)_{n = 0}^{\infty} \),令\( L = \lim_{n \to \infty} a_n \),我们打算证明该序列的元素最终会变成常数,具体的,我们打算证明\( \exists e^{(i)} \in E, \exists N \geq 0, \forall n \geq N, a_n = e^{(i)} \),如果证明了这点,则明显目标序列的极限\( L \)会是该常数,而该常数存在于\( E \)中,这意味着目标序列收敛于\( E \),从而可以利用定理1.2.15(b)证明\( E \)为闭集。
假设该序列的元素最终不会变成常数,即\( \forall e^{(i)} \in E, \forall N \geq 0, \exists n \geq N, a_n \neq e^{(i)} \) (1) ,则我们打算证明\( (a_n)_{n = 0}^{\infty} \)非柯西序列,然而收敛序列必是柯西序列,从而产生矛盾,由\( (a_n)_{n = 0}^{\infty} \)收敛,可得该序列为柯西序列,从而可得,\( \exists N \geq 0, \forall i, j \geq N, d_{l^1}(a_j, a_k) < 1 \) (2) ,取\( e^{(i)} = a_N \in E \),则根据(1),可得\( \exists n_0 \geq N, a_{n_0} \neq e^{(i)} = a_N \),而\( a_{n_0} \in E \),可得\( \exists e^{(l)} = a_{n_0} \in E \),这意味着\( e^{(l)} \neq e^{(i)} \),进而有\( i \neq l \),于是根据\( e^{(l)}, e^{(i)} \)的定义以及\( d_{l^1} \)的定义,可得\( d_{l^1}(e^{(l)}, e^{(i)}) = 2 \),也就是\( d_{l^1}(a_{n_0}, a_N) = 2 \geq 1 \),然而这和(2)矛盾,因此假设不成立,有该序列的元素最终会变成常数,也就是\( \exists e^{(i)} \in E, \exists N \geq 0, \forall n \geq N, a_n = e^{(i)} \),于是明显目标序列的极限\( L \)会是该常数,而该常数存在于\( E \)中,这意味着目标序列收敛于\( E \),使用定理1.2.15(b),可得\( E \)为闭集。
接着证明\( E \)有界:
\( \forall i \geq 0 \),针对\( e^{(i)} \),我们得证明存在一个半径\( r \),使得\( E \subseteq B(e^{(i)}, r) \),这点是很容易做到的,毕竟\( E \)中所有元素的距离都是固定的2(不同元素)或者0(相同元素), \( r \)取个3就行了,具体的,\( \forall i \geq 0 \),则\( \forall e^{(j)} \in E \),如果\( j = i \),则\( d_{l^1}(e^{(j)}, e^{(i)}) = 0 \),否则\( d_{l^1}(e^{(j)}, e^{(i)}) = 2 \),可得\( d_{l^1}(e^{(j)}, e^{(i)}) < 3 \),于是有\( e^{(j)} \in B(e^{(i)}, 3) \),综上,有\( E \subseteq B(e^{(i)}, 3) \),可得\( E \)有界。
最后证明\( E \)非紧集:
我们断言序列\( (e^{(n)})_{n = 0}^{\infty} \)没有收敛的子序列,从而说明\( E \)非紧集。
给定任意\( (e^{(n)})_{n = 0}^{\infty} \)的子序列\( (e^{(n_j)})_{j = 0}^{\infty} \),明显有\( \forall i \neq j \geq 0, d(e^{(n_i)}, e^{(n_j)}) = 2 \),这意味着\( (e^{(n_j)})_{j = 0}^{\infty} \)非柯西序列,然而收敛序列必是柯西序列,这说明该子序列非收敛序列,综上,有\( (e^{(n)})_{n = 0}^{\infty} \)没有收敛的子序列,这说明\( E \)非紧集。
证毕。
练习1.5.9
题目:
Show that a metric space \( (X, d) \) is compact if and only if every sequence in \( X \) has at least one limit point.
证明:
必要性:
如果\( (X, d) \)是紧度量空间,则给定任意\( X \)上的序列\( (x_n)_{n = 1}^{\infty} \),由\( X \)是紧集,可得\( (x_n)_{n = 1}^{\infty} \)存在收敛的子序列 \( (x_{n_j})_{j = 1}^{\infty} \),令\( L = \lim_{j \to \infty} x_{n_j} \), 则根据定理1.4.5,可得\( L \)为\( (x_n)_{n = 1}^{\infty} \)的极限点。
充分性:
如果给定任意\( X \)上的序列,该序列均至少有一个极限点,则任给一个\( X \)上的序列\( (x_n)_{n = 1}^{\infty} \),可得该序列存在一个极限点\( L \),根据定理1.4.5,可得\( (x_n)_{n = 1}^{\infty} \)存在子序列收敛于\( L \),重点在于\( (x_n)_{n = 1}^{\infty} \)存在收敛的子序列,综上,可得\( (X, d) \)是紧度量空间。
证毕。
练习1.5.10
题目:
A metric space \( (X, d) \) is called totally bounded if for every \( \epsilon > 0 \), there exists a natural number \( n \) and a finite number of balls \( B(x^{(1)} , \epsilon), \dots , B(x^{(n)} , \epsilon) \) which cover \( X \). (i.e., \( X = \bigcup_{i = 1}^{n} B(x^{(i)} , \epsilon) \).)
- Show that every totally bounded space is bounded.
- Show the following stronger version of Proposition 1.5.5: if \( (X, d) \) is compact, then complete and totally bounded. (Hint: if \( X \) is not totally bounded, then there is some \( \epsilon > 0 \) such that \( X \) cannot be covered by finitely many \( \epsilon \)-balls. Then use Exercise 8.5.20 to find an infinite sequence of balls \( B(x^{(n)}, \epsilon / 2) \) which are disjoint from each other. Use this to then construct a sequence which has no convergent subsequence.)
- Conversely, show that if \( X \) is complete and totally bounded, then \( X \) is compact. (Hint: if \( (x^{(n)})_{n = 1}^{\infty} \) is a sequence in \( X \), use the total boundedness hypothesis to recursively construct a sequence of subsequences \( (x^{(n; j)})_{n = 1}^{\infty} \) of \( (x^{(n)})_{n = 1}^{\infty} \) for each positive integer \( j \), such that for each \( j \), the elements of the sequence \( (x^{(n; j)})_{n = 1}^{\infty} \) are contained in a single ball of radius \( 1/j \), and also that each sequence \( (x^{(n; j+1)})_{n = 1}^{\infty} \) is a subsequence of the previous one \( (x^{(n; j)})_{n = 1}^{\infty} \). Then show that the “diagonal” sequence \( (x^{(n; n)})_{n = 1}^{\infty} \) is a Cauchy sequence, and then use the completeness hypothesis.)
证明:
证明1:
如果\( X \)为空集,则命题显然成立,下面考虑\( X \)非空的情况。
\( \forall x \in X \),由\( (X, d) \)完全有界,可得存在\( n \)个球\( B(x^{(1)} , 1), \dots , B(x^{(n)} , 1) \)覆盖\( X \),我们打算找一个以\( x \)为中心的、足够大的球,覆盖这\( n \)个小球,从而达到覆盖整个\( X \),考虑\( n = 1 \)的情况,则令\( r = d(x, x^{(1)}) + 1 \),易得\( B(x, r) \supseteq B(x^{(1)}, 1) \)(脑中想下那个画面, \( x \)在\( x^{(1)} \)的左边,首先半径\( r \)扩大到\( d(x, x^{(1)}) \),则以\( x \)为中心的球\( B(x, r) \)会刚好触碰到\( x^{(1)} \),但是还没完全包含\( B(x^{(1)}, 1) \)的半径范围,进一步将半径\( r \)额外扩大\( 1 \),则\( B(x, r) \)就可以完全包含\( B(x^{(1)}, 1) \)了),进一步考虑\( n = 2 \)的情况,此时之前令\( r = d(x, x^{(1)}) + 1 \)只能刚好包含\( B(x^{(1)}, 1) \),我们可以在此基础上,进一步扩大\( r \),以让它还包含\( B(x^{(2)}, 1) \),具体的,令\( r = (d(x, x^{(1)}) + 1) + (d(x, x^{(2)}) + 1) \),则易得\( B(x, r) \supseteq B(x^{(2)}, 1) \),以此类推,令\( r = \sum_{i = 1}^{n} (d(x, x^{(i)} + 1)) \),则\( \forall 1 \leq i \leq n, B(x, r) \supseteq B(x^{(i)}, 1) \),进而可得\( B(x, r) \supseteq X = \bigcup_{i = 1}^{n} B(x^{(i)} , \epsilon) \)。综上,\( (X, d) \)有界。
证明2:
定理1.5.5已经证明了,如果\( (X, d) \)为紧度量空间,则\( (X, d) \)为完备度量空间,我们仅需要额外证明:如果\( (X, d) \)为紧度量空间,则\( (X, d) \)为完全有界度量空间。
接下来我们会用两种方法来证明,一种使用定理1.5.8,证明会比较简单,另外一种就是按提示来,也简单,但是写的内容会比较长。
证明2(方法1,使用定理1.5.8):
如果\( (X, d) \)为紧度量空间,则\( \forall \epsilon > 0 \),显然有\( X \subseteq \bigcup_{x \in X} B(x, \epsilon) \),这里\( \forall x \in X, B(x, \epsilon) \)均为\( X \)上的开集,于是根据定理1.5.8,可得\( \exists A \)的有限子集\( F \)满足 \( X \subseteq \bigcup_{x \in F} B(x, \epsilon) \),令\( n = \#(F) \),容易构建双射\( f: \{ 1, \dots, n \} \to F \),令\( x^{(1)} = f(1), \dots, x^{(n)} = f(n) \),可得\( X \subseteq \bigcup_{i = 1}^{n} B(x^{(i)}, \epsilon) \),到这里实际上已经证明了存在有限个球覆盖\( X \)了,但是题目多了句\( X = \bigcup_{i = 1}^{n} B(x^{(i)} , \epsilon) \),而不是\( X \subseteq \bigcup_{i = 1}^{n} B(x^{(i)}, \epsilon) \),这是因为\( (X, d) \)是整个度量空间,因此也有\( \bigcup_{i = 1}^{n} B(x^{(i)}, \epsilon) \subseteq X \),于是可得\( X = \bigcup_{i = 1}^{n} B(x^{(i)} , \epsilon) \)。
证明2(方法2,按提示来):
如果\( X \)为空集,则命题显然成立,下面考虑\( X \)非空的情况。
如果\( (X, d) \)为紧度量空间,假设\( (X, d) \)非完全有界,则存在\( \epsilon > 0 \)满足\( X \)无法被任何有限个\( \epsilon \)-球覆盖,任取一个\( x^{(1)} \in X \),我们知道\( B(x^{(1)}, \epsilon) \)无法覆盖\( X \),因此\( \exists x^{(2)} \in X \setminus B(x^{(1)}, \epsilon) \),这里由于\( x^{(2)} \notin B(x^{(1)}, \epsilon) \),可得\( B(x^{(2)}, \epsilon / 2) \cap B(x^{(1)}, \epsilon / 2) = \empty \),同理,\( B(x^{(1)}, \epsilon), B(x^{(2)}, \epsilon) \)无法覆盖\( X \),因此\( \exists x^{(3)} \in X \setminus \bigcup_{i = 1}^{2} (B(x^{(i)}, \epsilon)) \),同理易得\( B(x^{(1)}, \epsilon / 2), \dots B(x^{(3)}, \epsilon / 2) \)之间互不相交(两个两个考虑过去),以此类推,我们可以通过选择公理,构造出一个序列\( (x^{(n)})_{n = 1}^{\infty} \) 满足\( \forall n \geq 1 \),\( x^{(n)} \in X \setminus \bigcup_{i = 1}^{n} (B(x^{(i)}, \epsilon)) \),且\( B(x^{(1)}, \epsilon), \dots, B(x^{(n)}, \epsilon) \)无法覆盖\( X \),且\( B(x^{(1)}, \epsilon / 2), \dots B(x^{(n)}, \epsilon / 2) \)之间互不相交,我们证明\( (x^{(n)})_{n = 1}^{\infty} \)没有收敛子序列,而这只要证明它没有子序列为柯西序列即可,这是因为收敛序列必为柯西序列,会产生矛盾,这里我们证明它的任意子序列的元素均不会最终\( \epsilon / 3 \)-接近,进而不是柯西序列,给定任意\( (x^{(n)})_{n = 1}^{\infty} \)的子序列\( (x^{(n_j)})_{n = 1}^{\infty} \), \( \forall N \geq 1 \),\( \forall j, k \geq N \),均有\( B(x^{(j)}, \epsilon, 2) \cap B(x^{(k)}, \epsilon, 2) = \empty \),假设\( d(x^{(j)}, x^{(k)}) < \epsilon / 2 \),则\( x^{(k)} \in (B(x^{(j)}, \epsilon, 2) \cap B(x^{(k)}, \epsilon, 2)) \),这和\( B(x^{(j)}, \epsilon, 2) \cap B(x^{(k)}, \epsilon, 2) = \empty \)矛盾,可得\( d(x^{(j)}, x^{(k)}) \geq \epsilon / 2 > \epsilon / 3 \),综上,可得\( (x^{(n_j)})_{n = 1}^{\infty} \)非柯西序列,进而不会是收敛序列,也就是说\( (x^{(n)})_{n = 1}^{\infty} \)没有收敛子序列,这和\( X \)为紧集矛盾,因此一开始的假设不成立,可得\( (X, d) \)完全有界。
证明3:
按提示来证明,给定任意\( X \)上的序列\( (x^{(n)})_{n = 1}^{\infty} \),当\( j = 1 \)时,\( \dfrac{1}{j} = 1 \),由\( X \)完全有界,可得存在有限个半径为\( \dfrac{1}{j} = 1 \)的球覆盖\( X \),这有限个球中,必然存在某个球包含\( (x^{(n)})_{n = 1}^{\infty} \)中无限个元素(假设所有球都仅包含\( (x^{(n)})_{n = 1}^{\infty} \)中有限个元素,则\( (x^{(n)})_{n = 1}^{\infty} \)中存在元素不被这有限个球包含,进而这有限个球无法覆盖\( X \),矛盾),我们记这个包含无限个序列元素的球为\( B_j = B_1 \),现在我们递归定义\( (x^{(n)})_{n = 1}^{\infty} \)的子序列\( (x^{(n; j)})_{n = 1}^{\infty} = (x^{(n; 1)})_{n = 1}^{\infty} \),令\( x^{(1; 1)} \)为\( (x^{(n)})_{n = 1}^{\infty} \)中第一个被\( B_1 \)包含的元素,归纳假设\( x^{(k; 1)} \)已经定义,我们定义\( x^{(k + 1; 1)} \)为 \( (x^{(n)})_{n = k + 1}^{\infty} \)中第一个被\( B_1 \)包含的元素,至此,我们递归定义了\( (x^{(n)})_{n = 1}^{\infty} \)的子序列\( (x^{(n; j)})_{n = 1}^{\infty} = (x^{(n; 1)})_{n = 1}^{\infty} \),在基础上,我们进一步定义一系列额外的子序列,其中每个子序列均为前一个子序列的子序列,归纳假设\( (x^{(n; j)})_{n = 1}^{\infty} \)已定义,我们定义\( (x^{(n; j + 1)})_{n = 1}^{\infty} \),由\( X \)完全有界,可得存在有限个半径为\( \dfrac{1}{j + 1} \)的球覆盖\( X \),这有限个球中,必然存在某个球包含\( (x^{(n; j)})_{n = 1}^{\infty} \)中无限个元素,我们记这个包含无限个序列元素的球为\( B_{j + 1} \),我们递归定义\( (x^{(n; j)})_{n = 1}^{\infty} \)的子序列\( (x^{(n; j + 1)})_{n = 1}^{\infty} \),令\( x^{(1; j + 1)} \)为\( (x^{(n; j)})_{n = 1}^{\infty} \)中第一个被\( B_{j + 1} \)包含的元素,归纳假设\( x^{(k; j + 1)} \)已经定义,我们定义\( x^{(k + 1; j + 1)} \)为 \( (x^{(n; j)})_{n = k + 1}^{\infty} \)中第一个被\( B_{j + 1} \)包含的元素,至此,我们定义好了\( (x^{(n; j + 1)})_{n = 1}^{\infty} \)。
显然的,通过按顺序在第\( n \)个序列中取第\( n \)个元素,以此组成的序列\( (x^{(n; n)})_{n = 1}^{\infty} \)为\( (x^{(n)})_{n = 1}^{\infty} \)的子序列,我们证明该子序列为柯西序列, \( \forall \epsilon > 0, \exists N \geq 1 \)满足\( \dfrac{1}{N} < \epsilon \),于是\( \forall j, k \geq N \),由\( x^{(j, j)}, x^{(k, k)} \)均在半径为\( \dfrac{1}{N} \)的球\( B_{N} \)中,可得\( d(x^{(j, j)}, x^{(k, k)}) \leq \dfrac{1}{N} < \epsilon \),综上,可得\( (x^{(n; n)})_{n = 1}^{\infty} \)为柯西序列,进而根据\( X \)完备,可得\( (x^{(n; n)})_{n = 1}^{\infty} \)收敛于\( X \)。
至此,我们证明了任意\( X \)上的序列,均存在收敛于\( X \)的子序列,这说明了\( X \)是紧集。
证毕。
练习1.5.11
题目:
Let \( (X, d) \) have the property that every open cover of \( X \) has a finite subcover. Show that \( X \) is compact. (Hint: if \( X \) is not compact, then by Exercise 1.5.9, there is a sequence \( (x^{(n)})_{n = 1}^{\infty} \) with no limit points. Then for every \( x ∈ X \) there exists a ball \( B(x, \epsilon) \) containing \( x \) which contains at most finitely many elements of this sequence. Now use the hypothesis.)
证明:
假设\( X \)非紧集,则根据练习1.5.9,可得存在\( X \)上的序列\( (x^{(n)})_{n = 1}^{\infty} \),该序列没有极限点,我们断言,给定任意\( x \in X \),均存在半径\( \epsilon_x \),使得\( B(x, \epsilon_x) \)仅包含\( (x^{(n)})_{n = 1}^{\infty} \)中有限个元素(读者得注意什么叫包含有限个元素,特别的,如果序列是常数序列,则半球只要包含这个常数,就包含序列无限个元素了),假设这点不成立,即存在\( x_0 \in X \),不存在半径\( \epsilon \)使得 \( B(x_0, \epsilon) \)仅包含\( (x^{(n)})_{n = 1}^{\infty} \)中有限个元素,即针对所有半径\( \epsilon \), \( B(x_0, \epsilon) \)都包含\( (x^{(n)})_{n = 1}^{\infty} \)中无限个元素,则容易证明\( x_0 \)为\( (x^{(n)})_{n = 1}^{\infty} \)的极限点,矛盾,因此假设不成立,我们前面的断言是成立的。显然\( \{ B(x, \epsilon_x) : x \in X \} \)为\( X \)的开覆盖,可得该覆盖存在有限的子覆盖,即\( \exists \)有限集\( F \subseteq X \)满足 \( \{ B(x, \epsilon_x) : x \in F \} \)为\( X \)的覆盖,但是针对所有\( x \in F \),\( B(x, \epsilon_x) \)都仅包含了\( (x^{(n)})_{n = 1}^{\infty} \)中有限个元素,进而\( \{ B(x, \epsilon_x) : x \in F \} \)也必仅覆盖了\( (x^{(n)})_{n = 1}^{\infty} \)中有限个元素,这意味着\( \{ B(x, \epsilon_x) : x \in F \} \)连\( (x^{(n)})_{n = 1}^{\infty} \)都覆盖不了,更别说覆盖\( X \)了,但前面说了\( \{ B(x, \epsilon_x) : x \in F \} \)为\( X \)的覆盖,矛盾,因此一开始的假设不成立,有\( X \)为紧集。
证毕。
练习1.5.12
题目:
Let \( (X, d_{\text{disc}}) \) be a metric space with the discrete metric \( d_{\text{disc}} \).
- Show that \( X \) is always complete.
- When is X\( X \) compact, and when is \( X \) not compact? Prove your claim. (Hint: the Heine–Borel theorem will be useless here since that only applies to Euclidean spaces.)
解答:
证明1:
给定任意\( X \)上的柯西序列\( (x_n)_{n = 1}^{\infty} \),取距离\( \epsilon = \dfrac{1}{2} \),由于度量为\( d_{\text{disc}} \)以及柯西序列的元素必须最终无限接近,可得\( (x_n)_{n = 1}^{\infty} \)尾部必须是常数\( x_0 \),易得\( (x_n)_{n = 1}^{\infty} \)收敛于\( x_0 \),综上,\( X \)完备。
解答2:
根据练习1.5.9,\( X \)为紧集当且仅当所有\( X \)上的序列均有极限点,而\( x_0 \)为\( X \)上的序列\( (x_n)_{n = 1}^{\infty} \)的极限点当且仅当,不管你限定多么小的距离\( \epsilon \),在此基础上,你随便取个起始下标\( N \geq 1 \),我都能在\( (x_n)_{n = N}^{\infty} \)中找到一个元素\( x_1 \),它和\( x_0 \)的距离在\( \epsilon \)之内,特别的,我们取距离\( \epsilon = \dfrac{1}{2} \),可得\( \forall N \geq 1 \),均得能在\( (x_n)_{n = N}^{\infty} \)中找到一个元素\( x_1 \),它和\( x_0 \)的距离在\( \epsilon \)之内,由于度量为\( d_{\text{disc}} \),这意味着\( x_0 = x_1 \),综上,我们通过不断取起始下标\( N \)为越来越大的数,可得序列\( (x_n)_{n = 1}^{\infty} \)必须包含无限个极限点\( x_0 \),而这个条件必须对任何序列成立,可得\( X \)必须是有限集合,这样你随便构造个序列,你必须取无限个元素作为序列元素,而\( X \)中仅有有限个元素,这意味着你取的这无限个元素中,必至少有一个元素会重复无限次,这个元素就是序列的极限点,进而该序列包含无限个极限点。
这里,我们断言,当\( X \)为有限集时,\( X \)为紧集,当\( X \)为无限集时,\( X \)非紧集,我们下面去证明这两点。
当\( X \)为有限集时,给定任意\( X \)上的序列\( (x_n)_{n = 1}^{\infty} \),由于\( X \)仅有有限个元素,这意味着\( (x_n)_{n = 1}^{\infty} \)中存在一个\( X \)的元素\( x_0 \)重复无限多次,易得\( x_0 \)为\( (x_n)_{n = 1}^{\infty} \)的极限点,综上,\( X \)上的所有序列均有极限点,进而根据练习1.5.9,可得\( X \)为紧集。
当\( X \)为无限集时,则可构造一个序列\( (x_n)_{n = 1}^{\infty} \),\( \forall n \geq 1, x_n \)均与前面的所有序列元素不同,易得该序列中所有元素都互不相同,而根据前面的讨论,该序列要想有个极限点\( x_0 \),则它必须包含\( x_0 \)无限多次,这和该序列所有元素互不相同矛盾,因此该序列没有极限点,根据练习1.5.9,可得\( X \)非紧集。
练习1.5.13
题目:
Let \( E \) and \( F \) be two compact subsets of \( \mathbf{R} \) (with the standard metric \( d(x, y) = |x - y| \)). Show that the Cartesian product \( E \times F := \{ (x, y) : x \in E, y \in F \} \) is a compact subset of \( \mathbf{R}^2 \) (with the Euclidean metric \( d_{l^2} \)).
证明:
由Heine–Borel定理(定理1.5.7),可得\( E, F \)均是封闭且有界的。
下面,我们证明\( E \times F \)也是封闭且有界的,进一步使用Heine–Borel定理,就能说明\( E \times F \)为紧集了。
先证明\( E \times F \)是闭集,这里通过1.2.15(b)来证明,我们只要证明针对\( E \times F \)上的任意收敛序列,该序列的极限都在\( E \times F \)内,就能说明\( E \times F \)是闭集了:给定任意\( E \times F \)上的收敛序列\( (p^{(n)})_{n = 1}^{\infty} \),这里\( \forall n \geq 1 \),我们记\( p^{(n)} = (p_x^{(n)}, p_y^{(n)}) \),我们还记\( \lim_{n \to \infty} p^{(n)} = (a, b) \),根据定理1.1.18,可得\( (p_x^{(n)})_{n = 1}^{\infty}, (p_y^{(n)})_{n = 1}^{\infty} \)分别收敛于\( a, b \),进而根据\( E, F \)是闭集,可得\( a \in E, b \in F \),进而\( \lim_{n \to \infty} p^{(n)} = (a, b) \in E \times F \),此时根据1.2.15(b),可得\( E \times F \)为闭集。
接着证明\( E \times F \)有界,由\( E, F \)均有界,可得存在\( M_1, M_2 > 0 \)满足\( E \subseteq [-M_1, M_1], F \subseteq [-M_2, M_2] \),给定任意\( (x, y) \in E \times F \),有\( \forall (a, b) \in E \times F, d_{l^2}((x, y), (a, b)) \leq d_{l^2}((x, y), (0, 0)) + d_{l^2}((0, 0), (a, b)) = \sqrt{x^2 + y^2} + \sqrt{a^2 + b^2} \),而由\( E \subseteq [-M_1, M_1], F \subseteq [-M_2, M_2] \),可得\( \sqrt{x^2 + y^2} \leq \sqrt{M_1^2 + M_2^2}, \sqrt{a^2 + b^2} \leq \sqrt{M_1^2 + M_2^2} \),进而可得\( d_{l^2}((x, y), (a, b)) \leq 2 \sqrt{M_1^2 + M_2^2} < 2 \sqrt{M_1^2 + M_2^2} + 1 \),令\( M = 2 \sqrt{M_1^2 + M_2^2} + 1 \),可得\( E \times F \subseteq B((x, y), M) \),至此可得\( E \times F \)有界。
综上,由\( E \times F \)封闭且有界以及Heine–Borel定理,可得\( E \times F \)为紧集。
证毕。
练习1.5.14
题目:
Let \( (X, d) \) be a metric space, let \( E \) be a non-empty compact subset of \( X \), and let \( x_0 \) be a point in \( X \). Show that there exists a point \( x \in E \) such that \[ d(x_0, x) = \inf \{ d(x_0, y) : y \in E \}, \] i.e., \( x \) is the closest point in \( E \) to \( x_0 \). (Hint: let \( R \) be the quantity \( R := \inf \{ d(x_0, y) : y \in E \} \). Construct a sequence \( (x^{(n)})_{n = 1}^{\infty} \) in \( E \) such that \( d(x_0, x^{(n)}) \leq R + \dfrac{1}{n} \), and then use the compactness of \( E \).)
证明:
根据提示来做。
令\( R = \inf \{ d(x_0, y) : y \in E \} \),定义序列\( x^{(n)})_{n = 1}^{\infty} \), \( \forall n \geq 1 \),由于\( R + \dfrac{1}{n} > R \),而\( R \)是\( \{ d(x_0, y) : y \in E \} \)的下界,可得\( \exists x' \in \{ d(x_0, y) : y \in E \}, x' < R + \dfrac{1}{n} \),可得\( \exists y' \in E, d(x_0, y') = x' < R + \dfrac{1}{n} \),令\( x^{(n)} = y' \),有\( x^{(n)} \in E, d(x_0, x^{(n)}) < R + \dfrac{1}{n} \)。由\( x^{(n)})_{n = 1}^{\infty} \)是\( E \)上的序列以及\( E \)为紧集,可得\( x^{(n)})_{n = 1}^{\infty} \)存在收敛的子序列 \( x^{(n_j)})_{j = 1}^{\infty} \),令\( x = \lim_{j \to \infty} x^{(n_j)} \),我们打算证明\( d(x_0, x) = R \)。
根据推论1.5.6,可得\( E \)是闭集,进而根据1.2.15(b),可得\( x \in E \),而\( R \)是\( \{ d(x_0, y) : y \in E \} \)的下界,这意味着\( R \leq d(x_0, x) \)。
\( \forall \epsilon > 0 \),由\( x = \lim_{j \to \infty} x^{(n_j)} \),可得\( \exists J \geq 1, \forall j \geq J, d(x^{(n_j)}, x) \leq \dfrac{\epsilon}{2} \),又存在\( j \geq J \)满足\( \dfrac{1}{j} \leq \dfrac{\epsilon}{2} \),于是可得\( d(x_0, x) \leq d(x_0, x^{(n_j)}) + d(x^{(n_j)}, x) \leq (R + \dfrac{1}{n_j}) + \dfrac{\epsilon}{2} \leq (R + \dfrac{1}{j}) + \dfrac{\epsilon}{2} \leq (R + \dfrac{\epsilon}{2}) + \dfrac{\epsilon}{2} = R + \epsilon \),也就是\( d(x_0, x) \leq R + \epsilon \),加上前面得到的\( R \leq d(x_0, x) \),有\( R \leq d(x_0, x) \leq R + \epsilon \),而这两个不等式对所有\( \epsilon > 0 \)都成立,可得\( d(x_0, x) = R \),也就是说, \( x \)就是我们要找的点。
证毕。
练习1.5.15
题目:
Let \( (X, d) \) be a compact metric space. Suppose that \( (K_{\alpha})_{\alpha \in I} \) is a collection of closed sets in \( X \) with the property that any finite subcollection of these sets necessarily has non-empty intersection, thus \( \bigcap_{\alpha \in F} K_{\alpha} \neq \empty \) for all finite \( F \subseteq I \). (This property is known as the finite intersection property.) Show that the entire collection has non-empty intersection, thus \( \bigcap_{\alpha \in I} K_{\alpha} \neq \empty \). Show by counterexample that this statement fails if \( X \) is not compact.
注:
需要清楚明显有限下标子集\( F \)不能是空集。
证明:
假设\( \bigcap_{\alpha \in I} K_{\alpha} = \empty \),则\( X \setminus (\bigcap_{\alpha \in I} K_{\alpha}) = X \setminus \empty = X \),而\( X \setminus (\bigcap_{\alpha \in I} K_{\alpha}) = \bigcup_{\alpha \in I} (X \setminus K_{\alpha}) \),这里\( \forall \alpha \in I, K_{\alpha} \)为闭集,因此\( X \setminus K_{\alpha} \)为开集(注:定理1.2.15(e)),这说明\( \{ X \setminus K_{\alpha} : \alpha \in I \} \)为\( X \)的开覆盖,由\( X \)是紧集以及定理1.5.8,可得存在有限集\( F \subseteq I \),使得\( X \subseteq \bigcup_{\alpha \in F} (X \setminus K_{\alpha}) = X \setminus (\bigcap_{\alpha \in F} K_{\alpha}) \),这意味着\( \bigcap_{\alpha \in F} K_{\alpha} = \empty \),和题设矛盾,因此假设不成立,有\( \bigcap_{\alpha \in I} K_{\alpha} \neq \empty \)。
证毕。